diff --git "a/checkpoint-45820/trainer_state.json" "b/checkpoint-45820/trainer_state.json" new file mode 100644--- /dev/null +++ "b/checkpoint-45820/trainer_state.json" @@ -0,0 +1,275026 @@ +{ + "best_metric": 1.1250884532928467, + "best_model_checkpoint": "bird_sounds_classification/checkpoint-45820", + "epoch": 10.0, + "global_step": 45820, + "is_hyper_param_search": false, + "is_local_process_zero": true, + "is_world_process_zero": true, + "log_history": [ + { + "epoch": 0.0, + "learning_rate": 1.0000000000000001e-07, + "loss": 3.9271, + "step": 1 + }, + { + "epoch": 0.0, + "learning_rate": 2.0000000000000002e-07, + "loss": 3.9014, + "step": 2 + }, + { + "epoch": 0.0, + "learning_rate": 3.0000000000000004e-07, + "loss": 3.9174, + "step": 3 + }, + { + "epoch": 0.0, + "learning_rate": 4.0000000000000003e-07, + "loss": 3.9122, + "step": 4 + }, + { + "epoch": 0.0, + "learning_rate": 5.000000000000001e-07, + "loss": 3.926, + "step": 5 + }, + { + "epoch": 0.0, + "learning_rate": 6.000000000000001e-07, + "loss": 3.9155, + "step": 6 + }, + { + "epoch": 0.0, + "learning_rate": 7.000000000000001e-07, + "loss": 3.9036, + "step": 7 + }, + { + "epoch": 0.0, + "learning_rate": 8.000000000000001e-07, + "loss": 3.9038, + "step": 8 + }, + { + "epoch": 0.0, + "learning_rate": 9.000000000000001e-07, + "loss": 3.9065, + "step": 9 + }, + { + "epoch": 0.0, + "learning_rate": 1.0000000000000002e-06, + "loss": 3.9121, + "step": 10 + }, + { + "epoch": 0.0, + "learning_rate": 1.1e-06, + "loss": 3.8959, + "step": 11 + }, + { + "epoch": 0.0, + "learning_rate": 1.2000000000000002e-06, + "loss": 3.9163, + "step": 12 + }, + { + "epoch": 0.0, + "learning_rate": 1.3e-06, + "loss": 3.92, + "step": 13 + }, + { + "epoch": 0.0, + "learning_rate": 1.4000000000000001e-06, + "loss": 3.9013, + "step": 14 + }, + { + "epoch": 0.0, + "learning_rate": 1.5e-06, + "loss": 3.9173, + "step": 15 + }, + { + "epoch": 0.0, + "learning_rate": 1.6000000000000001e-06, + "loss": 3.889, + "step": 16 + }, + { + "epoch": 0.0, + "learning_rate": 1.7000000000000002e-06, + "loss": 4.0255, + "step": 17 + }, + { + "epoch": 0.0, + "learning_rate": 1.8000000000000001e-06, + "loss": 3.9129, + "step": 18 + }, + { + "epoch": 0.0, + "learning_rate": 1.9000000000000002e-06, + "loss": 4.0031, + "step": 19 + }, + { + "epoch": 0.0, + "learning_rate": 2.0000000000000003e-06, + "loss": 3.9127, + "step": 20 + }, + { + "epoch": 0.0, + "learning_rate": 2.1000000000000002e-06, + "loss": 3.9188, + "step": 21 + }, + { + "epoch": 0.0, + "learning_rate": 2.2e-06, + "loss": 3.9234, + "step": 22 + }, + { + "epoch": 0.01, + "learning_rate": 2.3000000000000004e-06, + "loss": 3.9108, + "step": 23 + }, + { + "epoch": 0.01, + "learning_rate": 2.4000000000000003e-06, + "loss": 3.9206, + "step": 24 + }, + { + "epoch": 0.01, + "learning_rate": 2.5e-06, + "loss": 3.9094, + "step": 25 + }, + { + "epoch": 0.01, + "learning_rate": 2.6e-06, + "loss": 3.9151, + "step": 26 + }, + { + "epoch": 0.01, + "learning_rate": 2.7000000000000004e-06, + "loss": 3.9078, + "step": 27 + }, + { + "epoch": 0.01, + "learning_rate": 2.8000000000000003e-06, + "loss": 3.8972, + "step": 28 + }, + { + "epoch": 0.01, + "learning_rate": 2.9e-06, + "loss": 3.9182, + "step": 29 + }, + { + "epoch": 0.01, + "learning_rate": 3e-06, + "loss": 3.9191, + "step": 30 + }, + { + "epoch": 0.01, + "learning_rate": 3.1000000000000004e-06, + "loss": 3.9135, + "step": 31 + }, + { + "epoch": 0.01, + "learning_rate": 3.2000000000000003e-06, + "loss": 3.9081, + "step": 32 + }, + { + "epoch": 0.01, + "learning_rate": 3.3000000000000006e-06, + "loss": 3.9081, + "step": 33 + }, + { + "epoch": 0.01, + "learning_rate": 3.4000000000000005e-06, + "loss": 3.9161, + "step": 34 + }, + { + "epoch": 0.01, + "learning_rate": 3.5e-06, + "loss": 3.907, + "step": 35 + }, + { + "epoch": 0.01, + "learning_rate": 3.6000000000000003e-06, + "loss": 3.9181, + "step": 36 + }, + { + "epoch": 0.01, + "learning_rate": 3.7e-06, + "loss": 3.8976, + "step": 37 + }, + { + "epoch": 0.01, + "learning_rate": 3.8000000000000005e-06, + "loss": 3.9186, + "step": 38 + }, + { + "epoch": 0.01, + "learning_rate": 3.900000000000001e-06, + "loss": 3.9134, + "step": 39 + }, + { + "epoch": 0.01, + "learning_rate": 4.000000000000001e-06, + "loss": 3.9043, + "step": 40 + }, + { + "epoch": 0.01, + "learning_rate": 4.1e-06, + "loss": 3.9145, + "step": 41 + }, + { + "epoch": 0.01, + "learning_rate": 4.2000000000000004e-06, + "loss": 3.9696, + "step": 42 + }, + { + "epoch": 0.01, + "learning_rate": 4.3e-06, + "loss": 3.9018, + "step": 43 + }, + { + "epoch": 0.01, + "learning_rate": 4.4e-06, + "loss": 3.8972, + "step": 44 + }, + { + "epoch": 0.01, + "learning_rate": 4.5e-06, + "loss": 3.9046, + "step": 45 + }, + { + "epoch": 0.01, + "learning_rate": 4.600000000000001e-06, + "loss": 3.9193, + "step": 46 + }, + { + "epoch": 0.01, + "learning_rate": 4.7e-06, + "loss": 3.9209, + "step": 47 + }, + { + "epoch": 0.01, + "learning_rate": 4.800000000000001e-06, + "loss": 3.9045, + "step": 48 + }, + { + "epoch": 0.01, + "learning_rate": 4.9000000000000005e-06, + "loss": 3.9313, + "step": 49 + }, + { + "epoch": 0.01, + "learning_rate": 5e-06, + "loss": 3.9337, + "step": 50 + }, + { + "epoch": 0.01, + "learning_rate": 4.9998907581385194e-06, + "loss": 3.9053, + "step": 51 + }, + { + "epoch": 0.01, + "learning_rate": 4.999781516277038e-06, + "loss": 3.9205, + "step": 52 + }, + { + "epoch": 0.01, + "learning_rate": 4.999672274415557e-06, + "loss": 3.9054, + "step": 53 + }, + { + "epoch": 0.01, + "learning_rate": 4.999563032554075e-06, + "loss": 3.9218, + "step": 54 + }, + { + "epoch": 0.01, + "learning_rate": 4.999453790692594e-06, + "loss": 3.8872, + "step": 55 + }, + { + "epoch": 0.01, + "learning_rate": 4.999344548831113e-06, + "loss": 3.9214, + "step": 56 + }, + { + "epoch": 0.01, + "learning_rate": 4.999235306969631e-06, + "loss": 3.9098, + "step": 57 + }, + { + "epoch": 0.01, + "learning_rate": 4.999126065108149e-06, + "loss": 3.9025, + "step": 58 + }, + { + "epoch": 0.01, + "learning_rate": 4.999016823246668e-06, + "loss": 3.9157, + "step": 59 + }, + { + "epoch": 0.01, + "learning_rate": 4.998907581385187e-06, + "loss": 3.9133, + "step": 60 + }, + { + "epoch": 0.01, + "learning_rate": 4.998798339523706e-06, + "loss": 3.9179, + "step": 61 + }, + { + "epoch": 0.01, + "learning_rate": 4.998689097662225e-06, + "loss": 3.9255, + "step": 62 + }, + { + "epoch": 0.01, + "learning_rate": 4.998579855800743e-06, + "loss": 3.9156, + "step": 63 + }, + { + "epoch": 0.01, + "learning_rate": 4.998470613939262e-06, + "loss": 3.9097, + "step": 64 + }, + { + "epoch": 0.01, + "learning_rate": 4.998361372077781e-06, + "loss": 3.9163, + "step": 65 + }, + { + "epoch": 0.01, + "learning_rate": 4.9982521302163e-06, + "loss": 3.8202, + "step": 66 + }, + { + "epoch": 0.01, + "learning_rate": 4.998142888354818e-06, + "loss": 3.8851, + "step": 67 + }, + { + "epoch": 0.01, + "learning_rate": 4.998033646493336e-06, + "loss": 3.8884, + "step": 68 + }, + { + "epoch": 0.02, + "learning_rate": 4.997924404631855e-06, + "loss": 3.9022, + "step": 69 + }, + { + "epoch": 0.02, + "learning_rate": 4.997815162770374e-06, + "loss": 3.9226, + "step": 70 + }, + { + "epoch": 0.02, + "learning_rate": 4.997705920908893e-06, + "loss": 3.9096, + "step": 71 + }, + { + "epoch": 0.02, + "learning_rate": 4.9975966790474115e-06, + "loss": 3.8879, + "step": 72 + }, + { + "epoch": 0.02, + "learning_rate": 4.99748743718593e-06, + "loss": 3.9084, + "step": 73 + }, + { + "epoch": 0.02, + "learning_rate": 4.997378195324449e-06, + "loss": 3.8948, + "step": 74 + }, + { + "epoch": 0.02, + "learning_rate": 4.997268953462968e-06, + "loss": 3.9101, + "step": 75 + }, + { + "epoch": 0.02, + "learning_rate": 4.997159711601486e-06, + "loss": 3.9441, + "step": 76 + }, + { + "epoch": 0.02, + "learning_rate": 4.997050469740005e-06, + "loss": 3.9194, + "step": 77 + }, + { + "epoch": 0.02, + "learning_rate": 4.996941227878523e-06, + "loss": 3.8928, + "step": 78 + }, + { + "epoch": 0.02, + "learning_rate": 4.996831986017042e-06, + "loss": 3.8979, + "step": 79 + }, + { + "epoch": 0.02, + "learning_rate": 4.996722744155561e-06, + "loss": 3.9075, + "step": 80 + }, + { + "epoch": 0.02, + "learning_rate": 4.9966135022940795e-06, + "loss": 3.9035, + "step": 81 + }, + { + "epoch": 0.02, + "learning_rate": 4.996504260432598e-06, + "loss": 3.8888, + "step": 82 + }, + { + "epoch": 0.02, + "learning_rate": 4.996395018571117e-06, + "loss": 3.8978, + "step": 83 + }, + { + "epoch": 0.02, + "learning_rate": 4.996285776709636e-06, + "loss": 3.8975, + "step": 84 + }, + { + "epoch": 0.02, + "learning_rate": 4.996176534848155e-06, + "loss": 3.9036, + "step": 85 + }, + { + "epoch": 0.02, + "learning_rate": 4.996067292986673e-06, + "loss": 3.9115, + "step": 86 + }, + { + "epoch": 0.02, + "learning_rate": 4.995958051125191e-06, + "loss": 3.8913, + "step": 87 + }, + { + "epoch": 0.02, + "learning_rate": 4.99584880926371e-06, + "loss": 3.8857, + "step": 88 + }, + { + "epoch": 0.02, + "learning_rate": 4.995739567402229e-06, + "loss": 3.9053, + "step": 89 + }, + { + "epoch": 0.02, + "learning_rate": 4.995630325540748e-06, + "loss": 3.9077, + "step": 90 + }, + { + "epoch": 0.02, + "learning_rate": 4.9955210836792664e-06, + "loss": 3.928, + "step": 91 + }, + { + "epoch": 0.02, + "learning_rate": 4.995411841817785e-06, + "loss": 3.9188, + "step": 92 + }, + { + "epoch": 0.02, + "learning_rate": 4.995302599956304e-06, + "loss": 3.9185, + "step": 93 + }, + { + "epoch": 0.02, + "learning_rate": 4.995193358094823e-06, + "loss": 3.8987, + "step": 94 + }, + { + "epoch": 0.02, + "learning_rate": 4.995084116233341e-06, + "loss": 3.9226, + "step": 95 + }, + { + "epoch": 0.02, + "learning_rate": 4.99497487437186e-06, + "loss": 3.8836, + "step": 96 + }, + { + "epoch": 0.02, + "learning_rate": 4.994865632510378e-06, + "loss": 3.9074, + "step": 97 + }, + { + "epoch": 0.02, + "learning_rate": 4.994756390648897e-06, + "loss": 3.9178, + "step": 98 + }, + { + "epoch": 0.02, + "learning_rate": 4.994647148787416e-06, + "loss": 3.9016, + "step": 99 + }, + { + "epoch": 0.02, + "learning_rate": 4.994537906925934e-06, + "loss": 3.9274, + "step": 100 + }, + { + "epoch": 0.02, + "learning_rate": 4.994428665064453e-06, + "loss": 3.8926, + "step": 101 + }, + { + "epoch": 0.02, + "learning_rate": 4.9943194232029716e-06, + "loss": 3.9168, + "step": 102 + }, + { + "epoch": 0.02, + "learning_rate": 4.994210181341491e-06, + "loss": 3.9134, + "step": 103 + }, + { + "epoch": 0.02, + "learning_rate": 4.99410093948001e-06, + "loss": 3.9158, + "step": 104 + }, + { + "epoch": 0.02, + "learning_rate": 4.993991697618528e-06, + "loss": 3.924, + "step": 105 + }, + { + "epoch": 0.02, + "learning_rate": 4.993882455757047e-06, + "loss": 3.9193, + "step": 106 + }, + { + "epoch": 0.02, + "learning_rate": 4.993773213895565e-06, + "loss": 3.9231, + "step": 107 + }, + { + "epoch": 0.02, + "learning_rate": 4.993663972034084e-06, + "loss": 3.8917, + "step": 108 + }, + { + "epoch": 0.02, + "learning_rate": 4.993554730172602e-06, + "loss": 3.9131, + "step": 109 + }, + { + "epoch": 0.02, + "learning_rate": 4.993445488311121e-06, + "loss": 3.8954, + "step": 110 + }, + { + "epoch": 0.02, + "learning_rate": 4.9933362464496395e-06, + "loss": 3.9034, + "step": 111 + }, + { + "epoch": 0.02, + "learning_rate": 4.9932270045881585e-06, + "loss": 3.9162, + "step": 112 + }, + { + "epoch": 0.02, + "learning_rate": 4.9931177627266776e-06, + "loss": 3.88, + "step": 113 + }, + { + "epoch": 0.02, + "learning_rate": 4.993008520865196e-06, + "loss": 3.8891, + "step": 114 + }, + { + "epoch": 0.03, + "learning_rate": 4.992899279003715e-06, + "loss": 3.8806, + "step": 115 + }, + { + "epoch": 0.03, + "learning_rate": 4.992790037142233e-06, + "loss": 3.9014, + "step": 116 + }, + { + "epoch": 0.03, + "learning_rate": 4.992680795280752e-06, + "loss": 3.8544, + "step": 117 + }, + { + "epoch": 0.03, + "learning_rate": 4.992571553419271e-06, + "loss": 3.9015, + "step": 118 + }, + { + "epoch": 0.03, + "learning_rate": 4.992462311557789e-06, + "loss": 3.9053, + "step": 119 + }, + { + "epoch": 0.03, + "learning_rate": 4.992353069696308e-06, + "loss": 3.8913, + "step": 120 + }, + { + "epoch": 0.03, + "learning_rate": 4.9922438278348264e-06, + "loss": 3.8953, + "step": 121 + }, + { + "epoch": 0.03, + "learning_rate": 4.9921345859733455e-06, + "loss": 3.8766, + "step": 122 + }, + { + "epoch": 0.03, + "learning_rate": 4.9920253441118645e-06, + "loss": 3.9048, + "step": 123 + }, + { + "epoch": 0.03, + "learning_rate": 4.991916102250383e-06, + "loss": 3.8831, + "step": 124 + }, + { + "epoch": 0.03, + "learning_rate": 4.991806860388902e-06, + "loss": 3.9077, + "step": 125 + }, + { + "epoch": 0.03, + "learning_rate": 4.99169761852742e-06, + "loss": 3.9036, + "step": 126 + }, + { + "epoch": 0.03, + "learning_rate": 4.991588376665939e-06, + "loss": 3.9077, + "step": 127 + }, + { + "epoch": 0.03, + "learning_rate": 4.991479134804457e-06, + "loss": 3.8896, + "step": 128 + }, + { + "epoch": 0.03, + "learning_rate": 4.991369892942976e-06, + "loss": 3.8895, + "step": 129 + }, + { + "epoch": 0.03, + "learning_rate": 4.991260651081495e-06, + "loss": 3.8813, + "step": 130 + }, + { + "epoch": 0.03, + "learning_rate": 4.991151409220013e-06, + "loss": 3.923, + "step": 131 + }, + { + "epoch": 0.03, + "learning_rate": 4.9910421673585324e-06, + "loss": 3.9078, + "step": 132 + }, + { + "epoch": 0.03, + "learning_rate": 4.990932925497051e-06, + "loss": 3.9173, + "step": 133 + }, + { + "epoch": 0.03, + "learning_rate": 4.99082368363557e-06, + "loss": 3.9016, + "step": 134 + }, + { + "epoch": 0.03, + "learning_rate": 4.990714441774088e-06, + "loss": 3.9235, + "step": 135 + }, + { + "epoch": 0.03, + "learning_rate": 4.990605199912607e-06, + "loss": 3.9182, + "step": 136 + }, + { + "epoch": 0.03, + "learning_rate": 4.990495958051126e-06, + "loss": 3.8883, + "step": 137 + }, + { + "epoch": 0.03, + "learning_rate": 4.990386716189644e-06, + "loss": 3.9019, + "step": 138 + }, + { + "epoch": 0.03, + "learning_rate": 4.990277474328163e-06, + "loss": 3.8711, + "step": 139 + }, + { + "epoch": 0.03, + "learning_rate": 4.990168232466681e-06, + "loss": 3.9007, + "step": 140 + }, + { + "epoch": 0.03, + "learning_rate": 4.9900589906052e-06, + "loss": 3.8936, + "step": 141 + }, + { + "epoch": 0.03, + "learning_rate": 4.9899497487437185e-06, + "loss": 3.9037, + "step": 142 + }, + { + "epoch": 0.03, + "learning_rate": 4.989840506882238e-06, + "loss": 3.8914, + "step": 143 + }, + { + "epoch": 0.03, + "learning_rate": 4.989731265020757e-06, + "loss": 3.8986, + "step": 144 + }, + { + "epoch": 0.03, + "learning_rate": 4.989622023159275e-06, + "loss": 3.9842, + "step": 145 + }, + { + "epoch": 0.03, + "learning_rate": 4.989512781297794e-06, + "loss": 3.8875, + "step": 146 + }, + { + "epoch": 0.03, + "learning_rate": 4.989403539436312e-06, + "loss": 3.8769, + "step": 147 + }, + { + "epoch": 0.03, + "learning_rate": 4.989294297574831e-06, + "loss": 3.8379, + "step": 148 + }, + { + "epoch": 0.03, + "learning_rate": 4.98918505571335e-06, + "loss": 3.9105, + "step": 149 + }, + { + "epoch": 0.03, + "learning_rate": 4.989075813851868e-06, + "loss": 3.8692, + "step": 150 + }, + { + "epoch": 0.03, + "learning_rate": 4.988966571990387e-06, + "loss": 3.9108, + "step": 151 + }, + { + "epoch": 0.03, + "learning_rate": 4.9888573301289055e-06, + "loss": 3.9173, + "step": 152 + }, + { + "epoch": 0.03, + "learning_rate": 4.9887480882674245e-06, + "loss": 3.9106, + "step": 153 + }, + { + "epoch": 0.03, + "learning_rate": 4.9886388464059436e-06, + "loss": 3.884, + "step": 154 + }, + { + "epoch": 0.03, + "learning_rate": 4.988529604544462e-06, + "loss": 3.813, + "step": 155 + }, + { + "epoch": 0.03, + "learning_rate": 4.988420362682981e-06, + "loss": 3.8905, + "step": 156 + }, + { + "epoch": 0.03, + "learning_rate": 4.988311120821499e-06, + "loss": 3.9261, + "step": 157 + }, + { + "epoch": 0.03, + "learning_rate": 4.988201878960018e-06, + "loss": 3.8742, + "step": 158 + }, + { + "epoch": 0.03, + "learning_rate": 4.988092637098537e-06, + "loss": 3.8807, + "step": 159 + }, + { + "epoch": 0.03, + "learning_rate": 4.987983395237055e-06, + "loss": 3.8504, + "step": 160 + }, + { + "epoch": 0.04, + "learning_rate": 4.9878741533755734e-06, + "loss": 3.9277, + "step": 161 + }, + { + "epoch": 0.04, + "learning_rate": 4.9877649115140925e-06, + "loss": 3.8757, + "step": 162 + }, + { + "epoch": 0.04, + "learning_rate": 4.9876556696526115e-06, + "loss": 3.8657, + "step": 163 + }, + { + "epoch": 0.04, + "learning_rate": 4.98754642779113e-06, + "loss": 3.9214, + "step": 164 + }, + { + "epoch": 0.04, + "learning_rate": 4.987437185929649e-06, + "loss": 3.8738, + "step": 165 + }, + { + "epoch": 0.04, + "learning_rate": 4.987327944068167e-06, + "loss": 3.9489, + "step": 166 + }, + { + "epoch": 0.04, + "learning_rate": 4.987218702206686e-06, + "loss": 3.9121, + "step": 167 + }, + { + "epoch": 0.04, + "learning_rate": 4.987109460345205e-06, + "loss": 3.9072, + "step": 168 + }, + { + "epoch": 0.04, + "learning_rate": 4.987000218483723e-06, + "loss": 3.9235, + "step": 169 + }, + { + "epoch": 0.04, + "learning_rate": 4.986890976622242e-06, + "loss": 3.9071, + "step": 170 + }, + { + "epoch": 0.04, + "learning_rate": 4.98678173476076e-06, + "loss": 3.8641, + "step": 171 + }, + { + "epoch": 0.04, + "learning_rate": 4.986672492899279e-06, + "loss": 3.9307, + "step": 172 + }, + { + "epoch": 0.04, + "learning_rate": 4.9865632510377985e-06, + "loss": 3.8349, + "step": 173 + }, + { + "epoch": 0.04, + "learning_rate": 4.986454009176317e-06, + "loss": 3.8962, + "step": 174 + }, + { + "epoch": 0.04, + "learning_rate": 4.986344767314835e-06, + "loss": 3.9341, + "step": 175 + }, + { + "epoch": 0.04, + "learning_rate": 4.986235525453354e-06, + "loss": 3.908, + "step": 176 + }, + { + "epoch": 0.04, + "learning_rate": 4.986126283591873e-06, + "loss": 3.9195, + "step": 177 + }, + { + "epoch": 0.04, + "learning_rate": 4.986017041730392e-06, + "loss": 3.8767, + "step": 178 + }, + { + "epoch": 0.04, + "learning_rate": 4.98590779986891e-06, + "loss": 3.9371, + "step": 179 + }, + { + "epoch": 0.04, + "learning_rate": 4.985798558007428e-06, + "loss": 3.8691, + "step": 180 + }, + { + "epoch": 0.04, + "learning_rate": 4.985689316145947e-06, + "loss": 3.9115, + "step": 181 + }, + { + "epoch": 0.04, + "learning_rate": 4.985580074284466e-06, + "loss": 3.9081, + "step": 182 + }, + { + "epoch": 0.04, + "learning_rate": 4.985470832422985e-06, + "loss": 3.913, + "step": 183 + }, + { + "epoch": 0.04, + "learning_rate": 4.985361590561504e-06, + "loss": 3.881, + "step": 184 + }, + { + "epoch": 0.04, + "learning_rate": 4.985252348700022e-06, + "loss": 3.8632, + "step": 185 + }, + { + "epoch": 0.04, + "learning_rate": 4.985143106838541e-06, + "loss": 3.8501, + "step": 186 + }, + { + "epoch": 0.04, + "learning_rate": 4.98503386497706e-06, + "loss": 3.9037, + "step": 187 + }, + { + "epoch": 0.04, + "learning_rate": 4.984924623115578e-06, + "loss": 3.8822, + "step": 188 + }, + { + "epoch": 0.04, + "learning_rate": 4.984815381254097e-06, + "loss": 3.885, + "step": 189 + }, + { + "epoch": 0.04, + "learning_rate": 4.984706139392615e-06, + "loss": 3.8575, + "step": 190 + }, + { + "epoch": 0.04, + "learning_rate": 4.984596897531134e-06, + "loss": 3.8944, + "step": 191 + }, + { + "epoch": 0.04, + "learning_rate": 4.984487655669653e-06, + "loss": 3.899, + "step": 192 + }, + { + "epoch": 0.04, + "learning_rate": 4.9843784138081715e-06, + "loss": 3.9157, + "step": 193 + }, + { + "epoch": 0.04, + "learning_rate": 4.9842691719466906e-06, + "loss": 3.9182, + "step": 194 + }, + { + "epoch": 0.04, + "learning_rate": 4.984159930085209e-06, + "loss": 3.8652, + "step": 195 + }, + { + "epoch": 0.04, + "learning_rate": 4.984050688223728e-06, + "loss": 3.8929, + "step": 196 + }, + { + "epoch": 0.04, + "learning_rate": 4.983941446362247e-06, + "loss": 3.878, + "step": 197 + }, + { + "epoch": 0.04, + "learning_rate": 4.983832204500765e-06, + "loss": 3.8796, + "step": 198 + }, + { + "epoch": 0.04, + "learning_rate": 4.983722962639283e-06, + "loss": 3.9458, + "step": 199 + }, + { + "epoch": 0.04, + "learning_rate": 4.983613720777802e-06, + "loss": 3.9394, + "step": 200 + }, + { + "epoch": 0.04, + "learning_rate": 4.983504478916321e-06, + "loss": 3.8487, + "step": 201 + }, + { + "epoch": 0.04, + "learning_rate": 4.98339523705484e-06, + "loss": 3.8652, + "step": 202 + }, + { + "epoch": 0.04, + "learning_rate": 4.9832859951933585e-06, + "loss": 3.8027, + "step": 203 + }, + { + "epoch": 0.04, + "learning_rate": 4.983176753331877e-06, + "loss": 3.8715, + "step": 204 + }, + { + "epoch": 0.04, + "learning_rate": 4.983067511470396e-06, + "loss": 3.8716, + "step": 205 + }, + { + "epoch": 0.04, + "learning_rate": 4.982958269608915e-06, + "loss": 3.9324, + "step": 206 + }, + { + "epoch": 0.05, + "learning_rate": 4.982849027747434e-06, + "loss": 3.9429, + "step": 207 + }, + { + "epoch": 0.05, + "learning_rate": 4.982739785885952e-06, + "loss": 3.9214, + "step": 208 + }, + { + "epoch": 0.05, + "learning_rate": 4.98263054402447e-06, + "loss": 3.8385, + "step": 209 + }, + { + "epoch": 0.05, + "learning_rate": 4.982521302162989e-06, + "loss": 3.9021, + "step": 210 + }, + { + "epoch": 0.05, + "learning_rate": 4.982412060301508e-06, + "loss": 3.9145, + "step": 211 + }, + { + "epoch": 0.05, + "learning_rate": 4.982302818440027e-06, + "loss": 3.9283, + "step": 212 + }, + { + "epoch": 0.05, + "learning_rate": 4.9821935765785454e-06, + "loss": 3.8882, + "step": 213 + }, + { + "epoch": 0.05, + "learning_rate": 4.982084334717064e-06, + "loss": 3.9427, + "step": 214 + }, + { + "epoch": 0.05, + "learning_rate": 4.981975092855583e-06, + "loss": 3.9169, + "step": 215 + }, + { + "epoch": 0.05, + "learning_rate": 4.981865850994102e-06, + "loss": 3.8722, + "step": 216 + }, + { + "epoch": 0.05, + "learning_rate": 4.98175660913262e-06, + "loss": 3.8373, + "step": 217 + }, + { + "epoch": 0.05, + "learning_rate": 4.981647367271139e-06, + "loss": 3.9287, + "step": 218 + }, + { + "epoch": 0.05, + "learning_rate": 4.981538125409657e-06, + "loss": 3.8213, + "step": 219 + }, + { + "epoch": 0.05, + "learning_rate": 4.981428883548176e-06, + "loss": 3.828, + "step": 220 + }, + { + "epoch": 0.05, + "learning_rate": 4.981319641686695e-06, + "loss": 3.9053, + "step": 221 + }, + { + "epoch": 0.05, + "learning_rate": 4.981210399825213e-06, + "loss": 3.9053, + "step": 222 + }, + { + "epoch": 0.05, + "learning_rate": 4.9811011579637315e-06, + "loss": 3.9271, + "step": 223 + }, + { + "epoch": 0.05, + "learning_rate": 4.980991916102251e-06, + "loss": 3.8761, + "step": 224 + }, + { + "epoch": 0.05, + "learning_rate": 4.98088267424077e-06, + "loss": 3.8784, + "step": 225 + }, + { + "epoch": 0.05, + "learning_rate": 4.980773432379289e-06, + "loss": 3.7882, + "step": 226 + }, + { + "epoch": 0.05, + "learning_rate": 4.980664190517807e-06, + "loss": 3.8748, + "step": 227 + }, + { + "epoch": 0.05, + "learning_rate": 4.980554948656325e-06, + "loss": 3.9188, + "step": 228 + }, + { + "epoch": 0.05, + "learning_rate": 4.980445706794844e-06, + "loss": 3.8851, + "step": 229 + }, + { + "epoch": 0.05, + "learning_rate": 4.980336464933363e-06, + "loss": 3.8799, + "step": 230 + }, + { + "epoch": 0.05, + "learning_rate": 4.980227223071882e-06, + "loss": 3.8676, + "step": 231 + }, + { + "epoch": 0.05, + "learning_rate": 4.9801179812104e-06, + "loss": 3.9049, + "step": 232 + }, + { + "epoch": 0.05, + "learning_rate": 4.9800087393489185e-06, + "loss": 3.8414, + "step": 233 + }, + { + "epoch": 0.05, + "learning_rate": 4.9798994974874375e-06, + "loss": 3.8065, + "step": 234 + }, + { + "epoch": 0.05, + "learning_rate": 4.9797902556259566e-06, + "loss": 3.9521, + "step": 235 + }, + { + "epoch": 0.05, + "learning_rate": 4.979681013764476e-06, + "loss": 3.898, + "step": 236 + }, + { + "epoch": 0.05, + "learning_rate": 4.979571771902994e-06, + "loss": 3.9184, + "step": 237 + }, + { + "epoch": 0.05, + "learning_rate": 4.979462530041512e-06, + "loss": 3.8848, + "step": 238 + }, + { + "epoch": 0.05, + "learning_rate": 4.979353288180031e-06, + "loss": 3.9581, + "step": 239 + }, + { + "epoch": 0.05, + "learning_rate": 4.97924404631855e-06, + "loss": 3.7667, + "step": 240 + }, + { + "epoch": 0.05, + "learning_rate": 4.979134804457068e-06, + "loss": 3.959, + "step": 241 + }, + { + "epoch": 0.05, + "learning_rate": 4.979025562595587e-06, + "loss": 3.8652, + "step": 242 + }, + { + "epoch": 0.05, + "learning_rate": 4.9789163207341055e-06, + "loss": 3.9067, + "step": 243 + }, + { + "epoch": 0.05, + "learning_rate": 4.9788070788726245e-06, + "loss": 3.9441, + "step": 244 + }, + { + "epoch": 0.05, + "learning_rate": 4.9786978370111435e-06, + "loss": 3.8516, + "step": 245 + }, + { + "epoch": 0.05, + "learning_rate": 4.978588595149662e-06, + "loss": 3.8068, + "step": 246 + }, + { + "epoch": 0.05, + "learning_rate": 4.978479353288181e-06, + "loss": 3.94, + "step": 247 + }, + { + "epoch": 0.05, + "learning_rate": 4.978370111426699e-06, + "loss": 3.8879, + "step": 248 + }, + { + "epoch": 0.05, + "learning_rate": 4.978260869565218e-06, + "loss": 3.9437, + "step": 249 + }, + { + "epoch": 0.05, + "learning_rate": 4.978151627703737e-06, + "loss": 3.8733, + "step": 250 + }, + { + "epoch": 0.05, + "learning_rate": 4.978042385842255e-06, + "loss": 3.924, + "step": 251 + }, + { + "epoch": 0.05, + "learning_rate": 4.977933143980773e-06, + "loss": 3.9021, + "step": 252 + }, + { + "epoch": 0.06, + "learning_rate": 4.977823902119292e-06, + "loss": 3.8216, + "step": 253 + }, + { + "epoch": 0.06, + "learning_rate": 4.9777146602578115e-06, + "loss": 3.8705, + "step": 254 + }, + { + "epoch": 0.06, + "learning_rate": 4.9776054183963305e-06, + "loss": 3.7649, + "step": 255 + }, + { + "epoch": 0.06, + "learning_rate": 4.977496176534849e-06, + "loss": 3.8719, + "step": 256 + }, + { + "epoch": 0.06, + "learning_rate": 4.977386934673367e-06, + "loss": 3.8791, + "step": 257 + }, + { + "epoch": 0.06, + "learning_rate": 4.977277692811886e-06, + "loss": 3.9098, + "step": 258 + }, + { + "epoch": 0.06, + "learning_rate": 4.977168450950405e-06, + "loss": 3.9263, + "step": 259 + }, + { + "epoch": 0.06, + "learning_rate": 4.977059209088923e-06, + "loss": 3.8973, + "step": 260 + }, + { + "epoch": 0.06, + "learning_rate": 4.976949967227442e-06, + "loss": 3.8097, + "step": 261 + }, + { + "epoch": 0.06, + "learning_rate": 4.97684072536596e-06, + "loss": 3.8889, + "step": 262 + }, + { + "epoch": 0.06, + "learning_rate": 4.976731483504479e-06, + "loss": 3.9102, + "step": 263 + }, + { + "epoch": 0.06, + "learning_rate": 4.976622241642998e-06, + "loss": 3.7812, + "step": 264 + }, + { + "epoch": 0.06, + "learning_rate": 4.976512999781517e-06, + "loss": 3.8757, + "step": 265 + }, + { + "epoch": 0.06, + "learning_rate": 4.976403757920036e-06, + "loss": 3.8996, + "step": 266 + }, + { + "epoch": 0.06, + "learning_rate": 4.976294516058554e-06, + "loss": 3.8829, + "step": 267 + }, + { + "epoch": 0.06, + "learning_rate": 4.976185274197073e-06, + "loss": 3.9417, + "step": 268 + }, + { + "epoch": 0.06, + "learning_rate": 4.976076032335592e-06, + "loss": 3.943, + "step": 269 + }, + { + "epoch": 0.06, + "learning_rate": 4.97596679047411e-06, + "loss": 3.8052, + "step": 270 + }, + { + "epoch": 0.06, + "learning_rate": 4.975857548612629e-06, + "loss": 3.9308, + "step": 271 + }, + { + "epoch": 0.06, + "learning_rate": 4.975748306751147e-06, + "loss": 3.9197, + "step": 272 + }, + { + "epoch": 0.06, + "learning_rate": 4.975639064889666e-06, + "loss": 3.9079, + "step": 273 + }, + { + "epoch": 0.06, + "learning_rate": 4.9755298230281845e-06, + "loss": 3.8667, + "step": 274 + }, + { + "epoch": 0.06, + "learning_rate": 4.9754205811667036e-06, + "loss": 3.9865, + "step": 275 + }, + { + "epoch": 0.06, + "learning_rate": 4.975311339305222e-06, + "loss": 3.8828, + "step": 276 + }, + { + "epoch": 0.06, + "learning_rate": 4.975202097443741e-06, + "loss": 3.9143, + "step": 277 + }, + { + "epoch": 0.06, + "learning_rate": 4.97509285558226e-06, + "loss": 3.8621, + "step": 278 + }, + { + "epoch": 0.06, + "learning_rate": 4.974983613720778e-06, + "loss": 3.8058, + "step": 279 + }, + { + "epoch": 0.06, + "learning_rate": 4.974874371859297e-06, + "loss": 3.8596, + "step": 280 + }, + { + "epoch": 0.06, + "learning_rate": 4.974765129997815e-06, + "loss": 3.9481, + "step": 281 + }, + { + "epoch": 0.06, + "learning_rate": 4.974655888136334e-06, + "loss": 3.834, + "step": 282 + }, + { + "epoch": 0.06, + "learning_rate": 4.974546646274853e-06, + "loss": 3.8394, + "step": 283 + }, + { + "epoch": 0.06, + "learning_rate": 4.9744374044133715e-06, + "loss": 3.8192, + "step": 284 + }, + { + "epoch": 0.06, + "learning_rate": 4.9743281625518905e-06, + "loss": 3.9124, + "step": 285 + }, + { + "epoch": 0.06, + "learning_rate": 4.974218920690409e-06, + "loss": 3.8719, + "step": 286 + }, + { + "epoch": 0.06, + "learning_rate": 4.974109678828928e-06, + "loss": 3.8188, + "step": 287 + }, + { + "epoch": 0.06, + "learning_rate": 4.974000436967446e-06, + "loss": 3.8723, + "step": 288 + }, + { + "epoch": 0.06, + "learning_rate": 4.973891195105965e-06, + "loss": 3.783, + "step": 289 + }, + { + "epoch": 0.06, + "learning_rate": 4.973781953244484e-06, + "loss": 3.8891, + "step": 290 + }, + { + "epoch": 0.06, + "learning_rate": 4.973672711383002e-06, + "loss": 3.9242, + "step": 291 + }, + { + "epoch": 0.06, + "learning_rate": 4.973563469521521e-06, + "loss": 3.826, + "step": 292 + }, + { + "epoch": 0.06, + "learning_rate": 4.973454227660039e-06, + "loss": 3.9141, + "step": 293 + }, + { + "epoch": 0.06, + "learning_rate": 4.9733449857985584e-06, + "loss": 3.9384, + "step": 294 + }, + { + "epoch": 0.06, + "learning_rate": 4.9732357439370775e-06, + "loss": 3.8704, + "step": 295 + }, + { + "epoch": 0.06, + "learning_rate": 4.973126502075596e-06, + "loss": 3.8117, + "step": 296 + }, + { + "epoch": 0.06, + "learning_rate": 4.973017260214115e-06, + "loss": 3.8656, + "step": 297 + }, + { + "epoch": 0.07, + "learning_rate": 4.972908018352633e-06, + "loss": 3.9082, + "step": 298 + }, + { + "epoch": 0.07, + "learning_rate": 4.972798776491152e-06, + "loss": 3.8649, + "step": 299 + }, + { + "epoch": 0.07, + "learning_rate": 4.972689534629671e-06, + "loss": 3.9288, + "step": 300 + }, + { + "epoch": 0.07, + "learning_rate": 4.972580292768189e-06, + "loss": 3.9474, + "step": 301 + }, + { + "epoch": 0.07, + "learning_rate": 4.972471050906708e-06, + "loss": 3.9051, + "step": 302 + }, + { + "epoch": 0.07, + "learning_rate": 4.972361809045226e-06, + "loss": 3.9372, + "step": 303 + }, + { + "epoch": 0.07, + "learning_rate": 4.972252567183745e-06, + "loss": 3.9302, + "step": 304 + }, + { + "epoch": 0.07, + "learning_rate": 4.9721433253222636e-06, + "loss": 3.8036, + "step": 305 + }, + { + "epoch": 0.07, + "learning_rate": 4.972034083460783e-06, + "loss": 3.9017, + "step": 306 + }, + { + "epoch": 0.07, + "learning_rate": 4.971924841599301e-06, + "loss": 3.8454, + "step": 307 + }, + { + "epoch": 0.07, + "learning_rate": 4.97181559973782e-06, + "loss": 3.9022, + "step": 308 + }, + { + "epoch": 0.07, + "learning_rate": 4.971706357876339e-06, + "loss": 3.913, + "step": 309 + }, + { + "epoch": 0.07, + "learning_rate": 4.971597116014857e-06, + "loss": 3.8254, + "step": 310 + }, + { + "epoch": 0.07, + "learning_rate": 4.971487874153376e-06, + "loss": 3.8835, + "step": 311 + }, + { + "epoch": 0.07, + "learning_rate": 4.971378632291894e-06, + "loss": 3.9061, + "step": 312 + }, + { + "epoch": 0.07, + "learning_rate": 4.971269390430413e-06, + "loss": 3.9886, + "step": 313 + }, + { + "epoch": 0.07, + "learning_rate": 4.971160148568932e-06, + "loss": 3.8795, + "step": 314 + }, + { + "epoch": 0.07, + "learning_rate": 4.9710509067074505e-06, + "loss": 3.919, + "step": 315 + }, + { + "epoch": 0.07, + "learning_rate": 4.9709416648459696e-06, + "loss": 3.885, + "step": 316 + }, + { + "epoch": 0.07, + "learning_rate": 4.970832422984488e-06, + "loss": 3.8932, + "step": 317 + }, + { + "epoch": 0.07, + "learning_rate": 4.970723181123007e-06, + "loss": 3.8612, + "step": 318 + }, + { + "epoch": 0.07, + "learning_rate": 4.970613939261526e-06, + "loss": 3.8062, + "step": 319 + }, + { + "epoch": 0.07, + "learning_rate": 4.970504697400044e-06, + "loss": 3.88, + "step": 320 + }, + { + "epoch": 0.07, + "learning_rate": 4.970395455538562e-06, + "loss": 3.9184, + "step": 321 + }, + { + "epoch": 0.07, + "learning_rate": 4.970286213677081e-06, + "loss": 3.878, + "step": 322 + }, + { + "epoch": 0.07, + "learning_rate": 4.9701769718156e-06, + "loss": 3.9048, + "step": 323 + }, + { + "epoch": 0.07, + "learning_rate": 4.970067729954119e-06, + "loss": 3.8056, + "step": 324 + }, + { + "epoch": 0.07, + "learning_rate": 4.9699584880926375e-06, + "loss": 3.9224, + "step": 325 + }, + { + "epoch": 0.07, + "learning_rate": 4.969849246231156e-06, + "loss": 3.8566, + "step": 326 + }, + { + "epoch": 0.07, + "learning_rate": 4.969740004369675e-06, + "loss": 3.984, + "step": 327 + }, + { + "epoch": 0.07, + "learning_rate": 4.969630762508194e-06, + "loss": 3.7776, + "step": 328 + }, + { + "epoch": 0.07, + "learning_rate": 4.969521520646712e-06, + "loss": 3.8337, + "step": 329 + }, + { + "epoch": 0.07, + "learning_rate": 4.969412278785231e-06, + "loss": 3.8938, + "step": 330 + }, + { + "epoch": 0.07, + "learning_rate": 4.969303036923749e-06, + "loss": 3.8472, + "step": 331 + }, + { + "epoch": 0.07, + "learning_rate": 4.969193795062268e-06, + "loss": 3.8943, + "step": 332 + }, + { + "epoch": 0.07, + "learning_rate": 4.969084553200787e-06, + "loss": 3.8563, + "step": 333 + }, + { + "epoch": 0.07, + "learning_rate": 4.968975311339305e-06, + "loss": 3.887, + "step": 334 + }, + { + "epoch": 0.07, + "learning_rate": 4.9688660694778244e-06, + "loss": 3.9107, + "step": 335 + }, + { + "epoch": 0.07, + "learning_rate": 4.968756827616343e-06, + "loss": 3.7904, + "step": 336 + }, + { + "epoch": 0.07, + "learning_rate": 4.968647585754862e-06, + "loss": 3.8892, + "step": 337 + }, + { + "epoch": 0.07, + "learning_rate": 4.968538343893381e-06, + "loss": 3.9228, + "step": 338 + }, + { + "epoch": 0.07, + "learning_rate": 4.968429102031899e-06, + "loss": 3.8192, + "step": 339 + }, + { + "epoch": 0.07, + "learning_rate": 4.968319860170417e-06, + "loss": 3.8472, + "step": 340 + }, + { + "epoch": 0.07, + "learning_rate": 4.968210618308936e-06, + "loss": 3.756, + "step": 341 + }, + { + "epoch": 0.07, + "learning_rate": 4.968101376447455e-06, + "loss": 3.8806, + "step": 342 + }, + { + "epoch": 0.07, + "learning_rate": 4.967992134585974e-06, + "loss": 3.8383, + "step": 343 + }, + { + "epoch": 0.08, + "learning_rate": 4.967882892724492e-06, + "loss": 3.9254, + "step": 344 + }, + { + "epoch": 0.08, + "learning_rate": 4.9677736508630106e-06, + "loss": 3.87, + "step": 345 + }, + { + "epoch": 0.08, + "learning_rate": 4.96766440900153e-06, + "loss": 3.9253, + "step": 346 + }, + { + "epoch": 0.08, + "learning_rate": 4.967555167140049e-06, + "loss": 3.7368, + "step": 347 + }, + { + "epoch": 0.08, + "learning_rate": 4.967445925278568e-06, + "loss": 3.8778, + "step": 348 + }, + { + "epoch": 0.08, + "learning_rate": 4.967336683417086e-06, + "loss": 3.928, + "step": 349 + }, + { + "epoch": 0.08, + "learning_rate": 4.967227441555604e-06, + "loss": 3.866, + "step": 350 + }, + { + "epoch": 0.08, + "learning_rate": 4.967118199694123e-06, + "loss": 3.8656, + "step": 351 + }, + { + "epoch": 0.08, + "learning_rate": 4.967008957832642e-06, + "loss": 3.9004, + "step": 352 + }, + { + "epoch": 0.08, + "learning_rate": 4.966899715971161e-06, + "loss": 3.9211, + "step": 353 + }, + { + "epoch": 0.08, + "learning_rate": 4.966790474109679e-06, + "loss": 3.888, + "step": 354 + }, + { + "epoch": 0.08, + "learning_rate": 4.9666812322481975e-06, + "loss": 3.9377, + "step": 355 + }, + { + "epoch": 0.08, + "learning_rate": 4.9665719903867165e-06, + "loss": 3.9564, + "step": 356 + }, + { + "epoch": 0.08, + "learning_rate": 4.966462748525236e-06, + "loss": 3.8508, + "step": 357 + }, + { + "epoch": 0.08, + "learning_rate": 4.966353506663754e-06, + "loss": 3.841, + "step": 358 + }, + { + "epoch": 0.08, + "learning_rate": 4.966244264802273e-06, + "loss": 3.9218, + "step": 359 + }, + { + "epoch": 0.08, + "learning_rate": 4.966135022940791e-06, + "loss": 3.903, + "step": 360 + }, + { + "epoch": 0.08, + "learning_rate": 4.96602578107931e-06, + "loss": 3.9153, + "step": 361 + }, + { + "epoch": 0.08, + "learning_rate": 4.965916539217829e-06, + "loss": 3.8502, + "step": 362 + }, + { + "epoch": 0.08, + "learning_rate": 4.965807297356347e-06, + "loss": 3.8147, + "step": 363 + }, + { + "epoch": 0.08, + "learning_rate": 4.9656980554948654e-06, + "loss": 3.872, + "step": 364 + }, + { + "epoch": 0.08, + "learning_rate": 4.9655888136333845e-06, + "loss": 3.9082, + "step": 365 + }, + { + "epoch": 0.08, + "learning_rate": 4.9654795717719035e-06, + "loss": 3.8529, + "step": 366 + }, + { + "epoch": 0.08, + "learning_rate": 4.9653703299104225e-06, + "loss": 3.9068, + "step": 367 + }, + { + "epoch": 0.08, + "learning_rate": 4.965261088048941e-06, + "loss": 3.8715, + "step": 368 + }, + { + "epoch": 0.08, + "learning_rate": 4.965151846187459e-06, + "loss": 3.818, + "step": 369 + }, + { + "epoch": 0.08, + "learning_rate": 4.965042604325978e-06, + "loss": 3.9518, + "step": 370 + }, + { + "epoch": 0.08, + "learning_rate": 4.964933362464497e-06, + "loss": 3.951, + "step": 371 + }, + { + "epoch": 0.08, + "learning_rate": 4.964824120603016e-06, + "loss": 3.881, + "step": 372 + }, + { + "epoch": 0.08, + "learning_rate": 4.964714878741534e-06, + "loss": 3.9405, + "step": 373 + }, + { + "epoch": 0.08, + "learning_rate": 4.964605636880052e-06, + "loss": 3.8381, + "step": 374 + }, + { + "epoch": 0.08, + "learning_rate": 4.9644963950185714e-06, + "loss": 3.7342, + "step": 375 + }, + { + "epoch": 0.08, + "learning_rate": 4.9643871531570905e-06, + "loss": 3.9393, + "step": 376 + }, + { + "epoch": 0.08, + "learning_rate": 4.9642779112956095e-06, + "loss": 3.8646, + "step": 377 + }, + { + "epoch": 0.08, + "learning_rate": 4.964168669434128e-06, + "loss": 3.9563, + "step": 378 + }, + { + "epoch": 0.08, + "learning_rate": 4.964059427572646e-06, + "loss": 3.9384, + "step": 379 + }, + { + "epoch": 0.08, + "learning_rate": 4.963950185711165e-06, + "loss": 3.9279, + "step": 380 + }, + { + "epoch": 0.08, + "learning_rate": 4.963840943849684e-06, + "loss": 3.8869, + "step": 381 + }, + { + "epoch": 0.08, + "learning_rate": 4.963731701988202e-06, + "loss": 3.9836, + "step": 382 + }, + { + "epoch": 0.08, + "learning_rate": 4.963622460126721e-06, + "loss": 3.8187, + "step": 383 + }, + { + "epoch": 0.08, + "learning_rate": 4.963513218265239e-06, + "loss": 3.8279, + "step": 384 + }, + { + "epoch": 0.08, + "learning_rate": 4.963403976403758e-06, + "loss": 3.6946, + "step": 385 + }, + { + "epoch": 0.08, + "learning_rate": 4.963294734542277e-06, + "loss": 3.9541, + "step": 386 + }, + { + "epoch": 0.08, + "learning_rate": 4.963185492680796e-06, + "loss": 3.8754, + "step": 387 + }, + { + "epoch": 0.08, + "learning_rate": 4.963076250819314e-06, + "loss": 3.9014, + "step": 388 + }, + { + "epoch": 0.08, + "learning_rate": 4.962967008957833e-06, + "loss": 3.8851, + "step": 389 + }, + { + "epoch": 0.09, + "learning_rate": 4.962857767096352e-06, + "loss": 3.9365, + "step": 390 + }, + { + "epoch": 0.09, + "learning_rate": 4.962748525234871e-06, + "loss": 3.8112, + "step": 391 + }, + { + "epoch": 0.09, + "learning_rate": 4.962639283373389e-06, + "loss": 3.7862, + "step": 392 + }, + { + "epoch": 0.09, + "learning_rate": 4.962530041511907e-06, + "loss": 3.8744, + "step": 393 + }, + { + "epoch": 0.09, + "learning_rate": 4.962420799650426e-06, + "loss": 3.8345, + "step": 394 + }, + { + "epoch": 0.09, + "learning_rate": 4.962311557788945e-06, + "loss": 3.8435, + "step": 395 + }, + { + "epoch": 0.09, + "learning_rate": 4.962202315927464e-06, + "loss": 3.8548, + "step": 396 + }, + { + "epoch": 0.09, + "learning_rate": 4.9620930740659826e-06, + "loss": 3.7793, + "step": 397 + }, + { + "epoch": 0.09, + "learning_rate": 4.961983832204501e-06, + "loss": 3.887, + "step": 398 + }, + { + "epoch": 0.09, + "learning_rate": 4.96187459034302e-06, + "loss": 3.8834, + "step": 399 + }, + { + "epoch": 0.09, + "learning_rate": 4.961765348481539e-06, + "loss": 3.9159, + "step": 400 + }, + { + "epoch": 0.09, + "learning_rate": 4.961656106620058e-06, + "loss": 3.7931, + "step": 401 + }, + { + "epoch": 0.09, + "learning_rate": 4.961546864758576e-06, + "loss": 3.7805, + "step": 402 + }, + { + "epoch": 0.09, + "learning_rate": 4.961437622897094e-06, + "loss": 3.7271, + "step": 403 + }, + { + "epoch": 0.09, + "learning_rate": 4.961328381035613e-06, + "loss": 3.7942, + "step": 404 + }, + { + "epoch": 0.09, + "learning_rate": 4.961219139174132e-06, + "loss": 3.8552, + "step": 405 + }, + { + "epoch": 0.09, + "learning_rate": 4.9611098973126505e-06, + "loss": 3.955, + "step": 406 + }, + { + "epoch": 0.09, + "learning_rate": 4.9610006554511695e-06, + "loss": 3.8918, + "step": 407 + }, + { + "epoch": 0.09, + "learning_rate": 4.960891413589688e-06, + "loss": 3.8062, + "step": 408 + }, + { + "epoch": 0.09, + "learning_rate": 4.960782171728207e-06, + "loss": 3.8602, + "step": 409 + }, + { + "epoch": 0.09, + "learning_rate": 4.960672929866726e-06, + "loss": 3.8166, + "step": 410 + }, + { + "epoch": 0.09, + "learning_rate": 4.960563688005244e-06, + "loss": 3.8115, + "step": 411 + }, + { + "epoch": 0.09, + "learning_rate": 4.960454446143763e-06, + "loss": 3.8122, + "step": 412 + }, + { + "epoch": 0.09, + "learning_rate": 4.960345204282281e-06, + "loss": 3.8951, + "step": 413 + }, + { + "epoch": 0.09, + "learning_rate": 4.9602359624208e-06, + "loss": 3.9196, + "step": 414 + }, + { + "epoch": 0.09, + "learning_rate": 4.960126720559319e-06, + "loss": 3.862, + "step": 415 + }, + { + "epoch": 0.09, + "learning_rate": 4.9600174786978374e-06, + "loss": 3.8131, + "step": 416 + }, + { + "epoch": 0.09, + "learning_rate": 4.959908236836356e-06, + "loss": 3.8874, + "step": 417 + }, + { + "epoch": 0.09, + "learning_rate": 4.959798994974875e-06, + "loss": 3.9959, + "step": 418 + }, + { + "epoch": 0.09, + "learning_rate": 4.959689753113394e-06, + "loss": 3.8583, + "step": 419 + }, + { + "epoch": 0.09, + "learning_rate": 4.959580511251912e-06, + "loss": 3.876, + "step": 420 + }, + { + "epoch": 0.09, + "learning_rate": 4.959471269390431e-06, + "loss": 3.8364, + "step": 421 + }, + { + "epoch": 0.09, + "learning_rate": 4.959362027528949e-06, + "loss": 3.9542, + "step": 422 + }, + { + "epoch": 0.09, + "learning_rate": 4.959252785667468e-06, + "loss": 3.9016, + "step": 423 + }, + { + "epoch": 0.09, + "learning_rate": 4.959143543805987e-06, + "loss": 3.9403, + "step": 424 + }, + { + "epoch": 0.09, + "learning_rate": 4.959034301944505e-06, + "loss": 3.8938, + "step": 425 + }, + { + "epoch": 0.09, + "learning_rate": 4.958925060083024e-06, + "loss": 3.8308, + "step": 426 + }, + { + "epoch": 0.09, + "learning_rate": 4.958815818221543e-06, + "loss": 3.871, + "step": 427 + }, + { + "epoch": 0.09, + "learning_rate": 4.958706576360062e-06, + "loss": 3.803, + "step": 428 + }, + { + "epoch": 0.09, + "learning_rate": 4.958597334498581e-06, + "loss": 3.8369, + "step": 429 + }, + { + "epoch": 0.09, + "learning_rate": 4.958488092637099e-06, + "loss": 3.8588, + "step": 430 + }, + { + "epoch": 0.09, + "learning_rate": 4.958378850775618e-06, + "loss": 3.9304, + "step": 431 + }, + { + "epoch": 0.09, + "learning_rate": 4.958269608914136e-06, + "loss": 3.8948, + "step": 432 + }, + { + "epoch": 0.09, + "learning_rate": 4.958160367052655e-06, + "loss": 3.7965, + "step": 433 + }, + { + "epoch": 0.09, + "learning_rate": 4.958051125191174e-06, + "loss": 3.7043, + "step": 434 + }, + { + "epoch": 0.09, + "learning_rate": 4.957941883329692e-06, + "loss": 3.8702, + "step": 435 + }, + { + "epoch": 0.1, + "learning_rate": 4.957832641468211e-06, + "loss": 3.8355, + "step": 436 + }, + { + "epoch": 0.1, + "learning_rate": 4.9577233996067295e-06, + "loss": 3.8284, + "step": 437 + }, + { + "epoch": 0.1, + "learning_rate": 4.957614157745249e-06, + "loss": 3.8311, + "step": 438 + }, + { + "epoch": 0.1, + "learning_rate": 4.957504915883767e-06, + "loss": 3.8504, + "step": 439 + }, + { + "epoch": 0.1, + "learning_rate": 4.957395674022286e-06, + "loss": 3.8746, + "step": 440 + }, + { + "epoch": 0.1, + "learning_rate": 4.957286432160804e-06, + "loss": 3.778, + "step": 441 + }, + { + "epoch": 0.1, + "learning_rate": 4.957177190299323e-06, + "loss": 3.7511, + "step": 442 + }, + { + "epoch": 0.1, + "learning_rate": 4.957067948437842e-06, + "loss": 3.8755, + "step": 443 + }, + { + "epoch": 0.1, + "learning_rate": 4.95695870657636e-06, + "loss": 3.8044, + "step": 444 + }, + { + "epoch": 0.1, + "learning_rate": 4.956849464714879e-06, + "loss": 3.8279, + "step": 445 + }, + { + "epoch": 0.1, + "learning_rate": 4.9567402228533975e-06, + "loss": 3.7934, + "step": 446 + }, + { + "epoch": 0.1, + "learning_rate": 4.9566309809919165e-06, + "loss": 3.7727, + "step": 447 + }, + { + "epoch": 0.1, + "learning_rate": 4.9565217391304355e-06, + "loss": 3.7752, + "step": 448 + }, + { + "epoch": 0.1, + "learning_rate": 4.956412497268954e-06, + "loss": 3.8558, + "step": 449 + }, + { + "epoch": 0.1, + "learning_rate": 4.956303255407473e-06, + "loss": 3.7912, + "step": 450 + }, + { + "epoch": 0.1, + "learning_rate": 4.956194013545991e-06, + "loss": 3.9126, + "step": 451 + }, + { + "epoch": 0.1, + "learning_rate": 4.95608477168451e-06, + "loss": 3.8728, + "step": 452 + }, + { + "epoch": 0.1, + "learning_rate": 4.955975529823028e-06, + "loss": 4.0069, + "step": 453 + }, + { + "epoch": 0.1, + "learning_rate": 4.955866287961547e-06, + "loss": 3.9835, + "step": 454 + }, + { + "epoch": 0.1, + "learning_rate": 4.955757046100066e-06, + "loss": 3.7845, + "step": 455 + }, + { + "epoch": 0.1, + "learning_rate": 4.955647804238584e-06, + "loss": 3.8281, + "step": 456 + }, + { + "epoch": 0.1, + "learning_rate": 4.9555385623771035e-06, + "loss": 3.7964, + "step": 457 + }, + { + "epoch": 0.1, + "learning_rate": 4.955429320515622e-06, + "loss": 3.8627, + "step": 458 + }, + { + "epoch": 0.1, + "learning_rate": 4.955320078654141e-06, + "loss": 3.7874, + "step": 459 + }, + { + "epoch": 0.1, + "learning_rate": 4.95521083679266e-06, + "loss": 3.8567, + "step": 460 + }, + { + "epoch": 0.1, + "learning_rate": 4.955101594931178e-06, + "loss": 3.8086, + "step": 461 + }, + { + "epoch": 0.1, + "learning_rate": 4.954992353069697e-06, + "loss": 3.8443, + "step": 462 + }, + { + "epoch": 0.1, + "learning_rate": 4.954883111208215e-06, + "loss": 3.7993, + "step": 463 + }, + { + "epoch": 0.1, + "learning_rate": 4.954773869346734e-06, + "loss": 3.943, + "step": 464 + }, + { + "epoch": 0.1, + "learning_rate": 4.954664627485253e-06, + "loss": 3.7649, + "step": 465 + }, + { + "epoch": 0.1, + "learning_rate": 4.954555385623771e-06, + "loss": 3.797, + "step": 466 + }, + { + "epoch": 0.1, + "learning_rate": 4.95444614376229e-06, + "loss": 3.883, + "step": 467 + }, + { + "epoch": 0.1, + "learning_rate": 4.954336901900809e-06, + "loss": 3.8053, + "step": 468 + }, + { + "epoch": 0.1, + "learning_rate": 4.954227660039328e-06, + "loss": 3.8987, + "step": 469 + }, + { + "epoch": 0.1, + "learning_rate": 4.954118418177846e-06, + "loss": 3.9734, + "step": 470 + }, + { + "epoch": 0.1, + "learning_rate": 4.954009176316365e-06, + "loss": 3.8458, + "step": 471 + }, + { + "epoch": 0.1, + "learning_rate": 4.953899934454883e-06, + "loss": 3.8338, + "step": 472 + }, + { + "epoch": 0.1, + "learning_rate": 4.953790692593402e-06, + "loss": 3.7679, + "step": 473 + }, + { + "epoch": 0.1, + "learning_rate": 4.953681450731921e-06, + "loss": 3.8421, + "step": 474 + }, + { + "epoch": 0.1, + "learning_rate": 4.953572208870439e-06, + "loss": 3.8078, + "step": 475 + }, + { + "epoch": 0.1, + "learning_rate": 4.953462967008958e-06, + "loss": 3.8655, + "step": 476 + }, + { + "epoch": 0.1, + "learning_rate": 4.9533537251474765e-06, + "loss": 3.9105, + "step": 477 + }, + { + "epoch": 0.1, + "learning_rate": 4.9532444832859956e-06, + "loss": 3.7929, + "step": 478 + }, + { + "epoch": 0.1, + "learning_rate": 4.953135241424515e-06, + "loss": 3.9715, + "step": 479 + }, + { + "epoch": 0.1, + "learning_rate": 4.953025999563033e-06, + "loss": 3.8892, + "step": 480 + }, + { + "epoch": 0.1, + "learning_rate": 4.952916757701552e-06, + "loss": 3.6726, + "step": 481 + }, + { + "epoch": 0.11, + "learning_rate": 4.95280751584007e-06, + "loss": 3.935, + "step": 482 + }, + { + "epoch": 0.11, + "learning_rate": 4.952698273978589e-06, + "loss": 3.8559, + "step": 483 + }, + { + "epoch": 0.11, + "learning_rate": 4.952589032117108e-06, + "loss": 3.827, + "step": 484 + }, + { + "epoch": 0.11, + "learning_rate": 4.952479790255626e-06, + "loss": 3.7043, + "step": 485 + }, + { + "epoch": 0.11, + "learning_rate": 4.9523705483941444e-06, + "loss": 3.7923, + "step": 486 + }, + { + "epoch": 0.11, + "learning_rate": 4.9522613065326635e-06, + "loss": 3.8379, + "step": 487 + }, + { + "epoch": 0.11, + "learning_rate": 4.9521520646711825e-06, + "loss": 3.8425, + "step": 488 + }, + { + "epoch": 0.11, + "learning_rate": 4.9520428228097015e-06, + "loss": 3.8222, + "step": 489 + }, + { + "epoch": 0.11, + "learning_rate": 4.95193358094822e-06, + "loss": 3.8596, + "step": 490 + }, + { + "epoch": 0.11, + "learning_rate": 4.951824339086738e-06, + "loss": 3.7561, + "step": 491 + }, + { + "epoch": 0.11, + "learning_rate": 4.951715097225257e-06, + "loss": 3.8586, + "step": 492 + }, + { + "epoch": 0.11, + "learning_rate": 4.951605855363776e-06, + "loss": 3.9116, + "step": 493 + }, + { + "epoch": 0.11, + "learning_rate": 4.951496613502294e-06, + "loss": 3.8903, + "step": 494 + }, + { + "epoch": 0.11, + "learning_rate": 4.951387371640813e-06, + "loss": 3.7608, + "step": 495 + }, + { + "epoch": 0.11, + "learning_rate": 4.951278129779331e-06, + "loss": 3.8002, + "step": 496 + }, + { + "epoch": 0.11, + "learning_rate": 4.9511688879178504e-06, + "loss": 3.7934, + "step": 497 + }, + { + "epoch": 0.11, + "learning_rate": 4.9510596460563695e-06, + "loss": 3.8812, + "step": 498 + }, + { + "epoch": 0.11, + "learning_rate": 4.950950404194888e-06, + "loss": 3.7617, + "step": 499 + }, + { + "epoch": 0.11, + "learning_rate": 4.950841162333407e-06, + "loss": 3.8683, + "step": 500 + }, + { + "epoch": 0.11, + "learning_rate": 4.950731920471925e-06, + "loss": 3.8939, + "step": 501 + }, + { + "epoch": 0.11, + "learning_rate": 4.950622678610444e-06, + "loss": 3.9072, + "step": 502 + }, + { + "epoch": 0.11, + "learning_rate": 4.950513436748963e-06, + "loss": 3.7809, + "step": 503 + }, + { + "epoch": 0.11, + "learning_rate": 4.950404194887481e-06, + "loss": 3.8285, + "step": 504 + }, + { + "epoch": 0.11, + "learning_rate": 4.950294953025999e-06, + "loss": 3.7568, + "step": 505 + }, + { + "epoch": 0.11, + "learning_rate": 4.950185711164518e-06, + "loss": 3.8911, + "step": 506 + }, + { + "epoch": 0.11, + "learning_rate": 4.950076469303037e-06, + "loss": 3.8826, + "step": 507 + }, + { + "epoch": 0.11, + "learning_rate": 4.9499672274415564e-06, + "loss": 3.8121, + "step": 508 + }, + { + "epoch": 0.11, + "learning_rate": 4.949857985580075e-06, + "loss": 3.7619, + "step": 509 + }, + { + "epoch": 0.11, + "learning_rate": 4.949748743718593e-06, + "loss": 3.9082, + "step": 510 + }, + { + "epoch": 0.11, + "learning_rate": 4.949639501857112e-06, + "loss": 3.9492, + "step": 511 + }, + { + "epoch": 0.11, + "learning_rate": 4.949530259995631e-06, + "loss": 3.8916, + "step": 512 + }, + { + "epoch": 0.11, + "learning_rate": 4.94942101813415e-06, + "loss": 3.8755, + "step": 513 + }, + { + "epoch": 0.11, + "learning_rate": 4.949311776272668e-06, + "loss": 3.9189, + "step": 514 + }, + { + "epoch": 0.11, + "learning_rate": 4.949202534411186e-06, + "loss": 3.8247, + "step": 515 + }, + { + "epoch": 0.11, + "learning_rate": 4.949093292549705e-06, + "loss": 3.812, + "step": 516 + }, + { + "epoch": 0.11, + "learning_rate": 4.948984050688224e-06, + "loss": 3.8781, + "step": 517 + }, + { + "epoch": 0.11, + "learning_rate": 4.948874808826743e-06, + "loss": 3.828, + "step": 518 + }, + { + "epoch": 0.11, + "learning_rate": 4.9487655669652616e-06, + "loss": 3.7898, + "step": 519 + }, + { + "epoch": 0.11, + "learning_rate": 4.94865632510378e-06, + "loss": 3.9069, + "step": 520 + }, + { + "epoch": 0.11, + "learning_rate": 4.948547083242299e-06, + "loss": 3.9009, + "step": 521 + }, + { + "epoch": 0.11, + "learning_rate": 4.948437841380818e-06, + "loss": 3.7183, + "step": 522 + }, + { + "epoch": 0.11, + "learning_rate": 4.948328599519336e-06, + "loss": 3.8485, + "step": 523 + }, + { + "epoch": 0.11, + "learning_rate": 4.948219357657855e-06, + "loss": 3.8757, + "step": 524 + }, + { + "epoch": 0.11, + "learning_rate": 4.948110115796373e-06, + "loss": 3.7873, + "step": 525 + }, + { + "epoch": 0.11, + "learning_rate": 4.948000873934892e-06, + "loss": 3.9355, + "step": 526 + }, + { + "epoch": 0.12, + "learning_rate": 4.947891632073411e-06, + "loss": 3.9181, + "step": 527 + }, + { + "epoch": 0.12, + "learning_rate": 4.9477823902119295e-06, + "loss": 3.8591, + "step": 528 + }, + { + "epoch": 0.12, + "learning_rate": 4.947673148350448e-06, + "loss": 3.8655, + "step": 529 + }, + { + "epoch": 0.12, + "learning_rate": 4.947563906488967e-06, + "loss": 3.9489, + "step": 530 + }, + { + "epoch": 0.12, + "learning_rate": 4.947454664627486e-06, + "loss": 3.6412, + "step": 531 + }, + { + "epoch": 0.12, + "learning_rate": 4.947345422766005e-06, + "loss": 3.8959, + "step": 532 + }, + { + "epoch": 0.12, + "learning_rate": 4.947236180904523e-06, + "loss": 3.7979, + "step": 533 + }, + { + "epoch": 0.12, + "learning_rate": 4.947126939043041e-06, + "loss": 3.9539, + "step": 534 + }, + { + "epoch": 0.12, + "learning_rate": 4.94701769718156e-06, + "loss": 3.8113, + "step": 535 + }, + { + "epoch": 0.12, + "learning_rate": 4.946908455320079e-06, + "loss": 3.9261, + "step": 536 + }, + { + "epoch": 0.12, + "learning_rate": 4.946799213458598e-06, + "loss": 3.905, + "step": 537 + }, + { + "epoch": 0.12, + "learning_rate": 4.9466899715971165e-06, + "loss": 3.7982, + "step": 538 + }, + { + "epoch": 0.12, + "learning_rate": 4.946580729735635e-06, + "loss": 3.7408, + "step": 539 + }, + { + "epoch": 0.12, + "learning_rate": 4.946471487874154e-06, + "loss": 3.9598, + "step": 540 + }, + { + "epoch": 0.12, + "learning_rate": 4.946362246012673e-06, + "loss": 3.7688, + "step": 541 + }, + { + "epoch": 0.12, + "learning_rate": 4.946253004151192e-06, + "loss": 3.9289, + "step": 542 + }, + { + "epoch": 0.12, + "learning_rate": 4.94614376228971e-06, + "loss": 3.8769, + "step": 543 + }, + { + "epoch": 0.12, + "learning_rate": 4.946034520428228e-06, + "loss": 3.974, + "step": 544 + }, + { + "epoch": 0.12, + "learning_rate": 4.945925278566747e-06, + "loss": 3.9452, + "step": 545 + }, + { + "epoch": 0.12, + "learning_rate": 4.945816036705266e-06, + "loss": 3.6661, + "step": 546 + }, + { + "epoch": 0.12, + "learning_rate": 4.945706794843784e-06, + "loss": 3.9051, + "step": 547 + }, + { + "epoch": 0.12, + "learning_rate": 4.945597552982303e-06, + "loss": 3.7771, + "step": 548 + }, + { + "epoch": 0.12, + "learning_rate": 4.945488311120822e-06, + "loss": 3.8807, + "step": 549 + }, + { + "epoch": 0.12, + "learning_rate": 4.945379069259341e-06, + "loss": 3.7085, + "step": 550 + }, + { + "epoch": 0.12, + "learning_rate": 4.94526982739786e-06, + "loss": 3.7862, + "step": 551 + }, + { + "epoch": 0.12, + "learning_rate": 4.945160585536378e-06, + "loss": 3.9446, + "step": 552 + }, + { + "epoch": 0.12, + "learning_rate": 4.945051343674897e-06, + "loss": 3.869, + "step": 553 + }, + { + "epoch": 0.12, + "learning_rate": 4.944942101813415e-06, + "loss": 3.8282, + "step": 554 + }, + { + "epoch": 0.12, + "learning_rate": 4.944832859951934e-06, + "loss": 3.7538, + "step": 555 + }, + { + "epoch": 0.12, + "learning_rate": 4.944723618090453e-06, + "loss": 3.7202, + "step": 556 + }, + { + "epoch": 0.12, + "learning_rate": 4.944614376228971e-06, + "loss": 3.7572, + "step": 557 + }, + { + "epoch": 0.12, + "learning_rate": 4.9445051343674895e-06, + "loss": 3.8389, + "step": 558 + }, + { + "epoch": 0.12, + "learning_rate": 4.9443958925060086e-06, + "loss": 3.8367, + "step": 559 + }, + { + "epoch": 0.12, + "learning_rate": 4.944286650644528e-06, + "loss": 3.6748, + "step": 560 + }, + { + "epoch": 0.12, + "learning_rate": 4.944177408783047e-06, + "loss": 3.9369, + "step": 561 + }, + { + "epoch": 0.12, + "learning_rate": 4.944068166921565e-06, + "loss": 3.9236, + "step": 562 + }, + { + "epoch": 0.12, + "learning_rate": 4.943958925060083e-06, + "loss": 3.9772, + "step": 563 + }, + { + "epoch": 0.12, + "learning_rate": 4.943849683198602e-06, + "loss": 3.7884, + "step": 564 + }, + { + "epoch": 0.12, + "learning_rate": 4.943740441337121e-06, + "loss": 3.8446, + "step": 565 + }, + { + "epoch": 0.12, + "learning_rate": 4.94363119947564e-06, + "loss": 3.9579, + "step": 566 + }, + { + "epoch": 0.12, + "learning_rate": 4.943521957614158e-06, + "loss": 3.7957, + "step": 567 + }, + { + "epoch": 0.12, + "learning_rate": 4.9434127157526765e-06, + "loss": 3.7494, + "step": 568 + }, + { + "epoch": 0.12, + "learning_rate": 4.9433034738911955e-06, + "loss": 3.8932, + "step": 569 + }, + { + "epoch": 0.12, + "learning_rate": 4.9431942320297145e-06, + "loss": 3.8748, + "step": 570 + }, + { + "epoch": 0.12, + "learning_rate": 4.943084990168233e-06, + "loss": 3.983, + "step": 571 + }, + { + "epoch": 0.12, + "learning_rate": 4.942975748306752e-06, + "loss": 3.8601, + "step": 572 + }, + { + "epoch": 0.13, + "learning_rate": 4.94286650644527e-06, + "loss": 3.9684, + "step": 573 + }, + { + "epoch": 0.13, + "learning_rate": 4.942757264583789e-06, + "loss": 3.9593, + "step": 574 + }, + { + "epoch": 0.13, + "learning_rate": 4.942648022722308e-06, + "loss": 3.8985, + "step": 575 + }, + { + "epoch": 0.13, + "learning_rate": 4.942538780860826e-06, + "loss": 3.8876, + "step": 576 + }, + { + "epoch": 0.13, + "learning_rate": 4.942429538999345e-06, + "loss": 3.8261, + "step": 577 + }, + { + "epoch": 0.13, + "learning_rate": 4.9423202971378634e-06, + "loss": 3.8805, + "step": 578 + }, + { + "epoch": 0.13, + "learning_rate": 4.9422110552763825e-06, + "loss": 3.7585, + "step": 579 + }, + { + "epoch": 0.13, + "learning_rate": 4.9421018134149015e-06, + "loss": 3.9254, + "step": 580 + }, + { + "epoch": 0.13, + "learning_rate": 4.94199257155342e-06, + "loss": 3.9284, + "step": 581 + }, + { + "epoch": 0.13, + "learning_rate": 4.941883329691938e-06, + "loss": 3.8341, + "step": 582 + }, + { + "epoch": 0.13, + "learning_rate": 4.941774087830457e-06, + "loss": 3.7959, + "step": 583 + }, + { + "epoch": 0.13, + "learning_rate": 4.941664845968976e-06, + "loss": 3.8646, + "step": 584 + }, + { + "epoch": 0.13, + "learning_rate": 4.941555604107494e-06, + "loss": 3.9035, + "step": 585 + }, + { + "epoch": 0.13, + "learning_rate": 4.941446362246013e-06, + "loss": 3.9723, + "step": 586 + }, + { + "epoch": 0.13, + "learning_rate": 4.941337120384531e-06, + "loss": 3.8546, + "step": 587 + }, + { + "epoch": 0.13, + "learning_rate": 4.94122787852305e-06, + "loss": 3.7705, + "step": 588 + }, + { + "epoch": 0.13, + "learning_rate": 4.9411186366615694e-06, + "loss": 3.8279, + "step": 589 + }, + { + "epoch": 0.13, + "learning_rate": 4.941009394800088e-06, + "loss": 3.7052, + "step": 590 + }, + { + "epoch": 0.13, + "learning_rate": 4.940900152938607e-06, + "loss": 3.8491, + "step": 591 + }, + { + "epoch": 0.13, + "learning_rate": 4.940790911077125e-06, + "loss": 3.9418, + "step": 592 + }, + { + "epoch": 0.13, + "learning_rate": 4.940681669215644e-06, + "loss": 3.7845, + "step": 593 + }, + { + "epoch": 0.13, + "learning_rate": 4.940572427354163e-06, + "loss": 3.9187, + "step": 594 + }, + { + "epoch": 0.13, + "learning_rate": 4.940463185492681e-06, + "loss": 3.8022, + "step": 595 + }, + { + "epoch": 0.13, + "learning_rate": 4.9403539436312e-06, + "loss": 3.864, + "step": 596 + }, + { + "epoch": 0.13, + "learning_rate": 4.940244701769718e-06, + "loss": 3.9414, + "step": 597 + }, + { + "epoch": 0.13, + "learning_rate": 4.940135459908237e-06, + "loss": 3.8482, + "step": 598 + }, + { + "epoch": 0.13, + "learning_rate": 4.940026218046756e-06, + "loss": 3.8572, + "step": 599 + }, + { + "epoch": 0.13, + "learning_rate": 4.9399169761852746e-06, + "loss": 3.8525, + "step": 600 + }, + { + "epoch": 0.13, + "learning_rate": 4.939807734323794e-06, + "loss": 3.8392, + "step": 601 + }, + { + "epoch": 0.13, + "learning_rate": 4.939698492462312e-06, + "loss": 3.7869, + "step": 602 + }, + { + "epoch": 0.13, + "learning_rate": 4.939589250600831e-06, + "loss": 3.6809, + "step": 603 + }, + { + "epoch": 0.13, + "learning_rate": 4.939480008739349e-06, + "loss": 3.5324, + "step": 604 + }, + { + "epoch": 0.13, + "learning_rate": 4.939370766877868e-06, + "loss": 3.8853, + "step": 605 + }, + { + "epoch": 0.13, + "learning_rate": 4.939261525016387e-06, + "loss": 3.9694, + "step": 606 + }, + { + "epoch": 0.13, + "learning_rate": 4.939152283154905e-06, + "loss": 3.969, + "step": 607 + }, + { + "epoch": 0.13, + "learning_rate": 4.939043041293424e-06, + "loss": 3.9097, + "step": 608 + }, + { + "epoch": 0.13, + "learning_rate": 4.9389337994319425e-06, + "loss": 3.9023, + "step": 609 + }, + { + "epoch": 0.13, + "learning_rate": 4.9388245575704615e-06, + "loss": 3.7454, + "step": 610 + }, + { + "epoch": 0.13, + "learning_rate": 4.93871531570898e-06, + "loss": 3.9239, + "step": 611 + }, + { + "epoch": 0.13, + "learning_rate": 4.938606073847499e-06, + "loss": 3.8946, + "step": 612 + }, + { + "epoch": 0.13, + "learning_rate": 4.938496831986018e-06, + "loss": 3.9316, + "step": 613 + }, + { + "epoch": 0.13, + "learning_rate": 4.938387590124536e-06, + "loss": 3.7021, + "step": 614 + }, + { + "epoch": 0.13, + "learning_rate": 4.938278348263055e-06, + "loss": 3.842, + "step": 615 + }, + { + "epoch": 0.13, + "learning_rate": 4.938169106401573e-06, + "loss": 3.94, + "step": 616 + }, + { + "epoch": 0.13, + "learning_rate": 4.938059864540092e-06, + "loss": 3.8811, + "step": 617 + }, + { + "epoch": 0.13, + "learning_rate": 4.93795062267861e-06, + "loss": 3.8147, + "step": 618 + }, + { + "epoch": 0.14, + "learning_rate": 4.9378413808171294e-06, + "loss": 3.9189, + "step": 619 + }, + { + "epoch": 0.14, + "learning_rate": 4.9377321389556485e-06, + "loss": 3.8276, + "step": 620 + }, + { + "epoch": 0.14, + "learning_rate": 4.937622897094167e-06, + "loss": 3.7751, + "step": 621 + }, + { + "epoch": 0.14, + "learning_rate": 4.937513655232686e-06, + "loss": 3.7743, + "step": 622 + }, + { + "epoch": 0.14, + "learning_rate": 4.937404413371204e-06, + "loss": 3.9601, + "step": 623 + }, + { + "epoch": 0.14, + "learning_rate": 4.937295171509723e-06, + "loss": 3.7893, + "step": 624 + }, + { + "epoch": 0.14, + "learning_rate": 4.937185929648242e-06, + "loss": 3.8943, + "step": 625 + }, + { + "epoch": 0.14, + "learning_rate": 4.93707668778676e-06, + "loss": 3.979, + "step": 626 + }, + { + "epoch": 0.14, + "learning_rate": 4.936967445925279e-06, + "loss": 3.8791, + "step": 627 + }, + { + "epoch": 0.14, + "learning_rate": 4.936858204063797e-06, + "loss": 4.0092, + "step": 628 + }, + { + "epoch": 0.14, + "learning_rate": 4.936748962202316e-06, + "loss": 3.6584, + "step": 629 + }, + { + "epoch": 0.14, + "learning_rate": 4.9366397203408354e-06, + "loss": 3.8703, + "step": 630 + }, + { + "epoch": 0.14, + "learning_rate": 4.936530478479354e-06, + "loss": 3.8824, + "step": 631 + }, + { + "epoch": 0.14, + "learning_rate": 4.936421236617873e-06, + "loss": 3.7775, + "step": 632 + }, + { + "epoch": 0.14, + "learning_rate": 4.936311994756391e-06, + "loss": 3.8761, + "step": 633 + }, + { + "epoch": 0.14, + "learning_rate": 4.93620275289491e-06, + "loss": 3.9425, + "step": 634 + }, + { + "epoch": 0.14, + "learning_rate": 4.936093511033428e-06, + "loss": 3.9676, + "step": 635 + }, + { + "epoch": 0.14, + "learning_rate": 4.935984269171947e-06, + "loss": 3.9041, + "step": 636 + }, + { + "epoch": 0.14, + "learning_rate": 4.935875027310465e-06, + "loss": 3.8414, + "step": 637 + }, + { + "epoch": 0.14, + "learning_rate": 4.935765785448984e-06, + "loss": 3.8152, + "step": 638 + }, + { + "epoch": 0.14, + "learning_rate": 4.935656543587503e-06, + "loss": 3.9176, + "step": 639 + }, + { + "epoch": 0.14, + "learning_rate": 4.9355473017260215e-06, + "loss": 4.0028, + "step": 640 + }, + { + "epoch": 0.14, + "learning_rate": 4.935438059864541e-06, + "loss": 3.6564, + "step": 641 + }, + { + "epoch": 0.14, + "learning_rate": 4.935328818003059e-06, + "loss": 3.9149, + "step": 642 + }, + { + "epoch": 0.14, + "learning_rate": 4.935219576141578e-06, + "loss": 3.7229, + "step": 643 + }, + { + "epoch": 0.14, + "learning_rate": 4.935110334280097e-06, + "loss": 3.8357, + "step": 644 + }, + { + "epoch": 0.14, + "learning_rate": 4.935001092418615e-06, + "loss": 3.7082, + "step": 645 + }, + { + "epoch": 0.14, + "learning_rate": 4.934891850557134e-06, + "loss": 3.7392, + "step": 646 + }, + { + "epoch": 0.14, + "learning_rate": 4.934782608695652e-06, + "loss": 4.0507, + "step": 647 + }, + { + "epoch": 0.14, + "learning_rate": 4.934673366834171e-06, + "loss": 3.8931, + "step": 648 + }, + { + "epoch": 0.14, + "learning_rate": 4.93456412497269e-06, + "loss": 3.852, + "step": 649 + }, + { + "epoch": 0.14, + "learning_rate": 4.9344548831112085e-06, + "loss": 3.7887, + "step": 650 + }, + { + "epoch": 0.14, + "learning_rate": 4.934345641249727e-06, + "loss": 3.8955, + "step": 651 + }, + { + "epoch": 0.14, + "learning_rate": 4.934236399388246e-06, + "loss": 4.0058, + "step": 652 + }, + { + "epoch": 0.14, + "learning_rate": 4.934127157526765e-06, + "loss": 3.9005, + "step": 653 + }, + { + "epoch": 0.14, + "learning_rate": 4.934017915665284e-06, + "loss": 3.962, + "step": 654 + }, + { + "epoch": 0.14, + "learning_rate": 4.933908673803802e-06, + "loss": 3.8725, + "step": 655 + }, + { + "epoch": 0.14, + "learning_rate": 4.93379943194232e-06, + "loss": 3.8739, + "step": 656 + }, + { + "epoch": 0.14, + "learning_rate": 4.933690190080839e-06, + "loss": 3.8608, + "step": 657 + }, + { + "epoch": 0.14, + "learning_rate": 4.933580948219358e-06, + "loss": 3.8214, + "step": 658 + }, + { + "epoch": 0.14, + "learning_rate": 4.933471706357877e-06, + "loss": 3.8908, + "step": 659 + }, + { + "epoch": 0.14, + "learning_rate": 4.9333624644963955e-06, + "loss": 3.8616, + "step": 660 + }, + { + "epoch": 0.14, + "learning_rate": 4.933253222634914e-06, + "loss": 3.8455, + "step": 661 + }, + { + "epoch": 0.14, + "learning_rate": 4.933143980773433e-06, + "loss": 3.7644, + "step": 662 + }, + { + "epoch": 0.14, + "learning_rate": 4.933034738911952e-06, + "loss": 3.8432, + "step": 663 + }, + { + "epoch": 0.14, + "learning_rate": 4.93292549705047e-06, + "loss": 3.7214, + "step": 664 + }, + { + "epoch": 0.15, + "learning_rate": 4.932816255188989e-06, + "loss": 3.8704, + "step": 665 + }, + { + "epoch": 0.15, + "learning_rate": 4.932707013327507e-06, + "loss": 3.8626, + "step": 666 + }, + { + "epoch": 0.15, + "learning_rate": 4.932597771466026e-06, + "loss": 3.891, + "step": 667 + }, + { + "epoch": 0.15, + "learning_rate": 4.932488529604545e-06, + "loss": 3.7303, + "step": 668 + }, + { + "epoch": 0.15, + "learning_rate": 4.932379287743063e-06, + "loss": 3.8876, + "step": 669 + }, + { + "epoch": 0.15, + "learning_rate": 4.9322700458815816e-06, + "loss": 3.7111, + "step": 670 + }, + { + "epoch": 0.15, + "learning_rate": 4.932160804020101e-06, + "loss": 3.9144, + "step": 671 + }, + { + "epoch": 0.15, + "learning_rate": 4.93205156215862e-06, + "loss": 3.7925, + "step": 672 + }, + { + "epoch": 0.15, + "learning_rate": 4.931942320297139e-06, + "loss": 3.76, + "step": 673 + }, + { + "epoch": 0.15, + "learning_rate": 4.931833078435657e-06, + "loss": 3.6397, + "step": 674 + }, + { + "epoch": 0.15, + "learning_rate": 4.931723836574175e-06, + "loss": 3.9235, + "step": 675 + }, + { + "epoch": 0.15, + "learning_rate": 4.931614594712694e-06, + "loss": 3.8603, + "step": 676 + }, + { + "epoch": 0.15, + "learning_rate": 4.931505352851213e-06, + "loss": 4.0188, + "step": 677 + }, + { + "epoch": 0.15, + "learning_rate": 4.931396110989732e-06, + "loss": 3.9328, + "step": 678 + }, + { + "epoch": 0.15, + "learning_rate": 4.93128686912825e-06, + "loss": 3.8754, + "step": 679 + }, + { + "epoch": 0.15, + "learning_rate": 4.9311776272667685e-06, + "loss": 3.9608, + "step": 680 + }, + { + "epoch": 0.15, + "learning_rate": 4.9310683854052876e-06, + "loss": 3.8134, + "step": 681 + }, + { + "epoch": 0.15, + "learning_rate": 4.930959143543807e-06, + "loss": 3.8074, + "step": 682 + }, + { + "epoch": 0.15, + "learning_rate": 4.930849901682326e-06, + "loss": 3.9118, + "step": 683 + }, + { + "epoch": 0.15, + "learning_rate": 4.930740659820844e-06, + "loss": 3.855, + "step": 684 + }, + { + "epoch": 0.15, + "learning_rate": 4.930631417959362e-06, + "loss": 3.7173, + "step": 685 + }, + { + "epoch": 0.15, + "learning_rate": 4.930522176097881e-06, + "loss": 3.813, + "step": 686 + }, + { + "epoch": 0.15, + "learning_rate": 4.9304129342364e-06, + "loss": 3.7954, + "step": 687 + }, + { + "epoch": 0.15, + "learning_rate": 4.930303692374918e-06, + "loss": 3.8139, + "step": 688 + }, + { + "epoch": 0.15, + "learning_rate": 4.930194450513437e-06, + "loss": 3.8282, + "step": 689 + }, + { + "epoch": 0.15, + "learning_rate": 4.9300852086519555e-06, + "loss": 3.7585, + "step": 690 + }, + { + "epoch": 0.15, + "learning_rate": 4.9299759667904745e-06, + "loss": 3.9308, + "step": 691 + }, + { + "epoch": 0.15, + "learning_rate": 4.9298667249289936e-06, + "loss": 4.1196, + "step": 692 + }, + { + "epoch": 0.15, + "learning_rate": 4.929757483067512e-06, + "loss": 3.8152, + "step": 693 + }, + { + "epoch": 0.15, + "learning_rate": 4.929648241206031e-06, + "loss": 3.7834, + "step": 694 + }, + { + "epoch": 0.15, + "learning_rate": 4.929538999344549e-06, + "loss": 3.8716, + "step": 695 + }, + { + "epoch": 0.15, + "learning_rate": 4.929429757483068e-06, + "loss": 3.8554, + "step": 696 + }, + { + "epoch": 0.15, + "learning_rate": 4.929320515621587e-06, + "loss": 3.8151, + "step": 697 + }, + { + "epoch": 0.15, + "learning_rate": 4.929211273760105e-06, + "loss": 3.8714, + "step": 698 + }, + { + "epoch": 0.15, + "learning_rate": 4.929102031898623e-06, + "loss": 3.8651, + "step": 699 + }, + { + "epoch": 0.15, + "learning_rate": 4.9289927900371424e-06, + "loss": 3.8322, + "step": 700 + }, + { + "epoch": 0.15, + "learning_rate": 4.9288835481756615e-06, + "loss": 3.9937, + "step": 701 + }, + { + "epoch": 0.15, + "learning_rate": 4.9287743063141805e-06, + "loss": 3.8711, + "step": 702 + }, + { + "epoch": 0.15, + "learning_rate": 4.928665064452699e-06, + "loss": 3.8751, + "step": 703 + }, + { + "epoch": 0.15, + "learning_rate": 4.928555822591217e-06, + "loss": 3.8837, + "step": 704 + }, + { + "epoch": 0.15, + "learning_rate": 4.928446580729736e-06, + "loss": 3.8688, + "step": 705 + }, + { + "epoch": 0.15, + "learning_rate": 4.928337338868255e-06, + "loss": 3.8235, + "step": 706 + }, + { + "epoch": 0.15, + "learning_rate": 4.928228097006774e-06, + "loss": 3.9324, + "step": 707 + }, + { + "epoch": 0.15, + "learning_rate": 4.928118855145292e-06, + "loss": 3.9041, + "step": 708 + }, + { + "epoch": 0.15, + "learning_rate": 4.92800961328381e-06, + "loss": 3.8697, + "step": 709 + }, + { + "epoch": 0.15, + "learning_rate": 4.927900371422329e-06, + "loss": 3.9312, + "step": 710 + }, + { + "epoch": 0.16, + "learning_rate": 4.9277911295608484e-06, + "loss": 3.8382, + "step": 711 + }, + { + "epoch": 0.16, + "learning_rate": 4.9276818876993675e-06, + "loss": 3.8182, + "step": 712 + }, + { + "epoch": 0.16, + "learning_rate": 4.927572645837886e-06, + "loss": 3.6371, + "step": 713 + }, + { + "epoch": 0.16, + "learning_rate": 4.927463403976404e-06, + "loss": 4.0313, + "step": 714 + }, + { + "epoch": 0.16, + "learning_rate": 4.927354162114923e-06, + "loss": 3.5796, + "step": 715 + }, + { + "epoch": 0.16, + "learning_rate": 4.927244920253442e-06, + "loss": 3.9439, + "step": 716 + }, + { + "epoch": 0.16, + "learning_rate": 4.92713567839196e-06, + "loss": 3.8871, + "step": 717 + }, + { + "epoch": 0.16, + "learning_rate": 4.927026436530479e-06, + "loss": 3.9328, + "step": 718 + }, + { + "epoch": 0.16, + "learning_rate": 4.926917194668997e-06, + "loss": 3.8757, + "step": 719 + }, + { + "epoch": 0.16, + "learning_rate": 4.926807952807516e-06, + "loss": 3.7794, + "step": 720 + }, + { + "epoch": 0.16, + "learning_rate": 4.926698710946035e-06, + "loss": 3.8428, + "step": 721 + }, + { + "epoch": 0.16, + "learning_rate": 4.926589469084554e-06, + "loss": 3.8271, + "step": 722 + }, + { + "epoch": 0.16, + "learning_rate": 4.926480227223072e-06, + "loss": 3.9522, + "step": 723 + }, + { + "epoch": 0.16, + "learning_rate": 4.926370985361591e-06, + "loss": 3.8165, + "step": 724 + }, + { + "epoch": 0.16, + "learning_rate": 4.92626174350011e-06, + "loss": 3.8434, + "step": 725 + }, + { + "epoch": 0.16, + "learning_rate": 4.926152501638629e-06, + "loss": 3.8059, + "step": 726 + }, + { + "epoch": 0.16, + "learning_rate": 4.926043259777147e-06, + "loss": 3.7602, + "step": 727 + }, + { + "epoch": 0.16, + "learning_rate": 4.925934017915665e-06, + "loss": 3.8696, + "step": 728 + }, + { + "epoch": 0.16, + "learning_rate": 4.925824776054184e-06, + "loss": 3.8698, + "step": 729 + }, + { + "epoch": 0.16, + "learning_rate": 4.925715534192703e-06, + "loss": 3.7996, + "step": 730 + }, + { + "epoch": 0.16, + "learning_rate": 4.9256062923312215e-06, + "loss": 3.8292, + "step": 731 + }, + { + "epoch": 0.16, + "learning_rate": 4.9254970504697405e-06, + "loss": 3.907, + "step": 732 + }, + { + "epoch": 0.16, + "learning_rate": 4.925387808608259e-06, + "loss": 3.8066, + "step": 733 + }, + { + "epoch": 0.16, + "learning_rate": 4.925278566746778e-06, + "loss": 3.9668, + "step": 734 + }, + { + "epoch": 0.16, + "learning_rate": 4.925169324885297e-06, + "loss": 3.7983, + "step": 735 + }, + { + "epoch": 0.16, + "learning_rate": 4.925060083023815e-06, + "loss": 3.7922, + "step": 736 + }, + { + "epoch": 0.16, + "learning_rate": 4.924950841162334e-06, + "loss": 3.976, + "step": 737 + }, + { + "epoch": 0.16, + "learning_rate": 4.924841599300852e-06, + "loss": 3.868, + "step": 738 + }, + { + "epoch": 0.16, + "learning_rate": 4.924732357439371e-06, + "loss": 3.9403, + "step": 739 + }, + { + "epoch": 0.16, + "learning_rate": 4.92462311557789e-06, + "loss": 3.7494, + "step": 740 + }, + { + "epoch": 0.16, + "learning_rate": 4.9245138737164085e-06, + "loss": 3.743, + "step": 741 + }, + { + "epoch": 0.16, + "learning_rate": 4.9244046318549275e-06, + "loss": 3.8007, + "step": 742 + }, + { + "epoch": 0.16, + "learning_rate": 4.924295389993446e-06, + "loss": 3.892, + "step": 743 + }, + { + "epoch": 0.16, + "learning_rate": 4.924186148131965e-06, + "loss": 3.8561, + "step": 744 + }, + { + "epoch": 0.16, + "learning_rate": 4.924076906270484e-06, + "loss": 3.8263, + "step": 745 + }, + { + "epoch": 0.16, + "learning_rate": 4.923967664409002e-06, + "loss": 3.872, + "step": 746 + }, + { + "epoch": 0.16, + "learning_rate": 4.923858422547521e-06, + "loss": 3.6888, + "step": 747 + }, + { + "epoch": 0.16, + "learning_rate": 4.923749180686039e-06, + "loss": 3.5876, + "step": 748 + }, + { + "epoch": 0.16, + "learning_rate": 4.923639938824558e-06, + "loss": 3.7958, + "step": 749 + }, + { + "epoch": 0.16, + "learning_rate": 4.923530696963076e-06, + "loss": 3.8065, + "step": 750 + }, + { + "epoch": 0.16, + "learning_rate": 4.923421455101595e-06, + "loss": 3.8813, + "step": 751 + }, + { + "epoch": 0.16, + "learning_rate": 4.923312213240114e-06, + "loss": 3.8389, + "step": 752 + }, + { + "epoch": 0.16, + "learning_rate": 4.923202971378633e-06, + "loss": 3.7929, + "step": 753 + }, + { + "epoch": 0.16, + "learning_rate": 4.923093729517152e-06, + "loss": 3.8574, + "step": 754 + }, + { + "epoch": 0.16, + "learning_rate": 4.92298448765567e-06, + "loss": 3.8565, + "step": 755 + }, + { + "epoch": 0.16, + "learning_rate": 4.922875245794189e-06, + "loss": 3.8215, + "step": 756 + }, + { + "epoch": 0.17, + "learning_rate": 4.922766003932707e-06, + "loss": 3.8238, + "step": 757 + }, + { + "epoch": 0.17, + "learning_rate": 4.922656762071226e-06, + "loss": 3.8618, + "step": 758 + }, + { + "epoch": 0.17, + "learning_rate": 4.922547520209745e-06, + "loss": 3.9056, + "step": 759 + }, + { + "epoch": 0.17, + "learning_rate": 4.922438278348263e-06, + "loss": 3.7858, + "step": 760 + }, + { + "epoch": 0.17, + "learning_rate": 4.922329036486782e-06, + "loss": 3.8625, + "step": 761 + }, + { + "epoch": 0.17, + "learning_rate": 4.9222197946253006e-06, + "loss": 3.9666, + "step": 762 + }, + { + "epoch": 0.17, + "learning_rate": 4.92211055276382e-06, + "loss": 3.9034, + "step": 763 + }, + { + "epoch": 0.17, + "learning_rate": 4.922001310902338e-06, + "loss": 3.8253, + "step": 764 + }, + { + "epoch": 0.17, + "learning_rate": 4.921892069040857e-06, + "loss": 3.8266, + "step": 765 + }, + { + "epoch": 0.17, + "learning_rate": 4.921782827179376e-06, + "loss": 3.7116, + "step": 766 + }, + { + "epoch": 0.17, + "learning_rate": 4.921673585317894e-06, + "loss": 3.9431, + "step": 767 + }, + { + "epoch": 0.17, + "learning_rate": 4.921564343456413e-06, + "loss": 3.8859, + "step": 768 + }, + { + "epoch": 0.17, + "learning_rate": 4.921455101594931e-06, + "loss": 3.8661, + "step": 769 + }, + { + "epoch": 0.17, + "learning_rate": 4.92134585973345e-06, + "loss": 3.8554, + "step": 770 + }, + { + "epoch": 0.17, + "learning_rate": 4.921236617871969e-06, + "loss": 3.7519, + "step": 771 + }, + { + "epoch": 0.17, + "learning_rate": 4.9211273760104875e-06, + "loss": 3.8518, + "step": 772 + }, + { + "epoch": 0.17, + "learning_rate": 4.9210181341490066e-06, + "loss": 3.822, + "step": 773 + }, + { + "epoch": 0.17, + "learning_rate": 4.920908892287525e-06, + "loss": 3.8661, + "step": 774 + }, + { + "epoch": 0.17, + "learning_rate": 4.920799650426044e-06, + "loss": 3.9095, + "step": 775 + }, + { + "epoch": 0.17, + "learning_rate": 4.920690408564562e-06, + "loss": 3.7336, + "step": 776 + }, + { + "epoch": 0.17, + "learning_rate": 4.920581166703081e-06, + "loss": 3.6363, + "step": 777 + }, + { + "epoch": 0.17, + "learning_rate": 4.9204719248416e-06, + "loss": 3.7676, + "step": 778 + }, + { + "epoch": 0.17, + "learning_rate": 4.920362682980118e-06, + "loss": 3.9324, + "step": 779 + }, + { + "epoch": 0.17, + "learning_rate": 4.920253441118637e-06, + "loss": 3.9266, + "step": 780 + }, + { + "epoch": 0.17, + "learning_rate": 4.9201441992571554e-06, + "loss": 3.6871, + "step": 781 + }, + { + "epoch": 0.17, + "learning_rate": 4.9200349573956745e-06, + "loss": 3.8937, + "step": 782 + }, + { + "epoch": 0.17, + "learning_rate": 4.919925715534193e-06, + "loss": 4.0023, + "step": 783 + }, + { + "epoch": 0.17, + "learning_rate": 4.919816473672712e-06, + "loss": 3.6869, + "step": 784 + }, + { + "epoch": 0.17, + "learning_rate": 4.919707231811231e-06, + "loss": 3.7317, + "step": 785 + }, + { + "epoch": 0.17, + "learning_rate": 4.919597989949749e-06, + "loss": 3.775, + "step": 786 + }, + { + "epoch": 0.17, + "learning_rate": 4.919488748088268e-06, + "loss": 3.8538, + "step": 787 + }, + { + "epoch": 0.17, + "learning_rate": 4.919379506226786e-06, + "loss": 3.9178, + "step": 788 + }, + { + "epoch": 0.17, + "learning_rate": 4.919270264365305e-06, + "loss": 3.7921, + "step": 789 + }, + { + "epoch": 0.17, + "learning_rate": 4.919161022503824e-06, + "loss": 3.9349, + "step": 790 + }, + { + "epoch": 0.17, + "learning_rate": 4.919051780642342e-06, + "loss": 3.7931, + "step": 791 + }, + { + "epoch": 0.17, + "learning_rate": 4.9189425387808614e-06, + "loss": 3.7758, + "step": 792 + }, + { + "epoch": 0.17, + "learning_rate": 4.91883329691938e-06, + "loss": 3.794, + "step": 793 + }, + { + "epoch": 0.17, + "learning_rate": 4.918724055057899e-06, + "loss": 3.8281, + "step": 794 + }, + { + "epoch": 0.17, + "learning_rate": 4.918614813196418e-06, + "loss": 3.8495, + "step": 795 + }, + { + "epoch": 0.17, + "learning_rate": 4.918505571334936e-06, + "loss": 3.8939, + "step": 796 + }, + { + "epoch": 0.17, + "learning_rate": 4.918396329473454e-06, + "loss": 3.8158, + "step": 797 + }, + { + "epoch": 0.17, + "learning_rate": 4.918287087611973e-06, + "loss": 3.767, + "step": 798 + }, + { + "epoch": 0.17, + "learning_rate": 4.918177845750492e-06, + "loss": 3.8712, + "step": 799 + }, + { + "epoch": 0.17, + "learning_rate": 4.918068603889011e-06, + "loss": 3.6911, + "step": 800 + }, + { + "epoch": 0.17, + "learning_rate": 4.917959362027529e-06, + "loss": 3.9274, + "step": 801 + }, + { + "epoch": 0.18, + "learning_rate": 4.9178501201660475e-06, + "loss": 3.8215, + "step": 802 + }, + { + "epoch": 0.18, + "learning_rate": 4.9177408783045666e-06, + "loss": 3.8299, + "step": 803 + }, + { + "epoch": 0.18, + "learning_rate": 4.917631636443086e-06, + "loss": 3.8859, + "step": 804 + }, + { + "epoch": 0.18, + "learning_rate": 4.917522394581604e-06, + "loss": 3.6591, + "step": 805 + }, + { + "epoch": 0.18, + "learning_rate": 4.917413152720123e-06, + "loss": 3.891, + "step": 806 + }, + { + "epoch": 0.18, + "learning_rate": 4.917303910858641e-06, + "loss": 3.6382, + "step": 807 + }, + { + "epoch": 0.18, + "learning_rate": 4.91719466899716e-06, + "loss": 3.7583, + "step": 808 + }, + { + "epoch": 0.18, + "learning_rate": 4.917085427135679e-06, + "loss": 3.9677, + "step": 809 + }, + { + "epoch": 0.18, + "learning_rate": 4.916976185274197e-06, + "loss": 3.7433, + "step": 810 + }, + { + "epoch": 0.18, + "learning_rate": 4.916866943412716e-06, + "loss": 3.6355, + "step": 811 + }, + { + "epoch": 0.18, + "learning_rate": 4.9167577015512345e-06, + "loss": 3.7758, + "step": 812 + }, + { + "epoch": 0.18, + "learning_rate": 4.9166484596897535e-06, + "loss": 3.7885, + "step": 813 + }, + { + "epoch": 0.18, + "learning_rate": 4.9165392178282726e-06, + "loss": 3.8145, + "step": 814 + }, + { + "epoch": 0.18, + "learning_rate": 4.916429975966791e-06, + "loss": 3.662, + "step": 815 + }, + { + "epoch": 0.18, + "learning_rate": 4.916320734105309e-06, + "loss": 4.028, + "step": 816 + }, + { + "epoch": 0.18, + "learning_rate": 4.916211492243828e-06, + "loss": 3.9631, + "step": 817 + }, + { + "epoch": 0.18, + "learning_rate": 4.916102250382347e-06, + "loss": 3.7208, + "step": 818 + }, + { + "epoch": 0.18, + "learning_rate": 4.915993008520866e-06, + "loss": 3.7869, + "step": 819 + }, + { + "epoch": 0.18, + "learning_rate": 4.915883766659384e-06, + "loss": 3.9345, + "step": 820 + }, + { + "epoch": 0.18, + "learning_rate": 4.915774524797902e-06, + "loss": 3.9706, + "step": 821 + }, + { + "epoch": 0.18, + "learning_rate": 4.9156652829364215e-06, + "loss": 3.8824, + "step": 822 + }, + { + "epoch": 0.18, + "learning_rate": 4.9155560410749405e-06, + "loss": 3.9638, + "step": 823 + }, + { + "epoch": 0.18, + "learning_rate": 4.9154467992134595e-06, + "loss": 3.6332, + "step": 824 + }, + { + "epoch": 0.18, + "learning_rate": 4.915337557351978e-06, + "loss": 3.8557, + "step": 825 + }, + { + "epoch": 0.18, + "learning_rate": 4.915228315490496e-06, + "loss": 3.8614, + "step": 826 + }, + { + "epoch": 0.18, + "learning_rate": 4.915119073629015e-06, + "loss": 3.658, + "step": 827 + }, + { + "epoch": 0.18, + "learning_rate": 4.915009831767534e-06, + "loss": 3.7587, + "step": 828 + }, + { + "epoch": 0.18, + "learning_rate": 4.914900589906052e-06, + "loss": 3.7848, + "step": 829 + }, + { + "epoch": 0.18, + "learning_rate": 4.914791348044571e-06, + "loss": 3.7421, + "step": 830 + }, + { + "epoch": 0.18, + "learning_rate": 4.914682106183089e-06, + "loss": 3.8869, + "step": 831 + }, + { + "epoch": 0.18, + "learning_rate": 4.914572864321608e-06, + "loss": 3.6626, + "step": 832 + }, + { + "epoch": 0.18, + "learning_rate": 4.9144636224601274e-06, + "loss": 3.6879, + "step": 833 + }, + { + "epoch": 0.18, + "learning_rate": 4.914354380598646e-06, + "loss": 3.8029, + "step": 834 + }, + { + "epoch": 0.18, + "learning_rate": 4.914245138737165e-06, + "loss": 3.7634, + "step": 835 + }, + { + "epoch": 0.18, + "learning_rate": 4.914135896875683e-06, + "loss": 3.7513, + "step": 836 + }, + { + "epoch": 0.18, + "learning_rate": 4.914026655014202e-06, + "loss": 3.8029, + "step": 837 + }, + { + "epoch": 0.18, + "learning_rate": 4.913917413152721e-06, + "loss": 3.8314, + "step": 838 + }, + { + "epoch": 0.18, + "learning_rate": 4.913808171291239e-06, + "loss": 3.8354, + "step": 839 + }, + { + "epoch": 0.18, + "learning_rate": 4.913698929429757e-06, + "loss": 3.936, + "step": 840 + }, + { + "epoch": 0.18, + "learning_rate": 4.913589687568276e-06, + "loss": 3.7252, + "step": 841 + }, + { + "epoch": 0.18, + "learning_rate": 4.913480445706795e-06, + "loss": 3.899, + "step": 842 + }, + { + "epoch": 0.18, + "learning_rate": 4.913371203845314e-06, + "loss": 3.8345, + "step": 843 + }, + { + "epoch": 0.18, + "learning_rate": 4.913261961983833e-06, + "loss": 3.8389, + "step": 844 + }, + { + "epoch": 0.18, + "learning_rate": 4.913152720122351e-06, + "loss": 3.8455, + "step": 845 + }, + { + "epoch": 0.18, + "learning_rate": 4.91304347826087e-06, + "loss": 3.8691, + "step": 846 + }, + { + "epoch": 0.18, + "learning_rate": 4.912934236399389e-06, + "loss": 3.6509, + "step": 847 + }, + { + "epoch": 0.19, + "learning_rate": 4.912824994537908e-06, + "loss": 3.9009, + "step": 848 + }, + { + "epoch": 0.19, + "learning_rate": 4.912715752676426e-06, + "loss": 3.6176, + "step": 849 + }, + { + "epoch": 0.19, + "learning_rate": 4.912606510814944e-06, + "loss": 3.8363, + "step": 850 + }, + { + "epoch": 0.19, + "learning_rate": 4.912497268953463e-06, + "loss": 3.9237, + "step": 851 + }, + { + "epoch": 0.19, + "learning_rate": 4.912388027091982e-06, + "loss": 3.8796, + "step": 852 + }, + { + "epoch": 0.19, + "learning_rate": 4.912278785230501e-06, + "loss": 3.7891, + "step": 853 + }, + { + "epoch": 0.19, + "learning_rate": 4.9121695433690195e-06, + "loss": 3.8016, + "step": 854 + }, + { + "epoch": 0.19, + "learning_rate": 4.912060301507538e-06, + "loss": 4.0203, + "step": 855 + }, + { + "epoch": 0.19, + "learning_rate": 4.911951059646057e-06, + "loss": 3.811, + "step": 856 + }, + { + "epoch": 0.19, + "learning_rate": 4.911841817784576e-06, + "loss": 3.8466, + "step": 857 + }, + { + "epoch": 0.19, + "learning_rate": 4.911732575923094e-06, + "loss": 3.8501, + "step": 858 + }, + { + "epoch": 0.19, + "learning_rate": 4.911623334061613e-06, + "loss": 3.7916, + "step": 859 + }, + { + "epoch": 0.19, + "learning_rate": 4.911514092200131e-06, + "loss": 3.7952, + "step": 860 + }, + { + "epoch": 0.19, + "learning_rate": 4.91140485033865e-06, + "loss": 3.6833, + "step": 861 + }, + { + "epoch": 0.19, + "learning_rate": 4.911295608477169e-06, + "loss": 3.7497, + "step": 862 + }, + { + "epoch": 0.19, + "learning_rate": 4.9111863666156875e-06, + "loss": 4.0897, + "step": 863 + }, + { + "epoch": 0.19, + "learning_rate": 4.911077124754206e-06, + "loss": 3.8067, + "step": 864 + }, + { + "epoch": 0.19, + "learning_rate": 4.910967882892725e-06, + "loss": 3.7162, + "step": 865 + }, + { + "epoch": 0.19, + "learning_rate": 4.910858641031244e-06, + "loss": 3.8044, + "step": 866 + }, + { + "epoch": 0.19, + "learning_rate": 4.910749399169763e-06, + "loss": 3.8065, + "step": 867 + }, + { + "epoch": 0.19, + "learning_rate": 4.910640157308281e-06, + "loss": 3.8479, + "step": 868 + }, + { + "epoch": 0.19, + "learning_rate": 4.910530915446799e-06, + "loss": 3.7624, + "step": 869 + }, + { + "epoch": 0.19, + "learning_rate": 4.910421673585318e-06, + "loss": 3.738, + "step": 870 + }, + { + "epoch": 0.19, + "learning_rate": 4.910312431723837e-06, + "loss": 3.96, + "step": 871 + }, + { + "epoch": 0.19, + "learning_rate": 4.910203189862356e-06, + "loss": 3.7141, + "step": 872 + }, + { + "epoch": 0.19, + "learning_rate": 4.9100939480008744e-06, + "loss": 3.7873, + "step": 873 + }, + { + "epoch": 0.19, + "learning_rate": 4.909984706139393e-06, + "loss": 3.7932, + "step": 874 + }, + { + "epoch": 0.19, + "learning_rate": 4.909875464277912e-06, + "loss": 3.5059, + "step": 875 + }, + { + "epoch": 0.19, + "learning_rate": 4.909766222416431e-06, + "loss": 3.7984, + "step": 876 + }, + { + "epoch": 0.19, + "learning_rate": 4.90965698055495e-06, + "loss": 3.8633, + "step": 877 + }, + { + "epoch": 0.19, + "learning_rate": 4.909547738693468e-06, + "loss": 3.8874, + "step": 878 + }, + { + "epoch": 0.19, + "learning_rate": 4.909438496831986e-06, + "loss": 3.8368, + "step": 879 + }, + { + "epoch": 0.19, + "learning_rate": 4.909329254970505e-06, + "loss": 3.6912, + "step": 880 + }, + { + "epoch": 0.19, + "learning_rate": 4.909220013109024e-06, + "loss": 3.7788, + "step": 881 + }, + { + "epoch": 0.19, + "learning_rate": 4.909110771247542e-06, + "loss": 3.7525, + "step": 882 + }, + { + "epoch": 0.19, + "learning_rate": 4.909001529386061e-06, + "loss": 3.8333, + "step": 883 + }, + { + "epoch": 0.19, + "learning_rate": 4.9088922875245796e-06, + "loss": 3.9745, + "step": 884 + }, + { + "epoch": 0.19, + "learning_rate": 4.908783045663099e-06, + "loss": 3.891, + "step": 885 + }, + { + "epoch": 0.19, + "learning_rate": 4.908673803801618e-06, + "loss": 3.7273, + "step": 886 + }, + { + "epoch": 0.19, + "learning_rate": 4.908564561940136e-06, + "loss": 3.7352, + "step": 887 + }, + { + "epoch": 0.19, + "learning_rate": 4.908455320078655e-06, + "loss": 3.7656, + "step": 888 + }, + { + "epoch": 0.19, + "learning_rate": 4.908346078217173e-06, + "loss": 3.7939, + "step": 889 + }, + { + "epoch": 0.19, + "learning_rate": 4.908236836355692e-06, + "loss": 3.751, + "step": 890 + }, + { + "epoch": 0.19, + "learning_rate": 4.908127594494211e-06, + "loss": 3.6963, + "step": 891 + }, + { + "epoch": 0.19, + "learning_rate": 4.908018352632729e-06, + "loss": 3.8949, + "step": 892 + }, + { + "epoch": 0.19, + "learning_rate": 4.9079091107712475e-06, + "loss": 3.7552, + "step": 893 + }, + { + "epoch": 0.2, + "learning_rate": 4.9077998689097665e-06, + "loss": 3.8749, + "step": 894 + }, + { + "epoch": 0.2, + "learning_rate": 4.9076906270482856e-06, + "loss": 3.9012, + "step": 895 + }, + { + "epoch": 0.2, + "learning_rate": 4.907581385186804e-06, + "loss": 3.6155, + "step": 896 + }, + { + "epoch": 0.2, + "learning_rate": 4.907472143325323e-06, + "loss": 3.5561, + "step": 897 + }, + { + "epoch": 0.2, + "learning_rate": 4.907362901463841e-06, + "loss": 3.9041, + "step": 898 + }, + { + "epoch": 0.2, + "learning_rate": 4.90725365960236e-06, + "loss": 3.8927, + "step": 899 + }, + { + "epoch": 0.2, + "learning_rate": 4.907144417740879e-06, + "loss": 3.784, + "step": 900 + }, + { + "epoch": 0.2, + "learning_rate": 4.907035175879397e-06, + "loss": 3.9203, + "step": 901 + }, + { + "epoch": 0.2, + "learning_rate": 4.906925934017916e-06, + "loss": 3.7629, + "step": 902 + }, + { + "epoch": 0.2, + "learning_rate": 4.9068166921564345e-06, + "loss": 3.8931, + "step": 903 + }, + { + "epoch": 0.2, + "learning_rate": 4.9067074502949535e-06, + "loss": 3.8749, + "step": 904 + }, + { + "epoch": 0.2, + "learning_rate": 4.9065982084334725e-06, + "loss": 3.8367, + "step": 905 + }, + { + "epoch": 0.2, + "learning_rate": 4.906488966571991e-06, + "loss": 3.9871, + "step": 906 + }, + { + "epoch": 0.2, + "learning_rate": 4.90637972471051e-06, + "loss": 3.8508, + "step": 907 + }, + { + "epoch": 0.2, + "learning_rate": 4.906270482849028e-06, + "loss": 4.0029, + "step": 908 + }, + { + "epoch": 0.2, + "learning_rate": 4.906161240987547e-06, + "loss": 3.8569, + "step": 909 + }, + { + "epoch": 0.2, + "learning_rate": 4.906051999126066e-06, + "loss": 3.8162, + "step": 910 + }, + { + "epoch": 0.2, + "learning_rate": 4.905942757264584e-06, + "loss": 3.9608, + "step": 911 + }, + { + "epoch": 0.2, + "learning_rate": 4.905833515403103e-06, + "loss": 3.9625, + "step": 912 + }, + { + "epoch": 0.2, + "learning_rate": 4.905724273541621e-06, + "loss": 3.8854, + "step": 913 + }, + { + "epoch": 0.2, + "learning_rate": 4.9056150316801404e-06, + "loss": 3.7548, + "step": 914 + }, + { + "epoch": 0.2, + "learning_rate": 4.905505789818659e-06, + "loss": 3.8908, + "step": 915 + }, + { + "epoch": 0.2, + "learning_rate": 4.905396547957178e-06, + "loss": 4.0928, + "step": 916 + }, + { + "epoch": 0.2, + "learning_rate": 4.905287306095696e-06, + "loss": 3.916, + "step": 917 + }, + { + "epoch": 0.2, + "learning_rate": 4.905178064234215e-06, + "loss": 3.9659, + "step": 918 + }, + { + "epoch": 0.2, + "learning_rate": 4.905068822372734e-06, + "loss": 3.9033, + "step": 919 + }, + { + "epoch": 0.2, + "learning_rate": 4.904959580511252e-06, + "loss": 3.7767, + "step": 920 + }, + { + "epoch": 0.2, + "learning_rate": 4.904850338649771e-06, + "loss": 3.8745, + "step": 921 + }, + { + "epoch": 0.2, + "learning_rate": 4.904741096788289e-06, + "loss": 3.9231, + "step": 922 + }, + { + "epoch": 0.2, + "learning_rate": 4.904631854926808e-06, + "loss": 3.7917, + "step": 923 + }, + { + "epoch": 0.2, + "learning_rate": 4.904522613065327e-06, + "loss": 3.658, + "step": 924 + }, + { + "epoch": 0.2, + "learning_rate": 4.904413371203846e-06, + "loss": 3.7532, + "step": 925 + }, + { + "epoch": 0.2, + "learning_rate": 4.904304129342365e-06, + "loss": 3.8025, + "step": 926 + }, + { + "epoch": 0.2, + "learning_rate": 4.904194887480883e-06, + "loss": 3.7659, + "step": 927 + }, + { + "epoch": 0.2, + "learning_rate": 4.904085645619402e-06, + "loss": 3.9516, + "step": 928 + }, + { + "epoch": 0.2, + "learning_rate": 4.90397640375792e-06, + "loss": 3.8799, + "step": 929 + }, + { + "epoch": 0.2, + "learning_rate": 4.903867161896439e-06, + "loss": 3.9015, + "step": 930 + }, + { + "epoch": 0.2, + "learning_rate": 4.903757920034958e-06, + "loss": 3.7862, + "step": 931 + }, + { + "epoch": 0.2, + "learning_rate": 4.903648678173476e-06, + "loss": 3.8303, + "step": 932 + }, + { + "epoch": 0.2, + "learning_rate": 4.903539436311995e-06, + "loss": 4.0391, + "step": 933 + }, + { + "epoch": 0.2, + "learning_rate": 4.9034301944505135e-06, + "loss": 3.6192, + "step": 934 + }, + { + "epoch": 0.2, + "learning_rate": 4.9033209525890325e-06, + "loss": 3.8644, + "step": 935 + }, + { + "epoch": 0.2, + "learning_rate": 4.903211710727552e-06, + "loss": 3.8557, + "step": 936 + }, + { + "epoch": 0.2, + "learning_rate": 4.90310246886607e-06, + "loss": 3.8571, + "step": 937 + }, + { + "epoch": 0.2, + "learning_rate": 4.902993227004589e-06, + "loss": 3.679, + "step": 938 + }, + { + "epoch": 0.2, + "learning_rate": 4.902883985143107e-06, + "loss": 3.7852, + "step": 939 + }, + { + "epoch": 0.21, + "learning_rate": 4.902774743281626e-06, + "loss": 3.7378, + "step": 940 + }, + { + "epoch": 0.21, + "learning_rate": 4.902665501420145e-06, + "loss": 3.6006, + "step": 941 + }, + { + "epoch": 0.21, + "learning_rate": 4.902556259558663e-06, + "loss": 3.9206, + "step": 942 + }, + { + "epoch": 0.21, + "learning_rate": 4.902447017697182e-06, + "loss": 3.848, + "step": 943 + }, + { + "epoch": 0.21, + "learning_rate": 4.9023377758357005e-06, + "loss": 3.7247, + "step": 944 + }, + { + "epoch": 0.21, + "learning_rate": 4.9022285339742195e-06, + "loss": 3.7971, + "step": 945 + }, + { + "epoch": 0.21, + "learning_rate": 4.902119292112738e-06, + "loss": 3.7623, + "step": 946 + }, + { + "epoch": 0.21, + "learning_rate": 4.902010050251257e-06, + "loss": 3.7144, + "step": 947 + }, + { + "epoch": 0.21, + "learning_rate": 4.901900808389775e-06, + "loss": 3.857, + "step": 948 + }, + { + "epoch": 0.21, + "learning_rate": 4.901791566528294e-06, + "loss": 3.883, + "step": 949 + }, + { + "epoch": 0.21, + "learning_rate": 4.901682324666813e-06, + "loss": 3.8014, + "step": 950 + }, + { + "epoch": 0.21, + "learning_rate": 4.901573082805331e-06, + "loss": 3.8916, + "step": 951 + }, + { + "epoch": 0.21, + "learning_rate": 4.90146384094385e-06, + "loss": 3.8032, + "step": 952 + }, + { + "epoch": 0.21, + "learning_rate": 4.901354599082368e-06, + "loss": 3.6714, + "step": 953 + }, + { + "epoch": 0.21, + "learning_rate": 4.901245357220887e-06, + "loss": 3.88, + "step": 954 + }, + { + "epoch": 0.21, + "learning_rate": 4.9011361153594065e-06, + "loss": 3.9171, + "step": 955 + }, + { + "epoch": 0.21, + "learning_rate": 4.901026873497925e-06, + "loss": 3.944, + "step": 956 + }, + { + "epoch": 0.21, + "learning_rate": 4.900917631636444e-06, + "loss": 3.8677, + "step": 957 + }, + { + "epoch": 0.21, + "learning_rate": 4.900808389774962e-06, + "loss": 3.7508, + "step": 958 + }, + { + "epoch": 0.21, + "learning_rate": 4.900699147913481e-06, + "loss": 3.8504, + "step": 959 + }, + { + "epoch": 0.21, + "learning_rate": 4.900589906052e-06, + "loss": 3.6886, + "step": 960 + }, + { + "epoch": 0.21, + "learning_rate": 4.900480664190518e-06, + "loss": 3.9835, + "step": 961 + }, + { + "epoch": 0.21, + "learning_rate": 4.900371422329036e-06, + "loss": 3.8676, + "step": 962 + }, + { + "epoch": 0.21, + "learning_rate": 4.900262180467555e-06, + "loss": 3.9883, + "step": 963 + }, + { + "epoch": 0.21, + "learning_rate": 4.900152938606074e-06, + "loss": 3.8986, + "step": 964 + }, + { + "epoch": 0.21, + "learning_rate": 4.900043696744593e-06, + "loss": 3.772, + "step": 965 + }, + { + "epoch": 0.21, + "learning_rate": 4.899934454883112e-06, + "loss": 3.9982, + "step": 966 + }, + { + "epoch": 0.21, + "learning_rate": 4.89982521302163e-06, + "loss": 3.8976, + "step": 967 + }, + { + "epoch": 0.21, + "learning_rate": 4.899715971160149e-06, + "loss": 3.8691, + "step": 968 + }, + { + "epoch": 0.21, + "learning_rate": 4.899606729298668e-06, + "loss": 3.8346, + "step": 969 + }, + { + "epoch": 0.21, + "learning_rate": 4.899497487437186e-06, + "loss": 3.6221, + "step": 970 + }, + { + "epoch": 0.21, + "learning_rate": 4.899388245575705e-06, + "loss": 3.746, + "step": 971 + }, + { + "epoch": 0.21, + "learning_rate": 4.899279003714223e-06, + "loss": 3.8542, + "step": 972 + }, + { + "epoch": 0.21, + "learning_rate": 4.899169761852742e-06, + "loss": 3.9176, + "step": 973 + }, + { + "epoch": 0.21, + "learning_rate": 4.899060519991261e-06, + "loss": 3.9239, + "step": 974 + }, + { + "epoch": 0.21, + "learning_rate": 4.8989512781297795e-06, + "loss": 4.0019, + "step": 975 + }, + { + "epoch": 0.21, + "learning_rate": 4.8988420362682986e-06, + "loss": 3.8464, + "step": 976 + }, + { + "epoch": 0.21, + "learning_rate": 4.898732794406817e-06, + "loss": 3.8742, + "step": 977 + }, + { + "epoch": 0.21, + "learning_rate": 4.898623552545336e-06, + "loss": 3.7445, + "step": 978 + }, + { + "epoch": 0.21, + "learning_rate": 4.898514310683855e-06, + "loss": 3.78, + "step": 979 + }, + { + "epoch": 0.21, + "learning_rate": 4.898405068822373e-06, + "loss": 3.843, + "step": 980 + }, + { + "epoch": 0.21, + "learning_rate": 4.898295826960891e-06, + "loss": 3.9601, + "step": 981 + }, + { + "epoch": 0.21, + "learning_rate": 4.89818658509941e-06, + "loss": 3.8846, + "step": 982 + }, + { + "epoch": 0.21, + "learning_rate": 4.898077343237929e-06, + "loss": 4.0123, + "step": 983 + }, + { + "epoch": 0.21, + "learning_rate": 4.897968101376448e-06, + "loss": 3.7367, + "step": 984 + }, + { + "epoch": 0.21, + "learning_rate": 4.8978588595149665e-06, + "loss": 3.9067, + "step": 985 + }, + { + "epoch": 0.22, + "learning_rate": 4.897749617653485e-06, + "loss": 3.6954, + "step": 986 + }, + { + "epoch": 0.22, + "learning_rate": 4.897640375792004e-06, + "loss": 3.7484, + "step": 987 + }, + { + "epoch": 0.22, + "learning_rate": 4.897531133930523e-06, + "loss": 3.7209, + "step": 988 + }, + { + "epoch": 0.22, + "learning_rate": 4.897421892069042e-06, + "loss": 3.6354, + "step": 989 + }, + { + "epoch": 0.22, + "learning_rate": 4.89731265020756e-06, + "loss": 3.7647, + "step": 990 + }, + { + "epoch": 0.22, + "learning_rate": 4.897203408346078e-06, + "loss": 3.5622, + "step": 991 + }, + { + "epoch": 0.22, + "learning_rate": 4.897094166484597e-06, + "loss": 3.7289, + "step": 992 + }, + { + "epoch": 0.22, + "learning_rate": 4.896984924623116e-06, + "loss": 3.8551, + "step": 993 + }, + { + "epoch": 0.22, + "learning_rate": 4.896875682761635e-06, + "loss": 3.9923, + "step": 994 + }, + { + "epoch": 0.22, + "learning_rate": 4.8967664409001534e-06, + "loss": 3.8703, + "step": 995 + }, + { + "epoch": 0.22, + "learning_rate": 4.896657199038672e-06, + "loss": 3.7031, + "step": 996 + }, + { + "epoch": 0.22, + "learning_rate": 4.896547957177191e-06, + "loss": 3.8073, + "step": 997 + }, + { + "epoch": 0.22, + "learning_rate": 4.89643871531571e-06, + "loss": 3.9128, + "step": 998 + }, + { + "epoch": 0.22, + "learning_rate": 4.896329473454228e-06, + "loss": 3.7479, + "step": 999 + }, + { + "epoch": 0.22, + "learning_rate": 4.896220231592747e-06, + "loss": 3.8503, + "step": 1000 + }, + { + "epoch": 0.22, + "learning_rate": 4.896110989731265e-06, + "loss": 3.962, + "step": 1001 + }, + { + "epoch": 0.22, + "learning_rate": 4.896001747869784e-06, + "loss": 3.8572, + "step": 1002 + }, + { + "epoch": 0.22, + "learning_rate": 4.895892506008303e-06, + "loss": 3.817, + "step": 1003 + }, + { + "epoch": 0.22, + "learning_rate": 4.895783264146821e-06, + "loss": 3.8057, + "step": 1004 + }, + { + "epoch": 0.22, + "learning_rate": 4.8956740222853395e-06, + "loss": 3.7017, + "step": 1005 + }, + { + "epoch": 0.22, + "learning_rate": 4.895564780423859e-06, + "loss": 3.8402, + "step": 1006 + }, + { + "epoch": 0.22, + "learning_rate": 4.895455538562378e-06, + "loss": 3.8482, + "step": 1007 + }, + { + "epoch": 0.22, + "learning_rate": 4.895346296700897e-06, + "loss": 3.8109, + "step": 1008 + }, + { + "epoch": 0.22, + "learning_rate": 4.895237054839415e-06, + "loss": 3.9977, + "step": 1009 + }, + { + "epoch": 0.22, + "learning_rate": 4.895127812977933e-06, + "loss": 3.4665, + "step": 1010 + }, + { + "epoch": 0.22, + "learning_rate": 4.895018571116452e-06, + "loss": 3.627, + "step": 1011 + }, + { + "epoch": 0.22, + "learning_rate": 4.894909329254971e-06, + "loss": 3.9823, + "step": 1012 + }, + { + "epoch": 0.22, + "learning_rate": 4.89480008739349e-06, + "loss": 4.1037, + "step": 1013 + }, + { + "epoch": 0.22, + "learning_rate": 4.894690845532008e-06, + "loss": 3.7571, + "step": 1014 + }, + { + "epoch": 0.22, + "learning_rate": 4.8945816036705265e-06, + "loss": 3.7827, + "step": 1015 + }, + { + "epoch": 0.22, + "learning_rate": 4.8944723618090455e-06, + "loss": 3.8467, + "step": 1016 + }, + { + "epoch": 0.22, + "learning_rate": 4.8943631199475646e-06, + "loss": 3.9897, + "step": 1017 + }, + { + "epoch": 0.22, + "learning_rate": 4.894253878086084e-06, + "loss": 3.7966, + "step": 1018 + }, + { + "epoch": 0.22, + "learning_rate": 4.894144636224602e-06, + "loss": 3.5509, + "step": 1019 + }, + { + "epoch": 0.22, + "learning_rate": 4.89403539436312e-06, + "loss": 3.7836, + "step": 1020 + }, + { + "epoch": 0.22, + "learning_rate": 4.893926152501639e-06, + "loss": 3.8003, + "step": 1021 + }, + { + "epoch": 0.22, + "learning_rate": 4.893816910640158e-06, + "loss": 3.9758, + "step": 1022 + }, + { + "epoch": 0.22, + "learning_rate": 4.893707668778676e-06, + "loss": 3.8947, + "step": 1023 + }, + { + "epoch": 0.22, + "learning_rate": 4.893598426917195e-06, + "loss": 3.7218, + "step": 1024 + }, + { + "epoch": 0.22, + "learning_rate": 4.8934891850557135e-06, + "loss": 3.8249, + "step": 1025 + }, + { + "epoch": 0.22, + "learning_rate": 4.8933799431942325e-06, + "loss": 3.8318, + "step": 1026 + }, + { + "epoch": 0.22, + "learning_rate": 4.8932707013327515e-06, + "loss": 3.609, + "step": 1027 + }, + { + "epoch": 0.22, + "learning_rate": 4.89316145947127e-06, + "loss": 3.7192, + "step": 1028 + }, + { + "epoch": 0.22, + "learning_rate": 4.893052217609788e-06, + "loss": 3.715, + "step": 1029 + }, + { + "epoch": 0.22, + "learning_rate": 4.892942975748307e-06, + "loss": 3.76, + "step": 1030 + }, + { + "epoch": 0.23, + "learning_rate": 4.892833733886826e-06, + "loss": 3.8112, + "step": 1031 + }, + { + "epoch": 0.23, + "learning_rate": 4.892724492025345e-06, + "loss": 3.7116, + "step": 1032 + }, + { + "epoch": 0.23, + "learning_rate": 4.892615250163863e-06, + "loss": 3.7414, + "step": 1033 + }, + { + "epoch": 0.23, + "learning_rate": 4.892506008302381e-06, + "loss": 3.627, + "step": 1034 + }, + { + "epoch": 0.23, + "learning_rate": 4.8923967664409e-06, + "loss": 3.6989, + "step": 1035 + }, + { + "epoch": 0.23, + "learning_rate": 4.8922875245794195e-06, + "loss": 3.8156, + "step": 1036 + }, + { + "epoch": 0.23, + "learning_rate": 4.8921782827179385e-06, + "loss": 3.7818, + "step": 1037 + }, + { + "epoch": 0.23, + "learning_rate": 4.892069040856457e-06, + "loss": 3.957, + "step": 1038 + }, + { + "epoch": 0.23, + "learning_rate": 4.891959798994975e-06, + "loss": 3.8293, + "step": 1039 + }, + { + "epoch": 0.23, + "learning_rate": 4.891850557133494e-06, + "loss": 4.0383, + "step": 1040 + }, + { + "epoch": 0.23, + "learning_rate": 4.891741315272013e-06, + "loss": 3.7989, + "step": 1041 + }, + { + "epoch": 0.23, + "learning_rate": 4.891632073410532e-06, + "loss": 3.7456, + "step": 1042 + }, + { + "epoch": 0.23, + "learning_rate": 4.89152283154905e-06, + "loss": 3.8195, + "step": 1043 + }, + { + "epoch": 0.23, + "learning_rate": 4.891413589687568e-06, + "loss": 3.8071, + "step": 1044 + }, + { + "epoch": 0.23, + "learning_rate": 4.891304347826087e-06, + "loss": 3.7081, + "step": 1045 + }, + { + "epoch": 0.23, + "learning_rate": 4.891195105964606e-06, + "loss": 3.7012, + "step": 1046 + }, + { + "epoch": 0.23, + "learning_rate": 4.891085864103125e-06, + "loss": 3.7939, + "step": 1047 + }, + { + "epoch": 0.23, + "learning_rate": 4.890976622241644e-06, + "loss": 3.8187, + "step": 1048 + }, + { + "epoch": 0.23, + "learning_rate": 4.890867380380162e-06, + "loss": 3.7419, + "step": 1049 + }, + { + "epoch": 0.23, + "learning_rate": 4.890758138518681e-06, + "loss": 3.9267, + "step": 1050 + }, + { + "epoch": 0.23, + "learning_rate": 4.8906488966572e-06, + "loss": 3.9358, + "step": 1051 + }, + { + "epoch": 0.23, + "learning_rate": 4.890539654795718e-06, + "loss": 4.0404, + "step": 1052 + }, + { + "epoch": 0.23, + "learning_rate": 4.890430412934237e-06, + "loss": 3.7041, + "step": 1053 + }, + { + "epoch": 0.23, + "learning_rate": 4.890321171072755e-06, + "loss": 3.9037, + "step": 1054 + }, + { + "epoch": 0.23, + "learning_rate": 4.890211929211274e-06, + "loss": 3.7126, + "step": 1055 + }, + { + "epoch": 0.23, + "learning_rate": 4.890102687349793e-06, + "loss": 3.8305, + "step": 1056 + }, + { + "epoch": 0.23, + "learning_rate": 4.8899934454883116e-06, + "loss": 3.6624, + "step": 1057 + }, + { + "epoch": 0.23, + "learning_rate": 4.88988420362683e-06, + "loss": 3.9394, + "step": 1058 + }, + { + "epoch": 0.23, + "learning_rate": 4.889774961765349e-06, + "loss": 3.7976, + "step": 1059 + }, + { + "epoch": 0.23, + "learning_rate": 4.889665719903868e-06, + "loss": 4.0151, + "step": 1060 + }, + { + "epoch": 0.23, + "learning_rate": 4.889556478042386e-06, + "loss": 3.9064, + "step": 1061 + }, + { + "epoch": 0.23, + "learning_rate": 4.889447236180905e-06, + "loss": 3.8418, + "step": 1062 + }, + { + "epoch": 0.23, + "learning_rate": 4.889337994319423e-06, + "loss": 3.9353, + "step": 1063 + }, + { + "epoch": 0.23, + "learning_rate": 4.889228752457942e-06, + "loss": 3.9975, + "step": 1064 + }, + { + "epoch": 0.23, + "learning_rate": 4.889119510596461e-06, + "loss": 3.7685, + "step": 1065 + }, + { + "epoch": 0.23, + "learning_rate": 4.8890102687349795e-06, + "loss": 3.8778, + "step": 1066 + }, + { + "epoch": 0.23, + "learning_rate": 4.8889010268734985e-06, + "loss": 3.6842, + "step": 1067 + }, + { + "epoch": 0.23, + "learning_rate": 4.888791785012017e-06, + "loss": 3.6604, + "step": 1068 + }, + { + "epoch": 0.23, + "learning_rate": 4.888682543150536e-06, + "loss": 3.6296, + "step": 1069 + }, + { + "epoch": 0.23, + "learning_rate": 4.888573301289055e-06, + "loss": 3.8103, + "step": 1070 + }, + { + "epoch": 0.23, + "learning_rate": 4.888464059427573e-06, + "loss": 3.8508, + "step": 1071 + }, + { + "epoch": 0.23, + "learning_rate": 4.888354817566092e-06, + "loss": 3.8921, + "step": 1072 + }, + { + "epoch": 0.23, + "learning_rate": 4.88824557570461e-06, + "loss": 3.7029, + "step": 1073 + }, + { + "epoch": 0.23, + "learning_rate": 4.888136333843129e-06, + "loss": 3.8002, + "step": 1074 + }, + { + "epoch": 0.23, + "learning_rate": 4.888027091981648e-06, + "loss": 3.5784, + "step": 1075 + }, + { + "epoch": 0.23, + "learning_rate": 4.8879178501201664e-06, + "loss": 3.8287, + "step": 1076 + }, + { + "epoch": 0.24, + "learning_rate": 4.8878086082586855e-06, + "loss": 3.8976, + "step": 1077 + }, + { + "epoch": 0.24, + "learning_rate": 4.887699366397204e-06, + "loss": 3.9072, + "step": 1078 + }, + { + "epoch": 0.24, + "learning_rate": 4.887590124535723e-06, + "loss": 3.7498, + "step": 1079 + }, + { + "epoch": 0.24, + "learning_rate": 4.887480882674241e-06, + "loss": 3.9375, + "step": 1080 + }, + { + "epoch": 0.24, + "learning_rate": 4.88737164081276e-06, + "loss": 3.8719, + "step": 1081 + }, + { + "epoch": 0.24, + "learning_rate": 4.887262398951278e-06, + "loss": 3.8071, + "step": 1082 + }, + { + "epoch": 0.24, + "learning_rate": 4.887153157089797e-06, + "loss": 3.7952, + "step": 1083 + }, + { + "epoch": 0.24, + "learning_rate": 4.887043915228316e-06, + "loss": 3.8137, + "step": 1084 + }, + { + "epoch": 0.24, + "learning_rate": 4.886934673366834e-06, + "loss": 3.8434, + "step": 1085 + }, + { + "epoch": 0.24, + "learning_rate": 4.886825431505353e-06, + "loss": 3.6457, + "step": 1086 + }, + { + "epoch": 0.24, + "learning_rate": 4.886716189643872e-06, + "loss": 4.0461, + "step": 1087 + }, + { + "epoch": 0.24, + "learning_rate": 4.886606947782391e-06, + "loss": 3.949, + "step": 1088 + }, + { + "epoch": 0.24, + "learning_rate": 4.88649770592091e-06, + "loss": 4.0906, + "step": 1089 + }, + { + "epoch": 0.24, + "learning_rate": 4.886388464059428e-06, + "loss": 3.6909, + "step": 1090 + }, + { + "epoch": 0.24, + "learning_rate": 4.886279222197947e-06, + "loss": 3.7391, + "step": 1091 + }, + { + "epoch": 0.24, + "learning_rate": 4.886169980336465e-06, + "loss": 3.8957, + "step": 1092 + }, + { + "epoch": 0.24, + "learning_rate": 4.886060738474984e-06, + "loss": 3.9008, + "step": 1093 + }, + { + "epoch": 0.24, + "learning_rate": 4.885951496613502e-06, + "loss": 3.6861, + "step": 1094 + }, + { + "epoch": 0.24, + "learning_rate": 4.885842254752021e-06, + "loss": 3.7347, + "step": 1095 + }, + { + "epoch": 0.24, + "learning_rate": 4.88573301289054e-06, + "loss": 3.7992, + "step": 1096 + }, + { + "epoch": 0.24, + "learning_rate": 4.8856237710290585e-06, + "loss": 3.9184, + "step": 1097 + }, + { + "epoch": 0.24, + "learning_rate": 4.8855145291675776e-06, + "loss": 3.9338, + "step": 1098 + }, + { + "epoch": 0.24, + "learning_rate": 4.885405287306096e-06, + "loss": 3.8939, + "step": 1099 + }, + { + "epoch": 0.24, + "learning_rate": 4.885296045444615e-06, + "loss": 3.6438, + "step": 1100 + }, + { + "epoch": 0.24, + "learning_rate": 4.885186803583134e-06, + "loss": 3.8783, + "step": 1101 + }, + { + "epoch": 0.24, + "learning_rate": 4.885077561721652e-06, + "loss": 3.8972, + "step": 1102 + }, + { + "epoch": 0.24, + "learning_rate": 4.884968319860171e-06, + "loss": 3.9669, + "step": 1103 + }, + { + "epoch": 0.24, + "learning_rate": 4.884859077998689e-06, + "loss": 3.7097, + "step": 1104 + }, + { + "epoch": 0.24, + "learning_rate": 4.884749836137208e-06, + "loss": 4.0338, + "step": 1105 + }, + { + "epoch": 0.24, + "learning_rate": 4.884640594275727e-06, + "loss": 3.823, + "step": 1106 + }, + { + "epoch": 0.24, + "learning_rate": 4.8845313524142455e-06, + "loss": 3.9039, + "step": 1107 + }, + { + "epoch": 0.24, + "learning_rate": 4.8844221105527645e-06, + "loss": 3.8022, + "step": 1108 + }, + { + "epoch": 0.24, + "learning_rate": 4.884312868691283e-06, + "loss": 3.8447, + "step": 1109 + }, + { + "epoch": 0.24, + "learning_rate": 4.884203626829802e-06, + "loss": 3.9291, + "step": 1110 + }, + { + "epoch": 0.24, + "learning_rate": 4.88409438496832e-06, + "loss": 3.753, + "step": 1111 + }, + { + "epoch": 0.24, + "learning_rate": 4.883985143106839e-06, + "loss": 3.6819, + "step": 1112 + }, + { + "epoch": 0.24, + "learning_rate": 4.883875901245357e-06, + "loss": 3.8196, + "step": 1113 + }, + { + "epoch": 0.24, + "learning_rate": 4.883766659383876e-06, + "loss": 3.8452, + "step": 1114 + }, + { + "epoch": 0.24, + "learning_rate": 4.883657417522395e-06, + "loss": 3.6633, + "step": 1115 + }, + { + "epoch": 0.24, + "learning_rate": 4.883548175660913e-06, + "loss": 3.8777, + "step": 1116 + }, + { + "epoch": 0.24, + "learning_rate": 4.8834389337994324e-06, + "loss": 3.8379, + "step": 1117 + }, + { + "epoch": 0.24, + "learning_rate": 4.883329691937951e-06, + "loss": 3.8642, + "step": 1118 + }, + { + "epoch": 0.24, + "learning_rate": 4.88322045007647e-06, + "loss": 3.7725, + "step": 1119 + }, + { + "epoch": 0.24, + "learning_rate": 4.883111208214989e-06, + "loss": 3.9601, + "step": 1120 + }, + { + "epoch": 0.24, + "learning_rate": 4.883001966353507e-06, + "loss": 3.9699, + "step": 1121 + }, + { + "epoch": 0.24, + "learning_rate": 4.882892724492026e-06, + "loss": 3.8512, + "step": 1122 + }, + { + "epoch": 0.25, + "learning_rate": 4.882783482630544e-06, + "loss": 3.9156, + "step": 1123 + }, + { + "epoch": 0.25, + "learning_rate": 4.882674240769063e-06, + "loss": 3.7994, + "step": 1124 + }, + { + "epoch": 0.25, + "learning_rate": 4.882564998907582e-06, + "loss": 3.9497, + "step": 1125 + }, + { + "epoch": 0.25, + "learning_rate": 4.8824557570461e-06, + "loss": 3.8206, + "step": 1126 + }, + { + "epoch": 0.25, + "learning_rate": 4.8823465151846186e-06, + "loss": 3.8695, + "step": 1127 + }, + { + "epoch": 0.25, + "learning_rate": 4.882237273323138e-06, + "loss": 3.7417, + "step": 1128 + }, + { + "epoch": 0.25, + "learning_rate": 4.882128031461657e-06, + "loss": 3.862, + "step": 1129 + }, + { + "epoch": 0.25, + "learning_rate": 4.882018789600176e-06, + "loss": 3.6283, + "step": 1130 + }, + { + "epoch": 0.25, + "learning_rate": 4.881909547738694e-06, + "loss": 3.7525, + "step": 1131 + }, + { + "epoch": 0.25, + "learning_rate": 4.881800305877212e-06, + "loss": 3.9565, + "step": 1132 + }, + { + "epoch": 0.25, + "learning_rate": 4.881691064015731e-06, + "loss": 3.6977, + "step": 1133 + }, + { + "epoch": 0.25, + "learning_rate": 4.88158182215425e-06, + "loss": 3.9077, + "step": 1134 + }, + { + "epoch": 0.25, + "learning_rate": 4.881472580292768e-06, + "loss": 3.8709, + "step": 1135 + }, + { + "epoch": 0.25, + "learning_rate": 4.881363338431287e-06, + "loss": 3.7416, + "step": 1136 + }, + { + "epoch": 0.25, + "learning_rate": 4.8812540965698055e-06, + "loss": 3.8066, + "step": 1137 + }, + { + "epoch": 0.25, + "learning_rate": 4.8811448547083245e-06, + "loss": 3.7422, + "step": 1138 + }, + { + "epoch": 0.25, + "learning_rate": 4.881035612846844e-06, + "loss": 3.7937, + "step": 1139 + }, + { + "epoch": 0.25, + "learning_rate": 4.880926370985362e-06, + "loss": 3.704, + "step": 1140 + }, + { + "epoch": 0.25, + "learning_rate": 4.880817129123881e-06, + "loss": 3.9692, + "step": 1141 + }, + { + "epoch": 0.25, + "learning_rate": 4.880707887262399e-06, + "loss": 3.734, + "step": 1142 + }, + { + "epoch": 0.25, + "learning_rate": 4.880598645400918e-06, + "loss": 3.741, + "step": 1143 + }, + { + "epoch": 0.25, + "learning_rate": 4.880489403539437e-06, + "loss": 3.8282, + "step": 1144 + }, + { + "epoch": 0.25, + "learning_rate": 4.880380161677955e-06, + "loss": 3.6574, + "step": 1145 + }, + { + "epoch": 0.25, + "learning_rate": 4.8802709198164734e-06, + "loss": 3.8738, + "step": 1146 + }, + { + "epoch": 0.25, + "learning_rate": 4.8801616779549925e-06, + "loss": 3.8334, + "step": 1147 + }, + { + "epoch": 0.25, + "learning_rate": 4.8800524360935115e-06, + "loss": 3.8584, + "step": 1148 + }, + { + "epoch": 0.25, + "learning_rate": 4.8799431942320305e-06, + "loss": 3.7734, + "step": 1149 + }, + { + "epoch": 0.25, + "learning_rate": 4.879833952370549e-06, + "loss": 3.8442, + "step": 1150 + }, + { + "epoch": 0.25, + "learning_rate": 4.879724710509067e-06, + "loss": 3.8508, + "step": 1151 + }, + { + "epoch": 0.25, + "learning_rate": 4.879615468647586e-06, + "loss": 3.7811, + "step": 1152 + }, + { + "epoch": 0.25, + "learning_rate": 4.879506226786105e-06, + "loss": 3.9083, + "step": 1153 + }, + { + "epoch": 0.25, + "learning_rate": 4.879396984924624e-06, + "loss": 3.8371, + "step": 1154 + }, + { + "epoch": 0.25, + "learning_rate": 4.879287743063142e-06, + "loss": 3.7129, + "step": 1155 + }, + { + "epoch": 0.25, + "learning_rate": 4.87917850120166e-06, + "loss": 3.8524, + "step": 1156 + }, + { + "epoch": 0.25, + "learning_rate": 4.8790692593401794e-06, + "loss": 4.0367, + "step": 1157 + }, + { + "epoch": 0.25, + "learning_rate": 4.8789600174786985e-06, + "loss": 3.7917, + "step": 1158 + }, + { + "epoch": 0.25, + "learning_rate": 4.8788507756172175e-06, + "loss": 3.8774, + "step": 1159 + }, + { + "epoch": 0.25, + "learning_rate": 4.878741533755736e-06, + "loss": 3.8334, + "step": 1160 + }, + { + "epoch": 0.25, + "learning_rate": 4.878632291894254e-06, + "loss": 3.8, + "step": 1161 + }, + { + "epoch": 0.25, + "learning_rate": 4.878523050032773e-06, + "loss": 3.7011, + "step": 1162 + }, + { + "epoch": 0.25, + "learning_rate": 4.878413808171292e-06, + "loss": 3.6438, + "step": 1163 + }, + { + "epoch": 0.25, + "learning_rate": 4.87830456630981e-06, + "loss": 3.9147, + "step": 1164 + }, + { + "epoch": 0.25, + "learning_rate": 4.878195324448329e-06, + "loss": 3.7532, + "step": 1165 + }, + { + "epoch": 0.25, + "learning_rate": 4.878086082586847e-06, + "loss": 3.5586, + "step": 1166 + }, + { + "epoch": 0.25, + "learning_rate": 4.877976840725366e-06, + "loss": 3.7001, + "step": 1167 + }, + { + "epoch": 0.25, + "learning_rate": 4.877867598863885e-06, + "loss": 3.7447, + "step": 1168 + }, + { + "epoch": 0.26, + "learning_rate": 4.877758357002404e-06, + "loss": 3.8657, + "step": 1169 + }, + { + "epoch": 0.26, + "learning_rate": 4.877649115140922e-06, + "loss": 3.86, + "step": 1170 + }, + { + "epoch": 0.26, + "learning_rate": 4.877539873279441e-06, + "loss": 3.8937, + "step": 1171 + }, + { + "epoch": 0.26, + "learning_rate": 4.87743063141796e-06, + "loss": 3.8704, + "step": 1172 + }, + { + "epoch": 0.26, + "learning_rate": 4.877321389556479e-06, + "loss": 3.7101, + "step": 1173 + }, + { + "epoch": 0.26, + "learning_rate": 4.877212147694997e-06, + "loss": 3.8959, + "step": 1174 + }, + { + "epoch": 0.26, + "learning_rate": 4.877102905833515e-06, + "loss": 3.5963, + "step": 1175 + }, + { + "epoch": 0.26, + "learning_rate": 4.876993663972034e-06, + "loss": 3.7222, + "step": 1176 + }, + { + "epoch": 0.26, + "learning_rate": 4.876884422110553e-06, + "loss": 3.6546, + "step": 1177 + }, + { + "epoch": 0.26, + "learning_rate": 4.876775180249072e-06, + "loss": 4.0285, + "step": 1178 + }, + { + "epoch": 0.26, + "learning_rate": 4.8766659383875906e-06, + "loss": 3.8256, + "step": 1179 + }, + { + "epoch": 0.26, + "learning_rate": 4.876556696526109e-06, + "loss": 3.8373, + "step": 1180 + }, + { + "epoch": 0.26, + "learning_rate": 4.876447454664628e-06, + "loss": 3.8103, + "step": 1181 + }, + { + "epoch": 0.26, + "learning_rate": 4.876338212803147e-06, + "loss": 3.8663, + "step": 1182 + }, + { + "epoch": 0.26, + "learning_rate": 4.876228970941666e-06, + "loss": 3.8506, + "step": 1183 + }, + { + "epoch": 0.26, + "learning_rate": 4.876119729080184e-06, + "loss": 3.7471, + "step": 1184 + }, + { + "epoch": 0.26, + "learning_rate": 4.876010487218702e-06, + "loss": 3.9615, + "step": 1185 + }, + { + "epoch": 0.26, + "learning_rate": 4.875901245357221e-06, + "loss": 3.6604, + "step": 1186 + }, + { + "epoch": 0.26, + "learning_rate": 4.87579200349574e-06, + "loss": 3.8932, + "step": 1187 + }, + { + "epoch": 0.26, + "learning_rate": 4.8756827616342585e-06, + "loss": 3.6873, + "step": 1188 + }, + { + "epoch": 0.26, + "learning_rate": 4.8755735197727775e-06, + "loss": 3.7199, + "step": 1189 + }, + { + "epoch": 0.26, + "learning_rate": 4.875464277911296e-06, + "loss": 3.7912, + "step": 1190 + }, + { + "epoch": 0.26, + "learning_rate": 4.875355036049815e-06, + "loss": 3.8371, + "step": 1191 + }, + { + "epoch": 0.26, + "learning_rate": 4.875245794188334e-06, + "loss": 3.8085, + "step": 1192 + }, + { + "epoch": 0.26, + "learning_rate": 4.875136552326852e-06, + "loss": 3.7513, + "step": 1193 + }, + { + "epoch": 0.26, + "learning_rate": 4.875027310465371e-06, + "loss": 3.81, + "step": 1194 + }, + { + "epoch": 0.26, + "learning_rate": 4.874918068603889e-06, + "loss": 3.637, + "step": 1195 + }, + { + "epoch": 0.26, + "learning_rate": 4.874808826742408e-06, + "loss": 3.7535, + "step": 1196 + }, + { + "epoch": 0.26, + "learning_rate": 4.874699584880927e-06, + "loss": 3.7769, + "step": 1197 + }, + { + "epoch": 0.26, + "learning_rate": 4.8745903430194454e-06, + "loss": 3.6843, + "step": 1198 + }, + { + "epoch": 0.26, + "learning_rate": 4.874481101157964e-06, + "loss": 3.8456, + "step": 1199 + }, + { + "epoch": 0.26, + "learning_rate": 4.874371859296483e-06, + "loss": 3.9739, + "step": 1200 + }, + { + "epoch": 0.26, + "learning_rate": 4.874262617435002e-06, + "loss": 3.7445, + "step": 1201 + }, + { + "epoch": 0.26, + "learning_rate": 4.874153375573521e-06, + "loss": 3.7345, + "step": 1202 + }, + { + "epoch": 0.26, + "learning_rate": 4.874044133712039e-06, + "loss": 3.6947, + "step": 1203 + }, + { + "epoch": 0.26, + "learning_rate": 4.873934891850557e-06, + "loss": 3.7041, + "step": 1204 + }, + { + "epoch": 0.26, + "learning_rate": 4.873825649989076e-06, + "loss": 3.8235, + "step": 1205 + }, + { + "epoch": 0.26, + "learning_rate": 4.873716408127595e-06, + "loss": 3.7317, + "step": 1206 + }, + { + "epoch": 0.26, + "learning_rate": 4.873607166266113e-06, + "loss": 3.801, + "step": 1207 + }, + { + "epoch": 0.26, + "learning_rate": 4.873497924404632e-06, + "loss": 3.7397, + "step": 1208 + }, + { + "epoch": 0.26, + "learning_rate": 4.873388682543151e-06, + "loss": 3.4952, + "step": 1209 + }, + { + "epoch": 0.26, + "learning_rate": 4.87327944068167e-06, + "loss": 3.8596, + "step": 1210 + }, + { + "epoch": 0.26, + "learning_rate": 4.873170198820189e-06, + "loss": 3.7899, + "step": 1211 + }, + { + "epoch": 0.26, + "learning_rate": 4.873060956958707e-06, + "loss": 3.8413, + "step": 1212 + }, + { + "epoch": 0.26, + "learning_rate": 4.872951715097226e-06, + "loss": 3.7096, + "step": 1213 + }, + { + "epoch": 0.26, + "learning_rate": 4.872842473235744e-06, + "loss": 3.6861, + "step": 1214 + }, + { + "epoch": 0.27, + "learning_rate": 4.872733231374263e-06, + "loss": 3.853, + "step": 1215 + }, + { + "epoch": 0.27, + "learning_rate": 4.872623989512782e-06, + "loss": 3.8987, + "step": 1216 + }, + { + "epoch": 0.27, + "learning_rate": 4.8725147476513e-06, + "loss": 3.9274, + "step": 1217 + }, + { + "epoch": 0.27, + "learning_rate": 4.872405505789819e-06, + "loss": 3.8742, + "step": 1218 + }, + { + "epoch": 0.27, + "learning_rate": 4.8722962639283375e-06, + "loss": 3.7417, + "step": 1219 + }, + { + "epoch": 0.27, + "learning_rate": 4.872187022066857e-06, + "loss": 3.7601, + "step": 1220 + }, + { + "epoch": 0.27, + "learning_rate": 4.872077780205376e-06, + "loss": 3.6531, + "step": 1221 + }, + { + "epoch": 0.27, + "learning_rate": 4.871968538343894e-06, + "loss": 3.7702, + "step": 1222 + }, + { + "epoch": 0.27, + "learning_rate": 4.871859296482412e-06, + "loss": 3.8559, + "step": 1223 + }, + { + "epoch": 0.27, + "learning_rate": 4.871750054620931e-06, + "loss": 3.6905, + "step": 1224 + }, + { + "epoch": 0.27, + "learning_rate": 4.87164081275945e-06, + "loss": 3.781, + "step": 1225 + }, + { + "epoch": 0.27, + "learning_rate": 4.871531570897968e-06, + "loss": 4.0107, + "step": 1226 + }, + { + "epoch": 0.27, + "learning_rate": 4.871422329036487e-06, + "loss": 3.8186, + "step": 1227 + }, + { + "epoch": 0.27, + "learning_rate": 4.8713130871750055e-06, + "loss": 3.7902, + "step": 1228 + }, + { + "epoch": 0.27, + "learning_rate": 4.8712038453135245e-06, + "loss": 3.739, + "step": 1229 + }, + { + "epoch": 0.27, + "learning_rate": 4.8710946034520435e-06, + "loss": 3.6587, + "step": 1230 + }, + { + "epoch": 0.27, + "learning_rate": 4.870985361590562e-06, + "loss": 3.6808, + "step": 1231 + }, + { + "epoch": 0.27, + "learning_rate": 4.870876119729081e-06, + "loss": 3.9174, + "step": 1232 + }, + { + "epoch": 0.27, + "learning_rate": 4.870766877867599e-06, + "loss": 3.8432, + "step": 1233 + }, + { + "epoch": 0.27, + "learning_rate": 4.870657636006118e-06, + "loss": 3.6404, + "step": 1234 + }, + { + "epoch": 0.27, + "learning_rate": 4.870548394144637e-06, + "loss": 3.6694, + "step": 1235 + }, + { + "epoch": 0.27, + "learning_rate": 4.870439152283155e-06, + "loss": 3.8245, + "step": 1236 + }, + { + "epoch": 0.27, + "learning_rate": 4.870329910421674e-06, + "loss": 3.7043, + "step": 1237 + }, + { + "epoch": 0.27, + "learning_rate": 4.8702206685601924e-06, + "loss": 3.8435, + "step": 1238 + }, + { + "epoch": 0.27, + "learning_rate": 4.8701114266987115e-06, + "loss": 3.7851, + "step": 1239 + }, + { + "epoch": 0.27, + "learning_rate": 4.87000218483723e-06, + "loss": 3.8991, + "step": 1240 + }, + { + "epoch": 0.27, + "learning_rate": 4.869892942975749e-06, + "loss": 3.882, + "step": 1241 + }, + { + "epoch": 0.27, + "learning_rate": 4.869783701114268e-06, + "loss": 3.8532, + "step": 1242 + }, + { + "epoch": 0.27, + "learning_rate": 4.869674459252786e-06, + "loss": 3.937, + "step": 1243 + }, + { + "epoch": 0.27, + "learning_rate": 4.869565217391305e-06, + "loss": 3.7915, + "step": 1244 + }, + { + "epoch": 0.27, + "learning_rate": 4.869455975529823e-06, + "loss": 3.6787, + "step": 1245 + }, + { + "epoch": 0.27, + "learning_rate": 4.869346733668342e-06, + "loss": 4.085, + "step": 1246 + }, + { + "epoch": 0.27, + "learning_rate": 4.869237491806861e-06, + "loss": 3.7633, + "step": 1247 + }, + { + "epoch": 0.27, + "learning_rate": 4.869128249945379e-06, + "loss": 3.9194, + "step": 1248 + }, + { + "epoch": 0.27, + "learning_rate": 4.869019008083898e-06, + "loss": 3.5564, + "step": 1249 + }, + { + "epoch": 0.27, + "learning_rate": 4.868909766222417e-06, + "loss": 3.8783, + "step": 1250 + }, + { + "epoch": 0.27, + "learning_rate": 4.868800524360936e-06, + "loss": 4.0178, + "step": 1251 + }, + { + "epoch": 0.27, + "learning_rate": 4.868691282499454e-06, + "loss": 3.78, + "step": 1252 + }, + { + "epoch": 0.27, + "learning_rate": 4.868582040637973e-06, + "loss": 3.7034, + "step": 1253 + }, + { + "epoch": 0.27, + "learning_rate": 4.868472798776492e-06, + "loss": 3.5928, + "step": 1254 + }, + { + "epoch": 0.27, + "learning_rate": 4.86836355691501e-06, + "loss": 3.9647, + "step": 1255 + }, + { + "epoch": 0.27, + "learning_rate": 4.868254315053529e-06, + "loss": 3.7753, + "step": 1256 + }, + { + "epoch": 0.27, + "learning_rate": 4.868145073192047e-06, + "loss": 3.6929, + "step": 1257 + }, + { + "epoch": 0.27, + "learning_rate": 4.868035831330566e-06, + "loss": 3.7176, + "step": 1258 + }, + { + "epoch": 0.27, + "learning_rate": 4.8679265894690845e-06, + "loss": 3.9248, + "step": 1259 + }, + { + "epoch": 0.27, + "learning_rate": 4.8678173476076036e-06, + "loss": 3.683, + "step": 1260 + }, + { + "epoch": 0.28, + "learning_rate": 4.867708105746123e-06, + "loss": 3.7056, + "step": 1261 + }, + { + "epoch": 0.28, + "learning_rate": 4.867598863884641e-06, + "loss": 3.8301, + "step": 1262 + }, + { + "epoch": 0.28, + "learning_rate": 4.86748962202316e-06, + "loss": 3.7614, + "step": 1263 + }, + { + "epoch": 0.28, + "learning_rate": 4.867380380161678e-06, + "loss": 3.6658, + "step": 1264 + }, + { + "epoch": 0.28, + "learning_rate": 4.867271138300197e-06, + "loss": 3.6047, + "step": 1265 + }, + { + "epoch": 0.28, + "learning_rate": 4.867161896438716e-06, + "loss": 3.7523, + "step": 1266 + }, + { + "epoch": 0.28, + "learning_rate": 4.867052654577234e-06, + "loss": 3.7517, + "step": 1267 + }, + { + "epoch": 0.28, + "learning_rate": 4.866943412715753e-06, + "loss": 3.5484, + "step": 1268 + }, + { + "epoch": 0.28, + "learning_rate": 4.8668341708542715e-06, + "loss": 3.7863, + "step": 1269 + }, + { + "epoch": 0.28, + "learning_rate": 4.8667249289927905e-06, + "loss": 3.7747, + "step": 1270 + }, + { + "epoch": 0.28, + "learning_rate": 4.8666156871313096e-06, + "loss": 3.7893, + "step": 1271 + }, + { + "epoch": 0.28, + "learning_rate": 4.866506445269828e-06, + "loss": 3.5829, + "step": 1272 + }, + { + "epoch": 0.28, + "learning_rate": 4.866397203408346e-06, + "loss": 3.8182, + "step": 1273 + }, + { + "epoch": 0.28, + "learning_rate": 4.866287961546865e-06, + "loss": 3.8062, + "step": 1274 + }, + { + "epoch": 0.28, + "learning_rate": 4.866178719685384e-06, + "loss": 3.7896, + "step": 1275 + }, + { + "epoch": 0.28, + "learning_rate": 4.866069477823902e-06, + "loss": 3.8684, + "step": 1276 + }, + { + "epoch": 0.28, + "learning_rate": 4.865960235962421e-06, + "loss": 3.7831, + "step": 1277 + }, + { + "epoch": 0.28, + "learning_rate": 4.865850994100939e-06, + "loss": 3.8351, + "step": 1278 + }, + { + "epoch": 0.28, + "learning_rate": 4.8657417522394584e-06, + "loss": 3.7473, + "step": 1279 + }, + { + "epoch": 0.28, + "learning_rate": 4.8656325103779775e-06, + "loss": 3.7758, + "step": 1280 + }, + { + "epoch": 0.28, + "learning_rate": 4.865523268516496e-06, + "loss": 3.7624, + "step": 1281 + }, + { + "epoch": 0.28, + "learning_rate": 4.865414026655015e-06, + "loss": 3.9416, + "step": 1282 + }, + { + "epoch": 0.28, + "learning_rate": 4.865304784793533e-06, + "loss": 3.9855, + "step": 1283 + }, + { + "epoch": 0.28, + "learning_rate": 4.865195542932052e-06, + "loss": 3.8019, + "step": 1284 + }, + { + "epoch": 0.28, + "learning_rate": 4.865086301070571e-06, + "loss": 3.5895, + "step": 1285 + }, + { + "epoch": 0.28, + "learning_rate": 4.864977059209089e-06, + "loss": 3.9698, + "step": 1286 + }, + { + "epoch": 0.28, + "learning_rate": 4.864867817347608e-06, + "loss": 4.1051, + "step": 1287 + }, + { + "epoch": 0.28, + "learning_rate": 4.864758575486126e-06, + "loss": 3.6996, + "step": 1288 + }, + { + "epoch": 0.28, + "learning_rate": 4.864649333624645e-06, + "loss": 3.8049, + "step": 1289 + }, + { + "epoch": 0.28, + "learning_rate": 4.8645400917631644e-06, + "loss": 3.9868, + "step": 1290 + }, + { + "epoch": 0.28, + "learning_rate": 4.864430849901683e-06, + "loss": 3.9767, + "step": 1291 + }, + { + "epoch": 0.28, + "learning_rate": 4.864321608040201e-06, + "loss": 3.7559, + "step": 1292 + }, + { + "epoch": 0.28, + "learning_rate": 4.86421236617872e-06, + "loss": 3.5948, + "step": 1293 + }, + { + "epoch": 0.28, + "learning_rate": 4.864103124317239e-06, + "loss": 3.9312, + "step": 1294 + }, + { + "epoch": 0.28, + "learning_rate": 4.863993882455758e-06, + "loss": 3.6852, + "step": 1295 + }, + { + "epoch": 0.28, + "learning_rate": 4.863884640594276e-06, + "loss": 3.6197, + "step": 1296 + }, + { + "epoch": 0.28, + "learning_rate": 4.863775398732794e-06, + "loss": 3.6934, + "step": 1297 + }, + { + "epoch": 0.28, + "learning_rate": 4.863666156871313e-06, + "loss": 3.9017, + "step": 1298 + }, + { + "epoch": 0.28, + "learning_rate": 4.863556915009832e-06, + "loss": 3.8952, + "step": 1299 + }, + { + "epoch": 0.28, + "learning_rate": 4.863447673148351e-06, + "loss": 3.5423, + "step": 1300 + }, + { + "epoch": 0.28, + "learning_rate": 4.8633384312868696e-06, + "loss": 3.9493, + "step": 1301 + }, + { + "epoch": 0.28, + "learning_rate": 4.863229189425388e-06, + "loss": 3.5357, + "step": 1302 + }, + { + "epoch": 0.28, + "learning_rate": 4.863119947563907e-06, + "loss": 3.998, + "step": 1303 + }, + { + "epoch": 0.28, + "learning_rate": 4.863010705702426e-06, + "loss": 3.8841, + "step": 1304 + }, + { + "epoch": 0.28, + "learning_rate": 4.862901463840944e-06, + "loss": 3.7548, + "step": 1305 + }, + { + "epoch": 0.29, + "learning_rate": 4.862792221979463e-06, + "loss": 3.8295, + "step": 1306 + }, + { + "epoch": 0.29, + "learning_rate": 4.862682980117981e-06, + "loss": 3.8094, + "step": 1307 + }, + { + "epoch": 0.29, + "learning_rate": 4.8625737382565e-06, + "loss": 3.9038, + "step": 1308 + }, + { + "epoch": 0.29, + "learning_rate": 4.862464496395019e-06, + "loss": 3.5713, + "step": 1309 + }, + { + "epoch": 0.29, + "learning_rate": 4.8623552545335375e-06, + "loss": 3.7129, + "step": 1310 + }, + { + "epoch": 0.29, + "learning_rate": 4.862246012672056e-06, + "loss": 3.6332, + "step": 1311 + }, + { + "epoch": 0.29, + "learning_rate": 4.862136770810575e-06, + "loss": 3.7656, + "step": 1312 + }, + { + "epoch": 0.29, + "learning_rate": 4.862027528949094e-06, + "loss": 3.528, + "step": 1313 + }, + { + "epoch": 0.29, + "learning_rate": 4.861918287087613e-06, + "loss": 3.8446, + "step": 1314 + }, + { + "epoch": 0.29, + "learning_rate": 4.861809045226131e-06, + "loss": 3.7813, + "step": 1315 + }, + { + "epoch": 0.29, + "learning_rate": 4.861699803364649e-06, + "loss": 3.7623, + "step": 1316 + }, + { + "epoch": 0.29, + "learning_rate": 4.861590561503168e-06, + "loss": 3.9087, + "step": 1317 + }, + { + "epoch": 0.29, + "learning_rate": 4.861481319641687e-06, + "loss": 3.6358, + "step": 1318 + }, + { + "epoch": 0.29, + "learning_rate": 4.861372077780206e-06, + "loss": 3.7172, + "step": 1319 + }, + { + "epoch": 0.29, + "learning_rate": 4.8612628359187245e-06, + "loss": 3.8747, + "step": 1320 + }, + { + "epoch": 0.29, + "learning_rate": 4.861153594057243e-06, + "loss": 3.6403, + "step": 1321 + }, + { + "epoch": 0.29, + "learning_rate": 4.861044352195762e-06, + "loss": 3.5926, + "step": 1322 + }, + { + "epoch": 0.29, + "learning_rate": 4.860935110334281e-06, + "loss": 3.5635, + "step": 1323 + }, + { + "epoch": 0.29, + "learning_rate": 4.8608258684728e-06, + "loss": 3.9419, + "step": 1324 + }, + { + "epoch": 0.29, + "learning_rate": 4.860716626611318e-06, + "loss": 3.7237, + "step": 1325 + }, + { + "epoch": 0.29, + "learning_rate": 4.860607384749836e-06, + "loss": 3.9111, + "step": 1326 + }, + { + "epoch": 0.29, + "learning_rate": 4.860498142888355e-06, + "loss": 3.7408, + "step": 1327 + }, + { + "epoch": 0.29, + "learning_rate": 4.860388901026874e-06, + "loss": 3.8238, + "step": 1328 + }, + { + "epoch": 0.29, + "learning_rate": 4.860279659165392e-06, + "loss": 3.7238, + "step": 1329 + }, + { + "epoch": 0.29, + "learning_rate": 4.860170417303911e-06, + "loss": 3.7102, + "step": 1330 + }, + { + "epoch": 0.29, + "learning_rate": 4.86006117544243e-06, + "loss": 3.8678, + "step": 1331 + }, + { + "epoch": 0.29, + "learning_rate": 4.859951933580949e-06, + "loss": 3.9838, + "step": 1332 + }, + { + "epoch": 0.29, + "learning_rate": 4.859842691719468e-06, + "loss": 3.7556, + "step": 1333 + }, + { + "epoch": 0.29, + "learning_rate": 4.859733449857986e-06, + "loss": 3.9747, + "step": 1334 + }, + { + "epoch": 0.29, + "learning_rate": 4.859624207996505e-06, + "loss": 3.9773, + "step": 1335 + }, + { + "epoch": 0.29, + "learning_rate": 4.859514966135023e-06, + "loss": 3.6973, + "step": 1336 + }, + { + "epoch": 0.29, + "learning_rate": 4.859405724273542e-06, + "loss": 3.5889, + "step": 1337 + }, + { + "epoch": 0.29, + "learning_rate": 4.859296482412061e-06, + "loss": 3.8393, + "step": 1338 + }, + { + "epoch": 0.29, + "learning_rate": 4.859187240550579e-06, + "loss": 3.6884, + "step": 1339 + }, + { + "epoch": 0.29, + "learning_rate": 4.8590779986890975e-06, + "loss": 3.9394, + "step": 1340 + }, + { + "epoch": 0.29, + "learning_rate": 4.8589687568276166e-06, + "loss": 3.7833, + "step": 1341 + }, + { + "epoch": 0.29, + "learning_rate": 4.858859514966136e-06, + "loss": 3.8721, + "step": 1342 + }, + { + "epoch": 0.29, + "learning_rate": 4.858750273104655e-06, + "loss": 3.7207, + "step": 1343 + }, + { + "epoch": 0.29, + "learning_rate": 4.858641031243173e-06, + "loss": 3.7669, + "step": 1344 + }, + { + "epoch": 0.29, + "learning_rate": 4.858531789381691e-06, + "loss": 3.85, + "step": 1345 + }, + { + "epoch": 0.29, + "learning_rate": 4.85842254752021e-06, + "loss": 3.6927, + "step": 1346 + }, + { + "epoch": 0.29, + "learning_rate": 4.858313305658729e-06, + "loss": 4.1096, + "step": 1347 + }, + { + "epoch": 0.29, + "learning_rate": 4.858204063797248e-06, + "loss": 3.6775, + "step": 1348 + }, + { + "epoch": 0.29, + "learning_rate": 4.858094821935766e-06, + "loss": 3.543, + "step": 1349 + }, + { + "epoch": 0.29, + "learning_rate": 4.8579855800742845e-06, + "loss": 3.7754, + "step": 1350 + }, + { + "epoch": 0.29, + "learning_rate": 4.8578763382128035e-06, + "loss": 3.7727, + "step": 1351 + }, + { + "epoch": 0.3, + "learning_rate": 4.8577670963513225e-06, + "loss": 3.7146, + "step": 1352 + }, + { + "epoch": 0.3, + "learning_rate": 4.857657854489842e-06, + "loss": 3.8646, + "step": 1353 + }, + { + "epoch": 0.3, + "learning_rate": 4.85754861262836e-06, + "loss": 3.6221, + "step": 1354 + }, + { + "epoch": 0.3, + "learning_rate": 4.857439370766878e-06, + "loss": 3.6119, + "step": 1355 + }, + { + "epoch": 0.3, + "learning_rate": 4.857330128905397e-06, + "loss": 3.6047, + "step": 1356 + }, + { + "epoch": 0.3, + "learning_rate": 4.857220887043916e-06, + "loss": 3.8465, + "step": 1357 + }, + { + "epoch": 0.3, + "learning_rate": 4.857111645182434e-06, + "loss": 3.9055, + "step": 1358 + }, + { + "epoch": 0.3, + "learning_rate": 4.857002403320953e-06, + "loss": 3.7843, + "step": 1359 + }, + { + "epoch": 0.3, + "learning_rate": 4.8568931614594714e-06, + "loss": 3.7626, + "step": 1360 + }, + { + "epoch": 0.3, + "learning_rate": 4.8567839195979905e-06, + "loss": 3.6484, + "step": 1361 + }, + { + "epoch": 0.3, + "learning_rate": 4.8566746777365095e-06, + "loss": 3.6422, + "step": 1362 + }, + { + "epoch": 0.3, + "learning_rate": 4.856565435875028e-06, + "loss": 3.9593, + "step": 1363 + }, + { + "epoch": 0.3, + "learning_rate": 4.856456194013546e-06, + "loss": 3.4953, + "step": 1364 + }, + { + "epoch": 0.3, + "learning_rate": 4.856346952152065e-06, + "loss": 3.4476, + "step": 1365 + }, + { + "epoch": 0.3, + "learning_rate": 4.856237710290584e-06, + "loss": 3.8115, + "step": 1366 + }, + { + "epoch": 0.3, + "learning_rate": 4.856128468429103e-06, + "loss": 3.6642, + "step": 1367 + }, + { + "epoch": 0.3, + "learning_rate": 4.856019226567621e-06, + "loss": 3.667, + "step": 1368 + }, + { + "epoch": 0.3, + "learning_rate": 4.855909984706139e-06, + "loss": 3.7842, + "step": 1369 + }, + { + "epoch": 0.3, + "learning_rate": 4.855800742844658e-06, + "loss": 3.5643, + "step": 1370 + }, + { + "epoch": 0.3, + "learning_rate": 4.8556915009831774e-06, + "loss": 3.8969, + "step": 1371 + }, + { + "epoch": 0.3, + "learning_rate": 4.855582259121696e-06, + "loss": 3.9302, + "step": 1372 + }, + { + "epoch": 0.3, + "learning_rate": 4.855473017260215e-06, + "loss": 3.928, + "step": 1373 + }, + { + "epoch": 0.3, + "learning_rate": 4.855363775398733e-06, + "loss": 3.9572, + "step": 1374 + }, + { + "epoch": 0.3, + "learning_rate": 4.855254533537252e-06, + "loss": 3.8251, + "step": 1375 + }, + { + "epoch": 0.3, + "learning_rate": 4.855145291675771e-06, + "loss": 3.776, + "step": 1376 + }, + { + "epoch": 0.3, + "learning_rate": 4.855036049814289e-06, + "loss": 3.7969, + "step": 1377 + }, + { + "epoch": 0.3, + "learning_rate": 4.854926807952808e-06, + "loss": 3.7102, + "step": 1378 + }, + { + "epoch": 0.3, + "learning_rate": 4.854817566091326e-06, + "loss": 3.6825, + "step": 1379 + }, + { + "epoch": 0.3, + "learning_rate": 4.854708324229845e-06, + "loss": 3.772, + "step": 1380 + }, + { + "epoch": 0.3, + "learning_rate": 4.854599082368364e-06, + "loss": 3.6714, + "step": 1381 + }, + { + "epoch": 0.3, + "learning_rate": 4.8544898405068826e-06, + "loss": 3.8814, + "step": 1382 + }, + { + "epoch": 0.3, + "learning_rate": 4.854380598645402e-06, + "loss": 3.5449, + "step": 1383 + }, + { + "epoch": 0.3, + "learning_rate": 4.85427135678392e-06, + "loss": 3.545, + "step": 1384 + }, + { + "epoch": 0.3, + "learning_rate": 4.854162114922439e-06, + "loss": 3.9282, + "step": 1385 + }, + { + "epoch": 0.3, + "learning_rate": 4.854052873060958e-06, + "loss": 3.8901, + "step": 1386 + }, + { + "epoch": 0.3, + "learning_rate": 4.853943631199476e-06, + "loss": 3.6276, + "step": 1387 + }, + { + "epoch": 0.3, + "learning_rate": 4.853834389337995e-06, + "loss": 3.6704, + "step": 1388 + }, + { + "epoch": 0.3, + "learning_rate": 4.853725147476513e-06, + "loss": 3.5129, + "step": 1389 + }, + { + "epoch": 0.3, + "learning_rate": 4.853615905615032e-06, + "loss": 3.7955, + "step": 1390 + }, + { + "epoch": 0.3, + "learning_rate": 4.8535066637535505e-06, + "loss": 3.6866, + "step": 1391 + }, + { + "epoch": 0.3, + "learning_rate": 4.8533974218920695e-06, + "loss": 3.6185, + "step": 1392 + }, + { + "epoch": 0.3, + "learning_rate": 4.853288180030588e-06, + "loss": 3.7008, + "step": 1393 + }, + { + "epoch": 0.3, + "learning_rate": 4.853178938169107e-06, + "loss": 3.8731, + "step": 1394 + }, + { + "epoch": 0.3, + "learning_rate": 4.853069696307626e-06, + "loss": 3.7104, + "step": 1395 + }, + { + "epoch": 0.3, + "learning_rate": 4.852960454446144e-06, + "loss": 3.642, + "step": 1396 + }, + { + "epoch": 0.3, + "learning_rate": 4.852851212584663e-06, + "loss": 3.7784, + "step": 1397 + }, + { + "epoch": 0.31, + "learning_rate": 4.852741970723181e-06, + "loss": 3.7658, + "step": 1398 + }, + { + "epoch": 0.31, + "learning_rate": 4.8526327288617e-06, + "loss": 3.4617, + "step": 1399 + }, + { + "epoch": 0.31, + "learning_rate": 4.852523487000219e-06, + "loss": 3.9536, + "step": 1400 + }, + { + "epoch": 0.31, + "learning_rate": 4.8524142451387375e-06, + "loss": 3.7015, + "step": 1401 + }, + { + "epoch": 0.31, + "learning_rate": 4.8523050032772565e-06, + "loss": 3.7889, + "step": 1402 + }, + { + "epoch": 0.31, + "learning_rate": 4.852195761415775e-06, + "loss": 3.5904, + "step": 1403 + }, + { + "epoch": 0.31, + "learning_rate": 4.852086519554294e-06, + "loss": 3.5399, + "step": 1404 + }, + { + "epoch": 0.31, + "learning_rate": 4.851977277692812e-06, + "loss": 3.7099, + "step": 1405 + }, + { + "epoch": 0.31, + "learning_rate": 4.851868035831331e-06, + "loss": 3.6747, + "step": 1406 + }, + { + "epoch": 0.31, + "learning_rate": 4.85175879396985e-06, + "loss": 3.5686, + "step": 1407 + }, + { + "epoch": 0.31, + "learning_rate": 4.851649552108368e-06, + "loss": 3.629, + "step": 1408 + }, + { + "epoch": 0.31, + "learning_rate": 4.851540310246887e-06, + "loss": 3.6434, + "step": 1409 + }, + { + "epoch": 0.31, + "learning_rate": 4.851431068385405e-06, + "loss": 3.7598, + "step": 1410 + }, + { + "epoch": 0.31, + "learning_rate": 4.851321826523924e-06, + "loss": 3.7398, + "step": 1411 + }, + { + "epoch": 0.31, + "learning_rate": 4.8512125846624434e-06, + "loss": 3.8209, + "step": 1412 + }, + { + "epoch": 0.31, + "learning_rate": 4.851103342800962e-06, + "loss": 3.8025, + "step": 1413 + }, + { + "epoch": 0.31, + "learning_rate": 4.850994100939481e-06, + "loss": 3.688, + "step": 1414 + }, + { + "epoch": 0.31, + "learning_rate": 4.850884859077999e-06, + "loss": 3.56, + "step": 1415 + }, + { + "epoch": 0.31, + "learning_rate": 4.850775617216518e-06, + "loss": 3.777, + "step": 1416 + }, + { + "epoch": 0.31, + "learning_rate": 4.850666375355036e-06, + "loss": 3.7379, + "step": 1417 + }, + { + "epoch": 0.31, + "learning_rate": 4.850557133493555e-06, + "loss": 3.6667, + "step": 1418 + }, + { + "epoch": 0.31, + "learning_rate": 4.850447891632074e-06, + "loss": 3.7156, + "step": 1419 + }, + { + "epoch": 0.31, + "learning_rate": 4.850338649770592e-06, + "loss": 3.759, + "step": 1420 + }, + { + "epoch": 0.31, + "learning_rate": 4.850229407909111e-06, + "loss": 3.7407, + "step": 1421 + }, + { + "epoch": 0.31, + "learning_rate": 4.8501201660476296e-06, + "loss": 3.8336, + "step": 1422 + }, + { + "epoch": 0.31, + "learning_rate": 4.850010924186149e-06, + "loss": 3.6209, + "step": 1423 + }, + { + "epoch": 0.31, + "learning_rate": 4.849901682324667e-06, + "loss": 3.7566, + "step": 1424 + }, + { + "epoch": 0.31, + "learning_rate": 4.849792440463186e-06, + "loss": 3.7821, + "step": 1425 + }, + { + "epoch": 0.31, + "learning_rate": 4.849683198601705e-06, + "loss": 3.6111, + "step": 1426 + }, + { + "epoch": 0.31, + "learning_rate": 4.849573956740223e-06, + "loss": 3.8188, + "step": 1427 + }, + { + "epoch": 0.31, + "learning_rate": 4.849464714878742e-06, + "loss": 3.6108, + "step": 1428 + }, + { + "epoch": 0.31, + "learning_rate": 4.84935547301726e-06, + "loss": 3.6472, + "step": 1429 + }, + { + "epoch": 0.31, + "learning_rate": 4.849246231155779e-06, + "loss": 4.0051, + "step": 1430 + }, + { + "epoch": 0.31, + "learning_rate": 4.849136989294298e-06, + "loss": 3.7036, + "step": 1431 + }, + { + "epoch": 0.31, + "learning_rate": 4.8490277474328165e-06, + "loss": 3.6965, + "step": 1432 + }, + { + "epoch": 0.31, + "learning_rate": 4.8489185055713355e-06, + "loss": 3.7413, + "step": 1433 + }, + { + "epoch": 0.31, + "learning_rate": 4.848809263709854e-06, + "loss": 3.6796, + "step": 1434 + }, + { + "epoch": 0.31, + "learning_rate": 4.848700021848373e-06, + "loss": 3.7765, + "step": 1435 + }, + { + "epoch": 0.31, + "learning_rate": 4.848590779986892e-06, + "loss": 3.6902, + "step": 1436 + }, + { + "epoch": 0.31, + "learning_rate": 4.84848153812541e-06, + "loss": 3.6609, + "step": 1437 + }, + { + "epoch": 0.31, + "learning_rate": 4.848372296263928e-06, + "loss": 3.5762, + "step": 1438 + }, + { + "epoch": 0.31, + "learning_rate": 4.848263054402447e-06, + "loss": 3.6651, + "step": 1439 + }, + { + "epoch": 0.31, + "learning_rate": 4.848153812540966e-06, + "loss": 3.8225, + "step": 1440 + }, + { + "epoch": 0.31, + "learning_rate": 4.848044570679485e-06, + "loss": 3.7922, + "step": 1441 + }, + { + "epoch": 0.31, + "learning_rate": 4.8479353288180035e-06, + "loss": 3.897, + "step": 1442 + }, + { + "epoch": 0.31, + "learning_rate": 4.847826086956522e-06, + "loss": 3.7736, + "step": 1443 + }, + { + "epoch": 0.32, + "learning_rate": 4.847716845095041e-06, + "loss": 3.7476, + "step": 1444 + }, + { + "epoch": 0.32, + "learning_rate": 4.84760760323356e-06, + "loss": 3.5143, + "step": 1445 + }, + { + "epoch": 0.32, + "learning_rate": 4.847498361372078e-06, + "loss": 3.7188, + "step": 1446 + }, + { + "epoch": 0.32, + "learning_rate": 4.847389119510597e-06, + "loss": 3.8854, + "step": 1447 + }, + { + "epoch": 0.32, + "learning_rate": 4.847279877649115e-06, + "loss": 3.6463, + "step": 1448 + }, + { + "epoch": 0.32, + "learning_rate": 4.847170635787634e-06, + "loss": 3.7153, + "step": 1449 + }, + { + "epoch": 0.32, + "learning_rate": 4.847061393926153e-06, + "loss": 3.4279, + "step": 1450 + }, + { + "epoch": 0.32, + "learning_rate": 4.846952152064671e-06, + "loss": 3.8447, + "step": 1451 + }, + { + "epoch": 0.32, + "learning_rate": 4.84684291020319e-06, + "loss": 3.8947, + "step": 1452 + }, + { + "epoch": 0.32, + "learning_rate": 4.846733668341709e-06, + "loss": 3.9435, + "step": 1453 + }, + { + "epoch": 0.32, + "learning_rate": 4.846624426480228e-06, + "loss": 3.7394, + "step": 1454 + }, + { + "epoch": 0.32, + "learning_rate": 4.846515184618747e-06, + "loss": 3.7077, + "step": 1455 + }, + { + "epoch": 0.32, + "learning_rate": 4.846405942757265e-06, + "loss": 3.9567, + "step": 1456 + }, + { + "epoch": 0.32, + "learning_rate": 4.846296700895783e-06, + "loss": 3.542, + "step": 1457 + }, + { + "epoch": 0.32, + "learning_rate": 4.846187459034302e-06, + "loss": 3.717, + "step": 1458 + }, + { + "epoch": 0.32, + "learning_rate": 4.846078217172821e-06, + "loss": 3.9051, + "step": 1459 + }, + { + "epoch": 0.32, + "learning_rate": 4.84596897531134e-06, + "loss": 3.8337, + "step": 1460 + }, + { + "epoch": 0.32, + "learning_rate": 4.845859733449858e-06, + "loss": 3.5226, + "step": 1461 + }, + { + "epoch": 0.32, + "learning_rate": 4.8457504915883765e-06, + "loss": 3.6482, + "step": 1462 + }, + { + "epoch": 0.32, + "learning_rate": 4.8456412497268956e-06, + "loss": 3.6374, + "step": 1463 + }, + { + "epoch": 0.32, + "learning_rate": 4.845532007865415e-06, + "loss": 3.6533, + "step": 1464 + }, + { + "epoch": 0.32, + "learning_rate": 4.845422766003934e-06, + "loss": 4.0278, + "step": 1465 + }, + { + "epoch": 0.32, + "learning_rate": 4.845313524142452e-06, + "loss": 3.7638, + "step": 1466 + }, + { + "epoch": 0.32, + "learning_rate": 4.84520428228097e-06, + "loss": 3.5476, + "step": 1467 + }, + { + "epoch": 0.32, + "learning_rate": 4.845095040419489e-06, + "loss": 3.8271, + "step": 1468 + }, + { + "epoch": 0.32, + "learning_rate": 4.844985798558008e-06, + "loss": 3.7168, + "step": 1469 + }, + { + "epoch": 0.32, + "learning_rate": 4.844876556696526e-06, + "loss": 3.8332, + "step": 1470 + }, + { + "epoch": 0.32, + "learning_rate": 4.844767314835045e-06, + "loss": 3.7267, + "step": 1471 + }, + { + "epoch": 0.32, + "learning_rate": 4.8446580729735635e-06, + "loss": 3.9044, + "step": 1472 + }, + { + "epoch": 0.32, + "learning_rate": 4.8445488311120825e-06, + "loss": 3.7606, + "step": 1473 + }, + { + "epoch": 0.32, + "learning_rate": 4.8444395892506016e-06, + "loss": 3.695, + "step": 1474 + }, + { + "epoch": 0.32, + "learning_rate": 4.84433034738912e-06, + "loss": 3.7827, + "step": 1475 + }, + { + "epoch": 0.32, + "learning_rate": 4.844221105527638e-06, + "loss": 3.7476, + "step": 1476 + }, + { + "epoch": 0.32, + "learning_rate": 4.844111863666157e-06, + "loss": 3.7231, + "step": 1477 + }, + { + "epoch": 0.32, + "learning_rate": 4.844002621804676e-06, + "loss": 3.7576, + "step": 1478 + }, + { + "epoch": 0.32, + "learning_rate": 4.843893379943195e-06, + "loss": 3.7514, + "step": 1479 + }, + { + "epoch": 0.32, + "learning_rate": 4.843784138081713e-06, + "loss": 3.851, + "step": 1480 + }, + { + "epoch": 0.32, + "learning_rate": 4.843674896220231e-06, + "loss": 3.9691, + "step": 1481 + }, + { + "epoch": 0.32, + "learning_rate": 4.8435656543587504e-06, + "loss": 3.7557, + "step": 1482 + }, + { + "epoch": 0.32, + "learning_rate": 4.8434564124972695e-06, + "loss": 3.5779, + "step": 1483 + }, + { + "epoch": 0.32, + "learning_rate": 4.8433471706357885e-06, + "loss": 3.604, + "step": 1484 + }, + { + "epoch": 0.32, + "learning_rate": 4.843237928774307e-06, + "loss": 3.6907, + "step": 1485 + }, + { + "epoch": 0.32, + "learning_rate": 4.843128686912825e-06, + "loss": 3.6242, + "step": 1486 + }, + { + "epoch": 0.32, + "learning_rate": 4.843019445051344e-06, + "loss": 3.5746, + "step": 1487 + }, + { + "epoch": 0.32, + "learning_rate": 4.842910203189863e-06, + "loss": 3.6949, + "step": 1488 + }, + { + "epoch": 0.32, + "learning_rate": 4.842800961328382e-06, + "loss": 3.6198, + "step": 1489 + }, + { + "epoch": 0.33, + "learning_rate": 4.8426917194669e-06, + "loss": 3.8766, + "step": 1490 + }, + { + "epoch": 0.33, + "learning_rate": 4.842582477605418e-06, + "loss": 3.8244, + "step": 1491 + }, + { + "epoch": 0.33, + "learning_rate": 4.842473235743937e-06, + "loss": 3.499, + "step": 1492 + }, + { + "epoch": 0.33, + "learning_rate": 4.8423639938824564e-06, + "loss": 3.8955, + "step": 1493 + }, + { + "epoch": 0.33, + "learning_rate": 4.8422547520209755e-06, + "loss": 3.6991, + "step": 1494 + }, + { + "epoch": 0.33, + "learning_rate": 4.842145510159494e-06, + "loss": 3.9498, + "step": 1495 + }, + { + "epoch": 0.33, + "learning_rate": 4.842036268298012e-06, + "loss": 3.8573, + "step": 1496 + }, + { + "epoch": 0.33, + "learning_rate": 4.841927026436531e-06, + "loss": 3.6281, + "step": 1497 + }, + { + "epoch": 0.33, + "learning_rate": 4.84181778457505e-06, + "loss": 3.8841, + "step": 1498 + }, + { + "epoch": 0.33, + "learning_rate": 4.841708542713568e-06, + "loss": 3.4978, + "step": 1499 + }, + { + "epoch": 0.33, + "learning_rate": 4.841599300852087e-06, + "loss": 3.8289, + "step": 1500 + }, + { + "epoch": 0.33, + "learning_rate": 4.841490058990605e-06, + "loss": 3.8734, + "step": 1501 + }, + { + "epoch": 0.33, + "learning_rate": 4.841380817129124e-06, + "loss": 3.5542, + "step": 1502 + }, + { + "epoch": 0.33, + "learning_rate": 4.841271575267643e-06, + "loss": 3.8102, + "step": 1503 + }, + { + "epoch": 0.33, + "learning_rate": 4.841162333406162e-06, + "loss": 3.607, + "step": 1504 + }, + { + "epoch": 0.33, + "learning_rate": 4.84105309154468e-06, + "loss": 3.8187, + "step": 1505 + }, + { + "epoch": 0.33, + "learning_rate": 4.840943849683199e-06, + "loss": 3.6827, + "step": 1506 + }, + { + "epoch": 0.33, + "learning_rate": 4.840834607821718e-06, + "loss": 3.9298, + "step": 1507 + }, + { + "epoch": 0.33, + "learning_rate": 4.840725365960237e-06, + "loss": 3.7278, + "step": 1508 + }, + { + "epoch": 0.33, + "learning_rate": 4.840616124098755e-06, + "loss": 3.9047, + "step": 1509 + }, + { + "epoch": 0.33, + "learning_rate": 4.840506882237273e-06, + "loss": 3.4682, + "step": 1510 + }, + { + "epoch": 0.33, + "learning_rate": 4.840397640375792e-06, + "loss": 3.7216, + "step": 1511 + }, + { + "epoch": 0.33, + "learning_rate": 4.840288398514311e-06, + "loss": 3.8947, + "step": 1512 + }, + { + "epoch": 0.33, + "learning_rate": 4.84017915665283e-06, + "loss": 3.7844, + "step": 1513 + }, + { + "epoch": 0.33, + "learning_rate": 4.8400699147913485e-06, + "loss": 3.5588, + "step": 1514 + }, + { + "epoch": 0.33, + "learning_rate": 4.839960672929867e-06, + "loss": 3.965, + "step": 1515 + }, + { + "epoch": 0.33, + "learning_rate": 4.839851431068386e-06, + "loss": 3.4965, + "step": 1516 + }, + { + "epoch": 0.33, + "learning_rate": 4.839742189206905e-06, + "loss": 3.6623, + "step": 1517 + }, + { + "epoch": 0.33, + "learning_rate": 4.839632947345424e-06, + "loss": 3.6055, + "step": 1518 + }, + { + "epoch": 0.33, + "learning_rate": 4.839523705483942e-06, + "loss": 3.5225, + "step": 1519 + }, + { + "epoch": 0.33, + "learning_rate": 4.83941446362246e-06, + "loss": 3.3884, + "step": 1520 + }, + { + "epoch": 0.33, + "learning_rate": 4.839305221760979e-06, + "loss": 3.8721, + "step": 1521 + }, + { + "epoch": 0.33, + "learning_rate": 4.839195979899498e-06, + "loss": 3.6812, + "step": 1522 + }, + { + "epoch": 0.33, + "learning_rate": 4.8390867380380165e-06, + "loss": 3.7714, + "step": 1523 + }, + { + "epoch": 0.33, + "learning_rate": 4.8389774961765355e-06, + "loss": 3.4451, + "step": 1524 + }, + { + "epoch": 0.33, + "learning_rate": 4.838868254315054e-06, + "loss": 3.5867, + "step": 1525 + }, + { + "epoch": 0.33, + "learning_rate": 4.838759012453573e-06, + "loss": 3.7716, + "step": 1526 + }, + { + "epoch": 0.33, + "learning_rate": 4.838649770592092e-06, + "loss": 3.8283, + "step": 1527 + }, + { + "epoch": 0.33, + "learning_rate": 4.83854052873061e-06, + "loss": 4.0143, + "step": 1528 + }, + { + "epoch": 0.33, + "learning_rate": 4.838431286869129e-06, + "loss": 3.771, + "step": 1529 + }, + { + "epoch": 0.33, + "learning_rate": 4.838322045007647e-06, + "loss": 3.7249, + "step": 1530 + }, + { + "epoch": 0.33, + "learning_rate": 4.838212803146166e-06, + "loss": 3.8485, + "step": 1531 + }, + { + "epoch": 0.33, + "learning_rate": 4.838103561284685e-06, + "loss": 3.9023, + "step": 1532 + }, + { + "epoch": 0.33, + "learning_rate": 4.837994319423203e-06, + "loss": 3.7074, + "step": 1533 + }, + { + "epoch": 0.33, + "learning_rate": 4.837885077561722e-06, + "loss": 3.6409, + "step": 1534 + }, + { + "epoch": 0.34, + "learning_rate": 4.837775835700241e-06, + "loss": 4.1062, + "step": 1535 + }, + { + "epoch": 0.34, + "learning_rate": 4.83766659383876e-06, + "loss": 3.4936, + "step": 1536 + }, + { + "epoch": 0.34, + "learning_rate": 4.837557351977278e-06, + "loss": 3.722, + "step": 1537 + }, + { + "epoch": 0.34, + "learning_rate": 4.837448110115797e-06, + "loss": 3.8119, + "step": 1538 + }, + { + "epoch": 0.34, + "learning_rate": 4.837338868254315e-06, + "loss": 3.7094, + "step": 1539 + }, + { + "epoch": 0.34, + "learning_rate": 4.837229626392834e-06, + "loss": 3.7391, + "step": 1540 + }, + { + "epoch": 0.34, + "learning_rate": 4.837120384531353e-06, + "loss": 3.5298, + "step": 1541 + }, + { + "epoch": 0.34, + "learning_rate": 4.837011142669871e-06, + "loss": 3.6244, + "step": 1542 + }, + { + "epoch": 0.34, + "learning_rate": 4.83690190080839e-06, + "loss": 3.7161, + "step": 1543 + }, + { + "epoch": 0.34, + "learning_rate": 4.8367926589469086e-06, + "loss": 3.5507, + "step": 1544 + }, + { + "epoch": 0.34, + "learning_rate": 4.836683417085428e-06, + "loss": 3.7964, + "step": 1545 + }, + { + "epoch": 0.34, + "learning_rate": 4.836574175223947e-06, + "loss": 3.7846, + "step": 1546 + }, + { + "epoch": 0.34, + "learning_rate": 4.836464933362465e-06, + "loss": 3.7534, + "step": 1547 + }, + { + "epoch": 0.34, + "learning_rate": 4.836355691500984e-06, + "loss": 3.8869, + "step": 1548 + }, + { + "epoch": 0.34, + "learning_rate": 4.836246449639502e-06, + "loss": 3.7519, + "step": 1549 + }, + { + "epoch": 0.34, + "learning_rate": 4.836137207778021e-06, + "loss": 3.6586, + "step": 1550 + }, + { + "epoch": 0.34, + "learning_rate": 4.83602796591654e-06, + "loss": 3.8449, + "step": 1551 + }, + { + "epoch": 0.34, + "learning_rate": 4.835918724055058e-06, + "loss": 3.7563, + "step": 1552 + }, + { + "epoch": 0.34, + "learning_rate": 4.835809482193577e-06, + "loss": 3.6597, + "step": 1553 + }, + { + "epoch": 0.34, + "learning_rate": 4.8357002403320955e-06, + "loss": 3.916, + "step": 1554 + }, + { + "epoch": 0.34, + "learning_rate": 4.8355909984706146e-06, + "loss": 3.5971, + "step": 1555 + }, + { + "epoch": 0.34, + "learning_rate": 4.835481756609133e-06, + "loss": 3.6854, + "step": 1556 + }, + { + "epoch": 0.34, + "learning_rate": 4.835372514747652e-06, + "loss": 3.6715, + "step": 1557 + }, + { + "epoch": 0.34, + "learning_rate": 4.83526327288617e-06, + "loss": 3.5202, + "step": 1558 + }, + { + "epoch": 0.34, + "learning_rate": 4.835154031024689e-06, + "loss": 3.8539, + "step": 1559 + }, + { + "epoch": 0.34, + "learning_rate": 