diff --git "a/trainer_state.json" "b/trainer_state.json" deleted file mode 100644--- "a/trainer_state.json" +++ /dev/null @@ -1,19276 +0,0 @@ -{ - "best_metric": null, - "best_model_checkpoint": null, - "epoch": 1.0, - "eval_steps": 500, - "global_step": 3208, - "is_hyper_param_search": false, - "is_local_process_zero": true, - "is_world_process_zero": true, - "log_history": [ - { - "epoch": 0.0, - "learning_rate": 2e-08, - "loss": 2.0757, - "step": 1 - }, - { - "epoch": 0.0, - "learning_rate": 4e-08, - "loss": 2.108, - "step": 2 - }, - { - "epoch": 0.0, - "learning_rate": 6e-08, - "loss": 2.0124, - "step": 3 - }, - { - "epoch": 0.0, - "learning_rate": 8e-08, - "loss": 2.1088, - "step": 4 - }, - { - "epoch": 0.0, - "learning_rate": 1e-07, - "loss": 2.0745, - "step": 5 - }, - { - "epoch": 0.0, - "learning_rate": 1.2e-07, - "loss": 2.0838, - "step": 6 - }, - { - "epoch": 0.0, - "learning_rate": 1.4e-07, - "loss": 2.0833, - "step": 7 - }, - { - "epoch": 0.0, - "learning_rate": 1.6e-07, - "loss": 2.084, - "step": 8 - }, - { - "epoch": 0.0, - "learning_rate": 1.8e-07, - "loss": 2.0779, - "step": 9 - }, - { - "epoch": 0.0, - "learning_rate": 2e-07, - "loss": 2.1059, - "step": 10 - }, - { - "epoch": 0.0, - "learning_rate": 2.1999999999999998e-07, - "loss": 2.051, - "step": 11 - }, - { - "epoch": 0.0, - "learning_rate": 2.4e-07, - "loss": 2.0392, - "step": 12 - }, - { - "epoch": 0.0, - "learning_rate": 2.6e-07, - "loss": 2.0617, - "step": 13 - }, - { - "epoch": 0.0, - "learning_rate": 2.8e-07, - "loss": 2.1052, - "step": 14 - }, - { - "epoch": 0.0, - "learning_rate": 3e-07, - "loss": 2.1029, - "step": 15 - }, - { - "epoch": 0.0, - "learning_rate": 3.2e-07, - "loss": 2.037, - "step": 16 - }, - { - "epoch": 0.01, - "learning_rate": 3.4000000000000003e-07, - "loss": 2.0572, - "step": 17 - }, - { - "epoch": 0.01, - "learning_rate": 3.6e-07, - "loss": 2.049, - "step": 18 - }, - { - "epoch": 0.01, - "learning_rate": 3.7999999999999996e-07, - "loss": 2.0519, - "step": 19 - }, - { - "epoch": 0.01, - "learning_rate": 4e-07, - "loss": 2.0995, - "step": 20 - }, - { - "epoch": 0.01, - "learning_rate": 4.1999999999999995e-07, - "loss": 2.0938, - "step": 21 - }, - { - "epoch": 0.01, - "learning_rate": 4.3999999999999997e-07, - "loss": 2.0736, - "step": 22 - }, - { - "epoch": 0.01, - "learning_rate": 4.6e-07, - "loss": 2.0531, - "step": 23 - }, - { - "epoch": 0.01, - "learning_rate": 4.8e-07, - "loss": 2.0576, - "step": 24 - }, - { - "epoch": 0.01, - "learning_rate": 5e-07, - "loss": 2.0346, - "step": 25 - }, - { - "epoch": 0.01, - "learning_rate": 5.2e-07, - "loss": 1.9808, - "step": 26 - }, - { - "epoch": 0.01, - "learning_rate": 5.4e-07, - "loss": 2.0266, - "step": 27 - }, - { - "epoch": 0.01, - "learning_rate": 5.6e-07, - "loss": 2.0269, - "step": 28 - }, - { - "epoch": 0.01, - "learning_rate": 5.8e-07, - "loss": 1.9745, - "step": 29 - }, - { - "epoch": 0.01, - "learning_rate": 6e-07, - "loss": 2.0107, - "step": 30 - }, - { - "epoch": 0.01, - "learning_rate": 6.2e-07, - "loss": 1.9893, - "step": 31 - }, - { - "epoch": 0.01, - "learning_rate": 6.4e-07, - "loss": 1.9282, - "step": 32 - }, - { - "epoch": 0.01, - "learning_rate": 6.6e-07, - "loss": 1.9669, - "step": 33 - }, - { - "epoch": 0.01, - "learning_rate": 6.800000000000001e-07, - "loss": 1.9762, - "step": 34 - }, - { - "epoch": 0.01, - "learning_rate": 7e-07, - "loss": 1.882, - "step": 35 - }, - { - "epoch": 0.01, - "learning_rate": 7.2e-07, - "loss": 1.8934, - "step": 36 - }, - { - "epoch": 0.01, - "learning_rate": 7.4e-07, - "loss": 1.8637, - "step": 37 - }, - { - "epoch": 0.01, - "learning_rate": 7.599999999999999e-07, - "loss": 1.8751, - "step": 38 - }, - { - "epoch": 0.01, - "learning_rate": 7.799999999999999e-07, - "loss": 1.8834, - "step": 39 - }, - { - "epoch": 0.01, - "learning_rate": 8e-07, - "loss": 1.8791, - "step": 40 - }, - { - "epoch": 0.01, - "learning_rate": 8.199999999999999e-07, - "loss": 1.8564, - "step": 41 - }, - { - "epoch": 0.01, - "learning_rate": 8.399999999999999e-07, - "loss": 1.8322, - "step": 42 - }, - { - "epoch": 0.01, - "learning_rate": 8.599999999999999e-07, - "loss": 1.786, - "step": 43 - }, - { - "epoch": 0.01, - "learning_rate": 8.799999999999999e-07, - "loss": 1.7528, - "step": 44 - }, - { - "epoch": 0.01, - "learning_rate": 9e-07, - "loss": 1.7296, - "step": 45 - }, - { - "epoch": 0.01, - "learning_rate": 9.2e-07, - "loss": 1.6692, - "step": 46 - }, - { - "epoch": 0.01, - "learning_rate": 9.399999999999999e-07, - "loss": 1.6781, - "step": 47 - }, - { - "epoch": 0.01, - "learning_rate": 9.6e-07, - "loss": 1.6542, - "step": 48 - }, - { - "epoch": 0.02, - "learning_rate": 9.8e-07, - "loss": 1.6239, - "step": 49 - }, - { - "epoch": 0.02, - "learning_rate": 1e-06, - "loss": 1.6107, - "step": 50 - }, - { - "epoch": 0.02, - "learning_rate": 9.999997525910152e-07, - "loss": 1.6193, - "step": 51 - }, - { - "epoch": 0.02, - "learning_rate": 9.999990103643063e-07, - "loss": 1.5841, - "step": 52 - }, - { - "epoch": 0.02, - "learning_rate": 9.999977733206075e-07, - "loss": 1.5725, - "step": 53 - }, - { - "epoch": 0.02, - "learning_rate": 9.99996041461143e-07, - "loss": 1.5456, - "step": 54 - }, - { - "epoch": 0.02, - "learning_rate": 9.99993814787627e-07, - "loss": 1.6147, - "step": 55 - }, - { - "epoch": 0.02, - "learning_rate": 9.999910933022626e-07, - "loss": 1.585, - "step": 56 - }, - { - "epoch": 0.02, - "learning_rate": 9.999878770077437e-07, - "loss": 1.5471, - "step": 57 - }, - { - "epoch": 0.02, - "learning_rate": 9.999841659072525e-07, - "loss": 1.5124, - "step": 58 - }, - { - "epoch": 0.02, - "learning_rate": 9.999799600044625e-07, - "loss": 1.5382, - "step": 59 - }, - { - "epoch": 0.02, - "learning_rate": 9.999752593035354e-07, - "loss": 1.5136, - "step": 60 - }, - { - "epoch": 0.02, - "learning_rate": 9.999700638091232e-07, - "loss": 1.4647, - "step": 61 - }, - { - "epoch": 0.02, - "learning_rate": 9.99964373526368e-07, - "loss": 1.4128, - "step": 62 - }, - { - "epoch": 0.02, - "learning_rate": 9.999581884609006e-07, - "loss": 1.4234, - "step": 63 - }, - { - "epoch": 0.02, - "learning_rate": 9.999515086188423e-07, - "loss": 1.3618, - "step": 64 - }, - { - "epoch": 0.02, - "learning_rate": 9.999443340068036e-07, - "loss": 1.3327, - "step": 65 - }, - { - "epoch": 0.02, - "learning_rate": 9.999366646318848e-07, - "loss": 1.3721, - "step": 66 - }, - { - "epoch": 0.02, - "learning_rate": 9.999285005016756e-07, - "loss": 1.3082, - "step": 67 - }, - { - "epoch": 0.02, - "learning_rate": 9.999198416242555e-07, - "loss": 1.2808, - "step": 68 - }, - { - "epoch": 0.02, - "learning_rate": 9.999106880081942e-07, - "loss": 1.2859, - "step": 69 - }, - { - "epoch": 0.02, - "learning_rate": 9.999010396625498e-07, - "loss": 1.2491, - "step": 70 - }, - { - "epoch": 0.02, - "learning_rate": 9.998908965968708e-07, - "loss": 1.2633, - "step": 71 - }, - { - "epoch": 0.02, - "learning_rate": 9.998802588211954e-07, - "loss": 1.2479, - "step": 72 - }, - { - "epoch": 0.02, - "learning_rate": 9.99869126346051e-07, - "loss": 1.2733, - "step": 73 - }, - { - "epoch": 0.02, - "learning_rate": 9.998574991824545e-07, - "loss": 1.2285, - "step": 74 - }, - { - "epoch": 0.02, - "learning_rate": 9.998453773419128e-07, - "loss": 1.2153, - "step": 75 - }, - { - "epoch": 0.02, - "learning_rate": 9.99832760836422e-07, - "loss": 1.1915, - "step": 76 - }, - { - "epoch": 0.02, - "learning_rate": 9.998196496784678e-07, - "loss": 1.1936, - "step": 77 - }, - { - "epoch": 0.02, - "learning_rate": 9.998060438810256e-07, - "loss": 1.1851, - "step": 78 - }, - { - "epoch": 0.02, - "learning_rate": 9.9979194345756e-07, - "loss": 1.1862, - "step": 79 - }, - { - "epoch": 0.02, - "learning_rate": 9.997773484220256e-07, - "loss": 1.187, - "step": 80 - }, - { - "epoch": 0.03, - "learning_rate": 9.99762258788866e-07, - "loss": 1.1626, - "step": 81 - }, - { - "epoch": 0.03, - "learning_rate": 9.997466745730142e-07, - "loss": 1.1692, - "step": 82 - }, - { - "epoch": 0.03, - "learning_rate": 9.997305957898934e-07, - "loss": 1.158, - "step": 83 - }, - { - "epoch": 0.03, - "learning_rate": 9.997140224554153e-07, - "loss": 1.1727, - "step": 84 - }, - { - "epoch": 0.03, - "learning_rate": 9.996969545859817e-07, - "loss": 1.0935, - "step": 85 - }, - { - "epoch": 0.03, - "learning_rate": 9.996793921984837e-07, - "loss": 1.1423, - "step": 86 - }, - { - "epoch": 0.03, - "learning_rate": 9.996613353103012e-07, - "loss": 1.1355, - "step": 87 - }, - { - "epoch": 0.03, - "learning_rate": 9.996427839393043e-07, - "loss": 1.1198, - "step": 88 - }, - { - "epoch": 0.03, - "learning_rate": 9.99623738103852e-07, - "loss": 1.166, - "step": 89 - }, - { - "epoch": 0.03, - "learning_rate": 9.996041978227928e-07, - "loss": 1.1377, - "step": 90 - }, - { - "epoch": 0.03, - "learning_rate": 9.995841631154645e-07, - "loss": 1.0944, - "step": 91 - }, - { - "epoch": 0.03, - "learning_rate": 9.99563634001694e-07, - "loss": 1.127, - "step": 92 - }, - { - "epoch": 0.03, - "learning_rate": 9.995426105017979e-07, - "loss": 1.123, - "step": 93 - }, - { - "epoch": 0.03, - "learning_rate": 9.995210926365817e-07, - "loss": 1.0766, - "step": 94 - }, - { - "epoch": 0.03, - "learning_rate": 9.9949908042734e-07, - "loss": 1.0756, - "step": 95 - }, - { - "epoch": 0.03, - "learning_rate": 9.99476573895857e-07, - "loss": 1.1163, - "step": 96 - }, - { - "epoch": 0.03, - "learning_rate": 9.994535730644064e-07, - "loss": 1.0828, - "step": 97 - }, - { - "epoch": 0.03, - "learning_rate": 9.994300779557501e-07, - "loss": 1.0854, - "step": 98 - }, - { - "epoch": 0.03, - "learning_rate": 9.9940608859314e-07, - "loss": 1.0979, - "step": 99 - }, - { - "epoch": 0.03, - "learning_rate": 9.993816050003168e-07, - "loss": 1.0783, - "step": 100 - }, - { - "epoch": 0.03, - "learning_rate": 9.993566272015102e-07, - "loss": 1.073, - "step": 101 - }, - { - "epoch": 0.03, - "learning_rate": 9.993311552214393e-07, - "loss": 1.0393, - "step": 102 - }, - { - "epoch": 0.03, - "learning_rate": 9.99305189085312e-07, - "loss": 1.0539, - "step": 103 - }, - { - "epoch": 0.03, - "learning_rate": 9.992787288188253e-07, - "loss": 1.0198, - "step": 104 - }, - { - "epoch": 0.03, - "learning_rate": 9.992517744481652e-07, - "loss": 1.0335, - "step": 105 - }, - { - "epoch": 0.03, - "learning_rate": 9.99224326000007e-07, - "loss": 1.0205, - "step": 106 - }, - { - "epoch": 0.03, - "learning_rate": 9.991963835015142e-07, - "loss": 1.0248, - "step": 107 - }, - { - "epoch": 0.03, - "learning_rate": 9.991679469803401e-07, - "loss": 1.0122, - "step": 108 - }, - { - "epoch": 0.03, - "learning_rate": 9.991390164646263e-07, - "loss": 1.0211, - "step": 109 - }, - { - "epoch": 0.03, - "learning_rate": 9.991095919830034e-07, - "loss": 0.9916, - "step": 110 - }, - { - "epoch": 0.03, - "learning_rate": 9.990796735645912e-07, - "loss": 1.0104, - "step": 111 - }, - { - "epoch": 0.03, - "learning_rate": 9.99049261238998e-07, - "loss": 0.9783, - "step": 112 - }, - { - "epoch": 0.04, - "learning_rate": 9.990183550363206e-07, - "loss": 1.0295, - "step": 113 - }, - { - "epoch": 0.04, - "learning_rate": 9.989869549871453e-07, - "loss": 0.9861, - "step": 114 - }, - { - "epoch": 0.04, - "learning_rate": 9.989550611225466e-07, - "loss": 1.0223, - "step": 115 - }, - { - "epoch": 0.04, - "learning_rate": 9.989226734740876e-07, - "loss": 0.9758, - "step": 116 - }, - { - "epoch": 0.04, - "learning_rate": 9.988897920738203e-07, - "loss": 0.9953, - "step": 117 - }, - { - "epoch": 0.04, - "learning_rate": 9.988564169542857e-07, - "loss": 0.961, - "step": 118 - }, - { - "epoch": 0.04, - "learning_rate": 9.988225481485125e-07, - "loss": 0.9865, - "step": 119 - }, - { - "epoch": 0.04, - "learning_rate": 9.98788185690019e-07, - "loss": 0.9974, - "step": 120 - }, - { - "epoch": 0.04, - "learning_rate": 9.98753329612811e-07, - "loss": 0.9533, - "step": 121 - }, - { - "epoch": 0.04, - "learning_rate": 9.987179799513838e-07, - "loss": 0.9635, - "step": 122 - }, - { - "epoch": 0.04, - "learning_rate": 9.986821367407206e-07, - "loss": 0.9598, - "step": 123 - }, - { - "epoch": 0.04, - "learning_rate": 9.986458000162929e-07, - "loss": 0.9645, - "step": 124 - }, - { - "epoch": 0.04, - "learning_rate": 9.98608969814061e-07, - "loss": 0.9262, - "step": 125 - }, - { - "epoch": 0.04, - "learning_rate": 9.985716461704731e-07, - "loss": 0.9116, - "step": 126 - }, - { - "epoch": 0.04, - "learning_rate": 9.985338291224666e-07, - "loss": 0.9839, - "step": 127 - }, - { - "epoch": 0.04, - "learning_rate": 9.98495518707466e-07, - "loss": 0.9367, - "step": 128 - }, - { - "epoch": 0.04, - "learning_rate": 9.984567149633852e-07, - "loss": 0.9253, - "step": 129 - }, - { - "epoch": 0.04, - "learning_rate": 9.984174179286253e-07, - "loss": 0.9197, - "step": 130 - }, - { - "epoch": 0.04, - "learning_rate": 9.983776276420763e-07, - "loss": 0.9174, - "step": 131 - }, - { - "epoch": 0.04, - "learning_rate": 9.983373441431161e-07, - "loss": 0.9025, - "step": 132 - }, - { - "epoch": 0.04, - "learning_rate": 9.982965674716108e-07, - "loss": 0.8936, - "step": 133 - }, - { - "epoch": 0.04, - "learning_rate": 9.982552976679143e-07, - "loss": 0.9193, - "step": 134 - }, - { - "epoch": 0.04, - "learning_rate": 9.982135347728684e-07, - "loss": 0.8924, - "step": 135 - }, - { - "epoch": 0.04, - "learning_rate": 9.981712788278038e-07, - "loss": 0.8799, - "step": 136 - }, - { - "epoch": 0.04, - "learning_rate": 9.98128529874538e-07, - "loss": 0.8717, - "step": 137 - }, - { - "epoch": 0.04, - "learning_rate": 9.980852879553773e-07, - "loss": 0.8848, - "step": 138 - }, - { - "epoch": 0.04, - "learning_rate": 9.98041553113115e-07, - "loss": 0.8862, - "step": 139 - }, - { - "epoch": 0.04, - "learning_rate": 9.97997325391033e-07, - "loss": 0.8698, - "step": 140 - }, - { - "epoch": 0.04, - "learning_rate": 9.979526048329007e-07, - "loss": 0.8706, - "step": 141 - }, - { - "epoch": 0.04, - "learning_rate": 9.979073914829746e-07, - "loss": 0.896, - "step": 142 - }, - { - "epoch": 0.04, - "learning_rate": 9.978616853860002e-07, - "loss": 0.8758, - "step": 143 - }, - { - "epoch": 0.04, - "learning_rate": 9.978154865872095e-07, - "loss": 0.872, - "step": 144 - }, - { - "epoch": 0.05, - "learning_rate": 9.977687951323226e-07, - "loss": 0.8485, - "step": 145 - }, - { - "epoch": 0.05, - "learning_rate": 9.977216110675468e-07, - "loss": 0.8587, - "step": 146 - }, - { - "epoch": 0.05, - "learning_rate": 9.976739344395776e-07, - "loss": 0.8533, - "step": 147 - }, - { - "epoch": 0.05, - "learning_rate": 9.97625765295597e-07, - "loss": 0.8496, - "step": 148 - }, - { - "epoch": 0.05, - "learning_rate": 9.975771036832755e-07, - "loss": 0.8594, - "step": 149 - }, - { - "epoch": 0.05, - "learning_rate": 9.9752794965077e-07, - "loss": 0.8727, - "step": 150 - }, - { - "epoch": 0.05, - "learning_rate": 9.97478303246725e-07, - "loss": 0.8579, - "step": 151 - }, - { - "epoch": 0.05, - "learning_rate": 9.974281645202725e-07, - "loss": 0.8385, - "step": 152 - }, - { - "epoch": 0.05, - "learning_rate": 9.973775335210317e-07, - "loss": 0.8555, - "step": 153 - }, - { - "epoch": 0.05, - "learning_rate": 9.973264102991088e-07, - "loss": 0.8538, - "step": 154 - }, - { - "epoch": 0.05, - "learning_rate": 9.972747949050969e-07, - "loss": 0.8596, - "step": 155 - }, - { - "epoch": 0.05, - "learning_rate": 9.972226873900769e-07, - "loss": 0.8389, - "step": 156 - }, - { - "epoch": 0.05, - "learning_rate": 9.97170087805616e-07, - "loss": 0.8469, - "step": 157 - }, - { - "epoch": 0.05, - "learning_rate": 9.971169962037686e-07, - "loss": 0.8492, - "step": 158 - }, - { - "epoch": 0.05, - "learning_rate": 9.970634126370763e-07, - "loss": 0.8361, - "step": 159 - }, - { - "epoch": 0.05, - "learning_rate": 9.970093371585671e-07, - "loss": 0.8257, - "step": 160 - }, - { - "epoch": 0.05, - "learning_rate": 9.969547698217559e-07, - "loss": 0.8359, - "step": 161 - }, - { - "epoch": 0.05, - "learning_rate": 9.96899710680645e-07, - "loss": 0.8265, - "step": 162 - }, - { - "epoch": 0.05, - "learning_rate": 9.968441597897226e-07, - "loss": 0.7988, - "step": 163 - }, - { - "epoch": 0.05, - "learning_rate": 9.967881172039638e-07, - "loss": 0.8226, - "step": 164 - }, - { - "epoch": 0.05, - "learning_rate": 9.967315829788303e-07, - "loss": 0.8403, - "step": 165 - }, - { - "epoch": 0.05, - "learning_rate": 9.966745571702706e-07, - "loss": 0.8304, - "step": 166 - }, - { - "epoch": 0.05, - "learning_rate": 9.966170398347196e-07, - "loss": 0.7968, - "step": 167 - }, - { - "epoch": 0.05, - "learning_rate": 9.965590310290982e-07, - "loss": 0.8184, - "step": 168 - }, - { - "epoch": 0.05, - "learning_rate": 9.965005308108142e-07, - "loss": 0.8002, - "step": 169 - }, - { - "epoch": 0.05, - "learning_rate": 9.964415392377612e-07, - "loss": 0.8085, - "step": 170 - }, - { - "epoch": 0.05, - "learning_rate": 9.963820563683198e-07, - "loss": 0.8079, - "step": 171 - }, - { - "epoch": 0.05, - "learning_rate": 9.963220822613562e-07, - "loss": 0.8008, - "step": 172 - }, - { - "epoch": 0.05, - "learning_rate": 9.96261616976223e-07, - "loss": 0.8046, - "step": 173 - }, - { - "epoch": 0.05, - "learning_rate": 9.96200660572759e-07, - "loss": 0.7712, - "step": 174 - }, - { - "epoch": 0.05, - "learning_rate": 9.961392131112882e-07, - "loss": 0.7761, - "step": 175 - }, - { - "epoch": 0.05, - "learning_rate": 9.96077274652622e-07, - "loss": 0.7897, - "step": 176 - }, - { - "epoch": 0.06, - "learning_rate": 9.960148452580562e-07, - "loss": 0.7934, - "step": 177 - }, - { - "epoch": 0.06, - "learning_rate": 9.959519249893738e-07, - "loss": 0.7875, - "step": 178 - }, - { - "epoch": 0.06, - "learning_rate": 9.958885139088426e-07, - "loss": 0.7831, - "step": 179 - }, - { - "epoch": 0.06, - "learning_rate": 9.958246120792166e-07, - "loss": 0.8056, - "step": 180 - }, - { - "epoch": 0.06, - "learning_rate": 9.957602195637354e-07, - "loss": 0.7743, - "step": 181 - }, - { - "epoch": 0.06, - "learning_rate": 9.95695336426124e-07, - "loss": 0.7677, - "step": 182 - }, - { - "epoch": 0.06, - "learning_rate": 9.956299627305931e-07, - "loss": 0.772, - "step": 183 - }, - { - "epoch": 0.06, - "learning_rate": 9.95564098541839e-07, - "loss": 0.7893, - "step": 184 - }, - { - "epoch": 0.06, - "learning_rate": 9.954977439250432e-07, - "loss": 0.7413, - "step": 185 - }, - { - "epoch": 0.06, - "learning_rate": 9.954308989458725e-07, - "loss": 0.7822, - "step": 186 - }, - { - "epoch": 0.06, - "learning_rate": 9.953635636704792e-07, - "loss": 0.7944, - "step": 187 - }, - { - "epoch": 0.06, - "learning_rate": 9.952957381655007e-07, - "loss": 0.7657, - "step": 188 - }, - { - "epoch": 0.06, - "learning_rate": 9.952274224980598e-07, - "loss": 0.7727, - "step": 189 - }, - { - "epoch": 0.06, - "learning_rate": 9.951586167357636e-07, - "loss": 0.7471, - "step": 190 - }, - { - "epoch": 0.06, - "learning_rate": 9.95089320946705e-07, - "loss": 0.7897, - "step": 191 - }, - { - "epoch": 0.06, - "learning_rate": 9.950195351994619e-07, - "loss": 0.7864, - "step": 192 - }, - { - "epoch": 0.06, - "learning_rate": 9.949492595630963e-07, - "loss": 0.7727, - "step": 193 - }, - { - "epoch": 0.06, - "learning_rate": 9.94878494107156e-07, - "loss": 0.7582, - "step": 194 - }, - { - "epoch": 0.06, - "learning_rate": 9.948072389016727e-07, - "loss": 0.7725, - "step": 195 - }, - { - "epoch": 0.06, - "learning_rate": 9.94735494017163e-07, - "loss": 0.7795, - "step": 196 - }, - { - "epoch": 0.06, - "learning_rate": 9.946632595246285e-07, - "loss": 0.7981, - "step": 197 - }, - { - "epoch": 0.06, - "learning_rate": 9.94590535495555e-07, - "loss": 0.7402, - "step": 198 - }, - { - "epoch": 0.06, - "learning_rate": 9.945173220019127e-07, - "loss": 0.7373, - "step": 199 - }, - { - "epoch": 0.06, - "learning_rate": 9.944436191161565e-07, - "loss": 0.7234, - "step": 200 - }, - { - "epoch": 0.06, - "learning_rate": 9.943694269112251e-07, - "loss": 0.7582, - "step": 201 - }, - { - "epoch": 0.06, - "learning_rate": 9.942947454605422e-07, - "loss": 0.7385, - "step": 202 - }, - { - "epoch": 0.06, - "learning_rate": 9.942195748380148e-07, - "loss": 0.7543, - "step": 203 - }, - { - "epoch": 0.06, - "learning_rate": 9.941439151180346e-07, - "loss": 0.7336, - "step": 204 - }, - { - "epoch": 0.06, - "learning_rate": 9.940677663754775e-07, - "loss": 0.7354, - "step": 205 - }, - { - "epoch": 0.06, - "learning_rate": 9.939911286857025e-07, - "loss": 0.7611, - "step": 206 - }, - { - "epoch": 0.06, - "learning_rate": 9.939140021245535e-07, - "loss": 0.7743, - "step": 207 - }, - { - "epoch": 0.06, - "learning_rate": 9.938363867683575e-07, - "loss": 0.7315, - "step": 208 - }, - { - "epoch": 0.07, - "learning_rate": 9.937582826939252e-07, - "loss": 0.7343, - "step": 209 - }, - { - "epoch": 0.07, - "learning_rate": 9.936796899785518e-07, - "loss": 0.7201, - "step": 210 - }, - { - "epoch": 0.07, - "learning_rate": 9.93600608700015e-07, - "loss": 0.7385, - "step": 211 - }, - { - "epoch": 0.07, - "learning_rate": 9.935210389365764e-07, - "loss": 0.727, - "step": 212 - }, - { - "epoch": 0.07, - "learning_rate": 9.934409807669814e-07, - "loss": 0.76, - "step": 213 - }, - { - "epoch": 0.07, - "learning_rate": 9.933604342704584e-07, - "loss": 0.7198, - "step": 214 - }, - { - "epoch": 0.07, - "learning_rate": 9.93279399526719e-07, - "loss": 0.747, - "step": 215 - }, - { - "epoch": 0.07, - "learning_rate": 9.931978766159581e-07, - "loss": 0.7619, - "step": 216 - }, - { - "epoch": 0.07, - "learning_rate": 9.931158656188538e-07, - "loss": 0.733, - "step": 217 - }, - { - "epoch": 0.07, - "learning_rate": 9.93033366616567e-07, - "loss": 0.7517, - "step": 218 - }, - { - "epoch": 0.07, - "learning_rate": 9.929503796907418e-07, - "loss": 0.7119, - "step": 219 - }, - { - "epoch": 0.07, - "learning_rate": 9.92866904923505e-07, - "loss": 0.7258, - "step": 220 - }, - { - "epoch": 0.07, - "learning_rate": 9.92782942397466e-07, - "loss": 0.7147, - "step": 221 - }, - { - "epoch": 0.07, - "learning_rate": 9.926984921957178e-07, - "loss": 0.733, - "step": 222 - }, - { - "epoch": 0.07, - "learning_rate": 9.926135544018346e-07, - "loss": 0.7049, - "step": 223 - }, - { - "epoch": 0.07, - "learning_rate": 9.925281290998742e-07, - "loss": 0.7368, - "step": 224 - }, - { - "epoch": 0.07, - "learning_rate": 9.924422163743765e-07, - "loss": 0.7102, - "step": 225 - }, - { - "epoch": 0.07, - "learning_rate": 9.923558163103641e-07, - "loss": 0.7081, - "step": 226 - }, - { - "epoch": 0.07, - "learning_rate": 9.922689289933413e-07, - "loss": 0.6959, - "step": 227 - }, - { - "epoch": 0.07, - "learning_rate": 9.921815545092949e-07, - "loss": 0.7129, - "step": 228 - }, - { - "epoch": 0.07, - "learning_rate": 9.920936929446941e-07, - "loss": 0.71, - "step": 229 - }, - { - "epoch": 0.07, - "learning_rate": 9.920053443864896e-07, - "loss": 0.7064, - "step": 230 - }, - { - "epoch": 0.07, - "learning_rate": 9.919165089221143e-07, - "loss": 0.6966, - "step": 231 - }, - { - "epoch": 0.07, - "learning_rate": 9.91827186639483e-07, - "loss": 0.7206, - "step": 232 - }, - { - "epoch": 0.07, - "learning_rate": 9.917373776269926e-07, - "loss": 0.7244, - "step": 233 - }, - { - "epoch": 0.07, - "learning_rate": 9.916470819735208e-07, - "loss": 0.7038, - "step": 234 - }, - { - "epoch": 0.07, - "learning_rate": 9.915562997684278e-07, - "loss": 0.7015, - "step": 235 - }, - { - "epoch": 0.07, - "learning_rate": 9.914650311015546e-07, - "loss": 0.7322, - "step": 236 - }, - { - "epoch": 0.07, - "learning_rate": 9.913732760632242e-07, - "loss": 0.7055, - "step": 237 - }, - { - "epoch": 0.07, - "learning_rate": 9.912810347442406e-07, - "loss": 0.6942, - "step": 238 - }, - { - "epoch": 0.07, - "learning_rate": 9.911883072358893e-07, - "loss": 0.6995, - "step": 239 - }, - { - "epoch": 0.07, - "learning_rate": 9.910950936299365e-07, - "loss": 0.6911, - "step": 240 - }, - { - "epoch": 0.08, - "learning_rate": 9.910013940186297e-07, - "loss": 0.7117, - "step": 241 - }, - { - "epoch": 0.08, - "learning_rate": 9.909072084946977e-07, - "loss": 0.7138, - "step": 242 - }, - { - "epoch": 0.08, - "learning_rate": 9.9081253715135e-07, - "loss": 0.7006, - "step": 243 - }, - { - "epoch": 0.08, - "learning_rate": 9.907173800822762e-07, - "loss": 0.7184, - "step": 244 - }, - { - "epoch": 0.08, - "learning_rate": 9.906217373816475e-07, - "loss": 0.6881, - "step": 245 - }, - { - "epoch": 0.08, - "learning_rate": 9.905256091441152e-07, - "loss": 0.6736, - "step": 246 - }, - { - "epoch": 0.08, - "learning_rate": 9.904289954648116e-07, - "loss": 0.6994, - "step": 247 - }, - { - "epoch": 0.08, - "learning_rate": 9.903318964393485e-07, - "loss": 0.6798, - "step": 248 - }, - { - "epoch": 0.08, - "learning_rate": 9.902343121638192e-07, - "loss": 0.696, - "step": 249 - }, - { - "epoch": 0.08, - "learning_rate": 9.901362427347964e-07, - "loss": 0.667, - "step": 250 - }, - { - "epoch": 0.08, - "learning_rate": 9.900376882493328e-07, - "loss": 0.6954, - "step": 251 - }, - { - "epoch": 0.08, - "learning_rate": 9.89938648804962e-07, - "loss": 0.6929, - "step": 252 - }, - { - "epoch": 0.08, - "learning_rate": 9.898391244996964e-07, - "loss": 0.6865, - "step": 253 - }, - { - "epoch": 0.08, - "learning_rate": 9.897391154320293e-07, - "loss": 0.6724, - "step": 254 - }, - { - "epoch": 0.08, - "learning_rate": 9.896386217009331e-07, - "loss": 0.6871, - "step": 255 - }, - { - "epoch": 0.08, - "learning_rate": 9.8953764340586e-07, - "loss": 0.6882, - "step": 256 - }, - { - "epoch": 0.08, - "learning_rate": 9.894361806467418e-07, - "loss": 0.6974, - "step": 257 - }, - { - "epoch": 0.08, - "learning_rate": 9.893342335239896e-07, - "loss": 0.667, - "step": 258 - }, - { - "epoch": 0.08, - "learning_rate": 9.89231802138494e-07, - "loss": 0.708, - "step": 259 - }, - { - "epoch": 0.08, - "learning_rate": 9.89128886591625e-07, - "loss": 0.697, - "step": 260 - }, - { - "epoch": 0.08, - "learning_rate": 9.890254869852311e-07, - "loss": 0.7021, - "step": 261 - }, - { - "epoch": 0.08, - "learning_rate": 9.889216034216404e-07, - "loss": 0.6814, - "step": 262 - }, - { - "epoch": 0.08, - "learning_rate": 9.8881723600366e-07, - "loss": 0.6736, - "step": 263 - }, - { - "epoch": 0.08, - "learning_rate": 9.887123848345756e-07, - "loss": 0.6734, - "step": 264 - }, - { - "epoch": 0.08, - "learning_rate": 9.886070500181514e-07, - "loss": 0.6539, - "step": 265 - }, - { - "epoch": 0.08, - "learning_rate": 9.885012316586309e-07, - "loss": 0.6803, - "step": 266 - }, - { - "epoch": 0.08, - "learning_rate": 9.883949298607355e-07, - "loss": 0.6974, - "step": 267 - }, - { - "epoch": 0.08, - "learning_rate": 9.882881447296654e-07, - "loss": 0.6652, - "step": 268 - }, - { - "epoch": 0.08, - "learning_rate": 9.88180876371099e-07, - "loss": 0.6677, - "step": 269 - }, - { - "epoch": 0.08, - "learning_rate": 9.880731248911926e-07, - "loss": 0.7055, - "step": 270 - }, - { - "epoch": 0.08, - "learning_rate": 9.879648903965815e-07, - "loss": 0.6952, - "step": 271 - }, - { - "epoch": 0.08, - "learning_rate": 9.87856172994378e-07, - "loss": 0.6793, - "step": 272 - }, - { - "epoch": 0.09, - "learning_rate": 9.877469727921732e-07, - "loss": 0.6784, - "step": 273 - }, - { - "epoch": 0.09, - "learning_rate": 9.87637289898035e-07, - "loss": 0.6919, - "step": 274 - }, - { - "epoch": 0.09, - "learning_rate": 9.8752712442051e-07, - "loss": 0.6703, - "step": 275 - }, - { - "epoch": 0.09, - "learning_rate": 9.874164764686215e-07, - "loss": 0.6716, - "step": 276 - }, - { - "epoch": 0.09, - "learning_rate": 9.87305346151871e-07, - "loss": 0.6695, - "step": 277 - }, - { - "epoch": 0.09, - "learning_rate": 9.87193733580237e-07, - "loss": 0.6709, - "step": 278 - }, - { - "epoch": 0.09, - "learning_rate": 9.870816388641751e-07, - "loss": 0.649, - "step": 279 - }, - { - "epoch": 0.09, - "learning_rate": 9.869690621146184e-07, - "loss": 0.673, - "step": 280 - }, - { - "epoch": 0.09, - "learning_rate": 9.86856003442977e-07, - "loss": 0.6812, - "step": 281 - }, - { - "epoch": 0.09, - "learning_rate": 9.86742462961138e-07, - "loss": 0.6747, - "step": 282 - }, - { - "epoch": 0.09, - "learning_rate": 9.866284407814646e-07, - "loss": 0.6587, - "step": 283 - }, - { - "epoch": 0.09, - "learning_rate": 9.865139370167976e-07, - "loss": 0.6712, - "step": 284 - }, - { - "epoch": 0.09, - "learning_rate": 9.863989517804541e-07, - "loss": 0.6451, - "step": 285 - }, - { - "epoch": 0.09, - "learning_rate": 9.862834851862273e-07, - "loss": 0.6881, - "step": 286 - }, - { - "epoch": 0.09, - "learning_rate": 9.861675373483875e-07, - "loss": 0.6748, - "step": 287 - }, - { - "epoch": 0.09, - "learning_rate": 9.860511083816805e-07, - "loss": 0.6579, - "step": 288 - }, - { - "epoch": 0.09, - "learning_rate": 9.859341984013288e-07, - "loss": 0.6812, - "step": 289 - }, - { - "epoch": 0.09, - "learning_rate": 9.858168075230306e-07, - "loss": 0.6442, - "step": 290 - }, - { - "epoch": 0.09, - "learning_rate": 9.856989358629601e-07, - "loss": 0.6521, - "step": 291 - }, - { - "epoch": 0.09, - "learning_rate": 9.855805835377676e-07, - "loss": 0.7038, - "step": 292 - }, - { - "epoch": 0.09, - "learning_rate": 9.854617506645785e-07, - "loss": 0.6513, - "step": 293 - }, - { - "epoch": 0.09, - "learning_rate": 9.853424373609942e-07, - "loss": 0.6696, - "step": 294 - }, - { - "epoch": 0.09, - "learning_rate": 9.852226437450915e-07, - "loss": 0.673, - "step": 295 - }, - { - "epoch": 0.09, - "learning_rate": 9.851023699354224e-07, - "loss": 0.6557, - "step": 296 - }, - { - "epoch": 0.09, - "learning_rate": 9.849816160510142e-07, - "loss": 0.6504, - "step": 297 - }, - { - "epoch": 0.09, - "learning_rate": 9.848603822113692e-07, - "loss": 0.6992, - "step": 298 - }, - { - "epoch": 0.09, - "learning_rate": 9.847386685364649e-07, - "loss": 0.6706, - "step": 299 - }, - { - "epoch": 0.09, - "learning_rate": 9.846164751467533e-07, - "loss": 0.6586, - "step": 300 - }, - { - "epoch": 0.09, - "learning_rate": 9.844938021631619e-07, - "loss": 0.6475, - "step": 301 - }, - { - "epoch": 0.09, - "learning_rate": 9.843706497070917e-07, - "loss": 0.6495, - "step": 302 - }, - { - "epoch": 0.09, - "learning_rate": 9.84247017900419e-07, - "loss": 0.6313, - "step": 303 - }, - { - "epoch": 0.09, - "learning_rate": 9.841229068654944e-07, - "loss": 0.6478, - "step": 304 - }, - { - "epoch": 0.1, - "learning_rate": 9.839983167251424e-07, - "loss": 0.6583, - "step": 305 - }, - { - "epoch": 0.1, - "learning_rate": 9.83873247602662e-07, - "loss": 0.6366, - "step": 306 - }, - { - "epoch": 0.1, - "learning_rate": 9.837476996218263e-07, - "loss": 0.6389, - "step": 307 - }, - { - "epoch": 0.1, - "learning_rate": 9.836216729068818e-07, - "loss": 0.6731, - "step": 308 - }, - { - "epoch": 0.1, - "learning_rate": 9.834951675825492e-07, - "loss": 0.6422, - "step": 309 - }, - { - "epoch": 0.1, - "learning_rate": 9.833681837740228e-07, - "loss": 0.6368, - "step": 310 - }, - { - "epoch": 0.1, - "learning_rate": 9.832407216069702e-07, - "loss": 0.6419, - "step": 311 - }, - { - "epoch": 0.1, - "learning_rate": 9.831127812075327e-07, - "loss": 0.6469, - "step": 312 - }, - { - "epoch": 0.1, - "learning_rate": 9.829843627023241e-07, - "loss": 0.6562, - "step": 313 - }, - { - "epoch": 0.1, - "learning_rate": 9.82855466218433e-07, - "loss": 0.6317, - "step": 314 - }, - { - "epoch": 0.1, - "learning_rate": 9.827260918834192e-07, - "loss": 0.6351, - "step": 315 - }, - { - "epoch": 0.1, - "learning_rate": 9.825962398253167e-07, - "loss": 0.6432, - "step": 316 - }, - { - "epoch": 0.1, - "learning_rate": 9.824659101726312e-07, - "loss": 0.6515, - "step": 317 - }, - { - "epoch": 0.1, - "learning_rate": 9.82335103054342e-07, - "loss": 0.6497, - "step": 318 - }, - { - "epoch": 0.1, - "learning_rate": 9.822038185999007e-07, - "loss": 0.6674, - "step": 319 - }, - { - "epoch": 0.1, - "learning_rate": 9.820720569392307e-07, - "loss": 0.6435, - "step": 320 - }, - { - "epoch": 0.1, - "learning_rate": 9.819398182027281e-07, - "loss": 0.6196, - "step": 321 - }, - { - "epoch": 0.1, - "learning_rate": 9.818071025212613e-07, - "loss": 0.6698, - "step": 322 - }, - { - "epoch": 0.1, - "learning_rate": 9.816739100261706e-07, - "loss": 0.6348, - "step": 323 - }, - { - "epoch": 0.1, - "learning_rate": 9.815402408492677e-07, - "loss": 0.6443, - "step": 324 - }, - { - "epoch": 0.1, - "learning_rate": 9.81406095122837e-07, - "loss": 0.6325, - "step": 325 - }, - { - "epoch": 0.1, - "learning_rate": 9.812714729796333e-07, - "loss": 0.6311, - "step": 326 - }, - { - "epoch": 0.1, - "learning_rate": 9.811363745528839e-07, - "loss": 0.6373, - "step": 327 - }, - { - "epoch": 0.1, - "learning_rate": 9.810007999762868e-07, - "loss": 0.6372, - "step": 328 - }, - { - "epoch": 0.1, - "learning_rate": 9.808647493840118e-07, - "loss": 0.6143, - "step": 329 - }, - { - "epoch": 0.1, - "learning_rate": 9.807282229106992e-07, - "loss": 0.6451, - "step": 330 - }, - { - "epoch": 0.1, - "learning_rate": 9.805912206914606e-07, - "loss": 0.6347, - "step": 331 - }, - { - "epoch": 0.1, - "learning_rate": 9.804537428618781e-07, - "loss": 0.6349, - "step": 332 - }, - { - "epoch": 0.1, - "learning_rate": 9.803157895580054e-07, - "loss": 0.6222, - "step": 333 - }, - { - "epoch": 0.1, - "learning_rate": 9.801773609163653e-07, - "loss": 0.6269, - "step": 334 - }, - { - "epoch": 0.1, - "learning_rate": 9.80038457073952e-07, - "loss": 0.6324, - "step": 335 - }, - { - "epoch": 0.1, - "learning_rate": 9.798990781682296e-07, - "loss": 0.6164, - "step": 336 - }, - { - "epoch": 0.11, - "learning_rate": 9.797592243371327e-07, - "loss": 0.6379, - "step": 337 - }, - { - "epoch": 0.11, - "learning_rate": 9.796188957190657e-07, - "loss": 0.6414, - "step": 338 - }, - { - "epoch": 0.11, - "learning_rate": 9.794780924529026e-07, - "loss": 0.6698, - "step": 339 - }, - { - "epoch": 0.11, - "learning_rate": 9.793368146779876e-07, - "loss": 0.6051, - "step": 340 - }, - { - "epoch": 0.11, - "learning_rate": 9.791950625341343e-07, - "loss": 0.6594, - "step": 341 - }, - { - "epoch": 0.11, - "learning_rate": 9.790528361616254e-07, - "loss": 0.6119, - "step": 342 - }, - { - "epoch": 0.11, - "learning_rate": 9.789101357012136e-07, - "loss": 0.603, - "step": 343 - }, - { - "epoch": 0.11, - "learning_rate": 9.7876696129412e-07, - "loss": 0.6761, - "step": 344 - }, - { - "epoch": 0.11, - "learning_rate": 9.786233130820356e-07, - "loss": 0.6283, - "step": 345 - }, - { - "epoch": 0.11, - "learning_rate": 9.784791912071195e-07, - "loss": 0.6319, - "step": 346 - }, - { - "epoch": 0.11, - "learning_rate": 9.783345958120001e-07, - "loss": 0.6487, - "step": 347 - }, - { - "epoch": 0.11, - "learning_rate": 9.78189527039774e-07, - "loss": 0.6536, - "step": 348 - }, - { - "epoch": 0.11, - "learning_rate": 9.780439850340065e-07, - "loss": 0.6325, - "step": 349 - }, - { - "epoch": 0.11, - "learning_rate": 9.778979699387314e-07, - "loss": 0.6123, - "step": 350 - }, - { - "epoch": 0.11, - "learning_rate": 9.7775148189845e-07, - "loss": 0.6162, - "step": 351 - }, - { - "epoch": 0.11, - "learning_rate": 9.776045210581327e-07, - "loss": 0.6291, - "step": 352 - }, - { - "epoch": 0.11, - "learning_rate": 9.77457087563217e-07, - "loss": 0.6227, - "step": 353 - }, - { - "epoch": 0.11, - "learning_rate": 9.773091815596085e-07, - "loss": 0.6324, - "step": 354 - }, - { - "epoch": 0.11, - "learning_rate": 9.7716080319368e-07, - "loss": 0.6141, - "step": 355 - }, - { - "epoch": 0.11, - "learning_rate": 9.770119526122723e-07, - "loss": 0.6294, - "step": 356 - }, - { - "epoch": 0.11, - "learning_rate": 9.768626299626934e-07, - "loss": 0.6462, - "step": 357 - }, - { - "epoch": 0.11, - "learning_rate": 9.76712835392718e-07, - "loss": 0.6133, - "step": 358 - }, - { - "epoch": 0.11, - "learning_rate": 9.765625690505884e-07, - "loss": 0.604, - "step": 359 - }, - { - "epoch": 0.11, - "learning_rate": 9.764118310850136e-07, - "loss": 0.6287, - "step": 360 - }, - { - "epoch": 0.11, - "learning_rate": 9.762606216451692e-07, - "loss": 0.6254, - "step": 361 - }, - { - "epoch": 0.11, - "learning_rate": 9.761089408806975e-07, - "loss": 0.6424, - "step": 362 - }, - { - "epoch": 0.11, - "learning_rate": 9.759567889417075e-07, - "loss": 0.628, - "step": 363 - }, - { - "epoch": 0.11, - "learning_rate": 9.75804165978774e-07, - "loss": 0.6284, - "step": 364 - }, - { - "epoch": 0.11, - "learning_rate": 9.756510721429379e-07, - "loss": 0.6075, - "step": 365 - }, - { - "epoch": 0.11, - "learning_rate": 9.75497507585707e-07, - "loss": 0.6254, - "step": 366 - }, - { - "epoch": 0.11, - "learning_rate": 9.753434724590537e-07, - "loss": 0.6077, - "step": 367 - }, - { - "epoch": 0.11, - "learning_rate": 9.75188966915417e-07, - "loss": 0.6366, - "step": 368 - }, - { - "epoch": 0.12, - "learning_rate": 9.750339911077012e-07, - "loss": 0.6194, - "step": 369 - }, - { - "epoch": 0.12, - "learning_rate": 9.748785451892758e-07, - "loss": 0.6336, - "step": 370 - }, - { - "epoch": 0.12, - "learning_rate": 9.747226293139756e-07, - "loss": 0.6296, - "step": 371 - }, - { - "epoch": 0.12, - "learning_rate": 9.745662436361008e-07, - "loss": 0.6274, - "step": 372 - }, - { - "epoch": 0.12, - "learning_rate": 9.744093883104159e-07, - "loss": 0.6151, - "step": 373 - }, - { - "epoch": 0.12, - "learning_rate": 9.74252063492151e-07, - "loss": 0.6132, - "step": 374 - }, - { - "epoch": 0.12, - "learning_rate": 9.74094269337e-07, - "loss": 0.5981, - "step": 375 - }, - { - "epoch": 0.12, - "learning_rate": 9.73936006001122e-07, - "loss": 0.6046, - "step": 376 - }, - { - "epoch": 0.12, - "learning_rate": 9.7377727364114e-07, - "loss": 0.6109, - "step": 377 - }, - { - "epoch": 0.12, - "learning_rate": 9.73618072414141e-07, - "loss": 0.6307, - "step": 378 - }, - { - "epoch": 0.12, - "learning_rate": 9.734584024776766e-07, - "loss": 0.6131, - "step": 379 - }, - { - "epoch": 0.12, - "learning_rate": 9.732982639897616e-07, - "loss": 0.612, - "step": 380 - }, - { - "epoch": 0.12, - "learning_rate": 9.731376571088752e-07, - "loss": 0.6184, - "step": 381 - }, - { - "epoch": 0.12, - "learning_rate": 9.729765819939592e-07, - "loss": 0.6096, - "step": 382 - }, - { - "epoch": 0.12, - "learning_rate": 9.7281503880442e-07, - "loss": 0.5891, - "step": 383 - }, - { - "epoch": 0.12, - "learning_rate": 9.726530277001257e-07, - "loss": 0.6192, - "step": 384 - }, - { - "epoch": 0.12, - "learning_rate": 9.72490548841409e-07, - "loss": 0.6304, - "step": 385 - }, - { - "epoch": 0.12, - "learning_rate": 9.723276023890648e-07, - "loss": 0.6151, - "step": 386 - }, - { - "epoch": 0.12, - "learning_rate": 9.721641885043502e-07, - "loss": 0.6127, - "step": 387 - }, - { - "epoch": 0.12, - "learning_rate": 9.72000307348986e-07, - "loss": 0.626, - "step": 388 - }, - { - "epoch": 0.12, - "learning_rate": 9.718359590851546e-07, - "loss": 0.6321, - "step": 389 - }, - { - "epoch": 0.12, - "learning_rate": 9.716711438755013e-07, - "loss": 0.613, - "step": 390 - }, - { - "epoch": 0.12, - "learning_rate": 9.715058618831326e-07, - "loss": 0.6157, - "step": 391 - }, - { - "epoch": 0.12, - "learning_rate": 9.71340113271618e-07, - "loss": 0.6185, - "step": 392 - }, - { - "epoch": 0.12, - "learning_rate": 9.71173898204988e-07, - "loss": 0.6326, - "step": 393 - }, - { - "epoch": 0.12, - "learning_rate": 9.71007216847735e-07, - "loss": 0.6149, - "step": 394 - }, - { - "epoch": 0.12, - "learning_rate": 9.70840069364813e-07, - "loss": 0.6103, - "step": 395 - }, - { - "epoch": 0.12, - "learning_rate": 9.70672455921637e-07, - "loss": 0.6041, - "step": 396 - }, - { - "epoch": 0.12, - "learning_rate": 9.705043766840836e-07, - "loss": 0.617, - "step": 397 - }, - { - "epoch": 0.12, - "learning_rate": 9.703358318184897e-07, - "loss": 0.6095, - "step": 398 - }, - { - "epoch": 0.12, - "learning_rate": 9.701668214916536e-07, - "loss": 0.5898, - "step": 399 - }, - { - "epoch": 0.12, - "learning_rate": 9.699973458708338e-07, - "loss": 0.6208, - "step": 400 - }, - { - "epoch": 0.12, - "learning_rate": 9.698274051237494e-07, - "loss": 0.6086, - "step": 401 - }, - { - "epoch": 0.13, - "learning_rate": 9.696569994185804e-07, - "loss": 0.5914, - "step": 402 - }, - { - "epoch": 0.13, - "learning_rate": 9.694861289239658e-07, - "loss": 0.5913, - "step": 403 - }, - { - "epoch": 0.13, - "learning_rate": 9.693147938090054e-07, - "loss": 0.6092, - "step": 404 - }, - { - "epoch": 0.13, - "learning_rate": 9.691429942432588e-07, - "loss": 0.6458, - "step": 405 - }, - { - "epoch": 0.13, - "learning_rate": 9.689707303967446e-07, - "loss": 0.6043, - "step": 406 - }, - { - "epoch": 0.13, - "learning_rate": 9.687980024399416e-07, - "loss": 0.5955, - "step": 407 - }, - { - "epoch": 0.13, - "learning_rate": 9.686248105437875e-07, - "loss": 0.6159, - "step": 408 - }, - { - "epoch": 