4.835044789163208e-06, + "loss": 3.6622, + "step": 1560 + }, + { + "epoch": 0.34, + "learning_rate": 4.834935547301726e-06, + "loss": 3.9618, + "step": 1561 + }, + { + "epoch": 0.34, + "learning_rate": 4.834826305440245e-06, + "loss": 3.8436, + "step": 1562 + }, + { + "epoch": 0.34, + "learning_rate": 4.8347170635787634e-06, + "loss": 3.8112, + "step": 1563 + }, + { + "epoch": 0.34, + "learning_rate": 4.8346078217172825e-06, + "loss": 3.7106, + "step": 1564 + }, + { + "epoch": 0.34, + "learning_rate": 4.8344985798558015e-06, + "loss": 3.8633, + "step": 1565 + }, + { + "epoch": 0.34, + "learning_rate": 4.83438933799432e-06, + "loss": 3.7242, + "step": 1566 + }, + { + "epoch": 0.34, + "learning_rate": 4.834280096132839e-06, + "loss": 3.7821, + "step": 1567 + }, + { + "epoch": 0.34, + "learning_rate": 4.834170854271357e-06, + "loss": 3.9429, + "step": 1568 + }, + { + "epoch": 0.34, + "learning_rate": 4.834061612409876e-06, + "loss": 3.7879, + "step": 1569 + }, + { + "epoch": 0.34, + "learning_rate": 4.833952370548394e-06, + "loss": 3.6748, + "step": 1570 + }, + { + "epoch": 0.34, + "learning_rate": 4.833843128686913e-06, + "loss": 3.9087, + "step": 1571 + }, + { + "epoch": 0.34, + "learning_rate": 4.833733886825432e-06, + "loss": 3.8225, + "step": 1572 + }, + { + "epoch": 0.34, + "learning_rate": 4.83362464496395e-06, + "loss": 3.7982, + "step": 1573 + }, + { + "epoch": 0.34, + "learning_rate": 4.8335154031024694e-06, + "loss": 3.598, + "step": 1574 + }, + { + "epoch": 0.34, + "learning_rate": 4.833406161240988e-06, + "loss": 3.7789, + "step": 1575 + }, + { + "epoch": 0.34, + "learning_rate": 4.833296919379507e-06, + "loss": 3.6101, + "step": 1576 + }, + { + "epoch": 0.34, + "learning_rate": 4.833187677518026e-06, + "loss": 3.5705, + "step": 1577 + }, + { + "epoch": 0.34, + "learning_rate": 4.833078435656544e-06, + "loss": 3.8358, + "step": 1578 + }, + { + "epoch": 0.34, + "learning_rate": 4.832969193795063e-06, + "loss": 3.8354, + "step": 1579 + }, + { + "epoch": 0.34, + "learning_rate": 4.832859951933581e-06, + "loss": 3.5214, + "step": 1580 + }, + { + "epoch": 0.35, + "learning_rate": 4.8327507100721e-06, + "loss": 3.6454, + "step": 1581 + }, + { + "epoch": 0.35, + "learning_rate": 4.832641468210619e-06, + "loss": 3.7476, + "step": 1582 + }, + { + "epoch": 0.35, + "learning_rate": 4.832532226349137e-06, + "loss": 3.8113, + "step": 1583 + }, + { + "epoch": 0.35, + "learning_rate": 4.8324229844876555e-06, + "loss": 3.889, + "step": 1584 + }, + { + "epoch": 0.35, + "learning_rate": 4.832313742626175e-06, + "loss": 3.79, + "step": 1585 + }, + { + "epoch": 0.35, + "learning_rate": 4.832204500764694e-06, + "loss": 3.6159, + "step": 1586 + }, + { + "epoch": 0.35, + "learning_rate": 4.832095258903212e-06, + "loss": 3.7379, + "step": 1587 + }, + { + "epoch": 0.35, + "learning_rate": 4.831986017041731e-06, + "loss": 3.6083, + "step": 1588 + }, + { + "epoch": 0.35, + "learning_rate": 4.831876775180249e-06, + "loss": 3.7317, + "step": 1589 + }, + { + "epoch": 0.35, + "learning_rate": 4.831767533318768e-06, + "loss": 3.9552, + "step": 1590 + }, + { + "epoch": 0.35, + "learning_rate": 4.831658291457287e-06, + "loss": 3.8184, + "step": 1591 + }, + { + "epoch": 0.35, + "learning_rate": 4.831549049595805e-06, + "loss": 3.5094, + "step": 1592 + }, + { + "epoch": 0.35, + "learning_rate": 4.831439807734324e-06, + "loss": 3.8145, + "step": 1593 + }, + { + "epoch": 0.35, + "learning_rate": 4.8313305658728425e-06, + "loss": 3.7717, + "step": 1594 + }, + { + "epoch": 0.35, + "learning_rate": 4.8312213240113615e-06, + "loss": 3.9141, + "step": 1595 + }, + { + "epoch": 0.35, + "learning_rate": 4.8311120821498806e-06, + "loss": 3.9597, + "step": 1596 + }, + { + "epoch": 0.35, + "learning_rate": 4.831002840288399e-06, + "loss": 4.0133, + "step": 1597 + }, + { + "epoch": 0.35, + "learning_rate": 4.830893598426918e-06, + "loss": 3.737, + "step": 1598 + }, + { + "epoch": 0.35, + "learning_rate": 4.830784356565436e-06, + "loss": 3.7582, + "step": 1599 + }, + { + "epoch": 0.35, + "learning_rate": 4.830675114703955e-06, + "loss": 3.6533, + "step": 1600 + }, + { + "epoch": 0.35, + "learning_rate": 4.830565872842474e-06, + "loss": 3.8127, + "step": 1601 + }, + { + "epoch": 0.35, + "learning_rate": 4.830456630980992e-06, + "loss": 3.6288, + "step": 1602 + }, + { + "epoch": 0.35, + "learning_rate": 4.83034738911951e-06, + "loss": 3.764, + "step": 1603 + }, + { + "epoch": 0.35, + "learning_rate": 4.8302381472580295e-06, + "loss": 3.8473, + "step": 1604 + }, + { + "epoch": 0.35, + "learning_rate": 4.8301289053965485e-06, + "loss": 3.8346, + "step": 1605 + }, + { + "epoch": 0.35, + "learning_rate": 4.8300196635350675e-06, + "loss": 3.7782, + "step": 1606 + }, + { + "epoch": 0.35, + "learning_rate": 4.829910421673586e-06, + "loss": 3.7297, + "step": 1607 + }, + { + "epoch": 0.35, + "learning_rate": 4.829801179812104e-06, + "loss": 3.7168, + "step": 1608 + }, + { + "epoch": 0.35, + "learning_rate": 4.829691937950623e-06, + "loss": 3.7437, + "step": 1609 + }, + { + "epoch": 0.35, + "learning_rate": 4.829582696089142e-06, + "loss": 3.68, + "step": 1610 + }, + { + "epoch": 0.35, + "learning_rate": 4.82947345422766e-06, + "loss": 3.6131, + "step": 1611 + }, + { + "epoch": 0.35, + "learning_rate": 4.829364212366179e-06, + "loss": 3.855, + "step": 1612 + }, + { + "epoch": 0.35, + "learning_rate": 4.829254970504697e-06, + "loss": 3.6658, + "step": 1613 + }, + { + "epoch": 0.35, + "learning_rate": 4.829145728643216e-06, + "loss": 3.5119, + "step": 1614 + }, + { + "epoch": 0.35, + "learning_rate": 4.8290364867817354e-06, + "loss": 3.6968, + "step": 1615 + }, + { + "epoch": 0.35, + "learning_rate": 4.828927244920254e-06, + "loss": 3.936, + "step": 1616 + }, + { + "epoch": 0.35, + "learning_rate": 4.828818003058772e-06, + "loss": 3.8482, + "step": 1617 + }, + { + "epoch": 0.35, + "learning_rate": 4.828708761197291e-06, + "loss": 3.5419, + "step": 1618 + }, + { + "epoch": 0.35, + "learning_rate": 4.82859951933581e-06, + "loss": 3.8896, + "step": 1619 + }, + { + "epoch": 0.35, + "learning_rate": 4.828490277474329e-06, + "loss": 3.6928, + "step": 1620 + }, + { + "epoch": 0.35, + "learning_rate": 4.828381035612847e-06, + "loss": 3.7345, + "step": 1621 + }, + { + "epoch": 0.35, + "learning_rate": 4.828271793751365e-06, + "loss": 3.7323, + "step": 1622 + }, + { + "epoch": 0.35, + "learning_rate": 4.828162551889884e-06, + "loss": 3.8286, + "step": 1623 + }, + { + "epoch": 0.35, + "learning_rate": 4.828053310028403e-06, + "loss": 3.835, + "step": 1624 + }, + { + "epoch": 0.35, + "learning_rate": 4.827944068166922e-06, + "loss": 3.7393, + "step": 1625 + }, + { + "epoch": 0.35, + "learning_rate": 4.827834826305441e-06, + "loss": 3.7918, + "step": 1626 + }, + { + "epoch": 0.36, + "learning_rate": 4.827725584443959e-06, + "loss": 3.8471, + "step": 1627 + }, + { + "epoch": 0.36, + "learning_rate": 4.827616342582478e-06, + "loss": 3.8109, + "step": 1628 + }, + { + "epoch": 0.36, + "learning_rate": 4.827507100720997e-06, + "loss": 3.8728, + "step": 1629 + }, + { + "epoch": 0.36, + "learning_rate": 4.827397858859516e-06, + "loss": 3.8121, + "step": 1630 + }, + { + "epoch": 0.36, + "learning_rate": 4.827288616998034e-06, + "loss": 3.6943, + "step": 1631 + }, + { + "epoch": 0.36, + "learning_rate": 4.827179375136552e-06, + "loss": 3.4566, + "step": 1632 + }, + { + "epoch": 0.36, + "learning_rate": 4.827070133275071e-06, + "loss": 3.5071, + "step": 1633 + }, + { + "epoch": 0.36, + "learning_rate": 4.82696089141359e-06, + "loss": 3.9201, + "step": 1634 + }, + { + "epoch": 0.36, + "learning_rate": 4.826851649552109e-06, + "loss": 3.7224, + "step": 1635 + }, + { + "epoch": 0.36, + "learning_rate": 4.8267424076906275e-06, + "loss": 3.7312, + "step": 1636 + }, + { + "epoch": 0.36, + "learning_rate": 4.826633165829146e-06, + "loss": 3.2861, + "step": 1637 + }, + { + "epoch": 0.36, + "learning_rate": 4.826523923967665e-06, + "loss": 3.7556, + "step": 1638 + }, + { + "epoch": 0.36, + "learning_rate": 4.826414682106184e-06, + "loss": 3.6274, + "step": 1639 + }, + { + "epoch": 0.36, + "learning_rate": 4.826305440244702e-06, + "loss": 3.5059, + "step": 1640 + }, + { + "epoch": 0.36, + "learning_rate": 4.826196198383221e-06, + "loss": 3.8223, + "step": 1641 + }, + { + "epoch": 0.36, + "learning_rate": 4.826086956521739e-06, + "loss": 3.8677, + "step": 1642 + }, + { + "epoch": 0.36, + "learning_rate": 4.825977714660258e-06, + "loss": 3.6338, + "step": 1643 + }, + { + "epoch": 0.36, + "learning_rate": 4.825868472798777e-06, + "loss": 3.7472, + "step": 1644 + }, + { + "epoch": 0.36, + "learning_rate": 4.8257592309372955e-06, + "loss": 3.8191, + "step": 1645 + }, + { + "epoch": 0.36, + "learning_rate": 4.825649989075814e-06, + "loss": 3.8378, + "step": 1646 + }, + { + "epoch": 0.36, + "learning_rate": 4.825540747214333e-06, + "loss": 3.6955, + "step": 1647 + }, + { + "epoch": 0.36, + "learning_rate": 4.825431505352852e-06, + "loss": 3.9555, + "step": 1648 + }, + { + "epoch": 0.36, + "learning_rate": 4.825322263491371e-06, + "loss": 3.6706, + "step": 1649 + }, + { + "epoch": 0.36, + "learning_rate": 4.825213021629889e-06, + "loss": 4.0465, + "step": 1650 + }, + { + "epoch": 0.36, + "learning_rate": 4.825103779768407e-06, + "loss": 3.8842, + "step": 1651 + }, + { + "epoch": 0.36, + "learning_rate": 4.824994537906926e-06, + "loss": 3.7399, + "step": 1652 + }, + { + "epoch": 0.36, + "learning_rate": 4.824885296045445e-06, + "loss": 3.8419, + "step": 1653 + }, + { + "epoch": 0.36, + "learning_rate": 4.824776054183964e-06, + "loss": 3.9552, + "step": 1654 + }, + { + "epoch": 0.36, + "learning_rate": 4.8246668123224824e-06, + "loss": 3.6743, + "step": 1655 + }, + { + "epoch": 0.36, + "learning_rate": 4.824557570461001e-06, + "loss": 3.4528, + "step": 1656 + }, + { + "epoch": 0.36, + "learning_rate": 4.82444832859952e-06, + "loss": 3.8931, + "step": 1657 + }, + { + "epoch": 0.36, + "learning_rate": 4.824339086738039e-06, + "loss": 3.5572, + "step": 1658 + }, + { + "epoch": 0.36, + "learning_rate": 4.824229844876558e-06, + "loss": 3.6559, + "step": 1659 + }, + { + "epoch": 0.36, + "learning_rate": 4.824120603015076e-06, + "loss": 3.9183, + "step": 1660 + }, + { + "epoch": 0.36, + "learning_rate": 4.824011361153594e-06, + "loss": 3.5353, + "step": 1661 + }, + { + "epoch": 0.36, + "learning_rate": 4.823902119292113e-06, + "loss": 3.8576, + "step": 1662 + }, + { + "epoch": 0.36, + "learning_rate": 4.823792877430632e-06, + "loss": 3.8432, + "step": 1663 + }, + { + "epoch": 0.36, + "learning_rate": 4.82368363556915e-06, + "loss": 3.6464, + "step": 1664 + }, + { + "epoch": 0.36, + "learning_rate": 4.823574393707669e-06, + "loss": 3.5679, + "step": 1665 + }, + { + "epoch": 0.36, + "learning_rate": 4.8234651518461876e-06, + "loss": 3.5837, + "step": 1666 + }, + { + "epoch": 0.36, + "learning_rate": 4.823355909984707e-06, + "loss": 3.51, + "step": 1667 + }, + { + "epoch": 0.36, + "learning_rate": 4.823246668123226e-06, + "loss": 3.7715, + "step": 1668 + }, + { + "epoch": 0.36, + "learning_rate": 4.823137426261744e-06, + "loss": 3.6561, + "step": 1669 + }, + { + "epoch": 0.36, + "learning_rate": 4.823028184400262e-06, + "loss": 3.5608, + "step": 1670 + }, + { + "epoch": 0.36, + "learning_rate": 4.822918942538781e-06, + "loss": 3.7513, + "step": 1671 + }, + { + "epoch": 0.36, + "learning_rate": 4.8228097006773e-06, + "loss": 3.6189, + "step": 1672 + }, + { + "epoch": 0.37, + "learning_rate": 4.822700458815819e-06, + "loss": 3.3341, + "step": 1673 + }, + { + "epoch": 0.37, + "learning_rate": 4.822591216954337e-06, + "loss": 3.5284, + "step": 1674 + }, + { + "epoch": 0.37, + "learning_rate": 4.8224819750928555e-06, + "loss": 3.5679, + "step": 1675 + }, + { + "epoch": 0.37, + "learning_rate": 4.8223727332313745e-06, + "loss": 3.8, + "step": 1676 + }, + { + "epoch": 0.37, + "learning_rate": 4.8222634913698936e-06, + "loss": 3.7302, + "step": 1677 + }, + { + "epoch": 0.37, + "learning_rate": 4.822154249508413e-06, + "loss": 3.6458, + "step": 1678 + }, + { + "epoch": 0.37, + "learning_rate": 4.822045007646931e-06, + "loss": 3.8975, + "step": 1679 + }, + { + "epoch": 0.37, + "learning_rate": 4.821935765785449e-06, + "loss": 3.5341, + "step": 1680 + }, + { + "epoch": 0.37, + "learning_rate": 4.821826523923968e-06, + "loss": 3.7153, + "step": 1681 + }, + { + "epoch": 0.37, + "learning_rate": 4.821717282062487e-06, + "loss": 3.6707, + "step": 1682 + }, + { + "epoch": 0.37, + "learning_rate": 4.821608040201005e-06, + "loss": 3.6631, + "step": 1683 + }, + { + "epoch": 0.37, + "learning_rate": 4.821498798339524e-06, + "loss": 3.777, + "step": 1684 + }, + { + "epoch": 0.37, + "learning_rate": 4.8213895564780425e-06, + "loss": 3.6542, + "step": 1685 + }, + { + "epoch": 0.37, + "learning_rate": 4.8212803146165615e-06, + "loss": 3.5969, + "step": 1686 + }, + { + "epoch": 0.37, + "learning_rate": 4.8211710727550805e-06, + "loss": 3.6294, + "step": 1687 + }, + { + "epoch": 0.37, + "learning_rate": 4.821061830893599e-06, + "loss": 4.0633, + "step": 1688 + }, + { + "epoch": 0.37, + "learning_rate": 4.820952589032118e-06, + "loss": 3.872, + "step": 1689 + }, + { + "epoch": 0.37, + "learning_rate": 4.820843347170636e-06, + "loss": 3.6629, + "step": 1690 + }, + { + "epoch": 0.37, + "learning_rate": 4.820734105309155e-06, + "loss": 3.4608, + "step": 1691 + }, + { + "epoch": 0.37, + "learning_rate": 4.820624863447674e-06, + "loss": 3.757, + "step": 1692 + }, + { + "epoch": 0.37, + "learning_rate": 4.820515621586192e-06, + "loss": 3.4951, + "step": 1693 + }, + { + "epoch": 0.37, + "learning_rate": 4.820406379724711e-06, + "loss": 3.8873, + "step": 1694 + }, + { + "epoch": 0.37, + "learning_rate": 4.820297137863229e-06, + "loss": 3.4916, + "step": 1695 + }, + { + "epoch": 0.37, + "learning_rate": 4.8201878960017484e-06, + "loss": 3.4587, + "step": 1696 + }, + { + "epoch": 0.37, + "learning_rate": 4.8200786541402675e-06, + "loss": 3.547, + "step": 1697 + }, + { + "epoch": 0.37, + "learning_rate": 4.819969412278786e-06, + "loss": 3.5855, + "step": 1698 + }, + { + "epoch": 0.37, + "learning_rate": 4.819860170417304e-06, + "loss": 3.5541, + "step": 1699 + }, + { + "epoch": 0.37, + "learning_rate": 4.819750928555823e-06, + "loss": 3.7772, + "step": 1700 + }, + { + "epoch": 0.37, + "learning_rate": 4.819641686694342e-06, + "loss": 3.8535, + "step": 1701 + }, + { + "epoch": 0.37, + "learning_rate": 4.81953244483286e-06, + "loss": 3.5424, + "step": 1702 + }, + { + "epoch": 0.37, + "learning_rate": 4.819423202971379e-06, + "loss": 3.6528, + "step": 1703 + }, + { + "epoch": 0.37, + "learning_rate": 4.819313961109897e-06, + "loss": 3.6069, + "step": 1704 + }, + { + "epoch": 0.37, + "learning_rate": 4.819204719248416e-06, + "loss": 3.6886, + "step": 1705 + }, + { + "epoch": 0.37, + "learning_rate": 4.819095477386935e-06, + "loss": 3.7286, + "step": 1706 + }, + { + "epoch": 0.37, + "learning_rate": 4.818986235525454e-06, + "loss": 3.7637, + "step": 1707 + }, + { + "epoch": 0.37, + "learning_rate": 4.818876993663973e-06, + "loss": 3.7933, + "step": 1708 + }, + { + "epoch": 0.37, + "learning_rate": 4.818767751802491e-06, + "loss": 3.5615, + "step": 1709 + }, + { + "epoch": 0.37, + "learning_rate": 4.81865850994101e-06, + "loss": 3.6717, + "step": 1710 + }, + { + "epoch": 0.37, + "learning_rate": 4.818549268079529e-06, + "loss": 3.9453, + "step": 1711 + }, + { + "epoch": 0.37, + "learning_rate": 4.818440026218047e-06, + "loss": 3.6173, + "step": 1712 + }, + { + "epoch": 0.37, + "learning_rate": 4.818330784356566e-06, + "loss": 3.5514, + "step": 1713 + }, + { + "epoch": 0.37, + "learning_rate": 4.818221542495084e-06, + "loss": 3.6066, + "step": 1714 + }, + { + "epoch": 0.37, + "learning_rate": 4.818112300633603e-06, + "loss": 3.7285, + "step": 1715 + }, + { + "epoch": 0.37, + "learning_rate": 4.8180030587721215e-06, + "loss": 3.7479, + "step": 1716 + }, + { + "epoch": 0.37, + "learning_rate": 4.8178938169106405e-06, + "loss": 3.9138, + "step": 1717 + }, + { + "epoch": 0.37, + "learning_rate": 4.81778457504916e-06, + "loss": 3.5096, + "step": 1718 + }, + { + "epoch": 0.38, + "learning_rate": 4.817675333187678e-06, + "loss": 3.8314, + "step": 1719 + }, + { + "epoch": 0.38, + "learning_rate": 4.817566091326197e-06, + "loss": 3.69, + "step": 1720 + }, + { + "epoch": 0.38, + "learning_rate": 4.817456849464715e-06, + "loss": 3.8611, + "step": 1721 + }, + { + "epoch": 0.38, + "learning_rate": 4.817347607603234e-06, + "loss": 3.5695, + "step": 1722 + }, + { + "epoch": 0.38, + "learning_rate": 4.817238365741752e-06, + "loss": 3.7354, + "step": 1723 + }, + { + "epoch": 0.38, + "learning_rate": 4.817129123880271e-06, + "loss": 3.864, + "step": 1724 + }, + { + "epoch": 0.38, + "learning_rate": 4.81701988201879e-06, + "loss": 3.8289, + "step": 1725 + }, + { + "epoch": 0.38, + "learning_rate": 4.8169106401573085e-06, + "loss": 3.8194, + "step": 1726 + }, + { + "epoch": 0.38, + "learning_rate": 4.8168013982958275e-06, + "loss": 3.5707, + "step": 1727 + }, + { + "epoch": 0.38, + "learning_rate": 4.816692156434346e-06, + "loss": 3.8149, + "step": 1728 + }, + { + "epoch": 0.38, + "learning_rate": 4.816582914572865e-06, + "loss": 3.6228, + "step": 1729 + }, + { + "epoch": 0.38, + "learning_rate": 4.816473672711384e-06, + "loss": 4.0357, + "step": 1730 + }, + { + "epoch": 0.38, + "learning_rate": 4.816364430849902e-06, + "loss": 3.6146, + "step": 1731 + }, + { + "epoch": 0.38, + "learning_rate": 4.816255188988421e-06, + "loss": 3.696, + "step": 1732 + }, + { + "epoch": 0.38, + "learning_rate": 4.816145947126939e-06, + "loss": 3.7328, + "step": 1733 + }, + { + "epoch": 0.38, + "learning_rate": 4.816036705265458e-06, + "loss": 3.8288, + "step": 1734 + }, + { + "epoch": 0.38, + "learning_rate": 4.815927463403976e-06, + "loss": 3.6335, + "step": 1735 + }, + { + "epoch": 0.38, + "learning_rate": 4.8158182215424954e-06, + "loss": 3.6176, + "step": 1736 + }, + { + "epoch": 0.38, + "learning_rate": 4.8157089796810145e-06, + "loss": 3.6807, + "step": 1737 + }, + { + "epoch": 0.38, + "learning_rate": 4.815599737819533e-06, + "loss": 3.4324, + "step": 1738 + }, + { + "epoch": 0.38, + "learning_rate": 4.815490495958052e-06, + "loss": 3.5295, + "step": 1739 + }, + { + "epoch": 0.38, + "learning_rate": 4.81538125409657e-06, + "loss": 3.6453, + "step": 1740 + }, + { + "epoch": 0.38, + "learning_rate": 4.815272012235089e-06, + "loss": 3.7023, + "step": 1741 + }, + { + "epoch": 0.38, + "learning_rate": 4.815162770373608e-06, + "loss": 3.8459, + "step": 1742 + }, + { + "epoch": 0.38, + "learning_rate": 4.815053528512126e-06, + "loss": 3.7901, + "step": 1743 + }, + { + "epoch": 0.38, + "learning_rate": 4.814944286650645e-06, + "loss": 3.8872, + "step": 1744 + }, + { + "epoch": 0.38, + "learning_rate": 4.814835044789163e-06, + "loss": 3.678, + "step": 1745 + }, + { + "epoch": 0.38, + "learning_rate": 4.814725802927682e-06, + "loss": 3.5315, + "step": 1746 + }, + { + "epoch": 0.38, + "learning_rate": 4.814616561066201e-06, + "loss": 3.5403, + "step": 1747 + }, + { + "epoch": 0.38, + "learning_rate": 4.81450731920472e-06, + "loss": 3.9074, + "step": 1748 + }, + { + "epoch": 0.38, + "learning_rate": 4.814398077343238e-06, + "loss": 3.5119, + "step": 1749 + }, + { + "epoch": 0.38, + "learning_rate": 4.814288835481757e-06, + "loss": 3.5807, + "step": 1750 + }, + { + "epoch": 0.38, + "learning_rate": 4.814179593620276e-06, + "loss": 3.8486, + "step": 1751 + }, + { + "epoch": 0.38, + "learning_rate": 4.814070351758794e-06, + "loss": 3.5085, + "step": 1752 + }, + { + "epoch": 0.38, + "learning_rate": 4.813961109897313e-06, + "loss": 3.4339, + "step": 1753 + }, + { + "epoch": 0.38, + "learning_rate": 4.813851868035831e-06, + "loss": 3.5212, + "step": 1754 + }, + { + "epoch": 0.38, + "learning_rate": 4.81374262617435e-06, + "loss": 3.9362, + "step": 1755 + }, + { + "epoch": 0.38, + "learning_rate": 4.813633384312869e-06, + "loss": 3.5144, + "step": 1756 + }, + { + "epoch": 0.38, + "learning_rate": 4.8135241424513875e-06, + "loss": 3.5993, + "step": 1757 + }, + { + "epoch": 0.38, + "learning_rate": 4.8134149005899066e-06, + "loss": 3.638, + "step": 1758 + }, + { + "epoch": 0.38, + "learning_rate": 4.813305658728425e-06, + "loss": 3.7806, + "step": 1759 + }, + { + "epoch": 0.38, + "learning_rate": 4.813196416866944e-06, + "loss": 3.7423, + "step": 1760 + }, + { + "epoch": 0.38, + "learning_rate": 4.813087175005463e-06, + "loss": 3.9446, + "step": 1761 + }, + { + "epoch": 0.38, + "learning_rate": 4.812977933143981e-06, + "loss": 4.0902, + "step": 1762 + }, + { + "epoch": 0.38, + "learning_rate": 4.8128686912825e-06, + "loss": 3.7585, + "step": 1763 + }, + { + "epoch": 0.38, + "learning_rate": 4.812759449421018e-06, + "loss": 3.7436, + "step": 1764 + }, + { + "epoch": 0.39, + "learning_rate": 4.812650207559537e-06, + "loss": 3.9128, + "step": 1765 + }, + { + "epoch": 0.39, + "learning_rate": 4.812540965698056e-06, + "loss": 3.7438, + "step": 1766 + }, + { + "epoch": 0.39, + "learning_rate": 4.8124317238365745e-06, + "loss": 3.8503, + "step": 1767 + }, + { + "epoch": 0.39, + "learning_rate": 4.812322481975093e-06, + "loss": 3.8886, + "step": 1768 + }, + { + "epoch": 0.39, + "learning_rate": 4.812213240113612e-06, + "loss": 3.6544, + "step": 1769 + }, + { + "epoch": 0.39, + "learning_rate": 4.812103998252131e-06, + "loss": 3.6976, + "step": 1770 + }, + { + "epoch": 0.39, + "learning_rate": 4.81199475639065e-06, + "loss": 3.5725, + "step": 1771 + }, + { + "epoch": 0.39, + "learning_rate": 4.811885514529168e-06, + "loss": 4.0122, + "step": 1772 + }, + { + "epoch": 0.39, + "learning_rate": 4.811776272667686e-06, + "loss": 3.5435, + "step": 1773 + }, + { + "epoch": 0.39, + "learning_rate": 4.811667030806205e-06, + "loss": 3.6285, + "step": 1774 + }, + { + "epoch": 0.39, + "learning_rate": 4.811557788944724e-06, + "loss": 3.6532, + "step": 1775 + }, + { + "epoch": 0.39, + "learning_rate": 4.811448547083242e-06, + "loss": 3.7826, + "step": 1776 + }, + { + "epoch": 0.39, + "learning_rate": 4.8113393052217614e-06, + "loss": 3.6646, + "step": 1777 + }, + { + "epoch": 0.39, + "learning_rate": 4.81123006336028e-06, + "loss": 3.8134, + "step": 1778 + }, + { + "epoch": 0.39, + "learning_rate": 4.811120821498799e-06, + "loss": 3.7072, + "step": 1779 + }, + { + "epoch": 0.39, + "learning_rate": 4.811011579637318e-06, + "loss": 3.7391, + "step": 1780 + }, + { + "epoch": 0.39, + "learning_rate": 4.810902337775836e-06, + "loss": 3.5749, + "step": 1781 + }, + { + "epoch": 0.39, + "learning_rate": 4.810793095914355e-06, + "loss": 3.765, + "step": 1782 + }, + { + "epoch": 0.39, + "learning_rate": 4.810683854052873e-06, + "loss": 3.6875, + "step": 1783 + }, + { + "epoch": 0.39, + "learning_rate": 4.810574612191392e-06, + "loss": 3.6834, + "step": 1784 + }, + { + "epoch": 0.39, + "learning_rate": 4.810465370329911e-06, + "loss": 3.6171, + "step": 1785 + }, + { + "epoch": 0.39, + "learning_rate": 4.810356128468429e-06, + "loss": 3.4616, + "step": 1786 + }, + { + "epoch": 0.39, + "learning_rate": 4.8102468866069475e-06, + "loss": 3.6531, + "step": 1787 + }, + { + "epoch": 0.39, + "learning_rate": 4.810137644745467e-06, + "loss": 4.0333, + "step": 1788 + }, + { + "epoch": 0.39, + "learning_rate": 4.810028402883986e-06, + "loss": 3.8174, + "step": 1789 + }, + { + "epoch": 0.39, + "learning_rate": 4.809919161022505e-06, + "loss": 3.8836, + "step": 1790 + }, + { + "epoch": 0.39, + "learning_rate": 4.809809919161023e-06, + "loss": 3.622, + "step": 1791 + }, + { + "epoch": 0.39, + "learning_rate": 4.809700677299541e-06, + "loss": 3.6808, + "step": 1792 + }, + { + "epoch": 0.39, + "learning_rate": 4.80959143543806e-06, + "loss": 3.757, + "step": 1793 + }, + { + "epoch": 0.39, + "learning_rate": 4.809482193576579e-06, + "loss": 3.6144, + "step": 1794 + }, + { + "epoch": 0.39, + "learning_rate": 4.809372951715098e-06, + "loss": 3.9496, + "step": 1795 + }, + { + "epoch": 0.39, + "learning_rate": 4.809263709853616e-06, + "loss": 3.7728, + "step": 1796 + }, + { + "epoch": 0.39, + "learning_rate": 4.8091544679921345e-06, + "loss": 3.7481, + "step": 1797 + }, + { + "epoch": 0.39, + "learning_rate": 4.8090452261306535e-06, + "loss": 3.5402, + "step": 1798 + }, + { + "epoch": 0.39, + "learning_rate": 4.8089359842691726e-06, + "loss": 3.6644, + "step": 1799 + }, + { + "epoch": 0.39, + "learning_rate": 4.808826742407692e-06, + "loss": 3.946, + "step": 1800 + }, + { + "epoch": 0.39, + "learning_rate": 4.80871750054621e-06, + "loss": 3.8725, + "step": 1801 + }, + { + "epoch": 0.39, + "learning_rate": 4.808608258684728e-06, + "loss": 3.8928, + "step": 1802 + }, + { + "epoch": 0.39, + "learning_rate": 4.808499016823247e-06, + "loss": 3.5428, + "step": 1803 + }, + { + "epoch": 0.39, + "learning_rate": 4.808389774961766e-06, + "loss": 3.8139, + "step": 1804 + }, + { + "epoch": 0.39, + "learning_rate": 4.808280533100284e-06, + "loss": 3.713, + "step": 1805 + }, + { + "epoch": 0.39, + "learning_rate": 4.808171291238803e-06, + "loss": 3.7627, + "step": 1806 + }, + { + "epoch": 0.39, + "learning_rate": 4.8080620493773215e-06, + "loss": 3.5283, + "step": 1807 + }, + { + "epoch": 0.39, + "learning_rate": 4.8079528075158405e-06, + "loss": 3.5233, + "step": 1808 + }, + { + "epoch": 0.39, + "learning_rate": 4.8078435656543595e-06, + "loss": 3.6674, + "step": 1809 + }, + { + "epoch": 0.4, + "learning_rate": 4.807734323792878e-06, + "loss": 3.872, + "step": 1810 + }, + { + "epoch": 0.4, + "learning_rate": 4.807625081931396e-06, + "loss": 3.9331, + "step": 1811 + }, + { + "epoch": 0.4, + "learning_rate": 4.807515840069915e-06, + "loss": 3.6372, + "step": 1812 + }, + { + "epoch": 0.4, + "learning_rate": 4.807406598208434e-06, + "loss": 3.8295, + "step": 1813 + }, + { + "epoch": 0.4, + "learning_rate": 4.807297356346953e-06, + "loss": 3.6842, + "step": 1814 + }, + { + "epoch": 0.4, + "learning_rate": 4.807188114485471e-06, + "loss": 3.6763, + "step": 1815 + }, + { + "epoch": 0.4, + "learning_rate": 4.807078872623989e-06, + "loss": 3.8795, + "step": 1816 + }, + { + "epoch": 0.4, + "learning_rate": 4.806969630762508e-06, + "loss": 3.8726, + "step": 1817 + }, + { + "epoch": 0.4, + "learning_rate": 4.8068603889010275e-06, + "loss": 3.7805, + "step": 1818 + }, + { + "epoch": 0.4, + "learning_rate": 4.8067511470395465e-06, + "loss": 3.6553, + "step": 1819 + }, + { + "epoch": 0.4, + "learning_rate": 4.806641905178065e-06, + "loss": 4.0518, + "step": 1820 + }, + { + "epoch": 0.4, + "learning_rate": 4.806532663316583e-06, + "loss": 3.9571, + "step": 1821 + }, + { + "epoch": 0.4, + "learning_rate": 4.806423421455102e-06, + "loss": 3.5812, + "step": 1822 + }, + { + "epoch": 0.4, + "learning_rate": 4.806314179593621e-06, + "loss": 3.6491, + "step": 1823 + }, + { + "epoch": 0.4, + "learning_rate": 4.80620493773214e-06, + "loss": 3.8022, + "step": 1824 + }, + { + "epoch": 0.4, + "learning_rate": 4.806095695870658e-06, + "loss": 3.8592, + "step": 1825 + }, + { + "epoch": 0.4, + "learning_rate": 4.805986454009176e-06, + "loss": 3.5063, + "step": 1826 + }, + { + "epoch": 0.4, + "learning_rate": 4.805877212147695e-06, + "loss": 3.7234, + "step": 1827 + }, + { + "epoch": 0.4, + "learning_rate": 4.805767970286214e-06, + "loss": 3.9858, + "step": 1828 + }, + { + "epoch": 0.4, + "learning_rate": 4.805658728424733e-06, + "loss": 3.7993, + "step": 1829 + }, + { + "epoch": 0.4, + "learning_rate": 4.805549486563252e-06, + "loss": 3.7937, + "step": 1830 + }, + { + "epoch": 0.4, + "learning_rate": 4.80544024470177e-06, + "loss": 3.7429, + "step": 1831 + }, + { + "epoch": 0.4, + "learning_rate": 4.805331002840289e-06, + "loss": 3.6498, + "step": 1832 + }, + { + "epoch": 0.4, + "learning_rate": 4.805221760978808e-06, + "loss": 3.7667, + "step": 1833 + }, + { + "epoch": 0.4, + "learning_rate": 4.805112519117326e-06, + "loss": 3.7496, + "step": 1834 + }, + { + "epoch": 0.4, + "learning_rate": 4.805003277255845e-06, + "loss": 3.7406, + "step": 1835 + }, + { + "epoch": 0.4, + "learning_rate": 4.804894035394363e-06, + "loss": 3.6541, + "step": 1836 + }, + { + "epoch": 0.4, + "learning_rate": 4.804784793532882e-06, + "loss": 3.752, + "step": 1837 + }, + { + "epoch": 0.4, + "learning_rate": 4.804675551671401e-06, + "loss": 3.8294, + "step": 1838 + }, + { + "epoch": 0.4, + "learning_rate": 4.8045663098099196e-06, + "loss": 3.5076, + "step": 1839 + }, + { + "epoch": 0.4, + "learning_rate": 4.804457067948438e-06, + "loss": 3.7057, + "step": 1840 + }, + { + "epoch": 0.4, + "learning_rate": 4.804347826086957e-06, + "loss": 3.7806, + "step": 1841 + }, + { + "epoch": 0.4, + "learning_rate": 4.804238584225476e-06, + "loss": 3.6894, + "step": 1842 + }, + { + "epoch": 0.4, + "learning_rate": 4.804129342363995e-06, + "loss": 3.9355, + "step": 1843 + }, + { + "epoch": 0.4, + "learning_rate": 4.804020100502513e-06, + "loss": 3.9596, + "step": 1844 + }, + { + "epoch": 0.4, + "learning_rate": 4.803910858641031e-06, + "loss": 3.444, + "step": 1845 + }, + { + "epoch": 0.4, + "learning_rate": 4.80380161677955e-06, + "loss": 3.5855, + "step": 1846 + }, + { + "epoch": 0.4, + "learning_rate": 4.803692374918069e-06, + "loss": 3.5833, + "step": 1847 + }, + { + "epoch": 0.4, + "learning_rate": 4.8035831330565875e-06, + "loss": 3.477, + "step": 1848 + }, + { + "epoch": 0.4, + "learning_rate": 4.8034738911951065e-06, + "loss": 3.9538, + "step": 1849 + }, + { + "epoch": 0.4, + "learning_rate": 4.803364649333625e-06, + "loss": 3.6891, + "step": 1850 + }, + { + "epoch": 0.4, + "learning_rate": 4.803255407472144e-06, + "loss": 3.4017, + "step": 1851 + }, + { + "epoch": 0.4, + "learning_rate": 4.803146165610663e-06, + "loss": 3.7232, + "step": 1852 + }, + { + "epoch": 0.4, + "learning_rate": 4.803036923749181e-06, + "loss": 3.5847, + "step": 1853 + }, + { + "epoch": 0.4, + "learning_rate": 4.8029276818877e-06, + "loss": 3.6668, + "step": 1854 + }, + { + "epoch": 0.4, + "learning_rate": 4.802818440026218e-06, + "loss": 3.5849, + "step": 1855 + }, + { + "epoch": 0.41, + "learning_rate": 4.802709198164737e-06, + "loss": 3.7259, + "step": 1856 + }, + { + "epoch": 0.41, + "learning_rate": 4.802599956303256e-06, + "loss": 3.8792, + "step": 1857 + }, + { + "epoch": 0.41, + "learning_rate": 4.8024907144417744e-06, + "loss": 3.7213, + "step": 1858 + }, + { + "epoch": 0.41, + "learning_rate": 4.8023814725802935e-06, + "loss": 3.4857, + "step": 1859 + }, + { + "epoch": 0.41, + "learning_rate": 4.802272230718812e-06, + "loss": 3.3764, + "step": 1860 + }, + { + "epoch": 0.41, + "learning_rate": 4.802162988857331e-06, + "loss": 3.549, + "step": 1861 + }, + { + "epoch": 0.41, + "learning_rate": 4.80205374699585e-06, + "loss": 3.6632, + "step": 1862 + }, + { + "epoch": 0.41, + "learning_rate": 4.801944505134368e-06, + "loss": 3.9071, + "step": 1863 + }, + { + "epoch": 0.41, + "learning_rate": 4.801835263272886e-06, + "loss": 3.8611, + "step": 1864 + }, + { + "epoch": 0.41, + "learning_rate": 4.801726021411405e-06, + "loss": 3.3627, + "step": 1865 + }, + { + "epoch": 0.41, + "learning_rate": 4.801616779549924e-06, + "loss": 3.6916, + "step": 1866 + }, + { + "epoch": 0.41, + "learning_rate": 4.801507537688442e-06, + "loss": 3.7226, + "step": 1867 + }, + { + "epoch": 0.41, + "learning_rate": 4.801398295826961e-06, + "loss": 3.8999, + "step": 1868 + }, + { + "epoch": 0.41, + "learning_rate": 4.80128905396548e-06, + "loss": 3.8737, + "step": 1869 + }, + { + "epoch": 0.41, + "learning_rate": 4.801179812103999e-06, + "loss": 3.7383, + "step": 1870 + }, + { + "epoch": 0.41, + "learning_rate": 4.801070570242518e-06, + "loss": 3.6323, + "step": 1871 + }, + { + "epoch": 0.41, + "learning_rate": 4.800961328381036e-06, + "loss": 3.5851, + "step": 1872 + }, + { + "epoch": 0.41, + "learning_rate": 4.800852086519555e-06, + "loss": 3.4459, + "step": 1873 + }, + { + "epoch": 0.41, + "learning_rate": 4.800742844658073e-06, + "loss": 3.3922, + "step": 1874 + }, + { + "epoch": 0.41, + "learning_rate": 4.800633602796592e-06, + "loss": 3.6423, + "step": 1875 + }, + { + "epoch": 0.41, + "learning_rate": 4.800524360935111e-06, + "loss": 3.5793, + "step": 1876 + }, + { + "epoch": 0.41, + "learning_rate": 4.800415119073629e-06, + "loss": 3.6706, + "step": 1877 + }, + { + "epoch": 0.41, + "learning_rate": 4.800305877212148e-06, + "loss": 3.6608, + "step": 1878 + }, + { + "epoch": 0.41, + "learning_rate": 4.8001966353506665e-06, + "loss": 3.5461, + "step": 1879 + }, + { + "epoch": 0.41, + "learning_rate": 4.8000873934891856e-06, + "loss": 3.5426, + "step": 1880 + }, + { + "epoch": 0.41, + "learning_rate": 4.799978151627704e-06, + "loss": 3.7819, + "step": 1881 + }, + { + "epoch": 0.41, + "learning_rate": 4.799868909766223e-06, + "loss": 3.7698, + "step": 1882 + }, + { + "epoch": 0.41, + "learning_rate": 4.799759667904742e-06, + "loss": 3.6956, + "step": 1883 + }, + { + "epoch": 0.41, + "learning_rate": 4.79965042604326e-06, + "loss": 3.8811, + "step": 1884 + }, + { + "epoch": 0.41, + "learning_rate": 4.799541184181779e-06, + "loss": 3.6329, + "step": 1885 + }, + { + "epoch": 0.41, + "learning_rate": 4.799431942320297e-06, + "loss": 3.8543, + "step": 1886 + }, + { + "epoch": 0.41, + "learning_rate": 4.799322700458816e-06, + "loss": 3.473, + "step": 1887 + }, + { + "epoch": 0.41, + "learning_rate": 4.799213458597335e-06, + "loss": 3.8996, + "step": 1888 + }, + { + "epoch": 0.41, + "learning_rate": 4.7991042167358535e-06, + "loss": 3.6722, + "step": 1889 + }, + { + "epoch": 0.41, + "learning_rate": 4.7989949748743725e-06, + "loss": 3.5235, + "step": 1890 + }, + { + "epoch": 0.41, + "learning_rate": 4.798885733012891e-06, + "loss": 3.7304, + "step": 1891 + }, + { + "epoch": 0.41, + "learning_rate": 4.79877649115141e-06, + "loss": 3.8086, + "step": 1892 + }, + { + "epoch": 0.41, + "learning_rate": 4.798667249289928e-06, + "loss": 3.8309, + "step": 1893 + }, + { + "epoch": 0.41, + "learning_rate": 4.798558007428447e-06, + "loss": 3.8625, + "step": 1894 + }, + { + "epoch": 0.41, + "learning_rate": 4.798448765566966e-06, + "loss": 3.6984, + "step": 1895 + }, + { + "epoch": 0.41, + "learning_rate": 4.798339523705484e-06, + "loss": 3.8385, + "step": 1896 + }, + { + "epoch": 0.41, + "learning_rate": 4.798230281844003e-06, + "loss": 3.628, + "step": 1897 + }, + { + "epoch": 0.41, + "learning_rate": 4.798121039982521e-06, + "loss": 3.7314, + "step": 1898 + }, + { + "epoch": 0.41, + "learning_rate": 4.7980117981210405e-06, + "loss": 3.4865, + "step": 1899 + }, + { + "epoch": 0.41, + "learning_rate": 4.797902556259559e-06, + "loss": 3.7457, + "step": 1900 + }, + { + "epoch": 0.41, + "learning_rate": 4.797793314398078e-06, + "loss": 3.8934, + "step": 1901 + }, + { + "epoch": 0.42, + "learning_rate": 4.797684072536597e-06, + "loss": 3.6738, + "step": 1902 + }, + { + "epoch": 0.42, + "learning_rate": 4.797574830675115e-06, + "loss": 3.5252, + "step": 1903 + }, + { + "epoch": 0.42, + "learning_rate": 4.797465588813634e-06, + "loss": 3.5736, + "step": 1904 + }, + { + "epoch": 0.42, + "learning_rate": 4.797356346952152e-06, + "loss": 3.9307, + "step": 1905 + }, + { + "epoch": 0.42, + "learning_rate": 4.797247105090671e-06, + "loss": 3.6724, + "step": 1906 + }, + { + "epoch": 0.42, + "learning_rate": 4.79713786322919e-06, + "loss": 3.9047, + "step": 1907 + }, + { + "epoch": 0.42, + "learning_rate": 4.797028621367708e-06, + "loss": 3.2996, + "step": 1908 + }, + { + "epoch": 0.42, + "learning_rate": 4.796919379506227e-06, + "loss": 3.7227, + "step": 1909 + }, + { + "epoch": 0.42, + "learning_rate": 4.796810137644746e-06, + "loss": 3.9574, + "step": 1910 + }, + { + "epoch": 0.42, + "learning_rate": 4.796700895783265e-06, + "loss": 3.7589, + "step": 1911 + }, + { + "epoch": 0.42, + "learning_rate": 4.796591653921784e-06, + "loss": 3.7701, + "step": 1912 + }, + { + "epoch": 0.42, + "learning_rate": 4.796482412060302e-06, + "loss": 3.3733, + "step": 1913 + }, + { + "epoch": 0.42, + "learning_rate": 4.79637317019882e-06, + "loss": 3.7271, + "step": 1914 + }, + { + "epoch": 0.42, + "learning_rate": 4.796263928337339e-06, + "loss": 3.67, + "step": 1915 + }, + { + "epoch": 0.42, + "learning_rate": 4.796154686475858e-06, + "loss": 3.6075, + "step": 1916 + }, + { + "epoch": 0.42, + "learning_rate": 4.796045444614376e-06, + "loss": 4.0008, + "step": 1917 + }, + { + "epoch": 0.42, + "learning_rate": 4.795936202752895e-06, + "loss": 3.361, + "step": 1918 + }, + { + "epoch": 0.42, + "learning_rate": 4.7958269608914135e-06, + "loss": 3.4671, + "step": 1919 + }, + { + "epoch": 0.42, + "learning_rate": 4.7957177190299326e-06, + "loss": 3.6017, + "step": 1920 + }, + { + "epoch": 0.42, + "learning_rate": 4.795608477168452e-06, + "loss": 3.769, + "step": 1921 + }, + { + "epoch": 0.42, + "learning_rate": 4.79549923530697e-06, + "loss": 3.7384, + "step": 1922 + }, + { + "epoch": 0.42, + "learning_rate": 4.795389993445489e-06, + "loss": 3.8366, + "step": 1923 + }, + { + "epoch": 0.42, + "learning_rate": 4.795280751584007e-06, + "loss": 3.9211, + "step": 1924 + }, + { + "epoch": 0.42, + "learning_rate": 4.795171509722526e-06, + "loss": 4.0038, + "step": 1925 + }, + { + "epoch": 0.42, + "learning_rate": 4.795062267861045e-06, + "loss": 3.5202, + "step": 1926 + }, + { + "epoch": 0.42, + "learning_rate": 4.794953025999563e-06, + "loss": 3.8756, + "step": 1927 + }, + { + "epoch": 0.42, + "learning_rate": 4.794843784138082e-06, + "loss": 3.7233, + "step": 1928 + }, + { + "epoch": 0.42, + "learning_rate": 4.7947345422766005e-06, + "loss": 3.8828, + "step": 1929 + }, + { + "epoch": 0.42, + "learning_rate": 4.7946253004151195e-06, + "loss": 4.0111, + "step": 1930 + }, + { + "epoch": 0.42, + "learning_rate": 4.7945160585536385e-06, + "loss": 3.6936, + "step": 1931 + }, + { + "epoch": 0.42, + "learning_rate": 4.794406816692157e-06, + "loss": 3.6654, + "step": 1932 + }, + { + "epoch": 0.42, + "learning_rate": 4.794297574830675e-06, + "loss": 3.9779, + "step": 1933 + }, + { + "epoch": 0.42, + "learning_rate": 4.794188332969194e-06, + "loss": 3.6199, + "step": 1934 + }, + { + "epoch": 0.42, + "learning_rate": 4.794079091107713e-06, + "loss": 3.3086, + "step": 1935 + }, + { + "epoch": 0.42, + "learning_rate": 4.793969849246232e-06, + "loss": 3.613, + "step": 1936 + }, + { + "epoch": 0.42, + "learning_rate": 4.79386060738475e-06, + "loss": 3.6442, + "step": 1937 + }, + { + "epoch": 0.42, + "learning_rate": 4.793751365523268e-06, + "loss": 3.637, + "step": 1938 + }, + { + "epoch": 0.42, + "learning_rate": 4.7936421236617874e-06, + "loss": 3.5155, + "step": 1939 + }, + { + "epoch": 0.42, + "learning_rate": 4.7935328818003065e-06, + "loss": 4.0141, + "step": 1940 + }, + { + "epoch": 0.42, + "learning_rate": 4.7934236399388255e-06, + "loss": 3.9107, + "step": 1941 + }, + { + "epoch": 0.42, + "learning_rate": 4.793314398077344e-06, + "loss": 3.7101, + "step": 1942 + }, + { + "epoch": 0.42, + "learning_rate": 4.793205156215862e-06, + "loss": 3.4651, + "step": 1943 + }, + { + "epoch": 0.42, + "learning_rate": 4.793095914354381e-06, + "loss": 3.8269, + "step": 1944 + }, + { + "epoch": 0.42, + "learning_rate": 4.7929866724929e-06, + "loss": 3.3608, + "step": 1945 + }, + { + "epoch": 0.42, + "learning_rate": 4.792877430631418e-06, + "loss": 3.994, + "step": 1946 + }, + { + "epoch": 0.42, + "learning_rate": 4.792768188769937e-06, + "loss": 3.9285, + "step": 1947 + }, + { + "epoch": 0.43, + "learning_rate": 4.792658946908455e-06, + "loss": 3.9658, + "step": 1948 + }, + { + "epoch": 0.43, + "learning_rate": 4.792549705046974e-06, + "loss": 3.6561, + "step": 1949 + }, + { + "epoch": 0.43, + "learning_rate": 4.792440463185493e-06, + "loss": 3.7412, + "step": 1950 + }, + { + "epoch": 0.43, + "learning_rate": 4.792331221324012e-06, + "loss": 3.878, + "step": 1951 + }, + { + "epoch": 0.43, + "learning_rate": 4.79222197946253e-06, + "loss": 3.6483, + "step": 1952 + }, + { + "epoch": 0.43, + "learning_rate": 4.792112737601049e-06, + "loss": 3.3416, + "step": 1953 + }, + { + "epoch": 0.43, + "learning_rate": 4.792003495739568e-06, + "loss": 3.6041, + "step": 1954 + }, + { + "epoch": 0.43, + "learning_rate": 4.791894253878087e-06, + "loss": 3.6311, + "step": 1955 + }, + { + "epoch": 0.43, + "learning_rate": 4.791785012016605e-06, + "loss": 3.8336, + "step": 1956 + }, + { + "epoch": 0.43, + "learning_rate": 4.791675770155123e-06, + "loss": 3.7413, + "step": 1957 + }, + { + "epoch": 0.43, + "learning_rate": 4.791566528293642e-06, + "loss": 3.8789, + "step": 1958 + }, + { + "epoch": 0.43, + "learning_rate": 4.791457286432161e-06, + "loss": 3.6225, + "step": 1959 + }, + { + "epoch": 0.43, + "learning_rate": 4.79134804457068e-06, + "loss": 3.9474, + "step": 1960 + }, + { + "epoch": 0.43, + "learning_rate": 4.7912388027091986e-06, + "loss": 3.4333, + "step": 1961 + }, + { + "epoch": 0.43, + "learning_rate": 4.791129560847717e-06, + "loss": 3.6914, + "step": 1962 + }, + { + "epoch": 0.43, + "learning_rate": 4.791020318986236e-06, + "loss": 3.6188, + "step": 1963 + }, + { + "epoch": 0.43, + "learning_rate": 4.790911077124755e-06, + "loss": 3.9211, + "step": 1964 + }, + { + "epoch": 0.43, + "learning_rate": 4.790801835263274e-06, + "loss": 3.4545, + "step": 1965 + }, + { + "epoch": 0.43, + "learning_rate": 4.790692593401792e-06, + "loss": 3.6362, + "step": 1966 + }, + { + "epoch": 0.43, + "learning_rate": 4.79058335154031e-06, + "loss": 3.5544, + "step": 1967 + }, + { + "epoch": 0.43, + "learning_rate": 4.790474109678829e-06, + "loss": 3.865, + "step": 1968 + }, + { + "epoch": 0.43, + "learning_rate": 4.790364867817348e-06, + "loss": 3.3942, + "step": 1969 + }, + { + "epoch": 0.43, + "learning_rate": 4.7902556259558665e-06, + "loss": 3.7673, + "step": 1970 + }, + { + "epoch": 0.43, + "learning_rate": 4.7901463840943855e-06, + "loss": 3.7756, + "step": 1971 + }, + { + "epoch": 0.43, + "learning_rate": 4.790037142232904e-06, + "loss": 3.9377, + "step": 1972 + }, + { + "epoch": 0.43, + "learning_rate": 4.789927900371423e-06, + "loss": 3.5474, + "step": 1973 + }, + { + "epoch": 0.43, + "learning_rate": 4.789818658509942e-06, + "loss": 3.7065, + "step": 1974 + }, + { + "epoch": 0.43, + "learning_rate": 4.78970941664846e-06, + "loss": 3.7649, + "step": 1975 + }, + { + "epoch": 0.43, + "learning_rate": 4.789600174786979e-06, + "loss": 3.5755, + "step": 1976 + }, + { + "epoch": 0.43, + "learning_rate": 4.789490932925497e-06, + "loss": 3.4659, + "step": 1977 + }, + { + "epoch": 0.43, + "learning_rate": 4.789381691064016e-06, + "loss": 3.516, + "step": 1978 + }, + { + "epoch": 0.43, + "learning_rate": 4.789272449202535e-06, + "loss": 3.3901, + "step": 1979 + }, + { + "epoch": 0.43, + "learning_rate": 4.7891632073410534e-06, + "loss": 3.2167, + "step": 1980 + }, + { + "epoch": 0.43, + "learning_rate": 4.789053965479572e-06, + "loss": 3.5535, + "step": 1981 + }, + { + "epoch": 0.43, + "learning_rate": 4.788944723618091e-06, + "loss": 3.3973, + "step": 1982 + }, + { + "epoch": 0.43, + "learning_rate": 4.78883548175661e-06, + "loss": 3.7672, + "step": 1983 + }, + { + "epoch": 0.43, + "learning_rate": 4.788726239895129e-06, + "loss": 3.693, + "step": 1984 + }, + { + "epoch": 0.43, + "learning_rate": 4.788616998033647e-06, + "loss": 3.5518, + "step": 1985 + }, + { + "epoch": 0.43, + "learning_rate": 4.788507756172165e-06, + "loss": 3.7326, + "step": 1986 + }, + { + "epoch": 0.43, + "learning_rate": 4.788398514310684e-06, + "loss": 3.6127, + "step": 1987 + }, + { + "epoch": 0.43, + "learning_rate": 4.788289272449203e-06, + "loss": 3.7596, + "step": 1988 + }, + { + "epoch": 0.43, + "learning_rate": 4.788180030587722e-06, + "loss": 3.6096, + "step": 1989 + }, + { + "epoch": 0.43, + "learning_rate": 4.78807078872624e-06, + "loss": 3.6163, + "step": 1990 + }, + { + "epoch": 0.43, + "learning_rate": 4.787961546864759e-06, + "loss": 3.7606, + "step": 1991 + }, + { + "epoch": 0.43, + "learning_rate": 4.787852305003278e-06, + "loss": 3.6199, + "step": 1992 + }, + { + "epoch": 0.43, + "learning_rate": 4.787743063141797e-06, + "loss": 3.5305, + "step": 1993 + }, + { + "epoch": 0.44, + "learning_rate": 4.787633821280315e-06, + "loss": 3.9147, + "step": 1994 + }, + { + "epoch": 0.44, + "learning_rate": 4.787524579418834e-06, + "loss": 3.6611, + "step": 1995 + }, + { + "epoch": 0.44, + "learning_rate": 4.787415337557352e-06, + "loss": 3.8533, + "step": 1996 + }, + { + "epoch": 0.44, + "learning_rate": 4.787306095695871e-06, + "loss": 3.308, + "step": 1997 + }, + { + "epoch": 0.44, + "learning_rate": 4.78719685383439e-06, + "loss": 3.6913, + "step": 1998 + }, + { + "epoch": 0.44, + "learning_rate": 4.787087611972908e-06, + "loss": 3.4596, + "step": 1999 + }, + { + "epoch": 0.44, + "learning_rate": 4.786978370111427e-06, + "loss": 3.3734, + "step": 2000 + }, + { + "epoch": 0.44, + "learning_rate": 4.7868691282499455e-06, + "loss": 3.9831, + "step": 2001 + }, + { + "epoch": 0.44, + "learning_rate": 4.786759886388465e-06, + "loss": 3.7496, + "step": 2002 + }, + { + "epoch": 0.44, + "learning_rate": 4.786650644526984e-06, + "loss": 3.8033, + "step": 2003 + }, + { + "epoch": 0.44, + "learning_rate": 4.786541402665502e-06, + "loss": 3.5251, + "step": 2004 + }, + { + "epoch": 0.44, + "learning_rate": 4.78643216080402e-06, + "loss": 3.7208, + "step": 2005 + }, + { + "epoch": 0.44, + "learning_rate": 4.786322918942539e-06, + "loss": 3.6896, + "step": 2006 + }, + { + "epoch": 0.44, + "learning_rate": 4.786213677081058e-06, + "loss": 3.4983, + "step": 2007 + }, + { + "epoch": 0.44, + "learning_rate": 4.786104435219577e-06, + "loss": 3.5403, + "step": 2008 + }, + { + "epoch": 0.44, + "learning_rate": 4.785995193358095e-06, + "loss": 3.7387, + "step": 2009 + }, + { + "epoch": 0.44, + "learning_rate": 4.7858859514966135e-06, + "loss": 3.7154, + "step": 2010 + }, + { + "epoch": 0.44, + "learning_rate": 4.7857767096351325e-06, + "loss": 3.7895, + "step": 2011 + }, + { + "epoch": 0.44, + "learning_rate": 4.7856674677736515e-06, + "loss": 3.5525, + "step": 2012 + }, + { + "epoch": 0.44, + "learning_rate": 4.78555822591217e-06, + "loss": 3.3083, + "step": 2013 + }, + { + "epoch": 0.44, + "learning_rate": 4.785448984050689e-06, + "loss": 3.5619, + "step": 2014 + }, + { + "epoch": 0.44, + "learning_rate": 4.785339742189207e-06, + "loss": 3.807, + "step": 2015 + }, + { + "epoch": 0.44, + "learning_rate": 4.785230500327726e-06, + "loss": 3.8566, + "step": 2016 + }, + { + "epoch": 0.44, + "learning_rate": 4.785121258466245e-06, + "loss": 3.6786, + "step": 2017 + }, + { + "epoch": 0.44, + "learning_rate": 4.785012016604763e-06, + "loss": 3.7227, + "step": 2018 + }, + { + "epoch": 0.44, + "learning_rate": 4.784902774743282e-06, + "loss": 3.3718, + "step": 2019 + }, + { + "epoch": 0.44, + "learning_rate": 4.7847935328818004e-06, + "loss": 3.8221, + "step": 2020 + }, + { + "epoch": 0.44, + "learning_rate": 4.7846842910203195e-06, + "loss": 3.4667, + "step": 2021 + }, + { + "epoch": 0.44, + "learning_rate": 4.7845750491588385e-06, + "loss": 3.8141, + "step": 2022 + }, + { + "epoch": 0.44, + "learning_rate": 4.784465807297357e-06, + "loss": 4.0037, + "step": 2023 + }, + { + "epoch": 0.44, + "learning_rate": 4.784356565435876e-06, + "loss": 3.3635, + "step": 2024 + }, + { + "epoch": 0.44, + "learning_rate": 4.784247323574394e-06, + "loss": 3.7227, + "step": 2025 + }, + { + "epoch": 0.44, + "learning_rate": 4.784138081712913e-06, + "loss": 3.8582, + "step": 2026 + }, + { + "epoch": 0.44, + "learning_rate": 4.784028839851431e-06, + "loss": 3.7251, + "step": 2027 + }, + { + "epoch": 0.44, + "learning_rate": 4.78391959798995e-06, + "loss": 3.6986, + "step": 2028 + }, + { + "epoch": 0.44, + "learning_rate": 4.783810356128469e-06, + "loss": 3.7194, + "step": 2029 + }, + { + "epoch": 0.44, + "learning_rate": 4.783701114266987e-06, + "loss": 3.6375, + "step": 2030 + }, + { + "epoch": 0.44, + "learning_rate": 4.783591872405506e-06, + "loss": 3.5657, + "step": 2031 + }, + { + "epoch": 0.44, + "learning_rate": 4.783482630544025e-06, + "loss": 3.6008, + "step": 2032 + }, + { + "epoch": 0.44, + "learning_rate": 4.783373388682544e-06, + "loss": 3.855, + "step": 2033 + }, + { + "epoch": 0.44, + "learning_rate": 4.783264146821062e-06, + "loss": 3.7086, + "step": 2034 + }, + { + "epoch": 0.44, + "learning_rate": 4.783154904959581e-06, + "loss": 3.8625, + "step": 2035 + }, + { + "epoch": 0.44, + "learning_rate": 4.7830456630981e-06, + "loss": 3.7976, + "step": 2036 + }, + { + "epoch": 0.44, + "learning_rate": 4.782936421236618e-06, + "loss": 3.4625, + "step": 2037 + }, + { + "epoch": 0.44, + "learning_rate": 4.782827179375137e-06, + "loss": 4.0359, + "step": 2038 + }, + { + "epoch": 0.45, + "learning_rate": 4.782717937513655e-06, + "loss": 3.3721, + "step": 2039 + }, + { + "epoch": 0.45, + "learning_rate": 4.782608695652174e-06, + "loss": 3.3049, + "step": 2040 + }, + { + "epoch": 0.45, + "learning_rate": 4.782499453790693e-06, + "loss": 3.6503, + "step": 2041 + }, + { + "epoch": 0.45, + "learning_rate": 4.7823902119292116e-06, + "loss": 3.55, + "step": 2042 + }, + { + "epoch": 0.45, + "learning_rate": 4.782280970067731e-06, + "loss": 3.833, + "step": 2043 + }, + { + "epoch": 0.45, + "learning_rate": 4.782171728206249e-06, + "loss": 3.6169, + "step": 2044 + }, + { + "epoch": 0.45, + "learning_rate": 4.782062486344768e-06, + "loss": 3.5712, + "step": 2045 + }, + { + "epoch": 0.45, + "learning_rate": 4.781953244483286e-06, + "loss": 4.1276, + "step": 2046 + }, + { + "epoch": 0.45, + "learning_rate": 4.781844002621805e-06, + "loss": 3.6989, + "step": 2047 + }, + { + "epoch": 0.45, + "learning_rate": 4.781734760760324e-06, + "loss": 3.9072, + "step": 2048 + }, + { + "epoch": 0.45, + "learning_rate": 4.781625518898842e-06, + "loss": 3.7558, + "step": 2049 + }, + { + "epoch": 0.45, + "learning_rate": 4.781516277037361e-06, + "loss": 3.7152, + "step": 2050 + }, + { + "epoch": 0.45, + "learning_rate": 4.7814070351758795e-06, + "loss": 4.1028, + "step": 2051 + }, + { + "epoch": 0.45, + "learning_rate": 4.7812977933143985e-06, + "loss": 3.5447, + "step": 2052 + }, + { + "epoch": 0.45, + "learning_rate": 4.7811885514529176e-06, + "loss": 3.6417, + "step": 2053 + }, + { + "epoch": 0.45, + "learning_rate": 4.781079309591436e-06, + "loss": 3.5674, + "step": 2054 + }, + { + "epoch": 0.45, + "learning_rate": 4.780970067729955e-06, + "loss": 3.6073, + "step": 2055 + }, + { + "epoch": 0.45, + "learning_rate": 4.780860825868473e-06, + "loss": 3.5077, + "step": 2056 + }, + { + "epoch": 0.45, + "learning_rate": 4.780751584006992e-06, + "loss": 3.5276, + "step": 2057 + }, + { + "epoch": 0.45, + "learning_rate": 4.78064234214551e-06, + "loss": 3.5766, + "step": 2058 + }, + { + "epoch": 0.45, + "learning_rate": 4.780533100284029e-06, + "loss": 3.6406, + "step": 2059 + }, + { + "epoch": 0.45, + "learning_rate": 4.780423858422547e-06, + "loss": 3.5114, + "step": 2060 + }, + { + "epoch": 0.45, + "learning_rate": 4.7803146165610664e-06, + "loss": 3.7842, + "step": 2061 + }, + { + "epoch": 0.45, + "learning_rate": 4.7802053746995855e-06, + "loss": 3.5822, + "step": 2062 + }, + { + "epoch": 0.45, + "learning_rate": 4.780096132838104e-06, + "loss": 4.0181, + "step": 2063 + }, + { + "epoch": 0.45, + "learning_rate": 4.779986890976623e-06, + "loss": 3.8477, + "step": 2064 + }, + { + "epoch": 0.45, + "learning_rate": 4.779877649115141e-06, + "loss": 3.8464, + "step": 2065 + }, + { + "epoch": 0.45, + "learning_rate": 4.77976840725366e-06, + "loss": 3.2565, + "step": 2066 + }, + { + "epoch": 0.45, + "learning_rate": 4.779659165392179e-06, + "loss": 3.6228, + "step": 2067 + }, + { + "epoch": 0.45, + "learning_rate": 4.779549923530697e-06, + "loss": 3.5436, + "step": 2068 + }, + { + "epoch": 0.45, + "learning_rate": 4.779440681669216e-06, + "loss": 3.8551, + "step": 2069 + }, + { + "epoch": 0.45, + "learning_rate": 4.779331439807734e-06, + "loss": 3.7802, + "step": 2070 + }, + { + "epoch": 0.45, + "learning_rate": 4.779222197946253e-06, + "loss": 3.737, + "step": 2071 + }, + { + "epoch": 0.45, + "learning_rate": 4.7791129560847724e-06, + "loss": 3.7842, + "step": 2072 + }, + { + "epoch": 0.45, + "learning_rate": 4.779003714223291e-06, + "loss": 3.5861, + "step": 2073 + }, + { + "epoch": 0.45, + "learning_rate": 4.77889447236181e-06, + "loss": 3.1371, + "step": 2074 + }, + { + "epoch": 0.45, + "learning_rate": 4.778785230500328e-06, + "loss": 3.5208, + "step": 2075 + }, + { + "epoch": 0.45, + "learning_rate": 4.778675988638847e-06, + "loss": 3.7007, + "step": 2076 + }, + { + "epoch": 0.45, + "learning_rate": 4.778566746777366e-06, + "loss": 3.1767, + "step": 2077 + }, + { + "epoch": 0.45, + "learning_rate": 4.778457504915884e-06, + "loss": 3.7272, + "step": 2078 + }, + { + "epoch": 0.45, + "learning_rate": 4.778348263054402e-06, + "loss": 3.8603, + "step": 2079 + }, + { + "epoch": 0.45, + "learning_rate": 4.778239021192921e-06, + "loss": 3.5723, + "step": 2080 + }, + { + "epoch": 0.45, + "learning_rate": 4.77812977933144e-06, + "loss": 3.4587, + "step": 2081 + }, + { + "epoch": 0.45, + "learning_rate": 4.778020537469959e-06, + "loss": 3.8894, + "step": 2082 + }, + { + "epoch": 0.45, + "learning_rate": 4.777911295608478e-06, + "loss": 4.0743, + "step": 2083 + }, + { + "epoch": 0.45, + "learning_rate": 4.777802053746996e-06, + "loss": 3.6067, + "step": 2084 + }, + { + "epoch": 0.46, + "learning_rate": 4.777692811885515e-06, + "loss": 3.5012, + "step": 2085 + }, + { + "epoch": 0.46, + "learning_rate": 4.777583570024034e-06, + "loss": 3.9401, + "step": 2086 + }, + { + "epoch": 0.46, + "learning_rate": 4.777474328162552e-06, + "loss": 3.5781, + "step": 2087 + }, + { + "epoch": 0.46, + "learning_rate": 4.777365086301071e-06, + "loss": 3.1711, + "step": 2088 + }, + { + "epoch": 0.46, + "learning_rate": 4.777255844439589e-06, + "loss": 3.6433, + "step": 2089 + }, + { + "epoch": 0.46, + "learning_rate": 4.777146602578108e-06, + "loss": 3.7477, + "step": 2090 + }, + { + "epoch": 0.46, + "learning_rate": 4.777037360716627e-06, + "loss": 3.9581, + "step": 2091 + }, + { + "epoch": 0.46, + "learning_rate": 4.7769281188551455e-06, + "loss": 3.6005, + "step": 2092 + }, + { + "epoch": 0.46, + "learning_rate": 4.776818876993664e-06, + "loss": 3.7559, + "step": 2093 + }, + { + "epoch": 0.46, + "learning_rate": 4.776709635132183e-06, + "loss": 3.7786, + "step": 2094 + }, + { + "epoch": 0.46, + "learning_rate": 4.776600393270702e-06, + "loss": 3.4429, + "step": 2095 + }, + { + "epoch": 0.46, + "learning_rate": 4.776491151409221e-06, + "loss": 3.5669, + "step": 2096 + }, + { + "epoch": 0.46, + "learning_rate": 4.776381909547739e-06, + "loss": 4.1573, + "step": 2097 + }, + { + "epoch": 0.46, + "learning_rate": 4.776272667686257e-06, + "loss": 3.4658, + "step": 2098 + }, + { + "epoch": 0.46, + "learning_rate": 4.776163425824776e-06, + "loss": 4.0312, + "step": 2099 + }, + { + "epoch": 0.46, + "learning_rate": 4.776054183963295e-06, + "loss": 3.4851, + "step": 2100 + }, + { + "epoch": 0.46, + "learning_rate": 4.775944942101814e-06, + "loss": 3.8233, + "step": 2101 + }, + { + "epoch": 0.46, + "learning_rate": 4.7758357002403325e-06, + "loss": 3.6393, + "step": 2102 + }, + { + "epoch": 0.46, + "learning_rate": 4.775726458378851e-06, + "loss": 3.8795, + "step": 2103 + }, + { + "epoch": 0.46, + "learning_rate": 4.77561721651737e-06, + "loss": 3.7787, + "step": 2104 + }, + { + "epoch": 0.46, + "learning_rate": 4.775507974655889e-06, + "loss": 3.368, + "step": 2105 + }, + { + "epoch": 0.46, + "learning_rate": 4.775398732794408e-06, + "loss": 3.7844, + "step": 2106 + }, + { + "epoch": 0.46, + "learning_rate": 4.775289490932926e-06, + "loss": 4.0252, + "step": 2107 + }, + { + "epoch": 0.46, + "learning_rate": 4.775180249071444e-06, + "loss": 3.7253, + "step": 2108 + }, + { + "epoch": 0.46, + "learning_rate": 4.775071007209963e-06, + "loss": 3.5597, + "step": 2109 + }, + { + "epoch": 0.46, + "learning_rate": 4.774961765348482e-06, + "loss": 3.5737, + "step": 2110 + }, + { + "epoch": 0.46, + "learning_rate": 4.774852523487e-06, + "loss": 3.7669, + "step": 2111 + }, + { + "epoch": 0.46, + "learning_rate": 4.774743281625519e-06, + "loss": 3.8708, + "step": 2112 + }, + { + "epoch": 0.46, + "learning_rate": 4.774634039764038e-06, + "loss": 3.5219, + "step": 2113 + }, + { + "epoch": 0.46, + "learning_rate": 4.774524797902557e-06, + "loss": 3.8066, + "step": 2114 + }, + { + "epoch": 0.46, + "learning_rate": 4.774415556041076e-06, + "loss": 3.4237, + "step": 2115 + }, + { + "epoch": 0.46, + "learning_rate": 4.774306314179594e-06, + "loss": 3.6657, + "step": 2116 + }, + { + "epoch": 0.46, + "learning_rate": 4.774197072318112e-06, + "loss": 3.8032, + "step": 2117 + }, + { + "epoch": 0.46, + "learning_rate": 4.774087830456631e-06, + "loss": 3.7163, + "step": 2118 + }, + { + "epoch": 0.46, + "learning_rate": 4.77397858859515e-06, + "loss": 3.7734, + "step": 2119 + }, + { + "epoch": 0.46, + "learning_rate": 4.773869346733669e-06, + "loss": 3.5779, + "step": 2120 + }, + { + "epoch": 0.46, + "learning_rate": 4.773760104872187e-06, + "loss": 3.4709, + "step": 2121 + }, + { + "epoch": 0.46, + "learning_rate": 4.7736508630107055e-06, + "loss": 3.6277, + "step": 2122 + }, + { + "epoch": 0.46, + "learning_rate": 4.7735416211492246e-06, + "loss": 3.5385, + "step": 2123 + }, + { + "epoch": 0.46, + "learning_rate": 4.773432379287744e-06, + "loss": 3.5115, + "step": 2124 + }, + { + "epoch": 0.46, + "learning_rate": 4.773323137426263e-06, + "loss": 3.8755, + "step": 2125 + }, + { + "epoch": 0.46, + "learning_rate": 4.773213895564781e-06, + "loss": 3.5269, + "step": 2126 + }, + { + "epoch": 0.46, + "learning_rate": 4.773104653703299e-06, + "loss": 3.989, + "step": 2127 + }, + { + "epoch": 0.46, + "learning_rate": 4.772995411841818e-06, + "loss": 3.6683, + "step": 2128 + }, + { + "epoch": 0.46, + "learning_rate": 4.772886169980337e-06, + "loss": 3.4677, + "step": 2129 + }, + { + "epoch": 0.46, + "learning_rate": 4.772776928118856e-06, + "loss": 3.1856, + "step": 2130 + }, + { + "epoch": 0.47, + "learning_rate": 4.772667686257374e-06, + "loss": 3.8118, + "step": 2131 + }, + { + "epoch": 0.47, + "learning_rate": 4.7725584443958925e-06, + "loss": 3.5423, + "step": 2132 + }, + { + "epoch": 0.47, + "learning_rate": 4.7724492025344115e-06, + "loss": 3.7051, + "step": 2133 + }, + { + "epoch": 0.47, + "learning_rate": 4.7723399606729306e-06, + "loss": 3.7559, + "step": 2134 + }, + { + "epoch": 0.47, + "learning_rate": 4.77223071881145e-06, + "loss": 3.6436, + "step": 2135 + }, + { + "epoch": 0.47, + "learning_rate": 4.772121476949968e-06, + "loss": 3.3364, + "step": 2136 + }, + { + "epoch": 0.47, + "learning_rate": 4.772012235088486e-06, + "loss": 3.3969, + "step": 2137 + }, + { + "epoch": 0.47, + "learning_rate": 4.771902993227005e-06, + "loss": 3.9413, + "step": 2138 + }, + { + "epoch": 0.47, + "learning_rate": 4.771793751365524e-06, + "loss": 3.2908, + "step": 2139 + }, + { + "epoch": 0.47, + "learning_rate": 4.771684509504042e-06, + "loss": 3.7335, + "step": 2140 + }, + { + "epoch": 0.47, + "learning_rate": 4.771575267642561e-06, + "loss": 3.671, + "step": 2141 + }, + { + "epoch": 0.47, + "learning_rate": 4.7714660257810794e-06, + "loss": 3.284, + "step": 2142 + }, + { + "epoch": 0.47, + "learning_rate": 4.7713567839195985e-06, + "loss": 3.5297, + "step": 2143 + }, + { + "epoch": 0.47, + "learning_rate": 4.7712475420581175e-06, + "loss": 3.7964, + "step": 2144 + }, + { + "epoch": 0.47, + "learning_rate": 4.771138300196636e-06, + "loss": 3.7975, + "step": 2145 + }, + { + "epoch": 0.47, + "learning_rate": 4.771029058335154e-06, + "loss": 3.5577, + "step": 2146 + }, + { + "epoch": 0.47, + "learning_rate": 4.770919816473673e-06, + "loss": 3.7752, + "step": 2147 + }, + { + "epoch": 0.47, + "learning_rate": 4.770810574612192e-06, + "loss": 3.5415, + "step": 2148 + }, + { + "epoch": 0.47, + "learning_rate": 4.770701332750711e-06, + "loss": 3.6071, + "step": 2149 + }, + { + "epoch": 0.47, + "learning_rate": 4.770592090889229e-06, + "loss": 3.685, + "step": 2150 + }, + { + "epoch": 0.47, + "learning_rate": 4.770482849027747e-06, + "loss": 3.7067, + "step": 2151 + }, + { + "epoch": 0.47, + "learning_rate": 4.770373607166266e-06, + "loss": 3.9756, + "step": 2152 + }, + { + "epoch": 0.47, + "learning_rate": 4.7702643653047854e-06, + "loss": 3.799, + "step": 2153 + }, + { + "epoch": 0.47, + "learning_rate": 4.7701551234433045e-06, + "loss": 3.841, + "step": 2154 + }, + { + "epoch": 0.47, + "learning_rate": 4.770045881581823e-06, + "loss": 3.6629, + "step": 2155 + }, + { + "epoch": 0.47, + "learning_rate": 4.769936639720341e-06, + "loss": 3.7562, + "step": 2156 + }, + { + "epoch": 0.47, + "learning_rate": 4.76982739785886e-06, + "loss": 3.8906, + "step": 2157 + }, + { + "epoch": 0.47, + "learning_rate": 4.769718155997379e-06, + "loss": 3.6567, + "step": 2158 + }, + { + "epoch": 0.47, + "learning_rate": 4.769608914135897e-06, + "loss": 3.7091, + "step": 2159 + }, + { + "epoch": 0.47, + "learning_rate": 4.769499672274416e-06, + "loss": 3.8881, + "step": 2160 + }, + { + "epoch": 0.47, + "learning_rate": 4.769390430412934e-06, + "loss": 3.5619, + "step": 2161 + }, + { + "epoch": 0.47, + "learning_rate": 4.769281188551453e-06, + "loss": 3.8291, + "step": 2162 + }, + { + "epoch": 0.47, + "learning_rate": 4.769171946689972e-06, + "loss": 3.6639, + "step": 2163 + }, + { + "epoch": 0.47, + "learning_rate": 4.7690627048284906e-06, + "loss": 3.7541, + "step": 2164 + }, + { + "epoch": 0.47, + "learning_rate": 4.76895346296701e-06, + "loss": 3.3847, + "step": 2165 + }, + { + "epoch": 0.47, + "learning_rate": 4.768844221105528e-06, + "loss": 3.5896, + "step": 2166 + }, + { + "epoch": 0.47, + "learning_rate": 4.768734979244047e-06, + "loss": 3.662, + "step": 2167 + }, + { + "epoch": 0.47, + "learning_rate": 4.768625737382566e-06, + "loss": 3.2028, + "step": 2168 + }, + { + "epoch": 0.47, + "learning_rate": 4.768516495521084e-06, + "loss": 4.0833, + "step": 2169 + }, + { + "epoch": 0.47, + "learning_rate": 4.768407253659603e-06, + "loss": 3.8001, + "step": 2170 + }, + { + "epoch": 0.47, + "learning_rate": 4.768298011798121e-06, + "loss": 3.5344, + "step": 2171 + }, + { + "epoch": 0.47, + "learning_rate": 4.76818876993664e-06, + "loss": 3.6466, + "step": 2172 + }, + { + "epoch": 0.47, + "learning_rate": 4.768079528075159e-06, + "loss": 3.5952, + "step": 2173 + }, + { + "epoch": 0.47, + "learning_rate": 4.7679702862136775e-06, + "loss": 3.5887, + "step": 2174 + }, + { + "epoch": 0.47, + "learning_rate": 4.767861044352196e-06, + "loss": 3.6696, + "step": 2175 + }, + { + "epoch": 0.47, + "learning_rate": 4.767751802490715e-06, + "loss": 3.5521, + "step": 2176 + }, + { + "epoch": 0.48, + "learning_rate": 4.767642560629234e-06, + "loss": 3.6797, + "step": 2177 + }, + { + "epoch": 0.48, + "learning_rate": 4.767533318767752e-06, + "loss": 3.8583, + "step": 2178 + }, + { + "epoch": 0.48, + "learning_rate": 4.767424076906271e-06, + "loss": 3.7599, + "step": 2179 + }, + { + "epoch": 0.48, + "learning_rate": 4.767314835044789e-06, + "loss": 4.0056, + "step": 2180 + }, + { + "epoch": 0.48, + "learning_rate": 4.767205593183308e-06, + "loss": 3.6527, + "step": 2181 + }, + { + "epoch": 0.48, + "learning_rate": 4.767096351321827e-06, + "loss": 3.465, + "step": 2182 + }, + { + "epoch": 0.48, + "learning_rate": 4.7669871094603455e-06, + "loss": 3.633, + "step": 2183 + }, + { + "epoch": 0.48, + "learning_rate": 4.7668778675988645e-06, + "loss": 3.7425, + "step": 2184 + }, + { + "epoch": 0.48, + "learning_rate": 4.766768625737383e-06, + "loss": 3.8548, + "step": 2185 + }, + { + "epoch": 0.48, + "learning_rate": 4.766659383875902e-06, + "loss": 3.9234, + "step": 2186 + }, + { + "epoch": 0.48, + "learning_rate": 4.766550142014421e-06, + "loss": 3.4989, + "step": 2187 + }, + { + "epoch": 0.48, + "learning_rate": 4.766440900152939e-06, + "loss": 3.4174, + "step": 2188 + }, + { + "epoch": 0.48, + "learning_rate": 4.766331658291458e-06, + "loss": 3.435, + "step": 2189 + }, + { + "epoch": 0.48, + "learning_rate": 4.766222416429976e-06, + "loss": 3.3671, + "step": 2190 + }, + { + "epoch": 0.48, + "learning_rate": 4.766113174568495e-06, + "loss": 3.6667, + "step": 2191 + }, + { + "epoch": 0.48, + "learning_rate": 4.766003932707013e-06, + "loss": 3.6586, + "step": 2192 + }, + { + "epoch": 0.48, + "learning_rate": 4.765894690845532e-06, + "loss": 3.279, + "step": 2193 + }, + { + "epoch": 0.48, + "learning_rate": 4.7657854489840514e-06, + "loss": 3.6459, + "step": 2194 + }, + { + "epoch": 0.48, + "learning_rate": 4.76567620712257e-06, + "loss": 3.9227, + "step": 2195 + }, + { + "epoch": 0.48, + "learning_rate": 4.765566965261089e-06, + "loss": 3.4158, + "step": 2196 + }, + { + "epoch": 0.48, + "learning_rate": 4.765457723399607e-06, + "loss": 3.3744, + "step": 2197 + }, + { + "epoch": 0.48, + "learning_rate": 4.765348481538126e-06, + "loss": 3.603, + "step": 2198 + }, + { + "epoch": 0.48, + "learning_rate": 4.765239239676644e-06, + "loss": 3.3638, + "step": 2199 + }, + { + "epoch": 0.48, + "learning_rate": 4.765129997815163e-06, + "loss": 3.8136, + "step": 2200 + }, + { + "epoch": 0.48, + "learning_rate": 4.765020755953682e-06, + "loss": 3.8311, + "step": 2201 + }, + { + "epoch": 0.48, + "learning_rate": 4.7649115140922e-06, + "loss": 3.4152, + "step": 2202 + }, + { + "epoch": 0.48, + "learning_rate": 4.764802272230719e-06, + "loss": 3.8499, + "step": 2203 + }, + { + "epoch": 0.48, + "learning_rate": 4.7646930303692376e-06, + "loss": 3.3662, + "step": 2204 + }, + { + "epoch": 0.48, + "learning_rate": 4.764583788507757e-06, + "loss": 4.0637, + "step": 2205 + }, + { + "epoch": 0.48, + "learning_rate": 4.764474546646276e-06, + "loss": 3.6272, + "step": 2206 + }, + { + "epoch": 0.48, + "learning_rate": 4.764365304784794e-06, + "loss": 3.3737, + "step": 2207 + }, + { + "epoch": 0.48, + "learning_rate": 4.764256062923313e-06, + "loss": 3.6541, + "step": 2208 + }, + { + "epoch": 0.48, + "learning_rate": 4.764146821061831e-06, + "loss": 3.6218, + "step": 2209 + }, + { + "epoch": 0.48, + "learning_rate": 4.76403757920035e-06, + "loss": 4.051, + "step": 2210 + }, + { + "epoch": 0.48, + "learning_rate": 4.763928337338868e-06, + "loss": 3.8324, + "step": 2211 + }, + { + "epoch": 0.48, + "learning_rate": 4.763819095477387e-06, + "loss": 3.3972, + "step": 2212 + }, + { + "epoch": 0.48, + "learning_rate": 4.763709853615906e-06, + "loss": 3.4646, + "step": 2213 + }, + { + "epoch": 0.48, + "learning_rate": 4.7636006117544245e-06, + "loss": 3.7447, + "step": 2214 + }, + { + "epoch": 0.48, + "learning_rate": 4.7634913698929435e-06, + "loss": 3.3296, + "step": 2215 + }, + { + "epoch": 0.48, + "learning_rate": 4.763382128031462e-06, + "loss": 3.5134, + "step": 2216 + }, + { + "epoch": 0.48, + "learning_rate": 4.763272886169981e-06, + "loss": 3.6783, + "step": 2217 + }, + { + "epoch": 0.48, + "learning_rate": 4.7631636443085e-06, + "loss": 3.9814, + "step": 2218 + }, + { + "epoch": 0.48, + "learning_rate": 4.763054402447018e-06, + "loss": 3.8981, + "step": 2219 + }, + { + "epoch": 0.48, + "learning_rate": 4.762945160585537e-06, + "loss": 3.6452, + "step": 2220 + }, + { + "epoch": 0.48, + "learning_rate": 4.762835918724055e-06, + "loss": 3.9084, + "step": 2221 + }, + { + "epoch": 0.48, + "learning_rate": 4.762726676862574e-06, + "loss": 2.9984, + "step": 2222 + }, + { + "epoch": 0.49, + "learning_rate": 4.762617435001093e-06, + "loss": 3.5918, + "step": 2223 + }, + { + "epoch": 0.49, + "learning_rate": 4.7625081931396115e-06, + "loss": 3.51, + "step": 2224 + }, + { + "epoch": 0.49, + "learning_rate": 4.76239895127813e-06, + "loss": 3.6199, + "step": 2225 + }, + { + "epoch": 0.49, + "learning_rate": 4.762289709416649e-06, + "loss": 3.4239, + "step": 2226 + }, + { + "epoch": 0.49, + "learning_rate": 4.762180467555168e-06, + "loss": 3.3828, + "step": 2227 + }, + { + "epoch": 0.49, + "learning_rate": 4.762071225693686e-06, + "loss": 3.5117, + "step": 2228 + }, + { + "epoch": 0.49, + "learning_rate": 4.761961983832205e-06, + "loss": 3.5195, + "step": 2229 + }, + { + "epoch": 0.49, + "learning_rate": 4.761852741970723e-06, + "loss": 3.5826, + "step": 2230 + }, + { + "epoch": 0.49, + "learning_rate": 4.761743500109242e-06, + "loss": 3.9016, + "step": 2231 + }, + { + "epoch": 0.49, + "learning_rate": 4.761634258247761e-06, + "loss": 3.2607, + "step": 2232 + }, + { + "epoch": 0.49, + "learning_rate": 4.761525016386279e-06, + "loss": 3.7309, + "step": 2233 + }, + { + "epoch": 0.49, + "learning_rate": 4.7614157745247984e-06, + "loss": 3.514, + "step": 2234 + }, + { + "epoch": 0.49, + "learning_rate": 4.761306532663317e-06, + "loss": 3.5089, + "step": 2235 + }, + { + "epoch": 0.49, + "learning_rate": 4.761197290801836e-06, + "loss": 3.8117, + "step": 2236 + }, + { + "epoch": 0.49, + "learning_rate": 4.761088048940355e-06, + "loss": 3.9239, + "step": 2237 + }, + { + "epoch": 0.49, + "learning_rate": 4.760978807078873e-06, + "loss": 3.7301, + "step": 2238 + }, + { + "epoch": 0.49, + "learning_rate": 4.760869565217392e-06, + "loss": 3.5705, + "step": 2239 + }, + { + "epoch": 0.49, + "learning_rate": 4.76076032335591e-06, + "loss": 3.6101, + "step": 2240 + }, + { + "epoch": 0.49, + "learning_rate": 4.760651081494429e-06, + "loss": 3.7485, + "step": 2241 + }, + { + "epoch": 0.49, + "learning_rate": 4.760541839632948e-06, + "loss": 3.7909, + "step": 2242 + }, + { + "epoch": 0.49, + "learning_rate": 4.760432597771466e-06, + "loss": 3.727, + "step": 2243 + }, + { + "epoch": 0.49, + "learning_rate": 4.7603233559099845e-06, + "loss": 3.5494, + "step": 2244 + }, + { + "epoch": 0.49, + "learning_rate": 4.7602141140485036e-06, + "loss": 3.518, + "step": 2245 + }, + { + "epoch": 0.49, + "learning_rate": 4.760104872187023e-06, + "loss": 3.6373, + "step": 2246 + }, + { + "epoch": 0.49, + "learning_rate": 4.759995630325542e-06, + "loss": 3.5489, + "step": 2247 + }, + { + "epoch": 0.49, + "learning_rate": 4.75988638846406e-06, + "loss": 3.6314, + "step": 2248 + }, + { + "epoch": 0.49, + "learning_rate": 4.759777146602578e-06, + "loss": 3.8255, + "step": 2249 + }, + { + "epoch": 0.49, + "learning_rate": 4.759667904741097e-06, + "loss": 3.7834, + "step": 2250 + }, + { + "epoch": 0.49, + "learning_rate": 4.759558662879616e-06, + "loss": 3.4578, + "step": 2251 + }, + { + "epoch": 0.49, + "learning_rate": 4.759449421018134e-06, + "loss": 3.493, + "step": 2252 + }, + { + "epoch": 0.49, + "learning_rate": 4.759340179156653e-06, + "loss": 3.5567, + "step": 2253 + }, + { + "epoch": 0.49, + "learning_rate": 4.7592309372951715e-06, + "loss": 4.0084, + "step": 2254 + }, + { + "epoch": 0.49, + "learning_rate": 4.7591216954336905e-06, + "loss": 3.392, + "step": 2255 + }, + { + "epoch": 0.49, + "learning_rate": 4.7590124535722096e-06, + "loss": 3.8227, + "step": 2256 + }, + { + "epoch": 0.49, + "learning_rate": 4.758903211710728e-06, + "loss": 3.416, + "step": 2257 + }, + { + "epoch": 0.49, + "learning_rate": 4.758793969849246e-06, + "loss": 3.7412, + "step": 2258 + }, + { + "epoch": 0.49, + "learning_rate": 4.758684727987765e-06, + "loss": 3.8743, + "step": 2259 + }, + { + "epoch": 0.49, + "learning_rate": 4.758575486126284e-06, + "loss": 3.2015, + "step": 2260 + }, + { + "epoch": 0.49, + "learning_rate": 4.758466244264803e-06, + "loss": 3.7294, + "step": 2261 + }, + { + "epoch": 0.49, + "learning_rate": 4.758357002403321e-06, + "loss": 3.6877, + "step": 2262 + }, + { + "epoch": 0.49, + "learning_rate": 4.758247760541839e-06, + "loss": 3.0962, + "step": 2263 + }, + { + "epoch": 0.49, + "learning_rate": 4.7581385186803584e-06, + "loss": 3.799, + "step": 2264 + }, + { + "epoch": 0.49, + "learning_rate": 4.7580292768188775e-06, + "loss": 3.7233, + "step": 2265 + }, + { + "epoch": 0.49, + "learning_rate": 4.7579200349573965e-06, + "loss": 3.213, + "step": 2266 + }, + { + "epoch": 0.49, + "learning_rate": 4.757810793095915e-06, + "loss": 3.845, + "step": 2267 + }, + { + "epoch": 0.49, + "learning_rate": 4.757701551234433e-06, + "loss": 3.8366, + "step": 2268 + }, + { + "epoch": 0.5, + "learning_rate": 4.757592309372952e-06, + "loss": 3.6493, + "step": 2269 + }, + { + "epoch": 0.5, + "learning_rate": 4.757483067511471e-06, + "loss": 3.6263, + "step": 2270 + }, + { + "epoch": 0.5, + "learning_rate": 4.75737382564999e-06, + "loss": 3.1532, + "step": 2271 + }, + { + "epoch": 0.5, + "learning_rate": 4.757264583788508e-06, + "loss": 4.0771, + "step": 2272 + }, + { + "epoch": 0.5, + "learning_rate": 4.757155341927026e-06, + "loss": 3.9329, + "step": 2273 + }, + { + "epoch": 0.5, + "learning_rate": 4.757046100065545e-06, + "loss": 3.5867, + "step": 2274 + }, + { + "epoch": 0.5, + "learning_rate": 4.7569368582040644e-06, + "loss": 3.5383, + "step": 2275 + }, + { + "epoch": 0.5, + "learning_rate": 4.7568276163425835e-06, + "loss": 3.3353, + "step": 2276 + }, + { + "epoch": 0.5, + "learning_rate": 4.756718374481102e-06, + "loss": 3.5056, + "step": 2277 + }, + { + "epoch": 0.5, + "learning_rate": 4.75660913261962e-06, + "loss": 3.5987, + "step": 2278 + }, + { + "epoch": 0.5, + "learning_rate": 4.756499890758139e-06, + "loss": 3.8082, + "step": 2279 + }, + { + "epoch": 0.5, + "learning_rate": 4.756390648896658e-06, + "loss": 3.9316, + "step": 2280 + }, + { + "epoch": 0.5, + "learning_rate": 4.756281407035176e-06, + "loss": 4.0946, + "step": 2281 + }, + { + "epoch": 0.5, + "learning_rate": 4.756172165173695e-06, + "loss": 3.9196, + "step": 2282 + }, + { + "epoch": 0.5, + "learning_rate": 4.756062923312213e-06, + "loss": 3.6739, + "step": 2283 + }, + { + "epoch": 0.5, + "learning_rate": 4.755953681450732e-06, + "loss": 3.3008, + "step": 2284 + }, + { + "epoch": 0.5, + "learning_rate": 4.755844439589251e-06, + "loss": 3.6261, + "step": 2285 + }, + { + "epoch": 0.5, + "learning_rate": 4.75573519772777e-06, + "loss": 3.5845, + "step": 2286 + }, + { + "epoch": 0.5, + "learning_rate": 4.755625955866288e-06, + "loss": 3.8895, + "step": 2287 + }, + { + "epoch": 0.5, + "learning_rate": 4.755516714004807e-06, + "loss": 3.4421, + "step": 2288 + }, + { + "epoch": 0.5, + "learning_rate": 4.755407472143326e-06, + "loss": 3.6269, + "step": 2289 + }, + { + "epoch": 0.5, + "learning_rate": 4.755298230281845e-06, + "loss": 3.7405, + "step": 2290 + }, + { + "epoch": 0.5, + "learning_rate": 4.755188988420363e-06, + "loss": 3.4672, + "step": 2291 + }, + { + "epoch": 0.5, + "learning_rate": 4.755079746558881e-06, + "loss": 3.4951, + "step": 2292 + }, + { + "epoch": 0.5, + "learning_rate": 4.7549705046974e-06, + "loss": 3.8199, + "step": 2293 + }, + { + "epoch": 0.5, + "learning_rate": 4.754861262835919e-06, + "loss": 3.8825, + "step": 2294 + }, + { + "epoch": 0.5, + "learning_rate": 4.754752020974438e-06, + "loss": 3.4814, + "step": 2295 + }, + { + "epoch": 0.5, + "learning_rate": 4.7546427791129565e-06, + "loss": 3.7054, + "step": 2296 + }, + { + "epoch": 0.5, + "learning_rate": 4.754533537251475e-06, + "loss": 3.7163, + "step": 2297 + }, + { + "epoch": 0.5, + "learning_rate": 4.754424295389994e-06, + "loss": 3.4984, + "step": 2298 + }, + { + "epoch": 0.5, + "learning_rate": 4.754315053528513e-06, + "loss": 3.6129, + "step": 2299 + }, + { + "epoch": 0.5, + "learning_rate": 4.754205811667032e-06, + "loss": 3.9559, + "step": 2300 + }, + { + "epoch": 0.5, + "learning_rate": 4.75409656980555e-06, + "loss": 3.4171, + "step": 2301 + }, + { + "epoch": 0.5, + "learning_rate": 4.753987327944068e-06, + "loss": 3.5877, + "step": 2302 + }, + { + "epoch": 0.5, + "learning_rate": 4.753878086082587e-06, + "loss": 3.7576, + "step": 2303 + }, + { + "epoch": 0.5, + "learning_rate": 4.753768844221106e-06, + "loss": 3.4068, + "step": 2304 + }, + { + "epoch": 0.5, + "learning_rate": 4.7536596023596245e-06, + "loss": 3.8016, + "step": 2305 + }, + { + "epoch": 0.5, + "learning_rate": 4.7535503604981435e-06, + "loss": 3.8309, + "step": 2306 + }, + { + "epoch": 0.5, + "learning_rate": 4.753441118636662e-06, + "loss": 3.6503, + "step": 2307 + }, + { + "epoch": 0.5, + "learning_rate": 4.753331876775181e-06, + "loss": 3.8034, + "step": 2308 + }, + { + "epoch": 0.5, + "learning_rate": 4.7532226349137e-06, + "loss": 3.4667, + "step": 2309 + }, + { + "epoch": 0.5, + "learning_rate": 4.753113393052218e-06, + "loss": 3.9654, + "step": 2310 + }, + { + "epoch": 0.5, + "learning_rate": 4.753004151190736e-06, + "loss": 3.9639, + "step": 2311 + }, + { + "epoch": 0.5, + "learning_rate": 4.752894909329255e-06, + "loss": 3.5736, + "step": 2312 + }, + { + "epoch": 0.5, + "learning_rate": 4.752785667467774e-06, + "loss": 3.6681, + "step": 2313 + }, + { + "epoch": 0.51, + "learning_rate": 4.752676425606293e-06, + "loss": 3.5676, + "step": 2314 + }, + { + "epoch": 0.51, + "learning_rate": 4.752567183744811e-06, + "loss": 3.7144, + "step": 2315 + }, + { + "epoch": 0.51, + "learning_rate": 4.75245794188333e-06, + "loss": 3.941, + "step": 2316 + }, + { + "epoch": 0.51, + "learning_rate": 4.752348700021849e-06, + "loss": 3.5857, + "step": 2317 + }, + { + "epoch": 0.51, + "learning_rate": 4.752239458160368e-06, + "loss": 3.4684, + "step": 2318 + }, + { + "epoch": 0.51, + "learning_rate": 4.752130216298887e-06, + "loss": 3.5287, + "step": 2319 + }, + { + "epoch": 0.51, + "learning_rate": 4.752020974437405e-06, + "loss": 3.7431, + "step": 2320 + }, + { + "epoch": 0.51, + "learning_rate": 4.751911732575923e-06, + "loss": 3.611, + "step": 2321 + }, + { + "epoch": 0.51, + "learning_rate": 4.751802490714442e-06, + "loss": 3.1768, + "step": 2322 + }, + { + "epoch": 0.51, + "learning_rate": 4.751693248852961e-06, + "loss": 3.6719, + "step": 2323 + }, + { + "epoch": 0.51, + "learning_rate": 4.751584006991479e-06, + "loss": 3.7509, + "step": 2324 + }, + { + "epoch": 0.51, + "learning_rate": 4.751474765129998e-06, + "loss": 3.5792, + "step": 2325 + }, + { + "epoch": 0.51, + "learning_rate": 4.7513655232685166e-06, + "loss": 3.5481, + "step": 2326 + }, + { + "epoch": 0.51, + "learning_rate": 4.751256281407036e-06, + "loss": 3.8801, + "step": 2327 + }, + { + "epoch": 0.51, + "learning_rate": 4.751147039545555e-06, + "loss": 3.8585, + "step": 2328 + }, + { + "epoch": 0.51, + "learning_rate": 4.751037797684073e-06, + "loss": 3.4941, + "step": 2329 + }, + { + "epoch": 0.51, + "learning_rate": 4.750928555822592e-06, + "loss": 3.7056, + "step": 2330 + }, + { + "epoch": 0.51, + "learning_rate": 4.75081931396111e-06, + "loss": 3.3504, + "step": 2331 + }, + { + "epoch": 0.51, + "learning_rate": 4.750710072099629e-06, + "loss": 3.8478, + "step": 2332 + }, + { + "epoch": 0.51, + "learning_rate": 4.750600830238148e-06, + "loss": 3.5795, + "step": 2333 + }, + { + "epoch": 0.51, + "learning_rate": 4.750491588376666e-06, + "loss": 3.8506, + "step": 2334 + }, + { + "epoch": 0.51, + "learning_rate": 4.750382346515185e-06, + "loss": 3.6729, + "step": 2335 + }, + { + "epoch": 0.51, + "learning_rate": 4.7502731046537035e-06, + "loss": 3.5921, + "step": 2336 + }, + { + "epoch": 0.51, + "learning_rate": 4.7501638627922226e-06, + "loss": 3.6962, + "step": 2337 + }, + { + "epoch": 0.51, + "learning_rate": 4.750054620930742e-06, + "loss": 3.9137, + "step": 2338 + }, + { + "epoch": 0.51, + "learning_rate": 4.74994537906926e-06, + "loss": 3.6969, + "step": 2339 + }, + { + "epoch": 0.51, + "learning_rate": 4.749836137207778e-06, + "loss": 3.41, + "step": 2340 + }, + { + "epoch": 0.51, + "learning_rate": 4.749726895346297e-06, + "loss": 3.9295, + "step": 2341 + }, + { + "epoch": 0.51, + "learning_rate": 4.749617653484816e-06, + "loss": 3.5066, + "step": 2342 + }, + { + "epoch": 0.51, + "learning_rate": 4.749508411623334e-06, + "loss": 3.7361, + "step": 2343 + }, + { + "epoch": 0.51, + "learning_rate": 4.749399169761853e-06, + "loss": 3.6226, + "step": 2344 + }, + { + "epoch": 0.51, + "learning_rate": 4.7492899279003714e-06, + "loss": 3.7989, + "step": 2345 + }, + { + "epoch": 0.51, + "learning_rate": 4.7491806860388905e-06, + "loss": 3.764, + "step": 2346 + }, + { + "epoch": 0.51, + "learning_rate": 4.7490714441774095e-06, + "loss": 3.5563, + "step": 2347 + }, + { + "epoch": 0.51, + "learning_rate": 4.748962202315928e-06, + "loss": 3.9007, + "step": 2348 + }, + { + "epoch": 0.51, + "learning_rate": 4.748852960454447e-06, + "loss": 3.2587, + "step": 2349 + }, + { + "epoch": 0.51, + "learning_rate": 4.748743718592965e-06, + "loss": 3.6104, + "step": 2350 + }, + { + "epoch": 0.51, + "learning_rate": 4.748634476731484e-06, + "loss": 3.7826, + "step": 2351 + }, + { + "epoch": 0.51, + "learning_rate": 4.748525234870003e-06, + "loss": 3.7714, + "step": 2352 + }, + { + "epoch": 0.51, + "learning_rate": 4.748415993008521e-06, + "loss": 3.4352, + "step": 2353 + }, + { + "epoch": 0.51, + "learning_rate": 4.74830675114704e-06, + "loss": 3.8165, + "step": 2354 + }, + { + "epoch": 0.51, + "learning_rate": 4.748197509285558e-06, + "loss": 3.6973, + "step": 2355 + }, + { + "epoch": 0.51, + "learning_rate": 4.7480882674240774e-06, + "loss": 3.3249, + "step": 2356 + }, + { + "epoch": 0.51, + "learning_rate": 4.747979025562596e-06, + "loss": 3.4098, + "step": 2357 + }, + { + "epoch": 0.51, + "learning_rate": 4.747869783701115e-06, + "loss": 3.6457, + "step": 2358 + }, + { + "epoch": 0.51, + "learning_rate": 4.747760541839634e-06, + "loss": 3.5972, + "step": 2359 + }, + { + "epoch": 0.52, + "learning_rate": 4.747651299978152e-06, + "loss": 3.8284, + "step": 2360 + }, + { + "epoch": 0.52, + "learning_rate": 4.747542058116671e-06, + "loss": 3.5018, + "step": 2361 + }, + { + "epoch": 0.52, + "learning_rate": 4.747432816255189e-06, + "loss": 3.3714, + "step": 2362 + }, + { + "epoch": 0.52, + "learning_rate": 4.747323574393708e-06, + "loss": 3.6182, + "step": 2363 + }, + { + "epoch": 0.52, + "learning_rate": 4.747214332532226e-06, + "loss": 3.593, + "step": 2364 + }, + { + "epoch": 0.52, + "learning_rate": 4.747105090670745e-06, + "loss": 3.6185, + "step": 2365 + }, + { + "epoch": 0.52, + "learning_rate": 4.746995848809264e-06, + "loss": 3.4452, + "step": 2366 + }, + { + "epoch": 0.52, + "learning_rate": 4.746886606947783e-06, + "loss": 4.0011, + "step": 2367 + }, + { + "epoch": 0.52, + "learning_rate": 4.746777365086302e-06, + "loss": 3.5813, + "step": 2368 + }, + { + "epoch": 0.52, + "learning_rate": 4.74666812322482e-06, + "loss": 3.7322, + "step": 2369 + }, + { + "epoch": 0.52, + "learning_rate": 4.746558881363339e-06, + "loss": 3.4045, + "step": 2370 + }, + { + "epoch": 0.52, + "learning_rate": 4.746449639501858e-06, + "loss": 3.5661, + "step": 2371 + }, + { + "epoch": 0.52, + "learning_rate": 4.746340397640376e-06, + "loss": 3.4826, + "step": 2372 + }, + { + "epoch": 0.52, + "learning_rate": 4.746231155778895e-06, + "loss": 3.8129, + "step": 2373 + }, + { + "epoch": 0.52, + "learning_rate": 4.746121913917413e-06, + "loss": 3.6034, + "step": 2374 + }, + { + "epoch": 0.52, + "learning_rate": 4.746012672055932e-06, + "loss": 3.5798, + "step": 2375 + }, + { + "epoch": 0.52, + "learning_rate": 4.7459034301944505e-06, + "loss": 4.0216, + "step": 2376 + }, + { + "epoch": 0.52, + "learning_rate": 4.7457941883329695e-06, + "loss": 3.9198, + "step": 2377 + }, + { + "epoch": 0.52, + "learning_rate": 4.7456849464714886e-06, + "loss": 3.741, + "step": 2378 + }, + { + "epoch": 0.52, + "learning_rate": 4.745575704610007e-06, + "loss": 3.7807, + "step": 2379 + }, + { + "epoch": 0.52, + "learning_rate": 4.745466462748526e-06, + "loss": 4.0445, + "step": 2380 + }, + { + "epoch": 0.52, + "learning_rate": 4.745357220887044e-06, + "loss": 3.5389, + "step": 2381 + }, + { + "epoch": 0.52, + "learning_rate": 4.745247979025563e-06, + "loss": 3.5164, + "step": 2382 + }, + { + "epoch": 0.52, + "learning_rate": 4.745138737164082e-06, + "loss": 3.6726, + "step": 2383 + }, + { + "epoch": 0.52, + "learning_rate": 4.7450294953026e-06, + "loss": 3.4489, + "step": 2384 + }, + { + "epoch": 0.52, + "learning_rate": 4.744920253441119e-06, + "loss": 3.9162, + "step": 2385 + }, + { + "epoch": 0.52, + "learning_rate": 4.7448110115796375e-06, + "loss": 3.6656, + "step": 2386 + }, + { + "epoch": 0.52, + "learning_rate": 4.7447017697181565e-06, + "loss": 3.6356, + "step": 2387 + }, + { + "epoch": 0.52, + "learning_rate": 4.7445925278566755e-06, + "loss": 3.8048, + "step": 2388 + }, + { + "epoch": 0.52, + "learning_rate": 4.744483285995194e-06, + "loss": 3.7304, + "step": 2389 + }, + { + "epoch": 0.52, + "learning_rate": 4.744374044133712e-06, + "loss": 3.2463, + "step": 2390 + }, + { + "epoch": 0.52, + "learning_rate": 4.744264802272231e-06, + "loss": 3.5287, + "step": 2391 + }, + { + "epoch": 0.52, + "learning_rate": 4.74415556041075e-06, + "loss": 3.4318, + "step": 2392 + }, + { + "epoch": 0.52, + "learning_rate": 4.744046318549268e-06, + "loss": 4.0171, + "step": 2393 + }, + { + "epoch": 0.52, + "learning_rate": 4.743937076687787e-06, + "loss": 3.5271, + "step": 2394 + }, + { + "epoch": 0.52, + "learning_rate": 4.743827834826305e-06, + "loss": 3.845, + "step": 2395 + }, + { + "epoch": 0.52, + "learning_rate": 4.743718592964824e-06, + "loss": 3.9024, + "step": 2396 + }, + { + "epoch": 0.52, + "learning_rate": 4.7436093511033435e-06, + "loss": 3.4783, + "step": 2397 + }, + { + "epoch": 0.52, + "learning_rate": 4.743500109241862e-06, + "loss": 3.767, + "step": 2398 + }, + { + "epoch": 0.52, + "learning_rate": 4.743390867380381e-06, + "loss": 3.594, + "step": 2399 + }, + { + "epoch": 0.52, + "learning_rate": 4.743281625518899e-06, + "loss": 3.7838, + "step": 2400 + }, + { + "epoch": 0.52, + "learning_rate": 4.743172383657418e-06, + "loss": 3.9186, + "step": 2401 + }, + { + "epoch": 0.52, + "learning_rate": 4.743063141795937e-06, + "loss": 3.6628, + "step": 2402 + }, + { + "epoch": 0.52, + "learning_rate": 4.742953899934455e-06, + "loss": 3.5923, + "step": 2403 + }, + { + "epoch": 0.52, + "learning_rate": 4.742844658072974e-06, + "loss": 3.3989, + "step": 2404 + }, + { + "epoch": 0.52, + "learning_rate": 4.742735416211492e-06, + "loss": 3.5856, + "step": 2405 + }, + { + "epoch": 0.53, + "learning_rate": 4.742626174350011e-06, + "loss": 3.8813, + "step": 2406 + }, + { + "epoch": 0.53, + "learning_rate": 4.74251693248853e-06, + "loss": 3.7124, + "step": 2407 + }, + { + "epoch": 0.53, + "learning_rate": 4.742407690627049e-06, + "loss": 3.4638, + "step": 2408 + }, + { + "epoch": 0.53, + "learning_rate": 4.742298448765567e-06, + "loss": 3.4625, + "step": 2409 + }, + { + "epoch": 0.53, + "learning_rate": 4.742189206904086e-06, + "loss": 3.1787, + "step": 2410 + }, + { + "epoch": 0.53, + "learning_rate": 4.742079965042605e-06, + "loss": 3.5547, + "step": 2411 + }, + { + "epoch": 0.53, + "learning_rate": 4.741970723181124e-06, + "loss": 3.8622, + "step": 2412 + }, + { + "epoch": 0.53, + "learning_rate": 4.741861481319642e-06, + "loss": 3.8127, + "step": 2413 + }, + { + "epoch": 0.53, + "learning_rate": 4.74175223945816e-06, + "loss": 3.7863, + "step": 2414 + }, + { + "epoch": 0.53, + "learning_rate": 4.741642997596679e-06, + "loss": 3.5352, + "step": 2415 + }, + { + "epoch": 0.53, + "learning_rate": 4.741533755735198e-06, + "loss": 3.1789, + "step": 2416 + }, + { + "epoch": 0.53, + "learning_rate": 4.7414245138737165e-06, + "loss": 3.724, + "step": 2417 + }, + { + "epoch": 0.53, + "learning_rate": 4.7413152720122356e-06, + "loss": 4.033, + "step": 2418 + }, + { + "epoch": 0.53, + "learning_rate": 4.741206030150754e-06, + "loss": 3.6725, + "step": 2419 + }, + { + "epoch": 0.53, + "learning_rate": 4.741096788289273e-06, + "loss": 3.7096, + "step": 2420 + }, + { + "epoch": 0.53, + "learning_rate": 4.740987546427792e-06, + "loss": 3.8004, + "step": 2421 + }, + { + "epoch": 0.53, + "learning_rate": 4.74087830456631e-06, + "loss": 3.6869, + "step": 2422 + }, + { + "epoch": 0.53, + "learning_rate": 4.740769062704829e-06, + "loss": 4.2027, + "step": 2423 + }, + { + "epoch": 0.53, + "learning_rate": 4.740659820843347e-06, + "loss": 3.4923, + "step": 2424 + }, + { + "epoch": 0.53, + "learning_rate": 4.740550578981866e-06, + "loss": 3.5434, + "step": 2425 + }, + { + "epoch": 0.53, + "learning_rate": 4.740441337120385e-06, + "loss": 3.1379, + "step": 2426 + }, + { + "epoch": 0.53, + "learning_rate": 4.7403320952589035e-06, + "loss": 3.4621, + "step": 2427 + }, + { + "epoch": 0.53, + "learning_rate": 4.740222853397422e-06, + "loss": 3.4048, + "step": 2428 + }, + { + "epoch": 0.53, + "learning_rate": 4.740113611535941e-06, + "loss": 3.5398, + "step": 2429 + }, + { + "epoch": 0.53, + "learning_rate": 4.74000436967446e-06, + "loss": 3.7486, + "step": 2430 + }, + { + "epoch": 0.53, + "learning_rate": 4.739895127812979e-06, + "loss": 3.4314, + "step": 2431 + }, + { + "epoch": 0.53, + "learning_rate": 4.739785885951497e-06, + "loss": 3.9906, + "step": 2432 + }, + { + "epoch": 0.53, + "learning_rate": 4.739676644090015e-06, + "loss": 3.2514, + "step": 2433 + }, + { + "epoch": 0.53, + "learning_rate": 4.739567402228534e-06, + "loss": 3.7255, + "step": 2434 + }, + { + "epoch": 0.53, + "learning_rate": 4.739458160367053e-06, + "loss": 3.9542, + "step": 2435 + }, + { + "epoch": 0.53, + "learning_rate": 4.739348918505572e-06, + "loss": 3.6221, + "step": 2436 + }, + { + "epoch": 0.53, + "learning_rate": 4.7392396766440904e-06, + "loss": 2.9622, + "step": 2437 + }, + { + "epoch": 0.53, + "learning_rate": 4.739130434782609e-06, + "loss": 3.9292, + "step": 2438 + }, + { + "epoch": 0.53, + "learning_rate": 4.739021192921128e-06, + "loss": 3.9503, + "step": 2439 + }, + { + "epoch": 0.53, + "learning_rate": 4.738911951059647e-06, + "loss": 3.5868, + "step": 2440 + }, + { + "epoch": 0.53, + "learning_rate": 4.738802709198166e-06, + "loss": 3.7501, + "step": 2441 + }, + { + "epoch": 0.53, + "learning_rate": 4.738693467336684e-06, + "loss": 3.8732, + "step": 2442 + }, + { + "epoch": 0.53, + "learning_rate": 4.738584225475202e-06, + "loss": 3.129, + "step": 2443 + }, + { + "epoch": 0.53, + "learning_rate": 4.738474983613721e-06, + "loss": 3.7945, + "step": 2444 + }, + { + "epoch": 0.53, + "learning_rate": 4.73836574175224e-06, + "loss": 3.2615, + "step": 2445 + }, + { + "epoch": 0.53, + "learning_rate": 4.738256499890758e-06, + "loss": 4.0561, + "step": 2446 + }, + { + "epoch": 0.53, + "learning_rate": 4.738147258029277e-06, + "loss": 3.3656, + "step": 2447 + }, + { + "epoch": 0.53, + "learning_rate": 4.7380380161677956e-06, + "loss": 3.1202, + "step": 2448 + }, + { + "epoch": 0.53, + "learning_rate": 4.737928774306315e-06, + "loss": 3.6643, + "step": 2449 + }, + { + "epoch": 0.53, + "learning_rate": 4.737819532444834e-06, + "loss": 3.3711, + "step": 2450 + }, + { + "epoch": 0.53, + "learning_rate": 4.737710290583352e-06, + "loss": 3.5292, + "step": 2451 + }, + { + "epoch": 0.54, + "learning_rate": 4.73760104872187e-06, + "loss": 3.9019, + "step": 2452 + }, + { + "epoch": 0.54, + "learning_rate": 4.737491806860389e-06, + "loss": 3.5503, + "step": 2453 + }, + { + "epoch": 0.54, + "learning_rate": 4.737382564998908e-06, + "loss": 3.4823, + "step": 2454 + }, + { + "epoch": 0.54, + "learning_rate": 4.737273323137427e-06, + "loss": 3.5765, + "step": 2455 + }, + { + "epoch": 0.54, + "learning_rate": 4.737164081275945e-06, + "loss": 3.4904, + "step": 2456 + }, + { + "epoch": 0.54, + "learning_rate": 4.7370548394144635e-06, + "loss": 3.5721, + "step": 2457 + }, + { + "epoch": 0.54, + "learning_rate": 4.7369455975529825e-06, + "loss": 3.447, + "step": 2458 + }, + { + "epoch": 0.54, + "learning_rate": 4.7368363556915016e-06, + "loss": 3.3429, + "step": 2459 + }, + { + "epoch": 0.54, + "learning_rate": 4.736727113830021e-06, + "loss": 3.8154, + "step": 2460 + }, + { + "epoch": 0.54, + "learning_rate": 4.736617871968539e-06, + "loss": 3.6505, + "step": 2461 + }, + { + "epoch": 0.54, + "learning_rate": 4.736508630107057e-06, + "loss": 3.8809, + "step": 2462 + }, + { + "epoch": 0.54, + "learning_rate": 4.736399388245576e-06, + "loss": 3.7999, + "step": 2463 + }, + { + "epoch": 0.54, + "learning_rate": 4.736290146384095e-06, + "loss": 3.5945, + "step": 2464 + }, + { + "epoch": 0.54, + "learning_rate": 4.736180904522614e-06, + "loss": 3.204, + "step": 2465 + }, + { + "epoch": 0.54, + "learning_rate": 4.736071662661132e-06, + "loss": 3.6632, + "step": 2466 + }, + { + "epoch": 0.54, + "learning_rate": 4.7359624207996505e-06, + "loss": 3.5476, + "step": 2467 + }, + { + "epoch": 0.54, + "learning_rate": 4.7358531789381695e-06, + "loss": 3.6839, + "step": 2468 + }, + { + "epoch": 0.54, + "learning_rate": 4.7357439370766885e-06, + "loss": 3.88, + "step": 2469 + }, + { + "epoch": 0.54, + "learning_rate": 4.735634695215207e-06, + "loss": 3.6137, + "step": 2470 + }, + { + "epoch": 0.54, + "learning_rate": 4.735525453353726e-06, + "loss": 3.6695, + "step": 2471 + }, + { + "epoch": 0.54, + "learning_rate": 4.735416211492244e-06, + "loss": 3.6342, + "step": 2472 + }, + { + "epoch": 0.54, + "learning_rate": 4.735306969630763e-06, + "loss": 3.3854, + "step": 2473 + }, + { + "epoch": 0.54, + "learning_rate": 4.735197727769282e-06, + "loss": 3.7556, + "step": 2474 + }, + { + "epoch": 0.54, + "learning_rate": 4.7350884859078e-06, + "loss": 3.6533, + "step": 2475 + }, + { + "epoch": 0.54, + "learning_rate": 4.734979244046319e-06, + "loss": 3.6269, + "step": 2476 + }, + { + "epoch": 0.54, + "learning_rate": 4.734870002184837e-06, + "loss": 3.4703, + "step": 2477 + }, + { + "epoch": 0.54, + "learning_rate": 4.7347607603233564e-06, + "loss": 3.7792, + "step": 2478 + }, + { + "epoch": 0.54, + "learning_rate": 4.7346515184618755e-06, + "loss": 3.69, + "step": 2479 + }, + { + "epoch": 0.54, + "learning_rate": 4.734542276600394e-06, + "loss": 3.6064, + "step": 2480 + }, + { + "epoch": 0.54, + "learning_rate": 4.734433034738912e-06, + "loss": 3.7919, + "step": 2481 + }, + { + "epoch": 0.54, + "learning_rate": 4.734323792877431e-06, + "loss": 3.5569, + "step": 2482 + }, + { + "epoch": 0.54, + "learning_rate": 4.73421455101595e-06, + "loss": 3.4289, + "step": 2483 + }, + { + "epoch": 0.54, + "learning_rate": 4.734105309154469e-06, + "loss": 3.8429, + "step": 2484 + }, + { + "epoch": 0.54, + "learning_rate": 4.733996067292987e-06, + "loss": 3.5485, + "step": 2485 + }, + { + "epoch": 0.54, + "learning_rate": 4.733886825431505e-06, + "loss": 3.8862, + "step": 2486 + }, + { + "epoch": 0.54, + "learning_rate": 4.733777583570024e-06, + "loss": 3.6429, + "step": 2487 + }, + { + "epoch": 0.54, + "learning_rate": 4.733668341708543e-06, + "loss": 3.864, + "step": 2488 + }, + { + "epoch": 0.54, + "learning_rate": 4.733559099847062e-06, + "loss": 3.0215, + "step": 2489 + }, + { + "epoch": 0.54, + "learning_rate": 4.733449857985581e-06, + "loss": 3.3055, + "step": 2490 + }, + { + "epoch": 0.54, + "learning_rate": 4.733340616124099e-06, + "loss": 3.9711, + "step": 2491 + }, + { + "epoch": 0.54, + "learning_rate": 4.733231374262618e-06, + "loss": 3.8482, + "step": 2492 + }, + { + "epoch": 0.54, + "learning_rate": 4.733122132401137e-06, + "loss": 4.0019, + "step": 2493 + }, + { + "epoch": 0.54, + "learning_rate": 4.733012890539655e-06, + "loss": 3.4642, + "step": 2494 + }, + { + "epoch": 0.54, + "learning_rate": 4.732903648678174e-06, + "loss": 3.9749, + "step": 2495 + }, + { + "epoch": 0.54, + "learning_rate": 4.732794406816692e-06, + "loss": 3.6603, + "step": 2496 + }, + { + "epoch": 0.54, + "learning_rate": 4.732685164955211e-06, + "loss": 3.6325, + "step": 2497 + }, + { + "epoch": 0.55, + "learning_rate": 4.73257592309373e-06, + "loss": 3.7895, + "step": 2498 + }, + { + "epoch": 0.55, + "learning_rate": 4.7324666812322485e-06, + "loss": 3.7181, + "step": 2499 + }, + { + "epoch": 0.55, + "learning_rate": 4.732357439370768e-06, + "loss": 3.4952, + "step": 2500 + }, + { + "epoch": 0.55, + "learning_rate": 4.732248197509286e-06, + "loss": 3.2619, + "step": 2501 + }, + { + "epoch": 0.55, + "learning_rate": 4.732138955647805e-06, + "loss": 3.7911, + "step": 2502 + }, + { + "epoch": 0.55, + "learning_rate": 4.732029713786323e-06, + "loss": 3.6642, + "step": 2503 + }, + { + "epoch": 0.55, + "learning_rate": 4.731920471924842e-06, + "loss": 3.3307, + "step": 2504 + }, + { + "epoch": 0.55, + "learning_rate": 4.73181123006336e-06, + "loss": 3.9229, + "step": 2505 + }, + { + "epoch": 0.55, + "learning_rate": 4.731701988201879e-06, + "loss": 3.8652, + "step": 2506 + }, + { + "epoch": 0.55, + "learning_rate": 4.731592746340398e-06, + "loss": 3.9714, + "step": 2507 + }, + { + "epoch": 0.55, + "learning_rate": 4.7314835044789165e-06, + "loss": 3.3254, + "step": 2508 + }, + { + "epoch": 0.55, + "learning_rate": 4.7313742626174355e-06, + "loss": 3.7651, + "step": 2509 + }, + { + "epoch": 0.55, + "learning_rate": 4.731265020755954e-06, + "loss": 3.7255, + "step": 2510 + }, + { + "epoch": 0.55, + "learning_rate": 4.731155778894473e-06, + "loss": 4.0627, + "step": 2511 + }, + { + "epoch": 0.55, + "learning_rate": 4.731046537032992e-06, + "loss": 3.526, + "step": 2512 + }, + { + "epoch": 0.55, + "learning_rate": 4.73093729517151e-06, + "loss": 3.6729, + "step": 2513 + }, + { + "epoch": 0.55, + "learning_rate": 4.730828053310029e-06, + "loss": 3.7156, + "step": 2514 + }, + { + "epoch": 0.55, + "learning_rate": 4.730718811448547e-06, + "loss": 3.5497, + "step": 2515 + }, + { + "epoch": 0.55, + "learning_rate": 4.730609569587066e-06, + "loss": 3.9251, + "step": 2516 + }, + { + "epoch": 0.55, + "learning_rate": 4.730500327725585e-06, + "loss": 3.3895, + "step": 2517 + }, + { + "epoch": 0.55, + "learning_rate": 4.7303910858641034e-06, + "loss": 3.2201, + "step": 2518 + }, + { + "epoch": 0.55, + "learning_rate": 4.7302818440026225e-06, + "loss": 3.8746, + "step": 2519 + }, + { + "epoch": 0.55, + "learning_rate": 4.730172602141141e-06, + "loss": 3.9095, + "step": 2520 + }, + { + "epoch": 0.55, + "learning_rate": 4.73006336027966e-06, + "loss": 3.7625, + "step": 2521 + }, + { + "epoch": 0.55, + "learning_rate": 4.729954118418178e-06, + "loss": 3.7268, + "step": 2522 + }, + { + "epoch": 0.55, + "learning_rate": 4.729844876556697e-06, + "loss": 3.7182, + "step": 2523 + }, + { + "epoch": 0.55, + "learning_rate": 4.729735634695216e-06, + "loss": 3.908, + "step": 2524 + }, + { + "epoch": 0.55, + "learning_rate": 4.729626392833734e-06, + "loss": 3.9019, + "step": 2525 + }, + { + "epoch": 0.55, + "learning_rate": 4.729517150972253e-06, + "loss": 3.4192, + "step": 2526 + }, + { + "epoch": 0.55, + "learning_rate": 4.729407909110771e-06, + "loss": 3.5467, + "step": 2527 + }, + { + "epoch": 0.55, + "learning_rate": 4.72929866724929e-06, + "loss": 3.5686, + "step": 2528 + }, + { + "epoch": 0.55, + "learning_rate": 4.729189425387809e-06, + "loss": 3.9284, + "step": 2529 + }, + { + "epoch": 0.55, + "learning_rate": 4.729080183526328e-06, + "loss": 3.7367, + "step": 2530 + }, + { + "epoch": 0.55, + "learning_rate": 4.728970941664847e-06, + "loss": 3.4336, + "step": 2531 + }, + { + "epoch": 0.55, + "learning_rate": 4.728861699803365e-06, + "loss": 3.5938, + "step": 2532 + }, + { + "epoch": 0.55, + "learning_rate": 4.728752457941884e-06, + "loss": 3.1087, + "step": 2533 + }, + { + "epoch": 0.55, + "learning_rate": 4.728643216080402e-06, + "loss": 3.7833, + "step": 2534 + }, + { + "epoch": 0.55, + "learning_rate": 4.728533974218921e-06, + "loss": 3.7961, + "step": 2535 + }, + { + "epoch": 0.55, + "learning_rate": 4.728424732357439e-06, + "loss": 3.7366, + "step": 2536 + }, + { + "epoch": 0.55, + "learning_rate": 4.728315490495958e-06, + "loss": 3.7441, + "step": 2537 + }, + { + "epoch": 0.55, + "learning_rate": 4.728206248634477e-06, + "loss": 3.7123, + "step": 2538 + }, + { + "epoch": 0.55, + "learning_rate": 4.7280970067729955e-06, + "loss": 3.5271, + "step": 2539 + }, + { + "epoch": 0.55, + "learning_rate": 4.7279877649115146e-06, + "loss": 3.7361, + "step": 2540 + }, + { + "epoch": 0.55, + "learning_rate": 4.727878523050033e-06, + "loss": 3.8163, + "step": 2541 + }, + { + "epoch": 0.55, + "learning_rate": 4.727769281188552e-06, + "loss": 4.0165, + "step": 2542 + }, + { + "epoch": 0.55, + "learning_rate": 4.727660039327071e-06, + "loss": 3.5689, + "step": 2543 + }, + { + "epoch": 0.56, + "learning_rate": 4.727550797465589e-06, + "loss": 3.4494, + "step": 2544 + }, + { + "epoch": 0.56, + "learning_rate": 4.727441555604108e-06, + "loss": 3.164, + "step": 2545 + }, + { + "epoch": 0.56, + "learning_rate": 4.727332313742626e-06, + "loss": 3.5561, + "step": 2546 + }, + { + "epoch": 0.56, + "learning_rate": 4.727223071881145e-06, + "loss": 3.5324, + "step": 2547 + }, + { + "epoch": 0.56, + "learning_rate": 4.727113830019664e-06, + "loss": 3.7486, + "step": 2548 + }, + { + "epoch": 0.56, + "learning_rate": 4.7270045881581825e-06, + "loss": 3.8484, + "step": 2549 + }, + { + "epoch": 0.56, + "learning_rate": 4.7268953462967015e-06, + "loss": 3.3517, + "step": 2550 + }, + { + "epoch": 0.56, + "learning_rate": 4.72678610443522e-06, + "loss": 3.6244, + "step": 2551 + }, + { + "epoch": 0.56, + "learning_rate": 4.726676862573739e-06, + "loss": 3.7266, + "step": 2552 + }, + { + "epoch": 0.56, + "learning_rate": 4.726567620712258e-06, + "loss": 3.5602, + "step": 2553 + }, + { + "epoch": 0.56, + "learning_rate": 4.726458378850776e-06, + "loss": 3.8789, + "step": 2554 + }, + { + "epoch": 0.56, + "learning_rate": 4.726349136989294e-06, + "loss": 3.635, + "step": 2555 + }, + { + "epoch": 0.56, + "learning_rate": 4.726239895127813e-06, + "loss": 3.7517, + "step": 2556 + }, + { + "epoch": 0.56, + "learning_rate": 4.726130653266332e-06, + "loss": 3.494, + "step": 2557 + }, + { + "epoch": 0.56, + "learning_rate": 4.72602141140485e-06, + "loss": 3.8895, + "step": 2558 + }, + { + "epoch": 0.56, + "learning_rate": 4.7259121695433694e-06, + "loss": 3.4712, + "step": 2559 + }, + { + "epoch": 0.56, + "learning_rate": 4.725802927681888e-06, + "loss": 3.4781, + "step": 2560 + }, + { + "epoch": 0.56, + "learning_rate": 4.725693685820407e-06, + "loss": 3.4198, + "step": 2561 + }, + { + "epoch": 0.56, + "learning_rate": 4.725584443958926e-06, + "loss": 3.5626, + "step": 2562 + }, + { + "epoch": 0.56, + "learning_rate": 4.725475202097444e-06, + "loss": 3.7831, + "step": 2563 + }, + { + "epoch": 0.56, + "learning_rate": 4.725365960235963e-06, + "loss": 3.6373, + "step": 2564 + }, + { + "epoch": 0.56, + "learning_rate": 4.725256718374481e-06, + "loss": 3.273, + "step": 2565 + }, + { + "epoch": 0.56, + "learning_rate": 4.725147476513e-06, + "loss": 3.546, + "step": 2566 + }, + { + "epoch": 0.56, + "learning_rate": 4.725038234651519e-06, + "loss": 3.8847, + "step": 2567 + }, + { + "epoch": 0.56, + "learning_rate": 4.724928992790037e-06, + "loss": 3.2938, + "step": 2568 + }, + { + "epoch": 0.56, + "learning_rate": 4.7248197509285556e-06, + "loss": 3.5018, + "step": 2569 + }, + { + "epoch": 0.56, + "learning_rate": 4.724710509067075e-06, + "loss": 3.3626, + "step": 2570 + }, + { + "epoch": 0.56, + "learning_rate": 4.724601267205594e-06, + "loss": 3.664, + "step": 2571 + }, + { + "epoch": 0.56, + "learning_rate": 4.724492025344113e-06, + "loss": 3.5489, + "step": 2572 + }, + { + "epoch": 0.56, + "learning_rate": 4.724382783482631e-06, + "loss": 3.9632, + "step": 2573 + }, + { + "epoch": 0.56, + "learning_rate": 4.724273541621149e-06, + "loss": 3.451, + "step": 2574 + }, + { + "epoch": 0.56, + "learning_rate": 4.724164299759668e-06, + "loss": 3.7818, + "step": 2575 + }, + { + "epoch": 0.56, + "learning_rate": 4.724055057898187e-06, + "loss": 3.716, + "step": 2576 + }, + { + "epoch": 0.56, + "learning_rate": 4.723945816036706e-06, + "loss": 3.8121, + "step": 2577 + }, + { + "epoch": 0.56, + "learning_rate": 4.723836574175224e-06, + "loss": 3.9556, + "step": 2578 + }, + { + "epoch": 0.56, + "learning_rate": 4.7237273323137425e-06, + "loss": 3.1469, + "step": 2579 + }, + { + "epoch": 0.56, + "learning_rate": 4.7236180904522615e-06, + "loss": 4.0124, + "step": 2580 + }, + { + "epoch": 0.56, + "learning_rate": 4.723508848590781e-06, + "loss": 4.0677, + "step": 2581 + }, + { + "epoch": 0.56, + "learning_rate": 4.7233996067293e-06, + "loss": 3.9745, + "step": 2582 + }, + { + "epoch": 0.56, + "learning_rate": 4.723290364867818e-06, + "loss": 3.2906, + "step": 2583 + }, + { + "epoch": 0.56, + "learning_rate": 4.723181123006336e-06, + "loss": 3.4775, + "step": 2584 + }, + { + "epoch": 0.56, + "learning_rate": 4.723071881144855e-06, + "loss": 3.8594, + "step": 2585 + }, + { + "epoch": 0.56, + "learning_rate": 4.722962639283374e-06, + "loss": 3.3883, + "step": 2586 + }, + { + "epoch": 0.56, + "learning_rate": 4.722853397421892e-06, + "loss": 3.6304, + "step": 2587 + }, + { + "epoch": 0.56, + "learning_rate": 4.722744155560411e-06, + "loss": 3.6639, + "step": 2588 + }, + { + "epoch": 0.57, + "learning_rate": 4.7226349136989295e-06, + "loss": 3.4942, + "step": 2589 + }, + { + "epoch": 0.57, + "learning_rate": 4.7225256718374485e-06, + "loss": 3.4413, + "step": 2590 + }, + { + "epoch": 0.57, + "learning_rate": 4.7224164299759675e-06, + "loss": 3.1037, + "step": 2591 + }, + { + "epoch": 0.57, + "learning_rate": 4.722307188114486e-06, + "loss": 3.7607, + "step": 2592 + }, + { + "epoch": 0.57, + "learning_rate": 4.722197946253004e-06, + "loss": 3.7322, + "step": 2593 + }, + { + "epoch": 0.57, + "learning_rate": 4.722088704391523e-06, + "loss": 3.6015, + "step": 2594 + }, + { + "epoch": 0.57, + "learning_rate": 4.721979462530042e-06, + "loss": 3.6771, + "step": 2595 + }, + { + "epoch": 0.57, + "learning_rate": 4.721870220668561e-06, + "loss": 3.5564, + "step": 2596 + }, + { + "epoch": 0.57, + "learning_rate": 4.721760978807079e-06, + "loss": 3.8489, + "step": 2597 + }, + { + "epoch": 0.57, + "learning_rate": 4.721651736945597e-06, + "loss": 3.8258, + "step": 2598 + }, + { + "epoch": 0.57, + "learning_rate": 4.721542495084116e-06, + "loss": 3.6004, + "step": 2599 + }, + { + "epoch": 0.57, + "learning_rate": 4.7214332532226355e-06, + "loss": 3.0851, + "step": 2600 + }, + { + "epoch": 0.57, + "learning_rate": 4.7213240113611545e-06, + "loss": 3.5884, + "step": 2601 + }, + { + "epoch": 0.57, + "learning_rate": 4.721214769499673e-06, + "loss": 3.6875, + "step": 2602 + }, + { + "epoch": 0.57, + "learning_rate": 4.721105527638191e-06, + "loss": 3.897, + "step": 2603 + }, + { + "epoch": 0.57, + "learning_rate": 4.72099628577671e-06, + "loss": 3.767, + "step": 2604 + }, + { + "epoch": 0.57, + "learning_rate": 4.720887043915229e-06, + "loss": 3.9241, + "step": 2605 + }, + { + "epoch": 0.57, + "learning_rate": 4.720777802053748e-06, + "loss": 3.5524, + "step": 2606 + }, + { + "epoch": 0.57, + "learning_rate": 4.720668560192266e-06, + "loss": 3.5152, + "step": 2607 + }, + { + "epoch": 0.57, + "learning_rate": 4.720559318330784e-06, + "loss": 3.7059, + "step": 2608 + }, + { + "epoch": 0.57, + "learning_rate": 4.720450076469303e-06, + "loss": 3.6465, + "step": 2609 + }, + { + "epoch": 0.57, + "learning_rate": 4.720340834607822e-06, + "loss": 3.6712, + "step": 2610 + }, + { + "epoch": 0.57, + "learning_rate": 4.720231592746341e-06, + "loss": 3.8181, + "step": 2611 + }, + { + "epoch": 0.57, + "learning_rate": 4.72012235088486e-06, + "loss": 3.6562, + "step": 2612 + }, + { + "epoch": 0.57, + "learning_rate": 4.720013109023378e-06, + "loss": 3.3646, + "step": 2613 + }, + { + "epoch": 0.57, + "learning_rate": 4.719903867161897e-06, + "loss": 3.7378, + "step": 2614 + }, + { + "epoch": 0.57, + "learning_rate": 4.719794625300416e-06, + "loss": 4.0281, + "step": 2615 + }, + { + "epoch": 0.57, + "learning_rate": 4.719685383438934e-06, + "loss": 3.7318, + "step": 2616 + }, + { + "epoch": 0.57, + "learning_rate": 4.719576141577453e-06, + "loss": 3.7563, + "step": 2617 + }, + { + "epoch": 0.57, + "learning_rate": 4.719466899715971e-06, + "loss": 3.2669, + "step": 2618 + }, + { + "epoch": 0.57, + "learning_rate": 4.71935765785449e-06, + "loss": 3.726, + "step": 2619 + }, + { + "epoch": 0.57, + "learning_rate": 4.719248415993009e-06, + "loss": 3.8241, + "step": 2620 + }, + { + "epoch": 0.57, + "learning_rate": 4.7191391741315276e-06, + "loss": 3.3452, + "step": 2621 + }, + { + "epoch": 0.57, + "learning_rate": 4.719029932270046e-06, + "loss": 3.6491, + "step": 2622 + }, + { + "epoch": 0.57, + "learning_rate": 4.718920690408565e-06, + "loss": 3.7502, + "step": 2623 + }, + { + "epoch": 0.57, + "learning_rate": 4.718811448547084e-06, + "loss": 3.7769, + "step": 2624 + }, + { + "epoch": 0.57, + "learning_rate": 4.718702206685603e-06, + "loss": 3.7745, + "step": 2625 + }, + { + "epoch": 0.57, + "learning_rate": 4.718592964824121e-06, + "loss": 3.4107, + "step": 2626 + }, + { + "epoch": 0.57, + "learning_rate": 4.718483722962639e-06, + "loss": 3.7312, + "step": 2627 + }, + { + "epoch": 0.57, + "learning_rate": 4.718374481101158e-06, + "loss": 3.3658, + "step": 2628 + }, + { + "epoch": 0.57, + "learning_rate": 4.718265239239677e-06, + "loss": 3.599, + "step": 2629 + }, + { + "epoch": 0.57, + "learning_rate": 4.718155997378196e-06, + "loss": 3.2678, + "step": 2630 + }, + { + "epoch": 0.57, + "learning_rate": 4.7180467555167145e-06, + "loss": 3.7717, + "step": 2631 + }, + { + "epoch": 0.57, + "learning_rate": 4.717937513655233e-06, + "loss": 3.4492, + "step": 2632 + }, + { + "epoch": 0.57, + "learning_rate": 4.717828271793752e-06, + "loss": 3.7741, + "step": 2633 + }, + { + "epoch": 0.57, + "learning_rate": 4.717719029932271e-06, + "loss": 3.9147, + "step": 2634 + }, + { + "epoch": 0.58, + "learning_rate": 4.717609788070789e-06, + "loss": 3.4758, + "step": 2635 + }, + { + "epoch": 0.58, + "learning_rate": 4.717500546209308e-06, + "loss": 3.5439, + "step": 2636 + }, + { + "epoch": 0.58, + "learning_rate": 4.717391304347826e-06, + "loss": 3.4166, + "step": 2637 + }, + { + "epoch": 0.58, + "learning_rate": 4.717282062486345e-06, + "loss": 3.7501, + "step": 2638 + }, + { + "epoch": 0.58, + "learning_rate": 4.717172820624864e-06, + "loss": 3.9185, + "step": 2639 + }, + { + "epoch": 0.58, + "learning_rate": 4.7170635787633824e-06, + "loss": 3.7462, + "step": 2640 + }, + { + "epoch": 0.58, + "learning_rate": 4.7169543369019015e-06, + "loss": 3.7121, + "step": 2641 + }, + { + "epoch": 0.58, + "learning_rate": 4.71684509504042e-06, + "loss": 3.1904, + "step": 2642 + }, + { + "epoch": 0.58, + "learning_rate": 4.716735853178939e-06, + "loss": 3.5963, + "step": 2643 + }, + { + "epoch": 0.58, + "learning_rate": 4.716626611317458e-06, + "loss": 3.5638, + "step": 2644 + }, + { + "epoch": 0.58, + "learning_rate": 4.716517369455976e-06, + "loss": 3.6741, + "step": 2645 + }, + { + "epoch": 0.58, + "learning_rate": 4.716408127594494e-06, + "loss": 3.8593, + "step": 2646 + }, + { + "epoch": 0.58, + "learning_rate": 4.716298885733013e-06, + "loss": 3.4359, + "step": 2647 + }, + { + "epoch": 0.58, + "learning_rate": 4.716189643871532e-06, + "loss": 3.742, + "step": 2648 + }, + { + "epoch": 0.58, + "learning_rate": 4.716080402010051e-06, + "loss": 3.7045, + "step": 2649 + }, + { + "epoch": 0.58, + "learning_rate": 4.715971160148569e-06, + "loss": 3.76, + "step": 2650 + }, + { + "epoch": 0.58, + "learning_rate": 4.715861918287088e-06, + "loss": 3.5893, + "step": 2651 + }, + { + "epoch": 0.58, + "learning_rate": 4.715752676425607e-06, + "loss": 3.5939, + "step": 2652 + }, + { + "epoch": 0.58, + "learning_rate": 4.715643434564126e-06, + "loss": 3.472, + "step": 2653 + }, + { + "epoch": 0.58, + "learning_rate": 4.715534192702644e-06, + "loss": 3.4023, + "step": 2654 + }, + { + "epoch": 0.58, + "learning_rate": 4.715424950841163e-06, + "loss": 3.6684, + "step": 2655 + }, + { + "epoch": 0.58, + "learning_rate": 4.715315708979681e-06, + "loss": 3.4939, + "step": 2656 + }, + { + "epoch": 0.58, + "learning_rate": 4.7152064671182e-06, + "loss": 3.8079, + "step": 2657 + }, + { + "epoch": 0.58, + "learning_rate": 4.715097225256719e-06, + "loss": 3.6674, + "step": 2658 + }, + { + "epoch": 0.58, + "learning_rate": 4.714987983395237e-06, + "loss": 3.3565, + "step": 2659 + }, + { + "epoch": 0.58, + "learning_rate": 4.714878741533756e-06, + "loss": 3.5207, + "step": 2660 + }, + { + "epoch": 0.58, + "learning_rate": 4.7147694996722745e-06, + "loss": 3.2976, + "step": 2661 + }, + { + "epoch": 0.58, + "learning_rate": 4.7146602578107936e-06, + "loss": 3.1849, + "step": 2662 + }, + { + "epoch": 0.58, + "learning_rate": 4.714551015949313e-06, + "loss": 3.594, + "step": 2663 + }, + { + "epoch": 0.58, + "learning_rate": 4.714441774087831e-06, + "loss": 3.463, + "step": 2664 + }, + { + "epoch": 0.58, + "learning_rate": 4.71433253222635e-06, + "loss": 3.8944, + "step": 2665 + }, + { + "epoch": 0.58, + "learning_rate": 4.714223290364868e-06, + "loss": 3.7034, + "step": 2666 + }, + { + "epoch": 0.58, + "learning_rate": 4.714114048503387e-06, + "loss": 3.6172, + "step": 2667 + }, + { + "epoch": 0.58, + "learning_rate": 4.714004806641905e-06, + "loss": 3.5542, + "step": 2668 + }, + { + "epoch": 0.58, + "learning_rate": 4.713895564780424e-06, + "loss": 3.2546, + "step": 2669 + }, + { + "epoch": 0.58, + "learning_rate": 4.713786322918943e-06, + "loss": 3.5008, + "step": 2670 + }, + { + "epoch": 0.58, + "learning_rate": 4.7136770810574615e-06, + "loss": 3.8077, + "step": 2671 + }, + { + "epoch": 0.58, + "learning_rate": 4.7135678391959805e-06, + "loss": 3.2333, + "step": 2672 + }, + { + "epoch": 0.58, + "learning_rate": 4.713458597334499e-06, + "loss": 3.5617, + "step": 2673 + }, + { + "epoch": 0.58, + "learning_rate": 4.713349355473018e-06, + "loss": 3.6272, + "step": 2674 + }, + { + "epoch": 0.58, + "learning_rate": 4.713240113611536e-06, + "loss": 3.2082, + "step": 2675 + }, + { + "epoch": 0.58, + "learning_rate": 4.713130871750055e-06, + "loss": 3.6401, + "step": 2676 + }, + { + "epoch": 0.58, + "learning_rate": 4.713021629888574e-06, + "loss": 3.5734, + "step": 2677 + }, + { + "epoch": 0.58, + "learning_rate": 4.712912388027092e-06, + "loss": 3.4868, + "step": 2678 + }, + { + "epoch": 0.58, + "learning_rate": 4.712803146165611e-06, + "loss": 3.9728, + "step": 2679 + }, + { + "epoch": 0.58, + "learning_rate": 4.712693904304129e-06, + "loss": 3.6276, + "step": 2680 + }, + { + "epoch": 0.59, + "learning_rate": 4.7125846624426485e-06, + "loss": 3.3697, + "step": 2681 + }, + { + "epoch": 0.59, + "learning_rate": 4.7124754205811675e-06, + "loss": 3.8124, + "step": 2682 + }, + { + "epoch": 0.59, + "learning_rate": 4.712366178719686e-06, + "loss": 3.0437, + "step": 2683 + }, + { + "epoch": 0.59, + "learning_rate": 4.712256936858205e-06, + "loss": 3.2529, + "step": 2684 + }, + { + "epoch": 0.59, + "learning_rate": 4.712147694996723e-06, + "loss": 3.458, + "step": 2685 + }, + { + "epoch": 0.59, + "learning_rate": 4.712038453135242e-06, + "loss": 3.6296, + "step": 2686 + }, + { + "epoch": 0.59, + "learning_rate": 4.71192921127376e-06, + "loss": 3.8187, + "step": 2687 + }, + { + "epoch": 0.59, + "learning_rate": 4.711819969412279e-06, + "loss": 3.4462, + "step": 2688 + }, + { + "epoch": 0.59, + "learning_rate": 4.711710727550798e-06, + "loss": 3.6815, + "step": 2689 + }, + { + "epoch": 0.59, + "learning_rate": 4.711601485689316e-06, + "loss": 3.4219, + "step": 2690 + }, + { + "epoch": 0.59, + "learning_rate": 4.711492243827835e-06, + "loss": 3.3003, + "step": 2691 + }, + { + "epoch": 0.59, + "learning_rate": 4.711383001966354e-06, + "loss": 3.321, + "step": 2692 + }, + { + "epoch": 0.59, + "learning_rate": 4.711273760104873e-06, + "loss": 3.7685, + "step": 2693 + }, + { + "epoch": 0.59, + "learning_rate": 4.711164518243392e-06, + "loss": 3.6315, + "step": 2694 + }, + { + "epoch": 0.59, + "learning_rate": 4.71105527638191e-06, + "loss": 3.7039, + "step": 2695 + }, + { + "epoch": 0.59, + "learning_rate": 4.710946034520429e-06, + "loss": 3.8506, + "step": 2696 + }, + { + "epoch": 0.59, + "learning_rate": 4.710836792658947e-06, + "loss": 3.5176, + "step": 2697 + }, + { + "epoch": 0.59, + "learning_rate": 4.710727550797466e-06, + "loss": 3.6258, + "step": 2698 + }, + { + "epoch": 0.59, + "learning_rate": 4.710618308935984e-06, + "loss": 3.5653, + "step": 2699 + }, + { + "epoch": 0.59, + "learning_rate": 4.710509067074503e-06, + "loss": 3.6227, + "step": 2700 + }, + { + "epoch": 0.59, + "learning_rate": 4.7103998252130215e-06, + "loss": 3.8928, + "step": 2701 + }, + { + "epoch": 0.59, + "learning_rate": 4.7102905833515406e-06, + "loss": 3.4539, + "step": 2702 + }, + { + "epoch": 0.59, + "learning_rate": 4.71018134149006e-06, + "loss": 3.9098, + "step": 2703 + }, + { + "epoch": 0.59, + "learning_rate": 4.710072099628578e-06, + "loss": 3.3781, + "step": 2704 + }, + { + "epoch": 0.59, + "learning_rate": 4.709962857767097e-06, + "loss": 3.2776, + "step": 2705 + }, + { + "epoch": 0.59, + "learning_rate": 4.709853615905615e-06, + "loss": 3.5443, + "step": 2706 + }, + { + "epoch": 0.59, + "learning_rate": 4.709744374044134e-06, + "loss": 3.7643, + "step": 2707 + }, + { + "epoch": 0.59, + "learning_rate": 4.709635132182653e-06, + "loss": 4.17, + "step": 2708 + }, + { + "epoch": 0.59, + "learning_rate": 4.709525890321171e-06, + "loss": 3.7066, + "step": 2709 + }, + { + "epoch": 0.59, + "learning_rate": 4.70941664845969e-06, + "loss": 3.6436, + "step": 2710 + }, + { + "epoch": 0.59, + "learning_rate": 4.7093074065982085e-06, + "loss": 3.5839, + "step": 2711 + }, + { + "epoch": 0.59, + "learning_rate": 4.7091981647367275e-06, + "loss": 3.7198, + "step": 2712 + }, + { + "epoch": 0.59, + "learning_rate": 4.7090889228752465e-06, + "loss": 3.6343, + "step": 2713 + }, + { + "epoch": 0.59, + "learning_rate": 4.708979681013765e-06, + "loss": 3.1862, + "step": 2714 + }, + { + "epoch": 0.59, + "learning_rate": 4.708870439152284e-06, + "loss": 3.4485, + "step": 2715 + }, + { + "epoch": 0.59, + "learning_rate": 4.708761197290802e-06, + "loss": 3.5025, + "step": 2716 + }, + { + "epoch": 0.59, + "learning_rate": 4.708651955429321e-06, + "loss": 4.0674, + "step": 2717 + }, + { + "epoch": 0.59, + "learning_rate": 4.70854271356784e-06, + "loss": 3.6649, + "step": 2718 + }, + { + "epoch": 0.59, + "learning_rate": 4.708433471706358e-06, + "loss": 3.2432, + "step": 2719 + }, + { + "epoch": 0.59, + "learning_rate": 4.708324229844876e-06, + "loss": 3.3664, + "step": 2720 + }, + { + "epoch": 0.59, + "learning_rate": 4.7082149879833954e-06, + "loss": 3.8274, + "step": 2721 + }, + { + "epoch": 0.59, + "learning_rate": 4.7081057461219145e-06, + "loss": 3.1343, + "step": 2722 + }, + { + "epoch": 0.59, + "learning_rate": 4.7079965042604335e-06, + "loss": 3.6415, + "step": 2723 + }, + { + "epoch": 0.59, + "learning_rate": 4.707887262398952e-06, + "loss": 3.6469, + "step": 2724 + }, + { + "epoch": 0.59, + "learning_rate": 4.70777802053747e-06, + "loss": 3.5082, + "step": 2725 + }, + { + "epoch": 0.59, + "learning_rate": 4.707668778675989e-06, + "loss": 3.093, + "step": 2726 + }, + { + "epoch": 0.6, + "learning_rate": 4.707559536814508e-06, + "loss": 3.7967, + "step": 2727 + }, + { + "epoch": 0.6, + "learning_rate": 4.707450294953026e-06, + "loss": 3.2967, + "step": 2728 + }, + { + "epoch": 0.6, + "learning_rate": 4.707341053091545e-06, + "loss": 3.4644, + "step": 2729 + }, + { + "epoch": 0.6, + "learning_rate": 4.707231811230063e-06, + "loss": 3.8161, + "step": 2730 + }, + { + "epoch": 0.6, + "learning_rate": 4.707122569368582e-06, + "loss": 3.4825, + "step": 2731 + }, + { + "epoch": 0.6, + "learning_rate": 4.7070133275071014e-06, + "loss": 3.3452, + "step": 2732 + }, + { + "epoch": 0.6, + "learning_rate": 4.70690408564562e-06, + "loss": 3.527, + "step": 2733 + }, + { + "epoch": 0.6, + "learning_rate": 4.706794843784138e-06, + "loss": 3.5707, + "step": 2734 + }, + { + "epoch": 0.6, + "learning_rate": 4.706685601922657e-06, + "loss": 3.6158, + "step": 2735 + }, + { + "epoch": 0.6, + "learning_rate": 4.706576360061176e-06, + "loss": 3.8708, + "step": 2736 + }, + { + "epoch": 0.6, + "learning_rate": 4.706467118199695e-06, + "loss": 3.5877, + "step": 2737 + }, + { + "epoch": 0.6, + "learning_rate": 4.706357876338213e-06, + "loss": 3.4563, + "step": 2738 + }, + { + "epoch": 0.6, + "learning_rate": 4.706248634476731e-06, + "loss": 3.4111, + "step": 2739 + }, + { + "epoch": 0.6, + "learning_rate": 4.70613939261525e-06, + "loss": 3.7796, + "step": 2740 + }, + { + "epoch": 0.6, + "learning_rate": 4.706030150753769e-06, + "loss": 3.5586, + "step": 2741 + }, + { + "epoch": 0.6, + "learning_rate": 4.705920908892288e-06, + "loss": 3.7497, + "step": 2742 + }, + { + "epoch": 0.6, + "learning_rate": 4.7058116670308066e-06, + "loss": 3.7224, + "step": 2743 + }, + { + "epoch": 0.6, + "learning_rate": 4.705702425169325e-06, + "loss": 3.5256, + "step": 2744 + }, + { + "epoch": 0.6, + "learning_rate": 4.705593183307844e-06, + "loss": 3.5689, + "step": 2745 + }, + { + "epoch": 0.6, + "learning_rate": 4.705483941446363e-06, + "loss": 3.7133, + "step": 2746 + }, + { + "epoch": 0.6, + "learning_rate": 4.705374699584882e-06, + "loss": 3.3098, + "step": 2747 + }, + { + "epoch": 0.6, + "learning_rate": 4.7052654577234e-06, + "loss": 3.7083, + "step": 2748 + }, + { + "epoch": 0.6, + "learning_rate": 4.705156215861918e-06, + "loss": 3.2504, + "step": 2749 + }, + { + "epoch": 0.6, + "learning_rate": 4.705046974000437e-06, + "loss": 3.5967, + "step": 2750 + }, + { + "epoch": 0.6, + "learning_rate": 4.704937732138956e-06, + "loss": 3.4052, + "step": 2751 + }, + { + "epoch": 0.6, + "learning_rate": 4.7048284902774745e-06, + "loss": 2.8659, + "step": 2752 + }, + { + "epoch": 0.6, + "learning_rate": 4.7047192484159935e-06, + "loss": 3.3589, + "step": 2753 + }, + { + "epoch": 0.6, + "learning_rate": 4.704610006554512e-06, + "loss": 3.6522, + "step": 2754 + }, + { + "epoch": 0.6, + "learning_rate": 4.704500764693031e-06, + "loss": 3.3845, + "step": 2755 + }, + { + "epoch": 0.6, + "learning_rate": 4.70439152283155e-06, + "loss": 3.4323, + "step": 2756 + }, + { + "epoch": 0.6, + "learning_rate": 4.704282280970068e-06, + "loss": 3.2408, + "step": 2757 + }, + { + "epoch": 0.6, + "learning_rate": 4.704173039108586e-06, + "loss": 3.3422, + "step": 2758 + }, + { + "epoch": 0.6, + "learning_rate": 4.704063797247105e-06, + "loss": 3.4706, + "step": 2759 + }, + { + "epoch": 0.6, + "learning_rate": 4.703954555385624e-06, + "loss": 3.8706, + "step": 2760 + }, + { + "epoch": 0.6, + "learning_rate": 4.703845313524143e-06, + "loss": 3.2888, + "step": 2761 + }, + { + "epoch": 0.6, + "learning_rate": 4.7037360716626614e-06, + "loss": 3.2322, + "step": 2762 + }, + { + "epoch": 0.6, + "learning_rate": 4.70362682980118e-06, + "loss": 3.6813, + "step": 2763 + }, + { + "epoch": 0.6, + "learning_rate": 4.703517587939699e-06, + "loss": 3.2933, + "step": 2764 + }, + { + "epoch": 0.6, + "learning_rate": 4.703408346078218e-06, + "loss": 3.5367, + "step": 2765 + }, + { + "epoch": 0.6, + "learning_rate": 4.703299104216737e-06, + "loss": 3.2778, + "step": 2766 + }, + { + "epoch": 0.6, + "learning_rate": 4.703189862355255e-06, + "loss": 3.6816, + "step": 2767 + }, + { + "epoch": 0.6, + "learning_rate": 4.703080620493773e-06, + "loss": 3.6974, + "step": 2768 + }, + { + "epoch": 0.6, + "learning_rate": 4.702971378632292e-06, + "loss": 3.8412, + "step": 2769 + }, + { + "epoch": 0.6, + "learning_rate": 4.702862136770811e-06, + "loss": 3.2152, + "step": 2770 + }, + { + "epoch": 0.6, + "learning_rate": 4.70275289490933e-06, + "loss": 3.8799, + "step": 2771 + }, + { + "epoch": 0.6, + "learning_rate": 4.702643653047848e-06, + "loss": 4.0103, + "step": 2772 + }, + { + "epoch": 0.61, + "learning_rate": 4.702534411186367e-06, + "loss": 3.6672, + "step": 2773 + }, + { + "epoch": 0.61, + "learning_rate": 4.702425169324886e-06, + "loss": 3.2698, + "step": 2774 + }, + { + "epoch": 0.61, + "learning_rate": 4.702315927463405e-06, + "loss": 3.8334, + "step": 2775 + }, + { + "epoch": 0.61, + "learning_rate": 4.702206685601924e-06, + "loss": 3.6356, + "step": 2776 + }, + { + "epoch": 0.61, + "learning_rate": 4.702097443740442e-06, + "loss": 3.8256, + "step": 2777 + }, + { + "epoch": 0.61, + "learning_rate": 4.70198820187896e-06, + "loss": 3.5801, + "step": 2778 + }, + { + "epoch": 0.61, + "learning_rate": 4.701878960017479e-06, + "loss": 3.2765, + "step": 2779 + }, + { + "epoch": 0.61, + "learning_rate": 4.701769718155998e-06, + "loss": 3.4866, + "step": 2780 + }, + { + "epoch": 0.61, + "learning_rate": 4.701660476294516e-06, + "loss": 3.7642, + "step": 2781 + }, + { + "epoch": 0.61, + "learning_rate": 4.701551234433035e-06, + "loss": 3.6463, + "step": 2782 + }, + { + "epoch": 0.61, + "learning_rate": 4.7014419925715536e-06, + "loss": 3.4888, + "step": 2783 + }, + { + "epoch": 0.61, + "learning_rate": 4.701332750710073e-06, + "loss": 4.0724, + "step": 2784 + }, + { + "epoch": 0.61, + "learning_rate": 4.701223508848592e-06, + "loss": 3.4162, + "step": 2785 + }, + { + "epoch": 0.61, + "learning_rate": 4.70111426698711e-06, + "loss": 3.568, + "step": 2786 + }, + { + "epoch": 0.61, + "learning_rate": 4.701005025125628e-06, + "loss": 3.84, + "step": 2787 + }, + { + "epoch": 0.61, + "learning_rate": 4.700895783264147e-06, + "loss": 3.5209, + "step": 2788 + }, + { + "epoch": 0.61, + "learning_rate": 4.700786541402666e-06, + "loss": 3.4421, + "step": 2789 + }, + { + "epoch": 0.61, + "learning_rate": 4.700677299541185e-06, + "loss": 3.7719, + "step": 2790 + }, + { + "epoch": 0.61, + "learning_rate": 4.700568057679703e-06, + "loss": 3.7108, + "step": 2791 + }, + { + "epoch": 0.61, + "learning_rate": 4.7004588158182215e-06, + "loss": 3.3734, + "step": 2792 + }, + { + "epoch": 0.61, + "learning_rate": 4.7003495739567405e-06, + "loss": 3.2938, + "step": 2793 + }, + { + "epoch": 0.61, + "learning_rate": 4.7002403320952595e-06, + "loss": 3.3361, + "step": 2794 + }, + { + "epoch": 0.61, + "learning_rate": 4.7001310902337786e-06, + "loss": 3.5441, + "step": 2795 + }, + { + "epoch": 0.61, + "learning_rate": 4.700021848372297e-06, + "loss": 3.5596, + "step": 2796 + }, + { + "epoch": 0.61, + "learning_rate": 4.699912606510815e-06, + "loss": 3.1213, + "step": 2797 + }, + { + "epoch": 0.61, + "learning_rate": 4.699803364649334e-06, + "loss": 3.2737, + "step": 2798 + }, + { + "epoch": 0.61, + "learning_rate": 4.699694122787853e-06, + "loss": 3.6355, + "step": 2799 + }, + { + "epoch": 0.61, + "learning_rate": 4.699584880926371e-06, + "loss": 3.5168, + "step": 2800 + }, + { + "epoch": 0.61, + "learning_rate": 4.69947563906489e-06, + "loss": 3.6826, + "step": 2801 + }, + { + "epoch": 0.61, + "learning_rate": 4.6993663972034084e-06, + "loss": 3.7871, + "step": 2802 + }, + { + "epoch": 0.61, + "learning_rate": 4.6992571553419275e-06, + "loss": 3.2941, + "step": 2803 + }, + { + "epoch": 0.61, + "learning_rate": 4.6991479134804465e-06, + "loss": 3.7859, + "step": 2804 + }, + { + "epoch": 0.61, + "learning_rate": 4.699038671618965e-06, + "loss": 3.6844, + "step": 2805 + }, + { + "epoch": 0.61, + "learning_rate": 4.698929429757484e-06, + "loss": 3.7564, + "step": 2806 + }, + { + "epoch": 0.61, + "learning_rate": 4.698820187896002e-06, + "loss": 3.7764, + "step": 2807 + }, + { + "epoch": 0.61, + "learning_rate": 4.698710946034521e-06, + "loss": 3.8389, + "step": 2808 + }, + { + "epoch": 0.61, + "learning_rate": 4.69860170417304e-06, + "loss": 3.7269, + "step": 2809 + }, + { + "epoch": 0.61, + "learning_rate": 4.698492462311558e-06, + "loss": 3.4757, + "step": 2810 + }, + { + "epoch": 0.61, + "learning_rate": 4.698383220450077e-06, + "loss": 3.843, + "step": 2811 + }, + { + "epoch": 0.61, + "learning_rate": 4.698273978588595e-06, + "loss": 3.556, + "step": 2812 + }, + { + "epoch": 0.61, + "learning_rate": 4.698164736727114e-06, + "loss": 3.4115, + "step": 2813 + }, + { + "epoch": 0.61, + "learning_rate": 4.6980554948656335e-06, + "loss": 3.499, + "step": 2814 + }, + { + "epoch": 0.61, + "learning_rate": 4.697946253004152e-06, + "loss": 3.7359, + "step": 2815 + }, + { + "epoch": 0.61, + "learning_rate": 4.69783701114267e-06, + "loss": 3.8131, + "step": 2816 + }, + { + "epoch": 0.61, + "learning_rate": 4.697727769281189e-06, + "loss": 3.397, + "step": 2817 + }, + { + "epoch": 0.62, + "learning_rate": 4.697618527419708e-06, + "loss": 3.1895, + "step": 2818 + }, + { + "epoch": 0.62, + "learning_rate": 4.697509285558226e-06, + "loss": 3.877, + "step": 2819 + }, + { + "epoch": 0.62, + "learning_rate": 4.697400043696745e-06, + "loss": 3.3774, + "step": 2820 + }, + { + "epoch": 0.62, + "learning_rate": 4.697290801835263e-06, + "loss": 3.8198, + "step": 2821 + }, + { + "epoch": 0.62, + "learning_rate": 4.697181559973782e-06, + "loss": 3.5473, + "step": 2822 + }, + { + "epoch": 0.62, + "learning_rate": 4.697072318112301e-06, + "loss": 3.294, + "step": 2823 + }, + { + "epoch": 0.62, + "learning_rate": 4.6969630762508196e-06, + "loss": 3.5977, + "step": 2824 + }, + { + "epoch": 0.62, + "learning_rate": 4.696853834389339e-06, + "loss": 2.7819, + "step": 2825 + }, + { + "epoch": 0.62, + "learning_rate": 4.696744592527857e-06, + "loss": 3.0991, + "step": 2826 + }, + { + "epoch": 0.62, + "learning_rate": 4.696635350666376e-06, + "loss": 3.9242, + "step": 2827 + }, + { + "epoch": 0.62, + "learning_rate": 4.696526108804895e-06, + "loss": 3.7895, + "step": 2828 + }, + { + "epoch": 0.62, + "learning_rate": 4.696416866943413e-06, + "loss": 3.3005, + "step": 2829 + }, + { + "epoch": 0.62, + "learning_rate": 4.696307625081932e-06, + "loss": 3.4918, + "step": 2830 + }, + { + "epoch": 0.62, + "learning_rate": 4.69619838322045e-06, + "loss": 3.5871, + "step": 2831 + }, + { + "epoch": 0.62, + "learning_rate": 4.696089141358969e-06, + "loss": 3.7707, + "step": 2832 + }, + { + "epoch": 0.62, + "learning_rate": 4.6959798994974875e-06, + "loss": 3.9422, + "step": 2833 + }, + { + "epoch": 0.62, + "learning_rate": 4.6958706576360065e-06, + "loss": 3.6277, + "step": 2834 + }, + { + "epoch": 0.62, + "learning_rate": 4.6957614157745256e-06, + "loss": 3.5333, + "step": 2835 + }, + { + "epoch": 0.62, + "learning_rate": 4.695652173913044e-06, + "loss": 3.701, + "step": 2836 + }, + { + "epoch": 0.62, + "learning_rate": 4.695542932051563e-06, + "loss": 3.6881, + "step": 2837 + }, + { + "epoch": 0.62, + "learning_rate": 4.695433690190081e-06, + "loss": 2.7314, + "step": 2838 + }, + { + "epoch": 0.62, + "learning_rate": 4.6953244483286e-06, + "loss": 3.5633, + "step": 2839 + }, + { + "epoch": 0.62, + "learning_rate": 4.695215206467118e-06, + "loss": 3.8182, + "step": 2840 + }, + { + "epoch": 0.62, + "learning_rate": 4.695105964605637e-06, + "loss": 2.9897, + "step": 2841 + }, + { + "epoch": 0.62, + "learning_rate": 4.694996722744156e-06, + "loss": 3.6708, + "step": 2842 + }, + { + "epoch": 0.62, + "learning_rate": 4.6948874808826744e-06, + "loss": 3.5429, + "step": 2843 + }, + { + "epoch": 0.62, + "learning_rate": 4.6947782390211935e-06, + "loss": 3.8143, + "step": 2844 + }, + { + "epoch": 0.62, + "learning_rate": 4.694668997159712e-06, + "loss": 3.2581, + "step": 2845 + }, + { + "epoch": 0.62, + "learning_rate": 4.694559755298231e-06, + "loss": 2.8662, + "step": 2846 + }, + { + "epoch": 0.62, + "learning_rate": 4.69445051343675e-06, + "loss": 3.7088, + "step": 2847 + }, + { + "epoch": 0.62, + "learning_rate": 4.694341271575268e-06, + "loss": 3.5235, + "step": 2848 + }, + { + "epoch": 0.62, + "learning_rate": 4.694232029713787e-06, + "loss": 3.6472, + "step": 2849 + }, + { + "epoch": 0.62, + "learning_rate": 4.694122787852305e-06, + "loss": 3.6803, + "step": 2850 + }, + { + "epoch": 0.62, + "learning_rate": 4.694013545990824e-06, + "loss": 3.437, + "step": 2851 + }, + { + "epoch": 0.62, + "learning_rate": 4.693904304129342e-06, + "loss": 3.8294, + "step": 2852 + }, + { + "epoch": 0.62, + "learning_rate": 4.693795062267861e-06, + "loss": 3.6935, + "step": 2853 + }, + { + "epoch": 0.62, + "learning_rate": 4.6936858204063804e-06, + "loss": 3.6326, + "step": 2854 + }, + { + "epoch": 0.62, + "learning_rate": 4.693576578544899e-06, + "loss": 3.2423, + "step": 2855 + }, + { + "epoch": 0.62, + "learning_rate": 4.693467336683418e-06, + "loss": 3.4809, + "step": 2856 + }, + { + "epoch": 0.62, + "learning_rate": 4.693358094821936e-06, + "loss": 3.5796, + "step": 2857 + }, + { + "epoch": 0.62, + "learning_rate": 4.693248852960455e-06, + "loss": 3.57, + "step": 2858 + }, + { + "epoch": 0.62, + "learning_rate": 4.693139611098974e-06, + "loss": 3.8457, + "step": 2859 + }, + { + "epoch": 0.62, + "learning_rate": 4.693030369237492e-06, + "loss": 3.3983, + "step": 2860 + }, + { + "epoch": 0.62, + "learning_rate": 4.692921127376011e-06, + "loss": 3.2749, + "step": 2861 + }, + { + "epoch": 0.62, + "learning_rate": 4.692811885514529e-06, + "loss": 3.4825, + "step": 2862 + }, + { + "epoch": 0.62, + "learning_rate": 4.692702643653048e-06, + "loss": 3.296, + "step": 2863 + }, + { + "epoch": 0.63, + "learning_rate": 4.692593401791567e-06, + "loss": 3.5779, + "step": 2864 + }, + { + "epoch": 0.63, + "learning_rate": 4.692484159930086e-06, + "loss": 3.8879, + "step": 2865 + }, + { + "epoch": 0.63, + "learning_rate": 4.692374918068604e-06, + "loss": 3.6383, + "step": 2866 + }, + { + "epoch": 0.63, + "learning_rate": 4.692265676207123e-06, + "loss": 3.7902, + "step": 2867 + }, + { + "epoch": 0.63, + "learning_rate": 4.692156434345642e-06, + "loss": 3.7288, + "step": 2868 + }, + { + "epoch": 0.63, + "learning_rate": 4.69204719248416e-06, + "loss": 3.4158, + "step": 2869 + }, + { + "epoch": 0.63, + "learning_rate": 4.691937950622679e-06, + "loss": 3.6563, + "step": 2870 + }, + { + "epoch": 0.63, + "learning_rate": 4.691828708761197e-06, + "loss": 3.1806, + "step": 2871 + }, + { + "epoch": 0.63, + "learning_rate": 4.691719466899716e-06, + "loss": 3.4791, + "step": 2872 + }, + { + "epoch": 0.63, + "learning_rate": 4.691610225038235e-06, + "loss": 3.7919, + "step": 2873 + }, + { + "epoch": 0.63, + "learning_rate": 4.6915009831767535e-06, + "loss": 3.7762, + "step": 2874 + }, + { + "epoch": 0.63, + "learning_rate": 4.6913917413152725e-06, + "loss": 3.5742, + "step": 2875 + }, + { + "epoch": 0.63, + "learning_rate": 4.691282499453791e-06, + "loss": 3.1393, + "step": 2876 + }, + { + "epoch": 0.63, + "learning_rate": 4.69117325759231e-06, + "loss": 3.5969, + "step": 2877 + }, + { + "epoch": 0.63, + "learning_rate": 4.691064015730829e-06, + "loss": 3.1474, + "step": 2878 + }, + { + "epoch": 0.63, + "learning_rate": 4.690954773869347e-06, + "loss": 3.4239, + "step": 2879 + }, + { + "epoch": 0.63, + "learning_rate": 4.690845532007865e-06, + "loss": 3.4267, + "step": 2880 + }, + { + "epoch": 0.63, + "learning_rate": 4.690736290146384e-06, + "loss": 3.2553, + "step": 2881 + }, + { + "epoch": 0.63, + "learning_rate": 4.690627048284903e-06, + "loss": 3.173, + "step": 2882 + }, + { + "epoch": 0.63, + "learning_rate": 4.690517806423422e-06, + "loss": 3.1272, + "step": 2883 + }, + { + "epoch": 0.63, + "learning_rate": 4.6904085645619405e-06, + "loss": 3.0033, + "step": 2884 + }, + { + "epoch": 0.63, + "learning_rate": 4.690299322700459e-06, + "loss": 3.5931, + "step": 2885 + }, + { + "epoch": 0.63, + "learning_rate": 4.690190080838978e-06, + "loss": 3.7899, + "step": 2886 + }, + { + "epoch": 0.63, + "learning_rate": 4.690080838977497e-06, + "loss": 3.2643, + "step": 2887 + }, + { + "epoch": 0.63, + "learning_rate": 4.689971597116016e-06, + "loss": 3.5242, + "step": 2888 + }, + { + "epoch": 0.63, + "learning_rate": 4.689862355254534e-06, + "loss": 3.5048, + "step": 2889 + }, + { + "epoch": 0.63, + "learning_rate": 4.689753113393052e-06, + "loss": 3.8522, + "step": 2890 + }, + { + "epoch": 0.63, + "learning_rate": 4.689643871531571e-06, + "loss": 3.4142, + "step": 2891 + }, + { + "epoch": 0.63, + "learning_rate": 4.68953462967009e-06, + "loss": 3.2769, + "step": 2892 + }, + { + "epoch": 0.63, + "learning_rate": 4.689425387808608e-06, + "loss": 3.3137, + "step": 2893 + }, + { + "epoch": 0.63, + "learning_rate": 4.689316145947127e-06, + "loss": 3.7344, + "step": 2894 + }, + { + "epoch": 0.63, + "learning_rate": 4.689206904085646e-06, + "loss": 3.5031, + "step": 2895 + }, + { + "epoch": 0.63, + "learning_rate": 4.689097662224165e-06, + "loss": 3.6245, + "step": 2896 + }, + { + "epoch": 0.63, + "learning_rate": 4.688988420362684e-06, + "loss": 3.7283, + "step": 2897 + }, + { + "epoch": 0.63, + "learning_rate": 4.688879178501202e-06, + "loss": 3.5341, + "step": 2898 + }, + { + "epoch": 0.63, + "learning_rate": 4.68876993663972e-06, + "loss": 3.7629, + "step": 2899 + }, + { + "epoch": 0.63, + "learning_rate": 4.688660694778239e-06, + "loss": 3.7309, + "step": 2900 + }, + { + "epoch": 0.63, + "learning_rate": 4.688551452916758e-06, + "loss": 3.6958, + "step": 2901 + }, + { + "epoch": 0.63, + "learning_rate": 4.688442211055277e-06, + "loss": 3.3998, + "step": 2902 + }, + { + "epoch": 0.63, + "learning_rate": 4.688332969193795e-06, + "loss": 3.523, + "step": 2903 + }, + { + "epoch": 0.63, + "learning_rate": 4.6882237273323135e-06, + "loss": 3.7641, + "step": 2904 + }, + { + "epoch": 0.63, + "learning_rate": 4.6881144854708326e-06, + "loss": 3.3748, + "step": 2905 + }, + { + "epoch": 0.63, + "learning_rate": 4.688005243609352e-06, + "loss": 3.4694, + "step": 2906 + }, + { + "epoch": 0.63, + "learning_rate": 4.687896001747871e-06, + "loss": 3.8698, + "step": 2907 + }, + { + "epoch": 0.63, + "learning_rate": 4.687786759886389e-06, + "loss": 3.9488, + "step": 2908 + }, + { + "epoch": 0.63, + "learning_rate": 4.687677518024907e-06, + "loss": 3.4181, + "step": 2909 + }, + { + "epoch": 0.64, + "learning_rate": 4.687568276163426e-06, + "loss": 3.2149, + "step": 2910 + }, + { + "epoch": 0.64, + "learning_rate": 4.687459034301945e-06, + "loss": 3.5691, + "step": 2911 + }, + { + "epoch": 0.64, + "learning_rate": 4.687349792440464e-06, + "loss": 3.7656, + "step": 2912 + }, + { + "epoch": 0.64, + "learning_rate": 4.687240550578982e-06, + "loss": 3.5696, + "step": 2913 + }, + { + "epoch": 0.64, + "learning_rate": 4.6871313087175005e-06, + "loss": 3.5986, + "step": 2914 + }, + { + "epoch": 0.64, + "learning_rate": 4.6870220668560195e-06, + "loss": 3.6881, + "step": 2915 + }, + { + "epoch": 0.64, + "learning_rate": 4.6869128249945386e-06, + "loss": 3.6371, + "step": 2916 + }, + { + "epoch": 0.64, + "learning_rate": 4.686803583133058e-06, + "loss": 4.0811, + "step": 2917 + }, + { + "epoch": 0.64, + "learning_rate": 4.686694341271576e-06, + "loss": 3.5314, + "step": 2918 + }, + { + "epoch": 0.64, + "learning_rate": 4.686585099410094e-06, + "loss": 3.5129, + "step": 2919 + }, + { + "epoch": 0.64, + "learning_rate": 4.686475857548613e-06, + "loss": 3.8678, + "step": 2920 + }, + { + "epoch": 0.64, + "learning_rate": 4.686366615687132e-06, + "loss": 3.2844, + "step": 2921 + }, + { + "epoch": 0.64, + "learning_rate": 4.68625737382565e-06, + "loss": 3.481, + "step": 2922 + }, + { + "epoch": 0.64, + "learning_rate": 4.686148131964169e-06, + "loss": 3.7473, + "step": 2923 + }, + { + "epoch": 0.64, + "learning_rate": 4.6860388901026874e-06, + "loss": 3.7781, + "step": 2924 + }, + { + "epoch": 0.64, + "learning_rate": 4.6859296482412065e-06, + "loss": 3.9581, + "step": 2925 + }, + { + "epoch": 0.64, + "learning_rate": 4.6858204063797255e-06, + "loss": 3.7917, + "step": 2926 + }, + { + "epoch": 0.64, + "learning_rate": 4.685711164518244e-06, + "loss": 3.8397, + "step": 2927 + }, + { + "epoch": 0.64, + "learning_rate": 4.685601922656762e-06, + "loss": 3.6156, + "step": 2928 + }, + { + "epoch": 0.64, + "learning_rate": 4.685492680795281e-06, + "loss": 3.4224, + "step": 2929 + }, + { + "epoch": 0.64, + "learning_rate": 4.6853834389338e-06, + "loss": 2.9214, + "step": 2930 + }, + { + "epoch": 0.64, + "learning_rate": 4.685274197072319e-06, + "loss": 3.2717, + "step": 2931 + }, + { + "epoch": 0.64, + "learning_rate": 4.685164955210837e-06, + "loss": 3.5759, + "step": 2932 + }, + { + "epoch": 0.64, + "learning_rate": 4.685055713349355e-06, + "loss": 3.3633, + "step": 2933 + }, + { + "epoch": 0.64, + "learning_rate": 4.684946471487874e-06, + "loss": 3.0508, + "step": 2934 + }, + { + "epoch": 0.64, + "learning_rate": 4.6848372296263934e-06, + "loss": 3.3382, + "step": 2935 + }, + { + "epoch": 0.64, + "learning_rate": 4.6847279877649125e-06, + "loss": 3.8679, + "step": 2936 + }, + { + "epoch": 0.64, + "learning_rate": 4.684618745903431e-06, + "loss": 3.55, + "step": 2937 + }, + { + "epoch": 0.64, + "learning_rate": 4.684509504041949e-06, + "loss": 3.4025, + "step": 2938 + }, + { + "epoch": 0.64, + "learning_rate": 4.684400262180468e-06, + "loss": 3.1934, + "step": 2939 + }, + { + "epoch": 0.64, + "learning_rate": 4.684291020318987e-06, + "loss": 3.3644, + "step": 2940 + }, + { + "epoch": 0.64, + "learning_rate": 4.684181778457506e-06, + "loss": 3.4762, + "step": 2941 + }, + { + "epoch": 0.64, + "learning_rate": 4.684072536596024e-06, + "loss": 3.2909, + "step": 2942 + }, + { + "epoch": 0.64, + "learning_rate": 4.683963294734542e-06, + "loss": 3.7422, + "step": 2943 + }, + { + "epoch": 0.64, + "learning_rate": 4.683854052873061e-06, + "loss": 3.5301, + "step": 2944 + }, + { + "epoch": 0.64, + "learning_rate": 4.68374481101158e-06, + "loss": 2.9686, + "step": 2945 + }, + { + "epoch": 0.64, + "learning_rate": 4.683635569150099e-06, + "loss": 3.4855, + "step": 2946 + }, + { + "epoch": 0.64, + "learning_rate": 4.683526327288618e-06, + "loss": 3.8736, + "step": 2947 + }, + { + "epoch": 0.64, + "learning_rate": 4.683417085427136e-06, + "loss": 3.3581, + "step": 2948 + }, + { + "epoch": 0.64, + "learning_rate": 4.683307843565655e-06, + "loss": 3.8855, + "step": 2949 + }, + { + "epoch": 0.64, + "learning_rate": 4.683198601704174e-06, + "loss": 3.8137, + "step": 2950 + }, + { + "epoch": 0.64, + "learning_rate": 4.683089359842692e-06, + "loss": 3.1475, + "step": 2951 + }, + { + "epoch": 0.64, + "learning_rate": 4.68298011798121e-06, + "loss": 3.9371, + "step": 2952 + }, + { + "epoch": 0.64, + "learning_rate": 4.682870876119729e-06, + "loss": 3.7794, + "step": 2953 + }, + { + "epoch": 0.64, + "learning_rate": 4.682761634258248e-06, + "loss": 3.7073, + "step": 2954 + }, + { + "epoch": 0.64, + "learning_rate": 4.682652392396767e-06, + "loss": 3.5855, + "step": 2955 + }, + { + "epoch": 0.65, + "learning_rate": 4.6825431505352855e-06, + "loss": 3.4151, + "step": 2956 + }, + { + "epoch": 0.65, + "learning_rate": 4.682433908673804e-06, + "loss": 3.7941, + "step": 2957 + }, + { + "epoch": 0.65, + "learning_rate": 4.682324666812323e-06, + "loss": 3.2582, + "step": 2958 + }, + { + "epoch": 0.65, + "learning_rate": 4.682215424950842e-06, + "loss": 3.305, + "step": 2959 + }, + { + "epoch": 0.65, + "learning_rate": 4.682106183089361e-06, + "loss": 3.3416, + "step": 2960 + }, + { + "epoch": 0.65, + "learning_rate": 4.681996941227879e-06, + "loss": 3.828, + "step": 2961 + }, + { + "epoch": 0.65, + "learning_rate": 4.681887699366397e-06, + "loss": 3.788, + "step": 2962 + }, + { + "epoch": 0.65, + "learning_rate": 4.681778457504916e-06, + "loss": 3.5489, + "step": 2963 + }, + { + "epoch": 0.65, + "learning_rate": 4.681669215643435e-06, + "loss": 3.6026, + "step": 2964 + }, + { + "epoch": 0.65, + "learning_rate": 4.6815599737819535e-06, + "loss": 3.215, + "step": 2965 + }, + { + "epoch": 0.65, + "learning_rate": 4.6814507319204725e-06, + "loss": 3.7618, + "step": 2966 + }, + { + "epoch": 0.65, + "learning_rate": 4.681341490058991e-06, + "loss": 2.8754, + "step": 2967 + }, + { + "epoch": 0.65, + "learning_rate": 4.68123224819751e-06, + "loss": 3.5311, + "step": 2968 + }, + { + "epoch": 0.65, + "learning_rate": 4.681123006336029e-06, + "loss": 3.673, + "step": 2969 + }, + { + "epoch": 0.65, + "learning_rate": 4.681013764474547e-06, + "loss": 3.9015, + "step": 2970 + }, + { + "epoch": 0.65, + "learning_rate": 4.680904522613066e-06, + "loss": 3.4928, + "step": 2971 + }, + { + "epoch": 0.65, + "learning_rate": 4.680795280751584e-06, + "loss": 3.5673, + "step": 2972 + }, + { + "epoch": 0.65, + "learning_rate": 4.680686038890103e-06, + "loss": 3.4492, + "step": 2973 + }, + { + "epoch": 0.65, + "learning_rate": 4.680576797028622e-06, + "loss": 3.3615, + "step": 2974 + }, + { + "epoch": 0.65, + "learning_rate": 4.68046755516714e-06, + "loss": 3.5115, + "step": 2975 + }, + { + "epoch": 0.65, + "learning_rate": 4.6803583133056594e-06, + "loss": 3.4949, + "step": 2976 + }, + { + "epoch": 0.65, + "learning_rate": 4.680249071444178e-06, + "loss": 3.2031, + "step": 2977 + }, + { + "epoch": 0.65, + "learning_rate": 4.680139829582697e-06, + "loss": 2.991, + "step": 2978 + }, + { + "epoch": 0.65, + "learning_rate": 4.680030587721215e-06, + "loss": 3.3863, + "step": 2979 + }, + { + "epoch": 0.65, + "learning_rate": 4.679921345859734e-06, + "loss": 3.5691, + "step": 2980 + }, + { + "epoch": 0.65, + "learning_rate": 4.679812103998252e-06, + "loss": 3.8087, + "step": 2981 + }, + { + "epoch": 0.65, + "learning_rate": 4.679702862136771e-06, + "loss": 3.3053, + "step": 2982 + }, + { + "epoch": 0.65, + "learning_rate": 4.67959362027529e-06, + "loss": 3.4924, + "step": 2983 + }, + { + "epoch": 0.65, + "learning_rate": 4.679484378413808e-06, + "loss": 3.7846, + "step": 2984 + }, + { + "epoch": 0.65, + "learning_rate": 4.679375136552327e-06, + "loss": 3.1929, + "step": 2985 + }, + { + "epoch": 0.65, + "learning_rate": 4.6792658946908456e-06, + "loss": 3.8523, + "step": 2986 + }, + { + "epoch": 0.65, + "learning_rate": 4.679156652829365e-06, + "loss": 3.1296, + "step": 2987 + }, + { + "epoch": 0.65, + "learning_rate": 4.679047410967884e-06, + "loss": 3.521, + "step": 2988 + }, + { + "epoch": 0.65, + "learning_rate": 4.678938169106402e-06, + "loss": 3.6308, + "step": 2989 + }, + { + "epoch": 0.65, + "learning_rate": 4.678828927244921e-06, + "loss": 3.5132, + "step": 2990 + }, + { + "epoch": 0.65, + "learning_rate": 4.678719685383439e-06, + "loss": 3.3549, + "step": 2991 + }, + { + "epoch": 0.65, + "learning_rate": 4.678610443521958e-06, + "loss": 3.2071, + "step": 2992 + }, + { + "epoch": 0.65, + "learning_rate": 4.678501201660477e-06, + "loss": 3.6681, + "step": 2993 + }, + { + "epoch": 0.65, + "learning_rate": 4.678391959798995e-06, + "loss": 3.8363, + "step": 2994 + }, + { + "epoch": 0.65, + "learning_rate": 4.678282717937514e-06, + "loss": 3.8878, + "step": 2995 + }, + { + "epoch": 0.65, + "learning_rate": 4.6781734760760325e-06, + "loss": 3.7177, + "step": 2996 + }, + { + "epoch": 0.65, + "learning_rate": 4.6780642342145515e-06, + "loss": 3.8205, + "step": 2997 + }, + { + "epoch": 0.65, + "learning_rate": 4.67795499235307e-06, + "loss": 3.4288, + "step": 2998 + }, + { + "epoch": 0.65, + "learning_rate": 4.677845750491589e-06, + "loss": 4.0177, + "step": 2999 + }, + { + "epoch": 0.65, + "learning_rate": 4.677736508630108e-06, + "loss": 3.4613, + "step": 3000 + }, + { + "epoch": 0.65, + "learning_rate": 4.677627266768626e-06, + "loss": 3.4856, + "step": 3001 + }, + { + "epoch": 0.66, + "learning_rate": 4.677518024907145e-06, + "loss": 3.1294, + "step": 3002 + }, + { + "epoch": 0.66, + "learning_rate": 4.677408783045663e-06, + "loss": 3.7923, + "step": 3003 + }, + { + "epoch": 0.66, + "learning_rate": 4.677299541184182e-06, + "loss": 3.5224, + "step": 3004 + }, + { + "epoch": 0.66, + "learning_rate": 4.6771902993227004e-06, + "loss": 3.2905, + "step": 3005 + }, + { + "epoch": 0.66, + "learning_rate": 4.6770810574612195e-06, + "loss": 3.7159, + "step": 3006 + }, + { + "epoch": 0.66, + "learning_rate": 4.6769718155997385e-06, + "loss": 3.438, + "step": 3007 + }, + { + "epoch": 0.66, + "learning_rate": 4.676862573738257e-06, + "loss": 3.745, + "step": 3008 + }, + { + "epoch": 0.66, + "learning_rate": 4.676753331876776e-06, + "loss": 3.5847, + "step": 3009 + }, + { + "epoch": 0.66, + "learning_rate": 4.676644090015294e-06, + "loss": 3.5254, + "step": 3010 + }, + { + "epoch": 0.66, + "learning_rate": 4.676534848153813e-06, + "loss": 3.5023, + "step": 3011 + }, + { + "epoch": 0.66, + "learning_rate": 4.676425606292331e-06, + "loss": 3.7048, + "step": 3012 + }, + { + "epoch": 0.66, + "learning_rate": 4.67631636443085e-06, + "loss": 3.5679, + "step": 3013 + }, + { + "epoch": 0.66, + "learning_rate": 4.676207122569369e-06, + "loss": 3.7258, + "step": 3014 + }, + { + "epoch": 0.66, + "learning_rate": 4.676097880707887e-06, + "loss": 3.6989, + "step": 3015 + }, + { + "epoch": 0.66, + "learning_rate": 4.6759886388464064e-06, + "loss": 3.8495, + "step": 3016 + }, + { + "epoch": 0.66, + "learning_rate": 4.675879396984925e-06, + "loss": 3.771, + "step": 3017 + }, + { + "epoch": 0.66, + "learning_rate": 4.675770155123444e-06, + "loss": 3.6995, + "step": 3018 + }, + { + "epoch": 0.66, + "learning_rate": 4.675660913261963e-06, + "loss": 3.9136, + "step": 3019 + }, + { + "epoch": 0.66, + "learning_rate": 4.675551671400481e-06, + "loss": 3.387, + "step": 3020 + }, + { + "epoch": 0.66, + "learning_rate": 4.675442429539e-06, + "loss": 3.1314, + "step": 3021 + }, + { + "epoch": 0.66, + "learning_rate": 4.675333187677518e-06, + "loss": 3.4805, + "step": 3022 + }, + { + "epoch": 0.66, + "learning_rate": 4.675223945816037e-06, + "loss": 3.1692, + "step": 3023 + }, + { + "epoch": 0.66, + "learning_rate": 4.675114703954556e-06, + "loss": 3.8656, + "step": 3024 + }, + { + "epoch": 0.66, + "learning_rate": 4.675005462093074e-06, + "loss": 3.5442, + "step": 3025 + }, + { + "epoch": 0.66, + "learning_rate": 4.674896220231593e-06, + "loss": 3.4009, + "step": 3026 + }, + { + "epoch": 0.66, + "learning_rate": 4.6747869783701116e-06, + "loss": 3.439, + "step": 3027 + }, + { + "epoch": 0.66, + "learning_rate": 4.674677736508631e-06, + "loss": 3.4568, + "step": 3028 + }, + { + "epoch": 0.66, + "learning_rate": 4.67456849464715e-06, + "loss": 3.7412, + "step": 3029 + }, + { + "epoch": 0.66, + "learning_rate": 4.674459252785668e-06, + "loss": 3.7228, + "step": 3030 + }, + { + "epoch": 0.66, + "learning_rate": 4.674350010924186e-06, + "loss": 3.6193, + "step": 3031 + }, + { + "epoch": 0.66, + "learning_rate": 4.674240769062705e-06, + "loss": 3.4928, + "step": 3032 + }, + { + "epoch": 0.66, + "learning_rate": 4.674131527201224e-06, + "loss": 3.78, + "step": 3033 + }, + { + "epoch": 0.66, + "learning_rate": 4.674022285339742e-06, + "loss": 3.2787, + "step": 3034 + }, + { + "epoch": 0.66, + "learning_rate": 4.673913043478261e-06, + "loss": 2.9828, + "step": 3035 + }, + { + "epoch": 0.66, + "learning_rate": 4.6738038016167795e-06, + "loss": 3.3868, + "step": 3036 + }, + { + "epoch": 0.66, + "learning_rate": 4.6736945597552985e-06, + "loss": 3.449, + "step": 3037 + }, + { + "epoch": 0.66, + "learning_rate": 4.6735853178938176e-06, + "loss": 3.9076, + "step": 3038 + }, + { + "epoch": 0.66, + "learning_rate": 4.673476076032336e-06, + "loss": 3.7227, + "step": 3039 + }, + { + "epoch": 0.66, + "learning_rate": 4.673366834170855e-06, + "loss": 3.6434, + "step": 3040 + }, + { + "epoch": 0.66, + "learning_rate": 4.673257592309373e-06, + "loss": 3.398, + "step": 3041 + }, + { + "epoch": 0.66, + "learning_rate": 4.673148350447892e-06, + "loss": 3.4075, + "step": 3042 + }, + { + "epoch": 0.66, + "learning_rate": 4.673039108586411e-06, + "loss": 3.4679, + "step": 3043 + }, + { + "epoch": 0.66, + "learning_rate": 4.672929866724929e-06, + "loss": 3.5674, + "step": 3044 + }, + { + "epoch": 0.66, + "learning_rate": 4.672820624863447e-06, + "loss": 3.8662, + "step": 3045 + }, + { + "epoch": 0.66, + "learning_rate": 4.6727113830019665e-06, + "loss": 3.4923, + "step": 3046 + }, + { + "epoch": 0.66, + "learning_rate": 4.6726021411404855e-06, + "loss": 3.515, + "step": 3047 + }, + { + "epoch": 0.67, + "learning_rate": 4.6724928992790045e-06, + "loss": 3.6937, + "step": 3048 + }, + { + "epoch": 0.67, + "learning_rate": 4.672383657417523e-06, + "loss": 3.2058, + "step": 3049 + }, + { + "epoch": 0.67, + "learning_rate": 4.672274415556041e-06, + "loss": 3.4785, + "step": 3050 + }, + { + "epoch": 0.67, + "learning_rate": 4.67216517369456e-06, + "loss": 3.3092, + "step": 3051 + }, + { + "epoch": 0.67, + "learning_rate": 4.672055931833079e-06, + "loss": 3.5442, + "step": 3052 + }, + { + "epoch": 0.67, + "learning_rate": 4.671946689971598e-06, + "loss": 3.778, + "step": 3053 + }, + { + "epoch": 0.67, + "learning_rate": 4.671837448110116e-06, + "loss": 3.5535, + "step": 3054 + }, + { + "epoch": 0.67, + "learning_rate": 4.671728206248634e-06, + "loss": 3.2477, + "step": 3055 + }, + { + "epoch": 0.67, + "learning_rate": 4.671618964387153e-06, + "loss": 3.7833, + "step": 3056 + }, + { + "epoch": 0.67, + "learning_rate": 4.6715097225256724e-06, + "loss": 3.8869, + "step": 3057 + }, + { + "epoch": 0.67, + "learning_rate": 4.671400480664191e-06, + "loss": 3.396, + "step": 3058 + }, + { + "epoch": 0.67, + "learning_rate": 4.67129123880271e-06, + "loss": 3.6022, + "step": 3059 + }, + { + "epoch": 0.67, + "learning_rate": 4.671181996941228e-06, + "loss": 3.3051, + "step": 3060 + }, + { + "epoch": 0.67, + "learning_rate": 4.671072755079747e-06, + "loss": 3.7467, + "step": 3061 + }, + { + "epoch": 0.67, + "learning_rate": 4.670963513218266e-06, + "loss": 3.7392, + "step": 3062 + }, + { + "epoch": 0.67, + "learning_rate": 4.670854271356784e-06, + "loss": 3.5099, + "step": 3063 + }, + { + "epoch": 0.67, + "learning_rate": 4.670745029495303e-06, + "loss": 3.0734, + "step": 3064 + }, + { + "epoch": 0.67, + "learning_rate": 4.670635787633821e-06, + "loss": 3.9128, + "step": 3065 + }, + { + "epoch": 0.67, + "learning_rate": 4.67052654577234e-06, + "loss": 3.439, + "step": 3066 + }, + { + "epoch": 0.67, + "learning_rate": 4.670417303910859e-06, + "loss": 3.8747, + "step": 3067 + }, + { + "epoch": 0.67, + "learning_rate": 4.670308062049378e-06, + "loss": 3.1863, + "step": 3068 + }, + { + "epoch": 0.67, + "learning_rate": 4.670198820187896e-06, + "loss": 3.3764, + "step": 3069 + }, + { + "epoch": 0.67, + "learning_rate": 4.670089578326415e-06, + "loss": 3.1954, + "step": 3070 + }, + { + "epoch": 0.67, + "learning_rate": 4.669980336464934e-06, + "loss": 3.4142, + "step": 3071 + }, + { + "epoch": 0.67, + "learning_rate": 4.669871094603453e-06, + "loss": 3.5003, + "step": 3072 + }, + { + "epoch": 0.67, + "learning_rate": 4.669761852741971e-06, + "loss": 3.6389, + "step": 3073 + }, + { + "epoch": 0.67, + "learning_rate": 4.669652610880489e-06, + "loss": 3.9921, + "step": 3074 + }, + { + "epoch": 0.67, + "learning_rate": 4.669543369019008e-06, + "loss": 3.2896, + "step": 3075 + }, + { + "epoch": 0.67, + "learning_rate": 4.669434127157527e-06, + "loss": 3.6021, + "step": 3076 + }, + { + "epoch": 0.67, + "learning_rate": 4.669324885296046e-06, + "loss": 3.2833, + "step": 3077 + }, + { + "epoch": 0.67, + "learning_rate": 4.6692156434345645e-06, + "loss": 3.8543, + "step": 3078 + }, + { + "epoch": 0.67, + "learning_rate": 4.669106401573083e-06, + "loss": 3.4852, + "step": 3079 + }, + { + "epoch": 0.67, + "learning_rate": 4.668997159711602e-06, + "loss": 3.7175, + "step": 3080 + }, + { + "epoch": 0.67, + "learning_rate": 4.668887917850121e-06, + "loss": 3.7751, + "step": 3081 + }, + { + "epoch": 0.67, + "learning_rate": 4.66877867598864e-06, + "loss": 3.7951, + "step": 3082 + }, + { + "epoch": 0.67, + "learning_rate": 4.668669434127158e-06, + "loss": 3.3926, + "step": 3083 + }, + { + "epoch": 0.67, + "learning_rate": 4.668560192265676e-06, + "loss": 3.0868, + "step": 3084 + }, + { + "epoch": 0.67, + "learning_rate": 4.668450950404195e-06, + "loss": 3.5204, + "step": 3085 + }, + { + "epoch": 0.67, + "learning_rate": 4.668341708542714e-06, + "loss": 3.169, + "step": 3086 + }, + { + "epoch": 0.67, + "learning_rate": 4.6682324666812325e-06, + "loss": 3.5998, + "step": 3087 + }, + { + "epoch": 0.67, + "learning_rate": 4.6681232248197515e-06, + "loss": 3.5847, + "step": 3088 + }, + { + "epoch": 0.67, + "learning_rate": 4.66801398295827e-06, + "loss": 3.2457, + "step": 3089 + }, + { + "epoch": 0.67, + "learning_rate": 4.667904741096789e-06, + "loss": 3.7125, + "step": 3090 + }, + { + "epoch": 0.67, + "learning_rate": 4.667795499235308e-06, + "loss": 3.7724, + "step": 3091 + }, + { + "epoch": 0.67, + "learning_rate": 4.667686257373826e-06, + "loss": 3.0061, + "step": 3092 + }, + { + "epoch": 0.68, + "learning_rate": 4.667577015512344e-06, + "loss": 2.755, + "step": 3093 + }, + { + "epoch": 0.68, + "learning_rate": 4.667467773650863e-06, + "loss": 3.5291, + "step": 3094 + }, + { + "epoch": 0.68, + "learning_rate": 4.667358531789382e-06, + "loss": 3.6068, + "step": 3095 + }, + { + "epoch": 0.68, + "learning_rate": 4.667249289927901e-06, + "loss": 3.4983, + "step": 3096 + }, + { + "epoch": 0.68, + "learning_rate": 4.6671400480664194e-06, + "loss": 3.1971, + "step": 3097 + }, + { + "epoch": 0.68, + "learning_rate": 4.667030806204938e-06, + "loss": 3.6161, + "step": 3098 + }, + { + "epoch": 0.68, + "learning_rate": 4.666921564343457e-06, + "loss": 3.4222, + "step": 3099 + }, + { + "epoch": 0.68, + "learning_rate": 4.666812322481976e-06, + "loss": 3.7672, + "step": 3100 + }, + { + "epoch": 0.68, + "learning_rate": 4.666703080620495e-06, + "loss": 3.8165, + "step": 3101 + }, + { + "epoch": 0.68, + "learning_rate": 4.666593838759013e-06, + "loss": 3.7512, + "step": 3102 + }, + { + "epoch": 0.68, + "learning_rate": 4.666484596897531e-06, + "loss": 3.0788, + "step": 3103 + }, + { + "epoch": 0.68, + "learning_rate": 4.66637535503605e-06, + "loss": 3.0949, + "step": 3104 + }, + { + "epoch": 0.68, + "learning_rate": 4.666266113174569e-06, + "loss": 3.597, + "step": 3105 + }, + { + "epoch": 0.68, + "learning_rate": 4.666156871313088e-06, + "loss": 3.951, + "step": 3106 + }, + { + "epoch": 0.68, + "learning_rate": 4.666047629451606e-06, + "loss": 3.8105, + "step": 3107 + }, + { + "epoch": 0.68, + "learning_rate": 4.6659383875901246e-06, + "loss": 3.1878, + "step": 3108 + }, + { + "epoch": 0.68, + "learning_rate": 4.665829145728644e-06, + "loss": 3.7934, + "step": 3109 + }, + { + "epoch": 0.68, + "learning_rate": 4.665719903867163e-06, + "loss": 3.4269, + "step": 3110 + }, + { + "epoch": 0.68, + "learning_rate": 4.665610662005681e-06, + "loss": 3.6228, + "step": 3111 + }, + { + "epoch": 0.68, + "learning_rate": 4.6655014201442e-06, + "loss": 3.8132, + "step": 3112 + }, + { + "epoch": 0.68, + "learning_rate": 4.665392178282718e-06, + "loss": 3.5527, + "step": 3113 + }, + { + "epoch": 0.68, + "learning_rate": 4.665282936421237e-06, + "loss": 3.8364, + "step": 3114 + }, + { + "epoch": 0.68, + "learning_rate": 4.665173694559756e-06, + "loss": 3.5656, + "step": 3115 + }, + { + "epoch": 0.68, + "learning_rate": 4.665064452698274e-06, + "loss": 3.477, + "step": 3116 + }, + { + "epoch": 0.68, + "learning_rate": 4.664955210836793e-06, + "loss": 3.4678, + "step": 3117 + }, + { + "epoch": 0.68, + "learning_rate": 4.6648459689753115e-06, + "loss": 3.6188, + "step": 3118 + }, + { + "epoch": 0.68, + "learning_rate": 4.6647367271138306e-06, + "loss": 3.9206, + "step": 3119 + }, + { + "epoch": 0.68, + "learning_rate": 4.66462748525235e-06, + "loss": 3.2104, + "step": 3120 + }, + { + "epoch": 0.68, + "learning_rate": 4.664518243390868e-06, + "loss": 3.6832, + "step": 3121 + }, + { + "epoch": 0.68, + "learning_rate": 4.664409001529386e-06, + "loss": 3.4841, + "step": 3122 + }, + { + "epoch": 0.68, + "learning_rate": 4.664299759667905e-06, + "loss": 3.2365, + "step": 3123 + }, + { + "epoch": 0.68, + "learning_rate": 4.664190517806424e-06, + "loss": 3.1617, + "step": 3124 + }, + { + "epoch": 0.68, + "learning_rate": 4.664081275944943e-06, + "loss": 3.4051, + "step": 3125 + }, + { + "epoch": 0.68, + "learning_rate": 4.663972034083461e-06, + "loss": 3.6542, + "step": 3126 + }, + { + "epoch": 0.68, + "learning_rate": 4.6638627922219794e-06, + "loss": 3.7964, + "step": 3127 + }, + { + "epoch": 0.68, + "learning_rate": 4.6637535503604985e-06, + "loss": 3.3892, + "step": 3128 + }, + { + "epoch": 0.68, + "learning_rate": 4.6636443084990175e-06, + "loss": 3.3882, + "step": 3129 + }, + { + "epoch": 0.68, + "learning_rate": 4.663535066637536e-06, + "loss": 3.6952, + "step": 3130 + }, + { + "epoch": 0.68, + "learning_rate": 4.663425824776055e-06, + "loss": 3.721, + "step": 3131 + }, + { + "epoch": 0.68, + "learning_rate": 4.663316582914573e-06, + "loss": 3.1696, + "step": 3132 + }, + { + "epoch": 0.68, + "learning_rate": 4.663207341053092e-06, + "loss": 3.3856, + "step": 3133 + }, + { + "epoch": 0.68, + "learning_rate": 4.663098099191611e-06, + "loss": 3.2033, + "step": 3134 + }, + { + "epoch": 0.68, + "learning_rate": 4.662988857330129e-06, + "loss": 3.5358, + "step": 3135 + }, + { + "epoch": 0.68, + "learning_rate": 4.662879615468648e-06, + "loss": 3.183, + "step": 3136 + }, + { + "epoch": 0.68, + "learning_rate": 4.662770373607166e-06, + "loss": 3.6332, + "step": 3137 + }, + { + "epoch": 0.68, + "learning_rate": 4.6626611317456854e-06, + "loss": 3.6744, + "step": 3138 + }, + { + "epoch": 0.69, + "learning_rate": 4.6625518898842045e-06, + "loss": 3.8485, + "step": 3139 + }, + { + "epoch": 0.69, + "learning_rate": 4.662442648022723e-06, + "loss": 3.9191, + "step": 3140 + }, + { + "epoch": 0.69, + "learning_rate": 4.662333406161242e-06, + "loss": 3.3887, + "step": 3141 + }, + { + "epoch": 0.69, + "learning_rate": 4.66222416429976e-06, + "loss": 3.6351, + "step": 3142 + }, + { + "epoch": 0.69, + "learning_rate": 4.662114922438279e-06, + "loss": 3.5382, + "step": 3143 + }, + { + "epoch": 0.69, + "learning_rate": 4.662005680576797e-06, + "loss": 3.694, + "step": 3144 + }, + { + "epoch": 0.69, + "learning_rate": 4.661896438715316e-06, + "loss": 3.6437, + "step": 3145 + }, + { + "epoch": 0.69, + "learning_rate": 4.661787196853834e-06, + "loss": 3.2622, + "step": 3146 + }, + { + "epoch": 0.69, + "learning_rate": 4.661677954992353e-06, + "loss": 3.6091, + "step": 3147 + }, + { + "epoch": 0.69, + "learning_rate": 4.661568713130872e-06, + "loss": 3.735, + "step": 3148 + }, + { + "epoch": 0.69, + "learning_rate": 4.661459471269391e-06, + "loss": 3.375, + "step": 3149 + }, + { + "epoch": 0.69, + "learning_rate": 4.66135022940791e-06, + "loss": 3.6999, + "step": 3150 + }, + { + "epoch": 0.69, + "learning_rate": 4.661240987546428e-06, + "loss": 3.1676, + "step": 3151 + }, + { + "epoch": 0.69, + "learning_rate": 4.661131745684947e-06, + "loss": 3.4463, + "step": 3152 + }, + { + "epoch": 0.69, + "learning_rate": 4.661022503823466e-06, + "loss": 3.4002, + "step": 3153 + }, + { + "epoch": 0.69, + "learning_rate": 4.660913261961984e-06, + "loss": 3.844, + "step": 3154 + }, + { + "epoch": 0.69, + "learning_rate": 4.660804020100503e-06, + "loss": 3.8464, + "step": 3155 + }, + { + "epoch": 0.69, + "learning_rate": 4.660694778239021e-06, + "loss": 3.484, + "step": 3156 + }, + { + "epoch": 0.69, + "learning_rate": 4.66058553637754e-06, + "loss": 3.244, + "step": 3157 + }, + { + "epoch": 0.69, + "learning_rate": 4.660476294516059e-06, + "loss": 3.6584, + "step": 3158 + }, + { + "epoch": 0.69, + "learning_rate": 4.6603670526545775e-06, + "loss": 3.4714, + "step": 3159 + }, + { + "epoch": 0.69, + "learning_rate": 4.6602578107930966e-06, + "loss": 3.3204, + "step": 3160 + }, + { + "epoch": 0.69, + "learning_rate": 4.660148568931615e-06, + "loss": 3.1949, + "step": 3161 + }, + { + "epoch": 0.69, + "learning_rate": 4.660039327070134e-06, + "loss": 3.2882, + "step": 3162 + }, + { + "epoch": 0.69, + "learning_rate": 4.659930085208652e-06, + "loss": 3.2267, + "step": 3163 + }, + { + "epoch": 0.69, + "learning_rate": 4.659820843347171e-06, + "loss": 3.1528, + "step": 3164 + }, + { + "epoch": 0.69, + "learning_rate": 4.65971160148569e-06, + "loss": 3.7666, + "step": 3165 + }, + { + "epoch": 0.69, + "learning_rate": 4.659602359624208e-06, + "loss": 3.3555, + "step": 3166 + }, + { + "epoch": 0.69, + "learning_rate": 4.659493117762727e-06, + "loss": 3.6136, + "step": 3167 + }, + { + "epoch": 0.69, + "learning_rate": 4.6593838759012455e-06, + "loss": 3.64, + "step": 3168 + }, + { + "epoch": 0.69, + "learning_rate": 4.6592746340397645e-06, + "loss": 3.8496, + "step": 3169 + }, + { + "epoch": 0.69, + "learning_rate": 4.6591653921782835e-06, + "loss": 3.3874, + "step": 3170 + }, + { + "epoch": 0.69, + "learning_rate": 4.659056150316802e-06, + "loss": 3.2724, + "step": 3171 + }, + { + "epoch": 0.69, + "learning_rate": 4.658946908455321e-06, + "loss": 3.8452, + "step": 3172 + }, + { + "epoch": 0.69, + "learning_rate": 4.658837666593839e-06, + "loss": 3.1644, + "step": 3173 + }, + { + "epoch": 0.69, + "learning_rate": 4.658728424732358e-06, + "loss": 3.888, + "step": 3174 + }, + { + "epoch": 0.69, + "learning_rate": 4.658619182870876e-06, + "loss": 3.6732, + "step": 3175 + }, + { + "epoch": 0.69, + "learning_rate": 4.658509941009395e-06, + "loss": 3.329, + "step": 3176 + }, + { + "epoch": 0.69, + "learning_rate": 4.658400699147913e-06, + "loss": 3.7521, + "step": 3177 + }, + { + "epoch": 0.69, + "learning_rate": 4.658291457286432e-06, + "loss": 3.6338, + "step": 3178 + }, + { + "epoch": 0.69, + "learning_rate": 4.6581822154249515e-06, + "loss": 3.3387, + "step": 3179 + }, + { + "epoch": 0.69, + "learning_rate": 4.65807297356347e-06, + "loss": 3.766, + "step": 3180 + }, + { + "epoch": 0.69, + "learning_rate": 4.657963731701989e-06, + "loss": 3.7453, + "step": 3181 + }, + { + "epoch": 0.69, + "learning_rate": 4.657854489840507e-06, + "loss": 3.8583, + "step": 3182 + }, + { + "epoch": 0.69, + "learning_rate": 4.657745247979026e-06, + "loss": 3.6917, + "step": 3183 + }, + { + "epoch": 0.69, + "learning_rate": 4.657636006117545e-06, + "loss": 3.286, + "step": 3184 + }, + { + "epoch": 0.7, + "learning_rate": 4.657526764256063e-06, + "loss": 3.4838, + "step": 3185 + }, + { + "epoch": 0.7, + "learning_rate": 4.657417522394582e-06, + "loss": 3.7432, + "step": 3186 + }, + { + "epoch": 0.7, + "learning_rate": 4.6573082805331e-06, + "loss": 3.5435, + "step": 3187 + }, + { + "epoch": 0.7, + "learning_rate": 4.657199038671619e-06, + "loss": 3.815, + "step": 3188 + }, + { + "epoch": 0.7, + "learning_rate": 4.657089796810138e-06, + "loss": 3.4717, + "step": 3189 + }, + { + "epoch": 0.7, + "learning_rate": 4.656980554948657e-06, + "loss": 3.9579, + "step": 3190 + }, + { + "epoch": 0.7, + "learning_rate": 4.656871313087176e-06, + "loss": 3.2745, + "step": 3191 + }, + { + "epoch": 0.7, + "learning_rate": 4.656762071225694e-06, + "loss": 3.2746, + "step": 3192 + }, + { + "epoch": 0.7, + "learning_rate": 4.656652829364213e-06, + "loss": 3.5862, + "step": 3193 + }, + { + "epoch": 0.7, + "learning_rate": 4.656543587502732e-06, + "loss": 3.7461, + "step": 3194 + }, + { + "epoch": 0.7, + "learning_rate": 4.65643434564125e-06, + "loss": 3.198, + "step": 3195 + }, + { + "epoch": 0.7, + "learning_rate": 4.656325103779768e-06, + "loss": 3.3506, + "step": 3196 + }, + { + "epoch": 0.7, + "learning_rate": 4.656215861918287e-06, + "loss": 3.9525, + "step": 3197 + }, + { + "epoch": 0.7, + "learning_rate": 4.656106620056806e-06, + "loss": 3.6122, + "step": 3198 + }, + { + "epoch": 0.7, + "learning_rate": 4.6559973781953245e-06, + "loss": 3.3527, + "step": 3199 + }, + { + "epoch": 0.7, + "learning_rate": 4.6558881363338436e-06, + "loss": 3.1584, + "step": 3200 + }, + { + "epoch": 0.7, + "learning_rate": 4.655778894472362e-06, + "loss": 3.9097, + "step": 3201 + }, + { + "epoch": 0.7, + "learning_rate": 4.655669652610881e-06, + "loss": 3.464, + "step": 3202 + }, + { + "epoch": 0.7, + "learning_rate": 4.6555604107494e-06, + "loss": 3.409, + "step": 3203 + }, + { + "epoch": 0.7, + "learning_rate": 4.655451168887918e-06, + "loss": 3.3848, + "step": 3204 + }, + { + "epoch": 0.7, + "learning_rate": 4.655341927026437e-06, + "loss": 3.4662, + "step": 3205 + }, + { + "epoch": 0.7, + "learning_rate": 4.655232685164955e-06, + "loss": 3.402, + "step": 3206 + }, + { + "epoch": 0.7, + "learning_rate": 4.655123443303474e-06, + "loss": 2.9614, + "step": 3207 + }, + { + "epoch": 0.7, + "learning_rate": 4.655014201441993e-06, + "loss": 3.8323, + "step": 3208 + }, + { + "epoch": 0.7, + "learning_rate": 4.6549049595805115e-06, + "loss": 3.4736, + "step": 3209 + }, + { + "epoch": 0.7, + "learning_rate": 4.65479571771903e-06, + "loss": 3.5749, + "step": 3210 + }, + { + "epoch": 0.7, + "learning_rate": 4.654686475857549e-06, + "loss": 3.6197, + "step": 3211 + }, + { + "epoch": 0.7, + "learning_rate": 4.654577233996068e-06, + "loss": 3.4399, + "step": 3212 + }, + { + "epoch": 0.7, + "learning_rate": 4.654467992134587e-06, + "loss": 3.2717, + "step": 3213 + }, + { + "epoch": 0.7, + "learning_rate": 4.654358750273105e-06, + "loss": 3.5055, + "step": 3214 + }, + { + "epoch": 0.7, + "learning_rate": 4.654249508411623e-06, + "loss": 3.4407, + "step": 3215 + }, + { + "epoch": 0.7, + "learning_rate": 4.654140266550142e-06, + "loss": 3.3938, + "step": 3216 + }, + { + "epoch": 0.7, + "learning_rate": 4.654031024688661e-06, + "loss": 3.4935, + "step": 3217 + }, + { + "epoch": 0.7, + "learning_rate": 4.65392178282718e-06, + "loss": 3.6301, + "step": 3218 + }, + { + "epoch": 0.7, + "learning_rate": 4.6538125409656984e-06, + "loss": 3.1843, + "step": 3219 + }, + { + "epoch": 0.7, + "learning_rate": 4.653703299104217e-06, + "loss": 3.6525, + "step": 3220 + }, + { + "epoch": 0.7, + "learning_rate": 4.653594057242736e-06, + "loss": 3.5099, + "step": 3221 + }, + { + "epoch": 0.7, + "learning_rate": 4.653484815381255e-06, + "loss": 3.6633, + "step": 3222 + }, + { + "epoch": 0.7, + "learning_rate": 4.653375573519774e-06, + "loss": 3.8296, + "step": 3223 + }, + { + "epoch": 0.7, + "learning_rate": 4.653266331658292e-06, + "loss": 3.1319, + "step": 3224 + }, + { + "epoch": 0.7, + "learning_rate": 4.65315708979681e-06, + "loss": 3.4311, + "step": 3225 + }, + { + "epoch": 0.7, + "learning_rate": 4.653047847935329e-06, + "loss": 3.5129, + "step": 3226 + }, + { + "epoch": 0.7, + "learning_rate": 4.652938606073848e-06, + "loss": 3.32, + "step": 3227 + }, + { + "epoch": 0.7, + "learning_rate": 4.652829364212366e-06, + "loss": 3.7389, + "step": 3228 + }, + { + "epoch": 0.7, + "learning_rate": 4.652720122350885e-06, + "loss": 3.4246, + "step": 3229 + }, + { + "epoch": 0.7, + "learning_rate": 4.652610880489404e-06, + "loss": 3.2752, + "step": 3230 + }, + { + "epoch": 0.71, + "learning_rate": 4.652501638627923e-06, + "loss": 3.5135, + "step": 3231 + }, + { + "epoch": 0.71, + "learning_rate": 4.652392396766442e-06, + "loss": 3.7972, + "step": 3232 + }, + { + "epoch": 0.71, + "learning_rate": 4.65228315490496e-06, + "loss": 3.2114, + "step": 3233 + }, + { + "epoch": 0.71, + "learning_rate": 4.652173913043478e-06, + "loss": 3.5846, + "step": 3234 + }, + { + "epoch": 0.71, + "learning_rate": 4.652064671181997e-06, + "loss": 3.1353, + "step": 3235 + }, + { + "epoch": 0.71, + "learning_rate": 4.651955429320516e-06, + "loss": 3.1104, + "step": 3236 + }, + { + "epoch": 0.71, + "learning_rate": 4.651846187459035e-06, + "loss": 3.4847, + "step": 3237 + }, + { + "epoch": 0.71, + "learning_rate": 4.651736945597553e-06, + "loss": 3.2056, + "step": 3238 + }, + { + "epoch": 0.71, + "learning_rate": 4.6516277037360715e-06, + "loss": 3.1477, + "step": 3239 + }, + { + "epoch": 0.71, + "learning_rate": 4.6515184618745905e-06, + "loss": 3.4754, + "step": 3240 + }, + { + "epoch": 0.71, + "learning_rate": 4.6514092200131096e-06, + "loss": 3.5037, + "step": 3241 + }, + { + "epoch": 0.71, + "learning_rate": 4.651299978151629e-06, + "loss": 3.4066, + "step": 3242 + }, + { + "epoch": 0.71, + "learning_rate": 4.651190736290147e-06, + "loss": 3.0355, + "step": 3243 + }, + { + "epoch": 0.71, + "learning_rate": 4.651081494428665e-06, + "loss": 3.2717, + "step": 3244 + }, + { + "epoch": 0.71, + "learning_rate": 4.650972252567184e-06, + "loss": 3.259, + "step": 3245 + }, + { + "epoch": 0.71, + "learning_rate": 4.650863010705703e-06, + "loss": 3.463, + "step": 3246 + }, + { + "epoch": 0.71, + "learning_rate": 4.650753768844222e-06, + "loss": 3.6274, + "step": 3247 + }, + { + "epoch": 0.71, + "learning_rate": 4.65064452698274e-06, + "loss": 3.3785, + "step": 3248 + }, + { + "epoch": 0.71, + "learning_rate": 4.6505352851212585e-06, + "loss": 3.6999, + "step": 3249 + }, + { + "epoch": 0.71, + "learning_rate": 4.6504260432597775e-06, + "loss": 3.2986, + "step": 3250 + }, + { + "epoch": 0.71, + "learning_rate": 4.6503168013982965e-06, + "loss": 3.2238, + "step": 3251 + }, + { + "epoch": 0.71, + "learning_rate": 4.650207559536815e-06, + "loss": 3.6938, + "step": 3252 + }, + { + "epoch": 0.71, + "learning_rate": 4.650098317675334e-06, + "loss": 3.3127, + "step": 3253 + }, + { + "epoch": 0.71, + "learning_rate": 4.649989075813852e-06, + "loss": 3.3675, + "step": 3254 + }, + { + "epoch": 0.71, + "learning_rate": 4.649879833952371e-06, + "loss": 3.7217, + "step": 3255 + }, + { + "epoch": 0.71, + "learning_rate": 4.64977059209089e-06, + "loss": 2.957, + "step": 3256 + }, + { + "epoch": 0.71, + "learning_rate": 4.649661350229408e-06, + "loss": 3.3954, + "step": 3257 + }, + { + "epoch": 0.71, + "learning_rate": 4.649552108367927e-06, + "loss": 3.4593, + "step": 3258 + }, + { + "epoch": 0.71, + "learning_rate": 4.649442866506445e-06, + "loss": 3.9459, + "step": 3259 + }, + { + "epoch": 0.71, + "learning_rate": 4.6493336246449644e-06, + "loss": 3.9241, + "step": 3260 + }, + { + "epoch": 0.71, + "learning_rate": 4.6492243827834835e-06, + "loss": 3.7278, + "step": 3261 + }, + { + "epoch": 0.71, + "learning_rate": 4.649115140922002e-06, + "loss": 3.6243, + "step": 3262 + }, + { + "epoch": 0.71, + "learning_rate": 4.64900589906052e-06, + "loss": 2.9302, + "step": 3263 + }, + { + "epoch": 0.71, + "learning_rate": 4.648896657199039e-06, + "loss": 3.4821, + "step": 3264 + }, + { + "epoch": 0.71, + "learning_rate": 4.648787415337558e-06, + "loss": 3.7294, + "step": 3265 + }, + { + "epoch": 0.71, + "learning_rate": 4.648678173476077e-06, + "loss": 3.4926, + "step": 3266 + }, + { + "epoch": 0.71, + "learning_rate": 4.648568931614595e-06, + "loss": 3.1878, + "step": 3267 + }, + { + "epoch": 0.71, + "learning_rate": 4.648459689753113e-06, + "loss": 3.3955, + "step": 3268 + }, + { + "epoch": 0.71, + "learning_rate": 4.648350447891632e-06, + "loss": 3.3433, + "step": 3269 + }, + { + "epoch": 0.71, + "learning_rate": 4.648241206030151e-06, + "loss": 3.663, + "step": 3270 + }, + { + "epoch": 0.71, + "learning_rate": 4.6481319641686704e-06, + "loss": 2.9893, + "step": 3271 + }, + { + "epoch": 0.71, + "learning_rate": 4.648022722307189e-06, + "loss": 3.6545, + "step": 3272 + }, + { + "epoch": 0.71, + "learning_rate": 4.647913480445707e-06, + "loss": 3.6446, + "step": 3273 + }, + { + "epoch": 0.71, + "learning_rate": 4.647804238584226e-06, + "loss": 3.2663, + "step": 3274 + }, + { + "epoch": 0.71, + "learning_rate": 4.647694996722745e-06, + "loss": 3.3047, + "step": 3275 + }, + { + "epoch": 0.71, + "learning_rate": 4.647585754861263e-06, + "loss": 3.6128, + "step": 3276 + }, + { + "epoch": 0.72, + "learning_rate": 4.647476512999782e-06, + "loss": 3.4007, + "step": 3277 + }, + { + "epoch": 0.72, + "learning_rate": 4.6473672711383e-06, + "loss": 3.771, + "step": 3278 + }, + { + "epoch": 0.72, + "learning_rate": 4.647258029276819e-06, + "loss": 3.5369, + "step": 3279 + }, + { + "epoch": 0.72, + "learning_rate": 4.647148787415338e-06, + "loss": 3.175, + "step": 3280 + }, + { + "epoch": 0.72, + "learning_rate": 4.6470395455538566e-06, + "loss": 3.8728, + "step": 3281 + }, + { + "epoch": 0.72, + "learning_rate": 4.646930303692376e-06, + "loss": 3.4072, + "step": 3282 + }, + { + "epoch": 0.72, + "learning_rate": 4.646821061830894e-06, + "loss": 3.7569, + "step": 3283 + }, + { + "epoch": 0.72, + "learning_rate": 4.646711819969413e-06, + "loss": 3.2495, + "step": 3284 + }, + { + "epoch": 0.72, + "learning_rate": 4.646602578107932e-06, + "loss": 3.8007, + "step": 3285 + }, + { + "epoch": 0.72, + "learning_rate": 4.64649333624645e-06, + "loss": 4.0655, + "step": 3286 + }, + { + "epoch": 0.72, + "learning_rate": 4.646384094384968e-06, + "loss": 3.7237, + "step": 3287 + }, + { + "epoch": 0.72, + "learning_rate": 4.646274852523487e-06, + "loss": 3.7561, + "step": 3288 + }, + { + "epoch": 0.72, + "learning_rate": 4.646165610662006e-06, + "loss": 3.3878, + "step": 3289 + }, + { + "epoch": 0.72, + "learning_rate": 4.6460563688005245e-06, + "loss": 3.7952, + "step": 3290 + }, + { + "epoch": 0.72, + "learning_rate": 4.6459471269390435e-06, + "loss": 3.5557, + "step": 3291 + }, + { + "epoch": 0.72, + "learning_rate": 4.645837885077562e-06, + "loss": 3.2924, + "step": 3292 + }, + { + "epoch": 0.72, + "learning_rate": 4.645728643216081e-06, + "loss": 3.1465, + "step": 3293 + }, + { + "epoch": 0.72, + "learning_rate": 4.6456194013546e-06, + "loss": 3.8745, + "step": 3294 + }, + { + "epoch": 0.72, + "learning_rate": 4.645510159493118e-06, + "loss": 2.8733, + "step": 3295 + }, + { + "epoch": 0.72, + "learning_rate": 4.645400917631637e-06, + "loss": 3.2032, + "step": 3296 + }, + { + "epoch": 0.72, + "learning_rate": 4.645291675770155e-06, + "loss": 3.779, + "step": 3297 + }, + { + "epoch": 0.72, + "learning_rate": 4.645182433908674e-06, + "loss": 3.8276, + "step": 3298 + }, + { + "epoch": 0.72, + "learning_rate": 4.645073192047193e-06, + "loss": 3.9011, + "step": 3299 + }, + { + "epoch": 0.72, + "learning_rate": 4.6449639501857114e-06, + "loss": 2.9135, + "step": 3300 + }, + { + "epoch": 0.72, + "learning_rate": 4.6448547083242305e-06, + "loss": 3.7533, + "step": 3301 + }, + { + "epoch": 0.72, + "learning_rate": 4.644745466462749e-06, + "loss": 3.4973, + "step": 3302 + }, + { + "epoch": 0.72, + "learning_rate": 4.644636224601268e-06, + "loss": 3.3691, + "step": 3303 + }, + { + "epoch": 0.72, + "learning_rate": 4.644526982739787e-06, + "loss": 3.7968, + "step": 3304 + }, + { + "epoch": 0.72, + "learning_rate": 4.644417740878305e-06, + "loss": 3.4893, + "step": 3305 + }, + { + "epoch": 0.72, + "learning_rate": 4.644308499016824e-06, + "loss": 3.6452, + "step": 3306 + }, + { + "epoch": 0.72, + "learning_rate": 4.644199257155342e-06, + "loss": 3.7364, + "step": 3307 + }, + { + "epoch": 0.72, + "learning_rate": 4.644090015293861e-06, + "loss": 3.3847, + "step": 3308 + }, + { + "epoch": 0.72, + "learning_rate": 4.643980773432379e-06, + "loss": 3.233, + "step": 3309 + }, + { + "epoch": 0.72, + "learning_rate": 4.643871531570898e-06, + "loss": 3.4151, + "step": 3310 + }, + { + "epoch": 0.72, + "learning_rate": 4.643762289709417e-06, + "loss": 3.4925, + "step": 3311 + }, + { + "epoch": 0.72, + "learning_rate": 4.643653047847936e-06, + "loss": 3.5667, + "step": 3312 + }, + { + "epoch": 0.72, + "learning_rate": 4.643543805986455e-06, + "loss": 3.7821, + "step": 3313 + }, + { + "epoch": 0.72, + "learning_rate": 4.643434564124973e-06, + "loss": 3.043, + "step": 3314 + }, + { + "epoch": 0.72, + "learning_rate": 4.643325322263492e-06, + "loss": 3.1577, + "step": 3315 + }, + { + "epoch": 0.72, + "learning_rate": 4.64321608040201e-06, + "loss": 3.4964, + "step": 3316 + }, + { + "epoch": 0.72, + "learning_rate": 4.643106838540529e-06, + "loss": 3.6875, + "step": 3317 + }, + { + "epoch": 0.72, + "learning_rate": 4.642997596679048e-06, + "loss": 3.205, + "step": 3318 + }, + { + "epoch": 0.72, + "learning_rate": 4.642888354817566e-06, + "loss": 3.4726, + "step": 3319 + }, + { + "epoch": 0.72, + "learning_rate": 4.642779112956085e-06, + "loss": 3.6134, + "step": 3320 + }, + { + "epoch": 0.72, + "learning_rate": 4.6426698710946035e-06, + "loss": 3.5311, + "step": 3321 + }, + { + "epoch": 0.73, + "learning_rate": 4.6425606292331226e-06, + "loss": 3.5654, + "step": 3322 + }, + { + "epoch": 0.73, + "learning_rate": 4.642451387371641e-06, + "loss": 3.4038, + "step": 3323 + }, + { + "epoch": 0.73, + "learning_rate": 4.64234214551016e-06, + "loss": 3.5187, + "step": 3324 + }, + { + "epoch": 0.73, + "learning_rate": 4.642232903648679e-06, + "loss": 3.15, + "step": 3325 + }, + { + "epoch": 0.73, + "learning_rate": 4.642123661787197e-06, + "loss": 3.1726, + "step": 3326 + }, + { + "epoch": 0.73, + "learning_rate": 4.642014419925716e-06, + "loss": 3.5719, + "step": 3327 + }, + { + "epoch": 0.73, + "learning_rate": 4.641905178064234e-06, + "loss": 3.5221, + "step": 3328 + }, + { + "epoch": 0.73, + "learning_rate": 4.641795936202753e-06, + "loss": 3.4466, + "step": 3329 + }, + { + "epoch": 0.73, + "learning_rate": 4.641686694341272e-06, + "loss": 3.6043, + "step": 3330 + }, + { + "epoch": 0.73, + "learning_rate": 4.6415774524797905e-06, + "loss": 3.6882, + "step": 3331 + }, + { + "epoch": 0.73, + "learning_rate": 4.6414682106183095e-06, + "loss": 3.5735, + "step": 3332 + }, + { + "epoch": 0.73, + "learning_rate": 4.641358968756828e-06, + "loss": 3.4904, + "step": 3333 + }, + { + "epoch": 0.73, + "learning_rate": 4.641249726895347e-06, + "loss": 3.3572, + "step": 3334 + }, + { + "epoch": 0.73, + "learning_rate": 4.641140485033866e-06, + "loss": 4.0304, + "step": 3335 + }, + { + "epoch": 0.73, + "learning_rate": 4.641031243172384e-06, + "loss": 3.6801, + "step": 3336 + }, + { + "epoch": 0.73, + "learning_rate": 4.640922001310903e-06, + "loss": 3.3149, + "step": 3337 + }, + { + "epoch": 0.73, + "learning_rate": 4.640812759449421e-06, + "loss": 3.3414, + "step": 3338 + }, + { + "epoch": 0.73, + "learning_rate": 4.64070351758794e-06, + "loss": 3.0107, + "step": 3339 + }, + { + "epoch": 0.73, + "learning_rate": 4.640594275726458e-06, + "loss": 3.7949, + "step": 3340 + }, + { + "epoch": 0.73, + "learning_rate": 4.6404850338649774e-06, + "loss": 2.777, + "step": 3341 + }, + { + "epoch": 0.73, + "learning_rate": 4.640375792003496e-06, + "loss": 3.1855, + "step": 3342 + }, + { + "epoch": 0.73, + "learning_rate": 4.640266550142015e-06, + "loss": 3.2673, + "step": 3343 + }, + { + "epoch": 0.73, + "learning_rate": 4.640157308280534e-06, + "loss": 3.3273, + "step": 3344 + }, + { + "epoch": 0.73, + "learning_rate": 4.640048066419052e-06, + "loss": 3.3771, + "step": 3345 + }, + { + "epoch": 0.73, + "learning_rate": 4.639938824557571e-06, + "loss": 2.9801, + "step": 3346 + }, + { + "epoch": 0.73, + "learning_rate": 4.639829582696089e-06, + "loss": 3.5417, + "step": 3347 + }, + { + "epoch": 0.73, + "learning_rate": 4.639720340834608e-06, + "loss": 4.0042, + "step": 3348 + }, + { + "epoch": 0.73, + "learning_rate": 4.639611098973127e-06, + "loss": 3.5818, + "step": 3349 + }, + { + "epoch": 0.73, + "learning_rate": 4.639501857111645e-06, + "loss": 3.6496, + "step": 3350 + }, + { + "epoch": 0.73, + "learning_rate": 4.639392615250164e-06, + "loss": 3.6599, + "step": 3351 + }, + { + "epoch": 0.73, + "learning_rate": 4.639283373388683e-06, + "loss": 3.4786, + "step": 3352 + }, + { + "epoch": 0.73, + "learning_rate": 4.639174131527202e-06, + "loss": 3.6691, + "step": 3353 + }, + { + "epoch": 0.73, + "learning_rate": 4.639064889665721e-06, + "loss": 3.5542, + "step": 3354 + }, + { + "epoch": 0.73, + "learning_rate": 4.638955647804239e-06, + "loss": 3.5652, + "step": 3355 + }, + { + "epoch": 0.73, + "learning_rate": 4.638846405942757e-06, + "loss": 2.9777, + "step": 3356 + }, + { + "epoch": 0.73, + "learning_rate": 4.638737164081276e-06, + "loss": 3.6774, + "step": 3357 + }, + { + "epoch": 0.73, + "learning_rate": 4.638627922219795e-06, + "loss": 3.214, + "step": 3358 + }, + { + "epoch": 0.73, + "learning_rate": 4.638518680358314e-06, + "loss": 3.2678, + "step": 3359 + }, + { + "epoch": 0.73, + "learning_rate": 4.638409438496832e-06, + "loss": 3.2616, + "step": 3360 + }, + { + "epoch": 0.73, + "learning_rate": 4.6383001966353505e-06, + "loss": 3.7389, + "step": 3361 + }, + { + "epoch": 0.73, + "learning_rate": 4.6381909547738695e-06, + "loss": 3.6742, + "step": 3362 + }, + { + "epoch": 0.73, + "learning_rate": 4.638081712912389e-06, + "loss": 3.3876, + "step": 3363 + }, + { + "epoch": 0.73, + "learning_rate": 4.637972471050908e-06, + "loss": 3.1702, + "step": 3364 + }, + { + "epoch": 0.73, + "learning_rate": 4.637863229189426e-06, + "loss": 3.6138, + "step": 3365 + }, + { + "epoch": 0.73, + "learning_rate": 4.637753987327944e-06, + "loss": 3.8252, + "step": 3366 + }, + { + "epoch": 0.73, + "learning_rate": 4.637644745466463e-06, + "loss": 3.1255, + "step": 3367 + }, + { + "epoch": 0.74, + "learning_rate": 4.637535503604982e-06, + "loss": 3.7268, + "step": 3368 + }, + { + "epoch": 0.74, + "learning_rate": 4.6374262617435e-06, + "loss": 3.2967, + "step": 3369 + }, + { + "epoch": 0.74, + "learning_rate": 4.637317019882019e-06, + "loss": 3.5755, + "step": 3370 + }, + { + "epoch": 0.74, + "learning_rate": 4.6372077780205375e-06, + "loss": 3.4861, + "step": 3371 + }, + { + "epoch": 0.74, + "learning_rate": 4.6370985361590565e-06, + "loss": 3.7372, + "step": 3372 + }, + { + "epoch": 0.74, + "learning_rate": 4.6369892942975755e-06, + "loss": 3.2923, + "step": 3373 + }, + { + "epoch": 0.74, + "learning_rate": 4.636880052436094e-06, + "loss": 3.6595, + "step": 3374 + }, + { + "epoch": 0.74, + "learning_rate": 4.636770810574612e-06, + "loss": 3.3895, + "step": 3375 + }, + { + "epoch": 0.74, + "learning_rate": 4.636661568713131e-06, + "loss": 3.435, + "step": 3376 + }, + { + "epoch": 0.74, + "learning_rate": 4.63655232685165e-06, + "loss": 3.5296, + "step": 3377 + }, + { + "epoch": 0.74, + "learning_rate": 4.636443084990169e-06, + "loss": 3.5385, + "step": 3378 + }, + { + "epoch": 0.74, + "learning_rate": 4.636333843128687e-06, + "loss": 3.9074, + "step": 3379 + }, + { + "epoch": 0.74, + "learning_rate": 4.636224601267205e-06, + "loss": 3.2308, + "step": 3380 + }, + { + "epoch": 0.74, + "learning_rate": 4.6361153594057244e-06, + "loss": 3.3173, + "step": 3381 + }, + { + "epoch": 0.74, + "learning_rate": 4.6360061175442435e-06, + "loss": 3.2754, + "step": 3382 + }, + { + "epoch": 0.74, + "learning_rate": 4.6358968756827625e-06, + "loss": 3.3345, + "step": 3383 + }, + { + "epoch": 0.74, + "learning_rate": 4.635787633821281e-06, + "loss": 3.653, + "step": 3384 + }, + { + "epoch": 0.74, + "learning_rate": 4.635678391959799e-06, + "loss": 3.6304, + "step": 3385 + }, + { + "epoch": 0.74, + "learning_rate": 4.635569150098318e-06, + "loss": 3.6733, + "step": 3386 + }, + { + "epoch": 0.74, + "learning_rate": 4.635459908236837e-06, + "loss": 2.9378, + "step": 3387 + }, + { + "epoch": 0.74, + "learning_rate": 4.635350666375356e-06, + "loss": 2.7412, + "step": 3388 + }, + { + "epoch": 0.74, + "learning_rate": 4.635241424513874e-06, + "loss": 3.2713, + "step": 3389 + }, + { + "epoch": 0.74, + "learning_rate": 4.635132182652392e-06, + "loss": 3.3664, + "step": 3390 + }, + { + "epoch": 0.74, + "learning_rate": 4.635022940790911e-06, + "loss": 2.8995, + "step": 3391 + }, + { + "epoch": 0.74, + "learning_rate": 4.63491369892943e-06, + "loss": 4.0724, + "step": 3392 + }, + { + "epoch": 0.74, + "learning_rate": 4.634804457067949e-06, + "loss": 3.6444, + "step": 3393 + }, + { + "epoch": 0.74, + "learning_rate": 4.634695215206468e-06, + "loss": 3.6606, + "step": 3394 + }, + { + "epoch": 0.74, + "learning_rate": 4.634585973344986e-06, + "loss": 3.7447, + "step": 3395 + }, + { + "epoch": 0.74, + "learning_rate": 4.634476731483505e-06, + "loss": 3.5047, + "step": 3396 + }, + { + "epoch": 0.74, + "learning_rate": 4.634367489622024e-06, + "loss": 3.3571, + "step": 3397 + }, + { + "epoch": 0.74, + "learning_rate": 4.634258247760542e-06, + "loss": 2.9651, + "step": 3398 + }, + { + "epoch": 0.74, + "learning_rate": 4.63414900589906e-06, + "loss": 3.8484, + "step": 3399 + }, + { + "epoch": 0.74, + "learning_rate": 4.634039764037579e-06, + "loss": 3.5054, + "step": 3400 + }, + { + "epoch": 0.74, + "learning_rate": 4.633930522176098e-06, + "loss": 3.3183, + "step": 3401 + }, + { + "epoch": 0.74, + "learning_rate": 4.633821280314617e-06, + "loss": 3.7294, + "step": 3402 + }, + { + "epoch": 0.74, + "learning_rate": 4.6337120384531356e-06, + "loss": 2.7237, + "step": 3403 + }, + { + "epoch": 0.74, + "learning_rate": 4.633602796591654e-06, + "loss": 3.3984, + "step": 3404 + }, + { + "epoch": 0.74, + "learning_rate": 4.633493554730173e-06, + "loss": 2.9982, + "step": 3405 + }, + { + "epoch": 0.74, + "learning_rate": 4.633384312868692e-06, + "loss": 3.5599, + "step": 3406 + }, + { + "epoch": 0.74, + "learning_rate": 4.633275071007211e-06, + "loss": 3.9823, + "step": 3407 + }, + { + "epoch": 0.74, + "learning_rate": 4.633165829145729e-06, + "loss": 3.7302, + "step": 3408 + }, + { + "epoch": 0.74, + "learning_rate": 4.633056587284247e-06, + "loss": 3.3026, + "step": 3409 + }, + { + "epoch": 0.74, + "learning_rate": 4.632947345422766e-06, + "loss": 3.2142, + "step": 3410 + }, + { + "epoch": 0.74, + "learning_rate": 4.632838103561285e-06, + "loss": 3.1714, + "step": 3411 + }, + { + "epoch": 0.74, + "learning_rate": 4.632728861699804e-06, + "loss": 3.0646, + "step": 3412 + }, + { + "epoch": 0.74, + "learning_rate": 4.6326196198383225e-06, + "loss": 3.8468, + "step": 3413 + }, + { + "epoch": 0.75, + "learning_rate": 4.632510377976841e-06, + "loss": 3.2172, + "step": 3414 + }, + { + "epoch": 0.75, + "learning_rate": 4.63240113611536e-06, + "loss": 3.5694, + "step": 3415 + }, + { + "epoch": 0.75, + "learning_rate": 4.632291894253879e-06, + "loss": 3.5674, + "step": 3416 + }, + { + "epoch": 0.75, + "learning_rate": 4.632182652392398e-06, + "loss": 3.2172, + "step": 3417 + }, + { + "epoch": 0.75, + "learning_rate": 4.632073410530916e-06, + "loss": 3.758, + "step": 3418 + }, + { + "epoch": 0.75, + "learning_rate": 4.631964168669434e-06, + "loss": 3.4737, + "step": 3419 + }, + { + "epoch": 0.75, + "learning_rate": 4.631854926807953e-06, + "loss": 3.7221, + "step": 3420 + }, + { + "epoch": 0.75, + "learning_rate": 4.631745684946472e-06, + "loss": 3.575, + "step": 3421 + }, + { + "epoch": 0.75, + "learning_rate": 4.6316364430849904e-06, + "loss": 3.6735, + "step": 3422 + }, + { + "epoch": 0.75, + "learning_rate": 4.6315272012235095e-06, + "loss": 3.1044, + "step": 3423 + }, + { + "epoch": 0.75, + "learning_rate": 4.631417959362028e-06, + "loss": 3.1763, + "step": 3424 + }, + { + "epoch": 0.75, + "learning_rate": 4.631308717500547e-06, + "loss": 3.5466, + "step": 3425 + }, + { + "epoch": 0.75, + "learning_rate": 4.631199475639066e-06, + "loss": 3.7014, + "step": 3426 + }, + { + "epoch": 0.75, + "learning_rate": 4.631090233777584e-06, + "loss": 3.4867, + "step": 3427 + }, + { + "epoch": 0.75, + "learning_rate": 4.630980991916102e-06, + "loss": 3.7468, + "step": 3428 + }, + { + "epoch": 0.75, + "learning_rate": 4.630871750054621e-06, + "loss": 3.4666, + "step": 3429 + }, + { + "epoch": 0.75, + "learning_rate": 4.63076250819314e-06, + "loss": 3.7655, + "step": 3430 + }, + { + "epoch": 0.75, + "learning_rate": 4.630653266331659e-06, + "loss": 3.4526, + "step": 3431 + }, + { + "epoch": 0.75, + "learning_rate": 4.630544024470177e-06, + "loss": 3.7237, + "step": 3432 + }, + { + "epoch": 0.75, + "learning_rate": 4.630434782608696e-06, + "loss": 3.6339, + "step": 3433 + }, + { + "epoch": 0.75, + "learning_rate": 4.630325540747215e-06, + "loss": 2.7109, + "step": 3434 + }, + { + "epoch": 0.75, + "learning_rate": 4.630216298885734e-06, + "loss": 3.7846, + "step": 3435 + }, + { + "epoch": 0.75, + "learning_rate": 4.630107057024253e-06, + "loss": 4.1169, + "step": 3436 + }, + { + "epoch": 0.75, + "learning_rate": 4.629997815162771e-06, + "loss": 2.899, + "step": 3437 + }, + { + "epoch": 0.75, + "learning_rate": 4.629888573301289e-06, + "loss": 3.911, + "step": 3438 + }, + { + "epoch": 0.75, + "learning_rate": 4.629779331439808e-06, + "loss": 3.8507, + "step": 3439 + }, + { + "epoch": 0.75, + "learning_rate": 4.629670089578327e-06, + "loss": 2.8039, + "step": 3440 + }, + { + "epoch": 0.75, + "learning_rate": 4.629560847716845e-06, + "loss": 3.4984, + "step": 3441 + }, + { + "epoch": 0.75, + "learning_rate": 4.629451605855364e-06, + "loss": 3.8674, + "step": 3442 + }, + { + "epoch": 0.75, + "learning_rate": 4.6293423639938825e-06, + "loss": 3.4897, + "step": 3443 + }, + { + "epoch": 0.75, + "learning_rate": 4.629233122132402e-06, + "loss": 3.502, + "step": 3444 + }, + { + "epoch": 0.75, + "learning_rate": 4.629123880270921e-06, + "loss": 3.5029, + "step": 3445 + }, + { + "epoch": 0.75, + "learning_rate": 4.629014638409439e-06, + "loss": 3.5822, + "step": 3446 + }, + { + "epoch": 0.75, + "learning_rate": 4.628905396547958e-06, + "loss": 3.4204, + "step": 3447 + }, + { + "epoch": 0.75, + "learning_rate": 4.628796154686476e-06, + "loss": 3.4345, + "step": 3448 + }, + { + "epoch": 0.75, + "learning_rate": 4.628686912824995e-06, + "loss": 3.6124, + "step": 3449 + }, + { + "epoch": 0.75, + "learning_rate": 4.628577670963514e-06, + "loss": 3.2277, + "step": 3450 + }, + { + "epoch": 0.75, + "learning_rate": 4.628468429102032e-06, + "loss": 3.8055, + "step": 3451 + }, + { + "epoch": 0.75, + "learning_rate": 4.628359187240551e-06, + "loss": 3.709, + "step": 3452 + }, + { + "epoch": 0.75, + "learning_rate": 4.6282499453790695e-06, + "loss": 3.6714, + "step": 3453 + }, + { + "epoch": 0.75, + "learning_rate": 4.6281407035175885e-06, + "loss": 3.0376, + "step": 3454 + }, + { + "epoch": 0.75, + "learning_rate": 4.628031461656107e-06, + "loss": 2.8184, + "step": 3455 + }, + { + "epoch": 0.75, + "learning_rate": 4.627922219794626e-06, + "loss": 3.4068, + "step": 3456 + }, + { + "epoch": 0.75, + "learning_rate": 4.627812977933144e-06, + "loss": 3.2343, + "step": 3457 + }, + { + "epoch": 0.75, + "learning_rate": 4.627703736071663e-06, + "loss": 3.1949, + "step": 3458 + }, + { + "epoch": 0.75, + "learning_rate": 4.627594494210182e-06, + "loss": 3.8699, + "step": 3459 + }, + { + "epoch": 0.76, + "learning_rate": 4.6274852523487e-06, + "loss": 3.21, + "step": 3460 + }, + { + "epoch": 0.76, + "learning_rate": 4.627376010487219e-06, + "loss": 3.156, + "step": 3461 + }, + { + "epoch": 0.76, + "learning_rate": 4.627266768625737e-06, + "loss": 2.8513, + "step": 3462 + }, + { + "epoch": 0.76, + "learning_rate": 4.6271575267642565e-06, + "loss": 3.5088, + "step": 3463 + }, + { + "epoch": 0.76, + "learning_rate": 4.6270482849027755e-06, + "loss": 3.1771, + "step": 3464 + }, + { + "epoch": 0.76, + "learning_rate": 4.626939043041294e-06, + "loss": 3.47, + "step": 3465 + }, + { + "epoch": 0.76, + "learning_rate": 4.626829801179813e-06, + "loss": 3.6598, + "step": 3466 + }, + { + "epoch": 0.76, + "learning_rate": 4.626720559318331e-06, + "loss": 3.2132, + "step": 3467 + }, + { + "epoch": 0.76, + "learning_rate": 4.62661131745685e-06, + "loss": 3.0545, + "step": 3468 + }, + { + "epoch": 0.76, + "learning_rate": 4.626502075595369e-06, + "loss": 3.8674, + "step": 3469 + }, + { + "epoch": 0.76, + "learning_rate": 4.626392833733887e-06, + "loss": 3.3725, + "step": 3470 + }, + { + "epoch": 0.76, + "learning_rate": 4.626283591872406e-06, + "loss": 3.2109, + "step": 3471 + }, + { + "epoch": 0.76, + "learning_rate": 4.626174350010924e-06, + "loss": 3.5056, + "step": 3472 + }, + { + "epoch": 0.76, + "learning_rate": 4.626065108149443e-06, + "loss": 3.4874, + "step": 3473 + }, + { + "epoch": 0.76, + "learning_rate": 4.625955866287962e-06, + "loss": 3.317, + "step": 3474 + }, + { + "epoch": 0.76, + "learning_rate": 4.625846624426481e-06, + "loss": 3.6167, + "step": 3475 + }, + { + "epoch": 0.76, + "learning_rate": 4.625737382565e-06, + "loss": 3.5776, + "step": 3476 + }, + { + "epoch": 0.76, + "learning_rate": 4.625628140703518e-06, + "loss": 3.6177, + "step": 3477 + }, + { + "epoch": 0.76, + "learning_rate": 4.625518898842037e-06, + "loss": 2.9276, + "step": 3478 + }, + { + "epoch": 0.76, + "learning_rate": 4.625409656980555e-06, + "loss": 3.3765, + "step": 3479 + }, + { + "epoch": 0.76, + "learning_rate": 4.625300415119074e-06, + "loss": 3.9035, + "step": 3480 + }, + { + "epoch": 0.76, + "learning_rate": 4.625191173257592e-06, + "loss": 3.2732, + "step": 3481 + }, + { + "epoch": 0.76, + "learning_rate": 4.625081931396111e-06, + "loss": 3.8837, + "step": 3482 + }, + { + "epoch": 0.76, + "learning_rate": 4.62497268953463e-06, + "loss": 3.1668, + "step": 3483 + }, + { + "epoch": 0.76, + "learning_rate": 4.6248634476731486e-06, + "loss": 3.4245, + "step": 3484 + }, + { + "epoch": 0.76, + "learning_rate": 4.624754205811668e-06, + "loss": 3.0394, + "step": 3485 + }, + { + "epoch": 0.76, + "learning_rate": 4.624644963950186e-06, + "loss": 3.3608, + "step": 3486 + }, + { + "epoch": 0.76, + "learning_rate": 4.624535722088705e-06, + "loss": 3.7416, + "step": 3487 + }, + { + "epoch": 0.76, + "learning_rate": 4.624426480227223e-06, + "loss": 3.2106, + "step": 3488 + }, + { + "epoch": 0.76, + "learning_rate": 4.624317238365742e-06, + "loss": 3.3514, + "step": 3489 + }, + { + "epoch": 0.76, + "learning_rate": 4.624207996504261e-06, + "loss": 2.95, + "step": 3490 + }, + { + "epoch": 0.76, + "learning_rate": 4.624098754642779e-06, + "loss": 3.9152, + "step": 3491 + }, + { + "epoch": 0.76, + "learning_rate": 4.623989512781298e-06, + "loss": 3.2872, + "step": 3492 + }, + { + "epoch": 0.76, + "learning_rate": 4.6238802709198165e-06, + "loss": 3.4584, + "step": 3493 + }, + { + "epoch": 0.76, + "learning_rate": 4.6237710290583355e-06, + "loss": 3.4244, + "step": 3494 + }, + { + "epoch": 0.76, + "learning_rate": 4.6236617871968545e-06, + "loss": 3.2314, + "step": 3495 + }, + { + "epoch": 0.76, + "learning_rate": 4.623552545335373e-06, + "loss": 3.2822, + "step": 3496 + }, + { + "epoch": 0.76, + "learning_rate": 4.623443303473892e-06, + "loss": 3.856, + "step": 3497 + }, + { + "epoch": 0.76, + "learning_rate": 4.62333406161241e-06, + "loss": 3.8009, + "step": 3498 + }, + { + "epoch": 0.76, + "learning_rate": 4.623224819750929e-06, + "loss": 3.3717, + "step": 3499 + }, + { + "epoch": 0.76, + "learning_rate": 4.623115577889448e-06, + "loss": 3.1408, + "step": 3500 + }, + { + "epoch": 0.76, + "learning_rate": 4.623006336027966e-06, + "loss": 3.6177, + "step": 3501 + }, + { + "epoch": 0.76, + "learning_rate": 4.622897094166485e-06, + "loss": 3.5471, + "step": 3502 + }, + { + "epoch": 0.76, + "learning_rate": 4.6227878523050034e-06, + "loss": 3.2638, + "step": 3503 + }, + { + "epoch": 0.76, + "learning_rate": 4.6226786104435225e-06, + "loss": 3.1382, + "step": 3504 + }, + { + "epoch": 0.76, + "learning_rate": 4.6225693685820415e-06, + "loss": 3.4627, + "step": 3505 + }, + { + "epoch": 0.77, + "learning_rate": 4.62246012672056e-06, + "loss": 3.2408, + "step": 3506 + }, + { + "epoch": 0.77, + "learning_rate": 4.622350884859078e-06, + "loss": 3.7246, + "step": 3507 + }, + { + "epoch": 0.77, + "learning_rate": 4.622241642997597e-06, + "loss": 3.0642, + "step": 3508 + }, + { + "epoch": 0.77, + "learning_rate": 4.622132401136116e-06, + "loss": 3.6319, + "step": 3509 + }, + { + "epoch": 0.77, + "learning_rate": 4.622023159274634e-06, + "loss": 2.9822, + "step": 3510 + }, + { + "epoch": 0.77, + "learning_rate": 4.621913917413153e-06, + "loss": 3.4569, + "step": 3511 + }, + { + "epoch": 0.77, + "learning_rate": 4.621804675551671e-06, + "loss": 3.4323, + "step": 3512 + }, + { + "epoch": 0.77, + "learning_rate": 4.62169543369019e-06, + "loss": 3.6233, + "step": 3513 + }, + { + "epoch": 0.77, + "learning_rate": 4.6215861918287094e-06, + "loss": 3.3669, + "step": 3514 + }, + { + "epoch": 0.77, + "learning_rate": 4.621476949967228e-06, + "loss": 3.5438, + "step": 3515 + }, + { + "epoch": 0.77, + "learning_rate": 4.621367708105747e-06, + "loss": 3.7868, + "step": 3516 + }, + { + "epoch": 0.77, + "learning_rate": 4.621258466244265e-06, + "loss": 3.3127, + "step": 3517 + }, + { + "epoch": 0.77, + "learning_rate": 4.621149224382784e-06, + "loss": 3.0499, + "step": 3518 + }, + { + "epoch": 0.77, + "learning_rate": 4.621039982521303e-06, + "loss": 3.8953, + "step": 3519 + }, + { + "epoch": 0.77, + "learning_rate": 4.620930740659821e-06, + "loss": 3.5297, + "step": 3520 + }, + { + "epoch": 0.77, + "learning_rate": 4.620821498798339e-06, + "loss": 3.1561, + "step": 3521 + }, + { + "epoch": 0.77, + "learning_rate": 4.620712256936858e-06, + "loss": 3.3696, + "step": 3522 + }, + { + "epoch": 0.77, + "learning_rate": 4.620603015075377e-06, + "loss": 3.4371, + "step": 3523 + }, + { + "epoch": 0.77, + "learning_rate": 4.620493773213896e-06, + "loss": 3.3321, + "step": 3524 + }, + { + "epoch": 0.77, + "learning_rate": 4.6203845313524146e-06, + "loss": 3.6245, + "step": 3525 + }, + { + "epoch": 0.77, + "learning_rate": 4.620275289490933e-06, + "loss": 3.4444, + "step": 3526 + }, + { + "epoch": 0.77, + "learning_rate": 4.620166047629452e-06, + "loss": 3.7557, + "step": 3527 + }, + { + "epoch": 0.77, + "learning_rate": 4.620056805767971e-06, + "loss": 3.3957, + "step": 3528 + }, + { + "epoch": 0.77, + "learning_rate": 4.61994756390649e-06, + "loss": 3.5334, + "step": 3529 + }, + { + "epoch": 0.77, + "learning_rate": 4.619838322045008e-06, + "loss": 3.0565, + "step": 3530 + }, + { + "epoch": 0.77, + "learning_rate": 4.619729080183526e-06, + "loss": 3.5321, + "step": 3531 + }, + { + "epoch": 0.77, + "learning_rate": 4.619619838322045e-06, + "loss": 3.5745, + "step": 3532 + }, + { + "epoch": 0.77, + "learning_rate": 4.619510596460564e-06, + "loss": 3.4801, + "step": 3533 + }, + { + "epoch": 0.77, + "learning_rate": 4.6194013545990825e-06, + "loss": 3.4425, + "step": 3534 + }, + { + "epoch": 0.77, + "learning_rate": 4.6192921127376015e-06, + "loss": 3.3485, + "step": 3535 + }, + { + "epoch": 0.77, + "learning_rate": 4.61918287087612e-06, + "loss": 3.3824, + "step": 3536 + }, + { + "epoch": 0.77, + "learning_rate": 4.619073629014639e-06, + "loss": 2.7226, + "step": 3537 + }, + { + "epoch": 0.77, + "learning_rate": 4.618964387153158e-06, + "loss": 3.8537, + "step": 3538 + }, + { + "epoch": 0.77, + "learning_rate": 4.618855145291676e-06, + "loss": 3.2421, + "step": 3539 + }, + { + "epoch": 0.77, + "learning_rate": 4.618745903430194e-06, + "loss": 3.3551, + "step": 3540 + }, + { + "epoch": 0.77, + "learning_rate": 4.618636661568713e-06, + "loss": 2.9956, + "step": 3541 + }, + { + "epoch": 0.77, + "learning_rate": 4.618527419707232e-06, + "loss": 3.6706, + "step": 3542 + }, + { + "epoch": 0.77, + "learning_rate": 4.618418177845751e-06, + "loss": 3.3279, + "step": 3543 + }, + { + "epoch": 0.77, + "learning_rate": 4.6183089359842695e-06, + "loss": 3.2901, + "step": 3544 + }, + { + "epoch": 0.77, + "learning_rate": 4.618199694122788e-06, + "loss": 3.5528, + "step": 3545 + }, + { + "epoch": 0.77, + "learning_rate": 4.618090452261307e-06, + "loss": 3.4693, + "step": 3546 + }, + { + "epoch": 0.77, + "learning_rate": 4.617981210399826e-06, + "loss": 3.4318, + "step": 3547 + }, + { + "epoch": 0.77, + "learning_rate": 4.617871968538345e-06, + "loss": 3.1066, + "step": 3548 + }, + { + "epoch": 0.77, + "learning_rate": 4.617762726676863e-06, + "loss": 3.8125, + "step": 3549 + }, + { + "epoch": 0.77, + "learning_rate": 4.617653484815381e-06, + "loss": 3.7144, + "step": 3550 + }, + { + "epoch": 0.77, + "learning_rate": 4.6175442429539e-06, + "loss": 3.464, + "step": 3551 + }, + { + "epoch": 0.78, + "learning_rate": 4.617435001092419e-06, + "loss": 3.1939, + "step": 3552 + }, + { + "epoch": 0.78, + "learning_rate": 4.617325759230938e-06, + "loss": 3.4592, + "step": 3553 + }, + { + "epoch": 0.78, + "learning_rate": 4.617216517369456e-06, + "loss": 3.0468, + "step": 3554 + }, + { + "epoch": 0.78, + "learning_rate": 4.617107275507975e-06, + "loss": 3.8435, + "step": 3555 + }, + { + "epoch": 0.78, + "learning_rate": 4.616998033646494e-06, + "loss": 3.4622, + "step": 3556 + }, + { + "epoch": 0.78, + "learning_rate": 4.616888791785013e-06, + "loss": 3.3428, + "step": 3557 + }, + { + "epoch": 0.78, + "learning_rate": 4.616779549923532e-06, + "loss": 3.2817, + "step": 3558 + }, + { + "epoch": 0.78, + "learning_rate": 4.61667030806205e-06, + "loss": 3.6762, + "step": 3559 + }, + { + "epoch": 0.78, + "learning_rate": 4.616561066200568e-06, + "loss": 3.0832, + "step": 3560 + }, + { + "epoch": 0.78, + "learning_rate": 4.616451824339087e-06, + "loss": 2.8097, + "step": 3561 + }, + { + "epoch": 0.78, + "learning_rate": 4.616342582477606e-06, + "loss": 3.1911, + "step": 3562 + }, + { + "epoch": 0.78, + "learning_rate": 4.616233340616124e-06, + "loss": 2.9492, + "step": 3563 + }, + { + "epoch": 0.78, + "learning_rate": 4.616124098754643e-06, + "loss": 3.6844, + "step": 3564 + }, + { + "epoch": 0.78, + "learning_rate": 4.6160148568931616e-06, + "loss": 3.9489, + "step": 3565 + }, + { + "epoch": 0.78, + "learning_rate": 4.615905615031681e-06, + "loss": 3.2262, + "step": 3566 + }, + { + "epoch": 0.78, + "learning_rate": 4.6157963731702e-06, + "loss": 3.1179, + "step": 3567 + }, + { + "epoch": 0.78, + "learning_rate": 4.615687131308718e-06, + "loss": 3.73, + "step": 3568 + }, + { + "epoch": 0.78, + "learning_rate": 4.615577889447236e-06, + "loss": 3.0373, + "step": 3569 + }, + { + "epoch": 0.78, + "learning_rate": 4.615468647585755e-06, + "loss": 3.5462, + "step": 3570 + }, + { + "epoch": 0.78, + "learning_rate": 4.615359405724274e-06, + "loss": 3.3726, + "step": 3571 + }, + { + "epoch": 0.78, + "learning_rate": 4.615250163862793e-06, + "loss": 3.8272, + "step": 3572 + }, + { + "epoch": 0.78, + "learning_rate": 4.615140922001311e-06, + "loss": 3.4684, + "step": 3573 + }, + { + "epoch": 0.78, + "learning_rate": 4.6150316801398295e-06, + "loss": 3.3475, + "step": 3574 + }, + { + "epoch": 0.78, + "learning_rate": 4.6149224382783485e-06, + "loss": 3.3698, + "step": 3575 + }, + { + "epoch": 0.78, + "learning_rate": 4.6148131964168675e-06, + "loss": 3.4657, + "step": 3576 + }, + { + "epoch": 0.78, + "learning_rate": 4.614703954555387e-06, + "loss": 3.3768, + "step": 3577 + }, + { + "epoch": 0.78, + "learning_rate": 4.614594712693905e-06, + "loss": 3.8707, + "step": 3578 + }, + { + "epoch": 0.78, + "learning_rate": 4.614485470832423e-06, + "loss": 3.5396, + "step": 3579 + }, + { + "epoch": 0.78, + "learning_rate": 4.614376228970942e-06, + "loss": 3.1218, + "step": 3580 + }, + { + "epoch": 0.78, + "learning_rate": 4.614266987109461e-06, + "loss": 3.5747, + "step": 3581 + }, + { + "epoch": 0.78, + "learning_rate": 4.61415774524798e-06, + "loss": 3.5689, + "step": 3582 + }, + { + "epoch": 0.78, + "learning_rate": 4.614048503386498e-06, + "loss": 3.4018, + "step": 3583 + }, + { + "epoch": 0.78, + "learning_rate": 4.6139392615250164e-06, + "loss": 3.2647, + "step": 3584 + }, + { + "epoch": 0.78, + "learning_rate": 4.6138300196635355e-06, + "loss": 3.8298, + "step": 3585 + }, + { + "epoch": 0.78, + "learning_rate": 4.6137207778020545e-06, + "loss": 3.3743, + "step": 3586 + }, + { + "epoch": 0.78, + "learning_rate": 4.613611535940573e-06, + "loss": 3.5728, + "step": 3587 + }, + { + "epoch": 0.78, + "learning_rate": 4.613502294079092e-06, + "loss": 3.4524, + "step": 3588 + }, + { + "epoch": 0.78, + "learning_rate": 4.61339305221761e-06, + "loss": 4.1155, + "step": 3589 + }, + { + "epoch": 0.78, + "learning_rate": 4.613283810356129e-06, + "loss": 3.7149, + "step": 3590 + }, + { + "epoch": 0.78, + "learning_rate": 4.613174568494648e-06, + "loss": 3.7133, + "step": 3591 + }, + { + "epoch": 0.78, + "learning_rate": 4.613065326633166e-06, + "loss": 3.4679, + "step": 3592 + }, + { + "epoch": 0.78, + "learning_rate": 4.612956084771684e-06, + "loss": 3.7868, + "step": 3593 + }, + { + "epoch": 0.78, + "learning_rate": 4.612846842910203e-06, + "loss": 3.5349, + "step": 3594 + }, + { + "epoch": 0.78, + "learning_rate": 4.6127376010487224e-06, + "loss": 3.2605, + "step": 3595 + }, + { + "epoch": 0.78, + "learning_rate": 4.6126283591872415e-06, + "loss": 3.0824, + "step": 3596 + }, + { + "epoch": 0.79, + "learning_rate": 4.61251911732576e-06, + "loss": 3.315, + "step": 3597 + }, + { + "epoch": 0.79, + "learning_rate": 4.612409875464278e-06, + "loss": 3.1604, + "step": 3598 + }, + { + "epoch": 0.79, + "learning_rate": 4.612300633602797e-06, + "loss": 3.3246, + "step": 3599 + }, + { + "epoch": 0.79, + "learning_rate": 4.612191391741316e-06, + "loss": 3.485, + "step": 3600 + }, + { + "epoch": 0.79, + "learning_rate": 4.612082149879835e-06, + "loss": 2.9766, + "step": 3601 + }, + { + "epoch": 0.79, + "learning_rate": 4.611972908018353e-06, + "loss": 3.4838, + "step": 3602 + }, + { + "epoch": 0.79, + "learning_rate": 4.611863666156871e-06, + "loss": 3.62, + "step": 3603 + }, + { + "epoch": 0.79, + "learning_rate": 4.61175442429539e-06, + "loss": 3.5854, + "step": 3604 + }, + { + "epoch": 0.79, + "learning_rate": 4.611645182433909e-06, + "loss": 3.4171, + "step": 3605 + }, + { + "epoch": 0.79, + "learning_rate": 4.6115359405724276e-06, + "loss": 3.3462, + "step": 3606 + }, + { + "epoch": 0.79, + "learning_rate": 4.611426698710947e-06, + "loss": 3.4792, + "step": 3607 + }, + { + "epoch": 0.79, + "learning_rate": 4.611317456849465e-06, + "loss": 3.0401, + "step": 3608 + }, + { + "epoch": 0.79, + "learning_rate": 4.611208214987984e-06, + "loss": 3.0741, + "step": 3609 + }, + { + "epoch": 0.79, + "learning_rate": 4.611098973126503e-06, + "loss": 3.4542, + "step": 3610 + }, + { + "epoch": 0.79, + "learning_rate": 4.610989731265021e-06, + "loss": 3.1911, + "step": 3611 + }, + { + "epoch": 0.79, + "learning_rate": 4.61088048940354e-06, + "loss": 3.7333, + "step": 3612 + }, + { + "epoch": 0.79, + "learning_rate": 4.610771247542058e-06, + "loss": 3.3178, + "step": 3613 + }, + { + "epoch": 0.79, + "learning_rate": 4.610662005680577e-06, + "loss": 3.333, + "step": 3614 + }, + { + "epoch": 0.79, + "learning_rate": 4.610552763819096e-06, + "loss": 3.2768, + "step": 3615 + }, + { + "epoch": 0.79, + "learning_rate": 4.6104435219576145e-06, + "loss": 2.8282, + "step": 3616 + }, + { + "epoch": 0.79, + "learning_rate": 4.6103342800961336e-06, + "loss": 3.9263, + "step": 3617 + }, + { + "epoch": 0.79, + "learning_rate": 4.610225038234652e-06, + "loss": 3.4194, + "step": 3618 + }, + { + "epoch": 0.79, + "learning_rate": 4.610115796373171e-06, + "loss": 3.5109, + "step": 3619 + }, + { + "epoch": 0.79, + "learning_rate": 4.610006554511689e-06, + "loss": 3.3875, + "step": 3620 + }, + { + "epoch": 0.79, + "learning_rate": 4.609897312650208e-06, + "loss": 3.3399, + "step": 3621 + }, + { + "epoch": 0.79, + "learning_rate": 4.609788070788726e-06, + "loss": 3.2276, + "step": 3622 + }, + { + "epoch": 0.79, + "learning_rate": 4.609678828927245e-06, + "loss": 3.4343, + "step": 3623 + }, + { + "epoch": 0.79, + "learning_rate": 4.609569587065764e-06, + "loss": 3.5777, + "step": 3624 + }, + { + "epoch": 0.79, + "learning_rate": 4.6094603452042824e-06, + "loss": 3.212, + "step": 3625 + }, + { + "epoch": 0.79, + "learning_rate": 4.6093511033428015e-06, + "loss": 3.3398, + "step": 3626 + }, + { + "epoch": 0.79, + "learning_rate": 4.60924186148132e-06, + "loss": 3.3931, + "step": 3627 + }, + { + "epoch": 0.79, + "learning_rate": 4.609132619619839e-06, + "loss": 3.5415, + "step": 3628 + }, + { + "epoch": 0.79, + "learning_rate": 4.609023377758358e-06, + "loss": 3.9164, + "step": 3629 + }, + { + "epoch": 0.79, + "learning_rate": 4.608914135896876e-06, + "loss": 3.5989, + "step": 3630 + }, + { + "epoch": 0.79, + "learning_rate": 4.608804894035395e-06, + "loss": 3.4404, + "step": 3631 + }, + { + "epoch": 0.79, + "learning_rate": 4.608695652173913e-06, + "loss": 3.5961, + "step": 3632 + }, + { + "epoch": 0.79, + "learning_rate": 4.608586410312432e-06, + "loss": 3.2156, + "step": 3633 + }, + { + "epoch": 0.79, + "learning_rate": 4.608477168450951e-06, + "loss": 3.3624, + "step": 3634 + }, + { + "epoch": 0.79, + "learning_rate": 4.608367926589469e-06, + "loss": 3.2831, + "step": 3635 + }, + { + "epoch": 0.79, + "learning_rate": 4.6082586847279884e-06, + "loss": 3.1681, + "step": 3636 + }, + { + "epoch": 0.79, + "learning_rate": 4.608149442866507e-06, + "loss": 3.3749, + "step": 3637 + }, + { + "epoch": 0.79, + "learning_rate": 4.608040201005026e-06, + "loss": 3.3655, + "step": 3638 + }, + { + "epoch": 0.79, + "learning_rate": 4.607930959143544e-06, + "loss": 3.8232, + "step": 3639 + }, + { + "epoch": 0.79, + "learning_rate": 4.607821717282063e-06, + "loss": 3.4292, + "step": 3640 + }, + { + "epoch": 0.79, + "learning_rate": 4.607712475420582e-06, + "loss": 2.9689, + "step": 3641 + }, + { + "epoch": 0.79, + "learning_rate": 4.6076032335591e-06, + "loss": 2.8609, + "step": 3642 + }, + { + "epoch": 0.8, + "learning_rate": 4.607493991697619e-06, + "loss": 3.2209, + "step": 3643 + }, + { + "epoch": 0.8, + "learning_rate": 4.607384749836137e-06, + "loss": 3.8704, + "step": 3644 + }, + { + "epoch": 0.8, + "learning_rate": 4.607275507974656e-06, + "loss": 3.4589, + "step": 3645 + }, + { + "epoch": 0.8, + "learning_rate": 4.6071662661131745e-06, + "loss": 3.5554, + "step": 3646 + }, + { + "epoch": 0.8, + "learning_rate": 4.607057024251694e-06, + "loss": 3.4055, + "step": 3647 + }, + { + "epoch": 0.8, + "learning_rate": 4.606947782390213e-06, + "loss": 3.1427, + "step": 3648 + }, + { + "epoch": 0.8, + "learning_rate": 4.606838540528731e-06, + "loss": 3.3418, + "step": 3649 + }, + { + "epoch": 0.8, + "learning_rate": 4.60672929866725e-06, + "loss": 2.9188, + "step": 3650 + }, + { + "epoch": 0.8, + "learning_rate": 4.606620056805768e-06, + "loss": 3.3602, + "step": 3651 + }, + { + "epoch": 0.8, + "learning_rate": 4.606510814944287e-06, + "loss": 3.43, + "step": 3652 + }, + { + "epoch": 0.8, + "learning_rate": 4.606401573082805e-06, + "loss": 3.1638, + "step": 3653 + }, + { + "epoch": 0.8, + "learning_rate": 4.606292331221324e-06, + "loss": 3.4725, + "step": 3654 + }, + { + "epoch": 0.8, + "learning_rate": 4.606183089359843e-06, + "loss": 3.5315, + "step": 3655 + }, + { + "epoch": 0.8, + "learning_rate": 4.6060738474983615e-06, + "loss": 3.5433, + "step": 3656 + }, + { + "epoch": 0.8, + "learning_rate": 4.6059646056368805e-06, + "loss": 3.2024, + "step": 3657 + }, + { + "epoch": 0.8, + "learning_rate": 4.605855363775399e-06, + "loss": 3.2585, + "step": 3658 + }, + { + "epoch": 0.8, + "learning_rate": 4.605746121913918e-06, + "loss": 3.2682, + "step": 3659 + }, + { + "epoch": 0.8, + "learning_rate": 4.605636880052437e-06, + "loss": 3.8363, + "step": 3660 + }, + { + "epoch": 0.8, + "learning_rate": 4.605527638190955e-06, + "loss": 3.113, + "step": 3661 + }, + { + "epoch": 0.8, + "learning_rate": 4.605418396329474e-06, + "loss": 3.7039, + "step": 3662 + }, + { + "epoch": 0.8, + "learning_rate": 4.605309154467992e-06, + "loss": 3.46, + "step": 3663 + }, + { + "epoch": 0.8, + "learning_rate": 4.605199912606511e-06, + "loss": 4.0595, + "step": 3664 + }, + { + "epoch": 0.8, + "learning_rate": 4.60509067074503e-06, + "loss": 3.7645, + "step": 3665 + }, + { + "epoch": 0.8, + "learning_rate": 4.6049814288835485e-06, + "loss": 3.6619, + "step": 3666 + }, + { + "epoch": 0.8, + "learning_rate": 4.6048721870220675e-06, + "loss": 3.8878, + "step": 3667 + }, + { + "epoch": 0.8, + "learning_rate": 4.604762945160586e-06, + "loss": 3.1071, + "step": 3668 + }, + { + "epoch": 0.8, + "learning_rate": 4.604653703299105e-06, + "loss": 3.2883, + "step": 3669 + }, + { + "epoch": 0.8, + "learning_rate": 4.604544461437624e-06, + "loss": 3.3701, + "step": 3670 + }, + { + "epoch": 0.8, + "learning_rate": 4.604435219576142e-06, + "loss": 2.7489, + "step": 3671 + }, + { + "epoch": 0.8, + "learning_rate": 4.60432597771466e-06, + "loss": 3.4172, + "step": 3672 + }, + { + "epoch": 0.8, + "learning_rate": 4.604216735853179e-06, + "loss": 3.4787, + "step": 3673 + }, + { + "epoch": 0.8, + "learning_rate": 4.604107493991698e-06, + "loss": 3.3002, + "step": 3674 + }, + { + "epoch": 0.8, + "learning_rate": 4.603998252130216e-06, + "loss": 3.5153, + "step": 3675 + }, + { + "epoch": 0.8, + "learning_rate": 4.603889010268735e-06, + "loss": 3.3034, + "step": 3676 + }, + { + "epoch": 0.8, + "learning_rate": 4.603779768407254e-06, + "loss": 2.9295, + "step": 3677 + }, + { + "epoch": 0.8, + "learning_rate": 4.603670526545773e-06, + "loss": 3.3039, + "step": 3678 + }, + { + "epoch": 0.8, + "learning_rate": 4.603561284684292e-06, + "loss": 3.5136, + "step": 3679 + }, + { + "epoch": 0.8, + "learning_rate": 4.60345204282281e-06, + "loss": 3.369, + "step": 3680 + }, + { + "epoch": 0.8, + "learning_rate": 4.603342800961329e-06, + "loss": 3.0919, + "step": 3681 + }, + { + "epoch": 0.8, + "learning_rate": 4.603233559099847e-06, + "loss": 3.4685, + "step": 3682 + }, + { + "epoch": 0.8, + "learning_rate": 4.603124317238366e-06, + "loss": 3.4393, + "step": 3683 + }, + { + "epoch": 0.8, + "learning_rate": 4.603015075376885e-06, + "loss": 3.6482, + "step": 3684 + }, + { + "epoch": 0.8, + "learning_rate": 4.602905833515403e-06, + "loss": 3.5398, + "step": 3685 + }, + { + "epoch": 0.8, + "learning_rate": 4.6027965916539215e-06, + "loss": 3.7732, + "step": 3686 + }, + { + "epoch": 0.8, + "learning_rate": 4.6026873497924406e-06, + "loss": 3.6666, + "step": 3687 + }, + { + "epoch": 0.8, + "learning_rate": 4.60257810793096e-06, + "loss": 3.8307, + "step": 3688 + }, + { + "epoch": 0.81, + "learning_rate": 4.602468866069479e-06, + "loss": 3.23, + "step": 3689 + }, + { + "epoch": 0.81, + "learning_rate": 4.602359624207997e-06, + "loss": 3.7214, + "step": 3690 + }, + { + "epoch": 0.81, + "learning_rate": 4.602250382346515e-06, + "loss": 3.4369, + "step": 3691 + }, + { + "epoch": 0.81, + "learning_rate": 4.602141140485034e-06, + "loss": 3.5455, + "step": 3692 + }, + { + "epoch": 0.81, + "learning_rate": 4.602031898623553e-06, + "loss": 3.6663, + "step": 3693 + }, + { + "epoch": 0.81, + "learning_rate": 4.601922656762072e-06, + "loss": 3.8487, + "step": 3694 + }, + { + "epoch": 0.81, + "learning_rate": 4.60181341490059e-06, + "loss": 3.1521, + "step": 3695 + }, + { + "epoch": 0.81, + "learning_rate": 4.6017041730391085e-06, + "loss": 3.1706, + "step": 3696 + }, + { + "epoch": 0.81, + "learning_rate": 4.6015949311776275e-06, + "loss": 3.5039, + "step": 3697 + }, + { + "epoch": 0.81, + "learning_rate": 4.6014856893161466e-06, + "loss": 3.5614, + "step": 3698 + }, + { + "epoch": 0.81, + "learning_rate": 4.601376447454665e-06, + "loss": 3.3477, + "step": 3699 + }, + { + "epoch": 0.81, + "learning_rate": 4.601267205593184e-06, + "loss": 3.3361, + "step": 3700 + }, + { + "epoch": 0.81, + "learning_rate": 4.601157963731702e-06, + "loss": 3.6, + "step": 3701 + }, + { + "epoch": 0.81, + "learning_rate": 4.601048721870221e-06, + "loss": 3.4887, + "step": 3702 + }, + { + "epoch": 0.81, + "learning_rate": 4.60093948000874e-06, + "loss": 3.3802, + "step": 3703 + }, + { + "epoch": 0.81, + "learning_rate": 4.600830238147258e-06, + "loss": 3.7838, + "step": 3704 + }, + { + "epoch": 0.81, + "learning_rate": 4.600720996285777e-06, + "loss": 3.272, + "step": 3705 + }, + { + "epoch": 0.81, + "learning_rate": 4.6006117544242954e-06, + "loss": 3.8554, + "step": 3706 + }, + { + "epoch": 0.81, + "learning_rate": 4.6005025125628145e-06, + "loss": 3.3946, + "step": 3707 + }, + { + "epoch": 0.81, + "learning_rate": 4.6003932707013335e-06, + "loss": 3.8144, + "step": 3708 + }, + { + "epoch": 0.81, + "learning_rate": 4.600284028839852e-06, + "loss": 3.458, + "step": 3709 + }, + { + "epoch": 0.81, + "learning_rate": 4.60017478697837e-06, + "loss": 3.7582, + "step": 3710 + }, + { + "epoch": 0.81, + "learning_rate": 4.600065545116889e-06, + "loss": 3.884, + "step": 3711 + }, + { + "epoch": 0.81, + "learning_rate": 4.599956303255408e-06, + "loss": 3.7685, + "step": 3712 + }, + { + "epoch": 0.81, + "learning_rate": 4.599847061393927e-06, + "loss": 3.4254, + "step": 3713 + }, + { + "epoch": 0.81, + "learning_rate": 4.599737819532445e-06, + "loss": 3.5603, + "step": 3714 + }, + { + "epoch": 0.81, + "learning_rate": 4.599628577670963e-06, + "loss": 3.3212, + "step": 3715 + }, + { + "epoch": 0.81, + "learning_rate": 4.599519335809482e-06, + "loss": 3.5857, + "step": 3716 + }, + { + "epoch": 0.81, + "learning_rate": 4.5994100939480014e-06, + "loss": 2.7583, + "step": 3717 + }, + { + "epoch": 0.81, + "learning_rate": 4.5993008520865205e-06, + "loss": 3.6291, + "step": 3718 + }, + { + "epoch": 0.81, + "learning_rate": 4.599191610225039e-06, + "loss": 3.0152, + "step": 3719 + }, + { + "epoch": 0.81, + "learning_rate": 4.599082368363557e-06, + "loss": 3.8632, + "step": 3720 + }, + { + "epoch": 0.81, + "learning_rate": 4.598973126502076e-06, + "loss": 3.8432, + "step": 3721 + }, + { + "epoch": 0.81, + "learning_rate": 4.598863884640595e-06, + "loss": 3.3666, + "step": 3722 + }, + { + "epoch": 0.81, + "learning_rate": 4.598754642779114e-06, + "loss": 3.7678, + "step": 3723 + }, + { + "epoch": 0.81, + "learning_rate": 4.598645400917632e-06, + "loss": 3.6197, + "step": 3724 + }, + { + "epoch": 0.81, + "learning_rate": 4.59853615905615e-06, + "loss": 3.2098, + "step": 3725 + }, + { + "epoch": 0.81, + "learning_rate": 4.598426917194669e-06, + "loss": 3.5756, + "step": 3726 + }, + { + "epoch": 0.81, + "learning_rate": 4.598317675333188e-06, + "loss": 3.6114, + "step": 3727 + }, + { + "epoch": 0.81, + "learning_rate": 4.598208433471707e-06, + "loss": 3.2365, + "step": 3728 + }, + { + "epoch": 0.81, + "learning_rate": 4.598099191610226e-06, + "loss": 2.7791, + "step": 3729 + }, + { + "epoch": 0.81, + "learning_rate": 4.597989949748744e-06, + "loss": 3.5463, + "step": 3730 + }, + { + "epoch": 0.81, + "learning_rate": 4.597880707887263e-06, + "loss": 3.6974, + "step": 3731 + }, + { + "epoch": 0.81, + "learning_rate": 4.597771466025782e-06, + "loss": 3.2064, + "step": 3732 + }, + { + "epoch": 0.81, + "learning_rate": 4.5976622241643e-06, + "loss": 3.3707, + "step": 3733 + }, + { + "epoch": 0.81, + "learning_rate": 4.597552982302818e-06, + "loss": 3.565, + "step": 3734 + }, + { + "epoch": 0.82, + "learning_rate": 4.597443740441337e-06, + "loss": 3.6087, + "step": 3735 + }, + { + "epoch": 0.82, + "learning_rate": 4.597334498579856e-06, + "loss": 3.7547, + "step": 3736 + }, + { + "epoch": 0.82, + "learning_rate": 4.597225256718375e-06, + "loss": 3.4826, + "step": 3737 + }, + { + "epoch": 0.82, + "learning_rate": 4.5971160148568935e-06, + "loss": 2.8995, + "step": 3738 + }, + { + "epoch": 0.82, + "learning_rate": 4.597006772995412e-06, + "loss": 3.4742, + "step": 3739 + }, + { + "epoch": 0.82, + "learning_rate": 4.596897531133931e-06, + "loss": 3.3399, + "step": 3740 + }, + { + "epoch": 0.82, + "learning_rate": 4.59678828927245e-06, + "loss": 3.7676, + "step": 3741 + }, + { + "epoch": 0.82, + "learning_rate": 4.596679047410969e-06, + "loss": 3.3492, + "step": 3742 + }, + { + "epoch": 0.82, + "learning_rate": 4.596569805549487e-06, + "loss": 2.8683, + "step": 3743 + }, + { + "epoch": 0.82, + "learning_rate": 4.596460563688005e-06, + "loss": 3.1604, + "step": 3744 + }, + { + "epoch": 0.82, + "learning_rate": 4.596351321826524e-06, + "loss": 3.5333, + "step": 3745 + }, + { + "epoch": 0.82, + "learning_rate": 4.596242079965043e-06, + "loss": 3.7506, + "step": 3746 + }, + { + "epoch": 0.82, + "learning_rate": 4.596132838103562e-06, + "loss": 3.5405, + "step": 3747 + }, + { + "epoch": 0.82, + "learning_rate": 4.5960235962420805e-06, + "loss": 3.9675, + "step": 3748 + }, + { + "epoch": 0.82, + "learning_rate": 4.595914354380599e-06, + "loss": 2.9188, + "step": 3749 + }, + { + "epoch": 0.82, + "learning_rate": 4.595805112519118e-06, + "loss": 3.7583, + "step": 3750 + }, + { + "epoch": 0.82, + "learning_rate": 4.595695870657637e-06, + "loss": 3.1958, + "step": 3751 + }, + { + "epoch": 0.82, + "learning_rate": 4.595586628796155e-06, + "loss": 3.2747, + "step": 3752 + }, + { + "epoch": 0.82, + "learning_rate": 4.595477386934674e-06, + "loss": 3.1948, + "step": 3753 + }, + { + "epoch": 0.82, + "learning_rate": 4.595368145073192e-06, + "loss": 3.2697, + "step": 3754 + }, + { + "epoch": 0.82, + "learning_rate": 4.595258903211711e-06, + "loss": 3.1628, + "step": 3755 + }, + { + "epoch": 0.82, + "learning_rate": 4.59514966135023e-06, + "loss": 3.8806, + "step": 3756 + }, + { + "epoch": 0.82, + "learning_rate": 4.595040419488748e-06, + "loss": 3.4729, + "step": 3757 + }, + { + "epoch": 0.82, + "learning_rate": 4.5949311776272674e-06, + "loss": 3.4449, + "step": 3758 + }, + { + "epoch": 0.82, + "learning_rate": 4.594821935765786e-06, + "loss": 3.1644, + "step": 3759 + }, + { + "epoch": 0.82, + "learning_rate": 4.594712693904305e-06, + "loss": 3.0504, + "step": 3760 + }, + { + "epoch": 0.82, + "learning_rate": 4.594603452042824e-06, + "loss": 3.5136, + "step": 3761 + }, + { + "epoch": 0.82, + "learning_rate": 4.594494210181342e-06, + "loss": 2.4499, + "step": 3762 + }, + { + "epoch": 0.82, + "learning_rate": 4.59438496831986e-06, + "loss": 3.127, + "step": 3763 + }, + { + "epoch": 0.82, + "learning_rate": 4.594275726458379e-06, + "loss": 2.7779, + "step": 3764 + }, + { + "epoch": 0.82, + "learning_rate": 4.594166484596898e-06, + "loss": 3.2276, + "step": 3765 + }, + { + "epoch": 0.82, + "learning_rate": 4.594057242735416e-06, + "loss": 3.3716, + "step": 3766 + }, + { + "epoch": 0.82, + "learning_rate": 4.593948000873935e-06, + "loss": 3.6092, + "step": 3767 + }, + { + "epoch": 0.82, + "learning_rate": 4.5938387590124536e-06, + "loss": 3.371, + "step": 3768 + }, + { + "epoch": 0.82, + "learning_rate": 4.593729517150973e-06, + "loss": 3.2734, + "step": 3769 + }, + { + "epoch": 0.82, + "learning_rate": 4.593620275289492e-06, + "loss": 3.6023, + "step": 3770 + }, + { + "epoch": 0.82, + "learning_rate": 4.59351103342801e-06, + "loss": 2.8156, + "step": 3771 + }, + { + "epoch": 0.82, + "learning_rate": 4.593401791566529e-06, + "loss": 3.7583, + "step": 3772 + }, + { + "epoch": 0.82, + "learning_rate": 4.593292549705047e-06, + "loss": 3.1813, + "step": 3773 + }, + { + "epoch": 0.82, + "learning_rate": 4.593183307843566e-06, + "loss": 3.8143, + "step": 3774 + }, + { + "epoch": 0.82, + "learning_rate": 4.593074065982085e-06, + "loss": 3.301, + "step": 3775 + }, + { + "epoch": 0.82, + "learning_rate": 4.592964824120603e-06, + "loss": 3.2681, + "step": 3776 + }, + { + "epoch": 0.82, + "learning_rate": 4.592855582259122e-06, + "loss": 3.3323, + "step": 3777 + }, + { + "epoch": 0.82, + "learning_rate": 4.5927463403976405e-06, + "loss": 3.7829, + "step": 3778 + }, + { + "epoch": 0.82, + "learning_rate": 4.5926370985361596e-06, + "loss": 3.3495, + "step": 3779 + }, + { + "epoch": 0.82, + "learning_rate": 4.592527856674679e-06, + "loss": 3.2211, + "step": 3780 + }, + { + "epoch": 0.83, + "learning_rate": 4.592418614813197e-06, + "loss": 3.5172, + "step": 3781 + }, + { + "epoch": 0.83, + "learning_rate": 4.592309372951716e-06, + "loss": 3.7014, + "step": 3782 + }, + { + "epoch": 0.83, + "learning_rate": 4.592200131090234e-06, + "loss": 3.25, + "step": 3783 + }, + { + "epoch": 0.83, + "learning_rate": 4.592090889228753e-06, + "loss": 2.9373, + "step": 3784 + }, + { + "epoch": 0.83, + "learning_rate": 4.591981647367271e-06, + "loss": 3.8133, + "step": 3785 + }, + { + "epoch": 0.83, + "learning_rate": 4.59187240550579e-06, + "loss": 3.6505, + "step": 3786 + }, + { + "epoch": 0.83, + "learning_rate": 4.5917631636443084e-06, + "loss": 3.7554, + "step": 3787 + }, + { + "epoch": 0.83, + "learning_rate": 4.5916539217828275e-06, + "loss": 3.6645, + "step": 3788 + }, + { + "epoch": 0.83, + "learning_rate": 4.5915446799213465e-06, + "loss": 3.3644, + "step": 3789 + }, + { + "epoch": 0.83, + "learning_rate": 4.591435438059865e-06, + "loss": 3.7767, + "step": 3790 + }, + { + "epoch": 0.83, + "learning_rate": 4.591326196198384e-06, + "loss": 3.628, + "step": 3791 + }, + { + "epoch": 0.83, + "learning_rate": 4.591216954336902e-06, + "loss": 3.2234, + "step": 3792 + }, + { + "epoch": 0.83, + "learning_rate": 4.591107712475421e-06, + "loss": 3.7293, + "step": 3793 + }, + { + "epoch": 0.83, + "learning_rate": 4.59099847061394e-06, + "loss": 3.0683, + "step": 3794 + }, + { + "epoch": 0.83, + "learning_rate": 4.590889228752458e-06, + "loss": 3.6754, + "step": 3795 + }, + { + "epoch": 0.83, + "learning_rate": 4.590779986890977e-06, + "loss": 3.5231, + "step": 3796 + }, + { + "epoch": 0.83, + "learning_rate": 4.590670745029495e-06, + "loss": 3.4698, + "step": 3797 + }, + { + "epoch": 0.83, + "learning_rate": 4.5905615031680144e-06, + "loss": 3.1553, + "step": 3798 + }, + { + "epoch": 0.83, + "learning_rate": 4.590452261306533e-06, + "loss": 3.7345, + "step": 3799 + }, + { + "epoch": 0.83, + "learning_rate": 4.590343019445052e-06, + "loss": 2.6199, + "step": 3800 + }, + { + "epoch": 0.83, + "learning_rate": 4.590233777583571e-06, + "loss": 3.3688, + "step": 3801 + }, + { + "epoch": 0.83, + "learning_rate": 4.590124535722089e-06, + "loss": 3.1948, + "step": 3802 + }, + { + "epoch": 0.83, + "learning_rate": 4.590015293860608e-06, + "loss": 3.4216, + "step": 3803 + }, + { + "epoch": 0.83, + "learning_rate": 4.589906051999126e-06, + "loss": 3.7506, + "step": 3804 + }, + { + "epoch": 0.83, + "learning_rate": 4.589796810137645e-06, + "loss": 3.1609, + "step": 3805 + }, + { + "epoch": 0.83, + "learning_rate": 4.589687568276164e-06, + "loss": 3.3206, + "step": 3806 + }, + { + "epoch": 0.83, + "learning_rate": 4.589578326414682e-06, + "loss": 3.266, + "step": 3807 + }, + { + "epoch": 0.83, + "learning_rate": 4.589469084553201e-06, + "loss": 2.7218, + "step": 3808 + }, + { + "epoch": 0.83, + "learning_rate": 4.5893598426917196e-06, + "loss": 3.5833, + "step": 3809 + }, + { + "epoch": 0.83, + "learning_rate": 4.589250600830239e-06, + "loss": 3.4505, + "step": 3810 + }, + { + "epoch": 0.83, + "learning_rate": 4.589141358968758e-06, + "loss": 3.918, + "step": 3811 + }, + { + "epoch": 0.83, + "learning_rate": 4.589032117107276e-06, + "loss": 2.8954, + "step": 3812 + }, + { + "epoch": 0.83, + "learning_rate": 4.588922875245795e-06, + "loss": 3.4666, + "step": 3813 + }, + { + "epoch": 0.83, + "learning_rate": 4.588813633384313e-06, + "loss": 3.2273, + "step": 3814 + }, + { + "epoch": 0.83, + "learning_rate": 4.588704391522832e-06, + "loss": 3.0069, + "step": 3815 + }, + { + "epoch": 0.83, + "learning_rate": 4.58859514966135e-06, + "loss": 3.0596, + "step": 3816 + }, + { + "epoch": 0.83, + "learning_rate": 4.588485907799869e-06, + "loss": 3.6093, + "step": 3817 + }, + { + "epoch": 0.83, + "learning_rate": 4.5883766659383875e-06, + "loss": 3.6313, + "step": 3818 + }, + { + "epoch": 0.83, + "learning_rate": 4.5882674240769065e-06, + "loss": 3.1289, + "step": 3819 + }, + { + "epoch": 0.83, + "learning_rate": 4.5881581822154256e-06, + "loss": 3.3607, + "step": 3820 + }, + { + "epoch": 0.83, + "learning_rate": 4.588048940353944e-06, + "loss": 3.4215, + "step": 3821 + }, + { + "epoch": 0.83, + "learning_rate": 4.587939698492463e-06, + "loss": 3.5887, + "step": 3822 + }, + { + "epoch": 0.83, + "learning_rate": 4.587830456630981e-06, + "loss": 2.9983, + "step": 3823 + }, + { + "epoch": 0.83, + "learning_rate": 4.5877212147695e-06, + "loss": 3.5128, + "step": 3824 + }, + { + "epoch": 0.83, + "learning_rate": 4.587611972908019e-06, + "loss": 3.1652, + "step": 3825 + }, + { + "epoch": 0.84, + "learning_rate": 4.587502731046537e-06, + "loss": 2.9627, + "step": 3826 + }, + { + "epoch": 0.84, + "learning_rate": 4.587393489185056e-06, + "loss": 3.4628, + "step": 3827 + }, + { + "epoch": 0.84, + "learning_rate": 4.5872842473235745e-06, + "loss": 4.0723, + "step": 3828 + }, + { + "epoch": 0.84, + "learning_rate": 4.5871750054620935e-06, + "loss": 3.5819, + "step": 3829 + }, + { + "epoch": 0.84, + "learning_rate": 4.5870657636006125e-06, + "loss": 3.5501, + "step": 3830 + }, + { + "epoch": 0.84, + "learning_rate": 4.586956521739131e-06, + "loss": 2.5489, + "step": 3831 + }, + { + "epoch": 0.84, + "learning_rate": 4.586847279877649e-06, + "loss": 3.2451, + "step": 3832 + }, + { + "epoch": 0.84, + "learning_rate": 4.586738038016168e-06, + "loss": 3.5069, + "step": 3833 + }, + { + "epoch": 0.84, + "learning_rate": 4.586628796154687e-06, + "loss": 3.2626, + "step": 3834 + }, + { + "epoch": 0.84, + "learning_rate": 4.586519554293206e-06, + "loss": 3.3481, + "step": 3835 + }, + { + "epoch": 0.84, + "learning_rate": 4.586410312431724e-06, + "loss": 3.8403, + "step": 3836 + }, + { + "epoch": 0.84, + "learning_rate": 4.586301070570242e-06, + "loss": 3.4566, + "step": 3837 + }, + { + "epoch": 0.84, + "learning_rate": 4.586191828708761e-06, + "loss": 3.383, + "step": 3838 + }, + { + "epoch": 0.84, + "learning_rate": 4.5860825868472804e-06, + "loss": 2.9854, + "step": 3839 + }, + { + "epoch": 0.84, + "learning_rate": 4.585973344985799e-06, + "loss": 3.0525, + "step": 3840 + }, + { + "epoch": 0.84, + "learning_rate": 4.585864103124318e-06, + "loss": 3.8115, + "step": 3841 + }, + { + "epoch": 0.84, + "learning_rate": 4.585754861262836e-06, + "loss": 3.0709, + "step": 3842 + }, + { + "epoch": 0.84, + "learning_rate": 4.585645619401355e-06, + "loss": 3.3363, + "step": 3843 + }, + { + "epoch": 0.84, + "learning_rate": 4.585536377539874e-06, + "loss": 3.6148, + "step": 3844 + }, + { + "epoch": 0.84, + "learning_rate": 4.585427135678392e-06, + "loss": 3.1747, + "step": 3845 + }, + { + "epoch": 0.84, + "learning_rate": 4.585317893816911e-06, + "loss": 3.0373, + "step": 3846 + }, + { + "epoch": 0.84, + "learning_rate": 4.585208651955429e-06, + "loss": 3.0756, + "step": 3847 + }, + { + "epoch": 0.84, + "learning_rate": 4.585099410093948e-06, + "loss": 3.3781, + "step": 3848 + }, + { + "epoch": 0.84, + "learning_rate": 4.584990168232467e-06, + "loss": 3.525, + "step": 3849 + }, + { + "epoch": 0.84, + "learning_rate": 4.584880926370986e-06, + "loss": 3.2625, + "step": 3850 + }, + { + "epoch": 0.84, + "learning_rate": 4.584771684509504e-06, + "loss": 3.6961, + "step": 3851 + }, + { + "epoch": 0.84, + "learning_rate": 4.584662442648023e-06, + "loss": 3.5817, + "step": 3852 + }, + { + "epoch": 0.84, + "learning_rate": 4.584553200786542e-06, + "loss": 2.8877, + "step": 3853 + }, + { + "epoch": 0.84, + "learning_rate": 4.584443958925061e-06, + "loss": 3.558, + "step": 3854 + }, + { + "epoch": 0.84, + "learning_rate": 4.584334717063579e-06, + "loss": 3.528, + "step": 3855 + }, + { + "epoch": 0.84, + "learning_rate": 4.584225475202097e-06, + "loss": 3.2544, + "step": 3856 + }, + { + "epoch": 0.84, + "learning_rate": 4.584116233340616e-06, + "loss": 3.3115, + "step": 3857 + }, + { + "epoch": 0.84, + "learning_rate": 4.584006991479135e-06, + "loss": 3.4691, + "step": 3858 + }, + { + "epoch": 0.84, + "learning_rate": 4.583897749617654e-06, + "loss": 3.6226, + "step": 3859 + }, + { + "epoch": 0.84, + "learning_rate": 4.5837885077561725e-06, + "loss": 3.5279, + "step": 3860 + }, + { + "epoch": 0.84, + "learning_rate": 4.583679265894691e-06, + "loss": 3.4271, + "step": 3861 + }, + { + "epoch": 0.84, + "learning_rate": 4.58357002403321e-06, + "loss": 2.927, + "step": 3862 + }, + { + "epoch": 0.84, + "learning_rate": 4.583460782171729e-06, + "loss": 3.8135, + "step": 3863 + }, + { + "epoch": 0.84, + "learning_rate": 4.583351540310248e-06, + "loss": 2.5359, + "step": 3864 + }, + { + "epoch": 0.84, + "learning_rate": 4.583242298448766e-06, + "loss": 3.9241, + "step": 3865 + }, + { + "epoch": 0.84, + "learning_rate": 4.583133056587284e-06, + "loss": 3.1072, + "step": 3866 + }, + { + "epoch": 0.84, + "learning_rate": 4.583023814725803e-06, + "loss": 2.4736, + "step": 3867 + }, + { + "epoch": 0.84, + "learning_rate": 4.582914572864322e-06, + "loss": 3.464, + "step": 3868 + }, + { + "epoch": 0.84, + "learning_rate": 4.5828053310028405e-06, + "loss": 2.7933, + "step": 3869 + }, + { + "epoch": 0.84, + "learning_rate": 4.5826960891413595e-06, + "loss": 3.5238, + "step": 3870 + }, + { + "epoch": 0.84, + "learning_rate": 4.582586847279878e-06, + "loss": 3.4317, + "step": 3871 + }, + { + "epoch": 0.85, + "learning_rate": 4.582477605418397e-06, + "loss": 3.4198, + "step": 3872 + }, + { + "epoch": 0.85, + "learning_rate": 4.582368363556916e-06, + "loss": 3.1305, + "step": 3873 + }, + { + "epoch": 0.85, + "learning_rate": 4.582259121695434e-06, + "loss": 3.6733, + "step": 3874 + }, + { + "epoch": 0.85, + "learning_rate": 4.582149879833952e-06, + "loss": 3.1724, + "step": 3875 + }, + { + "epoch": 0.85, + "learning_rate": 4.582040637972471e-06, + "loss": 3.6256, + "step": 3876 + }, + { + "epoch": 0.85, + "learning_rate": 4.58193139611099e-06, + "loss": 2.5231, + "step": 3877 + }, + { + "epoch": 0.85, + "learning_rate": 4.581822154249509e-06, + "loss": 2.9308, + "step": 3878 + }, + { + "epoch": 0.85, + "learning_rate": 4.5817129123880274e-06, + "loss": 3.7387, + "step": 3879 + }, + { + "epoch": 0.85, + "learning_rate": 4.581603670526546e-06, + "loss": 3.5473, + "step": 3880 + }, + { + "epoch": 0.85, + "learning_rate": 4.581494428665065e-06, + "loss": 3.374, + "step": 3881 + }, + { + "epoch": 0.85, + "learning_rate": 4.581385186803584e-06, + "loss": 2.9096, + "step": 3882 + }, + { + "epoch": 0.85, + "learning_rate": 4.581275944942103e-06, + "loss": 3.8237, + "step": 3883 + }, + { + "epoch": 0.85, + "learning_rate": 4.581166703080621e-06, + "loss": 3.372, + "step": 3884 + }, + { + "epoch": 0.85, + "learning_rate": 4.581057461219139e-06, + "loss": 3.4806, + "step": 3885 + }, + { + "epoch": 0.85, + "learning_rate": 4.580948219357658e-06, + "loss": 3.4872, + "step": 3886 + }, + { + "epoch": 0.85, + "learning_rate": 4.580838977496177e-06, + "loss": 2.9763, + "step": 3887 + }, + { + "epoch": 0.85, + "learning_rate": 4.580729735634696e-06, + "loss": 3.5001, + "step": 3888 + }, + { + "epoch": 0.85, + "learning_rate": 4.580620493773214e-06, + "loss": 3.1693, + "step": 3889 + }, + { + "epoch": 0.85, + "learning_rate": 4.5805112519117326e-06, + "loss": 3.4212, + "step": 3890 + }, + { + "epoch": 0.85, + "learning_rate": 4.580402010050252e-06, + "loss": 3.445, + "step": 3891 + }, + { + "epoch": 0.85, + "learning_rate": 4.580292768188771e-06, + "loss": 3.2502, + "step": 3892 + }, + { + "epoch": 0.85, + "learning_rate": 4.580183526327289e-06, + "loss": 3.4316, + "step": 3893 + }, + { + "epoch": 0.85, + "learning_rate": 4.580074284465808e-06, + "loss": 3.8554, + "step": 3894 + }, + { + "epoch": 0.85, + "learning_rate": 4.579965042604326e-06, + "loss": 3.7297, + "step": 3895 + }, + { + "epoch": 0.85, + "learning_rate": 4.579855800742845e-06, + "loss": 3.4545, + "step": 3896 + }, + { + "epoch": 0.85, + "learning_rate": 4.579746558881364e-06, + "loss": 3.0316, + "step": 3897 + }, + { + "epoch": 0.85, + "learning_rate": 4.579637317019882e-06, + "loss": 3.0789, + "step": 3898 + }, + { + "epoch": 0.85, + "learning_rate": 4.579528075158401e-06, + "loss": 3.6309, + "step": 3899 + }, + { + "epoch": 0.85, + "learning_rate": 4.5794188332969195e-06, + "loss": 3.7848, + "step": 3900 + }, + { + "epoch": 0.85, + "learning_rate": 4.5793095914354386e-06, + "loss": 3.589, + "step": 3901 + }, + { + "epoch": 0.85, + "learning_rate": 4.579200349573958e-06, + "loss": 3.5854, + "step": 3902 + }, + { + "epoch": 0.85, + "learning_rate": 4.579091107712476e-06, + "loss": 3.7589, + "step": 3903 + }, + { + "epoch": 0.85, + "learning_rate": 4.578981865850994e-06, + "loss": 3.7843, + "step": 3904 + }, + { + "epoch": 0.85, + "learning_rate": 4.578872623989513e-06, + "loss": 3.2828, + "step": 3905 + }, + { + "epoch": 0.85, + "learning_rate": 4.578763382128032e-06, + "loss": 3.1698, + "step": 3906 + }, + { + "epoch": 0.85, + "learning_rate": 4.578654140266551e-06, + "loss": 2.9874, + "step": 3907 + }, + { + "epoch": 0.85, + "learning_rate": 4.578544898405069e-06, + "loss": 3.2802, + "step": 3908 + }, + { + "epoch": 0.85, + "learning_rate": 4.5784356565435874e-06, + "loss": 3.227, + "step": 3909 + }, + { + "epoch": 0.85, + "learning_rate": 4.5783264146821065e-06, + "loss": 2.5077, + "step": 3910 + }, + { + "epoch": 0.85, + "learning_rate": 4.5782171728206255e-06, + "loss": 3.2536, + "step": 3911 + }, + { + "epoch": 0.85, + "learning_rate": 4.5781079309591446e-06, + "loss": 3.2691, + "step": 3912 + }, + { + "epoch": 0.85, + "learning_rate": 4.577998689097663e-06, + "loss": 3.2475, + "step": 3913 + }, + { + "epoch": 0.85, + "learning_rate": 4.577889447236181e-06, + "loss": 2.9664, + "step": 3914 + }, + { + "epoch": 0.85, + "learning_rate": 4.5777802053747e-06, + "loss": 3.5629, + "step": 3915 + }, + { + "epoch": 0.85, + "learning_rate": 4.577670963513219e-06, + "loss": 3.874, + "step": 3916 + }, + { + "epoch": 0.85, + "learning_rate": 4.577561721651737e-06, + "loss": 3.8223, + "step": 3917 + }, + { + "epoch": 0.86, + "learning_rate": 4.577452479790256e-06, + "loss": 3.1666, + "step": 3918 + }, + { + "epoch": 0.86, + "learning_rate": 4.577343237928774e-06, + "loss": 3.2347, + "step": 3919 + }, + { + "epoch": 0.86, + "learning_rate": 4.5772339960672934e-06, + "loss": 3.6096, + "step": 3920 + }, + { + "epoch": 0.86, + "learning_rate": 4.5771247542058125e-06, + "loss": 3.7856, + "step": 3921 + }, + { + "epoch": 0.86, + "learning_rate": 4.577015512344331e-06, + "loss": 3.433, + "step": 3922 + }, + { + "epoch": 0.86, + "learning_rate": 4.57690627048285e-06, + "loss": 3.1164, + "step": 3923 + }, + { + "epoch": 0.86, + "learning_rate": 4.576797028621368e-06, + "loss": 3.7925, + "step": 3924 + }, + { + "epoch": 0.86, + "learning_rate": 4.576687786759887e-06, + "loss": 2.8858, + "step": 3925 + }, + { + "epoch": 0.86, + "learning_rate": 4.576578544898406e-06, + "loss": 3.3252, + "step": 3926 + }, + { + "epoch": 0.86, + "learning_rate": 4.576469303036924e-06, + "loss": 3.3758, + "step": 3927 + }, + { + "epoch": 0.86, + "learning_rate": 4.576360061175442e-06, + "loss": 3.5319, + "step": 3928 + }, + { + "epoch": 0.86, + "learning_rate": 4.576250819313961e-06, + "loss": 3.5076, + "step": 3929 + }, + { + "epoch": 0.86, + "learning_rate": 4.57614157745248e-06, + "loss": 3.383, + "step": 3930 + }, + { + "epoch": 0.86, + "learning_rate": 4.576032335590999e-06, + "loss": 2.9699, + "step": 3931 + }, + { + "epoch": 0.86, + "learning_rate": 4.575923093729518e-06, + "loss": 2.8676, + "step": 3932 + }, + { + "epoch": 0.86, + "learning_rate": 4.575813851868036e-06, + "loss": 2.597, + "step": 3933 + }, + { + "epoch": 0.86, + "learning_rate": 4.575704610006555e-06, + "loss": 3.6432, + "step": 3934 + }, + { + "epoch": 0.86, + "learning_rate": 4.575595368145074e-06, + "loss": 3.6851, + "step": 3935 + }, + { + "epoch": 0.86, + "learning_rate": 4.575486126283592e-06, + "loss": 2.9542, + "step": 3936 + }, + { + "epoch": 0.86, + "learning_rate": 4.575376884422111e-06, + "loss": 3.8853, + "step": 3937 + }, + { + "epoch": 0.86, + "learning_rate": 4.575267642560629e-06, + "loss": 3.2429, + "step": 3938 + }, + { + "epoch": 0.86, + "learning_rate": 4.575158400699148e-06, + "loss": 2.6661, + "step": 3939 + }, + { + "epoch": 0.86, + "learning_rate": 4.575049158837667e-06, + "loss": 3.1543, + "step": 3940 + }, + { + "epoch": 0.86, + "learning_rate": 4.5749399169761855e-06, + "loss": 3.4429, + "step": 3941 + }, + { + "epoch": 0.86, + "learning_rate": 4.574830675114705e-06, + "loss": 3.4793, + "step": 3942 + }, + { + "epoch": 0.86, + "learning_rate": 4.574721433253223e-06, + "loss": 3.7139, + "step": 3943 + }, + { + "epoch": 0.86, + "learning_rate": 4.574612191391742e-06, + "loss": 3.3196, + "step": 3944 + }, + { + "epoch": 0.86, + "learning_rate": 4.574502949530261e-06, + "loss": 3.3858, + "step": 3945 + }, + { + "epoch": 0.86, + "learning_rate": 4.574393707668779e-06, + "loss": 3.7838, + "step": 3946 + }, + { + "epoch": 0.86, + "learning_rate": 4.574284465807298e-06, + "loss": 3.6907, + "step": 3947 + }, + { + "epoch": 0.86, + "learning_rate": 4.574175223945816e-06, + "loss": 3.3575, + "step": 3948 + }, + { + "epoch": 0.86, + "learning_rate": 4.574065982084335e-06, + "loss": 2.5687, + "step": 3949 + }, + { + "epoch": 0.86, + "learning_rate": 4.5739567402228535e-06, + "loss": 3.4372, + "step": 3950 + }, + { + "epoch": 0.86, + "learning_rate": 4.5738474983613725e-06, + "loss": 3.4613, + "step": 3951 + }, + { + "epoch": 0.86, + "learning_rate": 4.5737382564998915e-06, + "loss": 3.2261, + "step": 3952 + }, + { + "epoch": 0.86, + "learning_rate": 4.57362901463841e-06, + "loss": 3.208, + "step": 3953 + }, + { + "epoch": 0.86, + "learning_rate": 4.573519772776929e-06, + "loss": 3.3207, + "step": 3954 + }, + { + "epoch": 0.86, + "learning_rate": 4.573410530915447e-06, + "loss": 3.1387, + "step": 3955 + }, + { + "epoch": 0.86, + "learning_rate": 4.573301289053966e-06, + "loss": 3.4196, + "step": 3956 + }, + { + "epoch": 0.86, + "learning_rate": 4.573192047192484e-06, + "loss": 3.8417, + "step": 3957 + }, + { + "epoch": 0.86, + "learning_rate": 4.573082805331003e-06, + "loss": 4.0003, + "step": 3958 + }, + { + "epoch": 0.86, + "learning_rate": 4.572973563469522e-06, + "loss": 3.224, + "step": 3959 + }, + { + "epoch": 0.86, + "learning_rate": 4.57286432160804e-06, + "loss": 3.3072, + "step": 3960 + }, + { + "epoch": 0.86, + "learning_rate": 4.5727550797465595e-06, + "loss": 4.1697, + "step": 3961 + }, + { + "epoch": 0.86, + "learning_rate": 4.572645837885078e-06, + "loss": 3.5811, + "step": 3962 + }, + { + "epoch": 0.86, + "learning_rate": 4.572536596023597e-06, + "loss": 3.4501, + "step": 3963 + }, + { + "epoch": 0.87, + "learning_rate": 4.572427354162115e-06, + "loss": 2.5286, + "step": 3964 + }, + { + "epoch": 0.87, + "learning_rate": 4.572318112300634e-06, + "loss": 3.4481, + "step": 3965 + }, + { + "epoch": 0.87, + "learning_rate": 4.572208870439153e-06, + "loss": 2.6976, + "step": 3966 + }, + { + "epoch": 0.87, + "learning_rate": 4.572099628577671e-06, + "loss": 4.0373, + "step": 3967 + }, + { + "epoch": 0.87, + "learning_rate": 4.57199038671619e-06, + "loss": 3.0738, + "step": 3968 + }, + { + "epoch": 0.87, + "learning_rate": 4.571881144854708e-06, + "loss": 3.1806, + "step": 3969 + }, + { + "epoch": 0.87, + "learning_rate": 4.571771902993227e-06, + "loss": 3.3687, + "step": 3970 + }, + { + "epoch": 0.87, + "learning_rate": 4.571662661131746e-06, + "loss": 3.174, + "step": 3971 + }, + { + "epoch": 0.87, + "learning_rate": 4.571553419270265e-06, + "loss": 3.4857, + "step": 3972 + }, + { + "epoch": 0.87, + "learning_rate": 4.571444177408784e-06, + "loss": 3.589, + "step": 3973 + }, + { + "epoch": 0.87, + "learning_rate": 4.571334935547302e-06, + "loss": 3.7285, + "step": 3974 + }, + { + "epoch": 0.87, + "learning_rate": 4.571225693685821e-06, + "loss": 3.4751, + "step": 3975 + }, + { + "epoch": 0.87, + "learning_rate": 4.57111645182434e-06, + "loss": 3.2177, + "step": 3976 + }, + { + "epoch": 0.87, + "learning_rate": 4.571007209962858e-06, + "loss": 3.7431, + "step": 3977 + }, + { + "epoch": 0.87, + "learning_rate": 4.570897968101377e-06, + "loss": 2.8024, + "step": 3978 + }, + { + "epoch": 0.87, + "learning_rate": 4.570788726239895e-06, + "loss": 3.0043, + "step": 3979 + }, + { + "epoch": 0.87, + "learning_rate": 4.570679484378414e-06, + "loss": 3.4377, + "step": 3980 + }, + { + "epoch": 0.87, + "learning_rate": 4.5705702425169325e-06, + "loss": 3.2076, + "step": 3981 + }, + { + "epoch": 0.87, + "learning_rate": 4.5704610006554516e-06, + "loss": 3.6986, + "step": 3982 + }, + { + "epoch": 0.87, + "learning_rate": 4.57035175879397e-06, + "loss": 2.6354, + "step": 3983 + }, + { + "epoch": 0.87, + "learning_rate": 4.570242516932489e-06, + "loss": 3.5587, + "step": 3984 + }, + { + "epoch": 0.87, + "learning_rate": 4.570133275071008e-06, + "loss": 2.8153, + "step": 3985 + }, + { + "epoch": 0.87, + "learning_rate": 4.570024033209526e-06, + "loss": 3.5037, + "step": 3986 + }, + { + "epoch": 0.87, + "learning_rate": 4.569914791348045e-06, + "loss": 3.3004, + "step": 3987 + }, + { + "epoch": 0.87, + "learning_rate": 4.569805549486563e-06, + "loss": 3.5965, + "step": 3988 + }, + { + "epoch": 0.87, + "learning_rate": 4.569696307625082e-06, + "loss": 3.3687, + "step": 3989 + }, + { + "epoch": 0.87, + "learning_rate": 4.569587065763601e-06, + "loss": 3.5603, + "step": 3990 + }, + { + "epoch": 0.87, + "learning_rate": 4.5694778239021195e-06, + "loss": 3.1373, + "step": 3991 + }, + { + "epoch": 0.87, + "learning_rate": 4.5693685820406385e-06, + "loss": 3.321, + "step": 3992 + }, + { + "epoch": 0.87, + "learning_rate": 4.569259340179157e-06, + "loss": 3.1075, + "step": 3993 + }, + { + "epoch": 0.87, + "learning_rate": 4.569150098317676e-06, + "loss": 3.2757, + "step": 3994 + }, + { + "epoch": 0.87, + "learning_rate": 4.569040856456195e-06, + "loss": 3.1462, + "step": 3995 + }, + { + "epoch": 0.87, + "learning_rate": 4.568931614594713e-06, + "loss": 3.256, + "step": 3996 + }, + { + "epoch": 0.87, + "learning_rate": 4.568822372733231e-06, + "loss": 3.0999, + "step": 3997 + }, + { + "epoch": 0.87, + "learning_rate": 4.56871313087175e-06, + "loss": 3.4607, + "step": 3998 + }, + { + "epoch": 0.87, + "learning_rate": 4.568603889010269e-06, + "loss": 3.7026, + "step": 3999 + }, + { + "epoch": 0.87, + "learning_rate": 4.568494647148788e-06, + "loss": 3.3258, + "step": 4000 + }, + { + "epoch": 0.87, + "learning_rate": 4.5683854052873064e-06, + "loss": 3.4925, + "step": 4001 + }, + { + "epoch": 0.87, + "learning_rate": 4.568276163425825e-06, + "loss": 3.6826, + "step": 4002 + }, + { + "epoch": 0.87, + "learning_rate": 4.568166921564344e-06, + "loss": 3.6877, + "step": 4003 + }, + { + "epoch": 0.87, + "learning_rate": 4.568057679702863e-06, + "loss": 3.9649, + "step": 4004 + }, + { + "epoch": 0.87, + "learning_rate": 4.567948437841382e-06, + "loss": 3.2983, + "step": 4005 + }, + { + "epoch": 0.87, + "learning_rate": 4.5678391959799e-06, + "loss": 3.0346, + "step": 4006 + }, + { + "epoch": 0.87, + "learning_rate": 4.567729954118418e-06, + "loss": 2.9156, + "step": 4007 + }, + { + "epoch": 0.87, + "learning_rate": 4.567620712256937e-06, + "loss": 3.7579, + "step": 4008 + }, + { + "epoch": 0.87, + "learning_rate": 4.567511470395456e-06, + "loss": 3.5981, + "step": 4009 + }, + { + "epoch": 0.88, + "learning_rate": 4.567402228533974e-06, + "loss": 3.7743, + "step": 4010 + }, + { + "epoch": 0.88, + "learning_rate": 4.567292986672493e-06, + "loss": 3.2988, + "step": 4011 + }, + { + "epoch": 0.88, + "learning_rate": 4.567183744811012e-06, + "loss": 3.5789, + "step": 4012 + }, + { + "epoch": 0.88, + "learning_rate": 4.567074502949531e-06, + "loss": 3.483, + "step": 4013 + }, + { + "epoch": 0.88, + "learning_rate": 4.56696526108805e-06, + "loss": 3.2797, + "step": 4014 + }, + { + "epoch": 0.88, + "learning_rate": 4.566856019226568e-06, + "loss": 3.4612, + "step": 4015 + }, + { + "epoch": 0.88, + "learning_rate": 4.566746777365086e-06, + "loss": 3.66, + "step": 4016 + }, + { + "epoch": 0.88, + "learning_rate": 4.566637535503605e-06, + "loss": 3.8963, + "step": 4017 + }, + { + "epoch": 0.88, + "learning_rate": 4.566528293642124e-06, + "loss": 3.2663, + "step": 4018 + }, + { + "epoch": 0.88, + "learning_rate": 4.566419051780643e-06, + "loss": 3.5817, + "step": 4019 + }, + { + "epoch": 0.88, + "learning_rate": 4.566309809919161e-06, + "loss": 3.2764, + "step": 4020 + }, + { + "epoch": 0.88, + "learning_rate": 4.5662005680576795e-06, + "loss": 3.7198, + "step": 4021 + }, + { + "epoch": 0.88, + "learning_rate": 4.5660913261961985e-06, + "loss": 2.9844, + "step": 4022 + }, + { + "epoch": 0.88, + "learning_rate": 4.5659820843347176e-06, + "loss": 3.719, + "step": 4023 + }, + { + "epoch": 0.88, + "learning_rate": 4.565872842473237e-06, + "loss": 3.4369, + "step": 4024 + }, + { + "epoch": 0.88, + "learning_rate": 4.565763600611755e-06, + "loss": 2.6926, + "step": 4025 + }, + { + "epoch": 0.88, + "learning_rate": 4.565654358750273e-06, + "loss": 3.5651, + "step": 4026 + }, + { + "epoch": 0.88, + "learning_rate": 4.565545116888792e-06, + "loss": 2.7282, + "step": 4027 + }, + { + "epoch": 0.88, + "learning_rate": 4.565435875027311e-06, + "loss": 3.589, + "step": 4028 + }, + { + "epoch": 0.88, + "learning_rate": 4.56532663316583e-06, + "loss": 3.0616, + "step": 4029 + }, + { + "epoch": 0.88, + "learning_rate": 4.565217391304348e-06, + "loss": 3.3305, + "step": 4030 + }, + { + "epoch": 0.88, + "learning_rate": 4.5651081494428665e-06, + "loss": 3.1317, + "step": 4031 + }, + { + "epoch": 0.88, + "learning_rate": 4.5649989075813855e-06, + "loss": 3.3234, + "step": 4032 + }, + { + "epoch": 0.88, + "learning_rate": 4.5648896657199045e-06, + "loss": 3.4428, + "step": 4033 + }, + { + "epoch": 0.88, + "learning_rate": 4.564780423858423e-06, + "loss": 3.1054, + "step": 4034 + }, + { + "epoch": 0.88, + "learning_rate": 4.564671181996942e-06, + "loss": 3.1641, + "step": 4035 + }, + { + "epoch": 0.88, + "learning_rate": 4.56456194013546e-06, + "loss": 2.8705, + "step": 4036 + }, + { + "epoch": 0.88, + "learning_rate": 4.564452698273979e-06, + "loss": 3.5448, + "step": 4037 + }, + { + "epoch": 0.88, + "learning_rate": 4.564343456412498e-06, + "loss": 3.4244, + "step": 4038 + }, + { + "epoch": 0.88, + "learning_rate": 4.564234214551016e-06, + "loss": 2.94, + "step": 4039 + }, + { + "epoch": 0.88, + "learning_rate": 4.564124972689534e-06, + "loss": 3.0357, + "step": 4040 + }, + { + "epoch": 0.88, + "learning_rate": 4.564015730828053e-06, + "loss": 4.0705, + "step": 4041 + }, + { + "epoch": 0.88, + "learning_rate": 4.5639064889665725e-06, + "loss": 3.9837, + "step": 4042 + }, + { + "epoch": 0.88, + "learning_rate": 4.5637972471050915e-06, + "loss": 3.6305, + "step": 4043 + }, + { + "epoch": 0.88, + "learning_rate": 4.56368800524361e-06, + "loss": 3.6048, + "step": 4044 + }, + { + "epoch": 0.88, + "learning_rate": 4.563578763382128e-06, + "loss": 3.5461, + "step": 4045 + }, + { + "epoch": 0.88, + "learning_rate": 4.563469521520647e-06, + "loss": 2.853, + "step": 4046 + }, + { + "epoch": 0.88, + "learning_rate": 4.563360279659166e-06, + "loss": 2.6306, + "step": 4047 + }, + { + "epoch": 0.88, + "learning_rate": 4.563251037797685e-06, + "loss": 3.17, + "step": 4048 + }, + { + "epoch": 0.88, + "learning_rate": 4.563141795936203e-06, + "loss": 3.0808, + "step": 4049 + }, + { + "epoch": 0.88, + "learning_rate": 4.563032554074721e-06, + "loss": 3.1025, + "step": 4050 + }, + { + "epoch": 0.88, + "learning_rate": 4.56292331221324e-06, + "loss": 2.8229, + "step": 4051 + }, + { + "epoch": 0.88, + "learning_rate": 4.562814070351759e-06, + "loss": 3.209, + "step": 4052 + }, + { + "epoch": 0.88, + "learning_rate": 4.5627048284902784e-06, + "loss": 3.801, + "step": 4053 + }, + { + "epoch": 0.88, + "learning_rate": 4.562595586628797e-06, + "loss": 3.6588, + "step": 4054 + }, + { + "epoch": 0.88, + "learning_rate": 4.562486344767315e-06, + "loss": 3.4449, + "step": 4055 + }, + { + "epoch": 0.89, + "learning_rate": 4.562377102905834e-06, + "loss": 3.5557, + "step": 4056 + }, + { + "epoch": 0.89, + "learning_rate": 4.562267861044353e-06, + "loss": 3.3616, + "step": 4057 + }, + { + "epoch": 0.89, + "learning_rate": 4.562158619182872e-06, + "loss": 2.931, + "step": 4058 + }, + { + "epoch": 0.89, + "learning_rate": 4.56204937732139e-06, + "loss": 3.1822, + "step": 4059 + }, + { + "epoch": 0.89, + "learning_rate": 4.561940135459908e-06, + "loss": 3.8351, + "step": 4060 + }, + { + "epoch": 0.89, + "learning_rate": 4.561830893598427e-06, + "loss": 3.2801, + "step": 4061 + }, + { + "epoch": 0.89, + "learning_rate": 4.561721651736946e-06, + "loss": 3.3871, + "step": 4062 + }, + { + "epoch": 0.89, + "learning_rate": 4.5616124098754646e-06, + "loss": 3.2773, + "step": 4063 + }, + { + "epoch": 0.89, + "learning_rate": 4.561503168013984e-06, + "loss": 3.2621, + "step": 4064 + }, + { + "epoch": 0.89, + "learning_rate": 4.561393926152502e-06, + "loss": 3.2562, + "step": 4065 + }, + { + "epoch": 0.89, + "learning_rate": 4.561284684291021e-06, + "loss": 3.0992, + "step": 4066 + }, + { + "epoch": 0.89, + "learning_rate": 4.56117544242954e-06, + "loss": 3.2518, + "step": 4067 + }, + { + "epoch": 0.89, + "learning_rate": 4.561066200568058e-06, + "loss": 3.004, + "step": 4068 + }, + { + "epoch": 0.89, + "learning_rate": 4.560956958706576e-06, + "loss": 3.5765, + "step": 4069 + }, + { + "epoch": 0.89, + "learning_rate": 4.560847716845095e-06, + "loss": 3.7244, + "step": 4070 + }, + { + "epoch": 0.89, + "learning_rate": 4.560738474983614e-06, + "loss": 3.468, + "step": 4071 + }, + { + "epoch": 0.89, + "learning_rate": 4.560629233122133e-06, + "loss": 3.2819, + "step": 4072 + }, + { + "epoch": 0.89, + "learning_rate": 4.5605199912606515e-06, + "loss": 2.7473, + "step": 4073 + }, + { + "epoch": 0.89, + "learning_rate": 4.56041074939917e-06, + "loss": 3.5946, + "step": 4074 + }, + { + "epoch": 0.89, + "learning_rate": 4.560301507537689e-06, + "loss": 3.2583, + "step": 4075 + }, + { + "epoch": 0.89, + "learning_rate": 4.560192265676208e-06, + "loss": 3.6488, + "step": 4076 + }, + { + "epoch": 0.89, + "learning_rate": 4.560083023814727e-06, + "loss": 3.1059, + "step": 4077 + }, + { + "epoch": 0.89, + "learning_rate": 4.559973781953245e-06, + "loss": 3.7294, + "step": 4078 + }, + { + "epoch": 0.89, + "learning_rate": 4.559864540091763e-06, + "loss": 4.1565, + "step": 4079 + }, + { + "epoch": 0.89, + "learning_rate": 4.559755298230282e-06, + "loss": 3.2638, + "step": 4080 + }, + { + "epoch": 0.89, + "learning_rate": 4.559646056368801e-06, + "loss": 2.8516, + "step": 4081 + }, + { + "epoch": 0.89, + "learning_rate": 4.5595368145073194e-06, + "loss": 2.768, + "step": 4082 + }, + { + "epoch": 0.89, + "learning_rate": 4.5594275726458385e-06, + "loss": 3.3489, + "step": 4083 + }, + { + "epoch": 0.89, + "learning_rate": 4.559318330784357e-06, + "loss": 3.0021, + "step": 4084 + }, + { + "epoch": 0.89, + "learning_rate": 4.559209088922876e-06, + "loss": 3.662, + "step": 4085 + }, + { + "epoch": 0.89, + "learning_rate": 4.559099847061395e-06, + "loss": 3.4388, + "step": 4086 + }, + { + "epoch": 0.89, + "learning_rate": 4.558990605199913e-06, + "loss": 3.4023, + "step": 4087 + }, + { + "epoch": 0.89, + "learning_rate": 4.558881363338432e-06, + "loss": 3.2786, + "step": 4088 + }, + { + "epoch": 0.89, + "learning_rate": 4.55877212147695e-06, + "loss": 3.5083, + "step": 4089 + }, + { + "epoch": 0.89, + "learning_rate": 4.558662879615469e-06, + "loss": 3.8017, + "step": 4090 + }, + { + "epoch": 0.89, + "learning_rate": 4.558553637753988e-06, + "loss": 3.4859, + "step": 4091 + }, + { + "epoch": 0.89, + "learning_rate": 4.558444395892506e-06, + "loss": 3.7311, + "step": 4092 + }, + { + "epoch": 0.89, + "learning_rate": 4.558335154031025e-06, + "loss": 2.9766, + "step": 4093 + }, + { + "epoch": 0.89, + "learning_rate": 4.558225912169544e-06, + "loss": 3.4353, + "step": 4094 + }, + { + "epoch": 0.89, + "learning_rate": 4.558116670308063e-06, + "loss": 3.3511, + "step": 4095 + }, + { + "epoch": 0.89, + "learning_rate": 4.558007428446581e-06, + "loss": 3.4213, + "step": 4096 + }, + { + "epoch": 0.89, + "learning_rate": 4.5578981865851e-06, + "loss": 3.272, + "step": 4097 + }, + { + "epoch": 0.89, + "learning_rate": 4.557788944723618e-06, + "loss": 3.1057, + "step": 4098 + }, + { + "epoch": 0.89, + "learning_rate": 4.557679702862137e-06, + "loss": 3.663, + "step": 4099 + }, + { + "epoch": 0.89, + "learning_rate": 4.557570461000656e-06, + "loss": 3.6796, + "step": 4100 + }, + { + "epoch": 0.9, + "learning_rate": 4.557461219139174e-06, + "loss": 3.3126, + "step": 4101 + }, + { + "epoch": 0.9, + "learning_rate": 4.557351977277693e-06, + "loss": 3.3982, + "step": 4102 + }, + { + "epoch": 0.9, + "learning_rate": 4.5572427354162115e-06, + "loss": 3.9291, + "step": 4103 + }, + { + "epoch": 0.9, + "learning_rate": 4.5571334935547306e-06, + "loss": 3.3249, + "step": 4104 + }, + { + "epoch": 0.9, + "learning_rate": 4.55702425169325e-06, + "loss": 3.3749, + "step": 4105 + }, + { + "epoch": 0.9, + "learning_rate": 4.556915009831768e-06, + "loss": 3.4164, + "step": 4106 + }, + { + "epoch": 0.9, + "learning_rate": 4.556805767970287e-06, + "loss": 3.7423, + "step": 4107 + }, + { + "epoch": 0.9, + "learning_rate": 4.556696526108805e-06, + "loss": 3.4404, + "step": 4108 + }, + { + "epoch": 0.9, + "learning_rate": 4.556587284247324e-06, + "loss": 3.1663, + "step": 4109 + }, + { + "epoch": 0.9, + "learning_rate": 4.556478042385843e-06, + "loss": 2.912, + "step": 4110 + }, + { + "epoch": 0.9, + "learning_rate": 4.556368800524361e-06, + "loss": 2.93, + "step": 4111 + }, + { + "epoch": 0.9, + "learning_rate": 4.55625955866288e-06, + "loss": 3.1301, + "step": 4112 + }, + { + "epoch": 0.9, + "learning_rate": 4.5561503168013985e-06, + "loss": 3.9734, + "step": 4113 + }, + { + "epoch": 0.9, + "learning_rate": 4.5560410749399175e-06, + "loss": 3.968, + "step": 4114 + }, + { + "epoch": 0.9, + "learning_rate": 4.555931833078436e-06, + "loss": 3.0423, + "step": 4115 + }, + { + "epoch": 0.9, + "learning_rate": 4.555822591216955e-06, + "loss": 3.5941, + "step": 4116 + }, + { + "epoch": 0.9, + "learning_rate": 4.555713349355474e-06, + "loss": 3.8669, + "step": 4117 + }, + { + "epoch": 0.9, + "learning_rate": 4.555604107493992e-06, + "loss": 3.4469, + "step": 4118 + }, + { + "epoch": 0.9, + "learning_rate": 4.555494865632511e-06, + "loss": 3.2143, + "step": 4119 + }, + { + "epoch": 0.9, + "learning_rate": 4.555385623771029e-06, + "loss": 3.3345, + "step": 4120 + }, + { + "epoch": 0.9, + "learning_rate": 4.555276381909548e-06, + "loss": 3.4505, + "step": 4121 + }, + { + "epoch": 0.9, + "learning_rate": 4.555167140048066e-06, + "loss": 3.6365, + "step": 4122 + }, + { + "epoch": 0.9, + "learning_rate": 4.5550578981865854e-06, + "loss": 3.2905, + "step": 4123 + }, + { + "epoch": 0.9, + "learning_rate": 4.5549486563251045e-06, + "loss": 3.6767, + "step": 4124 + }, + { + "epoch": 0.9, + "learning_rate": 4.554839414463623e-06, + "loss": 3.719, + "step": 4125 + }, + { + "epoch": 0.9, + "learning_rate": 4.554730172602142e-06, + "loss": 2.6789, + "step": 4126 + }, + { + "epoch": 0.9, + "learning_rate": 4.55462093074066e-06, + "loss": 3.2878, + "step": 4127 + }, + { + "epoch": 0.9, + "learning_rate": 4.554511688879179e-06, + "loss": 3.4821, + "step": 4128 + }, + { + "epoch": 0.9, + "learning_rate": 4.554402447017697e-06, + "loss": 3.7606, + "step": 4129 + }, + { + "epoch": 0.9, + "learning_rate": 4.554293205156216e-06, + "loss": 3.2267, + "step": 4130 + }, + { + "epoch": 0.9, + "learning_rate": 4.554183963294735e-06, + "loss": 3.9458, + "step": 4131 + }, + { + "epoch": 0.9, + "learning_rate": 4.554074721433253e-06, + "loss": 3.603, + "step": 4132 + }, + { + "epoch": 0.9, + "learning_rate": 4.553965479571772e-06, + "loss": 3.5757, + "step": 4133 + }, + { + "epoch": 0.9, + "learning_rate": 4.553856237710291e-06, + "loss": 3.4073, + "step": 4134 + }, + { + "epoch": 0.9, + "learning_rate": 4.55374699584881e-06, + "loss": 3.9884, + "step": 4135 + }, + { + "epoch": 0.9, + "learning_rate": 4.553637753987329e-06, + "loss": 3.7698, + "step": 4136 + }, + { + "epoch": 0.9, + "learning_rate": 4.553528512125847e-06, + "loss": 3.1213, + "step": 4137 + }, + { + "epoch": 0.9, + "learning_rate": 4.553419270264366e-06, + "loss": 3.2029, + "step": 4138 + }, + { + "epoch": 0.9, + "learning_rate": 4.553310028402884e-06, + "loss": 3.4297, + "step": 4139 + }, + { + "epoch": 0.9, + "learning_rate": 4.553200786541403e-06, + "loss": 3.1867, + "step": 4140 + }, + { + "epoch": 0.9, + "learning_rate": 4.553091544679922e-06, + "loss": 3.1565, + "step": 4141 + }, + { + "epoch": 0.9, + "learning_rate": 4.55298230281844e-06, + "loss": 3.8966, + "step": 4142 + }, + { + "epoch": 0.9, + "learning_rate": 4.552873060956959e-06, + "loss": 3.7791, + "step": 4143 + }, + { + "epoch": 0.9, + "learning_rate": 4.5527638190954775e-06, + "loss": 3.6443, + "step": 4144 + }, + { + "epoch": 0.9, + "learning_rate": 4.552654577233997e-06, + "loss": 3.0499, + "step": 4145 + }, + { + "epoch": 0.9, + "learning_rate": 4.552545335372516e-06, + "loss": 3.3974, + "step": 4146 + }, + { + "epoch": 0.91, + "learning_rate": 4.552436093511034e-06, + "loss": 2.6923, + "step": 4147 + }, + { + "epoch": 0.91, + "learning_rate": 4.552326851649552e-06, + "loss": 2.7491, + "step": 4148 + }, + { + "epoch": 0.91, + "learning_rate": 4.552217609788071e-06, + "loss": 3.4711, + "step": 4149 + }, + { + "epoch": 0.91, + "learning_rate": 4.55210836792659e-06, + "loss": 3.5692, + "step": 4150 + }, + { + "epoch": 0.91, + "learning_rate": 4.551999126065108e-06, + "loss": 3.8341, + "step": 4151 + }, + { + "epoch": 0.91, + "learning_rate": 4.551889884203627e-06, + "loss": 2.6485, + "step": 4152 + }, + { + "epoch": 0.91, + "learning_rate": 4.5517806423421455e-06, + "loss": 3.2923, + "step": 4153 + }, + { + "epoch": 0.91, + "learning_rate": 4.5516714004806645e-06, + "loss": 3.4978, + "step": 4154 + }, + { + "epoch": 0.91, + "learning_rate": 4.5515621586191835e-06, + "loss": 3.2954, + "step": 4155 + }, + { + "epoch": 0.91, + "learning_rate": 4.551452916757702e-06, + "loss": 3.9949, + "step": 4156 + }, + { + "epoch": 0.91, + "learning_rate": 4.551343674896221e-06, + "loss": 3.5546, + "step": 4157 + }, + { + "epoch": 0.91, + "learning_rate": 4.551234433034739e-06, + "loss": 2.9976, + "step": 4158 + }, + { + "epoch": 0.91, + "learning_rate": 4.551125191173258e-06, + "loss": 3.2257, + "step": 4159 + }, + { + "epoch": 0.91, + "learning_rate": 4.551015949311777e-06, + "loss": 3.8096, + "step": 4160 + }, + { + "epoch": 0.91, + "learning_rate": 4.550906707450295e-06, + "loss": 3.0363, + "step": 4161 + }, + { + "epoch": 0.91, + "learning_rate": 4.550797465588813e-06, + "loss": 3.1644, + "step": 4162 + }, + { + "epoch": 0.91, + "learning_rate": 4.5506882237273324e-06, + "loss": 3.1684, + "step": 4163 + }, + { + "epoch": 0.91, + "learning_rate": 4.5505789818658515e-06, + "loss": 2.7148, + "step": 4164 + }, + { + "epoch": 0.91, + "learning_rate": 4.5504697400043705e-06, + "loss": 3.4461, + "step": 4165 + }, + { + "epoch": 0.91, + "learning_rate": 4.550360498142889e-06, + "loss": 3.4909, + "step": 4166 + }, + { + "epoch": 0.91, + "learning_rate": 4.550251256281407e-06, + "loss": 3.9521, + "step": 4167 + }, + { + "epoch": 0.91, + "learning_rate": 4.550142014419926e-06, + "loss": 3.9481, + "step": 4168 + }, + { + "epoch": 0.91, + "learning_rate": 4.550032772558445e-06, + "loss": 3.2514, + "step": 4169 + }, + { + "epoch": 0.91, + "learning_rate": 4.549923530696964e-06, + "loss": 3.399, + "step": 4170 + }, + { + "epoch": 0.91, + "learning_rate": 4.549814288835482e-06, + "loss": 3.4915, + "step": 4171 + }, + { + "epoch": 0.91, + "learning_rate": 4.549705046974e-06, + "loss": 3.0615, + "step": 4172 + }, + { + "epoch": 0.91, + "learning_rate": 4.549595805112519e-06, + "loss": 3.5732, + "step": 4173 + }, + { + "epoch": 0.91, + "learning_rate": 4.549486563251038e-06, + "loss": 3.432, + "step": 4174 + }, + { + "epoch": 0.91, + "learning_rate": 4.549377321389557e-06, + "loss": 3.5903, + "step": 4175 + }, + { + "epoch": 0.91, + "learning_rate": 4.549268079528076e-06, + "loss": 3.1953, + "step": 4176 + }, + { + "epoch": 0.91, + "learning_rate": 4.549158837666594e-06, + "loss": 3.2444, + "step": 4177 + }, + { + "epoch": 0.91, + "learning_rate": 4.549049595805113e-06, + "loss": 3.649, + "step": 4178 + }, + { + "epoch": 0.91, + "learning_rate": 4.548940353943632e-06, + "loss": 2.9069, + "step": 4179 + }, + { + "epoch": 0.91, + "learning_rate": 4.54883111208215e-06, + "loss": 3.2639, + "step": 4180 + }, + { + "epoch": 0.91, + "learning_rate": 4.548721870220668e-06, + "loss": 3.2659, + "step": 4181 + }, + { + "epoch": 0.91, + "learning_rate": 4.548612628359187e-06, + "loss": 2.9636, + "step": 4182 + }, + { + "epoch": 0.91, + "learning_rate": 4.548503386497706e-06, + "loss": 2.888, + "step": 4183 + }, + { + "epoch": 0.91, + "learning_rate": 4.548394144636225e-06, + "loss": 3.5714, + "step": 4184 + }, + { + "epoch": 0.91, + "learning_rate": 4.5482849027747436e-06, + "loss": 3.2894, + "step": 4185 + }, + { + "epoch": 0.91, + "learning_rate": 4.548175660913262e-06, + "loss": 2.9382, + "step": 4186 + }, + { + "epoch": 0.91, + "learning_rate": 4.548066419051781e-06, + "loss": 3.3002, + "step": 4187 + }, + { + "epoch": 0.91, + "learning_rate": 4.5479571771903e-06, + "loss": 3.8879, + "step": 4188 + }, + { + "epoch": 0.91, + "learning_rate": 4.547847935328819e-06, + "loss": 3.5177, + "step": 4189 + }, + { + "epoch": 0.91, + "learning_rate": 4.547738693467337e-06, + "loss": 4.0903, + "step": 4190 + }, + { + "epoch": 0.91, + "learning_rate": 4.547629451605855e-06, + "loss": 3.5983, + "step": 4191 + }, + { + "epoch": 0.91, + "learning_rate": 4.547520209744374e-06, + "loss": 3.3384, + "step": 4192 + }, + { + "epoch": 0.92, + "learning_rate": 4.547410967882893e-06, + "loss": 3.3747, + "step": 4193 + }, + { + "epoch": 0.92, + "learning_rate": 4.547301726021412e-06, + "loss": 2.5179, + "step": 4194 + }, + { + "epoch": 0.92, + "learning_rate": 4.5471924841599305e-06, + "loss": 3.624, + "step": 4195 + }, + { + "epoch": 0.92, + "learning_rate": 4.547083242298449e-06, + "loss": 3.4632, + "step": 4196 + }, + { + "epoch": 0.92, + "learning_rate": 4.546974000436968e-06, + "loss": 3.2675, + "step": 4197 + }, + { + "epoch": 0.92, + "learning_rate": 4.546864758575487e-06, + "loss": 3.8854, + "step": 4198 + }, + { + "epoch": 0.92, + "learning_rate": 4.546755516714006e-06, + "loss": 3.3591, + "step": 4199 + }, + { + "epoch": 0.92, + "learning_rate": 4.546646274852524e-06, + "loss": 3.0436, + "step": 4200 + }, + { + "epoch": 0.92, + "learning_rate": 4.546537032991042e-06, + "loss": 3.3398, + "step": 4201 + }, + { + "epoch": 0.92, + "learning_rate": 4.546427791129561e-06, + "loss": 3.1818, + "step": 4202 + }, + { + "epoch": 0.92, + "learning_rate": 4.54631854926808e-06, + "loss": 3.3615, + "step": 4203 + }, + { + "epoch": 0.92, + "learning_rate": 4.5462093074065984e-06, + "loss": 3.5714, + "step": 4204 + }, + { + "epoch": 0.92, + "learning_rate": 4.5461000655451175e-06, + "loss": 3.0948, + "step": 4205 + }, + { + "epoch": 0.92, + "learning_rate": 4.545990823683636e-06, + "loss": 3.04, + "step": 4206 + }, + { + "epoch": 0.92, + "learning_rate": 4.545881581822155e-06, + "loss": 3.4696, + "step": 4207 + }, + { + "epoch": 0.92, + "learning_rate": 4.545772339960674e-06, + "loss": 3.4495, + "step": 4208 + }, + { + "epoch": 0.92, + "learning_rate": 4.545663098099192e-06, + "loss": 3.0129, + "step": 4209 + }, + { + "epoch": 0.92, + "learning_rate": 4.54555385623771e-06, + "loss": 2.7287, + "step": 4210 + }, + { + "epoch": 0.92, + "learning_rate": 4.545444614376229e-06, + "loss": 3.7134, + "step": 4211 + }, + { + "epoch": 0.92, + "learning_rate": 4.545335372514748e-06, + "loss": 3.7716, + "step": 4212 + }, + { + "epoch": 0.92, + "learning_rate": 4.545226130653267e-06, + "loss": 3.3211, + "step": 4213 + }, + { + "epoch": 0.92, + "learning_rate": 4.545116888791785e-06, + "loss": 3.5747, + "step": 4214 + }, + { + "epoch": 0.92, + "learning_rate": 4.545007646930304e-06, + "loss": 3.4342, + "step": 4215 + }, + { + "epoch": 0.92, + "learning_rate": 4.544898405068823e-06, + "loss": 3.6153, + "step": 4216 + }, + { + "epoch": 0.92, + "learning_rate": 4.544789163207342e-06, + "loss": 3.4866, + "step": 4217 + }, + { + "epoch": 0.92, + "learning_rate": 4.544679921345861e-06, + "loss": 3.5792, + "step": 4218 + }, + { + "epoch": 0.92, + "learning_rate": 4.544570679484379e-06, + "loss": 2.95, + "step": 4219 + }, + { + "epoch": 0.92, + "learning_rate": 4.544461437622897e-06, + "loss": 3.68, + "step": 4220 + }, + { + "epoch": 0.92, + "learning_rate": 4.544352195761416e-06, + "loss": 3.3997, + "step": 4221 + }, + { + "epoch": 0.92, + "learning_rate": 4.544242953899935e-06, + "loss": 3.4452, + "step": 4222 + }, + { + "epoch": 0.92, + "learning_rate": 4.544133712038454e-06, + "loss": 3.2668, + "step": 4223 + }, + { + "epoch": 0.92, + "learning_rate": 4.544024470176972e-06, + "loss": 3.3485, + "step": 4224 + }, + { + "epoch": 0.92, + "learning_rate": 4.5439152283154905e-06, + "loss": 3.4154, + "step": 4225 + }, + { + "epoch": 0.92, + "learning_rate": 4.54380598645401e-06, + "loss": 3.8524, + "step": 4226 + }, + { + "epoch": 0.92, + "learning_rate": 4.543696744592529e-06, + "loss": 3.249, + "step": 4227 + }, + { + "epoch": 0.92, + "learning_rate": 4.543587502731047e-06, + "loss": 3.1603, + "step": 4228 + }, + { + "epoch": 0.92, + "learning_rate": 4.543478260869566e-06, + "loss": 2.9129, + "step": 4229 + }, + { + "epoch": 0.92, + "learning_rate": 4.543369019008084e-06, + "loss": 3.8041, + "step": 4230 + }, + { + "epoch": 0.92, + "learning_rate": 4.543259777146603e-06, + "loss": 3.146, + "step": 4231 + }, + { + "epoch": 0.92, + "learning_rate": 4.543150535285122e-06, + "loss": 3.4072, + "step": 4232 + }, + { + "epoch": 0.92, + "learning_rate": 4.54304129342364e-06, + "loss": 3.3252, + "step": 4233 + }, + { + "epoch": 0.92, + "learning_rate": 4.5429320515621585e-06, + "loss": 3.3486, + "step": 4234 + }, + { + "epoch": 0.92, + "learning_rate": 4.5428228097006775e-06, + "loss": 3.4777, + "step": 4235 + }, + { + "epoch": 0.92, + "learning_rate": 4.5427135678391965e-06, + "loss": 3.1122, + "step": 4236 + }, + { + "epoch": 0.92, + "learning_rate": 4.5426043259777156e-06, + "loss": 3.7449, + "step": 4237 + }, + { + "epoch": 0.92, + "learning_rate": 4.542495084116234e-06, + "loss": 3.4175, + "step": 4238 + }, + { + "epoch": 0.93, + "learning_rate": 4.542385842254752e-06, + "loss": 3.4267, + "step": 4239 + }, + { + "epoch": 0.93, + "learning_rate": 4.542276600393271e-06, + "loss": 3.7713, + "step": 4240 + }, + { + "epoch": 0.93, + "learning_rate": 4.54216735853179e-06, + "loss": 3.8507, + "step": 4241 + }, + { + "epoch": 0.93, + "learning_rate": 4.542058116670308e-06, + "loss": 2.9558, + "step": 4242 + }, + { + "epoch": 0.93, + "learning_rate": 4.541948874808827e-06, + "loss": 3.0847, + "step": 4243 + }, + { + "epoch": 0.93, + "learning_rate": 4.5418396329473454e-06, + "loss": 3.7014, + "step": 4244 + }, + { + "epoch": 0.93, + "learning_rate": 4.5417303910858645e-06, + "loss": 2.723, + "step": 4245 + }, + { + "epoch": 0.93, + "learning_rate": 4.5416211492243835e-06, + "loss": 3.4721, + "step": 4246 + }, + { + "epoch": 0.93, + "learning_rate": 4.541511907362902e-06, + "loss": 3.3526, + "step": 4247 + }, + { + "epoch": 0.93, + "learning_rate": 4.541402665501421e-06, + "loss": 3.0635, + "step": 4248 + }, + { + "epoch": 0.93, + "learning_rate": 4.541293423639939e-06, + "loss": 3.7428, + "step": 4249 + }, + { + "epoch": 0.93, + "learning_rate": 4.541184181778458e-06, + "loss": 3.7807, + "step": 4250 + }, + { + "epoch": 0.93, + "learning_rate": 4.541074939916977e-06, + "loss": 3.3847, + "step": 4251 + }, + { + "epoch": 0.93, + "learning_rate": 4.540965698055495e-06, + "loss": 2.7094, + "step": 4252 + }, + { + "epoch": 0.93, + "learning_rate": 4.540856456194014e-06, + "loss": 3.2454, + "step": 4253 + }, + { + "epoch": 0.93, + "learning_rate": 4.540747214332532e-06, + "loss": 3.6629, + "step": 4254 + }, + { + "epoch": 0.93, + "learning_rate": 4.540637972471051e-06, + "loss": 2.8382, + "step": 4255 + }, + { + "epoch": 0.93, + "learning_rate": 4.5405287306095705e-06, + "loss": 3.0824, + "step": 4256 + }, + { + "epoch": 0.93, + "learning_rate": 4.540419488748089e-06, + "loss": 2.9996, + "step": 4257 + }, + { + "epoch": 0.93, + "learning_rate": 4.540310246886608e-06, + "loss": 3.0034, + "step": 4258 + }, + { + "epoch": 0.93, + "learning_rate": 4.540201005025126e-06, + "loss": 3.3932, + "step": 4259 + }, + { + "epoch": 0.93, + "learning_rate": 4.540091763163645e-06, + "loss": 2.7385, + "step": 4260 + }, + { + "epoch": 0.93, + "learning_rate": 4.539982521302163e-06, + "loss": 3.641, + "step": 4261 + }, + { + "epoch": 0.93, + "learning_rate": 4.539873279440682e-06, + "loss": 3.8256, + "step": 4262 + }, + { + "epoch": 0.93, + "learning_rate": 4.5397640375792e-06, + "loss": 3.1052, + "step": 4263 + }, + { + "epoch": 0.93, + "learning_rate": 4.539654795717719e-06, + "loss": 3.4786, + "step": 4264 + }, + { + "epoch": 0.93, + "learning_rate": 4.539545553856238e-06, + "loss": 3.5816, + "step": 4265 + }, + { + "epoch": 0.93, + "learning_rate": 4.5394363119947566e-06, + "loss": 3.1755, + "step": 4266 + }, + { + "epoch": 0.93, + "learning_rate": 4.539327070133276e-06, + "loss": 3.2494, + "step": 4267 + }, + { + "epoch": 0.93, + "learning_rate": 4.539217828271794e-06, + "loss": 4.0501, + "step": 4268 + }, + { + "epoch": 0.93, + "learning_rate": 4.539108586410313e-06, + "loss": 3.379, + "step": 4269 + }, + { + "epoch": 0.93, + "learning_rate": 4.538999344548832e-06, + "loss": 3.4683, + "step": 4270 + }, + { + "epoch": 0.93, + "learning_rate": 4.53889010268735e-06, + "loss": 2.7962, + "step": 4271 + }, + { + "epoch": 0.93, + "learning_rate": 4.538780860825869e-06, + "loss": 2.8309, + "step": 4272 + }, + { + "epoch": 0.93, + "learning_rate": 4.538671618964387e-06, + "loss": 3.2229, + "step": 4273 + }, + { + "epoch": 0.93, + "learning_rate": 4.538562377102906e-06, + "loss": 3.4036, + "step": 4274 + }, + { + "epoch": 0.93, + "learning_rate": 4.5384531352414245e-06, + "loss": 3.4449, + "step": 4275 + }, + { + "epoch": 0.93, + "learning_rate": 4.5383438933799435e-06, + "loss": 3.6177, + "step": 4276 + }, + { + "epoch": 0.93, + "learning_rate": 4.5382346515184626e-06, + "loss": 3.3461, + "step": 4277 + }, + { + "epoch": 0.93, + "learning_rate": 4.538125409656981e-06, + "loss": 2.9779, + "step": 4278 + }, + { + "epoch": 0.93, + "learning_rate": 4.5380161677955e-06, + "loss": 3.2313, + "step": 4279 + }, + { + "epoch": 0.93, + "learning_rate": 4.537906925934018e-06, + "loss": 3.57, + "step": 4280 + }, + { + "epoch": 0.93, + "learning_rate": 4.537797684072537e-06, + "loss": 3.6972, + "step": 4281 + }, + { + "epoch": 0.93, + "learning_rate": 4.537688442211056e-06, + "loss": 2.6915, + "step": 4282 + }, + { + "epoch": 0.93, + "learning_rate": 4.537579200349574e-06, + "loss": 2.8316, + "step": 4283 + }, + { + "epoch": 0.93, + "learning_rate": 4.537469958488093e-06, + "loss": 3.463, + "step": 4284 + }, + { + "epoch": 0.94, + "learning_rate": 4.5373607166266114e-06, + "loss": 3.5134, + "step": 4285 + }, + { + "epoch": 0.94, + "learning_rate": 4.5372514747651305e-06, + "loss": 3.3217, + "step": 4286 + }, + { + "epoch": 0.94, + "learning_rate": 4.537142232903649e-06, + "loss": 3.611, + "step": 4287 + }, + { + "epoch": 0.94, + "learning_rate": 4.537032991042168e-06, + "loss": 2.9754, + "step": 4288 + }, + { + "epoch": 0.94, + "learning_rate": 4.536923749180687e-06, + "loss": 2.4583, + "step": 4289 + }, + { + "epoch": 0.94, + "learning_rate": 4.536814507319205e-06, + "loss": 3.24, + "step": 4290 + }, + { + "epoch": 0.94, + "learning_rate": 4.536705265457724e-06, + "loss": 3.9144, + "step": 4291 + }, + { + "epoch": 0.94, + "learning_rate": 4.536596023596242e-06, + "loss": 2.6828, + "step": 4292 + }, + { + "epoch": 0.94, + "learning_rate": 4.536486781734761e-06, + "loss": 2.7353, + "step": 4293 + }, + { + "epoch": 0.94, + "learning_rate": 4.536377539873279e-06, + "loss": 3.2481, + "step": 4294 + }, + { + "epoch": 0.94, + "learning_rate": 4.536268298011798e-06, + "loss": 3.3971, + "step": 4295 + }, + { + "epoch": 0.94, + "learning_rate": 4.5361590561503174e-06, + "loss": 3.653, + "step": 4296 + }, + { + "epoch": 0.94, + "learning_rate": 4.536049814288836e-06, + "loss": 3.4938, + "step": 4297 + }, + { + "epoch": 0.94, + "learning_rate": 4.535940572427355e-06, + "loss": 3.128, + "step": 4298 + }, + { + "epoch": 0.94, + "learning_rate": 4.535831330565873e-06, + "loss": 3.4713, + "step": 4299 + }, + { + "epoch": 0.94, + "learning_rate": 4.535722088704392e-06, + "loss": 2.528, + "step": 4300 + }, + { + "epoch": 0.94, + "learning_rate": 4.535612846842911e-06, + "loss": 3.3485, + "step": 4301 + }, + { + "epoch": 0.94, + "learning_rate": 4.535503604981429e-06, + "loss": 2.4064, + "step": 4302 + }, + { + "epoch": 0.94, + "learning_rate": 4.535394363119948e-06, + "loss": 3.1372, + "step": 4303 + }, + { + "epoch": 0.94, + "learning_rate": 4.535285121258466e-06, + "loss": 3.0162, + "step": 4304 + }, + { + "epoch": 0.94, + "learning_rate": 4.535175879396985e-06, + "loss": 2.903, + "step": 4305 + }, + { + "epoch": 0.94, + "learning_rate": 4.535066637535504e-06, + "loss": 3.6408, + "step": 4306 + }, + { + "epoch": 0.94, + "learning_rate": 4.5349573956740226e-06, + "loss": 3.7637, + "step": 4307 + }, + { + "epoch": 0.94, + "learning_rate": 4.534848153812541e-06, + "loss": 3.6279, + "step": 4308 + }, + { + "epoch": 0.94, + "learning_rate": 4.53473891195106e-06, + "loss": 3.062, + "step": 4309 + }, + { + "epoch": 0.94, + "learning_rate": 4.534629670089579e-06, + "loss": 2.6007, + "step": 4310 + }, + { + "epoch": 0.94, + "learning_rate": 4.534520428228098e-06, + "loss": 3.4367, + "step": 4311 + }, + { + "epoch": 0.94, + "learning_rate": 4.534411186366616e-06, + "loss": 3.3565, + "step": 4312 + }, + { + "epoch": 0.94, + "learning_rate": 4.534301944505134e-06, + "loss": 3.4437, + "step": 4313 + }, + { + "epoch": 0.94, + "learning_rate": 4.534192702643653e-06, + "loss": 3.6793, + "step": 4314 + }, + { + "epoch": 0.94, + "learning_rate": 4.534083460782172e-06, + "loss": 3.4652, + "step": 4315 + }, + { + "epoch": 0.94, + "learning_rate": 4.5339742189206905e-06, + "loss": 3.363, + "step": 4316 + }, + { + "epoch": 0.94, + "learning_rate": 4.5338649770592095e-06, + "loss": 3.2038, + "step": 4317 + }, + { + "epoch": 0.94, + "learning_rate": 4.533755735197728e-06, + "loss": 3.1628, + "step": 4318 + }, + { + "epoch": 0.94, + "learning_rate": 4.533646493336247e-06, + "loss": 3.3235, + "step": 4319 + }, + { + "epoch": 0.94, + "learning_rate": 4.533537251474766e-06, + "loss": 3.8534, + "step": 4320 + }, + { + "epoch": 0.94, + "learning_rate": 4.533428009613284e-06, + "loss": 4.0096, + "step": 4321 + }, + { + "epoch": 0.94, + "learning_rate": 4.533318767751803e-06, + "loss": 3.1539, + "step": 4322 + }, + { + "epoch": 0.94, + "learning_rate": 4.533209525890321e-06, + "loss": 3.3623, + "step": 4323 + }, + { + "epoch": 0.94, + "learning_rate": 4.53310028402884e-06, + "loss": 3.636, + "step": 4324 + }, + { + "epoch": 0.94, + "learning_rate": 4.532991042167359e-06, + "loss": 3.5201, + "step": 4325 + }, + { + "epoch": 0.94, + "learning_rate": 4.5328818003058775e-06, + "loss": 2.967, + "step": 4326 + }, + { + "epoch": 0.94, + "learning_rate": 4.532772558444396e-06, + "loss": 3.7752, + "step": 4327 + }, + { + "epoch": 0.94, + "learning_rate": 4.532663316582915e-06, + "loss": 3.8412, + "step": 4328 + }, + { + "epoch": 0.94, + "learning_rate": 4.532554074721434e-06, + "loss": 2.9877, + "step": 4329 + }, + { + "epoch": 0.95, + "learning_rate": 4.532444832859953e-06, + "loss": 3.6325, + "step": 4330 + }, + { + "epoch": 0.95, + "learning_rate": 4.532335590998471e-06, + "loss": 3.1363, + "step": 4331 + }, + { + "epoch": 0.95, + "learning_rate": 4.532226349136989e-06, + "loss": 3.0753, + "step": 4332 + }, + { + "epoch": 0.95, + "learning_rate": 4.532117107275508e-06, + "loss": 3.3238, + "step": 4333 + }, + { + "epoch": 0.95, + "learning_rate": 4.532007865414027e-06, + "loss": 3.4637, + "step": 4334 + }, + { + "epoch": 0.95, + "learning_rate": 4.531898623552546e-06, + "loss": 3.2593, + "step": 4335 + }, + { + "epoch": 0.95, + "learning_rate": 4.531789381691064e-06, + "loss": 3.6883, + "step": 4336 + }, + { + "epoch": 0.95, + "learning_rate": 4.531680139829583e-06, + "loss": 3.3011, + "step": 4337 + }, + { + "epoch": 0.95, + "learning_rate": 4.531570897968102e-06, + "loss": 2.4513, + "step": 4338 + }, + { + "epoch": 0.95, + "learning_rate": 4.531461656106621e-06, + "loss": 2.4792, + "step": 4339 + }, + { + "epoch": 0.95, + "learning_rate": 4.531352414245139e-06, + "loss": 2.9643, + "step": 4340 + }, + { + "epoch": 0.95, + "learning_rate": 4.531243172383658e-06, + "loss": 3.7652, + "step": 4341 + }, + { + "epoch": 0.95, + "learning_rate": 4.531133930522176e-06, + "loss": 2.8454, + "step": 4342 + }, + { + "epoch": 0.95, + "learning_rate": 4.531024688660695e-06, + "loss": 3.3233, + "step": 4343 + }, + { + "epoch": 0.95, + "learning_rate": 4.530915446799214e-06, + "loss": 3.8717, + "step": 4344 + }, + { + "epoch": 0.95, + "learning_rate": 4.530806204937732e-06, + "loss": 3.2959, + "step": 4345 + }, + { + "epoch": 0.95, + "learning_rate": 4.530696963076251e-06, + "loss": 3.1725, + "step": 4346 + }, + { + "epoch": 0.95, + "learning_rate": 4.5305877212147696e-06, + "loss": 4.0919, + "step": 4347 + }, + { + "epoch": 0.95, + "learning_rate": 4.530478479353289e-06, + "loss": 3.55, + "step": 4348 + }, + { + "epoch": 0.95, + "learning_rate": 4.530369237491808e-06, + "loss": 3.115, + "step": 4349 + }, + { + "epoch": 0.95, + "learning_rate": 4.530259995630326e-06, + "loss": 3.3366, + "step": 4350 + }, + { + "epoch": 0.95, + "learning_rate": 4.530150753768844e-06, + "loss": 3.4909, + "step": 4351 + }, + { + "epoch": 0.95, + "learning_rate": 4.530041511907363e-06, + "loss": 3.2541, + "step": 4352 + }, + { + "epoch": 0.95, + "learning_rate": 4.529932270045882e-06, + "loss": 3.4501, + "step": 4353 + }, + { + "epoch": 0.95, + "learning_rate": 4.529823028184401e-06, + "loss": 3.3273, + "step": 4354 + }, + { + "epoch": 0.95, + "learning_rate": 4.529713786322919e-06, + "loss": 2.6638, + "step": 4355 + }, + { + "epoch": 0.95, + "learning_rate": 4.5296045444614375e-06, + "loss": 3.8233, + "step": 4356 + }, + { + "epoch": 0.95, + "learning_rate": 4.5294953025999565e-06, + "loss": 3.307, + "step": 4357 + }, + { + "epoch": 0.95, + "learning_rate": 4.5293860607384755e-06, + "loss": 3.4553, + "step": 4358 + }, + { + "epoch": 0.95, + "learning_rate": 4.529276818876995e-06, + "loss": 2.8021, + "step": 4359 + }, + { + "epoch": 0.95, + "learning_rate": 4.529167577015513e-06, + "loss": 3.1775, + "step": 4360 + }, + { + "epoch": 0.95, + "learning_rate": 4.529058335154031e-06, + "loss": 3.3159, + "step": 4361 + }, + { + "epoch": 0.95, + "learning_rate": 4.52894909329255e-06, + "loss": 3.213, + "step": 4362 + }, + { + "epoch": 0.95, + "learning_rate": 4.528839851431069e-06, + "loss": 2.8998, + "step": 4363 + }, + { + "epoch": 0.95, + "learning_rate": 4.528730609569588e-06, + "loss": 3.1791, + "step": 4364 + }, + { + "epoch": 0.95, + "learning_rate": 4.528621367708106e-06, + "loss": 3.3736, + "step": 4365 + }, + { + "epoch": 0.95, + "learning_rate": 4.5285121258466244e-06, + "loss": 3.3251, + "step": 4366 + }, + { + "epoch": 0.95, + "learning_rate": 4.5284028839851435e-06, + "loss": 3.3878, + "step": 4367 + }, + { + "epoch": 0.95, + "learning_rate": 4.5282936421236625e-06, + "loss": 3.3937, + "step": 4368 + }, + { + "epoch": 0.95, + "learning_rate": 4.528184400262181e-06, + "loss": 2.857, + "step": 4369 + }, + { + "epoch": 0.95, + "learning_rate": 4.5280751584007e-06, + "loss": 3.4445, + "step": 4370 + }, + { + "epoch": 0.95, + "learning_rate": 4.527965916539218e-06, + "loss": 3.7737, + "step": 4371 + }, + { + "epoch": 0.95, + "learning_rate": 4.527856674677737e-06, + "loss": 3.0536, + "step": 4372 + }, + { + "epoch": 0.95, + "learning_rate": 4.527747432816256e-06, + "loss": 3.1426, + "step": 4373 + }, + { + "epoch": 0.95, + "learning_rate": 4.527638190954774e-06, + "loss": 3.613, + "step": 4374 + }, + { + "epoch": 0.95, + "learning_rate": 4.527528949093292e-06, + "loss": 3.4325, + "step": 4375 + }, + { + "epoch": 0.96, + "learning_rate": 4.527419707231811e-06, + "loss": 3.5426, + "step": 4376 + }, + { + "epoch": 0.96, + "learning_rate": 4.5273104653703304e-06, + "loss": 3.36, + "step": 4377 + }, + { + "epoch": 0.96, + "learning_rate": 4.5272012235088495e-06, + "loss": 3.4142, + "step": 4378 + }, + { + "epoch": 0.96, + "learning_rate": 4.527091981647368e-06, + "loss": 3.0796, + "step": 4379 + }, + { + "epoch": 0.96, + "learning_rate": 4.526982739785886e-06, + "loss": 3.8315, + "step": 4380 + }, + { + "epoch": 0.96, + "learning_rate": 4.526873497924405e-06, + "loss": 2.9289, + "step": 4381 + }, + { + "epoch": 0.96, + "learning_rate": 4.526764256062924e-06, + "loss": 3.2837, + "step": 4382 + }, + { + "epoch": 0.96, + "learning_rate": 4.526655014201443e-06, + "loss": 3.2727, + "step": 4383 + }, + { + "epoch": 0.96, + "learning_rate": 4.526545772339961e-06, + "loss": 3.3952, + "step": 4384 + }, + { + "epoch": 0.96, + "learning_rate": 4.526436530478479e-06, + "loss": 3.3152, + "step": 4385 + }, + { + "epoch": 0.96, + "learning_rate": 4.526327288616998e-06, + "loss": 2.8189, + "step": 4386 + }, + { + "epoch": 0.96, + "learning_rate": 4.526218046755517e-06, + "loss": 3.5186, + "step": 4387 + }, + { + "epoch": 0.96, + "learning_rate": 4.526108804894036e-06, + "loss": 3.052, + "step": 4388 + }, + { + "epoch": 0.96, + "learning_rate": 4.525999563032555e-06, + "loss": 3.2559, + "step": 4389 + }, + { + "epoch": 0.96, + "learning_rate": 4.525890321171073e-06, + "loss": 3.0017, + "step": 4390 + }, + { + "epoch": 0.96, + "learning_rate": 4.525781079309592e-06, + "loss": 3.7527, + "step": 4391 + }, + { + "epoch": 0.96, + "learning_rate": 4.525671837448111e-06, + "loss": 2.9382, + "step": 4392 + }, + { + "epoch": 0.96, + "learning_rate": 4.525562595586629e-06, + "loss": 3.9459, + "step": 4393 + }, + { + "epoch": 0.96, + "learning_rate": 4.525453353725148e-06, + "loss": 2.5108, + "step": 4394 + }, + { + "epoch": 0.96, + "learning_rate": 4.525344111863666e-06, + "loss": 2.6044, + "step": 4395 + }, + { + "epoch": 0.96, + "learning_rate": 4.525234870002185e-06, + "loss": 3.0859, + "step": 4396 + }, + { + "epoch": 0.96, + "learning_rate": 4.525125628140704e-06, + "loss": 3.2615, + "step": 4397 + }, + { + "epoch": 0.96, + "learning_rate": 4.5250163862792225e-06, + "loss": 3.3996, + "step": 4398 + }, + { + "epoch": 0.96, + "learning_rate": 4.5249071444177416e-06, + "loss": 3.2284, + "step": 4399 + }, + { + "epoch": 0.96, + "learning_rate": 4.52479790255626e-06, + "loss": 3.7142, + "step": 4400 + }, + { + "epoch": 0.96, + "learning_rate": 4.524688660694779e-06, + "loss": 2.7736, + "step": 4401 + }, + { + "epoch": 0.96, + "learning_rate": 4.524579418833298e-06, + "loss": 3.8049, + "step": 4402 + }, + { + "epoch": 0.96, + "learning_rate": 4.524470176971816e-06, + "loss": 3.1374, + "step": 4403 + }, + { + "epoch": 0.96, + "learning_rate": 4.524360935110334e-06, + "loss": 3.5254, + "step": 4404 + }, + { + "epoch": 0.96, + "learning_rate": 4.524251693248853e-06, + "loss": 3.4548, + "step": 4405 + }, + { + "epoch": 0.96, + "learning_rate": 4.524142451387372e-06, + "loss": 2.968, + "step": 4406 + }, + { + "epoch": 0.96, + "learning_rate": 4.5240332095258905e-06, + "loss": 3.2096, + "step": 4407 + }, + { + "epoch": 0.96, + "learning_rate": 4.5239239676644095e-06, + "loss": 3.6211, + "step": 4408 + }, + { + "epoch": 0.96, + "learning_rate": 4.523814725802928e-06, + "loss": 2.9105, + "step": 4409 + }, + { + "epoch": 0.96, + "learning_rate": 4.523705483941447e-06, + "loss": 3.2946, + "step": 4410 + }, + { + "epoch": 0.96, + "learning_rate": 4.523596242079966e-06, + "loss": 3.4584, + "step": 4411 + }, + { + "epoch": 0.96, + "learning_rate": 4.523487000218484e-06, + "loss": 3.0439, + "step": 4412 + }, + { + "epoch": 0.96, + "learning_rate": 4.523377758357003e-06, + "loss": 3.0641, + "step": 4413 + }, + { + "epoch": 0.96, + "learning_rate": 4.523268516495521e-06, + "loss": 3.2065, + "step": 4414 + }, + { + "epoch": 0.96, + "learning_rate": 4.52315927463404e-06, + "loss": 3.6109, + "step": 4415 + }, + { + "epoch": 0.96, + "learning_rate": 4.523050032772559e-06, + "loss": 2.406, + "step": 4416 + }, + { + "epoch": 0.96, + "learning_rate": 4.522940790911077e-06, + "loss": 3.0599, + "step": 4417 + }, + { + "epoch": 0.96, + "learning_rate": 4.5228315490495964e-06, + "loss": 3.5602, + "step": 4418 + }, + { + "epoch": 0.96, + "learning_rate": 4.522722307188115e-06, + "loss": 3.4692, + "step": 4419 + }, + { + "epoch": 0.96, + "learning_rate": 4.522613065326634e-06, + "loss": 3.1968, + "step": 4420 + }, + { + "epoch": 0.96, + "learning_rate": 4.522503823465153e-06, + "loss": 3.341, + "step": 4421 + }, + { + "epoch": 0.97, + "learning_rate": 4.522394581603671e-06, + "loss": 2.9094, + "step": 4422 + }, + { + "epoch": 0.97, + "learning_rate": 4.52228533974219e-06, + "loss": 3.8865, + "step": 4423 + }, + { + "epoch": 0.97, + "learning_rate": 4.522176097880708e-06, + "loss": 2.8177, + "step": 4424 + }, + { + "epoch": 0.97, + "learning_rate": 4.522066856019227e-06, + "loss": 3.7629, + "step": 4425 + }, + { + "epoch": 0.97, + "learning_rate": 4.521957614157745e-06, + "loss": 3.9427, + "step": 4426 + }, + { + "epoch": 0.97, + "learning_rate": 4.521848372296264e-06, + "loss": 3.3488, + "step": 4427 + }, + { + "epoch": 0.97, + "learning_rate": 4.5217391304347826e-06, + "loss": 3.4097, + "step": 4428 + }, + { + "epoch": 0.97, + "learning_rate": 4.521629888573302e-06, + "loss": 3.2934, + "step": 4429 + }, + { + "epoch": 0.97, + "learning_rate": 4.521520646711821e-06, + "loss": 3.4386, + "step": 4430 + }, + { + "epoch": 0.97, + "learning_rate": 4.521411404850339e-06, + "loss": 2.9417, + "step": 4431 + }, + { + "epoch": 0.97, + "learning_rate": 4.521302162988858e-06, + "loss": 3.3191, + "step": 4432 + }, + { + "epoch": 0.97, + "learning_rate": 4.521192921127376e-06, + "loss": 3.5237, + "step": 4433 + }, + { + "epoch": 0.97, + "learning_rate": 4.521083679265895e-06, + "loss": 3.6559, + "step": 4434 + }, + { + "epoch": 0.97, + "learning_rate": 4.520974437404414e-06, + "loss": 3.6213, + "step": 4435 + }, + { + "epoch": 0.97, + "learning_rate": 4.520865195542932e-06, + "loss": 2.5944, + "step": 4436 + }, + { + "epoch": 0.97, + "learning_rate": 4.520755953681451e-06, + "loss": 3.303, + "step": 4437 + }, + { + "epoch": 0.97, + "learning_rate": 4.5206467118199695e-06, + "loss": 3.8781, + "step": 4438 + }, + { + "epoch": 0.97, + "learning_rate": 4.5205374699584885e-06, + "loss": 3.5104, + "step": 4439 + }, + { + "epoch": 0.97, + "learning_rate": 4.520428228097007e-06, + "loss": 3.0136, + "step": 4440 + }, + { + "epoch": 0.97, + "learning_rate": 4.520318986235526e-06, + "loss": 3.7065, + "step": 4441 + }, + { + "epoch": 0.97, + "learning_rate": 4.520209744374045e-06, + "loss": 3.4537, + "step": 4442 + }, + { + "epoch": 0.97, + "learning_rate": 4.520100502512563e-06, + "loss": 3.5053, + "step": 4443 + }, + { + "epoch": 0.97, + "learning_rate": 4.519991260651082e-06, + "loss": 3.286, + "step": 4444 + }, + { + "epoch": 0.97, + "learning_rate": 4.5198820187896e-06, + "loss": 2.8975, + "step": 4445 + }, + { + "epoch": 0.97, + "learning_rate": 4.519772776928119e-06, + "loss": 3.5251, + "step": 4446 + }, + { + "epoch": 0.97, + "learning_rate": 4.519663535066638e-06, + "loss": 3.5928, + "step": 4447 + }, + { + "epoch": 0.97, + "learning_rate": 4.5195542932051565e-06, + "loss": 3.5408, + "step": 4448 + }, + { + "epoch": 0.97, + "learning_rate": 4.5194450513436755e-06, + "loss": 2.9374, + "step": 4449 + }, + { + "epoch": 0.97, + "learning_rate": 4.519335809482194e-06, + "loss": 3.0158, + "step": 4450 + }, + { + "epoch": 0.97, + "learning_rate": 4.519226567620713e-06, + "loss": 3.3732, + "step": 4451 + }, + { + "epoch": 0.97, + "learning_rate": 4.519117325759232e-06, + "loss": 3.4208, + "step": 4452 + }, + { + "epoch": 0.97, + "learning_rate": 4.51900808389775e-06, + "loss": 3.2194, + "step": 4453 + }, + { + "epoch": 0.97, + "learning_rate": 4.518898842036269e-06, + "loss": 3.3197, + "step": 4454 + }, + { + "epoch": 0.97, + "learning_rate": 4.518789600174787e-06, + "loss": 3.6009, + "step": 4455 + }, + { + "epoch": 0.97, + "learning_rate": 4.518680358313306e-06, + "loss": 3.6396, + "step": 4456 + }, + { + "epoch": 0.97, + "learning_rate": 4.518571116451824e-06, + "loss": 2.5422, + "step": 4457 + }, + { + "epoch": 0.97, + "learning_rate": 4.518461874590343e-06, + "loss": 3.4619, + "step": 4458 + }, + { + "epoch": 0.97, + "learning_rate": 4.518352632728862e-06, + "loss": 2.9713, + "step": 4459 + }, + { + "epoch": 0.97, + "learning_rate": 4.518243390867381e-06, + "loss": 3.97, + "step": 4460 + }, + { + "epoch": 0.97, + "learning_rate": 4.5181341490059e-06, + "loss": 2.8561, + "step": 4461 + }, + { + "epoch": 0.97, + "learning_rate": 4.518024907144418e-06, + "loss": 3.3311, + "step": 4462 + }, + { + "epoch": 0.97, + "learning_rate": 4.517915665282937e-06, + "loss": 3.7577, + "step": 4463 + }, + { + "epoch": 0.97, + "learning_rate": 4.517806423421455e-06, + "loss": 2.5934, + "step": 4464 + }, + { + "epoch": 0.97, + "learning_rate": 4.517697181559974e-06, + "loss": 3.3743, + "step": 4465 + }, + { + "epoch": 0.97, + "learning_rate": 4.517587939698493e-06, + "loss": 2.8469, + "step": 4466 + }, + { + "epoch": 0.97, + "learning_rate": 4.517478697837011e-06, + "loss": 3.2112, + "step": 4467 + }, + { + "epoch": 0.98, + "learning_rate": 4.51736945597553e-06, + "loss": 3.1642, + "step": 4468 + }, + { + "epoch": 0.98, + "learning_rate": 4.5172602141140486e-06, + "loss": 3.6953, + "step": 4469 + }, + { + "epoch": 0.98, + "learning_rate": 4.517150972252568e-06, + "loss": 3.0422, + "step": 4470 + }, + { + "epoch": 0.98, + "learning_rate": 4.517041730391087e-06, + "loss": 3.7792, + "step": 4471 + }, + { + "epoch": 0.98, + "learning_rate": 4.516932488529605e-06, + "loss": 3.132, + "step": 4472 + }, + { + "epoch": 0.98, + "learning_rate": 4.516823246668123e-06, + "loss": 3.6526, + "step": 4473 + }, + { + "epoch": 0.98, + "learning_rate": 4.516714004806642e-06, + "loss": 3.4763, + "step": 4474 + }, + { + "epoch": 0.98, + "learning_rate": 4.516604762945161e-06, + "loss": 3.3002, + "step": 4475 + }, + { + "epoch": 0.98, + "learning_rate": 4.51649552108368e-06, + "loss": 3.5327, + "step": 4476 + }, + { + "epoch": 0.98, + "learning_rate": 4.516386279222198e-06, + "loss": 2.9849, + "step": 4477 + }, + { + "epoch": 0.98, + "learning_rate": 4.5162770373607165e-06, + "loss": 2.4812, + "step": 4478 + }, + { + "epoch": 0.98, + "learning_rate": 4.5161677954992355e-06, + "loss": 3.3771, + "step": 4479 + }, + { + "epoch": 0.98, + "learning_rate": 4.5160585536377546e-06, + "loss": 3.4894, + "step": 4480 + }, + { + "epoch": 0.98, + "learning_rate": 4.515949311776273e-06, + "loss": 3.7911, + "step": 4481 + }, + { + "epoch": 0.98, + "learning_rate": 4.515840069914792e-06, + "loss": 3.4201, + "step": 4482 + }, + { + "epoch": 0.98, + "learning_rate": 4.51573082805331e-06, + "loss": 3.1914, + "step": 4483 + }, + { + "epoch": 0.98, + "learning_rate": 4.515621586191829e-06, + "loss": 3.4372, + "step": 4484 + }, + { + "epoch": 0.98, + "learning_rate": 4.515512344330348e-06, + "loss": 3.8477, + "step": 4485 + }, + { + "epoch": 0.98, + "learning_rate": 4.515403102468866e-06, + "loss": 2.9635, + "step": 4486 + }, + { + "epoch": 0.98, + "learning_rate": 4.515293860607385e-06, + "loss": 3.912, + "step": 4487 + }, + { + "epoch": 0.98, + "learning_rate": 4.5151846187459034e-06, + "loss": 3.1816, + "step": 4488 + }, + { + "epoch": 0.98, + "learning_rate": 4.5150753768844225e-06, + "loss": 3.252, + "step": 4489 + }, + { + "epoch": 0.98, + "learning_rate": 4.5149661350229415e-06, + "loss": 3.3387, + "step": 4490 + }, + { + "epoch": 0.98, + "learning_rate": 4.51485689316146e-06, + "loss": 3.1322, + "step": 4491 + }, + { + "epoch": 0.98, + "learning_rate": 4.514747651299978e-06, + "loss": 3.5976, + "step": 4492 + }, + { + "epoch": 0.98, + "learning_rate": 4.514638409438497e-06, + "loss": 2.414, + "step": 4493 + }, + { + "epoch": 0.98, + "learning_rate": 4.514529167577016e-06, + "loss": 3.0313, + "step": 4494 + }, + { + "epoch": 0.98, + "learning_rate": 4.514419925715535e-06, + "loss": 2.6593, + "step": 4495 + }, + { + "epoch": 0.98, + "learning_rate": 4.514310683854053e-06, + "loss": 3.3501, + "step": 4496 + }, + { + "epoch": 0.98, + "learning_rate": 4.514201441992571e-06, + "loss": 2.9555, + "step": 4497 + }, + { + "epoch": 0.98, + "learning_rate": 4.51409220013109e-06, + "loss": 3.1193, + "step": 4498 + }, + { + "epoch": 0.98, + "learning_rate": 4.5139829582696094e-06, + "loss": 3.0321, + "step": 4499 + }, + { + "epoch": 0.98, + "learning_rate": 4.5138737164081285e-06, + "loss": 3.8249, + "step": 4500 + }, + { + "epoch": 0.98, + "learning_rate": 4.513764474546647e-06, + "loss": 3.2615, + "step": 4501 + }, + { + "epoch": 0.98, + "learning_rate": 4.513655232685165e-06, + "loss": 3.3255, + "step": 4502 + }, + { + "epoch": 0.98, + "learning_rate": 4.513545990823684e-06, + "loss": 3.0785, + "step": 4503 + }, + { + "epoch": 0.98, + "learning_rate": 4.513436748962203e-06, + "loss": 3.5663, + "step": 4504 + }, + { + "epoch": 0.98, + "learning_rate": 4.513327507100722e-06, + "loss": 3.6993, + "step": 4505 + }, + { + "epoch": 0.98, + "learning_rate": 4.51321826523924e-06, + "loss": 3.043, + "step": 4506 + }, + { + "epoch": 0.98, + "learning_rate": 4.513109023377758e-06, + "loss": 3.3999, + "step": 4507 + }, + { + "epoch": 0.98, + "learning_rate": 4.512999781516277e-06, + "loss": 2.9704, + "step": 4508 + }, + { + "epoch": 0.98, + "learning_rate": 4.512890539654796e-06, + "loss": 3.1356, + "step": 4509 + }, + { + "epoch": 0.98, + "learning_rate": 4.512781297793315e-06, + "loss": 3.1624, + "step": 4510 + }, + { + "epoch": 0.98, + "learning_rate": 4.512672055931834e-06, + "loss": 2.5133, + "step": 4511 + }, + { + "epoch": 0.98, + "learning_rate": 4.512562814070352e-06, + "loss": 2.9274, + "step": 4512 + }, + { + "epoch": 0.98, + "learning_rate": 4.512453572208871e-06, + "loss": 2.8787, + "step": 4513 + }, + { + "epoch": 0.99, + "learning_rate": 4.51234433034739e-06, + "loss": 3.381, + "step": 4514 + }, + { + "epoch": 0.99, + "learning_rate": 4.512235088485908e-06, + "loss": 3.1351, + "step": 4515 + }, + { + "epoch": 0.99, + "learning_rate": 4.512125846624426e-06, + "loss": 3.1564, + "step": 4516 + }, + { + "epoch": 0.99, + "learning_rate": 4.512016604762945e-06, + "loss": 2.6235, + "step": 4517 + }, + { + "epoch": 0.99, + "learning_rate": 4.511907362901464e-06, + "loss": 3.477, + "step": 4518 + }, + { + "epoch": 0.99, + "learning_rate": 4.511798121039983e-06, + "loss": 3.2498, + "step": 4519 + }, + { + "epoch": 0.99, + "learning_rate": 4.5116888791785015e-06, + "loss": 2.174, + "step": 4520 + }, + { + "epoch": 0.99, + "learning_rate": 4.51157963731702e-06, + "loss": 3.5345, + "step": 4521 + }, + { + "epoch": 0.99, + "learning_rate": 4.511470395455539e-06, + "loss": 3.6134, + "step": 4522 + }, + { + "epoch": 0.99, + "learning_rate": 4.511361153594058e-06, + "loss": 2.6853, + "step": 4523 + }, + { + "epoch": 0.99, + "learning_rate": 4.511251911732577e-06, + "loss": 3.6869, + "step": 4524 + }, + { + "epoch": 0.99, + "learning_rate": 4.511142669871095e-06, + "loss": 3.5623, + "step": 4525 + }, + { + "epoch": 0.99, + "learning_rate": 4.511033428009613e-06, + "loss": 3.0523, + "step": 4526 + }, + { + "epoch": 0.99, + "learning_rate": 4.510924186148132e-06, + "loss": 2.6558, + "step": 4527 + }, + { + "epoch": 0.99, + "learning_rate": 4.510814944286651e-06, + "loss": 3.5706, + "step": 4528 + }, + { + "epoch": 0.99, + "learning_rate": 4.51070570242517e-06, + "loss": 3.3338, + "step": 4529 + }, + { + "epoch": 0.99, + "learning_rate": 4.5105964605636885e-06, + "loss": 3.2937, + "step": 4530 + }, + { + "epoch": 0.99, + "learning_rate": 4.510487218702207e-06, + "loss": 3.4288, + "step": 4531 + }, + { + "epoch": 0.99, + "learning_rate": 4.510377976840726e-06, + "loss": 3.3548, + "step": 4532 + }, + { + "epoch": 0.99, + "learning_rate": 4.510268734979245e-06, + "loss": 3.1393, + "step": 4533 + }, + { + "epoch": 0.99, + "learning_rate": 4.510159493117763e-06, + "loss": 3.0499, + "step": 4534 + }, + { + "epoch": 0.99, + "learning_rate": 4.510050251256282e-06, + "loss": 3.0862, + "step": 4535 + }, + { + "epoch": 0.99, + "learning_rate": 4.5099410093948e-06, + "loss": 2.9141, + "step": 4536 + }, + { + "epoch": 0.99, + "learning_rate": 4.509831767533319e-06, + "loss": 3.4202, + "step": 4537 + }, + { + "epoch": 0.99, + "learning_rate": 4.509722525671838e-06, + "loss": 3.3978, + "step": 4538 + }, + { + "epoch": 0.99, + "learning_rate": 4.509613283810356e-06, + "loss": 2.5067, + "step": 4539 + }, + { + "epoch": 0.99, + "learning_rate": 4.5095040419488755e-06, + "loss": 3.3265, + "step": 4540 + }, + { + "epoch": 0.99, + "learning_rate": 4.509394800087394e-06, + "loss": 3.5234, + "step": 4541 + }, + { + "epoch": 0.99, + "learning_rate": 4.509285558225913e-06, + "loss": 3.6118, + "step": 4542 + }, + { + "epoch": 0.99, + "learning_rate": 4.509176316364432e-06, + "loss": 2.5409, + "step": 4543 + }, + { + "epoch": 0.99, + "learning_rate": 4.50906707450295e-06, + "loss": 3.6527, + "step": 4544 + }, + { + "epoch": 0.99, + "learning_rate": 4.508957832641468e-06, + "loss": 3.063, + "step": 4545 + }, + { + "epoch": 0.99, + "learning_rate": 4.508848590779987e-06, + "loss": 3.4908, + "step": 4546 + }, + { + "epoch": 0.99, + "learning_rate": 4.508739348918506e-06, + "loss": 4.152, + "step": 4547 + }, + { + "epoch": 0.99, + "learning_rate": 4.508630107057025e-06, + "loss": 3.3285, + "step": 4548 + }, + { + "epoch": 0.99, + "learning_rate": 4.508520865195543e-06, + "loss": 2.8365, + "step": 4549 + }, + { + "epoch": 0.99, + "learning_rate": 4.5084116233340616e-06, + "loss": 2.999, + "step": 4550 + }, + { + "epoch": 0.99, + "learning_rate": 4.508302381472581e-06, + "loss": 3.2758, + "step": 4551 + }, + { + "epoch": 0.99, + "learning_rate": 4.5081931396111e-06, + "loss": 3.3177, + "step": 4552 + }, + { + "epoch": 0.99, + "learning_rate": 4.508083897749619e-06, + "loss": 3.5823, + "step": 4553 + }, + { + "epoch": 0.99, + "learning_rate": 4.507974655888137e-06, + "loss": 2.5577, + "step": 4554 + }, + { + "epoch": 0.99, + "learning_rate": 4.507865414026655e-06, + "loss": 3.6135, + "step": 4555 + }, + { + "epoch": 0.99, + "learning_rate": 4.507756172165174e-06, + "loss": 3.1743, + "step": 4556 + }, + { + "epoch": 0.99, + "learning_rate": 4.507646930303693e-06, + "loss": 3.455, + "step": 4557 + }, + { + "epoch": 0.99, + "learning_rate": 4.507537688442211e-06, + "loss": 3.1287, + "step": 4558 + }, + { + "epoch": 0.99, + "learning_rate": 4.50742844658073e-06, + "loss": 3.0159, + "step": 4559 + }, + { + "epoch": 1.0, + "learning_rate": 4.5073192047192485e-06, + "loss": 3.0594, + "step": 4560 + }, + { + "epoch": 1.0, + "learning_rate": 4.5072099628577676e-06, + "loss": 3.317, + "step": 4561 + }, + { + "epoch": 1.0, + "learning_rate": 4.507100720996287e-06, + "loss": 2.7797, + "step": 4562 + }, + { + "epoch": 1.0, + "learning_rate": 4.506991479134805e-06, + "loss": 3.2567, + "step": 4563 + }, + { + "epoch": 1.0, + "learning_rate": 4.506882237273324e-06, + "loss": 3.0329, + "step": 4564 + }, + { + "epoch": 1.0, + "learning_rate": 4.506772995411842e-06, + "loss": 2.5891, + "step": 4565 + }, + { + "epoch": 1.0, + "learning_rate": 4.506663753550361e-06, + "loss": 2.3196, + "step": 4566 + }, + { + "epoch": 1.0, + "learning_rate": 4.50655451168888e-06, + "loss": 3.4132, + "step": 4567 + }, + { + "epoch": 1.0, + "learning_rate": 4.506445269827398e-06, + "loss": 3.1591, + "step": 4568 + }, + { + "epoch": 1.0, + "learning_rate": 4.5063360279659164e-06, + "loss": 3.4617, + "step": 4569 + }, + { + "epoch": 1.0, + "learning_rate": 4.5062267861044355e-06, + "loss": 3.1384, + "step": 4570 + }, + { + "epoch": 1.0, + "learning_rate": 4.5061175442429545e-06, + "loss": 3.436, + "step": 4571 + }, + { + "epoch": 1.0, + "learning_rate": 4.506008302381473e-06, + "loss": 2.8492, + "step": 4572 + }, + { + "epoch": 1.0, + "learning_rate": 4.505899060519992e-06, + "loss": 3.291, + "step": 4573 + }, + { + "epoch": 1.0, + "learning_rate": 4.50578981865851e-06, + "loss": 3.6726, + "step": 4574 + }, + { + "epoch": 1.0, + "learning_rate": 4.505680576797029e-06, + "loss": 2.9781, + "step": 4575 + }, + { + "epoch": 1.0, + "learning_rate": 4.505571334935548e-06, + "loss": 3.5258, + "step": 4576 + }, + { + "epoch": 1.0, + "learning_rate": 4.505462093074066e-06, + "loss": 3.1535, + "step": 4577 + }, + { + "epoch": 1.0, + "learning_rate": 4.505352851212585e-06, + "loss": 3.3934, + "step": 4578 + }, + { + "epoch": 1.0, + "learning_rate": 4.505243609351103e-06, + "loss": 2.6145, + "step": 4579 + }, + { + "epoch": 1.0, + "learning_rate": 4.5051343674896224e-06, + "loss": 3.7703, + "step": 4580 + }, + { + "epoch": 1.0, + "learning_rate": 4.5050251256281415e-06, + "loss": 3.4713, + "step": 4581 + }, + { + "epoch": 1.0, + "learning_rate": 4.50491588376666e-06, + "loss": 3.241, + "step": 4582 + }, + { + "epoch": 1.0, + "eval_accuracy": 0.24157119476268413, + "eval_loss": 3.1584739685058594, + "eval_runtime": 332.8778, + "eval_samples_per_second": 18.355, + "eval_steps_per_second": 4.59, + "step": 4582 + }, + { + "epoch": 1.0, + "learning_rate": 4.504806641905179e-06, + "loss": 2.6882, + "step": 4583 + }, + { + "epoch": 1.0, + "learning_rate": 4.504697400043697e-06, + "loss": 2.5185, + "step": 4584 + }, + { + "epoch": 1.0, + "learning_rate": 4.504588158182216e-06, + "loss": 3.3798, + "step": 4585 + }, + { + "epoch": 1.0, + "learning_rate": 4.504478916320734e-06, + "loss": 3.4712, + "step": 4586 + }, + { + "epoch": 1.0, + "learning_rate": 4.504369674459253e-06, + "loss": 2.6744, + "step": 4587 + }, + { + "epoch": 1.0, + "learning_rate": 4.504260432597772e-06, + "loss": 3.5582, + "step": 4588 + }, + { + "epoch": 1.0, + "learning_rate": 4.50415119073629e-06, + "loss": 3.6333, + "step": 4589 + }, + { + "epoch": 1.0, + "learning_rate": 4.504041948874809e-06, + "loss": 3.8503, + "step": 4590 + }, + { + "epoch": 1.0, + "learning_rate": 4.503932707013328e-06, + "loss": 3.0528, + "step": 4591 + }, + { + "epoch": 1.0, + "learning_rate": 4.503823465151847e-06, + "loss": 3.3787, + "step": 4592 + }, + { + "epoch": 1.0, + "learning_rate": 4.503714223290366e-06, + "loss": 4.032, + "step": 4593 + }, + { + "epoch": 1.0, + "learning_rate": 4.503604981428884e-06, + "loss": 3.0928, + "step": 4594 + }, + { + "epoch": 1.0, + "learning_rate": 4.503495739567403e-06, + "loss": 3.1518, + "step": 4595 + }, + { + "epoch": 1.0, + "learning_rate": 4.503386497705921e-06, + "loss": 2.8855, + "step": 4596 + }, + { + "epoch": 1.0, + "learning_rate": 4.50327725584444e-06, + "loss": 3.7176, + "step": 4597 + }, + { + "epoch": 1.0, + "learning_rate": 4.503168013982958e-06, + "loss": 2.8726, + "step": 4598 + }, + { + "epoch": 1.0, + "learning_rate": 4.503058772121477e-06, + "loss": 3.3512, + "step": 4599 + }, + { + "epoch": 1.0, + "learning_rate": 4.502949530259996e-06, + "loss": 3.5941, + "step": 4600 + }, + { + "epoch": 1.0, + "learning_rate": 4.5028402883985145e-06, + "loss": 2.7545, + "step": 4601 + }, + { + "epoch": 1.0, + "learning_rate": 4.5027310465370336e-06, + "loss": 2.9751, + "step": 4602 + }, + { + "epoch": 1.0, + "learning_rate": 4.502621804675552e-06, + "loss": 3.3298, + "step": 4603 + }, + { + "epoch": 1.0, + "learning_rate": 4.502512562814071e-06, + "loss": 3.962, + "step": 4604 + }, + { + "epoch": 1.01, + "learning_rate": 4.502403320952589e-06, + "loss": 3.1566, + "step": 4605 + }, + { + "epoch": 1.01, + "learning_rate": 4.502294079091108e-06, + "loss": 3.827, + "step": 4606 + }, + { + "epoch": 1.01, + "learning_rate": 4.502184837229627e-06, + "loss": 3.2934, + "step": 4607 + }, + { + "epoch": 1.01, + "learning_rate": 4.502075595368145e-06, + "loss": 3.2556, + "step": 4608 + }, + { + "epoch": 1.01, + "learning_rate": 4.501966353506664e-06, + "loss": 2.9953, + "step": 4609 + }, + { + "epoch": 1.01, + "learning_rate": 4.5018571116451825e-06, + "loss": 3.3524, + "step": 4610 + }, + { + "epoch": 1.01, + "learning_rate": 4.5017478697837015e-06, + "loss": 2.9749, + "step": 4611 + }, + { + "epoch": 1.01, + "learning_rate": 4.5016386279222205e-06, + "loss": 3.664, + "step": 4612 + }, + { + "epoch": 1.01, + "learning_rate": 4.501529386060739e-06, + "loss": 3.3102, + "step": 4613 + }, + { + "epoch": 1.01, + "learning_rate": 4.501420144199258e-06, + "loss": 3.6106, + "step": 4614 + }, + { + "epoch": 1.01, + "learning_rate": 4.501310902337776e-06, + "loss": 3.362, + "step": 4615 + }, + { + "epoch": 1.01, + "learning_rate": 4.501201660476295e-06, + "loss": 3.2568, + "step": 4616 + }, + { + "epoch": 1.01, + "learning_rate": 4.501092418614814e-06, + "loss": 2.744, + "step": 4617 + }, + { + "epoch": 1.01, + "learning_rate": 4.500983176753332e-06, + "loss": 3.2916, + "step": 4618 + }, + { + "epoch": 1.01, + "learning_rate": 4.50087393489185e-06, + "loss": 3.2122, + "step": 4619 + }, + { + "epoch": 1.01, + "learning_rate": 4.500764693030369e-06, + "loss": 2.928, + "step": 4620 + }, + { + "epoch": 1.01, + "learning_rate": 4.5006554511688884e-06, + "loss": 3.1706, + "step": 4621 + }, + { + "epoch": 1.01, + "learning_rate": 4.500546209307407e-06, + "loss": 2.937, + "step": 4622 + }, + { + "epoch": 1.01, + "learning_rate": 4.500436967445926e-06, + "loss": 3.4614, + "step": 4623 + }, + { + "epoch": 1.01, + "learning_rate": 4.500327725584444e-06, + "loss": 3.3609, + "step": 4624 + }, + { + "epoch": 1.01, + "learning_rate": 4.500218483722963e-06, + "loss": 3.0849, + "step": 4625 + }, + { + "epoch": 1.01, + "learning_rate": 4.500109241861482e-06, + "loss": 3.1397, + "step": 4626 + }, + { + "epoch": 1.01, + "learning_rate": 4.5e-06, + "loss": 3.1328, + "step": 4627 + }, + { + "epoch": 1.01, + "learning_rate": 4.499890758138519e-06, + "loss": 2.8873, + "step": 4628 + }, + { + "epoch": 1.01, + "learning_rate": 4.499781516277037e-06, + "loss": 3.3551, + "step": 4629 + }, + { + "epoch": 1.01, + "learning_rate": 4.499672274415556e-06, + "loss": 2.9858, + "step": 4630 + }, + { + "epoch": 1.01, + "learning_rate": 4.499563032554075e-06, + "loss": 3.6673, + "step": 4631 + }, + { + "epoch": 1.01, + "learning_rate": 4.499453790692594e-06, + "loss": 3.305, + "step": 4632 + }, + { + "epoch": 1.01, + "learning_rate": 4.499344548831113e-06, + "loss": 3.1421, + "step": 4633 + }, + { + "epoch": 1.01, + "learning_rate": 4.499235306969631e-06, + "loss": 3.0016, + "step": 4634 + }, + { + "epoch": 1.01, + "learning_rate": 4.49912606510815e-06, + "loss": 3.5489, + "step": 4635 + }, + { + "epoch": 1.01, + "learning_rate": 4.499016823246669e-06, + "loss": 3.6626, + "step": 4636 + }, + { + "epoch": 1.01, + "learning_rate": 4.498907581385187e-06, + "loss": 3.5074, + "step": 4637 + }, + { + "epoch": 1.01, + "learning_rate": 4.498798339523705e-06, + "loss": 3.3407, + "step": 4638 + }, + { + "epoch": 1.01, + "learning_rate": 4.498689097662224e-06, + "loss": 3.2486, + "step": 4639 + }, + { + "epoch": 1.01, + "learning_rate": 4.498579855800743e-06, + "loss": 2.9521, + "step": 4640 + }, + { + "epoch": 1.01, + "learning_rate": 4.498470613939262e-06, + "loss": 2.9313, + "step": 4641 + }, + { + "epoch": 1.01, + "learning_rate": 4.4983613720777805e-06, + "loss": 2.8258, + "step": 4642 + }, + { + "epoch": 1.01, + "learning_rate": 4.498252130216299e-06, + "loss": 3.2198, + "step": 4643 + }, + { + "epoch": 1.01, + "learning_rate": 4.498142888354818e-06, + "loss": 3.043, + "step": 4644 + }, + { + "epoch": 1.01, + "learning_rate": 4.498033646493337e-06, + "loss": 3.2834, + "step": 4645 + }, + { + "epoch": 1.01, + "learning_rate": 4.497924404631856e-06, + "loss": 3.2941, + "step": 4646 + }, + { + "epoch": 1.01, + "learning_rate": 4.497815162770374e-06, + "loss": 3.1951, + "step": 4647 + }, + { + "epoch": 1.01, + "learning_rate": 4.497705920908892e-06, + "loss": 3.2864, + "step": 4648 + }, + { + "epoch": 1.01, + "learning_rate": 4.497596679047411e-06, + "loss": 3.198, + "step": 4649 + }, + { + "epoch": 1.01, + "learning_rate": 4.49748743718593e-06, + "loss": 3.224, + "step": 4650 + }, + { + "epoch": 1.02, + "learning_rate": 4.4973781953244485e-06, + "loss": 3.9164, + "step": 4651 + }, + { + "epoch": 1.02, + "learning_rate": 4.4972689534629675e-06, + "loss": 3.1343, + "step": 4652 + }, + { + "epoch": 1.02, + "learning_rate": 4.497159711601486e-06, + "loss": 3.1631, + "step": 4653 + }, + { + "epoch": 1.02, + "learning_rate": 4.497050469740005e-06, + "loss": 3.4411, + "step": 4654 + }, + { + "epoch": 1.02, + "learning_rate": 4.496941227878524e-06, + "loss": 3.6485, + "step": 4655 + }, + { + "epoch": 1.02, + "learning_rate": 4.496831986017042e-06, + "loss": 3.3434, + "step": 4656 + }, + { + "epoch": 1.02, + "learning_rate": 4.49672274415556e-06, + "loss": 3.1575, + "step": 4657 + }, + { + "epoch": 1.02, + "learning_rate": 4.496613502294079e-06, + "loss": 3.0121, + "step": 4658 + }, + { + "epoch": 1.02, + "learning_rate": 4.496504260432598e-06, + "loss": 3.3806, + "step": 4659 + }, + { + "epoch": 1.02, + "learning_rate": 4.496395018571117e-06, + "loss": 