0.13, - "learning_rate": 9.684511548796795e-07, - "loss": 0.6031, - "step": 409 - }, - { - "epoch": 0.13, - "learning_rate": 9.68277035619473e-07, - "loss": 0.6089, - "step": 410 - }, - { - "epoch": 0.13, - "learning_rate": 9.681024529354828e-07, - "loss": 0.6065, - "step": 411 - }, - { - "epoch": 0.13, - "learning_rate": 9.679274070004825e-07, - "loss": 0.5887, - "step": 412 - }, - { - "epoch": 0.13, - "learning_rate": 9.677518979877035e-07, - "loss": 0.593, - "step": 413 - }, - { - "epoch": 0.13, - "learning_rate": 9.67575926070836e-07, - "loss": 0.5868, - "step": 414 - }, - { - "epoch": 0.13, - "learning_rate": 9.673994914240283e-07, - "loss": 0.5977, - "step": 415 - }, - { - "epoch": 0.13, - "learning_rate": 9.67222594221886e-07, - "loss": 0.6254, - "step": 416 - }, - { - "epoch": 0.13, - "learning_rate": 9.670452346394734e-07, - "loss": 0.6025, - "step": 417 - }, - { - "epoch": 0.13, - "learning_rate": 9.668674128523116e-07, - "loss": 0.614, - "step": 418 - }, - { - "epoch": 0.13, - "learning_rate": 9.666891290363797e-07, - "loss": 0.6083, - "step": 419 - }, - { - "epoch": 0.13, - "learning_rate": 9.665103833681136e-07, - "loss": 0.6197, - "step": 420 - }, - { - "epoch": 0.13, - "learning_rate": 9.663311760244064e-07, - "loss": 0.6342, - "step": 421 - }, - { - "epoch": 0.13, - "learning_rate": 9.661515071826081e-07, - "loss": 0.617, - "step": 422 - }, - { - "epoch": 0.13, - "learning_rate": 9.659713770205257e-07, - "loss": 0.5945, - "step": 423 - }, - { - "epoch": 0.13, - "learning_rate": 9.657907857164222e-07, - "loss": 0.6032, - "step": 424 - }, - { - "epoch": 0.13, - "learning_rate": 9.656097334490176e-07, - "loss": 0.585, - "step": 425 - }, - { - "epoch": 0.13, - "learning_rate": 9.654282203974873e-07, - "loss": 0.568, - "step": 426 - }, - { - "epoch": 0.13, - "learning_rate": 9.652462467414633e-07, - "loss": 0.6176, - "step": 427 - }, - { - "epoch": 0.13, - "learning_rate": 9.650638126610333e-07, - "loss": 0.5884, - "step": 428 - }, - { - "epoch": 0.13, - "learning_rate": 9.648809183367408e-07, - "loss": 0.5774, - "step": 429 - }, - { - "epoch": 0.13, - "learning_rate": 9.646975639495843e-07, - "loss": 0.6061, - "step": 430 - }, - { - "epoch": 0.13, - "learning_rate": 9.64513749681018e-07, - "loss": 0.5977, - "step": 431 - }, - { - "epoch": 0.13, - "learning_rate": 9.643294757129511e-07, - "loss": 0.5989, - "step": 432 - }, - { - "epoch": 0.13, - "learning_rate": 9.641447422277478e-07, - "loss": 0.5705, - "step": 433 - }, - { - "epoch": 0.14, - "learning_rate": 9.639595494082268e-07, - "loss": 0.5911, - "step": 434 - }, - { - "epoch": 0.14, - "learning_rate": 9.63773897437662e-07, - "loss": 0.601, - "step": 435 - }, - { - "epoch": 0.14, - "learning_rate": 9.635877864997807e-07, - "loss": 0.6196, - "step": 436 - }, - { - "epoch": 0.14, - "learning_rate": 9.634012167787656e-07, - "loss": 0.5891, - "step": 437 - }, - { - "epoch": 0.14, - "learning_rate": 9.632141884592521e-07, - "loss": 0.5981, - "step": 438 - }, - { - "epoch": 0.14, - "learning_rate": 9.630267017263307e-07, - "loss": 0.5917, - "step": 439 - }, - { - "epoch": 0.14, - "learning_rate": 9.628387567655449e-07, - "loss": 0.6013, - "step": 440 - }, - { - "epoch": 0.14, - "learning_rate": 9.626503537628916e-07, - "loss": 0.5938, - "step": 441 - }, - { - "epoch": 0.14, - "learning_rate": 9.624614929048213e-07, - "loss": 0.5814, - "step": 442 - }, - { - "epoch": 0.14, - "learning_rate": 9.622721743782374e-07, - "loss": 0.5946, - "step": 443 - }, - { - "epoch": 0.14, - "learning_rate": 9.620823983704964e-07, - "loss": 0.5919, - "step": 444 - }, - { - "epoch": 0.14, - "learning_rate": 9.618921650694077e-07, - "loss": 0.6103, - "step": 445 - }, - { - "epoch": 0.14, - "learning_rate": 9.617014746632324e-07, - "loss": 0.5999, - "step": 446 - }, - { - "epoch": 0.14, - "learning_rate": 9.615103273406852e-07, - "loss": 0.5893, - "step": 447 - }, - { - "epoch": 0.14, - "learning_rate": 9.61318723290932e-07, - "loss": 0.6015, - "step": 448 - }, - { - "epoch": 0.14, - "learning_rate": 9.611266627035912e-07, - "loss": 0.6025, - "step": 449 - }, - { - "epoch": 0.14, - "learning_rate": 9.609341457687329e-07, - "loss": 0.5893, - "step": 450 - }, - { - "epoch": 0.14, - "learning_rate": 9.607411726768785e-07, - "loss": 0.5857, - "step": 451 - }, - { - "epoch": 0.14, - "learning_rate": 9.605477436190013e-07, - "loss": 0.5778, - "step": 452 - }, - { - "epoch": 0.14, - "learning_rate": 9.603538587865258e-07, - "loss": 0.6025, - "step": 453 - }, - { - "epoch": 0.14, - "learning_rate": 9.601595183713272e-07, - "loss": 0.6109, - "step": 454 - }, - { - "epoch": 0.14, - "learning_rate": 9.599647225657317e-07, - "loss": 0.5852, - "step": 455 - }, - { - "epoch": 0.14, - "learning_rate": 9.597694715625165e-07, - "loss": 0.5523, - "step": 456 - }, - { - "epoch": 0.14, - "learning_rate": 9.595737655549088e-07, - "loss": 0.5873, - "step": 457 - }, - { - "epoch": 0.14, - "learning_rate": 9.593776047365861e-07, - "loss": 0.5773, - "step": 458 - }, - { - "epoch": 0.14, - "learning_rate": 9.591809893016767e-07, - "loss": 0.6038, - "step": 459 - }, - { - "epoch": 0.14, - "learning_rate": 9.58983919444758e-07, - "loss": 0.5861, - "step": 460 - }, - { - "epoch": 0.14, - "learning_rate": 9.587863953608576e-07, - "loss": 0.587, - "step": 461 - }, - { - "epoch": 0.14, - "learning_rate": 9.58588417245452e-07, - "loss": 0.5953, - "step": 462 - }, - { - "epoch": 0.14, - "learning_rate": 9.583899852944678e-07, - "loss": 0.5961, - "step": 463 - }, - { - "epoch": 0.14, - "learning_rate": 9.581910997042806e-07, - "loss": 0.6009, - "step": 464 - }, - { - "epoch": 0.14, - "learning_rate": 9.579917606717142e-07, - "loss": 0.5875, - "step": 465 - }, - { - "epoch": 0.15, - "learning_rate": 9.577919683940419e-07, - "loss": 0.5613, - "step": 466 - }, - { - "epoch": 0.15, - "learning_rate": 9.575917230689854e-07, - "loss": 0.6143, - "step": 467 - }, - { - "epoch": 0.15, - "learning_rate": 9.573910248947147e-07, - "loss": 0.5956, - "step": 468 - }, - { - "epoch": 0.15, - "learning_rate": 9.571898740698478e-07, - "loss": 0.5885, - "step": 469 - }, - { - "epoch": 0.15, - "learning_rate": 9.569882707934508e-07, - "loss": 0.5844, - "step": 470 - }, - { - "epoch": 0.15, - "learning_rate": 9.567862152650376e-07, - "loss": 0.5862, - "step": 471 - }, - { - "epoch": 0.15, - "learning_rate": 9.565837076845697e-07, - "loss": 0.5931, - "step": 472 - }, - { - "epoch": 0.15, - "learning_rate": 9.563807482524557e-07, - "loss": 0.5758, - "step": 473 - }, - { - "epoch": 0.15, - "learning_rate": 9.561773371695515e-07, - "loss": 0.586, - "step": 474 - }, - { - "epoch": 0.15, - "learning_rate": 9.559734746371604e-07, - "loss": 0.5608, - "step": 475 - }, - { - "epoch": 0.15, - "learning_rate": 9.557691608570315e-07, - "loss": 0.5748, - "step": 476 - }, - { - "epoch": 0.15, - "learning_rate": 9.555643960313616e-07, - "loss": 0.5922, - "step": 477 - }, - { - "epoch": 0.15, - "learning_rate": 9.55359180362793e-07, - "loss": 0.599, - "step": 478 - }, - { - "epoch": 0.15, - "learning_rate": 9.551535140544148e-07, - "loss": 0.5955, - "step": 479 - }, - { - "epoch": 0.15, - "learning_rate": 9.549473973097615e-07, - "loss": 0.581, - "step": 480 - }, - { - "epoch": 0.15, - "learning_rate": 9.547408303328135e-07, - "loss": 0.5991, - "step": 481 - }, - { - "epoch": 0.15, - "learning_rate": 9.545338133279975e-07, - "loss": 0.5871, - "step": 482 - }, - { - "epoch": 0.15, - "learning_rate": 9.543263465001844e-07, - "loss": 0.5829, - "step": 483 - }, - { - "epoch": 0.15, - "learning_rate": 9.541184300546908e-07, - "loss": 0.5793, - "step": 484 - }, - { - "epoch": 0.15, - "learning_rate": 9.539100641972788e-07, - "loss": 0.5982, - "step": 485 - }, - { - "epoch": 0.15, - "learning_rate": 9.537012491341544e-07, - "loss": 0.5539, - "step": 486 - }, - { - "epoch": 0.15, - "learning_rate": 9.534919850719686e-07, - "loss": 0.6131, - "step": 487 - }, - { - "epoch": 0.15, - "learning_rate": 9.532822722178164e-07, - "loss": 0.5933, - "step": 488 - }, - { - "epoch": 0.15, - "learning_rate": 9.530721107792376e-07, - "loss": 0.5648, - "step": 489 - }, - { - "epoch": 0.15, - "learning_rate": 9.528615009642151e-07, - "loss": 0.5934, - "step": 490 - }, - { - "epoch": 0.15, - "learning_rate": 9.526504429811762e-07, - "loss": 0.5971, - "step": 491 - }, - { - "epoch": 0.15, - "learning_rate": 9.524389370389912e-07, - "loss": 0.5792, - "step": 492 - }, - { - "epoch": 0.15, - "learning_rate": 9.522269833469742e-07, - "loss": 0.5941, - "step": 493 - }, - { - "epoch": 0.15, - "learning_rate": 9.520145821148823e-07, - "loss": 0.5788, - "step": 494 - }, - { - "epoch": 0.15, - "learning_rate": 9.518017335529151e-07, - "loss": 0.5791, - "step": 495 - }, - { - "epoch": 0.15, - "learning_rate": 9.515884378717156e-07, - "loss": 0.5948, - "step": 496 - }, - { - "epoch": 0.15, - "learning_rate": 9.513746952823684e-07, - "loss": 0.5534, - "step": 497 - }, - { - "epoch": 0.16, - "learning_rate": 9.511605059964009e-07, - "loss": 0.5899, - "step": 498 - }, - { - "epoch": 0.16, - "learning_rate": 9.509458702257829e-07, - "loss": 0.587, - "step": 499 - }, - { - "epoch": 0.16, - "learning_rate": 9.507307881829253e-07, - "loss": 0.6006, - "step": 500 - }, - { - "epoch": 0.16, - "learning_rate": 9.505152600806812e-07, - "loss": 0.5738, - "step": 501 - }, - { - "epoch": 0.16, - "learning_rate": 9.502992861323449e-07, - "loss": 0.5762, - "step": 502 - }, - { - "epoch": 0.16, - "learning_rate": 9.50082866551652e-07, - "loss": 0.5353, - "step": 503 - }, - { - "epoch": 0.16, - "learning_rate": 9.498660015527791e-07, - "loss": 0.5732, - "step": 504 - }, - { - "epoch": 0.16, - "learning_rate": 9.496486913503435e-07, - "loss": 0.5719, - "step": 505 - }, - { - "epoch": 0.16, - "learning_rate": 9.494309361594034e-07, - "loss": 0.5704, - "step": 506 - }, - { - "epoch": 0.16, - "learning_rate": 9.49212736195457e-07, - "loss": 0.5823, - "step": 507 - }, - { - "epoch": 0.16, - "learning_rate": 9.489940916744428e-07, - "loss": 0.5711, - "step": 508 - }, - { - "epoch": 0.16, - "learning_rate": 9.487750028127392e-07, - "loss": 0.5963, - "step": 509 - }, - { - "epoch": 0.16, - "learning_rate": 9.485554698271649e-07, - "loss": 0.5723, - "step": 510 - }, - { - "epoch": 0.16, - "learning_rate": 9.48335492934977e-07, - "loss": 0.564, - "step": 511 - }, - { - "epoch": 0.16, - "learning_rate": 9.481150723538729e-07, - "loss": 0.6019, - "step": 512 - }, - { - "epoch": 0.16, - "learning_rate": 9.478942083019885e-07, - "loss": 0.5824, - "step": 513 - }, - { - "epoch": 0.16, - "learning_rate": 9.476729009978993e-07, - "loss": 0.5705, - "step": 514 - }, - { - "epoch": 0.16, - "learning_rate": 9.474511506606182e-07, - "loss": 0.5576, - "step": 515 - }, - { - "epoch": 0.16, - "learning_rate": 9.47228957509598e-07, - "loss": 0.5621, - "step": 516 - }, - { - "epoch": 0.16, - "learning_rate": 9.470063217647286e-07, - "loss": 0.582, - "step": 517 - }, - { - "epoch": 0.16, - "learning_rate": 9.467832436463385e-07, - "loss": 0.5873, - "step": 518 - }, - { - "epoch": 0.16, - "learning_rate": 9.465597233751939e-07, - "loss": 0.5537, - "step": 519 - }, - { - "epoch": 0.16, - "learning_rate": 9.463357611724982e-07, - "loss": 0.569, - "step": 520 - }, - { - "epoch": 0.16, - "learning_rate": 9.461113572598926e-07, - "loss": 0.5698, - "step": 521 - }, - { - "epoch": 0.16, - "learning_rate": 9.458865118594554e-07, - "loss": 0.5827, - "step": 522 - }, - { - "epoch": 0.16, - "learning_rate": 9.456612251937018e-07, - "loss": 0.5823, - "step": 523 - }, - { - "epoch": 0.16, - "learning_rate": 9.454354974855831e-07, - "loss": 0.5409, - "step": 524 - }, - { - "epoch": 0.16, - "learning_rate": 9.45209328958488e-07, - "loss": 0.5843, - "step": 525 - }, - { - "epoch": 0.16, - "learning_rate": 9.449827198362407e-07, - "loss": 0.5923, - "step": 526 - }, - { - "epoch": 0.16, - "learning_rate": 9.447556703431021e-07, - "loss": 0.5469, - "step": 527 - }, - { - "epoch": 0.16, - "learning_rate": 9.445281807037682e-07, - "loss": 0.5561, - "step": 528 - }, - { - "epoch": 0.16, - "learning_rate": 9.443002511433709e-07, - "loss": 0.5667, - "step": 529 - }, - { - "epoch": 0.17, - "learning_rate": 9.440718818874777e-07, - "loss": 0.5595, - "step": 530 - }, - { - "epoch": 0.17, - "learning_rate": 9.438430731620909e-07, - "loss": 0.5747, - "step": 531 - }, - { - "epoch": 0.17, - "learning_rate": 9.436138251936478e-07, - "loss": 0.5861, - "step": 532 - }, - { - "epoch": 0.17, - "learning_rate": 9.433841382090206e-07, - "loss": 0.5645, - "step": 533 - }, - { - "epoch": 0.17, - "learning_rate": 9.431540124355157e-07, - "loss": 0.5833, - "step": 534 - }, - { - "epoch": 0.17, - "learning_rate": 9.429234481008738e-07, - "loss": 0.5903, - "step": 535 - }, - { - "epoch": 0.17, - "learning_rate": 9.426924454332699e-07, - "loss": 0.5868, - "step": 536 - }, - { - "epoch": 0.17, - "learning_rate": 9.424610046613119e-07, - "loss": 0.5542, - "step": 537 - }, - { - "epoch": 0.17, - "learning_rate": 9.422291260140427e-07, - "loss": 0.5702, - "step": 538 - }, - { - "epoch": 0.17, - "learning_rate": 9.419968097209372e-07, - "loss": 0.5734, - "step": 539 - }, - { - "epoch": 0.17, - "learning_rate": 9.417640560119042e-07, - "loss": 0.548, - "step": 540 - }, - { - "epoch": 0.17, - "learning_rate": 9.415308651172851e-07, - "loss": 0.5611, - "step": 541 - }, - { - "epoch": 0.17, - "learning_rate": 9.412972372678539e-07, - "loss": 0.5577, - "step": 542 - }, - { - "epoch": 0.17, - "learning_rate": 9.410631726948171e-07, - "loss": 0.5638, - "step": 543 - }, - { - "epoch": 0.17, - "learning_rate": 9.408286716298133e-07, - "loss": 0.5558, - "step": 544 - }, - { - "epoch": 0.17, - "learning_rate": 9.405937343049137e-07, - "loss": 0.5695, - "step": 545 - }, - { - "epoch": 0.17, - "learning_rate": 9.403583609526202e-07, - "loss": 0.5388, - "step": 546 - }, - { - "epoch": 0.17, - "learning_rate": 9.401225518058671e-07, - "loss": 0.5591, - "step": 547 - }, - { - "epoch": 0.17, - "learning_rate": 9.398863070980192e-07, - "loss": 0.5722, - "step": 548 - }, - { - "epoch": 0.17, - "learning_rate": 9.396496270628731e-07, - "loss": 0.5667, - "step": 549 - }, - { - "epoch": 0.17, - "learning_rate": 9.394125119346557e-07, - "loss": 0.5624, - "step": 550 - }, - { - "epoch": 0.17, - "learning_rate": 9.391749619480246e-07, - "loss": 0.5789, - "step": 551 - }, - { - "epoch": 0.17, - "learning_rate": 9.38936977338068e-07, - "loss": 0.5985, - "step": 552 - }, - { - "epoch": 0.17, - "learning_rate": 9.386985583403039e-07, - "loss": 0.5721, - "step": 553 - }, - { - "epoch": 0.17, - "learning_rate": 9.384597051906803e-07, - "loss": 0.5722, - "step": 554 - }, - { - "epoch": 0.17, - "learning_rate": 9.382204181255749e-07, - "loss": 0.5768, - "step": 555 - }, - { - "epoch": 0.17, - "learning_rate": 9.379806973817947e-07, - "loss": 0.5575, - "step": 556 - }, - { - "epoch": 0.17, - "learning_rate": 9.377405431965762e-07, - "loss": 0.5619, - "step": 557 - }, - { - "epoch": 0.17, - "learning_rate": 9.374999558075843e-07, - "loss": 0.5474, - "step": 558 - }, - { - "epoch": 0.17, - "learning_rate": 9.372589354529131e-07, - "loss": 0.5727, - "step": 559 - }, - { - "epoch": 0.17, - "learning_rate": 9.37017482371085e-07, - "loss": 0.5738, - "step": 560 - }, - { - "epoch": 0.17, - "learning_rate": 9.367755968010507e-07, - "loss": 0.5503, - "step": 561 - }, - { - "epoch": 0.18, - "learning_rate": 9.365332789821885e-07, - "loss": 0.5538, - "step": 562 - }, - { - "epoch": 0.18, - "learning_rate": 9.362905291543053e-07, - "loss": 0.5676, - "step": 563 - }, - { - "epoch": 0.18, - "learning_rate": 9.36047347557635e-07, - "loss": 0.5527, - "step": 564 - }, - { - "epoch": 0.18, - "learning_rate": 9.358037344328384e-07, - "loss": 0.558, - "step": 565 - }, - { - "epoch": 0.18, - "learning_rate": 9.355596900210042e-07, - "loss": 0.5695, - "step": 566 - }, - { - "epoch": 0.18, - "learning_rate": 9.353152145636474e-07, - "loss": 0.5343, - "step": 567 - }, - { - "epoch": 0.18, - "learning_rate": 9.350703083027098e-07, - "loss": 0.5548, - "step": 568 - }, - { - "epoch": 0.18, - "learning_rate": 9.348249714805593e-07, - "loss": 0.5521, - "step": 569 - }, - { - "epoch": 0.18, - "learning_rate": 9.345792043399901e-07, - "loss": 0.5608, - "step": 570 - }, - { - "epoch": 0.18, - "learning_rate": 9.343330071242221e-07, - "loss": 0.5592, - "step": 571 - }, - { - "epoch": 0.18, - "learning_rate": 9.340863800769012e-07, - "loss": 0.5501, - "step": 572 - }, - { - "epoch": 0.18, - "learning_rate": 9.33839323442098e-07, - "loss": 0.5644, - "step": 573 - }, - { - "epoch": 0.18, - "learning_rate": 9.335918374643089e-07, - "loss": 0.5565, - "step": 574 - }, - { - "epoch": 0.18, - "learning_rate": 9.333439223884548e-07, - "loss": 0.5621, - "step": 575 - }, - { - "epoch": 0.18, - "learning_rate": 9.330955784598816e-07, - "loss": 0.5454, - "step": 576 - }, - { - "epoch": 0.18, - "learning_rate": 9.328468059243588e-07, - "loss": 0.5628, - "step": 577 - }, - { - "epoch": 0.18, - "learning_rate": 9.325976050280812e-07, - "loss": 0.5684, - "step": 578 - }, - { - "epoch": 0.18, - "learning_rate": 9.323479760176669e-07, - "loss": 0.5536, - "step": 579 - }, - { - "epoch": 0.18, - "learning_rate": 9.320979191401575e-07, - "loss": 0.5737, - "step": 580 - }, - { - "epoch": 0.18, - "learning_rate": 9.318474346430185e-07, - "loss": 0.5515, - "step": 581 - }, - { - "epoch": 0.18, - "learning_rate": 9.315965227741381e-07, - "loss": 0.5665, - "step": 582 - }, - { - "epoch": 0.18, - "learning_rate": 9.31345183781828e-07, - "loss": 0.5655, - "step": 583 - }, - { - "epoch": 0.18, - "learning_rate": 9.310934179148221e-07, - "loss": 0.5513, - "step": 584 - }, - { - "epoch": 0.18, - "learning_rate": 9.308412254222769e-07, - "loss": 0.5549, - "step": 585 - }, - { - "epoch": 0.18, - "learning_rate": 9.305886065537714e-07, - "loss": 0.5478, - "step": 586 - }, - { - "epoch": 0.18, - "learning_rate": 9.303355615593062e-07, - "loss": 0.5672, - "step": 587 - }, - { - "epoch": 0.18, - "learning_rate": 9.300820906893037e-07, - "loss": 0.5508, - "step": 588 - }, - { - "epoch": 0.18, - "learning_rate": 9.298281941946076e-07, - "loss": 0.5818, - "step": 589 - }, - { - "epoch": 0.18, - "learning_rate": 9.295738723264832e-07, - "loss": 0.5814, - "step": 590 - }, - { - "epoch": 0.18, - "learning_rate": 9.293191253366166e-07, - "loss": 0.5442, - "step": 591 - }, - { - "epoch": 0.18, - "learning_rate": 9.290639534771145e-07, - "loss": 0.5642, - "step": 592 - }, - { - "epoch": 0.18, - "learning_rate": 9.288083570005043e-07, - "loss": 0.5642, - "step": 593 - }, - { - "epoch": 0.19, - "learning_rate": 9.285523361597331e-07, - "loss": 0.5664, - "step": 594 - }, - { - "epoch": 0.19, - "learning_rate": 9.282958912081686e-07, - "loss": 0.5831, - "step": 595 - }, - { - "epoch": 0.19, - "learning_rate": 9.28039022399598e-07, - "loss": 0.5784, - "step": 596 - }, - { - "epoch": 0.19, - "learning_rate": 9.277817299882276e-07, - "loss": 0.5516, - "step": 597 - }, - { - "epoch": 0.19, - "learning_rate": 9.275240142286834e-07, - "loss": 0.5588, - "step": 598 - }, - { - "epoch": 0.19, - "learning_rate": 9.272658753760103e-07, - "loss": 0.5656, - "step": 599 - }, - { - "epoch": 0.19, - "learning_rate": 9.270073136856715e-07, - "loss": 0.5761, - "step": 600 - }, - { - "epoch": 0.19, - "learning_rate": 9.267483294135492e-07, - "loss": 0.5601, - "step": 601 - }, - { - "epoch": 0.19, - "learning_rate": 9.264889228159434e-07, - "loss": 0.552, - "step": 602 - }, - { - "epoch": 0.19, - "learning_rate": 9.262290941495724e-07, - "loss": 0.5548, - "step": 603 - }, - { - "epoch": 0.19, - "learning_rate": 9.259688436715717e-07, - "loss": 0.562, - "step": 604 - }, - { - "epoch": 0.19, - "learning_rate": 9.257081716394945e-07, - "loss": 0.565, - "step": 605 - }, - { - "epoch": 0.19, - "learning_rate": 9.254470783113117e-07, - "loss": 0.5281, - "step": 606 - }, - { - "epoch": 0.19, - "learning_rate": 9.2518556394541e-07, - "loss": 0.567, - "step": 607 - }, - { - "epoch": 0.19, - "learning_rate": 9.249236288005938e-07, - "loss": 0.5574, - "step": 608 - }, - { - "epoch": 0.19, - "learning_rate": 9.246612731360834e-07, - "loss": 0.5474, - "step": 609 - }, - { - "epoch": 0.19, - "learning_rate": 9.243984972115155e-07, - "loss": 0.5541, - "step": 610 - }, - { - "epoch": 0.19, - "learning_rate": 9.241353012869426e-07, - "loss": 0.5732, - "step": 611 - }, - { - "epoch": 0.19, - "learning_rate": 9.238716856228328e-07, - "loss": 0.567, - "step": 612 - }, - { - "epoch": 0.19, - "learning_rate": 9.236076504800696e-07, - "loss": 0.547, - "step": 613 - }, - { - "epoch": 0.19, - "learning_rate": 9.233431961199516e-07, - "loss": 0.5638, - "step": 614 - }, - { - "epoch": 0.19, - "learning_rate": 9.230783228041925e-07, - "loss": 0.5549, - "step": 615 - }, - { - "epoch": 0.19, - "learning_rate": 9.228130307949202e-07, - "loss": 0.5552, - "step": 616 - }, - { - "epoch": 0.19, - "learning_rate": 9.225473203546777e-07, - "loss": 0.5536, - "step": 617 - }, - { - "epoch": 0.19, - "learning_rate": 9.22281191746421e-07, - "loss": 0.544, - "step": 618 - }, - { - "epoch": 0.19, - "learning_rate": 9.220146452335209e-07, - "loss": 0.5785, - "step": 619 - }, - { - "epoch": 0.19, - "learning_rate": 9.217476810797615e-07, - "loss": 0.5422, - "step": 620 - }, - { - "epoch": 0.19, - "learning_rate": 9.214802995493398e-07, - "loss": 0.5392, - "step": 621 - }, - { - "epoch": 0.19, - "learning_rate": 9.212125009068663e-07, - "loss": 0.5459, - "step": 622 - }, - { - "epoch": 0.19, - "learning_rate": 9.209442854173642e-07, - "loss": 0.5519, - "step": 623 - }, - { - "epoch": 0.19, - "learning_rate": 9.206756533462693e-07, - "loss": 0.5406, - "step": 624 - }, - { - "epoch": 0.19, - "learning_rate": 9.204066049594293e-07, - "loss": 0.5659, - "step": 625 - }, - { - "epoch": 0.2, - "learning_rate": 9.201371405231042e-07, - "loss": 0.5495, - "step": 626 - }, - { - "epoch": 0.2, - "learning_rate": 9.198672603039659e-07, - "loss": 0.5745, - "step": 627 - }, - { - "epoch": 0.2, - "learning_rate": 9.195969645690975e-07, - "loss": 0.5426, - "step": 628 - }, - { - "epoch": 0.2, - "learning_rate": 9.193262535859931e-07, - "loss": 0.5331, - "step": 629 - }, - { - "epoch": 0.2, - "learning_rate": 9.190551276225583e-07, - "loss": 0.5605, - "step": 630 - }, - { - "epoch": 0.2, - "learning_rate": 9.18783586947109e-07, - "loss": 0.5297, - "step": 631 - }, - { - "epoch": 0.2, - "learning_rate": 9.185116318283717e-07, - "loss": 0.5738, - "step": 632 - }, - { - "epoch": 0.2, - "learning_rate": 9.182392625354829e-07, - "loss": 0.5587, - "step": 633 - }, - { - "epoch": 0.2, - "learning_rate": 9.17966479337989e-07, - "loss": 0.5475, - "step": 634 - }, - { - "epoch": 0.2, - "learning_rate": 9.176932825058461e-07, - "loss": 0.5635, - "step": 635 - }, - { - "epoch": 0.2, - "learning_rate": 9.174196723094195e-07, - "loss": 0.5581, - "step": 636 - }, - { - "epoch": 0.2, - "learning_rate": 9.171456490194837e-07, - "loss": 0.579, - "step": 637 - }, - { - "epoch": 0.2, - "learning_rate": 9.168712129072223e-07, - "loss": 0.5586, - "step": 638 - }, - { - "epoch": 0.2, - "learning_rate": 9.165963642442268e-07, - "loss": 0.5759, - "step": 639 - }, - { - "epoch": 0.2, - "learning_rate": 9.163211033024973e-07, - "loss": 0.553, - "step": 640 - }, - { - "epoch": 0.2, - "learning_rate": 9.160454303544423e-07, - "loss": 0.5573, - "step": 641 - }, - { - "epoch": 0.2, - "learning_rate": 9.157693456728772e-07, - "loss": 0.5683, - "step": 642 - }, - { - "epoch": 0.2, - "learning_rate": 9.154928495310257e-07, - "loss": 0.5372, - "step": 643 - }, - { - "epoch": 0.2, - "learning_rate": 9.15215942202518e-07, - "loss": 0.557, - "step": 644 - }, - { - "epoch": 0.2, - "learning_rate": 9.149386239613917e-07, - "loss": 0.545, - "step": 645 - }, - { - "epoch": 0.2, - "learning_rate": 9.146608950820909e-07, - "loss": 0.5387, - "step": 646 - }, - { - "epoch": 0.2, - "learning_rate": 9.14382755839466e-07, - "loss": 0.5546, - "step": 647 - }, - { - "epoch": 0.2, - "learning_rate": 9.141042065087737e-07, - "loss": 0.5538, - "step": 648 - }, - { - "epoch": 0.2, - "learning_rate": 9.138252473656764e-07, - "loss": 0.5312, - "step": 649 - }, - { - "epoch": 0.2, - "learning_rate": 9.13545878686242e-07, - "loss": 0.5674, - "step": 650 - }, - { - "epoch": 0.2, - "learning_rate": 9.13266100746944e-07, - "loss": 0.5376, - "step": 651 - }, - { - "epoch": 0.2, - "learning_rate": 9.129859138246605e-07, - "loss": 0.5411, - "step": 652 - }, - { - "epoch": 0.2, - "learning_rate": 9.127053181966745e-07, - "loss": 0.5501, - "step": 653 - }, - { - "epoch": 0.2, - "learning_rate": 9.124243141406738e-07, - "loss": 0.5384, - "step": 654 - }, - { - "epoch": 0.2, - "learning_rate": 9.121429019347499e-07, - "loss": 0.5427, - "step": 655 - }, - { - "epoch": 0.2, - "learning_rate": 9.118610818573984e-07, - "loss": 0.5322, - "step": 656 - }, - { - "epoch": 0.2, - "learning_rate": 9.115788541875186e-07, - "loss": 0.545, - "step": 657 - }, - { - "epoch": 0.21, - "learning_rate": 9.112962192044133e-07, - "loss": 0.5249, - "step": 658 - }, - { - "epoch": 0.21, - "learning_rate": 9.110131771877881e-07, - "loss": 0.5286, - "step": 659 - }, - { - "epoch": 0.21, - "learning_rate": 9.107297284177516e-07, - "loss": 0.5505, - "step": 660 - }, - { - "epoch": 0.21, - "learning_rate": 9.10445873174815e-07, - "loss": 0.5463, - "step": 661 - }, - { - "epoch": 0.21, - "learning_rate": 9.101616117398914e-07, - "loss": 0.529, - "step": 662 - }, - { - "epoch": 0.21, - "learning_rate": 9.098769443942961e-07, - "loss": 0.5797, - "step": 663 - }, - { - "epoch": 0.21, - "learning_rate": 9.095918714197464e-07, - "loss": 0.5486, - "step": 664 - }, - { - "epoch": 0.21, - "learning_rate": 9.093063930983608e-07, - "loss": 0.5446, - "step": 665 - }, - { - "epoch": 0.21, - "learning_rate": 9.090205097126587e-07, - "loss": 0.5435, - "step": 666 - }, - { - "epoch": 0.21, - "learning_rate": 9.087342215455605e-07, - "loss": 0.5647, - "step": 667 - }, - { - "epoch": 0.21, - "learning_rate": 9.084475288803874e-07, - "loss": 0.5676, - "step": 668 - }, - { - "epoch": 0.21, - "learning_rate": 9.081604320008609e-07, - "loss": 0.5409, - "step": 669 - }, - { - "epoch": 0.21, - "learning_rate": 9.07872931191102e-07, - "loss": 0.549, - "step": 670 - }, - { - "epoch": 0.21, - "learning_rate": 9.075850267356323e-07, - "loss": 0.5272, - "step": 671 - }, - { - "epoch": 0.21, - "learning_rate": 9.072967189193721e-07, - "loss": 0.5446, - "step": 672 - }, - { - "epoch": 0.21, - "learning_rate": 9.070080080276413e-07, - "loss": 0.5417, - "step": 673 - }, - { - "epoch": 0.21, - "learning_rate": 9.067188943461585e-07, - "loss": 0.5449, - "step": 674 - }, - { - "epoch": 0.21, - "learning_rate": 9.06429378161041e-07, - "loss": 0.5251, - "step": 675 - }, - { - "epoch": 0.21, - "learning_rate": 9.061394597588044e-07, - "loss": 0.5428, - "step": 676 - }, - { - "epoch": 0.21, - "learning_rate": 9.058491394263625e-07, - "loss": 0.5552, - "step": 677 - }, - { - "epoch": 0.21, - "learning_rate": 9.055584174510265e-07, - "loss": 0.5567, - "step": 678 - }, - { - "epoch": 0.21, - "learning_rate": 9.052672941205056e-07, - "loss": 0.5448, - "step": 679 - }, - { - "epoch": 0.21, - "learning_rate": 9.049757697229057e-07, - "loss": 0.548, - "step": 680 - }, - { - "epoch": 0.21, - "learning_rate": 9.046838445467298e-07, - "loss": 0.5349, - "step": 681 - }, - { - "epoch": 0.21, - "learning_rate": 9.043915188808778e-07, - "loss": 0.5521, - "step": 682 - }, - { - "epoch": 0.21, - "learning_rate": 9.040987930146453e-07, - "loss": 0.5235, - "step": 683 - }, - { - "epoch": 0.21, - "learning_rate": 9.038056672377248e-07, - "loss": 0.5181, - "step": 684 - }, - { - "epoch": 0.21, - "learning_rate": 9.035121418402038e-07, - "loss": 0.5358, - "step": 685 - }, - { - "epoch": 0.21, - "learning_rate": 9.032182171125655e-07, - "loss": 0.5425, - "step": 686 - }, - { - "epoch": 0.21, - "learning_rate": 9.029238933456887e-07, - "loss": 0.5213, - "step": 687 - }, - { - "epoch": 0.21, - "learning_rate": 9.026291708308464e-07, - "loss": 0.5372, - "step": 688 - }, - { - "epoch": 0.21, - "learning_rate": 9.023340498597068e-07, - "loss": 0.5391, - "step": 689 - }, - { - "epoch": 0.22, - "learning_rate": 9.020385307243323e-07, - "loss": 0.5793, - "step": 690 - }, - { - "epoch": 0.22, - "learning_rate": 9.017426137171791e-07, - "loss": 0.5541, - "step": 691 - }, - { - "epoch": 0.22, - "learning_rate": 9.014462991310974e-07, - "loss": 0.5554, - "step": 692 - }, - { - "epoch": 0.22, - "learning_rate": 9.011495872593307e-07, - "loss": 0.5494, - "step": 693 - }, - { - "epoch": 0.22, - "learning_rate": 9.008524783955158e-07, - "loss": 0.5369, - "step": 694 - }, - { - "epoch": 0.22, - "learning_rate": 9.005549728336822e-07, - "loss": 0.5319, - "step": 695 - }, - { - "epoch": 0.22, - "learning_rate": 9.002570708682522e-07, - "loss": 0.5562, - "step": 696 - }, - { - "epoch": 0.22, - "learning_rate": 8.999587727940404e-07, - "loss": 0.5273, - "step": 697 - }, - { - "epoch": 0.22, - "learning_rate": 8.996600789062529e-07, - "loss": 0.5425, - "step": 698 - }, - { - "epoch": 0.22, - "learning_rate": 8.993609895004883e-07, - "loss": 0.5327, - "step": 699 - }, - { - "epoch": 0.22, - "learning_rate": 8.99061504872736e-07, - "loss": 0.5336, - "step": 700 - }, - { - "epoch": 0.22, - "learning_rate": 8.987616253193768e-07, - "loss": 0.5643, - "step": 701 - }, - { - "epoch": 0.22, - "learning_rate": 8.984613511371824e-07, - "loss": 0.5372, - "step": 702 - }, - { - "epoch": 0.22, - "learning_rate": 8.981606826233147e-07, - "loss": 0.5513, - "step": 703 - }, - { - "epoch": 0.22, - "learning_rate": 8.978596200753266e-07, - "loss": 0.5551, - "step": 704 - }, - { - "epoch": 0.22, - "learning_rate": 8.975581637911596e-07, - "loss": 0.5589, - "step": 705 - }, - { - "epoch": 0.22, - "learning_rate": 8.972563140691463e-07, - "loss": 0.5439, - "step": 706 - }, - { - "epoch": 0.22, - "learning_rate": 8.969540712080078e-07, - "loss": 0.5658, - "step": 707 - }, - { - "epoch": 0.22, - "learning_rate": 8.966514355068544e-07, - "loss": 0.521, - "step": 708 - }, - { - "epoch": 0.22, - "learning_rate": 8.963484072651854e-07, - "loss": 0.5386, - "step": 709 - }, - { - "epoch": 0.22, - "learning_rate": 8.960449867828886e-07, - "loss": 0.5183, - "step": 710 - }, - { - "epoch": 0.22, - "learning_rate": 8.957411743602396e-07, - "loss": 0.5099, - "step": 711 - }, - { - "epoch": 0.22, - "learning_rate": 8.954369702979021e-07, - "loss": 0.5456, - "step": 712 - }, - { - "epoch": 0.22, - "learning_rate": 8.951323748969273e-07, - "loss": 0.5375, - "step": 713 - }, - { - "epoch": 0.22, - "learning_rate": 8.94827388458754e-07, - "loss": 0.5583, - "step": 714 - }, - { - "epoch": 0.22, - "learning_rate": 8.945220112852076e-07, - "loss": 0.5435, - "step": 715 - }, - { - "epoch": 0.22, - "learning_rate": 8.942162436785003e-07, - "loss": 0.528, - "step": 716 - }, - { - "epoch": 0.22, - "learning_rate": 8.939100859412307e-07, - "loss": 0.5498, - "step": 717 - }, - { - "epoch": 0.22, - "learning_rate": 8.936035383763836e-07, - "loss": 0.5362, - "step": 718 - }, - { - "epoch": 0.22, - "learning_rate": 8.932966012873293e-07, - "loss": 0.5306, - "step": 719 - }, - { - "epoch": 0.22, - "learning_rate": 8.929892749778241e-07, - "loss": 0.5314, - "step": 720 - }, - { - "epoch": 0.22, - "learning_rate": 8.926815597520087e-07, - "loss": 0.5462, - "step": 721 - }, - { - "epoch": 0.23, - "learning_rate": 8.923734559144096e-07, - "loss": 0.5423, - "step": 722 - }, - { - "epoch": 0.23, - "learning_rate": 8.920649637699372e-07, - "loss": 0.5379, - "step": 723 - }, - { - "epoch": 0.23, - "learning_rate": 8.917560836238865e-07, - "loss": 0.5106, - "step": 724 - }, - { - "epoch": 0.23, - "learning_rate": 8.914468157819363e-07, - "loss": 0.5237, - "step": 725 - }, - { - "epoch": 0.23, - "learning_rate": 8.911371605501492e-07, - "loss": 0.5058, - "step": 726 - }, - { - "epoch": 0.23, - "learning_rate": 8.908271182349712e-07, - "loss": 0.5505, - "step": 727 - }, - { - "epoch": 0.23, - "learning_rate": 8.905166891432312e-07, - "loss": 0.5339, - "step": 728 - }, - { - "epoch": 0.23, - "learning_rate": 8.902058735821411e-07, - "loss": 0.5201, - "step": 729 - }, - { - "epoch": 0.23, - "learning_rate": 8.89894671859295e-07, - "loss": 0.552, - "step": 730 - }, - { - "epoch": 0.23, - "learning_rate": 8.895830842826696e-07, - "loss": 0.5554, - "step": 731 - }, - { - "epoch": 0.23, - "learning_rate": 8.892711111606229e-07, - "loss": 0.5564, - "step": 732 - }, - { - "epoch": 0.23, - "learning_rate": 8.889587528018949e-07, - "loss": 0.5503, - "step": 733 - }, - { - "epoch": 0.23, - "learning_rate": 8.886460095156065e-07, - "loss": 0.5358, - "step": 734 - }, - { - "epoch": 0.23, - "learning_rate": 8.883328816112596e-07, - "loss": 0.527, - "step": 735 - }, - { - "epoch": 0.23, - "learning_rate": 8.880193693987372e-07, - "loss": 0.5328, - "step": 736 - }, - { - "epoch": 0.23, - "learning_rate": 8.87705473188302e-07, - "loss": 0.5173, - "step": 737 - }, - { - "epoch": 0.23, - "learning_rate": 8.87391193290597e-07, - "loss": 0.5635, - "step": 738 - }, - { - "epoch": 0.23, - "learning_rate": 8.870765300166448e-07, - "loss": 0.5293, - "step": 739 - }, - { - "epoch": 0.23, - "learning_rate": 8.867614836778476e-07, - "loss": 0.5142, - "step": 740 - }, - { - "epoch": 0.23, - "learning_rate": 8.864460545859866e-07, - "loss": 0.5292, - "step": 741 - }, - { - "epoch": 0.23, - "learning_rate": 8.861302430532216e-07, - "loss": 0.5375, - "step": 742 - }, - { - "epoch": 0.23, - "learning_rate": 8.858140493920912e-07, - "loss": 0.5437, - "step": 743 - }, - { - "epoch": 0.23, - "learning_rate": 8.854974739155121e-07, - "loss": 0.5741, - "step": 744 - }, - { - "epoch": 0.23, - "learning_rate": 8.851805169367786e-07, - "loss": 0.5587, - "step": 745 - }, - { - "epoch": 0.23, - "learning_rate": 8.848631787695626e-07, - "loss": 0.545, - "step": 746 - }, - { - "epoch": 0.23, - "learning_rate": 8.845454597279137e-07, - "loss": 0.5323, - "step": 747 - }, - { - "epoch": 0.23, - "learning_rate": 8.842273601262576e-07, - "loss": 0.54, - "step": 748 - }, - { - "epoch": 0.23, - "learning_rate": 8.839088802793977e-07, - "loss": 0.5403, - "step": 749 - }, - { - "epoch": 0.23, - "learning_rate": 8.835900205025127e-07, - "loss": 0.536, - "step": 750 - }, - { - "epoch": 0.23, - "learning_rate": 8.832707811111577e-07, - "loss": 0.5403, - "step": 751 - }, - { - "epoch": 0.23, - "learning_rate": 8.829511624212635e-07, - "loss": 0.537, - "step": 752 - }, - { - "epoch": 0.23, - "learning_rate": 8.826311647491365e-07, - "loss": 0.537, - "step": 753 - }, - { - "epoch": 0.24, - "learning_rate": 8.823107884114575e-07, - "loss": 0.5353, - "step": 754 - }, - { - "epoch": 0.24, - "learning_rate": 8.819900337252828e-07, - "loss": 0.5354, - "step": 755 - }, - { - "epoch": 0.24, - "learning_rate": 8.816689010080426e-07, - "loss": 0.5205, - "step": 756 - }, - { - "epoch": 0.24, - "learning_rate": 8.813473905775413e-07, - "loss": 0.5198, - "step": 757 - }, - { - "epoch": 0.24, - "learning_rate": 8.810255027519573e-07, - "loss": 0.5361, - "step": 758 - }, - { - "epoch": 0.24, - "learning_rate": 8.807032378498422e-07, - "loss": 0.5476, - "step": 759 - }, - { - "epoch": 0.24, - "learning_rate": 8.803805961901212e-07, - "loss": 0.5304, - "step": 760 - }, - { - "epoch": 0.24, - "learning_rate": 8.800575780920918e-07, - "loss": 0.5332, - "step": 761 - }, - { - "epoch": 0.24, - "learning_rate": 8.797341838754247e-07, - "loss": 0.552, - "step": 762 - }, - { - "epoch": 0.24, - "learning_rate": 8.794104138601619e-07, - "loss": 0.5261, - "step": 763 - }, - { - "epoch": 0.24, - "learning_rate": 8.790862683667182e-07, - "loss": 0.5587, - "step": 764 - }, - { - "epoch": 0.24, - "learning_rate": 8.787617477158795e-07, - "loss": 0.527, - "step": 765 - }, - { - "epoch": 0.24, - "learning_rate": 8.784368522288033e-07, - "loss": 0.5083, - "step": 766 - }, - { - "epoch": 0.24, - "learning_rate": 8.781115822270176e-07, - "loss": 0.5271, - "step": 767 - }, - { - "epoch": 0.24, - "learning_rate": 8.777859380324215e-07, - "loss": 0.5214, - "step": 768 - }, - { - "epoch": 0.24, - "learning_rate": 8.774599199672839e-07, - "loss": 0.5274, - "step": 769 - }, - { - "epoch": 0.24, - "learning_rate": 8.771335283542442e-07, - "loss": 0.5065, - "step": 770 - }, - { - "epoch": 0.24, - "learning_rate": 8.768067635163115e-07, - "loss": 0.5321, - "step": 771 - }, - { - "epoch": 0.24, - "learning_rate": 8.764796257768637e-07, - "loss": 0.5197, - "step": 772 - }, - { - "epoch": 0.24, - "learning_rate": 8.76152115459648e-07, - "loss": 0.5122, - "step": 773 - }, - { - "epoch": 0.24, - "learning_rate": 8.758242328887806e-07, - "loss": 0.5408, - "step": 774 - }, - { - "epoch": 0.24, - "learning_rate": 8.754959783887457e-07, - "loss": 0.5221, - "step": 775 - }, - { - "epoch": 0.24, - "learning_rate": 8.75167352284396e-07, - "loss": 0.5206, - "step": 776 - }, - { - "epoch": 0.24, - "learning_rate": 8.748383549009514e-07, - "loss": 0.5517, - "step": 777 - }, - { - "epoch": 0.24, - "learning_rate": 8.745089865639999e-07, - "loss": 0.5339, - "step": 778 - }, - { - "epoch": 0.24, - "learning_rate": 8.741792475994958e-07, - "loss": 0.5303, - "step": 779 - }, - { - "epoch": 0.24, - "learning_rate": 8.73849138333761e-07, - "loss": 0.5384, - "step": 780 - }, - { - "epoch": 0.24, - "learning_rate": 8.735186590934832e-07, - "loss": 0.5423, - "step": 781 - }, - { - "epoch": 0.24, - "learning_rate": 8.731878102057169e-07, - "loss": 0.5418, - "step": 782 - }, - { - "epoch": 0.24, - "learning_rate": 8.728565919978816e-07, - "loss": 0.5411, - "step": 783 - }, - { - "epoch": 0.24, - "learning_rate": 8.725250047977631e-07, - "loss": 0.556, - "step": 784 - }, - { - "epoch": 0.24, - "learning_rate": 8.721930489335119e-07, - "loss": 0.5331, - "step": 785 - }, - { - "epoch": 0.25, - "learning_rate": 8.718607247336434e-07, - "loss": 0.5469, - "step": 786 - }, - { - "epoch": 0.25, - "learning_rate": 8.715280325270376e-07, - "loss": 0.5336, - "step": 787 - }, - { - "epoch": 0.25, - "learning_rate": 8.711949726429389e-07, - "loss": 0.5241, - "step": 788 - }, - { - "epoch": 0.25, - "learning_rate": 8.708615454109548e-07, - "loss": 0.5124, - "step": 789 - }, - { - "epoch": 0.25, - "learning_rate": 8.705277511610571e-07, - "loss": 0.5522, - "step": 790 - }, - { - "epoch": 0.25, - "learning_rate": 8.70193590223581e-07, - "loss": 0.5187, - "step": 791 - }, - { - "epoch": 0.25, - "learning_rate": 8.698590629292237e-07, - "loss": 0.5252, - "step": 792 - }, - { - "epoch": 0.25, - "learning_rate": 8.695241696090455e-07, - "loss": 0.5236, - "step": 793 - }, - { - "epoch": 0.25, - "learning_rate": 8.69188910594469e-07, - "loss": 0.5568, - "step": 794 - }, - { - "epoch": 0.25, - "learning_rate": 8.688532862172785e-07, - "loss": 0.527, - "step": 795 - }, - { - "epoch": 0.25, - "learning_rate": 8.685172968096199e-07, - "loss": 0.5191, - "step": 796 - }, - { - "epoch": 0.25, - "learning_rate": 8.681809427040005e-07, - "loss": 0.5322, - "step": 797 - }, - { - "epoch": 0.25, - "learning_rate": 8.678442242332884e-07, - "loss": 0.5501, - "step": 798 - }, - { - "epoch": 0.25, - "learning_rate": 8.67507141730712e-07, - "loss": 0.5449, - "step": 799 - }, - { - "epoch": 0.25, - "learning_rate": 8.671696955298608e-07, - "loss": 0.5091, - "step": 800 - }, - { - "epoch": 0.25, - "learning_rate": 8.668318859646832e-07, - "loss": 0.5292, - "step": 801 - }, - { - "epoch": 0.25, - "learning_rate": 8.664937133694881e-07, - "loss": 0.5044, - "step": 802 - }, - { - "epoch": 0.25, - "learning_rate": 8.661551780789429e-07, - "loss": 0.5086, - "step": 803 - }, - { - "epoch": 0.25, - "learning_rate": 8.658162804280744e-07, - "loss": 0.529, - "step": 804 - }, - { - "epoch": 0.25, - "learning_rate": 8.654770207522679e-07, - "loss": 0.5125, - "step": 805 - }, - { - "epoch": 0.25, - "learning_rate": 8.651373993872672e-07, - "loss": 0.5237, - "step": 806 - }, - { - "epoch": 0.25, - "learning_rate": 8.647974166691734e-07, - "loss": 0.5332, - "step": 807 - }, - { - "epoch": 0.25, - "learning_rate": 8.644570729344458e-07, - "loss": 0.5401, - "step": 808 - }, - { - "epoch": 0.25, - "learning_rate": 8.641163685199009e-07, - "loss": 0.5227, - "step": 809 - }, - { - "epoch": 0.25, - "learning_rate": 8.637753037627119e-07, - "loss": 0.5314, - "step": 810 - }, - { - "epoch": 0.25, - "learning_rate": 8.634338790004087e-07, - "loss": 0.5202, - "step": 811 - }, - { - "epoch": 0.25, - "learning_rate": 8.630920945708777e-07, - "loss": 0.5226, - "step": 812 - }, - { - "epoch": 0.25, - "learning_rate": 8.62749950812361e-07, - "loss": 0.5327, - "step": 813 - }, - { - "epoch": 0.25, - "learning_rate": 8.624074480634562e-07, - "loss": 0.5372, - "step": 814 - }, - { - "epoch": 0.25, - "learning_rate": 8.620645866631165e-07, - "loss": 0.5176, - "step": 815 - }, - { - "epoch": 0.25, - "learning_rate": 8.617213669506498e-07, - "loss": 0.5336, - "step": 816 - }, - { - "epoch": 0.25, - "learning_rate": 8.613777892657187e-07, - "loss": 0.5388, - "step": 817 - }, - { - "epoch": 0.25, - "learning_rate": 8.6103385394834e-07, - "loss": 0.5228, - "step": 818 - }, - { - "epoch": 0.26, - "learning_rate": 8.606895613388843e-07, - "loss": 0.5431, - "step": 819 - }, - { - "epoch": 0.26, - "learning_rate": 8.603449117780761e-07, - "loss": 0.5129, - "step": 820 - }, - { - "epoch": 0.26, - "learning_rate": 8.599999056069931e-07, - "loss": 0.5321, - "step": 821 - }, - { - "epoch": 0.26, - "learning_rate": 8.596545431670657e-07, - "loss": 0.5104, - "step": 822 - }, - { - "epoch": 0.26, - "learning_rate": 8.593088248000768e-07, - "loss": 0.5437, - "step": 823 - }, - { - "epoch": 0.26, - "learning_rate": 8.58962750848162e-07, - "loss": 0.4967, - "step": 824 - }, - { - "epoch": 0.26, - "learning_rate": 8.586163216538083e-07, - "loss": 0.528, - "step": 825 - }, - { - "epoch": 0.26, - "learning_rate": 8.582695375598546e-07, - "loss": 0.5406, - "step": 826 - }, - { - "epoch": 0.26, - "learning_rate": 8.57922398909491e-07, - "loss": 0.5295, - "step": 827 - }, - { - "epoch": 0.26, - "learning_rate": 8.575749060462581e-07, - "loss": 0.5314, - "step": 828 - }, - { - "epoch": 0.26, - "learning_rate": 8.572270593140476e-07, - "loss": 0.5126, - "step": 829 - }, - { - "epoch": 0.26, - "learning_rate": 8.56878859057101e-07, - "loss": 0.5409, - "step": 830 - }, - { - "epoch": 0.26, - "learning_rate": 8.565303056200098e-07, - "loss": 0.5281, - "step": 831 - }, - { - "epoch": 0.26, - "learning_rate": 8.561813993477149e-07, - "loss": 0.4975, - "step": 832 - }, - { - "epoch": 0.26, - "learning_rate": 8.558321405855069e-07, - "loss": 0.527, - "step": 833 - }, - { - "epoch": 0.26, - "learning_rate": 8.554825296790242e-07, - "loss": 0.5197, - "step": 834 - }, - { - "epoch": 0.26, - "learning_rate": 8.551325669742548e-07, - "loss": 0.5116, - "step": 835 - }, - { - "epoch": 0.26, - "learning_rate": 8.547822528175342e-07, - "loss": 0.5152, - "step": 836 - }, - { - "epoch": 0.26, - "learning_rate": 8.54431587555546e-07, - "loss": 0.5109, - "step": 837 - }, - { - "epoch": 0.26, - "learning_rate": 8.540805715353209e-07, - "loss": 0.517, - "step": 838 - }, - { - "epoch": 0.26, - "learning_rate": 8.537292051042372e-07, - "loss": 0.507, - "step": 839 - }, - { - "epoch": 0.26, - "learning_rate": 8.533774886100196e-07, - "loss": 0.5293, - "step": 840 - }, - { - "epoch": 0.26, - "learning_rate": 8.530254224007394e-07, - "loss": 0.5286, - "step": 841 - }, - { - "epoch": 0.26, - "learning_rate": 8.52673006824814e-07, - "loss": 0.5124, - "step": 842 - }, - { - "epoch": 0.26, - "learning_rate": 8.523202422310066e-07, - "loss": 0.5513, - "step": 843 - }, - { - "epoch": 0.26, - "learning_rate": 8.519671289684256e-07, - "loss": 0.5233, - "step": 844 - }, - { - "epoch": 0.26, - "learning_rate": 8.516136673865246e-07, - "loss": 0.4951, - "step": 845 - }, - { - "epoch": 0.26, - "learning_rate": 8.51259857835102e-07, - "loss": 0.5138, - "step": 846 - }, - { - "epoch": 0.26, - "learning_rate": 8.509057006643003e-07, - "loss": 0.4947, - "step": 847 - }, - { - "epoch": 0.26, - "learning_rate": 8.505511962246061e-07, - "loss": 0.5296, - "step": 848 - }, - { - "epoch": 0.26, - "learning_rate": 8.501963448668501e-07, - "loss": 0.5126, - "step": 849 - }, - { - "epoch": 0.26, - "learning_rate": 8.498411469422055e-07, - "loss": 0.506, - "step": 850 - }, - { - "epoch": 0.27, - "learning_rate": 8.494856028021893e-07, - "loss": 0.5151, - "step": 851 - }, - { - "epoch": 0.27, - "learning_rate": 8.491297127986605e-07, - "loss": 0.5202, - "step": 852 - }, - { - "epoch": 0.27, - "learning_rate": 8.487734772838209e-07, - "loss": 0.5264, - "step": 853 - }, - { - "epoch": 0.27, - "learning_rate": 8.484168966102136e-07, - "loss": 0.521, - "step": 854 - }, - { - "epoch": 0.27, - "learning_rate": 8.480599711307239e-07, - "loss": 0.4925, - "step": 855 - }, - { - "epoch": 0.27, - "learning_rate": 8.477027011985781e-07, - "loss": 0.5129, - "step": 856 - }, - { - "epoch": 0.27, - "learning_rate": 8.473450871673432e-07, - "loss": 0.5385, - "step": 857 - }, - { - "epoch": 0.27, - "learning_rate": 8.469871293909272e-07, - "loss": 0.5424, - "step": 858 - }, - { - "epoch": 0.27, - "learning_rate": 8.466288282235777e-07, - "loss": 0.512, - "step": 859 - }, - { - "epoch": 0.27, - "learning_rate": 8.462701840198823e-07, - "loss": 0.5308, - "step": 860 - }, - { - "epoch": 0.27, - "learning_rate": 8.459111971347685e-07, - "loss": 0.5207, - "step": 861 - }, - { - "epoch": 0.27, - "learning_rate": 8.455518679235025e-07, - "loss": 0.5283, - "step": 862 - }, - { - "epoch": 0.27, - "learning_rate": 8.451921967416895e-07, - "loss": 0.5262, - "step": 863 - }, - { - "epoch": 0.27, - "learning_rate": 8.44832183945273e-07, - "loss": 0.5267, - "step": 864 - }, - { - "epoch": 0.27, - "learning_rate": 8.444718298905342e-07, - "loss": 0.5286, - "step": 865 - }, - { - "epoch": 0.27, - "learning_rate": 8.44111134934093e-07, - "loss": 0.4955, - "step": 866 - }, - { - "epoch": 0.27, - "learning_rate": 8.437500994329058e-07, - "loss": 0.5185, - "step": 867 - }, - { - "epoch": 0.27, - "learning_rate": 8.433887237442665e-07, - "loss": 0.519, - "step": 868 - }, - { - "epoch": 0.27, - "learning_rate": 8.430270082258053e-07, - "loss": 0.5154, - "step": 869 - }, - { - "epoch": 0.27, - "learning_rate": 8.426649532354887e-07, - "loss": 0.5122, - "step": 870 - }, - { - "epoch": 0.27, - "learning_rate": 8.423025591316197e-07, - "loss": 0.5125, - "step": 871 - }, - { - "epoch": 0.27, - "learning_rate": 8.419398262728363e-07, - "loss": 0.4917, - "step": 872 - }, - { - "epoch": 0.27, - "learning_rate": 8.415767550181121e-07, - "loss": 0.5004, - "step": 873 - }, - { - "epoch": 0.27, - "learning_rate": 8.412133457267553e-07, - "loss": 0.5196, - "step": 874 - }, - { - "epoch": 0.27, - "learning_rate": 8.40849598758409e-07, - "loss": 0.536, - "step": 875 - }, - { - "epoch": 0.27, - "learning_rate": 8.4048551447305e-07, - "loss": 0.5052, - "step": 876 - }, - { - "epoch": 0.27, - "learning_rate": 8.401210932309895e-07, - "loss": 0.5203, - "step": 877 - }, - { - "epoch": 0.27, - "learning_rate": 8.397563353928719e-07, - "loss": 0.536, - "step": 878 - }, - { - "epoch": 0.27, - "learning_rate": 8.393912413196741e-07, - "loss": 0.5235, - "step": 879 - }, - { - "epoch": 0.27, - "learning_rate": 8.390258113727068e-07, - "loss": 0.5105, - "step": 880 - }, - { - "epoch": 0.27, - "learning_rate": 8.386600459136127e-07, - "loss": 0.5424, - "step": 881 - }, - { - "epoch": 0.27, - "learning_rate": 8.382939453043663e-07, - "loss": 0.5071, - "step": 882 - }, - { - "epoch": 0.28, - "learning_rate": 8.379275099072736e-07, - "loss": 0.5077, - "step": 883 - }, - { - "epoch": 0.28, - "learning_rate": 8.375607400849726e-07, - "loss": 0.5058, - "step": 884 - }, - { - "epoch": 0.28, - "learning_rate": 8.371936362004317e-07, - "loss": 0.4932, - "step": 885 - }, - { - "epoch": 0.28, - "learning_rate": 8.368261986169505e-07, - "loss": 0.5289, - "step": 886 - }, - { - "epoch": 0.28, - "learning_rate": 8.364584276981579e-07, - "loss": 0.5322, - "step": 887 - }, - { - "epoch": 0.28, - "learning_rate": 8.360903238080135e-07, - "loss": 0.517, - "step": 888 - }, - { - "epoch": 0.28, - "learning_rate": 8.35721887310806e-07, - "loss": 0.5087, - "step": 889 - }, - { - "epoch": 0.28, - "learning_rate": 8.353531185711537e-07, - "loss": 0.5123, - "step": 890 - }, - { - "epoch": 0.28, - "learning_rate": 8.349840179540032e-07, - "loss": 0.5196, - "step": 891 - }, - { - "epoch": 0.28, - "learning_rate": 8.346145858246297e-07, - "loss": 0.5339, - "step": 892 - }, - { - "epoch": 0.28, - "learning_rate": 8.342448225486365e-07, - "loss": 0.517, - "step": 893 - }, - { - "epoch": 0.28, - "learning_rate": 8.338747284919545e-07, - "loss": 0.5191, - "step": 894 - }, - { - "epoch": 0.28, - "learning_rate": 8.335043040208425e-07, - "loss": 0.496, - "step": 895 - }, - { - "epoch": 0.28, - "learning_rate": 8.331335495018854e-07, - "loss": 0.5104, - "step": 896 - }, - { - "epoch": 0.28, - "learning_rate": 8.327624653019956e-07, - "loss": 0.4942, - "step": 897 - }, - { - "epoch": 0.28, - "learning_rate": 8.323910517884112e-07, - "loss": 0.5025, - "step": 898 - }, - { - "epoch": 0.28, - "learning_rate": 8.320193093286962e-07, - "loss": 0.5394, - "step": 899 - }, - { - "epoch": 0.28, - "learning_rate": 8.316472382907405e-07, - "loss": 0.5056, - "step": 900 - }, - { - "epoch": 0.28, - "learning_rate": 8.312748390427588e-07, - "loss": 0.52, - "step": 901 - }, - { - "epoch": 0.28, - "learning_rate": 8.309021119532911e-07, - "loss": 0.4913, - "step": 902 - }, - { - "epoch": 0.28, - "learning_rate": 8.305290573912011e-07, - "loss": 0.514, - "step": 903 - }, - { - "epoch": 0.28, - "learning_rate": 8.301556757256773e-07, - "loss": 0.5072, - "step": 904 - }, - { - "epoch": 0.28, - "learning_rate": 8.297819673262315e-07, - "loss": 0.5321, - "step": 905 - }, - { - "epoch": 0.28, - "learning_rate": 8.294079325626989e-07, - "loss": 0.5124, - "step": 906 - }, - { - "epoch": 0.28, - "learning_rate": 8.29033571805238e-07, - "loss": 0.5075, - "step": 907 - }, - { - "epoch": 0.28, - "learning_rate": 8.286588854243293e-07, - "loss": 0.528, - "step": 908 - }, - { - "epoch": 0.28, - "learning_rate": 8.282838737907762e-07, - "loss": 0.4966, - "step": 909 - }, - { - "epoch": 0.28, - "learning_rate": 8.279085372757036e-07, - "loss": 0.4941, - "step": 910 - }, - { - "epoch": 0.28, - "learning_rate": 8.27532876250558e-07, - "loss": 0.5089, - "step": 911 - }, - { - "epoch": 0.28, - "learning_rate": 8.271568910871071e-07, - "loss": 0.5079, - "step": 912 - }, - { - "epoch": 0.28, - "learning_rate": 8.267805821574392e-07, - "loss": 0.5222, - "step": 913 - }, - { - "epoch": 0.28, - "learning_rate": 8.264039498339632e-07, - "loss": 0.5155, - "step": 914 - }, - { - "epoch": 0.29, - "learning_rate": 8.260269944894079e-07, - "loss": 0.5258, - "step": 915 - }, - { - "epoch": 0.29, - "learning_rate": 8.256497164968221e-07, - "loss": 0.5073, - "step": 916 - }, - { - "epoch": 0.29, - "learning_rate": 8.252721162295735e-07, - "loss": 0.5202, - "step": 917 - }, - { - "epoch": 0.29, - "learning_rate": 8.248941940613489e-07, - "loss": 0.5126, - "step": 918 - }, - { - "epoch": 0.29, - "learning_rate": 8.245159503661537e-07, - "loss": 0.5093, - "step": 919 - }, - { - "epoch": 0.29, - "learning_rate": 8.241373855183114e-07, - "loss": 0.5252, - "step": 920 - }, - { - "epoch": 0.29, - "learning_rate": 8.237584998924633e-07, - "loss": 0.526, - "step": 921 - }, - { - "epoch": 0.29, - "learning_rate": 8.233792938635684e-07, - "loss": 0.5336, - "step": 922 - }, - { - "epoch": 0.29, - "learning_rate": 8.229997678069025e-07, - "loss": 0.5175, - "step": 923 - }, - { - "epoch": 0.29, - "learning_rate": 8.226199220980583e-07, - "loss": 0.5244, - "step": 924 - }, - { - "epoch": 0.29, - "learning_rate": 8.222397571129448e-07, - "loss": 0.4949, - "step": 925 - }, - { - "epoch": 0.29, - "learning_rate": 8.218592732277868e-07, - "loss": 0.511, - "step": 926 - }, - { - "epoch": 0.29, - "learning_rate": 8.214784708191249e-07, - "loss": 0.5156, - "step": 927 - }, - { - "epoch": 0.29, - "learning_rate": 8.210973502638149e-07, - "loss": 0.5122, - "step": 928 - }, - { - "epoch": 0.29, - "learning_rate": 8.207159119390272e-07, - "loss": 0.5216, - "step": 929 - }, - { - "epoch": 0.29, - "learning_rate": 8.203341562222472e-07, - "loss": 0.516, - "step": 930 - }, - { - "epoch": 0.29, - "learning_rate": 8.199520834912738e-07, - "loss": 0.5417, - "step": 931 - }, - { - "epoch": 0.29, - "learning_rate": 8.195696941242201e-07, - "loss": 0.507, - "step": 932 - }, - { - "epoch": 0.29, - "learning_rate": 8.191869884995122e-07, - "loss": 0.5214, - "step": 933 - }, - { - "epoch": 0.29, - "learning_rate": 8.188039669958892e-07, - "loss": 0.5246, - "step": 934 - }, - { - "epoch": 0.29, - "learning_rate": 8.184206299924034e-07, - "loss": 0.526, - "step": 935 - }, - { - "epoch": 0.29, - "learning_rate": 8.180369778684185e-07, - "loss": 0.5115, - "step": 936 - }, - { - "epoch": 0.29, - "learning_rate": 8.176530110036106e-07, - "loss": 0.5136, - "step": 937 - }, - { - "epoch": 0.29, - "learning_rate": 8.17268729777967e-07, - "loss": 0.519, - "step": 938 - }, - { - "epoch": 0.29, - "learning_rate": 8.168841345717863e-07, - "loss": 0.5339, - "step": 939 - }, - { - "epoch": 0.29, - "learning_rate": 8.164992257656777e-07, - "loss": 0.4991, - "step": 940 - }, - { - "epoch": 0.29, - "learning_rate": 8.161140037405608e-07, - "loss": 0.5111, - "step": 941 - }, - { - "epoch": 0.29, - "learning_rate": 8.157284688776653e-07, - "loss": 0.5039, - "step": 942 - }, - { - "epoch": 0.29, - "learning_rate": 8.1534262155853e-07, - "loss": 0.516, - "step": 943 - }, - { - "epoch": 0.29, - "learning_rate": 8.149564621650036e-07, - "loss": 0.5066, - "step": 944 - }, - { - "epoch": 0.29, - "learning_rate": 8.145699910792432e-07, - "loss": 0.5042, - "step": 945 - }, - { - "epoch": 0.29, - "learning_rate": 8.141832086837143e-07, - "loss": 0.5245, - "step": 946 - }, - { - "epoch": 0.3, - "learning_rate": 8.13796115361191e-07, - "loss": 0.5038, - "step": 947 - }, - { - "epoch": 0.3, - "learning_rate": 8.134087114947546e-07, - "loss": 0.4965, - "step": 948 - }, - { - "epoch": 0.3, - "learning_rate": 8.130209974677938e-07, - "loss": 0.4997, - "step": 949 - }, - { - "epoch": 0.3, - "learning_rate": 8.126329736640045e-07, - "loss": 0.4698, - "step": 950 - }, - { - "epoch": 0.3, - "learning_rate": 8.122446404673888e-07, - "loss": 0.4948, - "step": 951 - }, - { - "epoch": 0.3, - "learning_rate": 8.118559982622554e-07, - "loss": 0.5019, - "step": 952 - }, - { - "epoch": 0.3, - "learning_rate": 8.114670474332186e-07, - "loss": 0.5124, - "step": 953 - }, - { - "epoch": 0.3, - "learning_rate": 8.110777883651978e-07, - "loss": 0.5036, - "step": 954 - }, - { - "epoch": 0.3, - "learning_rate": 8.106882214434182e-07, - "loss": 0.5101, - "step": 955 - }, - { - "epoch": 0.3, - "learning_rate": 8.102983470534089e-07, - "loss": 0.5035, - "step": 956 - }, - { - "epoch": 0.3, - "learning_rate": 8.099081655810038e-07, - "loss": 0.5215, - "step": 957 - }, - { - "epoch": 0.3, - "learning_rate": 8.095176774123402e-07, - "loss": 0.4956, - "step": 958 - }, - { - "epoch": 0.3, - "learning_rate": 8.091268829338597e-07, - "loss": 0.4868, - "step": 959 - }, - { - "epoch": 0.3, - "learning_rate": 8.087357825323062e-07, - "loss": 0.5065, - "step": 960 - }, - { - "epoch": 0.3, - "learning_rate": 8.083443765947269e-07, - "loss": 0.5137, - "step": 961 - }, - { - "epoch": 0.3, - "learning_rate": 8.07952665508471e-07, - "loss": 0.5052, - "step": 962 - }, - { - "epoch": 0.3, - "learning_rate": 8.0756064966119e-07, - "loss": 0.5143, - "step": 963 - }, - { - "epoch": 0.3, - "learning_rate": 8.071683294408369e-07, - "loss": 0.5193, - "step": 964 - }, - { - "epoch": 0.3, - "learning_rate": 8.067757052356659e-07, - "loss": 0.5301, - "step": 965 - }, - { - "epoch": 0.3, - "learning_rate": 8.063827774342318e-07, - "loss": 0.4867, - "step": 966 - }, - { - "epoch": 0.3, - "learning_rate": 8.059895464253903e-07, - "loss": 0.5071, - "step": 967 - }, - { - "epoch": 0.3, - "learning_rate": 8.055960125982967e-07, - "loss": 0.5137, - "step": 968 - }, - { - "epoch": 0.3, - "learning_rate": 8.052021763424066e-07, - "loss": 0.5052, - "step": 969 - }, - { - "epoch": 0.3, - "learning_rate": 8.048080380474743e-07, - "loss": 0.5138, - "step": 970 - }, - { - "epoch": 0.3, - "learning_rate": 8.044135981035532e-07, - "loss": 0.5129, - "step": 971 - }, - { - "epoch": 0.3, - "learning_rate": 8.040188569009954e-07, - "loss": 0.4896, - "step": 972 - }, - { - "epoch": 0.3, - "learning_rate": 8.036238148304506e-07, - "loss": 0.5035, - "step": 973 - }, - { - "epoch": 0.3, - "learning_rate": 8.032284722828669e-07, - "loss": 0.4972, - "step": 974 - }, - { - "epoch": 0.3, - "learning_rate": 8.028328296494897e-07, - "loss": 0.524, - "step": 975 - }, - { - "epoch": 0.3, - "learning_rate": 8.024368873218608e-07, - "loss": 0.5119, - "step": 976 - }, - { - "epoch": 0.3, - "learning_rate": 8.02040645691819e-07, - "loss": 0.5189, - "step": 977 - }, - { - "epoch": 0.3, - "learning_rate": 8.016441051514995e-07, - "loss": 0.5041, - "step": 978 - }, - { - "epoch": 0.31, - "learning_rate": 8.012472660933327e-07, - "loss": 0.5083, - "step": 979 - }, - { - "epoch": 0.31, - "learning_rate": 8.008501289100453e-07, - "loss": 0.5131, - "step": 980 - }, - { - "epoch": 0.31, - "learning_rate": 8.004526939946581e-07, - "loss": 0.4915, - "step": 981 - }, - { - "epoch": 0.31, - "learning_rate": 8.000549617404873e-07, - "loss": 0.5153, - "step": 982 - }, - { - "epoch": 0.31, - "learning_rate": 7.996569325411427e-07, - "loss": 0.5137, - "step": 983 - }, - { - "epoch": 0.31, - "learning_rate": 7.992586067905283e-07, - "loss": 0.5236, - "step": 984 - }, - { - "epoch": 0.31, - "learning_rate": 7.988599848828419e-07, - "loss": 0.502, - "step": 985 - }, - { - "epoch": 0.31, - "learning_rate": 7.984610672125739e-07, - "loss": 0.5069, - "step": 986 - }, - { - "epoch": 0.31, - "learning_rate": 7.980618541745077e-07, - "loss": 0.5207, - "step": 987 - }, - { - "epoch": 0.31, - "learning_rate": 7.976623461637184e-07, - "loss": 0.4924, - "step": 988 - }, - { - "epoch": 0.31, - "learning_rate": 7.972625435755741e-07, - "loss": 0.4693, - "step": 989 - }, - { - "epoch": 0.31, - "learning_rate": 7.968624468057334e-07, - "loss": 0.4906, - "step": 990 - }, - { - "epoch": 0.31, - "learning_rate": 7.964620562501467e-07, - "loss": 0.5127, - "step": 991 - }, - { - "epoch": 0.31, - "learning_rate": 7.960613723050545e-07, - "loss": 0.5183, - "step": 992 - }, - { - "epoch": 0.31, - "learning_rate": 7.956603953669884e-07, - "loss": 0.4972, - "step": 993 - }, - { - "epoch": 0.31, - "learning_rate": 7.952591258327696e-07, - "loss": 0.51, - "step": 994 - }, - { - "epoch": 0.31, - "learning_rate": 7.948575640995087e-07, - "loss": 0.4987, - "step": 995 - }, - { - "epoch": 0.31, - "learning_rate": 7.944557105646055e-07, - "loss": 0.5017, - "step": 996 - }, - { - "epoch": 0.31, - "learning_rate": 7.94053565625749e-07, - "loss": 0.5168, - "step": 997 - }, - { - "epoch": 0.31, - "learning_rate": 7.936511296809161e-07, - "loss": 0.5074, - "step": 998 - }, - { - "epoch": 0.31, - "learning_rate": 7.932484031283717e-07, - "loss": 0.5101, - "step": 999 - }, - { - "epoch": 0.31, - "learning_rate": 7.928453863666689e-07, - "loss": 0.5153, - "step": 1000 - }, - { - "epoch": 0.31, - "learning_rate": 7.924420797946472e-07, - "loss": 0.498, - "step": 1001 - }, - { - "epoch": 0.31, - "learning_rate": 7.920384838114336e-07, - "loss": 0.5069, - "step": 1002 - }, - { - "epoch": 0.31, - "learning_rate": 7.916345988164409e-07, - "loss": 0.5286, - "step": 1003 - }, - { - "epoch": 0.31, - "learning_rate": 7.912304252093684e-07, - "loss": 0.5292, - "step": 1004 - }, - { - "epoch": 0.31, - "learning_rate": 7.908259633902006e-07, - "loss": 0.4884, - "step": 1005 - }, - { - "epoch": 0.31, - "learning_rate": 7.904212137592077e-07, - "loss": 0.4956, - "step": 1006 - }, - { - "epoch": 0.31, - "learning_rate": 7.900161767169444e-07, - "loss": 0.5085, - "step": 1007 - }, - { - "epoch": 0.31, - "learning_rate": 7.896108526642498e-07, - "loss": 0.5139, - "step": 1008 - }, - { - "epoch": 0.31, - "learning_rate": 7.892052420022473e-07, - "loss": 0.5175, - "step": 1009 - }, - { - "epoch": 0.31, - "learning_rate": 7.887993451323438e-07, - "loss": 0.505, - "step": 1010 - }, - { - "epoch": 0.32, - "learning_rate": 7.883931624562292e-07, - "loss": 0.5145, - "step": 1011 - }, - { - "epoch": 0.32, - "learning_rate": 7.879866943758767e-07, - "loss": 0.5237, - "step": 1012 - }, - { - "epoch": 0.32, - "learning_rate": 7.875799412935418e-07, - "loss": 0.4926, - "step": 1013 - }, - { - "epoch": 0.32, - "learning_rate": 7.871729036117616e-07, - "loss": 0.4885, - "step": 1014 - }, - { - "epoch": 0.32, - "learning_rate": 7.867655817333556e-07, - "loss": 0.5035, - "step": 1015 - }, - { - "epoch": 0.32, - "learning_rate": 7.86357976061424e-07, - "loss": 0.4813, - "step": 1016 - }, - { - "epoch": 0.32, - "learning_rate": 7.859500869993479e-07, - "loss": 0.4963, - "step": 1017 - }, - { - "epoch": 0.32, - "learning_rate": 7.855419149507891e-07, - "loss": 0.5046, - "step": 1018 - }, - { - "epoch": 0.32, - "learning_rate": 7.851334603196896e-07, - "loss": 0.5197, - "step": 1019 - }, - { - "epoch": 0.32, - "learning_rate": 7.847247235102703e-07, - "loss": 0.5138, - "step": 1020 - }, - { - "epoch": 0.32, - "learning_rate": 7.84315704927032e-07, - "loss": 0.5145, - "step": 1021 - }, - { - "epoch": 0.32, - "learning_rate": 7.839064049747544e-07, - "loss": 0.5175, - "step": 1022 - }, - { - "epoch": 0.32, - "learning_rate": 7.834968240584952e-07, - "loss": 0.5078, - "step": 1023 - }, - { - "epoch": 0.32, - "learning_rate": 7.830869625835906e-07, - "loss": 0.4961, - "step": 1024 - }, - { - "epoch": 0.32, - "learning_rate": 7.826768209556538e-07, - "loss": 0.5129, - "step": 1025 - }, - { - "epoch": 0.32, - "learning_rate": 7.822663995805763e-07, - "loss": 0.511, - "step": 1026 - }, - { - "epoch": 0.32, - "learning_rate": 7.818556988645256e-07, - "loss": 0.5166, - "step": 1027 - }, - { - "epoch": 0.32, - "learning_rate": 7.814447192139457e-07, - "loss": 0.4931, - "step": 1028 - }, - { - "epoch": 0.32, - "learning_rate": 7.81033461035557e-07, - "loss": 0.5159, - "step": 1029 - }, - { - "epoch": 0.32, - "learning_rate": 7.806219247363554e-07, - "loss": 0.5096, - "step": 1030 - }, - { - "epoch": 0.32, - "learning_rate": 7.802101107236119e-07, - "loss": 0.4844, - "step": 1031 - }, - { - "epoch": 0.32, - "learning_rate": 7.797980194048726e-07, - "loss": 0.5006, - "step": 1032 - }, - { - "epoch": 0.32, - "learning_rate": 7.793856511879577e-07, - "loss": 0.4867, - "step": 1033 - }, - { - "epoch": 0.32, - "learning_rate": 7.789730064809619e-07, - "loss": 0.5114, - "step": 1034 - }, - { - "epoch": 0.32, - "learning_rate": 7.785600856922528e-07, - "loss": 0.5179, - "step": 1035 - }, - { - "epoch": 0.32, - "learning_rate": 7.781468892304719e-07, - "loss": 0.5184, - "step": 1036 - }, - { - "epoch": 0.32, - "learning_rate": 7.777334175045333e-07, - "loss": 0.4866, - "step": 1037 - }, - { - "epoch": 0.32, - "learning_rate": 7.773196709236236e-07, - "loss": 0.5054, - "step": 1038 - }, - { - "epoch": 0.32, - "learning_rate": 7.769056498972009e-07, - "loss": 0.5133, - "step": 1039 - }, - { - "epoch": 0.32, - "learning_rate": 7.764913548349956e-07, - "loss": 0.519, - "step": 1040 - }, - { - "epoch": 0.32, - "learning_rate": 7.760767861470089e-07, - "loss": 0.4959, - "step": 1041 - }, - { - "epoch": 0.32, - "learning_rate": 7.756619442435128e-07, - "loss": 0.4929, - "step": 1042 - }, - { - "epoch": 0.33, - "learning_rate": 7.752468295350498e-07, - "loss": 0.5009, - "step": 1043 - }, - { - "epoch": 0.33, - "learning_rate": 7.748314424324324e-07, - "loss": 0.4774, - "step": 1044 - }, - { - "epoch": 0.33, - "learning_rate": 7.744157833467426e-07, - "loss": 0.4926, - "step": 1045 - }, - { - "epoch": 0.33, - "learning_rate": 7.739998526893314e-07, - "loss": 0.474, - "step": 1046 - }, - { - "epoch": 0.33, - "learning_rate": 7.735836508718191e-07, - "loss": 0.4988, - "step": 1047 - }, - { - "epoch": 0.33, - "learning_rate": 7.731671783060935e-07, - "loss": 0.5147, - "step": 1048 - }, - { - "epoch": 0.33, - "learning_rate": 7.727504354043111e-07, - "loss": 0.4948, - "step": 1049 - }, - { - "epoch": 0.33, - "learning_rate": 7.723334225788956e-07, - "loss": 0.5048, - "step": 1050 - }, - { - "epoch": 0.33, - "learning_rate": 7.719161402425379e-07, - "loss": 0.5187, - "step": 1051 - }, - { - "epoch": 0.33, - "learning_rate": 7.714985888081955e-07, - "loss": 0.5075, - "step": 1052 - }, - { - "epoch": 0.33, - "learning_rate": 7.710807686890926e-07, - "loss": 0.5216, - "step": 1053 - }, - { - "epoch": 0.33, - "learning_rate": 7.706626802987186e-07, - "loss": 0.4992, - "step": 1054 - }, - { - "epoch": 0.33, - "learning_rate": 7.70244324050829e-07, - "loss": 0.4892, - "step": 1055 - }, - { - "epoch": 0.33, - "learning_rate": 7.698257003594441e-07, - "loss": 0.5197, - "step": 1056 - }, - { - "epoch": 0.33, - "learning_rate": 7.694068096388491e-07, - "loss": 0.5135, - "step": 1057 - }, - { - "epoch": 0.33, - "learning_rate": 7.689876523035932e-07, - "loss": 0.5008, - "step": 1058 - }, - { - "epoch": 0.33, - "learning_rate": 7.685682287684896e-07, - "loss": 0.5122, - "step": 1059 - }, - { - "epoch": 0.33, - "learning_rate": 7.681485394486149e-07, - "loss": 0.5011, - "step": 1060 - }, - { - "epoch": 0.33, - "learning_rate": 7.677285847593087e-07, - "loss": 0.5138, - "step": 1061 - }, - { - "epoch": 0.33, - "learning_rate": 7.673083651161733e-07, - "loss": 0.4852, - "step": 1062 - }, - { - "epoch": 0.33, - "learning_rate": 7.66887880935073e-07, - "loss": 0.5013, - "step": 1063 - }, - { - "epoch": 0.33, - "learning_rate": 7.664671326321343e-07, - "loss": 0.501, - "step": 1064 - }, - { - "epoch": 0.33, - "learning_rate": 7.660461206237448e-07, - "loss": 0.5095, - "step": 1065 - }, - { - "epoch": 0.33, - "learning_rate": 7.656248453265531e-07, - "loss": 0.5018, - "step": 1066 - }, - { - "epoch": 0.33, - "learning_rate": 7.652033071574681e-07, - "loss": 0.5004, - "step": 1067 - }, - { - "epoch": 0.33, - "learning_rate": 7.647815065336595e-07, - "loss": 0.507, - "step": 1068 - }, - { - "epoch": 0.33, - "learning_rate": 7.643594438725563e-07, - "loss": 0.4941, - "step": 1069 - }, - { - "epoch": 0.33, - "learning_rate": 7.639371195918465e-07, - "loss": 0.498, - "step": 1070 - }, - { - "epoch": 0.33, - "learning_rate": 7.635145341094777e-07, - "loss": 0.503, - "step": 1071 - }, - { - "epoch": 0.33, - "learning_rate": 7.630916878436558e-07, - "loss": 0.5035, - "step": 1072 - }, - { - "epoch": 0.33, - "learning_rate": 7.626685812128444e-07, - "loss": 0.5051, - "step": 1073 - }, - { - "epoch": 0.33, - "learning_rate": 7.622452146357652e-07, - "loss": 0.5181, - "step": 1074 - }, - { - "epoch": 0.34, - "learning_rate": 7.61821588531397e-07, - "loss": 0.4943, - "step": 1075 - }, - { - "epoch": 0.34, - "learning_rate": 7.613977033189751e-07, - "loss": 0.5033, - "step": 1076 - }, - { - "epoch": 0.34, - "learning_rate": 7.609735594179918e-07, - "loss": 0.4908, - "step": 1077 - }, - { - "epoch": 0.34, - "learning_rate": 7.605491572481952e-07, - "loss": 0.4997, - "step": 1078 - }, - { - "epoch": 0.34, - "learning_rate": 7.601244972295891e-07, - "loss": 0.5005, - "step": 1079 - }, - { - "epoch": 0.34, - "learning_rate": 7.596995797824317e-07, - "loss": 0.5252, - "step": 1080 - }, - { - "epoch": 0.34, - "learning_rate": 7.592744053272371e-07, - "loss": 0.4925, - "step": 1081 - }, - { - "epoch": 0.34, - "learning_rate": 7.588489742847733e-07, - "loss": 0.5044, - "step": 1082 - }, - { - "epoch": 0.34, - "learning_rate": 7.584232870760615e-07, - "loss": 0.4974, - "step": 1083 - }, - { - "epoch": 0.34, - "learning_rate": 7.579973441223777e-07, - "loss": 0.5044, - "step": 1084 - }, - { - "epoch": 0.34, - "learning_rate": 7.5757114584525e-07, - "loss": 0.5056, - "step": 1085 - }, - { - "epoch": 0.34, - "learning_rate": 7.571446926664598e-07, - "loss": 0.5033, - "step": 1086 - }, - { - "epoch": 0.34, - "learning_rate": 7.567179850080403e-07, - "loss": 0.4963, - "step": 1087 - }, - { - "epoch": 0.34, - "learning_rate": 7.562910232922768e-07, - "loss": 0.4883, - "step": 1088 - }, - { - "epoch": 0.34, - "learning_rate": 7.55863807941706e-07, - "loss": 0.5223, - "step": 1089 - }, - { - "epoch": 0.34, - "learning_rate": 7.554363393791154e-07, - "loss": 0.5293, - "step": 1090 - }, - { - "epoch": 0.34, - "learning_rate": 7.550086180275436e-07, - "loss": 0.4994, - "step": 1091 - }, - { - "epoch": 0.34, - "learning_rate": 7.545806443102787e-07, - "loss": 0.4958, - "step": 1092 - }, - { - "epoch": 0.34, - "learning_rate": 7.541524186508587e-07, - "loss": 0.5009, - "step": 1093 - }, - { - "epoch": 0.34, - "learning_rate": 7.537239414730716e-07, - "loss": 0.4945, - "step": 1094 - }, - { - "epoch": 0.34, - "learning_rate": 7.532952132009536e-07, - "loss": 0.4887, - "step": 1095 - }, - { - "epoch": 0.34, - "learning_rate": 7.528662342587895e-07, - "loss": 0.5095, - "step": 1096 - }, - { - "epoch": 0.34, - "learning_rate": 7.524370050711124e-07, - "loss": 0.5232, - "step": 1097 - }, - { - "epoch": 0.34, - "learning_rate": 7.520075260627029e-07, - "loss": 0.494, - "step": 1098 - }, - { - "epoch": 0.34, - "learning_rate": 7.515777976585887e-07, - "loss": 0.49, - "step": 1099 - }, - { - "epoch": 0.34, - "learning_rate": 7.511478202840449e-07, - "loss": 0.5063, - "step": 1100 - }, - { - "epoch": 0.34, - "learning_rate": 7.50717594364592e-07, - "loss": 0.4909, - "step": 1101 - }, - { - "epoch": 0.34, - "learning_rate": 7.502871203259975e-07, - "loss": 0.5062, - "step": 1102 - }, - { - "epoch": 0.34, - "learning_rate": 7.498563985942738e-07, - "loss": 0.5012, - "step": 1103 - }, - { - "epoch": 0.34, - "learning_rate": 7.494254295956786e-07, - "loss": 0.5039, - "step": 1104 - }, - { - "epoch": 0.34, - "learning_rate": 7.489942137567143e-07, - "loss": 0.5112, - "step": 1105 - }, - { - "epoch": 0.34, - "learning_rate": 7.485627515041274e-07, - "loss": 0.4986, - "step": 1106 - }, - { - "epoch": 0.35, - "learning_rate": 7.481310432649089e-07, - "loss": 0.5033, - "step": 1107 - }, - { - "epoch": 0.35, - "learning_rate": 7.476990894662924e-07, - "loss": 0.5021, - "step": 1108 - }, - { - "epoch": 0.35, - "learning_rate": 7.472668905357552e-07, - "loss": 0.4922, - "step": 1109 - }, - { - "epoch": 0.35, - "learning_rate": 7.468344469010166e-07, - "loss": 0.4878, - "step": 1110 - }, - { - "epoch": 0.35, - "learning_rate": 7.464017589900386e-07, - "loss": 0.4898, - "step": 1111 - }, - { - "epoch": 0.35, - "learning_rate": 7.459688272310245e-07, - "loss": 0.4916, - "step": 1112 - }, - { - "epoch": 0.35, - "learning_rate": 7.455356520524193e-07, - "loss": 0.4994, - "step": 1113 - }, - { - "epoch": 0.35, - "learning_rate": 7.451022338829086e-07, - "loss": 0.5085, - "step": 1114 - }, - { - "epoch": 0.35, - "learning_rate": 7.446685731514189e-07, - "loss": 0.516, - "step": 1115 - }, - { - "epoch": 0.35, - "learning_rate": 7.442346702871159e-07, - "loss": 0.4917, - "step": 1116 - }, - { - "epoch": 0.35, - "learning_rate": 7.438005257194059e-07, - "loss": 0.4956, - "step": 1117 - }, - { - "epoch": 0.35, - "learning_rate": 7.433661398779338e-07, - "loss": 0.4677, - "step": 1118 - }, - { - "epoch": 0.35, - "learning_rate": 7.429315131925835e-07, - "loss": 0.492, - "step": 1119 - }, - { - "epoch": 0.35, - "learning_rate": 7.424966460934772e-07, - "loss": 0.484, - "step": 1120 - }, - { - "epoch": 0.35, - "learning_rate": 7.420615390109749e-07, - "loss": 0.514, - "step": 1121 - }, - { - "epoch": 0.35, - "learning_rate": 7.416261923756743e-07, - "loss": 0.4939, - "step": 1122 - }, - { - "epoch": 0.35, - "learning_rate": 7.411906066184102e-07, - "loss": 0.5352, - "step": 1123 - }, - { - "epoch": 0.35, - "learning_rate": 7.407547821702537e-07, - "loss": 0.4823, - "step": 1124 - }, - { - "epoch": 0.35, - "learning_rate": 7.403187194625122e-07, - "loss": 0.4878, - "step": 1125 - }, - { - "epoch": 0.35, - "learning_rate": 7.398824189267295e-07, - "loss": 0.4862, - "step": 1126 - }, - { - "epoch": 0.35, - "learning_rate": 7.39445880994684e-07, - "loss": 0.4936, - "step": 1127 - }, - { - "epoch": 0.35, - "learning_rate": 7.390091060983892e-07, - "loss": 0.4931, - "step": 1128 - }, - { - "epoch": 0.35, - "learning_rate": 7.385720946700935e-07, - "loss": 0.4918, - "step": 1129 - }, - { - "epoch": 0.35, - "learning_rate": 7.38134847142279e-07, - "loss": 0.4832, - "step": 1130 - }, - { - "epoch": 0.35, - "learning_rate": 7.376973639476616e-07, - "loss": 0.5068, - "step": 1131 - }, - { - "epoch": 0.35, - "learning_rate": 7.372596455191904e-07, - "loss": 0.5108, - "step": 1132 - }, - { - "epoch": 0.35, - "learning_rate": 7.368216922900472e-07, - "loss": 0.5141, - "step": 1133 - }, - { - "epoch": 0.35, - "learning_rate": 7.363835046936463e-07, - "loss": 0.5074, - "step": 1134 - }, - { - "epoch": 0.35, - "learning_rate": 7.359450831636339e-07, - "loss": 0.5102, - "step": 1135 - }, - { - "epoch": 0.35, - "learning_rate": 7.355064281338877e-07, - "loss": 0.4775, - "step": 1136 - }, - { - "epoch": 0.35, - "learning_rate": 7.350675400385164e-07, - "loss": 0.521, - "step": 1137 - }, - { - "epoch": 0.35, - "learning_rate": 7.346284193118597e-07, - "loss": 0.5039, - "step": 1138 - }, - { - "epoch": 0.36, - "learning_rate": 7.341890663884868e-07, - "loss": 0.4792, - "step": 1139 - }, - { - "epoch": 0.36, - "learning_rate": 7.337494817031977e-07, - "loss": 0.5056, - "step": 1140 - }, - { - "epoch": 0.36, - "learning_rate": 7.333096656910209e-07, - "loss": 0.5062, - "step": 1141 - }, - { - "epoch": 0.36, - "learning_rate": 7.32869618787214e-07, - "loss": 0.4807, - "step": 1142 - }, - { - "epoch": 0.36, - "learning_rate": 7.324293414272635e-07, - "loss": 0.5009, - "step": 1143 - }, - { - "epoch": 0.36, - "learning_rate": 7.319888340468835e-07, - "loss": 0.5036, - "step": 1144 - }, - { - "epoch": 0.36, - "learning_rate": 7.315480970820161e-07, - "loss": 0.4876, - "step": 1145 - }, - { - "epoch": 0.36, - "learning_rate": 7.311071309688301e-07, - "loss": 0.4957, - "step": 1146 - }, - { - "epoch": 0.36, - "learning_rate": 7.30665936143722e-07, - "loss": 0.4657, - "step": 1147 - }, - { - "epoch": 0.36, - "learning_rate": 7.302245130433133e-07, - "loss": 0.4866, - "step": 1148 - }, - { - "epoch": 0.36, - "learning_rate": 7.297828621044529e-07, - "loss": 0.4971, - "step": 1149 - }, - { - "epoch": 0.36, - "learning_rate": 7.293409837642142e-07, - "loss": 0.4851, - "step": 1150 - }, - { - "epoch": 0.36, - "learning_rate": 7.288988784598957e-07, - "loss": 0.4776, - "step": 1151 - }, - { - "epoch": 0.36, - "learning_rate": 7.284565466290207e-07, - "loss": 0.5234, - "step": 1152 - }, - { - "epoch": 0.36, - "learning_rate": 7.280139887093368e-07, - "loss": 0.4967, - "step": 1153 - }, - { - "epoch": 0.36, - "learning_rate": 7.275712051388152e-07, - "loss": 0.4892, - "step": 1154 - }, - { - "epoch": 0.36, - "learning_rate": 7.271281963556506e-07, - "loss": 0.5167, - "step": 1155 - }, - { - "epoch": 0.36, - "learning_rate": 7.266849627982602e-07, - "loss": 0.4947, - "step": 1156 - }, - { - "epoch": 0.36, - "learning_rate": 7.26241504905284e-07, - "loss": 0.5092, - "step": 1157 - }, - { - "epoch": 0.36, - "learning_rate": 7.257978231155836e-07, - "loss": 0.4858, - "step": 1158 - }, - { - "epoch": 0.36, - "learning_rate": 7.253539178682426e-07, - "loss": 0.4729, - "step": 1159 - }, - { - "epoch": 0.36, - "learning_rate": 7.249097896025658e-07, - "loss": 0.497, - "step": 1160 - }, - { - "epoch": 0.36, - "learning_rate": 7.244654387580781e-07, - "loss": 0.5015, - "step": 1161 - }, - { - "epoch": 0.36, - "learning_rate": 7.240208657745254e-07, - "loss": 0.511, - "step": 1162 - }, - { - "epoch": 0.36, - "learning_rate": 7.235760710918728e-07, - "loss": 0.4797, - "step": 1163 - }, - { - "epoch": 0.36, - "learning_rate": 7.231310551503055e-07, - "loss": 0.5119, - "step": 1164 - }, - { - "epoch": 0.36, - "learning_rate": 7.226858183902269e-07, - "loss": 0.4786, - "step": 1165 - }, - { - "epoch": 0.36, - "learning_rate": 7.222403612522595e-07, - "loss": 0.4857, - "step": 1166 - }, - { - "epoch": 0.36, - "learning_rate": 7.217946841772436e-07, - "loss": 0.5245, - "step": 1167 - }, - { - "epoch": 0.36, - "learning_rate": 7.213487876062374e-07, - "loss": 0.5032, - "step": 1168 - }, - { - "epoch": 0.36, - "learning_rate": 7.209026719805159e-07, - "loss": 0.4988, - "step": 1169 - }, - { - "epoch": 0.36, - "learning_rate": 7.204563377415713e-07, - "loss": 0.5107, - "step": 1170 - }, - { - "epoch": 0.37, - "learning_rate": 7.200097853311122e-07, - "loss": 0.4905, - "step": 1171 - }, - { - "epoch": 0.37, - "learning_rate": 7.195630151910625e-07, - "loss": 0.4603, - "step": 1172 - }, - { - "epoch": 0.37, - "learning_rate": 7.191160277635624e-07, - "loss": 0.512, - "step": 1173 - }, - { - "epoch": 0.37, - "learning_rate": 7.186688234909664e-07, - "loss": 0.5025, - "step": 1174 - }, - { - "epoch": 0.37, - "learning_rate": 7.182214028158442e-07, - "loss": 0.491, - "step": 1175 - }, - { - "epoch": 0.37, - "learning_rate": 7.17773766180979e-07, - "loss": 0.5026, - "step": 1176 - }, - { - "epoch": 0.37, - "learning_rate": 7.173259140293686e-07, - "loss": 0.4817, - "step": 1177 - }, - { - "epoch": 0.37, - "learning_rate": 7.168778468042232e-07, - "loss": 0.4949, - "step": 1178 - }, - { - "epoch": 0.37, - "learning_rate": 7.164295649489663e-07, - "loss": 0.5106, - "step": 1179 - }, - { - "epoch": 0.37, - "learning_rate": 7.15981068907234e-07, - "loss": 0.5002, - "step": 1180 - }, - { - "epoch": 0.37, - "learning_rate": 7.155323591228737e-07, - "loss": 0.5136, - "step": 1181 - }, - { - "epoch": 0.37, - "learning_rate": 7.15083436039945e-07, - "loss": 0.5036, - "step": 1182 - }, - { - "epoch": 0.37, - "learning_rate": 7.146343001027182e-07, - "loss": 0.4926, - "step": 1183 - }, - { - "epoch": 0.37, - "learning_rate": 7.141849517556744e-07, - "loss": 0.4912, - "step": 1184 - }, - { - "epoch": 0.37, - "learning_rate": 7.137353914435049e-07, - "loss": 0.521, - "step": 1185 - }, - { - "epoch": 0.37, - "learning_rate": 7.132856196111107e-07, - "loss": 0.4804, - "step": 1186 - }, - { - "epoch": 0.37, - "learning_rate": 7.128356367036021e-07, - "loss": 0.5125, - "step": 1187 - }, - { - "epoch": 0.37, - "learning_rate": 7.123854431662984e-07, - "loss": 0.4692, - "step": 1188 - }, - { - "epoch": 0.37, - "learning_rate": 7.119350394447274e-07, - "loss": 0.4846, - "step": 1189 - }, - { - "epoch": 0.37, - "learning_rate": 7.114844259846248e-07, - "loss": 0.4922, - "step": 1190 - }, - { - "epoch": 0.37, - "learning_rate": 7.110336032319338e-07, - "loss": 0.4848, - "step": 1191 - }, - { - "epoch": 0.37, - "learning_rate": 7.105825716328049e-07, - "loss": 0.4751, - "step": 1192 - }, - { - "epoch": 0.37, - "learning_rate": 7.101313316335951e-07, - "loss": 0.4739, - "step": 1193 - }, - { - "epoch": 0.37, - "learning_rate": 7.096798836808676e-07, - "loss": 0.5005, - "step": 1194 - }, - { - "epoch": 0.37, - "learning_rate": 7.092282282213917e-07, - "loss": 0.4947, - "step": 1195 - }, - { - "epoch": 0.37, - "learning_rate": 7.087763657021418e-07, - "loss": 0.4927, - "step": 1196 - }, - { - "epoch": 0.37, - "learning_rate": 7.083242965702973e-07, - "loss": 0.4734, - "step": 1197 - }, - { - "epoch": 0.37, - "learning_rate": 7.078720212732421e-07, - "loss": 0.5015, - "step": 1198 - }, - { - "epoch": 0.37, - "learning_rate": 7.074195402585639e-07, - "loss": 0.4783, - "step": 1199 - }, - { - "epoch": 0.37, - "learning_rate": 7.069668539740546e-07, - "loss": 0.4942, - "step": 1200 - }, - { - "epoch": 0.37, - "learning_rate": 7.065139628677084e-07, - "loss": 0.5141, - "step": 1201 - }, - { - "epoch": 0.37, - "learning_rate": 7.060608673877228e-07, - "loss": 0.4867, - "step": 1202 - }, - { - "epoch": 0.38, - "learning_rate": 7.056075679824972e-07, - "loss": 0.474, - "step": 1203 - }, - { - "epoch": 0.38, - "learning_rate": 7.051540651006332e-07, - "loss": 0.4859, - "step": 1204 - }, - { - "epoch": 0.38, - "learning_rate": 7.047003591909335e-07, - "loss": 0.4859, - "step": 1205 - }, - { - "epoch": 0.38, - "learning_rate": 7.042464507024018e-07, - "loss": 0.4885, - "step": 1206 - }, - { - "epoch": 0.38, - "learning_rate": 7.037923400842421e-07, - "loss": 0.5128, - "step": 1207 - }, - { - "epoch": 0.38, - "learning_rate": 7.033380277858585e-07, - "loss": 0.5019, - "step": 1208 - }, - { - "epoch": 0.38, - "learning_rate": 7.028835142568552e-07, - "loss": 0.4959, - "step": 1209 - }, - { - "epoch": 0.38, - "learning_rate": 7.024287999470347e-07, - "loss": 0.4849, - "step": 1210 - }, - { - "epoch": 0.38, - "learning_rate": 7.01973885306399e-07, - "loss": 0.4831, - "step": 1211 - }, - { - "epoch": 0.38, - "learning_rate": 7.015187707851475e-07, - "loss": 0.4991, - "step": 1212 - }, - { - "epoch": 0.38, - "learning_rate": 7.010634568336784e-07, - "loss": 0.4915, - "step": 1213 - }, - { - "epoch": 0.38, - "learning_rate": 7.006079439025864e-07, - "loss": 0.5097, - "step": 1214 - }, - { - "epoch": 0.38, - "learning_rate": 7.001522324426637e-07, - "loss": 0.4881, - "step": 1215 - }, - { - "epoch": 0.38, - "learning_rate": 6.996963229048985e-07, - "loss": 