3.3859, + "step": 4660 + }, + { + "epoch": 1.02, + "learning_rate": 4.4962857767096354e-06, + "loss": 3.0681, + "step": 4661 + }, + { + "epoch": 1.02, + "learning_rate": 4.496176534848154e-06, + "loss": 3.9566, + "step": 4662 + }, + { + "epoch": 1.02, + "learning_rate": 4.496067292986673e-06, + "loss": 3.0682, + "step": 4663 + }, + { + "epoch": 1.02, + "learning_rate": 4.495958051125192e-06, + "loss": 3.1721, + "step": 4664 + }, + { + "epoch": 1.02, + "learning_rate": 4.495848809263711e-06, + "loss": 3.4055, + "step": 4665 + }, + { + "epoch": 1.02, + "learning_rate": 4.495739567402229e-06, + "loss": 2.7091, + "step": 4666 + }, + { + "epoch": 1.02, + "learning_rate": 4.495630325540747e-06, + "loss": 3.1028, + "step": 4667 + }, + { + "epoch": 1.02, + "learning_rate": 4.495521083679266e-06, + "loss": 3.2011, + "step": 4668 + }, + { + "epoch": 1.02, + "learning_rate": 4.495411841817785e-06, + "loss": 2.8795, + "step": 4669 + }, + { + "epoch": 1.02, + "learning_rate": 4.495302599956304e-06, + "loss": 2.9121, + "step": 4670 + }, + { + "epoch": 1.02, + "learning_rate": 4.495193358094822e-06, + "loss": 3.5369, + "step": 4671 + }, + { + "epoch": 1.02, + "learning_rate": 4.4950841162333406e-06, + "loss": 3.8878, + "step": 4672 + }, + { + "epoch": 1.02, + "learning_rate": 4.49497487437186e-06, + "loss": 3.504, + "step": 4673 + }, + { + "epoch": 1.02, + "learning_rate": 4.494865632510379e-06, + "loss": 3.317, + "step": 4674 + }, + { + "epoch": 1.02, + "learning_rate": 4.494756390648897e-06, + "loss": 2.7145, + "step": 4675 + }, + { + "epoch": 1.02, + "learning_rate": 4.494647148787416e-06, + "loss": 2.7855, + "step": 4676 + }, + { + "epoch": 1.02, + "learning_rate": 4.494537906925934e-06, + "loss": 2.7899, + "step": 4677 + }, + { + "epoch": 1.02, + "learning_rate": 4.494428665064453e-06, + "loss": 2.9526, + "step": 4678 + }, + { + "epoch": 1.02, + "learning_rate": 4.494319423202972e-06, + "loss": 3.705, + "step": 4679 + }, + { + "epoch": 1.02, + "learning_rate": 4.49421018134149e-06, + "loss": 3.5519, + "step": 4680 + }, + { + "epoch": 1.02, + "learning_rate": 4.4941009394800085e-06, + "loss": 3.1069, + "step": 4681 + }, + { + "epoch": 1.02, + "learning_rate": 4.4939916976185275e-06, + "loss": 2.8461, + "step": 4682 + }, + { + "epoch": 1.02, + "learning_rate": 4.4938824557570466e-06, + "loss": 3.0042, + "step": 4683 + }, + { + "epoch": 1.02, + "learning_rate": 4.493773213895566e-06, + "loss": 2.6754, + "step": 4684 + }, + { + "epoch": 1.02, + "learning_rate": 4.493663972034084e-06, + "loss": 3.7045, + "step": 4685 + }, + { + "epoch": 1.02, + "learning_rate": 4.493554730172602e-06, + "loss": 3.5783, + "step": 4686 + }, + { + "epoch": 1.02, + "learning_rate": 4.493445488311121e-06, + "loss": 3.8001, + "step": 4687 + }, + { + "epoch": 1.02, + "learning_rate": 4.49333624644964e-06, + "loss": 3.3721, + "step": 4688 + }, + { + "epoch": 1.02, + "learning_rate": 4.493227004588159e-06, + "loss": 2.6726, + "step": 4689 + }, + { + "epoch": 1.02, + "learning_rate": 4.493117762726677e-06, + "loss": 3.207, + "step": 4690 + }, + { + "epoch": 1.02, + "learning_rate": 4.4930085208651955e-06, + "loss": 3.2049, + "step": 4691 + }, + { + "epoch": 1.02, + "learning_rate": 4.4928992790037145e-06, + "loss": 2.8978, + "step": 4692 + }, + { + "epoch": 1.02, + "learning_rate": 4.4927900371422335e-06, + "loss": 3.1087, + "step": 4693 + }, + { + "epoch": 1.02, + "learning_rate": 4.4926807952807526e-06, + "loss": 3.0242, + "step": 4694 + }, + { + "epoch": 1.02, + "learning_rate": 4.492571553419271e-06, + "loss": 3.1726, + "step": 4695 + }, + { + "epoch": 1.02, + "learning_rate": 4.492462311557789e-06, + "loss": 3.2261, + "step": 4696 + }, + { + "epoch": 1.03, + "learning_rate": 4.492353069696308e-06, + "loss": 3.5039, + "step": 4697 + }, + { + "epoch": 1.03, + "learning_rate": 4.492243827834827e-06, + "loss": 2.5624, + "step": 4698 + }, + { + "epoch": 1.03, + "learning_rate": 4.492134585973346e-06, + "loss": 3.6041, + "step": 4699 + }, + { + "epoch": 1.03, + "learning_rate": 4.492025344111864e-06, + "loss": 2.6281, + "step": 4700 + }, + { + "epoch": 1.03, + "learning_rate": 4.491916102250382e-06, + "loss": 2.4841, + "step": 4701 + }, + { + "epoch": 1.03, + "learning_rate": 4.4918068603889014e-06, + "loss": 2.4041, + "step": 4702 + }, + { + "epoch": 1.03, + "learning_rate": 4.4916976185274205e-06, + "loss": 2.547, + "step": 4703 + }, + { + "epoch": 1.03, + "learning_rate": 4.491588376665939e-06, + "loss": 3.3579, + "step": 4704 + }, + { + "epoch": 1.03, + "learning_rate": 4.491479134804458e-06, + "loss": 3.6039, + "step": 4705 + }, + { + "epoch": 1.03, + "learning_rate": 4.491369892942976e-06, + "loss": 3.096, + "step": 4706 + }, + { + "epoch": 1.03, + "learning_rate": 4.491260651081495e-06, + "loss": 3.464, + "step": 4707 + }, + { + "epoch": 1.03, + "learning_rate": 4.491151409220014e-06, + "loss": 3.7919, + "step": 4708 + }, + { + "epoch": 1.03, + "learning_rate": 4.491042167358532e-06, + "loss": 3.5973, + "step": 4709 + }, + { + "epoch": 1.03, + "learning_rate": 4.49093292549705e-06, + "loss": 3.2445, + "step": 4710 + }, + { + "epoch": 1.03, + "learning_rate": 4.490823683635569e-06, + "loss": 2.4156, + "step": 4711 + }, + { + "epoch": 1.03, + "learning_rate": 4.490714441774088e-06, + "loss": 3.5847, + "step": 4712 + }, + { + "epoch": 1.03, + "learning_rate": 4.4906051999126074e-06, + "loss": 3.0727, + "step": 4713 + }, + { + "epoch": 1.03, + "learning_rate": 4.490495958051126e-06, + "loss": 3.6879, + "step": 4714 + }, + { + "epoch": 1.03, + "learning_rate": 4.490386716189644e-06, + "loss": 3.7427, + "step": 4715 + }, + { + "epoch": 1.03, + "learning_rate": 4.490277474328163e-06, + "loss": 3.0967, + "step": 4716 + }, + { + "epoch": 1.03, + "learning_rate": 4.490168232466682e-06, + "loss": 2.9358, + "step": 4717 + }, + { + "epoch": 1.03, + "learning_rate": 4.4900589906052e-06, + "loss": 3.0661, + "step": 4718 + }, + { + "epoch": 1.03, + "learning_rate": 4.489949748743719e-06, + "loss": 2.9115, + "step": 4719 + }, + { + "epoch": 1.03, + "learning_rate": 4.489840506882237e-06, + "loss": 3.5143, + "step": 4720 + }, + { + "epoch": 1.03, + "learning_rate": 4.489731265020756e-06, + "loss": 2.5366, + "step": 4721 + }, + { + "epoch": 1.03, + "learning_rate": 4.489622023159275e-06, + "loss": 2.8003, + "step": 4722 + }, + { + "epoch": 1.03, + "learning_rate": 4.4895127812977935e-06, + "loss": 3.6837, + "step": 4723 + }, + { + "epoch": 1.03, + "learning_rate": 4.489403539436313e-06, + "loss": 2.3814, + "step": 4724 + }, + { + "epoch": 1.03, + "learning_rate": 4.489294297574831e-06, + "loss": 3.2868, + "step": 4725 + }, + { + "epoch": 1.03, + "learning_rate": 4.48918505571335e-06, + "loss": 2.8553, + "step": 4726 + }, + { + "epoch": 1.03, + "learning_rate": 4.489075813851869e-06, + "loss": 3.8731, + "step": 4727 + }, + { + "epoch": 1.03, + "learning_rate": 4.488966571990387e-06, + "loss": 3.8587, + "step": 4728 + }, + { + "epoch": 1.03, + "learning_rate": 4.488857330128906e-06, + "loss": 3.0326, + "step": 4729 + }, + { + "epoch": 1.03, + "learning_rate": 4.488748088267424e-06, + "loss": 3.661, + "step": 4730 + }, + { + "epoch": 1.03, + "learning_rate": 4.488638846405943e-06, + "loss": 3.228, + "step": 4731 + }, + { + "epoch": 1.03, + "learning_rate": 4.488529604544462e-06, + "loss": 3.165, + "step": 4732 + }, + { + "epoch": 1.03, + "learning_rate": 4.4884203626829805e-06, + "loss": 2.9326, + "step": 4733 + }, + { + "epoch": 1.03, + "learning_rate": 4.488311120821499e-06, + "loss": 3.0206, + "step": 4734 + }, + { + "epoch": 1.03, + "learning_rate": 4.488201878960018e-06, + "loss": 3.1223, + "step": 4735 + }, + { + "epoch": 1.03, + "learning_rate": 4.488092637098537e-06, + "loss": 3.8008, + "step": 4736 + }, + { + "epoch": 1.03, + "learning_rate": 4.487983395237055e-06, + "loss": 3.2985, + "step": 4737 + }, + { + "epoch": 1.03, + "learning_rate": 4.487874153375574e-06, + "loss": 2.658, + "step": 4738 + }, + { + "epoch": 1.03, + "learning_rate": 4.487764911514092e-06, + "loss": 3.1906, + "step": 4739 + }, + { + "epoch": 1.03, + "learning_rate": 4.487655669652611e-06, + "loss": 3.0514, + "step": 4740 + }, + { + "epoch": 1.03, + "learning_rate": 4.48754642779113e-06, + "loss": 2.8201, + "step": 4741 + }, + { + "epoch": 1.03, + "learning_rate": 4.4874371859296484e-06, + "loss": 2.6421, + "step": 4742 + }, + { + "epoch": 1.04, + "learning_rate": 4.4873279440681675e-06, + "loss": 3.2414, + "step": 4743 + }, + { + "epoch": 1.04, + "learning_rate": 4.487218702206686e-06, + "loss": 2.9907, + "step": 4744 + }, + { + "epoch": 1.04, + "learning_rate": 4.487109460345205e-06, + "loss": 2.6018, + "step": 4745 + }, + { + "epoch": 1.04, + "learning_rate": 4.487000218483724e-06, + "loss": 3.367, + "step": 4746 + }, + { + "epoch": 1.04, + "learning_rate": 4.486890976622242e-06, + "loss": 3.4427, + "step": 4747 + }, + { + "epoch": 1.04, + "learning_rate": 4.486781734760761e-06, + "loss": 2.3712, + "step": 4748 + }, + { + "epoch": 1.04, + "learning_rate": 4.486672492899279e-06, + "loss": 3.2494, + "step": 4749 + }, + { + "epoch": 1.04, + "learning_rate": 4.486563251037798e-06, + "loss": 2.7361, + "step": 4750 + }, + { + "epoch": 1.04, + "learning_rate": 4.486454009176316e-06, + "loss": 3.6262, + "step": 4751 + }, + { + "epoch": 1.04, + "learning_rate": 4.486344767314835e-06, + "loss": 3.2546, + "step": 4752 + }, + { + "epoch": 1.04, + "learning_rate": 4.486235525453354e-06, + "loss": 3.355, + "step": 4753 + }, + { + "epoch": 1.04, + "learning_rate": 4.486126283591873e-06, + "loss": 2.892, + "step": 4754 + }, + { + "epoch": 1.04, + "learning_rate": 4.486017041730392e-06, + "loss": 2.4701, + "step": 4755 + }, + { + "epoch": 1.04, + "learning_rate": 4.48590779986891e-06, + "loss": 2.0306, + "step": 4756 + }, + { + "epoch": 1.04, + "learning_rate": 4.485798558007429e-06, + "loss": 2.46, + "step": 4757 + }, + { + "epoch": 1.04, + "learning_rate": 4.485689316145948e-06, + "loss": 3.2981, + "step": 4758 + }, + { + "epoch": 1.04, + "learning_rate": 4.485580074284466e-06, + "loss": 3.384, + "step": 4759 + }, + { + "epoch": 1.04, + "learning_rate": 4.485470832422985e-06, + "loss": 3.6026, + "step": 4760 + }, + { + "epoch": 1.04, + "learning_rate": 4.485361590561503e-06, + "loss": 3.6392, + "step": 4761 + }, + { + "epoch": 1.04, + "learning_rate": 4.485252348700022e-06, + "loss": 3.6466, + "step": 4762 + }, + { + "epoch": 1.04, + "learning_rate": 4.4851431068385405e-06, + "loss": 3.1003, + "step": 4763 + }, + { + "epoch": 1.04, + "learning_rate": 4.4850338649770596e-06, + "loss": 3.7302, + "step": 4764 + }, + { + "epoch": 1.04, + "learning_rate": 4.484924623115579e-06, + "loss": 3.4904, + "step": 4765 + }, + { + "epoch": 1.04, + "learning_rate": 4.484815381254097e-06, + "loss": 3.5111, + "step": 4766 + }, + { + "epoch": 1.04, + "learning_rate": 4.484706139392616e-06, + "loss": 3.2617, + "step": 4767 + }, + { + "epoch": 1.04, + "learning_rate": 4.484596897531134e-06, + "loss": 3.1169, + "step": 4768 + }, + { + "epoch": 1.04, + "learning_rate": 4.484487655669653e-06, + "loss": 3.1289, + "step": 4769 + }, + { + "epoch": 1.04, + "learning_rate": 4.484378413808171e-06, + "loss": 3.7455, + "step": 4770 + }, + { + "epoch": 1.04, + "learning_rate": 4.48426917194669e-06, + "loss": 2.8624, + "step": 4771 + }, + { + "epoch": 1.04, + "learning_rate": 4.484159930085209e-06, + "loss": 3.5859, + "step": 4772 + }, + { + "epoch": 1.04, + "learning_rate": 4.4840506882237275e-06, + "loss": 3.0672, + "step": 4773 + }, + { + "epoch": 1.04, + "learning_rate": 4.4839414463622465e-06, + "loss": 3.0721, + "step": 4774 + }, + { + "epoch": 1.04, + "learning_rate": 4.483832204500765e-06, + "loss": 3.6292, + "step": 4775 + }, + { + "epoch": 1.04, + "learning_rate": 4.483722962639284e-06, + "loss": 3.1838, + "step": 4776 + }, + { + "epoch": 1.04, + "learning_rate": 4.483613720777803e-06, + "loss": 3.0906, + "step": 4777 + }, + { + "epoch": 1.04, + "learning_rate": 4.483504478916321e-06, + "loss": 3.2601, + "step": 4778 + }, + { + "epoch": 1.04, + "learning_rate": 4.48339523705484e-06, + "loss": 3.2555, + "step": 4779 + }, + { + "epoch": 1.04, + "learning_rate": 4.483285995193358e-06, + "loss": 3.3608, + "step": 4780 + }, + { + "epoch": 1.04, + "learning_rate": 4.483176753331877e-06, + "loss": 3.294, + "step": 4781 + }, + { + "epoch": 1.04, + "learning_rate": 4.483067511470396e-06, + "loss": 3.5142, + "step": 4782 + }, + { + "epoch": 1.04, + "learning_rate": 4.4829582696089144e-06, + "loss": 3.5371, + "step": 4783 + }, + { + "epoch": 1.04, + "learning_rate": 4.482849027747433e-06, + "loss": 2.6687, + "step": 4784 + }, + { + "epoch": 1.04, + "learning_rate": 4.482739785885952e-06, + "loss": 3.0878, + "step": 4785 + }, + { + "epoch": 1.04, + "learning_rate": 4.482630544024471e-06, + "loss": 2.6119, + "step": 4786 + }, + { + "epoch": 1.04, + "learning_rate": 4.48252130216299e-06, + "loss": 3.2527, + "step": 4787 + }, + { + "epoch": 1.04, + "learning_rate": 4.482412060301508e-06, + "loss": 3.0822, + "step": 4788 + }, + { + "epoch": 1.05, + "learning_rate": 4.482302818440026e-06, + "loss": 3.5695, + "step": 4789 + }, + { + "epoch": 1.05, + "learning_rate": 4.482193576578545e-06, + "loss": 3.5458, + "step": 4790 + }, + { + "epoch": 1.05, + "learning_rate": 4.482084334717064e-06, + "loss": 2.9447, + "step": 4791 + }, + { + "epoch": 1.05, + "learning_rate": 4.481975092855582e-06, + "loss": 2.9228, + "step": 4792 + }, + { + "epoch": 1.05, + "learning_rate": 4.481865850994101e-06, + "loss": 3.3978, + "step": 4793 + }, + { + "epoch": 1.05, + "learning_rate": 4.48175660913262e-06, + "loss": 3.5581, + "step": 4794 + }, + { + "epoch": 1.05, + "learning_rate": 4.481647367271139e-06, + "loss": 2.6557, + "step": 4795 + }, + { + "epoch": 1.05, + "learning_rate": 4.481538125409658e-06, + "loss": 3.1399, + "step": 4796 + }, + { + "epoch": 1.05, + "learning_rate": 4.481428883548176e-06, + "loss": 3.2209, + "step": 4797 + }, + { + "epoch": 1.05, + "learning_rate": 4.481319641686695e-06, + "loss": 3.2226, + "step": 4798 + }, + { + "epoch": 1.05, + "learning_rate": 4.481210399825213e-06, + "loss": 3.5562, + "step": 4799 + }, + { + "epoch": 1.05, + "learning_rate": 4.481101157963732e-06, + "loss": 3.6015, + "step": 4800 + }, + { + "epoch": 1.05, + "learning_rate": 4.480991916102251e-06, + "loss": 3.71, + "step": 4801 + }, + { + "epoch": 1.05, + "learning_rate": 4.480882674240769e-06, + "loss": 3.1076, + "step": 4802 + }, + { + "epoch": 1.05, + "learning_rate": 4.4807734323792875e-06, + "loss": 3.2261, + "step": 4803 + }, + { + "epoch": 1.05, + "learning_rate": 4.4806641905178065e-06, + "loss": 3.3501, + "step": 4804 + }, + { + "epoch": 1.05, + "learning_rate": 4.4805549486563256e-06, + "loss": 3.7534, + "step": 4805 + }, + { + "epoch": 1.05, + "learning_rate": 4.480445706794845e-06, + "loss": 3.0369, + "step": 4806 + }, + { + "epoch": 1.05, + "learning_rate": 4.480336464933363e-06, + "loss": 3.692, + "step": 4807 + }, + { + "epoch": 1.05, + "learning_rate": 4.480227223071881e-06, + "loss": 3.1777, + "step": 4808 + }, + { + "epoch": 1.05, + "learning_rate": 4.4801179812104e-06, + "loss": 3.0285, + "step": 4809 + }, + { + "epoch": 1.05, + "learning_rate": 4.480008739348919e-06, + "loss": 2.9852, + "step": 4810 + }, + { + "epoch": 1.05, + "learning_rate": 4.479899497487438e-06, + "loss": 3.4125, + "step": 4811 + }, + { + "epoch": 1.05, + "learning_rate": 4.479790255625956e-06, + "loss": 2.9007, + "step": 4812 + }, + { + "epoch": 1.05, + "learning_rate": 4.4796810137644745e-06, + "loss": 3.2854, + "step": 4813 + }, + { + "epoch": 1.05, + "learning_rate": 4.4795717719029935e-06, + "loss": 3.4595, + "step": 4814 + }, + { + "epoch": 1.05, + "learning_rate": 4.4794625300415125e-06, + "loss": 3.3079, + "step": 4815 + }, + { + "epoch": 1.05, + "learning_rate": 4.479353288180031e-06, + "loss": 3.3346, + "step": 4816 + }, + { + "epoch": 1.05, + "learning_rate": 4.47924404631855e-06, + "loss": 3.0062, + "step": 4817 + }, + { + "epoch": 1.05, + "learning_rate": 4.479134804457068e-06, + "loss": 3.3377, + "step": 4818 + }, + { + "epoch": 1.05, + "learning_rate": 4.479025562595587e-06, + "loss": 3.1444, + "step": 4819 + }, + { + "epoch": 1.05, + "learning_rate": 4.478916320734106e-06, + "loss": 3.056, + "step": 4820 + }, + { + "epoch": 1.05, + "learning_rate": 4.478807078872624e-06, + "loss": 2.4893, + "step": 4821 + }, + { + "epoch": 1.05, + "learning_rate": 4.478697837011142e-06, + "loss": 2.872, + "step": 4822 + }, + { + "epoch": 1.05, + "learning_rate": 4.478588595149661e-06, + "loss": 3.9466, + "step": 4823 + }, + { + "epoch": 1.05, + "learning_rate": 4.4784793532881805e-06, + "loss": 2.3229, + "step": 4824 + }, + { + "epoch": 1.05, + "learning_rate": 4.4783701114266995e-06, + "loss": 2.794, + "step": 4825 + }, + { + "epoch": 1.05, + "learning_rate": 4.478260869565218e-06, + "loss": 2.2121, + "step": 4826 + }, + { + "epoch": 1.05, + "learning_rate": 4.478151627703736e-06, + "loss": 2.9694, + "step": 4827 + }, + { + "epoch": 1.05, + "learning_rate": 4.478042385842255e-06, + "loss": 2.6391, + "step": 4828 + }, + { + "epoch": 1.05, + "learning_rate": 4.477933143980774e-06, + "loss": 3.123, + "step": 4829 + }, + { + "epoch": 1.05, + "learning_rate": 4.477823902119293e-06, + "loss": 2.7713, + "step": 4830 + }, + { + "epoch": 1.05, + "learning_rate": 4.477714660257811e-06, + "loss": 3.1465, + "step": 4831 + }, + { + "epoch": 1.05, + "learning_rate": 4.477605418396329e-06, + "loss": 3.7957, + "step": 4832 + }, + { + "epoch": 1.05, + "learning_rate": 4.477496176534848e-06, + "loss": 3.2081, + "step": 4833 + }, + { + "epoch": 1.05, + "learning_rate": 4.477386934673367e-06, + "loss": 2.8752, + "step": 4834 + }, + { + "epoch": 1.06, + "learning_rate": 4.4772776928118864e-06, + "loss": 3.3754, + "step": 4835 + }, + { + "epoch": 1.06, + "learning_rate": 4.477168450950405e-06, + "loss": 3.2246, + "step": 4836 + }, + { + "epoch": 1.06, + "learning_rate": 4.477059209088923e-06, + "loss": 3.4082, + "step": 4837 + }, + { + "epoch": 1.06, + "learning_rate": 4.476949967227442e-06, + "loss": 3.1439, + "step": 4838 + }, + { + "epoch": 1.06, + "learning_rate": 4.476840725365961e-06, + "loss": 3.7791, + "step": 4839 + }, + { + "epoch": 1.06, + "learning_rate": 4.47673148350448e-06, + "loss": 3.5389, + "step": 4840 + }, + { + "epoch": 1.06, + "learning_rate": 4.476622241642998e-06, + "loss": 3.5593, + "step": 4841 + }, + { + "epoch": 1.06, + "learning_rate": 4.476512999781516e-06, + "loss": 3.3868, + "step": 4842 + }, + { + "epoch": 1.06, + "learning_rate": 4.476403757920035e-06, + "loss": 3.2748, + "step": 4843 + }, + { + "epoch": 1.06, + "learning_rate": 4.476294516058554e-06, + "loss": 3.737, + "step": 4844 + }, + { + "epoch": 1.06, + "learning_rate": 4.4761852741970726e-06, + "loss": 2.6399, + "step": 4845 + }, + { + "epoch": 1.06, + "learning_rate": 4.476076032335592e-06, + "loss": 2.9583, + "step": 4846 + }, + { + "epoch": 1.06, + "learning_rate": 4.47596679047411e-06, + "loss": 2.6739, + "step": 4847 + }, + { + "epoch": 1.06, + "learning_rate": 4.475857548612629e-06, + "loss": 3.4935, + "step": 4848 + }, + { + "epoch": 1.06, + "learning_rate": 4.475748306751148e-06, + "loss": 3.2172, + "step": 4849 + }, + { + "epoch": 1.06, + "learning_rate": 4.475639064889666e-06, + "loss": 2.7136, + "step": 4850 + }, + { + "epoch": 1.06, + "learning_rate": 4.475529823028184e-06, + "loss": 2.8389, + "step": 4851 + }, + { + "epoch": 1.06, + "learning_rate": 4.475420581166703e-06, + "loss": 3.4139, + "step": 4852 + }, + { + "epoch": 1.06, + "learning_rate": 4.475311339305222e-06, + "loss": 3.6841, + "step": 4853 + }, + { + "epoch": 1.06, + "learning_rate": 4.475202097443741e-06, + "loss": 3.5877, + "step": 4854 + }, + { + "epoch": 1.06, + "learning_rate": 4.4750928555822595e-06, + "loss": 3.5027, + "step": 4855 + }, + { + "epoch": 1.06, + "learning_rate": 4.474983613720778e-06, + "loss": 3.4598, + "step": 4856 + }, + { + "epoch": 1.06, + "learning_rate": 4.474874371859297e-06, + "loss": 2.953, + "step": 4857 + }, + { + "epoch": 1.06, + "learning_rate": 4.474765129997816e-06, + "loss": 3.596, + "step": 4858 + }, + { + "epoch": 1.06, + "learning_rate": 4.474655888136335e-06, + "loss": 3.8041, + "step": 4859 + }, + { + "epoch": 1.06, + "learning_rate": 4.474546646274853e-06, + "loss": 3.0164, + "step": 4860 + }, + { + "epoch": 1.06, + "learning_rate": 4.474437404413371e-06, + "loss": 2.9067, + "step": 4861 + }, + { + "epoch": 1.06, + "learning_rate": 4.47432816255189e-06, + "loss": 2.0277, + "step": 4862 + }, + { + "epoch": 1.06, + "learning_rate": 4.474218920690409e-06, + "loss": 3.524, + "step": 4863 + }, + { + "epoch": 1.06, + "learning_rate": 4.474109678828928e-06, + "loss": 2.9176, + "step": 4864 + }, + { + "epoch": 1.06, + "learning_rate": 4.4740004369674465e-06, + "loss": 3.8801, + "step": 4865 + }, + { + "epoch": 1.06, + "learning_rate": 4.473891195105965e-06, + "loss": 3.5072, + "step": 4866 + }, + { + "epoch": 1.06, + "learning_rate": 4.473781953244484e-06, + "loss": 3.6081, + "step": 4867 + }, + { + "epoch": 1.06, + "learning_rate": 4.473672711383003e-06, + "loss": 3.3444, + "step": 4868 + }, + { + "epoch": 1.06, + "learning_rate": 4.473563469521521e-06, + "loss": 3.7914, + "step": 4869 + }, + { + "epoch": 1.06, + "learning_rate": 4.47345422766004e-06, + "loss": 3.4657, + "step": 4870 + }, + { + "epoch": 1.06, + "learning_rate": 4.473344985798558e-06, + "loss": 3.3438, + "step": 4871 + }, + { + "epoch": 1.06, + "learning_rate": 4.473235743937077e-06, + "loss": 3.3424, + "step": 4872 + }, + { + "epoch": 1.06, + "learning_rate": 4.473126502075596e-06, + "loss": 3.7043, + "step": 4873 + }, + { + "epoch": 1.06, + "learning_rate": 4.473017260214114e-06, + "loss": 2.8571, + "step": 4874 + }, + { + "epoch": 1.06, + "learning_rate": 4.472908018352633e-06, + "loss": 2.8133, + "step": 4875 + }, + { + "epoch": 1.06, + "learning_rate": 4.472798776491152e-06, + "loss": 3.0922, + "step": 4876 + }, + { + "epoch": 1.06, + "learning_rate": 4.472689534629671e-06, + "loss": 3.0028, + "step": 4877 + }, + { + "epoch": 1.06, + "learning_rate": 4.47258029276819e-06, + "loss": 3.5327, + "step": 4878 + }, + { + "epoch": 1.06, + "learning_rate": 4.472471050906708e-06, + "loss": 2.9232, + "step": 4879 + }, + { + "epoch": 1.07, + "learning_rate": 4.472361809045226e-06, + "loss": 3.7331, + "step": 4880 + }, + { + "epoch": 1.07, + "learning_rate": 4.472252567183745e-06, + "loss": 3.8479, + "step": 4881 + }, + { + "epoch": 1.07, + "learning_rate": 4.472143325322264e-06, + "loss": 3.2395, + "step": 4882 + }, + { + "epoch": 1.07, + "learning_rate": 4.472034083460782e-06, + "loss": 2.9919, + "step": 4883 + }, + { + "epoch": 1.07, + "learning_rate": 4.471924841599301e-06, + "loss": 3.0679, + "step": 4884 + }, + { + "epoch": 1.07, + "learning_rate": 4.4718155997378195e-06, + "loss": 2.7795, + "step": 4885 + }, + { + "epoch": 1.07, + "learning_rate": 4.4717063578763386e-06, + "loss": 3.495, + "step": 4886 + }, + { + "epoch": 1.07, + "learning_rate": 4.471597116014858e-06, + "loss": 2.7891, + "step": 4887 + }, + { + "epoch": 1.07, + "learning_rate": 4.471487874153376e-06, + "loss": 3.2265, + "step": 4888 + }, + { + "epoch": 1.07, + "learning_rate": 4.471378632291895e-06, + "loss": 2.6327, + "step": 4889 + }, + { + "epoch": 1.07, + "learning_rate": 4.471269390430413e-06, + "loss": 3.0196, + "step": 4890 + }, + { + "epoch": 1.07, + "learning_rate": 4.471160148568932e-06, + "loss": 2.8518, + "step": 4891 + }, + { + "epoch": 1.07, + "learning_rate": 4.471050906707451e-06, + "loss": 2.7904, + "step": 4892 + }, + { + "epoch": 1.07, + "learning_rate": 4.470941664845969e-06, + "loss": 3.0408, + "step": 4893 + }, + { + "epoch": 1.07, + "learning_rate": 4.470832422984488e-06, + "loss": 2.3031, + "step": 4894 + }, + { + "epoch": 1.07, + "learning_rate": 4.4707231811230065e-06, + "loss": 3.8177, + "step": 4895 + }, + { + "epoch": 1.07, + "learning_rate": 4.4706139392615255e-06, + "loss": 2.3736, + "step": 4896 + }, + { + "epoch": 1.07, + "learning_rate": 4.4705046974000446e-06, + "loss": 2.877, + "step": 4897 + }, + { + "epoch": 1.07, + "learning_rate": 4.470395455538563e-06, + "loss": 3.1771, + "step": 4898 + }, + { + "epoch": 1.07, + "learning_rate": 4.470286213677082e-06, + "loss": 3.0432, + "step": 4899 + }, + { + "epoch": 1.07, + "learning_rate": 4.4701769718156e-06, + "loss": 2.7006, + "step": 4900 + }, + { + "epoch": 1.07, + "learning_rate": 4.470067729954119e-06, + "loss": 3.6245, + "step": 4901 + }, + { + "epoch": 1.07, + "learning_rate": 4.469958488092637e-06, + "loss": 2.7489, + "step": 4902 + }, + { + "epoch": 1.07, + "learning_rate": 4.469849246231156e-06, + "loss": 3.365, + "step": 4903 + }, + { + "epoch": 1.07, + "learning_rate": 4.469740004369674e-06, + "loss": 3.1468, + "step": 4904 + }, + { + "epoch": 1.07, + "learning_rate": 4.4696307625081935e-06, + "loss": 2.404, + "step": 4905 + }, + { + "epoch": 1.07, + "learning_rate": 4.4695215206467125e-06, + "loss": 3.3861, + "step": 4906 + }, + { + "epoch": 1.07, + "learning_rate": 4.469412278785231e-06, + "loss": 3.691, + "step": 4907 + }, + { + "epoch": 1.07, + "learning_rate": 4.46930303692375e-06, + "loss": 2.2794, + "step": 4908 + }, + { + "epoch": 1.07, + "learning_rate": 4.469193795062268e-06, + "loss": 3.3697, + "step": 4909 + }, + { + "epoch": 1.07, + "learning_rate": 4.469084553200787e-06, + "loss": 3.7071, + "step": 4910 + }, + { + "epoch": 1.07, + "learning_rate": 4.468975311339306e-06, + "loss": 3.4062, + "step": 4911 + }, + { + "epoch": 1.07, + "learning_rate": 4.468866069477824e-06, + "loss": 2.8846, + "step": 4912 + }, + { + "epoch": 1.07, + "learning_rate": 4.468756827616343e-06, + "loss": 3.8813, + "step": 4913 + }, + { + "epoch": 1.07, + "learning_rate": 4.468647585754861e-06, + "loss": 3.3369, + "step": 4914 + }, + { + "epoch": 1.07, + "learning_rate": 4.46853834389338e-06, + "loss": 3.4197, + "step": 4915 + }, + { + "epoch": 1.07, + "learning_rate": 4.468429102031899e-06, + "loss": 2.6955, + "step": 4916 + }, + { + "epoch": 1.07, + "learning_rate": 4.468319860170418e-06, + "loss": 2.9669, + "step": 4917 + }, + { + "epoch": 1.07, + "learning_rate": 4.468210618308937e-06, + "loss": 4.0432, + "step": 4918 + }, + { + "epoch": 1.07, + "learning_rate": 4.468101376447455e-06, + "loss": 3.5012, + "step": 4919 + }, + { + "epoch": 1.07, + "learning_rate": 4.467992134585974e-06, + "loss": 3.9176, + "step": 4920 + }, + { + "epoch": 1.07, + "learning_rate": 4.467882892724492e-06, + "loss": 4.3111, + "step": 4921 + }, + { + "epoch": 1.07, + "learning_rate": 4.467773650863011e-06, + "loss": 3.2776, + "step": 4922 + }, + { + "epoch": 1.07, + "learning_rate": 4.46766440900153e-06, + "loss": 2.9589, + "step": 4923 + }, + { + "epoch": 1.07, + "learning_rate": 4.467555167140048e-06, + "loss": 3.5765, + "step": 4924 + }, + { + "epoch": 1.07, + "learning_rate": 4.467445925278567e-06, + "loss": 3.3543, + "step": 4925 + }, + { + "epoch": 1.08, + "learning_rate": 4.4673366834170856e-06, + "loss": 3.7948, + "step": 4926 + }, + { + "epoch": 1.08, + "learning_rate": 4.467227441555605e-06, + "loss": 3.5756, + "step": 4927 + }, + { + "epoch": 1.08, + "learning_rate": 4.467118199694123e-06, + "loss": 2.9005, + "step": 4928 + }, + { + "epoch": 1.08, + "learning_rate": 4.467008957832642e-06, + "loss": 3.2786, + "step": 4929 + }, + { + "epoch": 1.08, + "learning_rate": 4.466899715971161e-06, + "loss": 3.1659, + "step": 4930 + }, + { + "epoch": 1.08, + "learning_rate": 4.466790474109679e-06, + "loss": 3.0824, + "step": 4931 + }, + { + "epoch": 1.08, + "learning_rate": 4.466681232248198e-06, + "loss": 2.653, + "step": 4932 + }, + { + "epoch": 1.08, + "learning_rate": 4.466571990386716e-06, + "loss": 3.0135, + "step": 4933 + }, + { + "epoch": 1.08, + "learning_rate": 4.466462748525235e-06, + "loss": 3.6574, + "step": 4934 + }, + { + "epoch": 1.08, + "learning_rate": 4.4663535066637535e-06, + "loss": 3.1656, + "step": 4935 + }, + { + "epoch": 1.08, + "learning_rate": 4.4662442648022725e-06, + "loss": 2.7619, + "step": 4936 + }, + { + "epoch": 1.08, + "learning_rate": 4.4661350229407915e-06, + "loss": 2.8655, + "step": 4937 + }, + { + "epoch": 1.08, + "learning_rate": 4.46602578107931e-06, + "loss": 3.7829, + "step": 4938 + }, + { + "epoch": 1.08, + "learning_rate": 4.465916539217829e-06, + "loss": 3.3544, + "step": 4939 + }, + { + "epoch": 1.08, + "learning_rate": 4.465807297356347e-06, + "loss": 3.2335, + "step": 4940 + }, + { + "epoch": 1.08, + "learning_rate": 4.465698055494866e-06, + "loss": 3.9331, + "step": 4941 + }, + { + "epoch": 1.08, + "learning_rate": 4.465588813633385e-06, + "loss": 3.321, + "step": 4942 + }, + { + "epoch": 1.08, + "learning_rate": 4.465479571771903e-06, + "loss": 3.3273, + "step": 4943 + }, + { + "epoch": 1.08, + "learning_rate": 4.465370329910422e-06, + "loss": 3.3721, + "step": 4944 + }, + { + "epoch": 1.08, + "learning_rate": 4.4652610880489404e-06, + "loss": 3.1992, + "step": 4945 + }, + { + "epoch": 1.08, + "learning_rate": 4.4651518461874595e-06, + "loss": 3.1355, + "step": 4946 + }, + { + "epoch": 1.08, + "learning_rate": 4.4650426043259785e-06, + "loss": 3.2603, + "step": 4947 + }, + { + "epoch": 1.08, + "learning_rate": 4.464933362464497e-06, + "loss": 2.4172, + "step": 4948 + }, + { + "epoch": 1.08, + "learning_rate": 4.464824120603015e-06, + "loss": 2.7869, + "step": 4949 + }, + { + "epoch": 1.08, + "learning_rate": 4.464714878741534e-06, + "loss": 3.2556, + "step": 4950 + }, + { + "epoch": 1.08, + "learning_rate": 4.464605636880053e-06, + "loss": 3.1876, + "step": 4951 + }, + { + "epoch": 1.08, + "learning_rate": 4.464496395018572e-06, + "loss": 2.7291, + "step": 4952 + }, + { + "epoch": 1.08, + "learning_rate": 4.46438715315709e-06, + "loss": 3.0742, + "step": 4953 + }, + { + "epoch": 1.08, + "learning_rate": 4.464277911295608e-06, + "loss": 3.7675, + "step": 4954 + }, + { + "epoch": 1.08, + "learning_rate": 4.464168669434127e-06, + "loss": 2.636, + "step": 4955 + }, + { + "epoch": 1.08, + "learning_rate": 4.464059427572646e-06, + "loss": 3.2645, + "step": 4956 + }, + { + "epoch": 1.08, + "learning_rate": 4.463950185711165e-06, + "loss": 3.3043, + "step": 4957 + }, + { + "epoch": 1.08, + "learning_rate": 4.463840943849684e-06, + "loss": 2.4105, + "step": 4958 + }, + { + "epoch": 1.08, + "learning_rate": 4.463731701988202e-06, + "loss": 3.0378, + "step": 4959 + }, + { + "epoch": 1.08, + "learning_rate": 4.463622460126721e-06, + "loss": 2.463, + "step": 4960 + }, + { + "epoch": 1.08, + "learning_rate": 4.46351321826524e-06, + "loss": 3.634, + "step": 4961 + }, + { + "epoch": 1.08, + "learning_rate": 4.463403976403758e-06, + "loss": 3.6383, + "step": 4962 + }, + { + "epoch": 1.08, + "learning_rate": 4.463294734542277e-06, + "loss": 3.3306, + "step": 4963 + }, + { + "epoch": 1.08, + "learning_rate": 4.463185492680795e-06, + "loss": 3.1278, + "step": 4964 + }, + { + "epoch": 1.08, + "learning_rate": 4.463076250819314e-06, + "loss": 3.2428, + "step": 4965 + }, + { + "epoch": 1.08, + "learning_rate": 4.462967008957833e-06, + "loss": 3.1956, + "step": 4966 + }, + { + "epoch": 1.08, + "learning_rate": 4.4628577670963516e-06, + "loss": 3.917, + "step": 4967 + }, + { + "epoch": 1.08, + "learning_rate": 4.46274852523487e-06, + "loss": 3.5724, + "step": 4968 + }, + { + "epoch": 1.08, + "learning_rate": 4.462639283373389e-06, + "loss": 2.6924, + "step": 4969 + }, + { + "epoch": 1.08, + "learning_rate": 4.462530041511908e-06, + "loss": 3.3659, + "step": 4970 + }, + { + "epoch": 1.08, + "learning_rate": 4.462420799650427e-06, + "loss": 3.1535, + "step": 4971 + }, + { + "epoch": 1.09, + "learning_rate": 4.462311557788945e-06, + "loss": 2.9996, + "step": 4972 + }, + { + "epoch": 1.09, + "learning_rate": 4.462202315927463e-06, + "loss": 2.7987, + "step": 4973 + }, + { + "epoch": 1.09, + "learning_rate": 4.462093074065982e-06, + "loss": 3.2604, + "step": 4974 + }, + { + "epoch": 1.09, + "learning_rate": 4.461983832204501e-06, + "loss": 3.4834, + "step": 4975 + }, + { + "epoch": 1.09, + "learning_rate": 4.46187459034302e-06, + "loss": 2.3659, + "step": 4976 + }, + { + "epoch": 1.09, + "learning_rate": 4.4617653484815385e-06, + "loss": 2.839, + "step": 4977 + }, + { + "epoch": 1.09, + "learning_rate": 4.461656106620057e-06, + "loss": 2.7837, + "step": 4978 + }, + { + "epoch": 1.09, + "learning_rate": 4.461546864758576e-06, + "loss": 3.7326, + "step": 4979 + }, + { + "epoch": 1.09, + "learning_rate": 4.461437622897095e-06, + "loss": 2.9989, + "step": 4980 + }, + { + "epoch": 1.09, + "learning_rate": 4.461328381035613e-06, + "loss": 3.0259, + "step": 4981 + }, + { + "epoch": 1.09, + "learning_rate": 4.461219139174132e-06, + "loss": 3.245, + "step": 4982 + }, + { + "epoch": 1.09, + "learning_rate": 4.46110989731265e-06, + "loss": 3.2345, + "step": 4983 + }, + { + "epoch": 1.09, + "learning_rate": 4.461000655451169e-06, + "loss": 2.9406, + "step": 4984 + }, + { + "epoch": 1.09, + "learning_rate": 4.460891413589688e-06, + "loss": 3.824, + "step": 4985 + }, + { + "epoch": 1.09, + "learning_rate": 4.4607821717282064e-06, + "loss": 3.5093, + "step": 4986 + }, + { + "epoch": 1.09, + "learning_rate": 4.4606729298667255e-06, + "loss": 2.6255, + "step": 4987 + }, + { + "epoch": 1.09, + "learning_rate": 4.460563688005244e-06, + "loss": 2.835, + "step": 4988 + }, + { + "epoch": 1.09, + "learning_rate": 4.460454446143763e-06, + "loss": 3.2738, + "step": 4989 + }, + { + "epoch": 1.09, + "learning_rate": 4.460345204282282e-06, + "loss": 3.2616, + "step": 4990 + }, + { + "epoch": 1.09, + "learning_rate": 4.4602359624208e-06, + "loss": 2.8059, + "step": 4991 + }, + { + "epoch": 1.09, + "learning_rate": 4.460126720559318e-06, + "loss": 3.6521, + "step": 4992 + }, + { + "epoch": 1.09, + "learning_rate": 4.460017478697837e-06, + "loss": 3.3653, + "step": 4993 + }, + { + "epoch": 1.09, + "learning_rate": 4.459908236836356e-06, + "loss": 3.0793, + "step": 4994 + }, + { + "epoch": 1.09, + "learning_rate": 4.459798994974875e-06, + "loss": 3.504, + "step": 4995 + }, + { + "epoch": 1.09, + "learning_rate": 4.459689753113393e-06, + "loss": 3.0541, + "step": 4996 + }, + { + "epoch": 1.09, + "learning_rate": 4.459580511251912e-06, + "loss": 3.155, + "step": 4997 + }, + { + "epoch": 1.09, + "learning_rate": 4.459471269390431e-06, + "loss": 2.9203, + "step": 4998 + }, + { + "epoch": 1.09, + "learning_rate": 4.45936202752895e-06, + "loss": 3.0295, + "step": 4999 + }, + { + "epoch": 1.09, + "learning_rate": 4.459252785667469e-06, + "loss": 3.8293, + "step": 5000 + }, + { + "epoch": 1.09, + "learning_rate": 4.459143543805987e-06, + "loss": 3.0476, + "step": 5001 + }, + { + "epoch": 1.09, + "learning_rate": 4.459034301944505e-06, + "loss": 3.3633, + "step": 5002 + }, + { + "epoch": 1.09, + "learning_rate": 4.458925060083024e-06, + "loss": 3.2607, + "step": 5003 + }, + { + "epoch": 1.09, + "learning_rate": 4.458815818221543e-06, + "loss": 3.7828, + "step": 5004 + }, + { + "epoch": 1.09, + "learning_rate": 4.458706576360062e-06, + "loss": 2.9017, + "step": 5005 + }, + { + "epoch": 1.09, + "learning_rate": 4.45859733449858e-06, + "loss": 3.3434, + "step": 5006 + }, + { + "epoch": 1.09, + "learning_rate": 4.4584880926370985e-06, + "loss": 3.0452, + "step": 5007 + }, + { + "epoch": 1.09, + "learning_rate": 4.458378850775618e-06, + "loss": 3.793, + "step": 5008 + }, + { + "epoch": 1.09, + "learning_rate": 4.458269608914137e-06, + "loss": 2.4512, + "step": 5009 + }, + { + "epoch": 1.09, + "learning_rate": 4.458160367052655e-06, + "loss": 3.505, + "step": 5010 + }, + { + "epoch": 1.09, + "learning_rate": 4.458051125191174e-06, + "loss": 3.3764, + "step": 5011 + }, + { + "epoch": 1.09, + "learning_rate": 4.457941883329692e-06, + "loss": 2.607, + "step": 5012 + }, + { + "epoch": 1.09, + "learning_rate": 4.457832641468211e-06, + "loss": 3.5017, + "step": 5013 + }, + { + "epoch": 1.09, + "learning_rate": 4.45772339960673e-06, + "loss": 3.7217, + "step": 5014 + }, + { + "epoch": 1.09, + "learning_rate": 4.457614157745248e-06, + "loss": 3.1996, + "step": 5015 + }, + { + "epoch": 1.09, + "learning_rate": 4.4575049158837665e-06, + "loss": 3.2654, + "step": 5016 + }, + { + "epoch": 1.09, + "learning_rate": 4.4573956740222855e-06, + "loss": 3.7364, + "step": 5017 + }, + { + "epoch": 1.1, + "learning_rate": 4.4572864321608045e-06, + "loss": 3.9621, + "step": 5018 + }, + { + "epoch": 1.1, + "learning_rate": 4.4571771902993236e-06, + "loss": 3.1735, + "step": 5019 + }, + { + "epoch": 1.1, + "learning_rate": 4.457067948437842e-06, + "loss": 4.0192, + "step": 5020 + }, + { + "epoch": 1.1, + "learning_rate": 4.45695870657636e-06, + "loss": 3.3233, + "step": 5021 + }, + { + "epoch": 1.1, + "learning_rate": 4.456849464714879e-06, + "loss": 3.7067, + "step": 5022 + }, + { + "epoch": 1.1, + "learning_rate": 4.456740222853398e-06, + "loss": 3.0439, + "step": 5023 + }, + { + "epoch": 1.1, + "learning_rate": 4.456630980991917e-06, + "loss": 3.0869, + "step": 5024 + }, + { + "epoch": 1.1, + "learning_rate": 4.456521739130435e-06, + "loss": 3.7164, + "step": 5025 + }, + { + "epoch": 1.1, + "learning_rate": 4.4564124972689534e-06, + "loss": 3.2653, + "step": 5026 + }, + { + "epoch": 1.1, + "learning_rate": 4.4563032554074725e-06, + "loss": 3.4146, + "step": 5027 + }, + { + "epoch": 1.1, + "learning_rate": 4.4561940135459915e-06, + "loss": 3.457, + "step": 5028 + }, + { + "epoch": 1.1, + "learning_rate": 4.45608477168451e-06, + "loss": 3.1008, + "step": 5029 + }, + { + "epoch": 1.1, + "learning_rate": 4.455975529823029e-06, + "loss": 3.1071, + "step": 5030 + }, + { + "epoch": 1.1, + "learning_rate": 4.455866287961547e-06, + "loss": 2.8494, + "step": 5031 + }, + { + "epoch": 1.1, + "learning_rate": 4.455757046100066e-06, + "loss": 3.2652, + "step": 5032 + }, + { + "epoch": 1.1, + "learning_rate": 4.455647804238585e-06, + "loss": 3.2627, + "step": 5033 + }, + { + "epoch": 1.1, + "learning_rate": 4.455538562377103e-06, + "loss": 2.8734, + "step": 5034 + }, + { + "epoch": 1.1, + "learning_rate": 4.455429320515622e-06, + "loss": 2.7409, + "step": 5035 + }, + { + "epoch": 1.1, + "learning_rate": 4.45532007865414e-06, + "loss": 3.4004, + "step": 5036 + }, + { + "epoch": 1.1, + "learning_rate": 4.455210836792659e-06, + "loss": 3.0168, + "step": 5037 + }, + { + "epoch": 1.1, + "learning_rate": 4.4551015949311785e-06, + "loss": 2.9509, + "step": 5038 + }, + { + "epoch": 1.1, + "learning_rate": 4.454992353069697e-06, + "loss": 2.7985, + "step": 5039 + }, + { + "epoch": 1.1, + "learning_rate": 4.454883111208216e-06, + "loss": 2.712, + "step": 5040 + }, + { + "epoch": 1.1, + "learning_rate": 4.454773869346734e-06, + "loss": 2.3606, + "step": 5041 + }, + { + "epoch": 1.1, + "learning_rate": 4.454664627485253e-06, + "loss": 2.8877, + "step": 5042 + }, + { + "epoch": 1.1, + "learning_rate": 4.454555385623772e-06, + "loss": 3.0416, + "step": 5043 + }, + { + "epoch": 1.1, + "learning_rate": 4.45444614376229e-06, + "loss": 3.3461, + "step": 5044 + }, + { + "epoch": 1.1, + "learning_rate": 4.454336901900808e-06, + "loss": 2.806, + "step": 5045 + }, + { + "epoch": 1.1, + "learning_rate": 4.454227660039327e-06, + "loss": 3.2346, + "step": 5046 + }, + { + "epoch": 1.1, + "learning_rate": 4.454118418177846e-06, + "loss": 3.3291, + "step": 5047 + }, + { + "epoch": 1.1, + "learning_rate": 4.4540091763163646e-06, + "loss": 3.3675, + "step": 5048 + }, + { + "epoch": 1.1, + "learning_rate": 4.453899934454884e-06, + "loss": 3.1502, + "step": 5049 + }, + { + "epoch": 1.1, + "learning_rate": 4.453790692593402e-06, + "loss": 2.9907, + "step": 5050 + }, + { + "epoch": 1.1, + "learning_rate": 4.453681450731921e-06, + "loss": 3.5687, + "step": 5051 + }, + { + "epoch": 1.1, + "learning_rate": 4.45357220887044e-06, + "loss": 3.0516, + "step": 5052 + }, + { + "epoch": 1.1, + "learning_rate": 4.453462967008958e-06, + "loss": 2.9822, + "step": 5053 + }, + { + "epoch": 1.1, + "learning_rate": 4.453353725147477e-06, + "loss": 2.3988, + "step": 5054 + }, + { + "epoch": 1.1, + "learning_rate": 4.453244483285995e-06, + "loss": 3.5434, + "step": 5055 + }, + { + "epoch": 1.1, + "learning_rate": 4.453135241424514e-06, + "loss": 3.2011, + "step": 5056 + }, + { + "epoch": 1.1, + "learning_rate": 4.453025999563033e-06, + "loss": 3.6134, + "step": 5057 + }, + { + "epoch": 1.1, + "learning_rate": 4.4529167577015515e-06, + "loss": 3.4947, + "step": 5058 + }, + { + "epoch": 1.1, + "learning_rate": 4.4528075158400706e-06, + "loss": 3.0934, + "step": 5059 + }, + { + "epoch": 1.1, + "learning_rate": 4.452698273978589e-06, + "loss": 3.1346, + "step": 5060 + }, + { + "epoch": 1.1, + "learning_rate": 4.452589032117108e-06, + "loss": 2.9887, + "step": 5061 + }, + { + "epoch": 1.1, + "learning_rate": 4.452479790255626e-06, + "loss": 3.5086, + "step": 5062 + }, + { + "epoch": 1.1, + "learning_rate": 4.452370548394145e-06, + "loss": 3.4687, + "step": 5063 + }, + { + "epoch": 1.11, + "learning_rate": 4.452261306532664e-06, + "loss": 3.2008, + "step": 5064 + }, + { + "epoch": 1.11, + "learning_rate": 4.452152064671182e-06, + "loss": 3.4744, + "step": 5065 + }, + { + "epoch": 1.11, + "learning_rate": 4.452042822809701e-06, + "loss": 3.6591, + "step": 5066 + }, + { + "epoch": 1.11, + "learning_rate": 4.4519335809482194e-06, + "loss": 3.7341, + "step": 5067 + }, + { + "epoch": 1.11, + "learning_rate": 4.4518243390867385e-06, + "loss": 3.1505, + "step": 5068 + }, + { + "epoch": 1.11, + "learning_rate": 4.451715097225257e-06, + "loss": 3.2863, + "step": 5069 + }, + { + "epoch": 1.11, + "learning_rate": 4.451605855363776e-06, + "loss": 2.7527, + "step": 5070 + }, + { + "epoch": 1.11, + "learning_rate": 4.451496613502295e-06, + "loss": 3.496, + "step": 5071 + }, + { + "epoch": 1.11, + "learning_rate": 4.451387371640813e-06, + "loss": 3.16, + "step": 5072 + }, + { + "epoch": 1.11, + "learning_rate": 4.451278129779332e-06, + "loss": 3.3455, + "step": 5073 + }, + { + "epoch": 1.11, + "learning_rate": 4.45116888791785e-06, + "loss": 3.1082, + "step": 5074 + }, + { + "epoch": 1.11, + "learning_rate": 4.451059646056369e-06, + "loss": 2.964, + "step": 5075 + }, + { + "epoch": 1.11, + "learning_rate": 4.450950404194888e-06, + "loss": 3.1952, + "step": 5076 + }, + { + "epoch": 1.11, + "learning_rate": 4.450841162333406e-06, + "loss": 3.4493, + "step": 5077 + }, + { + "epoch": 1.11, + "learning_rate": 4.4507319204719254e-06, + "loss": 3.1849, + "step": 5078 + }, + { + "epoch": 1.11, + "learning_rate": 4.450622678610444e-06, + "loss": 2.8298, + "step": 5079 + }, + { + "epoch": 1.11, + "learning_rate": 4.450513436748963e-06, + "loss": 2.5188, + "step": 5080 + }, + { + "epoch": 1.11, + "learning_rate": 4.450404194887481e-06, + "loss": 2.9615, + "step": 5081 + }, + { + "epoch": 1.11, + "learning_rate": 4.450294953026e-06, + "loss": 3.525, + "step": 5082 + }, + { + "epoch": 1.11, + "learning_rate": 4.450185711164519e-06, + "loss": 3.2981, + "step": 5083 + }, + { + "epoch": 1.11, + "learning_rate": 4.450076469303037e-06, + "loss": 2.5777, + "step": 5084 + }, + { + "epoch": 1.11, + "learning_rate": 4.449967227441556e-06, + "loss": 2.6242, + "step": 5085 + }, + { + "epoch": 1.11, + "learning_rate": 4.449857985580074e-06, + "loss": 3.6245, + "step": 5086 + }, + { + "epoch": 1.11, + "learning_rate": 4.449748743718593e-06, + "loss": 3.3469, + "step": 5087 + }, + { + "epoch": 1.11, + "learning_rate": 4.449639501857112e-06, + "loss": 3.0702, + "step": 5088 + }, + { + "epoch": 1.11, + "learning_rate": 4.449530259995631e-06, + "loss": 3.1525, + "step": 5089 + }, + { + "epoch": 1.11, + "learning_rate": 4.44942101813415e-06, + "loss": 3.2384, + "step": 5090 + }, + { + "epoch": 1.11, + "learning_rate": 4.449311776272668e-06, + "loss": 3.2298, + "step": 5091 + }, + { + "epoch": 1.11, + "learning_rate": 4.449202534411187e-06, + "loss": 3.3043, + "step": 5092 + }, + { + "epoch": 1.11, + "learning_rate": 4.449093292549706e-06, + "loss": 2.6013, + "step": 5093 + }, + { + "epoch": 1.11, + "learning_rate": 4.448984050688224e-06, + "loss": 3.722, + "step": 5094 + }, + { + "epoch": 1.11, + "learning_rate": 4.448874808826742e-06, + "loss": 2.8552, + "step": 5095 + }, + { + "epoch": 1.11, + "learning_rate": 4.448765566965261e-06, + "loss": 3.5726, + "step": 5096 + }, + { + "epoch": 1.11, + "learning_rate": 4.44865632510378e-06, + "loss": 3.0585, + "step": 5097 + }, + { + "epoch": 1.11, + "learning_rate": 4.4485470832422985e-06, + "loss": 3.7904, + "step": 5098 + }, + { + "epoch": 1.11, + "learning_rate": 4.4484378413808175e-06, + "loss": 3.0182, + "step": 5099 + }, + { + "epoch": 1.11, + "learning_rate": 4.448328599519336e-06, + "loss": 3.0373, + "step": 5100 + }, + { + "epoch": 1.11, + "learning_rate": 4.448219357657855e-06, + "loss": 3.0023, + "step": 5101 + }, + { + "epoch": 1.11, + "learning_rate": 4.448110115796374e-06, + "loss": 2.9044, + "step": 5102 + }, + { + "epoch": 1.11, + "learning_rate": 4.448000873934892e-06, + "loss": 3.0956, + "step": 5103 + }, + { + "epoch": 1.11, + "learning_rate": 4.447891632073411e-06, + "loss": 3.5432, + "step": 5104 + }, + { + "epoch": 1.11, + "learning_rate": 4.447782390211929e-06, + "loss": 3.3614, + "step": 5105 + }, + { + "epoch": 1.11, + "learning_rate": 4.447673148350448e-06, + "loss": 1.9312, + "step": 5106 + }, + { + "epoch": 1.11, + "learning_rate": 4.447563906488967e-06, + "loss": 3.5296, + "step": 5107 + }, + { + "epoch": 1.11, + "learning_rate": 4.4474546646274855e-06, + "loss": 3.6341, + "step": 5108 + }, + { + "epoch": 1.12, + "learning_rate": 4.4473454227660045e-06, + "loss": 2.7627, + "step": 5109 + }, + { + "epoch": 1.12, + "learning_rate": 4.447236180904523e-06, + "loss": 3.3762, + "step": 5110 + }, + { + "epoch": 1.12, + "learning_rate": 4.447126939043042e-06, + "loss": 2.6168, + "step": 5111 + }, + { + "epoch": 1.12, + "learning_rate": 4.447017697181561e-06, + "loss": 3.4885, + "step": 5112 + }, + { + "epoch": 1.12, + "learning_rate": 4.446908455320079e-06, + "loss": 2.4575, + "step": 5113 + }, + { + "epoch": 1.12, + "learning_rate": 4.446799213458597e-06, + "loss": 3.4976, + "step": 5114 + }, + { + "epoch": 1.12, + "learning_rate": 4.446689971597116e-06, + "loss": 2.8971, + "step": 5115 + }, + { + "epoch": 1.12, + "learning_rate": 4.446580729735635e-06, + "loss": 2.8521, + "step": 5116 + }, + { + "epoch": 1.12, + "learning_rate": 4.446471487874154e-06, + "loss": 3.6707, + "step": 5117 + }, + { + "epoch": 1.12, + "learning_rate": 4.446362246012672e-06, + "loss": 2.8036, + "step": 5118 + }, + { + "epoch": 1.12, + "learning_rate": 4.446253004151191e-06, + "loss": 3.7592, + "step": 5119 + }, + { + "epoch": 1.12, + "learning_rate": 4.44614376228971e-06, + "loss": 3.7308, + "step": 5120 + }, + { + "epoch": 1.12, + "learning_rate": 4.446034520428229e-06, + "loss": 3.5044, + "step": 5121 + }, + { + "epoch": 1.12, + "learning_rate": 4.445925278566747e-06, + "loss": 3.5806, + "step": 5122 + }, + { + "epoch": 1.12, + "learning_rate": 4.445816036705266e-06, + "loss": 3.4465, + "step": 5123 + }, + { + "epoch": 1.12, + "learning_rate": 4.445706794843784e-06, + "loss": 2.441, + "step": 5124 + }, + { + "epoch": 1.12, + "learning_rate": 4.445597552982303e-06, + "loss": 3.2642, + "step": 5125 + }, + { + "epoch": 1.12, + "learning_rate": 4.445488311120822e-06, + "loss": 3.0999, + "step": 5126 + }, + { + "epoch": 1.12, + "learning_rate": 4.44537906925934e-06, + "loss": 3.9667, + "step": 5127 + }, + { + "epoch": 1.12, + "learning_rate": 4.445269827397859e-06, + "loss": 3.6597, + "step": 5128 + }, + { + "epoch": 1.12, + "learning_rate": 4.4451605855363776e-06, + "loss": 3.4809, + "step": 5129 + }, + { + "epoch": 1.12, + "learning_rate": 4.445051343674897e-06, + "loss": 2.8802, + "step": 5130 + }, + { + "epoch": 1.12, + "learning_rate": 4.444942101813416e-06, + "loss": 2.9783, + "step": 5131 + }, + { + "epoch": 1.12, + "learning_rate": 4.444832859951934e-06, + "loss": 3.0844, + "step": 5132 + }, + { + "epoch": 1.12, + "learning_rate": 4.444723618090452e-06, + "loss": 3.8793, + "step": 5133 + }, + { + "epoch": 1.12, + "learning_rate": 4.444614376228971e-06, + "loss": 3.3223, + "step": 5134 + }, + { + "epoch": 1.12, + "learning_rate": 4.44450513436749e-06, + "loss": 2.081, + "step": 5135 + }, + { + "epoch": 1.12, + "learning_rate": 4.444395892506009e-06, + "loss": 2.957, + "step": 5136 + }, + { + "epoch": 1.12, + "learning_rate": 4.444286650644527e-06, + "loss": 2.4472, + "step": 5137 + }, + { + "epoch": 1.12, + "learning_rate": 4.4441774087830455e-06, + "loss": 2.616, + "step": 5138 + }, + { + "epoch": 1.12, + "learning_rate": 4.4440681669215645e-06, + "loss": 3.3069, + "step": 5139 + }, + { + "epoch": 1.12, + "learning_rate": 4.4439589250600835e-06, + "loss": 2.9542, + "step": 5140 + }, + { + "epoch": 1.12, + "learning_rate": 4.443849683198603e-06, + "loss": 3.4475, + "step": 5141 + }, + { + "epoch": 1.12, + "learning_rate": 4.443740441337121e-06, + "loss": 2.5835, + "step": 5142 + }, + { + "epoch": 1.12, + "learning_rate": 4.443631199475639e-06, + "loss": 2.4791, + "step": 5143 + }, + { + "epoch": 1.12, + "learning_rate": 4.443521957614158e-06, + "loss": 2.2716, + "step": 5144 + }, + { + "epoch": 1.12, + "learning_rate": 4.443412715752677e-06, + "loss": 3.133, + "step": 5145 + }, + { + "epoch": 1.12, + "learning_rate": 4.443303473891196e-06, + "loss": 2.5647, + "step": 5146 + }, + { + "epoch": 1.12, + "learning_rate": 4.443194232029714e-06, + "loss": 3.5281, + "step": 5147 + }, + { + "epoch": 1.12, + "learning_rate": 4.4430849901682324e-06, + "loss": 2.7178, + "step": 5148 + }, + { + "epoch": 1.12, + "learning_rate": 4.4429757483067515e-06, + "loss": 2.9731, + "step": 5149 + }, + { + "epoch": 1.12, + "learning_rate": 4.4428665064452705e-06, + "loss": 2.8827, + "step": 5150 + }, + { + "epoch": 1.12, + "learning_rate": 4.442757264583789e-06, + "loss": 3.3484, + "step": 5151 + }, + { + "epoch": 1.12, + "learning_rate": 4.442648022722308e-06, + "loss": 2.9427, + "step": 5152 + }, + { + "epoch": 1.12, + "learning_rate": 4.442538780860826e-06, + "loss": 3.4539, + "step": 5153 + }, + { + "epoch": 1.12, + "learning_rate": 4.442429538999345e-06, + "loss": 2.9451, + "step": 5154 + }, + { + "epoch": 1.13, + "learning_rate": 4.442320297137864e-06, + "loss": 2.2333, + "step": 5155 + }, + { + "epoch": 1.13, + "learning_rate": 4.442211055276382e-06, + "loss": 3.6006, + "step": 5156 + }, + { + "epoch": 1.13, + "learning_rate": 4.4421018134149e-06, + "loss": 2.6997, + "step": 5157 + }, + { + "epoch": 1.13, + "learning_rate": 4.441992571553419e-06, + "loss": 3.198, + "step": 5158 + }, + { + "epoch": 1.13, + "learning_rate": 4.4418833296919384e-06, + "loss": 2.6754, + "step": 5159 + }, + { + "epoch": 1.13, + "learning_rate": 4.4417740878304575e-06, + "loss": 2.8789, + "step": 5160 + }, + { + "epoch": 1.13, + "learning_rate": 4.441664845968976e-06, + "loss": 2.7611, + "step": 5161 + }, + { + "epoch": 1.13, + "learning_rate": 4.441555604107494e-06, + "loss": 3.1389, + "step": 5162 + }, + { + "epoch": 1.13, + "learning_rate": 4.441446362246013e-06, + "loss": 2.8015, + "step": 5163 + }, + { + "epoch": 1.13, + "learning_rate": 4.441337120384532e-06, + "loss": 2.8258, + "step": 5164 + }, + { + "epoch": 1.13, + "learning_rate": 4.441227878523051e-06, + "loss": 2.8361, + "step": 5165 + }, + { + "epoch": 1.13, + "learning_rate": 4.441118636661569e-06, + "loss": 2.9809, + "step": 5166 + }, + { + "epoch": 1.13, + "learning_rate": 4.441009394800087e-06, + "loss": 3.7748, + "step": 5167 + }, + { + "epoch": 1.13, + "learning_rate": 4.440900152938606e-06, + "loss": 3.4084, + "step": 5168 + }, + { + "epoch": 1.13, + "learning_rate": 4.440790911077125e-06, + "loss": 3.0927, + "step": 5169 + }, + { + "epoch": 1.13, + "learning_rate": 4.440681669215644e-06, + "loss": 3.5161, + "step": 5170 + }, + { + "epoch": 1.13, + "learning_rate": 4.440572427354163e-06, + "loss": 3.3154, + "step": 5171 + }, + { + "epoch": 1.13, + "learning_rate": 4.440463185492681e-06, + "loss": 2.5726, + "step": 5172 + }, + { + "epoch": 1.13, + "learning_rate": 4.4403539436312e-06, + "loss": 3.6077, + "step": 5173 + }, + { + "epoch": 1.13, + "learning_rate": 4.440244701769719e-06, + "loss": 2.755, + "step": 5174 + }, + { + "epoch": 1.13, + "learning_rate": 4.440135459908237e-06, + "loss": 3.2196, + "step": 5175 + }, + { + "epoch": 1.13, + "learning_rate": 4.440026218046756e-06, + "loss": 3.6374, + "step": 5176 + }, + { + "epoch": 1.13, + "learning_rate": 4.439916976185274e-06, + "loss": 2.7847, + "step": 5177 + }, + { + "epoch": 1.13, + "learning_rate": 4.439807734323793e-06, + "loss": 3.1944, + "step": 5178 + }, + { + "epoch": 1.13, + "learning_rate": 4.439698492462312e-06, + "loss": 3.0275, + "step": 5179 + }, + { + "epoch": 1.13, + "learning_rate": 4.4395892506008305e-06, + "loss": 3.5272, + "step": 5180 + }, + { + "epoch": 1.13, + "learning_rate": 4.4394800087393496e-06, + "loss": 3.3916, + "step": 5181 + }, + { + "epoch": 1.13, + "learning_rate": 4.439370766877868e-06, + "loss": 2.6741, + "step": 5182 + }, + { + "epoch": 1.13, + "learning_rate": 4.439261525016387e-06, + "loss": 3.4264, + "step": 5183 + }, + { + "epoch": 1.13, + "learning_rate": 4.439152283154906e-06, + "loss": 2.4648, + "step": 5184 + }, + { + "epoch": 1.13, + "learning_rate": 4.439043041293424e-06, + "loss": 3.0473, + "step": 5185 + }, + { + "epoch": 1.13, + "learning_rate": 4.438933799431942e-06, + "loss": 2.8772, + "step": 5186 + }, + { + "epoch": 1.13, + "learning_rate": 4.438824557570461e-06, + "loss": 3.1881, + "step": 5187 + }, + { + "epoch": 1.13, + "learning_rate": 4.43871531570898e-06, + "loss": 3.7373, + "step": 5188 + }, + { + "epoch": 1.13, + "learning_rate": 4.438606073847499e-06, + "loss": 2.727, + "step": 5189 + }, + { + "epoch": 1.13, + "learning_rate": 4.4384968319860175e-06, + "loss": 2.7882, + "step": 5190 + }, + { + "epoch": 1.13, + "learning_rate": 4.438387590124536e-06, + "loss": 2.9434, + "step": 5191 + }, + { + "epoch": 1.13, + "learning_rate": 4.438278348263055e-06, + "loss": 2.7797, + "step": 5192 + }, + { + "epoch": 1.13, + "learning_rate": 4.438169106401574e-06, + "loss": 2.9281, + "step": 5193 + }, + { + "epoch": 1.13, + "learning_rate": 4.438059864540092e-06, + "loss": 3.2548, + "step": 5194 + }, + { + "epoch": 1.13, + "learning_rate": 4.437950622678611e-06, + "loss": 3.1398, + "step": 5195 + }, + { + "epoch": 1.13, + "learning_rate": 4.437841380817129e-06, + "loss": 3.3412, + "step": 5196 + }, + { + "epoch": 1.13, + "learning_rate": 4.437732138955648e-06, + "loss": 3.3917, + "step": 5197 + }, + { + "epoch": 1.13, + "learning_rate": 4.437622897094167e-06, + "loss": 3.4297, + "step": 5198 + }, + { + "epoch": 1.13, + "learning_rate": 4.437513655232685e-06, + "loss": 3.2776, + "step": 5199 + }, + { + "epoch": 1.13, + "learning_rate": 4.4374044133712044e-06, + "loss": 3.269, + "step": 5200 + }, + { + "epoch": 1.14, + "learning_rate": 4.437295171509723e-06, + "loss": 2.7825, + "step": 5201 + }, + { + "epoch": 1.14, + "learning_rate": 4.437185929648242e-06, + "loss": 3.1525, + "step": 5202 + }, + { + "epoch": 1.14, + "learning_rate": 4.437076687786761e-06, + "loss": 3.0833, + "step": 5203 + }, + { + "epoch": 1.14, + "learning_rate": 4.436967445925279e-06, + "loss": 3.266, + "step": 5204 + }, + { + "epoch": 1.14, + "learning_rate": 4.436858204063798e-06, + "loss": 3.3075, + "step": 5205 + }, + { + "epoch": 1.14, + "learning_rate": 4.436748962202316e-06, + "loss": 3.4741, + "step": 5206 + }, + { + "epoch": 1.14, + "learning_rate": 4.436639720340835e-06, + "loss": 3.4941, + "step": 5207 + }, + { + "epoch": 1.14, + "learning_rate": 4.436530478479354e-06, + "loss": 3.1819, + "step": 5208 + }, + { + "epoch": 1.14, + "learning_rate": 4.436421236617872e-06, + "loss": 3.2496, + "step": 5209 + }, + { + "epoch": 1.14, + "learning_rate": 4.4363119947563906e-06, + "loss": 3.3703, + "step": 5210 + }, + { + "epoch": 1.14, + "learning_rate": 4.43620275289491e-06, + "loss": 3.3849, + "step": 5211 + }, + { + "epoch": 1.14, + "learning_rate": 4.436093511033429e-06, + "loss": 3.2922, + "step": 5212 + }, + { + "epoch": 1.14, + "learning_rate": 4.435984269171947e-06, + "loss": 3.4235, + "step": 5213 + }, + { + "epoch": 1.14, + "learning_rate": 4.435875027310466e-06, + "loss": 3.2493, + "step": 5214 + }, + { + "epoch": 1.14, + "learning_rate": 4.435765785448984e-06, + "loss": 3.1852, + "step": 5215 + }, + { + "epoch": 1.14, + "learning_rate": 4.435656543587503e-06, + "loss": 2.8933, + "step": 5216 + }, + { + "epoch": 1.14, + "learning_rate": 4.435547301726022e-06, + "loss": 3.2546, + "step": 5217 + }, + { + "epoch": 1.14, + "learning_rate": 4.43543805986454e-06, + "loss": 3.5054, + "step": 5218 + }, + { + "epoch": 1.14, + "learning_rate": 4.435328818003059e-06, + "loss": 3.8384, + "step": 5219 + }, + { + "epoch": 1.14, + "learning_rate": 4.4352195761415775e-06, + "loss": 3.0532, + "step": 5220 + }, + { + "epoch": 1.14, + "learning_rate": 4.4351103342800965e-06, + "loss": 2.5606, + "step": 5221 + }, + { + "epoch": 1.14, + "learning_rate": 4.435001092418616e-06, + "loss": 2.8031, + "step": 5222 + }, + { + "epoch": 1.14, + "learning_rate": 4.434891850557134e-06, + "loss": 3.8938, + "step": 5223 + }, + { + "epoch": 1.14, + "learning_rate": 4.434782608695653e-06, + "loss": 3.8295, + "step": 5224 + }, + { + "epoch": 1.14, + "learning_rate": 4.434673366834171e-06, + "loss": 3.2146, + "step": 5225 + }, + { + "epoch": 1.14, + "learning_rate": 4.43456412497269e-06, + "loss": 3.344, + "step": 5226 + }, + { + "epoch": 1.14, + "learning_rate": 4.434454883111208e-06, + "loss": 3.9566, + "step": 5227 + }, + { + "epoch": 1.14, + "learning_rate": 4.434345641249727e-06, + "loss": 3.6406, + "step": 5228 + }, + { + "epoch": 1.14, + "learning_rate": 4.434236399388246e-06, + "loss": 2.8555, + "step": 5229 + }, + { + "epoch": 1.14, + "learning_rate": 4.4341271575267645e-06, + "loss": 3.2403, + "step": 5230 + }, + { + "epoch": 1.14, + "learning_rate": 4.4340179156652835e-06, + "loss": 3.7879, + "step": 5231 + }, + { + "epoch": 1.14, + "learning_rate": 4.433908673803802e-06, + "loss": 3.2275, + "step": 5232 + }, + { + "epoch": 1.14, + "learning_rate": 4.433799431942321e-06, + "loss": 3.3275, + "step": 5233 + }, + { + "epoch": 1.14, + "learning_rate": 4.43369019008084e-06, + "loss": 3.4905, + "step": 5234 + }, + { + "epoch": 1.14, + "learning_rate": 4.433580948219358e-06, + "loss": 2.9286, + "step": 5235 + }, + { + "epoch": 1.14, + "learning_rate": 4.433471706357877e-06, + "loss": 3.3585, + "step": 5236 + }, + { + "epoch": 1.14, + "learning_rate": 4.433362464496395e-06, + "loss": 2.7745, + "step": 5237 + }, + { + "epoch": 1.14, + "learning_rate": 4.433253222634914e-06, + "loss": 3.6196, + "step": 5238 + }, + { + "epoch": 1.14, + "learning_rate": 4.433143980773432e-06, + "loss": 3.0214, + "step": 5239 + }, + { + "epoch": 1.14, + "learning_rate": 4.4330347389119514e-06, + "loss": 3.0602, + "step": 5240 + }, + { + "epoch": 1.14, + "learning_rate": 4.4329254970504705e-06, + "loss": 3.4488, + "step": 5241 + }, + { + "epoch": 1.14, + "learning_rate": 4.432816255188989e-06, + "loss": 3.1433, + "step": 5242 + }, + { + "epoch": 1.14, + "learning_rate": 4.432707013327508e-06, + "loss": 2.9128, + "step": 5243 + }, + { + "epoch": 1.14, + "learning_rate": 4.432597771466026e-06, + "loss": 3.1917, + "step": 5244 + }, + { + "epoch": 1.14, + "learning_rate": 4.432488529604545e-06, + "loss": 3.5685, + "step": 5245 + }, + { + "epoch": 1.14, + "learning_rate": 4.432379287743063e-06, + "loss": 2.699, + "step": 5246 + }, + { + "epoch": 1.15, + "learning_rate": 4.432270045881582e-06, + "loss": 3.626, + "step": 5247 + }, + { + "epoch": 1.15, + "learning_rate": 4.432160804020101e-06, + "loss": 2.9754, + "step": 5248 + }, + { + "epoch": 1.15, + "learning_rate": 4.432051562158619e-06, + "loss": 1.9708, + "step": 5249 + }, + { + "epoch": 1.15, + "learning_rate": 4.431942320297138e-06, + "loss": 3.3763, + "step": 5250 + }, + { + "epoch": 1.15, + "learning_rate": 4.4318330784356566e-06, + "loss": 3.3958, + "step": 5251 + }, + { + "epoch": 1.15, + "learning_rate": 4.431723836574176e-06, + "loss": 3.1789, + "step": 5252 + }, + { + "epoch": 1.15, + "learning_rate": 4.431614594712695e-06, + "loss": 3.6543, + "step": 5253 + }, + { + "epoch": 1.15, + "learning_rate": 4.431505352851213e-06, + "loss": 2.3502, + "step": 5254 + }, + { + "epoch": 1.15, + "learning_rate": 4.431396110989732e-06, + "loss": 3.4738, + "step": 5255 + }, + { + "epoch": 1.15, + "learning_rate": 4.43128686912825e-06, + "loss": 3.1362, + "step": 5256 + }, + { + "epoch": 1.15, + "learning_rate": 4.431177627266769e-06, + "loss": 3.0111, + "step": 5257 + }, + { + "epoch": 1.15, + "learning_rate": 4.431068385405288e-06, + "loss": 3.4033, + "step": 5258 + }, + { + "epoch": 1.15, + "learning_rate": 4.430959143543806e-06, + "loss": 2.7519, + "step": 5259 + }, + { + "epoch": 1.15, + "learning_rate": 4.4308499016823245e-06, + "loss": 2.2523, + "step": 5260 + }, + { + "epoch": 1.15, + "learning_rate": 4.4307406598208435e-06, + "loss": 3.4676, + "step": 5261 + }, + { + "epoch": 1.15, + "learning_rate": 4.4306314179593626e-06, + "loss": 2.8508, + "step": 5262 + }, + { + "epoch": 1.15, + "learning_rate": 4.430522176097881e-06, + "loss": 2.1419, + "step": 5263 + }, + { + "epoch": 1.15, + "learning_rate": 4.4304129342364e-06, + "loss": 3.2389, + "step": 5264 + }, + { + "epoch": 1.15, + "learning_rate": 4.430303692374918e-06, + "loss": 3.1424, + "step": 5265 + }, + { + "epoch": 1.15, + "learning_rate": 4.430194450513437e-06, + "loss": 2.4661, + "step": 5266 + }, + { + "epoch": 1.15, + "learning_rate": 4.430085208651956e-06, + "loss": 3.066, + "step": 5267 + }, + { + "epoch": 1.15, + "learning_rate": 4.429975966790474e-06, + "loss": 2.6297, + "step": 5268 + }, + { + "epoch": 1.15, + "learning_rate": 4.429866724928993e-06, + "loss": 3.6308, + "step": 5269 + }, + { + "epoch": 1.15, + "learning_rate": 4.4297574830675114e-06, + "loss": 2.8527, + "step": 5270 + }, + { + "epoch": 1.15, + "learning_rate": 4.4296482412060305e-06, + "loss": 2.7806, + "step": 5271 + }, + { + "epoch": 1.15, + "learning_rate": 4.4295389993445495e-06, + "loss": 2.8231, + "step": 5272 + }, + { + "epoch": 1.15, + "learning_rate": 4.429429757483068e-06, + "loss": 3.0864, + "step": 5273 + }, + { + "epoch": 1.15, + "learning_rate": 4.429320515621587e-06, + "loss": 3.3633, + "step": 5274 + }, + { + "epoch": 1.15, + "learning_rate": 4.429211273760105e-06, + "loss": 3.1094, + "step": 5275 + }, + { + "epoch": 1.15, + "learning_rate": 4.429102031898624e-06, + "loss": 3.0113, + "step": 5276 + }, + { + "epoch": 1.15, + "learning_rate": 4.428992790037143e-06, + "loss": 3.6294, + "step": 5277 + }, + { + "epoch": 1.15, + "learning_rate": 4.428883548175661e-06, + "loss": 3.5319, + "step": 5278 + }, + { + "epoch": 1.15, + "learning_rate": 4.428774306314179e-06, + "loss": 3.4475, + "step": 5279 + }, + { + "epoch": 1.15, + "learning_rate": 4.428665064452698e-06, + "loss": 3.8982, + "step": 5280 + }, + { + "epoch": 1.15, + "learning_rate": 4.4285558225912174e-06, + "loss": 3.8781, + "step": 5281 + }, + { + "epoch": 1.15, + "learning_rate": 4.4284465807297365e-06, + "loss": 3.1872, + "step": 5282 + }, + { + "epoch": 1.15, + "learning_rate": 4.428337338868255e-06, + "loss": 3.597, + "step": 5283 + }, + { + "epoch": 1.15, + "learning_rate": 4.428228097006773e-06, + "loss": 3.1566, + "step": 5284 + }, + { + "epoch": 1.15, + "learning_rate": 4.428118855145292e-06, + "loss": 3.6621, + "step": 5285 + }, + { + "epoch": 1.15, + "learning_rate": 4.428009613283811e-06, + "loss": 3.2843, + "step": 5286 + }, + { + "epoch": 1.15, + "learning_rate": 4.42790037142233e-06, + "loss": 3.264, + "step": 5287 + }, + { + "epoch": 1.15, + "learning_rate": 4.427791129560848e-06, + "loss": 3.3148, + "step": 5288 + }, + { + "epoch": 1.15, + "learning_rate": 4.427681887699366e-06, + "loss": 3.5961, + "step": 5289 + }, + { + "epoch": 1.15, + "learning_rate": 4.427572645837885e-06, + "loss": 2.4539, + "step": 5290 + }, + { + "epoch": 1.15, + "learning_rate": 4.427463403976404e-06, + "loss": 3.3354, + "step": 5291 + }, + { + "epoch": 1.15, + "learning_rate": 4.427354162114923e-06, + "loss": 3.4503, + "step": 5292 + }, + { + "epoch": 1.16, + "learning_rate": 4.427244920253442e-06, + "loss": 3.3103, + "step": 5293 + }, + { + "epoch": 1.16, + "learning_rate": 4.42713567839196e-06, + "loss": 3.8523, + "step": 5294 + }, + { + "epoch": 1.16, + "learning_rate": 4.427026436530479e-06, + "loss": 3.0508, + "step": 5295 + }, + { + "epoch": 1.16, + "learning_rate": 4.426917194668998e-06, + "loss": 2.7395, + "step": 5296 + }, + { + "epoch": 1.16, + "learning_rate": 4.426807952807516e-06, + "loss": 2.8419, + "step": 5297 + }, + { + "epoch": 1.16, + "learning_rate": 4.426698710946034e-06, + "loss": 3.0873, + "step": 5298 + }, + { + "epoch": 1.16, + "learning_rate": 4.426589469084553e-06, + "loss": 3.1385, + "step": 5299 + }, + { + "epoch": 1.16, + "learning_rate": 4.426480227223072e-06, + "loss": 3.3768, + "step": 5300 + }, + { + "epoch": 1.16, + "learning_rate": 4.426370985361591e-06, + "loss": 3.5171, + "step": 5301 + }, + { + "epoch": 1.16, + "learning_rate": 4.4262617435001095e-06, + "loss": 2.8159, + "step": 5302 + }, + { + "epoch": 1.16, + "learning_rate": 4.426152501638628e-06, + "loss": 2.6645, + "step": 5303 + }, + { + "epoch": 1.16, + "learning_rate": 4.426043259777147e-06, + "loss": 2.9755, + "step": 5304 + }, + { + "epoch": 1.16, + "learning_rate": 4.425934017915666e-06, + "loss": 3.1539, + "step": 5305 + }, + { + "epoch": 1.16, + "learning_rate": 4.425824776054185e-06, + "loss": 2.8138, + "step": 5306 + }, + { + "epoch": 1.16, + "learning_rate": 4.425715534192703e-06, + "loss": 3.1673, + "step": 5307 + }, + { + "epoch": 1.16, + "learning_rate": 4.425606292331221e-06, + "loss": 3.2156, + "step": 5308 + }, + { + "epoch": 1.16, + "learning_rate": 4.42549705046974e-06, + "loss": 2.8505, + "step": 5309 + }, + { + "epoch": 1.16, + "learning_rate": 4.425387808608259e-06, + "loss": 3.0844, + "step": 5310 + }, + { + "epoch": 1.16, + "learning_rate": 4.425278566746778e-06, + "loss": 3.3783, + "step": 5311 + }, + { + "epoch": 1.16, + "learning_rate": 4.4251693248852965e-06, + "loss": 2.9987, + "step": 5312 + }, + { + "epoch": 1.16, + "learning_rate": 4.425060083023815e-06, + "loss": 3.4923, + "step": 5313 + }, + { + "epoch": 1.16, + "learning_rate": 4.424950841162334e-06, + "loss": 3.5148, + "step": 5314 + }, + { + "epoch": 1.16, + "learning_rate": 4.424841599300853e-06, + "loss": 2.9022, + "step": 5315 + }, + { + "epoch": 1.16, + "learning_rate": 4.424732357439371e-06, + "loss": 3.5366, + "step": 5316 + }, + { + "epoch": 1.16, + "learning_rate": 4.42462311557789e-06, + "loss": 2.8349, + "step": 5317 + }, + { + "epoch": 1.16, + "learning_rate": 4.424513873716408e-06, + "loss": 3.5759, + "step": 5318 + }, + { + "epoch": 1.16, + "learning_rate": 4.424404631854927e-06, + "loss": 3.3319, + "step": 5319 + }, + { + "epoch": 1.16, + "learning_rate": 4.424295389993446e-06, + "loss": 3.1152, + "step": 5320 + }, + { + "epoch": 1.16, + "learning_rate": 4.424186148131964e-06, + "loss": 3.1756, + "step": 5321 + }, + { + "epoch": 1.16, + "learning_rate": 4.424076906270483e-06, + "loss": 3.5735, + "step": 5322 + }, + { + "epoch": 1.16, + "learning_rate": 4.423967664409002e-06, + "loss": 2.8498, + "step": 5323 + }, + { + "epoch": 1.16, + "learning_rate": 4.423858422547521e-06, + "loss": 2.9663, + "step": 5324 + }, + { + "epoch": 1.16, + "learning_rate": 4.42374918068604e-06, + "loss": 3.3324, + "step": 5325 + }, + { + "epoch": 1.16, + "learning_rate": 4.423639938824558e-06, + "loss": 3.2665, + "step": 5326 + }, + { + "epoch": 1.16, + "learning_rate": 4.423530696963076e-06, + "loss": 2.6129, + "step": 5327 + }, + { + "epoch": 1.16, + "learning_rate": 4.423421455101595e-06, + "loss": 2.7855, + "step": 5328 + }, + { + "epoch": 1.16, + "learning_rate": 4.423312213240114e-06, + "loss": 3.8852, + "step": 5329 + }, + { + "epoch": 1.16, + "learning_rate": 4.423202971378633e-06, + "loss": 2.2883, + "step": 5330 + }, + { + "epoch": 1.16, + "learning_rate": 4.423093729517151e-06, + "loss": 3.2461, + "step": 5331 + }, + { + "epoch": 1.16, + "learning_rate": 4.4229844876556696e-06, + "loss": 3.0304, + "step": 5332 + }, + { + "epoch": 1.16, + "learning_rate": 4.422875245794189e-06, + "loss": 2.2942, + "step": 5333 + }, + { + "epoch": 1.16, + "learning_rate": 4.422766003932708e-06, + "loss": 2.8046, + "step": 5334 + }, + { + "epoch": 1.16, + "learning_rate": 4.422656762071227e-06, + "loss": 3.6455, + "step": 5335 + }, + { + "epoch": 1.16, + "learning_rate": 4.422547520209745e-06, + "loss": 3.1307, + "step": 5336 + }, + { + "epoch": 1.16, + "learning_rate": 4.422438278348263e-06, + "loss": 3.4356, + "step": 5337 + }, + { + "epoch": 1.16, + "learning_rate": 4.422329036486782e-06, + "loss": 2.9718, + "step": 5338 + }, + { + "epoch": 1.17, + "learning_rate": 4.422219794625301e-06, + "loss": 3.4262, + "step": 5339 + }, + { + "epoch": 1.17, + "learning_rate": 4.42211055276382e-06, + "loss": 3.025, + "step": 5340 + }, + { + "epoch": 1.17, + "learning_rate": 4.422001310902338e-06, + "loss": 2.3051, + "step": 5341 + }, + { + "epoch": 1.17, + "learning_rate": 4.4218920690408565e-06, + "loss": 2.6436, + "step": 5342 + }, + { + "epoch": 1.17, + "learning_rate": 4.4217828271793756e-06, + "loss": 2.9738, + "step": 5343 + }, + { + "epoch": 1.17, + "learning_rate": 4.421673585317895e-06, + "loss": 2.5754, + "step": 5344 + }, + { + "epoch": 1.17, + "learning_rate": 4.421564343456413e-06, + "loss": 3.5015, + "step": 5345 + }, + { + "epoch": 1.17, + "learning_rate": 4.421455101594932e-06, + "loss": 2.7746, + "step": 5346 + }, + { + "epoch": 1.17, + "learning_rate": 4.42134585973345e-06, + "loss": 2.6394, + "step": 5347 + }, + { + "epoch": 1.17, + "learning_rate": 4.421236617871969e-06, + "loss": 3.326, + "step": 5348 + }, + { + "epoch": 1.17, + "learning_rate": 4.421127376010488e-06, + "loss": 2.9726, + "step": 5349 + }, + { + "epoch": 1.17, + "learning_rate": 4.421018134149006e-06, + "loss": 2.8705, + "step": 5350 + }, + { + "epoch": 1.17, + "learning_rate": 4.4209088922875244e-06, + "loss": 2.0317, + "step": 5351 + }, + { + "epoch": 1.17, + "learning_rate": 4.4207996504260435e-06, + "loss": 3.5872, + "step": 5352 + }, + { + "epoch": 1.17, + "learning_rate": 4.4206904085645625e-06, + "loss": 2.627, + "step": 5353 + }, + { + "epoch": 1.17, + "learning_rate": 4.4205811667030815e-06, + "loss": 3.0722, + "step": 5354 + }, + { + "epoch": 1.17, + "learning_rate": 4.4204719248416e-06, + "loss": 3.0015, + "step": 5355 + }, + { + "epoch": 1.17, + "learning_rate": 4.420362682980118e-06, + "loss": 3.6583, + "step": 5356 + }, + { + "epoch": 1.17, + "learning_rate": 4.420253441118637e-06, + "loss": 3.6864, + "step": 5357 + }, + { + "epoch": 1.17, + "learning_rate": 4.420144199257156e-06, + "loss": 3.6426, + "step": 5358 + }, + { + "epoch": 1.17, + "learning_rate": 4.420034957395674e-06, + "loss": 3.8318, + "step": 5359 + }, + { + "epoch": 1.17, + "learning_rate": 4.419925715534193e-06, + "loss": 3.299, + "step": 5360 + }, + { + "epoch": 1.17, + "learning_rate": 4.419816473672711e-06, + "loss": 3.1433, + "step": 5361 + }, + { + "epoch": 1.17, + "learning_rate": 4.4197072318112304e-06, + "loss": 3.2651, + "step": 5362 + }, + { + "epoch": 1.17, + "learning_rate": 4.4195979899497495e-06, + "loss": 3.1405, + "step": 5363 + }, + { + "epoch": 1.17, + "learning_rate": 4.419488748088268e-06, + "loss": 2.4124, + "step": 5364 + }, + { + "epoch": 1.17, + "learning_rate": 4.419379506226787e-06, + "loss": 3.1973, + "step": 5365 + }, + { + "epoch": 1.17, + "learning_rate": 4.419270264365305e-06, + "loss": 3.4074, + "step": 5366 + }, + { + "epoch": 1.17, + "learning_rate": 4.419161022503824e-06, + "loss": 3.0684, + "step": 5367 + }, + { + "epoch": 1.17, + "learning_rate": 4.419051780642343e-06, + "loss": 2.8803, + "step": 5368 + }, + { + "epoch": 1.17, + "learning_rate": 4.418942538780861e-06, + "loss": 3.5159, + "step": 5369 + }, + { + "epoch": 1.17, + "learning_rate": 4.41883329691938e-06, + "loss": 2.8449, + "step": 5370 + }, + { + "epoch": 1.17, + "learning_rate": 4.418724055057898e-06, + "loss": 2.5843, + "step": 5371 + }, + { + "epoch": 1.17, + "learning_rate": 4.418614813196417e-06, + "loss": 3.0508, + "step": 5372 + }, + { + "epoch": 1.17, + "learning_rate": 4.4185055713349364e-06, + "loss": 2.9384, + "step": 5373 + }, + { + "epoch": 1.17, + "learning_rate": 4.418396329473455e-06, + "loss": 3.1309, + "step": 5374 + }, + { + "epoch": 1.17, + "learning_rate": 4.418287087611973e-06, + "loss": 2.6568, + "step": 5375 + }, + { + "epoch": 1.17, + "learning_rate": 4.418177845750492e-06, + "loss": 3.3106, + "step": 5376 + }, + { + "epoch": 1.17, + "learning_rate": 4.418068603889011e-06, + "loss": 3.2929, + "step": 5377 + }, + { + "epoch": 1.17, + "learning_rate": 4.417959362027529e-06, + "loss": 3.2897, + "step": 5378 + }, + { + "epoch": 1.17, + "learning_rate": 4.417850120166048e-06, + "loss": 3.8896, + "step": 5379 + }, + { + "epoch": 1.17, + "learning_rate": 4.417740878304566e-06, + "loss": 2.9809, + "step": 5380 + }, + { + "epoch": 1.17, + "learning_rate": 4.417631636443085e-06, + "loss": 3.517, + "step": 5381 + }, + { + "epoch": 1.17, + "learning_rate": 4.417522394581604e-06, + "loss": 3.8415, + "step": 5382 + }, + { + "epoch": 1.17, + "learning_rate": 4.4174131527201225e-06, + "loss": 2.9573, + "step": 5383 + }, + { + "epoch": 1.18, + "learning_rate": 4.4173039108586416e-06, + "loss": 2.6418, + "step": 5384 + }, + { + "epoch": 1.18, + "learning_rate": 4.41719466899716e-06, + "loss": 3.8838, + "step": 5385 + }, + { + "epoch": 1.18, + "learning_rate": 4.417085427135679e-06, + "loss": 2.8078, + "step": 5386 + }, + { + "epoch": 1.18, + "learning_rate": 4.416976185274198e-06, + "loss": 2.7319, + "step": 5387 + }, + { + "epoch": 1.18, + "learning_rate": 4.416866943412716e-06, + "loss": 2.6815, + "step": 5388 + }, + { + "epoch": 1.18, + "learning_rate": 4.416757701551235e-06, + "loss": 3.6516, + "step": 5389 + }, + { + "epoch": 1.18, + "learning_rate": 4.416648459689753e-06, + "loss": 3.3915, + "step": 5390 + }, + { + "epoch": 1.18, + "learning_rate": 4.416539217828272e-06, + "loss": 3.4168, + "step": 5391 + }, + { + "epoch": 1.18, + "learning_rate": 4.4164299759667905e-06, + "loss": 2.9224, + "step": 5392 + }, + { + "epoch": 1.18, + "learning_rate": 4.4163207341053095e-06, + "loss": 3.449, + "step": 5393 + }, + { + "epoch": 1.18, + "learning_rate": 4.4162114922438285e-06, + "loss": 3.4166, + "step": 5394 + }, + { + "epoch": 1.18, + "learning_rate": 4.416102250382347e-06, + "loss": 3.5383, + "step": 5395 + }, + { + "epoch": 1.18, + "learning_rate": 4.415993008520866e-06, + "loss": 3.0143, + "step": 5396 + }, + { + "epoch": 1.18, + "learning_rate": 4.415883766659384e-06, + "loss": 3.585, + "step": 5397 + }, + { + "epoch": 1.18, + "learning_rate": 4.415774524797903e-06, + "loss": 3.5658, + "step": 5398 + }, + { + "epoch": 1.18, + "learning_rate": 4.415665282936422e-06, + "loss": 2.8126, + "step": 5399 + }, + { + "epoch": 1.18, + "learning_rate": 4.41555604107494e-06, + "loss": 2.9672, + "step": 5400 + }, + { + "epoch": 1.18, + "learning_rate": 4.415446799213459e-06, + "loss": 3.1714, + "step": 5401 + }, + { + "epoch": 1.18, + "learning_rate": 4.415337557351977e-06, + "loss": 3.1635, + "step": 5402 + }, + { + "epoch": 1.18, + "learning_rate": 4.4152283154904965e-06, + "loss": 3.5616, + "step": 5403 + }, + { + "epoch": 1.18, + "learning_rate": 4.415119073629015e-06, + "loss": 3.1189, + "step": 5404 + }, + { + "epoch": 1.18, + "learning_rate": 4.415009831767534e-06, + "loss": 2.7469, + "step": 5405 + }, + { + "epoch": 1.18, + "learning_rate": 4.414900589906053e-06, + "loss": 3.6825, + "step": 5406 + }, + { + "epoch": 1.18, + "learning_rate": 4.414791348044571e-06, + "loss": 3.0316, + "step": 5407 + }, + { + "epoch": 1.18, + "learning_rate": 4.41468210618309e-06, + "loss": 3.0309, + "step": 5408 + }, + { + "epoch": 1.18, + "learning_rate": 4.414572864321608e-06, + "loss": 2.4904, + "step": 5409 + }, + { + "epoch": 1.18, + "learning_rate": 4.414463622460127e-06, + "loss": 3.2791, + "step": 5410 + }, + { + "epoch": 1.18, + "learning_rate": 4.414354380598645e-06, + "loss": 3.2571, + "step": 5411 + }, + { + "epoch": 1.18, + "learning_rate": 4.414245138737164e-06, + "loss": 3.9052, + "step": 5412 + }, + { + "epoch": 1.18, + "learning_rate": 4.414135896875683e-06, + "loss": 2.5912, + "step": 5413 + }, + { + "epoch": 1.18, + "learning_rate": 4.414026655014202e-06, + "loss": 3.2413, + "step": 5414 + }, + { + "epoch": 1.18, + "learning_rate": 4.413917413152721e-06, + "loss": 2.9405, + "step": 5415 + }, + { + "epoch": 1.18, + "learning_rate": 4.413808171291239e-06, + "loss": 3.1821, + "step": 5416 + }, + { + "epoch": 1.18, + "learning_rate": 4.413698929429758e-06, + "loss": 3.6495, + "step": 5417 + }, + { + "epoch": 1.18, + "learning_rate": 4.413589687568277e-06, + "loss": 2.9378, + "step": 5418 + }, + { + "epoch": 1.18, + "learning_rate": 4.413480445706795e-06, + "loss": 3.1518, + "step": 5419 + }, + { + "epoch": 1.18, + "learning_rate": 4.413371203845314e-06, + "loss": 3.7898, + "step": 5420 + }, + { + "epoch": 1.18, + "learning_rate": 4.413261961983832e-06, + "loss": 3.0153, + "step": 5421 + }, + { + "epoch": 1.18, + "learning_rate": 4.413152720122351e-06, + "loss": 2.7863, + "step": 5422 + }, + { + "epoch": 1.18, + "learning_rate": 4.41304347826087e-06, + "loss": 2.5287, + "step": 5423 + }, + { + "epoch": 1.18, + "learning_rate": 4.4129342363993886e-06, + "loss": 3.1267, + "step": 5424 + }, + { + "epoch": 1.18, + "learning_rate": 4.412824994537907e-06, + "loss": 3.1305, + "step": 5425 + }, + { + "epoch": 1.18, + "learning_rate": 4.412715752676426e-06, + "loss": 2.883, + "step": 5426 + }, + { + "epoch": 1.18, + "learning_rate": 4.412606510814945e-06, + "loss": 3.0317, + "step": 5427 + }, + { + "epoch": 1.18, + "learning_rate": 4.412497268953464e-06, + "loss": 3.4769, + "step": 5428 + }, + { + "epoch": 1.18, + "learning_rate": 4.412388027091982e-06, + "loss": 3.1047, + "step": 5429 + }, + { + "epoch": 1.19, + "learning_rate": 4.4122787852305e-06, + "loss": 3.127, + "step": 5430 + }, + { + "epoch": 1.19, + "learning_rate": 4.412169543369019e-06, + "loss": 2.4514, + "step": 5431 + }, + { + "epoch": 1.19, + "learning_rate": 4.412060301507538e-06, + "loss": 3.173, + "step": 5432 + }, + { + "epoch": 1.19, + "learning_rate": 4.4119510596460565e-06, + "loss": 1.9696, + "step": 5433 + }, + { + "epoch": 1.19, + "learning_rate": 4.4118418177845755e-06, + "loss": 2.8929, + "step": 5434 + }, + { + "epoch": 1.19, + "learning_rate": 4.411732575923094e-06, + "loss": 2.8461, + "step": 5435 + }, + { + "epoch": 1.19, + "learning_rate": 4.411623334061613e-06, + "loss": 3.9358, + "step": 5436 + }, + { + "epoch": 1.19, + "learning_rate": 4.411514092200132e-06, + "loss": 3.8377, + "step": 5437 + }, + { + "epoch": 1.19, + "learning_rate": 4.41140485033865e-06, + "loss": 3.5041, + "step": 5438 + }, + { + "epoch": 1.19, + "learning_rate": 4.411295608477169e-06, + "loss": 3.2073, + "step": 5439 + }, + { + "epoch": 1.19, + "learning_rate": 4.411186366615687e-06, + "loss": 3.0974, + "step": 5440 + }, + { + "epoch": 1.19, + "learning_rate": 4.411077124754206e-06, + "loss": 2.5124, + "step": 5441 + }, + { + "epoch": 1.19, + "learning_rate": 4.410967882892725e-06, + "loss": 3.4602, + "step": 5442 + }, + { + "epoch": 1.19, + "learning_rate": 4.4108586410312434e-06, + "loss": 3.0005, + "step": 5443 + }, + { + "epoch": 1.19, + "learning_rate": 4.410749399169762e-06, + "loss": 3.1953, + "step": 5444 + }, + { + "epoch": 1.19, + "learning_rate": 4.410640157308281e-06, + "loss": 3.7095, + "step": 5445 + }, + { + "epoch": 1.19, + "learning_rate": 4.4105309154468e-06, + "loss": 3.3523, + "step": 5446 + }, + { + "epoch": 1.19, + "learning_rate": 4.410421673585319e-06, + "loss": 2.9225, + "step": 5447 + }, + { + "epoch": 1.19, + "learning_rate": 4.410312431723837e-06, + "loss": 2.6231, + "step": 5448 + }, + { + "epoch": 1.19, + "learning_rate": 4.410203189862355e-06, + "loss": 2.8842, + "step": 5449 + }, + { + "epoch": 1.19, + "learning_rate": 4.410093948000874e-06, + "loss": 2.4218, + "step": 5450 + }, + { + "epoch": 1.19, + "learning_rate": 4.409984706139393e-06, + "loss": 2.6971, + "step": 5451 + }, + { + "epoch": 1.19, + "learning_rate": 4.409875464277912e-06, + "loss": 2.9341, + "step": 5452 + }, + { + "epoch": 1.19, + "learning_rate": 4.40976622241643e-06, + "loss": 2.7905, + "step": 5453 + }, + { + "epoch": 1.19, + "learning_rate": 4.4096569805549486e-06, + "loss": 3.5499, + "step": 5454 + }, + { + "epoch": 1.19, + "learning_rate": 4.409547738693468e-06, + "loss": 3.9209, + "step": 5455 + }, + { + "epoch": 1.19, + "learning_rate": 4.409438496831987e-06, + "loss": 2.8863, + "step": 5456 + }, + { + "epoch": 1.19, + "learning_rate": 4.409329254970505e-06, + "loss": 3.8559, + "step": 5457 + }, + { + "epoch": 1.19, + "learning_rate": 4.409220013109024e-06, + "loss": 2.1009, + "step": 5458 + }, + { + "epoch": 1.19, + "learning_rate": 4.409110771247542e-06, + "loss": 3.3948, + "step": 5459 + }, + { + "epoch": 1.19, + "learning_rate": 4.409001529386061e-06, + "loss": 2.2897, + "step": 5460 + }, + { + "epoch": 1.19, + "learning_rate": 4.40889228752458e-06, + "loss": 3.5429, + "step": 5461 + }, + { + "epoch": 1.19, + "learning_rate": 4.408783045663098e-06, + "loss": 3.2926, + "step": 5462 + }, + { + "epoch": 1.19, + "learning_rate": 4.4086738038016165e-06, + "loss": 3.2655, + "step": 5463 + }, + { + "epoch": 1.19, + "learning_rate": 4.4085645619401355e-06, + "loss": 3.1056, + "step": 5464 + }, + { + "epoch": 1.19, + "learning_rate": 4.4084553200786546e-06, + "loss": 3.4796, + "step": 5465 + }, + { + "epoch": 1.19, + "learning_rate": 4.408346078217174e-06, + "loss": 3.1752, + "step": 5466 + }, + { + "epoch": 1.19, + "learning_rate": 4.408236836355692e-06, + "loss": 2.4495, + "step": 5467 + }, + { + "epoch": 1.19, + "learning_rate": 4.40812759449421e-06, + "loss": 2.5822, + "step": 5468 + }, + { + "epoch": 1.19, + "learning_rate": 4.408018352632729e-06, + "loss": 2.7965, + "step": 5469 + }, + { + "epoch": 1.19, + "learning_rate": 4.407909110771248e-06, + "loss": 3.6561, + "step": 5470 + }, + { + "epoch": 1.19, + "learning_rate": 4.407799868909767e-06, + "loss": 3.5009, + "step": 5471 + }, + { + "epoch": 1.19, + "learning_rate": 4.407690627048285e-06, + "loss": 3.1067, + "step": 5472 + }, + { + "epoch": 1.19, + "learning_rate": 4.4075813851868035e-06, + "loss": 4.1081, + "step": 5473 + }, + { + "epoch": 1.19, + "learning_rate": 4.4074721433253225e-06, + "loss": 2.8965, + "step": 5474 + }, + { + "epoch": 1.19, + "learning_rate": 4.4073629014638415e-06, + "loss": 2.5712, + "step": 5475 + }, + { + "epoch": 1.2, + "learning_rate": 4.4072536596023606e-06, + "loss": 3.3343, + "step": 5476 + }, + { + "epoch": 1.2, + "learning_rate": 4.407144417740879e-06, + "loss": 3.3906, + "step": 5477 + }, + { + "epoch": 1.2, + "learning_rate": 4.407035175879397e-06, + "loss": 2.5289, + "step": 5478 + }, + { + "epoch": 1.2, + "learning_rate": 4.406925934017916e-06, + "loss": 3.8027, + "step": 5479 + }, + { + "epoch": 1.2, + "learning_rate": 4.406816692156435e-06, + "loss": 2.8865, + "step": 5480 + }, + { + "epoch": 1.2, + "learning_rate": 4.406707450294954e-06, + "loss": 2.6865, + "step": 5481 + }, + { + "epoch": 1.2, + "learning_rate": 4.406598208433472e-06, + "loss": 3.2531, + "step": 5482 + }, + { + "epoch": 1.2, + "learning_rate": 4.40648896657199e-06, + "loss": 2.641, + "step": 5483 + }, + { + "epoch": 1.2, + "learning_rate": 4.4063797247105094e-06, + "loss": 3.1771, + "step": 5484 + }, + { + "epoch": 1.2, + "learning_rate": 4.4062704828490285e-06, + "loss": 2.8887, + "step": 5485 + }, + { + "epoch": 1.2, + "learning_rate": 4.406161240987547e-06, + "loss": 3.3747, + "step": 5486 + }, + { + "epoch": 1.2, + "learning_rate": 4.406051999126066e-06, + "loss": 2.3969, + "step": 5487 + }, + { + "epoch": 1.2, + "learning_rate": 4.405942757264584e-06, + "loss": 2.4296, + "step": 5488 + }, + { + "epoch": 1.2, + "learning_rate": 4.405833515403103e-06, + "loss": 3.5136, + "step": 5489 + }, + { + "epoch": 1.2, + "learning_rate": 4.405724273541622e-06, + "loss": 3.1011, + "step": 5490 + }, + { + "epoch": 1.2, + "learning_rate": 4.40561503168014e-06, + "loss": 3.4258, + "step": 5491 + }, + { + "epoch": 1.2, + "learning_rate": 4.405505789818658e-06, + "loss": 3.281, + "step": 5492 + }, + { + "epoch": 1.2, + "learning_rate": 4.405396547957177e-06, + "loss": 2.8346, + "step": 5493 + }, + { + "epoch": 1.2, + "learning_rate": 4.405287306095696e-06, + "loss": 3.6329, + "step": 5494 + }, + { + "epoch": 1.2, + "learning_rate": 4.4051780642342154e-06, + "loss": 2.4432, + "step": 5495 + }, + { + "epoch": 1.2, + "learning_rate": 4.405068822372734e-06, + "loss": 3.2944, + "step": 5496 + }, + { + "epoch": 1.2, + "learning_rate": 4.404959580511252e-06, + "loss": 2.8322, + "step": 5497 + }, + { + "epoch": 1.2, + "learning_rate": 4.404850338649771e-06, + "loss": 2.8728, + "step": 5498 + }, + { + "epoch": 1.2, + "learning_rate": 4.40474109678829e-06, + "loss": 3.285, + "step": 5499 + }, + { + "epoch": 1.2, + "learning_rate": 4.404631854926809e-06, + "loss": 2.7712, + "step": 5500 + }, + { + "epoch": 1.2, + "learning_rate": 4.404522613065327e-06, + "loss": 3.2884, + "step": 5501 + }, + { + "epoch": 1.2, + "learning_rate": 4.404413371203845e-06, + "loss": 3.2826, + "step": 5502 + }, + { + "epoch": 1.2, + "learning_rate": 4.404304129342364e-06, + "loss": 3.3238, + "step": 5503 + }, + { + "epoch": 1.2, + "learning_rate": 4.404194887480883e-06, + "loss": 3.6439, + "step": 5504 + }, + { + "epoch": 1.2, + "learning_rate": 4.4040856456194015e-06, + "loss": 2.1839, + "step": 5505 + }, + { + "epoch": 1.2, + "learning_rate": 4.403976403757921e-06, + "loss": 2.573, + "step": 5506 + }, + { + "epoch": 1.2, + "learning_rate": 4.403867161896439e-06, + "loss": 3.421, + "step": 5507 + }, + { + "epoch": 1.2, + "learning_rate": 4.403757920034958e-06, + "loss": 2.9479, + "step": 5508 + }, + { + "epoch": 1.2, + "learning_rate": 4.403648678173477e-06, + "loss": 3.4026, + "step": 5509 + }, + { + "epoch": 1.2, + "learning_rate": 4.403539436311995e-06, + "loss": 2.9459, + "step": 5510 + }, + { + "epoch": 1.2, + "learning_rate": 4.403430194450514e-06, + "loss": 2.6257, + "step": 5511 + }, + { + "epoch": 1.2, + "learning_rate": 4.403320952589032e-06, + "loss": 3.255, + "step": 5512 + }, + { + "epoch": 1.2, + "learning_rate": 4.403211710727551e-06, + "loss": 3.3972, + "step": 5513 + }, + { + "epoch": 1.2, + "learning_rate": 4.40310246886607e-06, + "loss": 3.0043, + "step": 5514 + }, + { + "epoch": 1.2, + "learning_rate": 4.4029932270045885e-06, + "loss": 2.16, + "step": 5515 + }, + { + "epoch": 1.2, + "learning_rate": 4.402883985143107e-06, + "loss": 2.9235, + "step": 5516 + }, + { + "epoch": 1.2, + "learning_rate": 4.402774743281626e-06, + "loss": 3.5486, + "step": 5517 + }, + { + "epoch": 1.2, + "learning_rate": 4.402665501420145e-06, + "loss": 3.3499, + "step": 5518 + }, + { + "epoch": 1.2, + "learning_rate": 4.402556259558664e-06, + "loss": 3.2694, + "step": 5519 + }, + { + "epoch": 1.2, + "learning_rate": 4.402447017697182e-06, + "loss": 2.3069, + "step": 5520 + }, + { + "epoch": 1.2, + "learning_rate": 4.4023377758357e-06, + "loss": 3.1881, + "step": 5521 + }, + { + "epoch": 1.21, + "learning_rate": 4.402228533974219e-06, + "loss": 2.7287, + "step": 5522 + }, + { + "epoch": 1.21, + "learning_rate": 4.402119292112738e-06, + "loss": 3.2266, + "step": 5523 + }, + { + "epoch": 1.21, + "learning_rate": 4.4020100502512564e-06, + "loss": 2.8924, + "step": 5524 + }, + { + "epoch": 1.21, + "learning_rate": 4.4019008083897755e-06, + "loss": 2.5226, + "step": 5525 + }, + { + "epoch": 1.21, + "learning_rate": 4.401791566528294e-06, + "loss": 3.8216, + "step": 5526 + }, + { + "epoch": 1.21, + "learning_rate": 4.401682324666813e-06, + "loss": 2.636, + "step": 5527 + }, + { + "epoch": 1.21, + "learning_rate": 4.401573082805332e-06, + "loss": 3.2682, + "step": 5528 + }, + { + "epoch": 1.21, + "learning_rate": 4.40146384094385e-06, + "loss": 3.2668, + "step": 5529 + }, + { + "epoch": 1.21, + "learning_rate": 4.401354599082369e-06, + "loss": 3.3707, + "step": 5530 + }, + { + "epoch": 1.21, + "learning_rate": 4.401245357220887e-06, + "loss": 3.2564, + "step": 5531 + }, + { + "epoch": 1.21, + "learning_rate": 4.401136115359406e-06, + "loss": 3.2321, + "step": 5532 + }, + { + "epoch": 1.21, + "learning_rate": 4.401026873497925e-06, + "loss": 2.5608, + "step": 5533 + }, + { + "epoch": 1.21, + "learning_rate": 4.400917631636443e-06, + "loss": 3.5334, + "step": 5534 + }, + { + "epoch": 1.21, + "learning_rate": 4.400808389774962e-06, + "loss": 2.9086, + "step": 5535 + }, + { + "epoch": 1.21, + "learning_rate": 4.400699147913481e-06, + "loss": 3.6511, + "step": 5536 + }, + { + "epoch": 1.21, + "learning_rate": 4.400589906052e-06, + "loss": 2.7208, + "step": 5537 + }, + { + "epoch": 1.21, + "learning_rate": 4.400480664190518e-06, + "loss": 2.5665, + "step": 5538 + }, + { + "epoch": 1.21, + "learning_rate": 4.400371422329037e-06, + "loss": 3.0545, + "step": 5539 + }, + { + "epoch": 1.21, + "learning_rate": 4.400262180467556e-06, + "loss": 2.4971, + "step": 5540 + }, + { + "epoch": 1.21, + "learning_rate": 4.400152938606074e-06, + "loss": 3.2276, + "step": 5541 + }, + { + "epoch": 1.21, + "learning_rate": 4.400043696744593e-06, + "loss": 3.6722, + "step": 5542 + }, + { + "epoch": 1.21, + "learning_rate": 4.399934454883111e-06, + "loss": 3.868, + "step": 5543 + }, + { + "epoch": 1.21, + "learning_rate": 4.39982521302163e-06, + "loss": 2.8766, + "step": 5544 + }, + { + "epoch": 1.21, + "learning_rate": 4.3997159711601485e-06, + "loss": 3.1958, + "step": 5545 + }, + { + "epoch": 1.21, + "learning_rate": 4.3996067292986676e-06, + "loss": 2.9432, + "step": 5546 + }, + { + "epoch": 1.21, + "learning_rate": 4.399497487437187e-06, + "loss": 1.97, + "step": 5547 + }, + { + "epoch": 1.21, + "learning_rate": 4.399388245575705e-06, + "loss": 3.0383, + "step": 5548 + }, + { + "epoch": 1.21, + "learning_rate": 4.399279003714224e-06, + "loss": 3.167, + "step": 5549 + }, + { + "epoch": 1.21, + "learning_rate": 4.399169761852742e-06, + "loss": 2.8748, + "step": 5550 + }, + { + "epoch": 1.21, + "learning_rate": 4.399060519991261e-06, + "loss": 3.0364, + "step": 5551 + }, + { + "epoch": 1.21, + "learning_rate": 4.39895127812978e-06, + "loss": 2.763, + "step": 5552 + }, + { + "epoch": 1.21, + "learning_rate": 4.398842036268298e-06, + "loss": 3.0692, + "step": 5553 + }, + { + "epoch": 1.21, + "learning_rate": 4.398732794406817e-06, + "loss": 2.5213, + "step": 5554 + }, + { + "epoch": 1.21, + "learning_rate": 4.3986235525453355e-06, + "loss": 3.4672, + "step": 5555 + }, + { + "epoch": 1.21, + "learning_rate": 4.3985143106838545e-06, + "loss": 3.3339, + "step": 5556 + }, + { + "epoch": 1.21, + "learning_rate": 4.398405068822373e-06, + "loss": 3.7948, + "step": 5557 + }, + { + "epoch": 1.21, + "learning_rate": 4.398295826960892e-06, + "loss": 2.4356, + "step": 5558 + }, + { + "epoch": 1.21, + "learning_rate": 4.398186585099411e-06, + "loss": 3.2364, + "step": 5559 + }, + { + "epoch": 1.21, + "learning_rate": 4.398077343237929e-06, + "loss": 3.1777, + "step": 5560 + }, + { + "epoch": 1.21, + "learning_rate": 4.397968101376448e-06, + "loss": 2.9856, + "step": 5561 + }, + { + "epoch": 1.21, + "learning_rate": 4.397858859514966e-06, + "loss": 3.6829, + "step": 5562 + }, + { + "epoch": 1.21, + "learning_rate": 4.397749617653485e-06, + "loss": 3.1916, + "step": 5563 + }, + { + "epoch": 1.21, + "learning_rate": 4.397640375792004e-06, + "loss": 2.9819, + "step": 5564 + }, + { + "epoch": 1.21, + "learning_rate": 4.3975311339305224e-06, + "loss": 3.9437, + "step": 5565 + }, + { + "epoch": 1.21, + "learning_rate": 4.3974218920690415e-06, + "loss": 2.829, + "step": 5566 + }, + { + "epoch": 1.21, + "learning_rate": 4.39731265020756e-06, + "loss": 3.675, + "step": 5567 + }, + { + "epoch": 1.22, + "learning_rate": 4.397203408346079e-06, + "loss": 2.7712, + "step": 5568 + }, + { + "epoch": 1.22, + "learning_rate": 4.397094166484597e-06, + "loss": 2.7223, + "step": 5569 + }, + { + "epoch": 1.22, + "learning_rate": 4.396984924623116e-06, + "loss": 3.2019, + "step": 5570 + }, + { + "epoch": 1.22, + "learning_rate": 4.396875682761634e-06, + "loss": 2.9857, + "step": 5571 + }, + { + "epoch": 1.22, + "learning_rate": 4.396766440900153e-06, + "loss": 3.2871, + "step": 5572 + }, + { + "epoch": 1.22, + "learning_rate": 4.396657199038672e-06, + "loss": 2.6118, + "step": 5573 + }, + { + "epoch": 1.22, + "learning_rate": 4.39654795717719e-06, + "loss": 2.4361, + "step": 5574 + }, + { + "epoch": 1.22, + "learning_rate": 4.396438715315709e-06, + "loss": 3.2956, + "step": 5575 + }, + { + "epoch": 1.22, + "learning_rate": 4.396329473454228e-06, + "loss": 3.2273, + "step": 5576 + }, + { + "epoch": 1.22, + "learning_rate": 4.396220231592747e-06, + "loss": 2.5176, + "step": 5577 + }, + { + "epoch": 1.22, + "learning_rate": 4.396110989731266e-06, + "loss": 2.7651, + "step": 5578 + }, + { + "epoch": 1.22, + "learning_rate": 4.396001747869784e-06, + "loss": 3.4591, + "step": 5579 + }, + { + "epoch": 1.22, + "learning_rate": 4.395892506008303e-06, + "loss": 3.6817, + "step": 5580 + }, + { + "epoch": 1.22, + "learning_rate": 4.395783264146821e-06, + "loss": 2.9703, + "step": 5581 + }, + { + "epoch": 1.22, + "learning_rate": 4.39567402228534e-06, + "loss": 2.9837, + "step": 5582 + }, + { + "epoch": 1.22, + "learning_rate": 4.395564780423859e-06, + "loss": 3.5048, + "step": 5583 + }, + { + "epoch": 1.22, + "learning_rate": 4.395455538562377e-06, + "loss": 2.7506, + "step": 5584 + }, + { + "epoch": 1.22, + "learning_rate": 4.395346296700896e-06, + "loss": 3.494, + "step": 5585 + }, + { + "epoch": 1.22, + "learning_rate": 4.3952370548394145e-06, + "loss": 3.037, + "step": 5586 + }, + { + "epoch": 1.22, + "learning_rate": 4.395127812977934e-06, + "loss": 3.3253, + "step": 5587 + }, + { + "epoch": 1.22, + "learning_rate": 4.395018571116453e-06, + "loss": 2.6295, + "step": 5588 + }, + { + "epoch": 1.22, + "learning_rate": 4.394909329254971e-06, + "loss": 2.4337, + "step": 5589 + }, + { + "epoch": 1.22, + "learning_rate": 4.394800087393489e-06, + "loss": 3.5674, + "step": 5590 + }, + { + "epoch": 1.22, + "learning_rate": 4.394690845532008e-06, + "loss": 3.0724, + "step": 5591 + }, + { + "epoch": 1.22, + "learning_rate": 4.394581603670527e-06, + "loss": 3.0265, + "step": 5592 + }, + { + "epoch": 1.22, + "learning_rate": 4.394472361809046e-06, + "loss": 3.1703, + "step": 5593 + }, + { + "epoch": 1.22, + "learning_rate": 4.394363119947564e-06, + "loss": 3.6967, + "step": 5594 + }, + { + "epoch": 1.22, + "learning_rate": 4.3942538780860825e-06, + "loss": 3.0641, + "step": 5595 + }, + { + "epoch": 1.22, + "learning_rate": 4.3941446362246015e-06, + "loss": 2.9168, + "step": 5596 + }, + { + "epoch": 1.22, + "learning_rate": 4.3940353943631205e-06, + "loss": 2.8793, + "step": 5597 + }, + { + "epoch": 1.22, + "learning_rate": 4.393926152501639e-06, + "loss": 2.6308, + "step": 5598 + }, + { + "epoch": 1.22, + "learning_rate": 4.393816910640158e-06, + "loss": 3.6843, + "step": 5599 + }, + { + "epoch": 1.22, + "learning_rate": 4.393707668778676e-06, + "loss": 3.4182, + "step": 5600 + }, + { + "epoch": 1.22, + "learning_rate": 4.393598426917195e-06, + "loss": 2.7533, + "step": 5601 + }, + { + "epoch": 1.22, + "learning_rate": 4.393489185055714e-06, + "loss": 3.198, + "step": 5602 + }, + { + "epoch": 1.22, + "learning_rate": 4.393379943194232e-06, + "loss": 2.964, + "step": 5603 + }, + { + "epoch": 1.22, + "learning_rate": 4.39327070133275e-06, + "loss": 2.8594, + "step": 5604 + }, + { + "epoch": 1.22, + "learning_rate": 4.393161459471269e-06, + "loss": 3.0422, + "step": 5605 + }, + { + "epoch": 1.22, + "learning_rate": 4.3930522176097885e-06, + "loss": 2.2662, + "step": 5606 + }, + { + "epoch": 1.22, + "learning_rate": 4.3929429757483075e-06, + "loss": 3.8164, + "step": 5607 + }, + { + "epoch": 1.22, + "learning_rate": 4.392833733886826e-06, + "loss": 2.3134, + "step": 5608 + }, + { + "epoch": 1.22, + "learning_rate": 4.392724492025344e-06, + "loss": 3.2741, + "step": 5609 + }, + { + "epoch": 1.22, + "learning_rate": 4.392615250163863e-06, + "loss": 2.9412, + "step": 5610 + }, + { + "epoch": 1.22, + "learning_rate": 4.392506008302382e-06, + "loss": 3.1472, + "step": 5611 + }, + { + "epoch": 1.22, + "learning_rate": 4.392396766440901e-06, + "loss": 3.356, + "step": 5612 + }, + { + "epoch": 1.23, + "learning_rate": 4.392287524579419e-06, + "loss": 3.3035, + "step": 5613 + }, + { + "epoch": 1.23, + "learning_rate": 4.392178282717937e-06, + "loss": 3.015, + "step": 5614 + }, + { + "epoch": 1.23, + "learning_rate": 4.392069040856456e-06, + "loss": 3.7161, + "step": 5615 + }, + { + "epoch": 1.23, + "learning_rate": 4.391959798994975e-06, + "loss": 2.8599, + "step": 5616 + }, + { + "epoch": 1.23, + "learning_rate": 4.3918505571334944e-06, + "loss": 3.4071, + "step": 5617 + }, + { + "epoch": 1.23, + "learning_rate": 4.391741315272013e-06, + "loss": 3.5868, + "step": 5618 + }, + { + "epoch": 1.23, + "learning_rate": 4.391632073410531e-06, + "loss": 3.2735, + "step": 5619 + }, + { + "epoch": 1.23, + "learning_rate": 4.39152283154905e-06, + "loss": 3.7356, + "step": 5620 + }, + { + "epoch": 1.23, + "learning_rate": 4.391413589687569e-06, + "loss": 2.6017, + "step": 5621 + }, + { + "epoch": 1.23, + "learning_rate": 4.391304347826087e-06, + "loss": 3.3444, + "step": 5622 + }, + { + "epoch": 1.23, + "learning_rate": 4.391195105964606e-06, + "loss": 2.7278, + "step": 5623 + }, + { + "epoch": 1.23, + "learning_rate": 4.391085864103124e-06, + "loss": 3.692, + "step": 5624 + }, + { + "epoch": 1.23, + "learning_rate": 4.390976622241643e-06, + "loss": 3.2081, + "step": 5625 + }, + { + "epoch": 1.23, + "learning_rate": 4.390867380380162e-06, + "loss": 2.4956, + "step": 5626 + }, + { + "epoch": 1.23, + "learning_rate": 4.3907581385186806e-06, + "loss": 3.4832, + "step": 5627 + }, + { + "epoch": 1.23, + "learning_rate": 4.3906488966572e-06, + "loss": 2.4114, + "step": 5628 + }, + { + "epoch": 1.23, + "learning_rate": 4.390539654795718e-06, + "loss": 3.53, + "step": 5629 + }, + { + "epoch": 1.23, + "learning_rate": 4.390430412934237e-06, + "loss": 3.4158, + "step": 5630 + }, + { + "epoch": 1.23, + "learning_rate": 4.390321171072756e-06, + "loss": 2.7157, + "step": 5631 + }, + { + "epoch": 1.23, + "learning_rate": 4.390211929211274e-06, + "loss": 2.9763, + "step": 5632 + }, + { + "epoch": 1.23, + "learning_rate": 4.390102687349792e-06, + "loss": 3.0831, + "step": 5633 + }, + { + "epoch": 1.23, + "learning_rate": 4.389993445488311e-06, + "loss": 2.9428, + "step": 5634 + }, + { + "epoch": 1.23, + "learning_rate": 4.38988420362683e-06, + "loss": 3.154, + "step": 5635 + }, + { + "epoch": 1.23, + "learning_rate": 4.389774961765349e-06, + "loss": 3.3057, + "step": 5636 + }, + { + "epoch": 1.23, + "learning_rate": 4.3896657199038675e-06, + "loss": 2.9576, + "step": 5637 + }, + { + "epoch": 1.23, + "learning_rate": 4.389556478042386e-06, + "loss": 2.8098, + "step": 5638 + }, + { + "epoch": 1.23, + "learning_rate": 4.389447236180905e-06, + "loss": 3.3439, + "step": 5639 + }, + { + "epoch": 1.23, + "learning_rate": 4.389337994319424e-06, + "loss": 3.1326, + "step": 5640 + }, + { + "epoch": 1.23, + "learning_rate": 4.389228752457943e-06, + "loss": 2.4243, + "step": 5641 + }, + { + "epoch": 1.23, + "learning_rate": 4.389119510596461e-06, + "loss": 3.1517, + "step": 5642 + }, + { + "epoch": 1.23, + "learning_rate": 4.389010268734979e-06, + "loss": 3.0433, + "step": 5643 + }, + { + "epoch": 1.23, + "learning_rate": 4.388901026873498e-06, + "loss": 2.9865, + "step": 5644 + }, + { + "epoch": 1.23, + "learning_rate": 4.388791785012017e-06, + "loss": 2.9028, + "step": 5645 + }, + { + "epoch": 1.23, + "learning_rate": 4.388682543150536e-06, + "loss": 3.1505, + "step": 5646 + }, + { + "epoch": 1.23, + "learning_rate": 4.3885733012890545e-06, + "loss": 2.7006, + "step": 5647 + }, + { + "epoch": 1.23, + "learning_rate": 4.388464059427573e-06, + "loss": 2.1416, + "step": 5648 + }, + { + "epoch": 1.23, + "learning_rate": 4.388354817566092e-06, + "loss": 2.9437, + "step": 5649 + }, + { + "epoch": 1.23, + "learning_rate": 4.388245575704611e-06, + "loss": 2.8925, + "step": 5650 + }, + { + "epoch": 1.23, + "learning_rate": 4.388136333843129e-06, + "loss": 2.312, + "step": 5651 + }, + { + "epoch": 1.23, + "learning_rate": 4.388027091981648e-06, + "loss": 2.6147, + "step": 5652 + }, + { + "epoch": 1.23, + "learning_rate": 4.387917850120166e-06, + "loss": 2.7814, + "step": 5653 + }, + { + "epoch": 1.23, + "learning_rate": 4.387808608258685e-06, + "loss": 3.3475, + "step": 5654 + }, + { + "epoch": 1.23, + "learning_rate": 4.387699366397204e-06, + "loss": 3.678, + "step": 5655 + }, + { + "epoch": 1.23, + "learning_rate": 4.387590124535722e-06, + "loss": 3.8092, + "step": 5656 + }, + { + "epoch": 1.23, + "learning_rate": 4.387480882674241e-06, + "loss": 2.9078, + "step": 5657 + }, + { + "epoch": 1.23, + "learning_rate": 4.38737164081276e-06, + "loss": 2.9421, + "step": 5658 + }, + { + "epoch": 1.24, + "learning_rate": 4.387262398951279e-06, + "loss": 3.5958, + "step": 5659 + }, + { + "epoch": 1.24, + "learning_rate": 4.387153157089798e-06, + "loss": 2.68, + "step": 5660 + }, + { + "epoch": 1.24, + "learning_rate": 4.387043915228316e-06, + "loss": 3.2721, + "step": 5661 + }, + { + "epoch": 1.24, + "learning_rate": 4.386934673366834e-06, + "loss": 2.1938, + "step": 5662 + }, + { + "epoch": 1.24, + "learning_rate": 4.386825431505353e-06, + "loss": 3.1816, + "step": 5663 + }, + { + "epoch": 1.24, + "learning_rate": 4.386716189643872e-06, + "loss": 3.0029, + "step": 5664 + }, + { + "epoch": 1.24, + "learning_rate": 4.386606947782391e-06, + "loss": 2.6972, + "step": 5665 + }, + { + "epoch": 1.24, + "learning_rate": 4.386497705920909e-06, + "loss": 2.5204, + "step": 5666 + }, + { + "epoch": 1.24, + "learning_rate": 4.3863884640594275e-06, + "loss": 3.3608, + "step": 5667 + }, + { + "epoch": 1.24, + "learning_rate": 4.3862792221979466e-06, + "loss": 2.3214, + "step": 5668 + }, + { + "epoch": 1.24, + "learning_rate": 4.386169980336466e-06, + "loss": 3.073, + "step": 5669 + }, + { + "epoch": 1.24, + "learning_rate": 4.386060738474984e-06, + "loss": 3.7742, + "step": 5670 + }, + { + "epoch": 1.24, + "learning_rate": 4.385951496613503e-06, + "loss": 3.2046, + "step": 5671 + }, + { + "epoch": 1.24, + "learning_rate": 4.385842254752021e-06, + "loss": 3.3197, + "step": 5672 + }, + { + "epoch": 1.24, + "learning_rate": 4.38573301289054e-06, + "loss": 2.0702, + "step": 5673 + }, + { + "epoch": 1.24, + "learning_rate": 4.385623771029059e-06, + "loss": 2.6655, + "step": 5674 + }, + { + "epoch": 1.24, + "learning_rate": 4.385514529167577e-06, + "loss": 3.2387, + "step": 5675 + }, + { + "epoch": 1.24, + "learning_rate": 4.385405287306096e-06, + "loss": 3.9079, + "step": 5676 + }, + { + "epoch": 1.24, + "learning_rate": 4.3852960454446145e-06, + "loss": 2.5924, + "step": 5677 + }, + { + "epoch": 1.24, + "learning_rate": 4.3851868035831335e-06, + "loss": 2.7401, + "step": 5678 + }, + { + "epoch": 1.24, + "learning_rate": 4.3850775617216526e-06, + "loss": 2.382, + "step": 5679 + }, + { + "epoch": 1.24, + "learning_rate": 4.384968319860171e-06, + "loss": 3.3724, + "step": 5680 + }, + { + "epoch": 1.24, + "learning_rate": 4.38485907799869e-06, + "loss": 3.7701, + "step": 5681 + }, + { + "epoch": 1.24, + "learning_rate": 4.384749836137208e-06, + "loss": 2.5772, + "step": 5682 + }, + { + "epoch": 1.24, + "learning_rate": 4.384640594275727e-06, + "loss": 3.3342, + "step": 5683 + }, + { + "epoch": 1.24, + "learning_rate": 4.384531352414246e-06, + "loss": 3.0171, + "step": 5684 + }, + { + "epoch": 1.24, + "learning_rate": 4.384422110552764e-06, + "loss": 2.2205, + "step": 5685 + }, + { + "epoch": 1.24, + "learning_rate": 4.384312868691282e-06, + "loss": 3.0102, + "step": 5686 + }, + { + "epoch": 1.24, + "learning_rate": 4.3842036268298015e-06, + "loss": 3.2265, + "step": 5687 + }, + { + "epoch": 1.24, + "learning_rate": 4.3840943849683205e-06, + "loss": 2.7952, + "step": 5688 + }, + { + "epoch": 1.24, + "learning_rate": 4.383985143106839e-06, + "loss": 3.7778, + "step": 5689 + }, + { + "epoch": 1.24, + "learning_rate": 4.383875901245358e-06, + "loss": 3.7569, + "step": 5690 + }, + { + "epoch": 1.24, + "learning_rate": 4.383766659383876e-06, + "loss": 3.2044, + "step": 5691 + }, + { + "epoch": 1.24, + "learning_rate": 4.383657417522395e-06, + "loss": 3.6358, + "step": 5692 + }, + { + "epoch": 1.24, + "learning_rate": 4.383548175660914e-06, + "loss": 2.8543, + "step": 5693 + }, + { + "epoch": 1.24, + "learning_rate": 4.383438933799432e-06, + "loss": 3.3891, + "step": 5694 + }, + { + "epoch": 1.24, + "learning_rate": 4.383329691937951e-06, + "loss": 3.5202, + "step": 5695 + }, + { + "epoch": 1.24, + "learning_rate": 4.383220450076469e-06, + "loss": 2.2281, + "step": 5696 + }, + { + "epoch": 1.24, + "learning_rate": 4.383111208214988e-06, + "loss": 2.9217, + "step": 5697 + }, + { + "epoch": 1.24, + "learning_rate": 4.3830019663535074e-06, + "loss": 2.7369, + "step": 5698 + }, + { + "epoch": 1.24, + "learning_rate": 4.382892724492026e-06, + "loss": 3.2629, + "step": 5699 + }, + { + "epoch": 1.24, + "learning_rate": 4.382783482630545e-06, + "loss": 2.7572, + "step": 5700 + }, + { + "epoch": 1.24, + "learning_rate": 4.382674240769063e-06, + "loss": 3.3544, + "step": 5701 + }, + { + "epoch": 1.24, + "learning_rate": 4.382564998907582e-06, + "loss": 3.4754, + "step": 5702 + }, + { + "epoch": 1.24, + "learning_rate": 4.3824557570461e-06, + "loss": 3.2299, + "step": 5703 + }, + { + "epoch": 1.24, + "learning_rate": 4.382346515184619e-06, + "loss": 2.195, + "step": 5704 + }, + { + "epoch": 1.25, + "learning_rate": 4.382237273323138e-06, + "loss": 3.4257, + "step": 5705 + }, + { + "epoch": 1.25, + "learning_rate": 4.382128031461656e-06, + "loss": 2.9788, + "step": 5706 + }, + { + "epoch": 1.25, + "learning_rate": 4.382018789600175e-06, + "loss": 2.5652, + "step": 5707 + }, + { + "epoch": 1.25, + "learning_rate": 4.3819095477386936e-06, + "loss": 2.9262, + "step": 5708 + }, + { + "epoch": 1.25, + "learning_rate": 4.381800305877213e-06, + "loss": 2.7105, + "step": 5709 + }, + { + "epoch": 1.25, + "learning_rate": 4.381691064015731e-06, + "loss": 2.5034, + "step": 5710 + }, + { + "epoch": 1.25, + "learning_rate": 4.38158182215425e-06, + "loss": 2.6041, + "step": 5711 + }, + { + "epoch": 1.25, + "learning_rate": 4.381472580292769e-06, + "loss": 2.5625, + "step": 5712 + }, + { + "epoch": 1.25, + "learning_rate": 4.381363338431287e-06, + "loss": 3.414, + "step": 5713 + }, + { + "epoch": 1.25, + "learning_rate": 4.381254096569806e-06, + "loss": 3.4359, + "step": 5714 + }, + { + "epoch": 1.25, + "learning_rate": 4.381144854708324e-06, + "loss": 2.3606, + "step": 5715 + }, + { + "epoch": 1.25, + "learning_rate": 4.381035612846843e-06, + "loss": 3.4568, + "step": 5716 + }, + { + "epoch": 1.25, + "learning_rate": 4.380926370985362e-06, + "loss": 3.0801, + "step": 5717 + }, + { + "epoch": 1.25, + "learning_rate": 4.3808171291238805e-06, + "loss": 3.118, + "step": 5718 + }, + { + "epoch": 1.25, + "learning_rate": 4.3807078872623995e-06, + "loss": 3.1332, + "step": 5719 + }, + { + "epoch": 1.25, + "learning_rate": 4.380598645400918e-06, + "loss": 2.842, + "step": 5720 + }, + { + "epoch": 1.25, + "learning_rate": 4.380489403539437e-06, + "loss": 3.3586, + "step": 5721 + }, + { + "epoch": 1.25, + "learning_rate": 4.380380161677955e-06, + "loss": 3.2274, + "step": 5722 + }, + { + "epoch": 1.25, + "learning_rate": 4.380270919816474e-06, + "loss": 2.7995, + "step": 5723 + }, + { + "epoch": 1.25, + "learning_rate": 4.380161677954993e-06, + "loss": 2.7084, + "step": 5724 + }, + { + "epoch": 1.25, + "learning_rate": 4.380052436093511e-06, + "loss": 3.7212, + "step": 5725 + }, + { + "epoch": 1.25, + "learning_rate": 4.37994319423203e-06, + "loss": 2.6508, + "step": 5726 + }, + { + "epoch": 1.25, + "learning_rate": 4.3798339523705484e-06, + "loss": 3.233, + "step": 5727 + }, + { + "epoch": 1.25, + "learning_rate": 4.3797247105090675e-06, + "loss": 3.6019, + "step": 5728 + }, + { + "epoch": 1.25, + "learning_rate": 4.3796154686475865e-06, + "loss": 3.4999, + "step": 5729 + }, + { + "epoch": 1.25, + "learning_rate": 4.379506226786105e-06, + "loss": 3.5859, + "step": 5730 + }, + { + "epoch": 1.25, + "learning_rate": 4.379396984924624e-06, + "loss": 3.0032, + "step": 5731 + }, + { + "epoch": 1.25, + "learning_rate": 4.379287743063142e-06, + "loss": 2.8049, + "step": 5732 + }, + { + "epoch": 1.25, + "learning_rate": 4.379178501201661e-06, + "loss": 3.751, + "step": 5733 + }, + { + "epoch": 1.25, + "learning_rate": 4.37906925934018e-06, + "loss": 3.2546, + "step": 5734 + }, + { + "epoch": 1.25, + "learning_rate": 4.378960017478698e-06, + "loss": 3.4995, + "step": 5735 + }, + { + "epoch": 1.25, + "learning_rate": 4.378850775617216e-06, + "loss": 1.8997, + "step": 5736 + }, + { + "epoch": 1.25, + "learning_rate": 4.378741533755735e-06, + "loss": 3.2024, + "step": 5737 + }, + { + "epoch": 1.25, + "learning_rate": 4.3786322918942544e-06, + "loss": 2.7754, + "step": 5738 + }, + { + "epoch": 1.25, + "learning_rate": 4.378523050032773e-06, + "loss": 2.7264, + "step": 5739 + }, + { + "epoch": 1.25, + "learning_rate": 4.378413808171292e-06, + "loss": 2.5399, + "step": 5740 + }, + { + "epoch": 1.25, + "learning_rate": 4.37830456630981e-06, + "loss": 2.8003, + "step": 5741 + }, + { + "epoch": 1.25, + "learning_rate": 4.378195324448329e-06, + "loss": 3.4703, + "step": 5742 + }, + { + "epoch": 1.25, + "learning_rate": 4.378086082586848e-06, + "loss": 4.1598, + "step": 5743 + }, + { + "epoch": 1.25, + "learning_rate": 4.377976840725366e-06, + "loss": 3.1384, + "step": 5744 + }, + { + "epoch": 1.25, + "learning_rate": 4.377867598863885e-06, + "loss": 2.748, + "step": 5745 + }, + { + "epoch": 1.25, + "learning_rate": 4.377758357002403e-06, + "loss": 3.2478, + "step": 5746 + }, + { + "epoch": 1.25, + "learning_rate": 4.377649115140922e-06, + "loss": 3.0401, + "step": 5747 + }, + { + "epoch": 1.25, + "learning_rate": 4.377539873279441e-06, + "loss": 3.428, + "step": 5748 + }, + { + "epoch": 1.25, + "learning_rate": 4.3774306314179596e-06, + "loss": 2.2563, + "step": 5749 + }, + { + "epoch": 1.25, + "learning_rate": 4.377321389556479e-06, + "loss": 3.078, + "step": 5750 + }, + { + "epoch": 1.26, + "learning_rate": 4.377212147694997e-06, + "loss": 2.2666, + "step": 5751 + }, + { + "epoch": 1.26, + "learning_rate": 4.377102905833516e-06, + "loss": 3.7048, + "step": 5752 + }, + { + "epoch": 1.26, + "learning_rate": 4.376993663972035e-06, + "loss": 3.0015, + "step": 5753 + }, + { + "epoch": 1.26, + "learning_rate": 4.376884422110553e-06, + "loss": 2.6723, + "step": 5754 + }, + { + "epoch": 1.26, + "learning_rate": 4.376775180249071e-06, + "loss": 2.9417, + "step": 5755 + }, + { + "epoch": 1.26, + "learning_rate": 4.37666593838759e-06, + "loss": 3.2253, + "step": 5756 + }, + { + "epoch": 1.26, + "learning_rate": 4.376556696526109e-06, + "loss": 3.0147, + "step": 5757 + }, + { + "epoch": 1.26, + "learning_rate": 4.376447454664628e-06, + "loss": 2.9832, + "step": 5758 + }, + { + "epoch": 1.26, + "learning_rate": 4.3763382128031465e-06, + "loss": 3.555, + "step": 5759 + }, + { + "epoch": 1.26, + "learning_rate": 4.376228970941665e-06, + "loss": 3.0831, + "step": 5760 + }, + { + "epoch": 1.26, + "learning_rate": 4.376119729080184e-06, + "loss": 2.6966, + "step": 5761 + }, + { + "epoch": 1.26, + "learning_rate": 4.376010487218703e-06, + "loss": 3.9255, + "step": 5762 + }, + { + "epoch": 1.26, + "learning_rate": 4.375901245357221e-06, + "loss": 2.799, + "step": 5763 + }, + { + "epoch": 1.26, + "learning_rate": 4.37579200349574e-06, + "loss": 2.3712, + "step": 5764 + }, + { + "epoch": 1.26, + "learning_rate": 4.375682761634258e-06, + "loss": 3.1021, + "step": 5765 + }, + { + "epoch": 1.26, + "learning_rate": 4.375573519772777e-06, + "loss": 3.1533, + "step": 5766 + }, + { + "epoch": 1.26, + "learning_rate": 4.375464277911296e-06, + "loss": 3.4653, + "step": 5767 + }, + { + "epoch": 1.26, + "learning_rate": 4.3753550360498144e-06, + "loss": 3.3307, + "step": 5768 + }, + { + "epoch": 1.26, + "learning_rate": 4.3752457941883335e-06, + "loss": 2.395, + "step": 5769 + }, + { + "epoch": 1.26, + "learning_rate": 4.375136552326852e-06, + "loss": 3.5447, + "step": 5770 + }, + { + "epoch": 1.26, + "learning_rate": 4.375027310465371e-06, + "loss": 3.5832, + "step": 5771 + }, + { + "epoch": 1.26, + "learning_rate": 4.37491806860389e-06, + "loss": 3.252, + "step": 5772 + }, + { + "epoch": 1.26, + "learning_rate": 4.374808826742408e-06, + "loss": 2.6629, + "step": 5773 + }, + { + "epoch": 1.26, + "learning_rate": 4.374699584880926e-06, + "loss": 3.0321, + "step": 5774 + }, + { + "epoch": 1.26, + "learning_rate": 4.374590343019445e-06, + "loss": 3.7811, + "step": 5775 + }, + { + "epoch": 1.26, + "learning_rate": 4.374481101157964e-06, + "loss": 2.7181, + "step": 5776 + }, + { + "epoch": 1.26, + "learning_rate": 4.374371859296483e-06, + "loss": 3.0537, + "step": 5777 + }, + { + "epoch": 1.26, + "learning_rate": 4.374262617435001e-06, + "loss": 2.568, + "step": 5778 + }, + { + "epoch": 1.26, + "learning_rate": 4.37415337557352e-06, + "loss": 3.3443, + "step": 5779 + }, + { + "epoch": 1.26, + "learning_rate": 4.374044133712039e-06, + "loss": 2.6652, + "step": 5780 + }, + { + "epoch": 1.26, + "learning_rate": 4.373934891850558e-06, + "loss": 2.6642, + "step": 5781 + }, + { + "epoch": 1.26, + "learning_rate": 4.373825649989077e-06, + "loss": 3.0393, + "step": 5782 + }, + { + "epoch": 1.26, + "learning_rate": 4.373716408127595e-06, + "loss": 4.0401, + "step": 5783 + }, + { + "epoch": 1.26, + "learning_rate": 4.373607166266113e-06, + "loss": 3.4641, + "step": 5784 + }, + { + "epoch": 1.26, + "learning_rate": 4.373497924404632e-06, + "loss": 2.6471, + "step": 5785 + }, + { + "epoch": 1.26, + "learning_rate": 4.373388682543151e-06, + "loss": 3.1833, + "step": 5786 + }, + { + "epoch": 1.26, + "learning_rate": 4.37327944068167e-06, + "loss": 3.1484, + "step": 5787 + }, + { + "epoch": 1.26, + "learning_rate": 4.373170198820188e-06, + "loss": 3.9994, + "step": 5788 + }, + { + "epoch": 1.26, + "learning_rate": 4.3730609569587065e-06, + "loss": 2.8527, + "step": 5789 + }, + { + "epoch": 1.26, + "learning_rate": 4.372951715097226e-06, + "loss": 3.0653, + "step": 5790 + }, + { + "epoch": 1.26, + "learning_rate": 4.372842473235745e-06, + "loss": 3.3281, + "step": 5791 + }, + { + "epoch": 1.26, + "learning_rate": 4.372733231374263e-06, + "loss": 3.5897, + "step": 5792 + }, + { + "epoch": 1.26, + "learning_rate": 4.372623989512782e-06, + "loss": 3.1462, + "step": 5793 + }, + { + "epoch": 1.26, + "learning_rate": 4.3725147476513e-06, + "loss": 3.3347, + "step": 5794 + }, + { + "epoch": 1.26, + "learning_rate": 4.372405505789819e-06, + "loss": 3.2658, + "step": 5795 + }, + { + "epoch": 1.26, + "learning_rate": 4.372296263928338e-06, + "loss": 3.4834, + "step": 5796 + }, + { + "epoch": 1.27, + "learning_rate": 4.372187022066856e-06, + "loss": 3.0609, + "step": 5797 + }, + { + "epoch": 1.27, + "learning_rate": 4.3720777802053745e-06, + "loss": 3.0737, + "step": 5798 + }, + { + "epoch": 1.27, + "learning_rate": 4.3719685383438935e-06, + "loss": 3.3903, + "step": 5799 + }, + { + "epoch": 1.27, + "learning_rate": 4.3718592964824125e-06, + "loss": 2.5923, + "step": 5800 + }, + { + "epoch": 1.27, + "learning_rate": 4.3717500546209316e-06, + "loss": 2.8501, + "step": 5801 + }, + { + "epoch": 1.27, + "learning_rate": 4.37164081275945e-06, + "loss": 2.5442, + "step": 5802 + }, + { + "epoch": 1.27, + "learning_rate": 4.371531570897968e-06, + "loss": 3.8519, + "step": 5803 + }, + { + "epoch": 1.27, + "learning_rate": 4.371422329036487e-06, + "loss": 2.8875, + "step": 5804 + }, + { + "epoch": 1.27, + "learning_rate": 4.371313087175006e-06, + "loss": 2.3935, + "step": 5805 + }, + { + "epoch": 1.27, + "learning_rate": 4.371203845313525e-06, + "loss": 3.2749, + "step": 5806 + }, + { + "epoch": 1.27, + "learning_rate": 4.371094603452043e-06, + "loss": 3.2886, + "step": 5807 + }, + { + "epoch": 1.27, + "learning_rate": 4.3709853615905614e-06, + "loss": 2.6982, + "step": 5808 + }, + { + "epoch": 1.27, + "learning_rate": 4.3708761197290805e-06, + "loss": 3.0056, + "step": 5809 + }, + { + "epoch": 1.27, + "learning_rate": 4.3707668778675995e-06, + "loss": 2.9296, + "step": 5810 + }, + { + "epoch": 1.27, + "learning_rate": 4.3706576360061185e-06, + "loss": 3.0346, + "step": 5811 + }, + { + "epoch": 1.27, + "learning_rate": 4.370548394144637e-06, + "loss": 2.6932, + "step": 5812 + }, + { + "epoch": 1.27, + "learning_rate": 4.370439152283155e-06, + "loss": 3.9025, + "step": 5813 + }, + { + "epoch": 1.27, + "learning_rate": 4.370329910421674e-06, + "loss": 2.9906, + "step": 5814 + }, + { + "epoch": 1.27, + "learning_rate": 4.370220668560193e-06, + "loss": 3.4002, + "step": 5815 + }, + { + "epoch": 1.27, + "learning_rate": 4.370111426698711e-06, + "loss": 3.2486, + "step": 5816 + }, + { + "epoch": 1.27, + "learning_rate": 4.37000218483723e-06, + "loss": 3.1177, + "step": 5817 + }, + { + "epoch": 1.27, + "learning_rate": 4.369892942975748e-06, + "loss": 3.002, + "step": 5818 + }, + { + "epoch": 1.27, + "learning_rate": 4.369783701114267e-06, + "loss": 3.1245, + "step": 5819 + }, + { + "epoch": 1.27, + "learning_rate": 4.3696744592527865e-06, + "loss": 3.4377, + "step": 5820 + }, + { + "epoch": 1.27, + "learning_rate": 4.369565217391305e-06, + "loss": 3.0649, + "step": 5821 + }, + { + "epoch": 1.27, + "learning_rate": 4.369455975529824e-06, + "loss": 3.1888, + "step": 5822 + }, + { + "epoch": 1.27, + "learning_rate": 4.369346733668342e-06, + "loss": 3.2523, + "step": 5823 + }, + { + "epoch": 1.27, + "learning_rate": 4.369237491806861e-06, + "loss": 3.7188, + "step": 5824 + }, + { + "epoch": 1.27, + "learning_rate": 4.36912824994538e-06, + "loss": 3.2164, + "step": 5825 + }, + { + "epoch": 1.27, + "learning_rate": 4.369019008083898e-06, + "loss": 2.6951, + "step": 5826 + }, + { + "epoch": 1.27, + "learning_rate": 4.368909766222416e-06, + "loss": 3.4773, + "step": 5827 + }, + { + "epoch": 1.27, + "learning_rate": 4.368800524360935e-06, + "loss": 2.5417, + "step": 5828 + }, + { + "epoch": 1.27, + "learning_rate": 4.368691282499454e-06, + "loss": 2.5131, + "step": 5829 + }, + { + "epoch": 1.27, + "learning_rate": 4.368582040637973e-06, + "loss": 3.2089, + "step": 5830 + }, + { + "epoch": 1.27, + "learning_rate": 4.368472798776492e-06, + "loss": 3.1448, + "step": 5831 + }, + { + "epoch": 1.27, + "learning_rate": 4.36836355691501e-06, + "loss": 3.1143, + "step": 5832 + }, + { + "epoch": 1.27, + "learning_rate": 4.368254315053529e-06, + "loss": 2.8789, + "step": 5833 + }, + { + "epoch": 1.27, + "learning_rate": 4.368145073192048e-06, + "loss": 2.5907, + "step": 5834 + }, + { + "epoch": 1.27, + "learning_rate": 4.368035831330566e-06, + "loss": 2.8111, + "step": 5835 + }, + { + "epoch": 1.27, + "learning_rate": 4.367926589469085e-06, + "loss": 3.2821, + "step": 5836 + }, + { + "epoch": 1.27, + "learning_rate": 4.367817347607603e-06, + "loss": 3.434, + "step": 5837 + }, + { + "epoch": 1.27, + "learning_rate": 4.367708105746122e-06, + "loss": 2.8951, + "step": 5838 + }, + { + "epoch": 1.27, + "learning_rate": 4.367598863884641e-06, + "loss": 3.1508, + "step": 5839 + }, + { + "epoch": 1.27, + "learning_rate": 4.3674896220231595e-06, + "loss": 2.9436, + "step": 5840 + }, + { + "epoch": 1.27, + "learning_rate": 4.3673803801616786e-06, + "loss": 2.1276, + "step": 5841 + }, + { + "epoch": 1.27, + "learning_rate": 4.367271138300197e-06, + "loss": 2.5536, + "step": 5842 + }, + { + "epoch": 1.28, + "learning_rate": 4.367161896438716e-06, + "loss": 2.5516, + "step": 5843 + }, + { + "epoch": 1.28, + "learning_rate": 4.367052654577235e-06, + "loss": 2.6816, + "step": 5844 + }, + { + "epoch": 1.28, + "learning_rate": 4.366943412715753e-06, + "loss": 3.2821, + "step": 5845 + }, + { + "epoch": 1.28, + "learning_rate": 4.366834170854272e-06, + "loss": 3.118, + "step": 5846 + }, + { + "epoch": 1.28, + "learning_rate": 4.36672492899279e-06, + "loss": 2.7846, + "step": 5847 + }, + { + "epoch": 1.28, + "learning_rate": 4.366615687131309e-06, + "loss": 2.918, + "step": 5848 + }, + { + "epoch": 1.28, + "learning_rate": 4.366506445269828e-06, + "loss": 3.3545, + "step": 5849 + }, + { + "epoch": 1.28, + "learning_rate": 4.3663972034083465e-06, + "loss": 4.0224, + "step": 5850 + }, + { + "epoch": 1.28, + "learning_rate": 4.366287961546865e-06, + "loss": 2.8323, + "step": 5851 + }, + { + "epoch": 1.28, + "learning_rate": 4.366178719685384e-06, + "loss": 2.8241, + "step": 5852 + }, + { + "epoch": 1.28, + "learning_rate": 4.366069477823903e-06, + "loss": 3.214, + "step": 5853 + }, + { + "epoch": 1.28, + "learning_rate": 4.365960235962421e-06, + "loss": 3.5995, + "step": 5854 + }, + { + "epoch": 1.28, + "learning_rate": 4.36585099410094e-06, + "loss": 3.6911, + "step": 5855 + }, + { + "epoch": 1.28, + "learning_rate": 4.365741752239458e-06, + "loss": 3.0759, + "step": 5856 + }, + { + "epoch": 1.28, + "learning_rate": 4.365632510377977e-06, + "loss": 3.4253, + "step": 5857 + }, + { + "epoch": 1.28, + "learning_rate": 4.365523268516496e-06, + "loss": 2.5056, + "step": 5858 + }, + { + "epoch": 1.28, + "learning_rate": 4.365414026655014e-06, + "loss": 3.2724, + "step": 5859 + }, + { + "epoch": 1.28, + "learning_rate": 4.3653047847935334e-06, + "loss": 2.917, + "step": 5860 + }, + { + "epoch": 1.28, + "learning_rate": 4.365195542932052e-06, + "loss": 3.0426, + "step": 5861 + }, + { + "epoch": 1.28, + "learning_rate": 4.365086301070571e-06, + "loss": 2.9541, + "step": 5862 + }, + { + "epoch": 1.28, + "learning_rate": 4.36497705920909e-06, + "loss": 3.0263, + "step": 5863 + }, + { + "epoch": 1.28, + "learning_rate": 4.364867817347608e-06, + "loss": 3.1856, + "step": 5864 + }, + { + "epoch": 1.28, + "learning_rate": 4.364758575486127e-06, + "loss": 3.2244, + "step": 5865 + }, + { + "epoch": 1.28, + "learning_rate": 4.364649333624645e-06, + "loss": 3.4254, + "step": 5866 + }, + { + "epoch": 1.28, + "learning_rate": 4.364540091763164e-06, + "loss": 2.7268, + "step": 5867 + }, + { + "epoch": 1.28, + "learning_rate": 4.364430849901682e-06, + "loss": 3.4438, + "step": 5868 + }, + { + "epoch": 1.28, + "learning_rate": 4.364321608040201e-06, + "loss": 3.0017, + "step": 5869 + }, + { + "epoch": 1.28, + "learning_rate": 4.36421236617872e-06, + "loss": 3.7978, + "step": 5870 + }, + { + "epoch": 1.28, + "learning_rate": 4.364103124317239e-06, + "loss": 2.7263, + "step": 5871 + }, + { + "epoch": 1.28, + "learning_rate": 4.363993882455758e-06, + "loss": 2.1307, + "step": 5872 + }, + { + "epoch": 1.28, + "learning_rate": 4.363884640594276e-06, + "loss": 3.0415, + "step": 5873 + }, + { + "epoch": 1.28, + "learning_rate": 4.363775398732795e-06, + "loss": 3.4586, + "step": 5874 + }, + { + "epoch": 1.28, + "learning_rate": 4.363666156871314e-06, + "loss": 2.5574, + "step": 5875 + }, + { + "epoch": 1.28, + "learning_rate": 4.363556915009832e-06, + "loss": 2.2837, + "step": 5876 + }, + { + "epoch": 1.28, + "learning_rate": 4.363447673148351e-06, + "loss": 2.6988, + "step": 5877 + }, + { + "epoch": 1.28, + "learning_rate": 4.363338431286869e-06, + "loss": 2.6027, + "step": 5878 + }, + { + "epoch": 1.28, + "learning_rate": 4.363229189425388e-06, + "loss": 2.9095, + "step": 5879 + }, + { + "epoch": 1.28, + "learning_rate": 4.3631199475639065e-06, + "loss": 3.1853, + "step": 5880 + }, + { + "epoch": 1.28, + "learning_rate": 4.3630107057024255e-06, + "loss": 2.84, + "step": 5881 + }, + { + "epoch": 1.28, + "learning_rate": 4.362901463840944e-06, + "loss": 2.8857, + "step": 5882 + }, + { + "epoch": 1.28, + "learning_rate": 4.362792221979463e-06, + "loss": 2.5266, + "step": 5883 + }, + { + "epoch": 1.28, + "learning_rate": 4.362682980117982e-06, + "loss": 3.1207, + "step": 5884 + }, + { + "epoch": 1.28, + "learning_rate": 4.3625737382565e-06, + "loss": 3.375, + "step": 5885 + }, + { + "epoch": 1.28, + "learning_rate": 4.362464496395019e-06, + "loss": 3.0251, + "step": 5886 + }, + { + "epoch": 1.28, + "learning_rate": 4.362355254533537e-06, + "loss": 2.689, + "step": 5887 + }, + { + "epoch": 1.29, + "learning_rate": 4.362246012672056e-06, + "loss": 2.2069, + "step": 5888 + }, + { + "epoch": 1.29, + "learning_rate": 4.362136770810575e-06, + "loss": 3.302, + "step": 5889 + }, + { + "epoch": 1.29, + "learning_rate": 4.3620275289490935e-06, + "loss": 3.7329, + "step": 5890 + }, + { + "epoch": 1.29, + "learning_rate": 4.3619182870876125e-06, + "loss": 2.633, + "step": 5891 + }, + { + "epoch": 1.29, + "learning_rate": 4.361809045226131e-06, + "loss": 2.3557, + "step": 5892 + }, + { + "epoch": 1.29, + "learning_rate": 4.36169980336465e-06, + "loss": 2.6805, + "step": 5893 + }, + { + "epoch": 1.29, + "learning_rate": 4.361590561503169e-06, + "loss": 3.4683, + "step": 5894 + }, + { + "epoch": 1.29, + "learning_rate": 4.361481319641687e-06, + "loss": 3.2849, + "step": 5895 + }, + { + "epoch": 1.29, + "learning_rate": 4.361372077780206e-06, + "loss": 3.0454, + "step": 5896 + }, + { + "epoch": 1.29, + "learning_rate": 4.361262835918724e-06, + "loss": 2.3764, + "step": 5897 + }, + { + "epoch": 1.29, + "learning_rate": 4.361153594057243e-06, + "loss": 3.4402, + "step": 5898 + }, + { + "epoch": 1.29, + "learning_rate": 4.361044352195762e-06, + "loss": 3.4392, + "step": 5899 + }, + { + "epoch": 1.29, + "learning_rate": 4.36093511033428e-06, + "loss": 2.5791, + "step": 5900 + }, + { + "epoch": 1.29, + "learning_rate": 4.360825868472799e-06, + "loss": 3.3168, + "step": 5901 + }, + { + "epoch": 1.29, + "learning_rate": 4.360716626611318e-06, + "loss": 2.5626, + "step": 5902 + }, + { + "epoch": 1.29, + "learning_rate": 4.360607384749837e-06, + "loss": 2.8371, + "step": 5903 + }, + { + "epoch": 1.29, + "learning_rate": 4.360498142888355e-06, + "loss": 3.7431, + "step": 5904 + }, + { + "epoch": 1.29, + "learning_rate": 4.360388901026874e-06, + "loss": 3.4871, + "step": 5905 + }, + { + "epoch": 1.29, + "learning_rate": 4.360279659165392e-06, + "loss": 2.8387, + "step": 5906 + }, + { + "epoch": 1.29, + "learning_rate": 4.360170417303911e-06, + "loss": 3.2416, + "step": 5907 + }, + { + "epoch": 1.29, + "learning_rate": 4.36006117544243e-06, + "loss": 3.1508, + "step": 5908 + }, + { + "epoch": 1.29, + "learning_rate": 4.359951933580948e-06, + "loss": 2.1595, + "step": 5909 + }, + { + "epoch": 1.29, + "learning_rate": 4.359842691719467e-06, + "loss": 2.679, + "step": 5910 + }, + { + "epoch": 1.29, + "learning_rate": 4.3597334498579856e-06, + "loss": 2.7494, + "step": 5911 + }, + { + "epoch": 1.29, + "learning_rate": 4.359624207996505e-06, + "loss": 2.1718, + "step": 5912 + }, + { + "epoch": 1.29, + "learning_rate": 4.359514966135024e-06, + "loss": 2.8575, + "step": 5913 + }, + { + "epoch": 1.29, + "learning_rate": 4.359405724273542e-06, + "loss": 2.5992, + "step": 5914 + }, + { + "epoch": 1.29, + "learning_rate": 4.35929648241206e-06, + "loss": 2.5929, + "step": 5915 + }, + { + "epoch": 1.29, + "learning_rate": 4.359187240550579e-06, + "loss": 2.3878, + "step": 5916 + }, + { + "epoch": 1.29, + "learning_rate": 4.359077998689098e-06, + "loss": 2.5305, + "step": 5917 + }, + { + "epoch": 1.29, + "learning_rate": 4.358968756827617e-06, + "loss": 3.4, + "step": 5918 + }, + { + "epoch": 1.29, + "learning_rate": 4.358859514966135e-06, + "loss": 3.334, + "step": 5919 + }, + { + "epoch": 1.29, + "learning_rate": 4.3587502731046535e-06, + "loss": 3.3694, + "step": 5920 + }, + { + "epoch": 1.29, + "learning_rate": 4.3586410312431725e-06, + "loss": 3.1977, + "step": 5921 + }, + { + "epoch": 1.29, + "learning_rate": 4.3585317893816916e-06, + "loss": 2.8527, + "step": 5922 + }, + { + "epoch": 1.29, + "learning_rate": 4.358422547520211e-06, + "loss": 2.808, + "step": 5923 + }, + { + "epoch": 1.29, + "learning_rate": 4.358313305658729e-06, + "loss": 3.7204, + "step": 5924 + }, + { + "epoch": 1.29, + "learning_rate": 4.358204063797247e-06, + "loss": 2.8163, + "step": 5925 + }, + { + "epoch": 1.29, + "learning_rate": 4.358094821935766e-06, + "loss": 3.0406, + "step": 5926 + }, + { + "epoch": 1.29, + "learning_rate": 4.357985580074285e-06, + "loss": 3.4406, + "step": 5927 + }, + { + "epoch": 1.29, + "learning_rate": 4.357876338212804e-06, + "loss": 3.0297, + "step": 5928 + }, + { + "epoch": 1.29, + "learning_rate": 4.357767096351322e-06, + "loss": 3.2641, + "step": 5929 + }, + { + "epoch": 1.29, + "learning_rate": 4.3576578544898404e-06, + "loss": 2.5151, + "step": 5930 + }, + { + "epoch": 1.29, + "learning_rate": 4.3575486126283595e-06, + "loss": 3.3209, + "step": 5931 + }, + { + "epoch": 1.29, + "learning_rate": 4.3574393707668785e-06, + "loss": 2.6071, + "step": 5932 + }, + { + "epoch": 1.29, + "learning_rate": 4.357330128905397e-06, + "loss": 2.6363, + "step": 5933 + }, + { + "epoch": 1.3, + "learning_rate": 4.357220887043916e-06, + "loss": 2.9924, + "step": 5934 + }, + { + "epoch": 1.3, + "learning_rate": 4.357111645182434e-06, + "loss": 2.9659, + "step": 5935 + }, + { + "epoch": 1.3, + "learning_rate": 4.357002403320953e-06, + "loss": 1.8045, + "step": 5936 + }, + { + "epoch": 1.3, + "learning_rate": 4.356893161459472e-06, + "loss": 3.2062, + "step": 5937 + }, + { + "epoch": 1.3, + "learning_rate": 4.35678391959799e-06, + "loss": 3.7325, + "step": 5938 + }, + { + "epoch": 1.3, + "learning_rate": 4.356674677736508e-06, + "loss": 3.5628, + "step": 5939 + }, + { + "epoch": 1.3, + "learning_rate": 4.356565435875027e-06, + "loss": 3.7053, + "step": 5940 + }, + { + "epoch": 1.3, + "learning_rate": 4.3564561940135464e-06, + "loss": 2.8459, + "step": 5941 + }, + { + "epoch": 1.3, + "learning_rate": 4.3563469521520655e-06, + "loss": 3.7668, + "step": 5942 + }, + { + "epoch": 1.3, + "learning_rate": 4.356237710290584e-06, + "loss": 3.1019, + "step": 5943 + }, + { + "epoch": 1.3, + "learning_rate": 4.356128468429102e-06, + "loss": 3.2439, + "step": 5944 + }, + { + "epoch": 1.3, + "learning_rate": 4.356019226567621e-06, + "loss": 4.3284, + "step": 5945 + }, + { + "epoch": 1.3, + "learning_rate": 4.35590998470614e-06, + "loss": 3.2121, + "step": 5946 + }, + { + "epoch": 1.3, + "learning_rate": 4.355800742844659e-06, + "loss": 2.828, + "step": 5947 + }, + { + "epoch": 1.3, + "learning_rate": 4.355691500983177e-06, + "loss": 3.0829, + "step": 5948 + }, + { + "epoch": 1.3, + "learning_rate": 4.355582259121695e-06, + "loss": 3.0039, + "step": 5949 + }, + { + "epoch": 1.3, + "learning_rate": 4.355473017260214e-06, + "loss": 2.4549, + "step": 5950 + }, + { + "epoch": 1.3, + "learning_rate": 4.355363775398733e-06, + "loss": 2.9398, + "step": 5951 + }, + { + "epoch": 1.3, + "learning_rate": 4.355254533537252e-06, + "loss": 2.4462, + "step": 5952 + }, + { + "epoch": 1.3, + "learning_rate": 4.355145291675771e-06, + "loss": 3.7691, + "step": 5953 + }, + { + "epoch": 1.3, + "learning_rate": 4.355036049814289e-06, + "loss": 2.7548, + "step": 5954 + }, + { + "epoch": 1.3, + "learning_rate": 4.354926807952808e-06, + "loss": 3.6614, + "step": 5955 + }, + { + "epoch": 1.3, + "learning_rate": 4.354817566091327e-06, + "loss": 3.0416, + "step": 5956 + }, + { + "epoch": 1.3, + "learning_rate": 4.354708324229845e-06, + "loss": 3.3262, + "step": 5957 + }, + { + "epoch": 1.3, + "learning_rate": 4.354599082368364e-06, + "loss": 3.4965, + "step": 5958 + }, + { + "epoch": 1.3, + "learning_rate": 4.354489840506882e-06, + "loss": 3.7161, + "step": 5959 + }, + { + "epoch": 1.3, + "learning_rate": 4.354380598645401e-06, + "loss": 3.3866, + "step": 5960 + }, + { + "epoch": 1.3, + "learning_rate": 4.35427135678392e-06, + "loss": 3.1789, + "step": 5961 + }, + { + "epoch": 1.3, + "learning_rate": 4.3541621149224385e-06, + "loss": 2.259, + "step": 5962 + }, + { + "epoch": 1.3, + "learning_rate": 4.354052873060957e-06, + "loss": 2.8538, + "step": 5963 + }, + { + "epoch": 1.3, + "learning_rate": 4.353943631199476e-06, + "loss": 2.8176, + "step": 5964 + }, + { + "epoch": 1.3, + "learning_rate": 4.353834389337995e-06, + "loss": 3.1829, + "step": 5965 + }, + { + "epoch": 1.3, + "learning_rate": 4.353725147476514e-06, + "loss": 3.4879, + "step": 5966 + }, + { + "epoch": 1.3, + "learning_rate": 4.353615905615032e-06, + "loss": 2.6477, + "step": 5967 + }, + { + "epoch": 1.3, + "learning_rate": 4.35350666375355e-06, + "loss": 3.5731, + "step": 5968 + }, + { + "epoch": 1.3, + "learning_rate": 4.353397421892069e-06, + "loss": 3.1188, + "step": 5969 + }, + { + "epoch": 1.3, + "learning_rate": 4.353288180030588e-06, + "loss": 3.4984, + "step": 5970 + }, + { + "epoch": 1.3, + "learning_rate": 4.353178938169107e-06, + "loss": 2.2182, + "step": 5971 + }, + { + "epoch": 1.3, + "learning_rate": 4.3530696963076255e-06, + "loss": 2.6523, + "step": 5972 + }, + { + "epoch": 1.3, + "learning_rate": 4.352960454446144e-06, + "loss": 3.2597, + "step": 5973 + }, + { + "epoch": 1.3, + "learning_rate": 4.352851212584663e-06, + "loss": 3.8044, + "step": 5974 + }, + { + "epoch": 1.3, + "learning_rate": 4.352741970723182e-06, + "loss": 3.3837, + "step": 5975 + }, + { + "epoch": 1.3, + "learning_rate": 4.352632728861701e-06, + "loss": 3.5665, + "step": 5976 + }, + { + "epoch": 1.3, + "learning_rate": 4.352523487000219e-06, + "loss": 2.9706, + "step": 5977 + }, + { + "epoch": 1.3, + "learning_rate": 4.352414245138737e-06, + "loss": 3.0357, + "step": 5978 + }, + { + "epoch": 1.3, + "learning_rate": 4.352305003277256e-06, + "loss": 3.0833, + "step": 5979 + }, + { + "epoch": 1.31, + "learning_rate": 4.352195761415775e-06, + "loss": 3.4778, + "step": 5980 + }, + { + "epoch": 1.31, + "learning_rate": 4.352086519554293e-06, + "loss": 2.6754, + "step": 5981 + }, + { + "epoch": 1.31, + "learning_rate": 4.3519772776928124e-06, + "loss": 3.3854, + "step": 5982 + }, + { + "epoch": 1.31, + "learning_rate": 4.351868035831331e-06, + "loss": 2.9387, + "step": 5983 + }, + { + "epoch": 1.31, + "learning_rate": 4.35175879396985e-06, + "loss": 3.2363, + "step": 5984 + }, + { + "epoch": 1.31, + "learning_rate": 4.351649552108369e-06, + "loss": 2.9616, + "step": 5985 + }, + { + "epoch": 1.31, + "learning_rate": 4.351540310246887e-06, + "loss": 3.245, + "step": 5986 + }, + { + "epoch": 1.31, + "learning_rate": 4.351431068385406e-06, + "loss": 2.4037, + "step": 5987 + }, + { + "epoch": 1.31, + "learning_rate": 4.351321826523924e-06, + "loss": 2.9986, + "step": 5988 + }, + { + "epoch": 1.31, + "learning_rate": 4.351212584662443e-06, + "loss": 3.1679, + "step": 5989 + }, + { + "epoch": 1.31, + "learning_rate": 4.351103342800962e-06, + "loss": 3.2383, + "step": 5990 + }, + { + "epoch": 1.31, + "learning_rate": 4.35099410093948e-06, + "loss": 3.2232, + "step": 5991 + }, + { + "epoch": 1.31, + "learning_rate": 4.3508848590779986e-06, + "loss": 2.5481, + "step": 5992 + }, + { + "epoch": 1.31, + "learning_rate": 4.350775617216518e-06, + "loss": 3.3417, + "step": 5993 + }, + { + "epoch": 1.31, + "learning_rate": 4.350666375355037e-06, + "loss": 3.3561, + "step": 5994 + }, + { + "epoch": 1.31, + "learning_rate": 4.350557133493556e-06, + "loss": 3.4535, + "step": 5995 + }, + { + "epoch": 1.31, + "learning_rate": 4.350447891632074e-06, + "loss": 2.678, + "step": 5996 + }, + { + "epoch": 1.31, + "learning_rate": 4.350338649770592e-06, + "loss": 2.3629, + "step": 5997 + }, + { + "epoch": 1.31, + "learning_rate": 4.350229407909111e-06, + "loss": 3.4098, + "step": 5998 + }, + { + "epoch": 1.31, + "learning_rate": 4.35012016604763e-06, + "loss": 2.5802, + "step": 5999 + }, + { + "epoch": 1.31, + "learning_rate": 4.350010924186148e-06, + "loss": 3.6601, + "step": 6000 + }, + { + "epoch": 1.31, + "learning_rate": 4.349901682324667e-06, + "loss": 2.6565, + "step": 6001 + }, + { + "epoch": 1.31, + "learning_rate": 4.3497924404631855e-06, + "loss": 3.4726, + "step": 6002 + }, + { + "epoch": 1.31, + "learning_rate": 4.3496831986017045e-06, + "loss": 3.5412, + "step": 6003 + }, + { + "epoch": 1.31, + "learning_rate": 4.349573956740224e-06, + "loss": 3.0743, + "step": 6004 + }, + { + "epoch": 1.31, + "learning_rate": 4.349464714878742e-06, + "loss": 3.5324, + "step": 6005 + }, + { + "epoch": 1.31, + "learning_rate": 4.349355473017261e-06, + "loss": 2.4587, + "step": 6006 + }, + { + "epoch": 1.31, + "learning_rate": 4.349246231155779e-06, + "loss": 3.7723, + "step": 6007 + }, + { + "epoch": 1.31, + "learning_rate": 4.349136989294298e-06, + "loss": 2.7223, + "step": 6008 + }, + { + "epoch": 1.31, + "learning_rate": 4.349027747432817e-06, + "loss": 3.1033, + "step": 6009 + }, + { + "epoch": 1.31, + "learning_rate": 4.348918505571335e-06, + "loss": 3.1029, + "step": 6010 + }, + { + "epoch": 1.31, + "learning_rate": 4.348809263709854e-06, + "loss": 3.0697, + "step": 6011 + }, + { + "epoch": 1.31, + "learning_rate": 4.3487000218483725e-06, + "loss": 3.3528, + "step": 6012 + }, + { + "epoch": 1.31, + "learning_rate": 4.3485907799868915e-06, + "loss": 3.2765, + "step": 6013 + }, + { + "epoch": 1.31, + "learning_rate": 4.34848153812541e-06, + "loss": 3.7147, + "step": 6014 + }, + { + "epoch": 1.31, + "learning_rate": 4.348372296263929e-06, + "loss": 2.6968, + "step": 6015 + }, + { + "epoch": 1.31, + "learning_rate": 4.348263054402447e-06, + "loss": 3.4809, + "step": 6016 + }, + { + "epoch": 1.31, + "learning_rate": 4.348153812540966e-06, + "loss": 3.4765, + "step": 6017 + }, + { + "epoch": 1.31, + "learning_rate": 4.348044570679485e-06, + "loss": 2.8417, + "step": 6018 + }, + { + "epoch": 1.31, + "learning_rate": 4.347935328818003e-06, + "loss": 3.5469, + "step": 6019 + }, + { + "epoch": 1.31, + "learning_rate": 4.347826086956522e-06, + "loss": 3.4256, + "step": 6020 + }, + { + "epoch": 1.31, + "learning_rate": 4.34771684509504e-06, + "loss": 2.8065, + "step": 6021 + }, + { + "epoch": 1.31, + "learning_rate": 4.3476076032335594e-06, + "loss": 3.4996, + "step": 6022 + }, + { + "epoch": 1.31, + "learning_rate": 4.3474983613720785e-06, + "loss": 3.351, + "step": 6023 + }, + { + "epoch": 1.31, + "learning_rate": 4.347389119510597e-06, + "loss": 2.8811, + "step": 6024 + }, + { + "epoch": 1.31, + "learning_rate": 4.347279877649116e-06, + "loss": 3.5686, + "step": 6025 + }, + { + "epoch": 1.32, + "learning_rate": 4.347170635787634e-06, + "loss": 3.3185, + "step": 6026 + }, + { + "epoch": 1.32, + "learning_rate": 4.347061393926153e-06, + "loss": 3.7587, + "step": 6027 + }, + { + "epoch": 1.32, + "learning_rate": 4.346952152064672e-06, + "loss": 3.2638, + "step": 6028 + }, + { + "epoch": 1.32, + "learning_rate": 4.34684291020319e-06, + "loss": 3.4634, + "step": 6029 + }, + { + "epoch": 1.32, + "learning_rate": 4.346733668341709e-06, + "loss": 2.4304, + "step": 6030 + }, + { + "epoch": 1.32, + "learning_rate": 4.346624426480227e-06, + "loss": 3.0268, + "step": 6031 + }, + { + "epoch": 1.32, + "learning_rate": 4.346515184618746e-06, + "loss": 2.7435, + "step": 6032 + }, + { + "epoch": 1.32, + "learning_rate": 4.3464059427572646e-06, + "loss": 2.7006, + "step": 6033 + }, + { + "epoch": 1.32, + "learning_rate": 4.346296700895784e-06, + "loss": 2.9641, + "step": 6034 + }, + { + "epoch": 1.32, + "learning_rate": 4.346187459034303e-06, + "loss": 3.1125, + "step": 6035 + }, + { + "epoch": 1.32, + "learning_rate": 4.346078217172821e-06, + "loss": 2.8504, + "step": 6036 + }, + { + "epoch": 1.32, + "learning_rate": 4.34596897531134e-06, + "loss": 3.3361, + "step": 6037 + }, + { + "epoch": 1.32, + "learning_rate": 4.345859733449858e-06, + "loss": 3.0162, + "step": 6038 + }, + { + "epoch": 1.32, + "learning_rate": 4.345750491588377e-06, + "loss": 3.455, + "step": 6039 + }, + { + "epoch": 1.32, + "learning_rate": 4.345641249726896e-06, + "loss": 3.3454, + "step": 6040 + }, + { + "epoch": 1.32, + "learning_rate": 4.345532007865414e-06, + "loss": 3.8344, + "step": 6041 + }, + { + "epoch": 1.32, + "learning_rate": 4.345422766003933e-06, + "loss": 2.7341, + "step": 6042 + }, + { + "epoch": 1.32, + "learning_rate": 4.3453135241424515e-06, + "loss": 3.1126, + "step": 6043 + }, + { + "epoch": 1.32, + "learning_rate": 4.3452042822809706e-06, + "loss": 3.5574, + "step": 6044 + }, + { + "epoch": 1.32, + "learning_rate": 4.345095040419489e-06, + "loss": 2.7463, + "step": 6045 + }, + { + "epoch": 1.32, + "learning_rate": 4.344985798558008e-06, + "loss": 2.7043, + "step": 6046 + }, + { + "epoch": 1.32, + "learning_rate": 4.344876556696526e-06, + "loss": 2.2272, + "step": 6047 + }, + { + "epoch": 1.32, + "learning_rate": 4.344767314835045e-06, + "loss": 3.1065, + "step": 6048 + }, + { + "epoch": 1.32, + "learning_rate": 4.344658072973564e-06, + "loss": 3.3012, + "step": 6049 + }, + { + "epoch": 1.32, + "learning_rate": 4.344548831112082e-06, + "loss": 2.9253, + "step": 6050 + }, + { + "epoch": 1.32, + "learning_rate": 4.344439589250601e-06, + "loss": 3.0823, + "step": 6051 + }, + { + "epoch": 1.32, + "learning_rate": 4.3443303473891195e-06, + "loss": 3.5889, + "step": 6052 + }, + { + "epoch": 1.32, + "learning_rate": 4.3442211055276385e-06, + "loss": 2.8452, + "step": 6053 + }, + { + "epoch": 1.32, + "learning_rate": 4.3441118636661575e-06, + "loss": 2.2721, + "step": 6054 + }, + { + "epoch": 1.32, + "learning_rate": 4.344002621804676e-06, + "loss": 3.1433, + "step": 6055 + }, + { + "epoch": 1.32, + "learning_rate": 4.343893379943195e-06, + "loss": 2.3238, + "step": 6056 + }, + { + "epoch": 1.32, + "learning_rate": 4.343784138081713e-06, + "loss": 3.6357, + "step": 6057 + }, + { + "epoch": 1.32, + "learning_rate": 4.343674896220232e-06, + "loss": 2.6314, + "step": 6058 + }, + { + "epoch": 1.32, + "learning_rate": 4.343565654358751e-06, + "loss": 2.3197, + "step": 6059 + }, + { + "epoch": 1.32, + "learning_rate": 4.343456412497269e-06, + "loss": 2.8467, + "step": 6060 + }, + { + "epoch": 1.32, + "learning_rate": 4.343347170635788e-06, + "loss": 2.4898, + "step": 6061 + }, + { + "epoch": 1.32, + "learning_rate": 4.343237928774306e-06, + "loss": 3.7724, + "step": 6062 + }, + { + "epoch": 1.32, + "learning_rate": 4.3431286869128254e-06, + "loss": 3.405, + "step": 6063 + }, + { + "epoch": 1.32, + "learning_rate": 4.3430194450513445e-06, + "loss": 2.5501, + "step": 6064 + }, + { + "epoch": 1.32, + "learning_rate": 4.342910203189863e-06, + "loss": 2.8685, + "step": 6065 + }, + { + "epoch": 1.32, + "learning_rate": 4.342800961328381e-06, + "loss": 3.1019, + "step": 6066 + }, + { + "epoch": 1.32, + "learning_rate": 4.3426917194669e-06, + "loss": 3.5112, + "step": 6067 + }, + { + "epoch": 1.32, + "learning_rate": 4.342582477605419e-06, + "loss": 2.5803, + "step": 6068 + }, + { + "epoch": 1.32, + "learning_rate": 4.342473235743938e-06, + "loss": 3.2991, + "step": 6069 + }, + { + "epoch": 1.32, + "learning_rate": 4.342363993882456e-06, + "loss": 3.0466, + "step": 6070 + }, + { + "epoch": 1.32, + "learning_rate": 4.342254752020974e-06, + "loss": 2.9059, + "step": 6071 + }, + { + "epoch": 1.33, + "learning_rate": 4.342145510159493e-06, + "loss": 3.1148, + "step": 6072 + }, + { + "epoch": 1.33, + "learning_rate": 4.342036268298012e-06, + "loss": 3.0331, + "step": 6073 + }, + { + "epoch": 1.33, + "learning_rate": 4.341927026436531e-06, + "loss": 3.0472, + "step": 6074 + }, + { + "epoch": 1.33, + "learning_rate": 4.34181778457505e-06, + "loss": 3.2147, + "step": 6075 + }, + { + "epoch": 1.33, + "learning_rate": 4.341708542713568e-06, + "loss": 3.7043, + "step": 6076 + }, + { + "epoch": 1.33, + "learning_rate": 4.341599300852087e-06, + "loss": 3.3875, + "step": 6077 + }, + { + "epoch": 1.33, + "learning_rate": 4.341490058990606e-06, + "loss": 2.6385, + "step": 6078 + }, + { + "epoch": 1.33, + "learning_rate": 4.341380817129124e-06, + "loss": 2.7349, + "step": 6079 + }, + { + "epoch": 1.33, + "learning_rate": 4.341271575267642e-06, + "loss": 2.6942, + "step": 6080 + }, + { + "epoch": 1.33, + "learning_rate": 4.341162333406161e-06, + "loss": 2.8461, + "step": 6081 + }, + { + "epoch": 1.33, + "learning_rate": 4.34105309154468e-06, + "loss": 3.6676, + "step": 6082 + }, + { + "epoch": 1.33, + "learning_rate": 4.340943849683199e-06, + "loss": 3.2417, + "step": 6083 + }, + { + "epoch": 1.33, + "learning_rate": 4.3408346078217175e-06, + "loss": 3.0217, + "step": 6084 + }, + { + "epoch": 1.33, + "learning_rate": 4.340725365960236e-06, + "loss": 3.0478, + "step": 6085 + }, + { + "epoch": 1.33, + "learning_rate": 4.340616124098755e-06, + "loss": 2.5985, + "step": 6086 + }, + { + "epoch": 1.33, + "learning_rate": 4.340506882237274e-06, + "loss": 3.065, + "step": 6087 + }, + { + "epoch": 1.33, + "learning_rate": 4.340397640375793e-06, + "loss": 2.8032, + "step": 6088 + }, + { + "epoch": 1.33, + "learning_rate": 4.340288398514311e-06, + "loss": 2.4243, + "step": 6089 + }, + { + "epoch": 1.33, + "learning_rate": 4.340179156652829e-06, + "loss": 2.607, + "step": 6090 + }, + { + "epoch": 1.33, + "learning_rate": 4.340069914791348e-06, + "loss": 2.5443, + "step": 6091 + }, + { + "epoch": 1.33, + "learning_rate": 4.339960672929867e-06, + "loss": 2.778, + "step": 6092 + }, + { + "epoch": 1.33, + "learning_rate": 4.339851431068386e-06, + "loss": 3.1271, + "step": 6093 + }, + { + "epoch": 1.33, + "learning_rate": 4.3397421892069045e-06, + "loss": 3.2792, + "step": 6094 + }, + { + "epoch": 1.33, + "learning_rate": 4.339632947345423e-06, + "loss": 2.8951, + "step": 6095 + }, + { + "epoch": 1.33, + "learning_rate": 4.339523705483942e-06, + "loss": 3.4037, + "step": 6096 + }, + { + "epoch": 1.33, + "learning_rate": 4.339414463622461e-06, + "loss": 3.0596, + "step": 6097 + }, + { + "epoch": 1.33, + "learning_rate": 4.339305221760979e-06, + "loss": 3.1813, + "step": 6098 + }, + { + "epoch": 1.33, + "learning_rate": 4.339195979899498e-06, + "loss": 2.9319, + "step": 6099 + }, + { + "epoch": 1.33, + "learning_rate": 4.339086738038016e-06, + "loss": 2.6483, + "step": 6100 + }, + { + "epoch": 1.33, + "learning_rate": 4.338977496176535e-06, + "loss": 3.2949, + "step": 6101 + }, + { + "epoch": 1.33, + "learning_rate": 4.338868254315054e-06, + "loss": 3.2229, + "step": 6102 + }, + { + "epoch": 1.33, + "learning_rate": 4.338759012453572e-06, + "loss": 2.7371, + "step": 6103 + }, + { + "epoch": 1.33, + "learning_rate": 4.338649770592091e-06, + "loss": 2.3615, + "step": 6104 + }, + { + "epoch": 1.33, + "learning_rate": 4.33854052873061e-06, + "loss": 2.7402, + "step": 6105 + }, + { + "epoch": 1.33, + "learning_rate": 4.338431286869129e-06, + "loss": 3.2399, + "step": 6106 + }, + { + "epoch": 1.33, + "learning_rate": 4.338322045007648e-06, + "loss": 2.6594, + "step": 6107 + }, + { + "epoch": 1.33, + "learning_rate": 4.338212803146166e-06, + "loss": 3.1658, + "step": 6108 + }, + { + "epoch": 1.33, + "learning_rate": 4.338103561284684e-06, + "loss": 3.2117, + "step": 6109 + }, + { + "epoch": 1.33, + "learning_rate": 4.337994319423203e-06, + "loss": 2.5689, + "step": 6110 + }, + { + "epoch": 1.33, + "learning_rate": 4.337885077561722e-06, + "loss": 2.3016, + "step": 6111 + }, + { + "epoch": 1.33, + "learning_rate": 4.337775835700241e-06, + "loss": 3.5101, + "step": 6112 + }, + { + "epoch": 1.33, + "learning_rate": 4.337666593838759e-06, + "loss": 3.5004, + "step": 6113 + }, + { + "epoch": 1.33, + "learning_rate": 4.3375573519772776e-06, + "loss": 2.7513, + "step": 6114 + }, + { + "epoch": 1.33, + "learning_rate": 4.337448110115797e-06, + "loss": 2.1896, + "step": 6115 + }, + { + "epoch": 1.33, + "learning_rate": 4.337338868254316e-06, + "loss": 2.0702, + "step": 6116 + }, + { + "epoch": 1.34, + "learning_rate": 4.337229626392835e-06, + "loss": 2.818, + "step": 6117 + }, + { + "epoch": 1.34, + "learning_rate": 4.337120384531353e-06, + "loss": 3.0793, + "step": 6118 + }, + { + "epoch": 1.34, + "learning_rate": 4.337011142669871e-06, + "loss": 3.1231, + "step": 6119 + }, + { + "epoch": 1.34, + "learning_rate": 4.33690190080839e-06, + "loss": 3.2724, + "step": 6120 + }, + { + "epoch": 1.34, + "learning_rate": 4.336792658946909e-06, + "loss": 3.2322, + "step": 6121 + }, + { + "epoch": 1.34, + "learning_rate": 4.336683417085428e-06, + "loss": 3.4178, + "step": 6122 + }, + { + "epoch": 1.34, + "learning_rate": 4.336574175223946e-06, + "loss": 3.7002, + "step": 6123 + }, + { + "epoch": 1.34, + "learning_rate": 4.3364649333624645e-06, + "loss": 2.5997, + "step": 6124 + }, + { + "epoch": 1.34, + "learning_rate": 4.3363556915009836e-06, + "loss": 2.7636, + "step": 6125 + }, + { + "epoch": 1.34, + "learning_rate": 4.336246449639503e-06, + "loss": 2.6728, + "step": 6126 + }, + { + "epoch": 1.34, + "learning_rate": 4.336137207778021e-06, + "loss": 2.4882, + "step": 6127 + }, + { + "epoch": 1.34, + "learning_rate": 4.33602796591654e-06, + "loss": 2.7705, + "step": 6128 + }, + { + "epoch": 1.34, + "learning_rate": 4.335918724055058e-06, + "loss": 2.822, + "step": 6129 + }, + { + "epoch": 1.34, + "learning_rate": 4.335809482193577e-06, + "loss": 2.8577, + "step": 6130 + }, + { + "epoch": 1.34, + "learning_rate": 4.335700240332096e-06, + "loss": 2.3045, + "step": 6131 + }, + { + "epoch": 1.34, + "learning_rate": 4.335590998470614e-06, + "loss": 3.0792, + "step": 6132 + }, + { + "epoch": 1.34, + "learning_rate": 4.3354817566091324e-06, + "loss": 2.0824, + "step": 6133 + }, + { + "epoch": 1.34, + "learning_rate": 4.3353725147476515e-06, + "loss": 2.8508, + "step": 6134 + }, + { + "epoch": 1.34, + "learning_rate": 4.3352632728861705e-06, + "loss": 2.5708, + "step": 6135 + }, + { + "epoch": 1.34, + "learning_rate": 4.3351540310246896e-06, + "loss": 3.2739, + "step": 6136 + }, + { + "epoch": 1.34, + "learning_rate": 4.335044789163208e-06, + "loss": 2.9232, + "step": 6137 + }, + { + "epoch": 1.34, + "learning_rate": 4.334935547301726e-06, + "loss": 2.9609, + "step": 6138 + }, + { + "epoch": 1.34, + "learning_rate": 4.334826305440245e-06, + "loss": 2.9438, + "step": 6139 + }, + { + "epoch": 1.34, + "learning_rate": 4.334717063578764e-06, + "loss": 3.6069, + "step": 6140 + }, + { + "epoch": 1.34, + "learning_rate": 4.334607821717283e-06, + "loss": 3.1164, + "step": 6141 + }, + { + "epoch": 1.34, + "learning_rate": 4.334498579855801e-06, + "loss": 3.9588, + "step": 6142 + }, + { + "epoch": 1.34, + "learning_rate": 4.334389337994319e-06, + "loss": 2.3394, + "step": 6143 + }, + { + "epoch": 1.34, + "learning_rate": 4.3342800961328384e-06, + "loss": 3.1977, + "step": 6144 + }, + { + "epoch": 1.34, + "learning_rate": 4.3341708542713575e-06, + "loss": 3.7243, + "step": 6145 + }, + { + "epoch": 1.34, + "learning_rate": 4.334061612409876e-06, + "loss": 2.7972, + "step": 6146 + }, + { + "epoch": 1.34, + "learning_rate": 4.333952370548395e-06, + "loss": 3.7122, + "step": 6147 + }, + { + "epoch": 1.34, + "learning_rate": 4.333843128686913e-06, + "loss": 3.0513, + "step": 6148 + }, + { + "epoch": 1.34, + "learning_rate": 4.333733886825432e-06, + "loss": 2.5966, + "step": 6149 + }, + { + "epoch": 1.34, + "learning_rate": 4.333624644963951e-06, + "loss": 2.6775, + "step": 6150 + }, + { + "epoch": 1.34, + "learning_rate": 4.333515403102469e-06, + "loss": 2.7642, + "step": 6151 + }, + { + "epoch": 1.34, + "learning_rate": 4.333406161240988e-06, + "loss": 3.1227, + "step": 6152 + }, + { + "epoch": 1.34, + "learning_rate": 4.333296919379506e-06, + "loss": 2.6662, + "step": 6153 + }, + { + "epoch": 1.34, + "learning_rate": 4.333187677518025e-06, + "loss": 2.3762, + "step": 6154 + }, + { + "epoch": 1.34, + "learning_rate": 4.3330784356565444e-06, + "loss": 3.2825, + "step": 6155 + }, + { + "epoch": 1.34, + "learning_rate": 4.332969193795063e-06, + "loss": 3.645, + "step": 6156 + }, + { + "epoch": 1.34, + "learning_rate": 4.332859951933581e-06, + "loss": 2.2696, + "step": 6157 + }, + { + "epoch": 1.34, + "learning_rate": 4.3327507100721e-06, + "loss": 3.2173, + "step": 6158 + }, + { + "epoch": 1.34, + "learning_rate": 4.332641468210619e-06, + "loss": 2.8347, + "step": 6159 + }, + { + "epoch": 1.34, + "learning_rate": 4.332532226349138e-06, + "loss": 2.0982, + "step": 6160 + }, + { + "epoch": 1.34, + "learning_rate": 4.332422984487656e-06, + "loss": 2.2338, + "step": 6161 + }, + { + "epoch": 1.34, + "learning_rate": 4.332313742626174e-06, + "loss": 3.6889, + "step": 6162 + }, + { + "epoch": 1.35, + "learning_rate": 4.332204500764693e-06, + "loss": 3.2595, + "step": 6163 + }, + { + "epoch": 1.35, + "learning_rate": 4.332095258903212e-06, + "loss": 3.3085, + "step": 6164 + }, + { + "epoch": 1.35, + "learning_rate": 4.3319860170417305e-06, + "loss": 3.6267, + "step": 6165 + }, + { + "epoch": 1.35, + "learning_rate": 4.3318767751802496e-06, + "loss": 2.7542, + "step": 6166 + }, + { + "epoch": 1.35, + "learning_rate": 4.331767533318768e-06, + "loss": 2.8078, + "step": 6167 + }, + { + "epoch": 1.35, + "learning_rate": 4.331658291457287e-06, + "loss": 3.5426, + "step": 6168 + }, + { + "epoch": 1.35, + "learning_rate": 4.331549049595806e-06, + "loss": 3.7071, + "step": 6169 + }, + { + "epoch": 1.35, + "learning_rate": 4.331439807734324e-06, + "loss": 3.5171, + "step": 6170 + }, + { + "epoch": 1.35, + "learning_rate": 4.331330565872843e-06, + "loss": 2.6504, + "step": 6171 + }, + { + "epoch": 1.35, + "learning_rate": 4.331221324011361e-06, + "loss": 3.5369, + "step": 6172 + }, + { + "epoch": 1.35, + "learning_rate": 4.33111208214988e-06, + "loss": 2.3384, + "step": 6173 + }, + { + "epoch": 1.35, + "learning_rate": 4.331002840288399e-06, + "loss": 3.1492, + "step": 6174 + }, + { + "epoch": 1.35, + "learning_rate": 4.3308935984269175e-06, + "loss": 3.7271, + "step": 6175 + }, + { + "epoch": 1.35, + "learning_rate": 4.3307843565654365e-06, + "loss": 2.5855, + "step": 6176 + }, + { + "epoch": 1.35, + "learning_rate": 4.330675114703955e-06, + "loss": 3.2834, + "step": 6177 + }, + { + "epoch": 1.35, + "learning_rate": 4.330565872842474e-06, + "loss": 2.1697, + "step": 6178 + }, + { + "epoch": 1.35, + "learning_rate": 4.330456630980992e-06, + "loss": 2.5237, + "step": 6179 + }, + { + "epoch": 1.35, + "learning_rate": 4.330347389119511e-06, + "loss": 2.7157, + "step": 6180 + }, + { + "epoch": 1.35, + "learning_rate": 4.33023814725803e-06, + "loss": 2.6629, + "step": 6181 + }, + { + "epoch": 1.35, + "learning_rate": 4.330128905396548e-06, + "loss": 3.0985, + "step": 6182 + }, + { + "epoch": 1.35, + "learning_rate": 4.330019663535067e-06, + "loss": 3.7863, + "step": 6183 + }, + { + "epoch": 1.35, + "learning_rate": 4.329910421673585e-06, + "loss": 3.9579, + "step": 6184 + }, + { + "epoch": 1.35, + "learning_rate": 4.3298011798121045e-06, + "loss": 2.7546, + "step": 6185 + }, + { + "epoch": 1.35, + "learning_rate": 4.329691937950623e-06, + "loss": 3.5153, + "step": 6186 + }, + { + "epoch": 1.35, + "learning_rate": 4.329582696089142e-06, + "loss": 2.9835, + "step": 6187 + }, + { + "epoch": 1.35, + "learning_rate": 4.329473454227661e-06, + "loss": 3.2504, + "step": 6188 + }, + { + "epoch": 1.35, + "learning_rate": 4.329364212366179e-06, + "loss": 2.7503, + "step": 6189 + }, + { + "epoch": 1.35, + "learning_rate": 4.329254970504698e-06, + "loss": 2.737, + "step": 6190 + }, + { + "epoch": 1.35, + "learning_rate": 4.329145728643216e-06, + "loss": 3.2967, + "step": 6191 + }, + { + "epoch": 1.35, + "learning_rate": 4.329036486781735e-06, + "loss": 2.9853, + "step": 6192 + }, + { + "epoch": 1.35, + "learning_rate": 4.328927244920254e-06, + "loss": 3.3481, + "step": 6193 + }, + { + "epoch": 1.35, + "learning_rate": 4.328818003058772e-06, + "loss": 3.1574, + "step": 6194 + }, + { + "epoch": 1.35, + "learning_rate": 4.328708761197291e-06, + "loss": 2.4759, + "step": 6195 + }, + { + "epoch": 1.35, + "learning_rate": 4.32859951933581e-06, + "loss": 3.0982, + "step": 6196 + }, + { + "epoch": 1.35, + "learning_rate": 4.328490277474329e-06, + "loss": 3.1624, + "step": 6197 + }, + { + "epoch": 1.35, + "learning_rate": 4.328381035612847e-06, + "loss": 2.9139, + "step": 6198 + }, + { + "epoch": 1.35, + "learning_rate": 4.328271793751366e-06, + "loss": 2.2374, + "step": 6199 + }, + { + "epoch": 1.35, + "learning_rate": 4.328162551889885e-06, + "loss": 2.7211, + "step": 6200 + }, + { + "epoch": 1.35, + "learning_rate": 4.328053310028403e-06, + "loss": 3.0571, + "step": 6201 + }, + { + "epoch": 1.35, + "learning_rate": 4.327944068166922e-06, + "loss": 3.3719, + "step": 6202 + }, + { + "epoch": 1.35, + "learning_rate": 4.32783482630544e-06, + "loss": 2.5562, + "step": 6203 + }, + { + "epoch": 1.35, + "learning_rate": 4.327725584443959e-06, + "loss": 2.2902, + "step": 6204 + }, + { + "epoch": 1.35, + "learning_rate": 4.327616342582478e-06, + "loss": 2.2342, + "step": 6205 + }, + { + "epoch": 1.35, + "learning_rate": 4.3275071007209966e-06, + "loss": 2.7341, + "step": 6206 + }, + { + "epoch": 1.35, + "learning_rate": 4.327397858859516e-06, + "loss": 3.7337, + "step": 6207 + }, + { + "epoch": 1.35, + "learning_rate": 4.327288616998034e-06, + "loss": 3.1801, + "step": 6208 + }, + { + "epoch": 1.36, + "learning_rate": 4.327179375136553e-06, + "loss": 3.0464, + "step": 6209 + }, + { + "epoch": 1.36, + "learning_rate": 4.327070133275071e-06, + "loss": 4.0556, + "step": 6210 + }, + { + "epoch": 1.36, + "learning_rate": 4.32696089141359e-06, + "loss": 3.2057, + "step": 6211 + }, + { + "epoch": 1.36, + "learning_rate": 4.326851649552108e-06, + "loss": 3.296, + "step": 6212 + }, + { + "epoch": 1.36, + "learning_rate": 4.326742407690627e-06, + "loss": 3.2344, + "step": 6213 + }, + { + "epoch": 1.36, + "learning_rate": 4.326633165829146e-06, + "loss": 2.1749, + "step": 6214 + }, + { + "epoch": 1.36, + "learning_rate": 4.3265239239676645e-06, + "loss": 2.8369, + "step": 6215 + }, + { + "epoch": 1.36, + "learning_rate": 4.3264146821061835e-06, + "loss": 2.5006, + "step": 6216 + }, + { + "epoch": 1.36, + "learning_rate": 4.326305440244702e-06, + "loss": 2.1805, + "step": 6217 + }, + { + "epoch": 1.36, + "learning_rate": 4.326196198383221e-06, + "loss": 3.4272, + "step": 6218 + }, + { + "epoch": 1.36, + "learning_rate": 4.32608695652174e-06, + "loss": 2.8894, + "step": 6219 + }, + { + "epoch": 1.36, + "learning_rate": 4.325977714660258e-06, + "loss": 3.359, + "step": 6220 + }, + { + "epoch": 1.36, + "learning_rate": 4.325868472798777e-06, + "loss": 2.4998, + "step": 6221 + }, + { + "epoch": 1.36, + "learning_rate": 4.325759230937295e-06, + "loss": 3.1596, + "step": 6222 + }, + { + "epoch": 1.36, + "learning_rate": 4.325649989075814e-06, + "loss": 2.6964, + "step": 6223 + }, + { + "epoch": 1.36, + "learning_rate": 4.325540747214333e-06, + "loss": 2.9841, + "step": 6224 + }, + { + "epoch": 1.36, + "learning_rate": 4.3254315053528514e-06, + "loss": 3.1427, + "step": 6225 + }, + { + "epoch": 1.36, + "learning_rate": 4.3253222634913705e-06, + "loss": 1.854, + "step": 6226 + }, + { + "epoch": 1.36, + "learning_rate": 4.325213021629889e-06, + "loss": 3.3311, + "step": 6227 + }, + { + "epoch": 1.36, + "learning_rate": 4.325103779768408e-06, + "loss": 2.1773, + "step": 6228 + }, + { + "epoch": 1.36, + "learning_rate": 4.324994537906927e-06, + "loss": 3.7025, + "step": 6229 + }, + { + "epoch": 1.36, + "learning_rate": 4.324885296045445e-06, + "loss": 3.1767, + "step": 6230 + }, + { + "epoch": 1.36, + "learning_rate": 4.324776054183963e-06, + "loss": 4.0921, + "step": 6231 + }, + { + "epoch": 1.36, + "learning_rate": 4.324666812322482e-06, + "loss": 2.7227, + "step": 6232 + }, + { + "epoch": 1.36, + "learning_rate": 4.324557570461001e-06, + "loss": 2.7695, + "step": 6233 + }, + { + "epoch": 1.36, + "learning_rate": 4.32444832859952e-06, + "loss": 3.046, + "step": 6234 + }, + { + "epoch": 1.36, + "learning_rate": 4.324339086738038e-06, + "loss": 3.2802, + "step": 6235 + }, + { + "epoch": 1.36, + "learning_rate": 4.324229844876557e-06, + "loss": 3.1588, + "step": 6236 + }, + { + "epoch": 1.36, + "learning_rate": 4.324120603015076e-06, + "loss": 2.4353, + "step": 6237 + }, + { + "epoch": 1.36, + "learning_rate": 4.324011361153595e-06, + "loss": 2.4455, + "step": 6238 + }, + { + "epoch": 1.36, + "learning_rate": 4.323902119292113e-06, + "loss": 2.9552, + "step": 6239 + }, + { + "epoch": 1.36, + "learning_rate": 4.323792877430632e-06, + "loss": 3.1993, + "step": 6240 + }, + { + "epoch": 1.36, + "learning_rate": 4.32368363556915e-06, + "loss": 3.2245, + "step": 6241 + }, + { + "epoch": 1.36, + "learning_rate": 4.323574393707669e-06, + "loss": 2.9594, + "step": 6242 + }, + { + "epoch": 1.36, + "learning_rate": 4.323465151846188e-06, + "loss": 3.8139, + "step": 6243 + }, + { + "epoch": 1.36, + "learning_rate": 4.323355909984706e-06, + "loss": 3.1981, + "step": 6244 + }, + { + "epoch": 1.36, + "learning_rate": 4.3232466681232245e-06, + "loss": 3.45, + "step": 6245 + }, + { + "epoch": 1.36, + "learning_rate": 4.3231374262617435e-06, + "loss": 3.3179, + "step": 6246 + }, + { + "epoch": 1.36, + "learning_rate": 4.3230281844002626e-06, + "loss": 2.2812, + "step": 6247 + }, + { + "epoch": 1.36, + "learning_rate": 4.322918942538782e-06, + "loss": 3.2205, + "step": 6248 + }, + { + "epoch": 1.36, + "learning_rate": 4.3228097006773e-06, + "loss": 3.5506, + "step": 6249 + }, + { + "epoch": 1.36, + "learning_rate": 4.322700458815818e-06, + "loss": 2.4599, + "step": 6250 + }, + { + "epoch": 1.36, + "learning_rate": 4.322591216954337e-06, + "loss": 3.3121, + "step": 6251 + }, + { + "epoch": 1.36, + "learning_rate": 4.322481975092856e-06, + "loss": 2.7374, + "step": 6252 + }, + { + "epoch": 1.36, + "learning_rate": 4.322372733231375e-06, + "loss": 3.919, + "step": 6253 + }, + { + "epoch": 1.36, + "learning_rate": 4.322263491369893e-06, + "loss": 2.2394, + "step": 6254 + }, + { + "epoch": 1.37, + "learning_rate": 4.3221542495084115e-06, + "loss": 2.9125, + "step": 6255 + }, + { + "epoch": 1.37, + "learning_rate": 4.3220450076469305e-06, + "loss": 3.0053, + "step": 6256 + }, + { + "epoch": 1.37, + "learning_rate": 4.3219357657854495e-06, + "loss": 2.994, + "step": 6257 + }, + { + "epoch": 1.37, + "learning_rate": 4.3218265239239686e-06, + "loss": 2.167, + "step": 6258 + }, + { + "epoch": 1.37, + "learning_rate": 4.321717282062487e-06, + "loss": 2.5993, + "step": 6259 + }, + { + "epoch": 1.37, + "learning_rate": 4.321608040201005e-06, + "loss": 3.1075, + "step": 6260 + }, + { + "epoch": 1.37, + "learning_rate": 4.321498798339524e-06, + "loss": 2.9517, + "step": 6261 + }, + { + "epoch": 1.37, + "learning_rate": 4.321389556478043e-06, + "loss": 2.5801, + "step": 6262 + }, + { + "epoch": 1.37, + "learning_rate": 4.321280314616561e-06, + "loss": 2.9182, + "step": 6263 + }, + { + "epoch": 1.37, + "learning_rate": 4.32117107275508e-06, + "loss": 2.7335, + "step": 6264 + }, + { + "epoch": 1.37, + "learning_rate": 4.321061830893598e-06, + "loss": 3.178, + "step": 6265 + }, + { + "epoch": 1.37, + "learning_rate": 4.3209525890321174e-06, + "loss": 2.4619, + "step": 6266 + }, + { + "epoch": 1.37, + "learning_rate": 4.3208433471706365e-06, + "loss": 3.5492, + "step": 6267 + }, + { + "epoch": 1.37, + "learning_rate": 4.320734105309155e-06, + "loss": 2.6895, + "step": 6268 + }, + { + "epoch": 1.37, + "learning_rate": 4.320624863447674e-06, + "loss": 3.5994, + "step": 6269 + }, + { + "epoch": 1.37, + "learning_rate": 4.320515621586192e-06, + "loss": 3.5054, + "step": 6270 + }, + { + "epoch": 1.37, + "learning_rate": 4.320406379724711e-06, + "loss": 3.9325, + "step": 6271 + }, + { + "epoch": 1.37, + "learning_rate": 4.32029713786323e-06, + "loss": 1.9145, + "step": 6272 + }, + { + "epoch": 1.37, + "learning_rate": 4.320187896001748e-06, + "loss": 2.917, + "step": 6273 + }, + { + "epoch": 1.37, + "learning_rate": 4.320078654140266e-06, + "loss": 2.6956, + "step": 6274 + }, + { + "epoch": 1.37, + "learning_rate": 4.319969412278785e-06, + "loss": 2.638, + "step": 6275 + }, + { + "epoch": 1.37, + "learning_rate": 4.319860170417304e-06, + "loss": 3.7212, + "step": 6276 + }, + { + "epoch": 1.37, + "learning_rate": 4.3197509285558234e-06, + "loss": 3.1883, + "step": 6277 + }, + { + "epoch": 1.37, + "learning_rate": 4.319641686694342e-06, + "loss": 2.7322, + "step": 6278 + }, + { + "epoch": 1.37, + "learning_rate": 4.31953244483286e-06, + "loss": 3.0061, + "step": 6279 + }, + { + "epoch": 1.37, + "learning_rate": 4.319423202971379e-06, + "loss": 3.0462, + "step": 6280 + }, + { + "epoch": 1.37, + "learning_rate": 4.319313961109898e-06, + "loss": 3.5626, + "step": 6281 + }, + { + "epoch": 1.37, + "learning_rate": 4.319204719248417e-06, + "loss": 2.9281, + "step": 6282 + }, + { + "epoch": 1.37, + "learning_rate": 4.319095477386935e-06, + "loss": 2.586, + "step": 6283 + }, + { + "epoch": 1.37, + "learning_rate": 4.318986235525453e-06, + "loss": 2.8322, + "step": 6284 + }, + { + "epoch": 1.37, + "learning_rate": 4.318876993663972e-06, + "loss": 2.6346, + "step": 6285 + }, + { + "epoch": 1.37, + "learning_rate": 4.318767751802491e-06, + "loss": 2.782, + "step": 6286 + }, + { + "epoch": 1.37, + "learning_rate": 4.31865850994101e-06, + "loss": 3.1276, + "step": 6287 + }, + { + "epoch": 1.37, + "learning_rate": 4.318549268079529e-06, + "loss": 3.8827, + "step": 6288 + }, + { + "epoch": 1.37, + "learning_rate": 4.318440026218047e-06, + "loss": 1.7375, + "step": 6289 + }, + { + "epoch": 1.37, + "learning_rate": 4.318330784356566e-06, + "loss": 3.0158, + "step": 6290 + }, + { + "epoch": 1.37, + "learning_rate": 4.318221542495085e-06, + "loss": 3.1569, + "step": 6291 + }, + { + "epoch": 1.37, + "learning_rate": 4.318112300633603e-06, + "loss": 3.2902, + "step": 6292 + }, + { + "epoch": 1.37, + "learning_rate": 4.318003058772122e-06, + "loss": 2.5869, + "step": 6293 + }, + { + "epoch": 1.37, + "learning_rate": 4.31789381691064e-06, + "loss": 3.2674, + "step": 6294 + }, + { + "epoch": 1.37, + "learning_rate": 4.317784575049159e-06, + "loss": 3.1233, + "step": 6295 + }, + { + "epoch": 1.37, + "learning_rate": 4.317675333187678e-06, + "loss": 2.539, + "step": 6296 + }, + { + "epoch": 1.37, + "learning_rate": 4.3175660913261965e-06, + "loss": 2.7491, + "step": 6297 + }, + { + "epoch": 1.37, + "learning_rate": 4.317456849464715e-06, + "loss": 3.1969, + "step": 6298 + }, + { + "epoch": 1.37, + "learning_rate": 4.317347607603234e-06, + "loss": 2.8175, + "step": 6299 + }, + { + "epoch": 1.37, + "learning_rate": 4.317238365741753e-06, + "loss": 1.7576, + "step": 6300 + }, + { + "epoch": 1.38, + "learning_rate": 4.317129123880272e-06, + "loss": 2.4978, + "step": 6301 + }, + { + "epoch": 1.38, + "learning_rate": 4.31701988201879e-06, + "loss": 3.2063, + "step": 6302 + }, + { + "epoch": 1.38, + "learning_rate": 4.316910640157308e-06, + "loss": 3.0786, + "step": 6303 + }, + { + "epoch": 1.38, + "learning_rate": 4.316801398295827e-06, + "loss": 2.8611, + "step": 6304 + }, + { + "epoch": 1.38, + "learning_rate": 4.316692156434346e-06, + "loss": 2.9603, + "step": 6305 + }, + { + "epoch": 1.38, + "learning_rate": 4.316582914572865e-06, + "loss": 3.8408, + "step": 6306 + }, + { + "epoch": 1.38, + "learning_rate": 4.3164736727113835e-06, + "loss": 3.6697, + "step": 6307 + }, + { + "epoch": 1.38, + "learning_rate": 4.316364430849902e-06, + "loss": 2.6646, + "step": 6308 + }, + { + "epoch": 1.38, + "learning_rate": 4.316255188988421e-06, + "loss": 3.8705, + "step": 6309 + }, + { + "epoch": 1.38, + "learning_rate": 4.31614594712694e-06, + "loss": 3.3517, + "step": 6310 + }, + { + "epoch": 1.38, + "learning_rate": 4.316036705265458e-06, + "loss": 3.844, + "step": 6311 + }, + { + "epoch": 1.38, + "learning_rate": 4.315927463403977e-06, + "loss": 2.5731, + "step": 6312 + }, + { + "epoch": 1.38, + "learning_rate": 4.315818221542495e-06, + "loss": 2.3572, + "step": 6313 + }, + { + "epoch": 1.38, + "learning_rate": 4.315708979681014e-06, + "loss": 3.0488, + "step": 6314 + }, + { + "epoch": 1.38, + "learning_rate": 4.315599737819533e-06, + "loss": 2.8816, + "step": 6315 + }, + { + "epoch": 1.38, + "learning_rate": 4.315490495958051e-06, + "loss": 3.2809, + "step": 6316 + }, + { + "epoch": 1.38, + "learning_rate": 4.31538125409657e-06, + "loss": 2.8945, + "step": 6317 + }, + { + "epoch": 1.38, + "learning_rate": 4.315272012235089e-06, + "loss": 3.1836, + "step": 6318 + }, + { + "epoch": 1.38, + "learning_rate": 4.315162770373608e-06, + "loss": 2.2396, + "step": 6319 + }, + { + "epoch": 1.38, + "learning_rate": 4.315053528512127e-06, + "loss": 3.5666, + "step": 6320 + }, + { + "epoch": 1.38, + "learning_rate": 4.314944286650645e-06, + "loss": 3.1843, + "step": 6321 + }, + { + "epoch": 1.38, + "learning_rate": 4.314835044789164e-06, + "loss": 3.0321, + "step": 6322 + }, + { + "epoch": 1.38, + "learning_rate": 4.314725802927682e-06, + "loss": 2.3072, + "step": 6323 + }, + { + "epoch": 1.38, + "learning_rate": 4.314616561066201e-06, + "loss": 3.1707, + "step": 6324 + }, + { + "epoch": 1.38, + "learning_rate": 4.314507319204719e-06, + "loss": 3.4965, + "step": 6325 + }, + { + "epoch": 1.38, + "learning_rate": 4.314398077343238e-06, + "loss": 2.7946, + "step": 6326 + }, + { + "epoch": 1.38, + "learning_rate": 4.3142888354817565e-06, + "loss": 2.646, + "step": 6327 + }, + { + "epoch": 1.38, + "learning_rate": 4.3141795936202756e-06, + "loss": 2.7632, + "step": 6328 + }, + { + "epoch": 1.38, + "learning_rate": 4.314070351758795e-06, + "loss": 1.55, + "step": 6329 + }, + { + "epoch": 1.38, + "learning_rate": 4.313961109897313e-06, + "loss": 2.5824, + "step": 6330 + }, + { + "epoch": 1.38, + "learning_rate": 4.313851868035832e-06, + "loss": 2.8845, + "step": 6331 + }, + { + "epoch": 1.38, + "learning_rate": 4.31374262617435e-06, + "loss": 2.4041, + "step": 6332 + }, + { + "epoch": 1.38, + "learning_rate": 4.313633384312869e-06, + "loss": 3.127, + "step": 6333 + }, + { + "epoch": 1.38, + "learning_rate": 4.313524142451388e-06, + "loss": 3.7011, + "step": 6334 + }, + { + "epoch": 1.38, + "learning_rate": 4.313414900589906e-06, + "loss": 2.9003, + "step": 6335 + }, + { + "epoch": 1.38, + "learning_rate": 4.313305658728425e-06, + "loss": 3.2507, + "step": 6336 + }, + { + "epoch": 1.38, + "learning_rate": 4.3131964168669435e-06, + "loss": 1.7904, + "step": 6337 + }, + { + "epoch": 1.38, + "learning_rate": 4.3130871750054625e-06, + "loss": 3.4803, + "step": 6338 + }, + { + "epoch": 1.38, + "learning_rate": 4.3129779331439816e-06, + "loss": 3.0304, + "step": 6339 + }, + { + "epoch": 1.38, + "learning_rate": 4.3128686912825e-06, + "loss": 1.716, + "step": 6340 + }, + { + "epoch": 1.38, + "learning_rate": 4.312759449421019e-06, + "loss": 2.461, + "step": 6341 + }, + { + "epoch": 1.38, + "learning_rate": 4.312650207559537e-06, + "loss": 3.3462, + "step": 6342 + }, + { + "epoch": 1.38, + "learning_rate": 4.312540965698056e-06, + "loss": 2.4194, + "step": 6343 + }, + { + "epoch": 1.38, + "learning_rate": 4.312431723836574e-06, + "loss": 2.7141, + "step": 6344 + }, + { + "epoch": 1.38, + "learning_rate": 4.312322481975093e-06, + "loss": 3.1132, + "step": 6345 + }, + { + "epoch": 1.38, + "learning_rate": 4.312213240113612e-06, + "loss": 3.6144, + "step": 6346 + }, + { + "epoch": 1.39, + "learning_rate": 4.3121039982521304e-06, + "loss": 3.085, + "step": 6347 + }, + { + "epoch": 1.39, + "learning_rate": 4.3119947563906495e-06, + "loss": 3.1164, + "step": 6348 + }, + { + "epoch": 1.39, + "learning_rate": 4.311885514529168e-06, + "loss": 2.9317, + "step": 6349 + }, + { + "epoch": 1.39, + "learning_rate": 4.311776272667687e-06, + "loss": 2.8499, + "step": 6350 + }, + { + "epoch": 1.39, + "learning_rate": 4.311667030806205e-06, + "loss": 3.3001, + "step": 6351 + }, + { + "epoch": 1.39, + "learning_rate": 4.311557788944724e-06, + "loss": 3.2547, + "step": 6352 + }, + { + "epoch": 1.39, + "learning_rate": 4.311448547083243e-06, + "loss": 2.1457, + "step": 6353 + }, + { + "epoch": 1.39, + "learning_rate": 4.311339305221761e-06, + "loss": 3.3788, + "step": 6354 + }, + { + "epoch": 1.39, + "learning_rate": 4.31123006336028e-06, + "loss": 3.1435, + "step": 6355 + }, + { + "epoch": 1.39, + "learning_rate": 4.311120821498798e-06, + "loss": 3.3134, + "step": 6356 + }, + { + "epoch": 1.39, + "learning_rate": 4.311011579637317e-06, + "loss": 2.0103, + "step": 6357 + }, + { + "epoch": 1.39, + "learning_rate": 4.310902337775836e-06, + "loss": 3.4407, + "step": 6358 + }, + { + "epoch": 1.39, + "learning_rate": 4.310793095914355e-06, + "loss": 2.9195, + "step": 6359 + }, + { + "epoch": 1.39, + "learning_rate": 4.310683854052874e-06, + "loss": 2.0412, + "step": 6360 + }, + { + "epoch": 1.39, + "learning_rate": 4.310574612191392e-06, + "loss": 2.6281, + "step": 6361 + }, + { + "epoch": 1.39, + "learning_rate": 4.310465370329911e-06, + "loss": 3.333, + "step": 6362 + }, + { + "epoch": 1.39, + "learning_rate": 4.310356128468429e-06, + "loss": 3.1502, + "step": 6363 + }, + { + "epoch": 1.39, + "learning_rate": 4.310246886606948e-06, + "loss": 2.4167, + "step": 6364 + }, + { + "epoch": 1.39, + "learning_rate": 4.310137644745467e-06, + "loss": 2.7167, + "step": 6365 + }, + { + "epoch": 1.39, + "learning_rate": 4.310028402883985e-06, + "loss": 3.0196, + "step": 6366 + }, + { + "epoch": 1.39, + "learning_rate": 4.309919161022504e-06, + "loss": 3.4063, + "step": 6367 + }, + { + "epoch": 1.39, + "learning_rate": 4.3098099191610225e-06, + "loss": 2.3662, + "step": 6368 + }, + { + "epoch": 1.39, + "learning_rate": 4.309700677299542e-06, + "loss": 3.2032, + "step": 6369 + }, + { + "epoch": 1.39, + "learning_rate": 4.309591435438061e-06, + "loss": 3.5041, + "step": 6370 + }, + { + "epoch": 1.39, + "learning_rate": 4.309482193576579e-06, + "loss": 3.1366, + "step": 6371 + }, + { + "epoch": 1.39, + "learning_rate": 4.309372951715098e-06, + "loss": 3.5554, + "step": 6372 + }, + { + "epoch": 1.39, + "learning_rate": 4.309263709853616e-06, + "loss": 2.8094, + "step": 6373 + }, + { + "epoch": 1.39, + "learning_rate": 4.309154467992135e-06, + "loss": 2.6467, + "step": 6374 + }, + { + "epoch": 1.39, + "learning_rate": 4.309045226130654e-06, + "loss": 2.0874, + "step": 6375 + }, + { + "epoch": 1.39, + "learning_rate": 4.308935984269172e-06, + "loss": 3.7792, + "step": 6376 + }, + { + "epoch": 1.39, + "learning_rate": 4.3088267424076905e-06, + "loss": 2.6237, + "step": 6377 + }, + { + "epoch": 1.39, + "learning_rate": 4.3087175005462095e-06, + "loss": 3.1037, + "step": 6378 + }, + { + "epoch": 1.39, + "learning_rate": 4.3086082586847285e-06, + "loss": 2.7127, + "step": 6379 + }, + { + "epoch": 1.39, + "learning_rate": 4.308499016823247e-06, + "loss": 3.2479, + "step": 6380 + }, + { + "epoch": 1.39, + "learning_rate": 4.308389774961766e-06, + "loss": 2.9015, + "step": 6381 + }, + { + "epoch": 1.39, + "learning_rate": 4.308280533100284e-06, + "loss": 3.3902, + "step": 6382 + }, + { + "epoch": 1.39, + "learning_rate": 4.308171291238803e-06, + "loss": 3.5631, + "step": 6383 + }, + { + "epoch": 1.39, + "learning_rate": 4.308062049377322e-06, + "loss": 3.1976, + "step": 6384 + }, + { + "epoch": 1.39, + "learning_rate": 4.30795280751584e-06, + "loss": 2.7744, + "step": 6385 + }, + { + "epoch": 1.39, + "learning_rate": 4.307843565654359e-06, + "loss": 2.2759, + "step": 6386 + }, + { + "epoch": 1.39, + "learning_rate": 4.3077343237928774e-06, + "loss": 3.0094, + "step": 6387 + }, + { + "epoch": 1.39, + "learning_rate": 4.3076250819313965e-06, + "loss": 3.4423, + "step": 6388 + }, + { + "epoch": 1.39, + "learning_rate": 4.3075158400699155e-06, + "loss": 2.9985, + "step": 6389 + }, + { + "epoch": 1.39, + "learning_rate": 4.307406598208434e-06, + "loss": 2.8228, + "step": 6390 + }, + { + "epoch": 1.39, + "learning_rate": 4.307297356346952e-06, + "loss": 1.8637, + "step": 6391 + }, + { + "epoch": 1.4, + "learning_rate": 4.307188114485471e-06, + "loss": 2.8321, + "step": 6392 + }, + { + "epoch": 1.4, + "learning_rate": 4.30707887262399e-06, + "loss": 3.0081, + "step": 6393 + }, + { + "epoch": 1.4, + "learning_rate": 4.306969630762509e-06, + "loss": 2.5459, + "step": 6394 + }, + { + "epoch": 1.4, + "learning_rate": 4.306860388901027e-06, + "loss": 3.3511, + "step": 6395 + }, + { + "epoch": 1.4, + "learning_rate": 4.306751147039545e-06, + "loss": 2.9244, + "step": 6396 + }, + { + "epoch": 1.4, + "learning_rate": 4.306641905178064e-06, + "loss": 3.3623, + "step": 6397 + }, + { + "epoch": 1.4, + "learning_rate": 4.306532663316583e-06, + "loss": 2.5091, + "step": 6398 + }, + { + "epoch": 1.4, + "learning_rate": 4.3064234214551025e-06, + "loss": 2.9944, + "step": 6399 + }, + { + "epoch": 1.4, + "learning_rate": 4.306314179593621e-06, + "loss": 2.7178, + "step": 6400 + }, + { + "epoch": 1.4, + "learning_rate": 4.306204937732139e-06, + "loss": 2.9748, + "step": 6401 + }, + { + "epoch": 1.4, + "learning_rate": 4.306095695870658e-06, + "loss": 3.1642, + "step": 6402 + }, + { + "epoch": 1.4, + "learning_rate": 4.305986454009177e-06, + "loss": 3.247, + "step": 6403 + }, + { + "epoch": 1.4, + "learning_rate": 4.305877212147695e-06, + "loss": 2.9131, + "step": 6404 + }, + { + "epoch": 1.4, + "learning_rate": 4.305767970286214e-06, + "loss": 2.9632, + "step": 6405 + }, + { + "epoch": 1.4, + "learning_rate": 4.305658728424732e-06, + "loss": 2.6468, + "step": 6406 + }, + { + "epoch": 1.4, + "learning_rate": 4.305549486563251e-06, + "loss": 2.4913, + "step": 6407 + }, + { + "epoch": 1.4, + "learning_rate": 4.30544024470177e-06, + "loss": 2.5384, + "step": 6408 + }, + { + "epoch": 1.4, + "learning_rate": 4.3053310028402886e-06, + "loss": 3.8535, + "step": 6409 + }, + { + "epoch": 1.4, + "learning_rate": 4.305221760978808e-06, + "loss": 2.6433, + "step": 6410 + }, + { + "epoch": 1.4, + "learning_rate": 4.305112519117326e-06, + "loss": 3.6905, + "step": 6411 + }, + { + "epoch": 1.4, + "learning_rate": 4.305003277255845e-06, + "loss": 3.7862, + "step": 6412 + }, + { + "epoch": 1.4, + "learning_rate": 4.304894035394364e-06, + "loss": 2.7371, + "step": 6413 + }, + { + "epoch": 1.4, + "learning_rate": 4.304784793532882e-06, + "loss": 3.1624, + "step": 6414 + }, + { + "epoch": 1.4, + "learning_rate": 4.3046755516714e-06, + "loss": 3.6837, + "step": 6415 + }, + { + "epoch": 1.4, + "learning_rate": 4.304566309809919e-06, + "loss": 2.688, + "step": 6416 + }, + { + "epoch": 1.4, + "learning_rate": 4.304457067948438e-06, + "loss": 2.8104, + "step": 6417 + }, + { + "epoch": 1.4, + "learning_rate": 4.304347826086957e-06, + "loss": 3.5803, + "step": 6418 + }, + { + "epoch": 1.4, + "learning_rate": 4.3042385842254755e-06, + "loss": 2.3115, + "step": 6419 + }, + { + "epoch": 1.4, + "learning_rate": 4.304129342363994e-06, + "loss": 2.7722, + "step": 6420 + }, + { + "epoch": 1.4, + "learning_rate": 4.304020100502513e-06, + "loss": 1.9265, + "step": 6421 + }, + { + "epoch": 1.4, + "learning_rate": 4.303910858641032e-06, + "loss": 3.0661, + "step": 6422 + }, + { + "epoch": 1.4, + "learning_rate": 4.303801616779551e-06, + "loss": 2.7491, + "step": 6423 + }, + { + "epoch": 1.4, + "learning_rate": 4.303692374918069e-06, + "loss": 2.8478, + "step": 6424 + }, + { + "epoch": 1.4, + "learning_rate": 4.303583133056587e-06, + "loss": 2.9355, + "step": 6425 + }, + { + "epoch": 1.4, + "learning_rate": 4.303473891195106e-06, + "loss": 3.307, + "step": 6426 + }, + { + "epoch": 1.4, + "learning_rate": 4.303364649333625e-06, + "loss": 2.5267, + "step": 6427 + }, + { + "epoch": 1.4, + "learning_rate": 4.303255407472144e-06, + "loss": 2.4056, + "step": 6428 + }, + { + "epoch": 1.4, + "learning_rate": 4.3031461656106625e-06, + "loss": 3.2213, + "step": 6429 + }, + { + "epoch": 1.4, + "learning_rate": 4.303036923749181e-06, + "loss": 2.8216, + "step": 6430 + }, + { + "epoch": 1.4, + "learning_rate": 4.3029276818877e-06, + "loss": 3.1357, + "step": 6431 + }, + { + "epoch": 1.4, + "learning_rate": 4.302818440026219e-06, + "loss": 3.1169, + "step": 6432 + }, + { + "epoch": 1.4, + "learning_rate": 4.302709198164737e-06, + "loss": 3.8539, + "step": 6433 + }, + { + "epoch": 1.4, + "learning_rate": 4.302599956303256e-06, + "loss": 2.3361, + "step": 6434 + }, + { + "epoch": 1.4, + "learning_rate": 4.302490714441774e-06, + "loss": 2.9159, + "step": 6435 + }, + { + "epoch": 1.4, + "learning_rate": 4.302381472580293e-06, + "loss": 2.2611, + "step": 6436 + }, + { + "epoch": 1.4, + "learning_rate": 4.302272230718812e-06, + "loss": 3.7955, + "step": 6437 + }, + { + "epoch": 1.41, + "learning_rate": 4.30216298885733e-06, + "loss": 2.9036, + "step": 6438 + }, + { + "epoch": 1.41, + "learning_rate": 4.302053746995849e-06, + "loss": 3.0002, + "step": 6439 + }, + { + "epoch": 1.41, + "learning_rate": 4.301944505134368e-06, + "loss": 2.5604, + "step": 6440 + }, + { + "epoch": 1.41, + "learning_rate": 4.301835263272887e-06, + "loss": 3.9844, + "step": 6441 + }, + { + "epoch": 1.41, + "learning_rate": 4.301726021411406e-06, + "loss": 3.234, + "step": 6442 + }, + { + "epoch": 1.41, + "learning_rate": 4.301616779549924e-06, + "loss": 3.234, + "step": 6443 + }, + { + "epoch": 1.41, + "learning_rate": 4.301507537688442e-06, + "loss": 2.5121, + "step": 6444 + }, + { + "epoch": 1.41, + "learning_rate": 4.301398295826961e-06, + "loss": 2.7035, + "step": 6445 + }, + { + "epoch": 1.41, + "learning_rate": 4.30128905396548e-06, + "loss": 3.6303, + "step": 6446 + }, + { + "epoch": 1.41, + "learning_rate": 4.301179812103999e-06, + "loss": 3.2964, + "step": 6447 + }, + { + "epoch": 1.41, + "learning_rate": 4.301070570242517e-06, + "loss": 2.7962, + "step": 6448 + }, + { + "epoch": 1.41, + "learning_rate": 4.3009613283810355e-06, + "loss": 3.1651, + "step": 6449 + }, + { + "epoch": 1.41, + "learning_rate": 4.3008520865195546e-06, + "loss": 2.0171, + "step": 6450 + }, + { + "epoch": 1.41, + "learning_rate": 4.300742844658074e-06, + "loss": 3.7882, + "step": 6451 + }, + { + "epoch": 1.41, + "learning_rate": 4.300633602796593e-06, + "loss": 2.7088, + "step": 6452 + }, + { + "epoch": 1.41, + "learning_rate": 4.300524360935111e-06, + "loss": 3.0848, + "step": 6453 + }, + { + "epoch": 1.41, + "learning_rate": 4.300415119073629e-06, + "loss": 3.0196, + "step": 6454 + }, + { + "epoch": 1.41, + "learning_rate": 4.300305877212148e-06, + "loss": 2.4594, + "step": 6455 + }, + { + "epoch": 1.41, + "learning_rate": 4.300196635350667e-06, + "loss": 2.4789, + "step": 6456 + }, + { + "epoch": 1.41, + "learning_rate": 4.300087393489185e-06, + "loss": 3.6748, + "step": 6457 + }, + { + "epoch": 1.41, + "learning_rate": 4.299978151627704e-06, + "loss": 2.9091, + "step": 6458 + }, + { + "epoch": 1.41, + "learning_rate": 4.2998689097662225e-06, + "loss": 2.8179, + "step": 6459 + }, + { + "epoch": 1.41, + "learning_rate": 4.2997596679047415e-06, + "loss": 3.328, + "step": 6460 + }, + { + "epoch": 1.41, + "learning_rate": 4.2996504260432606e-06, + "loss": 1.5491, + "step": 6461 + }, + { + "epoch": 1.41, + "learning_rate": 4.299541184181779e-06, + "loss": 2.6831, + "step": 6462 + }, + { + "epoch": 1.41, + "learning_rate": 4.299431942320298e-06, + "loss": 2.8342, + "step": 6463 + }, + { + "epoch": 1.41, + "learning_rate": 4.299322700458816e-06, + "loss": 2.4076, + "step": 6464 + }, + { + "epoch": 1.41, + "learning_rate": 4.299213458597335e-06, + "loss": 3.2304, + "step": 6465 + }, + { + "epoch": 1.41, + "learning_rate": 4.299104216735854e-06, + "loss": 3.9279, + "step": 6466 + }, + { + "epoch": 1.41, + "learning_rate": 4.298994974874372e-06, + "loss": 2.8209, + "step": 6467 + }, + { + "epoch": 1.41, + "learning_rate": 4.29888573301289e-06, + "loss": 2.9359, + "step": 6468 + }, + { + "epoch": 1.41, + "learning_rate": 4.2987764911514095e-06, + "loss": 1.9121, + "step": 6469 + }, + { + "epoch": 1.41, + "learning_rate": 4.2986672492899285e-06, + "loss": 3.994, + "step": 6470 + }, + { + "epoch": 1.41, + "learning_rate": 4.2985580074284475e-06, + "loss": 2.0126, + "step": 6471 + }, + { + "epoch": 1.41, + "learning_rate": 4.298448765566966e-06, + "loss": 2.7092, + "step": 6472 + }, + { + "epoch": 1.41, + "learning_rate": 4.298339523705484e-06, + "loss": 3.0055, + "step": 6473 + }, + { + "epoch": 1.41, + "learning_rate": 4.298230281844003e-06, + "loss": 2.5914, + "step": 6474 + }, + { + "epoch": 1.41, + "learning_rate": 4.298121039982522e-06, + "loss": 2.3591, + "step": 6475 + }, + { + "epoch": 1.41, + "learning_rate": 4.29801179812104e-06, + "loss": 3.4554, + "step": 6476 + }, + { + "epoch": 1.41, + "learning_rate": 4.297902556259559e-06, + "loss": 2.6663, + "step": 6477 + }, + { + "epoch": 1.41, + "learning_rate": 4.297793314398077e-06, + "loss": 3.5905, + "step": 6478 + }, + { + "epoch": 1.41, + "learning_rate": 4.297684072536596e-06, + "loss": 1.728, + "step": 6479 + }, + { + "epoch": 1.41, + "learning_rate": 4.2975748306751154e-06, + "loss": 2.732, + "step": 6480 + }, + { + "epoch": 1.41, + "learning_rate": 4.297465588813634e-06, + "loss": 3.556, + "step": 6481 + }, + { + "epoch": 1.41, + "learning_rate": 4.297356346952153e-06, + "loss": 1.4873, + "step": 6482 + }, + { + "epoch": 1.41, + "learning_rate": 4.297247105090671e-06, + "loss": 3.7988, + "step": 6483 + }, + { + "epoch": 1.42, + "learning_rate": 4.29713786322919e-06, + "loss": 2.9431, + "step": 6484 + }, + { + "epoch": 1.42, + "learning_rate": 4.297028621367709e-06, + "loss": 2.4507, + "step": 6485 + }, + { + "epoch": 1.42, + "learning_rate": 4.296919379506227e-06, + "loss": 2.6106, + "step": 6486 + }, + { + "epoch": 1.42, + "learning_rate": 4.296810137644746e-06, + "loss": 2.9304, + "step": 6487 + }, + { + "epoch": 1.42, + "learning_rate": 4.296700895783264e-06, + "loss": 2.9706, + "step": 6488 + }, + { + "epoch": 1.42, + "learning_rate": 4.296591653921783e-06, + "loss": 3.8669, + "step": 6489 + }, + { + "epoch": 1.42, + "learning_rate": 4.2964824120603016e-06, + "loss": 2.5566, + "step": 6490 + }, + { + "epoch": 1.42, + "learning_rate": 4.296373170198821e-06, + "loss": 2.3233, + "step": 6491 + }, + { + "epoch": 1.42, + "learning_rate": 4.296263928337339e-06, + "loss": 3.0267, + "step": 6492 + }, + { + "epoch": 1.42, + "learning_rate": 4.296154686475858e-06, + "loss": 2.6801, + "step": 6493 + }, + { + "epoch": 1.42, + "learning_rate": 4.296045444614377e-06, + "loss": 3.4883, + "step": 6494 + }, + { + "epoch": 1.42, + "learning_rate": 4.295936202752895e-06, + "loss": 3.9469, + "step": 6495 + }, + { + "epoch": 1.42, + "learning_rate": 4.295826960891414e-06, + "loss": 2.7023, + "step": 6496 + }, + { + "epoch": 1.42, + "learning_rate": 4.295717719029932e-06, + "loss": 3.4918, + "step": 6497 + }, + { + "epoch": 1.42, + "learning_rate": 4.295608477168451e-06, + "loss": 3.4544, + "step": 6498 + }, + { + "epoch": 1.42, + "learning_rate": 4.29549923530697e-06, + "loss": 3.5998, + "step": 6499 + }, + { + "epoch": 1.42, + "learning_rate": 4.2953899934454885e-06, + "loss": 3.0459, + "step": 6500 + }, + { + "epoch": 1.42, + "learning_rate": 4.2952807515840075e-06, + "loss": 2.3667, + "step": 6501 + }, + { + "epoch": 1.42, + "learning_rate": 4.295171509722526e-06, + "loss": 3.6148, + "step": 6502 + }, + { + "epoch": 1.42, + "learning_rate": 4.295062267861045e-06, + "loss": 3.0101, + "step": 6503 + }, + { + "epoch": 1.42, + "learning_rate": 4.294953025999564e-06, + "loss": 2.6789, + "step": 6504 + }, + { + "epoch": 1.42, + "learning_rate": 4.294843784138082e-06, + "loss": 2.1656, + "step": 6505 + }, + { + "epoch": 1.42, + "learning_rate": 4.294734542276601e-06, + "loss": 3.4308, + "step": 6506 + }, + { + "epoch": 1.42, + "learning_rate": 4.294625300415119e-06, + "loss": 2.7276, + "step": 6507 + }, + { + "epoch": 1.42, + "learning_rate": 4.294516058553638e-06, + "loss": 3.2111, + "step": 6508 + }, + { + "epoch": 1.42, + "learning_rate": 4.2944068166921564e-06, + "loss": 2.3119, + "step": 6509 + }, + { + "epoch": 1.42, + "learning_rate": 4.2942975748306755e-06, + "loss": 2.0825, + "step": 6510 + }, + { + "epoch": 1.42, + "learning_rate": 4.2941883329691945e-06, + "loss": 3.0272, + "step": 6511 + }, + { + "epoch": 1.42, + "learning_rate": 4.294079091107713e-06, + "loss": 3.5088, + "step": 6512 + }, + { + "epoch": 1.42, + "learning_rate": 4.293969849246232e-06, + "loss": 3.0401, + "step": 6513 + }, + { + "epoch": 1.42, + "learning_rate": 4.29386060738475e-06, + "loss": 3.2326, + "step": 6514 + }, + { + "epoch": 1.42, + "learning_rate": 4.293751365523269e-06, + "loss": 1.786, + "step": 6515 + }, + { + "epoch": 1.42, + "learning_rate": 4.293642123661788e-06, + "loss": 3.0049, + "step": 6516 + }, + { + "epoch": 1.42, + "learning_rate": 4.293532881800306e-06, + "loss": 2.8321, + "step": 6517 + }, + { + "epoch": 1.42, + "learning_rate": 4.293423639938825e-06, + "loss": 3.2956, + "step": 6518 + }, + { + "epoch": 1.42, + "learning_rate": 4.293314398077343e-06, + "loss": 2.5164, + "step": 6519 + }, + { + "epoch": 1.42, + "learning_rate": 4.2932051562158624e-06, + "loss": 3.3841, + "step": 6520 + }, + { + "epoch": 1.42, + "learning_rate": 4.293095914354381e-06, + "loss": 2.7744, + "step": 6521 + }, + { + "epoch": 1.42, + "learning_rate": 4.2929866724929e-06, + "loss": 2.8541, + "step": 6522 + }, + { + "epoch": 1.42, + "learning_rate": 4.292877430631418e-06, + "loss": 2.856, + "step": 6523 + }, + { + "epoch": 1.42, + "learning_rate": 4.292768188769937e-06, + "loss": 3.3583, + "step": 6524 + }, + { + "epoch": 1.42, + "learning_rate": 4.292658946908456e-06, + "loss": 2.7181, + "step": 6525 + }, + { + "epoch": 1.42, + "learning_rate": 4.292549705046974e-06, + "loss": 3.3504, + "step": 6526 + }, + { + "epoch": 1.42, + "learning_rate": 4.292440463185493e-06, + "loss": 2.6974, + "step": 6527 + }, + { + "epoch": 1.42, + "learning_rate": 4.292331221324011e-06, + "loss": 2.9931, + "step": 6528 + }, + { + "epoch": 1.42, + "learning_rate": 4.29222197946253e-06, + "loss": 2.7015, + "step": 6529 + }, + { + "epoch": 1.43, + "learning_rate": 4.292112737601049e-06, + "loss": 3.1159, + "step": 6530 + }, + { + "epoch": 1.43, + "learning_rate": 4.2920034957395676e-06, + "loss": 2.961, + "step": 6531 + }, + { + "epoch": 1.43, + "learning_rate": 4.291894253878087e-06, + "loss": 2.8092, + "step": 6532 + }, + { + "epoch": 1.43, + "learning_rate": 4.291785012016605e-06, + "loss": 3.3584, + "step": 6533 + }, + { + "epoch": 1.43, + "learning_rate": 4.291675770155124e-06, + "loss": 2.49, + "step": 6534 + }, + { + "epoch": 1.43, + "learning_rate": 4.291566528293643e-06, + "loss": 2.5887, + "step": 6535 + }, + { + "epoch": 1.43, + "learning_rate": 4.291457286432161e-06, + "loss": 3.4562, + "step": 6536 + }, + { + "epoch": 1.43, + "learning_rate": 4.29134804457068e-06, + "loss": 3.5249, + "step": 6537 + }, + { + "epoch": 1.43, + "learning_rate": 4.291238802709198e-06, + "loss": 3.0736, + "step": 6538 + }, + { + "epoch": 1.43, + "learning_rate": 4.291129560847717e-06, + "loss": 3.7549, + "step": 6539 + }, + { + "epoch": 1.43, + "learning_rate": 4.291020318986236e-06, + "loss": 2.2492, + "step": 6540 + }, + { + "epoch": 1.43, + "learning_rate": 4.2909110771247545e-06, + "loss": 3.3422, + "step": 6541 + }, + { + "epoch": 1.43, + "learning_rate": 4.290801835263273e-06, + "loss": 2.6849, + "step": 6542 + }, + { + "epoch": 1.43, + "learning_rate": 4.290692593401792e-06, + "loss": 1.5167, + "step": 6543 + }, + { + "epoch": 1.43, + "learning_rate": 4.290583351540311e-06, + "loss": 3.4472, + "step": 6544 + }, + { + "epoch": 1.43, + "learning_rate": 4.290474109678829e-06, + "loss": 3.5669, + "step": 6545 + }, + { + "epoch": 1.43, + "learning_rate": 4.290364867817348e-06, + "loss": 2.7461, + "step": 6546 + }, + { + "epoch": 1.43, + "learning_rate": 4.290255625955866e-06, + "loss": 3.8044, + "step": 6547 + }, + { + "epoch": 1.43, + "learning_rate": 4.290146384094385e-06, + "loss": 2.6532, + "step": 6548 + }, + { + "epoch": 1.43, + "learning_rate": 4.290037142232904e-06, + "loss": 3.0531, + "step": 6549 + }, + { + "epoch": 1.43, + "learning_rate": 4.2899279003714225e-06, + "loss": 2.5322, + "step": 6550 + }, + { + "epoch": 1.43, + "learning_rate": 4.2898186585099415e-06, + "loss": 1.9754, + "step": 6551 + }, + { + "epoch": 1.43, + "learning_rate": 4.28970941664846e-06, + "loss": 2.325, + "step": 6552 + }, + { + "epoch": 1.43, + "learning_rate": 4.289600174786979e-06, + "loss": 3.0073, + "step": 6553 + }, + { + "epoch": 1.43, + "learning_rate": 4.289490932925498e-06, + "loss": 3.1126, + "step": 6554 + }, + { + "epoch": 1.43, + "learning_rate": 4.289381691064016e-06, + "loss": 2.9957, + "step": 6555 + }, + { + "epoch": 1.43, + "learning_rate": 4.289272449202534e-06, + "loss": 3.281, + "step": 6556 + }, + { + "epoch": 1.43, + "learning_rate": 4.289163207341053e-06, + "loss": 3.7856, + "step": 6557 + }, + { + "epoch": 1.43, + "learning_rate": 4.289053965479572e-06, + "loss": 3.2769, + "step": 6558 + }, + { + "epoch": 1.43, + "learning_rate": 4.288944723618091e-06, + "loss": 2.8867, + "step": 6559 + }, + { + "epoch": 1.43, + "learning_rate": 4.288835481756609e-06, + "loss": 3.0307, + "step": 6560 + }, + { + "epoch": 1.43, + "learning_rate": 4.288726239895128e-06, + "loss": 3.5846, + "step": 6561 + }, + { + "epoch": 1.43, + "learning_rate": 4.288616998033647e-06, + "loss": 3.2618, + "step": 6562 + }, + { + "epoch": 1.43, + "learning_rate": 4.288507756172166e-06, + "loss": 2.6212, + "step": 6563 + }, + { + "epoch": 1.43, + "learning_rate": 4.288398514310685e-06, + "loss": 1.993, + "step": 6564 + }, + { + "epoch": 1.43, + "learning_rate": 4.288289272449203e-06, + "loss": 3.5102, + "step": 6565 + }, + { + "epoch": 1.43, + "learning_rate": 4.288180030587721e-06, + "loss": 3.0038, + "step": 6566 + }, + { + "epoch": 1.43, + "learning_rate": 4.28807078872624e-06, + "loss": 2.5118, + "step": 6567 + }, + { + "epoch": 1.43, + "learning_rate": 4.287961546864759e-06, + "loss": 2.8396, + "step": 6568 + }, + { + "epoch": 1.43, + "learning_rate": 4.287852305003278e-06, + "loss": 3.2042, + "step": 6569 + }, + { + "epoch": 1.43, + "learning_rate": 4.287743063141796e-06, + "loss": 2.0855, + "step": 6570 + }, + { + "epoch": 1.43, + "learning_rate": 4.2876338212803146e-06, + "loss": 3.2045, + "step": 6571 + }, + { + "epoch": 1.43, + "learning_rate": 4.287524579418834e-06, + "loss": 2.9965, + "step": 6572 + }, + { + "epoch": 1.43, + "learning_rate": 4.287415337557353e-06, + "loss": 2.8431, + "step": 6573 + }, + { + "epoch": 1.43, + "learning_rate": 4.287306095695871e-06, + "loss": 3.1271, + "step": 6574 + }, + { + "epoch": 1.43, + "learning_rate": 4.28719685383439e-06, + "loss": 3.7719, + "step": 6575 + }, + { + "epoch": 1.44, + "learning_rate": 4.287087611972908e-06, + "loss": 1.7096, + "step": 6576 + }, + { + "epoch": 1.44, + "learning_rate": 4.286978370111427e-06, + "loss": 3.3355, + "step": 6577 + }, + { + "epoch": 1.44, + "learning_rate": 4.286869128249946e-06, + "loss": 2.2136, + "step": 6578 + }, + { + "epoch": 1.44, + "learning_rate": 4.286759886388464e-06, + "loss": 3.1584, + "step": 6579 + }, + { + "epoch": 1.44, + "learning_rate": 4.2866506445269825e-06, + "loss": 2.5455, + "step": 6580 + }, + { + "epoch": 1.44, + "learning_rate": 4.2865414026655015e-06, + "loss": 3.218, + "step": 6581 + }, + { + "epoch": 1.44, + "learning_rate": 4.2864321608040205e-06, + "loss": 2.8689, + "step": 6582 + }, + { + "epoch": 1.44, + "learning_rate": 4.28632291894254e-06, + "loss": 2.8555, + "step": 6583 + }, + { + "epoch": 1.44, + "learning_rate": 4.286213677081058e-06, + "loss": 2.7338, + "step": 6584 + }, + { + "epoch": 1.44, + "learning_rate": 4.286104435219576e-06, + "loss": 3.4125, + "step": 6585 + }, + { + "epoch": 1.44, + "learning_rate": 4.285995193358095e-06, + "loss": 3.106, + "step": 6586 + }, + { + "epoch": 1.44, + "learning_rate": 4.285885951496614e-06, + "loss": 3.2804, + "step": 6587 + }, + { + "epoch": 1.44, + "learning_rate": 4.285776709635133e-06, + "loss": 2.1909, + "step": 6588 + }, + { + "epoch": 1.44, + "learning_rate": 4.285667467773651e-06, + "loss": 2.203, + "step": 6589 + }, + { + "epoch": 1.44, + "learning_rate": 4.2855582259121694e-06, + "loss": 2.6418, + "step": 6590 + }, + { + "epoch": 1.44, + "learning_rate": 4.2854489840506885e-06, + "loss": 3.3403, + "step": 6591 + }, + { + "epoch": 1.44, + "learning_rate": 4.2853397421892075e-06, + "loss": 3.4652, + "step": 6592 + }, + { + "epoch": 1.44, + "learning_rate": 4.2852305003277265e-06, + "loss": 2.6897, + "step": 6593 + }, + { + "epoch": 1.44, + "learning_rate": 4.285121258466245e-06, + "loss": 2.5025, + "step": 6594 + }, + { + "epoch": 1.44, + "learning_rate": 4.285012016604763e-06, + "loss": 2.6473, + "step": 6595 + }, + { + "epoch": 1.44, + "learning_rate": 4.284902774743282e-06, + "loss": 3.0338, + "step": 6596 + }, + { + "epoch": 1.44, + "learning_rate": 4.284793532881801e-06, + "loss": 3.3219, + "step": 6597 + }, + { + "epoch": 1.44, + "learning_rate": 4.284684291020319e-06, + "loss": 3.1976, + "step": 6598 + }, + { + "epoch": 1.44, + "learning_rate": 4.284575049158838e-06, + "loss": 1.634, + "step": 6599 + }, + { + "epoch": 1.44, + "learning_rate": 4.284465807297356e-06, + "loss": 2.5559, + "step": 6600 + }, + { + "epoch": 1.44, + "learning_rate": 4.284356565435875e-06, + "loss": 3.0281, + "step": 6601 + }, + { + "epoch": 1.44, + "learning_rate": 4.2842473235743945e-06, + "loss": 2.3547, + "step": 6602 + }, + { + "epoch": 1.44, + "learning_rate": 4.284138081712913e-06, + "loss": 3.5554, + "step": 6603 + }, + { + "epoch": 1.44, + "learning_rate": 4.284028839851431e-06, + "loss": 2.8474, + "step": 6604 + }, + { + "epoch": 1.44, + "learning_rate": 4.28391959798995e-06, + "loss": 2.4062, + "step": 6605 + }, + { + "epoch": 1.44, + "learning_rate": 4.283810356128469e-06, + "loss": 3.525, + "step": 6606 + }, + { + "epoch": 1.44, + "learning_rate": 4.283701114266988e-06, + "loss": 3.0053, + "step": 6607 + }, + { + "epoch": 1.44, + "learning_rate": 4.283591872405506e-06, + "loss": 2.7083, + "step": 6608 + }, + { + "epoch": 1.44, + "learning_rate": 4.283482630544024e-06, + "loss": 3.2178, + "step": 6609 + }, + { + "epoch": 1.44, + "learning_rate": 4.283373388682543e-06, + "loss": 3.1115, + "step": 6610 + }, + { + "epoch": 1.44, + "learning_rate": 4.283264146821062e-06, + "loss": 3.4333, + "step": 6611 + }, + { + "epoch": 1.44, + "learning_rate": 4.283154904959581e-06, + "loss": 2.2645, + "step": 6612 + }, + { + "epoch": 1.44, + "learning_rate": 4.2830456630981e-06, + "loss": 3.6741, + "step": 6613 + }, + { + "epoch": 1.44, + "learning_rate": 4.282936421236618e-06, + "loss": 3.6574, + "step": 6614 + }, + { + "epoch": 1.44, + "learning_rate": 4.282827179375137e-06, + "loss": 2.6784, + "step": 6615 + }, + { + "epoch": 1.44, + "learning_rate": 4.282717937513656e-06, + "loss": 3.3135, + "step": 6616 + }, + { + "epoch": 1.44, + "learning_rate": 4.282608695652175e-06, + "loss": 2.9301, + "step": 6617 + }, + { + "epoch": 1.44, + "learning_rate": 4.282499453790693e-06, + "loss": 2.5635, + "step": 6618 + }, + { + "epoch": 1.44, + "learning_rate": 4.282390211929211e-06, + "loss": 3.0731, + "step": 6619 + }, + { + "epoch": 1.44, + "learning_rate": 4.28228097006773e-06, + "loss": 3.1545, + "step": 6620 + }, + { + "epoch": 1.45, + "learning_rate": 4.282171728206249e-06, + "loss": 2.4422, + "step": 6621 + }, + { + "epoch": 1.45, + "learning_rate": 4.2820624863447675e-06, + "loss": 2.8248, + "step": 6622 + }, + { + "epoch": 1.45, + "learning_rate": 4.2819532444832866e-06, + "loss": 3.7481, + "step": 6623 + }, + { + "epoch": 1.45, + "learning_rate": 4.281844002621805e-06, + "loss": 3.2311, + "step": 6624 + }, + { + "epoch": 1.45, + "learning_rate": 4.281734760760324e-06, + "loss": 2.6828, + "step": 6625 + }, + { + "epoch": 1.45, + "learning_rate": 4.281625518898843e-06, + "loss": 3.0851, + "step": 6626 + }, + { + "epoch": 1.45, + "learning_rate": 4.281516277037361e-06, + "loss": 2.2523, + "step": 6627 + }, + { + "epoch": 1.45, + "learning_rate": 4.28140703517588e-06, + "loss": 3.8464, + "step": 6628 + }, + { + "epoch": 1.45, + "learning_rate": 4.281297793314398e-06, + "loss": 2.244, + "step": 6629 + }, + { + "epoch": 1.45, + "learning_rate": 4.281188551452917e-06, + "loss": 3.1233, + "step": 6630 + }, + { + "epoch": 1.45, + "learning_rate": 4.281079309591436e-06, + "loss": 3.0056, + "step": 6631 + }, + { + "epoch": 1.45, + "learning_rate": 4.2809700677299545e-06, + "loss": 3.686, + "step": 6632 + }, + { + "epoch": 1.45, + "learning_rate": 4.280860825868473e-06, + "loss": 3.3144, + "step": 6633 + }, + { + "epoch": 1.45, + "learning_rate": 4.280751584006992e-06, + "loss": 2.949, + "step": 6634 + }, + { + "epoch": 1.45, + "learning_rate": 4.280642342145511e-06, + "loss": 3.7315, + "step": 6635 + }, + { + "epoch": 1.45, + "learning_rate": 4.28053310028403e-06, + "loss": 2.844, + "step": 6636 + }, + { + "epoch": 1.45, + "learning_rate": 4.280423858422548e-06, + "loss": 2.772, + "step": 6637 + }, + { + "epoch": 1.45, + "learning_rate": 4.280314616561066e-06, + "loss": 2.7433, + "step": 6638 + }, + { + "epoch": 1.45, + "learning_rate": 4.280205374699585e-06, + "loss": 1.8384, + "step": 6639 + }, + { + "epoch": 1.45, + "learning_rate": 4.280096132838104e-06, + "loss": 3.019, + "step": 6640 + }, + { + "epoch": 1.45, + "learning_rate": 4.279986890976622e-06, + "loss": 2.9247, + "step": 6641 + }, + { + "epoch": 1.45, + "learning_rate": 4.2798776491151414e-06, + "loss": 3.5051, + "step": 6642 + }, + { + "epoch": 1.45, + "learning_rate": 4.27976840725366e-06, + "loss": 2.5479, + "step": 6643 + }, + { + "epoch": 1.45, + "learning_rate": 4.279659165392179e-06, + "loss": 2.6381, + "step": 6644 + }, + { + "epoch": 1.45, + "learning_rate": 4.279549923530698e-06, + "loss": 2.7459, + "step": 6645 + }, + { + "epoch": 1.45, + "learning_rate": 4.279440681669216e-06, + "loss": 2.5405, + "step": 6646 + }, + { + "epoch": 1.45, + "learning_rate": 4.279331439807735e-06, + "loss": 2.4838, + "step": 6647 + }, + { + "epoch": 1.45, + "learning_rate": 4.279222197946253e-06, + "loss": 2.5803, + "step": 6648 + }, + { + "epoch": 1.45, + "learning_rate": 4.279112956084772e-06, + "loss": 2.8576, + "step": 6649 + }, + { + "epoch": 1.45, + "learning_rate": 4.279003714223291e-06, + "loss": 3.1385, + "step": 6650 + }, + { + "epoch": 1.45, + "learning_rate": 4.278894472361809e-06, + "loss": 2.3384, + "step": 6651 + }, + { + "epoch": 1.45, + "learning_rate": 4.278785230500328e-06, + "loss": 2.3752, + "step": 6652 + }, + { + "epoch": 1.45, + "learning_rate": 4.278675988638847e-06, + "loss": 2.1406, + "step": 6653 + }, + { + "epoch": 1.45, + "learning_rate": 4.278566746777366e-06, + "loss": 2.9972, + "step": 6654 + }, + { + "epoch": 1.45, + "learning_rate": 4.278457504915884e-06, + "loss": 2.9226, + "step": 6655 + }, + { + "epoch": 1.45, + "learning_rate": 4.278348263054403e-06, + "loss": 2.1835, + "step": 6656 + }, + { + "epoch": 1.45, + "learning_rate": 4.278239021192921e-06, + "loss": 3.6835, + "step": 6657 + }, + { + "epoch": 1.45, + "learning_rate": 4.27812977933144e-06, + "loss": 3.6973, + "step": 6658 + }, + { + "epoch": 1.45, + "learning_rate": 4.278020537469959e-06, + "loss": 3.2592, + "step": 6659 + }, + { + "epoch": 1.45, + "learning_rate": 4.277911295608477e-06, + "loss": 3.0511, + "step": 6660 + }, + { + "epoch": 1.45, + "learning_rate": 4.277802053746996e-06, + "loss": 2.7016, + "step": 6661 + }, + { + "epoch": 1.45, + "learning_rate": 4.2776928118855145e-06, + "loss": 2.8475, + "step": 6662 + }, + { + "epoch": 1.45, + "learning_rate": 4.2775835700240335e-06, + "loss": 2.6981, + "step": 6663 + }, + { + "epoch": 1.45, + "learning_rate": 4.2774743281625526e-06, + "loss": 3.1747, + "step": 6664 + }, + { + "epoch": 1.45, + "learning_rate": 4.277365086301071e-06, + "loss": 4.0251, + "step": 6665 + }, + { + "epoch": 1.45, + "learning_rate": 4.27725584443959e-06, + "loss": 3.5012, + "step": 6666 + }, + { + "epoch": 1.46, + "learning_rate": 4.277146602578108e-06, + "loss": 2.4028, + "step": 6667 + }, + { + "epoch": 1.46, + "learning_rate": 4.277037360716627e-06, + "loss": 3.6662, + "step": 6668 + }, + { + "epoch": 1.46, + "learning_rate": 4.276928118855146e-06, + "loss": 2.5559, + "step": 6669 + }, + { + "epoch": 1.46, + "learning_rate": 4.276818876993664e-06, + "loss": 3.4697, + "step": 6670 + }, + { + "epoch": 1.46, + "learning_rate": 4.276709635132183e-06, + "loss": 3.527, + "step": 6671 + }, + { + "epoch": 1.46, + "learning_rate": 4.2766003932707015e-06, + "loss": 3.3601, + "step": 6672 + }, + { + "epoch": 1.46, + "learning_rate": 4.2764911514092205e-06, + "loss": 2.1886, + "step": 6673 + }, + { + "epoch": 1.46, + "learning_rate": 4.276381909547739e-06, + "loss": 2.7108, + "step": 6674 + }, + { + "epoch": 1.46, + "learning_rate": 4.276272667686258e-06, + "loss": 2.4467, + "step": 6675 + }, + { + "epoch": 1.46, + "learning_rate": 4.276163425824777e-06, + "loss": 2.9196, + "step": 6676 + }, + { + "epoch": 1.46, + "learning_rate": 4.276054183963295e-06, + "loss": 3.1807, + "step": 6677 + }, + { + "epoch": 1.46, + "learning_rate": 4.275944942101814e-06, + "loss": 2.8649, + "step": 6678 + }, + { + "epoch": 1.46, + "learning_rate": 4.275835700240332e-06, + "loss": 3.4725, + "step": 6679 + }, + { + "epoch": 1.46, + "learning_rate": 4.275726458378851e-06, + "loss": 2.1838, + "step": 6680 + }, + { + "epoch": 1.46, + "learning_rate": 4.27561721651737e-06, + "loss": 3.162, + "step": 6681 + }, + { + "epoch": 1.46, + "learning_rate": 4.275507974655888e-06, + "loss": 2.812, + "step": 6682 + }, + { + "epoch": 1.46, + "learning_rate": 4.2753987327944075e-06, + "loss": 2.9426, + "step": 6683 + }, + { + "epoch": 1.46, + "learning_rate": 4.275289490932926e-06, + "loss": 2.0377, + "step": 6684 + }, + { + "epoch": 1.46, + "learning_rate": 4.275180249071445e-06, + "loss": 3.4167, + "step": 6685 + }, + { + "epoch": 1.46, + "learning_rate": 4.275071007209963e-06, + "loss": 1.8185, + "step": 6686 + }, + { + "epoch": 1.46, + "learning_rate": 4.274961765348482e-06, + "loss": 1.9251, + "step": 6687 + }, + { + "epoch": 1.46, + "learning_rate": 4.274852523487e-06, + "loss": 3.2814, + "step": 6688 + }, + { + "epoch": 1.46, + "learning_rate": 4.274743281625519e-06, + "loss": 2.0969, + "step": 6689 + }, + { + "epoch": 1.46, + "learning_rate": 4.274634039764038e-06, + "loss": 3.2312, + "step": 6690 + }, + { + "epoch": 1.46, + "learning_rate": 4.274524797902556e-06, + "loss": 3.5072, + "step": 6691 + }, + { + "epoch": 1.46, + "learning_rate": 4.274415556041075e-06, + "loss": 2.6295, + "step": 6692 + }, + { + "epoch": 1.46, + "learning_rate": 4.2743063141795936e-06, + "loss": 3.7311, + "step": 6693 + }, + { + "epoch": 1.46, + "learning_rate": 4.274197072318113e-06, + "loss": 2.828, + "step": 6694 + }, + { + "epoch": 1.46, + "learning_rate": 4.274087830456632e-06, + "loss": 3.5942, + "step": 6695 + }, + { + "epoch": 1.46, + "learning_rate": 4.27397858859515e-06, + "loss": 3.3277, + "step": 6696 + }, + { + "epoch": 1.46, + "learning_rate": 4.273869346733669e-06, + "loss": 3.5047, + "step": 6697 + }, + { + "epoch": 1.46, + "learning_rate": 4.273760104872187e-06, + "loss": 2.3947, + "step": 6698 + }, + { + "epoch": 1.46, + "learning_rate": 4.273650863010706e-06, + "loss": 3.0934, + "step": 6699 + }, + { + "epoch": 1.46, + "learning_rate": 4.273541621149225e-06, + "loss": 3.0723, + "step": 6700 + }, + { + "epoch": 1.46, + "learning_rate": 4.273432379287743e-06, + "loss": 3.5035, + "step": 6701 + }, + { + "epoch": 1.46, + "learning_rate": 4.273323137426262e-06, + "loss": 2.546, + "step": 6702 + }, + { + "epoch": 1.46, + "learning_rate": 4.2732138955647805e-06, + "loss": 2.3616, + "step": 6703 + }, + { + "epoch": 1.46, + "learning_rate": 4.2731046537032996e-06, + "loss": 3.3584, + "step": 6704 + }, + { + "epoch": 1.46, + "learning_rate": 4.272995411841819e-06, + "loss": 3.491, + "step": 6705 + }, + { + "epoch": 1.46, + "learning_rate": 4.272886169980337e-06, + "loss": 3.1705, + "step": 6706 + }, + { + "epoch": 1.46, + "learning_rate": 4.272776928118855e-06, + "loss": 3.2473, + "step": 6707 + }, + { + "epoch": 1.46, + "learning_rate": 4.272667686257374e-06, + "loss": 3.016, + "step": 6708 + }, + { + "epoch": 1.46, + "learning_rate": 4.272558444395893e-06, + "loss": 3.5325, + "step": 6709 + }, + { + "epoch": 1.46, + "learning_rate": 4.272449202534411e-06, + "loss": 2.6964, + "step": 6710 + }, + { + "epoch": 1.46, + "learning_rate": 4.27233996067293e-06, + "loss": 3.4756, + "step": 6711 + }, + { + "epoch": 1.46, + "learning_rate": 4.2722307188114484e-06, + "loss": 2.9317, + "step": 6712 + }, + { + "epoch": 1.47, + "learning_rate": 4.2721214769499675e-06, + "loss": 3.8554, + "step": 6713 + }, + { + "epoch": 1.47, + "learning_rate": 4.2720122350884865e-06, + "loss": 2.9527, + "step": 6714 + }, + { + "epoch": 1.47, + "learning_rate": 4.271902993227005e-06, + "loss": 3.3504, + "step": 6715 + }, + { + "epoch": 1.47, + "learning_rate": 4.271793751365524e-06, + "loss": 2.8661, + "step": 6716 + }, + { + "epoch": 1.47, + "learning_rate": 4.271684509504042e-06, + "loss": 2.3591, + "step": 6717 + }, + { + "epoch": 1.47, + "learning_rate": 4.271575267642561e-06, + "loss": 2.3028, + "step": 6718 + }, + { + "epoch": 1.47, + "learning_rate": 4.27146602578108e-06, + "loss": 2.7129, + "step": 6719 + }, + { + "epoch": 1.47, + "learning_rate": 4.271356783919598e-06, + "loss": 2.6887, + "step": 6720 + }, + { + "epoch": 1.47, + "learning_rate": 4.271247542058116e-06, + "loss": 1.7388, + "step": 6721 + }, + { + "epoch": 1.47, + "learning_rate": 4.271138300196635e-06, + "loss": 2.9494, + "step": 6722 + }, + { + "epoch": 1.47, + "learning_rate": 4.2710290583351544e-06, + "loss": 3.041, + "step": 6723 + }, + { + "epoch": 1.47, + "learning_rate": 4.2709198164736735e-06, + "loss": 3.0713, + "step": 6724 + }, + { + "epoch": 1.47, + "learning_rate": 4.270810574612192e-06, + "loss": 2.7011, + "step": 6725 + }, + { + "epoch": 1.47, + "learning_rate": 4.27070133275071e-06, + "loss": 3.1437, + "step": 6726 + }, + { + "epoch": 1.47, + "learning_rate": 4.270592090889229e-06, + "loss": 2.018, + "step": 6727 + }, + { + "epoch": 1.47, + "learning_rate": 4.270482849027748e-06, + "loss": 2.8967, + "step": 6728 + }, + { + "epoch": 1.47, + "learning_rate": 4.270373607166267e-06, + "loss": 2.6809, + "step": 6729 + }, + { + "epoch": 1.47, + "learning_rate": 4.270264365304785e-06, + "loss": 3.745, + "step": 6730 + }, + { + "epoch": 1.47, + "learning_rate": 4.270155123443303e-06, + "loss": 3.6124, + "step": 6731 + }, + { + "epoch": 1.47, + "learning_rate": 4.270045881581822e-06, + "loss": 2.8871, + "step": 6732 + }, + { + "epoch": 1.47, + "learning_rate": 4.269936639720341e-06, + "loss": 2.5001, + "step": 6733 + }, + { + "epoch": 1.47, + "learning_rate": 4.2698273978588604e-06, + "loss": 2.385, + "step": 6734 + }, + { + "epoch": 1.47, + "learning_rate": 4.269718155997379e-06, + "loss": 3.2689, + "step": 6735 + }, + { + "epoch": 1.47, + "learning_rate": 4.269608914135897e-06, + "loss": 2.472, + "step": 6736 + }, + { + "epoch": 1.47, + "learning_rate": 4.269499672274416e-06, + "loss": 3.1689, + "step": 6737 + }, + { + "epoch": 1.47, + "learning_rate": 4.269390430412935e-06, + "loss": 2.1919, + "step": 6738 + }, + { + "epoch": 1.47, + "learning_rate": 4.269281188551453e-06, + "loss": 3.4114, + "step": 6739 + }, + { + "epoch": 1.47, + "learning_rate": 4.269171946689972e-06, + "loss": 3.6976, + "step": 6740 + }, + { + "epoch": 1.47, + "learning_rate": 4.26906270482849e-06, + "loss": 3.547, + "step": 6741 + }, + { + "epoch": 1.47, + "learning_rate": 4.268953462967009e-06, + "loss": 3.9502, + "step": 6742 + }, + { + "epoch": 1.47, + "learning_rate": 4.268844221105528e-06, + "loss": 2.7008, + "step": 6743 + }, + { + "epoch": 1.47, + "learning_rate": 4.2687349792440465e-06, + "loss": 2.2711, + "step": 6744 + }, + { + "epoch": 1.47, + "learning_rate": 4.268625737382565e-06, + "loss": 2.2705, + "step": 6745 + }, + { + "epoch": 1.47, + "learning_rate": 4.268516495521084e-06, + "loss": 3.2212, + "step": 6746 + }, + { + "epoch": 1.47, + "learning_rate": 4.268407253659603e-06, + "loss": 3.0337, + "step": 6747 + }, + { + "epoch": 1.47, + "learning_rate": 4.268298011798122e-06, + "loss": 3.9648, + "step": 6748 + }, + { + "epoch": 1.47, + "learning_rate": 4.26818876993664e-06, + "loss": 3.0223, + "step": 6749 + }, + { + "epoch": 1.47, + "learning_rate": 4.268079528075158e-06, + "loss": 3.4425, + "step": 6750 + }, + { + "epoch": 1.47, + "learning_rate": 4.267970286213677e-06, + "loss": 3.1886, + "step": 6751 + }, + { + "epoch": 1.47, + "learning_rate": 4.267861044352196e-06, + "loss": 3.0996, + "step": 6752 + }, + { + "epoch": 1.47, + "learning_rate": 4.267751802490715e-06, + "loss": 1.9322, + "step": 6753 + }, + { + "epoch": 1.47, + "learning_rate": 4.2676425606292335e-06, + "loss": 2.4327, + "step": 6754 + }, + { + "epoch": 1.47, + "learning_rate": 4.267533318767752e-06, + "loss": 3.2133, + "step": 6755 + }, + { + "epoch": 1.47, + "learning_rate": 4.267424076906271e-06, + "loss": 2.6616, + "step": 6756 + }, + { + "epoch": 1.47, + "learning_rate": 4.26731483504479e-06, + "loss": 2.7294, + "step": 6757 + }, + { + "epoch": 1.47, + "learning_rate": 4.267205593183309e-06, + "loss": 3.2249, + "step": 6758 + }, + { + "epoch": 1.48, + "learning_rate": 4.267096351321827e-06, + "loss": 3.1338, + "step": 6759 + }, + { + "epoch": 1.48, + "learning_rate": 4.266987109460345e-06, + "loss": 1.9542, + "step": 6760 + }, + { + "epoch": 1.48, + "learning_rate": 4.266877867598864e-06, + "loss": 3.0545, + "step": 6761 + }, + { + "epoch": 1.48, + "learning_rate": 4.266768625737383e-06, + "loss": 3.0721, + "step": 6762 + }, + { + "epoch": 1.48, + "learning_rate": 4.266659383875902e-06, + "loss": 2.9026, + "step": 6763 + }, + { + "epoch": 1.48, + "learning_rate": 4.2665501420144204e-06, + "loss": 2.7345, + "step": 6764 + }, + { + "epoch": 1.48, + "learning_rate": 4.266440900152939e-06, + "loss": 2.6591, + "step": 6765 + }, + { + "epoch": 1.48, + "learning_rate": 4.266331658291458e-06, + "loss": 2.9754, + "step": 6766 + }, + { + "epoch": 1.48, + "learning_rate": 4.266222416429977e-06, + "loss": 2.3285, + "step": 6767 + }, + { + "epoch": 1.48, + "learning_rate": 4.266113174568495e-06, + "loss": 2.8937, + "step": 6768 + }, + { + "epoch": 1.48, + "learning_rate": 4.266003932707014e-06, + "loss": 2.4356, + "step": 6769 + }, + { + "epoch": 1.48, + "learning_rate": 4.265894690845532e-06, + "loss": 3.0894, + "step": 6770 + }, + { + "epoch": 1.48, + "learning_rate": 4.265785448984051e-06, + "loss": 3.4516, + "step": 6771 + }, + { + "epoch": 1.48, + "learning_rate": 4.26567620712257e-06, + "loss": 2.746, + "step": 6772 + }, + { + "epoch": 1.48, + "learning_rate": 4.265566965261088e-06, + "loss": 2.6133, + "step": 6773 + }, + { + "epoch": 1.48, + "learning_rate": 4.2654577233996066e-06, + "loss": 3.1703, + "step": 6774 + }, + { + "epoch": 1.48, + "learning_rate": 4.265348481538126e-06, + "loss": 3.7608, + "step": 6775 + }, + { + "epoch": 1.48, + "learning_rate": 4.265239239676645e-06, + "loss": 3.4586, + "step": 6776 + }, + { + "epoch": 1.48, + "learning_rate": 4.265129997815164e-06, + "loss": 3.0086, + "step": 6777 + }, + { + "epoch": 1.48, + "learning_rate": 4.265020755953682e-06, + "loss": 2.654, + "step": 6778 + }, + { + "epoch": 1.48, + "learning_rate": 4.2649115140922e-06, + "loss": 3.4192, + "step": 6779 + }, + { + "epoch": 1.48, + "learning_rate": 4.264802272230719e-06, + "loss": 2.9698, + "step": 6780 + }, + { + "epoch": 1.48, + "learning_rate": 4.264693030369238e-06, + "loss": 2.757, + "step": 6781 + }, + { + "epoch": 1.48, + "learning_rate": 4.264583788507757e-06, + "loss": 2.1993, + "step": 6782 + }, + { + "epoch": 1.48, + "learning_rate": 4.264474546646275e-06, + "loss": 3.0662, + "step": 6783 + }, + { + "epoch": 1.48, + "learning_rate": 4.2643653047847935e-06, + "loss": 2.5889, + "step": 6784 + }, + { + "epoch": 1.48, + "learning_rate": 4.2642560629233126e-06, + "loss": 2.6379, + "step": 6785 + }, + { + "epoch": 1.48, + "learning_rate": 4.264146821061832e-06, + "loss": 3.1646, + "step": 6786 + }, + { + "epoch": 1.48, + "learning_rate": 4.26403757920035e-06, + "loss": 3.5524, + "step": 6787 + }, + { + "epoch": 1.48, + "learning_rate": 4.263928337338869e-06, + "loss": 2.7107, + "step": 6788 + }, + { + "epoch": 1.48, + "learning_rate": 4.263819095477387e-06, + "loss": 2.1617, + "step": 6789 + }, + { + "epoch": 1.48, + "learning_rate": 4.263709853615906e-06, + "loss": 2.7558, + "step": 6790 + }, + { + "epoch": 1.48, + "learning_rate": 4.263600611754425e-06, + "loss": 3.023, + "step": 6791 + }, + { + "epoch": 1.48, + "learning_rate": 4.263491369892943e-06, + "loss": 3.1836, + "step": 6792 + }, + { + "epoch": 1.48, + "learning_rate": 4.263382128031462e-06, + "loss": 2.5964, + "step": 6793 + }, + { + "epoch": 1.48, + "learning_rate": 4.2632728861699805e-06, + "loss": 3.2552, + "step": 6794 + }, + { + "epoch": 1.48, + "learning_rate": 4.2631636443084995e-06, + "loss": 2.7061, + "step": 6795 + }, + { + "epoch": 1.48, + "learning_rate": 4.2630544024470185e-06, + "loss": 3.6571, + "step": 6796 + }, + { + "epoch": 1.48, + "learning_rate": 4.262945160585537e-06, + "loss": 2.2034, + "step": 6797 + }, + { + "epoch": 1.48, + "learning_rate": 4.262835918724055e-06, + "loss": 3.0326, + "step": 6798 + }, + { + "epoch": 1.48, + "learning_rate": 4.262726676862574e-06, + "loss": 2.9592, + "step": 6799 + }, + { + "epoch": 1.48, + "learning_rate": 4.262617435001093e-06, + "loss": 2.5007, + "step": 6800 + }, + { + "epoch": 1.48, + "learning_rate": 4.262508193139611e-06, + "loss": 3.6212, + "step": 6801 + }, + { + "epoch": 1.48, + "learning_rate": 4.26239895127813e-06, + "loss": 2.2212, + "step": 6802 + }, + { + "epoch": 1.48, + "learning_rate": 4.262289709416648e-06, + "loss": 2.8966, + "step": 6803 + }, + { + "epoch": 1.48, + "learning_rate": 4.2621804675551674e-06, + "loss": 2.5301, + "step": 6804 + }, + { + "epoch": 1.49, + "learning_rate": 4.2620712256936865e-06, + "loss": 3.3023, + "step": 6805 + }, + { + "epoch": 1.49, + "learning_rate": 4.261961983832205e-06, + "loss": 2.2941, + "step": 6806 + }, + { + "epoch": 1.49, + "learning_rate": 4.261852741970724e-06, + "loss": 2.4222, + "step": 6807 + }, + { + "epoch": 1.49, + "learning_rate": 4.261743500109242e-06, + "loss": 2.9171, + "step": 6808 + }, + { + "epoch": 1.49, + "learning_rate": 4.261634258247761e-06, + "loss": 3.032, + "step": 6809 + }, + { + "epoch": 1.49, + "learning_rate": 4.26152501638628e-06, + "loss": 2.2948, + "step": 6810 + }, + { + "epoch": 1.49, + "learning_rate": 4.261415774524798e-06, + "loss": 2.4051, + "step": 6811 + }, + { + "epoch": 1.49, + "learning_rate": 4.261306532663317e-06, + "loss": 2.0788, + "step": 6812 + }, + { + "epoch": 1.49, + "learning_rate": 4.261197290801835e-06, + "loss": 1.6735, + "step": 6813 + }, + { + "epoch": 1.49, + "learning_rate": 4.261088048940354e-06, + "loss": 3.1976, + "step": 6814 + }, + { + "epoch": 1.49, + "learning_rate": 4.260978807078873e-06, + "loss": 2.5418, + "step": 6815 + }, + { + "epoch": 1.49, + "learning_rate": 4.260869565217392e-06, + "loss": 2.0124, + "step": 6816 + }, + { + "epoch": 1.49, + "learning_rate": 4.260760323355911e-06, + "loss": 3.2625, + "step": 6817 + }, + { + "epoch": 1.49, + "learning_rate": 4.260651081494429e-06, + "loss": 2.0867, + "step": 6818 + }, + { + "epoch": 1.49, + "learning_rate": 4.260541839632948e-06, + "loss": 1.6622, + "step": 6819 + }, + { + "epoch": 1.49, + "learning_rate": 4.260432597771466e-06, + "loss": 2.2135, + "step": 6820 + }, + { + "epoch": 1.49, + "learning_rate": 4.260323355909985e-06, + "loss": 3.3338, + "step": 6821 + }, + { + "epoch": 1.49, + "learning_rate": 4.260214114048504e-06, + "loss": 3.5641, + "step": 6822 + }, + { + "epoch": 1.49, + "learning_rate": 4.260104872187022e-06, + "loss": 2.4655, + "step": 6823 + }, + { + "epoch": 1.49, + "learning_rate": 4.259995630325541e-06, + "loss": 3.0444, + "step": 6824 + }, + { + "epoch": 1.49, + "learning_rate": 4.2598863884640595e-06, + "loss": 2.1554, + "step": 6825 + }, + { + "epoch": 1.49, + "learning_rate": 4.2597771466025786e-06, + "loss": 2.6912, + "step": 6826 + }, + { + "epoch": 1.49, + "learning_rate": 4.259667904741097e-06, + "loss": 2.1273, + "step": 6827 + }, + { + "epoch": 1.49, + "learning_rate": 4.259558662879616e-06, + "loss": 2.5812, + "step": 6828 + }, + { + "epoch": 1.49, + "learning_rate": 4.259449421018135e-06, + "loss": 3.0013, + "step": 6829 + }, + { + "epoch": 1.49, + "learning_rate": 4.259340179156653e-06, + "loss": 2.9419, + "step": 6830 + }, + { + "epoch": 1.49, + "learning_rate": 4.259230937295172e-06, + "loss": 3.3875, + "step": 6831 + }, + { + "epoch": 1.49, + "learning_rate": 4.25912169543369e-06, + "loss": 2.8956, + "step": 6832 + }, + { + "epoch": 1.49, + "learning_rate": 4.259012453572209e-06, + "loss": 2.0758, + "step": 6833 + }, + { + "epoch": 1.49, + "learning_rate": 4.2589032117107275e-06, + "loss": 2.4649, + "step": 6834 + }, + { + "epoch": 1.49, + "learning_rate": 4.2587939698492465e-06, + "loss": 3.8549, + "step": 6835 + }, + { + "epoch": 1.49, + "learning_rate": 4.2586847279877655e-06, + "loss": 3.8125, + "step": 6836 + }, + { + "epoch": 1.49, + "learning_rate": 4.258575486126284e-06, + "loss": 2.8962, + "step": 6837 + }, + { + "epoch": 1.49, + "learning_rate": 4.258466244264803e-06, + "loss": 3.1046, + "step": 6838 + }, + { + "epoch": 1.49, + "learning_rate": 4.258357002403321e-06, + "loss": 2.7953, + "step": 6839 + }, + { + "epoch": 1.49, + "learning_rate": 4.25824776054184e-06, + "loss": 2.9401, + "step": 6840 + }, + { + "epoch": 1.49, + "learning_rate": 4.258138518680359e-06, + "loss": 3.1952, + "step": 6841 + }, + { + "epoch": 1.49, + "learning_rate": 4.258029276818877e-06, + "loss": 2.7036, + "step": 6842 + }, + { + "epoch": 1.49, + "learning_rate": 4.257920034957396e-06, + "loss": 3.3042, + "step": 6843 + }, + { + "epoch": 1.49, + "learning_rate": 4.257810793095914e-06, + "loss": 3.4003, + "step": 6844 + }, + { + "epoch": 1.49, + "learning_rate": 4.2577015512344334e-06, + "loss": 2.4545, + "step": 6845 + }, + { + "epoch": 1.49, + "learning_rate": 4.2575923093729525e-06, + "loss": 2.8469, + "step": 6846 + }, + { + "epoch": 1.49, + "learning_rate": 4.257483067511471e-06, + "loss": 2.2027, + "step": 6847 + }, + { + "epoch": 1.49, + "learning_rate": 4.25737382564999e-06, + "loss": 2.2514, + "step": 6848 + }, + { + "epoch": 1.49, + "learning_rate": 4.257264583788508e-06, + "loss": 2.6915, + "step": 6849 + }, + { + "epoch": 1.49, + "learning_rate": 4.257155341927027e-06, + "loss": 3.2308, + "step": 6850 + }, + { + "epoch": 1.5, + "learning_rate": 4.257046100065545e-06, + "loss": 2.3252, + "step": 6851 + }, + { + "epoch": 1.5, + "learning_rate": 4.256936858204064e-06, + "loss": 2.2306, + "step": 6852 + }, + { + "epoch": 1.5, + "learning_rate": 4.256827616342582e-06, + "loss": 2.5949, + "step": 6853 + }, + { + "epoch": 1.5, + "learning_rate": 4.256718374481101e-06, + "loss": 3.0208, + "step": 6854 + }, + { + "epoch": 1.5, + "learning_rate": 4.25660913261962e-06, + "loss": 3.1571, + "step": 6855 + }, + { + "epoch": 1.5, + "learning_rate": 4.256499890758139e-06, + "loss": 3.0966, + "step": 6856 + }, + { + "epoch": 1.5, + "learning_rate": 4.256390648896658e-06, + "loss": 2.2538, + "step": 6857 + }, + { + "epoch": 1.5, + "learning_rate": 4.256281407035176e-06, + "loss": 3.5872, + "step": 6858 + }, + { + "epoch": 1.5, + "learning_rate": 4.256172165173695e-06, + "loss": 3.5335, + "step": 6859 + }, + { + "epoch": 1.5, + "learning_rate": 4.256062923312214e-06, + "loss": 2.4168, + "step": 6860 + }, + { + "epoch": 1.5, + "learning_rate": 4.255953681450732e-06, + "loss": 2.7929, + "step": 6861 + }, + { + "epoch": 1.5, + "learning_rate": 4.255844439589251e-06, + "loss": 2.9278, + "step": 6862 + }, + { + "epoch": 1.5, + "learning_rate": 4.255735197727769e-06, + "loss": 2.44, + "step": 6863 + }, + { + "epoch": 1.5, + "learning_rate": 4.255625955866288e-06, + "loss": 3.5093, + "step": 6864 + }, + { + "epoch": 1.5, + "learning_rate": 4.255516714004807e-06, + "loss": 3.0386, + "step": 6865 + }, + { + "epoch": 1.5, + "learning_rate": 4.2554074721433255e-06, + "loss": 3.1443, + "step": 6866 + }, + { + "epoch": 1.5, + "learning_rate": 4.255298230281844e-06, + "loss": 2.7582, + "step": 6867 + }, + { + "epoch": 1.5, + "learning_rate": 4.255188988420363e-06, + "loss": 2.5937, + "step": 6868 + }, + { + "epoch": 1.5, + "learning_rate": 4.255079746558882e-06, + "loss": 3.2997, + "step": 6869 + }, + { + "epoch": 1.5, + "learning_rate": 4.254970504697401e-06, + "loss": 3.4901, + "step": 6870 + }, + { + "epoch": 1.5, + "learning_rate": 4.254861262835919e-06, + "loss": 3.1088, + "step": 6871 + }, + { + "epoch": 1.5, + "learning_rate": 4.254752020974437e-06, + "loss": 1.779, + "step": 6872 + }, + { + "epoch": 1.5, + "learning_rate": 4.254642779112956e-06, + "loss": 2.5608, + "step": 6873 + }, + { + "epoch": 1.5, + "learning_rate": 4.254533537251475e-06, + "loss": 2.5875, + "step": 6874 + }, + { + "epoch": 1.5, + "learning_rate": 4.254424295389994e-06, + "loss": 2.7716, + "step": 6875 + }, + { + "epoch": 1.5, + "learning_rate": 4.2543150535285125e-06, + "loss": 3.8597, + "step": 6876 + }, + { + "epoch": 1.5, + "learning_rate": 4.254205811667031e-06, + "loss": 3.4301, + "step": 6877 + }, + { + "epoch": 1.5, + "learning_rate": 4.25409656980555e-06, + "loss": 3.2914, + "step": 6878 + }, + { + "epoch": 1.5, + "learning_rate": 4.253987327944069e-06, + "loss": 3.2062, + "step": 6879 + }, + { + "epoch": 1.5, + "learning_rate": 4.253878086082587e-06, + "loss": 3.2347, + "step": 6880 + }, + { + "epoch": 1.5, + "learning_rate": 4.253768844221106e-06, + "loss": 3.286, + "step": 6881 + }, + { + "epoch": 1.5, + "learning_rate": 4.253659602359624e-06, + "loss": 3.0501, + "step": 6882 + }, + { + "epoch": 1.5, + "learning_rate": 4.253550360498143e-06, + "loss": 1.9456, + "step": 6883 + }, + { + "epoch": 1.5, + "learning_rate": 4.253441118636662e-06, + "loss": 3.7335, + "step": 6884 + }, + { + "epoch": 1.5, + "learning_rate": 4.2533318767751804e-06, + "loss": 2.5211, + "step": 6885 + }, + { + "epoch": 1.5, + "learning_rate": 4.253222634913699e-06, + "loss": 3.3012, + "step": 6886 + }, + { + "epoch": 1.5, + "learning_rate": 4.253113393052218e-06, + "loss": 3.7342, + "step": 6887 + }, + { + "epoch": 1.5, + "learning_rate": 4.253004151190737e-06, + "loss": 3.4478, + "step": 6888 + }, + { + "epoch": 1.5, + "learning_rate": 4.252894909329256e-06, + "loss": 2.8281, + "step": 6889 + }, + { + "epoch": 1.5, + "learning_rate": 4.252785667467774e-06, + "loss": 2.8549, + "step": 6890 + }, + { + "epoch": 1.5, + "learning_rate": 4.252676425606292e-06, + "loss": 3.1336, + "step": 6891 + }, + { + "epoch": 1.5, + "learning_rate": 4.252567183744811e-06, + "loss": 2.9571, + "step": 6892 + }, + { + "epoch": 1.5, + "learning_rate": 4.25245794188333e-06, + "loss": 3.3571, + "step": 6893 + }, + { + "epoch": 1.5, + "learning_rate": 4.252348700021849e-06, + "loss": 3.2417, + "step": 6894 + }, + { + "epoch": 1.5, + "learning_rate": 4.252239458160367e-06, + "loss": 3.5923, + "step": 6895 + }, + { + "epoch": 1.51, + "learning_rate": 4.2521302162988856e-06, + "loss": 3.47, + "step": 6896 + }, + { + "epoch": 1.51, + "learning_rate": 4.252020974437405e-06, + "loss": 3.0198, + "step": 6897 + }, + { + "epoch": 1.51, + "learning_rate": 4.251911732575924e-06, + "loss": 2.7776, + "step": 6898 + }, + { + "epoch": 1.51, + "learning_rate": 4.251802490714443e-06, + "loss": 3.1795, + "step": 6899 + }, + { + "epoch": 1.51, + "learning_rate": 4.251693248852961e-06, + "loss": 2.4936, + "step": 6900 + }, + { + "epoch": 1.51, + "learning_rate": 4.251584006991479e-06, + "loss": 3.1218, + "step": 6901 + }, + { + "epoch": 1.51, + "learning_rate": 4.251474765129998e-06, + "loss": 2.9399, + "step": 6902 + }, + { + "epoch": 1.51, + "learning_rate": 4.251365523268517e-06, + "loss": 2.8281, + "step": 6903 + }, + { + "epoch": 1.51, + "learning_rate": 4.251256281407035e-06, + "loss": 3.1329, + "step": 6904 + }, + { + "epoch": 1.51, + "learning_rate": 4.251147039545554e-06, + "loss": 2.7155, + "step": 6905 + }, + { + "epoch": 1.51, + "learning_rate": 4.2510377976840725e-06, + "loss": 3.5777, + "step": 6906 + }, + { + "epoch": 1.51, + "learning_rate": 4.2509285558225916e-06, + "loss": 3.1961, + "step": 6907 + }, + { + "epoch": 1.51, + "learning_rate": 4.250819313961111e-06, + "loss": 2.3977, + "step": 6908 + }, + { + "epoch": 1.51, + "learning_rate": 4.250710072099629e-06, + "loss": 2.8697, + "step": 6909 + }, + { + "epoch": 1.51, + "learning_rate": 4.250600830238148e-06, + "loss": 2.6166, + "step": 6910 + }, + { + "epoch": 1.51, + "learning_rate": 4.250491588376666e-06, + "loss": 3.2135, + "step": 6911 + }, + { + "epoch": 1.51, + "learning_rate": 4.250382346515185e-06, + "loss": 2.5365, + "step": 6912 + }, + { + "epoch": 1.51, + "learning_rate": 4.250273104653704e-06, + "loss": 3.1357, + "step": 6913 + }, + { + "epoch": 1.51, + "learning_rate": 4.250163862792222e-06, + "loss": 3.1584, + "step": 6914 + }, + { + "epoch": 1.51, + "learning_rate": 4.2500546209307404e-06, + "loss": 2.4382, + "step": 6915 + }, + { + "epoch": 1.51, + "learning_rate": 4.2499453790692595e-06, + "loss": 3.0197, + "step": 6916 + }, + { + "epoch": 1.51, + "learning_rate": 4.2498361372077785e-06, + "loss": 3.2809, + "step": 6917 + }, + { + "epoch": 1.51, + "learning_rate": 4.2497268953462976e-06, + "loss": 2.8963, + "step": 6918 + }, + { + "epoch": 1.51, + "learning_rate": 4.249617653484816e-06, + "loss": 2.0525, + "step": 6919 + }, + { + "epoch": 1.51, + "learning_rate": 4.249508411623334e-06, + "loss": 2.7683, + "step": 6920 + }, + { + "epoch": 1.51, + "learning_rate": 4.249399169761853e-06, + "loss": 3.4737, + "step": 6921 + }, + { + "epoch": 1.51, + "learning_rate": 4.249289927900372e-06, + "loss": 3.711, + "step": 6922 + }, + { + "epoch": 1.51, + "learning_rate": 4.249180686038891e-06, + "loss": 2.9882, + "step": 6923 + }, + { + "epoch": 1.51, + "learning_rate": 4.249071444177409e-06, + "loss": 2.0626, + "step": 6924 + }, + { + "epoch": 1.51, + "learning_rate": 4.248962202315927e-06, + "loss": 2.2306, + "step": 6925 + }, + { + "epoch": 1.51, + "learning_rate": 4.2488529604544464e-06, + "loss": 2.6944, + "step": 6926 + }, + { + "epoch": 1.51, + "learning_rate": 4.2487437185929655e-06, + "loss": 2.5896, + "step": 6927 + }, + { + "epoch": 1.51, + "learning_rate": 4.2486344767314845e-06, + "loss": 3.2438, + "step": 6928 + }, + { + "epoch": 1.51, + "learning_rate": 4.248525234870003e-06, + "loss": 2.7012, + "step": 6929 + }, + { + "epoch": 1.51, + "learning_rate": 4.248415993008521e-06, + "loss": 3.1615, + "step": 6930 + }, + { + "epoch": 1.51, + "learning_rate": 4.24830675114704e-06, + "loss": 2.3882, + "step": 6931 + }, + { + "epoch": 1.51, + "learning_rate": 4.248197509285559e-06, + "loss": 2.9881, + "step": 6932 + }, + { + "epoch": 1.51, + "learning_rate": 4.248088267424077e-06, + "loss": 2.8096, + "step": 6933 + }, + { + "epoch": 1.51, + "learning_rate": 4.247979025562596e-06, + "loss": 2.0184, + "step": 6934 + }, + { + "epoch": 1.51, + "learning_rate": 4.247869783701114e-06, + "loss": 2.7444, + "step": 6935 + }, + { + "epoch": 1.51, + "learning_rate": 4.247760541839633e-06, + "loss": 2.4745, + "step": 6936 + }, + { + "epoch": 1.51, + "learning_rate": 4.2476512999781524e-06, + "loss": 2.648, + "step": 6937 + }, + { + "epoch": 1.51, + "learning_rate": 4.247542058116671e-06, + "loss": 2.4454, + "step": 6938 + }, + { + "epoch": 1.51, + "learning_rate": 4.247432816255189e-06, + "loss": 2.2981, + "step": 6939 + }, + { + "epoch": 1.51, + "learning_rate": 4.247323574393708e-06, + "loss": 2.5834, + "step": 6940 + }, + { + "epoch": 1.51, + "learning_rate": 4.247214332532227e-06, + "loss": 3.1089, + "step": 6941 + }, + { + "epoch": 1.52, + "learning_rate": 4.247105090670746e-06, + "loss": 2.2398, + "step": 6942 + }, + { + "epoch": 1.52, + "learning_rate": 4.246995848809264e-06, + "loss": 2.7517, + "step": 6943 + }, + { + "epoch": 1.52, + "learning_rate": 4.246886606947782e-06, + "loss": 2.7109, + "step": 6944 + }, + { + "epoch": 1.52, + "learning_rate": 4.246777365086301e-06, + "loss": 3.213, + "step": 6945 + }, + { + "epoch": 1.52, + "learning_rate": 4.24666812322482e-06, + "loss": 2.6078, + "step": 6946 + }, + { + "epoch": 1.52, + "learning_rate": 4.246558881363339e-06, + "loss": 2.4537, + "step": 6947 + }, + { + "epoch": 1.52, + "learning_rate": 4.2464496395018576e-06, + "loss": 2.9645, + "step": 6948 + }, + { + "epoch": 1.52, + "learning_rate": 4.246340397640376e-06, + "loss": 3.3652, + "step": 6949 + }, + { + "epoch": 1.52, + "learning_rate": 4.246231155778895e-06, + "loss": 3.1374, + "step": 6950 + }, + { + "epoch": 1.52, + "learning_rate": 4.246121913917414e-06, + "loss": 3.7241, + "step": 6951 + }, + { + "epoch": 1.52, + "learning_rate": 4.246012672055932e-06, + "loss": 2.6774, + "step": 6952 + }, + { + "epoch": 1.52, + "learning_rate": 4.245903430194451e-06, + "loss": 2.6897, + "step": 6953 + }, + { + "epoch": 1.52, + "learning_rate": 4.245794188332969e-06, + "loss": 2.2467, + "step": 6954 + }, + { + "epoch": 1.52, + "learning_rate": 4.245684946471488e-06, + "loss": 3.1159, + "step": 6955 + }, + { + "epoch": 1.52, + "learning_rate": 4.245575704610007e-06, + "loss": 3.3902, + "step": 6956 + }, + { + "epoch": 1.52, + "learning_rate": 4.2454664627485255e-06, + "loss": 3.4124, + "step": 6957 + }, + { + "epoch": 1.52, + "learning_rate": 4.2453572208870445e-06, + "loss": 2.7358, + "step": 6958 + }, + { + "epoch": 1.52, + "learning_rate": 4.245247979025563e-06, + "loss": 2.5789, + "step": 6959 + }, + { + "epoch": 1.52, + "learning_rate": 4.245138737164082e-06, + "loss": 2.089, + "step": 6960 + }, + { + "epoch": 1.52, + "learning_rate": 4.245029495302601e-06, + "loss": 3.5013, + "step": 6961 + }, + { + "epoch": 1.52, + "learning_rate": 4.244920253441119e-06, + "loss": 2.6868, + "step": 6962 + }, + { + "epoch": 1.52, + "learning_rate": 4.244811011579638e-06, + "loss": 2.867, + "step": 6963 + }, + { + "epoch": 1.52, + "learning_rate": 4.244701769718156e-06, + "loss": 2.3204, + "step": 6964 + }, + { + "epoch": 1.52, + "learning_rate": 4.244592527856675e-06, + "loss": 2.8308, + "step": 6965 + }, + { + "epoch": 1.52, + "learning_rate": 4.244483285995193e-06, + "loss": 3.6403, + "step": 6966 + }, + { + "epoch": 1.52, + "learning_rate": 4.2443740441337125e-06, + "loss": 2.9187, + "step": 6967 + }, + { + "epoch": 1.52, + "learning_rate": 4.244264802272231e-06, + "loss": 3.3906, + "step": 6968 + }, + { + "epoch": 1.52, + "learning_rate": 4.24415556041075e-06, + "loss": 2.31, + "step": 6969 + }, + { + "epoch": 1.52, + "learning_rate": 4.244046318549269e-06, + "loss": 2.9866, + "step": 6970 + }, + { + "epoch": 1.52, + "learning_rate": 4.243937076687787e-06, + "loss": 1.9803, + "step": 6971 + }, + { + "epoch": 1.52, + "learning_rate": 4.243827834826306e-06, + "loss": 2.4959, + "step": 6972 + }, + { + "epoch": 1.52, + "learning_rate": 4.243718592964824e-06, + "loss": 2.9451, + "step": 6973 + }, + { + "epoch": 1.52, + "learning_rate": 4.243609351103343e-06, + "loss": 2.6111, + "step": 6974 + }, + { + "epoch": 1.52, + "learning_rate": 4.243500109241862e-06, + "loss": 2.3553, + "step": 6975 + }, + { + "epoch": 1.52, + "learning_rate": 4.24339086738038e-06, + "loss": 2.4002, + "step": 6976 + }, + { + "epoch": 1.52, + "learning_rate": 4.243281625518899e-06, + "loss": 2.8639, + "step": 6977 + }, + { + "epoch": 1.52, + "learning_rate": 4.243172383657418e-06, + "loss": 2.1649, + "step": 6978 + }, + { + "epoch": 1.52, + "learning_rate": 4.243063141795937e-06, + "loss": 1.8162, + "step": 6979 + }, + { + "epoch": 1.52, + "learning_rate": 4.242953899934456e-06, + "loss": 2.5619, + "step": 6980 + }, + { + "epoch": 1.52, + "learning_rate": 4.242844658072974e-06, + "loss": 2.6935, + "step": 6981 + }, + { + "epoch": 1.52, + "learning_rate": 4.242735416211493e-06, + "loss": 2.4209, + "step": 6982 + }, + { + "epoch": 1.52, + "learning_rate": 4.242626174350011e-06, + "loss": 1.8913, + "step": 6983 + }, + { + "epoch": 1.52, + "learning_rate": 4.24251693248853e-06, + "loss": 3.2494, + "step": 6984 + }, + { + "epoch": 1.52, + "learning_rate": 4.242407690627048e-06, + "loss": 2.5075, + "step": 6985 + }, + { + "epoch": 1.52, + "learning_rate": 4.242298448765567e-06, + "loss": 3.104, + "step": 6986 + }, + { + "epoch": 1.52, + "learning_rate": 4.242189206904086e-06, + "loss": 2.9945, + "step": 6987 + }, + { + "epoch": 1.53, + "learning_rate": 4.2420799650426046e-06, + "loss": 3.3245, + "step": 6988 + }, + { + "epoch": 1.53, + "learning_rate": 4.241970723181124e-06, + "loss": 2.1885, + "step": 6989 + }, + { + "epoch": 1.53, + "learning_rate": 4.241861481319642e-06, + "loss": 3.4705, + "step": 6990 + }, + { + "epoch": 1.53, + "learning_rate": 4.241752239458161e-06, + "loss": 3.4004, + "step": 6991 + }, + { + "epoch": 1.53, + "learning_rate": 4.241642997596679e-06, + "loss": 2.9969, + "step": 6992 + }, + { + "epoch": 1.53, + "learning_rate": 4.241533755735198e-06, + "loss": 3.137, + "step": 6993 + }, + { + "epoch": 1.53, + "learning_rate": 4.241424513873717e-06, + "loss": 2.5943, + "step": 6994 + }, + { + "epoch": 1.53, + "learning_rate": 4.241315272012235e-06, + "loss": 3.1012, + "step": 6995 + }, + { + "epoch": 1.53, + "learning_rate": 4.241206030150754e-06, + "loss": 2.9123, + "step": 6996 + }, + { + "epoch": 1.53, + "learning_rate": 4.2410967882892725e-06, + "loss": 2.101, + "step": 6997 + }, + { + "epoch": 1.53, + "learning_rate": 4.2409875464277915e-06, + "loss": 3.8011, + "step": 6998 + }, + { + "epoch": 1.53, + "learning_rate": 4.24087830456631e-06, + "loss": 2.6329, + "step": 6999 + }, + { + "epoch": 1.53, + "learning_rate": 4.240769062704829e-06, + "loss": 2.7802, + "step": 7000 + }, + { + "epoch": 1.53, + "learning_rate": 4.240659820843348e-06, + "loss": 3.4504, + "step": 7001 + }, + { + "epoch": 1.53, + "learning_rate": 4.240550578981866e-06, + "loss": 3.1588, + "step": 7002 + }, + { + "epoch": 1.53, + "learning_rate": 4.240441337120385e-06, + "loss": 2.1871, + "step": 7003 + }, + { + "epoch": 1.53, + "learning_rate": 4.240332095258903e-06, + "loss": 2.6773, + "step": 7004 + }, + { + "epoch": 1.53, + "learning_rate": 4.240222853397422e-06, + "loss": 3.2007, + "step": 7005 + }, + { + "epoch": 1.53, + "learning_rate": 4.240113611535941e-06, + "loss": 1.9914, + "step": 7006 + }, + { + "epoch": 1.53, + "learning_rate": 4.2400043696744594e-06, + "loss": 2.2572, + "step": 7007 + }, + { + "epoch": 1.53, + "learning_rate": 4.2398951278129785e-06, + "loss": 2.9109, + "step": 7008 + }, + { + "epoch": 1.53, + "learning_rate": 4.239785885951497e-06, + "loss": 2.0528, + "step": 7009 + }, + { + "epoch": 1.53, + "learning_rate": 4.239676644090016e-06, + "loss": 2.3712, + "step": 7010 + }, + { + "epoch": 1.53, + "learning_rate": 4.239567402228535e-06, + "loss": 2.6454, + "step": 7011 + }, + { + "epoch": 1.53, + "learning_rate": 4.239458160367053e-06, + "loss": 3.4832, + "step": 7012 + }, + { + "epoch": 1.53, + "learning_rate": 4.239348918505572e-06, + "loss": 2.9936, + "step": 7013 + }, + { + "epoch": 1.53, + "learning_rate": 4.23923967664409e-06, + "loss": 3.5171, + "step": 7014 + }, + { + "epoch": 1.53, + "learning_rate": 4.239130434782609e-06, + "loss": 3.0037, + "step": 7015 + }, + { + "epoch": 1.53, + "learning_rate": 4.239021192921128e-06, + "loss": 2.9064, + "step": 7016 + }, + { + "epoch": 1.53, + "learning_rate": 4.238911951059646e-06, + "loss": 2.7876, + "step": 7017 + }, + { + "epoch": 1.53, + "learning_rate": 4.238802709198165e-06, + "loss": 2.7229, + "step": 7018 + }, + { + "epoch": 1.53, + "learning_rate": 4.238693467336684e-06, + "loss": 2.7031, + "step": 7019 + }, + { + "epoch": 1.53, + "learning_rate": 4.238584225475203e-06, + "loss": 3.469, + "step": 7020 + }, + { + "epoch": 1.53, + "learning_rate": 4.238474983613721e-06, + "loss": 2.5431, + "step": 7021 + }, + { + "epoch": 1.53, + "learning_rate": 4.23836574175224e-06, + "loss": 3.2215, + "step": 7022 + }, + { + "epoch": 1.53, + "learning_rate": 4.238256499890758e-06, + "loss": 2.7357, + "step": 7023 + }, + { + "epoch": 1.53, + "learning_rate": 4.238147258029277e-06, + "loss": 2.634, + "step": 7024 + }, + { + "epoch": 1.53, + "learning_rate": 4.238038016167796e-06, + "loss": 3.7219, + "step": 7025 + }, + { + "epoch": 1.53, + "learning_rate": 4.237928774306314e-06, + "loss": 3.0923, + "step": 7026 + }, + { + "epoch": 1.53, + "learning_rate": 4.237819532444833e-06, + "loss": 2.741, + "step": 7027 + }, + { + "epoch": 1.53, + "learning_rate": 4.2377102905833515e-06, + "loss": 3.365, + "step": 7028 + }, + { + "epoch": 1.53, + "learning_rate": 4.2376010487218706e-06, + "loss": 2.9383, + "step": 7029 + }, + { + "epoch": 1.53, + "learning_rate": 4.23749180686039e-06, + "loss": 3.3118, + "step": 7030 + }, + { + "epoch": 1.53, + "learning_rate": 4.237382564998908e-06, + "loss": 2.9005, + "step": 7031 + }, + { + "epoch": 1.53, + "learning_rate": 4.237273323137426e-06, + "loss": 3.9955, + "step": 7032 + }, + { + "epoch": 1.53, + "learning_rate": 4.237164081275945e-06, + "loss": 3.2873, + "step": 7033 + }, + { + "epoch": 1.54, + "learning_rate": 4.237054839414464e-06, + "loss": 3.0471, + "step": 7034 + }, + { + "epoch": 1.54, + "learning_rate": 4.236945597552983e-06, + "loss": 3.5366, + "step": 7035 + }, + { + "epoch": 1.54, + "learning_rate": 4.236836355691501e-06, + "loss": 2.4096, + "step": 7036 + }, + { + "epoch": 1.54, + "learning_rate": 4.2367271138300195e-06, + "loss": 3.1016, + "step": 7037 + }, + { + "epoch": 1.54, + "learning_rate": 4.2366178719685385e-06, + "loss": 2.8681, + "step": 7038 + }, + { + "epoch": 1.54, + "learning_rate": 4.2365086301070575e-06, + "loss": 2.1435, + "step": 7039 + }, + { + "epoch": 1.54, + "learning_rate": 4.2363993882455766e-06, + "loss": 2.828, + "step": 7040 + }, + { + "epoch": 1.54, + "learning_rate": 4.236290146384095e-06, + "loss": 2.3261, + "step": 7041 + }, + { + "epoch": 1.54, + "learning_rate": 4.236180904522613e-06, + "loss": 2.2375, + "step": 7042 + }, + { + "epoch": 1.54, + "learning_rate": 4.236071662661132e-06, + "loss": 3.1231, + "step": 7043 + }, + { + "epoch": 1.54, + "learning_rate": 4.235962420799651e-06, + "loss": 2.3973, + "step": 7044 + }, + { + "epoch": 1.54, + "learning_rate": 4.235853178938169e-06, + "loss": 3.3484, + "step": 7045 + }, + { + "epoch": 1.54, + "learning_rate": 4.235743937076688e-06, + "loss": 1.9619, + "step": 7046 + }, + { + "epoch": 1.54, + "learning_rate": 4.235634695215206e-06, + "loss": 3.4255, + "step": 7047 + }, + { + "epoch": 1.54, + "learning_rate": 4.2355254533537255e-06, + "loss": 3.3531, + "step": 7048 + }, + { + "epoch": 1.54, + "learning_rate": 4.2354162114922445e-06, + "loss": 3.4635, + "step": 7049 + }, + { + "epoch": 1.54, + "learning_rate": 4.235306969630763e-06, + "loss": 2.2933, + "step": 7050 + }, + { + "epoch": 1.54, + "learning_rate": 4.235197727769282e-06, + "loss": 3.3268, + "step": 7051 + }, + { + "epoch": 1.54, + "learning_rate": 4.2350884859078e-06, + "loss": 3.0889, + "step": 7052 + }, + { + "epoch": 1.54, + "learning_rate": 4.234979244046319e-06, + "loss": 3.4952, + "step": 7053 + }, + { + "epoch": 1.54, + "learning_rate": 4.234870002184838e-06, + "loss": 3.8468, + "step": 7054 + }, + { + "epoch": 1.54, + "learning_rate": 4.234760760323356e-06, + "loss": 3.2127, + "step": 7055 + }, + { + "epoch": 1.54, + "learning_rate": 4.234651518461874e-06, + "loss": 3.4381, + "step": 7056 + }, + { + "epoch": 1.54, + "learning_rate": 4.234542276600393e-06, + "loss": 2.1877, + "step": 7057 + }, + { + "epoch": 1.54, + "learning_rate": 4.234433034738912e-06, + "loss": 2.477, + "step": 7058 + }, + { + "epoch": 1.54, + "learning_rate": 4.2343237928774314e-06, + "loss": 2.4855, + "step": 7059 + }, + { + "epoch": 1.54, + "learning_rate": 4.23421455101595e-06, + "loss": 2.8753, + "step": 7060 + }, + { + "epoch": 1.54, + "learning_rate": 4.234105309154468e-06, + "loss": 3.5056, + "step": 7061 + }, + { + "epoch": 1.54, + "learning_rate": 4.233996067292987e-06, + "loss": 2.4992, + "step": 7062 + }, + { + "epoch": 1.54, + "learning_rate": 4.233886825431506e-06, + "loss": 4.4645, + "step": 7063 + }, + { + "epoch": 1.54, + "learning_rate": 4.233777583570025e-06, + "loss": 3.2593, + "step": 7064 + }, + { + "epoch": 1.54, + "learning_rate": 4.233668341708543e-06, + "loss": 2.8779, + "step": 7065 + }, + { + "epoch": 1.54, + "learning_rate": 4.233559099847061e-06, + "loss": 2.7713, + "step": 7066 + }, + { + "epoch": 1.54, + "learning_rate": 4.23344985798558e-06, + "loss": 2.7849, + "step": 7067 + }, + { + "epoch": 1.54, + "learning_rate": 4.233340616124099e-06, + "loss": 2.3664, + "step": 7068 + }, + { + "epoch": 1.54, + "learning_rate": 4.233231374262618e-06, + "loss": 3.1808, + "step": 7069 + }, + { + "epoch": 1.54, + "learning_rate": 4.233122132401137e-06, + "loss": 2.8193, + "step": 7070 + }, + { + "epoch": 1.54, + "learning_rate": 4.233012890539655e-06, + "loss": 3.5224, + "step": 7071 + }, + { + "epoch": 1.54, + "learning_rate": 4.232903648678174e-06, + "loss": 2.4656, + "step": 7072 + }, + { + "epoch": 1.54, + "learning_rate": 4.232794406816693e-06, + "loss": 3.0372, + "step": 7073 + }, + { + "epoch": 1.54, + "learning_rate": 4.232685164955211e-06, + "loss": 3.5231, + "step": 7074 + }, + { + "epoch": 1.54, + "learning_rate": 4.23257592309373e-06, + "loss": 3.3446, + "step": 7075 + }, + { + "epoch": 1.54, + "learning_rate": 4.232466681232248e-06, + "loss": 3.0224, + "step": 7076 + }, + { + "epoch": 1.54, + "learning_rate": 4.232357439370767e-06, + "loss": 3.0845, + "step": 7077 + }, + { + "epoch": 1.54, + "learning_rate": 4.232248197509286e-06, + "loss": 2.6396, + "step": 7078 + }, + { + "epoch": 1.54, + "learning_rate": 4.2321389556478045e-06, + "loss": 3.1618, + "step": 7079 + }, + { + "epoch": 1.55, + "learning_rate": 4.232029713786323e-06, + "loss": 3.0654, + "step": 7080 + }, + { + "epoch": 1.55, + "learning_rate": 4.231920471924842e-06, + "loss": 2.9774, + "step": 7081 + }, + { + "epoch": 1.55, + "learning_rate": 4.231811230063361e-06, + "loss": 2.7723, + "step": 7082 + }, + { + "epoch": 1.55, + "learning_rate": 4.23170198820188e-06, + "loss": 2.9783, + "step": 7083 + }, + { + "epoch": 1.55, + "learning_rate": 4.231592746340398e-06, + "loss": 2.9352, + "step": 7084 + }, + { + "epoch": 1.55, + "learning_rate": 4.231483504478916e-06, + "loss": 3.0001, + "step": 7085 + }, + { + "epoch": 1.55, + "learning_rate": 4.231374262617435e-06, + "loss": 3.5275, + "step": 7086 + }, + { + "epoch": 1.55, + "learning_rate": 4.231265020755954e-06, + "loss": 3.2321, + "step": 7087 + }, + { + "epoch": 1.55, + "learning_rate": 4.231155778894473e-06, + "loss": 2.7744, + "step": 7088 + }, + { + "epoch": 1.55, + "learning_rate": 4.2310465370329915e-06, + "loss": 1.981, + "step": 7089 + }, + { + "epoch": 1.55, + "learning_rate": 4.23093729517151e-06, + "loss": 2.3177, + "step": 7090 + }, + { + "epoch": 1.55, + "learning_rate": 4.230828053310029e-06, + "loss": 2.3076, + "step": 7091 + }, + { + "epoch": 1.55, + "learning_rate": 4.230718811448548e-06, + "loss": 3.7011, + "step": 7092 + }, + { + "epoch": 1.55, + "learning_rate": 4.230609569587067e-06, + "loss": 2.4945, + "step": 7093 + }, + { + "epoch": 1.55, + "learning_rate": 4.230500327725585e-06, + "loss": 2.6793, + "step": 7094 + }, + { + "epoch": 1.55, + "learning_rate": 4.230391085864103e-06, + "loss": 2.5272, + "step": 7095 + }, + { + "epoch": 1.55, + "learning_rate": 4.230281844002622e-06, + "loss": 3.6622, + "step": 7096 + }, + { + "epoch": 1.55, + "learning_rate": 4.230172602141141e-06, + "loss": 2.1322, + "step": 7097 + }, + { + "epoch": 1.55, + "learning_rate": 4.230063360279659e-06, + "loss": 2.2725, + "step": 7098 + }, + { + "epoch": 1.55, + "learning_rate": 4.2299541184181784e-06, + "loss": 2.5899, + "step": 7099 + }, + { + "epoch": 1.55, + "learning_rate": 4.229844876556697e-06, + "loss": 4.0867, + "step": 7100 + }, + { + "epoch": 1.55, + "learning_rate": 4.229735634695216e-06, + "loss": 2.9899, + "step": 7101 + }, + { + "epoch": 1.55, + "learning_rate": 4.229626392833735e-06, + "loss": 3.1661, + "step": 7102 + }, + { + "epoch": 1.55, + "learning_rate": 4.229517150972253e-06, + "loss": 3.1885, + "step": 7103 + }, + { + "epoch": 1.55, + "learning_rate": 4.229407909110772e-06, + "loss": 3.262, + "step": 7104 + }, + { + "epoch": 1.55, + "learning_rate": 4.22929866724929e-06, + "loss": 3.0212, + "step": 7105 + }, + { + "epoch": 1.55, + "learning_rate": 4.229189425387809e-06, + "loss": 3.6433, + "step": 7106 + }, + { + "epoch": 1.55, + "learning_rate": 4.229080183526328e-06, + "loss": 2.9605, + "step": 7107 + }, + { + "epoch": 1.55, + "learning_rate": 4.228970941664846e-06, + "loss": 1.2522, + "step": 7108 + }, + { + "epoch": 1.55, + "learning_rate": 4.2288616998033645e-06, + "loss": 3.139, + "step": 7109 + }, + { + "epoch": 1.55, + "learning_rate": 4.2287524579418836e-06, + "loss": 2.9566, + "step": 7110 + }, + { + "epoch": 1.55, + "learning_rate": 4.228643216080403e-06, + "loss": 3.1337, + "step": 7111 + }, + { + "epoch": 1.55, + "learning_rate": 4.228533974218922e-06, + "loss": 3.5555, + "step": 7112 + }, + { + "epoch": 1.55, + "learning_rate": 4.22842473235744e-06, + "loss": 3.1937, + "step": 7113 + }, + { + "epoch": 1.55, + "learning_rate": 4.228315490495958e-06, + "loss": 2.8413, + "step": 7114 + }, + { + "epoch": 1.55, + "learning_rate": 4.228206248634477e-06, + "loss": 3.1488, + "step": 7115 + }, + { + "epoch": 1.55, + "learning_rate": 4.228097006772996e-06, + "loss": 3.4277, + "step": 7116 + }, + { + "epoch": 1.55, + "learning_rate": 4.227987764911514e-06, + "loss": 2.1672, + "step": 7117 + }, + { + "epoch": 1.55, + "learning_rate": 4.227878523050033e-06, + "loss": 3.5212, + "step": 7118 + }, + { + "epoch": 1.55, + "learning_rate": 4.2277692811885515e-06, + "loss": 2.3649, + "step": 7119 + }, + { + "epoch": 1.55, + "learning_rate": 4.2276600393270705e-06, + "loss": 3.2833, + "step": 7120 + }, + { + "epoch": 1.55, + "learning_rate": 4.2275507974655896e-06, + "loss": 2.4226, + "step": 7121 + }, + { + "epoch": 1.55, + "learning_rate": 4.227441555604108e-06, + "loss": 2.0544, + "step": 7122 + }, + { + "epoch": 1.55, + "learning_rate": 4.227332313742627e-06, + "loss": 2.6854, + "step": 7123 + }, + { + "epoch": 1.55, + "learning_rate": 4.227223071881145e-06, + "loss": 2.6332, + "step": 7124 + }, + { + "epoch": 1.55, + "learning_rate": 4.227113830019664e-06, + "loss": 3.311, + "step": 7125 + }, + { + "epoch": 1.56, + "learning_rate": 4.227004588158183e-06, + "loss": 3.3074, + "step": 7126 + }, + { + "epoch": 1.56, + "learning_rate": 4.226895346296701e-06, + "loss": 3.1195, + "step": 7127 + }, + { + "epoch": 1.56, + "learning_rate": 4.22678610443522e-06, + "loss": 3.2801, + "step": 7128 + }, + { + "epoch": 1.56, + "learning_rate": 4.2266768625737384e-06, + "loss": 2.7572, + "step": 7129 + }, + { + "epoch": 1.56, + "learning_rate": 4.2265676207122575e-06, + "loss": 2.5023, + "step": 7130 + }, + { + "epoch": 1.56, + "learning_rate": 4.226458378850776e-06, + "loss": 3.4399, + "step": 7131 + }, + { + "epoch": 1.56, + "learning_rate": 4.226349136989295e-06, + "loss": 2.2506, + "step": 7132 + }, + { + "epoch": 1.56, + "learning_rate": 4.226239895127813e-06, + "loss": 3.588, + "step": 7133 + }, + { + "epoch": 1.56, + "learning_rate": 4.226130653266332e-06, + "loss": 3.079, + "step": 7134 + }, + { + "epoch": 1.56, + "learning_rate": 4.226021411404851e-06, + "loss": 3.6955, + "step": 7135 + }, + { + "epoch": 1.56, + "learning_rate": 4.225912169543369e-06, + "loss": 2.6696, + "step": 7136 + }, + { + "epoch": 1.56, + "learning_rate": 4.225802927681888e-06, + "loss": 3.7924, + "step": 7137 + }, + { + "epoch": 1.56, + "learning_rate": 4.225693685820406e-06, + "loss": 3.2907, + "step": 7138 + }, + { + "epoch": 1.56, + "learning_rate": 4.225584443958925e-06, + "loss": 3.3722, + "step": 7139 + }, + { + "epoch": 1.56, + "learning_rate": 4.2254752020974444e-06, + "loss": 2.9872, + "step": 7140 + }, + { + "epoch": 1.56, + "learning_rate": 4.225365960235963e-06, + "loss": 2.7079, + "step": 7141 + }, + { + "epoch": 1.56, + "learning_rate": 4.225256718374482e-06, + "loss": 2.7641, + "step": 7142 + }, + { + "epoch": 1.56, + "learning_rate": 4.225147476513e-06, + "loss": 3.4228, + "step": 7143 + }, + { + "epoch": 1.56, + "learning_rate": 4.225038234651519e-06, + "loss": 2.4987, + "step": 7144 + }, + { + "epoch": 1.56, + "learning_rate": 4.224928992790038e-06, + "loss": 3.0315, + "step": 7145 + }, + { + "epoch": 1.56, + "learning_rate": 4.224819750928556e-06, + "loss": 2.4642, + "step": 7146 + }, + { + "epoch": 1.56, + "learning_rate": 4.224710509067075e-06, + "loss": 2.2637, + "step": 7147 + }, + { + "epoch": 1.56, + "learning_rate": 4.224601267205593e-06, + "loss": 2.8294, + "step": 7148 + }, + { + "epoch": 1.56, + "learning_rate": 4.224492025344112e-06, + "loss": 1.9928, + "step": 7149 + }, + { + "epoch": 1.56, + "learning_rate": 4.2243827834826305e-06, + "loss": 2.9942, + "step": 7150 + }, + { + "epoch": 1.56, + "learning_rate": 4.22427354162115e-06, + "loss": 3.1355, + "step": 7151 + }, + { + "epoch": 1.56, + "learning_rate": 4.224164299759669e-06, + "loss": 2.8501, + "step": 7152 + }, + { + "epoch": 1.56, + "learning_rate": 4.224055057898187e-06, + "loss": 2.8162, + "step": 7153 + }, + { + "epoch": 1.56, + "learning_rate": 4.223945816036706e-06, + "loss": 2.5415, + "step": 7154 + }, + { + "epoch": 1.56, + "learning_rate": 4.223836574175224e-06, + "loss": 1.9295, + "step": 7155 + }, + { + "epoch": 1.56, + "learning_rate": 4.223727332313743e-06, + "loss": 2.9867, + "step": 7156 + }, + { + "epoch": 1.56, + "learning_rate": 4.223618090452262e-06, + "loss": 2.3639, + "step": 7157 + }, + { + "epoch": 1.56, + "learning_rate": 4.22350884859078e-06, + "loss": 2.6469, + "step": 7158 + }, + { + "epoch": 1.56, + "learning_rate": 4.223399606729299e-06, + "loss": 2.6408, + "step": 7159 + }, + { + "epoch": 1.56, + "learning_rate": 4.2232903648678175e-06, + "loss": 3.4346, + "step": 7160 + }, + { + "epoch": 1.56, + "learning_rate": 4.2231811230063365e-06, + "loss": 3.0505, + "step": 7161 + }, + { + "epoch": 1.56, + "learning_rate": 4.223071881144855e-06, + "loss": 3.0457, + "step": 7162 + }, + { + "epoch": 1.56, + "learning_rate": 4.222962639283374e-06, + "loss": 2.4391, + "step": 7163 + }, + { + "epoch": 1.56, + "learning_rate": 4.222853397421892e-06, + "loss": 3.3587, + "step": 7164 + }, + { + "epoch": 1.56, + "learning_rate": 4.222744155560411e-06, + "loss": 3.5301, + "step": 7165 + }, + { + "epoch": 1.56, + "learning_rate": 4.22263491369893e-06, + "loss": 1.8948, + "step": 7166 + }, + { + "epoch": 1.56, + "learning_rate": 4.222525671837448e-06, + "loss": 3.0656, + "step": 7167 + }, + { + "epoch": 1.56, + "learning_rate": 4.222416429975967e-06, + "loss": 3.2615, + "step": 7168 + }, + { + "epoch": 1.56, + "learning_rate": 4.2223071881144854e-06, + "loss": 2.2521, + "step": 7169 + }, + { + "epoch": 1.56, + "learning_rate": 4.2221979462530045e-06, + "loss": 3.0328, + "step": 7170 + }, + { + "epoch": 1.57, + "learning_rate": 4.2220887043915235e-06, + "loss": 2.2718, + "step": 7171 + }, + { + "epoch": 1.57, + "learning_rate": 4.221979462530042e-06, + "loss": 2.1044, + "step": 7172 + }, + { + "epoch": 1.57, + "learning_rate": 4.221870220668561e-06, + "loss": 2.6863, + "step": 7173 + }, + { + "epoch": 1.57, + "learning_rate": 4.221760978807079e-06, + "loss": 3.2868, + "step": 7174 + }, + { + "epoch": 1.57, + "learning_rate": 4.221651736945598e-06, + "loss": 1.7192, + "step": 7175 + }, + { + "epoch": 1.57, + "learning_rate": 4.221542495084117e-06, + "loss": 3.8161, + "step": 7176 + }, + { + "epoch": 1.57, + "learning_rate": 4.221433253222635e-06, + "loss": 3.0323, + "step": 7177 + }, + { + "epoch": 1.57, + "learning_rate": 4.221324011361153e-06, + "loss": 3.5826, + "step": 7178 + }, + { + "epoch": 1.57, + "learning_rate": 4.221214769499672e-06, + "loss": 3.2202, + "step": 7179 + }, + { + "epoch": 1.57, + "learning_rate": 4.221105527638191e-06, + "loss": 3.3151, + "step": 7180 + }, + { + "epoch": 1.57, + "learning_rate": 4.2209962857767105e-06, + "loss": 3.7361, + "step": 7181 + }, + { + "epoch": 1.57, + "learning_rate": 4.220887043915229e-06, + "loss": 3.4474, + "step": 7182 + }, + { + "epoch": 1.57, + "learning_rate": 4.220777802053747e-06, + "loss": 2.0956, + "step": 7183 + }, + { + "epoch": 1.57, + "learning_rate": 4.220668560192266e-06, + "loss": 3.5858, + "step": 7184 + }, + { + "epoch": 1.57, + "learning_rate": 4.220559318330785e-06, + "loss": 2.9316, + "step": 7185 + }, + { + "epoch": 1.57, + "learning_rate": 4.220450076469303e-06, + "loss": 2.3981, + "step": 7186 + }, + { + "epoch": 1.57, + "learning_rate": 4.220340834607822e-06, + "loss": 2.7127, + "step": 7187 + }, + { + "epoch": 1.57, + "learning_rate": 4.22023159274634e-06, + "loss": 2.9426, + "step": 7188 + }, + { + "epoch": 1.57, + "learning_rate": 4.220122350884859e-06, + "loss": 2.7269, + "step": 7189 + }, + { + "epoch": 1.57, + "learning_rate": 4.220013109023378e-06, + "loss": 2.6437, + "step": 7190 + }, + { + "epoch": 1.57, + "learning_rate": 4.2199038671618966e-06, + "loss": 2.7395, + "step": 7191 + }, + { + "epoch": 1.57, + "learning_rate": 4.219794625300416e-06, + "loss": 2.9934, + "step": 7192 + }, + { + "epoch": 1.57, + "learning_rate": 4.219685383438934e-06, + "loss": 2.4939, + "step": 7193 + }, + { + "epoch": 1.57, + "learning_rate": 4.219576141577453e-06, + "loss": 3.1421, + "step": 7194 + }, + { + "epoch": 1.57, + "learning_rate": 4.219466899715972e-06, + "loss": 2.7126, + "step": 7195 + }, + { + "epoch": 1.57, + "learning_rate": 4.21935765785449e-06, + "loss": 3.2397, + "step": 7196 + }, + { + "epoch": 1.57, + "learning_rate": 4.219248415993008e-06, + "loss": 2.8489, + "step": 7197 + }, + { + "epoch": 1.57, + "learning_rate": 4.219139174131527e-06, + "loss": 3.2875, + "step": 7198 + }, + { + "epoch": 1.57, + "learning_rate": 4.219029932270046e-06, + "loss": 3.0578, + "step": 7199 + }, + { + "epoch": 1.57, + "learning_rate": 4.218920690408565e-06, + "loss": 2.8882, + "step": 7200 + }, + { + "epoch": 1.57, + "learning_rate": 4.2188114485470835e-06, + "loss": 2.2451, + "step": 7201 + }, + { + "epoch": 1.57, + "learning_rate": 4.218702206685602e-06, + "loss": 2.8653, + "step": 7202 + }, + { + "epoch": 1.57, + "learning_rate": 4.218592964824121e-06, + "loss": 3.0711, + "step": 7203 + }, + { + "epoch": 1.57, + "learning_rate": 4.21848372296264e-06, + "loss": 2.1744, + "step": 7204 + }, + { + "epoch": 1.57, + "learning_rate": 4.218374481101159e-06, + "loss": 2.3607, + "step": 7205 + }, + { + "epoch": 1.57, + "learning_rate": 4.218265239239677e-06, + "loss": 3.1968, + "step": 7206 + }, + { + "epoch": 1.57, + "learning_rate": 4.218155997378195e-06, + "loss": 2.5773, + "step": 7207 + }, + { + "epoch": 1.57, + "learning_rate": 4.218046755516714e-06, + "loss": 2.8446, + "step": 7208 + }, + { + "epoch": 1.57, + "learning_rate": 4.217937513655233e-06, + "loss": 2.2074, + "step": 7209 + }, + { + "epoch": 1.57, + "learning_rate": 4.217828271793752e-06, + "loss": 3.3232, + "step": 7210 + }, + { + "epoch": 1.57, + "learning_rate": 4.2177190299322705e-06, + "loss": 3.0131, + "step": 7211 + }, + { + "epoch": 1.57, + "learning_rate": 4.217609788070789e-06, + "loss": 2.9176, + "step": 7212 + }, + { + "epoch": 1.57, + "learning_rate": 4.217500546209308e-06, + "loss": 3.4187, + "step": 7213 + }, + { + "epoch": 1.57, + "learning_rate": 4.217391304347827e-06, + "loss": 2.7486, + "step": 7214 + }, + { + "epoch": 1.57, + "learning_rate": 4.217282062486345e-06, + "loss": 1.5694, + "step": 7215 + }, + { + "epoch": 1.57, + "learning_rate": 4.217172820624864e-06, + "loss": 2.8647, + "step": 7216 + }, + { + "epoch": 1.58, + "learning_rate": 4.217063578763382e-06, + "loss": 1.7558, + "step": 7217 + }, + { + "epoch": 1.58, + "learning_rate": 4.216954336901901e-06, + "loss": 2.3149, + "step": 7218 + }, + { + "epoch": 1.58, + "learning_rate": 4.21684509504042e-06, + "loss": 2.9656, + "step": 7219 + }, + { + "epoch": 1.58, + "learning_rate": 4.216735853178938e-06, + "loss": 2.2299, + "step": 7220 + }, + { + "epoch": 1.58, + "learning_rate": 4.216626611317457e-06, + "loss": 2.3636, + "step": 7221 + }, + { + "epoch": 1.58, + "learning_rate": 4.216517369455976e-06, + "loss": 2.8247, + "step": 7222 + }, + { + "epoch": 1.58, + "learning_rate": 4.216408127594495e-06, + "loss": 3.0776, + "step": 7223 + }, + { + "epoch": 1.58, + "learning_rate": 4.216298885733014e-06, + "loss": 2.1537, + "step": 7224 + }, + { + "epoch": 1.58, + "learning_rate": 4.216189643871532e-06, + "loss": 3.1299, + "step": 7225 + }, + { + "epoch": 1.58, + "learning_rate": 4.21608040201005e-06, + "loss": 2.3312, + "step": 7226 + }, + { + "epoch": 1.58, + "learning_rate": 4.215971160148569e-06, + "loss": 2.0351, + "step": 7227 + }, + { + "epoch": 1.58, + "learning_rate": 4.215861918287088e-06, + "loss": 2.9551, + "step": 7228 + }, + { + "epoch": 1.58, + "learning_rate": 4.215752676425607e-06, + "loss": 3.3424, + "step": 7229 + }, + { + "epoch": 1.58, + "learning_rate": 4.215643434564125e-06, + "loss": 3.1863, + "step": 7230 + }, + { + "epoch": 1.58, + "learning_rate": 4.2155341927026435e-06, + "loss": 2.9511, + "step": 7231 + }, + { + "epoch": 1.58, + "learning_rate": 4.215424950841163e-06, + "loss": 3.216, + "step": 7232 + }, + { + "epoch": 1.58, + "learning_rate": 4.215315708979682e-06, + "loss": 2.4008, + "step": 7233 + }, + { + "epoch": 1.58, + "learning_rate": 4.215206467118201e-06, + "loss": 2.7455, + "step": 7234 + }, + { + "epoch": 1.58, + "learning_rate": 4.215097225256719e-06, + "loss": 2.659, + "step": 7235 + }, + { + "epoch": 1.58, + "learning_rate": 4.214987983395237e-06, + "loss": 2.6507, + "step": 7236 + }, + { + "epoch": 1.58, + "learning_rate": 4.214878741533756e-06, + "loss": 2.4628, + "step": 7237 + }, + { + "epoch": 1.58, + "learning_rate": 4.214769499672275e-06, + "loss": 2.5547, + "step": 7238 + }, + { + "epoch": 1.58, + "learning_rate": 4.214660257810793e-06, + "loss": 2.6073, + "step": 7239 + }, + { + "epoch": 1.58, + "learning_rate": 4.214551015949312e-06, + "loss": 3.0264, + "step": 7240 + }, + { + "epoch": 1.58, + "learning_rate": 4.2144417740878305e-06, + "loss": 3.779, + "step": 7241 + }, + { + "epoch": 1.58, + "learning_rate": 4.2143325322263495e-06, + "loss": 2.8341, + "step": 7242 + }, + { + "epoch": 1.58, + "learning_rate": 4.2142232903648686e-06, + "loss": 2.261, + "step": 7243 + }, + { + "epoch": 1.58, + "learning_rate": 4.214114048503387e-06, + "loss": 3.1031, + "step": 7244 + }, + { + "epoch": 1.58, + "learning_rate": 4.214004806641905e-06, + "loss": 2.7635, + "step": 7245 + }, + { + "epoch": 1.58, + "learning_rate": 4.213895564780424e-06, + "loss": 3.1164, + "step": 7246 + }, + { + "epoch": 1.58, + "learning_rate": 4.213786322918943e-06, + "loss": 2.1544, + "step": 7247 + }, + { + "epoch": 1.58, + "learning_rate": 4.213677081057462e-06, + "loss": 3.1161, + "step": 7248 + }, + { + "epoch": 1.58, + "learning_rate": 4.21356783919598e-06, + "loss": 2.3059, + "step": 7249 + }, + { + "epoch": 1.58, + "learning_rate": 4.2134585973344984e-06, + "loss": 3.4295, + "step": 7250 + }, + { + "epoch": 1.58, + "learning_rate": 4.2133493554730175e-06, + "loss": 2.916, + "step": 7251 + }, + { + "epoch": 1.58, + "learning_rate": 4.2132401136115365e-06, + "loss": 3.4132, + "step": 7252 + }, + { + "epoch": 1.58, + "learning_rate": 4.2131308717500555e-06, + "loss": 2.7867, + "step": 7253 + }, + { + "epoch": 1.58, + "learning_rate": 4.213021629888574e-06, + "loss": 2.9214, + "step": 7254 + }, + { + "epoch": 1.58, + "learning_rate": 4.212912388027092e-06, + "loss": 3.4457, + "step": 7255 + }, + { + "epoch": 1.58, + "learning_rate": 4.212803146165611e-06, + "loss": 3.4258, + "step": 7256 + }, + { + "epoch": 1.58, + "learning_rate": 4.21269390430413e-06, + "loss": 3.0152, + "step": 7257 + }, + { + "epoch": 1.58, + "learning_rate": 4.212584662442649e-06, + "loss": 2.7143, + "step": 7258 + }, + { + "epoch": 1.58, + "learning_rate": 4.212475420581167e-06, + "loss": 3.5601, + "step": 7259 + }, + { + "epoch": 1.58, + "learning_rate": 4.212366178719685e-06, + "loss": 3.236, + "step": 7260 + }, + { + "epoch": 1.58, + "learning_rate": 4.212256936858204e-06, + "loss": 2.5359, + "step": 7261 + }, + { + "epoch": 1.58, + "learning_rate": 4.2121476949967234e-06, + "loss": 2.3567, + "step": 7262 + }, + { + "epoch": 1.59, + "learning_rate": 4.212038453135242e-06, + "loss": 3.4511, + "step": 7263 + }, + { + "epoch": 1.59, + "learning_rate": 4.211929211273761e-06, + "loss": 2.2452, + "step": 7264 + }, + { + "epoch": 1.59, + "learning_rate": 4.211819969412279e-06, + "loss": 3.2663, + "step": 7265 + }, + { + "epoch": 1.59, + "learning_rate": 4.211710727550798e-06, + "loss": 2.8975, + "step": 7266 + }, + { + "epoch": 1.59, + "learning_rate": 4.211601485689317e-06, + "loss": 3.0978, + "step": 7267 + }, + { + "epoch": 1.59, + "learning_rate": 4.211492243827835e-06, + "loss": 2.7068, + "step": 7268 + }, + { + "epoch": 1.59, + "learning_rate": 4.211383001966354e-06, + "loss": 2.9539, + "step": 7269 + }, + { + "epoch": 1.59, + "learning_rate": 4.211273760104872e-06, + "loss": 3.9541, + "step": 7270 + }, + { + "epoch": 1.59, + "learning_rate": 4.211164518243391e-06, + "loss": 2.5229, + "step": 7271 + }, + { + "epoch": 1.59, + "learning_rate": 4.21105527638191e-06, + "loss": 3.5322, + "step": 7272 + }, + { + "epoch": 1.59, + "learning_rate": 4.210946034520429e-06, + "loss": 2.9735, + "step": 7273 + }, + { + "epoch": 1.59, + "learning_rate": 4.210836792658947e-06, + "loss": 3.2607, + "step": 7274 + }, + { + "epoch": 1.59, + "learning_rate": 4.210727550797466e-06, + "loss": 2.6457, + "step": 7275 + }, + { + "epoch": 1.59, + "learning_rate": 4.210618308935985e-06, + "loss": 2.8214, + "step": 7276 + }, + { + "epoch": 1.59, + "learning_rate": 4.210509067074503e-06, + "loss": 3.8217, + "step": 7277 + }, + { + "epoch": 1.59, + "learning_rate": 4.210399825213022e-06, + "loss": 3.6623, + "step": 7278 + }, + { + "epoch": 1.59, + "learning_rate": 4.21029058335154e-06, + "loss": 2.3717, + "step": 7279 + }, + { + "epoch": 1.59, + "learning_rate": 4.210181341490059e-06, + "loss": 2.2463, + "step": 7280 + }, + { + "epoch": 1.59, + "learning_rate": 4.210072099628578e-06, + "loss": 3.0245, + "step": 7281 + }, + { + "epoch": 1.59, + "learning_rate": 4.2099628577670965e-06, + "loss": 3.0074, + "step": 7282 + }, + { + "epoch": 1.59, + "learning_rate": 4.2098536159056156e-06, + "loss": 3.7026, + "step": 7283 + }, + { + "epoch": 1.59, + "learning_rate": 4.209744374044134e-06, + "loss": 3.6059, + "step": 7284 + }, + { + "epoch": 1.59, + "learning_rate": 4.209635132182653e-06, + "loss": 3.2798, + "step": 7285 + }, + { + "epoch": 1.59, + "learning_rate": 4.209525890321172e-06, + "loss": 2.7727, + "step": 7286 + }, + { + "epoch": 1.59, + "learning_rate": 4.20941664845969e-06, + "loss": 2.5356, + "step": 7287 + }, + { + "epoch": 1.59, + "learning_rate": 4.209307406598209e-06, + "loss": 3.622, + "step": 7288 + }, + { + "epoch": 1.59, + "learning_rate": 4.209198164736727e-06, + "loss": 2.8688, + "step": 7289 + }, + { + "epoch": 1.59, + "learning_rate": 4.209088922875246e-06, + "loss": 3.008, + "step": 7290 + }, + { + "epoch": 1.59, + "learning_rate": 4.208979681013765e-06, + "loss": 2.3792, + "step": 7291 + }, + { + "epoch": 1.59, + "learning_rate": 4.2088704391522835e-06, + "loss": 2.489, + "step": 7292 + }, + { + "epoch": 1.59, + "learning_rate": 4.2087611972908025e-06, + "loss": 3.3617, + "step": 7293 + }, + { + "epoch": 1.59, + "learning_rate": 4.208651955429321e-06, + "loss": 3.3953, + "step": 7294 + }, + { + "epoch": 1.59, + "learning_rate": 4.20854271356784e-06, + "loss": 2.4626, + "step": 7295 + }, + { + "epoch": 1.59, + "learning_rate": 4.208433471706358e-06, + "loss": 2.5151, + "step": 7296 + }, + { + "epoch": 1.59, + "learning_rate": 4.208324229844877e-06, + "loss": 2.9712, + "step": 7297 + }, + { + "epoch": 1.59, + "learning_rate": 4.208214987983395e-06, + "loss": 3.1904, + "step": 7298 + }, + { + "epoch": 1.59, + "learning_rate": 4.208105746121914e-06, + "loss": 3.6619, + "step": 7299 + }, + { + "epoch": 1.59, + "learning_rate": 4.207996504260433e-06, + "loss": 4.1277, + "step": 7300 + }, + { + "epoch": 1.59, + "learning_rate": 4.207887262398951e-06, + "loss": 3.0082, + "step": 7301 + }, + { + "epoch": 1.59, + "learning_rate": 4.2077780205374704e-06, + "loss": 2.7608, + "step": 7302 + }, + { + "epoch": 1.59, + "learning_rate": 4.207668778675989e-06, + "loss": 2.882, + "step": 7303 + }, + { + "epoch": 1.59, + "learning_rate": 4.207559536814508e-06, + "loss": 3.0847, + "step": 7304 + }, + { + "epoch": 1.59, + "learning_rate": 4.207450294953027e-06, + "loss": 2.5387, + "step": 7305 + }, + { + "epoch": 1.59, + "learning_rate": 4.207341053091545e-06, + "loss": 2.4637, + "step": 7306 + }, + { + "epoch": 1.59, + "learning_rate": 4.207231811230064e-06, + "loss": 2.0514, + "step": 7307 + }, + { + "epoch": 1.59, + "learning_rate": 4.207122569368582e-06, + "loss": 2.7846, + "step": 7308 + }, + { + "epoch": 1.6, + "learning_rate": 4.207013327507101e-06, + "loss": 2.9656, + "step": 7309 + }, + { + "epoch": 1.6, + "learning_rate": 4.206904085645619e-06, + "loss": 2.7045, + "step": 7310 + }, + { + "epoch": 1.6, + "learning_rate": 4.206794843784138e-06, + "loss": 2.9182, + "step": 7311 + }, + { + "epoch": 1.6, + "learning_rate": 4.206685601922657e-06, + "loss": 4.1197, + "step": 7312 + }, + { + "epoch": 1.6, + "learning_rate": 4.2065763600611756e-06, + "loss": 2.6479, + "step": 7313 + }, + { + "epoch": 1.6, + "learning_rate": 4.206467118199695e-06, + "loss": 3.6022, + "step": 7314 + }, + { + "epoch": 1.6, + "learning_rate": 4.206357876338213e-06, + "loss": 2.0276, + "step": 7315 + }, + { + "epoch": 1.6, + "learning_rate": 4.206248634476732e-06, + "loss": 2.9767, + "step": 7316 + }, + { + "epoch": 1.6, + "learning_rate": 4.206139392615251e-06, + "loss": 2.4511, + "step": 7317 + }, + { + "epoch": 1.6, + "learning_rate": 4.206030150753769e-06, + "loss": 2.6093, + "step": 7318 + }, + { + "epoch": 1.6, + "learning_rate": 4.205920908892288e-06, + "loss": 2.6546, + "step": 7319 + }, + { + "epoch": 1.6, + "learning_rate": 4.205811667030806e-06, + "loss": 1.6069, + "step": 7320 + }, + { + "epoch": 1.6, + "learning_rate": 4.205702425169325e-06, + "loss": 3.1064, + "step": 7321 + }, + { + "epoch": 1.6, + "learning_rate": 4.205593183307844e-06, + "loss": 2.633, + "step": 7322 + }, + { + "epoch": 1.6, + "learning_rate": 4.2054839414463625e-06, + "loss": 3.6196, + "step": 7323 + }, + { + "epoch": 1.6, + "learning_rate": 4.2053746995848816e-06, + "loss": 2.9078, + "step": 7324 + }, + { + "epoch": 1.6, + "learning_rate": 4.2052654577234e-06, + "loss": 2.4445, + "step": 7325 + }, + { + "epoch": 1.6, + "learning_rate": 4.205156215861919e-06, + "loss": 1.9793, + "step": 7326 + }, + { + "epoch": 1.6, + "learning_rate": 4.205046974000437e-06, + "loss": 2.3482, + "step": 7327 + }, + { + "epoch": 1.6, + "learning_rate": 4.204937732138956e-06, + "loss": 3.2276, + "step": 7328 + }, + { + "epoch": 1.6, + "learning_rate": 4.204828490277474e-06, + "loss": 3.1099, + "step": 7329 + }, + { + "epoch": 1.6, + "learning_rate": 4.204719248415993e-06, + "loss": 3.4631, + "step": 7330 + }, + { + "epoch": 1.6, + "learning_rate": 4.204610006554512e-06, + "loss": 3.1752, + "step": 7331 + }, + { + "epoch": 1.6, + "learning_rate": 4.2045007646930305e-06, + "loss": 3.1878, + "step": 7332 + }, + { + "epoch": 1.6, + "learning_rate": 4.2043915228315495e-06, + "loss": 3.3217, + "step": 7333 + }, + { + "epoch": 1.6, + "learning_rate": 4.204282280970068e-06, + "loss": 3.5378, + "step": 7334 + }, + { + "epoch": 1.6, + "learning_rate": 4.204173039108587e-06, + "loss": 3.2086, + "step": 7335 + }, + { + "epoch": 1.6, + "learning_rate": 4.204063797247106e-06, + "loss": 3.56, + "step": 7336 + }, + { + "epoch": 1.6, + "learning_rate": 4.203954555385624e-06, + "loss": 2.5544, + "step": 7337 + }, + { + "epoch": 1.6, + "learning_rate": 4.203845313524143e-06, + "loss": 2.8839, + "step": 7338 + }, + { + "epoch": 1.6, + "learning_rate": 4.203736071662661e-06, + "loss": 2.7144, + "step": 7339 + }, + { + "epoch": 1.6, + "learning_rate": 4.20362682980118e-06, + "loss": 3.097, + "step": 7340 + }, + { + "epoch": 1.6, + "learning_rate": 4.203517587939699e-06, + "loss": 3.046, + "step": 7341 + }, + { + "epoch": 1.6, + "learning_rate": 4.203408346078217e-06, + "loss": 3.5875, + "step": 7342 + }, + { + "epoch": 1.6, + "learning_rate": 4.203299104216736e-06, + "loss": 2.0515, + "step": 7343 + }, + { + "epoch": 1.6, + "learning_rate": 4.203189862355255e-06, + "loss": 3.4551, + "step": 7344 + }, + { + "epoch": 1.6, + "learning_rate": 4.203080620493774e-06, + "loss": 1.8425, + "step": 7345 + }, + { + "epoch": 1.6, + "learning_rate": 4.202971378632293e-06, + "loss": 2.4419, + "step": 7346 + }, + { + "epoch": 1.6, + "learning_rate": 4.202862136770811e-06, + "loss": 2.1628, + "step": 7347 + }, + { + "epoch": 1.6, + "learning_rate": 4.202752894909329e-06, + "loss": 2.1147, + "step": 7348 + }, + { + "epoch": 1.6, + "learning_rate": 4.202643653047848e-06, + "loss": 2.9875, + "step": 7349 + }, + { + "epoch": 1.6, + "learning_rate": 4.202534411186367e-06, + "loss": 2.701, + "step": 7350 + }, + { + "epoch": 1.6, + "learning_rate": 4.202425169324885e-06, + "loss": 2.5714, + "step": 7351 + }, + { + "epoch": 1.6, + "learning_rate": 4.202315927463404e-06, + "loss": 2.7518, + "step": 7352 + }, + { + "epoch": 1.6, + "learning_rate": 4.2022066856019226e-06, + "loss": 2.0171, + "step": 7353 + }, + { + "epoch": 1.6, + "learning_rate": 4.202097443740442e-06, + "loss": 4.1548, + "step": 7354 + }, + { + "epoch": 1.61, + "learning_rate": 4.201988201878961e-06, + "loss": 1.6953, + "step": 7355 + }, + { + "epoch": 1.61, + "learning_rate": 4.201878960017479e-06, + "loss": 2.756, + "step": 7356 + }, + { + "epoch": 1.61, + "learning_rate": 4.201769718155998e-06, + "loss": 2.9808, + "step": 7357 + }, + { + "epoch": 1.61, + "learning_rate": 4.201660476294516e-06, + "loss": 3.198, + "step": 7358 + }, + { + "epoch": 1.61, + "learning_rate": 4.201551234433035e-06, + "loss": 3.021, + "step": 7359 + }, + { + "epoch": 1.61, + "learning_rate": 4.201441992571554e-06, + "loss": 2.5054, + "step": 7360 + }, + { + "epoch": 1.61, + "learning_rate": 4.201332750710072e-06, + "loss": 3.0968, + "step": 7361 + }, + { + "epoch": 1.61, + "learning_rate": 4.2012235088485905e-06, + "loss": 2.7933, + "step": 7362 + }, + { + "epoch": 1.61, + "learning_rate": 4.2011142669871095e-06, + "loss": 3.1551, + "step": 7363 + }, + { + "epoch": 1.61, + "learning_rate": 4.2010050251256285e-06, + "loss": 2.1914, + "step": 7364 + }, + { + "epoch": 1.61, + "learning_rate": 4.200895783264148e-06, + "loss": 2.9837, + "step": 7365 + }, + { + "epoch": 1.61, + "learning_rate": 4.200786541402666e-06, + "loss": 3.5064, + "step": 7366 + }, + { + "epoch": 1.61, + "learning_rate": 4.200677299541184e-06, + "loss": 2.9035, + "step": 7367 + }, + { + "epoch": 1.61, + "learning_rate": 4.200568057679703e-06, + "loss": 3.0134, + "step": 7368 + }, + { + "epoch": 1.61, + "learning_rate": 4.200458815818222e-06, + "loss": 2.603, + "step": 7369 + }, + { + "epoch": 1.61, + "learning_rate": 4.200349573956741e-06, + "loss": 3.3933, + "step": 7370 + }, + { + "epoch": 1.61, + "learning_rate": 4.200240332095259e-06, + "loss": 3.8591, + "step": 7371 + }, + { + "epoch": 1.61, + "learning_rate": 4.2001310902337774e-06, + "loss": 2.4782, + "step": 7372 + }, + { + "epoch": 1.61, + "learning_rate": 4.2000218483722965e-06, + "loss": 2.6, + "step": 7373 + }, + { + "epoch": 1.61, + "learning_rate": 4.1999126065108155e-06, + "loss": 3.7023, + "step": 7374 + }, + { + "epoch": 1.61, + "learning_rate": 4.1998033646493345e-06, + "loss": 2.3881, + "step": 7375 + }, + { + "epoch": 1.61, + "learning_rate": 4.199694122787853e-06, + "loss": 3.6291, + "step": 7376 + }, + { + "epoch": 1.61, + "learning_rate": 4.199584880926371e-06, + "loss": 3.7928, + "step": 7377 + }, + { + "epoch": 1.61, + "learning_rate": 4.19947563906489e-06, + "loss": 2.178, + "step": 7378 + }, + { + "epoch": 1.61, + "learning_rate": 4.199366397203409e-06, + "loss": 2.4263, + "step": 7379 + }, + { + "epoch": 1.61, + "learning_rate": 4.199257155341927e-06, + "loss": 2.7671, + "step": 7380 + }, + { + "epoch": 1.61, + "learning_rate": 4.199147913480446e-06, + "loss": 3.5363, + "step": 7381 + }, + { + "epoch": 1.61, + "learning_rate": 4.199038671618964e-06, + "loss": 2.1168, + "step": 7382 + }, + { + "epoch": 1.61, + "learning_rate": 4.1989294297574834e-06, + "loss": 2.7454, + "step": 7383 + }, + { + "epoch": 1.61, + "learning_rate": 4.1988201878960025e-06, + "loss": 2.089, + "step": 7384 + }, + { + "epoch": 1.61, + "learning_rate": 4.198710946034521e-06, + "loss": 3.0831, + "step": 7385 + }, + { + "epoch": 1.61, + "learning_rate": 4.198601704173039e-06, + "loss": 2.1168, + "step": 7386 + }, + { + "epoch": 1.61, + "learning_rate": 4.198492462311558e-06, + "loss": 2.2832, + "step": 7387 + }, + { + "epoch": 1.61, + "learning_rate": 4.198383220450077e-06, + "loss": 1.8444, + "step": 7388 + }, + { + "epoch": 1.61, + "learning_rate": 4.198273978588596e-06, + "loss": 2.545, + "step": 7389 + }, + { + "epoch": 1.61, + "learning_rate": 4.198164736727114e-06, + "loss": 3.4405, + "step": 7390 + }, + { + "epoch": 1.61, + "learning_rate": 4.198055494865632e-06, + "loss": 3.1789, + "step": 7391 + }, + { + "epoch": 1.61, + "learning_rate": 4.197946253004151e-06, + "loss": 2.3062, + "step": 7392 + }, + { + "epoch": 1.61, + "learning_rate": 4.19783701114267e-06, + "loss": 2.3473, + "step": 7393 + }, + { + "epoch": 1.61, + "learning_rate": 4.197727769281189e-06, + "loss": 2.2407, + "step": 7394 + }, + { + "epoch": 1.61, + "learning_rate": 4.197618527419708e-06, + "loss": 3.3272, + "step": 7395 + }, + { + "epoch": 1.61, + "learning_rate": 4.197509285558226e-06, + "loss": 3.0508, + "step": 7396 + }, + { + "epoch": 1.61, + "learning_rate": 4.197400043696745e-06, + "loss": 2.7529, + "step": 7397 + }, + { + "epoch": 1.61, + "learning_rate": 4.197290801835264e-06, + "loss": 3.489, + "step": 7398 + }, + { + "epoch": 1.61, + "learning_rate": 4.197181559973783e-06, + "loss": 3.1598, + "step": 7399 + }, + { + "epoch": 1.62, + "learning_rate": 4.197072318112301e-06, + "loss": 3.1249, + "step": 7400 + }, + { + "epoch": 1.62, + "learning_rate": 4.196963076250819e-06, + "loss": 3.5551, + "step": 7401 + }, + { + "epoch": 1.62, + "learning_rate": 4.196853834389338e-06, + "loss": 2.689, + "step": 7402 + }, + { + "epoch": 1.62, + "learning_rate": 4.196744592527857e-06, + "loss": 3.3763, + "step": 7403 + }, + { + "epoch": 1.62, + "learning_rate": 4.196635350666376e-06, + "loss": 2.5143, + "step": 7404 + }, + { + "epoch": 1.62, + "learning_rate": 4.1965261088048946e-06, + "loss": 3.637, + "step": 7405 + }, + { + "epoch": 1.62, + "learning_rate": 4.196416866943413e-06, + "loss": 3.525, + "step": 7406 + }, + { + "epoch": 1.62, + "learning_rate": 4.196307625081932e-06, + "loss": 2.2298, + "step": 7407 + }, + { + "epoch": 1.62, + "learning_rate": 4.196198383220451e-06, + "loss": 2.529, + "step": 7408 + }, + { + "epoch": 1.62, + "learning_rate": 4.196089141358969e-06, + "loss": 1.7608, + "step": 7409 + }, + { + "epoch": 1.62, + "learning_rate": 4.195979899497488e-06, + "loss": 3.1092, + "step": 7410 + }, + { + "epoch": 1.62, + "learning_rate": 4.195870657636006e-06, + "loss": 2.9712, + "step": 7411 + }, + { + "epoch": 1.62, + "learning_rate": 4.195761415774525e-06, + "loss": 3.9105, + "step": 7412 + }, + { + "epoch": 1.62, + "learning_rate": 4.195652173913044e-06, + "loss": 3.7379, + "step": 7413 + }, + { + "epoch": 1.62, + "learning_rate": 4.1955429320515625e-06, + "loss": 2.9192, + "step": 7414 + }, + { + "epoch": 1.62, + "learning_rate": 4.195433690190081e-06, + "loss": 2.1208, + "step": 7415 + }, + { + "epoch": 1.62, + "learning_rate": 4.1953244483286e-06, + "loss": 2.4583, + "step": 7416 + }, + { + "epoch": 1.62, + "learning_rate": 4.195215206467119e-06, + "loss": 2.9443, + "step": 7417 + }, + { + "epoch": 1.62, + "learning_rate": 4.195105964605638e-06, + "loss": 2.495, + "step": 7418 + }, + { + "epoch": 1.62, + "learning_rate": 4.194996722744156e-06, + "loss": 2.5041, + "step": 7419 + }, + { + "epoch": 1.62, + "learning_rate": 4.194887480882674e-06, + "loss": 2.2015, + "step": 7420 + }, + { + "epoch": 1.62, + "learning_rate": 4.194778239021193e-06, + "loss": 3.1504, + "step": 7421 + }, + { + "epoch": 1.62, + "learning_rate": 4.194668997159712e-06, + "loss": 1.7954, + "step": 7422 + }, + { + "epoch": 1.62, + "learning_rate": 4.194559755298231e-06, + "loss": 2.6984, + "step": 7423 + }, + { + "epoch": 1.62, + "learning_rate": 4.1944505134367494e-06, + "loss": 2.5739, + "step": 7424 + }, + { + "epoch": 1.62, + "learning_rate": 4.194341271575268e-06, + "loss": 3.0253, + "step": 7425 + }, + { + "epoch": 1.62, + "learning_rate": 4.194232029713787e-06, + "loss": 3.4922, + "step": 7426 + }, + { + "epoch": 1.62, + "learning_rate": 4.194122787852306e-06, + "loss": 3.2747, + "step": 7427 + }, + { + "epoch": 1.62, + "learning_rate": 4.194013545990824e-06, + "loss": 2.8661, + "step": 7428 + }, + { + "epoch": 1.62, + "learning_rate": 4.193904304129343e-06, + "loss": 2.4659, + "step": 7429 + }, + { + "epoch": 1.62, + "learning_rate": 4.193795062267861e-06, + "loss": 3.5319, + "step": 7430 + }, + { + "epoch": 1.62, + "learning_rate": 4.19368582040638e-06, + "loss": 2.7067, + "step": 7431 + }, + { + "epoch": 1.62, + "learning_rate": 4.193576578544899e-06, + "loss": 3.3426, + "step": 7432 + }, + { + "epoch": 1.62, + "learning_rate": 4.193467336683417e-06, + "loss": 3.5959, + "step": 7433 + }, + { + "epoch": 1.62, + "learning_rate": 4.193358094821936e-06, + "loss": 2.6998, + "step": 7434 + }, + { + "epoch": 1.62, + "learning_rate": 4.193248852960455e-06, + "loss": 2.4208, + "step": 7435 + }, + { + "epoch": 1.62, + "learning_rate": 4.193139611098974e-06, + "loss": 2.9612, + "step": 7436 + }, + { + "epoch": 1.62, + "learning_rate": 4.193030369237493e-06, + "loss": 2.2005, + "step": 7437 + }, + { + "epoch": 1.62, + "learning_rate": 4.192921127376011e-06, + "loss": 2.8785, + "step": 7438 + }, + { + "epoch": 1.62, + "learning_rate": 4.192811885514529e-06, + "loss": 4.0286, + "step": 7439 + }, + { + "epoch": 1.62, + "learning_rate": 4.192702643653048e-06, + "loss": 2.0942, + "step": 7440 + }, + { + "epoch": 1.62, + "learning_rate": 4.192593401791567e-06, + "loss": 2.6237, + "step": 7441 + }, + { + "epoch": 1.62, + "learning_rate": 4.192484159930085e-06, + "loss": 3.3909, + "step": 7442 + }, + { + "epoch": 1.62, + "learning_rate": 4.192374918068604e-06, + "loss": 2.2933, + "step": 7443 + }, + { + "epoch": 1.62, + "learning_rate": 4.1922656762071225e-06, + "loss": 3.5389, + "step": 7444 + }, + { + "epoch": 1.62, + "learning_rate": 4.1921564343456415e-06, + "loss": 3.8644, + "step": 7445 + }, + { + "epoch": 1.63, + "learning_rate": 4.192047192484161e-06, + "loss": 2.0984, + "step": 7446 + }, + { + "epoch": 1.63, + "learning_rate": 4.191937950622679e-06, + "loss": 2.8515, + "step": 7447 + }, + { + "epoch": 1.63, + "learning_rate": 4.191828708761198e-06, + "loss": 2.951, + "step": 7448 + }, + { + "epoch": 1.63, + "learning_rate": 4.191719466899716e-06, + "loss": 2.4075, + "step": 7449 + }, + { + "epoch": 1.63, + "learning_rate": 4.191610225038235e-06, + "loss": 3.2817, + "step": 7450 + }, + { + "epoch": 1.63, + "learning_rate": 4.191500983176754e-06, + "loss": 2.579, + "step": 7451 + }, + { + "epoch": 1.63, + "learning_rate": 4.191391741315272e-06, + "loss": 2.7963, + "step": 7452 + }, + { + "epoch": 1.63, + "learning_rate": 4.191282499453791e-06, + "loss": 2.8208, + "step": 7453 + }, + { + "epoch": 1.63, + "learning_rate": 4.1911732575923095e-06, + "loss": 2.1632, + "step": 7454 + }, + { + "epoch": 1.63, + "learning_rate": 4.1910640157308285e-06, + "loss": 2.2213, + "step": 7455 + }, + { + "epoch": 1.63, + "learning_rate": 4.1909547738693475e-06, + "loss": 2.4358, + "step": 7456 + }, + { + "epoch": 1.63, + "learning_rate": 4.190845532007866e-06, + "loss": 2.7469, + "step": 7457 + }, + { + "epoch": 1.63, + "learning_rate": 4.190736290146385e-06, + "loss": 2.7252, + "step": 7458 + }, + { + "epoch": 1.63, + "learning_rate": 4.190627048284903e-06, + "loss": 2.3078, + "step": 7459 + }, + { + "epoch": 1.63, + "learning_rate": 4.190517806423422e-06, + "loss": 3.2498, + "step": 7460 + }, + { + "epoch": 1.63, + "learning_rate": 4.19040856456194e-06, + "loss": 2.5978, + "step": 7461 + }, + { + "epoch": 1.63, + "learning_rate": 4.190299322700459e-06, + "loss": 3.67, + "step": 7462 + }, + { + "epoch": 1.63, + "learning_rate": 4.190190080838978e-06, + "loss": 2.5067, + "step": 7463 + }, + { + "epoch": 1.63, + "learning_rate": 4.190080838977496e-06, + "loss": 3.4288, + "step": 7464 + }, + { + "epoch": 1.63, + "learning_rate": 4.1899715971160155e-06, + "loss": 2.7497, + "step": 7465 + }, + { + "epoch": 1.63, + "learning_rate": 4.189862355254534e-06, + "loss": 2.7618, + "step": 7466 + }, + { + "epoch": 1.63, + "learning_rate": 4.189753113393053e-06, + "loss": 3.3575, + "step": 7467 + }, + { + "epoch": 1.63, + "learning_rate": 4.189643871531571e-06, + "loss": 3.2352, + "step": 7468 + }, + { + "epoch": 1.63, + "learning_rate": 4.18953462967009e-06, + "loss": 2.1296, + "step": 7469 + }, + { + "epoch": 1.63, + "learning_rate": 4.189425387808609e-06, + "loss": 3.1356, + "step": 7470 + }, + { + "epoch": 1.63, + "learning_rate": 4.189316145947127e-06, + "loss": 2.5648, + "step": 7471 + }, + { + "epoch": 1.63, + "learning_rate": 4.189206904085646e-06, + "loss": 2.6961, + "step": 7472 + }, + { + "epoch": 1.63, + "learning_rate": 4.189097662224164e-06, + "loss": 2.6694, + "step": 7473 + }, + { + "epoch": 1.63, + "learning_rate": 4.188988420362683e-06, + "loss": 3.034, + "step": 7474 + }, + { + "epoch": 1.63, + "learning_rate": 4.1888791785012016e-06, + "loss": 1.8765, + "step": 7475 + }, + { + "epoch": 1.63, + "learning_rate": 4.188769936639721e-06, + "loss": 2.995, + "step": 7476 + }, + { + "epoch": 1.63, + "learning_rate": 4.18866069477824e-06, + "loss": 1.5894, + "step": 7477 + }, + { + "epoch": 1.63, + "learning_rate": 4.188551452916758e-06, + "loss": 3.5244, + "step": 7478 + }, + { + "epoch": 1.63, + "learning_rate": 4.188442211055277e-06, + "loss": 2.7093, + "step": 7479 + }, + { + "epoch": 1.63, + "learning_rate": 4.188332969193795e-06, + "loss": 3.4699, + "step": 7480 + }, + { + "epoch": 1.63, + "learning_rate": 4.188223727332314e-06, + "loss": 3.0298, + "step": 7481 + }, + { + "epoch": 1.63, + "learning_rate": 4.188114485470833e-06, + "loss": 2.6956, + "step": 7482 + }, + { + "epoch": 1.63, + "learning_rate": 4.188005243609351e-06, + "loss": 2.6781, + "step": 7483 + }, + { + "epoch": 1.63, + "learning_rate": 4.18789600174787e-06, + "loss": 2.4885, + "step": 7484 + }, + { + "epoch": 1.63, + "learning_rate": 4.1877867598863885e-06, + "loss": 3.0885, + "step": 7485 + }, + { + "epoch": 1.63, + "learning_rate": 4.1876775180249076e-06, + "loss": 3.0598, + "step": 7486 + }, + { + "epoch": 1.63, + "learning_rate": 4.187568276163427e-06, + "loss": 2.6392, + "step": 7487 + }, + { + "epoch": 1.63, + "learning_rate": 4.187459034301945e-06, + "loss": 3.3191, + "step": 7488 + }, + { + "epoch": 1.63, + "learning_rate": 4.187349792440464e-06, + "loss": 2.9909, + "step": 7489 + }, + { + "epoch": 1.63, + "learning_rate": 4.187240550578982e-06, + "loss": 2.9064, + "step": 7490 + }, + { + "epoch": 1.63, + "learning_rate": 4.187131308717501e-06, + "loss": 2.5572, + "step": 7491 + }, + { + "epoch": 1.64, + "learning_rate": 4.187022066856019e-06, + "loss": 2.9014, + "step": 7492 + }, + { + "epoch": 1.64, + "learning_rate": 4.186912824994538e-06, + "loss": 2.7929, + "step": 7493 + }, + { + "epoch": 1.64, + "learning_rate": 4.1868035831330564e-06, + "loss": 2.9408, + "step": 7494 + }, + { + "epoch": 1.64, + "learning_rate": 4.1866943412715755e-06, + "loss": 2.982, + "step": 7495 + }, + { + "epoch": 1.64, + "learning_rate": 4.1865850994100945e-06, + "loss": 2.6725, + "step": 7496 + }, + { + "epoch": 1.64, + "learning_rate": 4.186475857548613e-06, + "loss": 3.3687, + "step": 7497 + }, + { + "epoch": 1.64, + "learning_rate": 4.186366615687132e-06, + "loss": 3.1527, + "step": 7498 + }, + { + "epoch": 1.64, + "learning_rate": 4.18625737382565e-06, + "loss": 2.2703, + "step": 7499 + }, + { + "epoch": 1.64, + "learning_rate": 4.186148131964169e-06, + "loss": 3.6283, + "step": 7500 + }, + { + "epoch": 1.64, + "learning_rate": 4.186038890102688e-06, + "loss": 3.4082, + "step": 7501 + }, + { + "epoch": 1.64, + "learning_rate": 4.185929648241206e-06, + "loss": 2.4817, + "step": 7502 + }, + { + "epoch": 1.64, + "learning_rate": 4.185820406379725e-06, + "loss": 2.6348, + "step": 7503 + }, + { + "epoch": 1.64, + "learning_rate": 4.185711164518243e-06, + "loss": 2.9342, + "step": 7504 + }, + { + "epoch": 1.64, + "learning_rate": 4.1856019226567624e-06, + "loss": 2.9192, + "step": 7505 + }, + { + "epoch": 1.64, + "learning_rate": 4.1854926807952815e-06, + "loss": 3.1066, + "step": 7506 + }, + { + "epoch": 1.64, + "learning_rate": 4.1853834389338e-06, + "loss": 2.2873, + "step": 7507 + }, + { + "epoch": 1.64, + "learning_rate": 4.185274197072318e-06, + "loss": 3.1515, + "step": 7508 + }, + { + "epoch": 1.64, + "learning_rate": 4.185164955210837e-06, + "loss": 3.9782, + "step": 7509 + }, + { + "epoch": 1.64, + "learning_rate": 4.185055713349356e-06, + "loss": 2.657, + "step": 7510 + }, + { + "epoch": 1.64, + "learning_rate": 4.184946471487875e-06, + "loss": 2.6174, + "step": 7511 + }, + { + "epoch": 1.64, + "learning_rate": 4.184837229626393e-06, + "loss": 2.8569, + "step": 7512 + }, + { + "epoch": 1.64, + "learning_rate": 4.184727987764911e-06, + "loss": 2.5481, + "step": 7513 + }, + { + "epoch": 1.64, + "learning_rate": 4.18461874590343e-06, + "loss": 3.1736, + "step": 7514 + }, + { + "epoch": 1.64, + "learning_rate": 4.184509504041949e-06, + "loss": 3.3828, + "step": 7515 + }, + { + "epoch": 1.64, + "learning_rate": 4.1844002621804684e-06, + "loss": 3.0541, + "step": 7516 + }, + { + "epoch": 1.64, + "learning_rate": 4.184291020318987e-06, + "loss": 2.8046, + "step": 7517 + }, + { + "epoch": 1.64, + "learning_rate": 4.184181778457505e-06, + "loss": 2.8636, + "step": 7518 + }, + { + "epoch": 1.64, + "learning_rate": 4.184072536596024e-06, + "loss": 2.4321, + "step": 7519 + }, + { + "epoch": 1.64, + "learning_rate": 4.183963294734543e-06, + "loss": 2.8407, + "step": 7520 + }, + { + "epoch": 1.64, + "learning_rate": 4.183854052873061e-06, + "loss": 2.3302, + "step": 7521 + }, + { + "epoch": 1.64, + "learning_rate": 4.18374481101158e-06, + "loss": 1.7724, + "step": 7522 + }, + { + "epoch": 1.64, + "learning_rate": 4.183635569150098e-06, + "loss": 3.1833, + "step": 7523 + }, + { + "epoch": 1.64, + "learning_rate": 4.183526327288617e-06, + "loss": 3.7264, + "step": 7524 + }, + { + "epoch": 1.64, + "learning_rate": 4.183417085427136e-06, + "loss": 2.5854, + "step": 7525 + }, + { + "epoch": 1.64, + "learning_rate": 4.1833078435656545e-06, + "loss": 3.4854, + "step": 7526 + }, + { + "epoch": 1.64, + "learning_rate": 4.183198601704173e-06, + "loss": 2.9815, + "step": 7527 + }, + { + "epoch": 1.64, + "learning_rate": 4.183089359842692e-06, + "loss": 3.1111, + "step": 7528 + }, + { + "epoch": 1.64, + "learning_rate": 4.182980117981211e-06, + "loss": 2.1739, + "step": 7529 + }, + { + "epoch": 1.64, + "learning_rate": 4.18287087611973e-06, + "loss": 3.0905, + "step": 7530 + }, + { + "epoch": 1.64, + "learning_rate": 4.182761634258248e-06, + "loss": 1.802, + "step": 7531 + }, + { + "epoch": 1.64, + "learning_rate": 4.182652392396766e-06, + "loss": 2.1799, + "step": 7532 + }, + { + "epoch": 1.64, + "learning_rate": 4.182543150535285e-06, + "loss": 2.7367, + "step": 7533 + }, + { + "epoch": 1.64, + "learning_rate": 4.182433908673804e-06, + "loss": 3.1693, + "step": 7534 + }, + { + "epoch": 1.64, + "learning_rate": 4.182324666812323e-06, + "loss": 2.6056, + "step": 7535 + }, + { + "epoch": 1.64, + "learning_rate": 4.1822154249508415e-06, + "loss": 2.659, + "step": 7536 + }, + { + "epoch": 1.64, + "learning_rate": 4.18210618308936e-06, + "loss": 3.1944, + "step": 7537 + }, + { + "epoch": 1.65, + "learning_rate": 4.181996941227879e-06, + "loss": 4.0388, + "step": 7538 + }, + { + "epoch": 1.65, + "learning_rate": 4.181887699366398e-06, + "loss": 3.1879, + "step": 7539 + }, + { + "epoch": 1.65, + "learning_rate": 4.181778457504917e-06, + "loss": 1.7325, + "step": 7540 + }, + { + "epoch": 1.65, + "learning_rate": 4.181669215643435e-06, + "loss": 2.9139, + "step": 7541 + }, + { + "epoch": 1.65, + "learning_rate": 4.181559973781953e-06, + "loss": 2.6517, + "step": 7542 + }, + { + "epoch": 1.65, + "learning_rate": 4.181450731920472e-06, + "loss": 3.6124, + "step": 7543 + }, + { + "epoch": 1.65, + "learning_rate": 4.181341490058991e-06, + "loss": 1.6452, + "step": 7544 + }, + { + "epoch": 1.65, + "learning_rate": 4.181232248197509e-06, + "loss": 2.3365, + "step": 7545 + }, + { + "epoch": 1.65, + "learning_rate": 4.1811230063360285e-06, + "loss": 2.3635, + "step": 7546 + }, + { + "epoch": 1.65, + "learning_rate": 4.181013764474547e-06, + "loss": 2.9206, + "step": 7547 + }, + { + "epoch": 1.65, + "learning_rate": 4.180904522613066e-06, + "loss": 2.9163, + "step": 7548 + }, + { + "epoch": 1.65, + "learning_rate": 4.180795280751585e-06, + "loss": 2.5074, + "step": 7549 + }, + { + "epoch": 1.65, + "learning_rate": 4.180686038890103e-06, + "loss": 2.6732, + "step": 7550 + }, + { + "epoch": 1.65, + "learning_rate": 4.180576797028622e-06, + "loss": 2.3158, + "step": 7551 + }, + { + "epoch": 1.65, + "learning_rate": 4.18046755516714e-06, + "loss": 1.6141, + "step": 7552 + }, + { + "epoch": 1.65, + "learning_rate": 4.180358313305659e-06, + "loss": 1.8772, + "step": 7553 + }, + { + "epoch": 1.65, + "learning_rate": 4.180249071444178e-06, + "loss": 3.0899, + "step": 7554 + }, + { + "epoch": 1.65, + "learning_rate": 4.180139829582696e-06, + "loss": 2.3662, + "step": 7555 + }, + { + "epoch": 1.65, + "learning_rate": 4.1800305877212146e-06, + "loss": 2.2393, + "step": 7556 + }, + { + "epoch": 1.65, + "learning_rate": 4.179921345859734e-06, + "loss": 2.2452, + "step": 7557 + }, + { + "epoch": 1.65, + "learning_rate": 4.179812103998253e-06, + "loss": 2.8603, + "step": 7558 + }, + { + "epoch": 1.65, + "learning_rate": 4.179702862136772e-06, + "loss": 2.7349, + "step": 7559 + }, + { + "epoch": 1.65, + "learning_rate": 4.17959362027529e-06, + "loss": 3.2125, + "step": 7560 + }, + { + "epoch": 1.65, + "learning_rate": 4.179484378413808e-06, + "loss": 3.0045, + "step": 7561 + }, + { + "epoch": 1.65, + "learning_rate": 4.179375136552327e-06, + "loss": 3.2717, + "step": 7562 + }, + { + "epoch": 1.65, + "learning_rate": 4.179265894690846e-06, + "loss": 2.7679, + "step": 7563 + }, + { + "epoch": 1.65, + "learning_rate": 4.179156652829365e-06, + "loss": 2.4278, + "step": 7564 + }, + { + "epoch": 1.65, + "learning_rate": 4.179047410967883e-06, + "loss": 2.4246, + "step": 7565 + }, + { + "epoch": 1.65, + "learning_rate": 4.1789381691064015e-06, + "loss": 3.6095, + "step": 7566 + }, + { + "epoch": 1.65, + "learning_rate": 4.1788289272449206e-06, + "loss": 2.6266, + "step": 7567 + }, + { + "epoch": 1.65, + "learning_rate": 4.17871968538344e-06, + "loss": 3.4451, + "step": 7568 + }, + { + "epoch": 1.65, + "learning_rate": 4.178610443521959e-06, + "loss": 2.1684, + "step": 7569 + }, + { + "epoch": 1.65, + "learning_rate": 4.178501201660477e-06, + "loss": 2.1832, + "step": 7570 + }, + { + "epoch": 1.65, + "learning_rate": 4.178391959798995e-06, + "loss": 3.7638, + "step": 7571 + }, + { + "epoch": 1.65, + "learning_rate": 4.178282717937514e-06, + "loss": 3.4448, + "step": 7572 + }, + { + "epoch": 1.65, + "learning_rate": 4.178173476076033e-06, + "loss": 3.5251, + "step": 7573 + }, + { + "epoch": 1.65, + "learning_rate": 4.178064234214551e-06, + "loss": 2.6988, + "step": 7574 + }, + { + "epoch": 1.65, + "learning_rate": 4.17795499235307e-06, + "loss": 3.1557, + "step": 7575 + }, + { + "epoch": 1.65, + "learning_rate": 4.1778457504915885e-06, + "loss": 3.9311, + "step": 7576 + }, + { + "epoch": 1.65, + "learning_rate": 4.1777365086301075e-06, + "loss": 2.6461, + "step": 7577 + }, + { + "epoch": 1.65, + "learning_rate": 4.1776272667686265e-06, + "loss": 2.072, + "step": 7578 + }, + { + "epoch": 1.65, + "learning_rate": 4.177518024907145e-06, + "loss": 3.4715, + "step": 7579 + }, + { + "epoch": 1.65, + "learning_rate": 4.177408783045663e-06, + "loss": 2.5944, + "step": 7580 + }, + { + "epoch": 1.65, + "learning_rate": 4.177299541184182e-06, + "loss": 3.0271, + "step": 7581 + }, + { + "epoch": 1.65, + "learning_rate": 4.177190299322701e-06, + "loss": 3.296, + "step": 7582 + }, + { + "epoch": 1.65, + "learning_rate": 4.17708105746122e-06, + "loss": 2.4257, + "step": 7583 + }, + { + "epoch": 1.66, + "learning_rate": 4.176971815599738e-06, + "loss": 3.9238, + "step": 7584 + }, + { + "epoch": 1.66, + "learning_rate": 4.176862573738256e-06, + "loss": 3.451, + "step": 7585 + }, + { + "epoch": 1.66, + "learning_rate": 4.1767533318767754e-06, + "loss": 2.439, + "step": 7586 + }, + { + "epoch": 1.66, + "learning_rate": 4.1766440900152945e-06, + "loss": 2.6487, + "step": 7587 + }, + { + "epoch": 1.66, + "learning_rate": 4.1765348481538135e-06, + "loss": 1.9475, + "step": 7588 + }, + { + "epoch": 1.66, + "learning_rate": 4.176425606292332e-06, + "loss": 2.2757, + "step": 7589 + }, + { + "epoch": 1.66, + "learning_rate": 4.17631636443085e-06, + "loss": 2.3136, + "step": 7590 + }, + { + "epoch": 1.66, + "learning_rate": 4.176207122569369e-06, + "loss": 1.9987, + "step": 7591 + }, + { + "epoch": 1.66, + "learning_rate": 4.176097880707888e-06, + "loss": 2.9867, + "step": 7592 + }, + { + "epoch": 1.66, + "learning_rate": 4.175988638846406e-06, + "loss": 2.0952, + "step": 7593 + }, + { + "epoch": 1.66, + "learning_rate": 4.175879396984925e-06, + "loss": 1.9919, + "step": 7594 + }, + { + "epoch": 1.66, + "learning_rate": 4.175770155123443e-06, + "loss": 3.1596, + "step": 7595 + }, + { + "epoch": 1.66, + "learning_rate": 4.175660913261962e-06, + "loss": 3.1918, + "step": 7596 + }, + { + "epoch": 1.66, + "learning_rate": 4.1755516714004814e-06, + "loss": 2.1892, + "step": 7597 + }, + { + "epoch": 1.66, + "learning_rate": 4.175442429539e-06, + "loss": 2.1391, + "step": 7598 + }, + { + "epoch": 1.66, + "learning_rate": 4.175333187677519e-06, + "loss": 2.8782, + "step": 7599 + }, + { + "epoch": 1.66, + "learning_rate": 4.175223945816037e-06, + "loss": 2.6409, + "step": 7600 + }, + { + "epoch": 1.66, + "learning_rate": 4.175114703954556e-06, + "loss": 1.9432, + "step": 7601 + }, + { + "epoch": 1.66, + "learning_rate": 4.175005462093075e-06, + "loss": 1.6236, + "step": 7602 + }, + { + "epoch": 1.66, + "learning_rate": 4.174896220231593e-06, + "loss": 2.6701, + "step": 7603 + }, + { + "epoch": 1.66, + "learning_rate": 4.174786978370112e-06, + "loss": 2.34, + "step": 7604 + }, + { + "epoch": 1.66, + "learning_rate": 4.17467773650863e-06, + "loss": 2.2215, + "step": 7605 + }, + { + "epoch": 1.66, + "learning_rate": 4.174568494647149e-06, + "loss": 3.3084, + "step": 7606 + }, + { + "epoch": 1.66, + "learning_rate": 4.1744592527856675e-06, + "loss": 2.8, + "step": 7607 + }, + { + "epoch": 1.66, + "learning_rate": 4.1743500109241866e-06, + "loss": 2.5687, + "step": 7608 + }, + { + "epoch": 1.66, + "learning_rate": 4.174240769062705e-06, + "loss": 2.1716, + "step": 7609 + }, + { + "epoch": 1.66, + "learning_rate": 4.174131527201224e-06, + "loss": 2.9878, + "step": 7610 + }, + { + "epoch": 1.66, + "learning_rate": 4.174022285339743e-06, + "loss": 2.7156, + "step": 7611 + }, + { + "epoch": 1.66, + "learning_rate": 4.173913043478261e-06, + "loss": 2.8946, + "step": 7612 + }, + { + "epoch": 1.66, + "learning_rate": 4.17380380161678e-06, + "loss": 2.2772, + "step": 7613 + }, + { + "epoch": 1.66, + "learning_rate": 4.173694559755298e-06, + "loss": 3.3034, + "step": 7614 + }, + { + "epoch": 1.66, + "learning_rate": 4.173585317893817e-06, + "loss": 2.5884, + "step": 7615 + }, + { + "epoch": 1.66, + "learning_rate": 4.173476076032336e-06, + "loss": 2.6673, + "step": 7616 + }, + { + "epoch": 1.66, + "learning_rate": 4.1733668341708545e-06, + "loss": 2.614, + "step": 7617 + }, + { + "epoch": 1.66, + "learning_rate": 4.1732575923093735e-06, + "loss": 2.2968, + "step": 7618 + }, + { + "epoch": 1.66, + "learning_rate": 4.173148350447892e-06, + "loss": 2.5212, + "step": 7619 + }, + { + "epoch": 1.66, + "learning_rate": 4.173039108586411e-06, + "loss": 2.0396, + "step": 7620 + }, + { + "epoch": 1.66, + "learning_rate": 4.172929866724929e-06, + "loss": 2.4751, + "step": 7621 + }, + { + "epoch": 1.66, + "learning_rate": 4.172820624863448e-06, + "loss": 2.9806, + "step": 7622 + }, + { + "epoch": 1.66, + "learning_rate": 4.172711383001967e-06, + "loss": 2.7473, + "step": 7623 + }, + { + "epoch": 1.66, + "learning_rate": 4.172602141140485e-06, + "loss": 3.0708, + "step": 7624 + }, + { + "epoch": 1.66, + "learning_rate": 4.172492899279004e-06, + "loss": 2.6576, + "step": 7625 + }, + { + "epoch": 1.66, + "learning_rate": 4.172383657417522e-06, + "loss": 3.1982, + "step": 7626 + }, + { + "epoch": 1.66, + "learning_rate": 4.1722744155560414e-06, + "loss": 2.7539, + "step": 7627 + }, + { + "epoch": 1.66, + "learning_rate": 4.1721651736945605e-06, + "loss": 1.6154, + "step": 7628 + }, + { + "epoch": 1.66, + "learning_rate": 4.172055931833079e-06, + "loss": 2.0984, + "step": 7629 + }, + { + "epoch": 1.67, + "learning_rate": 4.171946689971598e-06, + "loss": 3.0242, + "step": 7630 + }, + { + "epoch": 1.67, + "learning_rate": 4.171837448110116e-06, + "loss": 2.7181, + "step": 7631 + }, + { + "epoch": 1.67, + "learning_rate": 4.171728206248635e-06, + "loss": 2.3621, + "step": 7632 + }, + { + "epoch": 1.67, + "learning_rate": 4.171618964387153e-06, + "loss": 2.4211, + "step": 7633 + }, + { + "epoch": 1.67, + "learning_rate": 4.171509722525672e-06, + "loss": 3.3842, + "step": 7634 + }, + { + "epoch": 1.67, + "learning_rate": 4.171400480664191e-06, + "loss": 3.6931, + "step": 7635 + }, + { + "epoch": 1.67, + "learning_rate": 4.171291238802709e-06, + "loss": 1.9405, + "step": 7636 + }, + { + "epoch": 1.67, + "learning_rate": 4.171181996941228e-06, + "loss": 3.0065, + "step": 7637 + }, + { + "epoch": 1.67, + "learning_rate": 4.171072755079747e-06, + "loss": 2.5627, + "step": 7638 + }, + { + "epoch": 1.67, + "learning_rate": 4.170963513218266e-06, + "loss": 2.6436, + "step": 7639 + }, + { + "epoch": 1.67, + "learning_rate": 4.170854271356784e-06, + "loss": 3.0209, + "step": 7640 + }, + { + "epoch": 1.67, + "learning_rate": 4.170745029495303e-06, + "loss": 2.7763, + "step": 7641 + }, + { + "epoch": 1.67, + "learning_rate": 4.170635787633822e-06, + "loss": 2.9411, + "step": 7642 + }, + { + "epoch": 1.67, + "learning_rate": 4.17052654577234e-06, + "loss": 1.9411, + "step": 7643 + }, + { + "epoch": 1.67, + "learning_rate": 4.170417303910859e-06, + "loss": 3.2282, + "step": 7644 + }, + { + "epoch": 1.67, + "learning_rate": 4.170308062049377e-06, + "loss": 2.8892, + "step": 7645 + }, + { + "epoch": 1.67, + "learning_rate": 4.170198820187896e-06, + "loss": 2.2928, + "step": 7646 + }, + { + "epoch": 1.67, + "learning_rate": 4.170089578326415e-06, + "loss": 3.2444, + "step": 7647 + }, + { + "epoch": 1.67, + "learning_rate": 4.1699803364649335e-06, + "loss": 1.8667, + "step": 7648 + }, + { + "epoch": 1.67, + "learning_rate": 4.169871094603453e-06, + "loss": 3.252, + "step": 7649 + }, + { + "epoch": 1.67, + "learning_rate": 4.169761852741971e-06, + "loss": 1.653, + "step": 7650 + }, + { + "epoch": 1.67, + "learning_rate": 4.16965261088049e-06, + "loss": 2.3052, + "step": 7651 + }, + { + "epoch": 1.67, + "learning_rate": 4.169543369019009e-06, + "loss": 1.8835, + "step": 7652 + }, + { + "epoch": 1.67, + "learning_rate": 4.169434127157527e-06, + "loss": 3.7265, + "step": 7653 + }, + { + "epoch": 1.67, + "learning_rate": 4.169324885296045e-06, + "loss": 3.6937, + "step": 7654 + }, + { + "epoch": 1.67, + "learning_rate": 4.169215643434564e-06, + "loss": 1.9132, + "step": 7655 + }, + { + "epoch": 1.67, + "learning_rate": 4.169106401573083e-06, + "loss": 3.9006, + "step": 7656 + }, + { + "epoch": 1.67, + "learning_rate": 4.168997159711602e-06, + "loss": 1.67, + "step": 7657 + }, + { + "epoch": 1.67, + "learning_rate": 4.1688879178501205e-06, + "loss": 2.2115, + "step": 7658 + }, + { + "epoch": 1.67, + "learning_rate": 4.168778675988639e-06, + "loss": 2.8821, + "step": 7659 + }, + { + "epoch": 1.67, + "learning_rate": 4.168669434127158e-06, + "loss": 2.5093, + "step": 7660 + }, + { + "epoch": 1.67, + "learning_rate": 4.168560192265677e-06, + "loss": 2.4302, + "step": 7661 + }, + { + "epoch": 1.67, + "learning_rate": 4.168450950404195e-06, + "loss": 3.9607, + "step": 7662 + }, + { + "epoch": 1.67, + "learning_rate": 4.168341708542714e-06, + "loss": 2.8659, + "step": 7663 + }, + { + "epoch": 1.67, + "learning_rate": 4.168232466681232e-06, + "loss": 3.9588, + "step": 7664 + }, + { + "epoch": 1.67, + "learning_rate": 4.168123224819751e-06, + "loss": 2.5343, + "step": 7665 + }, + { + "epoch": 1.67, + "learning_rate": 4.16801398295827e-06, + "loss": 2.5027, + "step": 7666 + }, + { + "epoch": 1.67, + "learning_rate": 4.1679047410967884e-06, + "loss": 3.1225, + "step": 7667 + }, + { + "epoch": 1.67, + "learning_rate": 4.1677954992353075e-06, + "loss": 2.523, + "step": 7668 + }, + { + "epoch": 1.67, + "learning_rate": 4.167686257373826e-06, + "loss": 3.1635, + "step": 7669 + }, + { + "epoch": 1.67, + "learning_rate": 4.167577015512345e-06, + "loss": 2.3374, + "step": 7670 + }, + { + "epoch": 1.67, + "learning_rate": 4.167467773650864e-06, + "loss": 2.5575, + "step": 7671 + }, + { + "epoch": 1.67, + "learning_rate": 4.167358531789382e-06, + "loss": 2.8888, + "step": 7672 + }, + { + "epoch": 1.67, + "learning_rate": 4.1672492899279e-06, + "loss": 2.2767, + "step": 7673 + }, + { + "epoch": 1.67, + "learning_rate": 4.167140048066419e-06, + "loss": 3.2808, + "step": 7674 + }, + { + "epoch": 1.68, + "learning_rate": 4.167030806204938e-06, + "loss": 2.8731, + "step": 7675 + }, + { + "epoch": 1.68, + "learning_rate": 4.166921564343457e-06, + "loss": 3.0912, + "step": 7676 + }, + { + "epoch": 1.68, + "learning_rate": 4.166812322481975e-06, + "loss": 2.5667, + "step": 7677 + }, + { + "epoch": 1.68, + "learning_rate": 4.1667030806204936e-06, + "loss": 3.5199, + "step": 7678 + }, + { + "epoch": 1.68, + "learning_rate": 4.166593838759013e-06, + "loss": 3.0929, + "step": 7679 + }, + { + "epoch": 1.68, + "learning_rate": 4.166484596897532e-06, + "loss": 2.6952, + "step": 7680 + }, + { + "epoch": 1.68, + "learning_rate": 4.166375355036051e-06, + "loss": 2.5896, + "step": 7681 + }, + { + "epoch": 1.68, + "learning_rate": 4.166266113174569e-06, + "loss": 3.1261, + "step": 7682 + }, + { + "epoch": 1.68, + "learning_rate": 4.166156871313087e-06, + "loss": 3.4242, + "step": 7683 + }, + { + "epoch": 1.68, + "learning_rate": 4.166047629451606e-06, + "loss": 2.337, + "step": 7684 + }, + { + "epoch": 1.68, + "learning_rate": 4.165938387590125e-06, + "loss": 3.303, + "step": 7685 + }, + { + "epoch": 1.68, + "learning_rate": 4.165829145728643e-06, + "loss": 1.7371, + "step": 7686 + }, + { + "epoch": 1.68, + "learning_rate": 4.165719903867162e-06, + "loss": 3.3055, + "step": 7687 + }, + { + "epoch": 1.68, + "learning_rate": 4.1656106620056805e-06, + "loss": 3.1287, + "step": 7688 + }, + { + "epoch": 1.68, + "learning_rate": 4.1655014201441996e-06, + "loss": 1.2273, + "step": 7689 + }, + { + "epoch": 1.68, + "learning_rate": 4.165392178282719e-06, + "loss": 3.2294, + "step": 7690 + }, + { + "epoch": 1.68, + "learning_rate": 4.165282936421237e-06, + "loss": 2.8588, + "step": 7691 + }, + { + "epoch": 1.68, + "learning_rate": 4.165173694559756e-06, + "loss": 2.9111, + "step": 7692 + }, + { + "epoch": 1.68, + "learning_rate": 4.165064452698274e-06, + "loss": 2.0514, + "step": 7693 + }, + { + "epoch": 1.68, + "learning_rate": 4.164955210836793e-06, + "loss": 2.522, + "step": 7694 + }, + { + "epoch": 1.68, + "learning_rate": 4.164845968975312e-06, + "loss": 2.0705, + "step": 7695 + }, + { + "epoch": 1.68, + "learning_rate": 4.16473672711383e-06, + "loss": 1.7242, + "step": 7696 + }, + { + "epoch": 1.68, + "learning_rate": 4.1646274852523485e-06, + "loss": 2.3285, + "step": 7697 + }, + { + "epoch": 1.68, + "learning_rate": 4.1645182433908675e-06, + "loss": 2.8787, + "step": 7698 + }, + { + "epoch": 1.68, + "learning_rate": 4.1644090015293865e-06, + "loss": 2.4535, + "step": 7699 + }, + { + "epoch": 1.68, + "learning_rate": 4.1642997596679056e-06, + "loss": 3.9213, + "step": 7700 + }, + { + "epoch": 1.68, + "learning_rate": 4.164190517806424e-06, + "loss": 2.2129, + "step": 7701 + }, + { + "epoch": 1.68, + "learning_rate": 4.164081275944942e-06, + "loss": 2.6453, + "step": 7702 + }, + { + "epoch": 1.68, + "learning_rate": 4.163972034083461e-06, + "loss": 2.3013, + "step": 7703 + }, + { + "epoch": 1.68, + "learning_rate": 4.16386279222198e-06, + "loss": 3.6685, + "step": 7704 + }, + { + "epoch": 1.68, + "learning_rate": 4.163753550360499e-06, + "loss": 2.2394, + "step": 7705 + }, + { + "epoch": 1.68, + "learning_rate": 4.163644308499017e-06, + "loss": 3.04, + "step": 7706 + }, + { + "epoch": 1.68, + "learning_rate": 4.163535066637535e-06, + "loss": 2.1671, + "step": 7707 + }, + { + "epoch": 1.68, + "learning_rate": 4.1634258247760544e-06, + "loss": 2.7389, + "step": 7708 + }, + { + "epoch": 1.68, + "learning_rate": 4.1633165829145735e-06, + "loss": 2.7023, + "step": 7709 + }, + { + "epoch": 1.68, + "learning_rate": 4.1632073410530925e-06, + "loss": 1.9452, + "step": 7710 + }, + { + "epoch": 1.68, + "learning_rate": 4.163098099191611e-06, + "loss": 2.37, + "step": 7711 + }, + { + "epoch": 1.68, + "learning_rate": 4.162988857330129e-06, + "loss": 2.2724, + "step": 7712 + }, + { + "epoch": 1.68, + "learning_rate": 4.162879615468648e-06, + "loss": 3.0651, + "step": 7713 + }, + { + "epoch": 1.68, + "learning_rate": 4.162770373607167e-06, + "loss": 2.911, + "step": 7714 + }, + { + "epoch": 1.68, + "learning_rate": 4.162661131745685e-06, + "loss": 2.1703, + "step": 7715 + }, + { + "epoch": 1.68, + "learning_rate": 4.162551889884204e-06, + "loss": 2.6649, + "step": 7716 + }, + { + "epoch": 1.68, + "learning_rate": 4.162442648022722e-06, + "loss": 2.6105, + "step": 7717 + }, + { + "epoch": 1.68, + "learning_rate": 4.162333406161241e-06, + "loss": 2.784, + "step": 7718 + }, + { + "epoch": 1.68, + "learning_rate": 4.1622241642997604e-06, + "loss": 2.2308, + "step": 7719 + }, + { + "epoch": 1.68, + "learning_rate": 4.162114922438279e-06, + "loss": 3.335, + "step": 7720 + }, + { + "epoch": 1.69, + "learning_rate": 4.162005680576797e-06, + "loss": 2.1721, + "step": 7721 + }, + { + "epoch": 1.69, + "learning_rate": 4.161896438715316e-06, + "loss": 1.7521, + "step": 7722 + }, + { + "epoch": 1.69, + "learning_rate": 4.161787196853835e-06, + "loss": 2.3561, + "step": 7723 + }, + { + "epoch": 1.69, + "learning_rate": 4.161677954992354e-06, + "loss": 2.2479, + "step": 7724 + }, + { + "epoch": 1.69, + "learning_rate": 4.161568713130872e-06, + "loss": 3.1487, + "step": 7725 + }, + { + "epoch": 1.69, + "learning_rate": 4.16145947126939e-06, + "loss": 3.0583, + "step": 7726 + }, + { + "epoch": 1.69, + "learning_rate": 4.161350229407909e-06, + "loss": 2.4144, + "step": 7727 + }, + { + "epoch": 1.69, + "learning_rate": 4.161240987546428e-06, + "loss": 3.7116, + "step": 7728 + }, + { + "epoch": 1.69, + "learning_rate": 4.161131745684947e-06, + "loss": 1.8091, + "step": 7729 + }, + { + "epoch": 1.69, + "learning_rate": 4.161022503823466e-06, + "loss": 2.1889, + "step": 7730 + }, + { + "epoch": 1.69, + "learning_rate": 4.160913261961984e-06, + "loss": 2.4351, + "step": 7731 + }, + { + "epoch": 1.69, + "learning_rate": 4.160804020100503e-06, + "loss": 2.9977, + "step": 7732 + }, + { + "epoch": 1.69, + "learning_rate": 4.160694778239022e-06, + "loss": 3.2153, + "step": 7733 + }, + { + "epoch": 1.69, + "learning_rate": 4.160585536377541e-06, + "loss": 2.7068, + "step": 7734 + }, + { + "epoch": 1.69, + "learning_rate": 4.160476294516059e-06, + "loss": 2.9835, + "step": 7735 + }, + { + "epoch": 1.69, + "learning_rate": 4.160367052654577e-06, + "loss": 4.2672, + "step": 7736 + }, + { + "epoch": 1.69, + "learning_rate": 4.160257810793096e-06, + "loss": 3.7132, + "step": 7737 + }, + { + "epoch": 1.69, + "learning_rate": 4.160148568931615e-06, + "loss": 3.6299, + "step": 7738 + }, + { + "epoch": 1.69, + "learning_rate": 4.1600393270701335e-06, + "loss": 2.1707, + "step": 7739 + }, + { + "epoch": 1.69, + "learning_rate": 4.1599300852086525e-06, + "loss": 2.2852, + "step": 7740 + }, + { + "epoch": 1.69, + "learning_rate": 4.159820843347171e-06, + "loss": 2.981, + "step": 7741 + }, + { + "epoch": 1.69, + "learning_rate": 4.15971160148569e-06, + "loss": 2.3351, + "step": 7742 + }, + { + "epoch": 1.69, + "learning_rate": 4.159602359624209e-06, + "loss": 3.494, + "step": 7743 + }, + { + "epoch": 1.69, + "learning_rate": 4.159493117762727e-06, + "loss": 3.1407, + "step": 7744 + }, + { + "epoch": 1.69, + "learning_rate": 4.159383875901246e-06, + "loss": 3.2955, + "step": 7745 + }, + { + "epoch": 1.69, + "learning_rate": 4.159274634039764e-06, + "loss": 2.3247, + "step": 7746 + }, + { + "epoch": 1.69, + "learning_rate": 4.159165392178283e-06, + "loss": 3.2839, + "step": 7747 + }, + { + "epoch": 1.69, + "learning_rate": 4.159056150316802e-06, + "loss": 2.3015, + "step": 7748 + }, + { + "epoch": 1.69, + "learning_rate": 4.1589469084553205e-06, + "loss": 3.0702, + "step": 7749 + }, + { + "epoch": 1.69, + "learning_rate": 4.158837666593839e-06, + "loss": 3.4646, + "step": 7750 + }, + { + "epoch": 1.69, + "learning_rate": 4.158728424732358e-06, + "loss": 3.2587, + "step": 7751 + }, + { + "epoch": 1.69, + "learning_rate": 4.158619182870877e-06, + "loss": 2.4863, + "step": 7752 + }, + { + "epoch": 1.69, + "learning_rate": 4.158509941009395e-06, + "loss": 2.9164, + "step": 7753 + }, + { + "epoch": 1.69, + "learning_rate": 4.158400699147914e-06, + "loss": 2.2667, + "step": 7754 + }, + { + "epoch": 1.69, + "learning_rate": 4.158291457286432e-06, + "loss": 4.055, + "step": 7755 + }, + { + "epoch": 1.69, + "learning_rate": 4.158182215424951e-06, + "loss": 3.3924, + "step": 7756 + }, + { + "epoch": 1.69, + "learning_rate": 4.15807297356347e-06, + "loss": 2.6911, + "step": 7757 + }, + { + "epoch": 1.69, + "learning_rate": 4.157963731701988e-06, + "loss": 2.8567, + "step": 7758 + }, + { + "epoch": 1.69, + "learning_rate": 4.157854489840507e-06, + "loss": 3.7663, + "step": 7759 + }, + { + "epoch": 1.69, + "learning_rate": 4.157745247979026e-06, + "loss": 3.1979, + "step": 7760 + }, + { + "epoch": 1.69, + "learning_rate": 4.157636006117545e-06, + "loss": 1.9821, + "step": 7761 + }, + { + "epoch": 1.69, + "learning_rate": 4.157526764256064e-06, + "loss": 2.8237, + "step": 7762 + }, + { + "epoch": 1.69, + "learning_rate": 4.157417522394582e-06, + "loss": 2.8618, + "step": 7763 + }, + { + "epoch": 1.69, + "learning_rate": 4.157308280533101e-06, + "loss": 3.5015, + "step": 7764 + }, + { + "epoch": 1.69, + "learning_rate": 4.157199038671619e-06, + "loss": 3.7278, + "step": 7765 + }, + { + "epoch": 1.69, + "learning_rate": 4.157089796810138e-06, + "loss": 3.4646, + "step": 7766 + }, + { + "epoch": 1.7, + "learning_rate": 4.156980554948657e-06, + "loss": 3.0369, + "step": 7767 + }, + { + "epoch": 1.7, + "learning_rate": 4.156871313087175e-06, + "loss": 3.3485, + "step": 7768 + }, + { + "epoch": 1.7, + "learning_rate": 4.156762071225694e-06, + "loss": 3.5112, + "step": 7769 + }, + { + "epoch": 1.7, + "learning_rate": 4.1566528293642126e-06, + "loss": 2.5664, + "step": 7770 + }, + { + "epoch": 1.7, + "learning_rate": 4.156543587502732e-06, + "loss": 3.57, + "step": 7771 + }, + { + "epoch": 1.7, + "learning_rate": 4.15643434564125e-06, + "loss": 3.0727, + "step": 7772 + }, + { + "epoch": 1.7, + "learning_rate": 4.156325103779769e-06, + "loss": 2.7818, + "step": 7773 + }, + { + "epoch": 1.7, + "learning_rate": 4.156215861918287e-06, + "loss": 3.2226, + "step": 7774 + }, + { + "epoch": 1.7, + "learning_rate": 4.156106620056806e-06, + "loss": 2.9893, + "step": 7775 + }, + { + "epoch": 1.7, + "learning_rate": 4.155997378195325e-06, + "loss": 4.1549, + "step": 7776 + }, + { + "epoch": 1.7, + "learning_rate": 4.155888136333843e-06, + "loss": 2.2562, + "step": 7777 + }, + { + "epoch": 1.7, + "learning_rate": 4.155778894472362e-06, + "loss": 2.1433, + "step": 7778 + }, + { + "epoch": 1.7, + "learning_rate": 4.1556696526108805e-06, + "loss": 2.4672, + "step": 7779 + }, + { + "epoch": 1.7, + "learning_rate": 4.1555604107493995e-06, + "loss": 3.7735, + "step": 7780 + }, + { + "epoch": 1.7, + "learning_rate": 4.1554511688879186e-06, + "loss": 1.8705, + "step": 7781 + }, + { + "epoch": 1.7, + "learning_rate": 4.155341927026437e-06, + "loss": 2.8894, + "step": 7782 + }, + { + "epoch": 1.7, + "learning_rate": 4.155232685164956e-06, + "loss": 2.7717, + "step": 7783 + }, + { + "epoch": 1.7, + "learning_rate": 4.155123443303474e-06, + "loss": 2.4771, + "step": 7784 + }, + { + "epoch": 1.7, + "learning_rate": 4.155014201441993e-06, + "loss": 2.3911, + "step": 7785 + }, + { + "epoch": 1.7, + "learning_rate": 4.154904959580511e-06, + "loss": 4.3776, + "step": 7786 + }, + { + "epoch": 1.7, + "learning_rate": 4.15479571771903e-06, + "loss": 3.2191, + "step": 7787 + }, + { + "epoch": 1.7, + "learning_rate": 4.154686475857549e-06, + "loss": 2.9982, + "step": 7788 + }, + { + "epoch": 1.7, + "learning_rate": 4.1545772339960674e-06, + "loss": 2.9484, + "step": 7789 + }, + { + "epoch": 1.7, + "learning_rate": 4.1544679921345865e-06, + "loss": 2.7654, + "step": 7790 + }, + { + "epoch": 1.7, + "learning_rate": 4.154358750273105e-06, + "loss": 2.1387, + "step": 7791 + }, + { + "epoch": 1.7, + "learning_rate": 4.154249508411624e-06, + "loss": 1.4713, + "step": 7792 + }, + { + "epoch": 1.7, + "learning_rate": 4.154140266550143e-06, + "loss": 3.2613, + "step": 7793 + }, + { + "epoch": 1.7, + "learning_rate": 4.154031024688661e-06, + "loss": 2.4867, + "step": 7794 + }, + { + "epoch": 1.7, + "learning_rate": 4.15392178282718e-06, + "loss": 2.9846, + "step": 7795 + }, + { + "epoch": 1.7, + "learning_rate": 4.153812540965698e-06, + "loss": 3.1868, + "step": 7796 + }, + { + "epoch": 1.7, + "learning_rate": 4.153703299104217e-06, + "loss": 2.2751, + "step": 7797 + }, + { + "epoch": 1.7, + "learning_rate": 4.153594057242736e-06, + "loss": 2.9496, + "step": 7798 + }, + { + "epoch": 1.7, + "learning_rate": 4.153484815381254e-06, + "loss": 2.0956, + "step": 7799 + }, + { + "epoch": 1.7, + "learning_rate": 4.1533755735197734e-06, + "loss": 3.2713, + "step": 7800 + }, + { + "epoch": 1.7, + "learning_rate": 4.153266331658292e-06, + "loss": 3.2965, + "step": 7801 + }, + { + "epoch": 1.7, + "learning_rate": 4.153157089796811e-06, + "loss": 2.3765, + "step": 7802 + }, + { + "epoch": 1.7, + "learning_rate": 4.153047847935329e-06, + "loss": 3.1905, + "step": 7803 + }, + { + "epoch": 1.7, + "learning_rate": 4.152938606073848e-06, + "loss": 3.2128, + "step": 7804 + }, + { + "epoch": 1.7, + "learning_rate": 4.152829364212366e-06, + "loss": 2.8492, + "step": 7805 + }, + { + "epoch": 1.7, + "learning_rate": 4.152720122350885e-06, + "loss": 3.9284, + "step": 7806 + }, + { + "epoch": 1.7, + "learning_rate": 4.152610880489404e-06, + "loss": 2.0872, + "step": 7807 + }, + { + "epoch": 1.7, + "learning_rate": 4.152501638627922e-06, + "loss": 2.3456, + "step": 7808 + }, + { + "epoch": 1.7, + "learning_rate": 4.152392396766441e-06, + "loss": 2.7004, + "step": 7809 + }, + { + "epoch": 1.7, + "learning_rate": 4.1522831549049595e-06, + "loss": 2.7625, + "step": 7810 + }, + { + "epoch": 1.7, + "learning_rate": 4.1521739130434786e-06, + "loss": 3.3786, + "step": 7811 + }, + { + "epoch": 1.7, + "learning_rate": 4.152064671181998e-06, + "loss": 1.7297, + "step": 7812 + }, + { + "epoch": 1.71, + "learning_rate": 4.151955429320516e-06, + "loss": 2.1973, + "step": 7813 + }, + { + "epoch": 1.71, + "learning_rate": 4.151846187459035e-06, + "loss": 2.6948, + "step": 7814 + }, + { + "epoch": 1.71, + "learning_rate": 4.151736945597553e-06, + "loss": 1.7932, + "step": 7815 + }, + { + "epoch": 1.71, + "learning_rate": 4.151627703736072e-06, + "loss": 3.096, + "step": 7816 + }, + { + "epoch": 1.71, + "learning_rate": 4.151518461874591e-06, + "loss": 3.164, + "step": 7817 + }, + { + "epoch": 1.71, + "learning_rate": 4.151409220013109e-06, + "loss": 3.2357, + "step": 7818 + }, + { + "epoch": 1.71, + "learning_rate": 4.1512999781516275e-06, + "loss": 3.0593, + "step": 7819 + }, + { + "epoch": 1.71, + "learning_rate": 4.1511907362901465e-06, + "loss": 2.9713, + "step": 7820 + }, + { + "epoch": 1.71, + "learning_rate": 4.1510814944286655e-06, + "loss": 2.7187, + "step": 7821 + }, + { + "epoch": 1.71, + "learning_rate": 4.1509722525671846e-06, + "loss": 2.2689, + "step": 7822 + }, + { + "epoch": 1.71, + "learning_rate": 4.150863010705703e-06, + "loss": 2.6523, + "step": 7823 + }, + { + "epoch": 1.71, + "learning_rate": 4.150753768844221e-06, + "loss": 2.1637, + "step": 7824 + }, + { + "epoch": 1.71, + "learning_rate": 4.15064452698274e-06, + "loss": 2.0518, + "step": 7825 + }, + { + "epoch": 1.71, + "learning_rate": 4.150535285121259e-06, + "loss": 3.8545, + "step": 7826 + }, + { + "epoch": 1.71, + "learning_rate": 4.150426043259777e-06, + "loss": 2.7328, + "step": 7827 + }, + { + "epoch": 1.71, + "learning_rate": 4.150316801398296e-06, + "loss": 3.1327, + "step": 7828 + }, + { + "epoch": 1.71, + "learning_rate": 4.150207559536814e-06, + "loss": 2.6374, + "step": 7829 + }, + { + "epoch": 1.71, + "learning_rate": 4.1500983176753335e-06, + "loss": 2.6893, + "step": 7830 + }, + { + "epoch": 1.71, + "learning_rate": 4.1499890758138525e-06, + "loss": 3.1058, + "step": 7831 + }, + { + "epoch": 1.71, + "learning_rate": 4.149879833952371e-06, + "loss": 2.2546, + "step": 7832 + }, + { + "epoch": 1.71, + "learning_rate": 4.14977059209089e-06, + "loss": 1.5029, + "step": 7833 + }, + { + "epoch": 1.71, + "learning_rate": 4.149661350229408e-06, + "loss": 2.8098, + "step": 7834 + }, + { + "epoch": 1.71, + "learning_rate": 4.149552108367927e-06, + "loss": 2.6296, + "step": 7835 + }, + { + "epoch": 1.71, + "learning_rate": 4.149442866506446e-06, + "loss": 2.949, + "step": 7836 + }, + { + "epoch": 1.71, + "learning_rate": 4.149333624644964e-06, + "loss": 2.556, + "step": 7837 + }, + { + "epoch": 1.71, + "learning_rate": 4.149224382783482e-06, + "loss": 2.9795, + "step": 7838 + }, + { + "epoch": 1.71, + "learning_rate": 4.149115140922001e-06, + "loss": 3.5951, + "step": 7839 + }, + { + "epoch": 1.71, + "learning_rate": 4.14900589906052e-06, + "loss": 3.2987, + "step": 7840 + }, + { + "epoch": 1.71, + "learning_rate": 4.1488966571990394e-06, + "loss": 2.1402, + "step": 7841 + }, + { + "epoch": 1.71, + "learning_rate": 4.148787415337558e-06, + "loss": 3.4883, + "step": 7842 + }, + { + "epoch": 1.71, + "learning_rate": 4.148678173476076e-06, + "loss": 3.2751, + "step": 7843 + }, + { + "epoch": 1.71, + "learning_rate": 4.148568931614595e-06, + "loss": 2.1043, + "step": 7844 + }, + { + "epoch": 1.71, + "learning_rate": 4.148459689753114e-06, + "loss": 3.9023, + "step": 7845 + }, + { + "epoch": 1.71, + "learning_rate": 4.148350447891633e-06, + "loss": 3.1208, + "step": 7846 + }, + { + "epoch": 1.71, + "learning_rate": 4.148241206030151e-06, + "loss": 2.2831, + "step": 7847 + }, + { + "epoch": 1.71, + "learning_rate": 4.148131964168669e-06, + "loss": 2.0428, + "step": 7848 + }, + { + "epoch": 1.71, + "learning_rate": 4.148022722307188e-06, + "loss": 2.7892, + "step": 7849 + }, + { + "epoch": 1.71, + "learning_rate": 4.147913480445707e-06, + "loss": 2.2946, + "step": 7850 + }, + { + "epoch": 1.71, + "learning_rate": 4.147804238584226e-06, + "loss": 2.6042, + "step": 7851 + }, + { + "epoch": 1.71, + "learning_rate": 4.147694996722745e-06, + "loss": 1.5143, + "step": 7852 + }, + { + "epoch": 1.71, + "learning_rate": 4.147585754861263e-06, + "loss": 3.3946, + "step": 7853 + }, + { + "epoch": 1.71, + "learning_rate": 4.147476512999782e-06, + "loss": 3.5967, + "step": 7854 + }, + { + "epoch": 1.71, + "learning_rate": 4.147367271138301e-06, + "loss": 2.7341, + "step": 7855 + }, + { + "epoch": 1.71, + "learning_rate": 4.147258029276819e-06, + "loss": 2.3314, + "step": 7856 + }, + { + "epoch": 1.71, + "learning_rate": 4.147148787415338e-06, + "loss": 2.334, + "step": 7857 + }, + { + "epoch": 1.71, + "learning_rate": 4.147039545553856e-06, + "loss": 2.7614, + "step": 7858 + }, + { + "epoch": 1.72, + "learning_rate": 4.146930303692375e-06, + "loss": 2.183, + "step": 7859 + }, + { + "epoch": 1.72, + "learning_rate": 4.146821061830894e-06, + "loss": 3.8541, + "step": 7860 + }, + { + "epoch": 1.72, + "learning_rate": 4.1467118199694125e-06, + "loss": 3.4286, + "step": 7861 + }, + { + "epoch": 1.72, + "learning_rate": 4.146602578107931e-06, + "loss": 3.4901, + "step": 7862 + }, + { + "epoch": 1.72, + "learning_rate": 4.14649333624645e-06, + "loss": 3.223, + "step": 7863 + }, + { + "epoch": 1.72, + "learning_rate": 4.146384094384969e-06, + "loss": 2.507, + "step": 7864 + }, + { + "epoch": 1.72, + "learning_rate": 4.146274852523488e-06, + "loss": 2.9155, + "step": 7865 + }, + { + "epoch": 1.72, + "learning_rate": 4.146165610662006e-06, + "loss": 3.1707, + "step": 7866 + }, + { + "epoch": 1.72, + "learning_rate": 4.146056368800524e-06, + "loss": 2.7497, + "step": 7867 + }, + { + "epoch": 1.72, + "learning_rate": 4.145947126939043e-06, + "loss": 2.3942, + "step": 7868 + }, + { + "epoch": 1.72, + "learning_rate": 4.145837885077562e-06, + "loss": 1.9386, + "step": 7869 + }, + { + "epoch": 1.72, + "learning_rate": 4.145728643216081e-06, + "loss": 2.7343, + "step": 7870 + }, + { + "epoch": 1.72, + "learning_rate": 4.1456194013545995e-06, + "loss": 3.1336, + "step": 7871 + }, + { + "epoch": 1.72, + "learning_rate": 4.145510159493118e-06, + "loss": 2.7739, + "step": 7872 + }, + { + "epoch": 1.72, + "learning_rate": 4.145400917631637e-06, + "loss": 3.7139, + "step": 7873 + }, + { + "epoch": 1.72, + "learning_rate": 4.145291675770156e-06, + "loss": 3.1065, + "step": 7874 + }, + { + "epoch": 1.72, + "learning_rate": 4.145182433908675e-06, + "loss": 1.7518, + "step": 7875 + }, + { + "epoch": 1.72, + "learning_rate": 4.145073192047193e-06, + "loss": 2.038, + "step": 7876 + }, + { + "epoch": 1.72, + "learning_rate": 4.144963950185711e-06, + "loss": 3.7249, + "step": 7877 + }, + { + "epoch": 1.72, + "learning_rate": 4.14485470832423e-06, + "loss": 2.3226, + "step": 7878 + }, + { + "epoch": 1.72, + "learning_rate": 4.144745466462749e-06, + "loss": 2.3138, + "step": 7879 + }, + { + "epoch": 1.72, + "learning_rate": 4.144636224601267e-06, + "loss": 3.0153, + "step": 7880 + }, + { + "epoch": 1.72, + "learning_rate": 4.1445269827397864e-06, + "loss": 2.1753, + "step": 7881 + }, + { + "epoch": 1.72, + "learning_rate": 4.144417740878305e-06, + "loss": 2.1492, + "step": 7882 + }, + { + "epoch": 1.72, + "learning_rate": 4.144308499016824e-06, + "loss": 4.0816, + "step": 7883 + }, + { + "epoch": 1.72, + "learning_rate": 4.144199257155343e-06, + "loss": 2.4166, + "step": 7884 + }, + { + "epoch": 1.72, + "learning_rate": 4.144090015293861e-06, + "loss": 2.8129, + "step": 7885 + }, + { + "epoch": 1.72, + "learning_rate": 4.143980773432379e-06, + "loss": 2.4601, + "step": 7886 + }, + { + "epoch": 1.72, + "learning_rate": 4.143871531570898e-06, + "loss": 2.312, + "step": 7887 + }, + { + "epoch": 1.72, + "learning_rate": 4.143762289709417e-06, + "loss": 3.1068, + "step": 7888 + }, + { + "epoch": 1.72, + "learning_rate": 4.143653047847936e-06, + "loss": 2.8027, + "step": 7889 + }, + { + "epoch": 1.72, + "learning_rate": 4.143543805986454e-06, + "loss": 2.7666, + "step": 7890 + }, + { + "epoch": 1.72, + "learning_rate": 4.1434345641249725e-06, + "loss": 3.0998, + "step": 7891 + }, + { + "epoch": 1.72, + "learning_rate": 4.1433253222634916e-06, + "loss": 2.3774, + "step": 7892 + }, + { + "epoch": 1.72, + "learning_rate": 4.143216080402011e-06, + "loss": 3.3162, + "step": 7893 + }, + { + "epoch": 1.72, + "learning_rate": 4.14310683854053e-06, + "loss": 2.6723, + "step": 7894 + }, + { + "epoch": 1.72, + "learning_rate": 4.142997596679048e-06, + "loss": 3.5272, + "step": 7895 + }, + { + "epoch": 1.72, + "learning_rate": 4.142888354817566e-06, + "loss": 2.2352, + "step": 7896 + }, + { + "epoch": 1.72, + "learning_rate": 4.142779112956085e-06, + "loss": 3.6997, + "step": 7897 + }, + { + "epoch": 1.72, + "learning_rate": 4.142669871094604e-06, + "loss": 3.5985, + "step": 7898 + }, + { + "epoch": 1.72, + "learning_rate": 4.142560629233123e-06, + "loss": 2.3183, + "step": 7899 + }, + { + "epoch": 1.72, + "learning_rate": 4.142451387371641e-06, + "loss": 2.6634, + "step": 7900 + }, + { + "epoch": 1.72, + "learning_rate": 4.1423421455101595e-06, + "loss": 2.1893, + "step": 7901 + }, + { + "epoch": 1.72, + "learning_rate": 4.1422329036486785e-06, + "loss": 2.5654, + "step": 7902 + }, + { + "epoch": 1.72, + "learning_rate": 4.1421236617871976e-06, + "loss": 2.6544, + "step": 7903 + }, + { + "epoch": 1.73, + "learning_rate": 4.142014419925716e-06, + "loss": 2.2599, + "step": 7904 + }, + { + "epoch": 1.73, + "learning_rate": 4.141905178064235e-06, + "loss": 3.2377, + "step": 7905 + }, + { + "epoch": 1.73, + "learning_rate": 4.141795936202753e-06, + "loss": 2.831, + "step": 7906 + }, + { + "epoch": 1.73, + "learning_rate": 4.141686694341272e-06, + "loss": 2.8751, + "step": 7907 + }, + { + "epoch": 1.73, + "learning_rate": 4.141577452479791e-06, + "loss": 2.703, + "step": 7908 + }, + { + "epoch": 1.73, + "learning_rate": 4.141468210618309e-06, + "loss": 2.713, + "step": 7909 + }, + { + "epoch": 1.73, + "learning_rate": 4.141358968756828e-06, + "loss": 2.3707, + "step": 7910 + }, + { + "epoch": 1.73, + "learning_rate": 4.1412497268953464e-06, + "loss": 2.1297, + "step": 7911 + }, + { + "epoch": 1.73, + "learning_rate": 4.1411404850338655e-06, + "loss": 1.7416, + "step": 7912 + }, + { + "epoch": 1.73, + "learning_rate": 4.1410312431723845e-06, + "loss": 2.897, + "step": 7913 + }, + { + "epoch": 1.73, + "learning_rate": 4.140922001310903e-06, + "loss": 2.4966, + "step": 7914 + }, + { + "epoch": 1.73, + "learning_rate": 4.140812759449421e-06, + "loss": 2.5986, + "step": 7915 + }, + { + "epoch": 1.73, + "learning_rate": 4.14070351758794e-06, + "loss": 3.1992, + "step": 7916 + }, + { + "epoch": 1.73, + "learning_rate": 4.140594275726459e-06, + "loss": 3.032, + "step": 7917 + }, + { + "epoch": 1.73, + "learning_rate": 4.140485033864977e-06, + "loss": 3.6957, + "step": 7918 + }, + { + "epoch": 1.73, + "learning_rate": 4.140375792003496e-06, + "loss": 3.2367, + "step": 7919 + }, + { + "epoch": 1.73, + "learning_rate": 4.140266550142014e-06, + "loss": 3.2674, + "step": 7920 + }, + { + "epoch": 1.73, + "learning_rate": 4.140157308280533e-06, + "loss": 3.1515, + "step": 7921 + }, + { + "epoch": 1.73, + "learning_rate": 4.1400480664190524e-06, + "loss": 2.8253, + "step": 7922 + }, + { + "epoch": 1.73, + "learning_rate": 4.139938824557571e-06, + "loss": 3.351, + "step": 7923 + }, + { + "epoch": 1.73, + "learning_rate": 4.13982958269609e-06, + "loss": 2.3743, + "step": 7924 + }, + { + "epoch": 1.73, + "learning_rate": 4.139720340834608e-06, + "loss": 2.4727, + "step": 7925 + }, + { + "epoch": 1.73, + "learning_rate": 4.139611098973127e-06, + "loss": 3.4622, + "step": 7926 + }, + { + "epoch": 1.73, + "learning_rate": 4.139501857111646e-06, + "loss": 2.8505, + "step": 7927 + }, + { + "epoch": 1.73, + "learning_rate": 4.139392615250164e-06, + "loss": 2.8017, + "step": 7928 + }, + { + "epoch": 1.73, + "learning_rate": 4.139283373388683e-06, + "loss": 2.7945, + "step": 7929 + }, + { + "epoch": 1.73, + "learning_rate": 4.139174131527201e-06, + "loss": 2.7885, + "step": 7930 + }, + { + "epoch": 1.73, + "learning_rate": 4.13906488966572e-06, + "loss": 2.5203, + "step": 7931 + }, + { + "epoch": 1.73, + "learning_rate": 4.138955647804239e-06, + "loss": 2.8601, + "step": 7932 + }, + { + "epoch": 1.73, + "learning_rate": 4.138846405942758e-06, + "loss": 2.0978, + "step": 7933 + }, + { + "epoch": 1.73, + "learning_rate": 4.138737164081277e-06, + "loss": 2.2486, + "step": 7934 + }, + { + "epoch": 1.73, + "learning_rate": 4.138627922219795e-06, + "loss": 2.9848, + "step": 7935 + }, + { + "epoch": 1.73, + "learning_rate": 4.138518680358314e-06, + "loss": 2.9029, + "step": 7936 + }, + { + "epoch": 1.73, + "learning_rate": 4.138409438496832e-06, + "loss": 2.5081, + "step": 7937 + }, + { + "epoch": 1.73, + "learning_rate": 4.138300196635351e-06, + "loss": 2.696, + "step": 7938 + }, + { + "epoch": 1.73, + "learning_rate": 4.138190954773869e-06, + "loss": 3.1613, + "step": 7939 + }, + { + "epoch": 1.73, + "learning_rate": 4.138081712912388e-06, + "loss": 2.8197, + "step": 7940 + }, + { + "epoch": 1.73, + "learning_rate": 4.137972471050907e-06, + "loss": 3.6922, + "step": 7941 + }, + { + "epoch": 1.73, + "learning_rate": 4.1378632291894255e-06, + "loss": 3.6209, + "step": 7942 + }, + { + "epoch": 1.73, + "learning_rate": 4.1377539873279445e-06, + "loss": 3.4703, + "step": 7943 + }, + { + "epoch": 1.73, + "learning_rate": 4.137644745466463e-06, + "loss": 3.2872, + "step": 7944 + }, + { + "epoch": 1.73, + "learning_rate": 4.137535503604982e-06, + "loss": 2.0067, + "step": 7945 + }, + { + "epoch": 1.73, + "learning_rate": 4.137426261743501e-06, + "loss": 2.6807, + "step": 7946 + }, + { + "epoch": 1.73, + "learning_rate": 4.137317019882019e-06, + "loss": 2.4971, + "step": 7947 + }, + { + "epoch": 1.73, + "learning_rate": 4.137207778020538e-06, + "loss": 1.9045, + "step": 7948 + }, + { + "epoch": 1.73, + "learning_rate": 4.137098536159056e-06, + "loss": 2.6716, + "step": 7949 + }, + { + "epoch": 1.74, + "learning_rate": 4.136989294297575e-06, + "loss": 2.0751, + "step": 7950 + }, + { + "epoch": 1.74, + "learning_rate": 4.1368800524360934e-06, + "loss": 2.1506, + "step": 7951 + }, + { + "epoch": 1.74, + "learning_rate": 4.1367708105746125e-06, + "loss": 3.9898, + "step": 7952 + }, + { + "epoch": 1.74, + "learning_rate": 4.1366615687131315e-06, + "loss": 3.0112, + "step": 7953 + }, + { + "epoch": 1.74, + "learning_rate": 4.13655232685165e-06, + "loss": 2.9439, + "step": 7954 + }, + { + "epoch": 1.74, + "learning_rate": 4.136443084990169e-06, + "loss": 3.5527, + "step": 7955 + }, + { + "epoch": 1.74, + "learning_rate": 4.136333843128687e-06, + "loss": 2.3158, + "step": 7956 + }, + { + "epoch": 1.74, + "learning_rate": 4.136224601267206e-06, + "loss": 2.155, + "step": 7957 + }, + { + "epoch": 1.74, + "learning_rate": 4.136115359405725e-06, + "loss": 1.9553, + "step": 7958 + }, + { + "epoch": 1.74, + "learning_rate": 4.136006117544243e-06, + "loss": 3.0945, + "step": 7959 + }, + { + "epoch": 1.74, + "learning_rate": 4.135896875682762e-06, + "loss": 3.2518, + "step": 7960 + }, + { + "epoch": 1.74, + "learning_rate": 4.13578763382128e-06, + "loss": 2.5417, + "step": 7961 + }, + { + "epoch": 1.74, + "learning_rate": 4.135678391959799e-06, + "loss": 2.5421, + "step": 7962 + }, + { + "epoch": 1.74, + "learning_rate": 4.1355691500983185e-06, + "loss": 3.3446, + "step": 7963 + }, + { + "epoch": 1.74, + "learning_rate": 4.135459908236837e-06, + "loss": 2.4155, + "step": 7964 + }, + { + "epoch": 1.74, + "learning_rate": 4.135350666375356e-06, + "loss": 2.077, + "step": 7965 + }, + { + "epoch": 1.74, + "learning_rate": 4.135241424513874e-06, + "loss": 2.7254, + "step": 7966 + }, + { + "epoch": 1.74, + "learning_rate": 4.135132182652393e-06, + "loss": 1.4817, + "step": 7967 + }, + { + "epoch": 1.74, + "learning_rate": 4.135022940790911e-06, + "loss": 2.1044, + "step": 7968 + }, + { + "epoch": 1.74, + "learning_rate": 4.13491369892943e-06, + "loss": 2.7683, + "step": 7969 + }, + { + "epoch": 1.74, + "learning_rate": 4.134804457067948e-06, + "loss": 2.1768, + "step": 7970 + }, + { + "epoch": 1.74, + "learning_rate": 4.134695215206467e-06, + "loss": 2.6944, + "step": 7971 + }, + { + "epoch": 1.74, + "learning_rate": 4.134585973344986e-06, + "loss": 1.8075, + "step": 7972 + }, + { + "epoch": 1.74, + "learning_rate": 4.1344767314835046e-06, + "loss": 3.2787, + "step": 7973 + }, + { + "epoch": 1.74, + "learning_rate": 4.134367489622024e-06, + "loss": 2.6918, + "step": 7974 + }, + { + "epoch": 1.74, + "learning_rate": 4.134258247760542e-06, + "loss": 3.0866, + "step": 7975 + }, + { + "epoch": 1.74, + "learning_rate": 4.134149005899061e-06, + "loss": 2.1403, + "step": 7976 + }, + { + "epoch": 1.74, + "learning_rate": 4.13403976403758e-06, + "loss": 2.4349, + "step": 7977 + }, + { + "epoch": 1.74, + "learning_rate": 4.133930522176098e-06, + "loss": 2.41, + "step": 7978 + }, + { + "epoch": 1.74, + "learning_rate": 4.133821280314617e-06, + "loss": 2.2459, + "step": 7979 + }, + { + "epoch": 1.74, + "learning_rate": 4.133712038453135e-06, + "loss": 2.8645, + "step": 7980 + }, + { + "epoch": 1.74, + "learning_rate": 4.133602796591654e-06, + "loss": 2.4286, + "step": 7981 + }, + { + "epoch": 1.74, + "learning_rate": 4.133493554730173e-06, + "loss": 2.2309, + "step": 7982 + }, + { + "epoch": 1.74, + "learning_rate": 4.1333843128686915e-06, + "loss": 3.8414, + "step": 7983 + }, + { + "epoch": 1.74, + "learning_rate": 4.13327507100721e-06, + "loss": 3.4604, + "step": 7984 + }, + { + "epoch": 1.74, + "learning_rate": 4.133165829145729e-06, + "loss": 2.763, + "step": 7985 + }, + { + "epoch": 1.74, + "learning_rate": 4.133056587284248e-06, + "loss": 3.4303, + "step": 7986 + }, + { + "epoch": 1.74, + "learning_rate": 4.132947345422767e-06, + "loss": 2.3455, + "step": 7987 + }, + { + "epoch": 1.74, + "learning_rate": 4.132838103561285e-06, + "loss": 2.5401, + "step": 7988 + }, + { + "epoch": 1.74, + "learning_rate": 4.132728861699803e-06, + "loss": 3.5047, + "step": 7989 + }, + { + "epoch": 1.74, + "learning_rate": 4.132619619838322e-06, + "loss": 2.2948, + "step": 7990 + }, + { + "epoch": 1.74, + "learning_rate": 4.132510377976841e-06, + "loss": 3.1211, + "step": 7991 + }, + { + "epoch": 1.74, + "learning_rate": 4.1324011361153594e-06, + "loss": 2.2242, + "step": 7992 + }, + { + "epoch": 1.74, + "learning_rate": 4.1322918942538785e-06, + "loss": 2.5584, + "step": 7993 + }, + { + "epoch": 1.74, + "learning_rate": 4.132182652392397e-06, + "loss": 2.4209, + "step": 7994 + }, + { + "epoch": 1.74, + "learning_rate": 4.132073410530916e-06, + "loss": 3.5455, + "step": 7995 + }, + { + "epoch": 1.75, + "learning_rate": 4.131964168669435e-06, + "loss": 3.7333, + "step": 7996 + }, + { + "epoch": 1.75, + "learning_rate": 4.131854926807953e-06, + "loss": 2.508, + "step": 7997 + }, + { + "epoch": 1.75, + "learning_rate": 4.131745684946472e-06, + "loss": 2.5999, + "step": 7998 + }, + { + "epoch": 1.75, + "learning_rate": 4.13163644308499e-06, + "loss": 2.1454, + "step": 7999 + }, + { + "epoch": 1.75, + "learning_rate": 4.131527201223509e-06, + "loss": 3.1673, + "step": 8000 + }, + { + "epoch": 1.75, + "learning_rate": 4.131417959362028e-06, + "loss": 2.5742, + "step": 8001 + }, + { + "epoch": 1.75, + "learning_rate": 4.131308717500546e-06, + "loss": 2.5901, + "step": 8002 + }, + { + "epoch": 1.75, + "learning_rate": 4.131199475639065e-06, + "loss": 3.168, + "step": 8003 + }, + { + "epoch": 1.75, + "learning_rate": 4.131090233777584e-06, + "loss": 2.8073, + "step": 8004 + }, + { + "epoch": 1.75, + "learning_rate": 4.130980991916103e-06, + "loss": 2.6938, + "step": 8005 + }, + { + "epoch": 1.75, + "learning_rate": 4.130871750054622e-06, + "loss": 1.8773, + "step": 8006 + }, + { + "epoch": 1.75, + "learning_rate": 4.13076250819314e-06, + "loss": 2.8284, + "step": 8007 + }, + { + "epoch": 1.75, + "learning_rate": 4.130653266331658e-06, + "loss": 2.9818, + "step": 8008 + }, + { + "epoch": 1.75, + "learning_rate": 4.130544024470177e-06, + "loss": 2.6113, + "step": 8009 + }, + { + "epoch": 1.75, + "learning_rate": 4.130434782608696e-06, + "loss": 2.2033, + "step": 8010 + }, + { + "epoch": 1.75, + "learning_rate": 4.130325540747215e-06, + "loss": 2.1153, + "step": 8011 + }, + { + "epoch": 1.75, + "learning_rate": 4.130216298885733e-06, + "loss": 2.2192, + "step": 8012 + }, + { + "epoch": 1.75, + "learning_rate": 4.1301070570242515e-06, + "loss": 2.4875, + "step": 8013 + }, + { + "epoch": 1.75, + "learning_rate": 4.129997815162771e-06, + "loss": 1.8855, + "step": 8014 + }, + { + "epoch": 1.75, + "learning_rate": 4.12988857330129e-06, + "loss": 3.298, + "step": 8015 + }, + { + "epoch": 1.75, + "learning_rate": 4.129779331439809e-06, + "loss": 2.8597, + "step": 8016 + }, + { + "epoch": 1.75, + "learning_rate": 4.129670089578327e-06, + "loss": 3.229, + "step": 8017 + }, + { + "epoch": 1.75, + "learning_rate": 4.129560847716845e-06, + "loss": 2.7418, + "step": 8018 + }, + { + "epoch": 1.75, + "learning_rate": 4.129451605855364e-06, + "loss": 2.8793, + "step": 8019 + }, + { + "epoch": 1.75, + "learning_rate": 4.129342363993883e-06, + "loss": 3.1432, + "step": 8020 + }, + { + "epoch": 1.75, + "learning_rate": 4.129233122132401e-06, + "loss": 2.4289, + "step": 8021 + }, + { + "epoch": 1.75, + "learning_rate": 4.12912388027092e-06, + "loss": 2.1466, + "step": 8022 + }, + { + "epoch": 1.75, + "learning_rate": 4.1290146384094385e-06, + "loss": 2.8166, + "step": 8023 + }, + { + "epoch": 1.75, + "learning_rate": 4.1289053965479575e-06, + "loss": 2.9789, + "step": 8024 + }, + { + "epoch": 1.75, + "learning_rate": 4.1287961546864766e-06, + "loss": 3.5802, + "step": 8025 + }, + { + "epoch": 1.75, + "learning_rate": 4.128686912824995e-06, + "loss": 2.7781, + "step": 8026 + }, + { + "epoch": 1.75, + "learning_rate": 4.128577670963513e-06, + "loss": 2.6198, + "step": 8027 + }, + { + "epoch": 1.75, + "learning_rate": 4.128468429102032e-06, + "loss": 2.7625, + "step": 8028 + }, + { + "epoch": 1.75, + "learning_rate": 4.128359187240551e-06, + "loss": 3.3185, + "step": 8029 + }, + { + "epoch": 1.75, + "learning_rate": 4.12824994537907e-06, + "loss": 2.4956, + "step": 8030 + }, + { + "epoch": 1.75, + "learning_rate": 4.128140703517588e-06, + "loss": 3.1698, + "step": 8031 + }, + { + "epoch": 1.75, + "learning_rate": 4.1280314616561064e-06, + "loss": 3.1811, + "step": 8032 + }, + { + "epoch": 1.75, + "learning_rate": 4.1279222197946255e-06, + "loss": 3.0409, + "step": 8033 + }, + { + "epoch": 1.75, + "learning_rate": 4.1278129779331445e-06, + "loss": 2.2329, + "step": 8034 + }, + { + "epoch": 1.75, + "learning_rate": 4.1277037360716635e-06, + "loss": 3.1415, + "step": 8035 + }, + { + "epoch": 1.75, + "learning_rate": 4.127594494210182e-06, + "loss": 2.7191, + "step": 8036 + }, + { + "epoch": 1.75, + "learning_rate": 4.1274852523487e-06, + "loss": 3.0851, + "step": 8037 + }, + { + "epoch": 1.75, + "learning_rate": 4.127376010487219e-06, + "loss": 2.6798, + "step": 8038 + }, + { + "epoch": 1.75, + "learning_rate": 4.127266768625738e-06, + "loss": 3.4185, + "step": 8039 + }, + { + "epoch": 1.75, + "learning_rate": 4.127157526764257e-06, + "loss": 2.4202, + "step": 8040 + }, + { + "epoch": 1.75, + "learning_rate": 4.127048284902775e-06, + "loss": 2.2418, + "step": 8041 + }, + { + "epoch": 1.76, + "learning_rate": 4.126939043041293e-06, + "loss": 2.5674, + "step": 8042 + }, + { + "epoch": 1.76, + "learning_rate": 4.126829801179812e-06, + "loss": 3.1102, + "step": 8043 + }, + { + "epoch": 1.76, + "learning_rate": 4.1267205593183315e-06, + "loss": 2.4722, + "step": 8044 + }, + { + "epoch": 1.76, + "learning_rate": 4.1266113174568505e-06, + "loss": 1.4469, + "step": 8045 + }, + { + "epoch": 1.76, + "learning_rate": 4.126502075595369e-06, + "loss": 2.8182, + "step": 8046 + }, + { + "epoch": 1.76, + "learning_rate": 4.126392833733887e-06, + "loss": 1.4083, + "step": 8047 + }, + { + "epoch": 1.76, + "learning_rate": 4.126283591872406e-06, + "loss": 2.9086, + "step": 8048 + }, + { + "epoch": 1.76, + "learning_rate": 4.126174350010925e-06, + "loss": 3.6724, + "step": 8049 + }, + { + "epoch": 1.76, + "learning_rate": 4.126065108149443e-06, + "loss": 1.7299, + "step": 8050 + }, + { + "epoch": 1.76, + "learning_rate": 4.125955866287962e-06, + "loss": 2.6853, + "step": 8051 + }, + { + "epoch": 1.76, + "learning_rate": 4.12584662442648e-06, + "loss": 3.359, + "step": 8052 + }, + { + "epoch": 1.76, + "learning_rate": 4.125737382564999e-06, + "loss": 3.0245, + "step": 8053 + }, + { + "epoch": 1.76, + "learning_rate": 4.125628140703518e-06, + "loss": 2.5454, + "step": 8054 + }, + { + "epoch": 1.76, + "learning_rate": 4.125518898842037e-06, + "loss": 3.2462, + "step": 8055 + }, + { + "epoch": 1.76, + "learning_rate": 4.125409656980555e-06, + "loss": 3.4589, + "step": 8056 + }, + { + "epoch": 1.76, + "learning_rate": 4.125300415119074e-06, + "loss": 3.642, + "step": 8057 + }, + { + "epoch": 1.76, + "learning_rate": 4.125191173257593e-06, + "loss": 2.7747, + "step": 8058 + }, + { + "epoch": 1.76, + "learning_rate": 4.125081931396112e-06, + "loss": 3.7196, + "step": 8059 + }, + { + "epoch": 1.76, + "learning_rate": 4.12497268953463e-06, + "loss": 2.0745, + "step": 8060 + }, + { + "epoch": 1.76, + "learning_rate": 4.124863447673148e-06, + "loss": 2.2597, + "step": 8061 + }, + { + "epoch": 1.76, + "learning_rate": 4.124754205811667e-06, + "loss": 1.9054, + "step": 8062 + }, + { + "epoch": 1.76, + "learning_rate": 4.124644963950186e-06, + "loss": 2.1302, + "step": 8063 + }, + { + "epoch": 1.76, + "learning_rate": 4.1245357220887045e-06, + "loss": 3.034, + "step": 8064 + }, + { + "epoch": 1.76, + "learning_rate": 4.1244264802272236e-06, + "loss": 3.1681, + "step": 8065 + }, + { + "epoch": 1.76, + "learning_rate": 4.124317238365742e-06, + "loss": 2.3027, + "step": 8066 + }, + { + "epoch": 1.76, + "learning_rate": 4.124207996504261e-06, + "loss": 2.1091, + "step": 8067 + }, + { + "epoch": 1.76, + "learning_rate": 4.12409875464278e-06, + "loss": 2.3833, + "step": 8068 + }, + { + "epoch": 1.76, + "learning_rate": 4.123989512781298e-06, + "loss": 3.1013, + "step": 8069 + }, + { + "epoch": 1.76, + "learning_rate": 4.123880270919817e-06, + "loss": 2.3247, + "step": 8070 + }, + { + "epoch": 1.76, + "learning_rate": 4.123771029058335e-06, + "loss": 2.1318, + "step": 8071 + }, + { + "epoch": 1.76, + "learning_rate": 4.123661787196854e-06, + "loss": 2.3679, + "step": 8072 + }, + { + "epoch": 1.76, + "learning_rate": 4.123552545335373e-06, + "loss": 2.2844, + "step": 8073 + }, + { + "epoch": 1.76, + "learning_rate": 4.1234433034738915e-06, + "loss": 2.4088, + "step": 8074 + }, + { + "epoch": 1.76, + "learning_rate": 4.1233340616124105e-06, + "loss": 3.3852, + "step": 8075 + }, + { + "epoch": 1.76, + "learning_rate": 4.123224819750929e-06, + "loss": 2.485, + "step": 8076 + }, + { + "epoch": 1.76, + "learning_rate": 4.123115577889448e-06, + "loss": 3.0446, + "step": 8077 + }, + { + "epoch": 1.76, + "learning_rate": 4.123006336027967e-06, + "loss": 2.2055, + "step": 8078 + }, + { + "epoch": 1.76, + "learning_rate": 4.122897094166485e-06, + "loss": 2.7528, + "step": 8079 + }, + { + "epoch": 1.76, + "learning_rate": 4.122787852305003e-06, + "loss": 2.8196, + "step": 8080 + }, + { + "epoch": 1.76, + "learning_rate": 4.122678610443522e-06, + "loss": 2.727, + "step": 8081 + }, + { + "epoch": 1.76, + "learning_rate": 4.122569368582041e-06, + "loss": 3.9113, + "step": 8082 + }, + { + "epoch": 1.76, + "learning_rate": 4.122460126720559e-06, + "loss": 3.9946, + "step": 8083 + }, + { + "epoch": 1.76, + "learning_rate": 4.1223508848590784e-06, + "loss": 3.6482, + "step": 8084 + }, + { + "epoch": 1.76, + "learning_rate": 4.122241642997597e-06, + "loss": 2.4552, + "step": 8085 + }, + { + "epoch": 1.76, + "learning_rate": 4.122132401136116e-06, + "loss": 2.6758, + "step": 8086 + }, + { + "epoch": 1.76, + "learning_rate": 4.122023159274635e-06, + "loss": 2.8212, + "step": 8087 + }, + { + "epoch": 1.77, + "learning_rate": 4.121913917413153e-06, + "loss": 3.3178, + "step": 8088 + }, + { + "epoch": 1.77, + "learning_rate": 4.121804675551672e-06, + "loss": 3.2671, + "step": 8089 + }, + { + "epoch": 1.77, + "learning_rate": 4.12169543369019e-06, + "loss": 2.694, + "step": 8090 + }, + { + "epoch": 1.77, + "learning_rate": 4.121586191828709e-06, + "loss": 1.9012, + "step": 8091 + }, + { + "epoch": 1.77, + "learning_rate": 4.121476949967228e-06, + "loss": 2.5194, + "step": 8092 + }, + { + "epoch": 1.77, + "learning_rate": 4.121367708105746e-06, + "loss": 3.8204, + "step": 8093 + }, + { + "epoch": 1.77, + "learning_rate": 4.121258466244265e-06, + "loss": 2.3148, + "step": 8094 + }, + { + "epoch": 1.77, + "learning_rate": 4.121149224382784e-06, + "loss": 2.1344, + "step": 8095 + }, + { + "epoch": 1.77, + "learning_rate": 4.121039982521303e-06, + "loss": 3.8936, + "step": 8096 + }, + { + "epoch": 1.77, + "learning_rate": 4.120930740659821e-06, + "loss": 3.0066, + "step": 8097 + }, + { + "epoch": 1.77, + "learning_rate": 4.12082149879834e-06, + "loss": 3.3787, + "step": 8098 + }, + { + "epoch": 1.77, + "learning_rate": 4.120712256936859e-06, + "loss": 2.9795, + "step": 8099 + }, + { + "epoch": 1.77, + "learning_rate": 4.120603015075377e-06, + "loss": 3.153, + "step": 8100 + }, + { + "epoch": 1.77, + "learning_rate": 4.120493773213896e-06, + "loss": 2.7694, + "step": 8101 + }, + { + "epoch": 1.77, + "learning_rate": 4.120384531352414e-06, + "loss": 2.82, + "step": 8102 + }, + { + "epoch": 1.77, + "learning_rate": 4.120275289490933e-06, + "loss": 2.8746, + "step": 8103 + }, + { + "epoch": 1.77, + "learning_rate": 4.120166047629452e-06, + "loss": 2.1218, + "step": 8104 + }, + { + "epoch": 1.77, + "learning_rate": 4.1200568057679705e-06, + "loss": 3.9638, + "step": 8105 + }, + { + "epoch": 1.77, + "learning_rate": 4.1199475639064896e-06, + "loss": 2.6504, + "step": 8106 + }, + { + "epoch": 1.77, + "learning_rate": 4.119838322045008e-06, + "loss": 2.4563, + "step": 8107 + }, + { + "epoch": 1.77, + "learning_rate": 4.119729080183527e-06, + "loss": 2.1516, + "step": 8108 + }, + { + "epoch": 1.77, + "learning_rate": 4.119619838322045e-06, + "loss": 2.8271, + "step": 8109 + }, + { + "epoch": 1.77, + "learning_rate": 4.119510596460564e-06, + "loss": 2.7255, + "step": 8110 + }, + { + "epoch": 1.77, + "learning_rate": 4.119401354599083e-06, + "loss": 3.0791, + "step": 8111 + }, + { + "epoch": 1.77, + "learning_rate": 4.119292112737601e-06, + "loss": 2.6749, + "step": 8112 + }, + { + "epoch": 1.77, + "learning_rate": 4.11918287087612e-06, + "loss": 3.2082, + "step": 8113 + }, + { + "epoch": 1.77, + "learning_rate": 4.1190736290146385e-06, + "loss": 2.4371, + "step": 8114 + }, + { + "epoch": 1.77, + "learning_rate": 4.1189643871531575e-06, + "loss": 2.288, + "step": 8115 + }, + { + "epoch": 1.77, + "learning_rate": 4.118855145291676e-06, + "loss": 2.7708, + "step": 8116 + }, + { + "epoch": 1.77, + "learning_rate": 4.118745903430195e-06, + "loss": 3.2135, + "step": 8117 + }, + { + "epoch": 1.77, + "learning_rate": 4.118636661568714e-06, + "loss": 1.9486, + "step": 8118 + }, + { + "epoch": 1.77, + "learning_rate": 4.118527419707232e-06, + "loss": 3.3537, + "step": 8119 + }, + { + "epoch": 1.77, + "learning_rate": 4.118418177845751e-06, + "loss": 2.9123, + "step": 8120 + }, + { + "epoch": 1.77, + "learning_rate": 4.118308935984269e-06, + "loss": 3.4936, + "step": 8121 + }, + { + "epoch": 1.77, + "learning_rate": 4.118199694122788e-06, + "loss": 2.8269, + "step": 8122 + }, + { + "epoch": 1.77, + "learning_rate": 4.118090452261307e-06, + "loss": 2.1302, + "step": 8123 + }, + { + "epoch": 1.77, + "learning_rate": 4.117981210399825e-06, + "loss": 2.3753, + "step": 8124 + }, + { + "epoch": 1.77, + "learning_rate": 4.1178719685383444e-06, + "loss": 3.206, + "step": 8125 + }, + { + "epoch": 1.77, + "learning_rate": 4.117762726676863e-06, + "loss": 2.8965, + "step": 8126 + }, + { + "epoch": 1.77, + "learning_rate": 4.117653484815382e-06, + "loss": 2.5874, + "step": 8127 + }, + { + "epoch": 1.77, + "learning_rate": 4.117544242953901e-06, + "loss": 3.1291, + "step": 8128 + }, + { + "epoch": 1.77, + "learning_rate": 4.117435001092419e-06, + "loss": 1.7647, + "step": 8129 + }, + { + "epoch": 1.77, + "learning_rate": 4.117325759230937e-06, + "loss": 2.1961, + "step": 8130 + }, + { + "epoch": 1.77, + "learning_rate": 4.117216517369456e-06, + "loss": 2.5122, + "step": 8131 + }, + { + "epoch": 1.77, + "learning_rate": 4.117107275507975e-06, + "loss": 2.1223, + "step": 8132 + }, + { + "epoch": 1.77, + "learning_rate": 4.116998033646493e-06, + "loss": 2.2747, + "step": 8133 + }, + { + "epoch": 1.78, + "learning_rate": 4.116888791785012e-06, + "loss": 2.694, + "step": 8134 + }, + { + "epoch": 1.78, + "learning_rate": 4.1167795499235306e-06, + "loss": 2.1947, + "step": 8135 + }, + { + "epoch": 1.78, + "learning_rate": 4.11667030806205e-06, + "loss": 3.8266, + "step": 8136 + }, + { + "epoch": 1.78, + "learning_rate": 4.116561066200569e-06, + "loss": 2.6635, + "step": 8137 + }, + { + "epoch": 1.78, + "learning_rate": 4.116451824339087e-06, + "loss": 3.3555, + "step": 8138 + }, + { + "epoch": 1.78, + "learning_rate": 4.116342582477606e-06, + "loss": 2.408, + "step": 8139 + }, + { + "epoch": 1.78, + "learning_rate": 4.116233340616124e-06, + "loss": 2.498, + "step": 8140 + }, + { + "epoch": 1.78, + "learning_rate": 4.116124098754643e-06, + "loss": 2.4445, + "step": 8141 + }, + { + "epoch": 1.78, + "learning_rate": 4.116014856893162e-06, + "loss": 4.2329, + "step": 8142 + }, + { + "epoch": 1.78, + "learning_rate": 4.11590561503168e-06, + "loss": 2.9259, + "step": 8143 + }, + { + "epoch": 1.78, + "learning_rate": 4.115796373170199e-06, + "loss": 2.5658, + "step": 8144 + }, + { + "epoch": 1.78, + "learning_rate": 4.1156871313087175e-06, + "loss": 2.1938, + "step": 8145 + }, + { + "epoch": 1.78, + "learning_rate": 4.1155778894472365e-06, + "loss": 2.3182, + "step": 8146 + }, + { + "epoch": 1.78, + "learning_rate": 4.115468647585756e-06, + "loss": 3.0708, + "step": 8147 + }, + { + "epoch": 1.78, + "learning_rate": 4.115359405724274e-06, + "loss": 1.7156, + "step": 8148 + }, + { + "epoch": 1.78, + "learning_rate": 4.115250163862792e-06, + "loss": 3.1835, + "step": 8149 + }, + { + "epoch": 1.78, + "learning_rate": 4.115140922001311e-06, + "loss": 3.1185, + "step": 8150 + }, + { + "epoch": 1.78, + "learning_rate": 4.11503168013983e-06, + "loss": 2.2662, + "step": 8151 + }, + { + "epoch": 1.78, + "learning_rate": 4.114922438278349e-06, + "loss": 2.1365, + "step": 8152 + }, + { + "epoch": 1.78, + "learning_rate": 4.114813196416867e-06, + "loss": 3.3787, + "step": 8153 + }, + { + "epoch": 1.78, + "learning_rate": 4.1147039545553854e-06, + "loss": 2.5666, + "step": 8154 + }, + { + "epoch": 1.78, + "learning_rate": 4.1145947126939045e-06, + "loss": 1.495, + "step": 8155 + }, + { + "epoch": 1.78, + "learning_rate": 4.1144854708324235e-06, + "loss": 3.1297, + "step": 8156 + }, + { + "epoch": 1.78, + "learning_rate": 4.1143762289709425e-06, + "loss": 1.9904, + "step": 8157 + }, + { + "epoch": 1.78, + "learning_rate": 4.114266987109461e-06, + "loss": 2.4676, + "step": 8158 + }, + { + "epoch": 1.78, + "learning_rate": 4.114157745247979e-06, + "loss": 2.6985, + "step": 8159 + }, + { + "epoch": 1.78, + "learning_rate": 4.114048503386498e-06, + "loss": 1.6265, + "step": 8160 + }, + { + "epoch": 1.78, + "learning_rate": 4.113939261525017e-06, + "loss": 3.0917, + "step": 8161 + }, + { + "epoch": 1.78, + "learning_rate": 4.113830019663535e-06, + "loss": 2.3908, + "step": 8162 + }, + { + "epoch": 1.78, + "learning_rate": 4.113720777802054e-06, + "loss": 3.7648, + "step": 8163 + }, + { + "epoch": 1.78, + "learning_rate": 4.113611535940572e-06, + "loss": 3.127, + "step": 8164 + }, + { + "epoch": 1.78, + "learning_rate": 4.1135022940790914e-06, + "loss": 3.4463, + "step": 8165 + }, + { + "epoch": 1.78, + "learning_rate": 4.1133930522176105e-06, + "loss": 2.6104, + "step": 8166 + }, + { + "epoch": 1.78, + "learning_rate": 4.113283810356129e-06, + "loss": 2.3066, + "step": 8167 + }, + { + "epoch": 1.78, + "learning_rate": 4.113174568494647e-06, + "loss": 2.3495, + "step": 8168 + }, + { + "epoch": 1.78, + "learning_rate": 4.113065326633166e-06, + "loss": 2.7519, + "step": 8169 + }, + { + "epoch": 1.78, + "learning_rate": 4.112956084771685e-06, + "loss": 3.1045, + "step": 8170 + }, + { + "epoch": 1.78, + "learning_rate": 4.112846842910204e-06, + "loss": 3.4069, + "step": 8171 + }, + { + "epoch": 1.78, + "learning_rate": 4.112737601048722e-06, + "loss": 2.683, + "step": 8172 + }, + { + "epoch": 1.78, + "learning_rate": 4.11262835918724e-06, + "loss": 3.0799, + "step": 8173 + }, + { + "epoch": 1.78, + "learning_rate": 4.112519117325759e-06, + "loss": 3.0671, + "step": 8174 + }, + { + "epoch": 1.78, + "learning_rate": 4.112409875464278e-06, + "loss": 1.7505, + "step": 8175 + }, + { + "epoch": 1.78, + "learning_rate": 4.112300633602797e-06, + "loss": 3.3224, + "step": 8176 + }, + { + "epoch": 1.78, + "learning_rate": 4.112191391741316e-06, + "loss": 2.2602, + "step": 8177 + }, + { + "epoch": 1.78, + "learning_rate": 4.112082149879834e-06, + "loss": 2.883, + "step": 8178 + }, + { + "epoch": 1.79, + "learning_rate": 4.111972908018353e-06, + "loss": 3.1759, + "step": 8179 + }, + { + "epoch": 1.79, + "learning_rate": 4.111863666156872e-06, + "loss": 3.0473, + "step": 8180 + }, + { + "epoch": 1.79, + "learning_rate": 4.111754424295391e-06, + "loss": 2.7093, + "step": 8181 + }, + { + "epoch": 1.79, + "learning_rate": 4.111645182433909e-06, + "loss": 2.9882, + "step": 8182 + }, + { + "epoch": 1.79, + "learning_rate": 4.111535940572427e-06, + "loss": 3.1011, + "step": 8183 + }, + { + "epoch": 1.79, + "learning_rate": 4.111426698710946e-06, + "loss": 2.4873, + "step": 8184 + }, + { + "epoch": 1.79, + "learning_rate": 4.111317456849465e-06, + "loss": 2.8465, + "step": 8185 + }, + { + "epoch": 1.79, + "learning_rate": 4.1112082149879835e-06, + "loss": 2.8387, + "step": 8186 + }, + { + "epoch": 1.79, + "learning_rate": 4.1110989731265026e-06, + "loss": 2.0706, + "step": 8187 + }, + { + "epoch": 1.79, + "learning_rate": 4.110989731265021e-06, + "loss": 1.8235, + "step": 8188 + }, + { + "epoch": 1.79, + "learning_rate": 4.11088048940354e-06, + "loss": 2.3444, + "step": 8189 + }, + { + "epoch": 1.79, + "learning_rate": 4.110771247542059e-06, + "loss": 2.2802, + "step": 8190 + }, + { + "epoch": 1.79, + "learning_rate": 4.110662005680577e-06, + "loss": 2.1921, + "step": 8191 + }, + { + "epoch": 1.79, + "learning_rate": 4.110552763819096e-06, + "loss": 2.4056, + "step": 8192 + }, + { + "epoch": 1.79, + "learning_rate": 4.110443521957614e-06, + "loss": 3.0727, + "step": 8193 + }, + { + "epoch": 1.79, + "learning_rate": 4.110334280096133e-06, + "loss": 3.444, + "step": 8194 + }, + { + "epoch": 1.79, + "learning_rate": 4.110225038234652e-06, + "loss": 2.3589, + "step": 8195 + }, + { + "epoch": 1.79, + "learning_rate": 4.1101157963731705e-06, + "loss": 3.0377, + "step": 8196 + }, + { + "epoch": 1.79, + "learning_rate": 4.110006554511689e-06, + "loss": 2.6418, + "step": 8197 + }, + { + "epoch": 1.79, + "learning_rate": 4.109897312650208e-06, + "loss": 3.1146, + "step": 8198 + }, + { + "epoch": 1.79, + "learning_rate": 4.109788070788727e-06, + "loss": 2.8763, + "step": 8199 + }, + { + "epoch": 1.79, + "learning_rate": 4.109678828927246e-06, + "loss": 2.6917, + "step": 8200 + }, + { + "epoch": 1.79, + "learning_rate": 4.109569587065764e-06, + "loss": 3.0076, + "step": 8201 + }, + { + "epoch": 1.79, + "learning_rate": 4.109460345204282e-06, + "loss": 2.9834, + "step": 8202 + }, + { + "epoch": 1.79, + "learning_rate": 4.109351103342801e-06, + "loss": 2.8951, + "step": 8203 + }, + { + "epoch": 1.79, + "learning_rate": 4.10924186148132e-06, + "loss": 3.5667, + "step": 8204 + }, + { + "epoch": 1.79, + "learning_rate": 4.109132619619839e-06, + "loss": 3.1264, + "step": 8205 + }, + { + "epoch": 1.79, + "learning_rate": 4.1090233777583574e-06, + "loss": 2.2717, + "step": 8206 + }, + { + "epoch": 1.79, + "learning_rate": 4.108914135896876e-06, + "loss": 3.2482, + "step": 8207 + }, + { + "epoch": 1.79, + "learning_rate": 4.108804894035395e-06, + "loss": 3.4802, + "step": 8208 + }, + { + "epoch": 1.79, + "learning_rate": 4.108695652173914e-06, + "loss": 2.2571, + "step": 8209 + }, + { + "epoch": 1.79, + "learning_rate": 4.108586410312433e-06, + "loss": 2.7315, + "step": 8210 + }, + { + "epoch": 1.79, + "learning_rate": 4.108477168450951e-06, + "loss": 3.3433, + "step": 8211 + }, + { + "epoch": 1.79, + "learning_rate": 4.108367926589469e-06, + "loss": 1.83, + "step": 8212 + }, + { + "epoch": 1.79, + "learning_rate": 4.108258684727988e-06, + "loss": 3.0822, + "step": 8213 + }, + { + "epoch": 1.79, + "learning_rate": 4.108149442866507e-06, + "loss": 2.1565, + "step": 8214 + }, + { + "epoch": 1.79, + "learning_rate": 4.108040201005025e-06, + "loss": 2.6944, + "step": 8215 + }, + { + "epoch": 1.79, + "learning_rate": 4.107930959143544e-06, + "loss": 3.5183, + "step": 8216 + }, + { + "epoch": 1.79, + "learning_rate": 4.107821717282063e-06, + "loss": 2.5335, + "step": 8217 + }, + { + "epoch": 1.79, + "learning_rate": 4.107712475420582e-06, + "loss": 2.7132, + "step": 8218 + }, + { + "epoch": 1.79, + "learning_rate": 4.107603233559101e-06, + "loss": 2.604, + "step": 8219 + }, + { + "epoch": 1.79, + "learning_rate": 4.107493991697619e-06, + "loss": 2.552, + "step": 8220 + }, + { + "epoch": 1.79, + "learning_rate": 4.107384749836137e-06, + "loss": 3.385, + "step": 8221 + }, + { + "epoch": 1.79, + "learning_rate": 4.107275507974656e-06, + "loss": 2.1309, + "step": 8222 + }, + { + "epoch": 1.79, + "learning_rate": 4.107166266113175e-06, + "loss": 2.3016, + "step": 8223 + }, + { + "epoch": 1.79, + "learning_rate": 4.107057024251694e-06, + "loss": 3.3451, + "step": 8224 + }, + { + "epoch": 1.8, + "learning_rate": 4.106947782390212e-06, + "loss": 2.2542, + "step": 8225 + }, + { + "epoch": 1.8, + "learning_rate": 4.1068385405287305e-06, + "loss": 3.2106, + "step": 8226 + }, + { + "epoch": 1.8, + "learning_rate": 4.1067292986672495e-06, + "loss": 3.4147, + "step": 8227 + }, + { + "epoch": 1.8, + "learning_rate": 4.106620056805769e-06, + "loss": 3.6851, + "step": 8228 + }, + { + "epoch": 1.8, + "learning_rate": 4.106510814944287e-06, + "loss": 3.2017, + "step": 8229 + }, + { + "epoch": 1.8, + "learning_rate": 4.106401573082806e-06, + "loss": 1.6085, + "step": 8230 + }, + { + "epoch": 1.8, + "learning_rate": 4.106292331221324e-06, + "loss": 1.5029, + "step": 8231 + }, + { + "epoch": 1.8, + "learning_rate": 4.106183089359843e-06, + "loss": 2.7565, + "step": 8232 + }, + { + "epoch": 1.8, + "learning_rate": 4.106073847498362e-06, + "loss": 1.9459, + "step": 8233 + }, + { + "epoch": 1.8, + "learning_rate": 4.10596460563688e-06, + "loss": 3.4891, + "step": 8234 + }, + { + "epoch": 1.8, + "learning_rate": 4.105855363775399e-06, + "loss": 2.059, + "step": 8235 + }, + { + "epoch": 1.8, + "learning_rate": 4.1057461219139175e-06, + "loss": 3.1543, + "step": 8236 + }, + { + "epoch": 1.8, + "learning_rate": 4.1056368800524365e-06, + "loss": 3.5114, + "step": 8237 + }, + { + "epoch": 1.8, + "learning_rate": 4.1055276381909555e-06, + "loss": 2.7883, + "step": 8238 + }, + { + "epoch": 1.8, + "learning_rate": 4.105418396329474e-06, + "loss": 2.8107, + "step": 8239 + }, + { + "epoch": 1.8, + "learning_rate": 4.105309154467993e-06, + "loss": 3.6074, + "step": 8240 + }, + { + "epoch": 1.8, + "learning_rate": 4.105199912606511e-06, + "loss": 2.4624, + "step": 8241 + }, + { + "epoch": 1.8, + "learning_rate": 4.10509067074503e-06, + "loss": 3.3152, + "step": 8242 + }, + { + "epoch": 1.8, + "learning_rate": 4.104981428883549e-06, + "loss": 3.0091, + "step": 8243 + }, + { + "epoch": 1.8, + "learning_rate": 4.104872187022067e-06, + "loss": 1.8996, + "step": 8244 + }, + { + "epoch": 1.8, + "learning_rate": 4.104762945160586e-06, + "loss": 2.416, + "step": 8245 + }, + { + "epoch": 1.8, + "learning_rate": 4.1046537032991044e-06, + "loss": 2.0892, + "step": 8246 + }, + { + "epoch": 1.8, + "learning_rate": 4.1045444614376235e-06, + "loss": 1.9258, + "step": 8247 + }, + { + "epoch": 1.8, + "learning_rate": 4.104435219576142e-06, + "loss": 3.4365, + "step": 8248 + }, + { + "epoch": 1.8, + "learning_rate": 4.104325977714661e-06, + "loss": 2.8636, + "step": 8249 + }, + { + "epoch": 1.8, + "learning_rate": 4.104216735853179e-06, + "loss": 3.3889, + "step": 8250 + }, + { + "epoch": 1.8, + "learning_rate": 4.104107493991698e-06, + "loss": 2.7571, + "step": 8251 + }, + { + "epoch": 1.8, + "learning_rate": 4.103998252130217e-06, + "loss": 3.0945, + "step": 8252 + }, + { + "epoch": 1.8, + "learning_rate": 4.103889010268735e-06, + "loss": 3.0083, + "step": 8253 + }, + { + "epoch": 1.8, + "learning_rate": 4.103779768407254e-06, + "loss": 3.3411, + "step": 8254 + }, + { + "epoch": 1.8, + "learning_rate": 4.103670526545772e-06, + "loss": 2.2686, + "step": 8255 + }, + { + "epoch": 1.8, + "learning_rate": 4.103561284684291e-06, + "loss": 2.6445, + "step": 8256 + }, + { + "epoch": 1.8, + "learning_rate": 4.10345204282281e-06, + "loss": 3.3346, + "step": 8257 + }, + { + "epoch": 1.8, + "learning_rate": 4.103342800961329e-06, + "loss": 3.1437, + "step": 8258 + }, + { + "epoch": 1.8, + "learning_rate": 4.103233559099848e-06, + "loss": 3.079, + "step": 8259 + }, + { + "epoch": 1.8, + "learning_rate": 4.103124317238366e-06, + "loss": 2.5106, + "step": 8260 + }, + { + "epoch": 1.8, + "learning_rate": 4.103015075376885e-06, + "loss": 2.7504, + "step": 8261 + }, + { + "epoch": 1.8, + "learning_rate": 4.102905833515403e-06, + "loss": 2.5989, + "step": 8262 + }, + { + "epoch": 1.8, + "learning_rate": 4.102796591653922e-06, + "loss": 2.1837, + "step": 8263 + }, + { + "epoch": 1.8, + "learning_rate": 4.102687349792441e-06, + "loss": 1.9802, + "step": 8264 + }, + { + "epoch": 1.8, + "learning_rate": 4.102578107930959e-06, + "loss": 3.5552, + "step": 8265 + }, + { + "epoch": 1.8, + "learning_rate": 4.102468866069478e-06, + "loss": 2.2835, + "step": 8266 + }, + { + "epoch": 1.8, + "learning_rate": 4.1023596242079965e-06, + "loss": 1.7065, + "step": 8267 + }, + { + "epoch": 1.8, + "learning_rate": 4.1022503823465156e-06, + "loss": 2.7778, + "step": 8268 + }, + { + "epoch": 1.8, + "learning_rate": 4.102141140485035e-06, + "loss": 2.2241, + "step": 8269 + }, + { + "epoch": 1.8, + "learning_rate": 4.102031898623553e-06, + "loss": 3.2351, + "step": 8270 + }, + { + "epoch": 1.81, + "learning_rate": 4.101922656762072e-06, + "loss": 2.505, + "step": 8271 + }, + { + "epoch": 1.81, + "learning_rate": 4.10181341490059e-06, + "loss": 1.4604, + "step": 8272 + }, + { + "epoch": 1.81, + "learning_rate": 4.101704173039109e-06, + "loss": 2.7855, + "step": 8273 + }, + { + "epoch": 1.81, + "learning_rate": 4.101594931177627e-06, + "loss": 3.4152, + "step": 8274 + }, + { + "epoch": 1.81, + "learning_rate": 4.101485689316146e-06, + "loss": 2.9834, + "step": 8275 + }, + { + "epoch": 1.81, + "learning_rate": 4.101376447454665e-06, + "loss": 2.9065, + "step": 8276 + }, + { + "epoch": 1.81, + "learning_rate": 4.1012672055931835e-06, + "loss": 3.6277, + "step": 8277 + }, + { + "epoch": 1.81, + "learning_rate": 4.1011579637317025e-06, + "loss": 2.5529, + "step": 8278 + }, + { + "epoch": 1.81, + "learning_rate": 4.101048721870221e-06, + "loss": 2.8005, + "step": 8279 + }, + { + "epoch": 1.81, + "learning_rate": 4.10093948000874e-06, + "loss": 2.2215, + "step": 8280 + }, + { + "epoch": 1.81, + "learning_rate": 4.100830238147258e-06, + "loss": 2.0945, + "step": 8281 + }, + { + "epoch": 1.81, + "learning_rate": 4.100720996285777e-06, + "loss": 2.6326, + "step": 8282 + }, + { + "epoch": 1.81, + "learning_rate": 4.100611754424296e-06, + "loss": 2.5112, + "step": 8283 + }, + { + "epoch": 1.81, + "learning_rate": 4.100502512562814e-06, + "loss": 3.3056, + "step": 8284 + }, + { + "epoch": 1.81, + "learning_rate": 4.100393270701333e-06, + "loss": 2.8225, + "step": 8285 + }, + { + "epoch": 1.81, + "learning_rate": 4.100284028839851e-06, + "loss": 2.2675, + "step": 8286 + }, + { + "epoch": 1.81, + "learning_rate": 4.1001747869783704e-06, + "loss": 2.2974, + "step": 8287 + }, + { + "epoch": 1.81, + "learning_rate": 4.1000655451168895e-06, + "loss": 3.7433, + "step": 8288 + }, + { + "epoch": 1.81, + "learning_rate": 4.099956303255408e-06, + "loss": 3.5832, + "step": 8289 + }, + { + "epoch": 1.81, + "learning_rate": 4.099847061393927e-06, + "loss": 3.1018, + "step": 8290 + }, + { + "epoch": 1.81, + "learning_rate": 4.099737819532445e-06, + "loss": 2.5918, + "step": 8291 + }, + { + "epoch": 1.81, + "learning_rate": 4.099628577670964e-06, + "loss": 2.0737, + "step": 8292 + }, + { + "epoch": 1.81, + "learning_rate": 4.099519335809483e-06, + "loss": 3.2498, + "step": 8293 + }, + { + "epoch": 1.81, + "learning_rate": 4.099410093948001e-06, + "loss": 2.7494, + "step": 8294 + }, + { + "epoch": 1.81, + "learning_rate": 4.099300852086519e-06, + "loss": 2.6281, + "step": 8295 + }, + { + "epoch": 1.81, + "learning_rate": 4.099191610225038e-06, + "loss": 3.4845, + "step": 8296 + }, + { + "epoch": 1.81, + "learning_rate": 4.099082368363557e-06, + "loss": 2.563, + "step": 8297 + }, + { + "epoch": 1.81, + "learning_rate": 4.0989731265020764e-06, + "loss": 2.7557, + "step": 8298 + }, + { + "epoch": 1.81, + "learning_rate": 4.098863884640595e-06, + "loss": 2.2231, + "step": 8299 + }, + { + "epoch": 1.81, + "learning_rate": 4.098754642779113e-06, + "loss": 2.8316, + "step": 8300 + }, + { + "epoch": 1.81, + "learning_rate": 4.098645400917632e-06, + "loss": 3.3302, + "step": 8301 + }, + { + "epoch": 1.81, + "learning_rate": 4.098536159056151e-06, + "loss": 3.1805, + "step": 8302 + }, + { + "epoch": 1.81, + "learning_rate": 4.098426917194669e-06, + "loss": 2.6079, + "step": 8303 + }, + { + "epoch": 1.81, + "learning_rate": 4.098317675333188e-06, + "loss": 3.0821, + "step": 8304 + }, + { + "epoch": 1.81, + "learning_rate": 4.098208433471706e-06, + "loss": 2.9596, + "step": 8305 + }, + { + "epoch": 1.81, + "learning_rate": 4.098099191610225e-06, + "loss": 2.3254, + "step": 8306 + }, + { + "epoch": 1.81, + "learning_rate": 4.097989949748744e-06, + "loss": 2.3612, + "step": 8307 + }, + { + "epoch": 1.81, + "learning_rate": 4.0978807078872625e-06, + "loss": 2.8719, + "step": 8308 + }, + { + "epoch": 1.81, + "learning_rate": 4.0977714660257816e-06, + "loss": 3.6148, + "step": 8309 + }, + { + "epoch": 1.81, + "learning_rate": 4.0976622241643e-06, + "loss": 2.2668, + "step": 8310 + }, + { + "epoch": 1.81, + "learning_rate": 4.097552982302819e-06, + "loss": 2.7748, + "step": 8311 + }, + { + "epoch": 1.81, + "learning_rate": 4.097443740441338e-06, + "loss": 3.2375, + "step": 8312 + }, + { + "epoch": 1.81, + "learning_rate": 4.097334498579856e-06, + "loss": 2.3363, + "step": 8313 + }, + { + "epoch": 1.81, + "learning_rate": 4.097225256718374e-06, + "loss": 3.4274, + "step": 8314 + }, + { + "epoch": 1.81, + "learning_rate": 4.097116014856893e-06, + "loss": 2.6114, + "step": 8315 + }, + { + "epoch": 1.81, + "learning_rate": 4.097006772995412e-06, + "loss": 2.6555, + "step": 8316 + }, + { + "epoch": 1.82, + "learning_rate": 4.096897531133931e-06, + "loss": 2.6415, + "step": 8317 + }, + { + "epoch": 1.82, + "learning_rate": 4.0967882892724495e-06, + "loss": 2.1407, + "step": 8318 + }, + { + "epoch": 1.82, + "learning_rate": 4.096679047410968e-06, + "loss": 2.6647, + "step": 8319 + }, + { + "epoch": 1.82, + "learning_rate": 4.096569805549487e-06, + "loss": 2.0892, + "step": 8320 + }, + { + "epoch": 1.82, + "learning_rate": 4.096460563688006e-06, + "loss": 2.8103, + "step": 8321 + }, + { + "epoch": 1.82, + "learning_rate": 4.096351321826525e-06, + "loss": 2.2439, + "step": 8322 + }, + { + "epoch": 1.82, + "learning_rate": 4.096242079965043e-06, + "loss": 2.6609, + "step": 8323 + }, + { + "epoch": 1.82, + "learning_rate": 4.096132838103561e-06, + "loss": 2.9114, + "step": 8324 + }, + { + "epoch": 1.82, + "learning_rate": 4.09602359624208e-06, + "loss": 2.5702, + "step": 8325 + }, + { + "epoch": 1.82, + "learning_rate": 4.095914354380599e-06, + "loss": 3.0044, + "step": 8326 + }, + { + "epoch": 1.82, + "learning_rate": 4.095805112519117e-06, + "loss": 2.8299, + "step": 8327 + }, + { + "epoch": 1.82, + "learning_rate": 4.0956958706576365e-06, + "loss": 2.759, + "step": 8328 + }, + { + "epoch": 1.82, + "learning_rate": 4.095586628796155e-06, + "loss": 2.5311, + "step": 8329 + }, + { + "epoch": 1.82, + "learning_rate": 4.095477386934674e-06, + "loss": 3.067, + "step": 8330 + }, + { + "epoch": 1.82, + "learning_rate": 4.095368145073193e-06, + "loss": 3.5205, + "step": 8331 + }, + { + "epoch": 1.82, + "learning_rate": 4.095258903211711e-06, + "loss": 3.6791, + "step": 8332 + }, + { + "epoch": 1.82, + "learning_rate": 4.09514966135023e-06, + "loss": 2.4362, + "step": 8333 + }, + { + "epoch": 1.82, + "learning_rate": 4.095040419488748e-06, + "loss": 2.8102, + "step": 8334 + }, + { + "epoch": 1.82, + "learning_rate": 4.094931177627267e-06, + "loss": 2.0665, + "step": 8335 + }, + { + "epoch": 1.82, + "learning_rate": 4.094821935765786e-06, + "loss": 2.8299, + "step": 8336 + }, + { + "epoch": 1.82, + "learning_rate": 4.094712693904304e-06, + "loss": 2.6084, + "step": 8337 + }, + { + "epoch": 1.82, + "learning_rate": 4.0946034520428226e-06, + "loss": 2.6621, + "step": 8338 + }, + { + "epoch": 1.82, + "learning_rate": 4.094494210181342e-06, + "loss": 3.3517, + "step": 8339 + }, + { + "epoch": 1.82, + "learning_rate": 4.094384968319861e-06, + "loss": 2.2311, + "step": 8340 + }, + { + "epoch": 1.82, + "learning_rate": 4.09427572645838e-06, + "loss": 3.1243, + "step": 8341 + }, + { + "epoch": 1.82, + "learning_rate": 4.094166484596898e-06, + "loss": 3.4561, + "step": 8342 + }, + { + "epoch": 1.82, + "learning_rate": 4.094057242735416e-06, + "loss": 3.1383, + "step": 8343 + }, + { + "epoch": 1.82, + "learning_rate": 4.093948000873935e-06, + "loss": 2.6963, + "step": 8344 + }, + { + "epoch": 1.82, + "learning_rate": 4.093838759012454e-06, + "loss": 2.7368, + "step": 8345 + }, + { + "epoch": 1.82, + "learning_rate": 4.093729517150973e-06, + "loss": 2.4613, + "step": 8346 + }, + { + "epoch": 1.82, + "learning_rate": 4.093620275289491e-06, + "loss": 1.7129, + "step": 8347 + }, + { + "epoch": 1.82, + "learning_rate": 4.0935110334280095e-06, + "loss": 3.6444, + "step": 8348 + }, + { + "epoch": 1.82, + "learning_rate": 4.0934017915665286e-06, + "loss": 2.6433, + "step": 8349 + }, + { + "epoch": 1.82, + "learning_rate": 4.093292549705048e-06, + "loss": 2.0737, + "step": 8350 + }, + { + "epoch": 1.82, + "learning_rate": 4.093183307843567e-06, + "loss": 1.6192, + "step": 8351 + }, + { + "epoch": 1.82, + "learning_rate": 4.093074065982085e-06, + "loss": 2.3704, + "step": 8352 + }, + { + "epoch": 1.82, + "learning_rate": 4.092964824120603e-06, + "loss": 2.3674, + "step": 8353 + }, + { + "epoch": 1.82, + "learning_rate": 4.092855582259122e-06, + "loss": 3.2041, + "step": 8354 + }, + { + "epoch": 1.82, + "learning_rate": 4.092746340397641e-06, + "loss": 2.9853, + "step": 8355 + }, + { + "epoch": 1.82, + "learning_rate": 4.092637098536159e-06, + "loss": 2.4082, + "step": 8356 + }, + { + "epoch": 1.82, + "learning_rate": 4.092527856674678e-06, + "loss": 2.9637, + "step": 8357 + }, + { + "epoch": 1.82, + "learning_rate": 4.0924186148131965e-06, + "loss": 2.2861, + "step": 8358 + }, + { + "epoch": 1.82, + "learning_rate": 4.0923093729517155e-06, + "loss": 2.052, + "step": 8359 + }, + { + "epoch": 1.82, + "learning_rate": 4.0922001310902345e-06, + "loss": 3.0565, + "step": 8360 + }, + { + "epoch": 1.82, + "learning_rate": 4.092090889228753e-06, + "loss": 2.588, + "step": 8361 + }, + { + "epoch": 1.82, + "learning_rate": 4.091981647367271e-06, + "loss": 2.1629, + "step": 8362 + }, + { + "epoch": 1.83, + "learning_rate": 4.09187240550579e-06, + "loss": 2.8108, + "step": 8363 + }, + { + "epoch": 1.83, + "learning_rate": 4.091763163644309e-06, + "loss": 2.1266, + "step": 8364 + }, + { + "epoch": 1.83, + "learning_rate": 4.091653921782828e-06, + "loss": 3.1411, + "step": 8365 + }, + { + "epoch": 1.83, + "learning_rate": 4.091544679921346e-06, + "loss": 2.1274, + "step": 8366 + }, + { + "epoch": 1.83, + "learning_rate": 4.091435438059864e-06, + "loss": 1.7856, + "step": 8367 + }, + { + "epoch": 1.83, + "learning_rate": 4.0913261961983834e-06, + "loss": 2.3484, + "step": 8368 + }, + { + "epoch": 1.83, + "learning_rate": 4.0912169543369025e-06, + "loss": 3.0351, + "step": 8369 + }, + { + "epoch": 1.83, + "learning_rate": 4.0911077124754215e-06, + "loss": 2.484, + "step": 8370 + }, + { + "epoch": 1.83, + "learning_rate": 4.09099847061394e-06, + "loss": 3.651, + "step": 8371 + }, + { + "epoch": 1.83, + "learning_rate": 4.090889228752458e-06, + "loss": 2.7782, + "step": 8372 + }, + { + "epoch": 1.83, + "learning_rate": 4.090779986890977e-06, + "loss": 2.4035, + "step": 8373 + }, + { + "epoch": 1.83, + "learning_rate": 4.090670745029496e-06, + "loss": 3.31, + "step": 8374 + }, + { + "epoch": 1.83, + "learning_rate": 4.090561503168015e-06, + "loss": 2.9913, + "step": 8375 + }, + { + "epoch": 1.83, + "learning_rate": 4.090452261306533e-06, + "loss": 3.6786, + "step": 8376 + }, + { + "epoch": 1.83, + "learning_rate": 4.090343019445051e-06, + "loss": 2.5868, + "step": 8377 + }, + { + "epoch": 1.83, + "learning_rate": 4.09023377758357e-06, + "loss": 1.966, + "step": 8378 + }, + { + "epoch": 1.83, + "learning_rate": 4.0901245357220894e-06, + "loss": 3.7992, + "step": 8379 + }, + { + "epoch": 1.83, + "learning_rate": 4.090015293860608e-06, + "loss": 2.2371, + "step": 8380 + }, + { + "epoch": 1.83, + "learning_rate": 4.089906051999127e-06, + "loss": 2.5823, + "step": 8381 + }, + { + "epoch": 1.83, + "learning_rate": 4.089796810137645e-06, + "loss": 1.8692, + "step": 8382 + }, + { + "epoch": 1.83, + "learning_rate": 4.089687568276164e-06, + "loss": 2.7571, + "step": 8383 + }, + { + "epoch": 1.83, + "learning_rate": 4.089578326414683e-06, + "loss": 2.9447, + "step": 8384 + }, + { + "epoch": 1.83, + "learning_rate": 4.089469084553201e-06, + "loss": 2.5319, + "step": 8385 + }, + { + "epoch": 1.83, + "learning_rate": 4.08935984269172e-06, + "loss": 3.8186, + "step": 8386 + }, + { + "epoch": 1.83, + "learning_rate": 4.089250600830238e-06, + "loss": 2.4675, + "step": 8387 + }, + { + "epoch": 1.83, + "learning_rate": 4.089141358968757e-06, + "loss": 3.4698, + "step": 8388 + }, + { + "epoch": 1.83, + "learning_rate": 4.089032117107276e-06, + "loss": 3.5138, + "step": 8389 + }, + { + "epoch": 1.83, + "learning_rate": 4.0889228752457946e-06, + "loss": 2.7043, + "step": 8390 + }, + { + "epoch": 1.83, + "learning_rate": 4.088813633384313e-06, + "loss": 3.2037, + "step": 8391 + }, + { + "epoch": 1.83, + "learning_rate": 4.088704391522832e-06, + "loss": 3.1984, + "step": 8392 + }, + { + "epoch": 1.83, + "learning_rate": 4.088595149661351e-06, + "loss": 2.5192, + "step": 8393 + }, + { + "epoch": 1.83, + "learning_rate": 4.088485907799869e-06, + "loss": 3.0554, + "step": 8394 + }, + { + "epoch": 1.83, + "learning_rate": 4.088376665938388e-06, + "loss": 2.8439, + "step": 8395 + }, + { + "epoch": 1.83, + "learning_rate": 4.088267424076906e-06, + "loss": 1.7791, + "step": 8396 + }, + { + "epoch": 1.83, + "learning_rate": 4.088158182215425e-06, + "loss": 2.7253, + "step": 8397 + }, + { + "epoch": 1.83, + "learning_rate": 4.088048940353944e-06, + "loss": 1.9212, + "step": 8398 + }, + { + "epoch": 1.83, + "learning_rate": 4.0879396984924625e-06, + "loss": 3.2118, + "step": 8399 + }, + { + "epoch": 1.83, + "learning_rate": 4.0878304566309815e-06, + "loss": 2.8603, + "step": 8400 + }, + { + "epoch": 1.83, + "learning_rate": 4.0877212147695e-06, + "loss": 2.6398, + "step": 8401 + }, + { + "epoch": 1.83, + "learning_rate": 4.087611972908019e-06, + "loss": 2.7002, + "step": 8402 + }, + { + "epoch": 1.83, + "learning_rate": 4.087502731046538e-06, + "loss": 3.121, + "step": 8403 + }, + { + "epoch": 1.83, + "learning_rate": 4.087393489185056e-06, + "loss": 2.3677, + "step": 8404 + }, + { + "epoch": 1.83, + "learning_rate": 4.087284247323575e-06, + "loss": 3.1918, + "step": 8405 + }, + { + "epoch": 1.83, + "learning_rate": 4.087175005462093e-06, + "loss": 2.1067, + "step": 8406 + }, + { + "epoch": 1.83, + "learning_rate": 4.087065763600612e-06, + "loss": 1.623, + "step": 8407 + }, + { + "epoch": 1.84, + "learning_rate": 4.086956521739131e-06, + "loss": 2.719, + "step": 8408 + }, + { + "epoch": 1.84, + "learning_rate": 4.0868472798776495e-06, + "loss": 3.6059, + "step": 8409 + }, + { + "epoch": 1.84, + "learning_rate": 4.0867380380161685e-06, + "loss": 2.6786, + "step": 8410 + }, + { + "epoch": 1.84, + "learning_rate": 4.086628796154687e-06, + "loss": 3.0325, + "step": 8411 + }, + { + "epoch": 1.84, + "learning_rate": 4.086519554293206e-06, + "loss": 3.0493, + "step": 8412 + }, + { + "epoch": 1.84, + "learning_rate": 4.086410312431724e-06, + "loss": 3.0394, + "step": 8413 + }, + { + "epoch": 1.84, + "learning_rate": 4.086301070570243e-06, + "loss": 2.618, + "step": 8414 + }, + { + "epoch": 1.84, + "learning_rate": 4.086191828708761e-06, + "loss": 4.0456, + "step": 8415 + }, + { + "epoch": 1.84, + "learning_rate": 4.08608258684728e-06, + "loss": 3.2457, + "step": 8416 + }, + { + "epoch": 1.84, + "learning_rate": 4.085973344985799e-06, + "loss": 3.1199, + "step": 8417 + }, + { + "epoch": 1.84, + "learning_rate": 4.085864103124317e-06, + "loss": 1.9965, + "step": 8418 + }, + { + "epoch": 1.84, + "learning_rate": 4.085754861262836e-06, + "loss": 2.6823, + "step": 8419 + }, + { + "epoch": 1.84, + "learning_rate": 4.085645619401355e-06, + "loss": 3.5474, + "step": 8420 + }, + { + "epoch": 1.84, + "learning_rate": 4.085536377539874e-06, + "loss": 3.2797, + "step": 8421 + }, + { + "epoch": 1.84, + "learning_rate": 4.085427135678393e-06, + "loss": 2.4927, + "step": 8422 + }, + { + "epoch": 1.84, + "learning_rate": 4.085317893816911e-06, + "loss": 2.8449, + "step": 8423 + }, + { + "epoch": 1.84, + "learning_rate": 4.08520865195543e-06, + "loss": 1.7408, + "step": 8424 + }, + { + "epoch": 1.84, + "learning_rate": 4.085099410093948e-06, + "loss": 1.8434, + "step": 8425 + }, + { + "epoch": 1.84, + "learning_rate": 4.084990168232467e-06, + "loss": 2.3424, + "step": 8426 + }, + { + "epoch": 1.84, + "learning_rate": 4.084880926370985e-06, + "loss": 2.7882, + "step": 8427 + }, + { + "epoch": 1.84, + "learning_rate": 4.084771684509504e-06, + "loss": 1.925, + "step": 8428 + }, + { + "epoch": 1.84, + "learning_rate": 4.084662442648023e-06, + "loss": 2.0345, + "step": 8429 + }, + { + "epoch": 1.84, + "learning_rate": 4.0845532007865416e-06, + "loss": 2.6469, + "step": 8430 + }, + { + "epoch": 1.84, + "learning_rate": 4.084443958925061e-06, + "loss": 2.5048, + "step": 8431 + }, + { + "epoch": 1.84, + "learning_rate": 4.084334717063579e-06, + "loss": 1.8857, + "step": 8432 + }, + { + "epoch": 1.84, + "learning_rate": 4.084225475202098e-06, + "loss": 3.2496, + "step": 8433 + }, + { + "epoch": 1.84, + "learning_rate": 4.084116233340617e-06, + "loss": 2.2325, + "step": 8434 + }, + { + "epoch": 1.84, + "learning_rate": 4.084006991479135e-06, + "loss": 2.7395, + "step": 8435 + }, + { + "epoch": 1.84, + "learning_rate": 4.083897749617654e-06, + "loss": 2.3851, + "step": 8436 + }, + { + "epoch": 1.84, + "learning_rate": 4.083788507756172e-06, + "loss": 2.4219, + "step": 8437 + }, + { + "epoch": 1.84, + "learning_rate": 4.083679265894691e-06, + "loss": 2.9246, + "step": 8438 + }, + { + "epoch": 1.84, + "learning_rate": 4.08357002403321e-06, + "loss": 3.4483, + "step": 8439 + }, + { + "epoch": 1.84, + "learning_rate": 4.0834607821717285e-06, + "loss": 3.6535, + "step": 8440 + }, + { + "epoch": 1.84, + "learning_rate": 4.0833515403102475e-06, + "loss": 3.1347, + "step": 8441 + }, + { + "epoch": 1.84, + "learning_rate": 4.083242298448766e-06, + "loss": 2.0626, + "step": 8442 + }, + { + "epoch": 1.84, + "learning_rate": 4.083133056587285e-06, + "loss": 2.4218, + "step": 8443 + }, + { + "epoch": 1.84, + "learning_rate": 4.083023814725803e-06, + "loss": 3.1604, + "step": 8444 + }, + { + "epoch": 1.84, + "learning_rate": 4.082914572864322e-06, + "loss": 3.0094, + "step": 8445 + }, + { + "epoch": 1.84, + "learning_rate": 4.08280533100284e-06, + "loss": 2.8635, + "step": 8446 + }, + { + "epoch": 1.84, + "learning_rate": 4.082696089141359e-06, + "loss": 2.5911, + "step": 8447 + }, + { + "epoch": 1.84, + "learning_rate": 4.082586847279878e-06, + "loss": 1.7921, + "step": 8448 + }, + { + "epoch": 1.84, + "learning_rate": 4.0824776054183964e-06, + "loss": 3.3493, + "step": 8449 + }, + { + "epoch": 1.84, + "learning_rate": 4.0823683635569155e-06, + "loss": 2.4571, + "step": 8450 + }, + { + "epoch": 1.84, + "learning_rate": 4.082259121695434e-06, + "loss": 2.3275, + "step": 8451 + }, + { + "epoch": 1.84, + "learning_rate": 4.082149879833953e-06, + "loss": 2.6337, + "step": 8452 + }, + { + "epoch": 1.84, + "learning_rate": 4.082040637972472e-06, + "loss": 3.1622, + "step": 8453 + }, + { + "epoch": 1.85, + "learning_rate": 4.08193139611099e-06, + "loss": 2.4947, + "step": 8454 + }, + { + "epoch": 1.85, + "learning_rate": 4.081822154249509e-06, + "loss": 2.6316, + "step": 8455 + }, + { + "epoch": 1.85, + "learning_rate": 4.081712912388027e-06, + "loss": 2.1133, + "step": 8456 + }, + { + "epoch": 1.85, + "learning_rate": 4.081603670526546e-06, + "loss": 3.0394, + "step": 8457 + }, + { + "epoch": 1.85, + "learning_rate": 4.081494428665065e-06, + "loss": 2.4435, + "step": 8458 + }, + { + "epoch": 1.85, + "learning_rate": 4.081385186803583e-06, + "loss": 2.7176, + "step": 8459 + }, + { + "epoch": 1.85, + "learning_rate": 4.0812759449421016e-06, + "loss": 1.9785, + "step": 8460 + }, + { + "epoch": 1.85, + "learning_rate": 4.081166703080621e-06, + "loss": 2.7813, + "step": 8461 + }, + { + "epoch": 1.85, + "learning_rate": 4.08105746121914e-06, + "loss": 3.3182, + "step": 8462 + }, + { + "epoch": 1.85, + "learning_rate": 4.080948219357659e-06, + "loss": 3.2669, + "step": 8463 + }, + { + "epoch": 1.85, + "learning_rate": 4.080838977496177e-06, + "loss": 2.8467, + "step": 8464 + }, + { + "epoch": 1.85, + "learning_rate": 4.080729735634695e-06, + "loss": 3.1751, + "step": 8465 + }, + { + "epoch": 1.85, + "learning_rate": 4.080620493773214e-06, + "loss": 2.4135, + "step": 8466 + }, + { + "epoch": 1.85, + "learning_rate": 4.080511251911733e-06, + "loss": 3.0699, + "step": 8467 + }, + { + "epoch": 1.85, + "learning_rate": 4.080402010050251e-06, + "loss": 3.0574, + "step": 8468 + }, + { + "epoch": 1.85, + "learning_rate": 4.08029276818877e-06, + "loss": 2.9505, + "step": 8469 + }, + { + "epoch": 1.85, + "learning_rate": 4.0801835263272885e-06, + "loss": 2.9597, + "step": 8470 + }, + { + "epoch": 1.85, + "learning_rate": 4.0800742844658076e-06, + "loss": 2.329, + "step": 8471 + }, + { + "epoch": 1.85, + "learning_rate": 4.079965042604327e-06, + "loss": 2.8191, + "step": 8472 + }, + { + "epoch": 1.85, + "learning_rate": 4.079855800742845e-06, + "loss": 1.9087, + "step": 8473 + }, + { + "epoch": 1.85, + "learning_rate": 4.079746558881363e-06, + "loss": 2.7383, + "step": 8474 + }, + { + "epoch": 1.85, + "learning_rate": 4.079637317019882e-06, + "loss": 2.8882, + "step": 8475 + }, + { + "epoch": 1.85, + "learning_rate": 4.079528075158401e-06, + "loss": 2.0614, + "step": 8476 + }, + { + "epoch": 1.85, + "learning_rate": 4.07941883329692e-06, + "loss": 3.202, + "step": 8477 + }, + { + "epoch": 1.85, + "learning_rate": 4.079309591435438e-06, + "loss": 3.5524, + "step": 8478 + }, + { + "epoch": 1.85, + "learning_rate": 4.0792003495739565e-06, + "loss": 3.5776, + "step": 8479 + }, + { + "epoch": 1.85, + "learning_rate": 4.0790911077124755e-06, + "loss": 2.8778, + "step": 8480 + }, + { + "epoch": 1.85, + "learning_rate": 4.0789818658509945e-06, + "loss": 2.8181, + "step": 8481 + }, + { + "epoch": 1.85, + "learning_rate": 4.0788726239895136e-06, + "loss": 2.4417, + "step": 8482 + }, + { + "epoch": 1.85, + "learning_rate": 4.078763382128032e-06, + "loss": 2.2738, + "step": 8483 + }, + { + "epoch": 1.85, + "learning_rate": 4.07865414026655e-06, + "loss": 2.4097, + "step": 8484 + }, + { + "epoch": 1.85, + "learning_rate": 4.078544898405069e-06, + "loss": 3.2353, + "step": 8485 + }, + { + "epoch": 1.85, + "learning_rate": 4.078435656543588e-06, + "loss": 3.1021, + "step": 8486 + }, + { + "epoch": 1.85, + "learning_rate": 4.078326414682107e-06, + "loss": 3.354, + "step": 8487 + }, + { + "epoch": 1.85, + "learning_rate": 4.078217172820625e-06, + "loss": 3.5131, + "step": 8488 + }, + { + "epoch": 1.85, + "learning_rate": 4.078107930959143e-06, + "loss": 3.248, + "step": 8489 + }, + { + "epoch": 1.85, + "learning_rate": 4.0779986890976624e-06, + "loss": 2.4952, + "step": 8490 + }, + { + "epoch": 1.85, + "learning_rate": 4.0778894472361815e-06, + "loss": 3.2486, + "step": 8491 + }, + { + "epoch": 1.85, + "learning_rate": 4.0777802053747005e-06, + "loss": 2.2826, + "step": 8492 + }, + { + "epoch": 1.85, + "learning_rate": 4.077670963513219e-06, + "loss": 1.906, + "step": 8493 + }, + { + "epoch": 1.85, + "learning_rate": 4.077561721651737e-06, + "loss": 3.2227, + "step": 8494 + }, + { + "epoch": 1.85, + "learning_rate": 4.077452479790256e-06, + "loss": 2.3946, + "step": 8495 + }, + { + "epoch": 1.85, + "learning_rate": 4.077343237928775e-06, + "loss": 2.2522, + "step": 8496 + }, + { + "epoch": 1.85, + "learning_rate": 4.077233996067293e-06, + "loss": 2.3203, + "step": 8497 + }, + { + "epoch": 1.85, + "learning_rate": 4.077124754205812e-06, + "loss": 1.9141, + "step": 8498 + }, + { + "epoch": 1.85, + "learning_rate": 4.07701551234433e-06, + "loss": 2.3063, + "step": 8499 + }, + { + "epoch": 1.86, + "learning_rate": 4.076906270482849e-06, + "loss": 3.0193, + "step": 8500 + }, + { + "epoch": 1.86, + "learning_rate": 4.0767970286213684e-06, + "loss": 2.3704, + "step": 8501 + }, + { + "epoch": 1.86, + "learning_rate": 4.076687786759887e-06, + "loss": 1.7032, + "step": 8502 + }, + { + "epoch": 1.86, + "learning_rate": 4.076578544898405e-06, + "loss": 2.2875, + "step": 8503 + }, + { + "epoch": 1.86, + "learning_rate": 4.076469303036924e-06, + "loss": 2.162, + "step": 8504 + }, + { + "epoch": 1.86, + "learning_rate": 4.076360061175443e-06, + "loss": 2.049, + "step": 8505 + }, + { + "epoch": 1.86, + "learning_rate": 4.076250819313962e-06, + "loss": 3.5082, + "step": 8506 + }, + { + "epoch": 1.86, + "learning_rate": 4.07614157745248e-06, + "loss": 2.9961, + "step": 8507 + }, + { + "epoch": 1.86, + "learning_rate": 4.076032335590998e-06, + "loss": 2.1165, + "step": 8508 + }, + { + "epoch": 1.86, + "learning_rate": 4.075923093729517e-06, + "loss": 2.8982, + "step": 8509 + }, + { + "epoch": 1.86, + "learning_rate": 4.075813851868036e-06, + "loss": 2.4493, + "step": 8510 + }, + { + "epoch": 1.86, + "learning_rate": 4.075704610006555e-06, + "loss": 3.5719, + "step": 8511 + }, + { + "epoch": 1.86, + "learning_rate": 4.075595368145074e-06, + "loss": 2.9695, + "step": 8512 + }, + { + "epoch": 1.86, + "learning_rate": 4.075486126283592e-06, + "loss": 1.9787, + "step": 8513 + }, + { + "epoch": 1.86, + "learning_rate": 4.075376884422111e-06, + "loss": 2.9635, + "step": 8514 + }, + { + "epoch": 1.86, + "learning_rate": 4.07526764256063e-06, + "loss": 2.6491, + "step": 8515 + }, + { + "epoch": 1.86, + "learning_rate": 4.075158400699149e-06, + "loss": 3.4026, + "step": 8516 + }, + { + "epoch": 1.86, + "learning_rate": 4.075049158837667e-06, + "loss": 2.2376, + "step": 8517 + }, + { + "epoch": 1.86, + "learning_rate": 4.074939916976185e-06, + "loss": 2.4119, + "step": 8518 + }, + { + "epoch": 1.86, + "learning_rate": 4.074830675114704e-06, + "loss": 2.1828, + "step": 8519 + }, + { + "epoch": 1.86, + "learning_rate": 4.074721433253223e-06, + "loss": 2.4375, + "step": 8520 + }, + { + "epoch": 1.86, + "learning_rate": 4.0746121913917415e-06, + "loss": 2.954, + "step": 8521 + }, + { + "epoch": 1.86, + "learning_rate": 4.0745029495302605e-06, + "loss": 2.3119, + "step": 8522 + }, + { + "epoch": 1.86, + "learning_rate": 4.074393707668779e-06, + "loss": 3.5484, + "step": 8523 + }, + { + "epoch": 1.86, + "learning_rate": 4.074284465807298e-06, + "loss": 2.4262, + "step": 8524 + }, + { + "epoch": 1.86, + "learning_rate": 4.074175223945817e-06, + "loss": 2.6895, + "step": 8525 + }, + { + "epoch": 1.86, + "learning_rate": 4.074065982084335e-06, + "loss": 2.9846, + "step": 8526 + }, + { + "epoch": 1.86, + "learning_rate": 4.073956740222853e-06, + "loss": 2.0261, + "step": 8527 + }, + { + "epoch": 1.86, + "learning_rate": 4.073847498361372e-06, + "loss": 2.7496, + "step": 8528 + }, + { + "epoch": 1.86, + "learning_rate": 4.073738256499891e-06, + "loss": 1.6103, + "step": 8529 + }, + { + "epoch": 1.86, + "learning_rate": 4.07362901463841e-06, + "loss": 2.8995, + "step": 8530 + }, + { + "epoch": 1.86, + "learning_rate": 4.0735197727769285e-06, + "loss": 2.0558, + "step": 8531 + }, + { + "epoch": 1.86, + "learning_rate": 4.073410530915447e-06, + "loss": 2.0871, + "step": 8532 + }, + { + "epoch": 1.86, + "learning_rate": 4.073301289053966e-06, + "loss": 2.7687, + "step": 8533 + }, + { + "epoch": 1.86, + "learning_rate": 4.073192047192485e-06, + "loss": 2.474, + "step": 8534 + }, + { + "epoch": 1.86, + "learning_rate": 4.073082805331004e-06, + "loss": 2.8373, + "step": 8535 + }, + { + "epoch": 1.86, + "learning_rate": 4.072973563469522e-06, + "loss": 2.8194, + "step": 8536 + }, + { + "epoch": 1.86, + "learning_rate": 4.07286432160804e-06, + "loss": 3.3778, + "step": 8537 + }, + { + "epoch": 1.86, + "learning_rate": 4.072755079746559e-06, + "loss": 2.0329, + "step": 8538 + }, + { + "epoch": 1.86, + "learning_rate": 4.072645837885078e-06, + "loss": 2.2938, + "step": 8539 + }, + { + "epoch": 1.86, + "learning_rate": 4.072536596023596e-06, + "loss": 2.3237, + "step": 8540 + }, + { + "epoch": 1.86, + "learning_rate": 4.072427354162115e-06, + "loss": 3.3169, + "step": 8541 + }, + { + "epoch": 1.86, + "learning_rate": 4.072318112300634e-06, + "loss": 1.4794, + "step": 8542 + }, + { + "epoch": 1.86, + "learning_rate": 4.072208870439153e-06, + "loss": 1.652, + "step": 8543 + }, + { + "epoch": 1.86, + "learning_rate": 4.072099628577672e-06, + "loss": 3.1396, + "step": 8544 + }, + { + "epoch": 1.86, + "learning_rate": 4.07199038671619e-06, + "loss": 2.6164, + "step": 8545 + }, + { + "epoch": 1.87, + "learning_rate": 4.071881144854709e-06, + "loss": 2.0686, + "step": 8546 + }, + { + "epoch": 1.87, + "learning_rate": 4.071771902993227e-06, + "loss": 1.7125, + "step": 8547 + }, + { + "epoch": 1.87, + "learning_rate": 4.071662661131746e-06, + "loss": 2.7714, + "step": 8548 + }, + { + "epoch": 1.87, + "learning_rate": 4.071553419270265e-06, + "loss": 3.1633, + "step": 8549 + }, + { + "epoch": 1.87, + "learning_rate": 4.071444177408783e-06, + "loss": 2.2393, + "step": 8550 + }, + { + "epoch": 1.87, + "learning_rate": 4.071334935547302e-06, + "loss": 2.5236, + "step": 8551 + }, + { + "epoch": 1.87, + "learning_rate": 4.0712256936858206e-06, + "loss": 2.4687, + "step": 8552 + }, + { + "epoch": 1.87, + "learning_rate": 4.07111645182434e-06, + "loss": 2.7935, + "step": 8553 + }, + { + "epoch": 1.87, + "learning_rate": 4.071007209962859e-06, + "loss": 2.64, + "step": 8554 + }, + { + "epoch": 1.87, + "learning_rate": 4.070897968101377e-06, + "loss": 2.6157, + "step": 8555 + }, + { + "epoch": 1.87, + "learning_rate": 4.070788726239895e-06, + "loss": 2.2865, + "step": 8556 + }, + { + "epoch": 1.87, + "learning_rate": 4.070679484378414e-06, + "loss": 2.8511, + "step": 8557 + }, + { + "epoch": 1.87, + "learning_rate": 4.070570242516933e-06, + "loss": 2.3905, + "step": 8558 + }, + { + "epoch": 1.87, + "learning_rate": 4.070461000655451e-06, + "loss": 2.6503, + "step": 8559 + }, + { + "epoch": 1.87, + "learning_rate": 4.07035175879397e-06, + "loss": 3.1191, + "step": 8560 + }, + { + "epoch": 1.87, + "learning_rate": 4.0702425169324885e-06, + "loss": 2.5325, + "step": 8561 + }, + { + "epoch": 1.87, + "learning_rate": 4.0701332750710075e-06, + "loss": 2.4609, + "step": 8562 + }, + { + "epoch": 1.87, + "learning_rate": 4.0700240332095266e-06, + "loss": 3.3587, + "step": 8563 + }, + { + "epoch": 1.87, + "learning_rate": 4.069914791348045e-06, + "loss": 3.6162, + "step": 8564 + }, + { + "epoch": 1.87, + "learning_rate": 4.069805549486564e-06, + "loss": 3.2259, + "step": 8565 + }, + { + "epoch": 1.87, + "learning_rate": 4.069696307625082e-06, + "loss": 2.138, + "step": 8566 + }, + { + "epoch": 1.87, + "learning_rate": 4.069587065763601e-06, + "loss": 3.0423, + "step": 8567 + }, + { + "epoch": 1.87, + "learning_rate": 4.06947782390212e-06, + "loss": 3.0124, + "step": 8568 + }, + { + "epoch": 1.87, + "learning_rate": 4.069368582040638e-06, + "loss": 3.4315, + "step": 8569 + }, + { + "epoch": 1.87, + "learning_rate": 4.069259340179157e-06, + "loss": 2.8635, + "step": 8570 + }, + { + "epoch": 1.87, + "learning_rate": 4.0691500983176754e-06, + "loss": 3.5715, + "step": 8571 + }, + { + "epoch": 1.87, + "learning_rate": 4.0690408564561945e-06, + "loss": 2.743, + "step": 8572 + }, + { + "epoch": 1.87, + "learning_rate": 4.068931614594713e-06, + "loss": 2.0099, + "step": 8573 + }, + { + "epoch": 1.87, + "learning_rate": 4.068822372733232e-06, + "loss": 3.1074, + "step": 8574 + }, + { + "epoch": 1.87, + "learning_rate": 4.068713130871751e-06, + "loss": 2.7001, + "step": 8575 + }, + { + "epoch": 1.87, + "learning_rate": 4.068603889010269e-06, + "loss": 2.7004, + "step": 8576 + }, + { + "epoch": 1.87, + "learning_rate": 4.068494647148788e-06, + "loss": 2.3546, + "step": 8577 + }, + { + "epoch": 1.87, + "learning_rate": 4.068385405287306e-06, + "loss": 3.2461, + "step": 8578 + }, + { + "epoch": 1.87, + "learning_rate": 4.068276163425825e-06, + "loss": 3.2493, + "step": 8579 + }, + { + "epoch": 1.87, + "learning_rate": 4.068166921564343e-06, + "loss": 3.2008, + "step": 8580 + }, + { + "epoch": 1.87, + "learning_rate": 4.068057679702862e-06, + "loss": 2.5421, + "step": 8581 + }, + { + "epoch": 1.87, + "learning_rate": 4.0679484378413814e-06, + "loss": 3.0091, + "step": 8582 + }, + { + "epoch": 1.87, + "learning_rate": 4.0678391959799e-06, + "loss": 2.7224, + "step": 8583 + }, + { + "epoch": 1.87, + "learning_rate": 4.067729954118419e-06, + "loss": 2.5875, + "step": 8584 + }, + { + "epoch": 1.87, + "learning_rate": 4.067620712256937e-06, + "loss": 2.3947, + "step": 8585 + }, + { + "epoch": 1.87, + "learning_rate": 4.067511470395456e-06, + "loss": 2.5919, + "step": 8586 + }, + { + "epoch": 1.87, + "learning_rate": 4.067402228533975e-06, + "loss": 2.5926, + "step": 8587 + }, + { + "epoch": 1.87, + "learning_rate": 4.067292986672493e-06, + "loss": 2.4889, + "step": 8588 + }, + { + "epoch": 1.87, + "learning_rate": 4.067183744811012e-06, + "loss": 2.8707, + "step": 8589 + }, + { + "epoch": 1.87, + "learning_rate": 4.06707450294953e-06, + "loss": 2.8114, + "step": 8590 + }, + { + "epoch": 1.87, + "learning_rate": 4.066965261088049e-06, + "loss": 2.4311, + "step": 8591 + }, + { + "epoch": 1.88, + "learning_rate": 4.0668560192265675e-06, + "loss": 2.8428, + "step": 8592 + }, + { + "epoch": 1.88, + "learning_rate": 4.066746777365087e-06, + "loss": 2.2453, + "step": 8593 + }, + { + "epoch": 1.88, + "learning_rate": 4.066637535503606e-06, + "loss": 1.6956, + "step": 8594 + }, + { + "epoch": 1.88, + "learning_rate": 4.066528293642124e-06, + "loss": 2.4945, + "step": 8595 + }, + { + "epoch": 1.88, + "learning_rate": 4.066419051780643e-06, + "loss": 2.9714, + "step": 8596 + }, + { + "epoch": 1.88, + "learning_rate": 4.066309809919161e-06, + "loss": 2.1552, + "step": 8597 + }, + { + "epoch": 1.88, + "learning_rate": 4.06620056805768e-06, + "loss": 3.2703, + "step": 8598 + }, + { + "epoch": 1.88, + "learning_rate": 4.066091326196199e-06, + "loss": 2.6929, + "step": 8599 + }, + { + "epoch": 1.88, + "learning_rate": 4.065982084334717e-06, + "loss": 3.0509, + "step": 8600 + }, + { + "epoch": 1.88, + "learning_rate": 4.065872842473236e-06, + "loss": 2.6784, + "step": 8601 + }, + { + "epoch": 1.88, + "learning_rate": 4.0657636006117545e-06, + "loss": 2.1213, + "step": 8602 + }, + { + "epoch": 1.88, + "learning_rate": 4.0656543587502735e-06, + "loss": 3.1131, + "step": 8603 + }, + { + "epoch": 1.88, + "learning_rate": 4.0655451168887926e-06, + "loss": 4.1427, + "step": 8604 + }, + { + "epoch": 1.88, + "learning_rate": 4.065435875027311e-06, + "loss": 2.2772, + "step": 8605 + }, + { + "epoch": 1.88, + "learning_rate": 4.065326633165829e-06, + "loss": 3.0602, + "step": 8606 + }, + { + "epoch": 1.88, + "learning_rate": 4.065217391304348e-06, + "loss": 2.9535, + "step": 8607 + }, + { + "epoch": 1.88, + "learning_rate": 4.065108149442867e-06, + "loss": 1.8222, + "step": 8608 + }, + { + "epoch": 1.88, + "learning_rate": 4.064998907581385e-06, + "loss": 3.5161, + "step": 8609 + }, + { + "epoch": 1.88, + "learning_rate": 4.064889665719904e-06, + "loss": 2.5261, + "step": 8610 + }, + { + "epoch": 1.88, + "learning_rate": 4.064780423858422e-06, + "loss": 2.7655, + "step": 8611 + }, + { + "epoch": 1.88, + "learning_rate": 4.0646711819969415e-06, + "loss": 3.2883, + "step": 8612 + }, + { + "epoch": 1.88, + "learning_rate": 4.0645619401354605e-06, + "loss": 2.5331, + "step": 8613 + }, + { + "epoch": 1.88, + "learning_rate": 4.064452698273979e-06, + "loss": 2.7827, + "step": 8614 + }, + { + "epoch": 1.88, + "learning_rate": 4.064343456412498e-06, + "loss": 3.7013, + "step": 8615 + }, + { + "epoch": 1.88, + "learning_rate": 4.064234214551016e-06, + "loss": 2.0954, + "step": 8616 + }, + { + "epoch": 1.88, + "learning_rate": 4.064124972689535e-06, + "loss": 2.7408, + "step": 8617 + }, + { + "epoch": 1.88, + "learning_rate": 4.064015730828054e-06, + "loss": 2.8083, + "step": 8618 + }, + { + "epoch": 1.88, + "learning_rate": 4.063906488966572e-06, + "loss": 3.9197, + "step": 8619 + }, + { + "epoch": 1.88, + "learning_rate": 4.063797247105091e-06, + "loss": 2.7873, + "step": 8620 + }, + { + "epoch": 1.88, + "learning_rate": 4.063688005243609e-06, + "loss": 2.2007, + "step": 8621 + }, + { + "epoch": 1.88, + "learning_rate": 4.063578763382128e-06, + "loss": 3.0988, + "step": 8622 + }, + { + "epoch": 1.88, + "learning_rate": 4.0634695215206474e-06, + "loss": 3.0287, + "step": 8623 + }, + { + "epoch": 1.88, + "learning_rate": 4.063360279659166e-06, + "loss": 2.8369, + "step": 8624 + }, + { + "epoch": 1.88, + "learning_rate": 4.063251037797684e-06, + "loss": 2.8687, + "step": 8625 + }, + { + "epoch": 1.88, + "learning_rate": 4.063141795936203e-06, + "loss": 2.7986, + "step": 8626 + }, + { + "epoch": 1.88, + "learning_rate": 4.063032554074722e-06, + "loss": 2.4863, + "step": 8627 + }, + { + "epoch": 1.88, + "learning_rate": 4.062923312213241e-06, + "loss": 2.7975, + "step": 8628 + }, + { + "epoch": 1.88, + "learning_rate": 4.062814070351759e-06, + "loss": 1.497, + "step": 8629 + }, + { + "epoch": 1.88, + "learning_rate": 4.062704828490277e-06, + "loss": 2.7316, + "step": 8630 + }, + { + "epoch": 1.88, + "learning_rate": 4.062595586628796e-06, + "loss": 2.7428, + "step": 8631 + }, + { + "epoch": 1.88, + "learning_rate": 4.062486344767315e-06, + "loss": 2.8904, + "step": 8632 + }, + { + "epoch": 1.88, + "learning_rate": 4.0623771029058336e-06, + "loss": 2.2553, + "step": 8633 + }, + { + "epoch": 1.88, + "learning_rate": 4.062267861044353e-06, + "loss": 2.8305, + "step": 8634 + }, + { + "epoch": 1.88, + "learning_rate": 4.062158619182871e-06, + "loss": 2.5766, + "step": 8635 + }, + { + "epoch": 1.88, + "learning_rate": 4.06204937732139e-06, + "loss": 3.142, + "step": 8636 + }, + { + "epoch": 1.88, + "learning_rate": 4.061940135459909e-06, + "loss": 2.4812, + "step": 8637 + }, + { + "epoch": 1.89, + "learning_rate": 4.061830893598427e-06, + "loss": 3.2762, + "step": 8638 + }, + { + "epoch": 1.89, + "learning_rate": 4.061721651736946e-06, + "loss": 3.7283, + "step": 8639 + }, + { + "epoch": 1.89, + "learning_rate": 4.061612409875464e-06, + "loss": 2.54, + "step": 8640 + }, + { + "epoch": 1.89, + "learning_rate": 4.061503168013983e-06, + "loss": 1.8067, + "step": 8641 + }, + { + "epoch": 1.89, + "learning_rate": 4.061393926152502e-06, + "loss": 2.5714, + "step": 8642 + }, + { + "epoch": 1.89, + "learning_rate": 4.0612846842910205e-06, + "loss": 2.4721, + "step": 8643 + }, + { + "epoch": 1.89, + "learning_rate": 4.061175442429539e-06, + "loss": 2.1737, + "step": 8644 + }, + { + "epoch": 1.89, + "learning_rate": 4.061066200568058e-06, + "loss": 3.1384, + "step": 8645 + }, + { + "epoch": 1.89, + "learning_rate": 4.060956958706577e-06, + "loss": 3.1572, + "step": 8646 + }, + { + "epoch": 1.89, + "learning_rate": 4.060847716845096e-06, + "loss": 3.0077, + "step": 8647 + }, + { + "epoch": 1.89, + "learning_rate": 4.060738474983614e-06, + "loss": 2.1035, + "step": 8648 + }, + { + "epoch": 1.89, + "learning_rate": 4.060629233122132e-06, + "loss": 2.8207, + "step": 8649 + }, + { + "epoch": 1.89, + "learning_rate": 4.060519991260651e-06, + "loss": 3.2215, + "step": 8650 + }, + { + "epoch": 1.89, + "learning_rate": 4.06041074939917e-06, + "loss": 3.1901, + "step": 8651 + }, + { + "epoch": 1.89, + "learning_rate": 4.060301507537689e-06, + "loss": 2.3976, + "step": 8652 + }, + { + "epoch": 1.89, + "learning_rate": 4.0601922656762075e-06, + "loss": 2.4205, + "step": 8653 + }, + { + "epoch": 1.89, + "learning_rate": 4.060083023814726e-06, + "loss": 2.7092, + "step": 8654 + }, + { + "epoch": 1.89, + "learning_rate": 4.059973781953245e-06, + "loss": 2.5085, + "step": 8655 + }, + { + "epoch": 1.89, + "learning_rate": 4.059864540091764e-06, + "loss": 3.1121, + "step": 8656 + }, + { + "epoch": 1.89, + "learning_rate": 4.059755298230283e-06, + "loss": 1.9718, + "step": 8657 + }, + { + "epoch": 1.89, + "learning_rate": 4.059646056368801e-06, + "loss": 3.1718, + "step": 8658 + }, + { + "epoch": 1.89, + "learning_rate": 4.059536814507319e-06, + "loss": 2.4047, + "step": 8659 + }, + { + "epoch": 1.89, + "learning_rate": 4.059427572645838e-06, + "loss": 2.4581, + "step": 8660 + }, + { + "epoch": 1.89, + "learning_rate": 4.059318330784357e-06, + "loss": 2.419, + "step": 8661 + }, + { + "epoch": 1.89, + "learning_rate": 4.059209088922875e-06, + "loss": 3.0225, + "step": 8662 + }, + { + "epoch": 1.89, + "learning_rate": 4.0590998470613944e-06, + "loss": 2.3004, + "step": 8663 + }, + { + "epoch": 1.89, + "learning_rate": 4.058990605199913e-06, + "loss": 1.6904, + "step": 8664 + }, + { + "epoch": 1.89, + "learning_rate": 4.058881363338432e-06, + "loss": 2.6036, + "step": 8665 + }, + { + "epoch": 1.89, + "learning_rate": 4.058772121476951e-06, + "loss": 1.7606, + "step": 8666 + }, + { + "epoch": 1.89, + "learning_rate": 4.058662879615469e-06, + "loss": 3.0514, + "step": 8667 + }, + { + "epoch": 1.89, + "learning_rate": 4.058553637753987e-06, + "loss": 2.3581, + "step": 8668 + }, + { + "epoch": 1.89, + "learning_rate": 4.058444395892506e-06, + "loss": 2.0402, + "step": 8669 + }, + { + "epoch": 1.89, + "learning_rate": 4.058335154031025e-06, + "loss": 3.4092, + "step": 8670 + }, + { + "epoch": 1.89, + "learning_rate": 4.058225912169544e-06, + "loss": 3.0251, + "step": 8671 + }, + { + "epoch": 1.89, + "learning_rate": 4.058116670308062e-06, + "loss": 3.355, + "step": 8672 + }, + { + "epoch": 1.89, + "learning_rate": 4.0580074284465805e-06, + "loss": 2.9487, + "step": 8673 + }, + { + "epoch": 1.89, + "learning_rate": 4.0578981865850996e-06, + "loss": 2.9076, + "step": 8674 + }, + { + "epoch": 1.89, + "learning_rate": 4.057788944723619e-06, + "loss": 2.9577, + "step": 8675 + }, + { + "epoch": 1.89, + "learning_rate": 4.057679702862138e-06, + "loss": 2.4609, + "step": 8676 + }, + { + "epoch": 1.89, + "learning_rate": 4.057570461000656e-06, + "loss": 2.6435, + "step": 8677 + }, + { + "epoch": 1.89, + "learning_rate": 4.057461219139174e-06, + "loss": 1.2267, + "step": 8678 + }, + { + "epoch": 1.89, + "learning_rate": 4.057351977277693e-06, + "loss": 3.4115, + "step": 8679 + }, + { + "epoch": 1.89, + "learning_rate": 4.057242735416212e-06, + "loss": 2.9034, + "step": 8680 + }, + { + "epoch": 1.89, + "learning_rate": 4.057133493554731e-06, + "loss": 3.4144, + "step": 8681 + }, + { + "epoch": 1.89, + "learning_rate": 4.057024251693249e-06, + "loss": 1.9112, + "step": 8682 + }, + { + "epoch": 1.9, + "learning_rate": 4.0569150098317675e-06, + "loss": 2.9724, + "step": 8683 + }, + { + "epoch": 1.9, + "learning_rate": 4.0568057679702865e-06, + "loss": 2.4544, + "step": 8684 + }, + { + "epoch": 1.9, + "learning_rate": 4.0566965261088056e-06, + "loss": 2.3209, + "step": 8685 + }, + { + "epoch": 1.9, + "learning_rate": 4.056587284247324e-06, + "loss": 3.373, + "step": 8686 + }, + { + "epoch": 1.9, + "learning_rate": 4.056478042385843e-06, + "loss": 2.9531, + "step": 8687 + }, + { + "epoch": 1.9, + "learning_rate": 4.056368800524361e-06, + "loss": 1.7468, + "step": 8688 + }, + { + "epoch": 1.9, + "learning_rate": 4.05625955866288e-06, + "loss": 2.9131, + "step": 8689 + }, + { + "epoch": 1.9, + "learning_rate": 4.056150316801399e-06, + "loss": 2.4737, + "step": 8690 + }, + { + "epoch": 1.9, + "learning_rate": 4.056041074939917e-06, + "loss": 2.5163, + "step": 8691 + }, + { + "epoch": 1.9, + "learning_rate": 4.055931833078436e-06, + "loss": 2.6831, + "step": 8692 + }, + { + "epoch": 1.9, + "learning_rate": 4.0558225912169545e-06, + "loss": 1.8664, + "step": 8693 + }, + { + "epoch": 1.9, + "learning_rate": 4.0557133493554735e-06, + "loss": 3.1806, + "step": 8694 + }, + { + "epoch": 1.9, + "learning_rate": 4.0556041074939925e-06, + "loss": 2.5978, + "step": 8695 + }, + { + "epoch": 1.9, + "learning_rate": 4.055494865632511e-06, + "loss": 1.9854, + "step": 8696 + }, + { + "epoch": 1.9, + "learning_rate": 4.055385623771029e-06, + "loss": 2.6762, + "step": 8697 + }, + { + "epoch": 1.9, + "learning_rate": 4.055276381909548e-06, + "loss": 2.0947, + "step": 8698 + }, + { + "epoch": 1.9, + "learning_rate": 4.055167140048067e-06, + "loss": 2.6565, + "step": 8699 + }, + { + "epoch": 1.9, + "learning_rate": 4.055057898186586e-06, + "loss": 3.331, + "step": 8700 + }, + { + "epoch": 1.9, + "learning_rate": 4.054948656325104e-06, + "loss": 3.9359, + "step": 8701 + }, + { + "epoch": 1.9, + "learning_rate": 4.054839414463622e-06, + "loss": 2.6058, + "step": 8702 + }, + { + "epoch": 1.9, + "learning_rate": 4.054730172602141e-06, + "loss": 1.4572, + "step": 8703 + }, + { + "epoch": 1.9, + "learning_rate": 4.0546209307406604e-06, + "loss": 2.8304, + "step": 8704 + }, + { + "epoch": 1.9, + "learning_rate": 4.054511688879179e-06, + "loss": 1.4505, + "step": 8705 + }, + { + "epoch": 1.9, + "learning_rate": 4.054402447017698e-06, + "loss": 3.1917, + "step": 8706 + }, + { + "epoch": 1.9, + "learning_rate": 4.054293205156216e-06, + "loss": 2.3268, + "step": 8707 + }, + { + "epoch": 1.9, + "learning_rate": 4.054183963294735e-06, + "loss": 3.0904, + "step": 8708 + }, + { + "epoch": 1.9, + "learning_rate": 4.054074721433254e-06, + "loss": 3.3053, + "step": 8709 + }, + { + "epoch": 1.9, + "learning_rate": 4.053965479571772e-06, + "loss": 2.51, + "step": 8710 + }, + { + "epoch": 1.9, + "learning_rate": 4.053856237710291e-06, + "loss": 1.6048, + "step": 8711 + }, + { + "epoch": 1.9, + "learning_rate": 4.053746995848809e-06, + "loss": 2.702, + "step": 8712 + }, + { + "epoch": 1.9, + "learning_rate": 4.053637753987328e-06, + "loss": 2.8635, + "step": 8713 + }, + { + "epoch": 1.9, + "learning_rate": 4.053528512125847e-06, + "loss": 2.4416, + "step": 8714 + }, + { + "epoch": 1.9, + "learning_rate": 4.053419270264366e-06, + "loss": 2.0693, + "step": 8715 + }, + { + "epoch": 1.9, + "learning_rate": 4.053310028402885e-06, + "loss": 2.9159, + "step": 8716 + }, + { + "epoch": 1.9, + "learning_rate": 4.053200786541403e-06, + "loss": 1.6199, + "step": 8717 + }, + { + "epoch": 1.9, + "learning_rate": 4.053091544679922e-06, + "loss": 1.7747, + "step": 8718 + }, + { + "epoch": 1.9, + "learning_rate": 4.052982302818441e-06, + "loss": 2.555, + "step": 8719 + }, + { + "epoch": 1.9, + "learning_rate": 4.052873060956959e-06, + "loss": 2.0793, + "step": 8720 + }, + { + "epoch": 1.9, + "learning_rate": 4.052763819095477e-06, + "loss": 3.7134, + "step": 8721 + }, + { + "epoch": 1.9, + "learning_rate": 4.052654577233996e-06, + "loss": 2.9385, + "step": 8722 + }, + { + "epoch": 1.9, + "learning_rate": 4.052545335372515e-06, + "loss": 2.917, + "step": 8723 + }, + { + "epoch": 1.9, + "learning_rate": 4.0524360935110335e-06, + "loss": 3.1581, + "step": 8724 + }, + { + "epoch": 1.9, + "learning_rate": 4.0523268516495525e-06, + "loss": 2.9667, + "step": 8725 + }, + { + "epoch": 1.9, + "learning_rate": 4.052217609788071e-06, + "loss": 2.8445, + "step": 8726 + }, + { + "epoch": 1.9, + "learning_rate": 4.05210836792659e-06, + "loss": 2.0174, + "step": 8727 + }, + { + "epoch": 1.9, + "learning_rate": 4.051999126065109e-06, + "loss": 1.5394, + "step": 8728 + }, + { + "epoch": 1.91, + "learning_rate": 4.051889884203627e-06, + "loss": 1.3333, + "step": 8729 + }, + { + "epoch": 1.91, + "learning_rate": 4.051780642342146e-06, + "loss": 2.3041, + "step": 8730 + }, + { + "epoch": 1.91, + "learning_rate": 4.051671400480664e-06, + "loss": 3.0324, + "step": 8731 + }, + { + "epoch": 1.91, + "learning_rate": 4.051562158619183e-06, + "loss": 2.6919, + "step": 8732 + }, + { + "epoch": 1.91, + "learning_rate": 4.051452916757702e-06, + "loss": 3.3172, + "step": 8733 + }, + { + "epoch": 1.91, + "learning_rate": 4.0513436748962205e-06, + "loss": 2.216, + "step": 8734 + }, + { + "epoch": 1.91, + "learning_rate": 4.0512344330347395e-06, + "loss": 3.1062, + "step": 8735 + }, + { + "epoch": 1.91, + "learning_rate": 4.051125191173258e-06, + "loss": 3.2432, + "step": 8736 + }, + { + "epoch": 1.91, + "learning_rate": 4.051015949311777e-06, + "loss": 3.297, + "step": 8737 + }, + { + "epoch": 1.91, + "learning_rate": 4.050906707450295e-06, + "loss": 2.2771, + "step": 8738 + }, + { + "epoch": 1.91, + "learning_rate": 4.050797465588814e-06, + "loss": 1.554, + "step": 8739 + }, + { + "epoch": 1.91, + "learning_rate": 4.050688223727333e-06, + "loss": 2.8748, + "step": 8740 + }, + { + "epoch": 1.91, + "learning_rate": 4.050578981865851e-06, + "loss": 2.1651, + "step": 8741 + }, + { + "epoch": 1.91, + "learning_rate": 4.05046974000437e-06, + "loss": 2.277, + "step": 8742 + }, + { + "epoch": 1.91, + "learning_rate": 4.050360498142888e-06, + "loss": 1.5227, + "step": 8743 + }, + { + "epoch": 1.91, + "learning_rate": 4.0502512562814074e-06, + "loss": 1.8136, + "step": 8744 + }, + { + "epoch": 1.91, + "learning_rate": 4.0501420144199265e-06, + "loss": 2.7854, + "step": 8745 + }, + { + "epoch": 1.91, + "learning_rate": 4.050032772558445e-06, + "loss": 2.2449, + "step": 8746 + }, + { + "epoch": 1.91, + "learning_rate": 4.049923530696964e-06, + "loss": 2.5277, + "step": 8747 + }, + { + "epoch": 1.91, + "learning_rate": 4.049814288835482e-06, + "loss": 3.2564, + "step": 8748 + }, + { + "epoch": 1.91, + "learning_rate": 4.049705046974001e-06, + "loss": 2.9244, + "step": 8749 + }, + { + "epoch": 1.91, + "learning_rate": 4.049595805112519e-06, + "loss": 3.4743, + "step": 8750 + }, + { + "epoch": 1.91, + "learning_rate": 4.049486563251038e-06, + "loss": 3.3083, + "step": 8751 + }, + { + "epoch": 1.91, + "learning_rate": 4.049377321389557e-06, + "loss": 1.9505, + "step": 8752 + }, + { + "epoch": 1.91, + "learning_rate": 4.049268079528075e-06, + "loss": 2.2991, + "step": 8753 + }, + { + "epoch": 1.91, + "learning_rate": 4.049158837666594e-06, + "loss": 3.0421, + "step": 8754 + }, + { + "epoch": 1.91, + "learning_rate": 4.0490495958051126e-06, + "loss": 3.0988, + "step": 8755 + }, + { + "epoch": 1.91, + "learning_rate": 4.048940353943632e-06, + "loss": 3.2085, + "step": 8756 + }, + { + "epoch": 1.91, + "learning_rate": 4.04883111208215e-06, + "loss": 2.1079, + "step": 8757 + }, + { + "epoch": 1.91, + "learning_rate": 4.048721870220669e-06, + "loss": 2.4856, + "step": 8758 + }, + { + "epoch": 1.91, + "learning_rate": 4.048612628359188e-06, + "loss": 3.2393, + "step": 8759 + }, + { + "epoch": 1.91, + "learning_rate": 4.048503386497706e-06, + "loss": 3.9468, + "step": 8760 + }, + { + "epoch": 1.91, + "learning_rate": 4.048394144636225e-06, + "loss": 2.1301, + "step": 8761 + }, + { + "epoch": 1.91, + "learning_rate": 4.048284902774743e-06, + "loss": 4.0761, + "step": 8762 + }, + { + "epoch": 1.91, + "learning_rate": 4.048175660913262e-06, + "loss": 3.1376, + "step": 8763 + }, + { + "epoch": 1.91, + "learning_rate": 4.048066419051781e-06, + "loss": 1.0713, + "step": 8764 + }, + { + "epoch": 1.91, + "learning_rate": 4.0479571771902995e-06, + "loss": 2.5067, + "step": 8765 + }, + { + "epoch": 1.91, + "learning_rate": 4.0478479353288186e-06, + "loss": 2.7586, + "step": 8766 + }, + { + "epoch": 1.91, + "learning_rate": 4.047738693467337e-06, + "loss": 3.6966, + "step": 8767 + }, + { + "epoch": 1.91, + "learning_rate": 4.047629451605856e-06, + "loss": 3.0812, + "step": 8768 + }, + { + "epoch": 1.91, + "learning_rate": 4.047520209744375e-06, + "loss": 3.1075, + "step": 8769 + }, + { + "epoch": 1.91, + "learning_rate": 4.047410967882893e-06, + "loss": 1.4928, + "step": 8770 + }, + { + "epoch": 1.91, + "learning_rate": 4.047301726021411e-06, + "loss": 2.3229, + "step": 8771 + }, + { + "epoch": 1.91, + "learning_rate": 4.04719248415993e-06, + "loss": 2.8442, + "step": 8772 + }, + { + "epoch": 1.91, + "learning_rate": 4.047083242298449e-06, + "loss": 3.0908, + "step": 8773 + }, + { + "epoch": 1.91, + "learning_rate": 4.0469740004369674e-06, + "loss": 2.8829, + "step": 8774 + }, + { + "epoch": 1.92, + "learning_rate": 4.0468647585754865e-06, + "loss": 2.1463, + "step": 8775 + }, + { + "epoch": 1.92, + "learning_rate": 4.046755516714005e-06, + "loss": 1.5039, + "step": 8776 + }, + { + "epoch": 1.92, + "learning_rate": 4.046646274852524e-06, + "loss": 2.3179, + "step": 8777 + }, + { + "epoch": 1.92, + "learning_rate": 4.046537032991043e-06, + "loss": 3.296, + "step": 8778 + }, + { + "epoch": 1.92, + "learning_rate": 4.046427791129561e-06, + "loss": 2.2521, + "step": 8779 + }, + { + "epoch": 1.92, + "learning_rate": 4.04631854926808e-06, + "loss": 3.5883, + "step": 8780 + }, + { + "epoch": 1.92, + "learning_rate": 4.046209307406598e-06, + "loss": 2.6832, + "step": 8781 + }, + { + "epoch": 1.92, + "learning_rate": 4.046100065545117e-06, + "loss": 3.3981, + "step": 8782 + }, + { + "epoch": 1.92, + "learning_rate": 4.045990823683636e-06, + "loss": 1.9613, + "step": 8783 + }, + { + "epoch": 1.92, + "learning_rate": 4.045881581822154e-06, + "loss": 3.5154, + "step": 8784 + }, + { + "epoch": 1.92, + "learning_rate": 4.0457723399606734e-06, + "loss": 3.8683, + "step": 8785 + }, + { + "epoch": 1.92, + "learning_rate": 4.045663098099192e-06, + "loss": 3.4451, + "step": 8786 + }, + { + "epoch": 1.92, + "learning_rate": 4.045553856237711e-06, + "loss": 2.1335, + "step": 8787 + }, + { + "epoch": 1.92, + "learning_rate": 4.04544461437623e-06, + "loss": 1.7954, + "step": 8788 + }, + { + "epoch": 1.92, + "learning_rate": 4.045335372514748e-06, + "loss": 3.1637, + "step": 8789 + }, + { + "epoch": 1.92, + "learning_rate": 4.045226130653266e-06, + "loss": 3.2886, + "step": 8790 + }, + { + "epoch": 1.92, + "learning_rate": 4.045116888791785e-06, + "loss": 1.612, + "step": 8791 + }, + { + "epoch": 1.92, + "learning_rate": 4.045007646930304e-06, + "loss": 2.3907, + "step": 8792 + }, + { + "epoch": 1.92, + "learning_rate": 4.044898405068823e-06, + "loss": 1.8254, + "step": 8793 + }, + { + "epoch": 1.92, + "learning_rate": 4.044789163207341e-06, + "loss": 2.9726, + "step": 8794 + }, + { + "epoch": 1.92, + "learning_rate": 4.0446799213458595e-06, + "loss": 1.0684, + "step": 8795 + }, + { + "epoch": 1.92, + "learning_rate": 4.044570679484379e-06, + "loss": 1.8438, + "step": 8796 + }, + { + "epoch": 1.92, + "learning_rate": 4.044461437622898e-06, + "loss": 2.2507, + "step": 8797 + }, + { + "epoch": 1.92, + "learning_rate": 4.044352195761417e-06, + "loss": 3.5021, + "step": 8798 + }, + { + "epoch": 1.92, + "learning_rate": 4.044242953899935e-06, + "loss": 2.755, + "step": 8799 + }, + { + "epoch": 1.92, + "learning_rate": 4.044133712038453e-06, + "loss": 2.6313, + "step": 8800 + }, + { + "epoch": 1.92, + "learning_rate": 4.044024470176972e-06, + "loss": 3.1432, + "step": 8801 + }, + { + "epoch": 1.92, + "learning_rate": 4.043915228315491e-06, + "loss": 1.6509, + "step": 8802 + }, + { + "epoch": 1.92, + "learning_rate": 4.043805986454009e-06, + "loss": 3.1665, + "step": 8803 + }, + { + "epoch": 1.92, + "learning_rate": 4.043696744592528e-06, + "loss": 2.782, + "step": 8804 + }, + { + "epoch": 1.92, + "learning_rate": 4.0435875027310465e-06, + "loss": 2.0482, + "step": 8805 + }, + { + "epoch": 1.92, + "learning_rate": 4.0434782608695655e-06, + "loss": 3.0529, + "step": 8806 + }, + { + "epoch": 1.92, + "learning_rate": 4.0433690190080846e-06, + "loss": 1.6568, + "step": 8807 + }, + { + "epoch": 1.92, + "learning_rate": 4.043259777146603e-06, + "loss": 3.6271, + "step": 8808 + }, + { + "epoch": 1.92, + "learning_rate": 4.043150535285121e-06, + "loss": 3.1515, + "step": 8809 + }, + { + "epoch": 1.92, + "learning_rate": 4.04304129342364e-06, + "loss": 2.6251, + "step": 8810 + }, + { + "epoch": 1.92, + "learning_rate": 4.042932051562159e-06, + "loss": 2.0431, + "step": 8811 + }, + { + "epoch": 1.92, + "learning_rate": 4.042822809700678e-06, + "loss": 3.3034, + "step": 8812 + }, + { + "epoch": 1.92, + "learning_rate": 4.042713567839196e-06, + "loss": 2.9787, + "step": 8813 + }, + { + "epoch": 1.92, + "learning_rate": 4.0426043259777144e-06, + "loss": 2.1929, + "step": 8814 + }, + { + "epoch": 1.92, + "learning_rate": 4.0424950841162335e-06, + "loss": 2.4314, + "step": 8815 + }, + { + "epoch": 1.92, + "learning_rate": 4.0423858422547525e-06, + "loss": 2.9182, + "step": 8816 + }, + { + "epoch": 1.92, + "learning_rate": 4.0422766003932715e-06, + "loss": 2.9603, + "step": 8817 + }, + { + "epoch": 1.92, + "learning_rate": 4.04216735853179e-06, + "loss": 1.8443, + "step": 8818 + }, + { + "epoch": 1.92, + "learning_rate": 4.042058116670308e-06, + "loss": 2.9749, + "step": 8819 + }, + { + "epoch": 1.92, + "learning_rate": 4.041948874808827e-06, + "loss": 2.0335, + "step": 8820 + }, + { + "epoch": 1.93, + "learning_rate": 4.041839632947346e-06, + "loss": 3.3686, + "step": 8821 + }, + { + "epoch": 1.93, + "learning_rate": 4.041730391085865e-06, + "loss": 1.5087, + "step": 8822 + }, + { + "epoch": 1.93, + "learning_rate": 4.041621149224383e-06, + "loss": 2.939, + "step": 8823 + }, + { + "epoch": 1.93, + "learning_rate": 4.041511907362901e-06, + "loss": 2.459, + "step": 8824 + }, + { + "epoch": 1.93, + "learning_rate": 4.04140266550142e-06, + "loss": 1.8261, + "step": 8825 + }, + { + "epoch": 1.93, + "learning_rate": 4.0412934236399395e-06, + "loss": 1.9505, + "step": 8826 + }, + { + "epoch": 1.93, + "learning_rate": 4.041184181778458e-06, + "loss": 3.38, + "step": 8827 + }, + { + "epoch": 1.93, + "learning_rate": 4.041074939916977e-06, + "loss": 2.2408, + "step": 8828 + }, + { + "epoch": 1.93, + "learning_rate": 4.040965698055495e-06, + "loss": 2.5412, + "step": 8829 + }, + { + "epoch": 1.93, + "learning_rate": 4.040856456194014e-06, + "loss": 3.4328, + "step": 8830 + }, + { + "epoch": 1.93, + "learning_rate": 4.040747214332533e-06, + "loss": 2.4694, + "step": 8831 + }, + { + "epoch": 1.93, + "learning_rate": 4.040637972471051e-06, + "loss": 2.9705, + "step": 8832 + }, + { + "epoch": 1.93, + "learning_rate": 4.04052873060957e-06, + "loss": 2.2861, + "step": 8833 + }, + { + "epoch": 1.93, + "learning_rate": 4.040419488748088e-06, + "loss": 3.1744, + "step": 8834 + }, + { + "epoch": 1.93, + "learning_rate": 4.040310246886607e-06, + "loss": 2.4496, + "step": 8835 + }, + { + "epoch": 1.93, + "learning_rate": 4.040201005025126e-06, + "loss": 2.6857, + "step": 8836 + }, + { + "epoch": 1.93, + "learning_rate": 4.040091763163645e-06, + "loss": 2.9323, + "step": 8837 + }, + { + "epoch": 1.93, + "learning_rate": 4.039982521302163e-06, + "loss": 3.1883, + "step": 8838 + }, + { + "epoch": 1.93, + "learning_rate": 4.039873279440682e-06, + "loss": 2.5065, + "step": 8839 + }, + { + "epoch": 1.93, + "learning_rate": 4.039764037579201e-06, + "loss": 3.6413, + "step": 8840 + }, + { + "epoch": 1.93, + "learning_rate": 4.03965479571772e-06, + "loss": 2.3973, + "step": 8841 + }, + { + "epoch": 1.93, + "learning_rate": 4.039545553856238e-06, + "loss": 2.3288, + "step": 8842 + }, + { + "epoch": 1.93, + "learning_rate": 4.039436311994756e-06, + "loss": 3.3427, + "step": 8843 + }, + { + "epoch": 1.93, + "learning_rate": 4.039327070133275e-06, + "loss": 2.1421, + "step": 8844 + }, + { + "epoch": 1.93, + "learning_rate": 4.039217828271794e-06, + "loss": 3.7246, + "step": 8845 + }, + { + "epoch": 1.93, + "learning_rate": 4.039108586410313e-06, + "loss": 2.0137, + "step": 8846 + }, + { + "epoch": 1.93, + "learning_rate": 4.0389993445488316e-06, + "loss": 3.0398, + "step": 8847 + }, + { + "epoch": 1.93, + "learning_rate": 4.03889010268735e-06, + "loss": 2.4041, + "step": 8848 + }, + { + "epoch": 1.93, + "learning_rate": 4.038780860825869e-06, + "loss": 1.8358, + "step": 8849 + }, + { + "epoch": 1.93, + "learning_rate": 4.038671618964388e-06, + "loss": 3.2065, + "step": 8850 + }, + { + "epoch": 1.93, + "learning_rate": 4.038562377102907e-06, + "loss": 2.5636, + "step": 8851 + }, + { + "epoch": 1.93, + "learning_rate": 4.038453135241425e-06, + "loss": 2.6042, + "step": 8852 + }, + { + "epoch": 1.93, + "learning_rate": 4.038343893379943e-06, + "loss": 2.3647, + "step": 8853 + }, + { + "epoch": 1.93, + "learning_rate": 4.038234651518462e-06, + "loss": 2.3165, + "step": 8854 + }, + { + "epoch": 1.93, + "learning_rate": 4.038125409656981e-06, + "loss": 2.9078, + "step": 8855 + }, + { + "epoch": 1.93, + "learning_rate": 4.0380161677954995e-06, + "loss": 3.5629, + "step": 8856 + }, + { + "epoch": 1.93, + "learning_rate": 4.0379069259340185e-06, + "loss": 2.5951, + "step": 8857 + }, + { + "epoch": 1.93, + "learning_rate": 4.037797684072537e-06, + "loss": 2.3792, + "step": 8858 + }, + { + "epoch": 1.93, + "learning_rate": 4.037688442211056e-06, + "loss": 2.041, + "step": 8859 + }, + { + "epoch": 1.93, + "learning_rate": 4.037579200349575e-06, + "loss": 2.7125, + "step": 8860 + }, + { + "epoch": 1.93, + "learning_rate": 4.037469958488093e-06, + "loss": 3.2447, + "step": 8861 + }, + { + "epoch": 1.93, + "learning_rate": 4.037360716626611e-06, + "loss": 1.9374, + "step": 8862 + }, + { + "epoch": 1.93, + "learning_rate": 4.03725147476513e-06, + "loss": 3.371, + "step": 8863 + }, + { + "epoch": 1.93, + "learning_rate": 4.037142232903649e-06, + "loss": 1.1241, + "step": 8864 + }, + { + "epoch": 1.93, + "learning_rate": 4.037032991042168e-06, + "loss": 3.2955, + "step": 8865 + }, + { + "epoch": 1.93, + "learning_rate": 4.0369237491806864e-06, + "loss": 2.0232, + "step": 8866 + }, + { + "epoch": 1.94, + "learning_rate": 4.036814507319205e-06, + "loss": 3.3282, + "step": 8867 + }, + { + "epoch": 1.94, + "learning_rate": 4.036705265457724e-06, + "loss": 2.1797, + "step": 8868 + }, + { + "epoch": 1.94, + "learning_rate": 4.036596023596243e-06, + "loss": 2.8278, + "step": 8869 + }, + { + "epoch": 1.94, + "learning_rate": 4.036486781734761e-06, + "loss": 3.0801, + "step": 8870 + }, + { + "epoch": 1.94, + "learning_rate": 4.03637753987328e-06, + "loss": 2.682, + "step": 8871 + }, + { + "epoch": 1.94, + "learning_rate": 4.036268298011798e-06, + "loss": 1.9278, + "step": 8872 + }, + { + "epoch": 1.94, + "learning_rate": 4.036159056150317e-06, + "loss": 2.3855, + "step": 8873 + }, + { + "epoch": 1.94, + "learning_rate": 4.036049814288836e-06, + "loss": 2.0345, + "step": 8874 + }, + { + "epoch": 1.94, + "learning_rate": 4.035940572427354e-06, + "loss": 2.6124, + "step": 8875 + }, + { + "epoch": 1.94, + "learning_rate": 4.035831330565873e-06, + "loss": 3.0151, + "step": 8876 + }, + { + "epoch": 1.94, + "learning_rate": 4.035722088704392e-06, + "loss": 1.6307, + "step": 8877 + }, + { + "epoch": 1.94, + "learning_rate": 4.035612846842911e-06, + "loss": 3.9906, + "step": 8878 + }, + { + "epoch": 1.94, + "learning_rate": 4.03550360498143e-06, + "loss": 2.2811, + "step": 8879 + }, + { + "epoch": 1.94, + "learning_rate": 4.035394363119948e-06, + "loss": 1.6961, + "step": 8880 + }, + { + "epoch": 1.94, + "learning_rate": 4.035285121258467e-06, + "loss": 2.4118, + "step": 8881 + }, + { + "epoch": 1.94, + "learning_rate": 4.035175879396985e-06, + "loss": 2.5376, + "step": 8882 + }, + { + "epoch": 1.94, + "learning_rate": 4.035066637535504e-06, + "loss": 2.811, + "step": 8883 + }, + { + "epoch": 1.94, + "learning_rate": 4.034957395674023e-06, + "loss": 2.6289, + "step": 8884 + }, + { + "epoch": 1.94, + "learning_rate": 4.034848153812541e-06, + "loss": 3.4821, + "step": 8885 + }, + { + "epoch": 1.94, + "learning_rate": 4.03473891195106e-06, + "loss": 2.5266, + "step": 8886 + }, + { + "epoch": 1.94, + "learning_rate": 4.0346296700895785e-06, + "loss": 2.3182, + "step": 8887 + }, + { + "epoch": 1.94, + "learning_rate": 4.0345204282280976e-06, + "loss": 3.0647, + "step": 8888 + }, + { + "epoch": 1.94, + "learning_rate": 4.034411186366616e-06, + "loss": 2.9844, + "step": 8889 + }, + { + "epoch": 1.94, + "learning_rate": 4.034301944505135e-06, + "loss": 2.7122, + "step": 8890 + }, + { + "epoch": 1.94, + "learning_rate": 4.034192702643653e-06, + "loss": 2.262, + "step": 8891 + }, + { + "epoch": 1.94, + "learning_rate": 4.034083460782172e-06, + "loss": 2.1741, + "step": 8892 + }, + { + "epoch": 1.94, + "learning_rate": 4.033974218920691e-06, + "loss": 2.992, + "step": 8893 + }, + { + "epoch": 1.94, + "learning_rate": 4.033864977059209e-06, + "loss": 2.5977, + "step": 8894 + }, + { + "epoch": 1.94, + "learning_rate": 4.033755735197728e-06, + "loss": 2.3252, + "step": 8895 + }, + { + "epoch": 1.94, + "learning_rate": 4.0336464933362465e-06, + "loss": 3.3455, + "step": 8896 + }, + { + "epoch": 1.94, + "learning_rate": 4.0335372514747655e-06, + "loss": 2.5856, + "step": 8897 + }, + { + "epoch": 1.94, + "learning_rate": 4.0334280096132845e-06, + "loss": 3.3969, + "step": 8898 + }, + { + "epoch": 1.94, + "learning_rate": 4.033318767751803e-06, + "loss": 2.7067, + "step": 8899 + }, + { + "epoch": 1.94, + "learning_rate": 4.033209525890322e-06, + "loss": 2.5142, + "step": 8900 + }, + { + "epoch": 1.94, + "learning_rate": 4.03310028402884e-06, + "loss": 2.4779, + "step": 8901 + }, + { + "epoch": 1.94, + "learning_rate": 4.032991042167359e-06, + "loss": 2.6432, + "step": 8902 + }, + { + "epoch": 1.94, + "learning_rate": 4.032881800305877e-06, + "loss": 2.3258, + "step": 8903 + }, + { + "epoch": 1.94, + "learning_rate": 4.032772558444396e-06, + "loss": 1.2261, + "step": 8904 + }, + { + "epoch": 1.94, + "learning_rate": 4.032663316582915e-06, + "loss": 2.8627, + "step": 8905 + }, + { + "epoch": 1.94, + "learning_rate": 4.032554074721433e-06, + "loss": 2.9642, + "step": 8906 + }, + { + "epoch": 1.94, + "learning_rate": 4.0324448328599525e-06, + "loss": 2.527, + "step": 8907 + }, + { + "epoch": 1.94, + "learning_rate": 4.032335590998471e-06, + "loss": 2.9567, + "step": 8908 + }, + { + "epoch": 1.94, + "learning_rate": 4.03222634913699e-06, + "loss": 2.5572, + "step": 8909 + }, + { + "epoch": 1.94, + "learning_rate": 4.032117107275509e-06, + "loss": 2.4518, + "step": 8910 + }, + { + "epoch": 1.94, + "learning_rate": 4.032007865414027e-06, + "loss": 2.4933, + "step": 8911 + }, + { + "epoch": 1.95, + "learning_rate": 4.031898623552546e-06, + "loss": 2.6825, + "step": 8912 + }, + { + "epoch": 1.95, + "learning_rate": 4.031789381691064e-06, + "loss": 2.3245, + "step": 8913 + }, + { + "epoch": 1.95, + "learning_rate": 4.031680139829583e-06, + "loss": 1.7632, + "step": 8914 + }, + { + "epoch": 1.95, + "learning_rate": 4.031570897968101e-06, + "loss": 3.0106, + "step": 8915 + }, + { + "epoch": 1.95, + "learning_rate": 4.03146165610662e-06, + "loss": 2.4424, + "step": 8916 + }, + { + "epoch": 1.95, + "learning_rate": 4.0313524142451386e-06, + "loss": 2.7162, + "step": 8917 + }, + { + "epoch": 1.95, + "learning_rate": 4.031243172383658e-06, + "loss": 1.9018, + "step": 8918 + }, + { + "epoch": 1.95, + "learning_rate": 4.031133930522177e-06, + "loss": 2.6817, + "step": 8919 + }, + { + "epoch": 1.95, + "learning_rate": 4.031024688660695e-06, + "loss": 2.093, + "step": 8920 + }, + { + "epoch": 1.95, + "learning_rate": 4.030915446799214e-06, + "loss": 2.5767, + "step": 8921 + }, + { + "epoch": 1.95, + "learning_rate": 4.030806204937732e-06, + "loss": 2.2668, + "step": 8922 + }, + { + "epoch": 1.95, + "learning_rate": 4.030696963076251e-06, + "loss": 2.4489, + "step": 8923 + }, + { + "epoch": 1.95, + "learning_rate": 4.03058772121477e-06, + "loss": 2.5917, + "step": 8924 + }, + { + "epoch": 1.95, + "learning_rate": 4.030478479353288e-06, + "loss": 3.2752, + "step": 8925 + }, + { + "epoch": 1.95, + "learning_rate": 4.030369237491807e-06, + "loss": 2.4674, + "step": 8926 + }, + { + "epoch": 1.95, + "learning_rate": 4.0302599956303255e-06, + "loss": 2.3724, + "step": 8927 + }, + { + "epoch": 1.95, + "learning_rate": 4.0301507537688446e-06, + "loss": 2.6383, + "step": 8928 + }, + { + "epoch": 1.95, + "learning_rate": 4.030041511907364e-06, + "loss": 3.1731, + "step": 8929 + }, + { + "epoch": 1.95, + "learning_rate": 4.029932270045882e-06, + "loss": 1.6776, + "step": 8930 + }, + { + "epoch": 1.95, + "learning_rate": 4.029823028184401e-06, + "loss": 2.9659, + "step": 8931 + }, + { + "epoch": 1.95, + "learning_rate": 4.029713786322919e-06, + "loss": 2.7352, + "step": 8932 + }, + { + "epoch": 1.95, + "learning_rate": 4.029604544461438e-06, + "loss": 3.1137, + "step": 8933 + }, + { + "epoch": 1.95, + "learning_rate": 4.029495302599957e-06, + "loss": 2.1701, + "step": 8934 + }, + { + "epoch": 1.95, + "learning_rate": 4.029386060738475e-06, + "loss": 2.2621, + "step": 8935 + }, + { + "epoch": 1.95, + "learning_rate": 4.0292768188769934e-06, + "loss": 3.6092, + "step": 8936 + }, + { + "epoch": 1.95, + "learning_rate": 4.0291675770155125e-06, + "loss": 2.4163, + "step": 8937 + }, + { + "epoch": 1.95, + "learning_rate": 4.0290583351540315e-06, + "loss": 1.6009, + "step": 8938 + }, + { + "epoch": 1.95, + "learning_rate": 4.0289490932925505e-06, + "loss": 2.7014, + "step": 8939 + }, + { + "epoch": 1.95, + "learning_rate": 4.028839851431069e-06, + "loss": 2.8678, + "step": 8940 + }, + { + "epoch": 1.95, + "learning_rate": 4.028730609569587e-06, + "loss": 3.2122, + "step": 8941 + }, + { + "epoch": 1.95, + "learning_rate": 4.028621367708106e-06, + "loss": 3.633, + "step": 8942 + }, + { + "epoch": 1.95, + "learning_rate": 4.028512125846625e-06, + "loss": 2.2419, + "step": 8943 + }, + { + "epoch": 1.95, + "learning_rate": 4.028402883985143e-06, + "loss": 2.4562, + "step": 8944 + }, + { + "epoch": 1.95, + "learning_rate": 4.028293642123662e-06, + "loss": 2.9222, + "step": 8945 + }, + { + "epoch": 1.95, + "learning_rate": 4.02818440026218e-06, + "loss": 3.6293, + "step": 8946 + }, + { + "epoch": 1.95, + "learning_rate": 4.0280751584006994e-06, + "loss": 3.0424, + "step": 8947 + }, + { + "epoch": 1.95, + "learning_rate": 4.0279659165392185e-06, + "loss": 2.0386, + "step": 8948 + }, + { + "epoch": 1.95, + "learning_rate": 4.027856674677737e-06, + "loss": 2.4696, + "step": 8949 + }, + { + "epoch": 1.95, + "learning_rate": 4.027747432816255e-06, + "loss": 1.9346, + "step": 8950 + }, + { + "epoch": 1.95, + "learning_rate": 4.027638190954774e-06, + "loss": 1.9308, + "step": 8951 + }, + { + "epoch": 1.95, + "learning_rate": 4.027528949093293e-06, + "loss": 2.8358, + "step": 8952 + }, + { + "epoch": 1.95, + "learning_rate": 4.027419707231812e-06, + "loss": 1.7892, + "step": 8953 + }, + { + "epoch": 1.95, + "learning_rate": 4.02731046537033e-06, + "loss": 2.158, + "step": 8954 + }, + { + "epoch": 1.95, + "learning_rate": 4.027201223508848e-06, + "loss": 2.0785, + "step": 8955 + }, + { + "epoch": 1.95, + "learning_rate": 4.027091981647367e-06, + "loss": 3.2775, + "step": 8956 + }, + { + "epoch": 1.95, + "learning_rate": 4.026982739785886e-06, + "loss": 2.4764, + "step": 8957 + }, + { + "epoch": 1.96, + "learning_rate": 4.026873497924405e-06, + "loss": 3.0805, + "step": 8958 + }, + { + "epoch": 1.96, + "learning_rate": 4.026764256062924e-06, + "loss": 2.7372, + "step": 8959 + }, + { + "epoch": 1.96, + "learning_rate": 4.026655014201442e-06, + "loss": 1.99, + "step": 8960 + }, + { + "epoch": 1.96, + "learning_rate": 4.026545772339961e-06, + "loss": 2.883, + "step": 8961 + }, + { + "epoch": 1.96, + "learning_rate": 4.02643653047848e-06, + "loss": 1.5102, + "step": 8962 + }, + { + "epoch": 1.96, + "learning_rate": 4.026327288616999e-06, + "loss": 2.6883, + "step": 8963 + }, + { + "epoch": 1.96, + "learning_rate": 4.026218046755517e-06, + "loss": 3.2881, + "step": 8964 + }, + { + "epoch": 1.96, + "learning_rate": 4.026108804894035e-06, + "loss": 2.1497, + "step": 8965 + }, + { + "epoch": 1.96, + "learning_rate": 4.025999563032554e-06, + "loss": 2.5545, + "step": 8966 + }, + { + "epoch": 1.96, + "learning_rate": 4.025890321171073e-06, + "loss": 3.6937, + "step": 8967 + }, + { + "epoch": 1.96, + "learning_rate": 4.0257810793095915e-06, + "loss": 3.3542, + "step": 8968 + }, + { + "epoch": 1.96, + "learning_rate": 4.0256718374481106e-06, + "loss": 1.9097, + "step": 8969 + }, + { + "epoch": 1.96, + "learning_rate": 4.025562595586629e-06, + "loss": 1.4944, + "step": 8970 + }, + { + "epoch": 1.96, + "learning_rate": 4.025453353725148e-06, + "loss": 1.6597, + "step": 8971 + }, + { + "epoch": 1.96, + "learning_rate": 4.025344111863667e-06, + "loss": 3.193, + "step": 8972 + }, + { + "epoch": 1.96, + "learning_rate": 4.025234870002185e-06, + "loss": 2.0552, + "step": 8973 + }, + { + "epoch": 1.96, + "learning_rate": 4.025125628140704e-06, + "loss": 1.8736, + "step": 8974 + }, + { + "epoch": 1.96, + "learning_rate": 4.025016386279222e-06, + "loss": 1.9836, + "step": 8975 + }, + { + "epoch": 1.96, + "learning_rate": 4.024907144417741e-06, + "loss": 2.894, + "step": 8976 + }, + { + "epoch": 1.96, + "learning_rate": 4.02479790255626e-06, + "loss": 2.291, + "step": 8977 + }, + { + "epoch": 1.96, + "learning_rate": 4.0246886606947785e-06, + "loss": 2.5129, + "step": 8978 + }, + { + "epoch": 1.96, + "learning_rate": 4.024579418833297e-06, + "loss": 2.4427, + "step": 8979 + }, + { + "epoch": 1.96, + "learning_rate": 4.024470176971816e-06, + "loss": 1.9057, + "step": 8980 + }, + { + "epoch": 1.96, + "learning_rate": 4.024360935110335e-06, + "loss": 3.0342, + "step": 8981 + }, + { + "epoch": 1.96, + "learning_rate": 4.024251693248854e-06, + "loss": 2.4087, + "step": 8982 + }, + { + "epoch": 1.96, + "learning_rate": 4.024142451387372e-06, + "loss": 3.1768, + "step": 8983 + }, + { + "epoch": 1.96, + "learning_rate": 4.02403320952589e-06, + "loss": 1.5573, + "step": 8984 + }, + { + "epoch": 1.96, + "learning_rate": 4.023923967664409e-06, + "loss": 2.4312, + "step": 8985 + }, + { + "epoch": 1.96, + "learning_rate": 4.023814725802928e-06, + "loss": 2.4863, + "step": 8986 + }, + { + "epoch": 1.96, + "learning_rate": 4.023705483941447e-06, + "loss": 2.6301, + "step": 8987 + }, + { + "epoch": 1.96, + "learning_rate": 4.0235962420799654e-06, + "loss": 2.9755, + "step": 8988 + }, + { + "epoch": 1.96, + "learning_rate": 4.023487000218484e-06, + "loss": 2.122, + "step": 8989 + }, + { + "epoch": 1.96, + "learning_rate": 4.023377758357003e-06, + "loss": 2.4407, + "step": 8990 + }, + { + "epoch": 1.96, + "learning_rate": 4.023268516495522e-06, + "loss": 2.0776, + "step": 8991 + }, + { + "epoch": 1.96, + "learning_rate": 4.023159274634041e-06, + "loss": 1.6514, + "step": 8992 + }, + { + "epoch": 1.96, + "learning_rate": 4.023050032772559e-06, + "loss": 2.9094, + "step": 8993 + }, + { + "epoch": 1.96, + "learning_rate": 4.022940790911077e-06, + "loss": 2.5965, + "step": 8994 + }, + { + "epoch": 1.96, + "learning_rate": 4.022831549049596e-06, + "loss": 2.3681, + "step": 8995 + }, + { + "epoch": 1.96, + "learning_rate": 4.022722307188115e-06, + "loss": 1.0053, + "step": 8996 + }, + { + "epoch": 1.96, + "learning_rate": 4.022613065326633e-06, + "loss": 2.5052, + "step": 8997 + }, + { + "epoch": 1.96, + "learning_rate": 4.022503823465152e-06, + "loss": 1.7237, + "step": 8998 + }, + { + "epoch": 1.96, + "learning_rate": 4.022394581603671e-06, + "loss": 2.0923, + "step": 8999 + }, + { + "epoch": 1.96, + "learning_rate": 4.02228533974219e-06, + "loss": 3.2918, + "step": 9000 + }, + { + "epoch": 1.96, + "learning_rate": 4.022176097880709e-06, + "loss": 2.1857, + "step": 9001 + }, + { + "epoch": 1.96, + "learning_rate": 4.022066856019227e-06, + "loss": 2.9184, + "step": 9002 + }, + { + "epoch": 1.96, + "learning_rate": 4.021957614157745e-06, + "loss": 2.7199, + "step": 9003 + }, + { + "epoch": 1.97, + "learning_rate": 4.021848372296264e-06, + "loss": 2.8412, + "step": 9004 + }, + { + "epoch": 1.97, + "learning_rate": 4.021739130434783e-06, + "loss": 2.1946, + "step": 9005 + }, + { + "epoch": 1.97, + "learning_rate": 4.021629888573302e-06, + "loss": 2.9591, + "step": 9006 + }, + { + "epoch": 1.97, + "learning_rate": 4.02152064671182e-06, + "loss": 2.286, + "step": 9007 + }, + { + "epoch": 1.97, + "learning_rate": 4.0214114048503385e-06, + "loss": 2.1844, + "step": 9008 + }, + { + "epoch": 1.97, + "learning_rate": 4.0213021629888575e-06, + "loss": 2.2456, + "step": 9009 + }, + { + "epoch": 1.97, + "learning_rate": 4.021192921127377e-06, + "loss": 2.5825, + "step": 9010 + }, + { + "epoch": 1.97, + "learning_rate": 4.021083679265896e-06, + "loss": 2.3632, + "step": 9011 + }, + { + "epoch": 1.97, + "learning_rate": 4.020974437404414e-06, + "loss": 2.4135, + "step": 9012 + }, + { + "epoch": 1.97, + "learning_rate": 4.020865195542932e-06, + "loss": 2.3808, + "step": 9013 + }, + { + "epoch": 1.97, + "learning_rate": 4.020755953681451e-06, + "loss": 3.2575, + "step": 9014 + }, + { + "epoch": 1.97, + "learning_rate": 4.02064671181997e-06, + "loss": 3.4457, + "step": 9015 + }, + { + "epoch": 1.97, + "learning_rate": 4.020537469958488e-06, + "loss": 2.9071, + "step": 9016 + }, + { + "epoch": 1.97, + "learning_rate": 4.020428228097007e-06, + "loss": 3.4442, + "step": 9017 + }, + { + "epoch": 1.97, + "learning_rate": 4.0203189862355255e-06, + "loss": 2.1226, + "step": 9018 + }, + { + "epoch": 1.97, + "learning_rate": 4.0202097443740445e-06, + "loss": 3.377, + "step": 9019 + }, + { + "epoch": 1.97, + "learning_rate": 4.0201005025125635e-06, + "loss": 1.5565, + "step": 9020 + }, + { + "epoch": 1.97, + "learning_rate": 4.019991260651082e-06, + "loss": 2.3755, + "step": 9021 + }, + { + "epoch": 1.97, + "learning_rate": 4.019882018789601e-06, + "loss": 1.77, + "step": 9022 + }, + { + "epoch": 1.97, + "learning_rate": 4.019772776928119e-06, + "loss": 2.7604, + "step": 9023 + }, + { + "epoch": 1.97, + "learning_rate": 4.019663535066638e-06, + "loss": 2.9564, + "step": 9024 + }, + { + "epoch": 1.97, + "learning_rate": 4.019554293205157e-06, + "loss": 2.0164, + "step": 9025 + }, + { + "epoch": 1.97, + "learning_rate": 4.019445051343675e-06, + "loss": 3.2358, + "step": 9026 + }, + { + "epoch": 1.97, + "learning_rate": 4.019335809482194e-06, + "loss": 2.6781, + "step": 9027 + }, + { + "epoch": 1.97, + "learning_rate": 4.0192265676207124e-06, + "loss": 1.8441, + "step": 9028 + }, + { + "epoch": 1.97, + "learning_rate": 4.0191173257592315e-06, + "loss": 3.1906, + "step": 9029 + }, + { + "epoch": 1.97, + "learning_rate": 4.0190080838977505e-06, + "loss": 1.9008, + "step": 9030 + }, + { + "epoch": 1.97, + "learning_rate": 4.018898842036269e-06, + "loss": 2.8224, + "step": 9031 + }, + { + "epoch": 1.97, + "learning_rate": 4.018789600174787e-06, + "loss": 3.3768, + "step": 9032 + }, + { + "epoch": 1.97, + "learning_rate": 4.018680358313306e-06, + "loss": 3.26, + "step": 9033 + }, + { + "epoch": 1.97, + "learning_rate": 4.018571116451825e-06, + "loss": 2.6136, + "step": 9034 + }, + { + "epoch": 1.97, + "learning_rate": 4.018461874590343e-06, + "loss": 2.2427, + "step": 9035 + }, + { + "epoch": 1.97, + "learning_rate": 4.018352632728862e-06, + "loss": 2.1561, + "step": 9036 + }, + { + "epoch": 1.97, + "learning_rate": 4.01824339086738e-06, + "loss": 3.8623, + "step": 9037 + }, + { + "epoch": 1.97, + "learning_rate": 4.018134149005899e-06, + "loss": 2.259, + "step": 9038 + }, + { + "epoch": 1.97, + "learning_rate": 4.018024907144418e-06, + "loss": 2.4838, + "step": 9039 + }, + { + "epoch": 1.97, + "learning_rate": 4.017915665282937e-06, + "loss": 2.7934, + "step": 9040 + }, + { + "epoch": 1.97, + "learning_rate": 4.017806423421456e-06, + "loss": 1.9394, + "step": 9041 + }, + { + "epoch": 1.97, + "learning_rate": 4.017697181559974e-06, + "loss": 2.7874, + "step": 9042 + }, + { + "epoch": 1.97, + "learning_rate": 4.017587939698493e-06, + "loss": 2.6979, + "step": 9043 + }, + { + "epoch": 1.97, + "learning_rate": 4.017478697837012e-06, + "loss": 3.9839, + "step": 9044 + }, + { + "epoch": 1.97, + "learning_rate": 4.01736945597553e-06, + "loss": 2.1479, + "step": 9045 + }, + { + "epoch": 1.97, + "learning_rate": 4.017260214114049e-06, + "loss": 2.1252, + "step": 9046 + }, + { + "epoch": 1.97, + "learning_rate": 4.017150972252567e-06, + "loss": 3.1695, + "step": 9047 + }, + { + "epoch": 1.97, + "learning_rate": 4.017041730391086e-06, + "loss": 1.4021, + "step": 9048 + }, + { + "epoch": 1.97, + "learning_rate": 4.0169324885296045e-06, + "loss": 2.7368, + "step": 9049 + }, + { + "epoch": 1.98, + "learning_rate": 4.0168232466681236e-06, + "loss": 2.8796, + "step": 9050 + }, + { + "epoch": 1.98, + "learning_rate": 4.016714004806643e-06, + "loss": 2.5539, + "step": 9051 + }, + { + "epoch": 1.98, + "learning_rate": 4.016604762945161e-06, + "loss": 1.3275, + "step": 9052 + }, + { + "epoch": 1.98, + "learning_rate": 4.01649552108368e-06, + "loss": 3.0821, + "step": 9053 + }, + { + "epoch": 1.98, + "learning_rate": 4.016386279222198e-06, + "loss": 3.1385, + "step": 9054 + }, + { + "epoch": 1.98, + "learning_rate": 4.016277037360717e-06, + "loss": 2.476, + "step": 9055 + }, + { + "epoch": 1.98, + "learning_rate": 4.016167795499235e-06, + "loss": 3.2165, + "step": 9056 + }, + { + "epoch": 1.98, + "learning_rate": 4.016058553637754e-06, + "loss": 2.7208, + "step": 9057 + }, + { + "epoch": 1.98, + "learning_rate": 4.015949311776273e-06, + "loss": 2.2342, + "step": 9058 + }, + { + "epoch": 1.98, + "learning_rate": 4.0158400699147915e-06, + "loss": 4.0304, + "step": 9059 + }, + { + "epoch": 1.98, + "learning_rate": 4.0157308280533105e-06, + "loss": 1.9647, + "step": 9060 + }, + { + "epoch": 1.98, + "learning_rate": 4.015621586191829e-06, + "loss": 2.7885, + "step": 9061 + }, + { + "epoch": 1.98, + "learning_rate": 4.015512344330348e-06, + "loss": 2.9245, + "step": 9062 + }, + { + "epoch": 1.98, + "learning_rate": 4.015403102468867e-06, + "loss": 3.1931, + "step": 9063 + }, + { + "epoch": 1.98, + "learning_rate": 4.015293860607385e-06, + "loss": 2.2434, + "step": 9064 + }, + { + "epoch": 1.98, + "learning_rate": 4.015184618745904e-06, + "loss": 2.2176, + "step": 9065 + }, + { + "epoch": 1.98, + "learning_rate": 4.015075376884422e-06, + "loss": 2.1254, + "step": 9066 + }, + { + "epoch": 1.98, + "learning_rate": 4.014966135022941e-06, + "loss": 2.6898, + "step": 9067 + }, + { + "epoch": 1.98, + "learning_rate": 4.014856893161459e-06, + "loss": 2.214, + "step": 9068 + }, + { + "epoch": 1.98, + "learning_rate": 4.0147476512999784e-06, + "loss": 3.6744, + "step": 9069 + }, + { + "epoch": 1.98, + "learning_rate": 4.0146384094384975e-06, + "loss": 2.3116, + "step": 9070 + }, + { + "epoch": 1.98, + "learning_rate": 4.014529167577016e-06, + "loss": 3.045, + "step": 9071 + }, + { + "epoch": 1.98, + "learning_rate": 4.014419925715535e-06, + "loss": 2.1008, + "step": 9072 + }, + { + "epoch": 1.98, + "learning_rate": 4.014310683854053e-06, + "loss": 2.3097, + "step": 9073 + }, + { + "epoch": 1.98, + "learning_rate": 4.014201441992572e-06, + "loss": 1.9563, + "step": 9074 + }, + { + "epoch": 1.98, + "learning_rate": 4.014092200131091e-06, + "loss": 1.4483, + "step": 9075 + }, + { + "epoch": 1.98, + "learning_rate": 4.013982958269609e-06, + "loss": 2.9567, + "step": 9076 + }, + { + "epoch": 1.98, + "learning_rate": 4.013873716408128e-06, + "loss": 1.9901, + "step": 9077 + }, + { + "epoch": 1.98, + "learning_rate": 4.013764474546646e-06, + "loss": 1.3699, + "step": 9078 + }, + { + "epoch": 1.98, + "learning_rate": 4.013655232685165e-06, + "loss": 1.3132, + "step": 9079 + }, + { + "epoch": 1.98, + "learning_rate": 4.0135459908236844e-06, + "loss": 3.68, + "step": 9080 + }, + { + "epoch": 1.98, + "learning_rate": 4.013436748962203e-06, + "loss": 1.8318, + "step": 9081 + }, + { + "epoch": 1.98, + "learning_rate": 4.013327507100721e-06, + "loss": 1.9442, + "step": 9082 + }, + { + "epoch": 1.98, + "learning_rate": 4.01321826523924e-06, + "loss": 2.4058, + "step": 9083 + }, + { + "epoch": 1.98, + "learning_rate": 4.013109023377759e-06, + "loss": 2.6862, + "step": 9084 + }, + { + "epoch": 1.98, + "learning_rate": 4.012999781516277e-06, + "loss": 3.3596, + "step": 9085 + }, + { + "epoch": 1.98, + "learning_rate": 4.012890539654796e-06, + "loss": 2.7383, + "step": 9086 + }, + { + "epoch": 1.98, + "learning_rate": 4.012781297793314e-06, + "loss": 2.3757, + "step": 9087 + }, + { + "epoch": 1.98, + "learning_rate": 4.012672055931833e-06, + "loss": 3.862, + "step": 9088 + }, + { + "epoch": 1.98, + "learning_rate": 4.012562814070352e-06, + "loss": 3.0848, + "step": 9089 + }, + { + "epoch": 1.98, + "learning_rate": 4.0124535722088705e-06, + "loss": 1.9964, + "step": 9090 + }, + { + "epoch": 1.98, + "learning_rate": 4.01234433034739e-06, + "loss": 2.7123, + "step": 9091 + }, + { + "epoch": 1.98, + "learning_rate": 4.012235088485908e-06, + "loss": 2.2131, + "step": 9092 + }, + { + "epoch": 1.98, + "learning_rate": 4.012125846624427e-06, + "loss": 1.9622, + "step": 9093 + }, + { + "epoch": 1.98, + "learning_rate": 4.012016604762946e-06, + "loss": 2.7717, + "step": 9094 + }, + { + "epoch": 1.98, + "learning_rate": 4.011907362901464e-06, + "loss": 2.3968, + "step": 9095 + }, + { + "epoch": 1.99, + "learning_rate": 4.011798121039983e-06, + "loss": 3.3617, + "step": 9096 + }, + { + "epoch": 1.99, + "learning_rate": 4.011688879178501e-06, + "loss": 2.4409, + "step": 9097 + }, + { + "epoch": 1.99, + "learning_rate": 4.01157963731702e-06, + "loss": 2.2655, + "step": 9098 + }, + { + "epoch": 1.99, + "learning_rate": 4.011470395455539e-06, + "loss": 3.9936, + "step": 9099 + }, + { + "epoch": 1.99, + "learning_rate": 4.0113611535940575e-06, + "loss": 2.8905, + "step": 9100 + }, + { + "epoch": 1.99, + "learning_rate": 4.011251911732576e-06, + "loss": 3.0198, + "step": 9101 + }, + { + "epoch": 1.99, + "learning_rate": 4.011142669871095e-06, + "loss": 3.6974, + "step": 9102 + }, + { + "epoch": 1.99, + "learning_rate": 4.011033428009614e-06, + "loss": 2.465, + "step": 9103 + }, + { + "epoch": 1.99, + "learning_rate": 4.010924186148133e-06, + "loss": 2.3084, + "step": 9104 + }, + { + "epoch": 1.99, + "learning_rate": 4.010814944286651e-06, + "loss": 1.8865, + "step": 9105 + }, + { + "epoch": 1.99, + "learning_rate": 4.010705702425169e-06, + "loss": 1.076, + "step": 9106 + }, + { + "epoch": 1.99, + "learning_rate": 4.010596460563688e-06, + "loss": 3.4908, + "step": 9107 + }, + { + "epoch": 1.99, + "learning_rate": 4.010487218702207e-06, + "loss": 3.0079, + "step": 9108 + }, + { + "epoch": 1.99, + "learning_rate": 4.010377976840725e-06, + "loss": 1.6602, + "step": 9109 + }, + { + "epoch": 1.99, + "learning_rate": 4.0102687349792445e-06, + "loss": 2.0679, + "step": 9110 + }, + { + "epoch": 1.99, + "learning_rate": 4.010159493117763e-06, + "loss": 1.9235, + "step": 9111 + }, + { + "epoch": 1.99, + "learning_rate": 4.010050251256282e-06, + "loss": 2.8475, + "step": 9112 + }, + { + "epoch": 1.99, + "learning_rate": 4.009941009394801e-06, + "loss": 3.0413, + "step": 9113 + }, + { + "epoch": 1.99, + "learning_rate": 4.009831767533319e-06, + "loss": 1.6132, + "step": 9114 + }, + { + "epoch": 1.99, + "learning_rate": 4.009722525671837e-06, + "loss": 2.8553, + "step": 9115 + }, + { + "epoch": 1.99, + "learning_rate": 4.009613283810356e-06, + "loss": 3.1532, + "step": 9116 + }, + { + "epoch": 1.99, + "learning_rate": 4.009504041948875e-06, + "loss": 1.8741, + "step": 9117 + }, + { + "epoch": 1.99, + "learning_rate": 4.009394800087394e-06, + "loss": 2.5172, + "step": 9118 + }, + { + "epoch": 1.99, + "learning_rate": 4.009285558225912e-06, + "loss": 1.7345, + "step": 9119 + }, + { + "epoch": 1.99, + "learning_rate": 4.0091763163644306e-06, + "loss": 2.929, + "step": 9120 + }, + { + "epoch": 1.99, + "learning_rate": 4.00906707450295e-06, + "loss": 2.8713, + "step": 9121 + }, + { + "epoch": 1.99, + "learning_rate": 4.008957832641469e-06, + "loss": 2.9361, + "step": 9122 + }, + { + "epoch": 1.99, + "learning_rate": 4.008848590779988e-06, + "loss": 2.4344, + "step": 9123 + }, + { + "epoch": 1.99, + "learning_rate": 4.008739348918506e-06, + "loss": 2.2187, + "step": 9124 + }, + { + "epoch": 1.99, + "learning_rate": 4.008630107057024e-06, + "loss": 3.3979, + "step": 9125 + }, + { + "epoch": 1.99, + "learning_rate": 4.008520865195543e-06, + "loss": 2.5951, + "step": 9126 + }, + { + "epoch": 1.99, + "learning_rate": 4.008411623334062e-06, + "loss": 2.9982, + "step": 9127 + }, + { + "epoch": 1.99, + "learning_rate": 4.008302381472581e-06, + "loss": 2.4427, + "step": 9128 + }, + { + "epoch": 1.99, + "learning_rate": 4.008193139611099e-06, + "loss": 3.2282, + "step": 9129 + }, + { + "epoch": 1.99, + "learning_rate": 4.0080838977496175e-06, + "loss": 3.352, + "step": 9130 + }, + { + "epoch": 1.99, + "learning_rate": 4.0079746558881366e-06, + "loss": 2.99, + "step": 9131 + }, + { + "epoch": 1.99, + "learning_rate": 4.007865414026656e-06, + "loss": 1.8555, + "step": 9132 + }, + { + "epoch": 1.99, + "learning_rate": 4.007756172165175e-06, + "loss": 2.7624, + "step": 9133 + }, + { + "epoch": 1.99, + "learning_rate": 4.007646930303693e-06, + "loss": 2.7232, + "step": 9134 + }, + { + "epoch": 1.99, + "learning_rate": 4.007537688442211e-06, + "loss": 2.8417, + "step": 9135 + }, + { + "epoch": 1.99, + "learning_rate": 4.00742844658073e-06, + "loss": 1.9048, + "step": 9136 + }, + { + "epoch": 1.99, + "learning_rate": 4.007319204719249e-06, + "loss": 2.5938, + "step": 9137 + }, + { + "epoch": 1.99, + "learning_rate": 4.007209962857767e-06, + "loss": 2.8706, + "step": 9138 + }, + { + "epoch": 1.99, + "learning_rate": 4.007100720996286e-06, + "loss": 1.853, + "step": 9139 + }, + { + "epoch": 1.99, + "learning_rate": 4.0069914791348045e-06, + "loss": 2.191, + "step": 9140 + }, + { + "epoch": 1.99, + "learning_rate": 4.0068822372733235e-06, + "loss": 2.5434, + "step": 9141 + }, + { + "epoch": 2.0, + "learning_rate": 4.0067729954118425e-06, + "loss": 2.2384, + "step": 9142 + }, + { + "epoch": 2.0, + "learning_rate": 4.006663753550361e-06, + "loss": 2.5669, + "step": 9143 + }, + { + "epoch": 2.0, + "learning_rate": 4.006554511688879e-06, + "loss": 2.9563, + "step": 9144 + }, + { + "epoch": 2.0, + "learning_rate": 4.006445269827398e-06, + "loss": 3.476, + "step": 9145 + }, + { + "epoch": 2.0, + "learning_rate": 4.006336027965917e-06, + "loss": 2.2611, + "step": 9146 + }, + { + "epoch": 2.0, + "learning_rate": 4.006226786104436e-06, + "loss": 3.2256, + "step": 9147 + }, + { + "epoch": 2.0, + "learning_rate": 4.006117544242954e-06, + "loss": 3.5914, + "step": 9148 + }, + { + "epoch": 2.0, + "learning_rate": 4.006008302381472e-06, + "loss": 2.8331, + "step": 9149 + }, + { + "epoch": 2.0, + "learning_rate": 4.0058990605199914e-06, + "loss": 2.3881, + "step": 9150 + }, + { + "epoch": 2.0, + "learning_rate": 4.0057898186585105e-06, + "loss": 3.4782, + "step": 9151 + }, + { + "epoch": 2.0, + "learning_rate": 4.0056805767970295e-06, + "loss": 2.9271, + "step": 9152 + }, + { + "epoch": 2.0, + "learning_rate": 4.005571334935548e-06, + "loss": 1.7706, + "step": 9153 + }, + { + "epoch": 2.0, + "learning_rate": 4.005462093074066e-06, + "loss": 2.1123, + "step": 9154 + }, + { + "epoch": 2.0, + "learning_rate": 4.005352851212585e-06, + "loss": 2.7775, + "step": 9155 + }, + { + "epoch": 2.0, + "learning_rate": 4.005243609351104e-06, + "loss": 3.2979, + "step": 9156 + }, + { + "epoch": 2.0, + "learning_rate": 4.005134367489623e-06, + "loss": 1.9374, + "step": 9157 + }, + { + "epoch": 2.0, + "learning_rate": 4.005025125628141e-06, + "loss": 2.8939, + "step": 9158 + }, + { + "epoch": 2.0, + "learning_rate": 4.004915883766659e-06, + "loss": 3.04, + "step": 9159 + }, + { + "epoch": 2.0, + "learning_rate": 4.004806641905178e-06, + "loss": 3.1028, + "step": 9160 + }, + { + "epoch": 2.0, + "learning_rate": 4.0046974000436974e-06, + "loss": 2.2819, + "step": 9161 + }, + { + "epoch": 2.0, + "learning_rate": 4.004588158182216e-06, + "loss": 1.9844, + "step": 9162 + }, + { + "epoch": 2.0, + "learning_rate": 4.004478916320735e-06, + "loss": 2.2599, + "step": 9163 + }, + { + "epoch": 2.0, + "learning_rate": 4.004369674459253e-06, + "loss": 1.4822, + "step": 9164 + }, + { + "epoch": 2.0, + "eval_accuracy": 0.3690671031096563, + "eval_loss": 2.4655795097351074, + "eval_runtime": 333.0975, + "eval_samples_per_second": 18.343, + "eval_steps_per_second": 4.587, + "step": 9164 + }, + { + "epoch": 2.0, + "learning_rate": 4.004260432597772e-06, + "loss": 1.5968, + "step": 9165 + }, + { + "epoch": 2.0, + "learning_rate": 4.004151190736291e-06, + "loss": 1.8661, + "step": 9166 + }, + { + "epoch": 2.0, + "learning_rate": 4.004041948874809e-06, + "loss": 3.2146, + "step": 9167 + }, + { + "epoch": 2.0, + "learning_rate": 4.003932707013327e-06, + "loss": 2.3943, + "step": 9168 + }, + { + "epoch": 2.0, + "learning_rate": 4.003823465151846e-06, + "loss": 3.7014, + "step": 9169 + }, + { + "epoch": 2.0, + "learning_rate": 4.003714223290365e-06, + "loss": 2.6331, + "step": 9170 + }, + { + "epoch": 2.0, + "learning_rate": 4.003604981428884e-06, + "loss": 3.0376, + "step": 9171 + }, + { + "epoch": 2.0, + "learning_rate": 4.0034957395674026e-06, + "loss": 2.7673, + "step": 9172 + }, + { + "epoch": 2.0, + "learning_rate": 4.003386497705921e-06, + "loss": 1.9538, + "step": 9173 + }, + { + "epoch": 2.0, + "learning_rate": 4.00327725584444e-06, + "loss": 2.6108, + "step": 9174 + }, + { + "epoch": 2.0, + "learning_rate": 4.003168013982959e-06, + "loss": 3.6352, + "step": 9175 + }, + { + "epoch": 2.0, + "learning_rate": 4.003058772121478e-06, + "loss": 1.7184, + "step": 9176 + }, + { + "epoch": 2.0, + "learning_rate": 4.002949530259996e-06, + "loss": 3.1339, + "step": 9177 + }, + { + "epoch": 2.0, + "learning_rate": 4.002840288398514e-06, + "loss": 2.9093, + "step": 9178 + }, + { + "epoch": 2.0, + "learning_rate": 4.002731046537033e-06, + "loss": 3.1657, + "step": 9179 + }, + { + "epoch": 2.0, + "learning_rate": 4.002621804675552e-06, + "loss": 1.6846, + "step": 9180 + }, + { + "epoch": 2.0, + "learning_rate": 4.0025125628140705e-06, + "loss": 2.3977, + "step": 9181 + }, + { + "epoch": 2.0, + "learning_rate": 4.0024033209525895e-06, + "loss": 2.145, + "step": 9182 + }, + { + "epoch": 2.0, + "learning_rate": 4.002294079091108e-06, + "loss": 3.5559, + "step": 9183 + }, + { + "epoch": 2.0, + "learning_rate": 4.002184837229627e-06, + "loss": 2.2074, + "step": 9184 + }, + { + "epoch": 2.0, + "learning_rate": 4.002075595368146e-06, + "loss": 2.4978, + "step": 9185 + }, + { + "epoch": 2.0, + "learning_rate": 4.001966353506664e-06, + "loss": 2.6971, + "step": 9186 + }, + { + "epoch": 2.01, + "learning_rate": 4.001857111645183e-06, + "loss": 2.915, + "step": 9187 + }, + { + "epoch": 2.01, + "learning_rate": 4.001747869783701e-06, + "loss": 2.0748, + "step": 9188 + }, + { + "epoch": 2.01, + "learning_rate": 4.00163862792222e-06, + "loss": 1.7863, + "step": 9189 + }, + { + "epoch": 2.01, + "learning_rate": 4.001529386060739e-06, + "loss": 3.4386, + "step": 9190 + }, + { + "epoch": 2.01, + "learning_rate": 4.0014201441992575e-06, + "loss": 3.5191, + "step": 9191 + }, + { + "epoch": 2.01, + "learning_rate": 4.0013109023377765e-06, + "loss": 2.4052, + "step": 9192 + }, + { + "epoch": 2.01, + "learning_rate": 4.001201660476295e-06, + "loss": 3.1793, + "step": 9193 + }, + { + "epoch": 2.01, + "learning_rate": 4.001092418614814e-06, + "loss": 2.7595, + "step": 9194 + }, + { + "epoch": 2.01, + "learning_rate": 4.000983176753333e-06, + "loss": 0.8743, + "step": 9195 + }, + { + "epoch": 2.01, + "learning_rate": 4.000873934891851e-06, + "loss": 2.9744, + "step": 9196 + }, + { + "epoch": 2.01, + "learning_rate": 4.000764693030369e-06, + "loss": 3.4106, + "step": 9197 + }, + { + "epoch": 2.01, + "learning_rate": 4.000655451168888e-06, + "loss": 1.7433, + "step": 9198 + }, + { + "epoch": 2.01, + "learning_rate": 4.000546209307407e-06, + "loss": 2.5775, + "step": 9199 + }, + { + "epoch": 2.01, + "learning_rate": 4.000436967445925e-06, + "loss": 2.7191, + "step": 9200 + }, + { + "epoch": 2.01, + "learning_rate": 4.000327725584444e-06, + "loss": 2.1294, + "step": 9201 + }, + { + "epoch": 2.01, + "learning_rate": 4.000218483722963e-06, + "loss": 2.9673, + "step": 9202 + }, + { + "epoch": 2.01, + "learning_rate": 4.000109241861482e-06, + "loss": 2.3176, + "step": 9203 + }, + { + "epoch": 2.01, + "learning_rate": 4.000000000000001e-06, + "loss": 3.2121, + "step": 9204 + }, + { + "epoch": 2.01, + "learning_rate": 3.999890758138519e-06, + "loss": 1.7308, + "step": 9205 + }, + { + "epoch": 2.01, + "learning_rate": 3.999781516277038e-06, + "loss": 2.2758, + "step": 9206 + }, + { + "epoch": 2.01, + "learning_rate": 3.999672274415556e-06, + "loss": 3.4442, + "step": 9207 + }, + { + "epoch": 2.01, + "learning_rate": 3.999563032554075e-06, + "loss": 3.338, + "step": 9208 + }, + { + "epoch": 2.01, + "learning_rate": 3.999453790692594e-06, + "loss": 2.2682, + "step": 9209 + }, + { + "epoch": 2.01, + "learning_rate": 3.999344548831112e-06, + "loss": 2.4326, + "step": 9210 + }, + { + "epoch": 2.01, + "learning_rate": 3.999235306969631e-06, + "loss": 2.9746, + "step": 9211 + }, + { + "epoch": 2.01, + "learning_rate": 3.9991260651081496e-06, + "loss": 2.5371, + "step": 9212 + }, + { + "epoch": 2.01, + "learning_rate": 3.999016823246669e-06, + "loss": 1.9332, + "step": 9213 + }, + { + "epoch": 2.01, + "learning_rate": 3.998907581385187e-06, + "loss": 3.5125, + "step": 9214 + }, + { + "epoch": 2.01, + "learning_rate": 3.998798339523706e-06, + "loss": 4.2734, + "step": 9215 + }, + { + "epoch": 2.01, + "learning_rate": 3.998689097662225e-06, + "loss": 3.6664, + "step": 9216 + }, + { + "epoch": 2.01, + "learning_rate": 3.998579855800743e-06, + "loss": 3.0637, + "step": 9217 + }, + { + "epoch": 2.01, + "learning_rate": 3.998470613939262e-06, + "loss": 2.8946, + "step": 9218 + }, + { + "epoch": 2.01, + "learning_rate": 3.99836137207778e-06, + "loss": 2.8336, + "step": 9219 + }, + { + "epoch": 2.01, + "learning_rate": 3.998252130216299e-06, + "loss": 2.3837, + "step": 9220 + }, + { + "epoch": 2.01, + "learning_rate": 3.9981428883548175e-06, + "loss": 2.0477, + "step": 9221 + }, + { + "epoch": 2.01, + "learning_rate": 3.9980336464933365e-06, + "loss": 2.9837, + "step": 9222 + }, + { + "epoch": 2.01, + "learning_rate": 3.9979244046318555e-06, + "loss": 2.9861, + "step": 9223 + }, + { + "epoch": 2.01, + "learning_rate": 3.997815162770374e-06, + "loss": 2.8716, + "step": 9224 + }, + { + "epoch": 2.01, + "learning_rate": 3.997705920908893e-06, + "loss": 3.1622, + "step": 9225 + }, + { + "epoch": 2.01, + "learning_rate": 3.997596679047411e-06, + "loss": 3.0063, + "step": 9226 + }, + { + "epoch": 2.01, + "learning_rate": 3.99748743718593e-06, + "loss": 2.8383, + "step": 9227 + }, + { + "epoch": 2.01, + "learning_rate": 3.997378195324449e-06, + "loss": 2.7573, + "step": 9228 + }, + { + "epoch": 2.01, + "learning_rate": 3.997268953462967e-06, + "loss": 3.004, + "step": 9229 + }, + { + "epoch": 2.01, + "learning_rate": 3.997159711601486e-06, + "loss": 2.0947, + "step": 9230 + }, + { + "epoch": 2.01, + "learning_rate": 3.9970504697400044e-06, + "loss": 3.0789, + "step": 9231 + }, + { + "epoch": 2.01, + "learning_rate": 3.9969412278785235e-06, + "loss": 3.5841, + "step": 9232 + }, + { + "epoch": 2.02, + "learning_rate": 3.996831986017042e-06, + "loss": 1.8924, + "step": 9233 + }, + { + "epoch": 2.02, + "learning_rate": 3.996722744155561e-06, + "loss": 2.1422, + "step": 9234 + }, + { + "epoch": 2.02, + "learning_rate": 3.99661350229408e-06, + "loss": 3.0755, + "step": 9235 + }, + { + "epoch": 2.02, + "learning_rate": 3.996504260432598e-06, + "loss": 3.0948, + "step": 9236 + }, + { + "epoch": 2.02, + "learning_rate": 3.996395018571117e-06, + "loss": 2.3772, + "step": 9237 + }, + { + "epoch": 2.02, + "learning_rate": 3.996285776709635e-06, + "loss": 2.6275, + "step": 9238 + }, + { + "epoch": 2.02, + "learning_rate": 3.996176534848154e-06, + "loss": 2.983, + "step": 9239 + }, + { + "epoch": 2.02, + "learning_rate": 3.996067292986673e-06, + "loss": 3.9402, + "step": 9240 + }, + { + "epoch": 2.02, + "learning_rate": 3.995958051125191e-06, + "loss": 3.0926, + "step": 9241 + }, + { + "epoch": 2.02, + "learning_rate": 3.9958488092637104e-06, + "loss": 3.1235, + "step": 9242 + }, + { + "epoch": 2.02, + "learning_rate": 3.995739567402229e-06, + "loss": 1.5842, + "step": 9243 + }, + { + "epoch": 2.02, + "learning_rate": 3.995630325540748e-06, + "loss": 2.5863, + "step": 9244 + }, + { + "epoch": 2.02, + "learning_rate": 3.995521083679267e-06, + "loss": 2.3523, + "step": 9245 + }, + { + "epoch": 2.02, + "learning_rate": 3.995411841817785e-06, + "loss": 2.9491, + "step": 9246 + }, + { + "epoch": 2.02, + "learning_rate": 3.995302599956303e-06, + "loss": 3.462, + "step": 9247 + }, + { + "epoch": 2.02, + "learning_rate": 3.995193358094822e-06, + "loss": 2.5331, + "step": 9248 + }, + { + "epoch": 2.02, + "learning_rate": 3.995084116233341e-06, + "loss": 2.454, + "step": 9249 + }, + { + "epoch": 2.02, + "learning_rate": 3.994974874371859e-06, + "loss": 2.2596, + "step": 9250 + }, + { + "epoch": 2.02, + "learning_rate": 3.994865632510378e-06, + "loss": 2.656, + "step": 9251 + }, + { + "epoch": 2.02, + "learning_rate": 3.9947563906488965e-06, + "loss": 2.9505, + "step": 9252 + }, + { + "epoch": 2.02, + "learning_rate": 3.9946471487874156e-06, + "loss": 2.7544, + "step": 9253 + }, + { + "epoch": 2.02, + "learning_rate": 3.994537906925935e-06, + "loss": 2.5114, + "step": 9254 + }, + { + "epoch": 2.02, + "learning_rate": 3.994428665064453e-06, + "loss": 2.9969, + "step": 9255 + }, + { + "epoch": 2.02, + "learning_rate": 3.994319423202972e-06, + "loss": 1.8785, + "step": 9256 + }, + { + "epoch": 2.02, + "learning_rate": 3.99421018134149e-06, + "loss": 3.5708, + "step": 9257 + }, + { + "epoch": 2.02, + "learning_rate": 3.994100939480009e-06, + "loss": 2.8698, + "step": 9258 + }, + { + "epoch": 2.02, + "learning_rate": 3.993991697618528e-06, + "loss": 3.0878, + "step": 9259 + }, + { + "epoch": 2.02, + "learning_rate": 3.993882455757046e-06, + "loss": 2.5244, + "step": 9260 + }, + { + "epoch": 2.02, + "learning_rate": 3.993773213895565e-06, + "loss": 2.7583, + "step": 9261 + }, + { + "epoch": 2.02, + "learning_rate": 3.9936639720340835e-06, + "loss": 1.9826, + "step": 9262 + }, + { + "epoch": 2.02, + "learning_rate": 3.9935547301726025e-06, + "loss": 2.6031, + "step": 9263 + }, + { + "epoch": 2.02, + "learning_rate": 3.9934454883111216e-06, + "loss": 2.6758, + "step": 9264 + }, + { + "epoch": 2.02, + "learning_rate": 3.99333624644964e-06, + "loss": 2.8724, + "step": 9265 + }, + { + "epoch": 2.02, + "learning_rate": 3.993227004588158e-06, + "loss": 2.4727, + "step": 9266 + }, + { + "epoch": 2.02, + "learning_rate": 3.993117762726677e-06, + "loss": 3.4911, + "step": 9267 + }, + { + "epoch": 2.02, + "learning_rate": 3.993008520865196e-06, + "loss": 3.5181, + "step": 9268 + }, + { + "epoch": 2.02, + "learning_rate": 3.992899279003715e-06, + "loss": 2.5892, + "step": 9269 + }, + { + "epoch": 2.02, + "learning_rate": 3.992790037142233e-06, + "loss": 1.9638, + "step": 9270 + }, + { + "epoch": 2.02, + "learning_rate": 3.992680795280751e-06, + "loss": 1.2439, + "step": 9271 + }, + { + "epoch": 2.02, + "learning_rate": 3.9925715534192704e-06, + "loss": 2.5112, + "step": 9272 + }, + { + "epoch": 2.02, + "learning_rate": 3.9924623115577895e-06, + "loss": 2.264, + "step": 9273 + }, + { + "epoch": 2.02, + "learning_rate": 3.992353069696308e-06, + "loss": 2.8771, + "step": 9274 + }, + { + "epoch": 2.02, + "learning_rate": 3.992243827834827e-06, + "loss": 1.8477, + "step": 9275 + }, + { + "epoch": 2.02, + "learning_rate": 3.992134585973345e-06, + "loss": 2.7476, + "step": 9276 + }, + { + "epoch": 2.02, + "learning_rate": 3.992025344111864e-06, + "loss": 3.2777, + "step": 9277 + }, + { + "epoch": 2.02, + "learning_rate": 3.991916102250383e-06, + "loss": 2.4006, + "step": 9278 + }, + { + "epoch": 2.03, + "learning_rate": 3.991806860388901e-06, + "loss": 3.1267, + "step": 9279 + }, + { + "epoch": 2.03, + "learning_rate": 3.99169761852742e-06, + "loss": 2.2779, + "step": 9280 + }, + { + "epoch": 2.03, + "learning_rate": 3.991588376665938e-06, + "loss": 2.5249, + "step": 9281 + }, + { + "epoch": 2.03, + "learning_rate": 3.991479134804457e-06, + "loss": 2.8273, + "step": 9282 + }, + { + "epoch": 2.03, + "learning_rate": 3.9913698929429764e-06, + "loss": 2.5866, + "step": 9283 + }, + { + "epoch": 2.03, + "learning_rate": 3.991260651081495e-06, + "loss": 3.7455, + "step": 9284 + }, + { + "epoch": 2.03, + "learning_rate": 3.991151409220013e-06, + "loss": 2.1835, + "step": 9285 + }, + { + "epoch": 2.03, + "learning_rate": 3.991042167358532e-06, + "loss": 1.9635, + "step": 9286 + }, + { + "epoch": 2.03, + "learning_rate": 3.990932925497051e-06, + "loss": 2.1628, + "step": 9287 + }, + { + "epoch": 2.03, + "learning_rate": 3.99082368363557e-06, + "loss": 2.1596, + "step": 9288 + }, + { + "epoch": 2.03, + "learning_rate": 3.990714441774088e-06, + "loss": 1.9363, + "step": 9289 + }, + { + "epoch": 2.03, + "learning_rate": 3.990605199912606e-06, + "loss": 2.3152, + "step": 9290 + }, + { + "epoch": 2.03, + "learning_rate": 3.990495958051125e-06, + "loss": 2.9624, + "step": 9291 + }, + { + "epoch": 2.03, + "learning_rate": 3.990386716189644e-06, + "loss": 3.2232, + "step": 9292 + }, + { + "epoch": 2.03, + "learning_rate": 3.990277474328163e-06, + "loss": 1.3591, + "step": 9293 + }, + { + "epoch": 2.03, + "learning_rate": 3.990168232466682e-06, + "loss": 3.0095, + "step": 9294 + }, + { + "epoch": 2.03, + "learning_rate": 3.9900589906052e-06, + "loss": 2.9803, + "step": 9295 + }, + { + "epoch": 2.03, + "learning_rate": 3.989949748743719e-06, + "loss": 2.2612, + "step": 9296 + }, + { + "epoch": 2.03, + "learning_rate": 3.989840506882238e-06, + "loss": 3.3336, + "step": 9297 + }, + { + "epoch": 2.03, + "learning_rate": 3.989731265020757e-06, + "loss": 2.9917, + "step": 9298 + }, + { + "epoch": 2.03, + "learning_rate": 3.989622023159275e-06, + "loss": 3.199, + "step": 9299 + }, + { + "epoch": 2.03, + "learning_rate": 3.989512781297793e-06, + "loss": 2.2799, + "step": 9300 + }, + { + "epoch": 2.03, + "learning_rate": 3.989403539436312e-06, + "loss": 1.6956, + "step": 9301 + }, + { + "epoch": 2.03, + "learning_rate": 3.989294297574831e-06, + "loss": 3.2864, + "step": 9302 + }, + { + "epoch": 2.03, + "learning_rate": 3.9891850557133495e-06, + "loss": 1.9667, + "step": 9303 + }, + { + "epoch": 2.03, + "learning_rate": 3.9890758138518685e-06, + "loss": 2.725, + "step": 9304 + }, + { + "epoch": 2.03, + "learning_rate": 3.988966571990387e-06, + "loss": 2.0351, + "step": 9305 + }, + { + "epoch": 2.03, + "learning_rate": 3.988857330128906e-06, + "loss": 3.0813, + "step": 9306 + }, + { + "epoch": 2.03, + "learning_rate": 3.988748088267425e-06, + "loss": 2.5946, + "step": 9307 + }, + { + "epoch": 2.03, + "learning_rate": 3.988638846405943e-06, + "loss": 2.7201, + "step": 9308 + }, + { + "epoch": 2.03, + "learning_rate": 3.988529604544461e-06, + "loss": 3.0417, + "step": 9309 + }, + { + "epoch": 2.03, + "learning_rate": 3.98842036268298e-06, + "loss": 