0.5081, - "step": 1216 - }, - { - "epoch": 0.38, - "learning_rate": 6.992402157404753e-07, - "loss": 0.4746, - "step": 1217 - }, - { - "epoch": 0.38, - "learning_rate": 6.987839114007743e-07, - "loss": 0.4855, - "step": 1218 - }, - { - "epoch": 0.38, - "learning_rate": 6.983274103373706e-07, - "loss": 0.4824, - "step": 1219 - }, - { - "epoch": 0.38, - "learning_rate": 6.978707130020342e-07, - "loss": 0.4824, - "step": 1220 - }, - { - "epoch": 0.38, - "learning_rate": 6.974138198467289e-07, - "loss": 0.4945, - "step": 1221 - }, - { - "epoch": 0.38, - "learning_rate": 6.969567313236128e-07, - "loss": 0.4835, - "step": 1222 - }, - { - "epoch": 0.38, - "learning_rate": 6.964994478850369e-07, - "loss": 0.4745, - "step": 1223 - }, - { - "epoch": 0.38, - "learning_rate": 6.960419699835457e-07, - "loss": 0.492, - "step": 1224 - }, - { - "epoch": 0.38, - "learning_rate": 6.955842980718754e-07, - "loss": 0.4994, - "step": 1225 - }, - { - "epoch": 0.38, - "learning_rate": 6.951264326029549e-07, - "loss": 0.4823, - "step": 1226 - }, - { - "epoch": 0.38, - "learning_rate": 6.94668374029904e-07, - "loss": 0.4738, - "step": 1227 - }, - { - "epoch": 0.38, - "learning_rate": 6.94210122806034e-07, - "loss": 0.497, - "step": 1228 - }, - { - "epoch": 0.38, - "learning_rate": 6.937516793848472e-07, - "loss": 0.5009, - "step": 1229 - }, - { - "epoch": 0.38, - "learning_rate": 6.93293044220035e-07, - "loss": 0.495, - "step": 1230 - }, - { - "epoch": 0.38, - "learning_rate": 6.928342177654798e-07, - "loss": 0.5117, - "step": 1231 - }, - { - "epoch": 0.38, - "learning_rate": 6.923752004752526e-07, - "loss": 0.4927, - "step": 1232 - }, - { - "epoch": 0.38, - "learning_rate": 6.919159928036133e-07, - "loss": 0.4831, - "step": 1233 - }, - { - "epoch": 0.38, - "learning_rate": 6.914565952050101e-07, - "loss": 0.4881, - "step": 1234 - }, - { - "epoch": 0.38, - "learning_rate": 6.909970081340799e-07, - "loss": 0.4725, - "step": 1235 - }, - { - "epoch": 0.39, - "learning_rate": 6.905372320456461e-07, - "loss": 0.4817, - "step": 1236 - }, - { - "epoch": 0.39, - "learning_rate": 6.900772673947201e-07, - "loss": 0.4828, - "step": 1237 - }, - { - "epoch": 0.39, - "learning_rate": 6.896171146364991e-07, - "loss": 0.4926, - "step": 1238 - }, - { - "epoch": 0.39, - "learning_rate": 6.891567742263669e-07, - "loss": 0.4723, - "step": 1239 - }, - { - "epoch": 0.39, - "learning_rate": 6.886962466198929e-07, - "loss": 0.5035, - "step": 1240 - }, - { - "epoch": 0.39, - "learning_rate": 6.882355322728318e-07, - "loss": 0.5111, - "step": 1241 - }, - { - "epoch": 0.39, - "learning_rate": 6.877746316411231e-07, - "loss": 0.4818, - "step": 1242 - }, - { - "epoch": 0.39, - "learning_rate": 6.873135451808906e-07, - "loss": 0.4726, - "step": 1243 - }, - { - "epoch": 0.39, - "learning_rate": 6.868522733484421e-07, - "loss": 0.5075, - "step": 1244 - }, - { - "epoch": 0.39, - "learning_rate": 6.863908166002688e-07, - "loss": 0.4921, - "step": 1245 - }, - { - "epoch": 0.39, - "learning_rate": 6.859291753930446e-07, - "loss": 0.4915, - "step": 1246 - }, - { - "epoch": 0.39, - "learning_rate": 6.854673501836265e-07, - "loss": 0.4878, - "step": 1247 - }, - { - "epoch": 0.39, - "learning_rate": 6.850053414290531e-07, - "loss": 0.4941, - "step": 1248 - }, - { - "epoch": 0.39, - "learning_rate": 6.845431495865452e-07, - "loss": 0.5056, - "step": 1249 - }, - { - "epoch": 0.39, - "learning_rate": 6.840807751135043e-07, - "loss": 0.5206, - "step": 1250 - }, - { - "epoch": 0.39, - "learning_rate": 6.836182184675127e-07, - "loss": 0.4833, - "step": 1251 - }, - { - "epoch": 0.39, - "learning_rate": 6.831554801063331e-07, - "loss": 0.4707, - "step": 1252 - }, - { - "epoch": 0.39, - "learning_rate": 6.826925604879081e-07, - "loss": 0.4943, - "step": 1253 - }, - { - "epoch": 0.39, - "learning_rate": 6.822294600703596e-07, - "loss": 0.5129, - "step": 1254 - }, - { - "epoch": 0.39, - "learning_rate": 6.817661793119884e-07, - "loss": 0.5104, - "step": 1255 - }, - { - "epoch": 0.39, - "learning_rate": 6.813027186712737e-07, - "loss": 0.4887, - "step": 1256 - }, - { - "epoch": 0.39, - "learning_rate": 6.808390786068728e-07, - "loss": 0.4576, - "step": 1257 - }, - { - "epoch": 0.39, - "learning_rate": 6.803752595776209e-07, - "loss": 0.4953, - "step": 1258 - }, - { - "epoch": 0.39, - "learning_rate": 6.799112620425293e-07, - "loss": 0.501, - "step": 1259 - }, - { - "epoch": 0.39, - "learning_rate": 6.794470864607874e-07, - "loss": 0.484, - "step": 1260 - }, - { - "epoch": 0.39, - "learning_rate": 6.789827332917595e-07, - "loss": 0.5016, - "step": 1261 - }, - { - "epoch": 0.39, - "learning_rate": 6.785182029949865e-07, - "loss": 0.4771, - "step": 1262 - }, - { - "epoch": 0.39, - "learning_rate": 6.78053496030184e-07, - "loss": 0.4809, - "step": 1263 - }, - { - "epoch": 0.39, - "learning_rate": 6.775886128572432e-07, - "loss": 0.486, - "step": 1264 - }, - { - "epoch": 0.39, - "learning_rate": 6.771235539362284e-07, - "loss": 0.504, - "step": 1265 - }, - { - "epoch": 0.39, - "learning_rate": 6.766583197273794e-07, - "loss": 0.5037, - "step": 1266 - }, - { - "epoch": 0.39, - "learning_rate": 6.761929106911081e-07, - "loss": 0.4708, - "step": 1267 - }, - { - "epoch": 0.4, - "learning_rate": 6.757273272880006e-07, - "loss": 0.488, - "step": 1268 - }, - { - "epoch": 0.4, - "learning_rate": 6.752615699788145e-07, - "loss": 0.4865, - "step": 1269 - }, - { - "epoch": 0.4, - "learning_rate": 6.747956392244801e-07, - "loss": 0.4893, - "step": 1270 - }, - { - "epoch": 0.4, - "learning_rate": 6.743295354860994e-07, - "loss": 0.4655, - "step": 1271 - }, - { - "epoch": 0.4, - "learning_rate": 6.73863259224945e-07, - "loss": 0.4759, - "step": 1272 - }, - { - "epoch": 0.4, - "learning_rate": 6.733968109024611e-07, - "loss": 0.495, - "step": 1273 - }, - { - "epoch": 0.4, - "learning_rate": 6.729301909802614e-07, - "loss": 0.5134, - "step": 1274 - }, - { - "epoch": 0.4, - "learning_rate": 6.724633999201299e-07, - "loss": 0.4974, - "step": 1275 - }, - { - "epoch": 0.4, - "learning_rate": 6.719964381840198e-07, - "loss": 0.4828, - "step": 1276 - }, - { - "epoch": 0.4, - "learning_rate": 6.715293062340532e-07, - "loss": 0.4783, - "step": 1277 - }, - { - "epoch": 0.4, - "learning_rate": 6.710620045325206e-07, - "loss": 0.4921, - "step": 1278 - }, - { - "epoch": 0.4, - "learning_rate": 6.705945335418806e-07, - "loss": 0.48, - "step": 1279 - }, - { - "epoch": 0.4, - "learning_rate": 6.701268937247594e-07, - "loss": 0.4759, - "step": 1280 - }, - { - "epoch": 0.4, - "learning_rate": 6.6965908554395e-07, - "loss": 0.4937, - "step": 1281 - }, - { - "epoch": 0.4, - "learning_rate": 6.691911094624124e-07, - "loss": 0.5037, - "step": 1282 - }, - { - "epoch": 0.4, - "learning_rate": 6.687229659432721e-07, - "loss": 0.4831, - "step": 1283 - }, - { - "epoch": 0.4, - "learning_rate": 6.682546554498214e-07, - "loss": 0.4839, - "step": 1284 - }, - { - "epoch": 0.4, - "learning_rate": 6.677861784455167e-07, - "loss": 0.4699, - "step": 1285 - }, - { - "epoch": 0.4, - "learning_rate": 6.673175353939798e-07, - "loss": 0.4994, - "step": 1286 - }, - { - "epoch": 0.4, - "learning_rate": 6.668487267589968e-07, - "loss": 0.489, - "step": 1287 - }, - { - "epoch": 0.4, - "learning_rate": 6.663797530045175e-07, - "loss": 0.5, - "step": 1288 - }, - { - "epoch": 0.4, - "learning_rate": 6.659106145946553e-07, - "loss": 0.4906, - "step": 1289 - }, - { - "epoch": 0.4, - "learning_rate": 6.654413119936862e-07, - "loss": 0.4693, - "step": 1290 - }, - { - "epoch": 0.4, - "learning_rate": 6.64971845666049e-07, - "loss": 0.4889, - "step": 1291 - }, - { - "epoch": 0.4, - "learning_rate": 6.645022160763446e-07, - "loss": 0.4857, - "step": 1292 - }, - { - "epoch": 0.4, - "learning_rate": 6.640324236893352e-07, - "loss": 0.4944, - "step": 1293 - }, - { - "epoch": 0.4, - "learning_rate": 6.635624689699444e-07, - "loss": 0.5062, - "step": 1294 - }, - { - "epoch": 0.4, - "learning_rate": 6.630923523832559e-07, - "loss": 0.4815, - "step": 1295 - }, - { - "epoch": 0.4, - "learning_rate": 6.626220743945143e-07, - "loss": 0.4938, - "step": 1296 - }, - { - "epoch": 0.4, - "learning_rate": 6.621516354691235e-07, - "loss": 0.4736, - "step": 1297 - }, - { - "epoch": 0.4, - "learning_rate": 6.616810360726467e-07, - "loss": 0.4928, - "step": 1298 - }, - { - "epoch": 0.4, - "learning_rate": 6.61210276670806e-07, - "loss": 0.4793, - "step": 1299 - }, - { - "epoch": 0.41, - "learning_rate": 6.60739357729482e-07, - "loss": 0.4968, - "step": 1300 - }, - { - "epoch": 0.41, - "learning_rate": 6.602682797147128e-07, - "loss": 0.4929, - "step": 1301 - }, - { - "epoch": 0.41, - "learning_rate": 6.597970430926942e-07, - "loss": 0.4816, - "step": 1302 - }, - { - "epoch": 0.41, - "learning_rate": 6.593256483297789e-07, - "loss": 0.4673, - "step": 1303 - }, - { - "epoch": 0.41, - "learning_rate": 6.588540958924761e-07, - "loss": 0.475, - "step": 1304 - }, - { - "epoch": 0.41, - "learning_rate": 6.58382386247451e-07, - "loss": 0.4861, - "step": 1305 - }, - { - "epoch": 0.41, - "learning_rate": 6.579105198615245e-07, - "loss": 0.4798, - "step": 1306 - }, - { - "epoch": 0.41, - "learning_rate": 6.574384972016726e-07, - "loss": 0.5113, - "step": 1307 - }, - { - "epoch": 0.41, - "learning_rate": 6.569663187350256e-07, - "loss": 0.4979, - "step": 1308 - }, - { - "epoch": 0.41, - "learning_rate": 6.564939849288686e-07, - "loss": 0.514, - "step": 1309 - }, - { - "epoch": 0.41, - "learning_rate": 6.560214962506397e-07, - "loss": 0.4955, - "step": 1310 - }, - { - "epoch": 0.41, - "learning_rate": 6.555488531679311e-07, - "loss": 0.4676, - "step": 1311 - }, - { - "epoch": 0.41, - "learning_rate": 6.550760561484874e-07, - "loss": 0.4925, - "step": 1312 - }, - { - "epoch": 0.41, - "learning_rate": 6.546031056602052e-07, - "loss": 0.4735, - "step": 1313 - }, - { - "epoch": 0.41, - "learning_rate": 6.541300021711334e-07, - "loss": 0.51, - "step": 1314 - }, - { - "epoch": 0.41, - "learning_rate": 6.536567461494723e-07, - "loss": 0.4702, - "step": 1315 - }, - { - "epoch": 0.41, - "learning_rate": 6.53183338063573e-07, - "loss": 0.4875, - "step": 1316 - }, - { - "epoch": 0.41, - "learning_rate": 6.527097783819372e-07, - "loss": 0.4764, - "step": 1317 - }, - { - "epoch": 0.41, - "learning_rate": 6.522360675732165e-07, - "loss": 0.4806, - "step": 1318 - }, - { - "epoch": 0.41, - "learning_rate": 6.517622061062124e-07, - "loss": 0.4995, - "step": 1319 - }, - { - "epoch": 0.41, - "learning_rate": 6.512881944498749e-07, - "loss": 0.4784, - "step": 1320 - }, - { - "epoch": 0.41, - "learning_rate": 6.50814033073303e-07, - "loss": 0.4779, - "step": 1321 - }, - { - "epoch": 0.41, - "learning_rate": 6.503397224457442e-07, - "loss": 0.4854, - "step": 1322 - }, - { - "epoch": 0.41, - "learning_rate": 6.498652630365929e-07, - "loss": 0.481, - "step": 1323 - }, - { - "epoch": 0.41, - "learning_rate": 6.493906553153914e-07, - "loss": 0.4792, - "step": 1324 - }, - { - "epoch": 0.41, - "learning_rate": 6.489158997518286e-07, - "loss": 0.5004, - "step": 1325 - }, - { - "epoch": 0.41, - "learning_rate": 6.484409968157396e-07, - "loss": 0.4717, - "step": 1326 - }, - { - "epoch": 0.41, - "learning_rate": 6.479659469771052e-07, - "loss": 0.501, - "step": 1327 - }, - { - "epoch": 0.41, - "learning_rate": 6.474907507060523e-07, - "loss": 0.4956, - "step": 1328 - }, - { - "epoch": 0.41, - "learning_rate": 6.470154084728519e-07, - "loss": 0.4722, - "step": 1329 - }, - { - "epoch": 0.41, - "learning_rate": 6.465399207479195e-07, - "loss": 0.4963, - "step": 1330 - }, - { - "epoch": 0.41, - "learning_rate": 6.460642880018153e-07, - "loss": 0.4962, - "step": 1331 - }, - { - "epoch": 0.42, - "learning_rate": 6.455885107052424e-07, - "loss": 0.4976, - "step": 1332 - }, - { - "epoch": 0.42, - "learning_rate": 6.451125893290472e-07, - "loss": 0.4947, - "step": 1333 - }, - { - "epoch": 0.42, - "learning_rate": 6.446365243442182e-07, - "loss": 0.4848, - "step": 1334 - }, - { - "epoch": 0.42, - "learning_rate": 6.441603162218869e-07, - "loss": 0.4797, - "step": 1335 - }, - { - "epoch": 0.42, - "learning_rate": 6.436839654333257e-07, - "loss": 0.4637, - "step": 1336 - }, - { - "epoch": 0.42, - "learning_rate": 6.432074724499486e-07, - "loss": 0.4905, - "step": 1337 - }, - { - "epoch": 0.42, - "learning_rate": 6.427308377433098e-07, - "loss": 0.4892, - "step": 1338 - }, - { - "epoch": 0.42, - "learning_rate": 6.422540617851047e-07, - "loss": 0.4875, - "step": 1339 - }, - { - "epoch": 0.42, - "learning_rate": 6.417771450471677e-07, - "loss": 0.4924, - "step": 1340 - }, - { - "epoch": 0.42, - "learning_rate": 6.413000880014726e-07, - "loss": 0.4933, - "step": 1341 - }, - { - "epoch": 0.42, - "learning_rate": 6.408228911201325e-07, - "loss": 0.5053, - "step": 1342 - }, - { - "epoch": 0.42, - "learning_rate": 6.403455548753981e-07, - "loss": 0.5046, - "step": 1343 - }, - { - "epoch": 0.42, - "learning_rate": 6.398680797396591e-07, - "loss": 0.4795, - "step": 1344 - }, - { - "epoch": 0.42, - "learning_rate": 6.393904661854416e-07, - "loss": 0.4835, - "step": 1345 - }, - { - "epoch": 0.42, - "learning_rate": 6.389127146854093e-07, - "loss": 0.4803, - "step": 1346 - }, - { - "epoch": 0.42, - "learning_rate": 6.384348257123621e-07, - "loss": 0.4653, - "step": 1347 - }, - { - "epoch": 0.42, - "learning_rate": 6.379567997392364e-07, - "loss": 0.4685, - "step": 1348 - }, - { - "epoch": 0.42, - "learning_rate": 6.374786372391038e-07, - "loss": 0.4968, - "step": 1349 - }, - { - "epoch": 0.42, - "learning_rate": 6.370003386851709e-07, - "loss": 0.4893, - "step": 1350 - }, - { - "epoch": 0.42, - "learning_rate": 6.365219045507793e-07, - "loss": 0.4765, - "step": 1351 - }, - { - "epoch": 0.42, - "learning_rate": 6.360433353094045e-07, - "loss": 0.5018, - "step": 1352 - }, - { - "epoch": 0.42, - "learning_rate": 6.35564631434656e-07, - "loss": 0.496, - "step": 1353 - }, - { - "epoch": 0.42, - "learning_rate": 6.350857934002761e-07, - "loss": 0.4981, - "step": 1354 - }, - { - "epoch": 0.42, - "learning_rate": 6.346068216801405e-07, - "loss": 0.4798, - "step": 1355 - }, - { - "epoch": 0.42, - "learning_rate": 6.341277167482564e-07, - "loss": 0.5303, - "step": 1356 - }, - { - "epoch": 0.42, - "learning_rate": 6.336484790787636e-07, - "loss": 0.4769, - "step": 1357 - }, - { - "epoch": 0.42, - "learning_rate": 6.331691091459328e-07, - "loss": 0.4882, - "step": 1358 - }, - { - "epoch": 0.42, - "learning_rate": 6.326896074241655e-07, - "loss": 0.479, - "step": 1359 - }, - { - "epoch": 0.42, - "learning_rate": 6.322099743879943e-07, - "loss": 0.4815, - "step": 1360 - }, - { - "epoch": 0.42, - "learning_rate": 6.317302105120809e-07, - "loss": 0.4836, - "step": 1361 - }, - { - "epoch": 0.42, - "learning_rate": 6.31250316271217e-07, - "loss": 0.4837, - "step": 1362 - }, - { - "epoch": 0.42, - "learning_rate": 6.307702921403231e-07, - "loss": 0.4857, - "step": 1363 - }, - { - "epoch": 0.43, - "learning_rate": 6.302901385944486e-07, - "loss": 0.4662, - "step": 1364 - }, - { - "epoch": 0.43, - "learning_rate": 6.298098561087703e-07, - "loss": 0.4683, - "step": 1365 - }, - { - "epoch": 0.43, - "learning_rate": 6.293294451585934e-07, - "loss": 0.4757, - "step": 1366 - }, - { - "epoch": 0.43, - "learning_rate": 6.288489062193496e-07, - "loss": 0.4892, - "step": 1367 - }, - { - "epoch": 0.43, - "learning_rate": 6.283682397665976e-07, - "loss": 0.4861, - "step": 1368 - }, - { - "epoch": 0.43, - "learning_rate": 6.278874462760221e-07, - "loss": 0.4931, - "step": 1369 - }, - { - "epoch": 0.43, - "learning_rate": 6.274065262234339e-07, - "loss": 0.494, - "step": 1370 - }, - { - "epoch": 0.43, - "learning_rate": 6.269254800847685e-07, - "loss": 0.476, - "step": 1371 - }, - { - "epoch": 0.43, - "learning_rate": 6.264443083360864e-07, - "loss": 0.4976, - "step": 1372 - }, - { - "epoch": 0.43, - "learning_rate": 6.259630114535726e-07, - "loss": 0.4911, - "step": 1373 - }, - { - "epoch": 0.43, - "learning_rate": 6.254815899135357e-07, - "loss": 0.4853, - "step": 1374 - }, - { - "epoch": 0.43, - "learning_rate": 6.25000044192408e-07, - "loss": 0.471, - "step": 1375 - }, - { - "epoch": 0.43, - "learning_rate": 6.245183747667439e-07, - "loss": 0.489, - "step": 1376 - }, - { - "epoch": 0.43, - "learning_rate": 6.240365821132214e-07, - "loss": 0.4803, - "step": 1377 - }, - { - "epoch": 0.43, - "learning_rate": 6.235546667086395e-07, - "loss": 0.4684, - "step": 1378 - }, - { - "epoch": 0.43, - "learning_rate": 6.230726290299189e-07, - "loss": 0.4711, - "step": 1379 - }, - { - "epoch": 0.43, - "learning_rate": 6.225904695541019e-07, - "loss": 0.4588, - "step": 1380 - }, - { - "epoch": 0.43, - "learning_rate": 6.221081887583504e-07, - "loss": 0.4957, - "step": 1381 - }, - { - "epoch": 0.43, - "learning_rate": 6.216257871199468e-07, - "loss": 0.4912, - "step": 1382 - }, - { - "epoch": 0.43, - "learning_rate": 6.211432651162935e-07, - "loss": 0.4899, - "step": 1383 - }, - { - "epoch": 0.43, - "learning_rate": 6.206606232249112e-07, - "loss": 0.4702, - "step": 1384 - }, - { - "epoch": 0.43, - "learning_rate": 6.201778619234397e-07, - "loss": 0.472, - "step": 1385 - }, - { - "epoch": 0.43, - "learning_rate": 6.19694981689637e-07, - "loss": 0.4901, - "step": 1386 - }, - { - "epoch": 0.43, - "learning_rate": 6.192119830013789e-07, - "loss": 0.4839, - "step": 1387 - }, - { - "epoch": 0.43, - "learning_rate": 6.18728866336658e-07, - "loss": 0.4813, - "step": 1388 - }, - { - "epoch": 0.43, - "learning_rate": 6.182456321735843e-07, - "loss": 0.4631, - "step": 1389 - }, - { - "epoch": 0.43, - "learning_rate": 6.177622809903831e-07, - "loss": 0.4878, - "step": 1390 - }, - { - "epoch": 0.43, - "learning_rate": 6.172788132653966e-07, - "loss": 0.4632, - "step": 1391 - }, - { - "epoch": 0.43, - "learning_rate": 6.167952294770817e-07, - "loss": 0.4967, - "step": 1392 - }, - { - "epoch": 0.43, - "learning_rate": 6.163115301040104e-07, - "loss": 0.4839, - "step": 1393 - }, - { - "epoch": 0.43, - "learning_rate": 6.158277156248685e-07, - "loss": 0.464, - "step": 1394 - }, - { - "epoch": 0.43, - "learning_rate": 6.15343786518457e-07, - "loss": 0.4685, - "step": 1395 - }, - { - "epoch": 0.44, - "learning_rate": 6.148597432636889e-07, - "loss": 0.4925, - "step": 1396 - }, - { - "epoch": 0.44, - "learning_rate": 6.143755863395909e-07, - "loss": 0.4771, - "step": 1397 - }, - { - "epoch": 0.44, - "learning_rate": 6.138913162253022e-07, - "loss": 0.4713, - "step": 1398 - }, - { - "epoch": 0.44, - "learning_rate": 6.134069334000739e-07, - "loss": 0.4948, - "step": 1399 - }, - { - "epoch": 0.44, - "learning_rate": 6.129224383432686e-07, - "loss": 0.4741, - "step": 1400 - }, - { - "epoch": 0.44, - "learning_rate": 6.124378315343601e-07, - "loss": 0.4795, - "step": 1401 - }, - { - "epoch": 0.44, - "learning_rate": 6.119531134529326e-07, - "loss": 0.5024, - "step": 1402 - }, - { - "epoch": 0.44, - "learning_rate": 6.114682845786808e-07, - "loss": 0.4706, - "step": 1403 - }, - { - "epoch": 0.44, - "learning_rate": 6.109833453914085e-07, - "loss": 0.4666, - "step": 1404 - }, - { - "epoch": 0.44, - "learning_rate": 6.104982963710289e-07, - "loss": 0.4556, - "step": 1405 - }, - { - "epoch": 0.44, - "learning_rate": 6.100131379975642e-07, - "loss": 0.5148, - "step": 1406 - }, - { - "epoch": 0.44, - "learning_rate": 6.095278707511444e-07, - "loss": 0.4729, - "step": 1407 - }, - { - "epoch": 0.44, - "learning_rate": 6.090424951120074e-07, - "loss": 0.4728, - "step": 1408 - }, - { - "epoch": 0.44, - "learning_rate": 6.085570115604986e-07, - "loss": 0.4675, - "step": 1409 - }, - { - "epoch": 0.44, - "learning_rate": 6.080714205770696e-07, - "loss": 0.488, - "step": 1410 - }, - { - "epoch": 0.44, - "learning_rate": 6.07585722642279e-07, - "loss": 0.4751, - "step": 1411 - }, - { - "epoch": 0.44, - "learning_rate": 6.070999182367907e-07, - "loss": 0.4551, - "step": 1412 - }, - { - "epoch": 0.44, - "learning_rate": 6.066140078413744e-07, - "loss": 0.472, - "step": 1413 - }, - { - "epoch": 0.44, - "learning_rate": 6.061279919369043e-07, - "loss": 0.4764, - "step": 1414 - }, - { - "epoch": 0.44, - "learning_rate": 6.056418710043595e-07, - "loss": 0.4701, - "step": 1415 - }, - { - "epoch": 0.44, - "learning_rate": 6.051556455248223e-07, - "loss": 0.4596, - "step": 1416 - }, - { - "epoch": 0.44, - "learning_rate": 6.046693159794793e-07, - "loss": 0.474, - "step": 1417 - }, - { - "epoch": 0.44, - "learning_rate": 6.041828828496196e-07, - "loss": 0.4855, - "step": 1418 - }, - { - "epoch": 0.44, - "learning_rate": 6.036963466166347e-07, - "loss": 0.468, - "step": 1419 - }, - { - "epoch": 0.44, - "learning_rate": 6.032097077620186e-07, - "loss": 0.4913, - "step": 1420 - }, - { - "epoch": 0.44, - "learning_rate": 6.027229667673666e-07, - "loss": 0.4724, - "step": 1421 - }, - { - "epoch": 0.44, - "learning_rate": 6.02236124114375e-07, - "loss": 0.4838, - "step": 1422 - }, - { - "epoch": 0.44, - "learning_rate": 6.017491802848405e-07, - "loss": 0.4856, - "step": 1423 - }, - { - "epoch": 0.44, - "learning_rate": 6.012621357606607e-07, - "loss": 0.4839, - "step": 1424 - }, - { - "epoch": 0.44, - "learning_rate": 6.007749910238321e-07, - "loss": 0.4849, - "step": 1425 - }, - { - "epoch": 0.44, - "learning_rate": 6.002877465564507e-07, - "loss": 0.4603, - "step": 1426 - }, - { - "epoch": 0.44, - "learning_rate": 5.99800402840711e-07, - "loss": 0.4895, - "step": 1427 - }, - { - "epoch": 0.45, - "learning_rate": 5.993129603589061e-07, - "loss": 0.4801, - "step": 1428 - }, - { - "epoch": 0.45, - "learning_rate": 5.988254195934265e-07, - "loss": 0.4851, - "step": 1429 - }, - { - "epoch": 0.45, - "learning_rate": 5.983377810267599e-07, - "loss": 0.4877, - "step": 1430 - }, - { - "epoch": 0.45, - "learning_rate": 5.978500451414913e-07, - "loss": 0.475, - "step": 1431 - }, - { - "epoch": 0.45, - "learning_rate": 5.973622124203013e-07, - "loss": 0.4983, - "step": 1432 - }, - { - "epoch": 0.45, - "learning_rate": 5.96874283345967e-07, - "loss": 0.4766, - "step": 1433 - }, - { - "epoch": 0.45, - "learning_rate": 5.963862584013601e-07, - "loss": 0.4742, - "step": 1434 - }, - { - "epoch": 0.45, - "learning_rate": 5.958981380694482e-07, - "loss": 0.4588, - "step": 1435 - }, - { - "epoch": 0.45, - "learning_rate": 5.954099228332923e-07, - "loss": 0.4914, - "step": 1436 - }, - { - "epoch": 0.45, - "learning_rate": 5.949216131760479e-07, - "loss": 0.4671, - "step": 1437 - }, - { - "epoch": 0.45, - "learning_rate": 5.944332095809638e-07, - "loss": 0.4611, - "step": 1438 - }, - { - "epoch": 0.45, - "learning_rate": 5.939447125313816e-07, - "loss": 0.4784, - "step": 1439 - }, - { - "epoch": 0.45, - "learning_rate": 5.934561225107358e-07, - "loss": 0.4886, - "step": 1440 - }, - { - "epoch": 0.45, - "learning_rate": 5.929674400025522e-07, - "loss": 0.4805, - "step": 1441 - }, - { - "epoch": 0.45, - "learning_rate": 5.924786654904492e-07, - "loss": 0.4524, - "step": 1442 - }, - { - "epoch": 0.45, - "learning_rate": 5.919897994581352e-07, - "loss": 0.4589, - "step": 1443 - }, - { - "epoch": 0.45, - "learning_rate": 5.915008423894097e-07, - "loss": 0.4619, - "step": 1444 - }, - { - "epoch": 0.45, - "learning_rate": 5.91011794768162e-07, - "loss": 0.45, - "step": 1445 - }, - { - "epoch": 0.45, - "learning_rate": 5.905226570783716e-07, - "loss": 0.4563, - "step": 1446 - }, - { - "epoch": 0.45, - "learning_rate": 5.900334298041067e-07, - "loss": 0.4784, - "step": 1447 - }, - { - "epoch": 0.45, - "learning_rate": 5.895441134295236e-07, - "loss": 0.4813, - "step": 1448 - }, - { - "epoch": 0.45, - "learning_rate": 5.890547084388679e-07, - "loss": 0.4707, - "step": 1449 - }, - { - "epoch": 0.45, - "learning_rate": 5.885652153164723e-07, - "loss": 0.4716, - "step": 1450 - }, - { - "epoch": 0.45, - "learning_rate": 5.880756345467566e-07, - "loss": 0.4613, - "step": 1451 - }, - { - "epoch": 0.45, - "learning_rate": 5.875859666142278e-07, - "loss": 0.4721, - "step": 1452 - }, - { - "epoch": 0.45, - "learning_rate": 5.870962120034787e-07, - "loss": 0.4869, - "step": 1453 - }, - { - "epoch": 0.45, - "learning_rate": 5.86606371199188e-07, - "loss": 0.4704, - "step": 1454 - }, - { - "epoch": 0.45, - "learning_rate": 5.8611644468612e-07, - "loss": 0.4665, - "step": 1455 - }, - { - "epoch": 0.45, - "learning_rate": 5.856264329491234e-07, - "loss": 0.5061, - "step": 1456 - }, - { - "epoch": 0.45, - "learning_rate": 5.851363364731314e-07, - "loss": 0.4597, - "step": 1457 - }, - { - "epoch": 0.45, - "learning_rate": 5.846461557431612e-07, - "loss": 0.4785, - "step": 1458 - }, - { - "epoch": 0.45, - "learning_rate": 5.841558912443133e-07, - "loss": 0.4789, - "step": 1459 - }, - { - "epoch": 0.46, - "learning_rate": 5.83665543461771e-07, - "loss": 0.47, - "step": 1460 - }, - { - "epoch": 0.46, - "learning_rate": 5.831751128808e-07, - "loss": 0.4737, - "step": 1461 - }, - { - "epoch": 0.46, - "learning_rate": 5.826845999867482e-07, - "loss": 0.5017, - "step": 1462 - }, - { - "epoch": 0.46, - "learning_rate": 5.821940052650446e-07, - "loss": 0.4745, - "step": 1463 - }, - { - "epoch": 0.46, - "learning_rate": 5.817033292011994e-07, - "loss": 0.4535, - "step": 1464 - }, - { - "epoch": 0.46, - "learning_rate": 5.812125722808034e-07, - "loss": 0.4849, - "step": 1465 - }, - { - "epoch": 0.46, - "learning_rate": 5.807217349895274e-07, - "loss": 0.4722, - "step": 1466 - }, - { - "epoch": 0.46, - "learning_rate": 5.802308178131212e-07, - "loss": 0.4624, - "step": 1467 - }, - { - "epoch": 0.46, - "learning_rate": 5.797398212374145e-07, - "loss": 0.482, - "step": 1468 - }, - { - "epoch": 0.46, - "learning_rate": 5.792487457483149e-07, - "loss": 0.4872, - "step": 1469 - }, - { - "epoch": 0.46, - "learning_rate": 5.787575918318082e-07, - "loss": 0.4841, - "step": 1470 - }, - { - "epoch": 0.46, - "learning_rate": 5.782663599739586e-07, - "loss": 0.4874, - "step": 1471 - }, - { - "epoch": 0.46, - "learning_rate": 5.777750506609062e-07, - "loss": 0.4622, - "step": 1472 - }, - { - "epoch": 0.46, - "learning_rate": 5.772836643788685e-07, - "loss": 0.4814, - "step": 1473 - }, - { - "epoch": 0.46, - "learning_rate": 5.767922016141392e-07, - "loss": 0.4672, - "step": 1474 - }, - { - "epoch": 0.46, - "learning_rate": 5.763006628530873e-07, - "loss": 0.5044, - "step": 1475 - }, - { - "epoch": 0.46, - "learning_rate": 5.758090485821574e-07, - "loss": 0.4792, - "step": 1476 - }, - { - "epoch": 0.46, - "learning_rate": 5.753173592878685e-07, - "loss": 0.4718, - "step": 1477 - }, - { - "epoch": 0.46, - "learning_rate": 5.748255954568141e-07, - "loss": 0.4591, - "step": 1478 - }, - { - "epoch": 0.46, - "learning_rate": 5.743337575756612e-07, - "loss": 0.4727, - "step": 1479 - }, - { - "epoch": 0.46, - "learning_rate": 5.738418461311505e-07, - "loss": 0.5038, - "step": 1480 - }, - { - "epoch": 0.46, - "learning_rate": 5.73349861610095e-07, - "loss": 0.4799, - "step": 1481 - }, - { - "epoch": 0.46, - "learning_rate": 5.728578044993804e-07, - "loss": 0.4511, - "step": 1482 - }, - { - "epoch": 0.46, - "learning_rate": 5.723656752859642e-07, - "loss": 0.4587, - "step": 1483 - }, - { - "epoch": 0.46, - "learning_rate": 5.71873474456875e-07, - "loss": 0.4888, - "step": 1484 - }, - { - "epoch": 0.46, - "learning_rate": 5.713812024992124e-07, - "loss": 0.4936, - "step": 1485 - }, - { - "epoch": 0.46, - "learning_rate": 5.708888599001465e-07, - "loss": 0.4787, - "step": 1486 - }, - { - "epoch": 0.46, - "learning_rate": 5.703964471469171e-07, - "loss": 0.4729, - "step": 1487 - }, - { - "epoch": 0.46, - "learning_rate": 5.699039647268338e-07, - "loss": 0.4629, - "step": 1488 - }, - { - "epoch": 0.46, - "learning_rate": 5.694114131272746e-07, - "loss": 0.4671, - "step": 1489 - }, - { - "epoch": 0.46, - "learning_rate": 5.689187928356866e-07, - "loss": 0.4771, - "step": 1490 - }, - { - "epoch": 0.46, - "learning_rate": 5.684261043395844e-07, - "loss": 0.4857, - "step": 1491 - }, - { - "epoch": 0.47, - "learning_rate": 5.679333481265501e-07, - "loss": 0.4881, - "step": 1492 - }, - { - "epoch": 0.47, - "learning_rate": 5.674405246842331e-07, - "loss": 0.4602, - "step": 1493 - }, - { - "epoch": 0.47, - "learning_rate": 5.669476345003491e-07, - "loss": 0.4779, - "step": 1494 - }, - { - "epoch": 0.47, - "learning_rate": 5.664546780626801e-07, - "loss": 0.4738, - "step": 1495 - }, - { - "epoch": 0.47, - "learning_rate": 5.659616558590734e-07, - "loss": 0.4943, - "step": 1496 - }, - { - "epoch": 0.47, - "learning_rate": 5.654685683774415e-07, - "loss": 0.473, - "step": 1497 - }, - { - "epoch": 0.47, - "learning_rate": 5.649754161057616e-07, - "loss": 0.4852, - "step": 1498 - }, - { - "epoch": 0.47, - "learning_rate": 5.644821995320746e-07, - "loss": 0.4776, - "step": 1499 - }, - { - "epoch": 0.47, - "learning_rate": 5.639889191444858e-07, - "loss": 0.4706, - "step": 1500 - }, - { - "epoch": 0.47, - "learning_rate": 5.634955754311627e-07, - "loss": 0.4837, - "step": 1501 - }, - { - "epoch": 0.47, - "learning_rate": 5.630021688803364e-07, - "loss": 0.4714, - "step": 1502 - }, - { - "epoch": 0.47, - "learning_rate": 5.625086999802994e-07, - "loss": 0.4506, - "step": 1503 - }, - { - "epoch": 0.47, - "learning_rate": 5.620151692194066e-07, - "loss": 0.4678, - "step": 1504 - }, - { - "epoch": 0.47, - "learning_rate": 5.615215770860735e-07, - "loss": 0.4799, - "step": 1505 - }, - { - "epoch": 0.47, - "learning_rate": 5.610279240687768e-07, - "loss": 0.474, - "step": 1506 - }, - { - "epoch": 0.47, - "learning_rate": 5.605342106560531e-07, - "loss": 0.4634, - "step": 1507 - }, - { - "epoch": 0.47, - "learning_rate": 5.600404373364992e-07, - "loss": 0.4644, - "step": 1508 - }, - { - "epoch": 0.47, - "learning_rate": 5.595466045987705e-07, - "loss": 0.4753, - "step": 1509 - }, - { - "epoch": 0.47, - "learning_rate": 5.590527129315821e-07, - "loss": 0.4798, - "step": 1510 - }, - { - "epoch": 0.47, - "learning_rate": 5.585587628237068e-07, - "loss": 0.4784, - "step": 1511 - }, - { - "epoch": 0.47, - "learning_rate": 5.580647547639751e-07, - "loss": 0.4813, - "step": 1512 - }, - { - "epoch": 0.47, - "learning_rate": 5.575706892412754e-07, - "loss": 0.4949, - "step": 1513 - }, - { - "epoch": 0.47, - "learning_rate": 5.570765667445526e-07, - "loss": 0.5095, - "step": 1514 - }, - { - "epoch": 0.47, - "learning_rate": 5.565823877628081e-07, - "loss": 0.4738, - "step": 1515 - }, - { - "epoch": 0.47, - "learning_rate": 5.560881527850991e-07, - "loss": 0.4566, - "step": 1516 - }, - { - "epoch": 0.47, - "learning_rate": 5.555938623005386e-07, - "loss": 0.4649, - "step": 1517 - }, - { - "epoch": 0.47, - "learning_rate": 5.550995167982939e-07, - "loss": 0.4726, - "step": 1518 - }, - { - "epoch": 0.47, - "learning_rate": 5.546051167675871e-07, - "loss": 0.4591, - "step": 1519 - }, - { - "epoch": 0.47, - "learning_rate": 5.541106626976944e-07, - "loss": 0.4516, - "step": 1520 - }, - { - "epoch": 0.47, - "learning_rate": 5.536161550779454e-07, - "loss": 0.4898, - "step": 1521 - }, - { - "epoch": 0.47, - "learning_rate": 5.531215943977222e-07, - "loss": 0.4626, - "step": 1522 - }, - { - "epoch": 0.47, - "learning_rate": 5.526269811464602e-07, - "loss": 0.4655, - "step": 1523 - }, - { - "epoch": 0.48, - "learning_rate": 5.521323158136464e-07, - "loss": 0.4749, - "step": 1524 - }, - { - "epoch": 0.48, - "learning_rate": 5.516375988888192e-07, - "loss": 0.4827, - "step": 1525 - }, - { - "epoch": 0.48, - "learning_rate": 5.511428308615683e-07, - "loss": 0.4976, - "step": 1526 - }, - { - "epoch": 0.48, - "learning_rate": 5.50648012221534e-07, - "loss": 0.4605, - "step": 1527 - }, - { - "epoch": 0.48, - "learning_rate": 5.501531434584068e-07, - "loss": 0.4713, - "step": 1528 - }, - { - "epoch": 0.48, - "learning_rate": 5.496582250619261e-07, - "loss": 0.4729, - "step": 1529 - }, - { - "epoch": 0.48, - "learning_rate": 5.491632575218814e-07, - "loss": 0.5031, - "step": 1530 - }, - { - "epoch": 0.48, - "learning_rate": 5.486682413281103e-07, - "loss": 0.4744, - "step": 1531 - }, - { - "epoch": 0.48, - "learning_rate": 5.481731769704984e-07, - "loss": 0.4544, - "step": 1532 - }, - { - "epoch": 0.48, - "learning_rate": 5.476780649389794e-07, - "loss": 0.4875, - "step": 1533 - }, - { - "epoch": 0.48, - "learning_rate": 5.471829057235339e-07, - "loss": 0.4614, - "step": 1534 - }, - { - "epoch": 0.48, - "learning_rate": 5.466876998141891e-07, - "loss": 0.4766, - "step": 1535 - }, - { - "epoch": 0.48, - "learning_rate": 5.461924477010187e-07, - "loss": 0.4733, - "step": 1536 - }, - { - "epoch": 0.48, - "learning_rate": 5.45697149874142e-07, - "loss": 0.4841, - "step": 1537 - }, - { - "epoch": 0.48, - "learning_rate": 5.452018068237234e-07, - "loss": 0.472, - "step": 1538 - }, - { - "epoch": 0.48, - "learning_rate": 5.447064190399723e-07, - "loss": 0.4737, - "step": 1539 - }, - { - "epoch": 0.48, - "learning_rate": 5.442109870131424e-07, - "loss": 0.477, - "step": 1540 - }, - { - "epoch": 0.48, - "learning_rate": 5.437155112335306e-07, - "loss": 0.4552, - "step": 1541 - }, - { - "epoch": 0.48, - "learning_rate": 5.432199921914781e-07, - "loss": 0.4739, - "step": 1542 - }, - { - "epoch": 0.48, - "learning_rate": 5.427244303773679e-07, - "loss": 0.486, - "step": 1543 - }, - { - "epoch": 0.48, - "learning_rate": 5.42228826281626e-07, - "loss": 0.4972, - "step": 1544 - }, - { - "epoch": 0.48, - "learning_rate": 5.4173318039472e-07, - "loss": 0.5005, - "step": 1545 - }, - { - "epoch": 0.48, - "learning_rate": 5.412374932071588e-07, - "loss": 0.4645, - "step": 1546 - }, - { - "epoch": 0.48, - "learning_rate": 5.407417652094925e-07, - "loss": 0.4675, - "step": 1547 - }, - { - "epoch": 0.48, - "learning_rate": 5.40245996892311e-07, - "loss": 0.4832, - "step": 1548 - }, - { - "epoch": 0.48, - "learning_rate": 5.397501887462445e-07, - "loss": 0.4805, - "step": 1549 - }, - { - "epoch": 0.48, - "learning_rate": 5.392543412619628e-07, - "loss": 0.4707, - "step": 1550 - }, - { - "epoch": 0.48, - "learning_rate": 5.387584549301742e-07, - "loss": 0.4781, - "step": 1551 - }, - { - "epoch": 0.48, - "learning_rate": 5.382625302416257e-07, - "loss": 0.4703, - "step": 1552 - }, - { - "epoch": 0.48, - "learning_rate": 5.377665676871021e-07, - "loss": 0.4825, - "step": 1553 - }, - { - "epoch": 0.48, - "learning_rate": 5.37270567757426e-07, - "loss": 0.4627, - "step": 1554 - }, - { - "epoch": 0.48, - "learning_rate": 5.367745309434565e-07, - "loss": 0.4679, - "step": 1555 - }, - { - "epoch": 0.49, - "learning_rate": 5.362784577360895e-07, - "loss": 0.4546, - "step": 1556 - }, - { - "epoch": 0.49, - "learning_rate": 5.35782348626257e-07, - "loss": 0.477, - "step": 1557 - }, - { - "epoch": 0.49, - "learning_rate": 5.352862041049263e-07, - "loss": 0.4613, - "step": 1558 - }, - { - "epoch": 0.49, - "learning_rate": 5.347900246630999e-07, - "loss": 0.4801, - "step": 1559 - }, - { - "epoch": 0.49, - "learning_rate": 5.34293810791815e-07, - "loss": 0.4611, - "step": 1560 - }, - { - "epoch": 0.49, - "learning_rate": 5.337975629821421e-07, - "loss": 0.4773, - "step": 1561 - }, - { - "epoch": 0.49, - "learning_rate": 5.333012817251867e-07, - "loss": 0.4533, - "step": 1562 - }, - { - "epoch": 0.49, - "learning_rate": 5.328049675120857e-07, - "loss": 0.4745, - "step": 1563 - }, - { - "epoch": 0.49, - "learning_rate": 5.3230862083401e-07, - "loss": 0.4743, - "step": 1564 - }, - { - "epoch": 0.49, - "learning_rate": 5.31812242182162e-07, - "loss": 0.4554, - "step": 1565 - }, - { - "epoch": 0.49, - "learning_rate": 5.31315832047776e-07, - "loss": 0.4759, - "step": 1566 - }, - { - "epoch": 0.49, - "learning_rate": 5.30819390922117e-07, - "loss": 0.4713, - "step": 1567 - }, - { - "epoch": 0.49, - "learning_rate": 5.303229192964812e-07, - "loss": 0.462, - "step": 1568 - }, - { - "epoch": 0.49, - "learning_rate": 5.298264176621947e-07, - "loss": 0.4805, - "step": 1569 - }, - { - "epoch": 0.49, - "learning_rate": 5.293298865106133e-07, - "loss": 0.4672, - "step": 1570 - }, - { - "epoch": 0.49, - "learning_rate": 5.288333263331223e-07, - "loss": 0.4777, - "step": 1571 - }, - { - "epoch": 0.49, - "learning_rate": 5.283367376211352e-07, - "loss": 0.4797, - "step": 1572 - }, - { - "epoch": 0.49, - "learning_rate": 5.278401208660943e-07, - "loss": 0.4911, - "step": 1573 - }, - { - "epoch": 0.49, - "learning_rate": 5.273434765594693e-07, - "loss": 0.4852, - "step": 1574 - }, - { - "epoch": 0.49, - "learning_rate": 5.26846805192757e-07, - "loss": 0.4769, - "step": 1575 - }, - { - "epoch": 0.49, - "learning_rate": 5.263501072574815e-07, - "loss": 0.4696, - "step": 1576 - }, - { - "epoch": 0.49, - "learning_rate": 5.258533832451931e-07, - "loss": 0.4718, - "step": 1577 - }, - { - "epoch": 0.49, - "learning_rate": 5.253566336474672e-07, - "loss": 0.4858, - "step": 1578 - }, - { - "epoch": 0.49, - "learning_rate": 5.248598589559054e-07, - "loss": 0.4942, - "step": 1579 - }, - { - "epoch": 0.49, - "learning_rate": 5.243630596621339e-07, - "loss": 0.452, - "step": 1580 - }, - { - "epoch": 0.49, - "learning_rate": 5.238662362578028e-07, - "loss": 0.4678, - "step": 1581 - }, - { - "epoch": 0.49, - "learning_rate": 5.233693892345868e-07, - "loss": 0.4691, - "step": 1582 - }, - { - "epoch": 0.49, - "learning_rate": 5.228725190841832e-07, - "loss": 0.4823, - "step": 1583 - }, - { - "epoch": 0.49, - "learning_rate": 5.223756262983129e-07, - "loss": 0.4718, - "step": 1584 - }, - { - "epoch": 0.49, - "learning_rate": 5.218787113687185e-07, - "loss": 0.4698, - "step": 1585 - }, - { - "epoch": 0.49, - "learning_rate": 5.213817747871651e-07, - "loss": 0.4814, - "step": 1586 - }, - { - "epoch": 0.49, - "learning_rate": 5.208848170454389e-07, - "loss": 0.4724, - "step": 1587 - }, - { - "epoch": 0.5, - "learning_rate": 5.203878386353474e-07, - "loss": 0.4579, - "step": 1588 - }, - { - "epoch": 0.5, - "learning_rate": 5.198908400487178e-07, - "loss": 0.4684, - "step": 1589 - }, - { - "epoch": 0.5, - "learning_rate": 5.193938217773981e-07, - "loss": 0.4614, - "step": 1590 - }, - { - "epoch": 0.5, - "learning_rate": 5.188967843132554e-07, - "loss": 0.4627, - "step": 1591 - }, - { - "epoch": 0.5, - "learning_rate": 5.183997281481759e-07, - "loss": 0.4666, - "step": 1592 - }, - { - "epoch": 0.5, - "learning_rate": 5.179026537740642e-07, - "loss": 0.4709, - "step": 1593 - }, - { - "epoch": 0.5, - "learning_rate": 5.174055616828428e-07, - "loss": 0.5113, - "step": 1594 - }, - { - "epoch": 0.5, - "learning_rate": 5.169084523664521e-07, - "loss": 0.4512, - "step": 1595 - }, - { - "epoch": 0.5, - "learning_rate": 5.164113263168491e-07, - "loss": 0.4778, - "step": 1596 - }, - { - "epoch": 0.5, - "learning_rate": 5.15914184026008e-07, - "loss": 0.479, - "step": 1597 - }, - { - "epoch": 0.5, - "learning_rate": 5.154170259859185e-07, - "loss": 0.4732, - "step": 1598 - }, - { - "epoch": 0.5, - "learning_rate": 5.149198526885859e-07, - "loss": 0.4619, - "step": 1599 - }, - { - "epoch": 0.5, - "learning_rate": 5.144226646260309e-07, - "loss": 0.4891, - "step": 1600 - }, - { - "epoch": 0.5, - "learning_rate": 5.139254622902887e-07, - "loss": 0.4589, - "step": 1601 - }, - { - "epoch": 0.5, - "learning_rate": 5.134282461734085e-07, - "loss": 0.4829, - "step": 1602 - }, - { - "epoch": 0.5, - "learning_rate": 5.129310167674532e-07, - "loss": 0.463, - "step": 1603 - }, - { - "epoch": 0.5, - "learning_rate": 5.124337745644993e-07, - "loss": 0.5055, - "step": 1604 - }, - { - "epoch": 0.5, - "learning_rate": 5.119365200566349e-07, - "loss": 0.4839, - "step": 1605 - }, - { - "epoch": 0.5, - "learning_rate": 5.114392537359614e-07, - "loss": 0.499, - "step": 1606 - }, - { - "epoch": 0.5, - "learning_rate": 5.109419760945912e-07, - "loss": 0.4513, - "step": 1607 - }, - { - "epoch": 0.5, - "learning_rate": 5.104446876246484e-07, - "loss": 0.4511, - "step": 1608 - }, - { - "epoch": 0.5, - "learning_rate": 5.099473888182672e-07, - "loss": 0.4869, - "step": 1609 - }, - { - "epoch": 0.5, - "learning_rate": 5.094500801675927e-07, - "loss": 0.4697, - "step": 1610 - }, - { - "epoch": 0.5, - "learning_rate": 5.089527621647792e-07, - "loss": 0.4942, - "step": 1611 - }, - { - "epoch": 0.5, - "learning_rate": 5.084554353019904e-07, - "loss": 0.4581, - "step": 1612 - }, - { - "epoch": 0.5, - "learning_rate": 5.079581000713992e-07, - "loss": 0.4669, - "step": 1613 - }, - { - "epoch": 0.5, - "learning_rate": 5.074607569651858e-07, - "loss": 0.4584, - "step": 1614 - }, - { - "epoch": 0.5, - "learning_rate": 5.069634064755394e-07, - "loss": 0.4684, - "step": 1615 - }, - { - "epoch": 0.5, - "learning_rate": 5.064660490946557e-07, - "loss": 0.4682, - "step": 1616 - }, - { - "epoch": 0.5, - "learning_rate": 5.059686853147375e-07, - "loss": 