2.8351, + "step": 9310 + }, + { + "epoch": 2.03, + "learning_rate": 3.988311120821499e-06, + "loss": 1.9328, + "step": 9311 + }, + { + "epoch": 2.03, + "learning_rate": 3.988201878960018e-06, + "loss": 1.2055, + "step": 9312 + }, + { + "epoch": 2.03, + "learning_rate": 3.9880926370985365e-06, + "loss": 2.1886, + "step": 9313 + }, + { + "epoch": 2.03, + "learning_rate": 3.987983395237055e-06, + "loss": 3.2469, + "step": 9314 + }, + { + "epoch": 2.03, + "learning_rate": 3.987874153375574e-06, + "loss": 2.8964, + "step": 9315 + }, + { + "epoch": 2.03, + "learning_rate": 3.987764911514093e-06, + "loss": 3.5665, + "step": 9316 + }, + { + "epoch": 2.03, + "learning_rate": 3.987655669652612e-06, + "loss": 3.241, + "step": 9317 + }, + { + "epoch": 2.03, + "learning_rate": 3.98754642779113e-06, + "loss": 2.8201, + "step": 9318 + }, + { + "epoch": 2.03, + "learning_rate": 3.987437185929648e-06, + "loss": 3.0083, + "step": 9319 + }, + { + "epoch": 2.03, + "learning_rate": 3.987327944068167e-06, + "loss": 3.4004, + "step": 9320 + }, + { + "epoch": 2.03, + "learning_rate": 3.987218702206686e-06, + "loss": 1.159, + "step": 9321 + }, + { + "epoch": 2.03, + "learning_rate": 3.987109460345205e-06, + "loss": 2.0474, + "step": 9322 + }, + { + "epoch": 2.03, + "learning_rate": 3.987000218483723e-06, + "loss": 1.8033, + "step": 9323 + }, + { + "epoch": 2.03, + "learning_rate": 3.986890976622242e-06, + "loss": 2.4245, + "step": 9324 + }, + { + "epoch": 2.04, + "learning_rate": 3.986781734760761e-06, + "loss": 2.8732, + "step": 9325 + }, + { + "epoch": 2.04, + "learning_rate": 3.98667249289928e-06, + "loss": 2.4406, + "step": 9326 + }, + { + "epoch": 2.04, + "learning_rate": 3.986563251037798e-06, + "loss": 2.4516, + "step": 9327 + }, + { + "epoch": 2.04, + "learning_rate": 3.986454009176317e-06, + "loss": 1.5679, + "step": 9328 + }, + { + "epoch": 2.04, + "learning_rate": 3.986344767314835e-06, + "loss": 1.4543, + "step": 9329 + }, + { + "epoch": 2.04, + "learning_rate": 3.986235525453354e-06, + "loss": 2.2764, + "step": 9330 + }, + { + "epoch": 2.04, + "learning_rate": 3.986126283591873e-06, + "loss": 1.7925, + "step": 9331 + }, + { + "epoch": 2.04, + "learning_rate": 3.986017041730391e-06, + "loss": 2.3532, + "step": 9332 + }, + { + "epoch": 2.04, + "learning_rate": 3.98590779986891e-06, + "loss": 2.2114, + "step": 9333 + }, + { + "epoch": 2.04, + "learning_rate": 3.9857985580074286e-06, + "loss": 1.0076, + "step": 9334 + }, + { + "epoch": 2.04, + "learning_rate": 3.985689316145948e-06, + "loss": 3.2123, + "step": 9335 + }, + { + "epoch": 2.04, + "learning_rate": 3.985580074284467e-06, + "loss": 2.439, + "step": 9336 + }, + { + "epoch": 2.04, + "learning_rate": 3.985470832422985e-06, + "loss": 1.5677, + "step": 9337 + }, + { + "epoch": 2.04, + "learning_rate": 3.985361590561503e-06, + "loss": 2.2531, + "step": 9338 + }, + { + "epoch": 2.04, + "learning_rate": 3.985252348700022e-06, + "loss": 1.4808, + "step": 9339 + }, + { + "epoch": 2.04, + "learning_rate": 3.985143106838541e-06, + "loss": 3.9172, + "step": 9340 + }, + { + "epoch": 2.04, + "learning_rate": 3.98503386497706e-06, + "loss": 2.6736, + "step": 9341 + }, + { + "epoch": 2.04, + "learning_rate": 3.984924623115578e-06, + "loss": 2.6861, + "step": 9342 + }, + { + "epoch": 2.04, + "learning_rate": 3.9848153812540965e-06, + "loss": 2.7412, + "step": 9343 + }, + { + "epoch": 2.04, + "learning_rate": 3.9847061393926155e-06, + "loss": 2.608, + "step": 9344 + }, + { + "epoch": 2.04, + "learning_rate": 3.9845968975311346e-06, + "loss": 1.3563, + "step": 9345 + }, + { + "epoch": 2.04, + "learning_rate": 3.984487655669653e-06, + "loss": 3.8124, + "step": 9346 + }, + { + "epoch": 2.04, + "learning_rate": 3.984378413808172e-06, + "loss": 2.101, + "step": 9347 + }, + { + "epoch": 2.04, + "learning_rate": 3.98426917194669e-06, + "loss": 2.212, + "step": 9348 + }, + { + "epoch": 2.04, + "learning_rate": 3.984159930085209e-06, + "loss": 3.2153, + "step": 9349 + }, + { + "epoch": 2.04, + "learning_rate": 3.984050688223728e-06, + "loss": 1.5222, + "step": 9350 + }, + { + "epoch": 2.04, + "learning_rate": 3.983941446362246e-06, + "loss": 2.3506, + "step": 9351 + }, + { + "epoch": 2.04, + "learning_rate": 3.983832204500765e-06, + "loss": 2.221, + "step": 9352 + }, + { + "epoch": 2.04, + "learning_rate": 3.9837229626392834e-06, + "loss": 3.2776, + "step": 9353 + }, + { + "epoch": 2.04, + "learning_rate": 3.9836137207778025e-06, + "loss": 1.8209, + "step": 9354 + }, + { + "epoch": 2.04, + "learning_rate": 3.9835044789163215e-06, + "loss": 2.5383, + "step": 9355 + }, + { + "epoch": 2.04, + "learning_rate": 3.98339523705484e-06, + "loss": 3.0322, + "step": 9356 + }, + { + "epoch": 2.04, + "learning_rate": 3.983285995193359e-06, + "loss": 3.626, + "step": 9357 + }, + { + "epoch": 2.04, + "learning_rate": 3.983176753331877e-06, + "loss": 1.9695, + "step": 9358 + }, + { + "epoch": 2.04, + "learning_rate": 3.983067511470396e-06, + "loss": 3.0321, + "step": 9359 + }, + { + "epoch": 2.04, + "learning_rate": 3.982958269608914e-06, + "loss": 1.6866, + "step": 9360 + }, + { + "epoch": 2.04, + "learning_rate": 3.982849027747433e-06, + "loss": 1.8257, + "step": 9361 + }, + { + "epoch": 2.04, + "learning_rate": 3.982739785885951e-06, + "loss": 2.3488, + "step": 9362 + }, + { + "epoch": 2.04, + "learning_rate": 3.98263054402447e-06, + "loss": 2.5192, + "step": 9363 + }, + { + "epoch": 2.04, + "learning_rate": 3.9825213021629894e-06, + "loss": 2.1629, + "step": 9364 + }, + { + "epoch": 2.04, + "learning_rate": 3.982412060301508e-06, + "loss": 1.9982, + "step": 9365 + }, + { + "epoch": 2.04, + "learning_rate": 3.982302818440027e-06, + "loss": 2.1241, + "step": 9366 + }, + { + "epoch": 2.04, + "learning_rate": 3.982193576578545e-06, + "loss": 2.8451, + "step": 9367 + }, + { + "epoch": 2.04, + "learning_rate": 3.982084334717064e-06, + "loss": 1.4363, + "step": 9368 + }, + { + "epoch": 2.04, + "learning_rate": 3.981975092855583e-06, + "loss": 3.6878, + "step": 9369 + }, + { + "epoch": 2.04, + "learning_rate": 3.981865850994101e-06, + "loss": 2.7946, + "step": 9370 + }, + { + "epoch": 2.05, + "learning_rate": 3.98175660913262e-06, + "loss": 4.0194, + "step": 9371 + }, + { + "epoch": 2.05, + "learning_rate": 3.981647367271138e-06, + "loss": 2.8458, + "step": 9372 + }, + { + "epoch": 2.05, + "learning_rate": 3.981538125409657e-06, + "loss": 1.5089, + "step": 9373 + }, + { + "epoch": 2.05, + "learning_rate": 3.981428883548176e-06, + "loss": 2.7411, + "step": 9374 + }, + { + "epoch": 2.05, + "learning_rate": 3.981319641686695e-06, + "loss": 1.8079, + "step": 9375 + }, + { + "epoch": 2.05, + "learning_rate": 3.981210399825214e-06, + "loss": 2.4085, + "step": 9376 + }, + { + "epoch": 2.05, + "learning_rate": 3.981101157963732e-06, + "loss": 2.5781, + "step": 9377 + }, + { + "epoch": 2.05, + "learning_rate": 3.980991916102251e-06, + "loss": 2.7255, + "step": 9378 + }, + { + "epoch": 2.05, + "learning_rate": 3.980882674240769e-06, + "loss": 2.9666, + "step": 9379 + }, + { + "epoch": 2.05, + "learning_rate": 3.980773432379288e-06, + "loss": 3.2951, + "step": 9380 + }, + { + "epoch": 2.05, + "learning_rate": 3.980664190517807e-06, + "loss": 2.0943, + "step": 9381 + }, + { + "epoch": 2.05, + "learning_rate": 3.980554948656325e-06, + "loss": 3.7273, + "step": 9382 + }, + { + "epoch": 2.05, + "learning_rate": 3.980445706794844e-06, + "loss": 2.2073, + "step": 9383 + }, + { + "epoch": 2.05, + "learning_rate": 3.9803364649333625e-06, + "loss": 1.6242, + "step": 9384 + }, + { + "epoch": 2.05, + "learning_rate": 3.9802272230718815e-06, + "loss": 2.1888, + "step": 9385 + }, + { + "epoch": 2.05, + "learning_rate": 3.9801179812104006e-06, + "loss": 1.6655, + "step": 9386 + }, + { + "epoch": 2.05, + "learning_rate": 3.980008739348919e-06, + "loss": 2.4416, + "step": 9387 + }, + { + "epoch": 2.05, + "learning_rate": 3.979899497487438e-06, + "loss": 3.4762, + "step": 9388 + }, + { + "epoch": 2.05, + "learning_rate": 3.979790255625956e-06, + "loss": 3.0378, + "step": 9389 + }, + { + "epoch": 2.05, + "learning_rate": 3.979681013764475e-06, + "loss": 3.2733, + "step": 9390 + }, + { + "epoch": 2.05, + "learning_rate": 3.979571771902993e-06, + "loss": 3.0027, + "step": 9391 + }, + { + "epoch": 2.05, + "learning_rate": 3.979462530041512e-06, + "loss": 2.7054, + "step": 9392 + }, + { + "epoch": 2.05, + "learning_rate": 3.9793532881800304e-06, + "loss": 2.4088, + "step": 9393 + }, + { + "epoch": 2.05, + "learning_rate": 3.9792440463185495e-06, + "loss": 2.8745, + "step": 9394 + }, + { + "epoch": 2.05, + "learning_rate": 3.9791348044570685e-06, + "loss": 2.626, + "step": 9395 + }, + { + "epoch": 2.05, + "learning_rate": 3.979025562595587e-06, + "loss": 2.0877, + "step": 9396 + }, + { + "epoch": 2.05, + "learning_rate": 3.978916320734106e-06, + "loss": 2.7106, + "step": 9397 + }, + { + "epoch": 2.05, + "learning_rate": 3.978807078872624e-06, + "loss": 3.3908, + "step": 9398 + }, + { + "epoch": 2.05, + "learning_rate": 3.978697837011143e-06, + "loss": 2.5821, + "step": 9399 + }, + { + "epoch": 2.05, + "learning_rate": 3.978588595149662e-06, + "loss": 1.7105, + "step": 9400 + }, + { + "epoch": 2.05, + "learning_rate": 3.97847935328818e-06, + "loss": 2.9444, + "step": 9401 + }, + { + "epoch": 2.05, + "learning_rate": 3.978370111426699e-06, + "loss": 3.2227, + "step": 9402 + }, + { + "epoch": 2.05, + "learning_rate": 3.978260869565217e-06, + "loss": 2.319, + "step": 9403 + }, + { + "epoch": 2.05, + "learning_rate": 3.978151627703736e-06, + "loss": 2.2487, + "step": 9404 + }, + { + "epoch": 2.05, + "learning_rate": 3.9780423858422555e-06, + "loss": 1.8785, + "step": 9405 + }, + { + "epoch": 2.05, + "learning_rate": 3.977933143980774e-06, + "loss": 2.3598, + "step": 9406 + }, + { + "epoch": 2.05, + "learning_rate": 3.977823902119293e-06, + "loss": 2.6016, + "step": 9407 + }, + { + "epoch": 2.05, + "learning_rate": 3.977714660257811e-06, + "loss": 3.1071, + "step": 9408 + }, + { + "epoch": 2.05, + "learning_rate": 3.97760541839633e-06, + "loss": 2.3872, + "step": 9409 + }, + { + "epoch": 2.05, + "learning_rate": 3.977496176534849e-06, + "loss": 0.936, + "step": 9410 + }, + { + "epoch": 2.05, + "learning_rate": 3.977386934673367e-06, + "loss": 2.4943, + "step": 9411 + }, + { + "epoch": 2.05, + "learning_rate": 3.977277692811885e-06, + "loss": 2.128, + "step": 9412 + }, + { + "epoch": 2.05, + "learning_rate": 3.977168450950404e-06, + "loss": 2.606, + "step": 9413 + }, + { + "epoch": 2.05, + "learning_rate": 3.977059209088923e-06, + "loss": 2.3798, + "step": 9414 + }, + { + "epoch": 2.05, + "learning_rate": 3.9769499672274416e-06, + "loss": 2.5212, + "step": 9415 + }, + { + "epoch": 2.05, + "learning_rate": 3.976840725365961e-06, + "loss": 2.1204, + "step": 9416 + }, + { + "epoch": 2.06, + "learning_rate": 3.976731483504479e-06, + "loss": 2.2827, + "step": 9417 + }, + { + "epoch": 2.06, + "learning_rate": 3.976622241642998e-06, + "loss": 2.0848, + "step": 9418 + }, + { + "epoch": 2.06, + "learning_rate": 3.976512999781517e-06, + "loss": 2.2117, + "step": 9419 + }, + { + "epoch": 2.06, + "learning_rate": 3.976403757920035e-06, + "loss": 2.8895, + "step": 9420 + }, + { + "epoch": 2.06, + "learning_rate": 3.976294516058554e-06, + "loss": 2.3425, + "step": 9421 + }, + { + "epoch": 2.06, + "learning_rate": 3.976185274197072e-06, + "loss": 3.0532, + "step": 9422 + }, + { + "epoch": 2.06, + "learning_rate": 3.976076032335591e-06, + "loss": 1.7675, + "step": 9423 + }, + { + "epoch": 2.06, + "learning_rate": 3.97596679047411e-06, + "loss": 1.9412, + "step": 9424 + }, + { + "epoch": 2.06, + "learning_rate": 3.9758575486126285e-06, + "loss": 2.1112, + "step": 9425 + }, + { + "epoch": 2.06, + "learning_rate": 3.975748306751147e-06, + "loss": 3.1769, + "step": 9426 + }, + { + "epoch": 2.06, + "learning_rate": 3.975639064889666e-06, + "loss": 3.3341, + "step": 9427 + }, + { + "epoch": 2.06, + "learning_rate": 3.975529823028185e-06, + "loss": 2.7589, + "step": 9428 + }, + { + "epoch": 2.06, + "learning_rate": 3.975420581166704e-06, + "loss": 1.7441, + "step": 9429 + }, + { + "epoch": 2.06, + "learning_rate": 3.975311339305222e-06, + "loss": 2.3789, + "step": 9430 + }, + { + "epoch": 2.06, + "learning_rate": 3.97520209744374e-06, + "loss": 2.851, + "step": 9431 + }, + { + "epoch": 2.06, + "learning_rate": 3.975092855582259e-06, + "loss": 2.7811, + "step": 9432 + }, + { + "epoch": 2.06, + "learning_rate": 3.974983613720778e-06, + "loss": 2.8995, + "step": 9433 + }, + { + "epoch": 2.06, + "learning_rate": 3.974874371859297e-06, + "loss": 1.7946, + "step": 9434 + }, + { + "epoch": 2.06, + "learning_rate": 3.9747651299978155e-06, + "loss": 2.4482, + "step": 9435 + }, + { + "epoch": 2.06, + "learning_rate": 3.974655888136334e-06, + "loss": 3.9357, + "step": 9436 + }, + { + "epoch": 2.06, + "learning_rate": 3.974546646274853e-06, + "loss": 2.3663, + "step": 9437 + }, + { + "epoch": 2.06, + "learning_rate": 3.974437404413372e-06, + "loss": 1.683, + "step": 9438 + }, + { + "epoch": 2.06, + "learning_rate": 3.974328162551891e-06, + "loss": 2.5478, + "step": 9439 + }, + { + "epoch": 2.06, + "learning_rate": 3.974218920690409e-06, + "loss": 2.5213, + "step": 9440 + }, + { + "epoch": 2.06, + "learning_rate": 3.974109678828927e-06, + "loss": 3.8155, + "step": 9441 + }, + { + "epoch": 2.06, + "learning_rate": 3.974000436967446e-06, + "loss": 3.2719, + "step": 9442 + }, + { + "epoch": 2.06, + "learning_rate": 3.973891195105965e-06, + "loss": 2.4329, + "step": 9443 + }, + { + "epoch": 2.06, + "learning_rate": 3.973781953244483e-06, + "loss": 2.9688, + "step": 9444 + }, + { + "epoch": 2.06, + "learning_rate": 3.9736727113830024e-06, + "loss": 2.2118, + "step": 9445 + }, + { + "epoch": 2.06, + "learning_rate": 3.973563469521521e-06, + "loss": 2.2973, + "step": 9446 + }, + { + "epoch": 2.06, + "learning_rate": 3.97345422766004e-06, + "loss": 2.1956, + "step": 9447 + }, + { + "epoch": 2.06, + "learning_rate": 3.973344985798559e-06, + "loss": 2.1897, + "step": 9448 + }, + { + "epoch": 2.06, + "learning_rate": 3.973235743937077e-06, + "loss": 3.4063, + "step": 9449 + }, + { + "epoch": 2.06, + "learning_rate": 3.973126502075595e-06, + "loss": 3.7566, + "step": 9450 + }, + { + "epoch": 2.06, + "learning_rate": 3.973017260214114e-06, + "loss": 2.6079, + "step": 9451 + }, + { + "epoch": 2.06, + "learning_rate": 3.972908018352633e-06, + "loss": 3.0574, + "step": 9452 + }, + { + "epoch": 2.06, + "learning_rate": 3.972798776491152e-06, + "loss": 1.7571, + "step": 9453 + }, + { + "epoch": 2.06, + "learning_rate": 3.97268953462967e-06, + "loss": 1.4384, + "step": 9454 + }, + { + "epoch": 2.06, + "learning_rate": 3.9725802927681885e-06, + "loss": 2.5521, + "step": 9455 + }, + { + "epoch": 2.06, + "learning_rate": 3.9724710509067076e-06, + "loss": 3.0981, + "step": 9456 + }, + { + "epoch": 2.06, + "learning_rate": 3.972361809045227e-06, + "loss": 2.2982, + "step": 9457 + }, + { + "epoch": 2.06, + "learning_rate": 3.972252567183746e-06, + "loss": 2.2328, + "step": 9458 + }, + { + "epoch": 2.06, + "learning_rate": 3.972143325322264e-06, + "loss": 3.5577, + "step": 9459 + }, + { + "epoch": 2.06, + "learning_rate": 3.972034083460782e-06, + "loss": 1.7595, + "step": 9460 + }, + { + "epoch": 2.06, + "learning_rate": 3.971924841599301e-06, + "loss": 3.1538, + "step": 9461 + }, + { + "epoch": 2.07, + "learning_rate": 3.97181559973782e-06, + "loss": 1.7089, + "step": 9462 + }, + { + "epoch": 2.07, + "learning_rate": 3.971706357876339e-06, + "loss": 2.5426, + "step": 9463 + }, + { + "epoch": 2.07, + "learning_rate": 3.971597116014857e-06, + "loss": 2.6441, + "step": 9464 + }, + { + "epoch": 2.07, + "learning_rate": 3.9714878741533755e-06, + "loss": 1.7758, + "step": 9465 + }, + { + "epoch": 2.07, + "learning_rate": 3.9713786322918945e-06, + "loss": 2.8983, + "step": 9466 + }, + { + "epoch": 2.07, + "learning_rate": 3.9712693904304136e-06, + "loss": 1.9347, + "step": 9467 + }, + { + "epoch": 2.07, + "learning_rate": 3.971160148568932e-06, + "loss": 2.7496, + "step": 9468 + }, + { + "epoch": 2.07, + "learning_rate": 3.971050906707451e-06, + "loss": 1.511, + "step": 9469 + }, + { + "epoch": 2.07, + "learning_rate": 3.970941664845969e-06, + "loss": 2.9095, + "step": 9470 + }, + { + "epoch": 2.07, + "learning_rate": 3.970832422984488e-06, + "loss": 2.2752, + "step": 9471 + }, + { + "epoch": 2.07, + "learning_rate": 3.970723181123007e-06, + "loss": 3.1325, + "step": 9472 + }, + { + "epoch": 2.07, + "learning_rate": 3.970613939261525e-06, + "loss": 2.1594, + "step": 9473 + }, + { + "epoch": 2.07, + "learning_rate": 3.970504697400044e-06, + "loss": 2.6197, + "step": 9474 + }, + { + "epoch": 2.07, + "learning_rate": 3.9703954555385625e-06, + "loss": 2.1029, + "step": 9475 + }, + { + "epoch": 2.07, + "learning_rate": 3.9702862136770815e-06, + "loss": 1.7839, + "step": 9476 + }, + { + "epoch": 2.07, + "learning_rate": 3.9701769718156005e-06, + "loss": 2.2067, + "step": 9477 + }, + { + "epoch": 2.07, + "learning_rate": 3.970067729954119e-06, + "loss": 2.66, + "step": 9478 + }, + { + "epoch": 2.07, + "learning_rate": 3.969958488092637e-06, + "loss": 2.785, + "step": 9479 + }, + { + "epoch": 2.07, + "learning_rate": 3.969849246231156e-06, + "loss": 3.0145, + "step": 9480 + }, + { + "epoch": 2.07, + "learning_rate": 3.969740004369675e-06, + "loss": 1.5481, + "step": 9481 + }, + { + "epoch": 2.07, + "learning_rate": 3.969630762508194e-06, + "loss": 2.3032, + "step": 9482 + }, + { + "epoch": 2.07, + "learning_rate": 3.969521520646712e-06, + "loss": 2.789, + "step": 9483 + }, + { + "epoch": 2.07, + "learning_rate": 3.96941227878523e-06, + "loss": 3.9228, + "step": 9484 + }, + { + "epoch": 2.07, + "learning_rate": 3.969303036923749e-06, + "loss": 2.3994, + "step": 9485 + }, + { + "epoch": 2.07, + "learning_rate": 3.9691937950622684e-06, + "loss": 2.7997, + "step": 9486 + }, + { + "epoch": 2.07, + "learning_rate": 3.9690845532007875e-06, + "loss": 3.1047, + "step": 9487 + }, + { + "epoch": 2.07, + "learning_rate": 3.968975311339306e-06, + "loss": 2.7948, + "step": 9488 + }, + { + "epoch": 2.07, + "learning_rate": 3.968866069477824e-06, + "loss": 2.1933, + "step": 9489 + }, + { + "epoch": 2.07, + "learning_rate": 3.968756827616343e-06, + "loss": 1.7202, + "step": 9490 + }, + { + "epoch": 2.07, + "learning_rate": 3.968647585754862e-06, + "loss": 2.8454, + "step": 9491 + }, + { + "epoch": 2.07, + "learning_rate": 3.96853834389338e-06, + "loss": 3.4171, + "step": 9492 + }, + { + "epoch": 2.07, + "learning_rate": 3.968429102031899e-06, + "loss": 2.5995, + "step": 9493 + }, + { + "epoch": 2.07, + "learning_rate": 3.968319860170417e-06, + "loss": 2.0484, + "step": 9494 + }, + { + "epoch": 2.07, + "learning_rate": 3.968210618308936e-06, + "loss": 1.6854, + "step": 9495 + }, + { + "epoch": 2.07, + "learning_rate": 3.968101376447455e-06, + "loss": 2.1691, + "step": 9496 + }, + { + "epoch": 2.07, + "learning_rate": 3.967992134585974e-06, + "loss": 2.8018, + "step": 9497 + }, + { + "epoch": 2.07, + "learning_rate": 3.967882892724493e-06, + "loss": 1.6416, + "step": 9498 + }, + { + "epoch": 2.07, + "learning_rate": 3.967773650863011e-06, + "loss": 2.6833, + "step": 9499 + }, + { + "epoch": 2.07, + "learning_rate": 3.96766440900153e-06, + "loss": 1.9585, + "step": 9500 + }, + { + "epoch": 2.07, + "learning_rate": 3.967555167140049e-06, + "loss": 2.5579, + "step": 9501 + }, + { + "epoch": 2.07, + "learning_rate": 3.967445925278567e-06, + "loss": 2.2425, + "step": 9502 + }, + { + "epoch": 2.07, + "learning_rate": 3.967336683417085e-06, + "loss": 2.5412, + "step": 9503 + }, + { + "epoch": 2.07, + "learning_rate": 3.967227441555604e-06, + "loss": 2.3461, + "step": 9504 + }, + { + "epoch": 2.07, + "learning_rate": 3.967118199694123e-06, + "loss": 2.0867, + "step": 9505 + }, + { + "epoch": 2.07, + "learning_rate": 3.967008957832642e-06, + "loss": 3.6053, + "step": 9506 + }, + { + "epoch": 2.07, + "learning_rate": 3.9668997159711605e-06, + "loss": 2.3556, + "step": 9507 + }, + { + "epoch": 2.08, + "learning_rate": 3.966790474109679e-06, + "loss": 2.722, + "step": 9508 + }, + { + "epoch": 2.08, + "learning_rate": 3.966681232248198e-06, + "loss": 2.9884, + "step": 9509 + }, + { + "epoch": 2.08, + "learning_rate": 3.966571990386717e-06, + "loss": 3.4761, + "step": 9510 + }, + { + "epoch": 2.08, + "learning_rate": 3.966462748525235e-06, + "loss": 2.2527, + "step": 9511 + }, + { + "epoch": 2.08, + "learning_rate": 3.966353506663754e-06, + "loss": 2.0124, + "step": 9512 + }, + { + "epoch": 2.08, + "learning_rate": 3.966244264802272e-06, + "loss": 3.1776, + "step": 9513 + }, + { + "epoch": 2.08, + "learning_rate": 3.966135022940791e-06, + "loss": 2.9255, + "step": 9514 + }, + { + "epoch": 2.08, + "learning_rate": 3.96602578107931e-06, + "loss": 3.7081, + "step": 9515 + }, + { + "epoch": 2.08, + "learning_rate": 3.9659165392178285e-06, + "loss": 2.9015, + "step": 9516 + }, + { + "epoch": 2.08, + "learning_rate": 3.9658072973563475e-06, + "loss": 2.0665, + "step": 9517 + }, + { + "epoch": 2.08, + "learning_rate": 3.965698055494866e-06, + "loss": 1.8029, + "step": 9518 + }, + { + "epoch": 2.08, + "learning_rate": 3.965588813633385e-06, + "loss": 3.1369, + "step": 9519 + }, + { + "epoch": 2.08, + "learning_rate": 3.965479571771904e-06, + "loss": 3.2267, + "step": 9520 + }, + { + "epoch": 2.08, + "learning_rate": 3.965370329910422e-06, + "loss": 2.1649, + "step": 9521 + }, + { + "epoch": 2.08, + "learning_rate": 3.965261088048941e-06, + "loss": 3.1825, + "step": 9522 + }, + { + "epoch": 2.08, + "learning_rate": 3.965151846187459e-06, + "loss": 2.3461, + "step": 9523 + }, + { + "epoch": 2.08, + "learning_rate": 3.965042604325978e-06, + "loss": 1.9837, + "step": 9524 + }, + { + "epoch": 2.08, + "learning_rate": 3.964933362464496e-06, + "loss": 2.6526, + "step": 9525 + }, + { + "epoch": 2.08, + "learning_rate": 3.9648241206030154e-06, + "loss": 2.6309, + "step": 9526 + }, + { + "epoch": 2.08, + "learning_rate": 3.9647148787415345e-06, + "loss": 3.3764, + "step": 9527 + }, + { + "epoch": 2.08, + "learning_rate": 3.964605636880053e-06, + "loss": 2.4438, + "step": 9528 + }, + { + "epoch": 2.08, + "learning_rate": 3.964496395018572e-06, + "loss": 2.6713, + "step": 9529 + }, + { + "epoch": 2.08, + "learning_rate": 3.96438715315709e-06, + "loss": 2.6843, + "step": 9530 + }, + { + "epoch": 2.08, + "learning_rate": 3.964277911295609e-06, + "loss": 3.3384, + "step": 9531 + }, + { + "epoch": 2.08, + "learning_rate": 3.964168669434127e-06, + "loss": 1.911, + "step": 9532 + }, + { + "epoch": 2.08, + "learning_rate": 3.964059427572646e-06, + "loss": 2.9105, + "step": 9533 + }, + { + "epoch": 2.08, + "learning_rate": 3.963950185711165e-06, + "loss": 3.0341, + "step": 9534 + }, + { + "epoch": 2.08, + "learning_rate": 3.963840943849683e-06, + "loss": 3.0874, + "step": 9535 + }, + { + "epoch": 2.08, + "learning_rate": 3.963731701988202e-06, + "loss": 2.0912, + "step": 9536 + }, + { + "epoch": 2.08, + "learning_rate": 3.9636224601267206e-06, + "loss": 2.1311, + "step": 9537 + }, + { + "epoch": 2.08, + "learning_rate": 3.96351321826524e-06, + "loss": 3.4273, + "step": 9538 + }, + { + "epoch": 2.08, + "learning_rate": 3.963403976403759e-06, + "loss": 2.121, + "step": 9539 + }, + { + "epoch": 2.08, + "learning_rate": 3.963294734542277e-06, + "loss": 2.9839, + "step": 9540 + }, + { + "epoch": 2.08, + "learning_rate": 3.963185492680796e-06, + "loss": 1.7064, + "step": 9541 + }, + { + "epoch": 2.08, + "learning_rate": 3.963076250819314e-06, + "loss": 1.9485, + "step": 9542 + }, + { + "epoch": 2.08, + "learning_rate": 3.962967008957833e-06, + "loss": 2.9011, + "step": 9543 + }, + { + "epoch": 2.08, + "learning_rate": 3.962857767096351e-06, + "loss": 3.4077, + "step": 9544 + }, + { + "epoch": 2.08, + "learning_rate": 3.96274852523487e-06, + "loss": 2.1621, + "step": 9545 + }, + { + "epoch": 2.08, + "learning_rate": 3.962639283373389e-06, + "loss": 2.6287, + "step": 9546 + }, + { + "epoch": 2.08, + "learning_rate": 3.9625300415119075e-06, + "loss": 2.6047, + "step": 9547 + }, + { + "epoch": 2.08, + "learning_rate": 3.9624207996504266e-06, + "loss": 2.8952, + "step": 9548 + }, + { + "epoch": 2.08, + "learning_rate": 3.962311557788945e-06, + "loss": 2.2984, + "step": 9549 + }, + { + "epoch": 2.08, + "learning_rate": 3.962202315927464e-06, + "loss": 2.1991, + "step": 9550 + }, + { + "epoch": 2.08, + "learning_rate": 3.962093074065983e-06, + "loss": 3.4213, + "step": 9551 + }, + { + "epoch": 2.08, + "learning_rate": 3.961983832204501e-06, + "loss": 2.0778, + "step": 9552 + }, + { + "epoch": 2.08, + "learning_rate": 3.96187459034302e-06, + "loss": 2.4418, + "step": 9553 + }, + { + "epoch": 2.09, + "learning_rate": 3.961765348481538e-06, + "loss": 2.1691, + "step": 9554 + }, + { + "epoch": 2.09, + "learning_rate": 3.961656106620057e-06, + "loss": 2.3514, + "step": 9555 + }, + { + "epoch": 2.09, + "learning_rate": 3.9615468647585755e-06, + "loss": 3.3799, + "step": 9556 + }, + { + "epoch": 2.09, + "learning_rate": 3.9614376228970945e-06, + "loss": 2.195, + "step": 9557 + }, + { + "epoch": 2.09, + "learning_rate": 3.961328381035613e-06, + "loss": 2.6696, + "step": 9558 + }, + { + "epoch": 2.09, + "learning_rate": 3.961219139174132e-06, + "loss": 2.048, + "step": 9559 + }, + { + "epoch": 2.09, + "learning_rate": 3.961109897312651e-06, + "loss": 2.8468, + "step": 9560 + }, + { + "epoch": 2.09, + "learning_rate": 3.961000655451169e-06, + "loss": 2.4563, + "step": 9561 + }, + { + "epoch": 2.09, + "learning_rate": 3.960891413589688e-06, + "loss": 2.5596, + "step": 9562 + }, + { + "epoch": 2.09, + "learning_rate": 3.960782171728206e-06, + "loss": 2.2072, + "step": 9563 + }, + { + "epoch": 2.09, + "learning_rate": 3.960672929866725e-06, + "loss": 2.9758, + "step": 9564 + }, + { + "epoch": 2.09, + "learning_rate": 3.960563688005244e-06, + "loss": 2.4144, + "step": 9565 + }, + { + "epoch": 2.09, + "learning_rate": 3.960454446143762e-06, + "loss": 2.2813, + "step": 9566 + }, + { + "epoch": 2.09, + "learning_rate": 3.9603452042822814e-06, + "loss": 2.7713, + "step": 9567 + }, + { + "epoch": 2.09, + "learning_rate": 3.9602359624208e-06, + "loss": 1.7573, + "step": 9568 + }, + { + "epoch": 2.09, + "learning_rate": 3.960126720559319e-06, + "loss": 2.4192, + "step": 9569 + }, + { + "epoch": 2.09, + "learning_rate": 3.960017478697838e-06, + "loss": 1.7624, + "step": 9570 + }, + { + "epoch": 2.09, + "learning_rate": 3.959908236836356e-06, + "loss": 2.2854, + "step": 9571 + }, + { + "epoch": 2.09, + "learning_rate": 3.959798994974875e-06, + "loss": 2.1473, + "step": 9572 + }, + { + "epoch": 2.09, + "learning_rate": 3.959689753113393e-06, + "loss": 2.6165, + "step": 9573 + }, + { + "epoch": 2.09, + "learning_rate": 3.959580511251912e-06, + "loss": 2.495, + "step": 9574 + }, + { + "epoch": 2.09, + "learning_rate": 3.959471269390431e-06, + "loss": 3.3141, + "step": 9575 + }, + { + "epoch": 2.09, + "learning_rate": 3.959362027528949e-06, + "loss": 2.271, + "step": 9576 + }, + { + "epoch": 2.09, + "learning_rate": 3.9592527856674676e-06, + "loss": 2.018, + "step": 9577 + }, + { + "epoch": 2.09, + "learning_rate": 3.959143543805987e-06, + "loss": 1.8449, + "step": 9578 + }, + { + "epoch": 2.09, + "learning_rate": 3.959034301944506e-06, + "loss": 3.6173, + "step": 9579 + }, + { + "epoch": 2.09, + "learning_rate": 3.958925060083025e-06, + "loss": 3.2619, + "step": 9580 + }, + { + "epoch": 2.09, + "learning_rate": 3.958815818221543e-06, + "loss": 2.549, + "step": 9581 + }, + { + "epoch": 2.09, + "learning_rate": 3.958706576360061e-06, + "loss": 1.9423, + "step": 9582 + }, + { + "epoch": 2.09, + "learning_rate": 3.95859733449858e-06, + "loss": 3.2979, + "step": 9583 + }, + { + "epoch": 2.09, + "learning_rate": 3.958488092637099e-06, + "loss": 2.6527, + "step": 9584 + }, + { + "epoch": 2.09, + "learning_rate": 3.958378850775617e-06, + "loss": 2.8558, + "step": 9585 + }, + { + "epoch": 2.09, + "learning_rate": 3.958269608914136e-06, + "loss": 2.3018, + "step": 9586 + }, + { + "epoch": 2.09, + "learning_rate": 3.9581603670526545e-06, + "loss": 2.4047, + "step": 9587 + }, + { + "epoch": 2.09, + "learning_rate": 3.9580511251911735e-06, + "loss": 2.3462, + "step": 9588 + }, + { + "epoch": 2.09, + "learning_rate": 3.957941883329693e-06, + "loss": 2.6941, + "step": 9589 + }, + { + "epoch": 2.09, + "learning_rate": 3.957832641468211e-06, + "loss": 1.4626, + "step": 9590 + }, + { + "epoch": 2.09, + "learning_rate": 3.957723399606729e-06, + "loss": 3.3855, + "step": 9591 + }, + { + "epoch": 2.09, + "learning_rate": 3.957614157745248e-06, + "loss": 2.035, + "step": 9592 + }, + { + "epoch": 2.09, + "learning_rate": 3.957504915883767e-06, + "loss": 2.9528, + "step": 9593 + }, + { + "epoch": 2.09, + "learning_rate": 3.957395674022286e-06, + "loss": 2.5627, + "step": 9594 + }, + { + "epoch": 2.09, + "learning_rate": 3.957286432160804e-06, + "loss": 2.3463, + "step": 9595 + }, + { + "epoch": 2.09, + "learning_rate": 3.9571771902993224e-06, + "loss": 2.1723, + "step": 9596 + }, + { + "epoch": 2.09, + "learning_rate": 3.9570679484378415e-06, + "loss": 2.9041, + "step": 9597 + }, + { + "epoch": 2.09, + "learning_rate": 3.9569587065763605e-06, + "loss": 2.9026, + "step": 9598 + }, + { + "epoch": 2.09, + "learning_rate": 3.9568494647148795e-06, + "loss": 2.8609, + "step": 9599 + }, + { + "epoch": 2.1, + "learning_rate": 3.956740222853398e-06, + "loss": 2.1843, + "step": 9600 + }, + { + "epoch": 2.1, + "learning_rate": 3.956630980991916e-06, + "loss": 1.6894, + "step": 9601 + }, + { + "epoch": 2.1, + "learning_rate": 3.956521739130435e-06, + "loss": 1.8202, + "step": 9602 + }, + { + "epoch": 2.1, + "learning_rate": 3.956412497268954e-06, + "loss": 3.4738, + "step": 9603 + }, + { + "epoch": 2.1, + "learning_rate": 3.956303255407473e-06, + "loss": 1.4327, + "step": 9604 + }, + { + "epoch": 2.1, + "learning_rate": 3.956194013545991e-06, + "loss": 2.7356, + "step": 9605 + }, + { + "epoch": 2.1, + "learning_rate": 3.956084771684509e-06, + "loss": 2.5005, + "step": 9606 + }, + { + "epoch": 2.1, + "learning_rate": 3.955975529823028e-06, + "loss": 3.1482, + "step": 9607 + }, + { + "epoch": 2.1, + "learning_rate": 3.9558662879615475e-06, + "loss": 2.549, + "step": 9608 + }, + { + "epoch": 2.1, + "learning_rate": 3.955757046100066e-06, + "loss": 3.6807, + "step": 9609 + }, + { + "epoch": 2.1, + "learning_rate": 3.955647804238585e-06, + "loss": 2.9987, + "step": 9610 + }, + { + "epoch": 2.1, + "learning_rate": 3.955538562377103e-06, + "loss": 1.1063, + "step": 9611 + }, + { + "epoch": 2.1, + "learning_rate": 3.955429320515622e-06, + "loss": 1.7254, + "step": 9612 + }, + { + "epoch": 2.1, + "learning_rate": 3.955320078654141e-06, + "loss": 1.9905, + "step": 9613 + }, + { + "epoch": 2.1, + "learning_rate": 3.955210836792659e-06, + "loss": 2.6576, + "step": 9614 + }, + { + "epoch": 2.1, + "learning_rate": 3.955101594931178e-06, + "loss": 3.1376, + "step": 9615 + }, + { + "epoch": 2.1, + "learning_rate": 3.954992353069696e-06, + "loss": 3.6546, + "step": 9616 + }, + { + "epoch": 2.1, + "learning_rate": 3.954883111208215e-06, + "loss": 2.9955, + "step": 9617 + }, + { + "epoch": 2.1, + "learning_rate": 3.954773869346734e-06, + "loss": 1.7182, + "step": 9618 + }, + { + "epoch": 2.1, + "learning_rate": 3.954664627485253e-06, + "loss": 1.8133, + "step": 9619 + }, + { + "epoch": 2.1, + "learning_rate": 3.954555385623771e-06, + "loss": 1.8693, + "step": 9620 + }, + { + "epoch": 2.1, + "learning_rate": 3.95444614376229e-06, + "loss": 3.616, + "step": 9621 + }, + { + "epoch": 2.1, + "learning_rate": 3.954336901900809e-06, + "loss": 1.6603, + "step": 9622 + }, + { + "epoch": 2.1, + "learning_rate": 3.954227660039328e-06, + "loss": 2.7918, + "step": 9623 + }, + { + "epoch": 2.1, + "learning_rate": 3.954118418177846e-06, + "loss": 3.1901, + "step": 9624 + }, + { + "epoch": 2.1, + "learning_rate": 3.954009176316364e-06, + "loss": 2.5591, + "step": 9625 + }, + { + "epoch": 2.1, + "learning_rate": 3.953899934454883e-06, + "loss": 2.8851, + "step": 9626 + }, + { + "epoch": 2.1, + "learning_rate": 3.953790692593402e-06, + "loss": 2.7299, + "step": 9627 + }, + { + "epoch": 2.1, + "learning_rate": 3.953681450731921e-06, + "loss": 1.7847, + "step": 9628 + }, + { + "epoch": 2.1, + "learning_rate": 3.9535722088704396e-06, + "loss": 1.6728, + "step": 9629 + }, + { + "epoch": 2.1, + "learning_rate": 3.953462967008958e-06, + "loss": 2.8765, + "step": 9630 + }, + { + "epoch": 2.1, + "learning_rate": 3.953353725147477e-06, + "loss": 3.1836, + "step": 9631 + }, + { + "epoch": 2.1, + "learning_rate": 3.953244483285996e-06, + "loss": 3.216, + "step": 9632 + }, + { + "epoch": 2.1, + "learning_rate": 3.953135241424515e-06, + "loss": 2.3246, + "step": 9633 + }, + { + "epoch": 2.1, + "learning_rate": 3.953025999563033e-06, + "loss": 2.3718, + "step": 9634 + }, + { + "epoch": 2.1, + "learning_rate": 3.952916757701551e-06, + "loss": 2.4179, + "step": 9635 + }, + { + "epoch": 2.1, + "learning_rate": 3.95280751584007e-06, + "loss": 3.3278, + "step": 9636 + }, + { + "epoch": 2.1, + "learning_rate": 3.952698273978589e-06, + "loss": 2.6972, + "step": 9637 + }, + { + "epoch": 2.1, + "learning_rate": 3.9525890321171075e-06, + "loss": 2.9171, + "step": 9638 + }, + { + "epoch": 2.1, + "learning_rate": 3.9524797902556265e-06, + "loss": 2.8432, + "step": 9639 + }, + { + "epoch": 2.1, + "learning_rate": 3.952370548394145e-06, + "loss": 1.7878, + "step": 9640 + }, + { + "epoch": 2.1, + "learning_rate": 3.952261306532664e-06, + "loss": 2.2113, + "step": 9641 + }, + { + "epoch": 2.1, + "learning_rate": 3.952152064671183e-06, + "loss": 1.6994, + "step": 9642 + }, + { + "epoch": 2.1, + "learning_rate": 3.952042822809701e-06, + "loss": 3.255, + "step": 9643 + }, + { + "epoch": 2.1, + "learning_rate": 3.951933580948219e-06, + "loss": 2.069, + "step": 9644 + }, + { + "epoch": 2.1, + "learning_rate": 3.951824339086738e-06, + "loss": 2.3962, + "step": 9645 + }, + { + "epoch": 2.11, + "learning_rate": 3.951715097225257e-06, + "loss": 3.797, + "step": 9646 + }, + { + "epoch": 2.11, + "learning_rate": 3.951605855363776e-06, + "loss": 1.7992, + "step": 9647 + }, + { + "epoch": 2.11, + "learning_rate": 3.9514966135022944e-06, + "loss": 2.5759, + "step": 9648 + }, + { + "epoch": 2.11, + "learning_rate": 3.951387371640813e-06, + "loss": 2.7685, + "step": 9649 + }, + { + "epoch": 2.11, + "learning_rate": 3.951278129779332e-06, + "loss": 3.2788, + "step": 9650 + }, + { + "epoch": 2.11, + "learning_rate": 3.951168887917851e-06, + "loss": 1.8206, + "step": 9651 + }, + { + "epoch": 2.11, + "learning_rate": 3.95105964605637e-06, + "loss": 1.9739, + "step": 9652 + }, + { + "epoch": 2.11, + "learning_rate": 3.950950404194888e-06, + "loss": 3.1707, + "step": 9653 + }, + { + "epoch": 2.11, + "learning_rate": 3.950841162333406e-06, + "loss": 1.9901, + "step": 9654 + }, + { + "epoch": 2.11, + "learning_rate": 3.950731920471925e-06, + "loss": 4.4493, + "step": 9655 + }, + { + "epoch": 2.11, + "learning_rate": 3.950622678610444e-06, + "loss": 2.6289, + "step": 9656 + }, + { + "epoch": 2.11, + "learning_rate": 3.950513436748962e-06, + "loss": 1.6855, + "step": 9657 + }, + { + "epoch": 2.11, + "learning_rate": 3.950404194887481e-06, + "loss": 2.2964, + "step": 9658 + }, + { + "epoch": 2.11, + "learning_rate": 3.950294953026e-06, + "loss": 2.8559, + "step": 9659 + }, + { + "epoch": 2.11, + "learning_rate": 3.950185711164519e-06, + "loss": 0.9728, + "step": 9660 + }, + { + "epoch": 2.11, + "learning_rate": 3.950076469303038e-06, + "loss": 1.9919, + "step": 9661 + }, + { + "epoch": 2.11, + "learning_rate": 3.949967227441556e-06, + "loss": 3.4144, + "step": 9662 + }, + { + "epoch": 2.11, + "learning_rate": 3.949857985580075e-06, + "loss": 2.8351, + "step": 9663 + }, + { + "epoch": 2.11, + "learning_rate": 3.949748743718593e-06, + "loss": 2.8175, + "step": 9664 + }, + { + "epoch": 2.11, + "learning_rate": 3.949639501857112e-06, + "loss": 3.2036, + "step": 9665 + }, + { + "epoch": 2.11, + "learning_rate": 3.949530259995631e-06, + "loss": 2.1196, + "step": 9666 + }, + { + "epoch": 2.11, + "learning_rate": 3.949421018134149e-06, + "loss": 2.8983, + "step": 9667 + }, + { + "epoch": 2.11, + "learning_rate": 3.949311776272668e-06, + "loss": 2.317, + "step": 9668 + }, + { + "epoch": 2.11, + "learning_rate": 3.9492025344111865e-06, + "loss": 2.3947, + "step": 9669 + }, + { + "epoch": 2.11, + "learning_rate": 3.9490932925497056e-06, + "loss": 2.9163, + "step": 9670 + }, + { + "epoch": 2.11, + "learning_rate": 3.948984050688225e-06, + "loss": 2.1371, + "step": 9671 + }, + { + "epoch": 2.11, + "learning_rate": 3.948874808826743e-06, + "loss": 1.5517, + "step": 9672 + }, + { + "epoch": 2.11, + "learning_rate": 3.948765566965261e-06, + "loss": 2.3104, + "step": 9673 + }, + { + "epoch": 2.11, + "learning_rate": 3.94865632510378e-06, + "loss": 2.2258, + "step": 9674 + }, + { + "epoch": 2.11, + "learning_rate": 3.948547083242299e-06, + "loss": 2.4511, + "step": 9675 + }, + { + "epoch": 2.11, + "learning_rate": 3.948437841380817e-06, + "loss": 2.2872, + "step": 9676 + }, + { + "epoch": 2.11, + "learning_rate": 3.948328599519336e-06, + "loss": 2.2683, + "step": 9677 + }, + { + "epoch": 2.11, + "learning_rate": 3.9482193576578545e-06, + "loss": 2.465, + "step": 9678 + }, + { + "epoch": 2.11, + "learning_rate": 3.9481101157963735e-06, + "loss": 2.569, + "step": 9679 + }, + { + "epoch": 2.11, + "learning_rate": 3.9480008739348925e-06, + "loss": 2.2114, + "step": 9680 + }, + { + "epoch": 2.11, + "learning_rate": 3.947891632073411e-06, + "loss": 2.338, + "step": 9681 + }, + { + "epoch": 2.11, + "learning_rate": 3.94778239021193e-06, + "loss": 2.4784, + "step": 9682 + }, + { + "epoch": 2.11, + "learning_rate": 3.947673148350448e-06, + "loss": 2.872, + "step": 9683 + }, + { + "epoch": 2.11, + "learning_rate": 3.947563906488967e-06, + "loss": 2.3269, + "step": 9684 + }, + { + "epoch": 2.11, + "learning_rate": 3.947454664627486e-06, + "loss": 2.2349, + "step": 9685 + }, + { + "epoch": 2.11, + "learning_rate": 3.947345422766004e-06, + "loss": 2.3834, + "step": 9686 + }, + { + "epoch": 2.11, + "learning_rate": 3.947236180904523e-06, + "loss": 1.6864, + "step": 9687 + }, + { + "epoch": 2.11, + "learning_rate": 3.947126939043041e-06, + "loss": 2.7628, + "step": 9688 + }, + { + "epoch": 2.11, + "learning_rate": 3.9470176971815605e-06, + "loss": 2.7087, + "step": 9689 + }, + { + "epoch": 2.11, + "learning_rate": 3.946908455320079e-06, + "loss": 1.6381, + "step": 9690 + }, + { + "epoch": 2.12, + "learning_rate": 3.946799213458598e-06, + "loss": 1.5994, + "step": 9691 + }, + { + "epoch": 2.12, + "learning_rate": 3.946689971597117e-06, + "loss": 2.2958, + "step": 9692 + }, + { + "epoch": 2.12, + "learning_rate": 3.946580729735635e-06, + "loss": 2.3561, + "step": 9693 + }, + { + "epoch": 2.12, + "learning_rate": 3.946471487874154e-06, + "loss": 2.9778, + "step": 9694 + }, + { + "epoch": 2.12, + "learning_rate": 3.946362246012672e-06, + "loss": 1.1637, + "step": 9695 + }, + { + "epoch": 2.12, + "learning_rate": 3.946253004151191e-06, + "loss": 3.6391, + "step": 9696 + }, + { + "epoch": 2.12, + "learning_rate": 3.946143762289709e-06, + "loss": 3.4839, + "step": 9697 + }, + { + "epoch": 2.12, + "learning_rate": 3.946034520428228e-06, + "loss": 2.9309, + "step": 9698 + }, + { + "epoch": 2.12, + "learning_rate": 3.945925278566747e-06, + "loss": 2.1882, + "step": 9699 + }, + { + "epoch": 2.12, + "learning_rate": 3.945816036705266e-06, + "loss": 2.9279, + "step": 9700 + }, + { + "epoch": 2.12, + "learning_rate": 3.945706794843785e-06, + "loss": 2.9882, + "step": 9701 + }, + { + "epoch": 2.12, + "learning_rate": 3.945597552982303e-06, + "loss": 2.8899, + "step": 9702 + }, + { + "epoch": 2.12, + "learning_rate": 3.945488311120822e-06, + "loss": 2.282, + "step": 9703 + }, + { + "epoch": 2.12, + "learning_rate": 3.945379069259341e-06, + "loss": 2.9732, + "step": 9704 + }, + { + "epoch": 2.12, + "learning_rate": 3.945269827397859e-06, + "loss": 3.2568, + "step": 9705 + }, + { + "epoch": 2.12, + "learning_rate": 3.945160585536378e-06, + "loss": 2.2643, + "step": 9706 + }, + { + "epoch": 2.12, + "learning_rate": 3.945051343674896e-06, + "loss": 2.2279, + "step": 9707 + }, + { + "epoch": 2.12, + "learning_rate": 3.944942101813415e-06, + "loss": 3.3302, + "step": 9708 + }, + { + "epoch": 2.12, + "learning_rate": 3.9448328599519335e-06, + "loss": 1.4333, + "step": 9709 + }, + { + "epoch": 2.12, + "learning_rate": 3.9447236180904526e-06, + "loss": 2.4315, + "step": 9710 + }, + { + "epoch": 2.12, + "learning_rate": 3.944614376228972e-06, + "loss": 2.6467, + "step": 9711 + }, + { + "epoch": 2.12, + "learning_rate": 3.94450513436749e-06, + "loss": 3.9883, + "step": 9712 + }, + { + "epoch": 2.12, + "learning_rate": 3.944395892506009e-06, + "loss": 2.0821, + "step": 9713 + }, + { + "epoch": 2.12, + "learning_rate": 3.944286650644527e-06, + "loss": 3.4017, + "step": 9714 + }, + { + "epoch": 2.12, + "learning_rate": 3.944177408783046e-06, + "loss": 1.6721, + "step": 9715 + }, + { + "epoch": 2.12, + "learning_rate": 3.944068166921565e-06, + "loss": 2.5434, + "step": 9716 + }, + { + "epoch": 2.12, + "learning_rate": 3.943958925060083e-06, + "loss": 2.043, + "step": 9717 + }, + { + "epoch": 2.12, + "learning_rate": 3.943849683198602e-06, + "loss": 1.9594, + "step": 9718 + }, + { + "epoch": 2.12, + "learning_rate": 3.9437404413371205e-06, + "loss": 3.2911, + "step": 9719 + }, + { + "epoch": 2.12, + "learning_rate": 3.9436311994756395e-06, + "loss": 2.9106, + "step": 9720 + }, + { + "epoch": 2.12, + "learning_rate": 3.9435219576141585e-06, + "loss": 3.3218, + "step": 9721 + }, + { + "epoch": 2.12, + "learning_rate": 3.943412715752677e-06, + "loss": 3.4385, + "step": 9722 + }, + { + "epoch": 2.12, + "learning_rate": 3.943303473891195e-06, + "loss": 2.3855, + "step": 9723 + }, + { + "epoch": 2.12, + "learning_rate": 3.943194232029714e-06, + "loss": 2.2083, + "step": 9724 + }, + { + "epoch": 2.12, + "learning_rate": 3.943084990168233e-06, + "loss": 2.8625, + "step": 9725 + }, + { + "epoch": 2.12, + "learning_rate": 3.942975748306751e-06, + "loss": 2.0412, + "step": 9726 + }, + { + "epoch": 2.12, + "learning_rate": 3.94286650644527e-06, + "loss": 1.3561, + "step": 9727 + }, + { + "epoch": 2.12, + "learning_rate": 3.942757264583788e-06, + "loss": 2.4131, + "step": 9728 + }, + { + "epoch": 2.12, + "learning_rate": 3.9426480227223074e-06, + "loss": 2.5336, + "step": 9729 + }, + { + "epoch": 2.12, + "learning_rate": 3.9425387808608265e-06, + "loss": 1.6873, + "step": 9730 + }, + { + "epoch": 2.12, + "learning_rate": 3.942429538999345e-06, + "loss": 2.3699, + "step": 9731 + }, + { + "epoch": 2.12, + "learning_rate": 3.942320297137864e-06, + "loss": 3.3199, + "step": 9732 + }, + { + "epoch": 2.12, + "learning_rate": 3.942211055276382e-06, + "loss": 2.2358, + "step": 9733 + }, + { + "epoch": 2.12, + "learning_rate": 3.942101813414901e-06, + "loss": 2.4942, + "step": 9734 + }, + { + "epoch": 2.12, + "learning_rate": 3.94199257155342e-06, + "loss": 3.1899, + "step": 9735 + }, + { + "epoch": 2.12, + "learning_rate": 3.941883329691938e-06, + "loss": 1.5262, + "step": 9736 + }, + { + "epoch": 2.13, + "learning_rate": 3.941774087830457e-06, + "loss": 3.6965, + "step": 9737 + }, + { + "epoch": 2.13, + "learning_rate": 3.941664845968975e-06, + "loss": 2.2113, + "step": 9738 + }, + { + "epoch": 2.13, + "learning_rate": 3.941555604107494e-06, + "loss": 2.6065, + "step": 9739 + }, + { + "epoch": 2.13, + "learning_rate": 3.9414463622460134e-06, + "loss": 2.3051, + "step": 9740 + }, + { + "epoch": 2.13, + "learning_rate": 3.941337120384532e-06, + "loss": 2.6684, + "step": 9741 + }, + { + "epoch": 2.13, + "learning_rate": 3.94122787852305e-06, + "loss": 2.8123, + "step": 9742 + }, + { + "epoch": 2.13, + "learning_rate": 3.941118636661569e-06, + "loss": 3.684, + "step": 9743 + }, + { + "epoch": 2.13, + "learning_rate": 3.941009394800088e-06, + "loss": 1.8312, + "step": 9744 + }, + { + "epoch": 2.13, + "learning_rate": 3.940900152938607e-06, + "loss": 2.0751, + "step": 9745 + }, + { + "epoch": 2.13, + "learning_rate": 3.940790911077125e-06, + "loss": 3.1772, + "step": 9746 + }, + { + "epoch": 2.13, + "learning_rate": 3.940681669215643e-06, + "loss": 3.022, + "step": 9747 + }, + { + "epoch": 2.13, + "learning_rate": 3.940572427354162e-06, + "loss": 3.1983, + "step": 9748 + }, + { + "epoch": 2.13, + "learning_rate": 3.940463185492681e-06, + "loss": 2.138, + "step": 9749 + }, + { + "epoch": 2.13, + "learning_rate": 3.9403539436311995e-06, + "loss": 3.0156, + "step": 9750 + }, + { + "epoch": 2.13, + "learning_rate": 3.9402447017697186e-06, + "loss": 1.7388, + "step": 9751 + }, + { + "epoch": 2.13, + "learning_rate": 3.940135459908237e-06, + "loss": 3.132, + "step": 9752 + }, + { + "epoch": 2.13, + "learning_rate": 3.940026218046756e-06, + "loss": 1.8886, + "step": 9753 + }, + { + "epoch": 2.13, + "learning_rate": 3.939916976185275e-06, + "loss": 1.9942, + "step": 9754 + }, + { + "epoch": 2.13, + "learning_rate": 3.939807734323793e-06, + "loss": 2.0941, + "step": 9755 + }, + { + "epoch": 2.13, + "learning_rate": 3.939698492462311e-06, + "loss": 2.4805, + "step": 9756 + }, + { + "epoch": 2.13, + "learning_rate": 3.93958925060083e-06, + "loss": 1.8653, + "step": 9757 + }, + { + "epoch": 2.13, + "learning_rate": 3.939480008739349e-06, + "loss": 2.147, + "step": 9758 + }, + { + "epoch": 2.13, + "learning_rate": 3.939370766877868e-06, + "loss": 2.3583, + "step": 9759 + }, + { + "epoch": 2.13, + "learning_rate": 3.9392615250163865e-06, + "loss": 2.2596, + "step": 9760 + }, + { + "epoch": 2.13, + "learning_rate": 3.939152283154905e-06, + "loss": 1.857, + "step": 9761 + }, + { + "epoch": 2.13, + "learning_rate": 3.939043041293424e-06, + "loss": 3.4533, + "step": 9762 + }, + { + "epoch": 2.13, + "learning_rate": 3.938933799431943e-06, + "loss": 2.3211, + "step": 9763 + }, + { + "epoch": 2.13, + "learning_rate": 3.938824557570462e-06, + "loss": 2.315, + "step": 9764 + }, + { + "epoch": 2.13, + "learning_rate": 3.93871531570898e-06, + "loss": 2.2232, + "step": 9765 + }, + { + "epoch": 2.13, + "learning_rate": 3.938606073847498e-06, + "loss": 2.9554, + "step": 9766 + }, + { + "epoch": 2.13, + "learning_rate": 3.938496831986017e-06, + "loss": 1.5563, + "step": 9767 + }, + { + "epoch": 2.13, + "learning_rate": 3.938387590124536e-06, + "loss": 2.0002, + "step": 9768 + }, + { + "epoch": 2.13, + "learning_rate": 3.938278348263055e-06, + "loss": 3.0227, + "step": 9769 + }, + { + "epoch": 2.13, + "learning_rate": 3.9381691064015734e-06, + "loss": 3.669, + "step": 9770 + }, + { + "epoch": 2.13, + "learning_rate": 3.938059864540092e-06, + "loss": 2.7558, + "step": 9771 + }, + { + "epoch": 2.13, + "learning_rate": 3.937950622678611e-06, + "loss": 1.157, + "step": 9772 + }, + { + "epoch": 2.13, + "learning_rate": 3.93784138081713e-06, + "loss": 1.6085, + "step": 9773 + }, + { + "epoch": 2.13, + "learning_rate": 3.937732138955649e-06, + "loss": 1.7883, + "step": 9774 + }, + { + "epoch": 2.13, + "learning_rate": 3.937622897094167e-06, + "loss": 3.9849, + "step": 9775 + }, + { + "epoch": 2.13, + "learning_rate": 3.937513655232685e-06, + "loss": 2.25, + "step": 9776 + }, + { + "epoch": 2.13, + "learning_rate": 3.937404413371204e-06, + "loss": 3.3263, + "step": 9777 + }, + { + "epoch": 2.13, + "learning_rate": 3.937295171509723e-06, + "loss": 1.8746, + "step": 9778 + }, + { + "epoch": 2.13, + "learning_rate": 3.937185929648241e-06, + "loss": 2.7128, + "step": 9779 + }, + { + "epoch": 2.13, + "learning_rate": 3.93707668778676e-06, + "loss": 2.5083, + "step": 9780 + }, + { + "epoch": 2.13, + "learning_rate": 3.936967445925279e-06, + "loss": 2.2941, + "step": 9781 + }, + { + "epoch": 2.13, + "learning_rate": 3.936858204063798e-06, + "loss": 2.7948, + "step": 9782 + }, + { + "epoch": 2.14, + "learning_rate": 3.936748962202317e-06, + "loss": 1.2665, + "step": 9783 + }, + { + "epoch": 2.14, + "learning_rate": 3.936639720340835e-06, + "loss": 3.6747, + "step": 9784 + }, + { + "epoch": 2.14, + "learning_rate": 3.936530478479353e-06, + "loss": 1.933, + "step": 9785 + }, + { + "epoch": 2.14, + "learning_rate": 3.936421236617872e-06, + "loss": 2.5612, + "step": 9786 + }, + { + "epoch": 2.14, + "learning_rate": 3.936311994756391e-06, + "loss": 3.2723, + "step": 9787 + }, + { + "epoch": 2.14, + "learning_rate": 3.93620275289491e-06, + "loss": 2.7593, + "step": 9788 + }, + { + "epoch": 2.14, + "learning_rate": 3.936093511033428e-06, + "loss": 2.2525, + "step": 9789 + }, + { + "epoch": 2.14, + "learning_rate": 3.9359842691719465e-06, + "loss": 2.3774, + "step": 9790 + }, + { + "epoch": 2.14, + "learning_rate": 3.9358750273104655e-06, + "loss": 2.8741, + "step": 9791 + }, + { + "epoch": 2.14, + "learning_rate": 3.935765785448985e-06, + "loss": 2.4267, + "step": 9792 + }, + { + "epoch": 2.14, + "learning_rate": 3.935656543587504e-06, + "loss": 0.9823, + "step": 9793 + }, + { + "epoch": 2.14, + "learning_rate": 3.935547301726022e-06, + "loss": 1.0174, + "step": 9794 + }, + { + "epoch": 2.14, + "learning_rate": 3.93543805986454e-06, + "loss": 3.3089, + "step": 9795 + }, + { + "epoch": 2.14, + "learning_rate": 3.935328818003059e-06, + "loss": 3.065, + "step": 9796 + }, + { + "epoch": 2.14, + "learning_rate": 3.935219576141578e-06, + "loss": 2.4671, + "step": 9797 + }, + { + "epoch": 2.14, + "learning_rate": 3.935110334280097e-06, + "loss": 2.0293, + "step": 9798 + }, + { + "epoch": 2.14, + "learning_rate": 3.935001092418615e-06, + "loss": 1.7614, + "step": 9799 + }, + { + "epoch": 2.14, + "learning_rate": 3.9348918505571335e-06, + "loss": 2.3529, + "step": 9800 + }, + { + "epoch": 2.14, + "learning_rate": 3.9347826086956525e-06, + "loss": 1.8749, + "step": 9801 + }, + { + "epoch": 2.14, + "learning_rate": 3.9346733668341715e-06, + "loss": 2.1846, + "step": 9802 + }, + { + "epoch": 2.14, + "learning_rate": 3.93456412497269e-06, + "loss": 1.7577, + "step": 9803 + }, + { + "epoch": 2.14, + "learning_rate": 3.934454883111209e-06, + "loss": 3.4344, + "step": 9804 + }, + { + "epoch": 2.14, + "learning_rate": 3.934345641249727e-06, + "loss": 3.3596, + "step": 9805 + }, + { + "epoch": 2.14, + "learning_rate": 3.934236399388246e-06, + "loss": 2.6705, + "step": 9806 + }, + { + "epoch": 2.14, + "learning_rate": 3.934127157526765e-06, + "loss": 1.9915, + "step": 9807 + }, + { + "epoch": 2.14, + "learning_rate": 3.934017915665283e-06, + "loss": 1.9415, + "step": 9808 + }, + { + "epoch": 2.14, + "learning_rate": 3.933908673803801e-06, + "loss": 2.4447, + "step": 9809 + }, + { + "epoch": 2.14, + "learning_rate": 3.9337994319423204e-06, + "loss": 2.0212, + "step": 9810 + }, + { + "epoch": 2.14, + "learning_rate": 3.9336901900808395e-06, + "loss": 2.2943, + "step": 9811 + }, + { + "epoch": 2.14, + "learning_rate": 3.9335809482193585e-06, + "loss": 2.2431, + "step": 9812 + }, + { + "epoch": 2.14, + "learning_rate": 3.933471706357877e-06, + "loss": 3.2961, + "step": 9813 + }, + { + "epoch": 2.14, + "learning_rate": 3.933362464496395e-06, + "loss": 1.9986, + "step": 9814 + }, + { + "epoch": 2.14, + "learning_rate": 3.933253222634914e-06, + "loss": 2.6867, + "step": 9815 + }, + { + "epoch": 2.14, + "learning_rate": 3.933143980773433e-06, + "loss": 2.1757, + "step": 9816 + }, + { + "epoch": 2.14, + "learning_rate": 3.933034738911952e-06, + "loss": 2.6463, + "step": 9817 + }, + { + "epoch": 2.14, + "learning_rate": 3.93292549705047e-06, + "loss": 3.2246, + "step": 9818 + }, + { + "epoch": 2.14, + "learning_rate": 3.932816255188988e-06, + "loss": 3.0147, + "step": 9819 + }, + { + "epoch": 2.14, + "learning_rate": 3.932707013327507e-06, + "loss": 1.6304, + "step": 9820 + }, + { + "epoch": 2.14, + "learning_rate": 3.932597771466026e-06, + "loss": 1.6134, + "step": 9821 + }, + { + "epoch": 2.14, + "learning_rate": 3.932488529604545e-06, + "loss": 3.2055, + "step": 9822 + }, + { + "epoch": 2.14, + "learning_rate": 3.932379287743064e-06, + "loss": 2.1293, + "step": 9823 + }, + { + "epoch": 2.14, + "learning_rate": 3.932270045881582e-06, + "loss": 1.6355, + "step": 9824 + }, + { + "epoch": 2.14, + "learning_rate": 3.932160804020101e-06, + "loss": 3.544, + "step": 9825 + }, + { + "epoch": 2.14, + "learning_rate": 3.93205156215862e-06, + "loss": 2.1469, + "step": 9826 + }, + { + "epoch": 2.14, + "learning_rate": 3.931942320297138e-06, + "loss": 1.3885, + "step": 9827 + }, + { + "epoch": 2.14, + "learning_rate": 3.931833078435657e-06, + "loss": 1.6013, + "step": 9828 + }, + { + "epoch": 2.15, + "learning_rate": 3.931723836574175e-06, + "loss": 2.533, + "step": 9829 + }, + { + "epoch": 2.15, + "learning_rate": 3.931614594712694e-06, + "loss": 2.3877, + "step": 9830 + }, + { + "epoch": 2.15, + "learning_rate": 3.931505352851213e-06, + "loss": 2.3854, + "step": 9831 + }, + { + "epoch": 2.15, + "learning_rate": 3.9313961109897316e-06, + "loss": 2.0415, + "step": 9832 + }, + { + "epoch": 2.15, + "learning_rate": 3.931286869128251e-06, + "loss": 1.4701, + "step": 9833 + }, + { + "epoch": 2.15, + "learning_rate": 3.931177627266769e-06, + "loss": 2.4321, + "step": 9834 + }, + { + "epoch": 2.15, + "learning_rate": 3.931068385405288e-06, + "loss": 2.3581, + "step": 9835 + }, + { + "epoch": 2.15, + "learning_rate": 3.930959143543806e-06, + "loss": 1.4838, + "step": 9836 + }, + { + "epoch": 2.15, + "learning_rate": 3.930849901682325e-06, + "loss": 2.2861, + "step": 9837 + }, + { + "epoch": 2.15, + "learning_rate": 3.930740659820843e-06, + "loss": 2.8559, + "step": 9838 + }, + { + "epoch": 2.15, + "learning_rate": 3.930631417959362e-06, + "loss": 2.5735, + "step": 9839 + }, + { + "epoch": 2.15, + "learning_rate": 3.930522176097881e-06, + "loss": 1.2451, + "step": 9840 + }, + { + "epoch": 2.15, + "learning_rate": 3.9304129342363995e-06, + "loss": 2.7402, + "step": 9841 + }, + { + "epoch": 2.15, + "learning_rate": 3.9303036923749185e-06, + "loss": 3.085, + "step": 9842 + }, + { + "epoch": 2.15, + "learning_rate": 3.930194450513437e-06, + "loss": 2.9281, + "step": 9843 + }, + { + "epoch": 2.15, + "learning_rate": 3.930085208651956e-06, + "loss": 3.191, + "step": 9844 + }, + { + "epoch": 2.15, + "learning_rate": 3.929975966790475e-06, + "loss": 2.8846, + "step": 9845 + }, + { + "epoch": 2.15, + "learning_rate": 3.929866724928993e-06, + "loss": 3.0284, + "step": 9846 + }, + { + "epoch": 2.15, + "learning_rate": 3.929757483067512e-06, + "loss": 2.858, + "step": 9847 + }, + { + "epoch": 2.15, + "learning_rate": 3.92964824120603e-06, + "loss": 1.7745, + "step": 9848 + }, + { + "epoch": 2.15, + "learning_rate": 3.929538999344549e-06, + "loss": 3.6134, + "step": 9849 + }, + { + "epoch": 2.15, + "learning_rate": 3.929429757483068e-06, + "loss": 2.8618, + "step": 9850 + }, + { + "epoch": 2.15, + "learning_rate": 3.9293205156215864e-06, + "loss": 3.7777, + "step": 9851 + }, + { + "epoch": 2.15, + "learning_rate": 3.9292112737601055e-06, + "loss": 1.9471, + "step": 9852 + }, + { + "epoch": 2.15, + "learning_rate": 3.929102031898624e-06, + "loss": 2.523, + "step": 9853 + }, + { + "epoch": 2.15, + "learning_rate": 3.928992790037143e-06, + "loss": 1.7273, + "step": 9854 + }, + { + "epoch": 2.15, + "learning_rate": 3.928883548175661e-06, + "loss": 2.9452, + "step": 9855 + }, + { + "epoch": 2.15, + "learning_rate": 3.92877430631418e-06, + "loss": 1.2156, + "step": 9856 + }, + { + "epoch": 2.15, + "learning_rate": 3.928665064452699e-06, + "loss": 3.0033, + "step": 9857 + }, + { + "epoch": 2.15, + "learning_rate": 3.928555822591217e-06, + "loss": 2.6488, + "step": 9858 + }, + { + "epoch": 2.15, + "learning_rate": 3.928446580729736e-06, + "loss": 2.948, + "step": 9859 + }, + { + "epoch": 2.15, + "learning_rate": 3.928337338868254e-06, + "loss": 2.7798, + "step": 9860 + }, + { + "epoch": 2.15, + "learning_rate": 3.928228097006773e-06, + "loss": 3.2648, + "step": 9861 + }, + { + "epoch": 2.15, + "learning_rate": 3.928118855145292e-06, + "loss": 1.9086, + "step": 9862 + }, + { + "epoch": 2.15, + "learning_rate": 3.928009613283811e-06, + "loss": 2.5061, + "step": 9863 + }, + { + "epoch": 2.15, + "learning_rate": 3.92790037142233e-06, + "loss": 1.9254, + "step": 9864 + }, + { + "epoch": 2.15, + "learning_rate": 3.927791129560848e-06, + "loss": 2.5016, + "step": 9865 + }, + { + "epoch": 2.15, + "learning_rate": 3.927681887699367e-06, + "loss": 2.0954, + "step": 9866 + }, + { + "epoch": 2.15, + "learning_rate": 3.927572645837885e-06, + "loss": 2.6156, + "step": 9867 + }, + { + "epoch": 2.15, + "learning_rate": 3.927463403976404e-06, + "loss": 2.6439, + "step": 9868 + }, + { + "epoch": 2.15, + "learning_rate": 3.927354162114922e-06, + "loss": 2.9424, + "step": 9869 + }, + { + "epoch": 2.15, + "learning_rate": 3.927244920253441e-06, + "loss": 1.3715, + "step": 9870 + }, + { + "epoch": 2.15, + "learning_rate": 3.92713567839196e-06, + "loss": 2.3792, + "step": 9871 + }, + { + "epoch": 2.15, + "learning_rate": 3.9270264365304785e-06, + "loss": 2.9899, + "step": 9872 + }, + { + "epoch": 2.15, + "learning_rate": 3.926917194668998e-06, + "loss": 2.6141, + "step": 9873 + }, + { + "epoch": 2.15, + "learning_rate": 3.926807952807516e-06, + "loss": 3.0131, + "step": 9874 + }, + { + "epoch": 2.16, + "learning_rate": 3.926698710946035e-06, + "loss": 2.719, + "step": 9875 + }, + { + "epoch": 2.16, + "learning_rate": 3.926589469084554e-06, + "loss": 2.2557, + "step": 9876 + }, + { + "epoch": 2.16, + "learning_rate": 3.926480227223072e-06, + "loss": 2.0245, + "step": 9877 + }, + { + "epoch": 2.16, + "learning_rate": 3.926370985361591e-06, + "loss": 2.3027, + "step": 9878 + }, + { + "epoch": 2.16, + "learning_rate": 3.926261743500109e-06, + "loss": 2.4865, + "step": 9879 + }, + { + "epoch": 2.16, + "learning_rate": 3.926152501638628e-06, + "loss": 3.9677, + "step": 9880 + }, + { + "epoch": 2.16, + "learning_rate": 3.926043259777147e-06, + "loss": 2.5492, + "step": 9881 + }, + { + "epoch": 2.16, + "learning_rate": 3.9259340179156655e-06, + "loss": 3.6268, + "step": 9882 + }, + { + "epoch": 2.16, + "learning_rate": 3.9258247760541845e-06, + "loss": 2.3174, + "step": 9883 + }, + { + "epoch": 2.16, + "learning_rate": 3.925715534192703e-06, + "loss": 1.8244, + "step": 9884 + }, + { + "epoch": 2.16, + "learning_rate": 3.925606292331222e-06, + "loss": 2.015, + "step": 9885 + }, + { + "epoch": 2.16, + "learning_rate": 3.925497050469741e-06, + "loss": 2.6857, + "step": 9886 + }, + { + "epoch": 2.16, + "learning_rate": 3.925387808608259e-06, + "loss": 2.2475, + "step": 9887 + }, + { + "epoch": 2.16, + "learning_rate": 3.925278566746777e-06, + "loss": 1.5674, + "step": 9888 + }, + { + "epoch": 2.16, + "learning_rate": 3.925169324885296e-06, + "loss": 1.8278, + "step": 9889 + }, + { + "epoch": 2.16, + "learning_rate": 3.925060083023815e-06, + "loss": 2.938, + "step": 9890 + }, + { + "epoch": 2.16, + "learning_rate": 3.9249508411623334e-06, + "loss": 2.3308, + "step": 9891 + }, + { + "epoch": 2.16, + "learning_rate": 3.9248415993008525e-06, + "loss": 2.0949, + "step": 9892 + }, + { + "epoch": 2.16, + "learning_rate": 3.924732357439371e-06, + "loss": 1.5349, + "step": 9893 + }, + { + "epoch": 2.16, + "learning_rate": 3.92462311557789e-06, + "loss": 3.4407, + "step": 9894 + }, + { + "epoch": 2.16, + "learning_rate": 3.924513873716409e-06, + "loss": 1.9112, + "step": 9895 + }, + { + "epoch": 2.16, + "learning_rate": 3.924404631854927e-06, + "loss": 2.6559, + "step": 9896 + }, + { + "epoch": 2.16, + "learning_rate": 3.924295389993446e-06, + "loss": 2.1879, + "step": 9897 + }, + { + "epoch": 2.16, + "learning_rate": 3.924186148131964e-06, + "loss": 2.6019, + "step": 9898 + }, + { + "epoch": 2.16, + "learning_rate": 3.924076906270483e-06, + "loss": 1.074, + "step": 9899 + }, + { + "epoch": 2.16, + "learning_rate": 3.923967664409002e-06, + "loss": 2.4374, + "step": 9900 + }, + { + "epoch": 2.16, + "learning_rate": 3.92385842254752e-06, + "loss": 2.4988, + "step": 9901 + }, + { + "epoch": 2.16, + "learning_rate": 3.9237491806860386e-06, + "loss": 3.1619, + "step": 9902 + }, + { + "epoch": 2.16, + "learning_rate": 3.923639938824558e-06, + "loss": 2.4432, + "step": 9903 + }, + { + "epoch": 2.16, + "learning_rate": 3.923530696963077e-06, + "loss": 1.9411, + "step": 9904 + }, + { + "epoch": 2.16, + "learning_rate": 3.923421455101596e-06, + "loss": 1.3903, + "step": 9905 + }, + { + "epoch": 2.16, + "learning_rate": 3.923312213240114e-06, + "loss": 2.5188, + "step": 9906 + }, + { + "epoch": 2.16, + "learning_rate": 3.923202971378632e-06, + "loss": 2.6029, + "step": 9907 + }, + { + "epoch": 2.16, + "learning_rate": 3.923093729517151e-06, + "loss": 1.6944, + "step": 9908 + }, + { + "epoch": 2.16, + "learning_rate": 3.92298448765567e-06, + "loss": 2.2018, + "step": 9909 + }, + { + "epoch": 2.16, + "learning_rate": 3.922875245794189e-06, + "loss": 2.9676, + "step": 9910 + }, + { + "epoch": 2.16, + "learning_rate": 3.922766003932707e-06, + "loss": 2.3044, + "step": 9911 + }, + { + "epoch": 2.16, + "learning_rate": 3.9226567620712255e-06, + "loss": 2.2786, + "step": 9912 + }, + { + "epoch": 2.16, + "learning_rate": 3.9225475202097446e-06, + "loss": 2.4237, + "step": 9913 + }, + { + "epoch": 2.16, + "learning_rate": 3.922438278348264e-06, + "loss": 2.434, + "step": 9914 + }, + { + "epoch": 2.16, + "learning_rate": 3.922329036486782e-06, + "loss": 1.9727, + "step": 9915 + }, + { + "epoch": 2.16, + "learning_rate": 3.922219794625301e-06, + "loss": 1.9043, + "step": 9916 + }, + { + "epoch": 2.16, + "learning_rate": 3.922110552763819e-06, + "loss": 2.2813, + "step": 9917 + }, + { + "epoch": 2.16, + "learning_rate": 3.922001310902338e-06, + "loss": 3.0838, + "step": 9918 + }, + { + "epoch": 2.16, + "learning_rate": 3.921892069040857e-06, + "loss": 3.2765, + "step": 9919 + }, + { + "epoch": 2.16, + "learning_rate": 3.921782827179375e-06, + "loss": 2.4536, + "step": 9920 + }, + { + "epoch": 2.17, + "learning_rate": 3.921673585317894e-06, + "loss": 1.775, + "step": 9921 + }, + { + "epoch": 2.17, + "learning_rate": 3.9215643434564125e-06, + "loss": 2.39, + "step": 9922 + }, + { + "epoch": 2.17, + "learning_rate": 3.9214551015949315e-06, + "loss": 1.3458, + "step": 9923 + }, + { + "epoch": 2.17, + "learning_rate": 3.9213458597334506e-06, + "loss": 2.7985, + "step": 9924 + }, + { + "epoch": 2.17, + "learning_rate": 3.921236617871969e-06, + "loss": 1.9701, + "step": 9925 + }, + { + "epoch": 2.17, + "learning_rate": 3.921127376010487e-06, + "loss": 2.8059, + "step": 9926 + }, + { + "epoch": 2.17, + "learning_rate": 3.921018134149006e-06, + "loss": 2.2539, + "step": 9927 + }, + { + "epoch": 2.17, + "learning_rate": 3.920908892287525e-06, + "loss": 2.551, + "step": 9928 + }, + { + "epoch": 2.17, + "learning_rate": 3.920799650426044e-06, + "loss": 3.0334, + "step": 9929 + }, + { + "epoch": 2.17, + "learning_rate": 3.920690408564562e-06, + "loss": 1.6087, + "step": 9930 + }, + { + "epoch": 2.17, + "learning_rate": 3.92058116670308e-06, + "loss": 2.0338, + "step": 9931 + }, + { + "epoch": 2.17, + "learning_rate": 3.9204719248415994e-06, + "loss": 2.5939, + "step": 9932 + }, + { + "epoch": 2.17, + "learning_rate": 3.9203626829801185e-06, + "loss": 1.6064, + "step": 9933 + }, + { + "epoch": 2.17, + "learning_rate": 3.9202534411186375e-06, + "loss": 3.1971, + "step": 9934 + }, + { + "epoch": 2.17, + "learning_rate": 3.920144199257156e-06, + "loss": 3.0564, + "step": 9935 + }, + { + "epoch": 2.17, + "learning_rate": 3.920034957395674e-06, + "loss": 2.4382, + "step": 9936 + }, + { + "epoch": 2.17, + "learning_rate": 3.919925715534193e-06, + "loss": 2.4243, + "step": 9937 + }, + { + "epoch": 2.17, + "learning_rate": 3.919816473672712e-06, + "loss": 2.059, + "step": 9938 + }, + { + "epoch": 2.17, + "learning_rate": 3.919707231811231e-06, + "loss": 2.3577, + "step": 9939 + }, + { + "epoch": 2.17, + "learning_rate": 3.919597989949749e-06, + "loss": 3.2164, + "step": 9940 + }, + { + "epoch": 2.17, + "learning_rate": 3.919488748088267e-06, + "loss": 2.4652, + "step": 9941 + }, + { + "epoch": 2.17, + "learning_rate": 3.919379506226786e-06, + "loss": 1.2778, + "step": 9942 + }, + { + "epoch": 2.17, + "learning_rate": 3.9192702643653054e-06, + "loss": 1.6776, + "step": 9943 + }, + { + "epoch": 2.17, + "learning_rate": 3.919161022503824e-06, + "loss": 2.7008, + "step": 9944 + }, + { + "epoch": 2.17, + "learning_rate": 3.919051780642343e-06, + "loss": 2.2711, + "step": 9945 + }, + { + "epoch": 2.17, + "learning_rate": 3.918942538780861e-06, + "loss": 1.8479, + "step": 9946 + }, + { + "epoch": 2.17, + "learning_rate": 3.91883329691938e-06, + "loss": 1.9353, + "step": 9947 + }, + { + "epoch": 2.17, + "learning_rate": 3.918724055057899e-06, + "loss": 2.0394, + "step": 9948 + }, + { + "epoch": 2.17, + "learning_rate": 3.918614813196417e-06, + "loss": 2.065, + "step": 9949 + }, + { + "epoch": 2.17, + "learning_rate": 3.918505571334935e-06, + "loss": 2.788, + "step": 9950 + }, + { + "epoch": 2.17, + "learning_rate": 3.918396329473454e-06, + "loss": 2.8746, + "step": 9951 + }, + { + "epoch": 2.17, + "learning_rate": 3.918287087611973e-06, + "loss": 2.5822, + "step": 9952 + }, + { + "epoch": 2.17, + "learning_rate": 3.918177845750492e-06, + "loss": 3.2607, + "step": 9953 + }, + { + "epoch": 2.17, + "learning_rate": 3.9180686038890106e-06, + "loss": 2.3055, + "step": 9954 + }, + { + "epoch": 2.17, + "learning_rate": 3.917959362027529e-06, + "loss": 1.9239, + "step": 9955 + }, + { + "epoch": 2.17, + "learning_rate": 3.917850120166048e-06, + "loss": 1.1353, + "step": 9956 + }, + { + "epoch": 2.17, + "learning_rate": 3.917740878304567e-06, + "loss": 2.6725, + "step": 9957 + }, + { + "epoch": 2.17, + "learning_rate": 3.917631636443086e-06, + "loss": 1.9272, + "step": 9958 + }, + { + "epoch": 2.17, + "learning_rate": 3.917522394581604e-06, + "loss": 1.9864, + "step": 9959 + }, + { + "epoch": 2.17, + "learning_rate": 3.917413152720122e-06, + "loss": 2.7503, + "step": 9960 + }, + { + "epoch": 2.17, + "learning_rate": 3.917303910858641e-06, + "loss": 3.4348, + "step": 9961 + }, + { + "epoch": 2.17, + "learning_rate": 3.91719466899716e-06, + "loss": 3.4161, + "step": 9962 + }, + { + "epoch": 2.17, + "learning_rate": 3.917085427135679e-06, + "loss": 2.2083, + "step": 9963 + }, + { + "epoch": 2.17, + "learning_rate": 3.9169761852741975e-06, + "loss": 1.7566, + "step": 9964 + }, + { + "epoch": 2.17, + "learning_rate": 3.916866943412716e-06, + "loss": 2.1075, + "step": 9965 + }, + { + "epoch": 2.18, + "learning_rate": 3.916757701551235e-06, + "loss": 2.2776, + "step": 9966 + }, + { + "epoch": 2.18, + "learning_rate": 3.916648459689754e-06, + "loss": 2.5604, + "step": 9967 + }, + { + "epoch": 2.18, + "learning_rate": 3.916539217828272e-06, + "loss": 2.107, + "step": 9968 + }, + { + "epoch": 2.18, + "learning_rate": 3.916429975966791e-06, + "loss": 2.1426, + "step": 9969 + }, + { + "epoch": 2.18, + "learning_rate": 3.916320734105309e-06, + "loss": 1.1505, + "step": 9970 + }, + { + "epoch": 2.18, + "learning_rate": 3.916211492243828e-06, + "loss": 3.3252, + "step": 9971 + }, + { + "epoch": 2.18, + "learning_rate": 3.916102250382347e-06, + "loss": 2.534, + "step": 9972 + }, + { + "epoch": 2.18, + "learning_rate": 3.9159930085208655e-06, + "loss": 3.1049, + "step": 9973 + }, + { + "epoch": 2.18, + "learning_rate": 3.9158837666593845e-06, + "loss": 2.774, + "step": 9974 + }, + { + "epoch": 2.18, + "learning_rate": 3.915774524797903e-06, + "loss": 3.2035, + "step": 9975 + }, + { + "epoch": 2.18, + "learning_rate": 3.915665282936422e-06, + "loss": 1.7118, + "step": 9976 + }, + { + "epoch": 2.18, + "learning_rate": 3.915556041074941e-06, + "loss": 2.3321, + "step": 9977 + }, + { + "epoch": 2.18, + "learning_rate": 3.915446799213459e-06, + "loss": 3.5847, + "step": 9978 + }, + { + "epoch": 2.18, + "learning_rate": 3.915337557351977e-06, + "loss": 3.1109, + "step": 9979 + }, + { + "epoch": 2.18, + "learning_rate": 3.915228315490496e-06, + "loss": 1.5574, + "step": 9980 + }, + { + "epoch": 2.18, + "learning_rate": 3.915119073629015e-06, + "loss": 2.7846, + "step": 9981 + }, + { + "epoch": 2.18, + "learning_rate": 3.915009831767534e-06, + "loss": 2.0492, + "step": 9982 + }, + { + "epoch": 2.18, + "learning_rate": 3.914900589906052e-06, + "loss": 1.3423, + "step": 9983 + }, + { + "epoch": 2.18, + "learning_rate": 3.914791348044571e-06, + "loss": 2.8126, + "step": 9984 + }, + { + "epoch": 2.18, + "learning_rate": 3.91468210618309e-06, + "loss": 3.1578, + "step": 9985 + }, + { + "epoch": 2.18, + "learning_rate": 3.914572864321609e-06, + "loss": 3.1954, + "step": 9986 + }, + { + "epoch": 2.18, + "learning_rate": 3.914463622460127e-06, + "loss": 2.3649, + "step": 9987 + }, + { + "epoch": 2.18, + "learning_rate": 3.914354380598646e-06, + "loss": 2.4891, + "step": 9988 + }, + { + "epoch": 2.18, + "learning_rate": 3.914245138737164e-06, + "loss": 2.0222, + "step": 9989 + }, + { + "epoch": 2.18, + "learning_rate": 3.914135896875683e-06, + "loss": 1.601, + "step": 9990 + }, + { + "epoch": 2.18, + "learning_rate": 3.914026655014202e-06, + "loss": 2.101, + "step": 9991 + }, + { + "epoch": 2.18, + "learning_rate": 3.91391741315272e-06, + "loss": 2.3783, + "step": 9992 + }, + { + "epoch": 2.18, + "learning_rate": 3.913808171291239e-06, + "loss": 2.8188, + "step": 9993 + }, + { + "epoch": 2.18, + "learning_rate": 3.9136989294297576e-06, + "loss": 3.6967, + "step": 9994 + }, + { + "epoch": 2.18, + "learning_rate": 3.913589687568277e-06, + "loss": 1.4593, + "step": 9995 + }, + { + "epoch": 2.18, + "learning_rate": 3.913480445706796e-06, + "loss": 2.5112, + "step": 9996 + }, + { + "epoch": 2.18, + "learning_rate": 3.913371203845314e-06, + "loss": 0.7581, + "step": 9997 + }, + { + "epoch": 2.18, + "learning_rate": 3.913261961983833e-06, + "loss": 1.935, + "step": 9998 + }, + { + "epoch": 2.18, + "learning_rate": 3.913152720122351e-06, + "loss": 2.5214, + "step": 9999 + }, + { + "epoch": 2.18, + "learning_rate": 3.91304347826087e-06, + "loss": 1.6796, + "step": 10000 + }, + { + "epoch": 2.18, + "learning_rate": 3.912934236399388e-06, + "loss": 3.2101, + "step": 10001 + }, + { + "epoch": 2.18, + "learning_rate": 3.912824994537907e-06, + "loss": 1.9244, + "step": 10002 + }, + { + "epoch": 2.18, + "learning_rate": 3.9127157526764255e-06, + "loss": 2.8677, + "step": 10003 + }, + { + "epoch": 2.18, + "learning_rate": 3.9126065108149445e-06, + "loss": 1.7782, + "step": 10004 + }, + { + "epoch": 2.18, + "learning_rate": 3.9124972689534635e-06, + "loss": 3.7754, + "step": 10005 + }, + { + "epoch": 2.18, + "learning_rate": 3.912388027091982e-06, + "loss": 2.7901, + "step": 10006 + }, + { + "epoch": 2.18, + "learning_rate": 3.912278785230501e-06, + "loss": 2.9691, + "step": 10007 + }, + { + "epoch": 2.18, + "learning_rate": 3.912169543369019e-06, + "loss": 2.7857, + "step": 10008 + }, + { + "epoch": 2.18, + "learning_rate": 3.912060301507538e-06, + "loss": 1.5192, + "step": 10009 + }, + { + "epoch": 2.18, + "learning_rate": 3.911951059646057e-06, + "loss": 2.4967, + "step": 10010 + }, + { + "epoch": 2.18, + "learning_rate": 3.911841817784575e-06, + "loss": 3.0071, + "step": 10011 + }, + { + "epoch": 2.19, + "learning_rate": 3.911732575923094e-06, + "loss": 2.2935, + "step": 10012 + }, + { + "epoch": 2.19, + "learning_rate": 3.9116233340616124e-06, + "loss": 2.8782, + "step": 10013 + }, + { + "epoch": 2.19, + "learning_rate": 3.9115140922001315e-06, + "loss": 2.2239, + "step": 10014 + }, + { + "epoch": 2.19, + "learning_rate": 3.9114048503386505e-06, + "loss": 2.1382, + "step": 10015 + }, + { + "epoch": 2.19, + "learning_rate": 3.911295608477169e-06, + "loss": 2.4471, + "step": 10016 + }, + { + "epoch": 2.19, + "learning_rate": 3.911186366615688e-06, + "loss": 2.5753, + "step": 10017 + }, + { + "epoch": 2.19, + "learning_rate": 3.911077124754206e-06, + "loss": 1.8242, + "step": 10018 + }, + { + "epoch": 2.19, + "learning_rate": 3.910967882892725e-06, + "loss": 2.0799, + "step": 10019 + }, + { + "epoch": 2.19, + "learning_rate": 3.910858641031243e-06, + "loss": 2.392, + "step": 10020 + }, + { + "epoch": 2.19, + "learning_rate": 3.910749399169762e-06, + "loss": 3.2335, + "step": 10021 + }, + { + "epoch": 2.19, + "learning_rate": 3.910640157308281e-06, + "loss": 2.3673, + "step": 10022 + }, + { + "epoch": 2.19, + "learning_rate": 3.910530915446799e-06, + "loss": 2.0618, + "step": 10023 + }, + { + "epoch": 2.19, + "learning_rate": 3.9104216735853184e-06, + "loss": 2.3244, + "step": 10024 + }, + { + "epoch": 2.19, + "learning_rate": 3.910312431723837e-06, + "loss": 2.852, + "step": 10025 + }, + { + "epoch": 2.19, + "learning_rate": 3.910203189862356e-06, + "loss": 2.1601, + "step": 10026 + }, + { + "epoch": 2.19, + "learning_rate": 3.910093948000875e-06, + "loss": 2.0638, + "step": 10027 + }, + { + "epoch": 2.19, + "learning_rate": 3.909984706139393e-06, + "loss": 2.4868, + "step": 10028 + }, + { + "epoch": 2.19, + "learning_rate": 3.909875464277912e-06, + "loss": 1.721, + "step": 10029 + }, + { + "epoch": 2.19, + "learning_rate": 3.90976622241643e-06, + "loss": 1.9026, + "step": 10030 + }, + { + "epoch": 2.19, + "learning_rate": 3.909656980554949e-06, + "loss": 2.4886, + "step": 10031 + }, + { + "epoch": 2.19, + "learning_rate": 3.909547738693467e-06, + "loss": 3.0644, + "step": 10032 + }, + { + "epoch": 2.19, + "learning_rate": 3.909438496831986e-06, + "loss": 3.2273, + "step": 10033 + }, + { + "epoch": 2.19, + "learning_rate": 3.9093292549705045e-06, + "loss": 2.8049, + "step": 10034 + }, + { + "epoch": 2.19, + "learning_rate": 3.9092200131090236e-06, + "loss": 1.4535, + "step": 10035 + }, + { + "epoch": 2.19, + "learning_rate": 3.909110771247543e-06, + "loss": 1.986, + "step": 10036 + }, + { + "epoch": 2.19, + "learning_rate": 3.909001529386061e-06, + "loss": 2.1874, + "step": 10037 + }, + { + "epoch": 2.19, + "learning_rate": 3.90889228752458e-06, + "loss": 2.7008, + "step": 10038 + }, + { + "epoch": 2.19, + "learning_rate": 3.908783045663098e-06, + "loss": 1.959, + "step": 10039 + }, + { + "epoch": 2.19, + "learning_rate": 3.908673803801617e-06, + "loss": 1.8697, + "step": 10040 + }, + { + "epoch": 2.19, + "learning_rate": 3.908564561940136e-06, + "loss": 2.6122, + "step": 10041 + }, + { + "epoch": 2.19, + "learning_rate": 3.908455320078654e-06, + "loss": 3.12, + "step": 10042 + }, + { + "epoch": 2.19, + "learning_rate": 3.908346078217173e-06, + "loss": 3.5202, + "step": 10043 + }, + { + "epoch": 2.19, + "learning_rate": 3.9082368363556915e-06, + "loss": 2.6505, + "step": 10044 + }, + { + "epoch": 2.19, + "learning_rate": 3.9081275944942105e-06, + "loss": 3.0077, + "step": 10045 + }, + { + "epoch": 2.19, + "learning_rate": 3.9080183526327296e-06, + "loss": 2.3959, + "step": 10046 + }, + { + "epoch": 2.19, + "learning_rate": 3.907909110771248e-06, + "loss": 2.5181, + "step": 10047 + }, + { + "epoch": 2.19, + "learning_rate": 3.907799868909767e-06, + "loss": 2.5851, + "step": 10048 + }, + { + "epoch": 2.19, + "learning_rate": 3.907690627048285e-06, + "loss": 2.1562, + "step": 10049 + }, + { + "epoch": 2.19, + "learning_rate": 3.907581385186804e-06, + "loss": 3.0931, + "step": 10050 + }, + { + "epoch": 2.19, + "learning_rate": 3.907472143325323e-06, + "loss": 2.5399, + "step": 10051 + }, + { + "epoch": 2.19, + "learning_rate": 3.907362901463841e-06, + "loss": 2.0631, + "step": 10052 + }, + { + "epoch": 2.19, + "learning_rate": 3.907253659602359e-06, + "loss": 2.1824, + "step": 10053 + }, + { + "epoch": 2.19, + "learning_rate": 3.9071444177408785e-06, + "loss": 1.624, + "step": 10054 + }, + { + "epoch": 2.19, + "learning_rate": 3.9070351758793975e-06, + "loss": 2.5974, + "step": 10055 + }, + { + "epoch": 2.19, + "learning_rate": 3.906925934017916e-06, + "loss": 2.2196, + "step": 10056 + }, + { + "epoch": 2.19, + "learning_rate": 3.906816692156435e-06, + "loss": 2.5211, + "step": 10057 + }, + { + "epoch": 2.2, + "learning_rate": 3.906707450294953e-06, + "loss": 2.7817, + "step": 10058 + }, + { + "epoch": 2.2, + "learning_rate": 3.906598208433472e-06, + "loss": 2.0639, + "step": 10059 + }, + { + "epoch": 2.2, + "learning_rate": 3.906488966571991e-06, + "loss": 2.2274, + "step": 10060 + }, + { + "epoch": 2.2, + "learning_rate": 3.906379724710509e-06, + "loss": 2.1992, + "step": 10061 + }, + { + "epoch": 2.2, + "learning_rate": 3.906270482849028e-06, + "loss": 1.9552, + "step": 10062 + }, + { + "epoch": 2.2, + "learning_rate": 3.906161240987546e-06, + "loss": 2.4543, + "step": 10063 + }, + { + "epoch": 2.2, + "learning_rate": 3.906051999126065e-06, + "loss": 2.4079, + "step": 10064 + }, + { + "epoch": 2.2, + "learning_rate": 3.9059427572645844e-06, + "loss": 1.3617, + "step": 10065 + }, + { + "epoch": 2.2, + "learning_rate": 3.905833515403103e-06, + "loss": 1.4816, + "step": 10066 + }, + { + "epoch": 2.2, + "learning_rate": 3.905724273541621e-06, + "loss": 2.5821, + "step": 10067 + }, + { + "epoch": 2.2, + "learning_rate": 3.90561503168014e-06, + "loss": 2.4075, + "step": 10068 + }, + { + "epoch": 2.2, + "learning_rate": 3.905505789818659e-06, + "loss": 2.2781, + "step": 10069 + }, + { + "epoch": 2.2, + "learning_rate": 3.905396547957178e-06, + "loss": 2.8144, + "step": 10070 + }, + { + "epoch": 2.2, + "learning_rate": 3.905287306095696e-06, + "loss": 3.6173, + "step": 10071 + }, + { + "epoch": 2.2, + "learning_rate": 3.905178064234214e-06, + "loss": 2.6356, + "step": 10072 + }, + { + "epoch": 2.2, + "learning_rate": 3.905068822372733e-06, + "loss": 3.2416, + "step": 10073 + }, + { + "epoch": 2.2, + "learning_rate": 3.904959580511252e-06, + "loss": 1.2078, + "step": 10074 + }, + { + "epoch": 2.2, + "learning_rate": 3.904850338649771e-06, + "loss": 2.4056, + "step": 10075 + }, + { + "epoch": 2.2, + "learning_rate": 3.90474109678829e-06, + "loss": 1.7213, + "step": 10076 + }, + { + "epoch": 2.2, + "learning_rate": 3.904631854926808e-06, + "loss": 2.2078, + "step": 10077 + }, + { + "epoch": 2.2, + "learning_rate": 3.904522613065327e-06, + "loss": 2.6939, + "step": 10078 + }, + { + "epoch": 2.2, + "learning_rate": 3.904413371203846e-06, + "loss": 2.8006, + "step": 10079 + }, + { + "epoch": 2.2, + "learning_rate": 3.904304129342365e-06, + "loss": 3.3194, + "step": 10080 + }, + { + "epoch": 2.2, + "learning_rate": 3.904194887480883e-06, + "loss": 1.611, + "step": 10081 + }, + { + "epoch": 2.2, + "learning_rate": 3.904085645619401e-06, + "loss": 1.1855, + "step": 10082 + }, + { + "epoch": 2.2, + "learning_rate": 3.90397640375792e-06, + "loss": 2.1251, + "step": 10083 + }, + { + "epoch": 2.2, + "learning_rate": 3.903867161896439e-06, + "loss": 2.2955, + "step": 10084 + }, + { + "epoch": 2.2, + "learning_rate": 3.9037579200349575e-06, + "loss": 3.3527, + "step": 10085 + }, + { + "epoch": 2.2, + "learning_rate": 3.9036486781734765e-06, + "loss": 3.1718, + "step": 10086 + }, + { + "epoch": 2.2, + "learning_rate": 3.903539436311995e-06, + "loss": 1.8633, + "step": 10087 + }, + { + "epoch": 2.2, + "learning_rate": 3.903430194450514e-06, + "loss": 2.571, + "step": 10088 + }, + { + "epoch": 2.2, + "learning_rate": 3.903320952589033e-06, + "loss": 2.804, + "step": 10089 + }, + { + "epoch": 2.2, + "learning_rate": 3.903211710727551e-06, + "loss": 2.192, + "step": 10090 + }, + { + "epoch": 2.2, + "learning_rate": 3.903102468866069e-06, + "loss": 3.0358, + "step": 10091 + }, + { + "epoch": 2.2, + "learning_rate": 3.902993227004588e-06, + "loss": 1.7536, + "step": 10092 + }, + { + "epoch": 2.2, + "learning_rate": 3.902883985143107e-06, + "loss": 2.187, + "step": 10093 + }, + { + "epoch": 2.2, + "learning_rate": 3.902774743281626e-06, + "loss": 1.8809, + "step": 10094 + }, + { + "epoch": 2.2, + "learning_rate": 3.9026655014201445e-06, + "loss": 3.0547, + "step": 10095 + }, + { + "epoch": 2.2, + "learning_rate": 3.902556259558663e-06, + "loss": 3.069, + "step": 10096 + }, + { + "epoch": 2.2, + "learning_rate": 3.902447017697182e-06, + "loss": 2.6234, + "step": 10097 + }, + { + "epoch": 2.2, + "learning_rate": 3.902337775835701e-06, + "loss": 2.0322, + "step": 10098 + }, + { + "epoch": 2.2, + "learning_rate": 3.90222853397422e-06, + "loss": 3.8567, + "step": 10099 + }, + { + "epoch": 2.2, + "learning_rate": 3.902119292112738e-06, + "loss": 1.9467, + "step": 10100 + }, + { + "epoch": 2.2, + "learning_rate": 3.902010050251256e-06, + "loss": 1.2508, + "step": 10101 + }, + { + "epoch": 2.2, + "learning_rate": 3.901900808389775e-06, + "loss": 2.8168, + "step": 10102 + }, + { + "epoch": 2.2, + "learning_rate": 3.901791566528294e-06, + "loss": 1.6561, + "step": 10103 + }, + { + "epoch": 2.21, + "learning_rate": 3.901682324666813e-06, + "loss": 2.611, + "step": 10104 + }, + { + "epoch": 2.21, + "learning_rate": 3.901573082805331e-06, + "loss": 1.871, + "step": 10105 + }, + { + "epoch": 2.21, + "learning_rate": 3.90146384094385e-06, + "loss": 2.7189, + "step": 10106 + }, + { + "epoch": 2.21, + "learning_rate": 3.901354599082369e-06, + "loss": 2.6186, + "step": 10107 + }, + { + "epoch": 2.21, + "learning_rate": 3.901245357220888e-06, + "loss": 2.4027, + "step": 10108 + }, + { + "epoch": 2.21, + "learning_rate": 3.901136115359406e-06, + "loss": 2.7571, + "step": 10109 + }, + { + "epoch": 2.21, + "learning_rate": 3.901026873497925e-06, + "loss": 3.639, + "step": 10110 + }, + { + "epoch": 2.21, + "learning_rate": 3.900917631636443e-06, + "loss": 3.0617, + "step": 10111 + }, + { + "epoch": 2.21, + "learning_rate": 3.900808389774962e-06, + "loss": 2.0123, + "step": 10112 + }, + { + "epoch": 2.21, + "learning_rate": 3.900699147913481e-06, + "loss": 2.3849, + "step": 10113 + }, + { + "epoch": 2.21, + "learning_rate": 3.900589906051999e-06, + "loss": 1.9086, + "step": 10114 + }, + { + "epoch": 2.21, + "learning_rate": 3.900480664190518e-06, + "loss": 2.5021, + "step": 10115 + }, + { + "epoch": 2.21, + "learning_rate": 3.9003714223290366e-06, + "loss": 2.0199, + "step": 10116 + }, + { + "epoch": 2.21, + "learning_rate": 3.900262180467556e-06, + "loss": 1.8392, + "step": 10117 + }, + { + "epoch": 2.21, + "learning_rate": 3.900152938606075e-06, + "loss": 1.874, + "step": 10118 + }, + { + "epoch": 2.21, + "learning_rate": 3.900043696744593e-06, + "loss": 2.2019, + "step": 10119 + }, + { + "epoch": 2.21, + "learning_rate": 3.899934454883111e-06, + "loss": 3.2006, + "step": 10120 + }, + { + "epoch": 2.21, + "learning_rate": 3.89982521302163e-06, + "loss": 2.0104, + "step": 10121 + }, + { + "epoch": 2.21, + "learning_rate": 3.899715971160149e-06, + "loss": 4.2587, + "step": 10122 + }, + { + "epoch": 2.21, + "learning_rate": 3.899606729298668e-06, + "loss": 1.8768, + "step": 10123 + }, + { + "epoch": 2.21, + "learning_rate": 3.899497487437186e-06, + "loss": 2.6043, + "step": 10124 + }, + { + "epoch": 2.21, + "learning_rate": 3.8993882455757045e-06, + "loss": 3.7466, + "step": 10125 + }, + { + "epoch": 2.21, + "learning_rate": 3.8992790037142235e-06, + "loss": 2.2822, + "step": 10126 + }, + { + "epoch": 2.21, + "learning_rate": 3.8991697618527426e-06, + "loss": 2.3031, + "step": 10127 + }, + { + "epoch": 2.21, + "learning_rate": 3.899060519991262e-06, + "loss": 3.2403, + "step": 10128 + }, + { + "epoch": 2.21, + "learning_rate": 3.89895127812978e-06, + "loss": 2.616, + "step": 10129 + }, + { + "epoch": 2.21, + "learning_rate": 3.898842036268298e-06, + "loss": 1.8108, + "step": 10130 + }, + { + "epoch": 2.21, + "learning_rate": 3.898732794406817e-06, + "loss": 1.7196, + "step": 10131 + }, + { + "epoch": 2.21, + "learning_rate": 3.898623552545336e-06, + "loss": 2.7434, + "step": 10132 + }, + { + "epoch": 2.21, + "learning_rate": 3.898514310683854e-06, + "loss": 2.1618, + "step": 10133 + }, + { + "epoch": 2.21, + "learning_rate": 3.898405068822373e-06, + "loss": 3.2043, + "step": 10134 + }, + { + "epoch": 2.21, + "learning_rate": 3.8982958269608914e-06, + "loss": 2.5231, + "step": 10135 + }, + { + "epoch": 2.21, + "learning_rate": 3.8981865850994105e-06, + "loss": 2.6459, + "step": 10136 + }, + { + "epoch": 2.21, + "learning_rate": 3.8980773432379295e-06, + "loss": 3.5362, + "step": 10137 + }, + { + "epoch": 2.21, + "learning_rate": 3.897968101376448e-06, + "loss": 3.5419, + "step": 10138 + }, + { + "epoch": 2.21, + "learning_rate": 3.897858859514967e-06, + "loss": 2.4178, + "step": 10139 + }, + { + "epoch": 2.21, + "learning_rate": 3.897749617653485e-06, + "loss": 2.5757, + "step": 10140 + }, + { + "epoch": 2.21, + "learning_rate": 3.897640375792004e-06, + "loss": 3.2281, + "step": 10141 + }, + { + "epoch": 2.21, + "learning_rate": 3.897531133930523e-06, + "loss": 3.821, + "step": 10142 + }, + { + "epoch": 2.21, + "learning_rate": 3.897421892069041e-06, + "loss": 2.739, + "step": 10143 + }, + { + "epoch": 2.21, + "learning_rate": 3.897312650207559e-06, + "loss": 2.7236, + "step": 10144 + }, + { + "epoch": 2.21, + "learning_rate": 3.897203408346078e-06, + "loss": 2.0933, + "step": 10145 + }, + { + "epoch": 2.21, + "learning_rate": 3.8970941664845974e-06, + "loss": 1.5871, + "step": 10146 + }, + { + "epoch": 2.21, + "learning_rate": 3.8969849246231165e-06, + "loss": 3.2346, + "step": 10147 + }, + { + "epoch": 2.21, + "learning_rate": 3.896875682761635e-06, + "loss": 2.6299, + "step": 10148 + }, + { + "epoch": 2.21, + "learning_rate": 3.896766440900153e-06, + "loss": 2.6372, + "step": 10149 + }, + { + "epoch": 2.22, + "learning_rate": 3.896657199038672e-06, + "loss": 3.1489, + "step": 10150 + }, + { + "epoch": 2.22, + "learning_rate": 3.896547957177191e-06, + "loss": 0.7998, + "step": 10151 + }, + { + "epoch": 2.22, + "learning_rate": 3.896438715315709e-06, + "loss": 2.8222, + "step": 10152 + }, + { + "epoch": 2.22, + "learning_rate": 3.896329473454228e-06, + "loss": 3.02, + "step": 10153 + }, + { + "epoch": 2.22, + "learning_rate": 3.896220231592746e-06, + "loss": 2.8914, + "step": 10154 + }, + { + "epoch": 2.22, + "learning_rate": 3.896110989731265e-06, + "loss": 2.3172, + "step": 10155 + }, + { + "epoch": 2.22, + "learning_rate": 3.896001747869784e-06, + "loss": 2.1497, + "step": 10156 + }, + { + "epoch": 2.22, + "learning_rate": 3.895892506008303e-06, + "loss": 3.256, + "step": 10157 + }, + { + "epoch": 2.22, + "learning_rate": 3.895783264146822e-06, + "loss": 2.5317, + "step": 10158 + }, + { + "epoch": 2.22, + "learning_rate": 3.89567402228534e-06, + "loss": 2.6837, + "step": 10159 + }, + { + "epoch": 2.22, + "learning_rate": 3.895564780423859e-06, + "loss": 3.3794, + "step": 10160 + }, + { + "epoch": 2.22, + "learning_rate": 3.895455538562378e-06, + "loss": 2.6692, + "step": 10161 + }, + { + "epoch": 2.22, + "learning_rate": 3.895346296700896e-06, + "loss": 2.76, + "step": 10162 + }, + { + "epoch": 2.22, + "learning_rate": 3.895237054839415e-06, + "loss": 2.2666, + "step": 10163 + }, + { + "epoch": 2.22, + "learning_rate": 3.895127812977933e-06, + "loss": 2.693, + "step": 10164 + }, + { + "epoch": 2.22, + "learning_rate": 3.895018571116452e-06, + "loss": 2.2923, + "step": 10165 + }, + { + "epoch": 2.22, + "learning_rate": 3.8949093292549705e-06, + "loss": 1.705, + "step": 10166 + }, + { + "epoch": 2.22, + "learning_rate": 3.8948000873934895e-06, + "loss": 3.2394, + "step": 10167 + }, + { + "epoch": 2.22, + "learning_rate": 3.8946908455320086e-06, + "loss": 2.9569, + "step": 10168 + }, + { + "epoch": 2.22, + "learning_rate": 3.894581603670527e-06, + "loss": 2.078, + "step": 10169 + }, + { + "epoch": 2.22, + "learning_rate": 3.894472361809046e-06, + "loss": 3.1978, + "step": 10170 + }, + { + "epoch": 2.22, + "learning_rate": 3.894363119947564e-06, + "loss": 3.4003, + "step": 10171 + }, + { + "epoch": 2.22, + "learning_rate": 3.894253878086083e-06, + "loss": 1.8219, + "step": 10172 + }, + { + "epoch": 2.22, + "learning_rate": 3.894144636224601e-06, + "loss": 2.2661, + "step": 10173 + }, + { + "epoch": 2.22, + "learning_rate": 3.89403539436312e-06, + "loss": 2.4716, + "step": 10174 + }, + { + "epoch": 2.22, + "learning_rate": 3.893926152501639e-06, + "loss": 3.3037, + "step": 10175 + }, + { + "epoch": 2.22, + "learning_rate": 3.8938169106401575e-06, + "loss": 2.9533, + "step": 10176 + }, + { + "epoch": 2.22, + "learning_rate": 3.8937076687786765e-06, + "loss": 2.8316, + "step": 10177 + }, + { + "epoch": 2.22, + "learning_rate": 3.893598426917195e-06, + "loss": 2.7728, + "step": 10178 + }, + { + "epoch": 2.22, + "learning_rate": 3.893489185055714e-06, + "loss": 3.2515, + "step": 10179 + }, + { + "epoch": 2.22, + "learning_rate": 3.893379943194233e-06, + "loss": 1.6928, + "step": 10180 + }, + { + "epoch": 2.22, + "learning_rate": 3.893270701332751e-06, + "loss": 3.3227, + "step": 10181 + }, + { + "epoch": 2.22, + "learning_rate": 3.89316145947127e-06, + "loss": 1.5529, + "step": 10182 + }, + { + "epoch": 2.22, + "learning_rate": 3.893052217609788e-06, + "loss": 2.2247, + "step": 10183 + }, + { + "epoch": 2.22, + "learning_rate": 3.892942975748307e-06, + "loss": 2.2715, + "step": 10184 + }, + { + "epoch": 2.22, + "learning_rate": 3.892833733886825e-06, + "loss": 2.3697, + "step": 10185 + }, + { + "epoch": 2.22, + "learning_rate": 3.892724492025344e-06, + "loss": 1.5119, + "step": 10186 + }, + { + "epoch": 2.22, + "learning_rate": 3.8926152501638635e-06, + "loss": 1.7684, + "step": 10187 + }, + { + "epoch": 2.22, + "learning_rate": 3.892506008302382e-06, + "loss": 3.21, + "step": 10188 + }, + { + "epoch": 2.22, + "learning_rate": 3.892396766440901e-06, + "loss": 2.8731, + "step": 10189 + }, + { + "epoch": 2.22, + "learning_rate": 3.892287524579419e-06, + "loss": 3.672, + "step": 10190 + }, + { + "epoch": 2.22, + "learning_rate": 3.892178282717938e-06, + "loss": 3.679, + "step": 10191 + }, + { + "epoch": 2.22, + "learning_rate": 3.892069040856457e-06, + "loss": 1.4121, + "step": 10192 + }, + { + "epoch": 2.22, + "learning_rate": 3.891959798994975e-06, + "loss": 2.9441, + "step": 10193 + }, + { + "epoch": 2.22, + "learning_rate": 3.891850557133494e-06, + "loss": 2.0199, + "step": 10194 + }, + { + "epoch": 2.23, + "learning_rate": 3.891741315272012e-06, + "loss": 2.0554, + "step": 10195 + }, + { + "epoch": 2.23, + "learning_rate": 3.891632073410531e-06, + "loss": 3.0935, + "step": 10196 + }, + { + "epoch": 2.23, + "learning_rate": 3.8915228315490496e-06, + "loss": 2.1525, + "step": 10197 + }, + { + "epoch": 2.23, + "learning_rate": 3.891413589687569e-06, + "loss": 2.609, + "step": 10198 + }, + { + "epoch": 2.23, + "learning_rate": 3.891304347826087e-06, + "loss": 2.1751, + "step": 10199 + }, + { + "epoch": 2.23, + "learning_rate": 3.891195105964606e-06, + "loss": 3.5777, + "step": 10200 + }, + { + "epoch": 2.23, + "learning_rate": 3.891085864103125e-06, + "loss": 2.7837, + "step": 10201 + }, + { + "epoch": 2.23, + "learning_rate": 3.890976622241643e-06, + "loss": 2.9495, + "step": 10202 + }, + { + "epoch": 2.23, + "learning_rate": 3.890867380380162e-06, + "loss": 3.7329, + "step": 10203 + }, + { + "epoch": 2.23, + "learning_rate": 3.89075813851868e-06, + "loss": 2.7083, + "step": 10204 + }, + { + "epoch": 2.23, + "learning_rate": 3.890648896657199e-06, + "loss": 2.4735, + "step": 10205 + }, + { + "epoch": 2.23, + "learning_rate": 3.890539654795718e-06, + "loss": 2.855, + "step": 10206 + }, + { + "epoch": 2.23, + "learning_rate": 3.8904304129342365e-06, + "loss": 2.4959, + "step": 10207 + }, + { + "epoch": 2.23, + "learning_rate": 3.8903211710727556e-06, + "loss": 2.0364, + "step": 10208 + }, + { + "epoch": 2.23, + "learning_rate": 3.890211929211274e-06, + "loss": 1.927, + "step": 10209 + }, + { + "epoch": 2.23, + "learning_rate": 3.890102687349793e-06, + "loss": 2.9082, + "step": 10210 + }, + { + "epoch": 2.23, + "learning_rate": 3.889993445488312e-06, + "loss": 3.2254, + "step": 10211 + }, + { + "epoch": 2.23, + "learning_rate": 3.88988420362683e-06, + "loss": 3.5466, + "step": 10212 + }, + { + "epoch": 2.23, + "learning_rate": 3.889774961765348e-06, + "loss": 2.5754, + "step": 10213 + }, + { + "epoch": 2.23, + "learning_rate": 3.889665719903867e-06, + "loss": 1.9297, + "step": 10214 + }, + { + "epoch": 2.23, + "learning_rate": 3.889556478042386e-06, + "loss": 2.3917, + "step": 10215 + }, + { + "epoch": 2.23, + "learning_rate": 3.889447236180905e-06, + "loss": 3.399, + "step": 10216 + }, + { + "epoch": 2.23, + "learning_rate": 3.8893379943194235e-06, + "loss": 2.8506, + "step": 10217 + }, + { + "epoch": 2.23, + "learning_rate": 3.889228752457942e-06, + "loss": 3.0506, + "step": 10218 + }, + { + "epoch": 2.23, + "learning_rate": 3.889119510596461e-06, + "loss": 2.3499, + "step": 10219 + }, + { + "epoch": 2.23, + "learning_rate": 3.88901026873498e-06, + "loss": 2.4563, + "step": 10220 + }, + { + "epoch": 2.23, + "learning_rate": 3.888901026873499e-06, + "loss": 2.5543, + "step": 10221 + }, + { + "epoch": 2.23, + "learning_rate": 3.888791785012017e-06, + "loss": 3.4329, + "step": 10222 + }, + { + "epoch": 2.23, + "learning_rate": 3.888682543150535e-06, + "loss": 2.6723, + "step": 10223 + }, + { + "epoch": 2.23, + "learning_rate": 3.888573301289054e-06, + "loss": 2.0422, + "step": 10224 + }, + { + "epoch": 2.23, + "learning_rate": 3.888464059427573e-06, + "loss": 2.1576, + "step": 10225 + }, + { + "epoch": 2.23, + "learning_rate": 3.888354817566091e-06, + "loss": 2.3801, + "step": 10226 + }, + { + "epoch": 2.23, + "learning_rate": 3.8882455757046104e-06, + "loss": 1.9592, + "step": 10227 + }, + { + "epoch": 2.23, + "learning_rate": 3.888136333843129e-06, + "loss": 2.4683, + "step": 10228 + }, + { + "epoch": 2.23, + "learning_rate": 3.888027091981648e-06, + "loss": 2.8089, + "step": 10229 + }, + { + "epoch": 2.23, + "learning_rate": 3.887917850120167e-06, + "loss": 2.5354, + "step": 10230 + }, + { + "epoch": 2.23, + "learning_rate": 3.887808608258685e-06, + "loss": 2.9132, + "step": 10231 + }, + { + "epoch": 2.23, + "learning_rate": 3.887699366397203e-06, + "loss": 2.1966, + "step": 10232 + }, + { + "epoch": 2.23, + "learning_rate": 3.887590124535722e-06, + "loss": 3.3261, + "step": 10233 + }, + { + "epoch": 2.23, + "learning_rate": 3.887480882674241e-06, + "loss": 2.2431, + "step": 10234 + }, + { + "epoch": 2.23, + "learning_rate": 3.88737164081276e-06, + "loss": 2.9269, + "step": 10235 + }, + { + "epoch": 2.23, + "learning_rate": 3.887262398951278e-06, + "loss": 1.4804, + "step": 10236 + }, + { + "epoch": 2.23, + "learning_rate": 3.8871531570897965e-06, + "loss": 1.2693, + "step": 10237 + }, + { + "epoch": 2.23, + "learning_rate": 3.887043915228316e-06, + "loss": 2.8516, + "step": 10238 + }, + { + "epoch": 2.23, + "learning_rate": 3.886934673366835e-06, + "loss": 3.0898, + "step": 10239 + }, + { + "epoch": 2.23, + "learning_rate": 3.886825431505354e-06, + "loss": 1.9572, + "step": 10240 + }, + { + "epoch": 2.24, + "learning_rate": 3.886716189643872e-06, + "loss": 1.5149, + "step": 10241 + }, + { + "epoch": 2.24, + "learning_rate": 3.88660694778239e-06, + "loss": 2.8623, + "step": 10242 + }, + { + "epoch": 2.24, + "learning_rate": 3.886497705920909e-06, + "loss": 3.266, + "step": 10243 + }, + { + "epoch": 2.24, + "learning_rate": 3.886388464059428e-06, + "loss": 2.117, + "step": 10244 + }, + { + "epoch": 2.24, + "learning_rate": 3.886279222197947e-06, + "loss": 2.7509, + "step": 10245 + }, + { + "epoch": 2.24, + "learning_rate": 3.886169980336465e-06, + "loss": 2.6923, + "step": 10246 + }, + { + "epoch": 2.24, + "learning_rate": 3.8860607384749835e-06, + "loss": 2.9, + "step": 10247 + }, + { + "epoch": 2.24, + "learning_rate": 3.8859514966135025e-06, + "loss": 2.1537, + "step": 10248 + }, + { + "epoch": 2.24, + "learning_rate": 3.8858422547520216e-06, + "loss": 3.1792, + "step": 10249 + }, + { + "epoch": 2.24, + "learning_rate": 3.88573301289054e-06, + "loss": 2.3957, + "step": 10250 + }, + { + "epoch": 2.24, + "learning_rate": 3.885623771029059e-06, + "loss": 2.584, + "step": 10251 + }, + { + "epoch": 2.24, + "learning_rate": 3.885514529167577e-06, + "loss": 2.6362, + "step": 10252 + }, + { + "epoch": 2.24, + "learning_rate": 3.885405287306096e-06, + "loss": 2.7921, + "step": 10253 + }, + { + "epoch": 2.24, + "learning_rate": 3.885296045444615e-06, + "loss": 2.004, + "step": 10254 + }, + { + "epoch": 2.24, + "learning_rate": 3.885186803583133e-06, + "loss": 1.5652, + "step": 10255 + }, + { + "epoch": 2.24, + "learning_rate": 3.885077561721652e-06, + "loss": 2.4145, + "step": 10256 + }, + { + "epoch": 2.24, + "learning_rate": 3.8849683198601705e-06, + "loss": 2.8874, + "step": 10257 + }, + { + "epoch": 2.24, + "learning_rate": 3.8848590779986895e-06, + "loss": 1.7845, + "step": 10258 + }, + { + "epoch": 2.24, + "learning_rate": 3.8847498361372085e-06, + "loss": 3.2102, + "step": 10259 + }, + { + "epoch": 2.24, + "learning_rate": 3.884640594275727e-06, + "loss": 2.8279, + "step": 10260 + }, + { + "epoch": 2.24, + "learning_rate": 3.884531352414245e-06, + "loss": 3.0502, + "step": 10261 + }, + { + "epoch": 2.24, + "learning_rate": 3.884422110552764e-06, + "loss": 2.8112, + "step": 10262 + }, + { + "epoch": 2.24, + "learning_rate": 3.884312868691283e-06, + "loss": 3.1136, + "step": 10263 + }, + { + "epoch": 2.24, + "learning_rate": 3.884203626829802e-06, + "loss": 3.1864, + "step": 10264 + }, + { + "epoch": 2.24, + "learning_rate": 3.88409438496832e-06, + "loss": 2.7763, + "step": 10265 + }, + { + "epoch": 2.24, + "learning_rate": 3.883985143106838e-06, + "loss": 1.8089, + "step": 10266 + }, + { + "epoch": 2.24, + "learning_rate": 3.883875901245357e-06, + "loss": 1.7708, + "step": 10267 + }, + { + "epoch": 2.24, + "learning_rate": 3.8837666593838764e-06, + "loss": 3.1513, + "step": 10268 + }, + { + "epoch": 2.24, + "learning_rate": 3.8836574175223955e-06, + "loss": 1.5631, + "step": 10269 + }, + { + "epoch": 2.24, + "learning_rate": 3.883548175660914e-06, + "loss": 2.8527, + "step": 10270 + }, + { + "epoch": 2.24, + "learning_rate": 3.883438933799432e-06, + "loss": 2.5139, + "step": 10271 + }, + { + "epoch": 2.24, + "learning_rate": 3.883329691937951e-06, + "loss": 1.7953, + "step": 10272 + }, + { + "epoch": 2.24, + "learning_rate": 3.88322045007647e-06, + "loss": 3.3899, + "step": 10273 + }, + { + "epoch": 2.24, + "learning_rate": 3.883111208214989e-06, + "loss": 2.491, + "step": 10274 + }, + { + "epoch": 2.24, + "learning_rate": 3.883001966353507e-06, + "loss": 2.6944, + "step": 10275 + }, + { + "epoch": 2.24, + "learning_rate": 3.882892724492025e-06, + "loss": 3.3447, + "step": 10276 + }, + { + "epoch": 2.24, + "learning_rate": 3.882783482630544e-06, + "loss": 2.7907, + "step": 10277 + }, + { + "epoch": 2.24, + "learning_rate": 3.882674240769063e-06, + "loss": 2.1082, + "step": 10278 + }, + { + "epoch": 2.24, + "learning_rate": 3.882564998907582e-06, + "loss": 1.957, + "step": 10279 + }, + { + "epoch": 2.24, + "learning_rate": 3.882455757046101e-06, + "loss": 2.936, + "step": 10280 + }, + { + "epoch": 2.24, + "learning_rate": 3.882346515184619e-06, + "loss": 1.9175, + "step": 10281 + }, + { + "epoch": 2.24, + "learning_rate": 3.882237273323138e-06, + "loss": 2.7209, + "step": 10282 + }, + { + "epoch": 2.24, + "learning_rate": 3.882128031461657e-06, + "loss": 2.6035, + "step": 10283 + }, + { + "epoch": 2.24, + "learning_rate": 3.882018789600175e-06, + "loss": 3.4658, + "step": 10284 + }, + { + "epoch": 2.24, + "learning_rate": 3.881909547738693e-06, + "loss": 3.1138, + "step": 10285 + }, + { + "epoch": 2.24, + "learning_rate": 3.881800305877212e-06, + "loss": 1.9874, + "step": 10286 + }, + { + "epoch": 2.25, + "learning_rate": 3.881691064015731e-06, + "loss": 2.0775, + "step": 10287 + }, + { + "epoch": 2.25, + "learning_rate": 3.88158182215425e-06, + "loss": 2.2, + "step": 10288 + }, + { + "epoch": 2.25, + "learning_rate": 3.8814725802927686e-06, + "loss": 3.3057, + "step": 10289 + }, + { + "epoch": 2.25, + "learning_rate": 3.881363338431287e-06, + "loss": 2.6899, + "step": 10290 + }, + { + "epoch": 2.25, + "learning_rate": 3.881254096569806e-06, + "loss": 3.0849, + "step": 10291 + }, + { + "epoch": 2.25, + "learning_rate": 3.881144854708325e-06, + "loss": 2.5003, + "step": 10292 + }, + { + "epoch": 2.25, + "learning_rate": 3.881035612846844e-06, + "loss": 3.4208, + "step": 10293 + }, + { + "epoch": 2.25, + "learning_rate": 3.880926370985362e-06, + "loss": 2.5597, + "step": 10294 + }, + { + "epoch": 2.25, + "learning_rate": 3.88081712912388e-06, + "loss": 1.8603, + "step": 10295 + }, + { + "epoch": 2.25, + "learning_rate": 3.880707887262399e-06, + "loss": 0.9239, + "step": 10296 + }, + { + "epoch": 2.25, + "learning_rate": 3.880598645400918e-06, + "loss": 2.5165, + "step": 10297 + }, + { + "epoch": 2.25, + "learning_rate": 3.8804894035394365e-06, + "loss": 2.7348, + "step": 10298 + }, + { + "epoch": 2.25, + "learning_rate": 3.8803801616779555e-06, + "loss": 2.5446, + "step": 10299 + }, + { + "epoch": 2.25, + "learning_rate": 3.880270919816474e-06, + "loss": 1.9911, + "step": 10300 + }, + { + "epoch": 2.25, + "learning_rate": 3.880161677954993e-06, + "loss": 2.3863, + "step": 10301 + }, + { + "epoch": 2.25, + "learning_rate": 3.880052436093512e-06, + "loss": 1.8907, + "step": 10302 + }, + { + "epoch": 2.25, + "learning_rate": 3.87994319423203e-06, + "loss": 2.8727, + "step": 10303 + }, + { + "epoch": 2.25, + "learning_rate": 3.879833952370549e-06, + "loss": 2.3624, + "step": 10304 + }, + { + "epoch": 2.25, + "learning_rate": 3.879724710509067e-06, + "loss": 2.5069, + "step": 10305 + }, + { + "epoch": 2.25, + "learning_rate": 3.879615468647586e-06, + "loss": 2.6698, + "step": 10306 + }, + { + "epoch": 2.25, + "learning_rate": 3.879506226786105e-06, + "loss": 3.173, + "step": 10307 + }, + { + "epoch": 2.25, + "learning_rate": 3.8793969849246234e-06, + "loss": 2.6334, + "step": 10308 + }, + { + "epoch": 2.25, + "learning_rate": 3.8792877430631425e-06, + "loss": 1.9577, + "step": 10309 + }, + { + "epoch": 2.25, + "learning_rate": 3.879178501201661e-06, + "loss": 3.0398, + "step": 10310 + }, + { + "epoch": 2.25, + "learning_rate": 3.87906925934018e-06, + "loss": 2.446, + "step": 10311 + }, + { + "epoch": 2.25, + "learning_rate": 3.878960017478698e-06, + "loss": 3.0123, + "step": 10312 + }, + { + "epoch": 2.25, + "learning_rate": 3.878850775617217e-06, + "loss": 0.7498, + "step": 10313 + }, + { + "epoch": 2.25, + "learning_rate": 3.878741533755735e-06, + "loss": 1.9009, + "step": 10314 + }, + { + "epoch": 2.25, + "learning_rate": 3.878632291894254e-06, + "loss": 2.0755, + "step": 10315 + }, + { + "epoch": 2.25, + "learning_rate": 3.878523050032773e-06, + "loss": 3.4152, + "step": 10316 + }, + { + "epoch": 2.25, + "learning_rate": 3.878413808171291e-06, + "loss": 2.3221, + "step": 10317 + }, + { + "epoch": 2.25, + "learning_rate": 3.87830456630981e-06, + "loss": 2.755, + "step": 10318 + }, + { + "epoch": 2.25, + "learning_rate": 3.8781953244483286e-06, + "loss": 2.2856, + "step": 10319 + }, + { + "epoch": 2.25, + "learning_rate": 3.878086082586848e-06, + "loss": 3.6602, + "step": 10320 + }, + { + "epoch": 2.25, + "learning_rate": 3.877976840725367e-06, + "loss": 1.7302, + "step": 10321 + }, + { + "epoch": 2.25, + "learning_rate": 3.877867598863885e-06, + "loss": 2.0996, + "step": 10322 + }, + { + "epoch": 2.25, + "learning_rate": 3.877758357002404e-06, + "loss": 2.4547, + "step": 10323 + }, + { + "epoch": 2.25, + "learning_rate": 3.877649115140922e-06, + "loss": 1.6724, + "step": 10324 + }, + { + "epoch": 2.25, + "learning_rate": 3.877539873279441e-06, + "loss": 2.7453, + "step": 10325 + }, + { + "epoch": 2.25, + "learning_rate": 3.87743063141796e-06, + "loss": 1.798, + "step": 10326 + }, + { + "epoch": 2.25, + "learning_rate": 3.877321389556478e-06, + "loss": 1.9158, + "step": 10327 + }, + { + "epoch": 2.25, + "learning_rate": 3.877212147694997e-06, + "loss": 1.3787, + "step": 10328 + }, + { + "epoch": 2.25, + "learning_rate": 3.8771029058335155e-06, + "loss": 2.8686, + "step": 10329 + }, + { + "epoch": 2.25, + "learning_rate": 3.8769936639720346e-06, + "loss": 2.371, + "step": 10330 + }, + { + "epoch": 2.25, + "learning_rate": 3.876884422110553e-06, + "loss": 1.9224, + "step": 10331 + }, + { + "epoch": 2.25, + "learning_rate": 3.876775180249072e-06, + "loss": 2.1655, + "step": 10332 + }, + { + "epoch": 2.26, + "learning_rate": 3.876665938387591e-06, + "loss": 2.3243, + "step": 10333 + }, + { + "epoch": 2.26, + "learning_rate": 3.876556696526109e-06, + "loss": 1.7727, + "step": 10334 + }, + { + "epoch": 2.26, + "learning_rate": 3.876447454664628e-06, + "loss": 2.8785, + "step": 10335 + }, + { + "epoch": 2.26, + "learning_rate": 3.876338212803146e-06, + "loss": 3.6608, + "step": 10336 + }, + { + "epoch": 2.26, + "learning_rate": 3.876228970941665e-06, + "loss": 2.3691, + "step": 10337 + }, + { + "epoch": 2.26, + "learning_rate": 3.8761197290801835e-06, + "loss": 1.5839, + "step": 10338 + }, + { + "epoch": 2.26, + "learning_rate": 3.8760104872187025e-06, + "loss": 2.8543, + "step": 10339 + }, + { + "epoch": 2.26, + "learning_rate": 3.8759012453572215e-06, + "loss": 2.0262, + "step": 10340 + }, + { + "epoch": 2.26, + "learning_rate": 3.87579200349574e-06, + "loss": 1.4747, + "step": 10341 + }, + { + "epoch": 2.26, + "learning_rate": 3.875682761634259e-06, + "loss": 2.1516, + "step": 10342 + }, + { + "epoch": 2.26, + "learning_rate": 3.875573519772777e-06, + "loss": 2.6811, + "step": 10343 + }, + { + "epoch": 2.26, + "learning_rate": 3.875464277911296e-06, + "loss": 1.8785, + "step": 10344 + }, + { + "epoch": 2.26, + "learning_rate": 3.875355036049814e-06, + "loss": 2.027, + "step": 10345 + }, + { + "epoch": 2.26, + "learning_rate": 3.875245794188333e-06, + "loss": 2.4839, + "step": 10346 + }, + { + "epoch": 2.26, + "learning_rate": 3.875136552326852e-06, + "loss": 2.8789, + "step": 10347 + }, + { + "epoch": 2.26, + "learning_rate": 3.87502731046537e-06, + "loss": 2.3713, + "step": 10348 + }, + { + "epoch": 2.26, + "learning_rate": 3.8749180686038894e-06, + "loss": 3.101, + "step": 10349 + }, + { + "epoch": 2.26, + "learning_rate": 3.874808826742408e-06, + "loss": 2.6823, + "step": 10350 + }, + { + "epoch": 2.26, + "learning_rate": 3.874699584880927e-06, + "loss": 2.3223, + "step": 10351 + }, + { + "epoch": 2.26, + "learning_rate": 3.874590343019446e-06, + "loss": 2.805, + "step": 10352 + }, + { + "epoch": 2.26, + "learning_rate": 3.874481101157964e-06, + "loss": 2.6718, + "step": 10353 + }, + { + "epoch": 2.26, + "learning_rate": 3.874371859296483e-06, + "loss": 2.096, + "step": 10354 + }, + { + "epoch": 2.26, + "learning_rate": 3.874262617435001e-06, + "loss": 1.2715, + "step": 10355 + }, + { + "epoch": 2.26, + "learning_rate": 3.87415337557352e-06, + "loss": 3.0328, + "step": 10356 + }, + { + "epoch": 2.26, + "learning_rate": 3.874044133712039e-06, + "loss": 2.7917, + "step": 10357 + }, + { + "epoch": 2.26, + "learning_rate": 3.873934891850557e-06, + "loss": 1.5146, + "step": 10358 + }, + { + "epoch": 2.26, + "learning_rate": 3.873825649989076e-06, + "loss": 3.4399, + "step": 10359 + }, + { + "epoch": 2.26, + "learning_rate": 3.873716408127595e-06, + "loss": 1.6685, + "step": 10360 + }, + { + "epoch": 2.26, + "learning_rate": 3.873607166266114e-06, + "loss": 1.762, + "step": 10361 + }, + { + "epoch": 2.26, + "learning_rate": 3.873497924404633e-06, + "loss": 2.525, + "step": 10362 + }, + { + "epoch": 2.26, + "learning_rate": 3.873388682543151e-06, + "loss": 2.6684, + "step": 10363 + }, + { + "epoch": 2.26, + "learning_rate": 3.873279440681669e-06, + "loss": 2.5272, + "step": 10364 + }, + { + "epoch": 2.26, + "learning_rate": 3.873170198820188e-06, + "loss": 2.3597, + "step": 10365 + }, + { + "epoch": 2.26, + "learning_rate": 3.873060956958707e-06, + "loss": 2.0681, + "step": 10366 + }, + { + "epoch": 2.26, + "learning_rate": 3.872951715097225e-06, + "loss": 1.4834, + "step": 10367 + }, + { + "epoch": 2.26, + "learning_rate": 3.872842473235744e-06, + "loss": 2.2526, + "step": 10368 + }, + { + "epoch": 2.26, + "learning_rate": 3.8727332313742625e-06, + "loss": 1.8516, + "step": 10369 + }, + { + "epoch": 2.26, + "learning_rate": 3.8726239895127815e-06, + "loss": 1.6124, + "step": 10370 + }, + { + "epoch": 2.26, + "learning_rate": 3.872514747651301e-06, + "loss": 1.7705, + "step": 10371 + }, + { + "epoch": 2.26, + "learning_rate": 3.872405505789819e-06, + "loss": 3.2676, + "step": 10372 + }, + { + "epoch": 2.26, + "learning_rate": 3.872296263928338e-06, + "loss": 2.08, + "step": 10373 + }, + { + "epoch": 2.26, + "learning_rate": 3.872187022066856e-06, + "loss": 1.8313, + "step": 10374 + }, + { + "epoch": 2.26, + "learning_rate": 3.872077780205375e-06, + "loss": 3.303, + "step": 10375 + }, + { + "epoch": 2.26, + "learning_rate": 3.871968538343894e-06, + "loss": 3.8505, + "step": 10376 + }, + { + "epoch": 2.26, + "learning_rate": 3.871859296482412e-06, + "loss": 1.9124, + "step": 10377 + }, + { + "epoch": 2.26, + "learning_rate": 3.8717500546209304e-06, + "loss": 1.3795, + "step": 10378 + }, + { + "epoch": 2.27, + "learning_rate": 3.8716408127594495e-06, + "loss": 3.0782, + "step": 10379 + }, + { + "epoch": 2.27, + "learning_rate": 3.8715315708979685e-06, + "loss": 2.808, + "step": 10380 + }, + { + "epoch": 2.27, + "learning_rate": 3.8714223290364875e-06, + "loss": 2.6648, + "step": 10381 + }, + { + "epoch": 2.27, + "learning_rate": 3.871313087175006e-06, + "loss": 2.6143, + "step": 10382 + }, + { + "epoch": 2.27, + "learning_rate": 3.871203845313524e-06, + "loss": 2.0895, + "step": 10383 + }, + { + "epoch": 2.27, + "learning_rate": 3.871094603452043e-06, + "loss": 2.854, + "step": 10384 + }, + { + "epoch": 2.27, + "learning_rate": 3.870985361590562e-06, + "loss": 2.1253, + "step": 10385 + }, + { + "epoch": 2.27, + "learning_rate": 3.870876119729081e-06, + "loss": 2.3124, + "step": 10386 + }, + { + "epoch": 2.27, + "learning_rate": 3.870766877867599e-06, + "loss": 3.3307, + "step": 10387 + }, + { + "epoch": 2.27, + "learning_rate": 3.870657636006117e-06, + "loss": 2.0616, + "step": 10388 + }, + { + "epoch": 2.27, + "learning_rate": 3.8705483941446364e-06, + "loss": 2.4989, + "step": 10389 + }, + { + "epoch": 2.27, + "learning_rate": 3.8704391522831555e-06, + "loss": 2.4379, + "step": 10390 + }, + { + "epoch": 2.27, + "learning_rate": 3.870329910421674e-06, + "loss": 1.7669, + "step": 10391 + }, + { + "epoch": 2.27, + "learning_rate": 3.870220668560193e-06, + "loss": 2.172, + "step": 10392 + }, + { + "epoch": 2.27, + "learning_rate": 3.870111426698711e-06, + "loss": 2.6159, + "step": 10393 + }, + { + "epoch": 2.27, + "learning_rate": 3.87000218483723e-06, + "loss": 1.6784, + "step": 10394 + }, + { + "epoch": 2.27, + "learning_rate": 3.869892942975749e-06, + "loss": 1.4067, + "step": 10395 + }, + { + "epoch": 2.27, + "learning_rate": 3.869783701114267e-06, + "loss": 2.314, + "step": 10396 + }, + { + "epoch": 2.27, + "learning_rate": 3.869674459252785e-06, + "loss": 1.9544, + "step": 10397 + }, + { + "epoch": 2.27, + "learning_rate": 3.869565217391304e-06, + "loss": 1.6728, + "step": 10398 + }, + { + "epoch": 2.27, + "learning_rate": 3.869455975529823e-06, + "loss": 2.6976, + "step": 10399 + }, + { + "epoch": 2.27, + "learning_rate": 3.869346733668342e-06, + "loss": 2.8758, + "step": 10400 + }, + { + "epoch": 2.27, + "learning_rate": 3.869237491806861e-06, + "loss": 2.916, + "step": 10401 + }, + { + "epoch": 2.27, + "learning_rate": 3.869128249945379e-06, + "loss": 3.007, + "step": 10402 + }, + { + "epoch": 2.27, + "learning_rate": 3.869019008083898e-06, + "loss": 1.7157, + "step": 10403 + }, + { + "epoch": 2.27, + "learning_rate": 3.868909766222417e-06, + "loss": 3.1309, + "step": 10404 + }, + { + "epoch": 2.27, + "learning_rate": 3.868800524360936e-06, + "loss": 2.5822, + "step": 10405 + }, + { + "epoch": 2.27, + "learning_rate": 3.868691282499454e-06, + "loss": 1.0352, + "step": 10406 + }, + { + "epoch": 2.27, + "learning_rate": 3.868582040637972e-06, + "loss": 2.3533, + "step": 10407 + }, + { + "epoch": 2.27, + "learning_rate": 3.868472798776491e-06, + "loss": 2.7894, + "step": 10408 + }, + { + "epoch": 2.27, + "learning_rate": 3.86836355691501e-06, + "loss": 1.8688, + "step": 10409 + }, + { + "epoch": 2.27, + "learning_rate": 3.868254315053529e-06, + "loss": 2.7719, + "step": 10410 + }, + { + "epoch": 2.27, + "learning_rate": 3.8681450731920476e-06, + "loss": 3.8211, + "step": 10411 + }, + { + "epoch": 2.27, + "learning_rate": 3.868035831330566e-06, + "loss": 3.535, + "step": 10412 + }, + { + "epoch": 2.27, + "learning_rate": 3.867926589469085e-06, + "loss": 3.6318, + "step": 10413 + }, + { + "epoch": 2.27, + "learning_rate": 3.867817347607604e-06, + "loss": 2.7234, + "step": 10414 + }, + { + "epoch": 2.27, + "learning_rate": 3.867708105746123e-06, + "loss": 2.2939, + "step": 10415 + }, + { + "epoch": 2.27, + "learning_rate": 3.867598863884641e-06, + "loss": 2.0347, + "step": 10416 + }, + { + "epoch": 2.27, + "learning_rate": 3.867489622023159e-06, + "loss": 3.6924, + "step": 10417 + }, + { + "epoch": 2.27, + "learning_rate": 3.867380380161678e-06, + "loss": 3.0617, + "step": 10418 + }, + { + "epoch": 2.27, + "learning_rate": 3.867271138300197e-06, + "loss": 1.8185, + "step": 10419 + }, + { + "epoch": 2.27, + "learning_rate": 3.8671618964387155e-06, + "loss": 2.1378, + "step": 10420 + }, + { + "epoch": 2.27, + "learning_rate": 3.8670526545772345e-06, + "loss": 2.9349, + "step": 10421 + }, + { + "epoch": 2.27, + "learning_rate": 3.866943412715753e-06, + "loss": 2.2261, + "step": 10422 + }, + { + "epoch": 2.27, + "learning_rate": 3.866834170854272e-06, + "loss": 2.4153, + "step": 10423 + }, + { + "epoch": 2.27, + "learning_rate": 3.866724928992791e-06, + "loss": 2.7993, + "step": 10424 + }, + { + "epoch": 2.28, + "learning_rate": 3.866615687131309e-06, + "loss": 2.0983, + "step": 10425 + }, + { + "epoch": 2.28, + "learning_rate": 3.866506445269827e-06, + "loss": 2.0774, + "step": 10426 + }, + { + "epoch": 2.28, + "learning_rate": 3.866397203408346e-06, + "loss": 3.308, + "step": 10427 + }, + { + "epoch": 2.28, + "learning_rate": 3.866287961546865e-06, + "loss": 3.7897, + "step": 10428 + }, + { + "epoch": 2.28, + "learning_rate": 3.866178719685384e-06, + "loss": 2.4605, + "step": 10429 + }, + { + "epoch": 2.28, + "learning_rate": 3.8660694778239024e-06, + "loss": 1.7174, + "step": 10430 + }, + { + "epoch": 2.28, + "learning_rate": 3.865960235962421e-06, + "loss": 1.1484, + "step": 10431 + }, + { + "epoch": 2.28, + "learning_rate": 3.86585099410094e-06, + "loss": 1.7409, + "step": 10432 + }, + { + "epoch": 2.28, + "learning_rate": 3.865741752239459e-06, + "loss": 1.8303, + "step": 10433 + }, + { + "epoch": 2.28, + "learning_rate": 3.865632510377978e-06, + "loss": 2.8967, + "step": 10434 + }, + { + "epoch": 2.28, + "learning_rate": 3.865523268516496e-06, + "loss": 1.8538, + "step": 10435 + }, + { + "epoch": 2.28, + "learning_rate": 3.865414026655014e-06, + "loss": 3.6157, + "step": 10436 + }, + { + "epoch": 2.28, + "learning_rate": 3.865304784793533e-06, + "loss": 2.379, + "step": 10437 + }, + { + "epoch": 2.28, + "learning_rate": 3.865195542932052e-06, + "loss": 2.5525, + "step": 10438 + }, + { + "epoch": 2.28, + "learning_rate": 3.865086301070571e-06, + "loss": 2.0271, + "step": 10439 + }, + { + "epoch": 2.28, + "learning_rate": 3.864977059209089e-06, + "loss": 2.3317, + "step": 10440 + }, + { + "epoch": 2.28, + "learning_rate": 3.864867817347608e-06, + "loss": 1.7181, + "step": 10441 + }, + { + "epoch": 2.28, + "learning_rate": 3.864758575486127e-06, + "loss": 2.82, + "step": 10442 + }, + { + "epoch": 2.28, + "learning_rate": 3.864649333624646e-06, + "loss": 3.0917, + "step": 10443 + }, + { + "epoch": 2.28, + "learning_rate": 3.864540091763164e-06, + "loss": 2.6377, + "step": 10444 + }, + { + "epoch": 2.28, + "learning_rate": 3.864430849901683e-06, + "loss": 3.0518, + "step": 10445 + }, + { + "epoch": 2.28, + "learning_rate": 3.864321608040201e-06, + "loss": 1.2488, + "step": 10446 + }, + { + "epoch": 2.28, + "learning_rate": 3.86421236617872e-06, + "loss": 0.7441, + "step": 10447 + }, + { + "epoch": 2.28, + "learning_rate": 3.864103124317239e-06, + "loss": 1.9728, + "step": 10448 + }, + { + "epoch": 2.28, + "learning_rate": 3.863993882455757e-06, + "loss": 2.0668, + "step": 10449 + }, + { + "epoch": 2.28, + "learning_rate": 3.8638846405942755e-06, + "loss": 2.8707, + "step": 10450 + }, + { + "epoch": 2.28, + "learning_rate": 3.8637753987327945e-06, + "loss": 3.833, + "step": 10451 + }, + { + "epoch": 2.28, + "learning_rate": 3.8636661568713136e-06, + "loss": 2.4923, + "step": 10452 + }, + { + "epoch": 2.28, + "learning_rate": 3.863556915009833e-06, + "loss": 2.6326, + "step": 10453 + }, + { + "epoch": 2.28, + "learning_rate": 3.863447673148351e-06, + "loss": 3.865, + "step": 10454 + }, + { + "epoch": 2.28, + "learning_rate": 3.863338431286869e-06, + "loss": 1.6, + "step": 10455 + }, + { + "epoch": 2.28, + "learning_rate": 3.863229189425388e-06, + "loss": 2.5175, + "step": 10456 + }, + { + "epoch": 2.28, + "learning_rate": 3.863119947563907e-06, + "loss": 2.1658, + "step": 10457 + }, + { + "epoch": 2.28, + "learning_rate": 3.863010705702426e-06, + "loss": 2.5793, + "step": 10458 + }, + { + "epoch": 2.28, + "learning_rate": 3.862901463840944e-06, + "loss": 2.4721, + "step": 10459 + }, + { + "epoch": 2.28, + "learning_rate": 3.8627922219794625e-06, + "loss": 2.7334, + "step": 10460 + }, + { + "epoch": 2.28, + "learning_rate": 3.8626829801179815e-06, + "loss": 2.3893, + "step": 10461 + }, + { + "epoch": 2.28, + "learning_rate": 3.8625737382565005e-06, + "loss": 2.4146, + "step": 10462 + }, + { + "epoch": 2.28, + "learning_rate": 3.862464496395019e-06, + "loss": 2.657, + "step": 10463 + }, + { + "epoch": 2.28, + "learning_rate": 3.862355254533538e-06, + "loss": 2.8127, + "step": 10464 + }, + { + "epoch": 2.28, + "learning_rate": 3.862246012672056e-06, + "loss": 3.3861, + "step": 10465 + }, + { + "epoch": 2.28, + "learning_rate": 3.862136770810575e-06, + "loss": 2.8778, + "step": 10466 + }, + { + "epoch": 2.28, + "learning_rate": 3.862027528949094e-06, + "loss": 2.4944, + "step": 10467 + }, + { + "epoch": 2.28, + "learning_rate": 3.861918287087612e-06, + "loss": 2.3595, + "step": 10468 + }, + { + "epoch": 2.28, + "learning_rate": 3.861809045226131e-06, + "loss": 2.327, + "step": 10469 + }, + { + "epoch": 2.29, + "learning_rate": 3.861699803364649e-06, + "loss": 1.8642, + "step": 10470 + }, + { + "epoch": 2.29, + "learning_rate": 3.8615905615031685e-06, + "loss": 2.8367, + "step": 10471 + }, + { + "epoch": 2.29, + "learning_rate": 3.8614813196416875e-06, + "loss": 2.2447, + "step": 10472 + }, + { + "epoch": 2.29, + "learning_rate": 3.861372077780206e-06, + "loss": 2.427, + "step": 10473 + }, + { + "epoch": 2.29, + "learning_rate": 3.861262835918725e-06, + "loss": 1.9814, + "step": 10474 + }, + { + "epoch": 2.29, + "learning_rate": 3.861153594057243e-06, + "loss": 2.4948, + "step": 10475 + }, + { + "epoch": 2.29, + "learning_rate": 3.861044352195762e-06, + "loss": 2.001, + "step": 10476 + }, + { + "epoch": 2.29, + "learning_rate": 3.86093511033428e-06, + "loss": 1.7354, + "step": 10477 + }, + { + "epoch": 2.29, + "learning_rate": 3.860825868472799e-06, + "loss": 4.1884, + "step": 10478 + }, + { + "epoch": 2.29, + "learning_rate": 3.860716626611317e-06, + "loss": 3.3319, + "step": 10479 + }, + { + "epoch": 2.29, + "learning_rate": 3.860607384749836e-06, + "loss": 2.3344, + "step": 10480 + }, + { + "epoch": 2.29, + "learning_rate": 3.860498142888355e-06, + "loss": 3.3851, + "step": 10481 + }, + { + "epoch": 2.29, + "learning_rate": 3.860388901026874e-06, + "loss": 2.1265, + "step": 10482 + }, + { + "epoch": 2.29, + "learning_rate": 3.860279659165393e-06, + "loss": 2.9398, + "step": 10483 + }, + { + "epoch": 2.29, + "learning_rate": 3.860170417303911e-06, + "loss": 1.8693, + "step": 10484 + }, + { + "epoch": 2.29, + "learning_rate": 3.86006117544243e-06, + "loss": 1.5851, + "step": 10485 + }, + { + "epoch": 2.29, + "learning_rate": 3.859951933580949e-06, + "loss": 2.2027, + "step": 10486 + }, + { + "epoch": 2.29, + "learning_rate": 3.859842691719467e-06, + "loss": 2.2749, + "step": 10487 + }, + { + "epoch": 2.29, + "learning_rate": 3.859733449857986e-06, + "loss": 3.3094, + "step": 10488 + }, + { + "epoch": 2.29, + "learning_rate": 3.859624207996504e-06, + "loss": 3.0231, + "step": 10489 + }, + { + "epoch": 2.29, + "learning_rate": 3.859514966135023e-06, + "loss": 1.6378, + "step": 10490 + }, + { + "epoch": 2.29, + "learning_rate": 3.859405724273542e-06, + "loss": 1.7685, + "step": 10491 + }, + { + "epoch": 2.29, + "learning_rate": 3.8592964824120606e-06, + "loss": 2.8072, + "step": 10492 + }, + { + "epoch": 2.29, + "learning_rate": 3.85918724055058e-06, + "loss": 2.1531, + "step": 10493 + }, + { + "epoch": 2.29, + "learning_rate": 3.859077998689098e-06, + "loss": 2.3628, + "step": 10494 + }, + { + "epoch": 2.29, + "learning_rate": 3.858968756827617e-06, + "loss": 2.0629, + "step": 10495 + }, + { + "epoch": 2.29, + "learning_rate": 3.858859514966135e-06, + "loss": 1.6476, + "step": 10496 + }, + { + "epoch": 2.29, + "learning_rate": 3.858750273104654e-06, + "loss": 3.228, + "step": 10497 + }, + { + "epoch": 2.29, + "learning_rate": 3.858641031243173e-06, + "loss": 2.4853, + "step": 10498 + }, + { + "epoch": 2.29, + "learning_rate": 3.858531789381691e-06, + "loss": 1.8485, + "step": 10499 + }, + { + "epoch": 2.29, + "learning_rate": 3.85842254752021e-06, + "loss": 1.7622, + "step": 10500 + }, + { + "epoch": 2.29, + "learning_rate": 3.8583133056587285e-06, + "loss": 3.5962, + "step": 10501 + }, + { + "epoch": 2.29, + "learning_rate": 3.8582040637972475e-06, + "loss": 1.956, + "step": 10502 + }, + { + "epoch": 2.29, + "learning_rate": 3.858094821935766e-06, + "loss": 1.6988, + "step": 10503 + }, + { + "epoch": 2.29, + "learning_rate": 3.857985580074285e-06, + "loss": 3.3066, + "step": 10504 + }, + { + "epoch": 2.29, + "learning_rate": 3.857876338212804e-06, + "loss": 0.9657, + "step": 10505 + }, + { + "epoch": 2.29, + "learning_rate": 3.857767096351322e-06, + "loss": 3.5032, + "step": 10506 + }, + { + "epoch": 2.29, + "learning_rate": 3.857657854489841e-06, + "loss": 1.6451, + "step": 10507 + }, + { + "epoch": 2.29, + "learning_rate": 3.857548612628359e-06, + "loss": 3.1938, + "step": 10508 + }, + { + "epoch": 2.29, + "learning_rate": 3.857439370766878e-06, + "loss": 1.6076, + "step": 10509 + }, + { + "epoch": 2.29, + "learning_rate": 3.857330128905396e-06, + "loss": 2.8177, + "step": 10510 + }, + { + "epoch": 2.29, + "learning_rate": 3.8572208870439154e-06, + "loss": 3.0817, + "step": 10511 + }, + { + "epoch": 2.29, + "learning_rate": 3.8571116451824345e-06, + "loss": 3.2956, + "step": 10512 + }, + { + "epoch": 2.29, + "learning_rate": 3.857002403320953e-06, + "loss": 2.1872, + "step": 10513 + }, + { + "epoch": 2.29, + "learning_rate": 3.856893161459472e-06, + "loss": 2.6586, + "step": 10514 + }, + { + "epoch": 2.29, + "learning_rate": 3.85678391959799e-06, + "loss": 1.5234, + "step": 10515 + }, + { + "epoch": 2.3, + "learning_rate": 3.856674677736509e-06, + "loss": 2.1531, + "step": 10516 + }, + { + "epoch": 2.3, + "learning_rate": 3.856565435875028e-06, + "loss": 3.9524, + "step": 10517 + }, + { + "epoch": 2.3, + "learning_rate": 3.856456194013546e-06, + "loss": 2.7024, + "step": 10518 + }, + { + "epoch": 2.3, + "learning_rate": 3.856346952152065e-06, + "loss": 2.5451, + "step": 10519 + }, + { + "epoch": 2.3, + "learning_rate": 3.856237710290583e-06, + "loss": 2.0219, + "step": 10520 + }, + { + "epoch": 2.3, + "learning_rate": 3.856128468429102e-06, + "loss": 2.5984, + "step": 10521 + }, + { + "epoch": 2.3, + "learning_rate": 3.8560192265676214e-06, + "loss": 3.0136, + "step": 10522 + }, + { + "epoch": 2.3, + "learning_rate": 3.85590998470614e-06, + "loss": 1.6551, + "step": 10523 + }, + { + "epoch": 2.3, + "learning_rate": 3.855800742844659e-06, + "loss": 3.4778, + "step": 10524 + }, + { + "epoch": 2.3, + "learning_rate": 3.855691500983177e-06, + "loss": 3.7673, + "step": 10525 + }, + { + "epoch": 2.3, + "learning_rate": 3.855582259121696e-06, + "loss": 2.6533, + "step": 10526 + }, + { + "epoch": 2.3, + "learning_rate": 3.855473017260215e-06, + "loss": 3.6523, + "step": 10527 + }, + { + "epoch": 2.3, + "learning_rate": 3.855363775398733e-06, + "loss": 2.8244, + "step": 10528 + }, + { + "epoch": 2.3, + "learning_rate": 3.855254533537251e-06, + "loss": 3.1895, + "step": 10529 + }, + { + "epoch": 2.3, + "learning_rate": 3.85514529167577e-06, + "loss": 2.1133, + "step": 10530 + }, + { + "epoch": 2.3, + "learning_rate": 3.855036049814289e-06, + "loss": 2.2565, + "step": 10531 + }, + { + "epoch": 2.3, + "learning_rate": 3.8549268079528075e-06, + "loss": 3.1415, + "step": 10532 + }, + { + "epoch": 2.3, + "learning_rate": 3.8548175660913266e-06, + "loss": 1.389, + "step": 10533 + }, + { + "epoch": 2.3, + "learning_rate": 3.854708324229845e-06, + "loss": 3.0945, + "step": 10534 + }, + { + "epoch": 2.3, + "learning_rate": 3.854599082368364e-06, + "loss": 3.2881, + "step": 10535 + }, + { + "epoch": 2.3, + "learning_rate": 3.854489840506883e-06, + "loss": 2.738, + "step": 10536 + }, + { + "epoch": 2.3, + "learning_rate": 3.854380598645401e-06, + "loss": 1.4966, + "step": 10537 + }, + { + "epoch": 2.3, + "learning_rate": 3.85427135678392e-06, + "loss": 2.1623, + "step": 10538 + }, + { + "epoch": 2.3, + "learning_rate": 3.854162114922438e-06, + "loss": 2.2862, + "step": 10539 + }, + { + "epoch": 2.3, + "learning_rate": 3.854052873060957e-06, + "loss": 2.7255, + "step": 10540 + }, + { + "epoch": 2.3, + "learning_rate": 3.853943631199476e-06, + "loss": 1.7782, + "step": 10541 + }, + { + "epoch": 2.3, + "learning_rate": 3.8538343893379945e-06, + "loss": 2.3902, + "step": 10542 + }, + { + "epoch": 2.3, + "learning_rate": 3.853725147476513e-06, + "loss": 3.439, + "step": 10543 + }, + { + "epoch": 2.3, + "learning_rate": 3.853615905615032e-06, + "loss": 2.0697, + "step": 10544 + }, + { + "epoch": 2.3, + "learning_rate": 3.853506663753551e-06, + "loss": 2.4434, + "step": 10545 + }, + { + "epoch": 2.3, + "learning_rate": 3.85339742189207e-06, + "loss": 2.0411, + "step": 10546 + }, + { + "epoch": 2.3, + "learning_rate": 3.853288180030588e-06, + "loss": 3.5741, + "step": 10547 + }, + { + "epoch": 2.3, + "learning_rate": 3.853178938169106e-06, + "loss": 2.5737, + "step": 10548 + }, + { + "epoch": 2.3, + "learning_rate": 3.853069696307625e-06, + "loss": 2.5262, + "step": 10549 + }, + { + "epoch": 2.3, + "learning_rate": 3.852960454446144e-06, + "loss": 3.2548, + "step": 10550 + }, + { + "epoch": 2.3, + "learning_rate": 3.852851212584663e-06, + "loss": 3.3808, + "step": 10551 + }, + { + "epoch": 2.3, + "learning_rate": 3.8527419707231815e-06, + "loss": 3.4059, + "step": 10552 + }, + { + "epoch": 2.3, + "learning_rate": 3.8526327288617e-06, + "loss": 2.9499, + "step": 10553 + }, + { + "epoch": 2.3, + "learning_rate": 3.852523487000219e-06, + "loss": 2.6662, + "step": 10554 + }, + { + "epoch": 2.3, + "learning_rate": 3.852414245138738e-06, + "loss": 2.2912, + "step": 10555 + }, + { + "epoch": 2.3, + "learning_rate": 3.852305003277256e-06, + "loss": 2.9785, + "step": 10556 + }, + { + "epoch": 2.3, + "learning_rate": 3.852195761415775e-06, + "loss": 2.7414, + "step": 10557 + }, + { + "epoch": 2.3, + "learning_rate": 3.852086519554293e-06, + "loss": 2.2518, + "step": 10558 + }, + { + "epoch": 2.3, + "learning_rate": 3.851977277692812e-06, + "loss": 2.791, + "step": 10559 + }, + { + "epoch": 2.3, + "learning_rate": 3.851868035831331e-06, + "loss": 2.4332, + "step": 10560 + }, + { + "epoch": 2.3, + "learning_rate": 3.851758793969849e-06, + "loss": 2.4255, + "step": 10561 + }, + { + "epoch": 2.31, + "learning_rate": 3.851649552108368e-06, + "loss": 2.1733, + "step": 10562 + }, + { + "epoch": 2.31, + "learning_rate": 3.851540310246887e-06, + "loss": 2.6533, + "step": 10563 + }, + { + "epoch": 2.31, + "learning_rate": 3.851431068385406e-06, + "loss": 2.0024, + "step": 10564 + }, + { + "epoch": 2.31, + "learning_rate": 3.851321826523925e-06, + "loss": 2.8899, + "step": 10565 + }, + { + "epoch": 2.31, + "learning_rate": 3.851212584662443e-06, + "loss": 2.5795, + "step": 10566 + }, + { + "epoch": 2.31, + "learning_rate": 3.851103342800961e-06, + "loss": 3.6812, + "step": 10567 + }, + { + "epoch": 2.31, + "learning_rate": 3.85099410093948e-06, + "loss": 1.6043, + "step": 10568 + }, + { + "epoch": 2.31, + "learning_rate": 3.850884859077999e-06, + "loss": 4.2903, + "step": 10569 + }, + { + "epoch": 2.31, + "learning_rate": 3.850775617216518e-06, + "loss": 3.047, + "step": 10570 + }, + { + "epoch": 2.31, + "learning_rate": 3.850666375355036e-06, + "loss": 2.1616, + "step": 10571 + }, + { + "epoch": 2.31, + "learning_rate": 3.8505571334935545e-06, + "loss": 2.5411, + "step": 10572 + }, + { + "epoch": 2.31, + "learning_rate": 3.8504478916320736e-06, + "loss": 2.721, + "step": 10573 + }, + { + "epoch": 2.31, + "learning_rate": 3.850338649770593e-06, + "loss": 2.0077, + "step": 10574 + }, + { + "epoch": 2.31, + "learning_rate": 3.850229407909112e-06, + "loss": 2.2441, + "step": 10575 + }, + { + "epoch": 2.31, + "learning_rate": 3.85012016604763e-06, + "loss": 3.0769, + "step": 10576 + }, + { + "epoch": 2.31, + "learning_rate": 3.850010924186148e-06, + "loss": 2.6171, + "step": 10577 + }, + { + "epoch": 2.31, + "learning_rate": 3.849901682324667e-06, + "loss": 2.4134, + "step": 10578 + }, + { + "epoch": 2.31, + "learning_rate": 3.849792440463186e-06, + "loss": 2.9263, + "step": 10579 + }, + { + "epoch": 2.31, + "learning_rate": 3.849683198601705e-06, + "loss": 2.3587, + "step": 10580 + }, + { + "epoch": 2.31, + "learning_rate": 3.849573956740223e-06, + "loss": 2.4623, + "step": 10581 + }, + { + "epoch": 2.31, + "learning_rate": 3.8494647148787415e-06, + "loss": 2.2568, + "step": 10582 + }, + { + "epoch": 2.31, + "learning_rate": 3.8493554730172605e-06, + "loss": 0.988, + "step": 10583 + }, + { + "epoch": 2.31, + "learning_rate": 3.8492462311557795e-06, + "loss": 2.4039, + "step": 10584 + }, + { + "epoch": 2.31, + "learning_rate": 3.849136989294298e-06, + "loss": 1.3225, + "step": 10585 + }, + { + "epoch": 2.31, + "learning_rate": 3.849027747432817e-06, + "loss": 3.2848, + "step": 10586 + }, + { + "epoch": 2.31, + "learning_rate": 3.848918505571335e-06, + "loss": 3.459, + "step": 10587 + }, + { + "epoch": 2.31, + "learning_rate": 3.848809263709854e-06, + "loss": 2.6592, + "step": 10588 + }, + { + "epoch": 2.31, + "learning_rate": 3.848700021848373e-06, + "loss": 1.9595, + "step": 10589 + }, + { + "epoch": 2.31, + "learning_rate": 3.848590779986891e-06, + "loss": 3.2181, + "step": 10590 + }, + { + "epoch": 2.31, + "learning_rate": 3.848481538125409e-06, + "loss": 3.5583, + "step": 10591 + }, + { + "epoch": 2.31, + "learning_rate": 3.8483722962639284e-06, + "loss": 2.174, + "step": 10592 + }, + { + "epoch": 2.31, + "learning_rate": 3.8482630544024475e-06, + "loss": 2.459, + "step": 10593 + }, + { + "epoch": 2.31, + "learning_rate": 3.8481538125409665e-06, + "loss": 1.9237, + "step": 10594 + }, + { + "epoch": 2.31, + "learning_rate": 3.848044570679485e-06, + "loss": 2.3697, + "step": 10595 + }, + { + "epoch": 2.31, + "learning_rate": 3.847935328818003e-06, + "loss": 2.5365, + "step": 10596 + }, + { + "epoch": 2.31, + "learning_rate": 3.847826086956522e-06, + "loss": 2.8705, + "step": 10597 + }, + { + "epoch": 2.31, + "learning_rate": 3.847716845095041e-06, + "loss": 2.659, + "step": 10598 + }, + { + "epoch": 2.31, + "learning_rate": 3.84760760323356e-06, + "loss": 2.043, + "step": 10599 + }, + { + "epoch": 2.31, + "learning_rate": 3.847498361372078e-06, + "loss": 3.3434, + "step": 10600 + }, + { + "epoch": 2.31, + "learning_rate": 3.847389119510596e-06, + "loss": 3.035, + "step": 10601 + }, + { + "epoch": 2.31, + "learning_rate": 3.847279877649115e-06, + "loss": 2.0543, + "step": 10602 + }, + { + "epoch": 2.31, + "learning_rate": 3.847170635787634e-06, + "loss": 2.4358, + "step": 10603 + }, + { + "epoch": 2.31, + "learning_rate": 3.8470613939261535e-06, + "loss": 3.3463, + "step": 10604 + }, + { + "epoch": 2.31, + "learning_rate": 3.846952152064672e-06, + "loss": 2.4956, + "step": 10605 + }, + { + "epoch": 2.31, + "learning_rate": 3.84684291020319e-06, + "loss": 2.1631, + "step": 10606 + }, + { + "epoch": 2.31, + "learning_rate": 3.846733668341709e-06, + "loss": 2.638, + "step": 10607 + }, + { + "epoch": 2.32, + "learning_rate": 3.846624426480228e-06, + "loss": 2.371, + "step": 10608 + }, + { + "epoch": 2.32, + "learning_rate": 3.846515184618746e-06, + "loss": 2.1541, + "step": 10609 + }, + { + "epoch": 2.32, + "learning_rate": 3.846405942757265e-06, + "loss": 3.4029, + "step": 10610 + }, + { + "epoch": 2.32, + "learning_rate": 3.846296700895783e-06, + "loss": 1.6496, + "step": 10611 + }, + { + "epoch": 2.32, + "learning_rate": 3.846187459034302e-06, + "loss": 2.0992, + "step": 10612 + }, + { + "epoch": 2.32, + "learning_rate": 3.846078217172821e-06, + "loss": 2.3722, + "step": 10613 + }, + { + "epoch": 2.32, + "learning_rate": 3.8459689753113396e-06, + "loss": 1.9029, + "step": 10614 + }, + { + "epoch": 2.32, + "learning_rate": 3.845859733449859e-06, + "loss": 2.8435, + "step": 10615 + }, + { + "epoch": 2.32, + "learning_rate": 3.845750491588377e-06, + "loss": 3.3149, + "step": 10616 + }, + { + "epoch": 2.32, + "learning_rate": 3.845641249726896e-06, + "loss": 3.022, + "step": 10617 + }, + { + "epoch": 2.32, + "learning_rate": 3.845532007865415e-06, + "loss": 1.3825, + "step": 10618 + }, + { + "epoch": 2.32, + "learning_rate": 3.845422766003933e-06, + "loss": 2.1266, + "step": 10619 + }, + { + "epoch": 2.32, + "learning_rate": 3.845313524142451e-06, + "loss": 2.5149, + "step": 10620 + }, + { + "epoch": 2.32, + "learning_rate": 3.84520428228097e-06, + "loss": 2.7803, + "step": 10621 + }, + { + "epoch": 2.32, + "learning_rate": 3.845095040419489e-06, + "loss": 2.2971, + "step": 10622 + }, + { + "epoch": 2.32, + "learning_rate": 3.844985798558008e-06, + "loss": 3.465, + "step": 10623 + }, + { + "epoch": 2.32, + "learning_rate": 3.8448765566965265e-06, + "loss": 3.4599, + "step": 10624 + }, + { + "epoch": 2.32, + "learning_rate": 3.844767314835045e-06, + "loss": 3.2593, + "step": 10625 + }, + { + "epoch": 2.32, + "learning_rate": 3.844658072973564e-06, + "loss": 1.7522, + "step": 10626 + }, + { + "epoch": 2.32, + "learning_rate": 3.844548831112083e-06, + "loss": 3.2414, + "step": 10627 + }, + { + "epoch": 2.32, + "learning_rate": 3.844439589250601e-06, + "loss": 3.1779, + "step": 10628 + }, + { + "epoch": 2.32, + "learning_rate": 3.84433034738912e-06, + "loss": 1.4892, + "step": 10629 + }, + { + "epoch": 2.32, + "learning_rate": 3.844221105527638e-06, + "loss": 2.3498, + "step": 10630 + }, + { + "epoch": 2.32, + "learning_rate": 3.844111863666157e-06, + "loss": 2.7897, + "step": 10631 + }, + { + "epoch": 2.32, + "learning_rate": 3.844002621804676e-06, + "loss": 1.9227, + "step": 10632 + }, + { + "epoch": 2.32, + "learning_rate": 3.8438933799431944e-06, + "loss": 2.2761, + "step": 10633 + }, + { + "epoch": 2.32, + "learning_rate": 3.8437841380817135e-06, + "loss": 3.5304, + "step": 10634 + }, + { + "epoch": 2.32, + "learning_rate": 3.843674896220232e-06, + "loss": 2.0746, + "step": 10635 + }, + { + "epoch": 2.32, + "learning_rate": 3.843565654358751e-06, + "loss": 2.9988, + "step": 10636 + }, + { + "epoch": 2.32, + "learning_rate": 3.84345641249727e-06, + "loss": 3.0131, + "step": 10637 + }, + { + "epoch": 2.32, + "learning_rate": 3.843347170635788e-06, + "loss": 2.3924, + "step": 10638 + }, + { + "epoch": 2.32, + "learning_rate": 3.843237928774307e-06, + "loss": 2.1777, + "step": 10639 + }, + { + "epoch": 2.32, + "learning_rate": 3.843128686912825e-06, + "loss": 2.5702, + "step": 10640 + }, + { + "epoch": 2.32, + "learning_rate": 3.843019445051344e-06, + "loss": 2.0882, + "step": 10641 + }, + { + "epoch": 2.32, + "learning_rate": 3.842910203189862e-06, + "loss": 1.9755, + "step": 10642 + }, + { + "epoch": 2.32, + "learning_rate": 3.842800961328381e-06, + "loss": 1.8728, + "step": 10643 + }, + { + "epoch": 2.32, + "learning_rate": 3.8426917194669e-06, + "loss": 3.3343, + "step": 10644 + }, + { + "epoch": 2.32, + "learning_rate": 3.842582477605419e-06, + "loss": 2.7145, + "step": 10645 + }, + { + "epoch": 2.32, + "learning_rate": 3.842473235743938e-06, + "loss": 1.8171, + "step": 10646 + }, + { + "epoch": 2.32, + "learning_rate": 3.842363993882456e-06, + "loss": 2.7493, + "step": 10647 + }, + { + "epoch": 2.32, + "learning_rate": 3.842254752020975e-06, + "loss": 2.8881, + "step": 10648 + }, + { + "epoch": 2.32, + "learning_rate": 3.842145510159493e-06, + "loss": 3.7687, + "step": 10649 + }, + { + "epoch": 2.32, + "learning_rate": 3.842036268298012e-06, + "loss": 2.4661, + "step": 10650 + }, + { + "epoch": 2.32, + "learning_rate": 3.841927026436531e-06, + "loss": 3.5703, + "step": 10651 + }, + { + "epoch": 2.32, + "learning_rate": 3.841817784575049e-06, + "loss": 3.2426, + "step": 10652 + }, + { + "epoch": 2.32, + "learning_rate": 3.841708542713568e-06, + "loss": 1.2631, + "step": 10653 + }, + { + "epoch": 2.33, + "learning_rate": 3.8415993008520865e-06, + "loss": 3.2186, + "step": 10654 + }, + { + "epoch": 2.33, + "learning_rate": 3.841490058990606e-06, + "loss": 2.0441, + "step": 10655 + }, + { + "epoch": 2.33, + "learning_rate": 3.841380817129124e-06, + "loss": 2.4713, + "step": 10656 + }, + { + "epoch": 2.33, + "learning_rate": 3.841271575267643e-06, + "loss": 2.4528, + "step": 10657 + }, + { + "epoch": 2.33, + "learning_rate": 3.841162333406162e-06, + "loss": 1.7296, + "step": 10658 + }, + { + "epoch": 2.33, + "learning_rate": 3.84105309154468e-06, + "loss": 3.0022, + "step": 10659 + }, + { + "epoch": 2.33, + "learning_rate": 3.840943849683199e-06, + "loss": 2.677, + "step": 10660 + }, + { + "epoch": 2.33, + "learning_rate": 3.840834607821717e-06, + "loss": 3.063, + "step": 10661 + }, + { + "epoch": 2.33, + "learning_rate": 3.840725365960236e-06, + "loss": 3.0762, + "step": 10662 + }, + { + "epoch": 2.33, + "learning_rate": 3.840616124098755e-06, + "loss": 2.135, + "step": 10663 + }, + { + "epoch": 2.33, + "learning_rate": 3.8405068822372735e-06, + "loss": 3.0955, + "step": 10664 + }, + { + "epoch": 2.33, + "learning_rate": 3.8403976403757925e-06, + "loss": 3.2923, + "step": 10665 + }, + { + "epoch": 2.33, + "learning_rate": 3.840288398514311e-06, + "loss": 2.7218, + "step": 10666 + }, + { + "epoch": 2.33, + "learning_rate": 3.84017915665283e-06, + "loss": 1.6177, + "step": 10667 + }, + { + "epoch": 2.33, + "learning_rate": 3.840069914791349e-06, + "loss": 2.2097, + "step": 10668 + }, + { + "epoch": 2.33, + "learning_rate": 3.839960672929867e-06, + "loss": 2.2895, + "step": 10669 + }, + { + "epoch": 2.33, + "learning_rate": 3.839851431068386e-06, + "loss": 2.3458, + "step": 10670 + }, + { + "epoch": 2.33, + "learning_rate": 3.839742189206904e-06, + "loss": 2.1787, + "step": 10671 + }, + { + "epoch": 2.33, + "learning_rate": 3.839632947345423e-06, + "loss": 2.5881, + "step": 10672 + }, + { + "epoch": 2.33, + "learning_rate": 3.8395237054839414e-06, + "loss": 1.4846, + "step": 10673 + }, + { + "epoch": 2.33, + "learning_rate": 3.8394144636224605e-06, + "loss": 2.7455, + "step": 10674 + }, + { + "epoch": 2.33, + "learning_rate": 3.839305221760979e-06, + "loss": 2.1834, + "step": 10675 + }, + { + "epoch": 2.33, + "learning_rate": 3.839195979899498e-06, + "loss": 2.8893, + "step": 10676 + }, + { + "epoch": 2.33, + "learning_rate": 3.839086738038017e-06, + "loss": 2.413, + "step": 10677 + }, + { + "epoch": 2.33, + "learning_rate": 3.838977496176535e-06, + "loss": 2.1187, + "step": 10678 + }, + { + "epoch": 2.33, + "learning_rate": 3.838868254315054e-06, + "loss": 2.7536, + "step": 10679 + }, + { + "epoch": 2.33, + "learning_rate": 3.838759012453572e-06, + "loss": 3.4217, + "step": 10680 + }, + { + "epoch": 2.33, + "learning_rate": 3.838649770592091e-06, + "loss": 3.5865, + "step": 10681 + }, + { + "epoch": 2.33, + "learning_rate": 3.83854052873061e-06, + "loss": 3.388, + "step": 10682 + }, + { + "epoch": 2.33, + "learning_rate": 3.838431286869128e-06, + "loss": 3.0946, + "step": 10683 + }, + { + "epoch": 2.33, + "learning_rate": 3.838322045007647e-06, + "loss": 1.9913, + "step": 10684 + }, + { + "epoch": 2.33, + "learning_rate": 3.838212803146166e-06, + "loss": 2.4813, + "step": 10685 + }, + { + "epoch": 2.33, + "learning_rate": 3.838103561284685e-06, + "loss": 1.8123, + "step": 10686 + }, + { + "epoch": 2.33, + "learning_rate": 3.837994319423204e-06, + "loss": 2.5023, + "step": 10687 + }, + { + "epoch": 2.33, + "learning_rate": 3.837885077561722e-06, + "loss": 1.0408, + "step": 10688 + }, + { + "epoch": 2.33, + "learning_rate": 3.83777583570024e-06, + "loss": 1.6319, + "step": 10689 + }, + { + "epoch": 2.33, + "learning_rate": 3.837666593838759e-06, + "loss": 1.1274, + "step": 10690 + }, + { + "epoch": 2.33, + "learning_rate": 3.837557351977278e-06, + "loss": 2.6609, + "step": 10691 + }, + { + "epoch": 2.33, + "learning_rate": 3.837448110115797e-06, + "loss": 1.7471, + "step": 10692 + }, + { + "epoch": 2.33, + "learning_rate": 3.837338868254315e-06, + "loss": 2.3734, + "step": 10693 + }, + { + "epoch": 2.33, + "learning_rate": 3.8372296263928335e-06, + "loss": 2.5982, + "step": 10694 + }, + { + "epoch": 2.33, + "learning_rate": 3.8371203845313526e-06, + "loss": 2.1477, + "step": 10695 + }, + { + "epoch": 2.33, + "learning_rate": 3.837011142669872e-06, + "loss": 2.3598, + "step": 10696 + }, + { + "epoch": 2.33, + "learning_rate": 3.83690190080839e-06, + "loss": 2.4898, + "step": 10697 + }, + { + "epoch": 2.33, + "learning_rate": 3.836792658946909e-06, + "loss": 1.8298, + "step": 10698 + }, + { + "epoch": 2.34, + "learning_rate": 3.836683417085427e-06, + "loss": 4.1938, + "step": 10699 + }, + { + "epoch": 2.34, + "learning_rate": 3.836574175223946e-06, + "loss": 2.5944, + "step": 10700 + }, + { + "epoch": 2.34, + "learning_rate": 3.836464933362465e-06, + "loss": 2.3139, + "step": 10701 + }, + { + "epoch": 2.34, + "learning_rate": 3.836355691500983e-06, + "loss": 2.137, + "step": 10702 + }, + { + "epoch": 2.34, + "learning_rate": 3.836246449639502e-06, + "loss": 2.0116, + "step": 10703 + }, + { + "epoch": 2.34, + "learning_rate": 3.8361372077780205e-06, + "loss": 1.7623, + "step": 10704 + }, + { + "epoch": 2.34, + "learning_rate": 3.8360279659165395e-06, + "loss": 1.9342, + "step": 10705 + }, + { + "epoch": 2.34, + "learning_rate": 3.8359187240550586e-06, + "loss": 2.1344, + "step": 10706 + }, + { + "epoch": 2.34, + "learning_rate": 3.835809482193577e-06, + "loss": 1.0859, + "step": 10707 + }, + { + "epoch": 2.34, + "learning_rate": 3.835700240332095e-06, + "loss": 3.0251, + "step": 10708 + }, + { + "epoch": 2.34, + "learning_rate": 3.835590998470614e-06, + "loss": 2.7561, + "step": 10709 + }, + { + "epoch": 2.34, + "learning_rate": 3.835481756609133e-06, + "loss": 2.3234, + "step": 10710 + }, + { + "epoch": 2.34, + "learning_rate": 3.835372514747652e-06, + "loss": 2.6508, + "step": 10711 + }, + { + "epoch": 2.34, + "learning_rate": 3.83526327288617e-06, + "loss": 1.3091, + "step": 10712 + }, + { + "epoch": 2.34, + "learning_rate": 3.835154031024688e-06, + "loss": 1.9655, + "step": 10713 + }, + { + "epoch": 2.34, + "learning_rate": 3.8350447891632074e-06, + "loss": 2.6505, + "step": 10714 + }, + { + "epoch": 2.34, + "learning_rate": 3.8349355473017265e-06, + "loss": 1.8644, + "step": 10715 + }, + { + "epoch": 2.34, + "learning_rate": 3.8348263054402455e-06, + "loss": 1.6333, + "step": 10716 + }, + { + "epoch": 2.34, + "learning_rate": 3.834717063578764e-06, + "loss": 2.9749, + "step": 10717 + }, + { + "epoch": 2.34, + "learning_rate": 3.834607821717282e-06, + "loss": 1.969, + "step": 10718 + }, + { + "epoch": 2.34, + "learning_rate": 3.834498579855801e-06, + "loss": 2.5425, + "step": 10719 + }, + { + "epoch": 2.34, + "learning_rate": 3.83438933799432e-06, + "loss": 2.4494, + "step": 10720 + }, + { + "epoch": 2.34, + "learning_rate": 3.834280096132839e-06, + "loss": 3.0209, + "step": 10721 + }, + { + "epoch": 2.34, + "learning_rate": 3.834170854271357e-06, + "loss": 1.9969, + "step": 10722 + }, + { + "epoch": 2.34, + "learning_rate": 3.834061612409875e-06, + "loss": 2.5118, + "step": 10723 + }, + { + "epoch": 2.34, + "learning_rate": 3.833952370548394e-06, + "loss": 1.032, + "step": 10724 + }, + { + "epoch": 2.34, + "learning_rate": 3.8338431286869134e-06, + "loss": 2.3411, + "step": 10725 + }, + { + "epoch": 2.34, + "learning_rate": 3.833733886825432e-06, + "loss": 3.1384, + "step": 10726 + }, + { + "epoch": 2.34, + "learning_rate": 3.833624644963951e-06, + "loss": 2.1587, + "step": 10727 + }, + { + "epoch": 2.34, + "learning_rate": 3.833515403102469e-06, + "loss": 1.3021, + "step": 10728 + }, + { + "epoch": 2.34, + "learning_rate": 3.833406161240988e-06, + "loss": 2.2515, + "step": 10729 + }, + { + "epoch": 2.34, + "learning_rate": 3.833296919379507e-06, + "loss": 2.7263, + "step": 10730 + }, + { + "epoch": 2.34, + "learning_rate": 3.833187677518025e-06, + "loss": 2.0148, + "step": 10731 + }, + { + "epoch": 2.34, + "learning_rate": 3.833078435656543e-06, + "loss": 1.9232, + "step": 10732 + }, + { + "epoch": 2.34, + "learning_rate": 3.832969193795062e-06, + "loss": 1.4456, + "step": 10733 + }, + { + "epoch": 2.34, + "learning_rate": 3.832859951933581e-06, + "loss": 3.1453, + "step": 10734 + }, + { + "epoch": 2.34, + "learning_rate": 3.8327507100721e-06, + "loss": 2.2379, + "step": 10735 + }, + { + "epoch": 2.34, + "learning_rate": 3.832641468210619e-06, + "loss": 2.2712, + "step": 10736 + }, + { + "epoch": 2.34, + "learning_rate": 3.832532226349137e-06, + "loss": 2.6259, + "step": 10737 + }, + { + "epoch": 2.34, + "learning_rate": 3.832422984487656e-06, + "loss": 1.6352, + "step": 10738 + }, + { + "epoch": 2.34, + "learning_rate": 3.832313742626175e-06, + "loss": 3.3405, + "step": 10739 + }, + { + "epoch": 2.34, + "learning_rate": 3.832204500764694e-06, + "loss": 2.9019, + "step": 10740 + }, + { + "epoch": 2.34, + "learning_rate": 3.832095258903212e-06, + "loss": 3.2388, + "step": 10741 + }, + { + "epoch": 2.34, + "learning_rate": 3.83198601704173e-06, + "loss": 1.7065, + "step": 10742 + }, + { + "epoch": 2.34, + "learning_rate": 3.831876775180249e-06, + "loss": 1.2608, + "step": 10743 + }, + { + "epoch": 2.34, + "learning_rate": 3.831767533318768e-06, + "loss": 3.2761, + "step": 10744 + }, + { + "epoch": 2.35, + "learning_rate": 3.831658291457287e-06, + "loss": 2.4416, + "step": 10745 + }, + { + "epoch": 2.35, + "learning_rate": 3.8315490495958055e-06, + "loss": 1.907, + "step": 10746 + }, + { + "epoch": 2.35, + "learning_rate": 3.831439807734324e-06, + "loss": 2.4287, + "step": 10747 + }, + { + "epoch": 2.35, + "learning_rate": 3.831330565872843e-06, + "loss": 1.5977, + "step": 10748 + }, + { + "epoch": 2.35, + "learning_rate": 3.831221324011362e-06, + "loss": 1.6564, + "step": 10749 + }, + { + "epoch": 2.35, + "learning_rate": 3.83111208214988e-06, + "loss": 1.8375, + "step": 10750 + }, + { + "epoch": 2.35, + "learning_rate": 3.831002840288399e-06, + "loss": 1.6176, + "step": 10751 + }, + { + "epoch": 2.35, + "learning_rate": 3.830893598426917e-06, + "loss": 2.5674, + "step": 10752 + }, + { + "epoch": 2.35, + "learning_rate": 3.830784356565436e-06, + "loss": 2.5706, + "step": 10753 + }, + { + "epoch": 2.35, + "learning_rate": 3.830675114703955e-06, + "loss": 1.4775, + "step": 10754 + }, + { + "epoch": 2.35, + "learning_rate": 3.8305658728424735e-06, + "loss": 2.1091, + "step": 10755 + }, + { + "epoch": 2.35, + "learning_rate": 3.8304566309809925e-06, + "loss": 2.0389, + "step": 10756 + }, + { + "epoch": 2.35, + "learning_rate": 3.830347389119511e-06, + "loss": 2.7355, + "step": 10757 + }, + { + "epoch": 2.35, + "learning_rate": 3.83023814725803e-06, + "loss": 3.0924, + "step": 10758 + }, + { + "epoch": 2.35, + "learning_rate": 3.830128905396549e-06, + "loss": 3.2638, + "step": 10759 + }, + { + "epoch": 2.35, + "learning_rate": 3.830019663535067e-06, + "loss": 1.3506, + "step": 10760 + }, + { + "epoch": 2.35, + "learning_rate": 3.829910421673585e-06, + "loss": 2.7723, + "step": 10761 + }, + { + "epoch": 2.35, + "learning_rate": 3.829801179812104e-06, + "loss": 2.2707, + "step": 10762 + }, + { + "epoch": 2.35, + "learning_rate": 3.829691937950623e-06, + "loss": 1.9772, + "step": 10763 + }, + { + "epoch": 2.35, + "learning_rate": 3.829582696089142e-06, + "loss": 2.7806, + "step": 10764 + }, + { + "epoch": 2.35, + "learning_rate": 3.82947345422766e-06, + "loss": 2.6492, + "step": 10765 + }, + { + "epoch": 2.35, + "learning_rate": 3.829364212366179e-06, + "loss": 2.9357, + "step": 10766 + }, + { + "epoch": 2.35, + "learning_rate": 3.829254970504698e-06, + "loss": 2.6516, + "step": 10767 + }, + { + "epoch": 2.35, + "learning_rate": 3.829145728643217e-06, + "loss": 2.2555, + "step": 10768 + }, + { + "epoch": 2.35, + "learning_rate": 3.829036486781736e-06, + "loss": 1.709, + "step": 10769 + }, + { + "epoch": 2.35, + "learning_rate": 3.828927244920254e-06, + "loss": 2.6093, + "step": 10770 + }, + { + "epoch": 2.35, + "learning_rate": 3.828818003058772e-06, + "loss": 2.1313, + "step": 10771 + }, + { + "epoch": 2.35, + "learning_rate": 3.828708761197291e-06, + "loss": 2.7275, + "step": 10772 + }, + { + "epoch": 2.35, + "learning_rate": 3.82859951933581e-06, + "loss": 1.5219, + "step": 10773 + }, + { + "epoch": 2.35, + "learning_rate": 3.828490277474328e-06, + "loss": 3.4889, + "step": 10774 + }, + { + "epoch": 2.35, + "learning_rate": 3.828381035612847e-06, + "loss": 1.5702, + "step": 10775 + }, + { + "epoch": 2.35, + "learning_rate": 3.8282717937513656e-06, + "loss": 2.5842, + "step": 10776 + }, + { + "epoch": 2.35, + "learning_rate": 3.828162551889885e-06, + "loss": 2.9312, + "step": 10777 + }, + { + "epoch": 2.35, + "learning_rate": 3.828053310028404e-06, + "loss": 1.8992, + "step": 10778 + }, + { + "epoch": 2.35, + "learning_rate": 3.827944068166922e-06, + "loss": 2.6854, + "step": 10779 + }, + { + "epoch": 2.35, + "learning_rate": 3.827834826305441e-06, + "loss": 1.5873, + "step": 10780 + }, + { + "epoch": 2.35, + "learning_rate": 3.827725584443959e-06, + "loss": 3.148, + "step": 10781 + }, + { + "epoch": 2.35, + "learning_rate": 3.827616342582478e-06, + "loss": 2.6132, + "step": 10782 + }, + { + "epoch": 2.35, + "learning_rate": 3.827507100720997e-06, + "loss": 2.2172, + "step": 10783 + }, + { + "epoch": 2.35, + "learning_rate": 3.827397858859515e-06, + "loss": 2.7111, + "step": 10784 + }, + { + "epoch": 2.35, + "learning_rate": 3.8272886169980335e-06, + "loss": 2.9098, + "step": 10785 + }, + { + "epoch": 2.35, + "learning_rate": 3.8271793751365525e-06, + "loss": 2.4514, + "step": 10786 + }, + { + "epoch": 2.35, + "learning_rate": 3.8270701332750716e-06, + "loss": 2.5838, + "step": 10787 + }, + { + "epoch": 2.35, + "learning_rate": 3.82696089141359e-06, + "loss": 2.3761, + "step": 10788 + }, + { + "epoch": 2.35, + "learning_rate": 3.826851649552109e-06, + "loss": 2.2604, + "step": 10789 + }, + { + "epoch": 2.35, + "learning_rate": 3.826742407690627e-06, + "loss": 0.7965, + "step": 10790 + }, + { + "epoch": 2.36, + "learning_rate": 3.826633165829146e-06, + "loss": 2.8103, + "step": 10791 + }, + { + "epoch": 2.36, + "learning_rate": 3.826523923967665e-06, + "loss": 2.846, + "step": 10792 + }, + { + "epoch": 2.36, + "learning_rate": 3.826414682106183e-06, + "loss": 1.7111, + "step": 10793 + }, + { + "epoch": 2.36, + "learning_rate": 3.826305440244702e-06, + "loss": 2.4847, + "step": 10794 + }, + { + "epoch": 2.36, + "learning_rate": 3.8261961983832204e-06, + "loss": 1.7846, + "step": 10795 + }, + { + "epoch": 2.36, + "learning_rate": 3.8260869565217395e-06, + "loss": 2.2415, + "step": 10796 + }, + { + "epoch": 2.36, + "learning_rate": 3.8259777146602585e-06, + "loss": 2.9903, + "step": 10797 + }, + { + "epoch": 2.36, + "learning_rate": 3.825868472798777e-06, + "loss": 2.9716, + "step": 10798 + }, + { + "epoch": 2.36, + "learning_rate": 3.825759230937296e-06, + "loss": 2.5105, + "step": 10799 + }, + { + "epoch": 2.36, + "learning_rate": 3.825649989075814e-06, + "loss": 2.1118, + "step": 10800 + }, + { + "epoch": 2.36, + "learning_rate": 3.825540747214333e-06, + "loss": 1.6683, + "step": 10801 + }, + { + "epoch": 2.36, + "learning_rate": 3.825431505352852e-06, + "loss": 2.9099, + "step": 10802 + }, + { + "epoch": 2.36, + "learning_rate": 3.82532226349137e-06, + "loss": 2.4637, + "step": 10803 + }, + { + "epoch": 2.36, + "learning_rate": 3.825213021629889e-06, + "loss": 2.1216, + "step": 10804 + }, + { + "epoch": 2.36, + "learning_rate": 3.825103779768407e-06, + "loss": 2.6079, + "step": 10805 + }, + { + "epoch": 2.36, + "learning_rate": 3.8249945379069264e-06, + "loss": 2.9011, + "step": 10806 + }, + { + "epoch": 2.36, + "learning_rate": 3.824885296045445e-06, + "loss": 2.3865, + "step": 10807 + }, + { + "epoch": 2.36, + "learning_rate": 3.824776054183964e-06, + "loss": 1.7367, + "step": 10808 + }, + { + "epoch": 2.36, + "learning_rate": 3.824666812322483e-06, + "loss": 2.441, + "step": 10809 + }, + { + "epoch": 2.36, + "learning_rate": 3.824557570461001e-06, + "loss": 2.3564, + "step": 10810 + }, + { + "epoch": 2.36, + "learning_rate": 3.82444832859952e-06, + "loss": 2.3934, + "step": 10811 + }, + { + "epoch": 2.36, + "learning_rate": 3.824339086738038e-06, + "loss": 2.0536, + "step": 10812 + }, + { + "epoch": 2.36, + "learning_rate": 3.824229844876557e-06, + "loss": 2.6185, + "step": 10813 + }, + { + "epoch": 2.36, + "learning_rate": 3.824120603015075e-06, + "loss": 2.3336, + "step": 10814 + }, + { + "epoch": 2.36, + "learning_rate": 3.824011361153594e-06, + "loss": 1.5426, + "step": 10815 + }, + { + "epoch": 2.36, + "learning_rate": 3.823902119292113e-06, + "loss": 2.101, + "step": 10816 + }, + { + "epoch": 2.36, + "learning_rate": 3.8237928774306316e-06, + "loss": 3.0418, + "step": 10817 + }, + { + "epoch": 2.36, + "learning_rate": 3.823683635569151e-06, + "loss": 2.2542, + "step": 10818 + }, + { + "epoch": 2.36, + "learning_rate": 3.823574393707669e-06, + "loss": 2.376, + "step": 10819 + }, + { + "epoch": 2.36, + "learning_rate": 3.823465151846188e-06, + "loss": 2.4131, + "step": 10820 + }, + { + "epoch": 2.36, + "learning_rate": 3.823355909984706e-06, + "loss": 2.2762, + "step": 10821 + }, + { + "epoch": 2.36, + "learning_rate": 3.823246668123225e-06, + "loss": 1.7118, + "step": 10822 + }, + { + "epoch": 2.36, + "learning_rate": 3.823137426261744e-06, + "loss": 2.2281, + "step": 10823 + }, + { + "epoch": 2.36, + "learning_rate": 3.823028184400262e-06, + "loss": 1.8579, + "step": 10824 + }, + { + "epoch": 2.36, + "learning_rate": 3.822918942538781e-06, + "loss": 2.3274, + "step": 10825 + }, + { + "epoch": 2.36, + "learning_rate": 3.8228097006772995e-06, + "loss": 2.2072, + "step": 10826 + }, + { + "epoch": 2.36, + "learning_rate": 3.8227004588158185e-06, + "loss": 1.8961, + "step": 10827 + }, + { + "epoch": 2.36, + "learning_rate": 3.8225912169543376e-06, + "loss": 2.3983, + "step": 10828 + }, + { + "epoch": 2.36, + "learning_rate": 3.822481975092856e-06, + "loss": 2.4512, + "step": 10829 + }, + { + "epoch": 2.36, + "learning_rate": 3.822372733231375e-06, + "loss": 1.7159, + "step": 10830 + }, + { + "epoch": 2.36, + "learning_rate": 3.822263491369893e-06, + "loss": 2.1117, + "step": 10831 + }, + { + "epoch": 2.36, + "learning_rate": 3.822154249508412e-06, + "loss": 1.9165, + "step": 10832 + }, + { + "epoch": 2.36, + "learning_rate": 3.822045007646931e-06, + "loss": 1.3562, + "step": 10833 + }, + { + "epoch": 2.36, + "learning_rate": 3.821935765785449e-06, + "loss": 3.0646, + "step": 10834 + }, + { + "epoch": 2.36, + "learning_rate": 3.821826523923968e-06, + "loss": 1.9333, + "step": 10835 + }, + { + "epoch": 2.36, + "learning_rate": 3.8217172820624865e-06, + "loss": 1.2058, + "step": 10836 + }, + { + "epoch": 2.37, + "learning_rate": 3.8216080402010055e-06, + "loss": 1.8858, + "step": 10837 + }, + { + "epoch": 2.37, + "learning_rate": 3.821498798339524e-06, + "loss": 2.6722, + "step": 10838 + }, + { + "epoch": 2.37, + "learning_rate": 3.821389556478043e-06, + "loss": 2.622, + "step": 10839 + }, + { + "epoch": 2.37, + "learning_rate": 3.821280314616561e-06, + "loss": 1.5188, + "step": 10840 + }, + { + "epoch": 2.37, + "learning_rate": 3.82117107275508e-06, + "loss": 2.5126, + "step": 10841 + }, + { + "epoch": 2.37, + "learning_rate": 3.821061830893599e-06, + "loss": 3.1654, + "step": 10842 + }, + { + "epoch": 2.37, + "learning_rate": 3.820952589032117e-06, + "loss": 1.6143, + "step": 10843 + }, + { + "epoch": 2.37, + "learning_rate": 3.820843347170636e-06, + "loss": 2.3789, + "step": 10844 + }, + { + "epoch": 2.37, + "learning_rate": 3.820734105309154e-06, + "loss": 1.7574, + "step": 10845 + }, + { + "epoch": 2.37, + "learning_rate": 3.820624863447673e-06, + "loss": 1.5792, + "step": 10846 + }, + { + "epoch": 2.37, + "learning_rate": 3.8205156215861924e-06, + "loss": 0.9806, + "step": 10847 + }, + { + "epoch": 2.37, + "learning_rate": 3.820406379724711e-06, + "loss": 3.3837, + "step": 10848 + }, + { + "epoch": 2.37, + "learning_rate": 3.82029713786323e-06, + "loss": 2.7541, + "step": 10849 + }, + { + "epoch": 2.37, + "learning_rate": 3.820187896001748e-06, + "loss": 2.7239, + "step": 10850 + }, + { + "epoch": 2.37, + "learning_rate": 3.820078654140267e-06, + "loss": 3.1008, + "step": 10851 + }, + { + "epoch": 2.37, + "learning_rate": 3.819969412278786e-06, + "loss": 2.8934, + "step": 