0.4764, - "step": 1617 - }, - { - "epoch": 0.5, - "learning_rate": 5.054713156279936e-07, - "loss": 0.4856, - "step": 1618 - }, - { - "epoch": 0.5, - "learning_rate": 5.049739405266392e-07, - "loss": 0.4685, - "step": 1619 - }, - { - "epoch": 0.5, - "learning_rate": 5.044765605028946e-07, - "loss": 0.4836, - "step": 1620 - }, - { - "epoch": 0.51, - "learning_rate": 5.039791760489847e-07, - "loss": 0.4841, - "step": 1621 - }, - { - "epoch": 0.51, - "learning_rate": 5.034817876571393e-07, - "loss": 0.4662, - "step": 1622 - }, - { - "epoch": 0.51, - "learning_rate": 5.029843958195918e-07, - "loss": 0.4981, - "step": 1623 - }, - { - "epoch": 0.51, - "learning_rate": 5.024870010285788e-07, - "loss": 0.4535, - "step": 1624 - }, - { - "epoch": 0.51, - "learning_rate": 5.019896037763401e-07, - "loss": 0.4709, - "step": 1625 - }, - { - "epoch": 0.51, - "learning_rate": 5.014922045551182e-07, - "loss": 0.4988, - "step": 1626 - }, - { - "epoch": 0.51, - "learning_rate": 5.009948038571568e-07, - "loss": 0.4564, - "step": 1627 - }, - { - "epoch": 0.51, - "learning_rate": 5.00497402174702e-07, - "loss": 0.4629, - "step": 1628 - }, - { - "epoch": 0.51, - "learning_rate": 5e-07, - "loss": 0.4814, - "step": 1629 - }, - { - "epoch": 0.51, - "learning_rate": 4.995025978252981e-07, - "loss": 0.4618, - "step": 1630 - }, - { - "epoch": 0.51, - "learning_rate": 4.99005196142843e-07, - "loss": 0.459, - "step": 1631 - }, - { - "epoch": 0.51, - "learning_rate": 4.985077954448818e-07, - "loss": 0.4644, - "step": 1632 - }, - { - "epoch": 0.51, - "learning_rate": 4.980103962236598e-07, - "loss": 0.4607, - "step": 1633 - }, - { - "epoch": 0.51, - "learning_rate": 4.975129989714214e-07, - "loss": 0.4906, - "step": 1634 - }, - { - "epoch": 0.51, - "learning_rate": 4.970156041804083e-07, - "loss": 0.4733, - "step": 1635 - }, - { - "epoch": 0.51, - "learning_rate": 4.965182123428608e-07, - "loss": 0.4652, - "step": 1636 - }, - { - "epoch": 0.51, - "learning_rate": 4.960208239510153e-07, - "loss": 0.4463, - "step": 1637 - }, - { - "epoch": 0.51, - "learning_rate": 4.955234394971056e-07, - "loss": 0.4585, - "step": 1638 - }, - { - "epoch": 0.51, - "learning_rate": 4.950260594733608e-07, - "loss": 0.4618, - "step": 1639 - }, - { - "epoch": 0.51, - "learning_rate": 4.945286843720063e-07, - "loss": 0.459, - "step": 1640 - }, - { - "epoch": 0.51, - "learning_rate": 4.940313146852626e-07, - "loss": 0.4714, - "step": 1641 - }, - { - "epoch": 0.51, - "learning_rate": 4.935339509053442e-07, - "loss": 0.4681, - "step": 1642 - }, - { - "epoch": 0.51, - "learning_rate": 4.930365935244605e-07, - "loss": 0.467, - "step": 1643 - }, - { - "epoch": 0.51, - "learning_rate": 4.925392430348141e-07, - "loss": 0.4928, - "step": 1644 - }, - { - "epoch": 0.51, - "learning_rate": 4.92041899928601e-07, - "loss": 0.4711, - "step": 1645 - }, - { - "epoch": 0.51, - "learning_rate": 4.915445646980094e-07, - "loss": 0.4788, - "step": 1646 - }, - { - "epoch": 0.51, - "learning_rate": 4.910472378352208e-07, - "loss": 0.4589, - "step": 1647 - }, - { - "epoch": 0.51, - "learning_rate": 4.905499198324073e-07, - "loss": 0.4824, - "step": 1648 - }, - { - "epoch": 0.51, - "learning_rate": 4.900526111817327e-07, - "loss": 0.4629, - "step": 1649 - }, - { - "epoch": 0.51, - "learning_rate": 4.895553123753516e-07, - "loss": 0.4887, - "step": 1650 - }, - { - "epoch": 0.51, - "learning_rate": 4.890580239054088e-07, - "loss": 0.4684, - "step": 1651 - }, - { - "epoch": 0.51, - "learning_rate": 4.885607462640387e-07, - "loss": 0.4629, - "step": 1652 - }, - { - "epoch": 0.52, - "learning_rate": 4.880634799433651e-07, - "loss": 0.4964, - "step": 1653 - }, - { - "epoch": 0.52, - "learning_rate": 4.875662254355008e-07, - "loss": 0.4453, - "step": 1654 - }, - { - "epoch": 0.52, - "learning_rate": 4.870689832325468e-07, - "loss": 0.482, - "step": 1655 - }, - { - "epoch": 0.52, - "learning_rate": 4.865717538265917e-07, - "loss": 0.4624, - "step": 1656 - }, - { - "epoch": 0.52, - "learning_rate": 4.860745377097113e-07, - "loss": 0.4745, - "step": 1657 - }, - { - "epoch": 0.52, - "learning_rate": 4.855773353739691e-07, - "loss": 0.476, - "step": 1658 - }, - { - "epoch": 0.52, - "learning_rate": 4.850801473114142e-07, - "loss": 0.4602, - "step": 1659 - }, - { - "epoch": 0.52, - "learning_rate": 4.845829740140815e-07, - "loss": 0.4623, - "step": 1660 - }, - { - "epoch": 0.52, - "learning_rate": 4.840858159739919e-07, - "loss": 0.5039, - "step": 1661 - }, - { - "epoch": 0.52, - "learning_rate": 4.835886736831508e-07, - "loss": 0.4671, - "step": 1662 - }, - { - "epoch": 0.52, - "learning_rate": 4.830915476335481e-07, - "loss": 0.4655, - "step": 1663 - }, - { - "epoch": 0.52, - "learning_rate": 4.825944383171572e-07, - "loss": 0.4666, - "step": 1664 - }, - { - "epoch": 0.52, - "learning_rate": 4.820973462259359e-07, - "loss": 0.4559, - "step": 1665 - }, - { - "epoch": 0.52, - "learning_rate": 4.816002718518241e-07, - "loss": 0.4883, - "step": 1666 - }, - { - "epoch": 0.52, - "learning_rate": 4.811032156867446e-07, - "loss": 0.4665, - "step": 1667 - }, - { - "epoch": 0.52, - "learning_rate": 4.806061782226018e-07, - "loss": 0.4709, - "step": 1668 - }, - { - "epoch": 0.52, - "learning_rate": 4.801091599512822e-07, - "loss": 0.4718, - "step": 1669 - }, - { - "epoch": 0.52, - "learning_rate": 4.796121613646528e-07, - "loss": 0.4737, - "step": 1670 - }, - { - "epoch": 0.52, - "learning_rate": 4.79115182954561e-07, - "loss": 0.4962, - "step": 1671 - }, - { - "epoch": 0.52, - "learning_rate": 4.786182252128348e-07, - "loss": 0.4612, - "step": 1672 - }, - { - "epoch": 0.52, - "learning_rate": 4.781212886312814e-07, - "loss": 0.4503, - "step": 1673 - }, - { - "epoch": 0.52, - "learning_rate": 4.776243737016872e-07, - "loss": 0.4574, - "step": 1674 - }, - { - "epoch": 0.52, - "learning_rate": 4.771274809158167e-07, - "loss": 0.4921, - "step": 1675 - }, - { - "epoch": 0.52, - "learning_rate": 4.766306107654132e-07, - "loss": 0.479, - "step": 1676 - }, - { - "epoch": 0.52, - "learning_rate": 4.7613376374219717e-07, - "loss": 0.4814, - "step": 1677 - }, - { - "epoch": 0.52, - "learning_rate": 4.7563694033786626e-07, - "loss": 0.4666, - "step": 1678 - }, - { - "epoch": 0.52, - "learning_rate": 4.7514014104409456e-07, - "loss": 0.4798, - "step": 1679 - }, - { - "epoch": 0.52, - "learning_rate": 4.746433663525329e-07, - "loss": 0.4624, - "step": 1680 - }, - { - "epoch": 0.52, - "learning_rate": 4.7414661675480715e-07, - "loss": 0.4807, - "step": 1681 - }, - { - "epoch": 0.52, - "learning_rate": 4.7364989274251844e-07, - "loss": 0.475, - "step": 1682 - }, - { - "epoch": 0.52, - "learning_rate": 4.73153194807243e-07, - "loss": 0.4698, - "step": 1683 - }, - { - "epoch": 0.52, - "learning_rate": 4.7265652344053086e-07, - "loss": 0.4757, - "step": 1684 - }, - { - "epoch": 0.53, - "learning_rate": 4.7215987913390574e-07, - "loss": 0.4905, - "step": 1685 - }, - { - "epoch": 0.53, - "learning_rate": 4.716632623788647e-07, - "loss": 0.4789, - "step": 1686 - }, - { - "epoch": 0.53, - "learning_rate": 4.711666736668777e-07, - "loss": 0.4678, - "step": 1687 - }, - { - "epoch": 0.53, - "learning_rate": 4.706701134893867e-07, - "loss": 0.4602, - "step": 1688 - }, - { - "epoch": 0.53, - "learning_rate": 4.701735823378054e-07, - "loss": 0.4377, - "step": 1689 - }, - { - "epoch": 0.53, - "learning_rate": 4.6967708070351873e-07, - "loss": 0.4531, - "step": 1690 - }, - { - "epoch": 0.53, - "learning_rate": 4.69180609077883e-07, - "loss": 0.4676, - "step": 1691 - }, - { - "epoch": 0.53, - "learning_rate": 4.686841679522241e-07, - "loss": 0.4843, - "step": 1692 - }, - { - "epoch": 0.53, - "learning_rate": 4.681877578178378e-07, - "loss": 0.4919, - "step": 1693 - }, - { - "epoch": 0.53, - "learning_rate": 4.676913791659899e-07, - "loss": 0.465, - "step": 1694 - }, - { - "epoch": 0.53, - "learning_rate": 4.671950324879143e-07, - "loss": 0.4982, - "step": 1695 - }, - { - "epoch": 0.53, - "learning_rate": 4.666987182748135e-07, - "loss": 0.4639, - "step": 1696 - }, - { - "epoch": 0.53, - "learning_rate": 4.6620243701785776e-07, - "loss": 0.4905, - "step": 1697 - }, - { - "epoch": 0.53, - "learning_rate": 4.6570618920818503e-07, - "loss": 0.4422, - "step": 1698 - }, - { - "epoch": 0.53, - "learning_rate": 4.652099753369e-07, - "loss": 0.4659, - "step": 1699 - }, - { - "epoch": 0.53, - "learning_rate": 4.6471379589507357e-07, - "loss": 0.4588, - "step": 1700 - }, - { - "epoch": 0.53, - "learning_rate": 4.64217651373743e-07, - "loss": 0.4591, - "step": 1701 - }, - { - "epoch": 0.53, - "learning_rate": 4.6372154226391054e-07, - "loss": 0.4739, - "step": 1702 - }, - { - "epoch": 0.53, - "learning_rate": 4.6322546905654366e-07, - "loss": 0.4389, - "step": 1703 - }, - { - "epoch": 0.53, - "learning_rate": 4.62729432242574e-07, - "loss": 0.4619, - "step": 1704 - }, - { - "epoch": 0.53, - "learning_rate": 4.622334323128979e-07, - "loss": 0.4784, - "step": 1705 - }, - { - "epoch": 0.53, - "learning_rate": 4.6173746975837436e-07, - "loss": 0.4637, - "step": 1706 - }, - { - "epoch": 0.53, - "learning_rate": 4.6124154506982586e-07, - "loss": 0.4726, - "step": 1707 - }, - { - "epoch": 0.53, - "learning_rate": 4.6074565873803714e-07, - "loss": 0.461, - "step": 1708 - }, - { - "epoch": 0.53, - "learning_rate": 4.6024981125375546e-07, - "loss": 0.4679, - "step": 1709 - }, - { - "epoch": 0.53, - "learning_rate": 4.5975400310768906e-07, - "loss": 0.4975, - "step": 1710 - }, - { - "epoch": 0.53, - "learning_rate": 4.592582347905075e-07, - "loss": 0.4905, - "step": 1711 - }, - { - "epoch": 0.53, - "learning_rate": 4.587625067928411e-07, - "loss": 0.4592, - "step": 1712 - }, - { - "epoch": 0.53, - "learning_rate": 4.5826681960528004e-07, - "loss": 0.4693, - "step": 1713 - }, - { - "epoch": 0.53, - "learning_rate": 4.57771173718374e-07, - "loss": 0.4756, - "step": 1714 - }, - { - "epoch": 0.53, - "learning_rate": 4.57275569622632e-07, - "loss": 0.478, - "step": 1715 - }, - { - "epoch": 0.53, - "learning_rate": 4.5678000780852187e-07, - "loss": 0.4883, - "step": 1716 - }, - { - "epoch": 0.54, - "learning_rate": 4.5628448876646935e-07, - "loss": 0.4491, - "step": 1717 - }, - { - "epoch": 0.54, - "learning_rate": 4.557890129868578e-07, - "loss": 0.4432, - "step": 1718 - }, - { - "epoch": 0.54, - "learning_rate": 4.552935809600276e-07, - "loss": 0.4618, - "step": 1719 - }, - { - "epoch": 0.54, - "learning_rate": 4.547981931762766e-07, - "loss": 0.4783, - "step": 1720 - }, - { - "epoch": 0.54, - "learning_rate": 4.5430285012585814e-07, - "loss": 0.4575, - "step": 1721 - }, - { - "epoch": 0.54, - "learning_rate": 4.538075522989813e-07, - "loss": 0.4617, - "step": 1722 - }, - { - "epoch": 0.54, - "learning_rate": 4.5331230018581103e-07, - "loss": 0.4955, - "step": 1723 - }, - { - "epoch": 0.54, - "learning_rate": 4.5281709427646615e-07, - "loss": 0.4836, - "step": 1724 - }, - { - "epoch": 0.54, - "learning_rate": 4.523219350610206e-07, - "loss": 0.4629, - "step": 1725 - }, - { - "epoch": 0.54, - "learning_rate": 4.518268230295015e-07, - "loss": 0.4589, - "step": 1726 - }, - { - "epoch": 0.54, - "learning_rate": 4.5133175867188965e-07, - "loss": 0.4502, - "step": 1727 - }, - { - "epoch": 0.54, - "learning_rate": 4.508367424781185e-07, - "loss": 0.4649, - "step": 1728 - }, - { - "epoch": 0.54, - "learning_rate": 4.503417749380739e-07, - "loss": 0.4494, - "step": 1729 - }, - { - "epoch": 0.54, - "learning_rate": 4.4984685654159324e-07, - "loss": 0.4797, - "step": 1730 - }, - { - "epoch": 0.54, - "learning_rate": 4.4935198777846594e-07, - "loss": 0.474, - "step": 1731 - }, - { - "epoch": 0.54, - "learning_rate": 4.4885716913843177e-07, - "loss": 0.4677, - "step": 1732 - }, - { - "epoch": 0.54, - "learning_rate": 4.483624011111808e-07, - "loss": 0.4556, - "step": 1733 - }, - { - "epoch": 0.54, - "learning_rate": 4.4786768418635367e-07, - "loss": 0.4838, - "step": 1734 - }, - { - "epoch": 0.54, - "learning_rate": 4.4737301885353986e-07, - "loss": 0.4528, - "step": 1735 - }, - { - "epoch": 0.54, - "learning_rate": 4.4687840560227783e-07, - "loss": 0.4708, - "step": 1736 - }, - { - "epoch": 0.54, - "learning_rate": 4.4638384492205473e-07, - "loss": 0.4651, - "step": 1737 - }, - { - "epoch": 0.54, - "learning_rate": 4.458893373023056e-07, - "loss": 0.462, - "step": 1738 - }, - { - "epoch": 0.54, - "learning_rate": 4.453948832324129e-07, - "loss": 0.4805, - "step": 1739 - }, - { - "epoch": 0.54, - "learning_rate": 4.4490048320170605e-07, - "loss": 0.4696, - "step": 1740 - }, - { - "epoch": 0.54, - "learning_rate": 4.444061376994614e-07, - "loss": 0.4612, - "step": 1741 - }, - { - "epoch": 0.54, - "learning_rate": 4.439118472149008e-07, - "loss": 0.4673, - "step": 1742 - }, - { - "epoch": 0.54, - "learning_rate": 4.43417612237192e-07, - "loss": 0.4577, - "step": 1743 - }, - { - "epoch": 0.54, - "learning_rate": 4.4292343325544736e-07, - "loss": 0.4779, - "step": 1744 - }, - { - "epoch": 0.54, - "learning_rate": 4.424293107587246e-07, - "loss": 0.4664, - "step": 1745 - }, - { - "epoch": 0.54, - "learning_rate": 4.4193524523602495e-07, - "loss": 0.4674, - "step": 1746 - }, - { - "epoch": 0.54, - "learning_rate": 4.4144123717629334e-07, - "loss": 0.4763, - "step": 1747 - }, - { - "epoch": 0.54, - "learning_rate": 4.409472870684178e-07, - "loss": 0.4834, - "step": 1748 - }, - { - "epoch": 0.55, - "learning_rate": 4.404533954012294e-07, - "loss": 0.4538, - "step": 1749 - }, - { - "epoch": 0.55, - "learning_rate": 4.3995956266350084e-07, - "loss": 0.4739, - "step": 1750 - }, - { - "epoch": 0.55, - "learning_rate": 4.3946578934394676e-07, - "loss": 0.4542, - "step": 1751 - }, - { - "epoch": 0.55, - "learning_rate": 4.389720759312232e-07, - "loss": 0.4744, - "step": 1752 - }, - { - "epoch": 0.55, - "learning_rate": 4.384784229139265e-07, - "loss": 0.4513, - "step": 1753 - }, - { - "epoch": 0.55, - "learning_rate": 4.379848307805935e-07, - "loss": 0.4742, - "step": 1754 - }, - { - "epoch": 0.55, - "learning_rate": 4.374913000197005e-07, - "loss": 0.478, - "step": 1755 - }, - { - "epoch": 0.55, - "learning_rate": 4.3699783111966366e-07, - "loss": 0.4656, - "step": 1756 - }, - { - "epoch": 0.55, - "learning_rate": 4.3650442456883734e-07, - "loss": 0.4907, - "step": 1757 - }, - { - "epoch": 0.55, - "learning_rate": 4.360110808555144e-07, - "loss": 0.465, - "step": 1758 - }, - { - "epoch": 0.55, - "learning_rate": 4.355178004679253e-07, - "loss": 0.4625, - "step": 1759 - }, - { - "epoch": 0.55, - "learning_rate": 4.3502458389423845e-07, - "loss": 0.4627, - "step": 1760 - }, - { - "epoch": 0.55, - "learning_rate": 4.3453143162255843e-07, - "loss": 0.4903, - "step": 1761 - }, - { - "epoch": 0.55, - "learning_rate": 4.3403834414092655e-07, - "loss": 0.4674, - "step": 1762 - }, - { - "epoch": 0.55, - "learning_rate": 4.3354532193731987e-07, - "loss": 0.4697, - "step": 1763 - }, - { - "epoch": 0.55, - "learning_rate": 4.330523654996509e-07, - "loss": 0.4706, - "step": 1764 - }, - { - "epoch": 0.55, - "learning_rate": 4.32559475315767e-07, - "loss": 0.4707, - "step": 1765 - }, - { - "epoch": 0.55, - "learning_rate": 4.320666518734499e-07, - "loss": 0.4517, - "step": 1766 - }, - { - "epoch": 0.55, - "learning_rate": 4.315738956604157e-07, - "loss": 0.462, - "step": 1767 - }, - { - "epoch": 0.55, - "learning_rate": 4.3108120716431344e-07, - "loss": 0.4795, - "step": 1768 - }, - { - "epoch": 0.55, - "learning_rate": 4.3058858687272547e-07, - "loss": 0.4635, - "step": 1769 - }, - { - "epoch": 0.55, - "learning_rate": 4.300960352731662e-07, - "loss": 0.4391, - "step": 1770 - }, - { - "epoch": 0.55, - "learning_rate": 4.2960355285308294e-07, - "loss": 0.4875, - "step": 1771 - }, - { - "epoch": 0.55, - "learning_rate": 4.291111400998536e-07, - "loss": 0.4771, - "step": 1772 - }, - { - "epoch": 0.55, - "learning_rate": 4.2861879750078765e-07, - "loss": 0.4972, - "step": 1773 - }, - { - "epoch": 0.55, - "learning_rate": 4.281265255431251e-07, - "loss": 0.4755, - "step": 1774 - }, - { - "epoch": 0.55, - "learning_rate": 4.276343247140359e-07, - "loss": 0.4718, - "step": 1775 - }, - { - "epoch": 0.55, - "learning_rate": 4.271421955006195e-07, - "loss": 0.4847, - "step": 1776 - }, - { - "epoch": 0.55, - "learning_rate": 4.2665013838990493e-07, - "loss": 0.463, - "step": 1777 - }, - { - "epoch": 0.55, - "learning_rate": 4.261581538688495e-07, - "loss": 0.4629, - "step": 1778 - }, - { - "epoch": 0.55, - "learning_rate": 4.2566624242433884e-07, - "loss": 0.4768, - "step": 1779 - }, - { - "epoch": 0.55, - "learning_rate": 4.2517440454318585e-07, - "loss": 0.4796, - "step": 1780 - }, - { - "epoch": 0.56, - "learning_rate": 4.246826407121315e-07, - "loss": 0.4702, - "step": 1781 - }, - { - "epoch": 0.56, - "learning_rate": 4.241909514178427e-07, - "loss": 0.4684, - "step": 1782 - }, - { - "epoch": 0.56, - "learning_rate": 4.236993371469128e-07, - "loss": 0.4551, - "step": 1783 - }, - { - "epoch": 0.56, - "learning_rate": 4.232077983858607e-07, - "loss": 0.453, - "step": 1784 - }, - { - "epoch": 0.56, - "learning_rate": 4.227163356211314e-07, - "loss": 0.467, - "step": 1785 - }, - { - "epoch": 0.56, - "learning_rate": 4.2222494933909385e-07, - "loss": 0.4682, - "step": 1786 - }, - { - "epoch": 0.56, - "learning_rate": 4.217336400260415e-07, - "loss": 0.4545, - "step": 1787 - }, - { - "epoch": 0.56, - "learning_rate": 4.2124240816819166e-07, - "loss": 0.4653, - "step": 1788 - }, - { - "epoch": 0.56, - "learning_rate": 4.2075125425168525e-07, - "loss": 0.4676, - "step": 1789 - }, - { - "epoch": 0.56, - "learning_rate": 4.2026017876258566e-07, - "loss": 0.4716, - "step": 1790 - }, - { - "epoch": 0.56, - "learning_rate": 4.197691821868787e-07, - "loss": 0.4678, - "step": 1791 - }, - { - "epoch": 0.56, - "learning_rate": 4.1927826501047265e-07, - "loss": 0.4598, - "step": 1792 - }, - { - "epoch": 0.56, - "learning_rate": 4.1878742771919655e-07, - "loss": 0.4538, - "step": 1793 - }, - { - "epoch": 0.56, - "learning_rate": 4.182966707988006e-07, - "loss": 0.4659, - "step": 1794 - }, - { - "epoch": 0.56, - "learning_rate": 4.178059947349554e-07, - "loss": 0.4703, - "step": 1795 - }, - { - "epoch": 0.56, - "learning_rate": 4.1731540001325185e-07, - "loss": 0.464, - "step": 1796 - }, - { - "epoch": 0.56, - "learning_rate": 4.168248871192001e-07, - "loss": 0.4605, - "step": 1797 - }, - { - "epoch": 0.56, - "learning_rate": 4.163344565382291e-07, - "loss": 0.4856, - "step": 1798 - }, - { - "epoch": 0.56, - "learning_rate": 4.1584410875568666e-07, - "loss": 0.4545, - "step": 1799 - }, - { - "epoch": 0.56, - "learning_rate": 4.153538442568388e-07, - "loss": 0.4615, - "step": 1800 - }, - { - "epoch": 0.56, - "learning_rate": 4.148636635268685e-07, - "loss": 0.4664, - "step": 1801 - }, - { - "epoch": 0.56, - "learning_rate": 4.1437356705087655e-07, - "loss": 0.4767, - "step": 1802 - }, - { - "epoch": 0.56, - "learning_rate": 4.1388355531388e-07, - "loss": 0.4707, - "step": 1803 - }, - { - "epoch": 0.56, - "learning_rate": 4.1339362880081195e-07, - "loss": 0.4759, - "step": 1804 - }, - { - "epoch": 0.56, - "learning_rate": 4.129037879965214e-07, - "loss": 0.4706, - "step": 1805 - }, - { - "epoch": 0.56, - "learning_rate": 4.1241403338577215e-07, - "loss": 0.4627, - "step": 1806 - }, - { - "epoch": 0.56, - "learning_rate": 4.119243654532433e-07, - "loss": 0.4854, - "step": 1807 - }, - { - "epoch": 0.56, - "learning_rate": 4.1143478468352776e-07, - "loss": 0.4507, - "step": 1808 - }, - { - "epoch": 0.56, - "learning_rate": 4.1094529156113214e-07, - "loss": 0.4792, - "step": 1809 - }, - { - "epoch": 0.56, - "learning_rate": 4.104558865704763e-07, - "loss": 0.4474, - "step": 1810 - }, - { - "epoch": 0.56, - "learning_rate": 4.099665701958934e-07, - "loss": 0.4521, - "step": 1811 - }, - { - "epoch": 0.56, - "learning_rate": 4.094773429216283e-07, - "loss": 0.46, - "step": 1812 - }, - { - "epoch": 0.57, - "learning_rate": 4.089882052318378e-07, - "loss": 0.4595, - "step": 1813 - }, - { - "epoch": 0.57, - "learning_rate": 4.0849915761059026e-07, - "loss": 0.4355, - "step": 1814 - }, - { - "epoch": 0.57, - "learning_rate": 4.080102005418648e-07, - "loss": 0.4745, - "step": 1815 - }, - { - "epoch": 0.57, - "learning_rate": 4.0752133450955083e-07, - "loss": 0.4893, - "step": 1816 - }, - { - "epoch": 0.57, - "learning_rate": 4.0703255999744755e-07, - "loss": 0.4631, - "step": 1817 - }, - { - "epoch": 0.57, - "learning_rate": 4.0654387748926425e-07, - "loss": 0.4578, - "step": 1818 - }, - { - "epoch": 0.57, - "learning_rate": 4.0605528746861846e-07, - "loss": 0.4732, - "step": 1819 - }, - { - "epoch": 0.57, - "learning_rate": 4.055667904190364e-07, - "loss": 0.4852, - "step": 1820 - }, - { - "epoch": 0.57, - "learning_rate": 4.0507838682395214e-07, - "loss": 0.4531, - "step": 1821 - }, - { - "epoch": 0.57, - "learning_rate": 4.0459007716670774e-07, - "loss": 0.4874, - "step": 1822 - }, - { - "epoch": 0.57, - "learning_rate": 4.041018619305519e-07, - "loss": 0.4592, - "step": 1823 - }, - { - "epoch": 0.57, - "learning_rate": 4.036137415986398e-07, - "loss": 0.4845, - "step": 1824 - }, - { - "epoch": 0.57, - "learning_rate": 4.0312571665403316e-07, - "loss": 0.4874, - "step": 1825 - }, - { - "epoch": 0.57, - "learning_rate": 4.026377875796988e-07, - "loss": 0.4735, - "step": 1826 - }, - { - "epoch": 0.57, - "learning_rate": 4.021499548585088e-07, - "loss": 0.4796, - "step": 1827 - }, - { - "epoch": 0.57, - "learning_rate": 4.0166221897323997e-07, - "loss": 0.4777, - "step": 1828 - }, - { - "epoch": 0.57, - "learning_rate": 4.011745804065735e-07, - "loss": 0.4601, - "step": 1829 - }, - { - "epoch": 0.57, - "learning_rate": 4.0068703964109396e-07, - "loss": 0.4518, - "step": 1830 - }, - { - "epoch": 0.57, - "learning_rate": 4.0019959715928885e-07, - "loss": 0.4625, - "step": 1831 - }, - { - "epoch": 0.57, - "learning_rate": 3.997122534435493e-07, - "loss": 0.4658, - "step": 1832 - }, - { - "epoch": 0.57, - "learning_rate": 3.9922500897616794e-07, - "loss": 0.4549, - "step": 1833 - }, - { - "epoch": 0.57, - "learning_rate": 3.9873786423933934e-07, - "loss": 0.4639, - "step": 1834 - }, - { - "epoch": 0.57, - "learning_rate": 3.9825081971515936e-07, - "loss": 0.4813, - "step": 1835 - }, - { - "epoch": 0.57, - "learning_rate": 3.9776387588562505e-07, - "loss": 0.4467, - "step": 1836 - }, - { - "epoch": 0.57, - "learning_rate": 3.972770332326334e-07, - "loss": 0.4651, - "step": 1837 - }, - { - "epoch": 0.57, - "learning_rate": 3.967902922379813e-07, - "loss": 0.4594, - "step": 1838 - }, - { - "epoch": 0.57, - "learning_rate": 3.963036533833652e-07, - "loss": 0.4728, - "step": 1839 - }, - { - "epoch": 0.57, - "learning_rate": 3.9581711715038046e-07, - "loss": 0.4442, - "step": 1840 - }, - { - "epoch": 0.57, - "learning_rate": 3.953306840205207e-07, - "loss": 0.4697, - "step": 1841 - }, - { - "epoch": 0.57, - "learning_rate": 3.9484435447517757e-07, - "loss": 0.4768, - "step": 1842 - }, - { - "epoch": 0.57, - "learning_rate": 3.943581289956405e-07, - "loss": 0.4687, - "step": 1843 - }, - { - "epoch": 0.57, - "learning_rate": 3.9387200806309564e-07, - "loss": 0.4583, - "step": 1844 - }, - { - "epoch": 0.58, - "learning_rate": 3.933859921586257e-07, - "loss": 0.458, - "step": 1845 - }, - { - "epoch": 0.58, - "learning_rate": 3.9290008176320914e-07, - "loss": 0.4666, - "step": 1846 - }, - { - "epoch": 0.58, - "learning_rate": 3.92414277357721e-07, - "loss": 0.4624, - "step": 1847 - }, - { - "epoch": 0.58, - "learning_rate": 3.919285794229304e-07, - "loss": 0.4617, - "step": 1848 - }, - { - "epoch": 0.58, - "learning_rate": 3.914429884395015e-07, - "loss": 0.4453, - "step": 1849 - }, - { - "epoch": 0.58, - "learning_rate": 3.909575048879925e-07, - "loss": 0.462, - "step": 1850 - }, - { - "epoch": 0.58, - "learning_rate": 3.9047212924885565e-07, - "loss": 0.4777, - "step": 1851 - }, - { - "epoch": 0.58, - "learning_rate": 3.899868620024359e-07, - "loss": 0.4738, - "step": 1852 - }, - { - "epoch": 0.58, - "learning_rate": 3.89501703628971e-07, - "loss": 0.4809, - "step": 1853 - }, - { - "epoch": 0.58, - "learning_rate": 3.8901665460859154e-07, - "loss": 0.4504, - "step": 1854 - }, - { - "epoch": 0.58, - "learning_rate": 3.885317154213193e-07, - "loss": 0.4774, - "step": 1855 - }, - { - "epoch": 0.58, - "learning_rate": 3.880468865470674e-07, - "loss": 0.4395, - "step": 1856 - }, - { - "epoch": 0.58, - "learning_rate": 3.8756216846563983e-07, - "loss": 0.4581, - "step": 1857 - }, - { - "epoch": 0.58, - "learning_rate": 3.8707756165673137e-07, - "loss": 0.4728, - "step": 1858 - }, - { - "epoch": 0.58, - "learning_rate": 3.865930665999262e-07, - "loss": 0.4614, - "step": 1859 - }, - { - "epoch": 0.58, - "learning_rate": 3.8610868377469797e-07, - "loss": 0.453, - "step": 1860 - }, - { - "epoch": 0.58, - "learning_rate": 3.856244136604091e-07, - "loss": 0.4623, - "step": 1861 - }, - { - "epoch": 0.58, - "learning_rate": 3.851402567363112e-07, - "loss": 0.4772, - "step": 1862 - }, - { - "epoch": 0.58, - "learning_rate": 3.846562134815431e-07, - "loss": 0.4744, - "step": 1863 - }, - { - "epoch": 0.58, - "learning_rate": 3.8417228437513126e-07, - "loss": 0.4508, - "step": 1864 - }, - { - "epoch": 0.58, - "learning_rate": 3.8368846989598967e-07, - "loss": 0.4779, - "step": 1865 - }, - { - "epoch": 0.58, - "learning_rate": 3.8320477052291826e-07, - "loss": 0.4661, - "step": 1866 - }, - { - "epoch": 0.58, - "learning_rate": 3.827211867346034e-07, - "loss": 0.4607, - "step": 1867 - }, - { - "epoch": 0.58, - "learning_rate": 3.822377190096168e-07, - "loss": 0.4774, - "step": 1868 - }, - { - "epoch": 0.58, - "learning_rate": 3.817543678264158e-07, - "loss": 0.4531, - "step": 1869 - }, - { - "epoch": 0.58, - "learning_rate": 3.8127113366334195e-07, - "loss": 0.4724, - "step": 1870 - }, - { - "epoch": 0.58, - "learning_rate": 3.80788016998621e-07, - "loss": 0.4671, - "step": 1871 - }, - { - "epoch": 0.58, - "learning_rate": 3.8030501831036296e-07, - "loss": 0.4723, - "step": 1872 - }, - { - "epoch": 0.58, - "learning_rate": 3.798221380765604e-07, - "loss": 0.467, - "step": 1873 - }, - { - "epoch": 0.58, - "learning_rate": 3.79339376775089e-07, - "loss": 0.4737, - "step": 1874 - }, - { - "epoch": 0.58, - "learning_rate": 3.788567348837066e-07, - "loss": 0.4646, - "step": 1875 - }, - { - "epoch": 0.58, - "learning_rate": 3.7837421288005315e-07, - "loss": 0.4661, - "step": 1876 - }, - { - "epoch": 0.59, - "learning_rate": 3.778918112416497e-07, - "loss": 0.4767, - "step": 1877 - }, - { - "epoch": 0.59, - "learning_rate": 3.774095304458982e-07, - "loss": 0.4586, - "step": 1878 - }, - { - "epoch": 0.59, - "learning_rate": 3.7692737097008086e-07, - "loss": 0.4818, - "step": 1879 - }, - { - "epoch": 0.59, - "learning_rate": 3.764453332913605e-07, - "loss": 0.4762, - "step": 1880 - }, - { - "epoch": 0.59, - "learning_rate": 3.7596341788677864e-07, - "loss": 0.4586, - "step": 1881 - }, - { - "epoch": 0.59, - "learning_rate": 3.754816252332559e-07, - "loss": 0.4764, - "step": 1882 - }, - { - "epoch": 0.59, - "learning_rate": 3.7499995580759213e-07, - "loss": 0.457, - "step": 1883 - }, - { - "epoch": 0.59, - "learning_rate": 3.7451841008646433e-07, - "loss": 0.4631, - "step": 1884 - }, - { - "epoch": 0.59, - "learning_rate": 3.7403698854642753e-07, - "loss": 0.4663, - "step": 1885 - }, - { - "epoch": 0.59, - "learning_rate": 3.7355569166391364e-07, - "loss": 0.4643, - "step": 1886 - }, - { - "epoch": 0.59, - "learning_rate": 3.730745199152315e-07, - "loss": 0.4812, - "step": 1887 - }, - { - "epoch": 0.59, - "learning_rate": 3.7259347377656613e-07, - "loss": 0.4358, - "step": 1888 - }, - { - "epoch": 0.59, - "learning_rate": 3.721125537239778e-07, - "loss": 0.4669, - "step": 1889 - }, - { - "epoch": 0.59, - "learning_rate": 3.716317602334024e-07, - "loss": 0.4673, - "step": 1890 - }, - { - "epoch": 0.59, - "learning_rate": 3.711510937806504e-07, - "loss": 0.4855, - "step": 1891 - }, - { - "epoch": 0.59, - "learning_rate": 3.706705548414067e-07, - "loss": 0.4814, - "step": 1892 - }, - { - "epoch": 0.59, - "learning_rate": 3.701901438912296e-07, - "loss": 0.4763, - "step": 1893 - }, - { - "epoch": 0.59, - "learning_rate": 3.6970986140555145e-07, - "loss": 0.4543, - "step": 1894 - }, - { - "epoch": 0.59, - "learning_rate": 3.692297078596769e-07, - "loss": 0.4493, - "step": 1895 - }, - { - "epoch": 0.59, - "learning_rate": 3.687496837287831e-07, - "loss": 0.4533, - "step": 1896 - }, - { - "epoch": 0.59, - "learning_rate": 3.6826978948791907e-07, - "loss": 0.4665, - "step": 1897 - }, - { - "epoch": 0.59, - "learning_rate": 3.6779002561200573e-07, - "loss": 0.4558, - "step": 1898 - }, - { - "epoch": 0.59, - "learning_rate": 3.673103925758344e-07, - "loss": 0.4616, - "step": 1899 - }, - { - "epoch": 0.59, - "learning_rate": 3.668308908540673e-07, - "loss": 0.483, - "step": 1900 - }, - { - "epoch": 0.59, - "learning_rate": 3.663515209212364e-07, - "loss": 0.4652, - "step": 1901 - }, - { - "epoch": 0.59, - "learning_rate": 3.658722832517436e-07, - "loss": 0.4688, - "step": 1902 - }, - { - "epoch": 0.59, - "learning_rate": 3.653931783198596e-07, - "loss": 0.4459, - "step": 1903 - }, - { - "epoch": 0.59, - "learning_rate": 3.649142065997238e-07, - "loss": 0.4687, - "step": 1904 - }, - { - "epoch": 0.59, - "learning_rate": 3.64435368565344e-07, - "loss": 0.4701, - "step": 1905 - }, - { - "epoch": 0.59, - "learning_rate": 3.639566646905955e-07, - "loss": 0.4492, - "step": 1906 - }, - { - "epoch": 0.59, - "learning_rate": 3.634780954492208e-07, - "loss": 0.4632, - "step": 1907 - }, - { - "epoch": 0.59, - "learning_rate": 3.6299966131482907e-07, - "loss": 0.4643, - "step": 1908 - }, - { - "epoch": 0.6, - "learning_rate": 3.6252136276089617e-07, - "loss": 0.4831, - "step": 1909 - }, - { - "epoch": 0.6, - "learning_rate": 3.620432002607636e-07, - "loss": 0.4731, - "step": 1910 - }, - { - "epoch": 0.6, - "learning_rate": 3.6156517428763777e-07, - "loss": 0.4822, - "step": 1911 - }, - { - "epoch": 0.6, - "learning_rate": 3.6108728531459064e-07, - "loss": 0.4838, - "step": 1912 - }, - { - "epoch": 0.6, - "learning_rate": 3.606095338145584e-07, - "loss": 0.4637, - "step": 1913 - }, - { - "epoch": 0.6, - "learning_rate": 3.60131920260341e-07, - "loss": 0.4571, - "step": 1914 - }, - { - "epoch": 0.6, - "learning_rate": 3.596544451246017e-07, - "loss": 0.4692, - "step": 1915 - }, - { - "epoch": 0.6, - "learning_rate": 3.5917710887986755e-07, - "loss": 0.4881, - "step": 1916 - }, - { - "epoch": 0.6, - "learning_rate": 3.586999119985274e-07, - "loss": 0.4675, - "step": 1917 - }, - { - "epoch": 0.6, - "learning_rate": 3.5822285495283246e-07, - "loss": 0.4633, - "step": 1918 - }, - { - "epoch": 0.6, - "learning_rate": 3.577459382148952e-07, - "loss": 0.4691, - "step": 1919 - }, - { - "epoch": 0.6, - "learning_rate": 3.572691622566901e-07, - "loss": 0.4583, - "step": 1920 - }, - { - "epoch": 0.6, - "learning_rate": 3.567925275500516e-07, - "loss": 0.4903, - "step": 1921 - }, - { - "epoch": 0.6, - "learning_rate": 3.5631603456667427e-07, - "loss": 0.4533, - "step": 1922 - }, - { - "epoch": 0.6, - "learning_rate": 3.558396837781131e-07, - "loss": 0.462, - "step": 1923 - }, - { - "epoch": 0.6, - "learning_rate": 3.553634756557818e-07, - "loss": 0.4829, - "step": 1924 - }, - { - "epoch": 0.6, - "learning_rate": 3.5488741067095295e-07, - "loss": 0.4547, - "step": 1925 - }, - { - "epoch": 0.6, - "learning_rate": 3.544114892947575e-07, - "loss": 0.4737, - "step": 1926 - }, - { - "epoch": 0.6, - "learning_rate": 3.539357119981846e-07, - "loss": 0.4721, - "step": 1927 - }, - { - "epoch": 0.6, - "learning_rate": 3.534600792520805e-07, - "loss": 0.4705, - "step": 1928 - }, - { - "epoch": 0.6, - "learning_rate": 3.529845915271483e-07, - "loss": 0.4724, - "step": 1929 - }, - { - "epoch": 0.6, - "learning_rate": 3.5250924929394764e-07, - "loss": 0.4662, - "step": 1930 - }, - { - "epoch": 0.6, - "learning_rate": 3.5203405302289466e-07, - "loss": 0.4906, - "step": 1931 - }, - { - "epoch": 0.6, - "learning_rate": 3.515590031842605e-07, - "loss": 0.4811, - "step": 1932 - }, - { - "epoch": 0.6, - "learning_rate": 3.510841002481714e-07, - "loss": 0.4531, - "step": 1933 - }, - { - "epoch": 0.6, - "learning_rate": 3.5060934468460855e-07, - "loss": 0.4572, - "step": 1934 - }, - { - "epoch": 0.6, - "learning_rate": 3.501347369634071e-07, - "loss": 0.4701, - "step": 1935 - }, - { - "epoch": 0.6, - "learning_rate": 3.49660277554256e-07, - "loss": 0.4537, - "step": 1936 - }, - { - "epoch": 0.6, - "learning_rate": 3.49185966926697e-07, - "loss": 0.4694, - "step": 1937 - }, - { - "epoch": 0.6, - "learning_rate": 3.487118055501251e-07, - "loss": 0.4616, - "step": 1938 - }, - { - "epoch": 0.6, - "learning_rate": 3.482377938937877e-07, - "loss": 0.466, - "step": 1939 - }, - { - "epoch": 0.6, - "learning_rate": 3.477639324267835e-07, - "loss": 0.4501, - "step": 1940 - }, - { - "epoch": 0.61, - "learning_rate": 3.4729022161806283e-07, - "loss": 0.4422, - "step": 1941 - }, - { - "epoch": 0.61, - "learning_rate": 3.468166619364271e-07, - "loss": 0.4545, - "step": 1942 - }, - { - "epoch": 0.61, - "learning_rate": 3.4634325385052787e-07, - "loss": 0.4719, - "step": 1943 - }, - { - "epoch": 0.61, - "learning_rate": 3.4586999782886663e-07, - "loss": 0.4774, - "step": 1944 - }, - { - "epoch": 0.61, - "learning_rate": 3.4539689433979486e-07, - "loss": 0.483, - "step": 1945 - }, - { - "epoch": 0.61, - "learning_rate": 3.449239438515127e-07, - "loss": 0.4581, - "step": 1946 - }, - { - "epoch": 0.61, - "learning_rate": 3.4445114683206886e-07, - "loss": 0.4585, - "step": 1947 - }, - { - "epoch": 0.61, - "learning_rate": 3.4397850374936016e-07, - "loss": 0.4643, - "step": 1948 - }, - { - "epoch": 0.61, - "learning_rate": 3.435060150711315e-07, - "loss": 0.4573, - "step": 1949 - }, - { - "epoch": 0.61, - "learning_rate": 3.4303368126497444e-07, - "loss": 0.4758, - "step": 1950 - }, - { - "epoch": 0.61, - "learning_rate": 3.425615027983274e-07, - "loss": 0.4693, - "step": 1951 - }, - { - "epoch": 0.61, - "learning_rate": 3.4208948013847547e-07, - "loss": 0.4577, - "step": 1952 - }, - { - "epoch": 0.61, - "learning_rate": 3.4161761375254905e-07, - "loss": 0.4932, - "step": 1953 - }, - { - "epoch": 0.61, - "learning_rate": 3.4114590410752403e-07, - "loss": 0.4742, - "step": 1954 - }, - { - "epoch": 0.61, - "learning_rate": 3.4067435167022107e-07, - "loss": 0.4509, - "step": 1955 - }, - { - "epoch": 0.61, - "learning_rate": 3.402029569073058e-07, - "loss": 0.4665, - "step": 1956 - }, - { - "epoch": 0.61, - "learning_rate": 3.397317202852873e-07, - "loss": 0.4608, - "step": 1957 - }, - { - "epoch": 0.61, - "learning_rate": 3.392606422705181e-07, - "loss": 0.4715, - "step": 1958 - }, - { - "epoch": 0.61, - "learning_rate": 3.3878972332919393e-07, - "loss": 0.4664, - "step": 1959 - }, - { - "epoch": 0.61, - "learning_rate": 3.3831896392735335e-07, - "loss": 0.4552, - "step": 1960 - }, - { - "epoch": 0.61, - "learning_rate": 3.378483645308766e-07, - "loss": 0.4619, - "step": 1961 - }, - { - "epoch": 0.61, - "learning_rate": 3.3737792560548574e-07, - "loss": 0.4425, - "step": 1962 - }, - { - "epoch": 0.61, - "learning_rate": 3.369076476167442e-07, - "loss": 0.4701, - "step": 1963 - }, - { - "epoch": 0.61, - "learning_rate": 3.364375310300557e-07, - "loss": 0.4756, - "step": 1964 - }, - { - "epoch": 0.61, - "learning_rate": 3.359675763106648e-07, - "loss": 0.4665, - "step": 1965 - }, - { - "epoch": 0.61, - "learning_rate": 3.3549778392365535e-07, - "loss": 0.4523, - "step": 1966 - }, - { - "epoch": 0.61, - "learning_rate": 3.35028154333951e-07, - "loss": 0.4663, - "step": 1967 - }, - { - "epoch": 0.61, - "learning_rate": 3.3455868800631395e-07, - "loss": 0.4561, - "step": 1968 - }, - { - "epoch": 0.61, - "learning_rate": 3.340893854053449e-07, - "loss": 0.4623, - "step": 1969 - }, - { - "epoch": 0.61, - "learning_rate": 3.336202469954824e-07, - "loss": 0.4508, - "step": 1970 - }, - { - "epoch": 0.61, - "learning_rate": 3.331512732410032e-07, - "loss": 0.47, - "step": 1971 - }, - { - "epoch": 0.61, - "learning_rate": 3.3268246460602023e-07, - "loss": 0.4675, - "step": 1972 - }, - { - "epoch": 0.62, - "learning_rate": 3.3221382155448325e-07, - "loss": 0.4452, - "step": 1973 - }, - { - "epoch": 0.62, - "learning_rate": 3.317453445501786e-07, - "loss": 0.4626, - "step": 1974 - }, - { - "epoch": 0.62, - "learning_rate": 3.312770340567278e-07, - "loss": 0.4674, - "step": 1975 - }, - { - "epoch": 0.62, - "learning_rate": 3.3080889053758776e-07, - "loss": 0.4347, - "step": 1976 - }, - { - "epoch": 0.62, - "learning_rate": 3.3034091445605e-07, - "loss": 0.4588, - "step": 1977 - }, - { - "epoch": 0.62, - "learning_rate": 3.298731062752407e-07, - "loss": 0.4543, - "step": 1978 - }, - { - "epoch": 0.62, - "learning_rate": 3.2940546645811946e-07, - "loss": 0.4915, - "step": 1979 - }, - { - "epoch": 0.62, - "learning_rate": 3.2893799546747956e-07, - "loss": 0.4758, - "step": 1980 - }, - { - "epoch": 0.62, - "learning_rate": 3.284706937659468e-07, - "loss": 0.4779, - "step": 1981 - }, - { - "epoch": 0.62, - "learning_rate": 3.280035618159802e-07, - "loss": 0.4492, - "step": 1982 - }, - { - "epoch": 0.62, - "learning_rate": 3.2753660007987016e-07, - "loss": 0.4647, - "step": 1983 - }, - { - "epoch": 0.62, - "learning_rate": 3.270698090197385e-07, - "loss": 0.4545, - "step": 1984 - }, - { - "epoch": 0.62, - "learning_rate": 3.266031890975389e-07, - "loss": 0.4394, - "step": 1985 - }, - { - "epoch": 0.62, - "learning_rate": 3.261367407750549e-07, - "loss": 0.4801, - "step": 1986 - }, - { - "epoch": 0.62, - "learning_rate": 3.2567046451390073e-07, - "loss": 0.4688, - "step": 1987 - }, - { - "epoch": 0.62, - "learning_rate": 3.252043607755198e-07, - "loss": 0.4631, - "step": 1988 - }, - { - "epoch": 0.62, - "learning_rate": 3.247384300211855e-07, - "loss": 0.4674, - "step": 1989 - }, - { - "epoch": 0.62, - "learning_rate": 3.2427267271199936e-07, - "loss": 0.4721, - "step": 1990 - }, - { - "epoch": 0.62, - "learning_rate": 3.2380708930889175e-07, - "loss": 0.4854, - "step": 1991 - }, - { - "epoch": 0.62, - "learning_rate": 3.233416802726206e-07, - "loss": 0.4626, - "step": 1992 - }, - { - "epoch": 0.62, - "learning_rate": 3.228764460637716e-07, - "loss": 0.4625, - "step": 1993 - }, - { - "epoch": 0.62, - "learning_rate": 3.2241138714275707e-07, - "loss": 0.4576, - "step": 1994 - }, - { - "epoch": 0.62, - "learning_rate": 3.2194650396981584e-07, - "loss": 0.4624, - "step": 1995 - }, - { - "epoch": 0.62, - "learning_rate": 3.214817970050135e-07, - "loss": 0.438, - "step": 1996 - }, - { - "epoch": 0.62, - "learning_rate": 3.210172667082405e-07, - "loss": 0.4571, - "step": 1997 - }, - { - "epoch": 0.62, - "learning_rate": 3.2055291353921273e-07, - "loss": 0.4704, - "step": 1998 - }, - { - "epoch": 0.62, - "learning_rate": 3.200887379574705e-07, - "loss": 0.4692, - "step": 1999 - }, - { - "epoch": 0.62, - "learning_rate": 3.196247404223792e-07, - "loss": 0.4557, - "step": 2000 - }, - { - "epoch": 0.62, - "learning_rate": 3.1916092139312717e-07, - "loss": 0.4429, - "step": 2001 - }, - { - "epoch": 0.62, - "learning_rate": 3.186972813287263e-07, - "loss": 0.4798, - "step": 2002 - }, - { - "epoch": 0.62, - "learning_rate": 3.182338206880116e-07, - "loss": 0.4765, - "step": 2003 - }, - { - "epoch": 0.62, - "learning_rate": 3.1777053992964043e-07, - "loss": 0.4729, - "step": 2004 - }, - { - "epoch": 0.62, - "learning_rate": 3.1730743951209193e-07, - "loss": 0.4521, - "step": 2005 - }, - { - "epoch": 0.63, - "learning_rate": 3.1684451989366683e-07, - "loss": 0.463, - "step": 2006 - }, - { - "epoch": 0.63, - "learning_rate": 3.163817815324873e-07, - "loss": 0.4624, - "step": 2007 - }, - { - "epoch": 0.63, - "learning_rate": 3.159192248864958e-07, - "loss": 0.4618, - "step": 2008 - }, - { - "epoch": 0.63, - "learning_rate": 3.1545685041345484e-07, - "loss": 0.454, - "step": 2009 - }, - { - "epoch": 0.63, - "learning_rate": 3.1499465857094674e-07, - "loss": 0.4564, - "step": 2010 - }, - { - "epoch": 0.63, - "learning_rate": 3.145326498163735e-07, - "loss": 0.4729, - "step": 2011 - }, - { - "epoch": 0.63, - "learning_rate": 3.1407082460695546e-07, - "loss": 0.4617, - "step": 2012 - }, - { - "epoch": 0.63, - "learning_rate": 3.1360918339973133e-07, - "loss": 0.46, - "step": 2013 - }, - { - "epoch": 0.63, - "learning_rate": 3.1314772665155784e-07, - "loss": 0.4739, - "step": 2014 - }, - { - "epoch": 0.63, - "learning_rate": 3.1268645481910937e-07, - "loss": 0.4472, - "step": 2015 - }, - { - "epoch": 0.63, - "learning_rate": 3.122253683588768e-07, - "loss": 0.473, - "step": 2016 - }, - { - "epoch": 0.63, - "learning_rate": 3.117644677271681e-07, - "loss": 0.466, - "step": 2017 - }, - { - "epoch": 0.63, - "learning_rate": 3.1130375338010704e-07, - "loss": 0.4622, - "step": 2018 - }, - { - "epoch": 0.63, - "learning_rate": 3.108432257736332e-07, - "loss": 0.4599, - "step": 2019 - }, - { - "epoch": 0.63, - "learning_rate": 3.1038288536350105e-07, - "loss": 0.465, - "step": 2020 - }, - { - "epoch": 0.63, - "learning_rate": 3.099227326052799e-07, - "loss": 0.4819, - "step": 2021 - }, - { - "epoch": 0.63, - "learning_rate": 3.094627679543538e-07, - "loss": 0.4475, - "step": 2022 - }, - { - "epoch": 0.63, - "learning_rate": 3.090029918659202e-07, - "loss": 0.4702, - "step": 2023 - }, - { - "epoch": 0.63, - "learning_rate": 3.0854340479498976e-07, - "loss": 0.4621, - "step": 2024 - }, - { - "epoch": 0.63, - "learning_rate": 3.0808400719638683e-07, - "loss": 0.4533, - "step": 2025 - }, - { - "epoch": 0.63, - "learning_rate": 3.076247995247475e-07, - "loss": 0.4619, - "step": 2026 - }, - { - "epoch": 0.63, - "learning_rate": 3.071657822345202e-07, - "loss": 0.4619, - "step": 2027 - }, - { - "epoch": 0.63, - "learning_rate": 3.067069557799649e-07, - "loss": 0.4514, - "step": 2028 - }, - { - "epoch": 0.63, - "learning_rate": 3.0624832061515294e-07, - "loss": 0.4611, - "step": 2029 - }, - { - "epoch": 0.63, - "learning_rate": 3.05789877193966e-07, - "loss": 0.4876, - "step": 2030 - }, - { - "epoch": 0.63, - "learning_rate": 3.0533162597009615e-07, - "loss": 0.4519, - "step": 2031 - }, - { - "epoch": 0.63, - "learning_rate": 3.048735673970452e-07, - "loss": 0.4459, - "step": 2032 - }, - { - "epoch": 0.63, - "learning_rate": 3.044157019281246e-07, - "loss": 0.4783, - "step": 2033 - }, - { - "epoch": 0.63, - "learning_rate": 3.0395803001645444e-07, - "loss": 0.449, - "step": 2034 - }, - { - "epoch": 0.63, - "learning_rate": 3.0350055211496294e-07, - "loss": 0.45, - "step": 2035 - }, - { - "epoch": 0.63, - "learning_rate": 3.0304326867638724e-07, - "loss": 0.4733, - "step": 2036 - }, - { - "epoch": 0.63, - "learning_rate": 3.0258618015327117e-07, - "loss": 0.4479, - "step": 2037 - }, - { - "epoch": 0.64, - "learning_rate": 3.0212928699796593e-07, - "loss": 0.4777, - "step": 2038 - }, - { - "epoch": 0.64, - "learning_rate": 3.016725896626293e-07, - "loss": 0.4893, - "step": 2039 - }, - { - "epoch": 0.64, - "learning_rate": 3.0121608859922567e-07, - "loss": 0.4561, - "step": 2040 - }, - { - "epoch": 0.64, - "learning_rate": 3.0075978425952477e-07, - "loss": 0.4685, - "step": 2041 - }, - { - "epoch": 0.64, - "learning_rate": 3.0030367709510153e-07, - "loss": 0.4827, - "step": 2042 - }, - { - "epoch": 0.64, - "learning_rate": 2.998477675573364e-07, - "loss": 0.4541, - "step": 2043 - }, - { - "epoch": 0.64, - "learning_rate": 2.993920560974136e-07, - "loss": 0.4703, - "step": 2044 - }, - { - "epoch": 0.64, - "learning_rate": 2.9893654316632165e-07, - "loss": 0.4608, - "step": 2045 - }, - { - "epoch": 0.64, - "learning_rate": 2.9848122921485236e-07, - "loss": 0.4415, - "step": 2046 - }, - { - "epoch": 0.64, - "learning_rate": 2.9802611469360104e-07, - "loss": 0.4474, - "step": 2047 - }, - { - "epoch": 0.64, - "learning_rate": 2.975712000529653e-07, - "loss": 0.4778, - "step": 2048 - }, - { - "epoch": 0.64, - "learning_rate": 2.971164857431449e-07, - "loss": 0.435, - "step": 2049 - }, - { - "epoch": 0.64, - "learning_rate": 2.9666197221414133e-07, - "loss": 0.4751, - "step": 2050 - }, - { - "epoch": 0.64, - "learning_rate": 2.96207659915758e-07, - "loss": 0.4709, - "step": 2051 - }, - { - "epoch": 0.64, - "learning_rate": 2.9575354929759835e-07, - "loss": 0.4742, - "step": 2052 - }, - { - "epoch": 0.64, - "learning_rate": 2.9529964080906646e-07, - "loss": 0.4453, - "step": 2053 - }, - { - "epoch": 0.64, - "learning_rate": 2.948459348993667e-07, - "loss": 0.4708, - "step": 2054 - }, - { - "epoch": 0.64, - "learning_rate": 2.9439243201750273e-07, - "loss": 0.4703, - "step": 2055 - }, - { - "epoch": 0.64, - "learning_rate": 2.939391326122773e-07, - "loss": 0.4824, - "step": 2056 - }, - { - "epoch": 0.64, - "learning_rate": 2.9348603713229155e-07, - "loss": 0.4433, - "step": 2057 - }, - { - "epoch": 0.64, - "learning_rate": 2.930331460259454e-07, - "loss": 0.4721, - "step": 2058 - }, - { - "epoch": 0.64, - "learning_rate": 2.9258045974143596e-07, - "loss": 0.4656, - "step": 2059 - }, - { - "epoch": 0.64, - "learning_rate": 2.92127978726758e-07, - "loss": 0.4863, - "step": 2060 - }, - { - "epoch": 0.64, - "learning_rate": 2.916757034297026e-07, - "loss": 0.471, - "step": 2061 - }, - { - "epoch": 0.64, - "learning_rate": 2.9122363429785827e-07, - "loss": 0.4783, - "step": 2062 - }, - { - "epoch": 0.64, - "learning_rate": 2.907717717786083e-07, - "loss": 0.4484, - "step": 2063 - }, - { - "epoch": 0.64, - "learning_rate": 2.903201163191324e-07, - "loss": 0.4654, - "step": 2064 - }, - { - "epoch": 0.64, - "learning_rate": 2.898686683664049e-07, - "loss": 0.4689, - "step": 2065 - }, - { - "epoch": 0.64, - "learning_rate": 2.894174283671952e-07, - "loss": 0.4659, - "step": 2066 - }, - { - "epoch": 0.64, - "learning_rate": 2.8896639676806616e-07, - "loss": 0.4544, - "step": 2067 - }, - { - "epoch": 0.64, - "learning_rate": 2.885155740153751e-07, - "loss": 0.4621, - "step": 2068 - }, - { - "epoch": 0.64, - "learning_rate": 2.8806496055527244e-07, - "loss": 0.4694, - "step": 2069 - }, - { - "epoch": 0.65, - "learning_rate": 2.8761455683370163e-07, - "loss": 0.4648, - "step": 2070 - }, - { - "epoch": 0.65, - "learning_rate": 2.8716436329639793e-07, - "loss": 0.4533, - "step": 2071 - }, - { - "epoch": 0.65, - "learning_rate": 2.867143803888894e-07, - "loss": 0.4719, - "step": 2072 - }, - { - "epoch": 0.65, - "learning_rate": 2.8626460855649526e-07, - "loss": 0.4442, - "step": 2073 - }, - { - "epoch": 0.65, - "learning_rate": 2.858150482443257e-07, - "loss": 0.4634, - "step": 2074 - }, - { - "epoch": 0.65, - "learning_rate": 2.8536569989728184e-07, - "loss": 0.4602, - "step": 2075 - }, - { - "epoch": 0.65, - "learning_rate": 2.84916563960055e-07, - "loss": 0.4821, - "step": 2076 - }, - { - "epoch": 0.65, - "learning_rate": 2.8446764087712635e-07, - "loss": 0.4575, - "step": 2077 - }, - { - "epoch": 0.65, - "learning_rate": 2.840189310927661e-07, - "loss": 0.4736, - "step": 2078 - }, - { - "epoch": 0.65, - "learning_rate": 2.8357043505103356e-07, - "loss": 0.448, - "step": 2079 - }, - { - "epoch": 0.65, - "learning_rate": 2.831221531957767e-07, - "loss": 0.4519, - "step": 2080 - }, - { - "epoch": 0.65, - "learning_rate": 2.826740859706315e-07, - "loss": 0.4623, - "step": 2081 - }, - { - "epoch": 0.65, - "learning_rate": 2.822262338190209e-07, - "loss": 0.4832, - "step": 2082 - }, - { - "epoch": 0.65, - "learning_rate": 2.8177859718415583e-07, - "loss": 0.4682, - "step": 2083 - }, - { - "epoch": 0.65, - "learning_rate": 2.8133117650903347e-07, - "loss": 0.4808, - "step": 2084 - }, - { - "epoch": 0.65, - "learning_rate": 2.8088397223643767e-07, - "loss": 0.4826, - "step": 2085 - }, - { - "epoch": 0.65, - "learning_rate": 2.8043698480893744e-07, - "loss": 0.4562, - "step": 2086 - }, - { - "epoch": 0.65, - "learning_rate": 2.799902146688878e-07, - "loss": 0.4469, - "step": 2087 - }, - { - "epoch": 0.65, - "learning_rate": 2.795436622584287e-07, - "loss": 0.4581, - "step": 2088 - }, - { - "epoch": 0.65, - "learning_rate": 2.7909732801948414e-07, - "loss": 0.4462, - "step": 2089 - }, - { - "epoch": 0.65, - "learning_rate": 2.7865121239376267e-07, - "loss": 0.4795, - "step": 2090 - }, - { - "epoch": 0.65, - "learning_rate": 2.782053158227563e-07, - "loss": 0.4671, - "step": 2091 - }, - { - "epoch": 0.65, - "learning_rate": 2.7775963874774055e-07, - "loss": 0.4718, - "step": 2092 - }, - { - "epoch": 0.65, - "learning_rate": 2.7731418160977294e-07, - "loss": 0.4716, - "step": 2093 - }, - { - "epoch": 0.65, - "learning_rate": 2.768689448496945e-07, - "loss": 0.4671, - "step": 2094 - }, - { - "epoch": 0.65, - "learning_rate": 2.7642392890812704e-07, - "loss": 0.4775, - "step": 2095 - }, - { - "epoch": 0.65, - "learning_rate": 2.759791342254747e-07, - "loss": 0.4669, - "step": 2096 - }, - { - "epoch": 0.65, - "learning_rate": 2.75534561241922e-07, - "loss": 0.4669, - "step": 2097 - }, - { - "epoch": 0.65, - "learning_rate": 2.750902103974343e-07, - "loss": 0.4546, - "step": 2098 - }, - { - "epoch": 0.65, - "learning_rate": 2.746460821317575e-07, - "loss": 0.4181, - "step": 2099 - }, - { - "epoch": 0.65, - "learning_rate": 2.7420217688441655e-07, - "loss": 0.4819, - "step": 2100 - }, - { - "epoch": 0.65, - "learning_rate": 2.7375849509471615e-07, - "loss": 0.4467, - "step": 2101 - }, - { - "epoch": 0.66, - "learning_rate": 2.7331503720173975e-07, - "loss": 0.4788, - "step": 2102 - }, - { - "epoch": 0.66, - "learning_rate": 2.7287180364434946e-07, - "loss": 0.4438, - "step": 2103 - }, - { - "epoch": 0.66, - "learning_rate": 2.724287948611845e-07, - "loss": 0.4819, - "step": 2104 - }, - { - "epoch": 0.66, - "learning_rate": 2.719860112906632e-07, - "loss": 0.4725, - "step": 2105 - }, - { - "epoch": 0.66, - "learning_rate": 2.715434533709793e-07, - "loss": 0.4601, - "step": 2106 - }, - { - "epoch": 0.66, - "learning_rate": 2.711011215401045e-07, - "loss": 0.4809, - "step": 2107 - }, - { - "epoch": 0.66, - "learning_rate": 2.706590162357859e-07, - "loss": 0.4741, - "step": 2108 - }, - { - "epoch": 0.66, - "learning_rate": 2.70217137895547e-07, - "loss": 0.4469, - "step": 2109 - }, - { - "epoch": 0.66, - "learning_rate": 2.697754869566865e-07, - "loss": 0.4661, - "step": 2110 - }, - { - "epoch": 0.66, - "learning_rate": 2.693340638562782e-07, - "loss": 0.4647, - "step": 2111 - }, - { - "epoch": 0.66, - "learning_rate": 2.688928690311699e-07, - "loss": 0.4754, - "step": 2112 - }, - { - "epoch": 0.66, - "learning_rate": 2.6845190291798404e-07, - "loss": 0.4442, - "step": 2113 - }, - { - "epoch": 0.66, - "learning_rate": 2.680111659531167e-07, - "loss": 0.4531, - "step": 2114 - }, - { - "epoch": 0.66, - "learning_rate": 2.675706585727364e-07, - "loss": 0.4681, - "step": 2115 - }, - { - "epoch": 0.66, - "learning_rate": 2.6713038121278605e-07, - "loss": 0.4617, - "step": 2116 - }, - { - "epoch": 0.66, - "learning_rate": 2.6669033430897915e-07, - "loss": 0.4668, - "step": 2117 - }, - { - "epoch": 0.66, - "learning_rate": 2.662505182968023e-07, - "loss": 0.4437, - "step": 2118 - }, - { - "epoch": 0.66, - "learning_rate": 2.658109336115129e-07, - "loss": 0.4536, - "step": 2119 - }, - { - "epoch": 0.66, - "learning_rate": 2.653715806881404e-07, - "loss": 0.4858, - "step": 2120 - }, - { - "epoch": 0.66, - "learning_rate": 2.649324599614835e-07, - "loss": 0.4635, - "step": 2121 - }, - { - "epoch": 0.66, - "learning_rate": 2.644935718661122e-07, - "loss": 0.4563, - "step": 2122 - }, - { - "epoch": 0.66, - "learning_rate": 2.6405491683636613e-07, - "loss": 0.4694, - "step": 2123 - }, - { - "epoch": 0.66, - "learning_rate": 2.636164953063538e-07, - "loss": 0.4723, - "step": 2124 - }, - { - "epoch": 0.66, - "learning_rate": 2.6317830770995295e-07, - "loss": 0.4836, - "step": 2125 - }, - { - "epoch": 0.66, - "learning_rate": 2.6274035448080957e-07, - "loss": 0.4466, - "step": 2126 - }, - { - "epoch": 0.66, - "learning_rate": 2.6230263605233837e-07, - "loss": 0.4561, - "step": 2127 - }, - { - "epoch": 0.66, - "learning_rate": 2.618651528577209e-07, - "loss": 0.461, - "step": 2128 - }, - { - "epoch": 0.66, - "learning_rate": 2.614279053299065e-07, - "loss": 0.4541, - "step": 2129 - }, - { - "epoch": 0.66, - "learning_rate": 2.609908939016107e-07, - "loss": 0.4509, - "step": 2130 - }, - { - "epoch": 0.66, - "learning_rate": 2.605541190053161e-07, - "loss": 0.4674, - "step": 2131 - }, - { - "epoch": 0.66, - "learning_rate": 2.6011758107327046e-07, - "loss": 0.4925, - "step": 2132 - }, - { - "epoch": 0.66, - "learning_rate": 2.596812805374876e-07, - "loss": 0.4868, - "step": 2133 - }, - { - "epoch": 0.67, - "learning_rate": 2.5924521782974644e-07, - "loss": 0.4471, - "step": 2134 - }, - { - "epoch": 0.67, - "learning_rate": 2.588093933815899e-07, - "loss": 0.435, - "step": 2135 - }, - { - "epoch": 0.67, - "learning_rate": 2.583738076243256e-07, - "loss": 0.4474, - "step": 2136 - }, - { - "epoch": 0.67, - "learning_rate": 2.57938460989025e-07, - "loss": 0.4526, - "step": 2137 - }, - { - "epoch": 0.67, - "learning_rate": 2.575033539065229e-07, - "loss": 0.4411, - "step": 2138 - }, - { - "epoch": 0.67, - "learning_rate": 2.570684868074165e-07, - "loss": 0.4864, - "step": 2139 - }, - { - "epoch": 0.67, - "learning_rate": 2.5663386012206636e-07, - "loss": 0.4598, - "step": 2140 - }, - { - "epoch": 0.67, - "learning_rate": 2.5619947428059404e-07, - "loss": 0.4512, - "step": 2141 - }, - { - "epoch": 0.67, - "learning_rate": 2.557653297128842e-07, - "loss": 0.4731, - "step": 2142 - }, - { - "epoch": 0.67, - "learning_rate": 2.553314268485812e-07, - "loss": 0.4498, - "step": 2143 - }, - { - "epoch": 0.67, - "learning_rate": 2.548977661170913e-07, - "loss": 0.4547, - "step": 2144 - }, - { - "epoch": 0.67, - "learning_rate": 2.544643479475806e-07, - "loss": 0.4645, - "step": 2145 - }, - { - "epoch": 0.67, - "learning_rate": 2.5403117276897555e-07, - "loss": 0.4566, - "step": 2146 - }, - { - "epoch": 0.67, - "learning_rate": 2.535982410099614e-07, - "loss": 0.4705, - "step": 2147 - }, - { - "epoch": 0.67, - "learning_rate": 2.5316555309898337e-07, - "loss": 0.4729, - "step": 2148 - }, - { - "epoch": 0.67, - "learning_rate": 2.5273310946424496e-07, - "loss": 0.4655, - "step": 2149 - }, - { - "epoch": 0.67, - "learning_rate": 2.523009105337076e-07, - "loss": 0.4777, - "step": 2150 - }, - { - "epoch": 0.67, - "learning_rate": 2.5186895673509137e-07, - "loss": 0.4644, - "step": 2151 - }, - { - "epoch": 0.67, - "learning_rate": 2.5143724849587247e-07, - "loss": 0.4642, - "step": 2152 - }, - { - "epoch": 0.67, - "learning_rate": 2.510057862432859e-07, - "loss": 0.4699, - "step": 2153 - }, - { - "epoch": 0.67, - "learning_rate": 2.505745704043215e-07, - "loss": 0.4617, - "step": 2154 - }, - { - "epoch": 0.67, - "learning_rate": 2.501436014057262e-07, - "loss": 0.46, - "step": 2155 - }, - { - "epoch": 0.67, - "learning_rate": 2.4971287967400237e-07, - "loss": 0.4314, - "step": 2156 - }, - { - "epoch": 0.67, - "learning_rate": 2.4928240563540797e-07, - "loss": 0.4663, - "step": 2157 - }, - { - "epoch": 0.67, - "learning_rate": 2.488521797159552e-07, - "loss": 0.4747, - "step": 2158 - }, - { - "epoch": 0.67, - "learning_rate": 2.484222023414111e-07, - "loss": 0.4548, - "step": 2159 - }, - { - "epoch": 0.67, - "learning_rate": 2.4799247393729723e-07, - "loss": 0.4744, - "step": 2160 - }, - { - "epoch": 0.67, - "learning_rate": 2.4756299492888765e-07, - "loss": 0.4732, - "step": 2161 - }, - { - "epoch": 0.67, - "learning_rate": 2.471337657412105e-07, - "loss": 0.4665, - "step": 2162 - }, - { - "epoch": 0.67, - "learning_rate": 2.467047867990463e-07, - "loss": 0.4683, - "step": 2163 - }, - { - "epoch": 0.67, - "learning_rate": 2.462760585269284e-07, - "loss": 0.4588, - "step": 2164 - }, - { - "epoch": 0.67, - "learning_rate": 2.458475813491412e-07, - "loss": 0.4352, - "step": 2165 - }, - { - "epoch": 0.68, - "learning_rate": 2.454193556897214e-07, - "loss": 0.4703, - "step": 2166 - }, - { - "epoch": 0.68, - "learning_rate": 2.449913819724564e-07, - "loss": 0.4584, - "step": 2167 - }, - { - "epoch": 0.68, - "learning_rate": 2.445636606208846e-07, - "loss": 0.4385, - "step": 2168 - }, - { - "epoch": 0.68, - "learning_rate": 2.44136192058294e-07, - "loss": 0.4542, - "step": 2169 - }, - { - "epoch": 0.68, - "learning_rate": 2.4370897670772313e-07, - "loss": 0.4462, - "step": 2170 - }, - { - "epoch": 0.68, - "learning_rate": 2.432820149919596e-07, - "loss": 0.4427, - "step": 2171 - }, - { - "epoch": 0.68, - "learning_rate": 2.4285530733354025e-07, - "loss": 0.4382, - "step": 2172 - }, - { - "epoch": 0.68, - "learning_rate": 2.4242885415474996e-07, - "loss": 0.4574, - "step": 2173 - }, - { - "epoch": 0.68, - "learning_rate": 2.420026558776223e-07, - "loss": 0.4583, - "step": 2174 - }, - { - "epoch": 0.68, - "learning_rate": 2.415767129239386e-07, - "loss": 0.4602, - "step": 2175 - }, - { - "epoch": 0.68, - "learning_rate": 2.411510257152269e-07, - "loss": 0.4515, - "step": 2176 - }, - { - "epoch": 0.68, - "learning_rate": 2.4072559467276287e-07, - "loss": 0.4398, - "step": 2177 - }, - { - "epoch": 0.68, - "learning_rate": 2.403004202175682e-07, - "loss": 0.4541, - "step": 2178 - }, - { - "epoch": 0.68, - "learning_rate": 2.398755027704111e-07, - "loss": 0.4582, - "step": 2179 - }, - { - "epoch": 0.68, - "learning_rate": 2.394508427518047e-07, - "loss": 0.4747, - "step": 2180 - }, - { - "epoch": 0.68, - "learning_rate": 2.3902644058200803e-07, - "loss": 0.4795, - "step": 2181 - }, - { - "epoch": 0.68, - "learning_rate": 2.386022966810248e-07, - "loss": 0.4627, - "step": 2182 - }, - { - "epoch": 0.68, - "learning_rate": 2.3817841146860318e-07, - "loss": 0.4643, - "step": 2183 - }, - { - "epoch": 0.68, - "learning_rate": 2.3775478536423482e-07, - "loss": 0.4587, - "step": 2184 - }, - { - "epoch": 0.68, - "learning_rate": 2.373314187871555e-07, - "loss": 0.4677, - "step": 2185 - }, - { - "epoch": 0.68, - "learning_rate": 2.3690831215634432e-07, - "loss": 0.4768, - "step": 2186 - }, - { - "epoch": 0.68, - "learning_rate": 2.3648546589052233e-07, - "loss": 0.4537, - "step": 2187 - }, - { - "epoch": 0.68, - "learning_rate": 2.3606288040815357e-07, - "loss": 0.4548, - "step": 2188 - }, - { - "epoch": 0.68, - "learning_rate": 2.3564055612744383e-07, - "loss": 0.4596, - "step": 2189 - }, - { - "epoch": 0.68, - "learning_rate": 2.3521849346634066e-07, - "loss": 0.475, - "step": 2190 - }, - { - "epoch": 0.68, - "learning_rate": 2.3479669284253195e-07, - "loss": 0.4411, - "step": 2191 - }, - { - "epoch": 0.68, - "learning_rate": 2.3437515467344697e-07, - "loss": 0.4545, - "step": 2192 - }, - { - "epoch": 0.68, - "learning_rate": 2.3395387937625515e-07, - "loss": 0.4759, - "step": 2193 - }, - { - "epoch": 0.68, - "learning_rate": 2.335328673678657e-07, - "loss": 0.4714, - "step": 2194 - }, - { - "epoch": 0.68, - "learning_rate": 2.3311211906492678e-07, - "loss": 0.4612, - "step": 2195 - }, - { - "epoch": 0.68, - "learning_rate": 2.326916348838267e-07, - "loss": 0.4619, - "step": 2196 - }, - { - "epoch": 0.68, - "learning_rate": 2.3227141524069127e-07, - "loss": 0.4589, - "step": 2197 - }, - { - "epoch": 0.69, - "learning_rate": 2.3185146055138515e-07, - "loss": 0.4656, - "step": 2198 - }, - { - "epoch": 0.69, - "learning_rate": 2.3143177123151043e-07, - "loss": 0.4572, - "step": 2199 - }, - { - "epoch": 0.69, - "learning_rate": 2.3101234769640676e-07, - "loss": 0.4844, - "step": 2200 - }, - { - "epoch": 0.69, - "learning_rate": 2.3059319036115098e-07, - "loss": 0.4764, - "step": 2201 - }, - { - "epoch": 0.69, - "learning_rate": 2.3017429964055591e-07, - "loss": 0.4604, - "step": 2202 - }, - { - "epoch": 0.69, - "learning_rate": 2.2975567594917106e-07, - "loss": 0.4673, - "step": 2203 - }, - { - "epoch": 0.69, - "learning_rate": 2.293373197012814e-07, - "loss": 0.4567, - "step": 2204 - }, - { - "epoch": 0.69, - "learning_rate": 2.2891923131090758e-07, - "loss": 0.467, - "step": 2205 - }, - { - "epoch": 0.69, - "learning_rate": 2.2850141119180428e-07, - "loss": 0.4487, - "step": 2206 - }, - { - "epoch": 0.69, - "learning_rate": 2.2808385975746208e-07, - "loss": 0.4622, - "step": 2207 - }, - { - "epoch": 0.69, - "learning_rate": 2.2766657742110428e-07, - "loss": 0.4554, - "step": 2208 - }, - { - "epoch": 0.69, - "learning_rate": 2.2724956459568895e-07, - "loss": 0.4644, - "step": 2209 - }, - { - "epoch": 0.69, - "learning_rate": 2.268328216939065e-07, - "loss": 0.4659, - "step": 2210 - }, - { - "epoch": 0.69, - "learning_rate": 2.2641634912818092e-07, - "loss": 0.4583, - "step": 2211 - }, - { - "epoch": 0.69, - "learning_rate": 2.2600014731066862e-07, - "loss": 0.4335, - "step": 2212 - }, - { - "epoch": 0.69, - "learning_rate": 2.2558421665325728e-07, - "loss": 0.4616, - "step": 2213 - }, - { - "epoch": 0.69, - "learning_rate": 2.2516855756756753e-07, - "loss": 0.4584, - "step": 2214 - }, - { - "epoch": 0.69, - "learning_rate": 2.2475317046495013e-07, - "loss": 0.4494, - "step": 2215 - }, - { - "epoch": 0.69, - "learning_rate": 2.243380557564873e-07, - "loss": 0.4647, - "step": 2216 - }, - { - "epoch": 0.69, - "learning_rate": 2.2392321385299102e-07, - "loss": 0.4863, - "step": 2217 - }, - { - "epoch": 0.69, - "learning_rate": 2.235086451650044e-07, - "loss": 0.4572, - "step": 2218 - }, - { - "epoch": 0.69, - "learning_rate": 2.2309435010279904e-07, - "loss": 0.465, - "step": 2219 - }, - { - "epoch": 0.69, - "learning_rate": 2.2268032907637657e-07, - "loss": 0.4835, - "step": 2220 - }, - { - "epoch": 0.69, - "learning_rate": 2.2226658249546648e-07, - "loss": 0.467, - "step": 2221 - }, - { - "epoch": 0.69, - "learning_rate": 2.2185311076952805e-07, - "loss": 0.4724, - "step": 2222 - }, - { - "epoch": 0.69, - "learning_rate": 2.2143991430774717e-07, - "loss": 0.4853, - "step": 2223 - }, - { - "epoch": 0.69, - "learning_rate": 2.2102699351903814e-07, - "loss": 0.4601, - "step": 2224 - }, - { - "epoch": 0.69, - "learning_rate": 2.2061434881204232e-07, - "loss": 0.4516, - "step": 2225 - }, - { - "epoch": 0.69, - "learning_rate": 2.202019805951274e-07, - "loss": 0.4742, - "step": 2226 - }, - { - "epoch": 0.69, - "learning_rate": 2.197898892763882e-07, - "loss": 0.4768, - "step": 2227 - }, - { - "epoch": 0.69, - "learning_rate": 2.1937807526364454e-07, - "loss": 0.4593, - "step": 2228 - }, - { - "epoch": 0.69, - "learning_rate": 2.1896653896444307e-07, - "loss": 0.4839, - "step": 2229 - }, - { - "epoch": 0.7, - "learning_rate": 2.1855528078605435e-07, - "loss": 0.4701, - "step": 2230 - }, - { - "epoch": 0.7, - "learning_rate": 2.181443011354746e-07, - "loss": 0.4691, - "step": 2231 - }, - { - "epoch": 0.7, - "learning_rate": 2.1773360041942357e-07, - "loss": 0.4369, - "step": 2232 - }, - { - "epoch": 0.7, - "learning_rate": 2.1732317904434615e-07, - "loss": 0.4637, - "step": 2233 - }, - { - "epoch": 0.7, - "learning_rate": 2.1691303741640953e-07, - "loss": 0.4716, - "step": 2234 - }, - { - "epoch": 0.7, - "learning_rate": 2.1650317594150474e-07, - "loss": 0.4494, - "step": 2235 - }, - { - "epoch": 0.7, - "learning_rate": 2.160935950252457e-07, - "loss": 0.4583, - "step": 2236 - }, - { - "epoch": 0.7, - "learning_rate": 2.1568429507296797e-07, - "loss": 0.4538, - "step": 2237 - }, - { - "epoch": 0.7, - "learning_rate": 2.1527527648972993e-07, - "loss": 0.4588, - "step": 2238 - }, - { - "epoch": 0.7, - "learning_rate": 2.1486653968031043e-07, - "loss": 0.4779, - "step": 2239 - }, - { - "epoch": 0.7, - "learning_rate": 2.1445808504921086e-07, - "loss": 0.462, - "step": 2240 - }, - { - "epoch": 0.7, - "learning_rate": 2.1404991300065213e-07, - "loss": 0.4547, - "step": 2241 - }, - { - "epoch": 0.7, - "learning_rate": 2.1364202393857622e-07, - "loss": 0.4487, - "step": 2242 - }, - { - "epoch": 0.7, - "learning_rate": 2.1323441826664435e-07, - "loss": 0.4568, - "step": 2243 - }, - { - "epoch": 0.7, - "learning_rate": 2.1282709638823843e-07, - "loss": 0.4718, - "step": 2244 - }, - { - "epoch": 0.7, - "learning_rate": 2.1242005870645824e-07, - "loss": 0.4639, - "step": 2245 - }, - { - "epoch": 0.7, - "learning_rate": 2.1201330562412318e-07, - "loss": 0.4475, - "step": 2246 - }, - { - "epoch": 0.7, - "learning_rate": 2.1160683754377063e-07, - "loss": 0.4674, - "step": 2247 - }, - { - "epoch": 0.7, - "learning_rate": 2.1120065486765627e-07, - "loss": 0.4438, - "step": 2248 - }, - { - "epoch": 0.7, - "learning_rate": 2.1079475799775266e-07, - "loss": 0.4551, - "step": 2249 - }, - { - "epoch": 0.7, - "learning_rate": 2.103891473357501e-07, - "loss": 0.4714, - "step": 2250 - }, - { - "epoch": 0.7, - "learning_rate": 2.099838232830557e-07, - "loss": 0.454, - "step": 2251 - }, - { - "epoch": 0.7, - "learning_rate": 2.095787862407923e-07, - "loss": 0.4643, - "step": 2252 - }, - { - "epoch": 0.7, - "learning_rate": 2.091740366097993e-07, - "loss": 0.4532, - "step": 2253 - }, - { - "epoch": 0.7, - "learning_rate": 2.0876957479063156e-07, - "loss": 0.4677, - "step": 2254 - }, - { - "epoch": 0.7, - "learning_rate": 2.0836540118355916e-07, - "loss": 0.461, - "step": 2255 - }, - { - "epoch": 0.7, - "learning_rate": 2.0796151618856638e-07, - "loss": 0.4513, - "step": 2256 - }, - { - "epoch": 0.7, - "learning_rate": 2.0755792020535267e-07, - "loss": 0.4465, - "step": 2257 - }, - { - "epoch": 0.7, - "learning_rate": 2.0715461363333099e-07, - "loss": 0.4534, - "step": 2258 - }, - { - "epoch": 0.7, - "learning_rate": 2.067515968716283e-07, - "loss": 0.4449, - "step": 2259 - }, - { - "epoch": 0.7, - "learning_rate": 2.0634887031908398e-07, - "loss": 0.457, - "step": 2260 - }, - { - "epoch": 0.7, - "learning_rate": 2.0594643437425102e-07, - "loss": 0.4689, - "step": 2261 - }, - { - "epoch": 0.71, - "learning_rate": 2.0554428943539454e-07, - "loss": 0.4675, - "step": 2262 - }, - { - "epoch": 0.71, - "learning_rate": 2.051424359004914e-07, - "loss": 0.4573, - "step": 2263 - }, - { - "epoch": 0.71, - "learning_rate": 2.0474087416723034e-07, - "loss": 0.4525, - "step": 2264 - }, - { - "epoch": 0.71, - "learning_rate": 2.0433960463301142e-07, - "loss": 0.4536, - "step": 2265 - }, - { - "epoch": 0.71, - "learning_rate": 2.0393862769494552e-07, - "loss": 0.4684, - "step": 2266 - }, - { - "epoch": 0.71, - "learning_rate": 2.035379437498534e-07, - "loss": 0.4528, - "step": 2267 - }, - { - "epoch": 0.71, - "learning_rate": 2.0313755319426655e-07, - "loss": 0.4769, - "step": 2268 - }, - { - "epoch": 0.71, - "learning_rate": 2.0273745642442585e-07, - "loss": 0.4694, - "step": 2269 - }, - { - "epoch": 0.71, - "learning_rate": 2.023376538362816e-07, - "loss": 0.4517, - "step": 2270 - }, - { - "epoch": 0.71, - "learning_rate": 2.019381458254924e-07, - "loss": 0.4717, - "step": 2271 - }, - { - "epoch": 0.71, - "learning_rate": 2.01538932787426e-07, - "loss": 0.4606, - "step": 2272 - }, - { - "epoch": 0.71, - "learning_rate": 2.01140015117158e-07, - "loss": 0.4767, - "step": 2273 - }, - { - "epoch": 0.71, - "learning_rate": 2.007413932094717e-07, - "loss": 0.4755, - "step": 2274 - }, - { - "epoch": 0.71, - "learning_rate": 2.0034306745885743e-07, - "loss": 0.4628, - "step": 2275 - }, - { - "epoch": 0.71, - "learning_rate": 1.9994503825951275e-07, - "loss": 0.4405, - "step": 2276 - }, - { - "epoch": 0.71, - "learning_rate": 1.9954730600534198e-07, - "loss": 0.4487, - "step": 2277 - }, - { - "epoch": 0.71, - "learning_rate": 1.9914987108995473e-07, - "loss": 0.4607, - "step": 2278 - }, - { - "epoch": 0.71, - "learning_rate": 1.9875273390666718e-07, - "loss": 0.4774, - "step": 2279 - }, - { - "epoch": 0.71, - "learning_rate": 1.983558948485005e-07, - "loss": 0.4757, - "step": 2280 - }, - { - "epoch": 0.71, - "learning_rate": 1.9795935430818105e-07, - "loss": 0.4733, - "step": 2281 - }, - { - "epoch": 0.71, - "learning_rate": 1.975631126781393e-07, - "loss": 0.4631, - "step": 2282 - }, - { - "epoch": 0.71, - "learning_rate": 1.9716717035051035e-07, - "loss": 0.4516, - "step": 2283 - }, - { - "epoch": 0.71, - "learning_rate": 1.9677152771713297e-07, - "loss": 0.4606, - "step": 2284 - }, - { - "epoch": 0.71, - "learning_rate": 1.9637618516954947e-07, - "loss": 0.4585, - "step": 2285 - }, - { - "epoch": 0.71, - "learning_rate": 1.959811430990047e-07, - "loss": 0.462, - "step": 2286 - }, - { - "epoch": 0.71, - "learning_rate": 1.9558640189644672e-07, - "loss": 0.442, - "step": 2287 - }, - { - "epoch": 0.71, - "learning_rate": 1.9519196195252575e-07, - "loss": 0.4598, - "step": 2288 - }, - { - "epoch": 0.71, - "learning_rate": 1.9479782365759333e-07, - "loss": 0.4612, - "step": 2289 - }, - { - "epoch": 0.71, - "learning_rate": 1.9440398740170316e-07, - "loss": 0.4569, - "step": 2290 - }, - { - "epoch": 0.71, - "learning_rate": 1.940104535746097e-07, - "loss": 0.4542, - "step": 2291 - }, - { - "epoch": 0.71, - "learning_rate": 1.9361722256576834e-07, - "loss": 0.4673, - "step": 2292 - }, - { - "epoch": 0.71, - "learning_rate": 1.9322429476433404e-07, - "loss": 0.4524, - "step": 2293 - }, - { - "epoch": 0.72, - "learning_rate": 1.9283167055916312e-07, - "loss": 0.4372, - "step": 2294 - }, - { - "epoch": 0.72, - "learning_rate": 1.9243935033880992e-07, - "loss": 0.4717, - "step": 2295 - }, - { - "epoch": 0.72, - "learning_rate": 1.920473344915291e-07, - "loss": 0.4615, - "step": 2296 - }, - { - "epoch": 0.72, - "learning_rate": 1.916556234052732e-07, - "loss": 0.4737, - "step": 2297 - }, - { - "epoch": 0.72, - "learning_rate": 1.9126421746769377e-07, - "loss": 0.4749, - "step": 2298 - }, - { - "epoch": 0.72, - "learning_rate": 1.9087311706614024e-07, - "loss": 0.4782, - "step": 2299 - }, - { - "epoch": 0.72, - "learning_rate": 1.904823225876598e-07, - "loss": 0.4658, - "step": 2300 - }, - { - "epoch": 0.72, - "learning_rate": 1.900918344189963e-07, - "loss": 0.4406, - "step": 2301 - }, - { - "epoch": 0.72, - "learning_rate": 1.8970165294659108e-07, - "loss": 0.4599, - "step": 2302 - }, - { - "epoch": 0.72, - "learning_rate": 1.8931177855658192e-07, - "loss": 0.4448, - "step": 2303 - }, - { - "epoch": 0.72, - "learning_rate": 1.88922211634802e-07, - "loss": 0.447, - "step": 2304 - }, - { - "epoch": 0.72, - "learning_rate": 1.8853295256678148e-07, - "loss": 0.4528, - "step": 2305 - }, - { - "epoch": 0.72, - "learning_rate": 1.8814400173774452e-07, - "loss": 0.4584, - "step": 2306 - }, - { - "epoch": 0.72, - "learning_rate": 1.8775535953261123e-07, - "loss": 0.4762, - "step": 2307 - }, - { - "epoch": 0.72, - "learning_rate": 1.873670263359954e-07, - "loss": 0.4592, - "step": 2308 - }, - { - "epoch": 0.72, - "learning_rate": 1.869790025322062e-07, - "loss": 0.4686, - "step": 2309 - }, - { - "epoch": 0.72, - "learning_rate": 1.865912885052454e-07, - "loss": 0.4504, - "step": 2310 - }, - { - "epoch": 0.72, - "learning_rate": 1.8620388463880908e-07, - "loss": 0.4557, - "step": 2311 - }, - { - "epoch": 0.72, - "learning_rate": 1.858167913162857e-07, - "loss": 0.4562, - "step": 2312 - }, - { - "epoch": 0.72, - "learning_rate": 1.8543000892075688e-07, - "loss": 0.4653, - "step": 2313 - }, - { - "epoch": 0.72, - "learning_rate": 1.8504353783499654e-07, - "loss": 0.4604, - "step": 2314 - }, - { - "epoch": 0.72, - "learning_rate": 1.846573784414699e-07, - "loss": 0.4586, - "step": 2315 - }, - { - "epoch": 0.72, - "learning_rate": 1.842715311223348e-07, - "loss": 0.4605, - "step": 2316 - }, - { - "epoch": 0.72, - "learning_rate": 1.8388599625943912e-07, - "loss": 0.4543, - "step": 2317 - }, - { - "epoch": 0.72, - "learning_rate": 1.8350077423432237e-07, - "loss": 0.4443, - "step": 2318 - }, - { - "epoch": 0.72, - "learning_rate": 1.8311586542821356e-07, - "loss": 0.4582, - "step": 2319 - }, - { - "epoch": 0.72, - "learning_rate": 1.82731270222033e-07, - "loss": 0.4844, - "step": 2320 - }, - { - "epoch": 0.72, - "learning_rate": 1.823469889963894e-07, - "loss": 0.4612, - "step": 2321 - }, - { - "epoch": 0.72, - "learning_rate": 1.819630221315816e-07, - "loss": 0.4822, - "step": 2322 - }, - { - "epoch": 0.72, - "learning_rate": 1.8157937000759665e-07, - "loss": 0.4675, - "step": 2323 - }, - { - "epoch": 0.72, - "learning_rate": 1.8119603300411075e-07, - "loss": 0.4625, - "step": 2324 - }, - { - "epoch": 0.72, - "learning_rate": 1.8081301150048785e-07, - "loss": 0.4632, - "step": 2325 - }, - { - "epoch": 0.73, - "learning_rate": 1.804303058757799e-07, - "loss": 0.453, - "step": 2326 - }, - { - "epoch": 0.73, - "learning_rate": 1.800479165087262e-07, - "loss": 0.4673, - "step": 2327 - }, - { - "epoch": 0.73, - "learning_rate": 1.796658437777528e-07, - "loss": 0.467, - "step": 2328 - }, - { - "epoch": 0.73, - "learning_rate": 1.7928408806097283e-07, - "loss": 0.4709, - "step": 2329 - }, - { - "epoch": 0.73, - "learning_rate": 1.7890264973618502e-07, - "loss": 0.455, - "step": 2330 - }, - { - "epoch": 0.73, - "learning_rate": 1.785215291808751e-07, - "loss": 0.4647, - "step": 2331 - }, - { - "epoch": 0.73, - "learning_rate": 1.7814072677221314e-07, - "loss": 0.4485, - "step": 2332 - }, - { - "epoch": 0.73, - "learning_rate": 1.777602428870553e-07, - "loss": 0.4611, - "step": 2333 - }, - { - "epoch": 0.73, - "learning_rate": 1.7738007790194155e-07, - "loss": 0.4482, - "step": 2334 - }, - { - "epoch": 0.73, - "learning_rate": 1.770002321930975e-07, - "loss": 0.4571, - "step": 2335 - }, - { - "epoch": 0.73, - "learning_rate": 1.766207061364316e-07, - "loss": 0.453, - "step": 2336 - }, - { - "epoch": 0.73, - "learning_rate": 1.7624150010753664e-07, - "loss": 0.469, - "step": 2337 - }, - { - "epoch": 0.73, - "learning_rate": 1.758626144816887e-07, - "loss": 0.4779, - "step": 2338 - }, - { - "epoch": 0.73, - "learning_rate": 1.7548404963384633e-07, - "loss": 0.4485, - "step": 2339 - }, - { - "epoch": 0.73, - "learning_rate": 1.751058059386512e-07, - "loss": 0.4678, - "step": 2340 - }, - { - "epoch": 0.73, - "learning_rate": 1.7472788377042641e-07, - "loss": 0.478, - "step": 2341 - }, - { - "epoch": 0.73, - "learning_rate": 1.7435028350317787e-07, - "loss": 0.457, - "step": 2342 - }, - { - "epoch": 0.73, - "learning_rate": 1.7397300551059203e-07, - "loss": 0.4681, - "step": 2343 - }, - { - "epoch": 0.73, - "learning_rate": 1.7359605016603679e-07, - "loss": 0.4494, - "step": 2344 - }, - { - "epoch": 0.73, - "learning_rate": 1.7321941784256073e-07, - "loss": 0.4669, - "step": 2345 - }, - { - "epoch": 0.73, - "learning_rate": 1.7284310891289294e-07, - "loss": 0.461, - "step": 2346 - }, - { - "epoch": 0.73, - "learning_rate": 1.7246712374944194e-07, - "loss": 0.4372, - "step": 2347 - }, - { - "epoch": 0.73, - "learning_rate": 1.7209146272429624e-07, - "loss": 0.4596, - "step": 2348 - }, - { - "epoch": 0.73, - "learning_rate": 1.7171612620922376e-07, - "loss": 0.4678, - "step": 2349 - }, - { - "epoch": 0.73, - "learning_rate": 1.7134111457567063e-07, - "loss": 0.4655, - "step": 2350 - }, - { - "epoch": 0.73, - "learning_rate": 1.7096642819476198e-07, - "loss": 0.4561, - "step": 2351 - }, - { - "epoch": 0.73, - "learning_rate": 1.7059206743730098e-07, - "loss": 0.4309, - "step": 2352 - }, - { - "epoch": 0.73, - "learning_rate": 1.702180326737686e-07, - "loss": 0.4633, - "step": 2353 - }, - { - "epoch": 0.73, - "learning_rate": 1.6984432427432277e-07, - "loss": 0.4575, - "step": 2354 - }, - { - "epoch": 0.73, - "learning_rate": 1.694709426087989e-07, - "loss": 0.441, - "step": 2355 - }, - { - "epoch": 0.73, - "learning_rate": 1.6909788804670893e-07, - "loss": 0.4792, - "step": 2356 - }, - { - "epoch": 0.73, - "learning_rate": 1.6872516095724122e-07, - "loss": 0.4929, - "step": 2357 - }, - { - "epoch": 0.74, - "learning_rate": 1.6835276170925955e-07, - "loss": 0.4512, - "step": 2358 - }, - { - "epoch": 0.74, - "learning_rate": 1.6798069067130377e-07, - "loss": 0.4538, - "step": 2359 - }, - { - "epoch": 0.74, - "learning_rate": 1.6760894821158878e-07, - "loss": 0.4709, - "step": 2360 - }, - { - "epoch": 0.74, - "learning_rate": 1.672375346980044e-07, - "loss": 0.4641, - "step": 2361 - }, - { - "epoch": 0.74, - "learning_rate": 1.6686645049811448e-07, - "loss": 0.4662, - "step": 2362 - }, - { - "epoch": 0.74, - "learning_rate": 1.6649569597915748e-07, - "loss": 0.4372, - "step": 2363 - }, - { - "epoch": 0.74, - "learning_rate": 1.661252715080455e-07, - "loss": 0.4633, - "step": 2364 - }, - { - "epoch": 0.74, - "learning_rate": 1.6575517745136363e-07, - "loss": 0.4598, - "step": 2365 - }, - { - "epoch": 0.74, - "learning_rate": 1.6538541417537038e-07, - "loss": 0.4621, - "step": 2366 - }, - { - "epoch": 0.74, - "learning_rate": 1.6501598204599676e-07, - "loss": 0.467, - "step": 2367 - }, - { - "epoch": 0.74, - "learning_rate": 1.6464688142884632e-07, - "loss": 0.4714, - "step": 2368 - }, - { - "epoch": 0.74, - "learning_rate": 1.6427811268919388e-07, - "loss": 0.4736, - "step": 2369 - }, - { - "epoch": 0.74, - "learning_rate": 1.6390967619198648e-07, - "loss": 0.4451, - "step": 2370 - }, - { - "epoch": 0.74, - "learning_rate": 1.6354157230184208e-07, - "loss": 0.4818, - "step": 2371 - }, - { - "epoch": 0.74, - "learning_rate": 1.6317380138304966e-07, - "loss": 0.4532, - "step": 2372 - }, - { - "epoch": 0.74, - "learning_rate": 1.6280636379956824e-07, - "loss": 0.4856, - "step": 2373 - }, - { - "epoch": 0.74, - "learning_rate": 1.6243925991502738e-07, - "loss": 0.4456, - "step": 2374 - }, - { - "epoch": 0.74, - "learning_rate": 1.6207249009272655e-07, - "loss": 0.4373, - "step": 2375 - }, - { - "epoch": 0.74, - "learning_rate": 1.6170605469563392e-07, - "loss": 0.4444, - "step": 2376 - }, - { - "epoch": 0.74, - "learning_rate": 1.6133995408638728e-07, - "loss": 0.4742, - "step": 2377 - }, - { - "epoch": 0.74, - "learning_rate": 1.6097418862729306e-07, - "loss": 0.454, - "step": 2378 - }, - { - "epoch": 0.74, - "learning_rate": 1.606087586803259e-07, - "loss": 0.4633, - "step": 2379 - }, - { - "epoch": 0.74, - "learning_rate": 1.602436646071283e-07, - "loss": 0.4744, - "step": 2380 - }, - { - "epoch": 0.74, - "learning_rate": 1.5987890676901044e-07, - "loss": 0.4651, - "step": 2381 - }, - { - "epoch": 0.74, - "learning_rate": 1.5951448552694986e-07, - "loss": 0.4642, - "step": 2382 - }, - { - "epoch": 0.74, - "learning_rate": 1.5915040124159107e-07, - "loss": 0.4341, - "step": 2383 - }, - { - "epoch": 0.74, - "learning_rate": 1.5878665427324456e-07, - "loss": 0.4622, - "step": 2384 - }, - { - "epoch": 0.74, - "learning_rate": 1.5842324498188792e-07, - "loss": 0.4657, - "step": 2385 - }, - { - "epoch": 0.74, - "learning_rate": 1.5806017372716362e-07, - "loss": 0.4686, - "step": 2386 - }, - { - "epoch": 0.74, - "learning_rate": 1.576974408683804e-07, - "loss": 0.4361, - "step": 2387 - }, - { - "epoch": 0.74, - "learning_rate": 1.5733504676451125e-07, - "loss": 0.453, - "step": 2388 - }, - { - "epoch": 0.74, - "learning_rate": 1.5697299177419476e-07, - "loss": 0.4449, - "step": 2389 - }, - { - "epoch": 0.75, - "learning_rate": 1.5661127625573361e-07, - "loss": 0.4689, - "step": 2390 - }, - { - "epoch": 0.75, - "learning_rate": 1.5624990056709414e-07, - "loss": 0.4411, - "step": 2391 - }, - { - "epoch": 0.75, - "learning_rate": 1.5588886506590693e-07, - "loss": 0.4615, - "step": 2392 - }, - { - "epoch": 0.75, - "learning_rate": 1.5552817010946566e-07, - "loss": 0.442, - "step": 2393 - }, - { - "epoch": 0.75, - "learning_rate": 1.5516781605472717e-07, - "loss": 0.4634, - "step": 2394 - }, - { - "epoch": 0.75, - "learning_rate": 1.548078032583104e-07, - "loss": 0.4683, - "step": 2395 - }, - { - "epoch": 0.75, - "learning_rate": 1.5444813207649743e-07, - "loss": 0.4438, - "step": 2396 - }, - { - "epoch": 0.75, - "learning_rate": 1.540888028652314e-07, - "loss": 0.4573, - "step": 2397 - }, - { - "epoch": 0.75, - "learning_rate": 1.5372981598011776e-07, - "loss": 0.436, - "step": 2398 - }, - { - "epoch": 0.75, - "learning_rate": 1.5337117177642244e-07, - "loss": 0.4692, - "step": 2399 - }, - { - "epoch": 0.75, - "learning_rate": 1.5301287060907286e-07, - "loss": 0.4574, - "step": 2400 - }, - { - "epoch": 0.75, - "learning_rate": 1.5265491283265685e-07, - "loss": 0.4461, - "step": 2401 - }, - { - "epoch": 0.75, - "learning_rate": 1.5229729880142194e-07, - "loss": 0.4897, - "step": 2402 - }, - { - "epoch": 0.75, - "learning_rate": 1.519400288692761e-07, - "loss": 0.4526, - "step": 2403 - }, - { - "epoch": 0.75, - "learning_rate": 1.5158310338978642e-07, - "loss": 0.4746, - "step": 2404 - }, - { - "epoch": 0.75, - "learning_rate": 1.512265227161793e-07, - "loss": 0.4568, - "step": 2405 - }, - { - "epoch": 0.75, - "learning_rate": 1.5087028720133937e-07, - "loss": 0.4588, - "step": 2406 - }, - { - "epoch": 0.75, - "learning_rate": 1.5051439719781068e-07, - "loss": 0.4525, - "step": 2407 - }, - { - "epoch": 0.75, - "learning_rate": 1.5015885305779442e-07, - "loss": 0.4691, - "step": 2408 - }, - { - "epoch": 0.75, - "learning_rate": 1.4980365513314995e-07, - "loss": 0.4487, - "step": 2409 - }, - { - "epoch": 0.75, - "learning_rate": 1.4944880377539366e-07, - "loss": 0.453, - "step": 2410 - }, - { - "epoch": 0.75, - "learning_rate": 1.4909429933569972e-07, - "loss": 0.4563, - "step": 2411 - }, - { - "epoch": 0.75, - "learning_rate": 1.48740142164898e-07, - "loss": 0.4499, - "step": 2412 - }, - { - "epoch": 0.75, - "learning_rate": 1.4838633261347545e-07, - "loss": 0.464, - "step": 2413 - }, - { - "epoch": 0.75, - "learning_rate": 1.4803287103157442e-07, - "loss": 0.4505, - "step": 2414 - }, - { - "epoch": 0.75, - "learning_rate": 1.4767975776899345e-07, - "loss": 0.4677, - "step": 2415 - }, - { - "epoch": 0.75, - "learning_rate": 1.473269931751861e-07, - "loss": 0.4778, - "step": 2416 - }, - { - "epoch": 0.75, - "learning_rate": 1.4697457759926052e-07, - "loss": 0.4463, - "step": 2417 - }, - { - "epoch": 0.75, - "learning_rate": 1.466225113899805e-07, - "loss": 0.4629, - "step": 2418 - }, - { - "epoch": 0.75, - "learning_rate": 1.4627079489576283e-07, - "loss": 0.4611, - "step": 2419 - }, - { - "epoch": 0.75, - "learning_rate": 1.459194284646792e-07, - "loss": 0.4725, - "step": 2420 - }, - { - "epoch": 0.75, - "learning_rate": 1.455684124444539e-07, - "loss": 0.4528, - "step": 2421 - }, - { - "epoch": 0.75, - "learning_rate": 1.452177471824657e-07, - "loss": 0.4518, - "step": 2422 - }, - { - "epoch": 0.76, - "learning_rate": 1.4486743302574512e-07, - "loss": 0.4602, - "step": 2423 - }, - { - "epoch": 0.76, - "learning_rate": 1.445174703209756e-07, - "loss": 0.4386, - "step": 2424 - }, - { - "epoch": 0.76, - "learning_rate": 1.4416785941449323e-07, - "loss": 0.4604, - "step": 2425 - }, - { - "epoch": 0.76, - "learning_rate": 1.4381860065228496e-07, - "loss": 0.4507, - "step": 2426 - }, - { - "epoch": 0.76, - "learning_rate": 1.4346969437999028e-07, - "loss": 0.4603, - "step": 2427 - }, - { - "epoch": 0.76, - "learning_rate": 1.431211409428989e-07, - "loss": 0.4597, - "step": 2428 - }, - { - "epoch": 0.76, - "learning_rate": 1.4277294068595243e-07, - "loss": 0.458, - "step": 2429 - }, - { - "epoch": 0.76, - "learning_rate": 1.4242509395374187e-07, - "loss": 0.4673, - "step": 2430 - }, - { - "epoch": 0.76, - "learning_rate": 1.4207760109050914e-07, - "loss": 0.4615, - "step": 2431 - }, - { - "epoch": 0.76, - "learning_rate": 1.4173046244014526e-07, - "loss": 0.4441, - "step": 2432 - }, - { - "epoch": 0.76, - "learning_rate": 1.413836783461917e-07, - "loss": 0.4542, - "step": 2433 - }, - { - "epoch": 0.76, - "learning_rate": 1.41037249151838e-07, - "loss": 0.4783, - "step": 2434 - }, - { - "epoch": 0.76, - "learning_rate": 1.4069117519992312e-07, - "loss": 0.4554, - "step": 2435 - }, - { - "epoch": 0.76, - "learning_rate": 1.4034545683293426e-07, - "loss": 0.4493, - "step": 2436 - }, - { - "epoch": 0.76, - "learning_rate": 1.4000009439300686e-07, - "loss": 0.4488, - "step": 2437 - }, - { - "epoch": 0.76, - "learning_rate": 1.396550882219238e-07, - "loss": 0.466, - "step": 2438 - }, - { - "epoch": 0.76, - "learning_rate": 1.393104386611156e-07, - "loss": 0.4607, - "step": 2439 - }, - { - "epoch": 0.76, - "learning_rate": 1.3896614605166006e-07, - "loss": 0.4495, - "step": 2440 - }, - { - "epoch": 0.76, - "learning_rate": 1.3862221073428132e-07, - "loss": 0.4622, - "step": 2441 - }, - { - "epoch": 0.76, - "learning_rate": 1.382786330493503e-07, - "loss": 0.4778, - "step": 2442 - }, - { - "epoch": 0.76, - "learning_rate": 1.3793541333688336e-07, - "loss": 0.4516, - "step": 2443 - }, - { - "epoch": 0.76, - "learning_rate": 1.3759255193654373e-07, - "loss": 0.4487, - "step": 2444 - }, - { - "epoch": 0.76, - "learning_rate": 1.3725004918763893e-07, - "loss": 0.4675, - "step": 2445 - }, - { - "epoch": 0.76, - "learning_rate": 1.3690790542912216e-07, - "loss": 0.4471, - "step": 2446 - }, - { - "epoch": 0.76, - "learning_rate": 1.365661209995911e-07, - "loss": 0.467, - "step": 2447 - }, - { - "epoch": 0.76, - "learning_rate": 1.362246962372881e-07, - "loss": 0.4656, - "step": 2448 - }, - { - "epoch": 0.76, - "learning_rate": 1.3588363148009908e-07, - "loss": 0.4444, - "step": 2449 - }, - { - "epoch": 0.76, - "learning_rate": 1.355429270655541e-07, - "loss": 0.4728, - "step": 2450 - }, - { - "epoch": 0.76, - "learning_rate": 1.3520258333082668e-07, - "loss": 0.4642, - "step": 2451 - }, - { - "epoch": 0.76, - "learning_rate": 1.348626006127329e-07, - "loss": 0.4486, - "step": 2452 - }, - { - "epoch": 0.76, - "learning_rate": 1.3452297924773215e-07, - "loss": 0.4387, - "step": 2453 - }, - { - "epoch": 0.76, - "learning_rate": 1.3418371957192553e-07, - "loss": 0.4548, - "step": 2454 - }, - { - "epoch": 0.77, - "learning_rate": 1.3384482192105717e-07, - "loss": 0.4583, - "step": 2455 - }, - { - "epoch": 0.77, - "learning_rate": 1.3350628663051193e-07, - "loss": 0.455, - "step": 2456 - }, - { - "epoch": 0.77, - "learning_rate": 1.331681140353167e-07, - "loss": 0.4772, - "step": 2457 - }, - { - "epoch": 0.77, - "learning_rate": 1.3283030447013915e-07, - "loss": 0.4513, - "step": 2458 - }, - { - "epoch": 0.77, - "learning_rate": 1.324928582692879e-07, - "loss": 0.4561, - "step": 2459 - }, - { - "epoch": 0.77, - "learning_rate": 1.3215577576671166e-07, - "loss": 0.487, - "step": 2460 - }, - { - "epoch": 0.77, - "learning_rate": 1.3181905729599947e-07, - "loss": 0.4477, - "step": 2461 - }, - { - "epoch": 0.77, - "learning_rate": 1.3148270319038002e-07, - "loss": 0.4569, - "step": 2462 - }, - { - "epoch": 0.77, - "learning_rate": 1.3114671378272151e-07, - "loss": 0.447, - "step": 2463 - }, - { - "epoch": 0.77, - "learning_rate": 1.3081108940553097e-07, - "loss": 0.455, - "step": 2464 - }, - { - "epoch": 0.77, - "learning_rate": 1.3047583039095444e-07, - "loss": 0.4503, - "step": 2465 - }, - { - "epoch": 0.77, - "learning_rate": 1.301409370707764e-07, - "loss": 0.4618, - "step": 2466 - }, - { - "epoch": 0.77, - "learning_rate": 1.2980640977641906e-07, - "loss": 0.4502, - "step": 2467 - }, - { - "epoch": 0.77, - "learning_rate": 1.2947224883894276e-07, - "loss": 0.4658, - "step": 2468 - }, - { - "epoch": 0.77, - "learning_rate": 1.2913845458904514e-07, - "loss": 0.4721, - "step": 2469 - }, - { - "epoch": 0.77, - "learning_rate": 1.2880502735706127e-07, - "loss": 0.4463, - "step": 2470 - }, - { - "epoch": 0.77, - "learning_rate": 1.2847196747296234e-07, - "loss": 0.466, - "step": 2471 - }, - { - "epoch": 0.77, - "learning_rate": 1.2813927526635655e-07, - "loss": 0.4674, - "step": 2472 - }, - { - "epoch": 0.77, - "learning_rate": 1.2780695106648804e-07, - "loss": 0.4546, - "step": 2473 - }, - { - "epoch": 0.77, - "learning_rate": 1.2747499520223693e-07, - "loss": 0.4561, - "step": 2474 - }, - { - "epoch": 0.77, - "learning_rate": 1.2714340800211836e-07, - "loss": 0.4584, - "step": 2475 - }, - { - "epoch": 0.77, - "learning_rate": 1.2681218979428316e-07, - "loss": 0.4382, - "step": 2476 - }, - { - "epoch": 0.77, - "learning_rate": 1.2648134090651686e-07, - "loss": 0.4521, - "step": 2477 - }, - { - "epoch": 0.77, - "learning_rate": 1.261508616662391e-07, - "loss": 0.4871, - "step": 2478 - }, - { - "epoch": 0.77, - "learning_rate": 1.258207524005042e-07, - "loss": 0.4635, - "step": 2479 - }, - { - "epoch": 0.77, - "learning_rate": 1.2549101343600017e-07, - "loss": 0.4341, - "step": 2480 - }, - { - "epoch": 0.77, - "learning_rate": 1.2516164509904865e-07, - "loss": 0.4611, - "step": 2481 - }, - { - "epoch": 0.77, - "learning_rate": 1.248326477156041e-07, - "loss": 0.4564, - "step": 2482 - }, - { - "epoch": 0.77, - "learning_rate": 1.2450402161125422e-07, - "loss": 0.4602, - "step": 2483 - }, - { - "epoch": 0.77, - "learning_rate": 1.241757671112194e-07, - "loss": 0.4671, - "step": 2484 - }, - { - "epoch": 0.77, - "learning_rate": 1.2384788454035206e-07, - "loss": 0.4606, - "step": 2485 - }, - { - "epoch": 0.77, - "learning_rate": 1.2352037422313637e-07, - "loss": 0.4609, - "step": 2486 - }, - { - "epoch": 0.78, - "learning_rate": 1.2319323648368847e-07, - "loss": 0.4559, - "step": 2487 - }, - { - "epoch": 0.78, - "learning_rate": 1.2286647164575558e-07, - "loss": 0.4613, - "step": 2488 - }, - { - "epoch": 0.78, - "learning_rate": 1.225400800327161e-07, - "loss": 0.4577, - "step": 2489 - }, - { - "epoch": 0.78, - "learning_rate": 1.2221406196757854e-07, - "loss": 0.442, - "step": 2490 - }, - { - "epoch": 0.78, - "learning_rate": 1.218884177729823e-07, - "loss": 0.4442, - "step": 2491 - }, - { - "epoch": 0.78, - "learning_rate": 1.2156314777119676e-07, - "loss": 0.4349, - "step": 2492 - }, - { - "epoch": 0.78, - "learning_rate": 1.2123825228412048e-07, - "loss": 0.4594, - "step": 2493 - }, - { - "epoch": 0.78, - "learning_rate": 1.2091373163328177e-07, - "loss": 0.4427, - "step": 2494 - }, - { - "epoch": 0.78, - "learning_rate": 1.205895861398381e-07, - "loss": 0.4504, - "step": 2495 - }, - { - "epoch": 0.78, - "learning_rate": 1.2026581612457549e-07, - "loss": 0.474, - "step": 2496 - }, - { - "epoch": 0.78, - "learning_rate": 1.1994242190790798e-07, - "loss": 0.4688, - "step": 2497 - }, - { - "epoch": 0.78, - "learning_rate": 1.1961940380987879e-07, - "loss": 0.4488, - "step": 2498 - }, - { - "epoch": 0.78, - "learning_rate": 1.1929676215015766e-07, - "loss": 0.4712, - "step": 2499 - }, - { - "epoch": 0.78, - "learning_rate": 1.1897449724804277e-07, - "loss": 0.47, - "step": 2500 - }, - { - "epoch": 0.78, - "learning_rate": 1.186526094224587e-07, - "loss": 0.4478, - "step": 2501 - }, - { - "epoch": 0.78, - "learning_rate": 1.1833109899195743e-07, - "loss": 0.4496, - "step": 2502 - }, - { - "epoch": 0.78, - "learning_rate": 1.1800996627471726e-07, - "loss": 0.4387, - "step": 2503 - }, - { - "epoch": 0.78, - "learning_rate": 1.1768921158854234e-07, - "loss": 0.4514, - "step": 2504 - }, - { - "epoch": 0.78, - "learning_rate": 1.1736883525086355e-07, - "loss": 0.4599, - "step": 2505 - }, - { - "epoch": 0.78, - "learning_rate": 1.1704883757873641e-07, - "loss": 0.4696, - "step": 2506 - }, - { - "epoch": 0.78, - "learning_rate": 1.1672921888884241e-07, - "loss": 0.4571, - "step": 2507 - }, - { - "epoch": 0.78, - "learning_rate": 1.1640997949748726e-07, - "loss": 0.4601, - "step": 2508 - }, - { - "epoch": 0.78, - "learning_rate": 1.1609111972060237e-07, - "loss": 0.4647, - "step": 2509 - }, - { - "epoch": 0.78, - "learning_rate": 1.157726398737423e-07, - "loss": 0.4473, - "step": 2510 - }, - { - "epoch": 0.78, - "learning_rate": 1.1545454027208645e-07, - "loss": 0.4661, - "step": 2511 - }, - { - "epoch": 0.78, - "learning_rate": 1.1513682123043744e-07, - "loss": 0.4336, - "step": 2512 - }, - { - "epoch": 0.78, - "learning_rate": 1.1481948306322148e-07, - "loss": 0.4655, - "step": 2513 - }, - { - "epoch": 0.78, - "learning_rate": 1.145025260844879e-07, - "loss": 0.4633, - "step": 2514 - }, - { - "epoch": 0.78, - "learning_rate": 1.1418595060790864e-07, - "loss": 0.4742, - "step": 2515 - }, - { - "epoch": 0.78, - "learning_rate": 1.1386975694677836e-07, - "loss": 0.4617, - "step": 2516 - }, - { - "epoch": 0.78, - "learning_rate": 1.1355394541401342e-07, - "loss": 0.4615, - "step": 2517 - }, - { - "epoch": 0.78, - "learning_rate": 1.1323851632215248e-07, - "loss": 0.4601, - "step": 2518 - }, - { - "epoch": 0.79, - "learning_rate": 1.1292346998335511e-07, - "loss": 0.4463, - "step": 2519 - }, - { - "epoch": 0.79, - "learning_rate": 1.1260880670940309e-07, - "loss": 0.4433, - "step": 2520 - }, - { - "epoch": 0.79, - "learning_rate": 1.1229452681169799e-07, - "loss": 0.4446, - "step": 2521 - }, - { - "epoch": 0.79, - "learning_rate": 1.1198063060126284e-07, - "loss": 0.4356, - "step": 2522 - }, - { - "epoch": 0.79, - "learning_rate": 1.1166711838874021e-07, - "loss": 0.4589, - "step": 2523 - }, - { - "epoch": 0.79, - "learning_rate": 1.1135399048439353e-07, - "loss": 0.467, - "step": 2524 - }, - { - "epoch": 0.79, - "learning_rate": 1.1104124719810509e-07, - "loss": 0.4713, - "step": 2525 - }, - { - "epoch": 0.79, - "learning_rate": 1.1072888883937698e-07, - "loss": 0.4437, - "step": 2526 - }, - { - "epoch": 0.79, - "learning_rate": 1.1041691571733042e-07, - "loss": 0.4707, - "step": 2527 - }, - { - "epoch": 0.79, - "learning_rate": 1.101053281407049e-07, - "loss": 0.4591, - "step": 2528 - }, - { - "epoch": 0.79, - "learning_rate": 1.0979412641785901e-07, - "loss": 0.4293, - "step": 2529 - }, - { - "epoch": 0.79, - "learning_rate": 1.0948331085676876e-07, - "loss": 0.4759, - "step": 2530 - }, - { - "epoch": 0.79, - "learning_rate": 1.0917288176502892e-07, - "loss": 0.4469, - "step": 2531 - }, - { - "epoch": 0.79, - "learning_rate": 1.0886283944985086e-07, - "loss": 0.4765, - "step": 2532 - }, - { - "epoch": 0.79, - "learning_rate": 1.0855318421806387e-07, - "loss": 0.4572, - "step": 2533 - }, - { - "epoch": 0.79, - "learning_rate": 1.0824391637611347e-07, - "loss": 0.4367, - "step": 2534 - }, - { - "epoch": 0.79, - "learning_rate": 1.0793503623006278e-07, - "loss": 0.4516, - "step": 2535 - }, - { - "epoch": 0.79, - "learning_rate": 1.0762654408559035e-07, - "loss": 0.4581, - "step": 2536 - }, - { - "epoch": 0.79, - "learning_rate": 1.073184402479912e-07, - "loss": 0.4695, - "step": 2537 - }, - { - "epoch": 0.79, - "learning_rate": 1.0701072502217601e-07, - "loss": 0.4564, - "step": 2538 - }, - { - "epoch": 0.79, - "learning_rate": 1.067033987126707e-07, - "loss": 0.4535, - "step": 2539 - }, - { - "epoch": 0.79, - "learning_rate": 1.0639646162361643e-07, - "loss": 0.4653, - "step": 2540 - }, - { - "epoch": 0.79, - "learning_rate": 1.0608991405876927e-07, - "loss": 0.4569, - "step": 2541 - }, - { - "epoch": 0.79, - "learning_rate": 1.0578375632149978e-07, - "loss": 0.4724, - "step": 2542 - }, - { - "epoch": 0.79, - "learning_rate": 1.0547798871479241e-07, - "loss": 0.47, - "step": 2543 - }, - { - "epoch": 0.79, - "learning_rate": 1.0517261154124608e-07, - "loss": 0.4499, - "step": 2544 - }, - { - "epoch": 0.79, - "learning_rate": 1.0486762510307257e-07, - "loss": 0.4678, - "step": 2545 - }, - { - "epoch": 0.79, - "learning_rate": 1.0456302970209795e-07, - "loss": 0.466, - "step": 2546 - }, - { - "epoch": 0.79, - "learning_rate": 1.042588256397604e-07, - "loss": 0.4535, - "step": 2547 - }, - { - "epoch": 0.79, - "learning_rate": 1.0395501321711131e-07, - "loss": 0.4624, - "step": 2548 - }, - { - "epoch": 0.79, - "learning_rate": 1.0365159273481439e-07, - "loss": 0.4648, - "step": 2549 - }, - { - "epoch": 0.79, - "learning_rate": 1.0334856449314555e-07, - "loss": 0.4593, - "step": 2550 - }, - { - "epoch": 0.8, - "learning_rate": 1.030459287919922e-07, - "loss": 0.4693, - "step": 2551 - }, - { - "epoch": 0.8, - "learning_rate": 1.0274368593085364e-07, - "loss": 0.4605, - "step": 2552 - }, - { - "epoch": 0.8, - "learning_rate": 1.0244183620884039e-07, - "loss": 0.47, - "step": 2553 - }, - { - "epoch": 0.8, - "learning_rate": 1.0214037992467351e-07, - "loss": 0.4729, - "step": 2554 - }, - { - "epoch": 0.8, - "learning_rate": 1.0183931737668511e-07, - "loss": 0.4669, - "step": 2555 - }, - { - "epoch": 0.8, - "learning_rate": 1.0153864886281749e-07, - "loss": 0.4677, - "step": 2556 - }, - { - "epoch": 0.8, - "learning_rate": 1.0123837468062319e-07, - "loss": 0.4575, - "step": 2557 - }, - { - "epoch": 0.8, - "learning_rate": 1.0093849512726398e-07, - "loss": 0.4584, - "step": 2558 - }, - { - "epoch": 0.8, - "learning_rate": 1.0063901049951173e-07, - "loss": 0.4565, - "step": 2559 - }, - { - "epoch": 0.8, - "learning_rate": 1.0033992109374705e-07, - "loss": 0.4651, - "step": 2560 - }, - { - "epoch": 0.8, - "learning_rate": 1.0004122720595976e-07, - "loss": 0.4639, - "step": 2561 - }, - { - "epoch": 0.8, - "learning_rate": 9.974292913174776e-08, - "loss": 0.4556, - "step": 2562 - }, - { - "epoch": 0.8, - "learning_rate": 9.944502716631776e-08, - "loss": 0.4472, - "step": 2563 - }, - { - "epoch": 0.8, - "learning_rate": 9.914752160448425e-08, - "loss": 0.4641, - "step": 2564 - }, - { - "epoch": 0.8, - "learning_rate": 9.88504127406693e-08, - "loss": 0.4668, - "step": 2565 - }, - { - "epoch": 0.8, - "learning_rate": 9.855370086890258e-08, - "loss": 0.4629, - "step": 2566 - }, - { - "epoch": 0.8, - "learning_rate": 9.825738628282082e-08, - "loss": 0.4712, - "step": 2567 - }, - { - "epoch": 0.8, - "learning_rate": 9.796146927566778e-08, - "loss": 0.4782, - "step": 2568 - }, - { - "epoch": 0.8, - "learning_rate": 9.766595014029317e-08, - "loss": 0.4609, - "step": 2569 - }, - { - "epoch": 0.8, - "learning_rate": 9.737082916915362e-08, - "loss": 0.4526, - "step": 2570 - }, - { - "epoch": 0.8, - "learning_rate": 9.707610665431137e-08, - "loss": 0.4537, - "step": 2571 - }, - { - "epoch": 0.8, - "learning_rate": 9.678178288743456e-08, - "loss": 0.4458, - "step": 2572 - }, - { - "epoch": 0.8, - "learning_rate": 9.648785815979632e-08, - "loss": 0.433, - "step": 2573 - }, - { - "epoch": 0.8, - "learning_rate": 9.619433276227517e-08, - "loss": 0.454, - "step": 2574 - }, - { - "epoch": 0.8, - "learning_rate": 9.590120698535454e-08, - "loss": 0.4807, - "step": 2575 - }, - { - "epoch": 0.8, - "learning_rate": 9.560848111912228e-08, - "loss": 0.4573, - "step": 2576 - }, - { - "epoch": 0.8, - "learning_rate": 9.531615545327016e-08, - "loss": 0.4456, - "step": 2577 - }, - { - "epoch": 0.8, - "learning_rate": 9.502423027709427e-08, - "loss": 0.4613, - "step": 2578 - }, - { - "epoch": 0.8, - "learning_rate": 9.47327058794945e-08, - "loss": 0.4882, - "step": 2579 - }, - { - "epoch": 0.8, - "learning_rate": 9.44415825489735e-08, - "loss": 0.4617, - "step": 2580 - }, - { - "epoch": 0.8, - "learning_rate": 9.415086057363753e-08, - "loss": 0.4645, - "step": 2581 - }, - { - "epoch": 0.8, - "learning_rate": 9.386054024119555e-08, - "loss": 0.4625, - "step": 2582 - }, - { - "epoch": 0.81, - "learning_rate": 9.357062183895903e-08, - "loss": 0.4779, - "step": 2583 - }, - { - "epoch": 0.81, - "learning_rate": 9.328110565384151e-08, - "loss": 0.4548, - "step": 2584 - }, - { - "epoch": 0.81, - "learning_rate": 9.299199197235868e-08, - "loss": 0.4735, - "step": 2585 - }, - { - "epoch": 0.81, - "learning_rate": 9.27032810806278e-08, - "loss": 0.4456, - "step": 2586 - }, - { - "epoch": 0.81, - "learning_rate": 9.24149732643677e-08, - "loss": 0.4711, - "step": 2587 - }, - { - "epoch": 0.81, - "learning_rate": 9.212706880889792e-08, - "loss": 0.4493, - "step": 2588 - }, - { - "epoch": 0.81, - "learning_rate": 9.183956799913911e-08, - "loss": 0.468, - "step": 2589 - }, - { - "epoch": 0.81, - "learning_rate": 9.155247111961262e-08, - "loss": 0.4666, - "step": 2590 - }, - { - "epoch": 0.81, - "learning_rate": 9.126577845443956e-08, - "loss": 0.4509, - "step": 2591 - }, - { - "epoch": 0.81, - "learning_rate": 9.097949028734142e-08, - "loss": 0.4432, - "step": 2592 - }, - { - "epoch": 0.81, - "learning_rate": 9.069360690163918e-08, - "loss": 0.4595, - "step": 2593 - }, - { - "epoch": 0.81, - "learning_rate": 9.040812858025359e-08, - "loss": 0.4537, - "step": 2594 - }, - { - "epoch": 0.81, - "learning_rate": 9.012305560570376e-08, - "loss": 0.4748, - "step": 2595 - }, - { - "epoch": 0.81, - "learning_rate": 8.98383882601087e-08, - "loss": 0.456, - "step": 2596 - }, - { - "epoch": 0.81, - "learning_rate": 8.955412682518505e-08, - "loss": 0.4321, - "step": 2597 - }, - { - "epoch": 0.81, - "learning_rate": 8.927027158224842e-08, - "loss": 0.4648, - "step": 2598 - }, - { - "epoch": 0.81, - "learning_rate": 8.898682281221176e-08, - "loss": 0.4686, - "step": 2599 - }, - { - "epoch": 0.81, - "learning_rate": 8.870378079558666e-08, - "loss": 0.4504, - "step": 2600 - }, - { - "epoch": 0.81, - "learning_rate": 8.842114581248127e-08, - "loss": 0.4458, - "step": 2601 - }, - { - "epoch": 0.81, - "learning_rate": 8.813891814260166e-08, - "loss": 0.4854, - "step": 2602 - }, - { - "epoch": 0.81, - "learning_rate": 8.785709806525015e-08, - "loss": 0.4625, - "step": 2603 - }, - { - "epoch": 0.81, - "learning_rate": 8.757568585932617e-08, - "loss": 0.4608, - "step": 2604 - }, - { - "epoch": 0.81, - "learning_rate": 8.72946818033255e-08, - "loss": 0.4503, - "step": 2605 - }, - { - "epoch": 0.81, - "learning_rate": 8.701408617533945e-08, - "loss": 0.4497, - "step": 2606 - }, - { - "epoch": 0.81, - "learning_rate": 8.673389925305602e-08, - "loss": 0.4718, - "step": 2607 - }, - { - "epoch": 0.81, - "learning_rate": 8.64541213137579e-08, - "loss": 0.4699, - "step": 2608 - }, - { - "epoch": 0.81, - "learning_rate": 8.617475263432367e-08, - "loss": 0.473, - "step": 2609 - }, - { - "epoch": 0.81, - "learning_rate": 8.589579349122616e-08, - "loss": 0.4595, - "step": 2610 - }, - { - "epoch": 0.81, - "learning_rate": 8.561724416053395e-08, - "loss": 0.4555, - "step": 2611 - }, - { - "epoch": 0.81, - "learning_rate": 8.533910491790908e-08, - "loss": 0.4723, - "step": 2612 - }, - { - "epoch": 0.81, - "learning_rate": 8.506137603860835e-08, - "loss": 0.4758, - "step": 2613 - }, - { - "epoch": 0.81, - "learning_rate": 8.478405779748204e-08, - "loss": 0.4694, - "step": 2614 - }, - { - "epoch": 0.82, - "learning_rate": 8.450715046897433e-08, - "loss": 0.4577, - "step": 2615 - }, - { - "epoch": 0.82, - "learning_rate": 8.423065432712284e-08, - "loss": 0.4517, - "step": 2616 - }, - { - "epoch": 0.82, - "learning_rate": 8.395456964555764e-08, - "loss": 0.4457, - "step": 2617 - }, - { - "epoch": 0.82, - "learning_rate": 8.367889669750255e-08, - "loss": 0.4318, - "step": 2618 - }, - { - "epoch": 0.82, - "learning_rate": 8.340363575577314e-08, - "loss": 0.4674, - "step": 2619 - }, - { - "epoch": 0.82, - "learning_rate": 8.312878709277777e-08, - "loss": 0.4646, - "step": 2620 - }, - { - "epoch": 0.82, - "learning_rate": 8.285435098051607e-08, - "loss": 0.4587, - "step": 2621 - }, - { - "epoch": 0.82, - "learning_rate": 8.258032769058049e-08, - "loss": 0.443, - "step": 2622 - }, - { - "epoch": 0.82, - "learning_rate": 8.230671749415392e-08, - "loss": 0.4395, - "step": 2623 - }, - { - "epoch": 0.82, - "learning_rate": 8.203352066201108e-08, - "loss": 0.4567, - "step": 2624 - }, - { - "epoch": 0.82, - "learning_rate": 8.176073746451701e-08, - "loss": 0.4664, - "step": 2625 - }, - { - "epoch": 0.82, - "learning_rate": 8.148836817162824e-08, - "loss": 0.4465, - "step": 2626 - }, - { - "epoch": 0.82, - "learning_rate": 8.12164130528909e-08, - "loss": 0.4549, - "step": 2627 - }, - { - "epoch": 0.82, - "learning_rate": 8.094487237744157e-08, - "loss": 0.4584, - "step": 2628 - }, - { - "epoch": 0.82, - "learning_rate": 8.067374641400688e-08, - "loss": 0.4497, - "step": 2629 - }, - { - "epoch": 0.82, - "learning_rate": 8.040303543090254e-08, - "loss": 0.4491, - "step": 2630 - }, - { - "epoch": 0.82, - "learning_rate": 8.013273969603413e-08, - "loss": 0.4776, - "step": 2631 - }, - { - "epoch": 0.82, - "learning_rate": 7.986285947689564e-08, - "loss": 0.4528, - "step": 2632 - }, - { - "epoch": 0.82, - "learning_rate": 7.959339504057071e-08, - "loss": 0.4415, - "step": 2633 - }, - { - "epoch": 0.82, - "learning_rate": 7.932434665373073e-08, - "loss": 0.455, - "step": 2634 - }, - { - "epoch": 0.82, - "learning_rate": 7.905571458263571e-08, - "loss": 0.454, - "step": 2635 - }, - { - "epoch": 0.82, - "learning_rate": 7.878749909313359e-08, - "loss": 0.4474, - "step": 2636 - }, - { - "epoch": 0.82, - "learning_rate": 7.85197004506602e-08, - "loss": 0.4602, - "step": 2637 - }, - { - "epoch": 0.82, - "learning_rate": 7.825231892023853e-08, - "loss": 0.4746, - "step": 2638 - }, - { - "epoch": 0.82, - "learning_rate": 7.798535476647894e-08, - "loss": 0.4687, - "step": 2639 - }, - { - "epoch": 0.82, - "learning_rate": 7.771880825357895e-08, - "loss": 0.4694, - "step": 2640 - }, - { - "epoch": 0.82, - "learning_rate": 7.745267964532237e-08, - "loss": 0.4498, - "step": 2641 - }, - { - "epoch": 0.82, - "learning_rate": 7.718696920507977e-08, - "loss": 0.4506, - "step": 2642 - }, - { - "epoch": 0.82, - "learning_rate": 7.692167719580749e-08, - "loss": 0.4504, - "step": 2643 - }, - { - "epoch": 0.82, - "learning_rate": 7.665680388004847e-08, - "loss": 0.4708, - "step": 2644 - }, - { - "epoch": 0.82, - "learning_rate": 7.639234951993046e-08, - "loss": 0.4629, - "step": 2645 - }, - { - "epoch": 0.82, - "learning_rate": 7.612831437716715e-08, - "loss": 0.4651, - "step": 2646 - }, - { - "epoch": 0.83, - "learning_rate": 7.586469871305728e-08, - "loss": 0.4361, - "step": 2647 - }, - { - "epoch": 0.83, - "learning_rate": 7.560150278848443e-08, - "loss": 0.4482, - "step": 2648 - }, - { - "epoch": 0.83, - "learning_rate": 7.533872686391651e-08, - "loss": 0.4794, - "step": 2649 - }, - { - "epoch": 0.83, - "learning_rate": 7.507637119940613e-08, - "loss": 0.4431, - "step": 2650 - }, - { - "epoch": 0.83, - "learning_rate": 7.481443605458993e-08, - "loss": 0.4503, - "step": 2651 - }, - { - "epoch": 0.83, - "learning_rate": 7.455292168868843e-08, - "loss": 0.4806, - "step": 2652 - }, - { - "epoch": 0.83, - "learning_rate": 7.429182836050535e-08, - "loss": 0.456, - "step": 2653 - }, - { - "epoch": 0.83, - "learning_rate": 7.403115632842827e-08, - "loss": 0.4584, - "step": 2654 - }, - { - "epoch": 0.83, - "learning_rate": 7.377090585042766e-08, - "loss": 0.4473, - "step": 2655 - }, - { - "epoch": 0.83, - "learning_rate": 7.351107718405653e-08, - "loss": 0.4594, - "step": 2656 - }, - { - "epoch": 0.83, - "learning_rate": 7.32516705864507e-08, - "loss": 0.4821, - "step": 2657 - }, - { - "epoch": 0.83, - "learning_rate": 7.299268631432836e-08, - "loss": 0.4694, - "step": 2658 - }, - { - "epoch": 0.83, - "learning_rate": 7.273412462398975e-08, - "loss": 0.4515, - "step": 2659 - }, - { - "epoch": 0.83, - "learning_rate": 7.247598577131658e-08, - "loss": 0.4599, - "step": 2660 - }, - { - "epoch": 0.83, - "learning_rate": 7.221827001177239e-08, - "loss": 0.471, - "step": 2661 - }, - { - "epoch": 0.83, - "learning_rate": 7.196097760040203e-08, - "loss": 0.4472, - "step": 2662 - }, - { - "epoch": 0.83, - "learning_rate": 7.170410879183143e-08, - "loss": 0.4703, - "step": 2663 - }, - { - "epoch": 0.83, - "learning_rate": 7.14476638402669e-08, - "loss": 0.4503, - "step": 2664 - }, - { - "epoch": 0.83, - "learning_rate": 7.11916429994957e-08, - "loss": 0.449, - "step": 2665 - }, - { - "epoch": 0.83, - "learning_rate": 7.093604652288548e-08, - "loss": 0.4611, - "step": 2666 - }, - { - "epoch": 0.83, - "learning_rate": 7.068087466338336e-08, - "loss": 0.4446, - "step": 2667 - }, - { - "epoch": 0.83, - "learning_rate": 7.042612767351675e-08, - "loss": 0.4577, - "step": 2668 - }, - { - "epoch": 0.83, - "learning_rate": 7.017180580539239e-08, - "loss": 0.462, - "step": 2669 - }, - { - "epoch": 0.83, - "learning_rate": 6.991790931069647e-08, - "loss": 0.4674, - "step": 2670 - }, - { - "epoch": 0.83, - "learning_rate": 6.966443844069386e-08, - "loss": 0.4429, - "step": 2671 - }, - { - "epoch": 0.83, - "learning_rate": 6.941139344622855e-08, - "loss": 0.4617, - "step": 2672 - }, - { - "epoch": 0.83, - "learning_rate": 6.915877457772301e-08, - "loss": 0.4585, - "step": 2673 - }, - { - "epoch": 0.83, - "learning_rate": 6.890658208517797e-08, - "loss": 0.4501, - "step": 2674 - }, - { - "epoch": 0.83, - "learning_rate": 6.865481621817205e-08, - "loss": 0.5079, - "step": 2675 - }, - { - "epoch": 0.83, - "learning_rate": 6.840347722586182e-08, - "loss": 0.4734, - "step": 2676 - }, - { - "epoch": 0.83, - "learning_rate": 6.815256535698149e-08, - "loss": 0.4554, - "step": 2677 - }, - { - "epoch": 0.83, - "learning_rate": 6.79020808598425e-08, - "loss": 0.439, - "step": 2678 - }, - { - "epoch": 0.84, - "learning_rate": 6.76520239823331e-08, - "loss": 0.4348, - "step": 2679 - }, - { - "epoch": 0.84, - "learning_rate": 6.740239497191869e-08, - "loss": 0.4504, - "step": 2680 - }, - { - "epoch": 0.84, - "learning_rate": 6.715319407564119e-08, - "loss": 0.4258, - "step": 2681 - }, - { - "epoch": 0.84, - "learning_rate": 6.690442154011855e-08, - "loss": 0.444, - "step": 2682 - }, - { - "epoch": 0.84, - "learning_rate": 6.665607761154512e-08, - "loss": 0.4536, - "step": 2683 - }, - { - "epoch": 0.84, - "learning_rate": 6.640816253569103e-08, - "loss": 0.4506, - "step": 2684 - }, - { - "epoch": 0.84, - "learning_rate": 6.616067655790203e-08, - "loss": 0.4581, - "step": 2685 - }, - { - "epoch": 0.84, - "learning_rate": 6.591361992309874e-08, - "loss": 0.4669, - "step": 2686 - }, - { - "epoch": 0.84, - "learning_rate": 6.566699287577782e-08, - "loss": 0.4586, - "step": 2687 - }, - { - "epoch": 0.84, - "learning_rate": 6.542079566000986e-08, - "loss": 0.4637, - "step": 2688 - }, - { - "epoch": 0.84, - "learning_rate": 6.517502851944074e-08, - "loss": 0.4516, - "step": 2689 - }, - { - "epoch": 0.84, - "learning_rate": 6.492969169729023e-08, - "loss": 0.4393, - "step": 2690 - }, - { - "epoch": 0.84, - "learning_rate": 6.468478543635258e-08, - "loss": 0.4622, - "step": 2691 - }, - { - "epoch": 0.84, - "learning_rate": 6.444030997899586e-08, - "loss": 0.4618, - "step": 2692 - }, - { - "epoch": 0.84, - "learning_rate": 6.419626556716168e-08, - "loss": 0.434, - "step": 2693 - }, - { - "epoch": 0.84, - "learning_rate": 6.395265244236514e-08, - "loss": 0.4472, - "step": 2694 - }, - { - "epoch": 0.84, - "learning_rate": 6.370947084569461e-08, - "loss": 0.4794, - "step": 2695 - }, - { - "epoch": 0.84, - "learning_rate": 6.346672101781148e-08, - "loss": 0.4618, - "step": 2696 - }, - { - "epoch": 0.84, - "learning_rate": 6.322440319894934e-08, - "loss": 0.465, - "step": 2697 - }, - { - "epoch": 0.84, - "learning_rate": 6.298251762891499e-08, - "loss": 0.4721, - "step": 2698 - }, - { - "epoch": 0.84, - "learning_rate": 6.274106454708683e-08, - "loss": 0.4775, - "step": 2699 - }, - { - "epoch": 0.84, - "learning_rate": 6.250004419241572e-08, - "loss": 0.4442, - "step": 2700 - }, - { - "epoch": 0.84, - "learning_rate": 6.225945680342381e-08, - "loss": 0.4882, - "step": 2701 - }, - { - "epoch": 0.84, - "learning_rate": 6.20193026182052e-08, - "loss": 0.4572, - "step": 2702 - }, - { - "epoch": 0.84, - "learning_rate": 6.177958187442506e-08, - "loss": 0.4762, - "step": 2703 - }, - { - "epoch": 0.84, - "learning_rate": 6.154029480931972e-08, - "loss": 0.4452, - "step": 2704 - }, - { - "epoch": 0.84, - "learning_rate": 6.130144165969614e-08, - "loss": 0.4578, - "step": 2705 - }, - { - "epoch": 0.84, - "learning_rate": 6.106302266193197e-08, - "loss": 0.4631, - "step": 2706 - }, - { - "epoch": 0.84, - "learning_rate": 6.082503805197542e-08, - "loss": 0.4498, - "step": 2707 - }, - { - "epoch": 0.84, - "learning_rate": 6.05874880653443e-08, - "loss": 0.4496, - "step": 2708 - }, - { - "epoch": 0.84, - "learning_rate": 6.035037293712697e-08, - "loss": 0.4641, - "step": 2709 - }, - { - "epoch": 0.84, - "learning_rate": 6.011369290198077e-08, - "loss": 0.471, - "step": 2710 - }, - { - "epoch": 0.85, - "learning_rate": 5.987744819413304e-08, - "loss": 0.4834, - "step": 2711 - }, - { - "epoch": 0.85, - "learning_rate": 5.964163904737967e-08, - "loss": 0.4725, - "step": 2712 - }, - { - "epoch": 0.85, - "learning_rate": 5.940626569508628e-08, - "loss": 0.4557, - "step": 2713 - }, - { - "epoch": 0.85, - "learning_rate": 5.917132837018657e-08, - "loss": 0.4461, - "step": 2714 - }, - { - "epoch": 0.85, - "learning_rate": 5.8936827305183014e-08, - "loss": 0.4773, - "step": 2715 - }, - { - "epoch": 0.85, - "learning_rate": 5.870276273214625e-08, - "loss": 0.4693, - "step": 2716 - }, - { - "epoch": 0.85, - "learning_rate": 5.8469134882714986e-08, - "loss": 0.4764, - "step": 2717 - }, - { - "epoch": 0.85, - "learning_rate": 5.823594398809589e-08, - "loss": 0.4632, - "step": 2718 - }, - { - "epoch": 0.85, - "learning_rate": 5.8003190279062695e-08, - "loss": 0.4356, - "step": 2719 - }, - { - "epoch": 0.85, - "learning_rate": 5.777087398595731e-08, - "loss": 0.4496, - "step": 2720 - }, - { - "epoch": 0.85, - "learning_rate": 5.753899533868795e-08, - "loss": 0.4622, - "step": 2721 - }, - { - "epoch": 0.85, - "learning_rate": 5.730755456673025e-08, - "loss": 0.4433, - "step": 2722 - }, - { - "epoch": 0.85, - "learning_rate": 5.7076551899126024e-08, - "loss": 0.4546, - "step": 2723 - }, - { - "epoch": 0.85, - "learning_rate": 5.684598756448422e-08, - "loss": 0.448, - "step": 2724 - }, - { - "epoch": 0.85, - "learning_rate": 5.661586179097927e-08, - "loss": 0.4467, - "step": 2725 - }, - { - "epoch": 0.85, - "learning_rate": 5.638617480635205e-08, - "loss": 0.4776, - "step": 2726 - }, - { - "epoch": 0.85, - "learning_rate": 5.6156926837909126e-08, - "loss": 0.4497, - "step": 2727 - }, - { - "epoch": 0.85, - "learning_rate": 5.5928118112522346e-08, - "loss": 0.4615, - "step": 2728 - }, - { - "epoch": 0.85, - "learning_rate": 5.569974885662909e-08, - "loss": 0.4665, - "step": 2729 - }, - { - "epoch": 0.85, - "learning_rate": 5.547181929623185e-08, - "loss": 0.4457, - "step": 2730 - }, - { - "epoch": 0.85, - "learning_rate": 5.5244329656897957e-08, - "loss": 0.452, - "step": 2731 - }, - { - "epoch": 0.85, - "learning_rate": 5.5017280163759175e-08, - "loss": 0.4628, - "step": 2732 - }, - { - "epoch": 0.85, - "learning_rate": 5.479067104151203e-08, - "loss": 0.4545, - "step": 2733 - }, - { - "epoch": 0.85, - "learning_rate": 5.456450251441675e-08, - "loss": 0.4677, - "step": 2734 - }, - { - "epoch": 0.85, - "learning_rate": 5.433877480629823e-08, - "loss": 0.4484, - "step": 2735 - }, - { - "epoch": 0.85, - "learning_rate": 5.411348814054445e-08, - "loss": 0.4596, - "step": 2736 - }, - { - "epoch": 0.85, - "learning_rate": 5.388864274010724e-08, - "loss": 0.4465, - "step": 2737 - }, - { - "epoch": 0.85, - "learning_rate": 5.3664238827501764e-08, - "loss": 0.4698, - "step": 2738 - }, - { - "epoch": 0.85, - "learning_rate": 5.344027662480621e-08, - "loss": 0.4537, - "step": 2739 - }, - { - "epoch": 0.85, - "learning_rate": 5.321675635366146e-08, - "loss": 0.474, - "step": 2740 - }, - { - "epoch": 0.85, - "learning_rate": 5.299367823527134e-08, - "loss": 0.4595, - "step": 2741 - }, - { - "epoch": 0.85, - "learning_rate": 5.2771042490402096e-08, - "loss": 0.4511, - "step": 2742 - }, - { - "epoch": 0.86, - "learning_rate": 5.2548849339381754e-08, - "loss": 0.4801, - "step": 2743 - }, - { - "epoch": 0.86, - "learning_rate": 5.232709900210092e-08, - "loss": 0.4586, - "step": 2744 - }, - { - "epoch": 0.86, - "learning_rate": 5.2105791698011324e-08, - "loss": 0.4539, - "step": 2745 - }, - { - "epoch": 0.86, - "learning_rate": 5.188492764612712e-08, - "loss": 0.4518, - "step": 2746 - }, - { - "epoch": 0.86, - "learning_rate": 5.166450706502301e-08, - "loss": 0.4721, - "step": 2747 - }, - { - "epoch": 0.86, - "learning_rate": 5.144453017283518e-08, - "loss": 0.4623, - "step": 2748 - }, - { - "epoch": 0.86, - "learning_rate": 5.122499718726064e-08, - "loss": 0.4642, - "step": 2749 - }, - { - "epoch": 0.86, - "learning_rate": 5.1005908325557236e-08, - "loss": 0.4687, - "step": 2750 - }, - { - "epoch": 0.86, - "learning_rate": 5.078726380454301e-08, - "loss": 0.4634, - "step": 2751 - }, - { - "epoch": 0.86, - "learning_rate": 5.056906384059656e-08, - "loss": 0.454, - "step": 2752 - }, - { - "epoch": 0.86, - "learning_rate": 5.0351308649656474e-08, - "loss": 0.4684, - "step": 2753 - }, - { - "epoch": 0.86, - "learning_rate": 5.013399844722088e-08, - "loss": 0.4383, - "step": 2754 - }, - { - "epoch": 0.86, - "learning_rate": 4.991713344834797e-08, - "loss": 0.4571, - "step": 2755 - }, - { - "epoch": 0.86, - "learning_rate": 4.970071386765501e-08, - "loss": 0.4701, - "step": 2756 - }, - { - "epoch": 0.86, - "learning_rate": 4.948473991931879e-08, - "loss": 0.4591, - "step": 2757 - }, - { - "epoch": 0.86, - "learning_rate": 4.926921181707466e-08, - "loss": 0.449, - "step": 2758 - }, - { - "epoch": 0.86, - "learning_rate": 4.905412977421708e-08, - "loss": 0.4471, - "step": 2759 - }, - { - "epoch": 0.86, - "learning_rate": 4.8839494003598993e-08, - "loss": 0.4596, - "step": 2760 - }, - { - "epoch": 0.86, - "learning_rate": 4.862530471763171e-08, - "loss": 0.4616, - "step": 2761 - }, - { - "epoch": 0.86, - "learning_rate": 4.841156212828445e-08, - "loss": 0.4488, - "step": 2762 - }, - { - "epoch": 0.86, - "learning_rate": 4.8198266447084746e-08, - "loss": 0.4667, - "step": 2763 - }, - { - "epoch": 0.86, - "learning_rate": 4.798541788511756e-08, - "loss": 0.4594, - "step": 2764 - }, - { - "epoch": 0.86, - "learning_rate": 4.777301665302569e-08, - "loss": 0.4596, - "step": 2765 - }, - { - "epoch": 0.86, - "learning_rate": 4.756106296100876e-08, - "loss": 0.4703, - "step": 2766 - }, - { - "epoch": 0.86, - "learning_rate": 4.7349557018823817e-08, - "loss": 0.4605, - "step": 2767 - }, - { - "epoch": 0.86, - "learning_rate": 4.713849903578493e-08, - "loss": 0.4687, - "step": 2768 - }, - { - "epoch": 0.86, - "learning_rate": 4.6927889220762464e-08, - "loss": 0.4417, - "step": 2769 - }, - { - "epoch": 0.86, - "learning_rate": 4.6717727782183527e-08, - "loss": 0.4653, - "step": 2770 - }, - { - "epoch": 0.86, - "learning_rate": 4.650801492803141e-08, - "loss": 0.4437, - "step": 2771 - }, - { - "epoch": 0.86, - "learning_rate": 4.629875086584562e-08, - "loss": 0.4538, - "step": 2772 - }, - { - "epoch": 0.86, - "learning_rate": 4.6089935802721134e-08, - "loss": 0.4443, - "step": 2773 - }, - { - "epoch": 0.86, - "learning_rate": 4.588156994530906e-08, - "loss": 0.4564, - "step": 2774 - }, - { - "epoch": 0.87, - "learning_rate": 4.567365349981567e-08, - "loss": 0.4468, - "step": 2775 - }, - { - "epoch": 0.87, - "learning_rate": 4.546618667200264e-08, - "loss": 0.4487, - "step": 2776 - }, - { - "epoch": 0.87, - "learning_rate": 4.5259169667186405e-08, - "loss": 0.474, - "step": 2777 - }, - { - "epoch": 0.87, - "learning_rate": 4.5052602690238584e-08, - "loss": 0.4441, - "step": 2778 - }, - { - "epoch": 0.87, - "learning_rate": 4.4846485945585266e-08, - "loss": 0.4542, - "step": 2779 - }, - { - "epoch": 0.87, - "learning_rate": 4.464081963720695e-08, - "loss": 0.4726, - "step": 2780 - }, - { - "epoch": 0.87, - "learning_rate": 4.443560396863838e-08, - "loss": 0.4509, - "step": 2781 - }, - { - "epoch": 0.87, - "learning_rate": 4.423083914296838e-08, - "loss": 0.4696, - "step": 2782 - }, - { - "epoch": 0.87, - "learning_rate": 4.4026525362839684e-08, - "loss": 0.4433, - "step": 2783 - }, - { - "epoch": 0.87, - "learning_rate": 4.3822662830448386e-08, - "loss": 0.4689, - "step": 2784 - }, - { - "epoch": 0.87, - "learning_rate": 4.3619251747544275e-08, - "loss": 0.4506, - "step": 2785 - }, - { - "epoch": 0.87, - "learning_rate": 4.3416292315430256e-08, - "loss": 0.4591, - "step": 2786 - }, - { - "epoch": 0.87, - "learning_rate": 4.3213784734962286e-08, - "loss": 0.5094, - "step": 2787 - }, - { - "epoch": 0.87, - "learning_rate": 4.3011729206548995e-08, - "loss": 0.4546, - "step": 2788 - }, - { - "epoch": 0.87, - "learning_rate": 4.28101259301521e-08, - "loss": 0.4614, - "step": 2789 - }, - { - "epoch": 0.87, - "learning_rate": 4.260897510528522e-08, - "loss": 0.4384, - "step": 2790 - }, - { - "epoch": 0.87, - "learning_rate": 4.240827693101451e-08, - "loss": 0.48, - "step": 2791 - }, - { - "epoch": 0.87, - "learning_rate": 4.2208031605958036e-08, - "loss": 0.4671, - "step": 2792 - }, - { - "epoch": 0.87, - "learning_rate": 4.2008239328285785e-08, - "loss": 0.4359, - "step": 2793 - }, - { - "epoch": 0.87, - "learning_rate": 4.180890029571954e-08, - "loss": 0.4784, - "step": 2794 - }, - { - "epoch": 0.87, - "learning_rate": 4.16100147055321e-08, - "loss": 0.462, - "step": 2795 - }, - { - "epoch": 0.87, - "learning_rate": 4.141158275454798e-08, - "loss": 0.4568, - "step": 2796 - }, - { - "epoch": 0.87, - "learning_rate": 4.1213604639142475e-08, - "loss": 0.4659, - "step": 2797 - }, - { - "epoch": 0.87, - "learning_rate": 4.101608055524197e-08, - "loss": 0.4819, - "step": 2798 - }, - { - "epoch": 0.87, - "learning_rate": 4.0819010698323164e-08, - "loss": 0.4424, - "step": 2799 - }, - { - "epoch": 0.87, - "learning_rate": 4.062239526341377e-08, - "loss": 0.4447, - "step": 2800 - }, - { - "epoch": 0.87, - "learning_rate": 4.0426234445091246e-08, - "loss": 0.4633, - "step": 2801 - }, - { - "epoch": 0.87, - "learning_rate": 4.023052843748359e-08, - "loss": 0.452, - "step": 2802 - }, - { - "epoch": 0.87, - "learning_rate": 4.0035277434268266e-08, - "loss": 0.4597, - "step": 2803 - }, - { - "epoch": 0.87, - "learning_rate": 3.984048162867282e-08, - "loss": 0.4533, - "step": 2804 - }, - { - "epoch": 0.87, - "learning_rate": 3.9646141213474205e-08, - "loss": 0.471, - "step": 2805 - }, - { - "epoch": 0.87, - "learning_rate": 3.945225638099864e-08, - "loss": 0.4599, - "step": 2806 - }, - { - "epoch": 0.88, - "learning_rate": 3.925882732312158e-08, - "loss": 0.4602, - "step": 2807 - }, - { - "epoch": 0.88, - "learning_rate": 3.906585423126724e-08, - "loss": 0.4725, - "step": 2808 - }, - { - "epoch": 0.88, - "learning_rate": 3.88733372964089e-08, - "loss": 0.4623, - "step": 2809 - }, - { - "epoch": 0.88, - "learning_rate": 3.868127670906796e-08, - "loss": 0.4615, - "step": 2810 - }, - { - "epoch": 0.88, - "learning_rate": 3.848967265931485e-08, - "loss": 0.4526, - "step": 2811 - }, - { - "epoch": 0.88, - "learning_rate": 3.829852533676753e-08, - "loss": 0.4729, - "step": 2812 - }, - { - "epoch": 0.88, - "learning_rate": 3.810783493059244e-08, - "loss": 0.4577, - "step": 2813 - }, - { - "epoch": 0.88, - "learning_rate": 3.791760162950347e-08, - "loss": 0.4567, - "step": 2814 - }, - { - "epoch": 0.88, - "learning_rate": 3.7727825621762575e-08, - "loss": 0.457, - "step": 2815 - }, - { - "epoch": 0.88, - "learning_rate": 3.75385070951787e-08, - "loss": 0.4388, - "step": 2816 - }, - { - "epoch": 0.88, - "learning_rate": 3.7349646237108345e-08, - "loss": 0.4755, - "step": 2817 - }, - { - "epoch": 0.88, - "learning_rate": 3.716124323445513e-08, - "loss": 0.455, - "step": 2818 - }, - { - "epoch": 0.88, - "learning_rate": 3.6973298273669184e-08, - "loss": 0.4833, - "step": 2819 - }, - { - "epoch": 0.88, - "learning_rate": 3.678581154074784e-08, - "loss": 0.4834, - "step": 2820 - }, - { - "epoch": 0.88, - "learning_rate": 3.659878322123439e-08, - "loss": 0.4387, - "step": 2821 - }, - { - "epoch": 0.88, - "learning_rate": 3.641221350021917e-08, - "loss": 0.4513, - "step": 2822 - }, - { - "epoch": 0.88, - "learning_rate": 3.622610256233799e-08, - "loss": 0.4571, - "step": 2823 - }, - { - "epoch": 0.88, - "learning_rate": 3.604045059177313e-08, - "loss": 0.4593, - "step": 2824 - }, - { - "epoch": 0.88, - "learning_rate": 3.5855257772252126e-08, - "loss": 0.4633, - "step": 2825 - }, - { - "epoch": 0.88, - "learning_rate": 3.567052428704892e-08, - "loss": 0.4757, - "step": 2826 - }, - { - "epoch": 0.88, - "learning_rate": 3.5486250318982005e-08, - "loss": 0.4612, - "step": 2827 - }, - { - "epoch": 0.88, - "learning_rate": 3.530243605041566e-08, - "loss": 0.4855, - "step": 2828 - }, - { - "epoch": 0.88, - "learning_rate": 3.511908166325928e-08, - "loss": 0.4572, - "step": 2829 - }, - { - "epoch": 0.88, - "learning_rate": 3.4936187338966584e-08, - "loss": 0.4561, - "step": 2830 - }, - { - "epoch": 0.88, - "learning_rate": 3.4753753258536636e-08, - "loss": 0.4808, - "step": 2831 - }, - { - "epoch": 0.88, - "learning_rate": 3.457177960251267e-08, - "loss": 0.446, - "step": 2832 - }, - { - "epoch": 0.88, - "learning_rate": 3.439026655098243e-08, - "loss": 0.4445, - "step": 2833 - }, - { - "epoch": 0.88, - "learning_rate": 3.4209214283577645e-08, - "loss": 0.4653, - "step": 2834 - }, - { - "epoch": 0.88, - "learning_rate": 3.402862297947429e-08, - "loss": 0.4645, - "step": 2835 - }, - { - "epoch": 0.88, - "learning_rate": 3.3848492817391715e-08, - "loss": 0.4464, - "step": 2836 - }, - { - "epoch": 0.88, - "learning_rate": 3.366882397559362e-08, - "loss": 0.4496, - "step": 2837 - }, - { - "epoch": 0.88, - "learning_rate": 3.348961663188643e-08, - "loss": 0.4567, - "step": 2838 - }, - { - "epoch": 0.88, - "learning_rate": 3.331087096362023e-08, - "loss": 0.4565, - "step": 2839 - }, - { - "epoch": 0.89, - "learning_rate": 3.313258714768824e-08, - "loss": 0.4379, - "step": 2840 - }, - { - "epoch": 0.89, - "learning_rate": 3.2954765360526605e-08, - "loss": 0.4358, - "step": 2841 - }, - { - "epoch": 0.89, - "learning_rate": 3.277740577811394e-08, - "loss": 0.4568, - "step": 2842 - }, - { - "epoch": 0.89, - "learning_rate": 3.260050857597174e-08, - "loss": 0.4552, - "step": 2843 - }, - { - "epoch": 0.89, - "learning_rate": 3.242407392916396e-08, - "loss": 0.4532, - "step": 2844 - }, - { - "epoch": 0.89, - "learning_rate": 3.2248102012296494e-08, - "loss": 0.465, - "step": 2845 - }, - { - "epoch": 0.89, - "learning_rate": 3.2072592999517533e-08, - "loss": 0.4565, - "step": 2846 - }, - { - "epoch": 0.89, - "learning_rate": 3.189754706451714e-08, - "loss": 0.4638, - "step": 2847 - }, - { - "epoch": 0.89, - "learning_rate": 3.172296438052713e-08, - "loss": 0.4495, - "step": 2848 - }, - { - "epoch": 0.89, - "learning_rate": 3.1548845120320633e-08, - "loss": 0.4574, - "step": 2849 - }, - { - "epoch": 0.89, - "learning_rate": 3.137518945621237e-08, - "loss": 0.4664, - "step": 2850 - }, - { - "epoch": 0.89, - "learning_rate": 3.120199756005831e-08, - "loss": 0.4365, - "step": 2851 - }, - { - "epoch": 0.89, - "learning_rate": 3.102926960325541e-08, - "loss": 0.4638, - "step": 2852 - }, - { - "epoch": 0.89, - "learning_rate": 3.0857005756741274e-08, - "loss": 0.4702, - "step": 2853 - }, - { - "epoch": 0.89, - "learning_rate": 3.0685206190994514e-08, - "loss": 0.4584, - "step": 2854 - }, - { - "epoch": 0.89, - "learning_rate": 3.0513871076034256e-08, - "loss": 0.4668, - "step": 2855 - }, - { - "epoch": 0.89, - "learning_rate": 3.0343000581419643e-08, - "loss": 0.4538, - "step": 2856 - }, - { - "epoch": 0.89, - "learning_rate": 3.017259487625046e-08, - "loss": 0.4402, - "step": 2857 - }, - { - "epoch": 0.89, - "learning_rate": 3.0002654129166194e-08, - "loss": 0.4489, - "step": 2858 - }, - { - "epoch": 0.89, - "learning_rate": 2.983317850834649e-08, - "loss": 0.4488, - "step": 2859 - }, - { - "epoch": 0.89, - "learning_rate": 2.9664168181510297e-08, - "loss": 0.4575, - "step": 2860 - }, - { - "epoch": 0.89, - "learning_rate": 2.949562331591643e-08, - "loss": 0.4532, - "step": 2861 - }, - { - "epoch": 0.89, - "learning_rate": 2.932754407836291e-08, - "loss": 0.4697, - "step": 2862 - }, - { - "epoch": 0.89, - "learning_rate": 2.9159930635187068e-08, - "loss": 0.4406, - "step": 2863 - }, - { - "epoch": 0.89, - "learning_rate": 2.8992783152265055e-08, - "loss": 0.4595, - "step": 2864 - }, - { - "epoch": 0.89, - "learning_rate": 2.8826101795012103e-08, - "loss": 0.4978, - "step": 2865 - }, - { - "epoch": 0.89, - "learning_rate": 2.8659886728382044e-08, - "loss": 0.4553, - "step": 2866 - }, - { - "epoch": 0.89, - "learning_rate": 2.8494138116867407e-08, - "loss": 0.4504, - "step": 2867 - }, - { - "epoch": 0.89, - "learning_rate": 2.8328856124498814e-08, - "loss": 0.4468, - "step": 2868 - }, - { - "epoch": 0.89, - "learning_rate": 2.8164040914845256e-08, - "loss": 0.4755, - "step": 2869 - }, - { - "epoch": 0.89, - "learning_rate": 2.7999692651013983e-08, - "loss": 0.4545, - "step": 2870 - }, - { - "epoch": 0.89, - "learning_rate": 2.783581149564973e-08, - "loss": 0.4543, - "step": 2871 - }, - { - "epoch": 0.9, - "learning_rate": 2.7672397610935317e-08, - "loss": 0.4834, - "step": 2872 - }, - { - "epoch": 0.9, - "learning_rate": 2.750945115859088e-08, - "loss": 0.4609, - "step": 2873 - }, - { - "epoch": 0.9, - "learning_rate": 2.734697229987426e-08, - "loss": 0.4623, - "step": 2874 - }, - { - "epoch": 0.9, - "learning_rate": 2.7184961195580168e-08, - "loss": 0.4494, - "step": 2875 - }, - { - "epoch": 0.9, - "learning_rate": 2.702341800604069e-08, - "loss": 0.4519, - "step": 2876 - }, - { - "epoch": 0.9, - "learning_rate": 2.6862342891124823e-08, - "loss": 0.4464, - "step": 2877 - }, - { - "epoch": 0.9, - "learning_rate": 2.670173601023834e-08, - "loss": 0.4543, - "step": 2878 - }, - { - "epoch": 0.9, - "learning_rate": 2.6541597522323377e-08, - "loss": 0.4456, - "step": 2879 - }, - { - "epoch": 0.9, - "learning_rate": 2.638192758585894e-08, - "loss": 0.4595, - "step": 2880 - }, - { - "epoch": 0.9, - "learning_rate": 2.622272635886008e-08, - "loss": 0.4363, - "step": 2881 - }, - { - "epoch": 0.9, - "learning_rate": 2.6063993998877932e-08, - "loss": 0.4672, - "step": 2882 - }, - { - "epoch": 0.9, - "learning_rate": 2.5905730662999893e-08, - "loss": 0.4469, - "step": 2883 - }, - { - "epoch": 0.9, - "learning_rate": 2.5747936507849012e-08, - "loss": 0.4694, - "step": 2884 - }, - { - "epoch": 0.9, - "learning_rate": 2.559061168958404e-08, - "loss": 0.4408, - "step": 2885 - }, - { - "epoch": 0.9, - "learning_rate": 2.5433756363899274e-08, - "loss": 0.4547, - "step": 2886 - }, - { - "epoch": 0.9, - "learning_rate": 2.5277370686024313e-08, - "loss": 0.4506, - "step": 2887 - }, - { - "epoch": 0.9, - "learning_rate": 2.5121454810724198e-08, - "loss": 0.4607, - "step": 2888 - }, - { - "epoch": 0.9, - "learning_rate": 2.4966008892298774e-08, - "loss": 0.4567, - "step": 2889 - }, - { - "epoch": 0.9, - "learning_rate": 2.481103308458282e-08, - "loss": 0.4754, - "step": 2890 - }, - { - "epoch": 0.9, - "learning_rate": 2.465652754094627e-08, - "loss": 0.4691, - "step": 2891 - }, - { - "epoch": 0.9, - "learning_rate": 2.450249241429303e-08, - "loss": 0.462, - "step": 2892 - }, - { - "epoch": 0.9, - "learning_rate": 2.434892785706205e-08, - "loss": 0.4486, - "step": 2893 - }, - { - "epoch": 0.9, - "learning_rate": 2.4195834021226103e-08, - "loss": 0.4451, - "step": 2894 - }, - { - "epoch": 0.9, - "learning_rate": 2.404321105829249e-08, - "loss": 0.4357, - "step": 2895 - }, - { - "epoch": 0.9, - "learning_rate": 2.3891059119302447e-08, - "loss": 0.4613, - "step": 2896 - }, - { - "epoch": 0.9, - "learning_rate": 2.3739378354830754e-08, - "loss": 0.4496, - "step": 2897 - }, - { - "epoch": 0.9, - "learning_rate": 2.358816891498644e-08, - "loss": 0.4513, - "step": 2898 - }, - { - "epoch": 0.9, - "learning_rate": 2.3437430949411584e-08, - "loss": 0.4365, - "step": 2899 - }, - { - "epoch": 0.9, - "learning_rate": 2.328716460728203e-08, - "loss": 0.4371, - "step": 2900 - }, - { - "epoch": 0.9, - "learning_rate": 2.3137370037306593e-08, - "loss": 0.4502, - "step": 2901 - }, - { - "epoch": 0.9, - "learning_rate": 2.2988047387727583e-08, - "loss": 0.4452, - "step": 2902 - }, - { - "epoch": 0.9, - "learning_rate": 2.2839196806319905e-08, - "loss": 0.4491, - "step": 2903 - }, - { - "epoch": 0.91, - "learning_rate": 2.26908184403915e-08, - "loss": 0.4596, - "step": 2904 - }, - { - "epoch": 0.91, - "learning_rate": 2.2542912436782913e-08, - "loss": 0.4715, - "step": 2905 - }, - { - "epoch": 0.91, - "learning_rate": 2.2395478941867163e-08, - "loss": 0.4475, - "step": 2906 - }, - { - "epoch": 0.91, - "learning_rate": 2.2248518101549928e-08, - "loss": 0.4558, - "step": 2907 - }, - { - "epoch": 0.91, - "learning_rate": 2.210203006126865e-08, - "loss": 0.4589, - "step": 2908 - }, - { - "epoch": 0.91, - "learning_rate": 2.195601496599353e-08, - "loss": 0.4484, - "step": 2909 - }, - { - "epoch": 0.91, - "learning_rate": 2.1810472960226033e-08, - "loss": 0.4763, - "step": 2910 - }, - { - "epoch": 0.91, - "learning_rate": 2.1665404187999948e-08, - "loss": 0.4627, - "step": 2911 - }, - { - "epoch": 0.91, - "learning_rate": 2.1520808792880374e-08, - "loss": 0.4565, - "step": 2912 - }, - { - "epoch": 0.91, - "learning_rate": 2.1376686917964293e-08, - "loss": 0.4501, - "step": 2913 - }, - { - "epoch": 0.91, - "learning_rate": 2.123303870587978e-08, - "loss": 0.4628, - "step": 2914 - }, - { - "epoch": 0.91, - "learning_rate": 2.1089864298786397e-08, - "loss": 0.4345, - "step": 2915 - }, - { - "epoch": 0.91, - "learning_rate": 2.0947163838374416e-08, - "loss": 0.4514, - "step": 2916 - }, - { - "epoch": 0.91, - "learning_rate": 2.0804937465865646e-08, - "loss": 0.4504, - "step": 2917 - }, - { - "epoch": 0.91, - "learning_rate": 2.066318532201222e-08, - "loss": 0.4439, - "step": 2918 - }, - { - "epoch": 0.91, - "learning_rate": 2.0521907547097204e-08, - "loss": 0.4561, - "step": 2919 - }, - { - "epoch": 0.91, - "learning_rate": 2.038110428093426e-08, - "loss": 0.4683, - "step": 2920 - }, - { - "epoch": 0.91, - "learning_rate": 2.0240775662867193e-08, - "loss": 0.4459, - "step": 2921 - }, - { - "epoch": 0.91, - "learning_rate": 2.0100921831770423e-08, - "loss": 0.4446, - "step": 2922 - }, - { - "epoch": 0.91, - "learning_rate": 1.9961542926048068e-08, - "loss": 0.4607, - "step": 2923 - }, - { - "epoch": 0.91, - "learning_rate": 1.9822639083634795e-08, - "loss": 0.4525, - "step": 2924 - }, - { - "epoch": 0.91, - "learning_rate": 1.9684210441994643e-08, - "loss": 0.4409, - "step": 2925 - }, - { - "epoch": 0.91, - "learning_rate": 1.9546257138121746e-08, - "loss": 0.4499, - "step": 2926 - }, - { - "epoch": 0.91, - "learning_rate": 1.9408779308539346e-08, - "loss": 0.4546, - "step": 2927 - }, - { - "epoch": 0.91, - "learning_rate": 1.9271777089300835e-08, - "loss": 0.4776, - "step": 2928 - }, - { - "epoch": 0.91, - "learning_rate": 1.9135250615988196e-08, - "loss": 0.456, - "step": 2929 - }, - { - "epoch": 0.91, - "learning_rate": 1.8999200023713135e-08, - "loss": 0.4893, - "step": 2930 - }, - { - "epoch": 0.91, - "learning_rate": 1.886362544711617e-08, - "loss": 0.4613, - "step": 2931 - }, - { - "epoch": 0.91, - "learning_rate": 1.8728527020366758e-08, - "loss": 0.4693, - "step": 2932 - }, - { - "epoch": 0.91, - "learning_rate": 1.859390487716317e-08, - "loss": 0.4789, - "step": 2933 - }, - { - "epoch": 0.91, - "learning_rate": 1.8459759150732167e-08, - "loss": 0.4474, - "step": 2934 - }, - { - "epoch": 0.91, - "learning_rate": 1.8326089973829452e-08, - "loss": 0.4541, - "step": 2935 - }, - { - "epoch": 0.92, - "learning_rate": 1.8192897478738644e-08, - "loss": 0.4436, - "step": 2936 - }, - { - "epoch": 0.92, - "learning_rate": 1.8060181797271868e-08, - "loss": 0.4548, - "step": 2937 - }, - { - "epoch": 0.92, - "learning_rate": 1.7927943060769335e-08, - "loss": 0.4548, - "step": 2938 - }, - { - "epoch": 0.92, - "learning_rate": 1.7796181400099365e-08, - "loss": 0.467, - "step": 2939 - }, - { - "epoch": 0.92, - "learning_rate": 1.766489694565787e-08, - "loss": 0.4785, - "step": 2940 - }, - { - "epoch": 0.92, - "learning_rate": 1.7534089827368702e-08, - "loss": 0.4481, - "step": 2941 - }, - { - "epoch": 0.92, - "learning_rate": 1.740376017468331e-08, - "loss": 0.4662, - "step": 2942 - }, - { - "epoch": 0.92, - "learning_rate": 1.727390811658069e-08, - "loss": 0.4543, - "step": 2943 - }, - { - "epoch": 0.92, - "learning_rate": 1.7144533781566938e-08, - "loss": 0.4687, - "step": 2944 - }, - { - "epoch": 0.92, - "learning_rate": 1.701563729767569e-08, - "loss": 0.4782, - "step": 2945 - }, - { - "epoch": 0.92, - "learning_rate": 1.6887218792467463e-08, - "loss": 0.4769, - "step": 2946 - }, - { - "epoch": 0.92, - "learning_rate": 1.675927839302982e-08, - "loss": 0.457, - "step": 2947 - }, - { - "epoch": 0.92, - "learning_rate": 1.6631816225977148e-08, - "loss": 0.4587, - "step": 2948 - }, - { - "epoch": 0.92, - "learning_rate": 1.6504832417450653e-08, - "loss": 0.4595, - "step": 2949 - }, - { - "epoch": 0.92, - "learning_rate": 1.6378327093118148e-08, - "loss": 0.4535, - "step": 2950 - }, - { - "epoch": 0.92, - "learning_rate": 1.6252300378173657e-08, - "loss": 0.4554, - "step": 2951 - }, - { - "epoch": 0.92, - "learning_rate": 1.6126752397337908e-08, - "loss": 0.4572, - "step": 2952 - }, - { - "epoch": 0.92, - "learning_rate": 1.600168327485757e-08, - "loss": 0.5025, - "step": 2953 - }, - { - "epoch": 0.92, - "learning_rate": 1.5877093134505692e-08, - "loss": 0.462, - "step": 2954 - }, - { - "epoch": 0.92, - "learning_rate": 1.5752982099581034e-08, - "loss": 0.4433, - "step": 2955 - }, - { - "epoch": 0.92, - "learning_rate": 1.562935029290835e-08, - "loss": 0.4552, - "step": 2956 - }, - { - "epoch": 0.92, - "learning_rate": 1.5506197836838208e-08, - "loss": 0.481, - "step": 2957 - }, - { - "epoch": 0.92, - "learning_rate": 1.5383524853246567e-08, - "loss": 0.4362, - "step": 2958 - }, - { - "epoch": 0.92, - "learning_rate": 1.5261331463535088e-08, - "loss": 0.4511, - "step": 2959 - }, - { - "epoch": 0.92, - "learning_rate": 1.5139617788630766e-08, - "loss": 0.4753, - "step": 2960 - }, - { - "epoch": 0.92, - "learning_rate": 1.5018383948985802e-08, - "loss": 0.461, - "step": 2961 - }, - { - "epoch": 0.92, - "learning_rate": 1.489763006457756e-08, - "loss": 0.464, - "step": 2962 - }, - { - "epoch": 0.92, - "learning_rate": 1.4777356254908446e-08, - "loss": 0.4595, - "step": 2963 - }, - { - "epoch": 0.92, - "learning_rate": 1.4657562639005694e-08, - "loss": 0.4427, - "step": 2964 - }, - { - "epoch": 0.92, - "learning_rate": 1.4538249335421471e-08, - "loss": 0.4582, - "step": 2965 - }, - { - "epoch": 0.92, - "learning_rate": 1.4419416462232381e-08, - "loss": 0.4805, - "step": 2966 - }, - { - "epoch": 0.92, - "learning_rate": 1.4301064137039742e-08, - "loss": 0.4418, - "step": 2967 - }, - { - "epoch": 0.93, - "learning_rate": 1.4183192476969309e-08, - "loss": 0.4469, - "step": 2968 - }, - { - "epoch": 0.93, - "learning_rate": 1.4065801598671157e-08, - "loss": 0.4652, - "step": 2969 - }, - { - "epoch": 0.93, - "learning_rate": 1.3948891618319413e-08, - "loss": 0.458, - "step": 2970 - }, - { - "epoch": 0.93, - "learning_rate": 1.3832462651612475e-08, - "loss": 0.4651, - "step": 2971 - }, - { - "epoch": 0.93, - "learning_rate": 1.3716514813772561e-08, - "loss": 0.44, - "step": 2972 - }, - { - "epoch": 0.93, - "learning_rate": 1.3601048219545885e-08, - "loss": 0.4698, - "step": 2973 - }, - { - "epoch": 0.93, - "learning_rate": 1.3486062983202262e-08, - "loss": 0.4522, - "step": 2974 - }, - { - "epoch": 0.93, - "learning_rate": 1.3371559218535277e-08, - "loss": 0.4606, - "step": 2975 - }, - { - "epoch": 0.93, - "learning_rate": 1.325753703886201e-08, - "loss": 0.4598, - "step": 2976 - }, - { - "epoch": 0.93, - "learning_rate": 1.3143996557022752e-08, - "loss": 0.4632, - "step": 2977 - }, - { - "epoch": 0.93, - "learning_rate": 1.3030937885381455e-08, - "loss": 0.4714, - "step": 2978 - }, - { - "epoch": 0.93, - "learning_rate": 1.2918361135824896e-08, - "loss": 0.4471, - "step": 2979 - }, - { - "epoch": 0.93, - "learning_rate": 1.2806266419763123e-08, - "loss": 0.4556, - "step": 2980 - }, - { - "epoch": 0.93, - "learning_rate": 1.2694653848129067e-08, - "loss": 0.431, - "step": 2981 - }, - { - "epoch": 0.93, - "learning_rate": 1.2583523531378536e-08, - "loss": 0.4669, - "step": 2982 - }, - { - "epoch": 0.93, - "learning_rate": 1.2472875579490117e-08, - "loss": 0.4312, - "step": 2983 - }, - { - "epoch": 0.93, - "learning_rate": 1.236271010196499e-08, - "loss": 0.4549, - "step": 2984 - }, - { - "epoch": 0.93, - "learning_rate": 1.2253027207826893e-08, - "loss": 0.4706, - "step": 2985 - }, - { - "epoch": 0.93, - "learning_rate": 1.2143827005621887e-08, - "loss": 0.4342, - "step": 2986 - }, - { - "epoch": 0.93, - "learning_rate": 1.2035109603418526e-08, - "loss": 0.4507, - "step": 2987 - }, - { - "epoch": 0.93, - "learning_rate": 1.19268751088073e-08, - "loss": 0.4528, - "step": 2988 - }, - { - "epoch": 0.93, - "learning_rate": 1.181912362890114e-08, - "loss": 0.4512, - "step": 2989 - }, - { - "epoch": 0.93, - "learning_rate": 1.1711855270334636e-08, - "loss": 0.4775, - "step": 2990 - }, - { - "epoch": 0.93, - "learning_rate": 1.160507013926454e-08, - "loss": 0.4502, - "step": 2991 - }, - { - "epoch": 0.93, - "learning_rate": 1.1498768341369147e-08, - "loss": 0.4402, - "step": 2992 - }, - { - "epoch": 0.93, - "learning_rate": 1.1392949981848533e-08, - "loss": 0.4424, - "step": 2993 - }, - { - "epoch": 0.93, - "learning_rate": 1.1287615165424424e-08, - "loss": 0.4654, - "step": 2994 - }, - { - "epoch": 0.93, - "learning_rate": 1.1182763996339938e-08, - "loss": 0.4524, - "step": 2995 - }, - { - "epoch": 0.93, - "learning_rate": 1.1078396578359517e-08, - "loss": 0.4717, - "step": 2996 - }, - { - "epoch": 0.93, - "learning_rate": 1.097451301476887e-08, - "loss": 0.4686, - "step": 2997 - }, - { - "epoch": 0.93, - "learning_rate": 1.0871113408375043e-08, - "loss": 0.4324, - "step": 2998 - }, - { - "epoch": 0.93, - "learning_rate": 1.0768197861505846e-08, - "loss": 0.4447, - "step": 2999 - }, - { - "epoch": 0.94, - "learning_rate": 1.0665766476010307e-08, - "loss": 0.4669, - "step": 3000 - }, - { - "epoch": 0.94, - "learning_rate": 1.056381935325812e-08, - "loss": 0.459, - "step": 3001 - }, - { - "epoch": 0.94, - "learning_rate": 1.0462356594139966e-08, - "loss": 0.4578, - "step": 3002 - }, - { - "epoch": 0.94, - "learning_rate": 1.0361378299066804e-08, - "loss": 0.4556, - "step": 3003 - }, - { - "epoch": 0.94, - "learning_rate": 1.026088456797064e-08, - "loss": 0.4639, - "step": 3004 - }, - { - "epoch": 0.94, - "learning_rate": 1.0160875500303535e-08, - "loss": 0.4438, - "step": 3005 - }, - { - "epoch": 0.94, - "learning_rate": 1.0061351195038148e-08, - "loss": 0.4633, - "step": 3006 - }, - { - "epoch": 0.94, - "learning_rate": 9.96231175066714e-09, - "loss": 0.4677, - "step": 3007 - }, - { - "epoch": 0.94, - "learning_rate": 9.863757265203665e-09, - "loss": 0.4511, - "step": 3008 - }, - { - "epoch": 0.94, - "learning_rate": 9.765687836180759e-09, - "loss": 0.4719, - "step": 3009 - }, - { - "epoch": 0.94, - "learning_rate": 9.668103560651342e-09, - "loss": 0.4745, - "step": 3010 - }, - { - "epoch": 0.94, - "learning_rate": 9.571004535188498e-09, - "loss": 0.4588, - "step": 3011 - }, - { - "epoch": 0.94, - "learning_rate": 9.47439085588475e-09, - "loss": 0.4607, - "step": 3012 - }, - { - "epoch": 0.94, - "learning_rate": 9.378262618352617e-09, - "loss": 0.4558, - "step": 3013 - }, - { - "epoch": 0.94, - "learning_rate": 9.282619917723833e-09, - "loss": 0.442, - "step": 3014 - }, - { - "epoch": 0.94, - "learning_rate": 9.18746284865013e-09, - "loss": 0.4786, - "step": 3015 - }, - { - "epoch": 0.94, - "learning_rate": 9.092791505302177e-09, - "loss": 0.4753, - "step": 3016 - }, - { - "epoch": 0.94, - "learning_rate": 8.998605981370144e-09, - "loss": 0.4683, - "step": 3017 - }, - { - "epoch": 0.94, - "learning_rate": 8.904906370063526e-09, - "loss": 0.4521, - "step": 3018 - }, - { - "epoch": 0.94, - "learning_rate": 8.811692764110757e-09, - "loss": 0.4503, - "step": 3019 - }, - { - "epoch": 0.94, - "learning_rate": 8.718965255759326e-09, - "loss": 0.4672, - "step": 3020 - }, - { - "epoch": 0.94, - "learning_rate": 8.626723936775715e-09, - "loss": 0.4512, - "step": 3021 - }, - { - "epoch": 0.94, - "learning_rate": 8.534968898445405e-09, - "loss": 0.4615, - "step": 3022 - }, - { - "epoch": 0.94, - "learning_rate": 8.44370023157226e-09, - "loss": 0.4353, - "step": 3023 - }, - { - "epoch": 0.94, - "learning_rate": 8.352918026479194e-09, - "loss": 0.4568, - "step": 3024 - }, - { - "epoch": 0.94, - "learning_rate": 8.262622373007399e-09, - "loss": 0.4722, - "step": 3025 - }, - { - "epoch": 0.94, - "learning_rate": 8.17281336051684e-09, - "loss": 0.4829, - "step": 3026 - }, - { - "epoch": 0.94, - "learning_rate": 8.08349107788564e-09, - "loss": 0.4725, - "step": 3027 - }, - { - "epoch": 0.94, - "learning_rate": 7.994655613510426e-09, - "loss": 0.4369, - "step": 3028 - }, - { - "epoch": 0.94, - "learning_rate": 7.906307055305872e-09, - "loss": 0.4607, - "step": 3029 - }, - { - "epoch": 0.94, - "learning_rate": 7.818445490704984e-09, - "loss": 0.4626, - "step": 3030 - }, - { - "epoch": 0.94, - "learning_rate": 7.731071006658651e-09, - "loss": 0.4502, - "step": 3031 - }, - { - "epoch": 0.95, - "learning_rate": 7.644183689635818e-09, - "loss": 0.441, - "step": 3032 - }, - { - "epoch": 0.95, - "learning_rate": 7.557783625623315e-09, - "loss": 0.4792, - "step": 3033 - }, - { - "epoch": 0.95, - "learning_rate": 7.471870900125742e-09, - "loss": 0.4756, - "step": 3034 - }, - { - "epoch": 0.95, - "learning_rate": 7.386445598165425e-09, - "loss": 0.4311, - "step": 3035 - }, - { - "epoch": 0.95, - "learning_rate": 7.301507804282236e-09, - "loss": 0.4812, - "step": 3036 - }, - { - "epoch": 0.95, - "learning_rate": 7.217057602533827e-09, - "loss": 0.4741, - "step": 3037 - }, - { - "epoch": 0.95, - "learning_rate": 7.13309507649501e-09, - "loss": 0.4667, - "step": 3038 - }, - { - "epoch": 0.95, - "learning_rate": 7.049620309258208e-09, - "loss": 0.483, - "step": 3039 - }, - { - "epoch": 0.95, - "learning_rate": 6.9666333834329515e-09, - "loss": 0.4739, - "step": 3040 - }, - { - "epoch": 0.95, - "learning_rate": 6.884134381146267e-09, - "loss": 0.4746, - "step": 3041 - }, - { - "epoch": 0.95, - "learning_rate": 6.8021233840419025e-09, - "loss": 0.456, - "step": 3042 - }, - { - "epoch": 0.95, - "learning_rate": 6.720600473281046e-09, - "loss": 0.4561, - "step": 3043 - }, - { - "epoch": 0.95, - "learning_rate": 6.639565729541663e-09, - "loss": 0.4341, - "step": 3044 - }, - { - "epoch": 0.95, - "learning_rate": 6.559019233018604e-09, - "loss": 0.4511, - "step": 3045 - }, - { - "epoch": 0.95, - "learning_rate": 6.478961063423605e-09, - "loss": 0.4813, - "step": 3046 - }, - { - "epoch": 0.95, - "learning_rate": 6.399391299985124e-09, - "loss": 0.457, - "step": 3047 - }, - { - "epoch": 0.95, - "learning_rate": 6.320310021448283e-09, - "loss": 0.4555, - "step": 3048 - }, - { - "epoch": 0.95, - "learning_rate": 6.241717306074701e-09, - "loss": 0.4543, - "step": 3049 - }, - { - "epoch": 0.95, - "learning_rate": 6.163613231642551e-09, - "loss": 0.4732, - "step": 3050 - }, - { - "epoch": 0.95, - "learning_rate": 6.085997875446447e-09, - "loss": 0.4267, - "step": 3051 - }, - { - "epoch": 0.95, - "learning_rate": 6.008871314297392e-09, - "loss": 0.4412, - "step": 3052 - }, - { - "epoch": 0.95, - "learning_rate": 5.932233624522498e-09, - "loss": 0.4609, - "step": 3053 - }, - { - "epoch": 0.95, - "learning_rate": 5.856084881965262e-09, - "loss": 0.4607, - "step": 3054 - }, - { - "epoch": 0.95, - "learning_rate": 5.7804251619851805e-09, - "loss": 0.4294, - "step": 3055 - }, - { - "epoch": 0.95, - "learning_rate": 5.705254539457915e-09, - "loss": 0.4592, - "step": 3056 - }, - { - "epoch": 0.95, - "learning_rate": 5.6305730887748486e-09, - "loss": 0.4444, - "step": 3057 - }, - { - "epoch": 0.95, - "learning_rate": 5.556380883843526e-09, - "loss": 0.4635, - "step": 3058 - }, - { - "epoch": 0.95, - "learning_rate": 5.482677998087215e-09, - "loss": 0.4813, - "step": 3059 - }, - { - "epoch": 0.95, - "learning_rate": 5.409464504444961e-09, - "loss": 0.4521, - "step": 3060 - }, - { - "epoch": 0.95, - "learning_rate": 5.336740475371415e-09, - "loss": 0.4586, - "step": 3061 - }, - { - "epoch": 0.95, - "learning_rate": 5.264505982836953e-09, - "loss": 0.4412, - "step": 3062 - }, - { - "epoch": 0.95, - "learning_rate": 5.192761098327392e-09, - "loss": 0.4582, - "step": 3063 - }, - { - "epoch": 0.96, - "learning_rate": 5.121505892843991e-09, - "loss": 0.4629, - "step": 3064 - }, - { - "epoch": 0.96, - "learning_rate": 5.050740436903567e-09, - "loss": 0.4675, - "step": 3065 - }, - { - "epoch": 0.96, - "learning_rate": 4.980464800538098e-09, - "loss": 0.4594, - "step": 3066 - }, - { - "epoch": 0.96, - "learning_rate": 4.910679053294897e-09, - "loss": 0.4553, - "step": 3067 - }, - { - "epoch": 0.96, - "learning_rate": 4.8413832642364425e-09, - "loss": 0.4514, - "step": 3068 - }, - { - "epoch": 0.96, - "learning_rate": 4.772577501940323e-09, - "loss": 0.4615, - "step": 3069 - }, - { - "epoch": 0.96, - "learning_rate": 4.7042618344992344e-09, - "loss": 0.4445, - "step": 3070 - }, - { - "epoch": 0.96, - "learning_rate": 4.636436329520765e-09, - "loss": 0.458, - "step": 3071 - }, - { - "epoch": 0.96, - "learning_rate": 4.5691010541275e-09, - "loss": 0.4571, - "step": 3072 - }, - { - "epoch": 0.96, - "learning_rate": 4.5022560749568e-09, - "loss": 0.4618, - "step": 3073 - }, - { - "epoch": 0.96, - "learning_rate": 4.435901458160973e-09, - "loss": 0.4723, - "step": 3074 - }, - { - "epoch": 0.96, - "learning_rate": 4.370037269406879e-09, - "loss": 0.4704, - "step": 3075 - }, - { - "epoch": 0.96, - "learning_rate": 4.304663573875988e-09, - "loss": 0.4644, - "step": 3076 - }, - { - "epoch": 0.96, - "learning_rate": 4.239780436264606e-09, - "loss": 0.4751, - "step": 3077 - }, - { - "epoch": 0.96, - "learning_rate": 4.175387920783313e-09, - "loss": 0.4588, - "step": 3078 - }, - { - "epoch": 0.96, - "learning_rate": 4.111486091157301e-09, - "loss": 0.4449, - "step": 3079 - }, - { - "epoch": 0.96, - "learning_rate": 4.048075010626151e-09, - "loss": 0.4597, - "step": 3080 - }, - { - "epoch": 0.96, - "learning_rate": 3.985154741943664e-09, - "loss": 0.4786, - "step": 3081 - }, - { - "epoch": 0.96, - "learning_rate": 3.922725347378086e-09, - "loss": 0.4287, - "step": 3082 - }, - { - "epoch": 0.96, - "learning_rate": 3.860786888711664e-09, - "loss": 0.45, - "step": 3083 - }, - { - "epoch": 0.96, - "learning_rate": 3.799339427241089e-09, - "loss": 0.4249, - "step": 3084 - }, - { - "epoch": 0.96, - "learning_rate": 3.738383023776881e-09, - "loss": 0.4627, - "step": 3085 - }, - { - "epoch": 0.96, - "learning_rate": 3.6779177386436764e-09, - "loss": 0.4417, - "step": 3086 - }, - { - "epoch": 0.96, - "learning_rate": 3.617943631680109e-09, - "loss": 0.466, - "step": 3087 - }, - { - "epoch": 0.96, - "learning_rate": 3.558460762238702e-09, - "loss": 0.4542, - "step": 3088 - }, - { - "epoch": 0.96, - "learning_rate": 3.499469189185922e-09, - "loss": 0.466, - "step": 3089 - }, - { - "epoch": 0.96, - "learning_rate": 3.4409689709018474e-09, - "loss": 0.4499, - "step": 3090 - }, - { - "epoch": 0.96, - "learning_rate": 3.3829601652804464e-09, - "loss": 0.4597, - "step": 3091 - }, - { - "epoch": 0.96, - "learning_rate": 3.325442829729297e-09, - "loss": 0.4577, - "step": 3092 - }, - { - "epoch": 0.96, - "learning_rate": 3.268417021169645e-09, - "loss": 0.4732, - "step": 3093 - }, - { - "epoch": 0.96, - "learning_rate": 3.211882796036236e-09, - "loss": 0.4595, - "step": 3094 - }, - { - "epoch": 0.96, - "learning_rate": 3.1558402102774273e-09, - "loss": 0.4672, - "step": 3095 - }, - { - "epoch": 0.97, - "learning_rate": 3.100289319354965e-09, - "loss": 0.4699, - "step": 3096 - }, - { - "epoch": 0.97, - "learning_rate": 3.0452301782439847e-09, - "loss": 0.4578, - "step": 3097 - }, - { - "epoch": 0.97, - "learning_rate": 2.990662841432956e-09, - "loss": 0.4772, - "step": 3098 - }, - { - "epoch": 0.97, - "learning_rate": 2.9365873629237368e-09, - "loss": 0.4574, - "step": 3099 - }, - { - "epoch": 0.97, - "learning_rate": 2.883003796231409e-09, - "loss": 0.4416, - "step": 3100 - }, - { - "epoch": 0.97, - "learning_rate": 2.8299121943839988e-09, - "loss": 0.4558, - "step": 3101 - }, - { - "epoch": 0.97, - "learning_rate": 2.777312609923144e-09, - "loss": 0.4703, - "step": 3102 - }, - { - "epoch": 0.97, - "learning_rate": 2.725205094903038e-09, - "loss": 0.4408, - "step": 3103 - }, - { - "epoch": 0.97, - "learning_rate": 2.673589700891321e-09, - "loss": 0.4659, - "step": 3104 - }, - { - "epoch": 0.97, - "learning_rate": 2.6224664789682994e-09, - "loss": 0.4397, - "step": 3105 - }, - { - "epoch": 0.97, - "learning_rate": 2.5718354797274467e-09, - "loss": 0.4574, - "step": 3106 - }, - { - "epoch": 0.97, - "learning_rate": 2.5216967532750155e-09, - "loss": 0.4427, - "step": 3107 - }, - { - "epoch": 0.97, - "learning_rate": 2.472050349230037e-09, - "loss": 0.4472, - "step": 3108 - }, - { - "epoch": 0.97, - "learning_rate": 2.422896316724432e-09, - "loss": 0.4804, - "step": 3109 - }, - { - "epoch": 0.97, - "learning_rate": 2.374234704402789e-09, - "loss": 0.4552, - "step": 3110 - }, - { - "epoch": 0.97, - "learning_rate": 2.3260655604223635e-09, - "loss": 0.4626, - "step": 3111 - }, - { - "epoch": 0.97, - "learning_rate": 2.278388932453079e-09, - "loss": 0.4598, - "step": 3112 - }, - { - "epoch": 0.97, - "learning_rate": 2.2312048676774165e-09, - "loss": 0.456, - "step": 3113 - }, - { - "epoch": 0.97, - "learning_rate": 2.1845134127904674e-09, - "loss": 0.4474, - "step": 3114 - }, - { - "epoch": 0.97, - "learning_rate": 2.1383146139998255e-09, - "loss": 0.4673, - "step": 3115 - }, - { - "epoch": 0.97, - "learning_rate": 2.092608517025307e-09, - "loss": 0.454, - "step": 3116 - }, - { - "epoch": 0.97, - "learning_rate": 2.0473951670994527e-09, - "loss": 0.46, - "step": 3117 - }, - { - "epoch": 0.97, - "learning_rate": 2.0026746089670253e-09, - "loss": 0.4832, - "step": 3118 - }, - { - "epoch": 0.97, - "learning_rate": 1.9584468868850123e-09, - "loss": 0.4578, - "step": 3119 - }, - { - "epoch": 0.97, - "learning_rate": 1.9147120446227906e-09, - "loss": 0.4621, - "step": 3120 - }, - { - "epoch": 0.97, - "learning_rate": 1.8714701254619047e-09, - "loss": 0.4373, - "step": 3121 - }, - { - "epoch": 0.97, - "learning_rate": 1.8287211721962347e-09, - "loss": 0.4737, - "step": 3122 - }, - { - "epoch": 0.97, - "learning_rate": 1.786465227131495e-09, - "loss": 0.4438, - "step": 3123 - }, - { - "epoch": 0.97, - "learning_rate": 1.7447023320858455e-09, - "loss": 0.4442, - "step": 3124 - }, - { - "epoch": 0.97, - "learning_rate": 1.7034325283892813e-09, - "loss": 0.4528, - "step": 3125 - }, - { - "epoch": 0.97, - "learning_rate": 1.6626558568838544e-09, - "loss": 0.4632, - "step": 3126 - }, - { - "epoch": 0.97, - "learning_rate": 1.6223723579236737e-09, - "loss": 0.4739, - "step": 3127 - }, - { - "epoch": 0.98, - "learning_rate": 1.5825820713746828e-09, - "loss": 0.451, - "step": 3128 - }, - { - "epoch": 0.98, - "learning_rate": 1.543285036614883e-09, - "loss": 0.4432, - "step": 3129 - }, - { - "epoch": 0.98, - "learning_rate": 1.5044812925339434e-09, - "loss": 0.458, - "step": 3130 - }, - { - "epoch": 0.98, - "learning_rate": 1.4661708775334236e-09, - "loss": 0.4582, - "step": 3131 - }, - { - "epoch": 0.98, - "learning_rate": 1.428353829526774e-09, - "loss": 0.4649, - "step": 3132 - }, - { - "epoch": 0.98, - "learning_rate": 1.3910301859390572e-09, - "loss": 0.4555, - "step": 3133 - }, - { - "epoch": 0.98, - "learning_rate": 1.354199983707116e-09, - "loss": 0.4582, - "step": 3134 - }, - { - "epoch": 0.98, - "learning_rate": 1.3178632592794058e-09, - "loss": 0.4475, - "step": 3135 - }, - { - "epoch": 0.98, - "learning_rate": 1.2820200486160504e-09, - "loss": 0.4327, - "step": 3136 - }, - { - "epoch": 0.98, - "learning_rate": 1.2466703871888419e-09, - "loss": 0.4596, - "step": 3137 - }, - { - "epoch": 0.98, - "learning_rate": 1.211814309980963e-09, - "loss": 0.4531, - "step": 3138 - }, - { - "epoch": 0.98, - "learning_rate": 1.1774518514873765e-09, - "loss": 0.4546, - "step": 3139 - }, - { - "epoch": 0.98, - "learning_rate": 1.1435830457143248e-09, - "loss": 0.462, - "step": 3140 - }, - { - "epoch": 0.98, - "learning_rate": 1.1102079261796071e-09, - "loss": 0.4458, - "step": 3141 - }, - { - "epoch": 0.98, - "learning_rate": 1.0773265259124143e-09, - "loss": 0.4674, - "step": 3142 - }, - { - "epoch": 0.98, - "learning_rate": 1.0449388774534385e-09, - "loss": 0.4676, - "step": 3143 - }, - { - "epoch": 0.98, - "learning_rate": 1.0130450128545964e-09, - "loss": 0.4509, - "step": 3144 - }, - { - "epoch": 0.98, - "learning_rate": 9.816449636792513e-10, - "loss": 0.457, - "step": 3145 - }, - { - "epoch": 0.98, - "learning_rate": 9.507387610019345e-10, - "loss": 0.4557, - "step": 3146 - }, - { - "epoch": 0.98, - "learning_rate": 9.203264354086249e-10, - "loss": 0.4597, - "step": 3147 - }, - { - "epoch": 0.98, - "learning_rate": 8.904080169964134e-10, - "loss": 0.4807, - "step": 3148 - }, - { - "epoch": 0.98, - "learning_rate": 8.609835353736716e-10, - "loss": 0.4456, - "step": 3149 - }, - { - "epoch": 0.98, - "learning_rate": 8.320530196598841e-10, - "loss": 0.4409, - "step": 3150 - }, - { - "epoch": 0.98, - "learning_rate": 8.036164984857041e-10, - "loss": 0.4621, - "step": 3151 - }, - { - "epoch": 0.98, - "learning_rate": 7.756739999930095e-10, - "loss": 0.4467, - "step": 3152 - }, - { - "epoch": 0.98, - "learning_rate": 7.482255518346803e-10, - "loss": 0.4543, - "step": 3153 - }, - { - "epoch": 0.98, - "learning_rate": 7.212711811746542e-10, - "loss": 0.4462, - "step": 3154 - }, - { - "epoch": 0.98, - "learning_rate": 6.948109146879266e-10, - "loss": 0.45, - "step": 3155 - }, - { - "epoch": 0.98, - "learning_rate": 6.688447785606066e-10, - "loss": 0.4579, - "step": 3156 - }, - { - "epoch": 0.98, - "learning_rate": 6.433727984896941e-10, - "loss": 0.4633, - "step": 3157 - }, - { - "epoch": 0.98, - "learning_rate": 6.183949996831362e-10, - "loss": 0.4493, - "step": 3158 - }, - { - "epoch": 0.98, - "learning_rate": 5.939114068598816e-10, - "loss": 0.4484, - "step": 3159 - }, - { - "epoch": 0.99, - "learning_rate": 5.69922044249771e-10, - "loss": 0.4803, - "step": 3160 - }, - { - "epoch": 0.99, - "learning_rate": 5.464269355935914e-10, - "loss": 0.4625, - "step": 3161 - }, - { - "epoch": 0.99, - "learning_rate": 5.234261041428545e-10, - "loss": 0.4895, - "step": 3162 - }, - { - "epoch": 0.99, - "learning_rate": 5.009195726600191e-10, - "loss": 0.4766, - "step": 3163 - }, - { - "epoch": 0.99, - "learning_rate": 4.789073634184348e-10, - "loss": 0.4504, - "step": 3164 - }, - { - "epoch": 0.99, - "learning_rate": 4.5738949820212093e-10, - "loss": 0.4403, - "step": 3165 - }, - { - "epoch": 0.99, - "learning_rate": 4.36365998305932e-10, - "loss": 0.4373, - "step": 3166 - }, - { - "epoch": 0.99, - "learning_rate": 4.1583688453550315e-10, - "loss": 0.4587, - "step": 3167 - }, - { - "epoch": 0.99, - "learning_rate": 3.958021772071385e-10, - "loss": 0.4653, - "step": 3168 - }, - { - "epoch": 0.99, - "learning_rate": 3.7626189614797797e-10, - "loss": 0.4592, - "step": 3169 - }, - { - "epoch": 0.99, - "learning_rate": 3.5721606069571975e-10, - "loss": 0.461, - "step": 3170 - }, - { - "epoch": 0.99, - "learning_rate": 3.386646896988421e-10, - "loss": 0.4408, - "step": 3171 - }, - { - "epoch": 0.99, - "learning_rate": 3.206078015164371e-10, - "loss": 0.4388, - "step": 3172 - }, - { - "epoch": 0.99, - "learning_rate": 3.030454140182104e-10, - "loss": 0.4627, - "step": 3173 - }, - { - "epoch": 0.99, - "learning_rate": 2.8597754458459243e-10, - "loss": 0.4539, - "step": 3174 - }, - { - "epoch": 0.99, - "learning_rate": 2.6940421010651637e-10, - "loss": 0.4505, - "step": 3175 - }, - { - "epoch": 0.99, - "learning_rate": 2.5332542698563996e-10, - "loss": 0.4531, - "step": 3176 - }, - { - "epoch": 0.99, - "learning_rate": 2.377412111339572e-10, - "loss": 0.459, - "step": 3177 - }, - { - "epoch": 0.99, - "learning_rate": 2.2265157797429768e-10, - "loss": 0.4506, - "step": 3178 - }, - { - "epoch": 0.99, - "learning_rate": 2.080565424398273e-10, - "loss": 0.4815, - "step": 3179 - }, - { - "epoch": 0.99, - "learning_rate": 1.939561189743255e-10, - "loss": 0.4568, - "step": 3180 - }, - { - "epoch": 0.99, - "learning_rate": 1.8035032153213005e-10, - "loss": 0.4655, - "step": 3181 - }, - { - "epoch": 0.99, - "learning_rate": 1.672391635779702e-10, - "loss": 0.4607, - "step": 3182 - }, - { - "epoch": 0.99, - "learning_rate": 1.54622658087189e-10, - "loss": 0.4699, - "step": 3183 - }, - { - "epoch": 0.99, - "learning_rate": 1.4250081754546562e-10, - "loss": 0.4491, - "step": 3184 - }, - { - "epoch": 0.99, - "learning_rate": 1.3087365394898186e-10, - "loss": 0.4456, - "step": 3185 - }, - { - "epoch": 0.99, - "learning_rate": 1.1974117880453327e-10, - "loss": 0.4604, - "step": 3186 - }, - { - "epoch": 0.99, - "learning_rate": 1.0910340312908494e-10, - "loss": 0.4471, - "step": 3187 - }, - { - "epoch": 0.99, - "learning_rate": 9.896033745016019e-11, - "loss": 0.5011, - "step": 3188 - }, - { - "epoch": 0.99, - "learning_rate": 8.931199180578497e-11, - "loss": 0.4753, - "step": 3189 - }, - { - "epoch": 0.99, - "learning_rate": 8.015837574432138e-11, - "loss": 0.4585, - "step": 3190 - }, - { - "epoch": 0.99, - "learning_rate": 7.149949832441215e-11, - "loss": 0.458, - "step": 3191 - }, - { - "epoch": 1.0, - "learning_rate": 6.33353681152582e-11, - "loss": 0.4597, - "step": 3192 - }, - { - "epoch": 1.0, - "learning_rate": 5.566599319634102e-11, - "loss": 0.4628, - "step": 3193 - }, - { - "epoch": 1.0, - "learning_rate": 4.849138115764484e-11, - "loss": 0.4553, - "step": 3194 - }, - { - "epoch": 1.0, - "learning_rate": 4.1811539099323446e-11, - "loss": 0.4591, - "step": 3195 - }, - { - "epoch": 1.0, - "learning_rate": 3.562647363197779e-11, - "loss": 0.4319, - "step": 3196 - }, - { - "epoch": 1.0, - "learning_rate": 2.9936190876656e-11, - "loss": 0.4392, - "step": 3197 - }, - { - "epoch": 1.0, - "learning_rate": 2.4740696464631284e-11, - "loss": 0.4838, - "step": 3198 - }, - { - "epoch": 1.0, - "learning_rate": 2.003999553751301e-11, - "loss": 0.4418, - "step": 3199 - }, - { - "epoch": 1.0, - "learning_rate": 1.5834092747302187e-11, - "loss": 0.4698, - "step": 3200 - }, - { - "epoch": 1.0, - "learning_rate": 1.2122992256335951e-11, - "loss": 0.4601, - "step": 3201 - }, - { - "epoch": 1.0, - "learning_rate": 8.906697737287583e-12, - "loss": 0.4582, - "step": 3202 - }, - { - "epoch": 1.0, - "learning_rate": 6.1852123729999595e-12, - "loss": 0.4434, - "step": 3203 - }, - { - "epoch": 1.0, - "learning_rate": 3.958538856874138e-12, - "loss": 0.4634, - "step": 3204 - }, - { - "epoch": 1.0, - "learning_rate": 2.226679392425268e-12, - "loss": 0.4695, - "step": 3205 - }, - { - "epoch": 1.0, - "learning_rate": 9.896356936156536e-13, - "loss": 0.4614, - "step": 3206 - }, - { - "epoch": 1.0, - "learning_rate": 2.474089846327132e-13, - "loss": 0.4417, - "step": 3207 - }, - { - "epoch": 1.0, - "learning_rate": 0.0, - "loss": 0.4459, - "step": 3208 - }, - { - "epoch": 1.0, - "step": 3208, - "total_flos": 6.708198816698335e+18, - "train_loss": 0.5434732109568363, - "train_runtime": 65335.2024, - "train_samples_per_second": 175.936, - "train_steps_per_second": 0.049 - } - ], - "logging_steps": 1.0, - "max_steps": 3208, - "num_train_epochs": 1, - "save_steps": 2000, - "total_flos": 6.708198816698335e+18, - "trial_name": null, - "trial_params": null -}