10852 + }, + { + "epoch": 2.37, + "learning_rate": 3.819860170417304e-06, + "loss": 3.2775, + "step": 10853 + }, + { + "epoch": 2.37, + "learning_rate": 3.819750928555822e-06, + "loss": 2.4982, + "step": 10854 + }, + { + "epoch": 2.37, + "learning_rate": 3.819641686694341e-06, + "loss": 2.572, + "step": 10855 + }, + { + "epoch": 2.37, + "learning_rate": 3.81953244483286e-06, + "loss": 1.9602, + "step": 10856 + }, + { + "epoch": 2.37, + "learning_rate": 3.819423202971379e-06, + "loss": 2.4101, + "step": 10857 + }, + { + "epoch": 2.37, + "learning_rate": 3.819313961109898e-06, + "loss": 2.5613, + "step": 10858 + }, + { + "epoch": 2.37, + "learning_rate": 3.819204719248416e-06, + "loss": 3.3028, + "step": 10859 + }, + { + "epoch": 2.37, + "learning_rate": 3.819095477386935e-06, + "loss": 2.854, + "step": 10860 + }, + { + "epoch": 2.37, + "learning_rate": 3.818986235525454e-06, + "loss": 3.0854, + "step": 10861 + }, + { + "epoch": 2.37, + "learning_rate": 3.818876993663973e-06, + "loss": 1.8692, + "step": 10862 + }, + { + "epoch": 2.37, + "learning_rate": 3.818767751802491e-06, + "loss": 1.9153, + "step": 10863 + }, + { + "epoch": 2.37, + "learning_rate": 3.818658509941009e-06, + "loss": 2.3875, + "step": 10864 + }, + { + "epoch": 2.37, + "learning_rate": 3.818549268079528e-06, + "loss": 1.2245, + "step": 10865 + }, + { + "epoch": 2.37, + "learning_rate": 3.818440026218047e-06, + "loss": 1.9956, + "step": 10866 + }, + { + "epoch": 2.37, + "learning_rate": 3.8183307843565655e-06, + "loss": 2.6749, + "step": 10867 + }, + { + "epoch": 2.37, + "learning_rate": 3.8182215424950845e-06, + "loss": 2.0098, + "step": 10868 + }, + { + "epoch": 2.37, + "learning_rate": 3.818112300633603e-06, + "loss": 1.9121, + "step": 10869 + }, + { + "epoch": 2.37, + "learning_rate": 3.818003058772122e-06, + "loss": 1.5164, + "step": 10870 + }, + { + "epoch": 2.37, + "learning_rate": 3.817893816910641e-06, + "loss": 2.1812, + "step": 10871 + }, + { + "epoch": 2.37, + "learning_rate": 3.817784575049159e-06, + "loss": 2.2147, + "step": 10872 + }, + { + "epoch": 2.37, + "learning_rate": 3.817675333187677e-06, + "loss": 3.0428, + "step": 10873 + }, + { + "epoch": 2.37, + "learning_rate": 3.817566091326196e-06, + "loss": 2.5741, + "step": 10874 + }, + { + "epoch": 2.37, + "learning_rate": 3.817456849464715e-06, + "loss": 2.6663, + "step": 10875 + }, + { + "epoch": 2.37, + "learning_rate": 3.817347607603234e-06, + "loss": 2.0319, + "step": 10876 + }, + { + "epoch": 2.37, + "learning_rate": 3.8172383657417525e-06, + "loss": 2.9795, + "step": 10877 + }, + { + "epoch": 2.37, + "learning_rate": 3.817129123880271e-06, + "loss": 1.9254, + "step": 10878 + }, + { + "epoch": 2.37, + "learning_rate": 3.81701988201879e-06, + "loss": 3.1311, + "step": 10879 + }, + { + "epoch": 2.37, + "learning_rate": 3.816910640157309e-06, + "loss": 2.3858, + "step": 10880 + }, + { + "epoch": 2.37, + "learning_rate": 3.816801398295828e-06, + "loss": 1.8195, + "step": 10881 + }, + { + "epoch": 2.37, + "learning_rate": 3.816692156434346e-06, + "loss": 2.4668, + "step": 10882 + }, + { + "epoch": 2.38, + "learning_rate": 3.816582914572864e-06, + "loss": 2.4334, + "step": 10883 + }, + { + "epoch": 2.38, + "learning_rate": 3.816473672711383e-06, + "loss": 2.99, + "step": 10884 + }, + { + "epoch": 2.38, + "learning_rate": 3.816364430849902e-06, + "loss": 2.9854, + "step": 10885 + }, + { + "epoch": 2.38, + "learning_rate": 3.816255188988421e-06, + "loss": 4.0584, + "step": 10886 + }, + { + "epoch": 2.38, + "learning_rate": 3.8161459471269394e-06, + "loss": 2.4796, + "step": 10887 + }, + { + "epoch": 2.38, + "learning_rate": 3.816036705265458e-06, + "loss": 2.2235, + "step": 10888 + }, + { + "epoch": 2.38, + "learning_rate": 3.815927463403977e-06, + "loss": 2.995, + "step": 10889 + }, + { + "epoch": 2.38, + "learning_rate": 3.815818221542496e-06, + "loss": 1.9771, + "step": 10890 + }, + { + "epoch": 2.38, + "learning_rate": 3.815708979681014e-06, + "loss": 3.2621, + "step": 10891 + }, + { + "epoch": 2.38, + "learning_rate": 3.815599737819533e-06, + "loss": 2.6327, + "step": 10892 + }, + { + "epoch": 2.38, + "learning_rate": 3.815490495958051e-06, + "loss": 0.9488, + "step": 10893 + }, + { + "epoch": 2.38, + "learning_rate": 3.81538125409657e-06, + "loss": 1.8249, + "step": 10894 + }, + { + "epoch": 2.38, + "learning_rate": 3.815272012235089e-06, + "loss": 2.832, + "step": 10895 + }, + { + "epoch": 2.38, + "learning_rate": 3.815162770373607e-06, + "loss": 1.8515, + "step": 10896 + }, + { + "epoch": 2.38, + "learning_rate": 3.8150535285121255e-06, + "loss": 1.5988, + "step": 10897 + }, + { + "epoch": 2.38, + "learning_rate": 3.8149442866506446e-06, + "loss": 1.5793, + "step": 10898 + }, + { + "epoch": 2.38, + "learning_rate": 3.814835044789164e-06, + "loss": 1.2401, + "step": 10899 + }, + { + "epoch": 2.38, + "learning_rate": 3.814725802927683e-06, + "loss": 2.228, + "step": 10900 + }, + { + "epoch": 2.38, + "learning_rate": 3.8146165610662012e-06, + "loss": 2.8051, + "step": 10901 + }, + { + "epoch": 2.38, + "learning_rate": 3.8145073192047194e-06, + "loss": 2.1165, + "step": 10902 + }, + { + "epoch": 2.38, + "learning_rate": 3.814398077343238e-06, + "loss": 2.9346, + "step": 10903 + }, + { + "epoch": 2.38, + "learning_rate": 3.814288835481757e-06, + "loss": 2.4166, + "step": 10904 + }, + { + "epoch": 2.38, + "learning_rate": 3.8141795936202757e-06, + "loss": 2.5487, + "step": 10905 + }, + { + "epoch": 2.38, + "learning_rate": 3.814070351758794e-06, + "loss": 3.6783, + "step": 10906 + }, + { + "epoch": 2.38, + "learning_rate": 3.813961109897313e-06, + "loss": 2.2504, + "step": 10907 + }, + { + "epoch": 2.38, + "learning_rate": 3.8138518680358315e-06, + "loss": 2.5985, + "step": 10908 + }, + { + "epoch": 2.38, + "learning_rate": 3.8137426261743506e-06, + "loss": 1.6564, + "step": 10909 + }, + { + "epoch": 2.38, + "learning_rate": 3.813633384312869e-06, + "loss": 2.996, + "step": 10910 + }, + { + "epoch": 2.38, + "learning_rate": 3.8135241424513874e-06, + "loss": 2.1891, + "step": 10911 + }, + { + "epoch": 2.38, + "learning_rate": 3.8134149005899064e-06, + "loss": 3.1405, + "step": 10912 + }, + { + "epoch": 2.38, + "learning_rate": 3.813305658728425e-06, + "loss": 1.9557, + "step": 10913 + }, + { + "epoch": 2.38, + "learning_rate": 3.813196416866944e-06, + "loss": 2.221, + "step": 10914 + }, + { + "epoch": 2.38, + "learning_rate": 3.8130871750054627e-06, + "loss": 2.1123, + "step": 10915 + }, + { + "epoch": 2.38, + "learning_rate": 3.812977933143981e-06, + "loss": 1.9681, + "step": 10916 + }, + { + "epoch": 2.38, + "learning_rate": 3.8128686912825e-06, + "loss": 1.4599, + "step": 10917 + }, + { + "epoch": 2.38, + "learning_rate": 3.8127594494210185e-06, + "loss": 2.9031, + "step": 10918 + }, + { + "epoch": 2.38, + "learning_rate": 3.812650207559537e-06, + "loss": 2.465, + "step": 10919 + }, + { + "epoch": 2.38, + "learning_rate": 3.8125409656980557e-06, + "loss": 2.5596, + "step": 10920 + }, + { + "epoch": 2.38, + "learning_rate": 3.8124317238365743e-06, + "loss": 2.8771, + "step": 10921 + }, + { + "epoch": 2.38, + "learning_rate": 3.812322481975093e-06, + "loss": 3.0151, + "step": 10922 + }, + { + "epoch": 2.38, + "learning_rate": 3.812213240113612e-06, + "loss": 1.4663, + "step": 10923 + }, + { + "epoch": 2.38, + "learning_rate": 3.8121039982521306e-06, + "loss": 2.1625, + "step": 10924 + }, + { + "epoch": 2.38, + "learning_rate": 3.811994756390649e-06, + "loss": 2.6904, + "step": 10925 + }, + { + "epoch": 2.38, + "learning_rate": 3.811885514529168e-06, + "loss": 2.1798, + "step": 10926 + }, + { + "epoch": 2.38, + "learning_rate": 3.8117762726676864e-06, + "loss": 2.8413, + "step": 10927 + }, + { + "epoch": 2.38, + "learning_rate": 3.8116670308062054e-06, + "loss": 3.0867, + "step": 10928 + }, + { + "epoch": 2.39, + "learning_rate": 3.811557788944724e-06, + "loss": 2.053, + "step": 10929 + }, + { + "epoch": 2.39, + "learning_rate": 3.8114485470832422e-06, + "loss": 3.1618, + "step": 10930 + }, + { + "epoch": 2.39, + "learning_rate": 3.8113393052217613e-06, + "loss": 2.9606, + "step": 10931 + }, + { + "epoch": 2.39, + "learning_rate": 3.81123006336028e-06, + "loss": 2.3021, + "step": 10932 + }, + { + "epoch": 2.39, + "learning_rate": 3.811120821498799e-06, + "loss": 2.4331, + "step": 10933 + }, + { + "epoch": 2.39, + "learning_rate": 3.8110115796373175e-06, + "loss": 2.9681, + "step": 10934 + }, + { + "epoch": 2.39, + "learning_rate": 3.8109023377758357e-06, + "loss": 2.5319, + "step": 10935 + }, + { + "epoch": 2.39, + "learning_rate": 3.8107930959143548e-06, + "loss": 2.745, + "step": 10936 + }, + { + "epoch": 2.39, + "learning_rate": 3.8106838540528734e-06, + "loss": 1.5452, + "step": 10937 + }, + { + "epoch": 2.39, + "learning_rate": 3.8105746121913924e-06, + "loss": 1.9746, + "step": 10938 + }, + { + "epoch": 2.39, + "learning_rate": 3.8104653703299106e-06, + "loss": 3.1536, + "step": 10939 + }, + { + "epoch": 2.39, + "learning_rate": 3.810356128468429e-06, + "loss": 1.7527, + "step": 10940 + }, + { + "epoch": 2.39, + "learning_rate": 3.8102468866069482e-06, + "loss": 1.196, + "step": 10941 + }, + { + "epoch": 2.39, + "learning_rate": 3.810137644745467e-06, + "loss": 2.1139, + "step": 10942 + }, + { + "epoch": 2.39, + "learning_rate": 3.810028402883986e-06, + "loss": 1.3952, + "step": 10943 + }, + { + "epoch": 2.39, + "learning_rate": 3.809919161022504e-06, + "loss": 2.2707, + "step": 10944 + }, + { + "epoch": 2.39, + "learning_rate": 3.8098099191610227e-06, + "loss": 2.773, + "step": 10945 + }, + { + "epoch": 2.39, + "learning_rate": 3.8097006772995417e-06, + "loss": 2.9531, + "step": 10946 + }, + { + "epoch": 2.39, + "learning_rate": 3.8095914354380603e-06, + "loss": 2.808, + "step": 10947 + }, + { + "epoch": 2.39, + "learning_rate": 3.809482193576579e-06, + "loss": 2.1775, + "step": 10948 + }, + { + "epoch": 2.39, + "learning_rate": 3.8093729517150975e-06, + "loss": 3.1831, + "step": 10949 + }, + { + "epoch": 2.39, + "learning_rate": 3.809263709853616e-06, + "loss": 2.3994, + "step": 10950 + }, + { + "epoch": 2.39, + "learning_rate": 3.8091544679921348e-06, + "loss": 2.6406, + "step": 10951 + }, + { + "epoch": 2.39, + "learning_rate": 3.809045226130654e-06, + "loss": 2.6355, + "step": 10952 + }, + { + "epoch": 2.39, + "learning_rate": 3.808935984269172e-06, + "loss": 3.8601, + "step": 10953 + }, + { + "epoch": 2.39, + "learning_rate": 3.8088267424076906e-06, + "loss": 2.7315, + "step": 10954 + }, + { + "epoch": 2.39, + "learning_rate": 3.8087175005462096e-06, + "loss": 3.353, + "step": 10955 + }, + { + "epoch": 2.39, + "learning_rate": 3.8086082586847282e-06, + "loss": 3.264, + "step": 10956 + }, + { + "epoch": 2.39, + "learning_rate": 3.8084990168232473e-06, + "loss": 2.731, + "step": 10957 + }, + { + "epoch": 2.39, + "learning_rate": 3.8083897749617655e-06, + "loss": 1.7714, + "step": 10958 + }, + { + "epoch": 2.39, + "learning_rate": 3.808280533100284e-06, + "loss": 1.6296, + "step": 10959 + }, + { + "epoch": 2.39, + "learning_rate": 3.808171291238803e-06, + "loss": 3.3119, + "step": 10960 + }, + { + "epoch": 2.39, + "learning_rate": 3.8080620493773217e-06, + "loss": 3.4985, + "step": 10961 + }, + { + "epoch": 2.39, + "learning_rate": 3.8079528075158408e-06, + "loss": 1.162, + "step": 10962 + }, + { + "epoch": 2.39, + "learning_rate": 3.807843565654359e-06, + "loss": 3.6759, + "step": 10963 + }, + { + "epoch": 2.39, + "learning_rate": 3.8077343237928776e-06, + "loss": 2.134, + "step": 10964 + }, + { + "epoch": 2.39, + "learning_rate": 3.8076250819313966e-06, + "loss": 2.8878, + "step": 10965 + }, + { + "epoch": 2.39, + "learning_rate": 3.807515840069915e-06, + "loss": 3.6028, + "step": 10966 + }, + { + "epoch": 2.39, + "learning_rate": 3.8074065982084342e-06, + "loss": 2.1548, + "step": 10967 + }, + { + "epoch": 2.39, + "learning_rate": 3.8072973563469524e-06, + "loss": 1.7713, + "step": 10968 + }, + { + "epoch": 2.39, + "learning_rate": 3.807188114485471e-06, + "loss": 1.9286, + "step": 10969 + }, + { + "epoch": 2.39, + "learning_rate": 3.80707887262399e-06, + "loss": 3.2215, + "step": 10970 + }, + { + "epoch": 2.39, + "learning_rate": 3.8069696307625087e-06, + "loss": 1.6188, + "step": 10971 + }, + { + "epoch": 2.39, + "learning_rate": 3.806860388901027e-06, + "loss": 2.8824, + "step": 10972 + }, + { + "epoch": 2.39, + "learning_rate": 3.806751147039546e-06, + "loss": 1.9432, + "step": 10973 + }, + { + "epoch": 2.4, + "learning_rate": 3.8066419051780645e-06, + "loss": 2.2899, + "step": 10974 + }, + { + "epoch": 2.4, + "learning_rate": 3.806532663316583e-06, + "loss": 2.0312, + "step": 10975 + }, + { + "epoch": 2.4, + "learning_rate": 3.806423421455102e-06, + "loss": 2.9678, + "step": 10976 + }, + { + "epoch": 2.4, + "learning_rate": 3.8063141795936203e-06, + "loss": 3.1049, + "step": 10977 + }, + { + "epoch": 2.4, + "learning_rate": 3.8062049377321394e-06, + "loss": 2.4177, + "step": 10978 + }, + { + "epoch": 2.4, + "learning_rate": 3.806095695870658e-06, + "loss": 3.8614, + "step": 10979 + }, + { + "epoch": 2.4, + "learning_rate": 3.8059864540091766e-06, + "loss": 2.6161, + "step": 10980 + }, + { + "epoch": 2.4, + "learning_rate": 3.8058772121476956e-06, + "loss": 2.2702, + "step": 10981 + }, + { + "epoch": 2.4, + "learning_rate": 3.805767970286214e-06, + "loss": 2.8492, + "step": 10982 + }, + { + "epoch": 2.4, + "learning_rate": 3.8056587284247324e-06, + "loss": 2.0579, + "step": 10983 + }, + { + "epoch": 2.4, + "learning_rate": 3.8055494865632515e-06, + "loss": 2.2926, + "step": 10984 + }, + { + "epoch": 2.4, + "learning_rate": 3.80544024470177e-06, + "loss": 2.513, + "step": 10985 + }, + { + "epoch": 2.4, + "learning_rate": 3.8053310028402883e-06, + "loss": 3.2438, + "step": 10986 + }, + { + "epoch": 2.4, + "learning_rate": 3.8052217609788073e-06, + "loss": 2.8303, + "step": 10987 + }, + { + "epoch": 2.4, + "learning_rate": 3.805112519117326e-06, + "loss": 1.6169, + "step": 10988 + }, + { + "epoch": 2.4, + "learning_rate": 3.805003277255845e-06, + "loss": 1.8583, + "step": 10989 + }, + { + "epoch": 2.4, + "learning_rate": 3.8048940353943636e-06, + "loss": 2.7661, + "step": 10990 + }, + { + "epoch": 2.4, + "learning_rate": 3.8047847935328817e-06, + "loss": 1.7362, + "step": 10991 + }, + { + "epoch": 2.4, + "learning_rate": 3.8046755516714008e-06, + "loss": 1.3777, + "step": 10992 + }, + { + "epoch": 2.4, + "learning_rate": 3.8045663098099194e-06, + "loss": 1.2875, + "step": 10993 + }, + { + "epoch": 2.4, + "learning_rate": 3.8044570679484384e-06, + "loss": 1.6242, + "step": 10994 + }, + { + "epoch": 2.4, + "learning_rate": 3.804347826086957e-06, + "loss": 2.9006, + "step": 10995 + }, + { + "epoch": 2.4, + "learning_rate": 3.8042385842254752e-06, + "loss": 1.2672, + "step": 10996 + }, + { + "epoch": 2.4, + "learning_rate": 3.8041293423639943e-06, + "loss": 1.3866, + "step": 10997 + }, + { + "epoch": 2.4, + "learning_rate": 3.804020100502513e-06, + "loss": 1.6852, + "step": 10998 + }, + { + "epoch": 2.4, + "learning_rate": 3.803910858641032e-06, + "loss": 1.9417, + "step": 10999 + }, + { + "epoch": 2.4, + "learning_rate": 3.8038016167795505e-06, + "loss": 1.8573, + "step": 11000 + }, + { + "epoch": 2.4, + "learning_rate": 3.8036923749180687e-06, + "loss": 1.9968, + "step": 11001 + }, + { + "epoch": 2.4, + "learning_rate": 3.8035831330565877e-06, + "loss": 1.8134, + "step": 11002 + }, + { + "epoch": 2.4, + "learning_rate": 3.8034738911951063e-06, + "loss": 1.6228, + "step": 11003 + }, + { + "epoch": 2.4, + "learning_rate": 3.803364649333625e-06, + "loss": 2.7732, + "step": 11004 + }, + { + "epoch": 2.4, + "learning_rate": 3.8032554074721436e-06, + "loss": 2.563, + "step": 11005 + }, + { + "epoch": 2.4, + "learning_rate": 3.803146165610662e-06, + "loss": 2.4901, + "step": 11006 + }, + { + "epoch": 2.4, + "learning_rate": 3.803036923749181e-06, + "loss": 1.7832, + "step": 11007 + }, + { + "epoch": 2.4, + "learning_rate": 3.8029276818877e-06, + "loss": 2.1197, + "step": 11008 + }, + { + "epoch": 2.4, + "learning_rate": 3.8028184400262184e-06, + "loss": 2.0865, + "step": 11009 + }, + { + "epoch": 2.4, + "learning_rate": 3.8027091981647366e-06, + "loss": 2.9574, + "step": 11010 + }, + { + "epoch": 2.4, + "learning_rate": 3.8025999563032557e-06, + "loss": 3.4762, + "step": 11011 + }, + { + "epoch": 2.4, + "learning_rate": 3.8024907144417743e-06, + "loss": 1.5702, + "step": 11012 + }, + { + "epoch": 2.4, + "learning_rate": 3.8023814725802933e-06, + "loss": 2.3644, + "step": 11013 + }, + { + "epoch": 2.4, + "learning_rate": 3.802272230718812e-06, + "loss": 2.0182, + "step": 11014 + }, + { + "epoch": 2.4, + "learning_rate": 3.80216298885733e-06, + "loss": 2.863, + "step": 11015 + }, + { + "epoch": 2.4, + "learning_rate": 3.802053746995849e-06, + "loss": 1.7023, + "step": 11016 + }, + { + "epoch": 2.4, + "learning_rate": 3.8019445051343677e-06, + "loss": 2.4037, + "step": 11017 + }, + { + "epoch": 2.4, + "learning_rate": 3.8018352632728868e-06, + "loss": 1.9687, + "step": 11018 + }, + { + "epoch": 2.4, + "learning_rate": 3.801726021411405e-06, + "loss": 2.2544, + "step": 11019 + }, + { + "epoch": 2.41, + "learning_rate": 3.8016167795499236e-06, + "loss": 1.6709, + "step": 11020 + }, + { + "epoch": 2.41, + "learning_rate": 3.8015075376884426e-06, + "loss": 2.2648, + "step": 11021 + }, + { + "epoch": 2.41, + "learning_rate": 3.8013982958269612e-06, + "loss": 2.3834, + "step": 11022 + }, + { + "epoch": 2.41, + "learning_rate": 3.8012890539654803e-06, + "loss": 1.0388, + "step": 11023 + }, + { + "epoch": 2.41, + "learning_rate": 3.8011798121039984e-06, + "loss": 2.4788, + "step": 11024 + }, + { + "epoch": 2.41, + "learning_rate": 3.801070570242517e-06, + "loss": 1.7538, + "step": 11025 + }, + { + "epoch": 2.41, + "learning_rate": 3.800961328381036e-06, + "loss": 2.0899, + "step": 11026 + }, + { + "epoch": 2.41, + "learning_rate": 3.8008520865195547e-06, + "loss": 2.6262, + "step": 11027 + }, + { + "epoch": 2.41, + "learning_rate": 3.8007428446580733e-06, + "loss": 3.339, + "step": 11028 + }, + { + "epoch": 2.41, + "learning_rate": 3.800633602796592e-06, + "loss": 1.9403, + "step": 11029 + }, + { + "epoch": 2.41, + "learning_rate": 3.8005243609351105e-06, + "loss": 3.1863, + "step": 11030 + }, + { + "epoch": 2.41, + "learning_rate": 3.8004151190736296e-06, + "loss": 1.8457, + "step": 11031 + }, + { + "epoch": 2.41, + "learning_rate": 3.800305877212148e-06, + "loss": 2.6326, + "step": 11032 + }, + { + "epoch": 2.41, + "learning_rate": 3.800196635350667e-06, + "loss": 2.2319, + "step": 11033 + }, + { + "epoch": 2.41, + "learning_rate": 3.8000873934891854e-06, + "loss": 2.4492, + "step": 11034 + }, + { + "epoch": 2.41, + "learning_rate": 3.799978151627704e-06, + "loss": 2.1373, + "step": 11035 + }, + { + "epoch": 2.41, + "learning_rate": 3.7998689097662226e-06, + "loss": 2.0068, + "step": 11036 + }, + { + "epoch": 2.41, + "learning_rate": 3.7997596679047417e-06, + "loss": 1.9153, + "step": 11037 + }, + { + "epoch": 2.41, + "learning_rate": 3.79965042604326e-06, + "loss": 2.8246, + "step": 11038 + }, + { + "epoch": 2.41, + "learning_rate": 3.7995411841817785e-06, + "loss": 2.839, + "step": 11039 + }, + { + "epoch": 2.41, + "learning_rate": 3.7994319423202975e-06, + "loss": 1.9223, + "step": 11040 + }, + { + "epoch": 2.41, + "learning_rate": 3.799322700458816e-06, + "loss": 3.5593, + "step": 11041 + }, + { + "epoch": 2.41, + "learning_rate": 3.799213458597335e-06, + "loss": 2.8464, + "step": 11042 + }, + { + "epoch": 2.41, + "learning_rate": 3.7991042167358533e-06, + "loss": 2.1722, + "step": 11043 + }, + { + "epoch": 2.41, + "learning_rate": 3.798994974874372e-06, + "loss": 2.4193, + "step": 11044 + }, + { + "epoch": 2.41, + "learning_rate": 3.798885733012891e-06, + "loss": 2.7185, + "step": 11045 + }, + { + "epoch": 2.41, + "learning_rate": 3.7987764911514096e-06, + "loss": 1.9935, + "step": 11046 + }, + { + "epoch": 2.41, + "learning_rate": 3.7986672492899286e-06, + "loss": 1.8475, + "step": 11047 + }, + { + "epoch": 2.41, + "learning_rate": 3.798558007428447e-06, + "loss": 1.2805, + "step": 11048 + }, + { + "epoch": 2.41, + "learning_rate": 3.7984487655669654e-06, + "loss": 2.2507, + "step": 11049 + }, + { + "epoch": 2.41, + "learning_rate": 3.7983395237054845e-06, + "loss": 1.9924, + "step": 11050 + }, + { + "epoch": 2.41, + "learning_rate": 3.798230281844003e-06, + "loss": 3.1866, + "step": 11051 + }, + { + "epoch": 2.41, + "learning_rate": 3.7981210399825212e-06, + "loss": 3.0179, + "step": 11052 + }, + { + "epoch": 2.41, + "learning_rate": 3.7980117981210403e-06, + "loss": 3.0901, + "step": 11053 + }, + { + "epoch": 2.41, + "learning_rate": 3.797902556259559e-06, + "loss": 1.8633, + "step": 11054 + }, + { + "epoch": 2.41, + "learning_rate": 3.797793314398078e-06, + "loss": 3.1278, + "step": 11055 + }, + { + "epoch": 2.41, + "learning_rate": 3.7976840725365965e-06, + "loss": 2.1295, + "step": 11056 + }, + { + "epoch": 2.41, + "learning_rate": 3.7975748306751147e-06, + "loss": 1.826, + "step": 11057 + }, + { + "epoch": 2.41, + "learning_rate": 3.7974655888136338e-06, + "loss": 3.3735, + "step": 11058 + }, + { + "epoch": 2.41, + "learning_rate": 3.7973563469521524e-06, + "loss": 3.1132, + "step": 11059 + }, + { + "epoch": 2.41, + "learning_rate": 3.797247105090671e-06, + "loss": 2.3527, + "step": 11060 + }, + { + "epoch": 2.41, + "learning_rate": 3.79713786322919e-06, + "loss": 1.6874, + "step": 11061 + }, + { + "epoch": 2.41, + "learning_rate": 3.797028621367708e-06, + "loss": 1.3638, + "step": 11062 + }, + { + "epoch": 2.41, + "learning_rate": 3.796919379506227e-06, + "loss": 2.5944, + "step": 11063 + }, + { + "epoch": 2.41, + "learning_rate": 3.796810137644746e-06, + "loss": 2.1182, + "step": 11064 + }, + { + "epoch": 2.41, + "learning_rate": 3.7967008957832645e-06, + "loss": 2.7986, + "step": 11065 + }, + { + "epoch": 2.42, + "learning_rate": 3.7965916539217827e-06, + "loss": 2.6174, + "step": 11066 + }, + { + "epoch": 2.42, + "learning_rate": 3.7964824120603017e-06, + "loss": 1.6227, + "step": 11067 + }, + { + "epoch": 2.42, + "learning_rate": 3.7963731701988203e-06, + "loss": 2.8732, + "step": 11068 + }, + { + "epoch": 2.42, + "learning_rate": 3.7962639283373393e-06, + "loss": 1.4225, + "step": 11069 + }, + { + "epoch": 2.42, + "learning_rate": 3.796154686475858e-06, + "loss": 2.2197, + "step": 11070 + }, + { + "epoch": 2.42, + "learning_rate": 3.796045444614376e-06, + "loss": 2.1958, + "step": 11071 + }, + { + "epoch": 2.42, + "learning_rate": 3.795936202752895e-06, + "loss": 2.1507, + "step": 11072 + }, + { + "epoch": 2.42, + "learning_rate": 3.7958269608914138e-06, + "loss": 3.0428, + "step": 11073 + }, + { + "epoch": 2.42, + "learning_rate": 3.795717719029933e-06, + "loss": 2.0206, + "step": 11074 + }, + { + "epoch": 2.42, + "learning_rate": 3.7956084771684514e-06, + "loss": 1.8274, + "step": 11075 + }, + { + "epoch": 2.42, + "learning_rate": 3.7954992353069696e-06, + "loss": 2.2535, + "step": 11076 + }, + { + "epoch": 2.42, + "learning_rate": 3.7953899934454886e-06, + "loss": 1.518, + "step": 11077 + }, + { + "epoch": 2.42, + "learning_rate": 3.7952807515840073e-06, + "loss": 3.0941, + "step": 11078 + }, + { + "epoch": 2.42, + "learning_rate": 3.7951715097225263e-06, + "loss": 1.8324, + "step": 11079 + }, + { + "epoch": 2.42, + "learning_rate": 3.795062267861045e-06, + "loss": 2.1443, + "step": 11080 + }, + { + "epoch": 2.42, + "learning_rate": 3.794953025999563e-06, + "loss": 2.7014, + "step": 11081 + }, + { + "epoch": 2.42, + "learning_rate": 3.794843784138082e-06, + "loss": 2.7698, + "step": 11082 + }, + { + "epoch": 2.42, + "learning_rate": 3.7947345422766007e-06, + "loss": 1.8507, + "step": 11083 + }, + { + "epoch": 2.42, + "learning_rate": 3.7946253004151198e-06, + "loss": 2.7895, + "step": 11084 + }, + { + "epoch": 2.42, + "learning_rate": 3.794516058553638e-06, + "loss": 2.0988, + "step": 11085 + }, + { + "epoch": 2.42, + "learning_rate": 3.7944068166921566e-06, + "loss": 2.0452, + "step": 11086 + }, + { + "epoch": 2.42, + "learning_rate": 3.7942975748306756e-06, + "loss": 2.7779, + "step": 11087 + }, + { + "epoch": 2.42, + "learning_rate": 3.794188332969194e-06, + "loss": 2.5661, + "step": 11088 + }, + { + "epoch": 2.42, + "learning_rate": 3.794079091107713e-06, + "loss": 2.2788, + "step": 11089 + }, + { + "epoch": 2.42, + "learning_rate": 3.7939698492462314e-06, + "loss": 2.9353, + "step": 11090 + }, + { + "epoch": 2.42, + "learning_rate": 3.79386060738475e-06, + "loss": 2.5055, + "step": 11091 + }, + { + "epoch": 2.42, + "learning_rate": 3.7937513655232687e-06, + "loss": 1.0499, + "step": 11092 + }, + { + "epoch": 2.42, + "learning_rate": 3.7936421236617877e-06, + "loss": 2.3193, + "step": 11093 + }, + { + "epoch": 2.42, + "learning_rate": 3.7935328818003063e-06, + "loss": 2.1791, + "step": 11094 + }, + { + "epoch": 2.42, + "learning_rate": 3.7934236399388245e-06, + "loss": 2.5557, + "step": 11095 + }, + { + "epoch": 2.42, + "learning_rate": 3.7933143980773435e-06, + "loss": 2.4841, + "step": 11096 + }, + { + "epoch": 2.42, + "learning_rate": 3.793205156215862e-06, + "loss": 2.5915, + "step": 11097 + }, + { + "epoch": 2.42, + "learning_rate": 3.793095914354381e-06, + "loss": 1.9883, + "step": 11098 + }, + { + "epoch": 2.42, + "learning_rate": 3.7929866724928994e-06, + "loss": 1.6644, + "step": 11099 + }, + { + "epoch": 2.42, + "learning_rate": 3.792877430631418e-06, + "loss": 2.4248, + "step": 11100 + }, + { + "epoch": 2.42, + "learning_rate": 3.792768188769937e-06, + "loss": 2.7467, + "step": 11101 + }, + { + "epoch": 2.42, + "learning_rate": 3.7926589469084556e-06, + "loss": 2.5853, + "step": 11102 + }, + { + "epoch": 2.42, + "learning_rate": 3.7925497050469746e-06, + "loss": 2.6116, + "step": 11103 + }, + { + "epoch": 2.42, + "learning_rate": 3.792440463185493e-06, + "loss": 1.9504, + "step": 11104 + }, + { + "epoch": 2.42, + "learning_rate": 3.7923312213240114e-06, + "loss": 2.581, + "step": 11105 + }, + { + "epoch": 2.42, + "learning_rate": 3.7922219794625305e-06, + "loss": 1.9987, + "step": 11106 + }, + { + "epoch": 2.42, + "learning_rate": 3.792112737601049e-06, + "loss": 3.0073, + "step": 11107 + }, + { + "epoch": 2.42, + "learning_rate": 3.792003495739568e-06, + "loss": 1.2334, + "step": 11108 + }, + { + "epoch": 2.42, + "learning_rate": 3.7918942538780863e-06, + "loss": 2.7349, + "step": 11109 + }, + { + "epoch": 2.42, + "learning_rate": 3.791785012016605e-06, + "loss": 2.8354, + "step": 11110 + }, + { + "epoch": 2.42, + "learning_rate": 3.791675770155124e-06, + "loss": 3.3775, + "step": 11111 + }, + { + "epoch": 2.43, + "learning_rate": 3.7915665282936426e-06, + "loss": 1.6398, + "step": 11112 + }, + { + "epoch": 2.43, + "learning_rate": 3.791457286432161e-06, + "loss": 1.5714, + "step": 11113 + }, + { + "epoch": 2.43, + "learning_rate": 3.79134804457068e-06, + "loss": 3.2321, + "step": 11114 + }, + { + "epoch": 2.43, + "learning_rate": 3.7912388027091984e-06, + "loss": 2.4519, + "step": 11115 + }, + { + "epoch": 2.43, + "learning_rate": 3.791129560847717e-06, + "loss": 2.4826, + "step": 11116 + }, + { + "epoch": 2.43, + "learning_rate": 3.791020318986236e-06, + "loss": 2.2792, + "step": 11117 + }, + { + "epoch": 2.43, + "learning_rate": 3.7909110771247542e-06, + "loss": 1.4696, + "step": 11118 + }, + { + "epoch": 2.43, + "learning_rate": 3.790801835263273e-06, + "loss": 2.4463, + "step": 11119 + }, + { + "epoch": 2.43, + "learning_rate": 3.790692593401792e-06, + "loss": 1.2198, + "step": 11120 + }, + { + "epoch": 2.43, + "learning_rate": 3.7905833515403105e-06, + "loss": 2.0164, + "step": 11121 + }, + { + "epoch": 2.43, + "learning_rate": 3.7904741096788295e-06, + "loss": 2.6749, + "step": 11122 + }, + { + "epoch": 2.43, + "learning_rate": 3.7903648678173477e-06, + "loss": 2.1578, + "step": 11123 + }, + { + "epoch": 2.43, + "learning_rate": 3.7902556259558663e-06, + "loss": 2.065, + "step": 11124 + }, + { + "epoch": 2.43, + "learning_rate": 3.7901463840943854e-06, + "loss": 2.3514, + "step": 11125 + }, + { + "epoch": 2.43, + "learning_rate": 3.790037142232904e-06, + "loss": 2.0473, + "step": 11126 + }, + { + "epoch": 2.43, + "learning_rate": 3.789927900371423e-06, + "loss": 3.1552, + "step": 11127 + }, + { + "epoch": 2.43, + "learning_rate": 3.789818658509941e-06, + "loss": 2.5485, + "step": 11128 + }, + { + "epoch": 2.43, + "learning_rate": 3.78970941664846e-06, + "loss": 3.5112, + "step": 11129 + }, + { + "epoch": 2.43, + "learning_rate": 3.789600174786979e-06, + "loss": 2.0884, + "step": 11130 + }, + { + "epoch": 2.43, + "learning_rate": 3.7894909329254974e-06, + "loss": 3.5642, + "step": 11131 + }, + { + "epoch": 2.43, + "learning_rate": 3.7893816910640156e-06, + "loss": 1.6588, + "step": 11132 + }, + { + "epoch": 2.43, + "learning_rate": 3.7892724492025347e-06, + "loss": 2.9543, + "step": 11133 + }, + { + "epoch": 2.43, + "learning_rate": 3.7891632073410533e-06, + "loss": 1.8772, + "step": 11134 + }, + { + "epoch": 2.43, + "learning_rate": 3.7890539654795723e-06, + "loss": 1.7898, + "step": 11135 + }, + { + "epoch": 2.43, + "learning_rate": 3.788944723618091e-06, + "loss": 1.7794, + "step": 11136 + }, + { + "epoch": 2.43, + "learning_rate": 3.788835481756609e-06, + "loss": 2.0846, + "step": 11137 + }, + { + "epoch": 2.43, + "learning_rate": 3.788726239895128e-06, + "loss": 2.6019, + "step": 11138 + }, + { + "epoch": 2.43, + "learning_rate": 3.7886169980336468e-06, + "loss": 2.1787, + "step": 11139 + }, + { + "epoch": 2.43, + "learning_rate": 3.788507756172166e-06, + "loss": 1.6943, + "step": 11140 + }, + { + "epoch": 2.43, + "learning_rate": 3.7883985143106844e-06, + "loss": 1.8076, + "step": 11141 + }, + { + "epoch": 2.43, + "learning_rate": 3.7882892724492026e-06, + "loss": 2.1565, + "step": 11142 + }, + { + "epoch": 2.43, + "learning_rate": 3.7881800305877216e-06, + "loss": 1.6645, + "step": 11143 + }, + { + "epoch": 2.43, + "learning_rate": 3.7880707887262402e-06, + "loss": 2.3755, + "step": 11144 + }, + { + "epoch": 2.43, + "learning_rate": 3.787961546864759e-06, + "loss": 2.6286, + "step": 11145 + }, + { + "epoch": 2.43, + "learning_rate": 3.787852305003278e-06, + "loss": 1.0927, + "step": 11146 + }, + { + "epoch": 2.43, + "learning_rate": 3.787743063141796e-06, + "loss": 3.7516, + "step": 11147 + }, + { + "epoch": 2.43, + "learning_rate": 3.7876338212803147e-06, + "loss": 3.2476, + "step": 11148 + }, + { + "epoch": 2.43, + "learning_rate": 3.7875245794188337e-06, + "loss": 2.3703, + "step": 11149 + }, + { + "epoch": 2.43, + "learning_rate": 3.7874153375573523e-06, + "loss": 2.0427, + "step": 11150 + }, + { + "epoch": 2.43, + "learning_rate": 3.7873060956958705e-06, + "loss": 2.4062, + "step": 11151 + }, + { + "epoch": 2.43, + "learning_rate": 3.7871968538343895e-06, + "loss": 1.3363, + "step": 11152 + }, + { + "epoch": 2.43, + "learning_rate": 3.787087611972908e-06, + "loss": 2.459, + "step": 11153 + }, + { + "epoch": 2.43, + "learning_rate": 3.786978370111427e-06, + "loss": 2.5625, + "step": 11154 + }, + { + "epoch": 2.43, + "learning_rate": 3.786869128249946e-06, + "loss": 3.0492, + "step": 11155 + }, + { + "epoch": 2.43, + "learning_rate": 3.786759886388464e-06, + "loss": 2.1324, + "step": 11156 + }, + { + "epoch": 2.43, + "learning_rate": 3.786650644526983e-06, + "loss": 1.6961, + "step": 11157 + }, + { + "epoch": 2.44, + "learning_rate": 3.7865414026655016e-06, + "loss": 1.8152, + "step": 11158 + }, + { + "epoch": 2.44, + "learning_rate": 3.7864321608040207e-06, + "loss": 3.483, + "step": 11159 + }, + { + "epoch": 2.44, + "learning_rate": 3.7863229189425393e-06, + "loss": 2.9984, + "step": 11160 + }, + { + "epoch": 2.44, + "learning_rate": 3.7862136770810575e-06, + "loss": 3.1724, + "step": 11161 + }, + { + "epoch": 2.44, + "learning_rate": 3.7861044352195765e-06, + "loss": 2.5019, + "step": 11162 + }, + { + "epoch": 2.44, + "learning_rate": 3.785995193358095e-06, + "loss": 2.6731, + "step": 11163 + }, + { + "epoch": 2.44, + "learning_rate": 3.785885951496614e-06, + "loss": 2.1816, + "step": 11164 + }, + { + "epoch": 2.44, + "learning_rate": 3.7857767096351323e-06, + "loss": 3.0894, + "step": 11165 + }, + { + "epoch": 2.44, + "learning_rate": 3.785667467773651e-06, + "loss": 2.8657, + "step": 11166 + }, + { + "epoch": 2.44, + "learning_rate": 3.78555822591217e-06, + "loss": 2.4456, + "step": 11167 + }, + { + "epoch": 2.44, + "learning_rate": 3.7854489840506886e-06, + "loss": 1.4226, + "step": 11168 + }, + { + "epoch": 2.44, + "learning_rate": 3.785339742189207e-06, + "loss": 3.3362, + "step": 11169 + }, + { + "epoch": 2.44, + "learning_rate": 3.785230500327726e-06, + "loss": 2.4624, + "step": 11170 + }, + { + "epoch": 2.44, + "learning_rate": 3.7851212584662444e-06, + "loss": 2.5009, + "step": 11171 + }, + { + "epoch": 2.44, + "learning_rate": 3.7850120166047635e-06, + "loss": 2.407, + "step": 11172 + }, + { + "epoch": 2.44, + "learning_rate": 3.784902774743282e-06, + "loss": 3.2974, + "step": 11173 + }, + { + "epoch": 2.44, + "learning_rate": 3.7847935328818007e-06, + "loss": 1.3989, + "step": 11174 + }, + { + "epoch": 2.44, + "learning_rate": 3.7846842910203193e-06, + "loss": 2.9944, + "step": 11175 + }, + { + "epoch": 2.44, + "learning_rate": 3.784575049158838e-06, + "loss": 2.1493, + "step": 11176 + }, + { + "epoch": 2.44, + "learning_rate": 3.7844658072973565e-06, + "loss": 2.3895, + "step": 11177 + }, + { + "epoch": 2.44, + "learning_rate": 3.7843565654358756e-06, + "loss": 2.1306, + "step": 11178 + }, + { + "epoch": 2.44, + "learning_rate": 3.784247323574394e-06, + "loss": 2.263, + "step": 11179 + }, + { + "epoch": 2.44, + "learning_rate": 3.7841380817129124e-06, + "loss": 2.9888, + "step": 11180 + }, + { + "epoch": 2.44, + "learning_rate": 3.7840288398514314e-06, + "loss": 1.7182, + "step": 11181 + }, + { + "epoch": 2.44, + "learning_rate": 3.78391959798995e-06, + "loss": 3.2577, + "step": 11182 + }, + { + "epoch": 2.44, + "learning_rate": 3.783810356128469e-06, + "loss": 1.981, + "step": 11183 + }, + { + "epoch": 2.44, + "learning_rate": 3.7837011142669872e-06, + "loss": 2.5143, + "step": 11184 + }, + { + "epoch": 2.44, + "learning_rate": 3.783591872405506e-06, + "loss": 1.684, + "step": 11185 + }, + { + "epoch": 2.44, + "learning_rate": 3.783482630544025e-06, + "loss": 1.4672, + "step": 11186 + }, + { + "epoch": 2.44, + "learning_rate": 3.7833733886825435e-06, + "loss": 2.046, + "step": 11187 + }, + { + "epoch": 2.44, + "learning_rate": 3.7832641468210625e-06, + "loss": 2.5281, + "step": 11188 + }, + { + "epoch": 2.44, + "learning_rate": 3.7831549049595807e-06, + "loss": 1.9294, + "step": 11189 + }, + { + "epoch": 2.44, + "learning_rate": 3.7830456630980993e-06, + "loss": 1.636, + "step": 11190 + }, + { + "epoch": 2.44, + "learning_rate": 3.7829364212366183e-06, + "loss": 2.5664, + "step": 11191 + }, + { + "epoch": 2.44, + "learning_rate": 3.782827179375137e-06, + "loss": 1.726, + "step": 11192 + }, + { + "epoch": 2.44, + "learning_rate": 3.782717937513656e-06, + "loss": 1.9101, + "step": 11193 + }, + { + "epoch": 2.44, + "learning_rate": 3.782608695652174e-06, + "loss": 2.3418, + "step": 11194 + }, + { + "epoch": 2.44, + "learning_rate": 3.7824994537906928e-06, + "loss": 3.193, + "step": 11195 + }, + { + "epoch": 2.44, + "learning_rate": 3.782390211929212e-06, + "loss": 1.6391, + "step": 11196 + }, + { + "epoch": 2.44, + "learning_rate": 3.7822809700677304e-06, + "loss": 2.8455, + "step": 11197 + }, + { + "epoch": 2.44, + "learning_rate": 3.7821717282062486e-06, + "loss": 3.0555, + "step": 11198 + }, + { + "epoch": 2.44, + "learning_rate": 3.7820624863447677e-06, + "loss": 3.3584, + "step": 11199 + }, + { + "epoch": 2.44, + "learning_rate": 3.7819532444832863e-06, + "loss": 1.7014, + "step": 11200 + }, + { + "epoch": 2.44, + "learning_rate": 3.781844002621805e-06, + "loss": 3.1853, + "step": 11201 + }, + { + "epoch": 2.44, + "learning_rate": 3.781734760760324e-06, + "loss": 3.0029, + "step": 11202 + }, + { + "epoch": 2.45, + "learning_rate": 3.781625518898842e-06, + "loss": 0.9453, + "step": 11203 + }, + { + "epoch": 2.45, + "learning_rate": 3.7815162770373607e-06, + "loss": 2.5927, + "step": 11204 + }, + { + "epoch": 2.45, + "learning_rate": 3.7814070351758797e-06, + "loss": 2.0546, + "step": 11205 + }, + { + "epoch": 2.45, + "learning_rate": 3.7812977933143984e-06, + "loss": 3.5095, + "step": 11206 + }, + { + "epoch": 2.45, + "learning_rate": 3.7811885514529174e-06, + "loss": 2.4103, + "step": 11207 + }, + { + "epoch": 2.45, + "learning_rate": 3.7810793095914356e-06, + "loss": 0.8951, + "step": 11208 + }, + { + "epoch": 2.45, + "learning_rate": 3.780970067729954e-06, + "loss": 1.6521, + "step": 11209 + }, + { + "epoch": 2.45, + "learning_rate": 3.7808608258684732e-06, + "loss": 3.2125, + "step": 11210 + }, + { + "epoch": 2.45, + "learning_rate": 3.780751584006992e-06, + "loss": 2.0037, + "step": 11211 + }, + { + "epoch": 2.45, + "learning_rate": 3.780642342145511e-06, + "loss": 2.6513, + "step": 11212 + }, + { + "epoch": 2.45, + "learning_rate": 3.780533100284029e-06, + "loss": 3.325, + "step": 11213 + }, + { + "epoch": 2.45, + "learning_rate": 3.7804238584225477e-06, + "loss": 3.321, + "step": 11214 + }, + { + "epoch": 2.45, + "learning_rate": 3.7803146165610667e-06, + "loss": 3.7166, + "step": 11215 + }, + { + "epoch": 2.45, + "learning_rate": 3.7802053746995853e-06, + "loss": 2.5206, + "step": 11216 + }, + { + "epoch": 2.45, + "learning_rate": 3.7800961328381035e-06, + "loss": 2.0019, + "step": 11217 + }, + { + "epoch": 2.45, + "learning_rate": 3.7799868909766225e-06, + "loss": 3.1137, + "step": 11218 + }, + { + "epoch": 2.45, + "learning_rate": 3.779877649115141e-06, + "loss": 1.7636, + "step": 11219 + }, + { + "epoch": 2.45, + "learning_rate": 3.77976840725366e-06, + "loss": 2.4519, + "step": 11220 + }, + { + "epoch": 2.45, + "learning_rate": 3.779659165392179e-06, + "loss": 2.5202, + "step": 11221 + }, + { + "epoch": 2.45, + "learning_rate": 3.779549923530697e-06, + "loss": 3.1027, + "step": 11222 + }, + { + "epoch": 2.45, + "learning_rate": 3.779440681669216e-06, + "loss": 3.4285, + "step": 11223 + }, + { + "epoch": 2.45, + "learning_rate": 3.7793314398077346e-06, + "loss": 1.8155, + "step": 11224 + }, + { + "epoch": 2.45, + "learning_rate": 3.7792221979462537e-06, + "loss": 2.3961, + "step": 11225 + }, + { + "epoch": 2.45, + "learning_rate": 3.7791129560847723e-06, + "loss": 2.4061, + "step": 11226 + }, + { + "epoch": 2.45, + "learning_rate": 3.7790037142232905e-06, + "loss": 1.8631, + "step": 11227 + }, + { + "epoch": 2.45, + "learning_rate": 3.7788944723618095e-06, + "loss": 1.9612, + "step": 11228 + }, + { + "epoch": 2.45, + "learning_rate": 3.778785230500328e-06, + "loss": 2.501, + "step": 11229 + }, + { + "epoch": 2.45, + "learning_rate": 3.7786759886388467e-06, + "loss": 3.6726, + "step": 11230 + }, + { + "epoch": 2.45, + "learning_rate": 3.7785667467773653e-06, + "loss": 2.1438, + "step": 11231 + }, + { + "epoch": 2.45, + "learning_rate": 3.778457504915884e-06, + "loss": 1.6465, + "step": 11232 + }, + { + "epoch": 2.45, + "learning_rate": 3.7783482630544025e-06, + "loss": 2.2078, + "step": 11233 + }, + { + "epoch": 2.45, + "learning_rate": 3.7782390211929216e-06, + "loss": 2.2659, + "step": 11234 + }, + { + "epoch": 2.45, + "learning_rate": 3.77812977933144e-06, + "loss": 2.5075, + "step": 11235 + }, + { + "epoch": 2.45, + "learning_rate": 3.7780205374699584e-06, + "loss": 2.8447, + "step": 11236 + }, + { + "epoch": 2.45, + "learning_rate": 3.7779112956084774e-06, + "loss": 1.176, + "step": 11237 + }, + { + "epoch": 2.45, + "learning_rate": 3.777802053746996e-06, + "loss": 1.6767, + "step": 11238 + }, + { + "epoch": 2.45, + "learning_rate": 3.777692811885515e-06, + "loss": 2.5642, + "step": 11239 + }, + { + "epoch": 2.45, + "learning_rate": 3.7775835700240337e-06, + "loss": 1.9054, + "step": 11240 + }, + { + "epoch": 2.45, + "learning_rate": 3.777474328162552e-06, + "loss": 2.3816, + "step": 11241 + }, + { + "epoch": 2.45, + "learning_rate": 3.777365086301071e-06, + "loss": 2.4164, + "step": 11242 + }, + { + "epoch": 2.45, + "learning_rate": 3.7772558444395895e-06, + "loss": 2.5991, + "step": 11243 + }, + { + "epoch": 2.45, + "learning_rate": 3.7771466025781085e-06, + "loss": 2.0153, + "step": 11244 + }, + { + "epoch": 2.45, + "learning_rate": 3.777037360716627e-06, + "loss": 2.2316, + "step": 11245 + }, + { + "epoch": 2.45, + "learning_rate": 3.7769281188551453e-06, + "loss": 1.5592, + "step": 11246 + }, + { + "epoch": 2.45, + "learning_rate": 3.7768188769936644e-06, + "loss": 3.0885, + "step": 11247 + }, + { + "epoch": 2.45, + "learning_rate": 3.776709635132183e-06, + "loss": 1.9389, + "step": 11248 + }, + { + "epoch": 2.46, + "learning_rate": 3.776600393270702e-06, + "loss": 2.7829, + "step": 11249 + }, + { + "epoch": 2.46, + "learning_rate": 3.77649115140922e-06, + "loss": 2.2688, + "step": 11250 + }, + { + "epoch": 2.46, + "learning_rate": 3.776381909547739e-06, + "loss": 3.0729, + "step": 11251 + }, + { + "epoch": 2.46, + "learning_rate": 3.776272667686258e-06, + "loss": 2.6966, + "step": 11252 + }, + { + "epoch": 2.46, + "learning_rate": 3.7761634258247765e-06, + "loss": 1.6119, + "step": 11253 + }, + { + "epoch": 2.46, + "learning_rate": 3.776054183963295e-06, + "loss": 2.5512, + "step": 11254 + }, + { + "epoch": 2.46, + "learning_rate": 3.7759449421018137e-06, + "loss": 2.7043, + "step": 11255 + }, + { + "epoch": 2.46, + "learning_rate": 3.7758357002403323e-06, + "loss": 2.0426, + "step": 11256 + }, + { + "epoch": 2.46, + "learning_rate": 3.775726458378851e-06, + "loss": 3.8325, + "step": 11257 + }, + { + "epoch": 2.46, + "learning_rate": 3.77561721651737e-06, + "loss": 2.4984, + "step": 11258 + }, + { + "epoch": 2.46, + "learning_rate": 3.7755079746558885e-06, + "loss": 2.7291, + "step": 11259 + }, + { + "epoch": 2.46, + "learning_rate": 3.7753987327944067e-06, + "loss": 2.8208, + "step": 11260 + }, + { + "epoch": 2.46, + "learning_rate": 3.7752894909329258e-06, + "loss": 2.3066, + "step": 11261 + }, + { + "epoch": 2.46, + "learning_rate": 3.7751802490714444e-06, + "loss": 3.3537, + "step": 11262 + }, + { + "epoch": 2.46, + "learning_rate": 3.7750710072099634e-06, + "loss": 3.0241, + "step": 11263 + }, + { + "epoch": 2.46, + "learning_rate": 3.7749617653484816e-06, + "loss": 2.3229, + "step": 11264 + }, + { + "epoch": 2.46, + "learning_rate": 3.7748525234870002e-06, + "loss": 3.9785, + "step": 11265 + }, + { + "epoch": 2.46, + "learning_rate": 3.7747432816255192e-06, + "loss": 3.1088, + "step": 11266 + }, + { + "epoch": 2.46, + "learning_rate": 3.774634039764038e-06, + "loss": 2.4791, + "step": 11267 + }, + { + "epoch": 2.46, + "learning_rate": 3.774524797902557e-06, + "loss": 1.6291, + "step": 11268 + }, + { + "epoch": 2.46, + "learning_rate": 3.774415556041075e-06, + "loss": 3.3807, + "step": 11269 + }, + { + "epoch": 2.46, + "learning_rate": 3.7743063141795937e-06, + "loss": 2.2798, + "step": 11270 + }, + { + "epoch": 2.46, + "learning_rate": 3.7741970723181127e-06, + "loss": 2.4897, + "step": 11271 + }, + { + "epoch": 2.46, + "learning_rate": 3.7740878304566313e-06, + "loss": 2.4995, + "step": 11272 + }, + { + "epoch": 2.46, + "learning_rate": 3.7739785885951504e-06, + "loss": 2.0991, + "step": 11273 + }, + { + "epoch": 2.46, + "learning_rate": 3.7738693467336686e-06, + "loss": 2.6392, + "step": 11274 + }, + { + "epoch": 2.46, + "learning_rate": 3.773760104872187e-06, + "loss": 1.0511, + "step": 11275 + }, + { + "epoch": 2.46, + "learning_rate": 3.773650863010706e-06, + "loss": 3.3724, + "step": 11276 + }, + { + "epoch": 2.46, + "learning_rate": 3.773541621149225e-06, + "loss": 1.8589, + "step": 11277 + }, + { + "epoch": 2.46, + "learning_rate": 3.773432379287744e-06, + "loss": 2.7907, + "step": 11278 + }, + { + "epoch": 2.46, + "learning_rate": 3.773323137426262e-06, + "loss": 2.2383, + "step": 11279 + }, + { + "epoch": 2.46, + "learning_rate": 3.7732138955647806e-06, + "loss": 3.7785, + "step": 11280 + }, + { + "epoch": 2.46, + "learning_rate": 3.7731046537032997e-06, + "loss": 1.7811, + "step": 11281 + }, + { + "epoch": 2.46, + "learning_rate": 3.7729954118418183e-06, + "loss": 2.0551, + "step": 11282 + }, + { + "epoch": 2.46, + "learning_rate": 3.7728861699803365e-06, + "loss": 2.3581, + "step": 11283 + }, + { + "epoch": 2.46, + "learning_rate": 3.7727769281188555e-06, + "loss": 2.0709, + "step": 11284 + }, + { + "epoch": 2.46, + "learning_rate": 3.772667686257374e-06, + "loss": 2.6369, + "step": 11285 + }, + { + "epoch": 2.46, + "learning_rate": 3.7725584443958927e-06, + "loss": 1.62, + "step": 11286 + }, + { + "epoch": 2.46, + "learning_rate": 3.7724492025344118e-06, + "loss": 3.7168, + "step": 11287 + }, + { + "epoch": 2.46, + "learning_rate": 3.77233996067293e-06, + "loss": 3.0498, + "step": 11288 + }, + { + "epoch": 2.46, + "learning_rate": 3.7722307188114486e-06, + "loss": 1.8474, + "step": 11289 + }, + { + "epoch": 2.46, + "learning_rate": 3.7721214769499676e-06, + "loss": 3.1434, + "step": 11290 + }, + { + "epoch": 2.46, + "learning_rate": 3.7720122350884862e-06, + "loss": 1.0164, + "step": 11291 + }, + { + "epoch": 2.46, + "learning_rate": 3.7719029932270053e-06, + "loss": 2.0786, + "step": 11292 + }, + { + "epoch": 2.46, + "learning_rate": 3.7717937513655234e-06, + "loss": 2.216, + "step": 11293 + }, + { + "epoch": 2.46, + "learning_rate": 3.771684509504042e-06, + "loss": 2.1001, + "step": 11294 + }, + { + "epoch": 2.47, + "learning_rate": 3.771575267642561e-06, + "loss": 2.0856, + "step": 11295 + }, + { + "epoch": 2.47, + "learning_rate": 3.7714660257810797e-06, + "loss": 2.2853, + "step": 11296 + }, + { + "epoch": 2.47, + "learning_rate": 3.771356783919598e-06, + "loss": 2.6883, + "step": 11297 + }, + { + "epoch": 2.47, + "learning_rate": 3.771247542058117e-06, + "loss": 2.3263, + "step": 11298 + }, + { + "epoch": 2.47, + "learning_rate": 3.7711383001966355e-06, + "loss": 3.5979, + "step": 11299 + }, + { + "epoch": 2.47, + "learning_rate": 3.7710290583351546e-06, + "loss": 2.1962, + "step": 11300 + }, + { + "epoch": 2.47, + "learning_rate": 3.770919816473673e-06, + "loss": 2.5212, + "step": 11301 + }, + { + "epoch": 2.47, + "learning_rate": 3.7708105746121914e-06, + "loss": 2.7662, + "step": 11302 + }, + { + "epoch": 2.47, + "learning_rate": 3.7707013327507104e-06, + "loss": 2.3829, + "step": 11303 + }, + { + "epoch": 2.47, + "learning_rate": 3.770592090889229e-06, + "loss": 2.8401, + "step": 11304 + }, + { + "epoch": 2.47, + "learning_rate": 3.770482849027748e-06, + "loss": 1.4197, + "step": 11305 + }, + { + "epoch": 2.47, + "learning_rate": 3.7703736071662667e-06, + "loss": 2.4706, + "step": 11306 + }, + { + "epoch": 2.47, + "learning_rate": 3.770264365304785e-06, + "loss": 2.6104, + "step": 11307 + }, + { + "epoch": 2.47, + "learning_rate": 3.770155123443304e-06, + "loss": 2.8404, + "step": 11308 + }, + { + "epoch": 2.47, + "learning_rate": 3.7700458815818225e-06, + "loss": 2.3399, + "step": 11309 + }, + { + "epoch": 2.47, + "learning_rate": 3.769936639720341e-06, + "loss": 2.2305, + "step": 11310 + }, + { + "epoch": 2.47, + "learning_rate": 3.76982739785886e-06, + "loss": 1.4767, + "step": 11311 + }, + { + "epoch": 2.47, + "learning_rate": 3.7697181559973783e-06, + "loss": 3.1339, + "step": 11312 + }, + { + "epoch": 2.47, + "learning_rate": 3.769608914135897e-06, + "loss": 3.4682, + "step": 11313 + }, + { + "epoch": 2.47, + "learning_rate": 3.769499672274416e-06, + "loss": 2.6486, + "step": 11314 + }, + { + "epoch": 2.47, + "learning_rate": 3.7693904304129346e-06, + "loss": 2.8855, + "step": 11315 + }, + { + "epoch": 2.47, + "learning_rate": 3.769281188551453e-06, + "loss": 2.8048, + "step": 11316 + }, + { + "epoch": 2.47, + "learning_rate": 3.769171946689972e-06, + "loss": 2.0727, + "step": 11317 + }, + { + "epoch": 2.47, + "learning_rate": 3.7690627048284904e-06, + "loss": 2.0459, + "step": 11318 + }, + { + "epoch": 2.47, + "learning_rate": 3.7689534629670094e-06, + "loss": 3.4632, + "step": 11319 + }, + { + "epoch": 2.47, + "learning_rate": 3.768844221105528e-06, + "loss": 2.7385, + "step": 11320 + }, + { + "epoch": 2.47, + "learning_rate": 3.7687349792440462e-06, + "loss": 2.2802, + "step": 11321 + }, + { + "epoch": 2.47, + "learning_rate": 3.7686257373825653e-06, + "loss": 2.2356, + "step": 11322 + }, + { + "epoch": 2.47, + "learning_rate": 3.768516495521084e-06, + "loss": 3.0841, + "step": 11323 + }, + { + "epoch": 2.47, + "learning_rate": 3.768407253659603e-06, + "loss": 3.8383, + "step": 11324 + }, + { + "epoch": 2.47, + "learning_rate": 3.7682980117981215e-06, + "loss": 1.5282, + "step": 11325 + }, + { + "epoch": 2.47, + "learning_rate": 3.7681887699366397e-06, + "loss": 1.6398, + "step": 11326 + }, + { + "epoch": 2.47, + "learning_rate": 3.7680795280751588e-06, + "loss": 2.6763, + "step": 11327 + }, + { + "epoch": 2.47, + "learning_rate": 3.7679702862136774e-06, + "loss": 1.5975, + "step": 11328 + }, + { + "epoch": 2.47, + "learning_rate": 3.7678610443521964e-06, + "loss": 2.1725, + "step": 11329 + }, + { + "epoch": 2.47, + "learning_rate": 3.7677518024907146e-06, + "loss": 2.1758, + "step": 11330 + }, + { + "epoch": 2.47, + "learning_rate": 3.767642560629233e-06, + "loss": 3.1988, + "step": 11331 + }, + { + "epoch": 2.47, + "learning_rate": 3.7675333187677522e-06, + "loss": 2.2859, + "step": 11332 + }, + { + "epoch": 2.47, + "learning_rate": 3.767424076906271e-06, + "loss": 2.2993, + "step": 11333 + }, + { + "epoch": 2.47, + "learning_rate": 3.76731483504479e-06, + "loss": 1.9159, + "step": 11334 + }, + { + "epoch": 2.47, + "learning_rate": 3.767205593183308e-06, + "loss": 3.1589, + "step": 11335 + }, + { + "epoch": 2.47, + "learning_rate": 3.7670963513218267e-06, + "loss": 2.5386, + "step": 11336 + }, + { + "epoch": 2.47, + "learning_rate": 3.7669871094603457e-06, + "loss": 3.1991, + "step": 11337 + }, + { + "epoch": 2.47, + "learning_rate": 3.7668778675988643e-06, + "loss": 2.5677, + "step": 11338 + }, + { + "epoch": 2.47, + "learning_rate": 3.766768625737383e-06, + "loss": 1.5628, + "step": 11339 + }, + { + "epoch": 2.47, + "learning_rate": 3.7666593838759015e-06, + "loss": 3.2517, + "step": 11340 + }, + { + "epoch": 2.48, + "learning_rate": 3.76655014201442e-06, + "loss": 2.8914, + "step": 11341 + }, + { + "epoch": 2.48, + "learning_rate": 3.7664409001529388e-06, + "loss": 3.0026, + "step": 11342 + }, + { + "epoch": 2.48, + "learning_rate": 3.766331658291458e-06, + "loss": 2.2516, + "step": 11343 + }, + { + "epoch": 2.48, + "learning_rate": 3.7662224164299764e-06, + "loss": 1.555, + "step": 11344 + }, + { + "epoch": 2.48, + "learning_rate": 3.7661131745684946e-06, + "loss": 1.6294, + "step": 11345 + }, + { + "epoch": 2.48, + "learning_rate": 3.7660039327070136e-06, + "loss": 1.9609, + "step": 11346 + }, + { + "epoch": 2.48, + "learning_rate": 3.7658946908455322e-06, + "loss": 1.777, + "step": 11347 + }, + { + "epoch": 2.48, + "learning_rate": 3.7657854489840513e-06, + "loss": 1.9766, + "step": 11348 + }, + { + "epoch": 2.48, + "learning_rate": 3.7656762071225695e-06, + "loss": 1.9011, + "step": 11349 + }, + { + "epoch": 2.48, + "learning_rate": 3.765566965261088e-06, + "loss": 2.7532, + "step": 11350 + }, + { + "epoch": 2.48, + "learning_rate": 3.765457723399607e-06, + "loss": 2.3092, + "step": 11351 + }, + { + "epoch": 2.48, + "learning_rate": 3.7653484815381257e-06, + "loss": 3.8118, + "step": 11352 + }, + { + "epoch": 2.48, + "learning_rate": 3.7652392396766448e-06, + "loss": 2.6688, + "step": 11353 + }, + { + "epoch": 2.48, + "learning_rate": 3.765129997815163e-06, + "loss": 2.4072, + "step": 11354 + }, + { + "epoch": 2.48, + "learning_rate": 3.7650207559536816e-06, + "loss": 3.3826, + "step": 11355 + }, + { + "epoch": 2.48, + "learning_rate": 3.7649115140922006e-06, + "loss": 3.0542, + "step": 11356 + }, + { + "epoch": 2.48, + "learning_rate": 3.764802272230719e-06, + "loss": 1.6902, + "step": 11357 + }, + { + "epoch": 2.48, + "learning_rate": 3.7646930303692382e-06, + "loss": 2.3952, + "step": 11358 + }, + { + "epoch": 2.48, + "learning_rate": 3.7645837885077564e-06, + "loss": 2.6269, + "step": 11359 + }, + { + "epoch": 2.48, + "learning_rate": 3.764474546646275e-06, + "loss": 3.3366, + "step": 11360 + }, + { + "epoch": 2.48, + "learning_rate": 3.764365304784794e-06, + "loss": 2.5883, + "step": 11361 + }, + { + "epoch": 2.48, + "learning_rate": 3.7642560629233127e-06, + "loss": 2.3115, + "step": 11362 + }, + { + "epoch": 2.48, + "learning_rate": 3.764146821061831e-06, + "loss": 3.1562, + "step": 11363 + }, + { + "epoch": 2.48, + "learning_rate": 3.76403757920035e-06, + "loss": 2.425, + "step": 11364 + }, + { + "epoch": 2.48, + "learning_rate": 3.7639283373388685e-06, + "loss": 1.8525, + "step": 11365 + }, + { + "epoch": 2.48, + "learning_rate": 3.763819095477387e-06, + "loss": 2.5526, + "step": 11366 + }, + { + "epoch": 2.48, + "learning_rate": 3.763709853615906e-06, + "loss": 2.5294, + "step": 11367 + }, + { + "epoch": 2.48, + "learning_rate": 3.7636006117544243e-06, + "loss": 1.5933, + "step": 11368 + }, + { + "epoch": 2.48, + "learning_rate": 3.7634913698929434e-06, + "loss": 1.3355, + "step": 11369 + }, + { + "epoch": 2.48, + "learning_rate": 3.763382128031462e-06, + "loss": 2.2248, + "step": 11370 + }, + { + "epoch": 2.48, + "learning_rate": 3.7632728861699806e-06, + "loss": 1.7021, + "step": 11371 + }, + { + "epoch": 2.48, + "learning_rate": 3.7631636443084996e-06, + "loss": 3.223, + "step": 11372 + }, + { + "epoch": 2.48, + "learning_rate": 3.763054402447018e-06, + "loss": 2.4411, + "step": 11373 + }, + { + "epoch": 2.48, + "learning_rate": 3.7629451605855364e-06, + "loss": 1.8252, + "step": 11374 + }, + { + "epoch": 2.48, + "learning_rate": 3.7628359187240555e-06, + "loss": 2.9662, + "step": 11375 + }, + { + "epoch": 2.48, + "learning_rate": 3.762726676862574e-06, + "loss": 3.3866, + "step": 11376 + }, + { + "epoch": 2.48, + "learning_rate": 3.762617435001093e-06, + "loss": 1.3966, + "step": 11377 + }, + { + "epoch": 2.48, + "learning_rate": 3.7625081931396113e-06, + "loss": 1.6946, + "step": 11378 + }, + { + "epoch": 2.48, + "learning_rate": 3.76239895127813e-06, + "loss": 1.3601, + "step": 11379 + }, + { + "epoch": 2.48, + "learning_rate": 3.762289709416649e-06, + "loss": 1.5769, + "step": 11380 + }, + { + "epoch": 2.48, + "learning_rate": 3.7621804675551676e-06, + "loss": 2.3994, + "step": 11381 + }, + { + "epoch": 2.48, + "learning_rate": 3.7620712256936857e-06, + "loss": 1.3968, + "step": 11382 + }, + { + "epoch": 2.48, + "learning_rate": 3.7619619838322048e-06, + "loss": 2.7062, + "step": 11383 + }, + { + "epoch": 2.48, + "learning_rate": 3.7618527419707234e-06, + "loss": 3.2178, + "step": 11384 + }, + { + "epoch": 2.48, + "learning_rate": 3.7617435001092424e-06, + "loss": 2.8044, + "step": 11385 + }, + { + "epoch": 2.48, + "learning_rate": 3.761634258247761e-06, + "loss": 3.0351, + "step": 11386 + }, + { + "epoch": 2.49, + "learning_rate": 3.7615250163862792e-06, + "loss": 2.4359, + "step": 11387 + }, + { + "epoch": 2.49, + "learning_rate": 3.7614157745247983e-06, + "loss": 3.4001, + "step": 11388 + }, + { + "epoch": 2.49, + "learning_rate": 3.761306532663317e-06, + "loss": 2.2966, + "step": 11389 + }, + { + "epoch": 2.49, + "learning_rate": 3.761197290801836e-06, + "loss": 1.4339, + "step": 11390 + }, + { + "epoch": 2.49, + "learning_rate": 3.7610880489403545e-06, + "loss": 2.9624, + "step": 11391 + }, + { + "epoch": 2.49, + "learning_rate": 3.7609788070788727e-06, + "loss": 2.3349, + "step": 11392 + }, + { + "epoch": 2.49, + "learning_rate": 3.7608695652173917e-06, + "loss": 2.0502, + "step": 11393 + }, + { + "epoch": 2.49, + "learning_rate": 3.7607603233559103e-06, + "loss": 2.3458, + "step": 11394 + }, + { + "epoch": 2.49, + "learning_rate": 3.760651081494429e-06, + "loss": 2.4671, + "step": 11395 + }, + { + "epoch": 2.49, + "learning_rate": 3.7605418396329476e-06, + "loss": 3.4245, + "step": 11396 + }, + { + "epoch": 2.49, + "learning_rate": 3.760432597771466e-06, + "loss": 2.5771, + "step": 11397 + }, + { + "epoch": 2.49, + "learning_rate": 3.760323355909985e-06, + "loss": 3.3048, + "step": 11398 + }, + { + "epoch": 2.49, + "learning_rate": 3.760214114048504e-06, + "loss": 1.3121, + "step": 11399 + }, + { + "epoch": 2.49, + "learning_rate": 3.7601048721870224e-06, + "loss": 2.5653, + "step": 11400 + }, + { + "epoch": 2.49, + "learning_rate": 3.7599956303255406e-06, + "loss": 2.4503, + "step": 11401 + }, + { + "epoch": 2.49, + "learning_rate": 3.7598863884640597e-06, + "loss": 3.1765, + "step": 11402 + }, + { + "epoch": 2.49, + "learning_rate": 3.7597771466025783e-06, + "loss": 2.6282, + "step": 11403 + }, + { + "epoch": 2.49, + "learning_rate": 3.7596679047410973e-06, + "loss": 0.8611, + "step": 11404 + }, + { + "epoch": 2.49, + "learning_rate": 3.759558662879616e-06, + "loss": 1.5375, + "step": 11405 + }, + { + "epoch": 2.49, + "learning_rate": 3.759449421018134e-06, + "loss": 3.0375, + "step": 11406 + }, + { + "epoch": 2.49, + "learning_rate": 3.759340179156653e-06, + "loss": 1.9361, + "step": 11407 + }, + { + "epoch": 2.49, + "learning_rate": 3.7592309372951717e-06, + "loss": 2.4992, + "step": 11408 + }, + { + "epoch": 2.49, + "learning_rate": 3.7591216954336908e-06, + "loss": 1.9689, + "step": 11409 + }, + { + "epoch": 2.49, + "learning_rate": 3.7590124535722094e-06, + "loss": 1.5298, + "step": 11410 + }, + { + "epoch": 2.49, + "learning_rate": 3.7589032117107276e-06, + "loss": 2.1316, + "step": 11411 + }, + { + "epoch": 2.49, + "learning_rate": 3.7587939698492466e-06, + "loss": 2.3544, + "step": 11412 + }, + { + "epoch": 2.49, + "learning_rate": 3.7586847279877652e-06, + "loss": 3.3777, + "step": 11413 + }, + { + "epoch": 2.49, + "learning_rate": 3.7585754861262843e-06, + "loss": 1.6639, + "step": 11414 + }, + { + "epoch": 2.49, + "learning_rate": 3.7584662442648024e-06, + "loss": 2.4862, + "step": 11415 + }, + { + "epoch": 2.49, + "learning_rate": 3.758357002403321e-06, + "loss": 2.1181, + "step": 11416 + }, + { + "epoch": 2.49, + "learning_rate": 3.75824776054184e-06, + "loss": 2.8283, + "step": 11417 + }, + { + "epoch": 2.49, + "learning_rate": 3.7581385186803587e-06, + "loss": 3.4087, + "step": 11418 + }, + { + "epoch": 2.49, + "learning_rate": 3.7580292768188773e-06, + "loss": 2.7102, + "step": 11419 + }, + { + "epoch": 2.49, + "learning_rate": 3.757920034957396e-06, + "loss": 3.0554, + "step": 11420 + }, + { + "epoch": 2.49, + "learning_rate": 3.7578107930959145e-06, + "loss": 2.1103, + "step": 11421 + }, + { + "epoch": 2.49, + "learning_rate": 3.7577015512344336e-06, + "loss": 3.7541, + "step": 11422 + }, + { + "epoch": 2.49, + "learning_rate": 3.757592309372952e-06, + "loss": 3.596, + "step": 11423 + }, + { + "epoch": 2.49, + "learning_rate": 3.757483067511471e-06, + "loss": 1.4179, + "step": 11424 + }, + { + "epoch": 2.49, + "learning_rate": 3.7573738256499894e-06, + "loss": 1.8042, + "step": 11425 + }, + { + "epoch": 2.49, + "learning_rate": 3.757264583788508e-06, + "loss": 3.7221, + "step": 11426 + }, + { + "epoch": 2.49, + "learning_rate": 3.7571553419270266e-06, + "loss": 2.0635, + "step": 11427 + }, + { + "epoch": 2.49, + "learning_rate": 3.7570461000655457e-06, + "loss": 2.2673, + "step": 11428 + }, + { + "epoch": 2.49, + "learning_rate": 3.756936858204064e-06, + "loss": 2.8051, + "step": 11429 + }, + { + "epoch": 2.49, + "learning_rate": 3.7568276163425825e-06, + "loss": 0.9697, + "step": 11430 + }, + { + "epoch": 2.49, + "learning_rate": 3.7567183744811015e-06, + "loss": 1.7404, + "step": 11431 + }, + { + "epoch": 2.49, + "learning_rate": 3.75660913261962e-06, + "loss": 2.5792, + "step": 11432 + }, + { + "epoch": 2.5, + "learning_rate": 3.756499890758139e-06, + "loss": 3.3284, + "step": 11433 + }, + { + "epoch": 2.5, + "learning_rate": 3.7563906488966573e-06, + "loss": 2.5521, + "step": 11434 + }, + { + "epoch": 2.5, + "learning_rate": 3.756281407035176e-06, + "loss": 3.3066, + "step": 11435 + }, + { + "epoch": 2.5, + "learning_rate": 3.756172165173695e-06, + "loss": 2.7474, + "step": 11436 + }, + { + "epoch": 2.5, + "learning_rate": 3.7560629233122136e-06, + "loss": 2.309, + "step": 11437 + }, + { + "epoch": 2.5, + "learning_rate": 3.7559536814507326e-06, + "loss": 1.7953, + "step": 11438 + }, + { + "epoch": 2.5, + "learning_rate": 3.755844439589251e-06, + "loss": 2.1735, + "step": 11439 + }, + { + "epoch": 2.5, + "learning_rate": 3.7557351977277694e-06, + "loss": 1.9126, + "step": 11440 + }, + { + "epoch": 2.5, + "learning_rate": 3.7556259558662885e-06, + "loss": 3.3327, + "step": 11441 + }, + { + "epoch": 2.5, + "learning_rate": 3.755516714004807e-06, + "loss": 3.0121, + "step": 11442 + }, + { + "epoch": 2.5, + "learning_rate": 3.755407472143326e-06, + "loss": 2.6462, + "step": 11443 + }, + { + "epoch": 2.5, + "learning_rate": 3.7552982302818443e-06, + "loss": 2.3, + "step": 11444 + }, + { + "epoch": 2.5, + "learning_rate": 3.755188988420363e-06, + "loss": 2.0236, + "step": 11445 + }, + { + "epoch": 2.5, + "learning_rate": 3.755079746558882e-06, + "loss": 2.7596, + "step": 11446 + }, + { + "epoch": 2.5, + "learning_rate": 3.7549705046974005e-06, + "loss": 2.7365, + "step": 11447 + }, + { + "epoch": 2.5, + "learning_rate": 3.7548612628359187e-06, + "loss": 1.0956, + "step": 11448 + }, + { + "epoch": 2.5, + "learning_rate": 3.7547520209744378e-06, + "loss": 3.0097, + "step": 11449 + }, + { + "epoch": 2.5, + "learning_rate": 3.7546427791129564e-06, + "loss": 1.8055, + "step": 11450 + }, + { + "epoch": 2.5, + "learning_rate": 3.754533537251475e-06, + "loss": 3.9353, + "step": 11451 + }, + { + "epoch": 2.5, + "learning_rate": 3.754424295389994e-06, + "loss": 2.3962, + "step": 11452 + }, + { + "epoch": 2.5, + "learning_rate": 3.754315053528512e-06, + "loss": 2.3328, + "step": 11453 + }, + { + "epoch": 2.5, + "learning_rate": 3.754205811667031e-06, + "loss": 1.2343, + "step": 11454 + }, + { + "epoch": 2.5, + "learning_rate": 3.75409656980555e-06, + "loss": 2.1232, + "step": 11455 + }, + { + "epoch": 2.5, + "learning_rate": 3.7539873279440685e-06, + "loss": 2.7882, + "step": 11456 + }, + { + "epoch": 2.5, + "learning_rate": 3.7538780860825875e-06, + "loss": 1.7229, + "step": 11457 + }, + { + "epoch": 2.5, + "learning_rate": 3.7537688442211057e-06, + "loss": 2.4735, + "step": 11458 + }, + { + "epoch": 2.5, + "learning_rate": 3.7536596023596243e-06, + "loss": 3.1807, + "step": 11459 + }, + { + "epoch": 2.5, + "learning_rate": 3.7535503604981433e-06, + "loss": 1.6608, + "step": 11460 + }, + { + "epoch": 2.5, + "learning_rate": 3.753441118636662e-06, + "loss": 2.5611, + "step": 11461 + }, + { + "epoch": 2.5, + "learning_rate": 3.75333187677518e-06, + "loss": 2.5079, + "step": 11462 + }, + { + "epoch": 2.5, + "learning_rate": 3.753222634913699e-06, + "loss": 2.4397, + "step": 11463 + }, + { + "epoch": 2.5, + "learning_rate": 3.7531133930522178e-06, + "loss": 1.3953, + "step": 11464 + }, + { + "epoch": 2.5, + "learning_rate": 3.753004151190737e-06, + "loss": 2.4796, + "step": 11465 + }, + { + "epoch": 2.5, + "learning_rate": 3.7528949093292554e-06, + "loss": 3.1955, + "step": 11466 + }, + { + "epoch": 2.5, + "learning_rate": 3.7527856674677736e-06, + "loss": 2.9502, + "step": 11467 + }, + { + "epoch": 2.5, + "learning_rate": 3.7526764256062926e-06, + "loss": 1.4044, + "step": 11468 + }, + { + "epoch": 2.5, + "learning_rate": 3.7525671837448113e-06, + "loss": 0.9212, + "step": 11469 + }, + { + "epoch": 2.5, + "learning_rate": 3.7524579418833303e-06, + "loss": 2.3528, + "step": 11470 + }, + { + "epoch": 2.5, + "learning_rate": 3.752348700021849e-06, + "loss": 1.6617, + "step": 11471 + }, + { + "epoch": 2.5, + "learning_rate": 3.752239458160367e-06, + "loss": 1.7385, + "step": 11472 + }, + { + "epoch": 2.5, + "learning_rate": 3.752130216298886e-06, + "loss": 1.8009, + "step": 11473 + }, + { + "epoch": 2.5, + "learning_rate": 3.7520209744374047e-06, + "loss": 1.6861, + "step": 11474 + }, + { + "epoch": 2.5, + "learning_rate": 3.7519117325759238e-06, + "loss": 1.8293, + "step": 11475 + }, + { + "epoch": 2.5, + "learning_rate": 3.7518024907144424e-06, + "loss": 2.242, + "step": 11476 + }, + { + "epoch": 2.5, + "learning_rate": 3.7516932488529606e-06, + "loss": 2.8135, + "step": 11477 + }, + { + "epoch": 2.51, + "learning_rate": 3.7515840069914796e-06, + "loss": 1.8796, + "step": 11478 + }, + { + "epoch": 2.51, + "learning_rate": 3.751474765129998e-06, + "loss": 1.898, + "step": 11479 + }, + { + "epoch": 2.51, + "learning_rate": 3.751365523268517e-06, + "loss": 2.3798, + "step": 11480 + }, + { + "epoch": 2.51, + "learning_rate": 3.7512562814070354e-06, + "loss": 1.704, + "step": 11481 + }, + { + "epoch": 2.51, + "learning_rate": 3.751147039545554e-06, + "loss": 1.4451, + "step": 11482 + }, + { + "epoch": 2.51, + "learning_rate": 3.7510377976840727e-06, + "loss": 3.6814, + "step": 11483 + }, + { + "epoch": 2.51, + "learning_rate": 3.7509285558225917e-06, + "loss": 2.9213, + "step": 11484 + }, + { + "epoch": 2.51, + "learning_rate": 3.7508193139611103e-06, + "loss": 2.0756, + "step": 11485 + }, + { + "epoch": 2.51, + "learning_rate": 3.7507100720996285e-06, + "loss": 2.0285, + "step": 11486 + }, + { + "epoch": 2.51, + "learning_rate": 3.7506008302381475e-06, + "loss": 2.8007, + "step": 11487 + }, + { + "epoch": 2.51, + "learning_rate": 3.750491588376666e-06, + "loss": 2.1859, + "step": 11488 + }, + { + "epoch": 2.51, + "learning_rate": 3.750382346515185e-06, + "loss": 2.8858, + "step": 11489 + }, + { + "epoch": 2.51, + "learning_rate": 3.7502731046537038e-06, + "loss": 1.2417, + "step": 11490 + }, + { + "epoch": 2.51, + "learning_rate": 3.750163862792222e-06, + "loss": 1.6678, + "step": 11491 + }, + { + "epoch": 2.51, + "learning_rate": 3.750054620930741e-06, + "loss": 2.4203, + "step": 11492 + }, + { + "epoch": 2.51, + "learning_rate": 3.7499453790692596e-06, + "loss": 2.3526, + "step": 11493 + }, + { + "epoch": 2.51, + "learning_rate": 3.7498361372077786e-06, + "loss": 1.0069, + "step": 11494 + }, + { + "epoch": 2.51, + "learning_rate": 3.749726895346297e-06, + "loss": 2.6728, + "step": 11495 + }, + { + "epoch": 2.51, + "learning_rate": 3.7496176534848154e-06, + "loss": 2.5731, + "step": 11496 + }, + { + "epoch": 2.51, + "learning_rate": 3.7495084116233345e-06, + "loss": 2.5406, + "step": 11497 + }, + { + "epoch": 2.51, + "learning_rate": 3.749399169761853e-06, + "loss": 2.2874, + "step": 11498 + }, + { + "epoch": 2.51, + "learning_rate": 3.749289927900372e-06, + "loss": 3.4147, + "step": 11499 + }, + { + "epoch": 2.51, + "learning_rate": 3.7491806860388903e-06, + "loss": 2.0872, + "step": 11500 + }, + { + "epoch": 2.51, + "learning_rate": 3.749071444177409e-06, + "loss": 2.3817, + "step": 11501 + }, + { + "epoch": 2.51, + "learning_rate": 3.748962202315928e-06, + "loss": 3.4249, + "step": 11502 + }, + { + "epoch": 2.51, + "learning_rate": 3.7488529604544466e-06, + "loss": 2.0968, + "step": 11503 + }, + { + "epoch": 2.51, + "learning_rate": 3.748743718592965e-06, + "loss": 3.4874, + "step": 11504 + }, + { + "epoch": 2.51, + "learning_rate": 3.748634476731484e-06, + "loss": 0.8314, + "step": 11505 + }, + { + "epoch": 2.51, + "learning_rate": 3.7485252348700024e-06, + "loss": 1.0873, + "step": 11506 + }, + { + "epoch": 2.51, + "learning_rate": 3.748415993008521e-06, + "loss": 1.1967, + "step": 11507 + }, + { + "epoch": 2.51, + "learning_rate": 3.74830675114704e-06, + "loss": 2.446, + "step": 11508 + }, + { + "epoch": 2.51, + "learning_rate": 3.7481975092855582e-06, + "loss": 3.0221, + "step": 11509 + }, + { + "epoch": 2.51, + "learning_rate": 3.748088267424077e-06, + "loss": 2.5711, + "step": 11510 + }, + { + "epoch": 2.51, + "learning_rate": 3.747979025562596e-06, + "loss": 2.6936, + "step": 11511 + }, + { + "epoch": 2.51, + "learning_rate": 3.7478697837011145e-06, + "loss": 4.0541, + "step": 11512 + }, + { + "epoch": 2.51, + "learning_rate": 3.7477605418396335e-06, + "loss": 1.7476, + "step": 11513 + }, + { + "epoch": 2.51, + "learning_rate": 3.7476512999781517e-06, + "loss": 3.0301, + "step": 11514 + }, + { + "epoch": 2.51, + "learning_rate": 3.7475420581166703e-06, + "loss": 2.875, + "step": 11515 + }, + { + "epoch": 2.51, + "learning_rate": 3.7474328162551894e-06, + "loss": 2.1106, + "step": 11516 + }, + { + "epoch": 2.51, + "learning_rate": 3.747323574393708e-06, + "loss": 2.9628, + "step": 11517 + }, + { + "epoch": 2.51, + "learning_rate": 3.747214332532227e-06, + "loss": 2.2354, + "step": 11518 + }, + { + "epoch": 2.51, + "learning_rate": 3.747105090670745e-06, + "loss": 2.3821, + "step": 11519 + }, + { + "epoch": 2.51, + "learning_rate": 3.746995848809264e-06, + "loss": 2.8309, + "step": 11520 + }, + { + "epoch": 2.51, + "learning_rate": 3.746886606947783e-06, + "loss": 1.5169, + "step": 11521 + }, + { + "epoch": 2.51, + "learning_rate": 3.7467773650863014e-06, + "loss": 3.1471, + "step": 11522 + }, + { + "epoch": 2.51, + "learning_rate": 3.7466681232248205e-06, + "loss": 1.8598, + "step": 11523 + }, + { + "epoch": 2.52, + "learning_rate": 3.7465588813633387e-06, + "loss": 2.6574, + "step": 11524 + }, + { + "epoch": 2.52, + "learning_rate": 3.7464496395018573e-06, + "loss": 3.2811, + "step": 11525 + }, + { + "epoch": 2.52, + "learning_rate": 3.7463403976403763e-06, + "loss": 1.6138, + "step": 11526 + }, + { + "epoch": 2.52, + "learning_rate": 3.746231155778895e-06, + "loss": 3.1715, + "step": 11527 + }, + { + "epoch": 2.52, + "learning_rate": 3.746121913917413e-06, + "loss": 3.437, + "step": 11528 + }, + { + "epoch": 2.52, + "learning_rate": 3.746012672055932e-06, + "loss": 3.881, + "step": 11529 + }, + { + "epoch": 2.52, + "learning_rate": 3.7459034301944508e-06, + "loss": 3.1338, + "step": 11530 + }, + { + "epoch": 2.52, + "learning_rate": 3.74579418833297e-06, + "loss": 1.4089, + "step": 11531 + }, + { + "epoch": 2.52, + "learning_rate": 3.7456849464714884e-06, + "loss": 1.8614, + "step": 11532 + }, + { + "epoch": 2.52, + "learning_rate": 3.7455757046100066e-06, + "loss": 3.2885, + "step": 11533 + }, + { + "epoch": 2.52, + "learning_rate": 3.7454664627485256e-06, + "loss": 1.7258, + "step": 11534 + }, + { + "epoch": 2.52, + "learning_rate": 3.7453572208870442e-06, + "loss": 3.1611, + "step": 11535 + }, + { + "epoch": 2.52, + "learning_rate": 3.745247979025563e-06, + "loss": 1.5953, + "step": 11536 + }, + { + "epoch": 2.52, + "learning_rate": 3.745138737164082e-06, + "loss": 2.5035, + "step": 11537 + }, + { + "epoch": 2.52, + "learning_rate": 3.7450294953026e-06, + "loss": 2.2905, + "step": 11538 + }, + { + "epoch": 2.52, + "learning_rate": 3.7449202534411187e-06, + "loss": 2.6296, + "step": 11539 + }, + { + "epoch": 2.52, + "learning_rate": 3.7448110115796377e-06, + "loss": 2.1351, + "step": 11540 + }, + { + "epoch": 2.52, + "learning_rate": 3.7447017697181563e-06, + "loss": 1.6522, + "step": 11541 + }, + { + "epoch": 2.52, + "learning_rate": 3.7445925278566745e-06, + "loss": 2.0005, + "step": 11542 + }, + { + "epoch": 2.52, + "learning_rate": 3.7444832859951935e-06, + "loss": 2.4586, + "step": 11543 + }, + { + "epoch": 2.52, + "learning_rate": 3.744374044133712e-06, + "loss": 1.6699, + "step": 11544 + }, + { + "epoch": 2.52, + "learning_rate": 3.744264802272231e-06, + "loss": 1.8906, + "step": 11545 + }, + { + "epoch": 2.52, + "learning_rate": 3.74415556041075e-06, + "loss": 1.9014, + "step": 11546 + }, + { + "epoch": 2.52, + "learning_rate": 3.744046318549268e-06, + "loss": 2.0075, + "step": 11547 + }, + { + "epoch": 2.52, + "learning_rate": 3.743937076687787e-06, + "loss": 1.6274, + "step": 11548 + }, + { + "epoch": 2.52, + "learning_rate": 3.7438278348263056e-06, + "loss": 2.8565, + "step": 11549 + }, + { + "epoch": 2.52, + "learning_rate": 3.7437185929648247e-06, + "loss": 1.633, + "step": 11550 + }, + { + "epoch": 2.52, + "learning_rate": 3.7436093511033433e-06, + "loss": 2.3168, + "step": 11551 + }, + { + "epoch": 2.52, + "learning_rate": 3.7435001092418615e-06, + "loss": 2.827, + "step": 11552 + }, + { + "epoch": 2.52, + "learning_rate": 3.7433908673803805e-06, + "loss": 3.1484, + "step": 11553 + }, + { + "epoch": 2.52, + "learning_rate": 3.743281625518899e-06, + "loss": 2.4211, + "step": 11554 + }, + { + "epoch": 2.52, + "learning_rate": 3.743172383657418e-06, + "loss": 2.5602, + "step": 11555 + }, + { + "epoch": 2.52, + "learning_rate": 3.7430631417959368e-06, + "loss": 2.695, + "step": 11556 + }, + { + "epoch": 2.52, + "learning_rate": 3.742953899934455e-06, + "loss": 1.5279, + "step": 11557 + }, + { + "epoch": 2.52, + "learning_rate": 3.742844658072974e-06, + "loss": 1.7487, + "step": 11558 + }, + { + "epoch": 2.52, + "learning_rate": 3.7427354162114926e-06, + "loss": 1.8985, + "step": 11559 + }, + { + "epoch": 2.52, + "learning_rate": 3.742626174350011e-06, + "loss": 2.2452, + "step": 11560 + }, + { + "epoch": 2.52, + "learning_rate": 3.74251693248853e-06, + "loss": 2.9926, + "step": 11561 + }, + { + "epoch": 2.52, + "learning_rate": 3.7424076906270484e-06, + "loss": 2.5272, + "step": 11562 + }, + { + "epoch": 2.52, + "learning_rate": 3.742298448765567e-06, + "loss": 1.6691, + "step": 11563 + }, + { + "epoch": 2.52, + "learning_rate": 3.742189206904086e-06, + "loss": 1.9617, + "step": 11564 + }, + { + "epoch": 2.52, + "learning_rate": 3.7420799650426047e-06, + "loss": 2.5563, + "step": 11565 + }, + { + "epoch": 2.52, + "learning_rate": 3.7419707231811233e-06, + "loss": 2.1056, + "step": 11566 + }, + { + "epoch": 2.52, + "learning_rate": 3.741861481319642e-06, + "loss": 1.7584, + "step": 11567 + }, + { + "epoch": 2.52, + "learning_rate": 3.7417522394581605e-06, + "loss": 3.6263, + "step": 11568 + }, + { + "epoch": 2.52, + "learning_rate": 3.7416429975966796e-06, + "loss": 3.3899, + "step": 11569 + }, + { + "epoch": 2.53, + "learning_rate": 3.741533755735198e-06, + "loss": 3.0787, + "step": 11570 + }, + { + "epoch": 2.53, + "learning_rate": 3.7414245138737164e-06, + "loss": 1.6544, + "step": 11571 + }, + { + "epoch": 2.53, + "learning_rate": 3.7413152720122354e-06, + "loss": 2.3835, + "step": 11572 + }, + { + "epoch": 2.53, + "learning_rate": 3.741206030150754e-06, + "loss": 4.2429, + "step": 11573 + }, + { + "epoch": 2.53, + "learning_rate": 3.741096788289273e-06, + "loss": 2.1889, + "step": 11574 + }, + { + "epoch": 2.53, + "learning_rate": 3.7409875464277912e-06, + "loss": 2.1709, + "step": 11575 + }, + { + "epoch": 2.53, + "learning_rate": 3.74087830456631e-06, + "loss": 2.0419, + "step": 11576 + }, + { + "epoch": 2.53, + "learning_rate": 3.740769062704829e-06, + "loss": 2.2686, + "step": 11577 + }, + { + "epoch": 2.53, + "learning_rate": 3.7406598208433475e-06, + "loss": 1.8069, + "step": 11578 + }, + { + "epoch": 2.53, + "learning_rate": 3.7405505789818665e-06, + "loss": 2.4233, + "step": 11579 + }, + { + "epoch": 2.53, + "learning_rate": 3.7404413371203847e-06, + "loss": 1.7883, + "step": 11580 + }, + { + "epoch": 2.53, + "learning_rate": 3.7403320952589033e-06, + "loss": 2.7881, + "step": 11581 + }, + { + "epoch": 2.53, + "learning_rate": 3.7402228533974223e-06, + "loss": 3.013, + "step": 11582 + }, + { + "epoch": 2.53, + "learning_rate": 3.740113611535941e-06, + "loss": 2.247, + "step": 11583 + }, + { + "epoch": 2.53, + "learning_rate": 3.74000436967446e-06, + "loss": 2.8906, + "step": 11584 + }, + { + "epoch": 2.53, + "learning_rate": 3.739895127812978e-06, + "loss": 1.3278, + "step": 11585 + }, + { + "epoch": 2.53, + "learning_rate": 3.7397858859514968e-06, + "loss": 1.9181, + "step": 11586 + }, + { + "epoch": 2.53, + "learning_rate": 3.739676644090016e-06, + "loss": 2.2731, + "step": 11587 + }, + { + "epoch": 2.53, + "learning_rate": 3.7395674022285344e-06, + "loss": 3.3471, + "step": 11588 + }, + { + "epoch": 2.53, + "learning_rate": 3.739458160367053e-06, + "loss": 1.5949, + "step": 11589 + }, + { + "epoch": 2.53, + "learning_rate": 3.7393489185055717e-06, + "loss": 2.3105, + "step": 11590 + }, + { + "epoch": 2.53, + "learning_rate": 3.7392396766440903e-06, + "loss": 2.4077, + "step": 11591 + }, + { + "epoch": 2.53, + "learning_rate": 3.739130434782609e-06, + "loss": 2.1786, + "step": 11592 + }, + { + "epoch": 2.53, + "learning_rate": 3.739021192921128e-06, + "loss": 3.3351, + "step": 11593 + }, + { + "epoch": 2.53, + "learning_rate": 3.738911951059646e-06, + "loss": 2.5285, + "step": 11594 + }, + { + "epoch": 2.53, + "learning_rate": 3.7388027091981647e-06, + "loss": 2.638, + "step": 11595 + }, + { + "epoch": 2.53, + "learning_rate": 3.7386934673366837e-06, + "loss": 2.8207, + "step": 11596 + }, + { + "epoch": 2.53, + "learning_rate": 3.7385842254752024e-06, + "loss": 3.2287, + "step": 11597 + }, + { + "epoch": 2.53, + "learning_rate": 3.7384749836137214e-06, + "loss": 2.0817, + "step": 11598 + }, + { + "epoch": 2.53, + "learning_rate": 3.7383657417522396e-06, + "loss": 2.4072, + "step": 11599 + }, + { + "epoch": 2.53, + "learning_rate": 3.738256499890758e-06, + "loss": 1.6356, + "step": 11600 + }, + { + "epoch": 2.53, + "learning_rate": 3.7381472580292772e-06, + "loss": 2.3748, + "step": 11601 + }, + { + "epoch": 2.53, + "learning_rate": 3.738038016167796e-06, + "loss": 1.6976, + "step": 11602 + }, + { + "epoch": 2.53, + "learning_rate": 3.737928774306315e-06, + "loss": 1.5847, + "step": 11603 + }, + { + "epoch": 2.53, + "learning_rate": 3.737819532444833e-06, + "loss": 2.0488, + "step": 11604 + }, + { + "epoch": 2.53, + "learning_rate": 3.7377102905833517e-06, + "loss": 2.0657, + "step": 11605 + }, + { + "epoch": 2.53, + "learning_rate": 3.7376010487218707e-06, + "loss": 2.9613, + "step": 11606 + }, + { + "epoch": 2.53, + "learning_rate": 3.7374918068603893e-06, + "loss": 2.6952, + "step": 11607 + }, + { + "epoch": 2.53, + "learning_rate": 3.7373825649989075e-06, + "loss": 1.5393, + "step": 11608 + }, + { + "epoch": 2.53, + "learning_rate": 3.7372733231374265e-06, + "loss": 1.9632, + "step": 11609 + }, + { + "epoch": 2.53, + "learning_rate": 3.737164081275945e-06, + "loss": 2.1682, + "step": 11610 + }, + { + "epoch": 2.53, + "learning_rate": 3.737054839414464e-06, + "loss": 3.451, + "step": 11611 + }, + { + "epoch": 2.53, + "learning_rate": 3.736945597552983e-06, + "loss": 3.0591, + "step": 11612 + }, + { + "epoch": 2.53, + "learning_rate": 3.736836355691501e-06, + "loss": 2.4481, + "step": 11613 + }, + { + "epoch": 2.53, + "learning_rate": 3.73672711383002e-06, + "loss": 1.51, + "step": 11614 + }, + { + "epoch": 2.53, + "learning_rate": 3.7366178719685386e-06, + "loss": 2.9384, + "step": 11615 + }, + { + "epoch": 2.54, + "learning_rate": 3.7365086301070572e-06, + "loss": 1.9526, + "step": 11616 + }, + { + "epoch": 2.54, + "learning_rate": 3.7363993882455763e-06, + "loss": 2.4729, + "step": 11617 + }, + { + "epoch": 2.54, + "learning_rate": 3.7362901463840945e-06, + "loss": 2.1982, + "step": 11618 + }, + { + "epoch": 2.54, + "learning_rate": 3.7361809045226135e-06, + "loss": 3.1268, + "step": 11619 + }, + { + "epoch": 2.54, + "learning_rate": 3.736071662661132e-06, + "loss": 2.6864, + "step": 11620 + }, + { + "epoch": 2.54, + "learning_rate": 3.7359624207996507e-06, + "loss": 2.2534, + "step": 11621 + }, + { + "epoch": 2.54, + "learning_rate": 3.7358531789381697e-06, + "loss": 2.3484, + "step": 11622 + }, + { + "epoch": 2.54, + "learning_rate": 3.735743937076688e-06, + "loss": 2.5868, + "step": 11623 + }, + { + "epoch": 2.54, + "learning_rate": 3.7356346952152065e-06, + "loss": 1.5957, + "step": 11624 + }, + { + "epoch": 2.54, + "learning_rate": 3.7355254533537256e-06, + "loss": 2.1467, + "step": 11625 + }, + { + "epoch": 2.54, + "learning_rate": 3.735416211492244e-06, + "loss": 3.1341, + "step": 11626 + }, + { + "epoch": 2.54, + "learning_rate": 3.7353069696307624e-06, + "loss": 2.3822, + "step": 11627 + }, + { + "epoch": 2.54, + "learning_rate": 3.7351977277692814e-06, + "loss": 3.4498, + "step": 11628 + }, + { + "epoch": 2.54, + "learning_rate": 3.7350884859078e-06, + "loss": 2.2772, + "step": 11629 + }, + { + "epoch": 2.54, + "learning_rate": 3.734979244046319e-06, + "loss": 2.2377, + "step": 11630 + }, + { + "epoch": 2.54, + "learning_rate": 3.7348700021848377e-06, + "loss": 0.9567, + "step": 11631 + }, + { + "epoch": 2.54, + "learning_rate": 3.734760760323356e-06, + "loss": 1.5761, + "step": 11632 + }, + { + "epoch": 2.54, + "learning_rate": 3.734651518461875e-06, + "loss": 2.827, + "step": 11633 + }, + { + "epoch": 2.54, + "learning_rate": 3.7345422766003935e-06, + "loss": 1.3471, + "step": 11634 + }, + { + "epoch": 2.54, + "learning_rate": 3.7344330347389125e-06, + "loss": 3.6689, + "step": 11635 + }, + { + "epoch": 2.54, + "learning_rate": 3.734323792877431e-06, + "loss": 1.4769, + "step": 11636 + }, + { + "epoch": 2.54, + "learning_rate": 3.7342145510159493e-06, + "loss": 1.1165, + "step": 11637 + }, + { + "epoch": 2.54, + "learning_rate": 3.7341053091544684e-06, + "loss": 2.9695, + "step": 11638 + }, + { + "epoch": 2.54, + "learning_rate": 3.733996067292987e-06, + "loss": 2.955, + "step": 11639 + }, + { + "epoch": 2.54, + "learning_rate": 3.733886825431506e-06, + "loss": 1.9004, + "step": 11640 + }, + { + "epoch": 2.54, + "learning_rate": 3.733777583570024e-06, + "loss": 1.9133, + "step": 11641 + }, + { + "epoch": 2.54, + "learning_rate": 3.733668341708543e-06, + "loss": 1.3596, + "step": 11642 + }, + { + "epoch": 2.54, + "learning_rate": 3.733559099847062e-06, + "loss": 1.6652, + "step": 11643 + }, + { + "epoch": 2.54, + "learning_rate": 3.7334498579855805e-06, + "loss": 2.3728, + "step": 11644 + }, + { + "epoch": 2.54, + "learning_rate": 3.733340616124099e-06, + "loss": 1.7193, + "step": 11645 + }, + { + "epoch": 2.54, + "learning_rate": 3.7332313742626177e-06, + "loss": 2.2189, + "step": 11646 + }, + { + "epoch": 2.54, + "learning_rate": 3.7331221324011363e-06, + "loss": 2.3687, + "step": 11647 + }, + { + "epoch": 2.54, + "learning_rate": 3.733012890539655e-06, + "loss": 2.4997, + "step": 11648 + }, + { + "epoch": 2.54, + "learning_rate": 3.732903648678174e-06, + "loss": 1.8923, + "step": 11649 + }, + { + "epoch": 2.54, + "learning_rate": 3.7327944068166925e-06, + "loss": 2.3405, + "step": 11650 + }, + { + "epoch": 2.54, + "learning_rate": 3.7326851649552107e-06, + "loss": 2.9744, + "step": 11651 + }, + { + "epoch": 2.54, + "learning_rate": 3.7325759230937298e-06, + "loss": 2.6493, + "step": 11652 + }, + { + "epoch": 2.54, + "learning_rate": 3.7324666812322484e-06, + "loss": 1.7927, + "step": 11653 + }, + { + "epoch": 2.54, + "learning_rate": 3.7323574393707674e-06, + "loss": 1.5165, + "step": 11654 + }, + { + "epoch": 2.54, + "learning_rate": 3.732248197509286e-06, + "loss": 1.9756, + "step": 11655 + }, + { + "epoch": 2.54, + "learning_rate": 3.7321389556478042e-06, + "loss": 2.6061, + "step": 11656 + }, + { + "epoch": 2.54, + "learning_rate": 3.7320297137863232e-06, + "loss": 3.7026, + "step": 11657 + }, + { + "epoch": 2.54, + "learning_rate": 3.731920471924842e-06, + "loss": 3.7657, + "step": 11658 + }, + { + "epoch": 2.54, + "learning_rate": 3.731811230063361e-06, + "loss": 2.9922, + "step": 11659 + }, + { + "epoch": 2.54, + "learning_rate": 3.731701988201879e-06, + "loss": 3.3856, + "step": 11660 + }, + { + "epoch": 2.54, + "learning_rate": 3.7315927463403977e-06, + "loss": 1.9227, + "step": 11661 + }, + { + "epoch": 2.55, + "learning_rate": 3.7314835044789167e-06, + "loss": 2.2438, + "step": 11662 + }, + { + "epoch": 2.55, + "learning_rate": 3.7313742626174353e-06, + "loss": 2.9312, + "step": 11663 + }, + { + "epoch": 2.55, + "learning_rate": 3.7312650207559544e-06, + "loss": 2.6539, + "step": 11664 + }, + { + "epoch": 2.55, + "learning_rate": 3.7311557788944726e-06, + "loss": 1.9479, + "step": 11665 + }, + { + "epoch": 2.55, + "learning_rate": 3.731046537032991e-06, + "loss": 3.0566, + "step": 11666 + }, + { + "epoch": 2.55, + "learning_rate": 3.73093729517151e-06, + "loss": 2.2088, + "step": 11667 + }, + { + "epoch": 2.55, + "learning_rate": 3.730828053310029e-06, + "loss": 0.8481, + "step": 11668 + }, + { + "epoch": 2.55, + "learning_rate": 3.7307188114485474e-06, + "loss": 2.9357, + "step": 11669 + }, + { + "epoch": 2.55, + "learning_rate": 3.730609569587066e-06, + "loss": 2.5592, + "step": 11670 + }, + { + "epoch": 2.55, + "learning_rate": 3.7305003277255847e-06, + "loss": 1.0936, + "step": 11671 + }, + { + "epoch": 2.55, + "learning_rate": 3.7303910858641037e-06, + "loss": 2.6182, + "step": 11672 + }, + { + "epoch": 2.55, + "learning_rate": 3.7302818440026223e-06, + "loss": 2.1128, + "step": 11673 + }, + { + "epoch": 2.55, + "learning_rate": 3.7301726021411405e-06, + "loss": 3.1774, + "step": 11674 + }, + { + "epoch": 2.55, + "learning_rate": 3.7300633602796595e-06, + "loss": 2.4594, + "step": 11675 + }, + { + "epoch": 2.55, + "learning_rate": 3.729954118418178e-06, + "loss": 1.982, + "step": 11676 + }, + { + "epoch": 2.55, + "learning_rate": 3.7298448765566967e-06, + "loss": 3.4404, + "step": 11677 + }, + { + "epoch": 2.55, + "learning_rate": 3.7297356346952158e-06, + "loss": 3.0026, + "step": 11678 + }, + { + "epoch": 2.55, + "learning_rate": 3.729626392833734e-06, + "loss": 3.6557, + "step": 11679 + }, + { + "epoch": 2.55, + "learning_rate": 3.7295171509722526e-06, + "loss": 0.8983, + "step": 11680 + }, + { + "epoch": 2.55, + "learning_rate": 3.7294079091107716e-06, + "loss": 1.4035, + "step": 11681 + }, + { + "epoch": 2.55, + "learning_rate": 3.7292986672492902e-06, + "loss": 3.1068, + "step": 11682 + }, + { + "epoch": 2.55, + "learning_rate": 3.7291894253878093e-06, + "loss": 1.9914, + "step": 11683 + }, + { + "epoch": 2.55, + "learning_rate": 3.7290801835263274e-06, + "loss": 1.8587, + "step": 11684 + }, + { + "epoch": 2.55, + "learning_rate": 3.728970941664846e-06, + "loss": 1.5187, + "step": 11685 + }, + { + "epoch": 2.55, + "learning_rate": 3.728861699803365e-06, + "loss": 2.874, + "step": 11686 + }, + { + "epoch": 2.55, + "learning_rate": 3.7287524579418837e-06, + "loss": 2.2266, + "step": 11687 + }, + { + "epoch": 2.55, + "learning_rate": 3.7286432160804027e-06, + "loss": 1.7851, + "step": 11688 + }, + { + "epoch": 2.55, + "learning_rate": 3.728533974218921e-06, + "loss": 2.451, + "step": 11689 + }, + { + "epoch": 2.55, + "learning_rate": 3.7284247323574395e-06, + "loss": 2.6769, + "step": 11690 + }, + { + "epoch": 2.55, + "learning_rate": 3.7283154904959586e-06, + "loss": 2.42, + "step": 11691 + }, + { + "epoch": 2.55, + "learning_rate": 3.728206248634477e-06, + "loss": 1.7996, + "step": 11692 + }, + { + "epoch": 2.55, + "learning_rate": 3.7280970067729954e-06, + "loss": 2.487, + "step": 11693 + }, + { + "epoch": 2.55, + "learning_rate": 3.7279877649115144e-06, + "loss": 2.3232, + "step": 11694 + }, + { + "epoch": 2.55, + "learning_rate": 3.727878523050033e-06, + "loss": 2.99, + "step": 11695 + }, + { + "epoch": 2.55, + "learning_rate": 3.727769281188552e-06, + "loss": 1.3714, + "step": 11696 + }, + { + "epoch": 2.55, + "learning_rate": 3.7276600393270707e-06, + "loss": 2.7299, + "step": 11697 + }, + { + "epoch": 2.55, + "learning_rate": 3.727550797465589e-06, + "loss": 2.1539, + "step": 11698 + }, + { + "epoch": 2.55, + "learning_rate": 3.727441555604108e-06, + "loss": 2.687, + "step": 11699 + }, + { + "epoch": 2.55, + "learning_rate": 3.7273323137426265e-06, + "loss": 1.8669, + "step": 11700 + }, + { + "epoch": 2.55, + "learning_rate": 3.727223071881145e-06, + "loss": 1.6077, + "step": 11701 + }, + { + "epoch": 2.55, + "learning_rate": 3.727113830019664e-06, + "loss": 1.9968, + "step": 11702 + }, + { + "epoch": 2.55, + "learning_rate": 3.7270045881581823e-06, + "loss": 1.255, + "step": 11703 + }, + { + "epoch": 2.55, + "learning_rate": 3.726895346296701e-06, + "loss": 2.6494, + "step": 11704 + }, + { + "epoch": 2.55, + "learning_rate": 3.72678610443522e-06, + "loss": 2.9537, + "step": 11705 + }, + { + "epoch": 2.55, + "learning_rate": 3.7266768625737386e-06, + "loss": 1.944, + "step": 11706 + }, + { + "epoch": 2.55, + "learning_rate": 3.7265676207122568e-06, + "loss": 3.059, + "step": 11707 + }, + { + "epoch": 2.56, + "learning_rate": 3.726458378850776e-06, + "loss": 3.0527, + "step": 11708 + }, + { + "epoch": 2.56, + "learning_rate": 3.7263491369892944e-06, + "loss": 1.4645, + "step": 11709 + }, + { + "epoch": 2.56, + "learning_rate": 3.7262398951278134e-06, + "loss": 2.6355, + "step": 11710 + }, + { + "epoch": 2.56, + "learning_rate": 3.726130653266332e-06, + "loss": 1.1073, + "step": 11711 + }, + { + "epoch": 2.56, + "learning_rate": 3.7260214114048502e-06, + "loss": 3.4024, + "step": 11712 + }, + { + "epoch": 2.56, + "learning_rate": 3.7259121695433693e-06, + "loss": 1.6263, + "step": 11713 + }, + { + "epoch": 2.56, + "learning_rate": 3.725802927681888e-06, + "loss": 3.3594, + "step": 11714 + }, + { + "epoch": 2.56, + "learning_rate": 3.725693685820407e-06, + "loss": 2.8552, + "step": 11715 + }, + { + "epoch": 2.56, + "learning_rate": 3.7255844439589255e-06, + "loss": 2.286, + "step": 11716 + }, + { + "epoch": 2.56, + "learning_rate": 3.7254752020974437e-06, + "loss": 0.9397, + "step": 11717 + }, + { + "epoch": 2.56, + "learning_rate": 3.7253659602359628e-06, + "loss": 2.5369, + "step": 11718 + }, + { + "epoch": 2.56, + "learning_rate": 3.7252567183744814e-06, + "loss": 1.6575, + "step": 11719 + }, + { + "epoch": 2.56, + "learning_rate": 3.7251474765130004e-06, + "loss": 3.0705, + "step": 11720 + }, + { + "epoch": 2.56, + "learning_rate": 3.725038234651519e-06, + "loss": 2.4855, + "step": 11721 + }, + { + "epoch": 2.56, + "learning_rate": 3.724928992790037e-06, + "loss": 2.2004, + "step": 11722 + }, + { + "epoch": 2.56, + "learning_rate": 3.7248197509285562e-06, + "loss": 3.0572, + "step": 11723 + }, + { + "epoch": 2.56, + "learning_rate": 3.724710509067075e-06, + "loss": 2.2837, + "step": 11724 + }, + { + "epoch": 2.56, + "learning_rate": 3.724601267205594e-06, + "loss": 2.0143, + "step": 11725 + }, + { + "epoch": 2.56, + "learning_rate": 3.724492025344112e-06, + "loss": 2.5879, + "step": 11726 + }, + { + "epoch": 2.56, + "learning_rate": 3.7243827834826307e-06, + "loss": 1.5814, + "step": 11727 + }, + { + "epoch": 2.56, + "learning_rate": 3.7242735416211497e-06, + "loss": 1.3872, + "step": 11728 + }, + { + "epoch": 2.56, + "learning_rate": 3.7241642997596683e-06, + "loss": 2.0111, + "step": 11729 + }, + { + "epoch": 2.56, + "learning_rate": 3.724055057898187e-06, + "loss": 1.8836, + "step": 11730 + }, + { + "epoch": 2.56, + "learning_rate": 3.7239458160367055e-06, + "loss": 2.1741, + "step": 11731 + }, + { + "epoch": 2.56, + "learning_rate": 3.723836574175224e-06, + "loss": 3.8811, + "step": 11732 + }, + { + "epoch": 2.56, + "learning_rate": 3.7237273323137428e-06, + "loss": 3.7041, + "step": 11733 + }, + { + "epoch": 2.56, + "learning_rate": 3.723618090452262e-06, + "loss": 1.8014, + "step": 11734 + }, + { + "epoch": 2.56, + "learning_rate": 3.7235088485907804e-06, + "loss": 1.0267, + "step": 11735 + }, + { + "epoch": 2.56, + "learning_rate": 3.7233996067292986e-06, + "loss": 4.2833, + "step": 11736 + }, + { + "epoch": 2.56, + "learning_rate": 3.7232903648678176e-06, + "loss": 3.5093, + "step": 11737 + }, + { + "epoch": 2.56, + "learning_rate": 3.7231811230063362e-06, + "loss": 0.573, + "step": 11738 + }, + { + "epoch": 2.56, + "learning_rate": 3.7230718811448553e-06, + "loss": 1.281, + "step": 11739 + }, + { + "epoch": 2.56, + "learning_rate": 3.7229626392833735e-06, + "loss": 1.3342, + "step": 11740 + }, + { + "epoch": 2.56, + "learning_rate": 3.722853397421892e-06, + "loss": 2.7074, + "step": 11741 + }, + { + "epoch": 2.56, + "learning_rate": 3.722744155560411e-06, + "loss": 2.1142, + "step": 11742 + }, + { + "epoch": 2.56, + "learning_rate": 3.7226349136989297e-06, + "loss": 2.7452, + "step": 11743 + }, + { + "epoch": 2.56, + "learning_rate": 3.7225256718374488e-06, + "loss": 2.4801, + "step": 11744 + }, + { + "epoch": 2.56, + "learning_rate": 3.722416429975967e-06, + "loss": 1.8764, + "step": 11745 + }, + { + "epoch": 2.56, + "learning_rate": 3.7223071881144856e-06, + "loss": 1.6365, + "step": 11746 + }, + { + "epoch": 2.56, + "learning_rate": 3.7221979462530046e-06, + "loss": 1.5623, + "step": 11747 + }, + { + "epoch": 2.56, + "learning_rate": 3.722088704391523e-06, + "loss": 2.7821, + "step": 11748 + }, + { + "epoch": 2.56, + "learning_rate": 3.7219794625300422e-06, + "loss": 2.431, + "step": 11749 + }, + { + "epoch": 2.56, + "learning_rate": 3.7218702206685604e-06, + "loss": 1.498, + "step": 11750 + }, + { + "epoch": 2.56, + "learning_rate": 3.721760978807079e-06, + "loss": 2.3951, + "step": 11751 + }, + { + "epoch": 2.56, + "learning_rate": 3.721651736945598e-06, + "loss": 0.4686, + "step": 11752 + }, + { + "epoch": 2.57, + "learning_rate": 3.7215424950841167e-06, + "loss": 3.0184, + "step": 11753 + }, + { + "epoch": 2.57, + "learning_rate": 3.7214332532226353e-06, + "loss": 2.6064, + "step": 11754 + }, + { + "epoch": 2.57, + "learning_rate": 3.721324011361154e-06, + "loss": 2.6305, + "step": 11755 + }, + { + "epoch": 2.57, + "learning_rate": 3.7212147694996725e-06, + "loss": 1.5455, + "step": 11756 + }, + { + "epoch": 2.57, + "learning_rate": 3.721105527638191e-06, + "loss": 2.1545, + "step": 11757 + }, + { + "epoch": 2.57, + "learning_rate": 3.72099628577671e-06, + "loss": 2.7387, + "step": 11758 + }, + { + "epoch": 2.57, + "learning_rate": 3.7208870439152283e-06, + "loss": 1.9202, + "step": 11759 + }, + { + "epoch": 2.57, + "learning_rate": 3.720777802053747e-06, + "loss": 2.7242, + "step": 11760 + }, + { + "epoch": 2.57, + "learning_rate": 3.720668560192266e-06, + "loss": 3.4966, + "step": 11761 + }, + { + "epoch": 2.57, + "learning_rate": 3.7205593183307846e-06, + "loss": 1.2635, + "step": 11762 + }, + { + "epoch": 2.57, + "learning_rate": 3.7204500764693036e-06, + "loss": 2.6662, + "step": 11763 + }, + { + "epoch": 2.57, + "learning_rate": 3.720340834607822e-06, + "loss": 3.4051, + "step": 11764 + }, + { + "epoch": 2.57, + "learning_rate": 3.7202315927463404e-06, + "loss": 2.3464, + "step": 11765 + }, + { + "epoch": 2.57, + "learning_rate": 3.7201223508848595e-06, + "loss": 1.7896, + "step": 11766 + }, + { + "epoch": 2.57, + "learning_rate": 3.720013109023378e-06, + "loss": 2.8383, + "step": 11767 + }, + { + "epoch": 2.57, + "learning_rate": 3.719903867161897e-06, + "loss": 1.6081, + "step": 11768 + }, + { + "epoch": 2.57, + "learning_rate": 3.7197946253004153e-06, + "loss": 3.0442, + "step": 11769 + }, + { + "epoch": 2.57, + "learning_rate": 3.719685383438934e-06, + "loss": 1.9598, + "step": 11770 + }, + { + "epoch": 2.57, + "learning_rate": 3.719576141577453e-06, + "loss": 3.2687, + "step": 11771 + }, + { + "epoch": 2.57, + "learning_rate": 3.7194668997159716e-06, + "loss": 2.0868, + "step": 11772 + }, + { + "epoch": 2.57, + "learning_rate": 3.7193576578544897e-06, + "loss": 2.3065, + "step": 11773 + }, + { + "epoch": 2.57, + "learning_rate": 3.7192484159930088e-06, + "loss": 1.6316, + "step": 11774 + }, + { + "epoch": 2.57, + "learning_rate": 3.7191391741315274e-06, + "loss": 1.8874, + "step": 11775 + }, + { + "epoch": 2.57, + "learning_rate": 3.7190299322700464e-06, + "loss": 1.5941, + "step": 11776 + }, + { + "epoch": 2.57, + "learning_rate": 3.718920690408565e-06, + "loss": 1.9986, + "step": 11777 + }, + { + "epoch": 2.57, + "learning_rate": 3.7188114485470832e-06, + "loss": 1.6687, + "step": 11778 + }, + { + "epoch": 2.57, + "learning_rate": 3.7187022066856023e-06, + "loss": 2.5153, + "step": 11779 + }, + { + "epoch": 2.57, + "learning_rate": 3.718592964824121e-06, + "loss": 2.231, + "step": 11780 + }, + { + "epoch": 2.57, + "learning_rate": 3.71848372296264e-06, + "loss": 1.8841, + "step": 11781 + }, + { + "epoch": 2.57, + "learning_rate": 3.7183744811011585e-06, + "loss": 2.6355, + "step": 11782 + }, + { + "epoch": 2.57, + "learning_rate": 3.7182652392396767e-06, + "loss": 2.1759, + "step": 11783 + }, + { + "epoch": 2.57, + "learning_rate": 3.7181559973781957e-06, + "loss": 3.1912, + "step": 11784 + }, + { + "epoch": 2.57, + "learning_rate": 3.7180467555167143e-06, + "loss": 2.9361, + "step": 11785 + }, + { + "epoch": 2.57, + "learning_rate": 3.717937513655233e-06, + "loss": 0.8395, + "step": 11786 + }, + { + "epoch": 2.57, + "learning_rate": 3.717828271793752e-06, + "loss": 3.1742, + "step": 11787 + }, + { + "epoch": 2.57, + "learning_rate": 3.71771902993227e-06, + "loss": 2.6079, + "step": 11788 + }, + { + "epoch": 2.57, + "learning_rate": 3.717609788070789e-06, + "loss": 3.0584, + "step": 11789 + }, + { + "epoch": 2.57, + "learning_rate": 3.717500546209308e-06, + "loss": 3.5941, + "step": 11790 + }, + { + "epoch": 2.57, + "learning_rate": 3.7173913043478264e-06, + "loss": 3.0192, + "step": 11791 + }, + { + "epoch": 2.57, + "learning_rate": 3.7172820624863446e-06, + "loss": 2.4334, + "step": 11792 + }, + { + "epoch": 2.57, + "learning_rate": 3.7171728206248637e-06, + "loss": 3.0974, + "step": 11793 + }, + { + "epoch": 2.57, + "learning_rate": 3.7170635787633823e-06, + "loss": 2.5323, + "step": 11794 + }, + { + "epoch": 2.57, + "learning_rate": 3.7169543369019013e-06, + "loss": 2.3521, + "step": 11795 + }, + { + "epoch": 2.57, + "learning_rate": 3.71684509504042e-06, + "loss": 1.244, + "step": 11796 + }, + { + "epoch": 2.57, + "learning_rate": 3.716735853178938e-06, + "loss": 2.4029, + "step": 11797 + }, + { + "epoch": 2.57, + "learning_rate": 3.716626611317457e-06, + "loss": 3.1248, + "step": 11798 + }, + { + "epoch": 2.58, + "learning_rate": 3.7165173694559758e-06, + "loss": 2.3681, + "step": 11799 + }, + { + "epoch": 2.58, + "learning_rate": 3.7164081275944948e-06, + "loss": 1.9819, + "step": 11800 + }, + { + "epoch": 2.58, + "learning_rate": 3.7162988857330134e-06, + "loss": 2.4931, + "step": 11801 + }, + { + "epoch": 2.58, + "learning_rate": 3.7161896438715316e-06, + "loss": 3.4443, + "step": 11802 + }, + { + "epoch": 2.58, + "learning_rate": 3.7160804020100506e-06, + "loss": 2.7173, + "step": 11803 + }, + { + "epoch": 2.58, + "learning_rate": 3.7159711601485692e-06, + "loss": 2.5612, + "step": 11804 + }, + { + "epoch": 2.58, + "learning_rate": 3.7158619182870883e-06, + "loss": 2.5102, + "step": 11805 + }, + { + "epoch": 2.58, + "learning_rate": 3.7157526764256065e-06, + "loss": 2.5194, + "step": 11806 + }, + { + "epoch": 2.58, + "learning_rate": 3.715643434564125e-06, + "loss": 2.5981, + "step": 11807 + }, + { + "epoch": 2.58, + "learning_rate": 3.715534192702644e-06, + "loss": 3.9972, + "step": 11808 + }, + { + "epoch": 2.58, + "learning_rate": 3.7154249508411627e-06, + "loss": 2.2664, + "step": 11809 + }, + { + "epoch": 2.58, + "learning_rate": 3.7153157089796813e-06, + "loss": 1.3046, + "step": 11810 + }, + { + "epoch": 2.58, + "learning_rate": 3.7152064671182e-06, + "loss": 1.6766, + "step": 11811 + }, + { + "epoch": 2.58, + "learning_rate": 3.7150972252567185e-06, + "loss": 1.1326, + "step": 11812 + }, + { + "epoch": 2.58, + "learning_rate": 3.714987983395237e-06, + "loss": 1.6142, + "step": 11813 + }, + { + "epoch": 2.58, + "learning_rate": 3.714878741533756e-06, + "loss": 0.6465, + "step": 11814 + }, + { + "epoch": 2.58, + "learning_rate": 3.714769499672275e-06, + "loss": 4.0432, + "step": 11815 + }, + { + "epoch": 2.58, + "learning_rate": 3.7146602578107934e-06, + "loss": 2.5475, + "step": 11816 + }, + { + "epoch": 2.58, + "learning_rate": 3.714551015949312e-06, + "loss": 1.9337, + "step": 11817 + }, + { + "epoch": 2.58, + "learning_rate": 3.7144417740878306e-06, + "loss": 2.9198, + "step": 11818 + }, + { + "epoch": 2.58, + "learning_rate": 3.7143325322263497e-06, + "loss": 2.0528, + "step": 11819 + }, + { + "epoch": 2.58, + "learning_rate": 3.7142232903648683e-06, + "loss": 2.3864, + "step": 11820 + }, + { + "epoch": 2.58, + "learning_rate": 3.7141140485033865e-06, + "loss": 3.2666, + "step": 11821 + }, + { + "epoch": 2.58, + "learning_rate": 3.7140048066419055e-06, + "loss": 2.4981, + "step": 11822 + }, + { + "epoch": 2.58, + "learning_rate": 3.713895564780424e-06, + "loss": 1.7645, + "step": 11823 + }, + { + "epoch": 2.58, + "learning_rate": 3.713786322918943e-06, + "loss": 2.2285, + "step": 11824 + }, + { + "epoch": 2.58, + "learning_rate": 3.7136770810574613e-06, + "loss": 1.9898, + "step": 11825 + }, + { + "epoch": 2.58, + "learning_rate": 3.71356783919598e-06, + "loss": 1.6016, + "step": 11826 + }, + { + "epoch": 2.58, + "learning_rate": 3.713458597334499e-06, + "loss": 2.9022, + "step": 11827 + }, + { + "epoch": 2.58, + "learning_rate": 3.7133493554730176e-06, + "loss": 2.5982, + "step": 11828 + }, + { + "epoch": 2.58, + "learning_rate": 3.7132401136115366e-06, + "loss": 2.4625, + "step": 11829 + }, + { + "epoch": 2.58, + "learning_rate": 3.713130871750055e-06, + "loss": 1.594, + "step": 11830 + }, + { + "epoch": 2.58, + "learning_rate": 3.7130216298885734e-06, + "loss": 3.6694, + "step": 11831 + }, + { + "epoch": 2.58, + "learning_rate": 3.7129123880270925e-06, + "loss": 2.6302, + "step": 11832 + }, + { + "epoch": 2.58, + "learning_rate": 3.712803146165611e-06, + "loss": 2.2275, + "step": 11833 + }, + { + "epoch": 2.58, + "learning_rate": 3.71269390430413e-06, + "loss": 2.3292, + "step": 11834 + }, + { + "epoch": 2.58, + "learning_rate": 3.7125846624426483e-06, + "loss": 1.954, + "step": 11835 + }, + { + "epoch": 2.58, + "learning_rate": 3.712475420581167e-06, + "loss": 0.8757, + "step": 11836 + }, + { + "epoch": 2.58, + "learning_rate": 3.712366178719686e-06, + "loss": 1.2224, + "step": 11837 + }, + { + "epoch": 2.58, + "learning_rate": 3.7122569368582045e-06, + "loss": 2.3956, + "step": 11838 + }, + { + "epoch": 2.58, + "learning_rate": 3.7121476949967227e-06, + "loss": 1.7362, + "step": 11839 + }, + { + "epoch": 2.58, + "learning_rate": 3.7120384531352418e-06, + "loss": 1.9004, + "step": 11840 + }, + { + "epoch": 2.58, + "learning_rate": 3.7119292112737604e-06, + "loss": 1.4947, + "step": 11841 + }, + { + "epoch": 2.58, + "learning_rate": 3.711819969412279e-06, + "loss": 2.2241, + "step": 11842 + }, + { + "epoch": 2.58, + "learning_rate": 3.711710727550798e-06, + "loss": 0.9521, + "step": 11843 + }, + { + "epoch": 2.58, + "learning_rate": 3.711601485689316e-06, + "loss": 2.0498, + "step": 11844 + }, + { + "epoch": 2.59, + "learning_rate": 3.711492243827835e-06, + "loss": 2.7519, + "step": 11845 + }, + { + "epoch": 2.59, + "learning_rate": 3.711383001966354e-06, + "loss": 1.4908, + "step": 11846 + }, + { + "epoch": 2.59, + "learning_rate": 3.7112737601048725e-06, + "loss": 1.7087, + "step": 11847 + }, + { + "epoch": 2.59, + "learning_rate": 3.7111645182433915e-06, + "loss": 3.1859, + "step": 11848 + }, + { + "epoch": 2.59, + "learning_rate": 3.7110552763819097e-06, + "loss": 1.3583, + "step": 11849 + }, + { + "epoch": 2.59, + "learning_rate": 3.7109460345204283e-06, + "loss": 1.8396, + "step": 11850 + }, + { + "epoch": 2.59, + "learning_rate": 3.7108367926589473e-06, + "loss": 1.6939, + "step": 11851 + }, + { + "epoch": 2.59, + "learning_rate": 3.710727550797466e-06, + "loss": 1.4935, + "step": 11852 + }, + { + "epoch": 2.59, + "learning_rate": 3.710618308935985e-06, + "loss": 2.5878, + "step": 11853 + }, + { + "epoch": 2.59, + "learning_rate": 3.710509067074503e-06, + "loss": 2.0973, + "step": 11854 + }, + { + "epoch": 2.59, + "learning_rate": 3.7103998252130218e-06, + "loss": 2.8019, + "step": 11855 + }, + { + "epoch": 2.59, + "learning_rate": 3.710290583351541e-06, + "loss": 1.7603, + "step": 11856 + }, + { + "epoch": 2.59, + "learning_rate": 3.7101813414900594e-06, + "loss": 1.5222, + "step": 11857 + }, + { + "epoch": 2.59, + "learning_rate": 3.7100720996285776e-06, + "loss": 2.0364, + "step": 11858 + }, + { + "epoch": 2.59, + "learning_rate": 3.7099628577670966e-06, + "loss": 1.7055, + "step": 11859 + }, + { + "epoch": 2.59, + "learning_rate": 3.7098536159056153e-06, + "loss": 2.4923, + "step": 11860 + }, + { + "epoch": 2.59, + "learning_rate": 3.7097443740441343e-06, + "loss": 1.8122, + "step": 11861 + }, + { + "epoch": 2.59, + "learning_rate": 3.709635132182653e-06, + "loss": 2.9615, + "step": 11862 + }, + { + "epoch": 2.59, + "learning_rate": 3.709525890321171e-06, + "loss": 2.0706, + "step": 11863 + }, + { + "epoch": 2.59, + "learning_rate": 3.70941664845969e-06, + "loss": 2.3941, + "step": 11864 + }, + { + "epoch": 2.59, + "learning_rate": 3.7093074065982087e-06, + "loss": 1.7058, + "step": 11865 + }, + { + "epoch": 2.59, + "learning_rate": 3.7091981647367278e-06, + "loss": 3.8795, + "step": 11866 + }, + { + "epoch": 2.59, + "learning_rate": 3.7090889228752464e-06, + "loss": 2.0257, + "step": 11867 + }, + { + "epoch": 2.59, + "learning_rate": 3.7089796810137646e-06, + "loss": 2.6216, + "step": 11868 + }, + { + "epoch": 2.59, + "learning_rate": 3.7088704391522836e-06, + "loss": 2.1985, + "step": 11869 + }, + { + "epoch": 2.59, + "learning_rate": 3.7087611972908022e-06, + "loss": 1.9659, + "step": 11870 + }, + { + "epoch": 2.59, + "learning_rate": 3.708651955429321e-06, + "loss": 1.7681, + "step": 11871 + }, + { + "epoch": 2.59, + "learning_rate": 3.7085427135678394e-06, + "loss": 2.6917, + "step": 11872 + }, + { + "epoch": 2.59, + "learning_rate": 3.708433471706358e-06, + "loss": 2.1705, + "step": 11873 + }, + { + "epoch": 2.59, + "learning_rate": 3.7083242298448767e-06, + "loss": 1.3875, + "step": 11874 + }, + { + "epoch": 2.59, + "learning_rate": 3.7082149879833957e-06, + "loss": 2.6466, + "step": 11875 + }, + { + "epoch": 2.59, + "learning_rate": 3.7081057461219143e-06, + "loss": 2.292, + "step": 11876 + }, + { + "epoch": 2.59, + "learning_rate": 3.7079965042604325e-06, + "loss": 1.706, + "step": 11877 + }, + { + "epoch": 2.59, + "learning_rate": 3.7078872623989515e-06, + "loss": 3.2113, + "step": 11878 + }, + { + "epoch": 2.59, + "learning_rate": 3.70777802053747e-06, + "loss": 2.0937, + "step": 11879 + }, + { + "epoch": 2.59, + "learning_rate": 3.707668778675989e-06, + "loss": 2.7979, + "step": 11880 + }, + { + "epoch": 2.59, + "learning_rate": 3.7075595368145078e-06, + "loss": 1.7621, + "step": 11881 + }, + { + "epoch": 2.59, + "learning_rate": 3.707450294953026e-06, + "loss": 2.3009, + "step": 11882 + }, + { + "epoch": 2.59, + "learning_rate": 3.707341053091545e-06, + "loss": 2.1411, + "step": 11883 + }, + { + "epoch": 2.59, + "learning_rate": 3.7072318112300636e-06, + "loss": 1.5251, + "step": 11884 + }, + { + "epoch": 2.59, + "learning_rate": 3.7071225693685826e-06, + "loss": 2.7252, + "step": 11885 + }, + { + "epoch": 2.59, + "learning_rate": 3.7070133275071013e-06, + "loss": 2.5789, + "step": 11886 + }, + { + "epoch": 2.59, + "learning_rate": 3.7069040856456194e-06, + "loss": 2.1829, + "step": 11887 + }, + { + "epoch": 2.59, + "learning_rate": 3.7067948437841385e-06, + "loss": 1.7519, + "step": 11888 + }, + { + "epoch": 2.59, + "learning_rate": 3.706685601922657e-06, + "loss": 3.0523, + "step": 11889 + }, + { + "epoch": 2.59, + "learning_rate": 3.706576360061176e-06, + "loss": 3.2796, + "step": 11890 + }, + { + "epoch": 2.6, + "learning_rate": 3.7064671181996943e-06, + "loss": 2.9873, + "step": 11891 + }, + { + "epoch": 2.6, + "learning_rate": 3.706357876338213e-06, + "loss": 2.2389, + "step": 11892 + }, + { + "epoch": 2.6, + "learning_rate": 3.706248634476732e-06, + "loss": 1.8586, + "step": 11893 + }, + { + "epoch": 2.6, + "learning_rate": 3.7061393926152506e-06, + "loss": 3.773, + "step": 11894 + }, + { + "epoch": 2.6, + "learning_rate": 3.706030150753769e-06, + "loss": 2.011, + "step": 11895 + }, + { + "epoch": 2.6, + "learning_rate": 3.705920908892288e-06, + "loss": 2.727, + "step": 11896 + }, + { + "epoch": 2.6, + "learning_rate": 3.7058116670308064e-06, + "loss": 2.2735, + "step": 11897 + }, + { + "epoch": 2.6, + "learning_rate": 3.705702425169325e-06, + "loss": 1.3848, + "step": 11898 + }, + { + "epoch": 2.6, + "learning_rate": 3.705593183307844e-06, + "loss": 1.6687, + "step": 11899 + }, + { + "epoch": 2.6, + "learning_rate": 3.7054839414463627e-06, + "loss": 1.1161, + "step": 11900 + }, + { + "epoch": 2.6, + "learning_rate": 3.705374699584881e-06, + "loss": 2.8308, + "step": 11901 + }, + { + "epoch": 2.6, + "learning_rate": 3.7052654577234e-06, + "loss": 1.8082, + "step": 11902 + }, + { + "epoch": 2.6, + "learning_rate": 3.7051562158619185e-06, + "loss": 2.737, + "step": 11903 + }, + { + "epoch": 2.6, + "learning_rate": 3.7050469740004375e-06, + "loss": 1.8795, + "step": 11904 + }, + { + "epoch": 2.6, + "learning_rate": 3.7049377321389557e-06, + "loss": 1.4703, + "step": 11905 + }, + { + "epoch": 2.6, + "learning_rate": 3.7048284902774743e-06, + "loss": 3.0045, + "step": 11906 + }, + { + "epoch": 2.6, + "learning_rate": 3.7047192484159934e-06, + "loss": 2.1902, + "step": 11907 + }, + { + "epoch": 2.6, + "learning_rate": 3.704610006554512e-06, + "loss": 1.9521, + "step": 11908 + }, + { + "epoch": 2.6, + "learning_rate": 3.704500764693031e-06, + "loss": 2.1527, + "step": 11909 + }, + { + "epoch": 2.6, + "learning_rate": 3.704391522831549e-06, + "loss": 2.647, + "step": 11910 + }, + { + "epoch": 2.6, + "learning_rate": 3.704282280970068e-06, + "loss": 1.6784, + "step": 11911 + }, + { + "epoch": 2.6, + "learning_rate": 3.704173039108587e-06, + "loss": 0.9459, + "step": 11912 + }, + { + "epoch": 2.6, + "learning_rate": 3.7040637972471054e-06, + "loss": 3.0029, + "step": 11913 + }, + { + "epoch": 2.6, + "learning_rate": 3.7039545553856245e-06, + "loss": 2.7671, + "step": 11914 + }, + { + "epoch": 2.6, + "learning_rate": 3.7038453135241427e-06, + "loss": 2.0984, + "step": 11915 + }, + { + "epoch": 2.6, + "learning_rate": 3.7037360716626613e-06, + "loss": 1.7468, + "step": 11916 + }, + { + "epoch": 2.6, + "learning_rate": 3.7036268298011803e-06, + "loss": 2.8311, + "step": 11917 + }, + { + "epoch": 2.6, + "learning_rate": 3.703517587939699e-06, + "loss": 2.8867, + "step": 11918 + }, + { + "epoch": 2.6, + "learning_rate": 3.703408346078218e-06, + "loss": 3.0243, + "step": 11919 + }, + { + "epoch": 2.6, + "learning_rate": 3.703299104216736e-06, + "loss": 3.6433, + "step": 11920 + }, + { + "epoch": 2.6, + "learning_rate": 3.7031898623552548e-06, + "loss": 1.2862, + "step": 11921 + }, + { + "epoch": 2.6, + "learning_rate": 3.703080620493774e-06, + "loss": 2.9464, + "step": 11922 + }, + { + "epoch": 2.6, + "learning_rate": 3.7029713786322924e-06, + "loss": 2.4718, + "step": 11923 + }, + { + "epoch": 2.6, + "learning_rate": 3.7028621367708106e-06, + "loss": 1.5478, + "step": 11924 + }, + { + "epoch": 2.6, + "learning_rate": 3.7027528949093296e-06, + "loss": 1.5626, + "step": 11925 + }, + { + "epoch": 2.6, + "learning_rate": 3.7026436530478482e-06, + "loss": 1.7857, + "step": 11926 + }, + { + "epoch": 2.6, + "learning_rate": 3.702534411186367e-06, + "loss": 1.9665, + "step": 11927 + }, + { + "epoch": 2.6, + "learning_rate": 3.702425169324886e-06, + "loss": 3.0405, + "step": 11928 + }, + { + "epoch": 2.6, + "learning_rate": 3.702315927463404e-06, + "loss": 1.6511, + "step": 11929 + }, + { + "epoch": 2.6, + "learning_rate": 3.7022066856019227e-06, + "loss": 1.5272, + "step": 11930 + }, + { + "epoch": 2.6, + "learning_rate": 3.7020974437404417e-06, + "loss": 2.3939, + "step": 11931 + }, + { + "epoch": 2.6, + "learning_rate": 3.7019882018789603e-06, + "loss": 1.7622, + "step": 11932 + }, + { + "epoch": 2.6, + "learning_rate": 3.7018789600174794e-06, + "loss": 2.9418, + "step": 11933 + }, + { + "epoch": 2.6, + "learning_rate": 3.7017697181559976e-06, + "loss": 1.074, + "step": 11934 + }, + { + "epoch": 2.6, + "learning_rate": 3.701660476294516e-06, + "loss": 2.4603, + "step": 11935 + }, + { + "epoch": 2.6, + "learning_rate": 3.701551234433035e-06, + "loss": 2.8526, + "step": 11936 + }, + { + "epoch": 2.61, + "learning_rate": 3.701441992571554e-06, + "loss": 2.661, + "step": 11937 + }, + { + "epoch": 2.61, + "learning_rate": 3.701332750710072e-06, + "loss": 2.2072, + "step": 11938 + }, + { + "epoch": 2.61, + "learning_rate": 3.701223508848591e-06, + "loss": 1.3279, + "step": 11939 + }, + { + "epoch": 2.61, + "learning_rate": 3.7011142669871096e-06, + "loss": 2.4484, + "step": 11940 + }, + { + "epoch": 2.61, + "learning_rate": 3.7010050251256287e-06, + "loss": 1.1324, + "step": 11941 + }, + { + "epoch": 2.61, + "learning_rate": 3.7008957832641473e-06, + "loss": 2.4846, + "step": 11942 + }, + { + "epoch": 2.61, + "learning_rate": 3.7007865414026655e-06, + "loss": 3.3698, + "step": 11943 + }, + { + "epoch": 2.61, + "learning_rate": 3.7006772995411845e-06, + "loss": 4.0609, + "step": 11944 + }, + { + "epoch": 2.61, + "learning_rate": 3.700568057679703e-06, + "loss": 2.2326, + "step": 11945 + }, + { + "epoch": 2.61, + "learning_rate": 3.700458815818222e-06, + "loss": 2.6973, + "step": 11946 + }, + { + "epoch": 2.61, + "learning_rate": 3.7003495739567408e-06, + "loss": 2.3406, + "step": 11947 + }, + { + "epoch": 2.61, + "learning_rate": 3.700240332095259e-06, + "loss": 2.2344, + "step": 11948 + }, + { + "epoch": 2.61, + "learning_rate": 3.700131090233778e-06, + "loss": 1.1346, + "step": 11949 + }, + { + "epoch": 2.61, + "learning_rate": 3.7000218483722966e-06, + "loss": 2.4302, + "step": 11950 + }, + { + "epoch": 2.61, + "learning_rate": 3.699912606510815e-06, + "loss": 2.4963, + "step": 11951 + }, + { + "epoch": 2.61, + "learning_rate": 3.699803364649334e-06, + "loss": 3.6038, + "step": 11952 + }, + { + "epoch": 2.61, + "learning_rate": 3.6996941227878524e-06, + "loss": 1.993, + "step": 11953 + }, + { + "epoch": 2.61, + "learning_rate": 3.699584880926371e-06, + "loss": 0.9067, + "step": 11954 + }, + { + "epoch": 2.61, + "learning_rate": 3.69947563906489e-06, + "loss": 2.8033, + "step": 11955 + }, + { + "epoch": 2.61, + "learning_rate": 3.6993663972034087e-06, + "loss": 3.0213, + "step": 11956 + }, + { + "epoch": 2.61, + "learning_rate": 3.6992571553419273e-06, + "loss": 2.4188, + "step": 11957 + }, + { + "epoch": 2.61, + "learning_rate": 3.699147913480446e-06, + "loss": 3.2249, + "step": 11958 + }, + { + "epoch": 2.61, + "learning_rate": 3.6990386716189645e-06, + "loss": 1.8479, + "step": 11959 + }, + { + "epoch": 2.61, + "learning_rate": 3.6989294297574836e-06, + "loss": 3.9509, + "step": 11960 + }, + { + "epoch": 2.61, + "learning_rate": 3.698820187896002e-06, + "loss": 1.6446, + "step": 11961 + }, + { + "epoch": 2.61, + "learning_rate": 3.6987109460345204e-06, + "loss": 2.8332, + "step": 11962 + }, + { + "epoch": 2.61, + "learning_rate": 3.6986017041730394e-06, + "loss": 1.5777, + "step": 11963 + }, + { + "epoch": 2.61, + "learning_rate": 3.698492462311558e-06, + "loss": 2.328, + "step": 11964 + }, + { + "epoch": 2.61, + "learning_rate": 3.698383220450077e-06, + "loss": 0.6595, + "step": 11965 + }, + { + "epoch": 2.61, + "learning_rate": 3.6982739785885956e-06, + "loss": 2.1709, + "step": 11966 + }, + { + "epoch": 2.61, + "learning_rate": 3.698164736727114e-06, + "loss": 1.028, + "step": 11967 + }, + { + "epoch": 2.61, + "learning_rate": 3.698055494865633e-06, + "loss": 1.8851, + "step": 11968 + }, + { + "epoch": 2.61, + "learning_rate": 3.6979462530041515e-06, + "loss": 2.6908, + "step": 11969 + }, + { + "epoch": 2.61, + "learning_rate": 3.6978370111426705e-06, + "loss": 2.1736, + "step": 11970 + }, + { + "epoch": 2.61, + "learning_rate": 3.6977277692811887e-06, + "loss": 2.9748, + "step": 11971 + }, + { + "epoch": 2.61, + "learning_rate": 3.6976185274197073e-06, + "loss": 2.3167, + "step": 11972 + }, + { + "epoch": 2.61, + "learning_rate": 3.6975092855582263e-06, + "loss": 2.6328, + "step": 11973 + }, + { + "epoch": 2.61, + "learning_rate": 3.697400043696745e-06, + "loss": 2.0235, + "step": 11974 + }, + { + "epoch": 2.61, + "learning_rate": 3.697290801835264e-06, + "loss": 2.9545, + "step": 11975 + }, + { + "epoch": 2.61, + "learning_rate": 3.697181559973782e-06, + "loss": 1.3058, + "step": 11976 + }, + { + "epoch": 2.61, + "learning_rate": 3.6970723181123008e-06, + "loss": 2.4874, + "step": 11977 + }, + { + "epoch": 2.61, + "learning_rate": 3.69696307625082e-06, + "loss": 2.9959, + "step": 11978 + }, + { + "epoch": 2.61, + "learning_rate": 3.6968538343893384e-06, + "loss": 2.8308, + "step": 11979 + }, + { + "epoch": 2.61, + "learning_rate": 3.696744592527857e-06, + "loss": 2.6517, + "step": 11980 + }, + { + "epoch": 2.61, + "learning_rate": 3.6966353506663757e-06, + "loss": 0.9672, + "step": 11981 + }, + { + "epoch": 2.62, + "learning_rate": 3.6965261088048943e-06, + "loss": 1.4351, + "step": 11982 + }, + { + "epoch": 2.62, + "learning_rate": 3.696416866943413e-06, + "loss": 2.4457, + "step": 11983 + }, + { + "epoch": 2.62, + "learning_rate": 3.696307625081932e-06, + "loss": 2.5707, + "step": 11984 + }, + { + "epoch": 2.62, + "learning_rate": 3.69619838322045e-06, + "loss": 2.7382, + "step": 11985 + }, + { + "epoch": 2.62, + "learning_rate": 3.6960891413589687e-06, + "loss": 2.5254, + "step": 11986 + }, + { + "epoch": 2.62, + "learning_rate": 3.6959798994974877e-06, + "loss": 1.7592, + "step": 11987 + }, + { + "epoch": 2.62, + "learning_rate": 3.6958706576360064e-06, + "loss": 1.5753, + "step": 11988 + }, + { + "epoch": 2.62, + "learning_rate": 3.6957614157745254e-06, + "loss": 2.8167, + "step": 11989 + }, + { + "epoch": 2.62, + "learning_rate": 3.6956521739130436e-06, + "loss": 2.7231, + "step": 11990 + }, + { + "epoch": 2.62, + "learning_rate": 3.695542932051562e-06, + "loss": 2.4658, + "step": 11991 + }, + { + "epoch": 2.62, + "learning_rate": 3.6954336901900812e-06, + "loss": 1.4818, + "step": 11992 + }, + { + "epoch": 2.62, + "learning_rate": 3.6953244483286e-06, + "loss": 1.7829, + "step": 11993 + }, + { + "epoch": 2.62, + "learning_rate": 3.695215206467119e-06, + "loss": 3.2543, + "step": 11994 + }, + { + "epoch": 2.62, + "learning_rate": 3.695105964605637e-06, + "loss": 1.8864, + "step": 11995 + }, + { + "epoch": 2.62, + "learning_rate": 3.6949967227441557e-06, + "loss": 2.2046, + "step": 11996 + }, + { + "epoch": 2.62, + "learning_rate": 3.6948874808826747e-06, + "loss": 3.5315, + "step": 11997 + }, + { + "epoch": 2.62, + "learning_rate": 3.6947782390211933e-06, + "loss": 1.7289, + "step": 11998 + }, + { + "epoch": 2.62, + "learning_rate": 3.6946689971597123e-06, + "loss": 2.2378, + "step": 11999 + }, + { + "epoch": 2.62, + "learning_rate": 3.6945597552982305e-06, + "loss": 0.6982, + "step": 12000 + }, + { + "epoch": 2.62, + "learning_rate": 3.694450513436749e-06, + "loss": 2.1294, + "step": 12001 + }, + { + "epoch": 2.62, + "learning_rate": 3.694341271575268e-06, + "loss": 2.8353, + "step": 12002 + }, + { + "epoch": 2.62, + "learning_rate": 3.694232029713787e-06, + "loss": 1.6123, + "step": 12003 + }, + { + "epoch": 2.62, + "learning_rate": 3.694122787852305e-06, + "loss": 1.8209, + "step": 12004 + }, + { + "epoch": 2.62, + "learning_rate": 3.694013545990824e-06, + "loss": 2.4914, + "step": 12005 + }, + { + "epoch": 2.62, + "learning_rate": 3.6939043041293426e-06, + "loss": 2.5766, + "step": 12006 + }, + { + "epoch": 2.62, + "learning_rate": 3.6937950622678612e-06, + "loss": 2.3286, + "step": 12007 + }, + { + "epoch": 2.62, + "learning_rate": 3.6936858204063803e-06, + "loss": 1.6907, + "step": 12008 + }, + { + "epoch": 2.62, + "learning_rate": 3.6935765785448985e-06, + "loss": 2.1081, + "step": 12009 + }, + { + "epoch": 2.62, + "learning_rate": 3.6934673366834175e-06, + "loss": 2.3017, + "step": 12010 + }, + { + "epoch": 2.62, + "learning_rate": 3.693358094821936e-06, + "loss": 2.151, + "step": 12011 + }, + { + "epoch": 2.62, + "learning_rate": 3.6932488529604547e-06, + "loss": 3.298, + "step": 12012 + }, + { + "epoch": 2.62, + "learning_rate": 3.6931396110989737e-06, + "loss": 1.9091, + "step": 12013 + }, + { + "epoch": 2.62, + "learning_rate": 3.693030369237492e-06, + "loss": 2.5943, + "step": 12014 + }, + { + "epoch": 2.62, + "learning_rate": 3.6929211273760105e-06, + "loss": 3.3705, + "step": 12015 + }, + { + "epoch": 2.62, + "learning_rate": 3.6928118855145296e-06, + "loss": 1.5235, + "step": 12016 + }, + { + "epoch": 2.62, + "learning_rate": 3.692702643653048e-06, + "loss": 2.9946, + "step": 12017 + }, + { + "epoch": 2.62, + "learning_rate": 3.6925934017915664e-06, + "loss": 2.6318, + "step": 12018 + }, + { + "epoch": 2.62, + "learning_rate": 3.6924841599300854e-06, + "loss": 2.6182, + "step": 12019 + }, + { + "epoch": 2.62, + "learning_rate": 3.692374918068604e-06, + "loss": 1.5605, + "step": 12020 + }, + { + "epoch": 2.62, + "learning_rate": 3.692265676207123e-06, + "loss": 0.7662, + "step": 12021 + }, + { + "epoch": 2.62, + "learning_rate": 3.6921564343456417e-06, + "loss": 2.8548, + "step": 12022 + }, + { + "epoch": 2.62, + "learning_rate": 3.69204719248416e-06, + "loss": 1.6295, + "step": 12023 + }, + { + "epoch": 2.62, + "learning_rate": 3.691937950622679e-06, + "loss": 1.6303, + "step": 12024 + }, + { + "epoch": 2.62, + "learning_rate": 3.6918287087611975e-06, + "loss": 3.2503, + "step": 12025 + }, + { + "epoch": 2.62, + "learning_rate": 3.6917194668997165e-06, + "loss": 2.4947, + "step": 12026 + }, + { + "epoch": 2.62, + "learning_rate": 3.691610225038235e-06, + "loss": 2.2124, + "step": 12027 + }, + { + "epoch": 2.63, + "learning_rate": 3.6915009831767533e-06, + "loss": 1.4048, + "step": 12028 + }, + { + "epoch": 2.63, + "learning_rate": 3.6913917413152724e-06, + "loss": 2.3501, + "step": 12029 + }, + { + "epoch": 2.63, + "learning_rate": 3.691282499453791e-06, + "loss": 1.8148, + "step": 12030 + }, + { + "epoch": 2.63, + "learning_rate": 3.69117325759231e-06, + "loss": 2.3352, + "step": 12031 + }, + { + "epoch": 2.63, + "learning_rate": 3.6910640157308286e-06, + "loss": 1.8186, + "step": 12032 + }, + { + "epoch": 2.63, + "learning_rate": 3.690954773869347e-06, + "loss": 2.4195, + "step": 12033 + }, + { + "epoch": 2.63, + "learning_rate": 3.690845532007866e-06, + "loss": 2.1273, + "step": 12034 + }, + { + "epoch": 2.63, + "learning_rate": 3.6907362901463845e-06, + "loss": 2.9234, + "step": 12035 + }, + { + "epoch": 2.63, + "learning_rate": 3.690627048284903e-06, + "loss": 2.2873, + "step": 12036 + }, + { + "epoch": 2.63, + "learning_rate": 3.6905178064234217e-06, + "loss": 2.5088, + "step": 12037 + }, + { + "epoch": 2.63, + "learning_rate": 3.6904085645619403e-06, + "loss": 1.3147, + "step": 12038 + }, + { + "epoch": 2.63, + "learning_rate": 3.690299322700459e-06, + "loss": 1.4737, + "step": 12039 + }, + { + "epoch": 2.63, + "learning_rate": 3.690190080838978e-06, + "loss": 2.5441, + "step": 12040 + }, + { + "epoch": 2.63, + "learning_rate": 3.6900808389774966e-06, + "loss": 1.7403, + "step": 12041 + }, + { + "epoch": 2.63, + "learning_rate": 3.6899715971160147e-06, + "loss": 1.4474, + "step": 12042 + }, + { + "epoch": 2.63, + "learning_rate": 3.6898623552545338e-06, + "loss": 1.7483, + "step": 12043 + }, + { + "epoch": 2.63, + "learning_rate": 3.6897531133930524e-06, + "loss": 2.3363, + "step": 12044 + }, + { + "epoch": 2.63, + "learning_rate": 3.6896438715315714e-06, + "loss": 2.1494, + "step": 12045 + }, + { + "epoch": 2.63, + "learning_rate": 3.68953462967009e-06, + "loss": 1.8248, + "step": 12046 + }, + { + "epoch": 2.63, + "learning_rate": 3.6894253878086082e-06, + "loss": 2.4323, + "step": 12047 + }, + { + "epoch": 2.63, + "learning_rate": 3.6893161459471273e-06, + "loss": 1.2572, + "step": 12048 + }, + { + "epoch": 2.63, + "learning_rate": 3.689206904085646e-06, + "loss": 2.6088, + "step": 12049 + }, + { + "epoch": 2.63, + "learning_rate": 3.689097662224165e-06, + "loss": 1.5853, + "step": 12050 + }, + { + "epoch": 2.63, + "learning_rate": 3.688988420362683e-06, + "loss": 2.1038, + "step": 12051 + }, + { + "epoch": 2.63, + "learning_rate": 3.6888791785012017e-06, + "loss": 3.2219, + "step": 12052 + }, + { + "epoch": 2.63, + "learning_rate": 3.6887699366397207e-06, + "loss": 2.6344, + "step": 12053 + }, + { + "epoch": 2.63, + "learning_rate": 3.6886606947782393e-06, + "loss": 2.6328, + "step": 12054 + }, + { + "epoch": 2.63, + "learning_rate": 3.6885514529167584e-06, + "loss": 2.444, + "step": 12055 + }, + { + "epoch": 2.63, + "learning_rate": 3.6884422110552766e-06, + "loss": 2.7117, + "step": 12056 + }, + { + "epoch": 2.63, + "learning_rate": 3.688332969193795e-06, + "loss": 1.6362, + "step": 12057 + }, + { + "epoch": 2.63, + "learning_rate": 3.688223727332314e-06, + "loss": 2.7698, + "step": 12058 + }, + { + "epoch": 2.63, + "learning_rate": 3.688114485470833e-06, + "loss": 2.9649, + "step": 12059 + }, + { + "epoch": 2.63, + "learning_rate": 3.6880052436093514e-06, + "loss": 0.5875, + "step": 12060 + }, + { + "epoch": 2.63, + "learning_rate": 3.68789600174787e-06, + "loss": 1.0583, + "step": 12061 + }, + { + "epoch": 2.63, + "learning_rate": 3.6877867598863887e-06, + "loss": 3.8372, + "step": 12062 + }, + { + "epoch": 2.63, + "learning_rate": 3.6876775180249077e-06, + "loss": 2.4626, + "step": 12063 + }, + { + "epoch": 2.63, + "learning_rate": 3.6875682761634263e-06, + "loss": 1.9911, + "step": 12064 + }, + { + "epoch": 2.63, + "learning_rate": 3.687459034301945e-06, + "loss": 2.3402, + "step": 12065 + }, + { + "epoch": 2.63, + "learning_rate": 3.6873497924404635e-06, + "loss": 2.2672, + "step": 12066 + }, + { + "epoch": 2.63, + "learning_rate": 3.687240550578982e-06, + "loss": 1.7284, + "step": 12067 + }, + { + "epoch": 2.63, + "learning_rate": 3.6871313087175007e-06, + "loss": 2.0768, + "step": 12068 + }, + { + "epoch": 2.63, + "learning_rate": 3.6870220668560198e-06, + "loss": 2.9819, + "step": 12069 + }, + { + "epoch": 2.63, + "learning_rate": 3.686912824994538e-06, + "loss": 3.0655, + "step": 12070 + }, + { + "epoch": 2.63, + "learning_rate": 3.6868035831330566e-06, + "loss": 2.5365, + "step": 12071 + }, + { + "epoch": 2.63, + "learning_rate": 3.6866943412715756e-06, + "loss": 1.7122, + "step": 12072 + }, + { + "epoch": 2.63, + "learning_rate": 3.6865850994100942e-06, + "loss": 2.5856, + "step": 12073 + }, + { + "epoch": 2.64, + "learning_rate": 3.6864758575486133e-06, + "loss": 2.3273, + "step": 12074 + }, + { + "epoch": 2.64, + "learning_rate": 3.6863666156871314e-06, + "loss": 2.4175, + "step": 12075 + }, + { + "epoch": 2.64, + "learning_rate": 3.68625737382565e-06, + "loss": 1.9304, + "step": 12076 + }, + { + "epoch": 2.64, + "learning_rate": 3.686148131964169e-06, + "loss": 2.3873, + "step": 12077 + }, + { + "epoch": 2.64, + "learning_rate": 3.6860388901026877e-06, + "loss": 1.0689, + "step": 12078 + }, + { + "epoch": 2.64, + "learning_rate": 3.6859296482412067e-06, + "loss": 0.6781, + "step": 12079 + }, + { + "epoch": 2.64, + "learning_rate": 3.685820406379725e-06, + "loss": 1.6006, + "step": 12080 + }, + { + "epoch": 2.64, + "learning_rate": 3.6857111645182435e-06, + "loss": 2.1796, + "step": 12081 + }, + { + "epoch": 2.64, + "learning_rate": 3.6856019226567626e-06, + "loss": 2.4411, + "step": 12082 + }, + { + "epoch": 2.64, + "learning_rate": 3.685492680795281e-06, + "loss": 1.9841, + "step": 12083 + }, + { + "epoch": 2.64, + "learning_rate": 3.6853834389337994e-06, + "loss": 1.9964, + "step": 12084 + }, + { + "epoch": 2.64, + "learning_rate": 3.6852741970723184e-06, + "loss": 2.1837, + "step": 12085 + }, + { + "epoch": 2.64, + "learning_rate": 3.685164955210837e-06, + "loss": 3.3313, + "step": 12086 + }, + { + "epoch": 2.64, + "learning_rate": 3.685055713349356e-06, + "loss": 3.8617, + "step": 12087 + }, + { + "epoch": 2.64, + "learning_rate": 3.6849464714878747e-06, + "loss": 1.7685, + "step": 12088 + }, + { + "epoch": 2.64, + "learning_rate": 3.684837229626393e-06, + "loss": 2.685, + "step": 12089 + }, + { + "epoch": 2.64, + "learning_rate": 3.684727987764912e-06, + "loss": 2.3228, + "step": 12090 + }, + { + "epoch": 2.64, + "learning_rate": 3.6846187459034305e-06, + "loss": 1.8321, + "step": 12091 + }, + { + "epoch": 2.64, + "learning_rate": 3.684509504041949e-06, + "loss": 2.3056, + "step": 12092 + }, + { + "epoch": 2.64, + "learning_rate": 3.684400262180468e-06, + "loss": 1.2964, + "step": 12093 + }, + { + "epoch": 2.64, + "learning_rate": 3.6842910203189863e-06, + "loss": 3.8591, + "step": 12094 + }, + { + "epoch": 2.64, + "learning_rate": 3.684181778457505e-06, + "loss": 2.2091, + "step": 12095 + }, + { + "epoch": 2.64, + "learning_rate": 3.684072536596024e-06, + "loss": 2.1315, + "step": 12096 + }, + { + "epoch": 2.64, + "learning_rate": 3.6839632947345426e-06, + "loss": 3.0947, + "step": 12097 + }, + { + "epoch": 2.64, + "learning_rate": 3.6838540528730616e-06, + "loss": 1.9357, + "step": 12098 + }, + { + "epoch": 2.64, + "learning_rate": 3.68374481101158e-06, + "loss": 2.7117, + "step": 12099 + }, + { + "epoch": 2.64, + "learning_rate": 3.6836355691500984e-06, + "loss": 2.1729, + "step": 12100 + }, + { + "epoch": 2.64, + "learning_rate": 3.6835263272886174e-06, + "loss": 2.7497, + "step": 12101 + }, + { + "epoch": 2.64, + "learning_rate": 3.683417085427136e-06, + "loss": 2.0725, + "step": 12102 + }, + { + "epoch": 2.64, + "learning_rate": 3.6833078435656542e-06, + "loss": 2.286, + "step": 12103 + }, + { + "epoch": 2.64, + "learning_rate": 3.6831986017041733e-06, + "loss": 2.422, + "step": 12104 + }, + { + "epoch": 2.64, + "learning_rate": 3.683089359842692e-06, + "loss": 3.817, + "step": 12105 + }, + { + "epoch": 2.64, + "learning_rate": 3.682980117981211e-06, + "loss": 2.7628, + "step": 12106 + }, + { + "epoch": 2.64, + "learning_rate": 3.6828708761197295e-06, + "loss": 2.4997, + "step": 12107 + }, + { + "epoch": 2.64, + "learning_rate": 3.6827616342582477e-06, + "loss": 2.4239, + "step": 12108 + }, + { + "epoch": 2.64, + "learning_rate": 3.6826523923967668e-06, + "loss": 1.736, + "step": 12109 + }, + { + "epoch": 2.64, + "learning_rate": 3.6825431505352854e-06, + "loss": 2.7582, + "step": 12110 + }, + { + "epoch": 2.64, + "learning_rate": 3.6824339086738044e-06, + "loss": 2.2161, + "step": 12111 + }, + { + "epoch": 2.64, + "learning_rate": 3.682324666812323e-06, + "loss": 1.4971, + "step": 12112 + }, + { + "epoch": 2.64, + "learning_rate": 3.682215424950841e-06, + "loss": 3.842, + "step": 12113 + }, + { + "epoch": 2.64, + "learning_rate": 3.6821061830893602e-06, + "loss": 2.3467, + "step": 12114 + }, + { + "epoch": 2.64, + "learning_rate": 3.681996941227879e-06, + "loss": 3.5022, + "step": 12115 + }, + { + "epoch": 2.64, + "learning_rate": 3.681887699366398e-06, + "loss": 2.1095, + "step": 12116 + }, + { + "epoch": 2.64, + "learning_rate": 3.681778457504916e-06, + "loss": 1.8068, + "step": 12117 + }, + { + "epoch": 2.64, + "learning_rate": 3.6816692156434347e-06, + "loss": 1.8941, + "step": 12118 + }, + { + "epoch": 2.64, + "learning_rate": 3.6815599737819537e-06, + "loss": 1.9307, + "step": 12119 + }, + { + "epoch": 2.65, + "learning_rate": 3.6814507319204723e-06, + "loss": 0.9867, + "step": 12120 + }, + { + "epoch": 2.65, + "learning_rate": 3.681341490058991e-06, + "loss": 3.5952, + "step": 12121 + }, + { + "epoch": 2.65, + "learning_rate": 3.6812322481975095e-06, + "loss": 1.7814, + "step": 12122 + }, + { + "epoch": 2.65, + "learning_rate": 3.681123006336028e-06, + "loss": 2.0448, + "step": 12123 + }, + { + "epoch": 2.65, + "learning_rate": 3.6810137644745468e-06, + "loss": 2.9024, + "step": 12124 + }, + { + "epoch": 2.65, + "learning_rate": 3.680904522613066e-06, + "loss": 2.2457, + "step": 12125 + }, + { + "epoch": 2.65, + "learning_rate": 3.6807952807515844e-06, + "loss": 2.5967, + "step": 12126 + }, + { + "epoch": 2.65, + "learning_rate": 3.6806860388901026e-06, + "loss": 2.3694, + "step": 12127 + }, + { + "epoch": 2.65, + "learning_rate": 3.6805767970286216e-06, + "loss": 2.0314, + "step": 12128 + }, + { + "epoch": 2.65, + "learning_rate": 3.6804675551671402e-06, + "loss": 1.578, + "step": 12129 + }, + { + "epoch": 2.65, + "learning_rate": 3.6803583133056593e-06, + "loss": 1.8444, + "step": 12130 + }, + { + "epoch": 2.65, + "learning_rate": 3.680249071444178e-06, + "loss": 2.8567, + "step": 12131 + }, + { + "epoch": 2.65, + "learning_rate": 3.680139829582696e-06, + "loss": 3.1816, + "step": 12132 + }, + { + "epoch": 2.65, + "learning_rate": 3.680030587721215e-06, + "loss": 2.4813, + "step": 12133 + }, + { + "epoch": 2.65, + "learning_rate": 3.6799213458597337e-06, + "loss": 2.9923, + "step": 12134 + }, + { + "epoch": 2.65, + "learning_rate": 3.6798121039982528e-06, + "loss": 1.8045, + "step": 12135 + }, + { + "epoch": 2.65, + "learning_rate": 3.679702862136771e-06, + "loss": 2.3786, + "step": 12136 + }, + { + "epoch": 2.65, + "learning_rate": 3.6795936202752896e-06, + "loss": 2.7588, + "step": 12137 + }, + { + "epoch": 2.65, + "learning_rate": 3.6794843784138086e-06, + "loss": 1.971, + "step": 12138 + }, + { + "epoch": 2.65, + "learning_rate": 3.679375136552327e-06, + "loss": 2.8407, + "step": 12139 + }, + { + "epoch": 2.65, + "learning_rate": 3.6792658946908462e-06, + "loss": 3.0368, + "step": 12140 + }, + { + "epoch": 2.65, + "learning_rate": 3.6791566528293644e-06, + "loss": 2.037, + "step": 12141 + }, + { + "epoch": 2.65, + "learning_rate": 3.679047410967883e-06, + "loss": 3.8513, + "step": 12142 + }, + { + "epoch": 2.65, + "learning_rate": 3.678938169106402e-06, + "loss": 2.2312, + "step": 12143 + }, + { + "epoch": 2.65, + "learning_rate": 3.6788289272449207e-06, + "loss": 2.0761, + "step": 12144 + }, + { + "epoch": 2.65, + "learning_rate": 3.6787196853834393e-06, + "loss": 1.8226, + "step": 12145 + }, + { + "epoch": 2.65, + "learning_rate": 3.678610443521958e-06, + "loss": 2.3086, + "step": 12146 + }, + { + "epoch": 2.65, + "learning_rate": 3.6785012016604765e-06, + "loss": 2.0092, + "step": 12147 + }, + { + "epoch": 2.65, + "learning_rate": 3.678391959798995e-06, + "loss": 1.3541, + "step": 12148 + }, + { + "epoch": 2.65, + "learning_rate": 3.678282717937514e-06, + "loss": 1.7535, + "step": 12149 + }, + { + "epoch": 2.65, + "learning_rate": 3.6781734760760323e-06, + "loss": 2.357, + "step": 12150 + }, + { + "epoch": 2.65, + "learning_rate": 3.678064234214551e-06, + "loss": 1.9157, + "step": 12151 + }, + { + "epoch": 2.65, + "learning_rate": 3.67795499235307e-06, + "loss": 1.5245, + "step": 12152 + }, + { + "epoch": 2.65, + "learning_rate": 3.6778457504915886e-06, + "loss": 2.974, + "step": 12153 + }, + { + "epoch": 2.65, + "learning_rate": 3.6777365086301076e-06, + "loss": 3.0995, + "step": 12154 + }, + { + "epoch": 2.65, + "learning_rate": 3.677627266768626e-06, + "loss": 2.2264, + "step": 12155 + }, + { + "epoch": 2.65, + "learning_rate": 3.6775180249071444e-06, + "loss": 2.6646, + "step": 12156 + }, + { + "epoch": 2.65, + "learning_rate": 3.6774087830456635e-06, + "loss": 3.3097, + "step": 12157 + }, + { + "epoch": 2.65, + "learning_rate": 3.677299541184182e-06, + "loss": 2.1774, + "step": 12158 + }, + { + "epoch": 2.65, + "learning_rate": 3.677190299322701e-06, + "loss": 1.9468, + "step": 12159 + }, + { + "epoch": 2.65, + "learning_rate": 3.6770810574612193e-06, + "loss": 1.4188, + "step": 12160 + }, + { + "epoch": 2.65, + "learning_rate": 3.676971815599738e-06, + "loss": 2.5075, + "step": 12161 + }, + { + "epoch": 2.65, + "learning_rate": 3.676862573738257e-06, + "loss": 2.5183, + "step": 12162 + }, + { + "epoch": 2.65, + "learning_rate": 3.6767533318767756e-06, + "loss": 1.8172, + "step": 12163 + }, + { + "epoch": 2.65, + "learning_rate": 3.6766440900152946e-06, + "loss": 1.0929, + "step": 12164 + }, + { + "epoch": 2.65, + "learning_rate": 3.6765348481538128e-06, + "loss": 2.3054, + "step": 12165 + }, + { + "epoch": 2.66, + "learning_rate": 3.6764256062923314e-06, + "loss": 2.8153, + "step": 12166 + }, + { + "epoch": 2.66, + "learning_rate": 3.6763163644308504e-06, + "loss": 2.8313, + "step": 12167 + }, + { + "epoch": 2.66, + "learning_rate": 3.676207122569369e-06, + "loss": 1.2077, + "step": 12168 + }, + { + "epoch": 2.66, + "learning_rate": 3.6760978807078872e-06, + "loss": 2.6486, + "step": 12169 + }, + { + "epoch": 2.66, + "learning_rate": 3.6759886388464063e-06, + "loss": 0.7996, + "step": 12170 + }, + { + "epoch": 2.66, + "learning_rate": 3.675879396984925e-06, + "loss": 1.6733, + "step": 12171 + }, + { + "epoch": 2.66, + "learning_rate": 3.675770155123444e-06, + "loss": 1.523, + "step": 12172 + }, + { + "epoch": 2.66, + "learning_rate": 3.6756609132619625e-06, + "loss": 2.9319, + "step": 12173 + }, + { + "epoch": 2.66, + "learning_rate": 3.6755516714004807e-06, + "loss": 2.9681, + "step": 12174 + }, + { + "epoch": 2.66, + "learning_rate": 3.6754424295389997e-06, + "loss": 2.6774, + "step": 12175 + }, + { + "epoch": 2.66, + "learning_rate": 3.6753331876775184e-06, + "loss": 1.9387, + "step": 12176 + }, + { + "epoch": 2.66, + "learning_rate": 3.675223945816037e-06, + "loss": 1.4424, + "step": 12177 + }, + { + "epoch": 2.66, + "learning_rate": 3.675114703954556e-06, + "loss": 2.3689, + "step": 12178 + }, + { + "epoch": 2.66, + "learning_rate": 3.675005462093074e-06, + "loss": 3.0899, + "step": 12179 + }, + { + "epoch": 2.66, + "learning_rate": 3.674896220231593e-06, + "loss": 2.1865, + "step": 12180 + }, + { + "epoch": 2.66, + "learning_rate": 3.674786978370112e-06, + "loss": 3.1115, + "step": 12181 + }, + { + "epoch": 2.66, + "learning_rate": 3.6746777365086304e-06, + "loss": 1.7481, + "step": 12182 + }, + { + "epoch": 2.66, + "learning_rate": 3.6745684946471486e-06, + "loss": 1.4016, + "step": 12183 + }, + { + "epoch": 2.66, + "learning_rate": 3.6744592527856677e-06, + "loss": 1.1631, + "step": 12184 + }, + { + "epoch": 2.66, + "learning_rate": 3.6743500109241863e-06, + "loss": 2.5183, + "step": 12185 + }, + { + "epoch": 2.66, + "learning_rate": 3.6742407690627053e-06, + "loss": 3.2446, + "step": 12186 + }, + { + "epoch": 2.66, + "learning_rate": 3.674131527201224e-06, + "loss": 3.8428, + "step": 12187 + }, + { + "epoch": 2.66, + "learning_rate": 3.674022285339742e-06, + "loss": 1.7533, + "step": 12188 + }, + { + "epoch": 2.66, + "learning_rate": 3.673913043478261e-06, + "loss": 1.3377, + "step": 12189 + }, + { + "epoch": 2.66, + "learning_rate": 3.6738038016167798e-06, + "loss": 2.6516, + "step": 12190 + }, + { + "epoch": 2.66, + "learning_rate": 3.6736945597552988e-06, + "loss": 2.6418, + "step": 12191 + }, + { + "epoch": 2.66, + "learning_rate": 3.6735853178938174e-06, + "loss": 1.8503, + "step": 12192 + }, + { + "epoch": 2.66, + "learning_rate": 3.6734760760323356e-06, + "loss": 1.7951, + "step": 12193 + }, + { + "epoch": 2.66, + "learning_rate": 3.6733668341708546e-06, + "loss": 2.7021, + "step": 12194 + }, + { + "epoch": 2.66, + "learning_rate": 3.6732575923093732e-06, + "loss": 3.4928, + "step": 12195 + }, + { + "epoch": 2.66, + "learning_rate": 3.6731483504478923e-06, + "loss": 2.0236, + "step": 12196 + }, + { + "epoch": 2.66, + "learning_rate": 3.673039108586411e-06, + "loss": 1.2615, + "step": 12197 + }, + { + "epoch": 2.66, + "learning_rate": 3.672929866724929e-06, + "loss": 2.6201, + "step": 12198 + }, + { + "epoch": 2.66, + "learning_rate": 3.672820624863448e-06, + "loss": 1.0918, + "step": 12199 + }, + { + "epoch": 2.66, + "learning_rate": 3.6727113830019667e-06, + "loss": 1.8901, + "step": 12200 + }, + { + "epoch": 2.66, + "learning_rate": 3.6726021411404853e-06, + "loss": 1.2208, + "step": 12201 + }, + { + "epoch": 2.66, + "learning_rate": 3.672492899279004e-06, + "loss": 2.1753, + "step": 12202 + }, + { + "epoch": 2.66, + "learning_rate": 3.6723836574175225e-06, + "loss": 2.7117, + "step": 12203 + }, + { + "epoch": 2.66, + "learning_rate": 3.672274415556041e-06, + "loss": 2.4863, + "step": 12204 + }, + { + "epoch": 2.66, + "learning_rate": 3.67216517369456e-06, + "loss": 2.0962, + "step": 12205 + }, + { + "epoch": 2.66, + "learning_rate": 3.672055931833079e-06, + "loss": 3.3891, + "step": 12206 + }, + { + "epoch": 2.66, + "learning_rate": 3.6719466899715974e-06, + "loss": 1.686, + "step": 12207 + }, + { + "epoch": 2.66, + "learning_rate": 3.671837448110116e-06, + "loss": 1.6666, + "step": 12208 + }, + { + "epoch": 2.66, + "learning_rate": 3.6717282062486346e-06, + "loss": 2.6322, + "step": 12209 + }, + { + "epoch": 2.66, + "learning_rate": 3.6716189643871537e-06, + "loss": 2.9343, + "step": 12210 + }, + { + "epoch": 2.66, + "learning_rate": 3.6715097225256723e-06, + "loss": 2.5144, + "step": 12211 + }, + { + "epoch": 2.67, + "learning_rate": 3.6714004806641905e-06, + "loss": 3.331, + "step": 12212 + }, + { + "epoch": 2.67, + "learning_rate": 3.6712912388027095e-06, + "loss": 1.5497, + "step": 12213 + }, + { + "epoch": 2.67, + "learning_rate": 3.671181996941228e-06, + "loss": 2.7145, + "step": 12214 + }, + { + "epoch": 2.67, + "learning_rate": 3.671072755079747e-06, + "loss": 1.8674, + "step": 12215 + }, + { + "epoch": 2.67, + "learning_rate": 3.6709635132182653e-06, + "loss": 2.0228, + "step": 12216 + }, + { + "epoch": 2.67, + "learning_rate": 3.670854271356784e-06, + "loss": 2.3325, + "step": 12217 + }, + { + "epoch": 2.67, + "learning_rate": 3.670745029495303e-06, + "loss": 2.7797, + "step": 12218 + }, + { + "epoch": 2.67, + "learning_rate": 3.6706357876338216e-06, + "loss": 2.3208, + "step": 12219 + }, + { + "epoch": 2.67, + "learning_rate": 3.6705265457723406e-06, + "loss": 2.0495, + "step": 12220 + }, + { + "epoch": 2.67, + "learning_rate": 3.670417303910859e-06, + "loss": 1.5902, + "step": 12221 + }, + { + "epoch": 2.67, + "learning_rate": 3.6703080620493774e-06, + "loss": 3.0824, + "step": 12222 + }, + { + "epoch": 2.67, + "learning_rate": 3.6701988201878965e-06, + "loss": 0.8687, + "step": 12223 + }, + { + "epoch": 2.67, + "learning_rate": 3.670089578326415e-06, + "loss": 2.1334, + "step": 12224 + }, + { + "epoch": 2.67, + "learning_rate": 3.669980336464934e-06, + "loss": 2.303, + "step": 12225 + }, + { + "epoch": 2.67, + "learning_rate": 3.6698710946034523e-06, + "loss": 2.0014, + "step": 12226 + }, + { + "epoch": 2.67, + "learning_rate": 3.669761852741971e-06, + "loss": 1.6219, + "step": 12227 + }, + { + "epoch": 2.67, + "learning_rate": 3.66965261088049e-06, + "loss": 3.2444, + "step": 12228 + }, + { + "epoch": 2.67, + "learning_rate": 3.6695433690190085e-06, + "loss": 1.0776, + "step": 12229 + }, + { + "epoch": 2.67, + "learning_rate": 3.669434127157527e-06, + "loss": 1.4691, + "step": 12230 + }, + { + "epoch": 2.67, + "learning_rate": 3.6693248852960458e-06, + "loss": 2.8787, + "step": 12231 + }, + { + "epoch": 2.67, + "learning_rate": 3.6692156434345644e-06, + "loss": 1.8019, + "step": 12232 + }, + { + "epoch": 2.67, + "learning_rate": 3.669106401573083e-06, + "loss": 2.4948, + "step": 12233 + }, + { + "epoch": 2.67, + "learning_rate": 3.668997159711602e-06, + "loss": 2.1809, + "step": 12234 + }, + { + "epoch": 2.67, + "learning_rate": 3.66888791785012e-06, + "loss": 1.9448, + "step": 12235 + }, + { + "epoch": 2.67, + "learning_rate": 3.668778675988639e-06, + "loss": 1.0084, + "step": 12236 + }, + { + "epoch": 2.67, + "learning_rate": 3.668669434127158e-06, + "loss": 3.2161, + "step": 12237 + }, + { + "epoch": 2.67, + "learning_rate": 3.6685601922656765e-06, + "loss": 2.7089, + "step": 12238 + }, + { + "epoch": 2.67, + "learning_rate": 3.6684509504041955e-06, + "loss": 2.8126, + "step": 12239 + }, + { + "epoch": 2.67, + "learning_rate": 3.6683417085427137e-06, + "loss": 3.0412, + "step": 12240 + }, + { + "epoch": 2.67, + "learning_rate": 3.6682324666812323e-06, + "loss": 2.3514, + "step": 12241 + }, + { + "epoch": 2.67, + "learning_rate": 3.6681232248197513e-06, + "loss": 3.5922, + "step": 12242 + }, + { + "epoch": 2.67, + "learning_rate": 3.66801398295827e-06, + "loss": 2.2567, + "step": 12243 + }, + { + "epoch": 2.67, + "learning_rate": 3.667904741096789e-06, + "loss": 2.5018, + "step": 12244 + }, + { + "epoch": 2.67, + "learning_rate": 3.667795499235307e-06, + "loss": 2.2609, + "step": 12245 + }, + { + "epoch": 2.67, + "learning_rate": 3.6676862573738258e-06, + "loss": 1.8458, + "step": 12246 + }, + { + "epoch": 2.67, + "learning_rate": 3.667577015512345e-06, + "loss": 2.3508, + "step": 12247 + }, + { + "epoch": 2.67, + "learning_rate": 3.6674677736508634e-06, + "loss": 2.3764, + "step": 12248 + }, + { + "epoch": 2.67, + "learning_rate": 3.6673585317893816e-06, + "loss": 2.3399, + "step": 12249 + }, + { + "epoch": 2.67, + "learning_rate": 3.6672492899279006e-06, + "loss": 2.9361, + "step": 12250 + }, + { + "epoch": 2.67, + "learning_rate": 3.6671400480664193e-06, + "loss": 2.6151, + "step": 12251 + }, + { + "epoch": 2.67, + "learning_rate": 3.6670308062049383e-06, + "loss": 2.8209, + "step": 12252 + }, + { + "epoch": 2.67, + "learning_rate": 3.666921564343457e-06, + "loss": 2.2171, + "step": 12253 + }, + { + "epoch": 2.67, + "learning_rate": 3.666812322481975e-06, + "loss": 2.8007, + "step": 12254 + }, + { + "epoch": 2.67, + "learning_rate": 3.666703080620494e-06, + "loss": 2.8756, + "step": 12255 + }, + { + "epoch": 2.67, + "learning_rate": 3.6665938387590127e-06, + "loss": 2.3631, + "step": 12256 + }, + { + "epoch": 2.68, + "learning_rate": 3.6664845968975313e-06, + "loss": 2.7743, + "step": 12257 + }, + { + "epoch": 2.68, + "learning_rate": 3.6663753550360504e-06, + "loss": 2.6094, + "step": 12258 + }, + { + "epoch": 2.68, + "learning_rate": 3.6662661131745686e-06, + "loss": 1.9132, + "step": 12259 + }, + { + "epoch": 2.68, + "learning_rate": 3.6661568713130876e-06, + "loss": 2.5828, + "step": 12260 + }, + { + "epoch": 2.68, + "learning_rate": 3.6660476294516062e-06, + "loss": 2.0872, + "step": 12261 + }, + { + "epoch": 2.68, + "learning_rate": 3.665938387590125e-06, + "loss": 2.3092, + "step": 12262 + }, + { + "epoch": 2.68, + "learning_rate": 3.665829145728644e-06, + "loss": 2.3612, + "step": 12263 + }, + { + "epoch": 2.68, + "learning_rate": 3.665719903867162e-06, + "loss": 2.6688, + "step": 12264 + }, + { + "epoch": 2.68, + "learning_rate": 3.6656106620056807e-06, + "loss": 1.7221, + "step": 12265 + }, + { + "epoch": 2.68, + "learning_rate": 3.6655014201441997e-06, + "loss": 2.6941, + "step": 12266 + }, + { + "epoch": 2.68, + "learning_rate": 3.6653921782827183e-06, + "loss": 3.4783, + "step": 12267 + }, + { + "epoch": 2.68, + "learning_rate": 3.6652829364212365e-06, + "loss": 1.6886, + "step": 12268 + }, + { + "epoch": 2.68, + "learning_rate": 3.6651736945597555e-06, + "loss": 1.4918, + "step": 12269 + }, + { + "epoch": 2.68, + "learning_rate": 3.665064452698274e-06, + "loss": 2.6071, + "step": 12270 + }, + { + "epoch": 2.68, + "learning_rate": 3.664955210836793e-06, + "loss": 2.5139, + "step": 12271 + }, + { + "epoch": 2.68, + "learning_rate": 3.6648459689753118e-06, + "loss": 2.4403, + "step": 12272 + }, + { + "epoch": 2.68, + "learning_rate": 3.66473672711383e-06, + "loss": 2.1529, + "step": 12273 + }, + { + "epoch": 2.68, + "learning_rate": 3.664627485252349e-06, + "loss": 2.5104, + "step": 12274 + }, + { + "epoch": 2.68, + "learning_rate": 3.6645182433908676e-06, + "loss": 2.0252, + "step": 12275 + }, + { + "epoch": 2.68, + "learning_rate": 3.6644090015293866e-06, + "loss": 2.9116, + "step": 12276 + }, + { + "epoch": 2.68, + "learning_rate": 3.6642997596679053e-06, + "loss": 2.997, + "step": 12277 + }, + { + "epoch": 2.68, + "learning_rate": 3.6641905178064234e-06, + "loss": 2.3419, + "step": 12278 + }, + { + "epoch": 2.68, + "learning_rate": 3.6640812759449425e-06, + "loss": 3.979, + "step": 12279 + }, + { + "epoch": 2.68, + "learning_rate": 3.663972034083461e-06, + "loss": 2.4938, + "step": 12280 + }, + { + "epoch": 2.68, + "learning_rate": 3.66386279222198e-06, + "loss": 2.8777, + "step": 12281 + }, + { + "epoch": 2.68, + "learning_rate": 3.6637535503604983e-06, + "loss": 2.0812, + "step": 12282 + }, + { + "epoch": 2.68, + "learning_rate": 3.663644308499017e-06, + "loss": 1.6798, + "step": 12283 + }, + { + "epoch": 2.68, + "learning_rate": 3.663535066637536e-06, + "loss": 2.2645, + "step": 12284 + }, + { + "epoch": 2.68, + "learning_rate": 3.6634258247760546e-06, + "loss": 1.6764, + "step": 12285 + }, + { + "epoch": 2.68, + "learning_rate": 3.663316582914573e-06, + "loss": 3.011, + "step": 12286 + }, + { + "epoch": 2.68, + "learning_rate": 3.663207341053092e-06, + "loss": 1.9867, + "step": 12287 + }, + { + "epoch": 2.68, + "learning_rate": 3.6630980991916104e-06, + "loss": 2.958, + "step": 12288 + }, + { + "epoch": 2.68, + "learning_rate": 3.662988857330129e-06, + "loss": 3.1373, + "step": 12289 + }, + { + "epoch": 2.68, + "learning_rate": 3.662879615468648e-06, + "loss": 3.0133, + "step": 12290 + }, + { + "epoch": 2.68, + "learning_rate": 3.6627703736071667e-06, + "loss": 2.7333, + "step": 12291 + }, + { + "epoch": 2.68, + "learning_rate": 3.662661131745685e-06, + "loss": 1.668, + "step": 12292 + }, + { + "epoch": 2.68, + "learning_rate": 3.662551889884204e-06, + "loss": 1.6586, + "step": 12293 + }, + { + "epoch": 2.68, + "learning_rate": 3.6624426480227225e-06, + "loss": 2.3252, + "step": 12294 + }, + { + "epoch": 2.68, + "learning_rate": 3.6623334061612415e-06, + "loss": 2.5678, + "step": 12295 + }, + { + "epoch": 2.68, + "learning_rate": 3.66222416429976e-06, + "loss": 1.7664, + "step": 12296 + }, + { + "epoch": 2.68, + "learning_rate": 3.6621149224382783e-06, + "loss": 3.0171, + "step": 12297 + }, + { + "epoch": 2.68, + "learning_rate": 3.6620056805767974e-06, + "loss": 1.7539, + "step": 12298 + }, + { + "epoch": 2.68, + "learning_rate": 3.661896438715316e-06, + "loss": 2.6783, + "step": 12299 + }, + { + "epoch": 2.68, + "learning_rate": 3.661787196853835e-06, + "loss": 1.9543, + "step": 12300 + }, + { + "epoch": 2.68, + "learning_rate": 3.661677954992353e-06, + "loss": 1.8959, + "step": 12301 + }, + { + "epoch": 2.68, + "learning_rate": 3.661568713130872e-06, + "loss": 1.9269, + "step": 12302 + }, + { + "epoch": 2.69, + "learning_rate": 3.661459471269391e-06, + "loss": 2.0943, + "step": 12303 + }, + { + "epoch": 2.69, + "learning_rate": 3.6613502294079095e-06, + "loss": 0.9691, + "step": 12304 + }, + { + "epoch": 2.69, + "learning_rate": 3.6612409875464285e-06, + "loss": 2.6661, + "step": 12305 + }, + { + "epoch": 2.69, + "learning_rate": 3.6611317456849467e-06, + "loss": 2.6173, + "step": 12306 + }, + { + "epoch": 2.69, + "learning_rate": 3.6610225038234653e-06, + "loss": 1.5516, + "step": 12307 + }, + { + "epoch": 2.69, + "learning_rate": 3.6609132619619843e-06, + "loss": 1.0101, + "step": 12308 + }, + { + "epoch": 2.69, + "learning_rate": 3.660804020100503e-06, + "loss": 2.3668, + "step": 12309 + }, + { + "epoch": 2.69, + "learning_rate": 3.6606947782390215e-06, + "loss": 3.3096, + "step": 12310 + }, + { + "epoch": 2.69, + "learning_rate": 3.66058553637754e-06, + "loss": 2.0076, + "step": 12311 + }, + { + "epoch": 2.69, + "learning_rate": 3.6604762945160588e-06, + "loss": 1.8953, + "step": 12312 + }, + { + "epoch": 2.69, + "learning_rate": 3.660367052654578e-06, + "loss": 1.3494, + "step": 12313 + }, + { + "epoch": 2.69, + "learning_rate": 3.6602578107930964e-06, + "loss": 1.7749, + "step": 12314 + }, + { + "epoch": 2.69, + "learning_rate": 3.6601485689316146e-06, + "loss": 2.0251, + "step": 12315 + }, + { + "epoch": 2.69, + "learning_rate": 3.6600393270701336e-06, + "loss": 1.7806, + "step": 12316 + }, + { + "epoch": 2.69, + "learning_rate": 3.6599300852086522e-06, + "loss": 2.3108, + "step": 12317 + }, + { + "epoch": 2.69, + "learning_rate": 3.659820843347171e-06, + "loss": 2.2418, + "step": 12318 + }, + { + "epoch": 2.69, + "learning_rate": 3.65971160148569e-06, + "loss": 2.2063, + "step": 12319 + }, + { + "epoch": 2.69, + "learning_rate": 3.659602359624208e-06, + "loss": 2.6219, + "step": 12320 + }, + { + "epoch": 2.69, + "learning_rate": 3.6594931177627267e-06, + "loss": 1.8892, + "step": 12321 + }, + { + "epoch": 2.69, + "learning_rate": 3.6593838759012457e-06, + "loss": 2.5162, + "step": 12322 + }, + { + "epoch": 2.69, + "learning_rate": 3.6592746340397643e-06, + "loss": 2.4614, + "step": 12323 + }, + { + "epoch": 2.69, + "learning_rate": 3.6591653921782834e-06, + "loss": 1.2339, + "step": 12324 + }, + { + "epoch": 2.69, + "learning_rate": 3.6590561503168016e-06, + "loss": 2.4224, + "step": 12325 + }, + { + "epoch": 2.69, + "learning_rate": 3.65894690845532e-06, + "loss": 2.4985, + "step": 12326 + }, + { + "epoch": 2.69, + "learning_rate": 3.658837666593839e-06, + "loss": 2.2196, + "step": 12327 + }, + { + "epoch": 2.69, + "learning_rate": 3.658728424732358e-06, + "loss": 2.2194, + "step": 12328 + }, + { + "epoch": 2.69, + "learning_rate": 3.658619182870877e-06, + "loss": 1.2185, + "step": 12329 + }, + { + "epoch": 2.69, + "learning_rate": 3.658509941009395e-06, + "loss": 2.2675, + "step": 12330 + }, + { + "epoch": 2.69, + "learning_rate": 3.6584006991479136e-06, + "loss": 2.0051, + "step": 12331 + }, + { + "epoch": 2.69, + "learning_rate": 3.6582914572864327e-06, + "loss": 1.6049, + "step": 12332 + }, + { + "epoch": 2.69, + "learning_rate": 3.6581822154249513e-06, + "loss": 1.6109, + "step": 12333 + }, + { + "epoch": 2.69, + "learning_rate": 3.6580729735634695e-06, + "loss": 1.807, + "step": 12334 + }, + { + "epoch": 2.69, + "learning_rate": 3.6579637317019885e-06, + "loss": 1.3087, + "step": 12335 + }, + { + "epoch": 2.69, + "learning_rate": 3.657854489840507e-06, + "loss": 2.7729, + "step": 12336 + }, + { + "epoch": 2.69, + "learning_rate": 3.657745247979026e-06, + "loss": 1.4522, + "step": 12337 + }, + { + "epoch": 2.69, + "learning_rate": 3.6576360061175448e-06, + "loss": 2.3713, + "step": 12338 + }, + { + "epoch": 2.69, + "learning_rate": 3.657526764256063e-06, + "loss": 2.8531, + "step": 12339 + }, + { + "epoch": 2.69, + "learning_rate": 3.657417522394582e-06, + "loss": 1.1654, + "step": 12340 + }, + { + "epoch": 2.69, + "learning_rate": 3.6573082805331006e-06, + "loss": 2.0735, + "step": 12341 + }, + { + "epoch": 2.69, + "learning_rate": 3.657199038671619e-06, + "loss": 2.3832, + "step": 12342 + }, + { + "epoch": 2.69, + "learning_rate": 3.6570897968101382e-06, + "loss": 1.9504, + "step": 12343 + }, + { + "epoch": 2.69, + "learning_rate": 3.6569805549486564e-06, + "loss": 2.8395, + "step": 12344 + }, + { + "epoch": 2.69, + "learning_rate": 3.656871313087175e-06, + "loss": 1.7165, + "step": 12345 + }, + { + "epoch": 2.69, + "learning_rate": 3.656762071225694e-06, + "loss": 2.6504, + "step": 12346 + }, + { + "epoch": 2.69, + "learning_rate": 3.6566528293642127e-06, + "loss": 1.5867, + "step": 12347 + }, + { + "epoch": 2.69, + "learning_rate": 3.656543587502731e-06, + "loss": 3.2571, + "step": 12348 + }, + { + "epoch": 2.7, + "learning_rate": 3.65643434564125e-06, + "loss": 2.2089, + "step": 12349 + }, + { + "epoch": 2.7, + "learning_rate": 3.6563251037797685e-06, + "loss": 1.4664, + "step": 12350 + }, + { + "epoch": 2.7, + "learning_rate": 3.6562158619182876e-06, + "loss": 1.9799, + "step": 12351 + }, + { + "epoch": 2.7, + "learning_rate": 3.656106620056806e-06, + "loss": 2.6333, + "step": 12352 + }, + { + "epoch": 2.7, + "learning_rate": 3.6559973781953244e-06, + "loss": 1.7206, + "step": 12353 + }, + { + "epoch": 2.7, + "learning_rate": 3.6558881363338434e-06, + "loss": 1.7806, + "step": 12354 + }, + { + "epoch": 2.7, + "learning_rate": 3.655778894472362e-06, + "loss": 2.1685, + "step": 12355 + }, + { + "epoch": 2.7, + "learning_rate": 3.655669652610881e-06, + "loss": 1.4554, + "step": 12356 + }, + { + "epoch": 2.7, + "learning_rate": 3.6555604107493996e-06, + "loss": 1.2864, + "step": 12357 + }, + { + "epoch": 2.7, + "learning_rate": 3.655451168887918e-06, + "loss": 2.7039, + "step": 12358 + }, + { + "epoch": 2.7, + "learning_rate": 3.655341927026437e-06, + "loss": 2.0864, + "step": 12359 + }, + { + "epoch": 2.7, + "learning_rate": 3.6552326851649555e-06, + "loss": 1.2446, + "step": 12360 + }, + { + "epoch": 2.7, + "learning_rate": 3.6551234433034745e-06, + "loss": 1.5975, + "step": 12361 + }, + { + "epoch": 2.7, + "learning_rate": 3.6550142014419927e-06, + "loss": 1.6981, + "step": 12362 + }, + { + "epoch": 2.7, + "learning_rate": 3.6549049595805113e-06, + "loss": 2.5269, + "step": 12363 + }, + { + "epoch": 2.7, + "learning_rate": 3.6547957177190303e-06, + "loss": 3.0766, + "step": 12364 + }, + { + "epoch": 2.7, + "learning_rate": 3.654686475857549e-06, + "loss": 1.9276, + "step": 12365 + }, + { + "epoch": 2.7, + "learning_rate": 3.654577233996068e-06, + "loss": 2.7185, + "step": 12366 + }, + { + "epoch": 2.7, + "learning_rate": 3.654467992134586e-06, + "loss": 2.0866, + "step": 12367 + }, + { + "epoch": 2.7, + "learning_rate": 3.654358750273105e-06, + "loss": 2.8064, + "step": 12368 + }, + { + "epoch": 2.7, + "learning_rate": 3.654249508411624e-06, + "loss": 2.3607, + "step": 12369 + }, + { + "epoch": 2.7, + "learning_rate": 3.6541402665501424e-06, + "loss": 1.9896, + "step": 12370 + }, + { + "epoch": 2.7, + "learning_rate": 3.654031024688661e-06, + "loss": 2.2703, + "step": 12371 + }, + { + "epoch": 2.7, + "learning_rate": 3.6539217828271797e-06, + "loss": 2.3388, + "step": 12372 + }, + { + "epoch": 2.7, + "learning_rate": 3.6538125409656983e-06, + "loss": 2.2041, + "step": 12373 + }, + { + "epoch": 2.7, + "learning_rate": 3.653703299104217e-06, + "loss": 2.1345, + "step": 12374 + }, + { + "epoch": 2.7, + "learning_rate": 3.653594057242736e-06, + "loss": 3.1652, + "step": 12375 + }, + { + "epoch": 2.7, + "learning_rate": 3.6534848153812545e-06, + "loss": 3.1188, + "step": 12376 + }, + { + "epoch": 2.7, + "learning_rate": 3.6533755735197727e-06, + "loss": 2.1642, + "step": 12377 + }, + { + "epoch": 2.7, + "learning_rate": 3.6532663316582917e-06, + "loss": 1.6485, + "step": 12378 + }, + { + "epoch": 2.7, + "learning_rate": 3.6531570897968104e-06, + "loss": 1.6382, + "step": 12379 + }, + { + "epoch": 2.7, + "learning_rate": 3.6530478479353294e-06, + "loss": 2.5431, + "step": 12380 + }, + { + "epoch": 2.7, + "learning_rate": 3.6529386060738476e-06, + "loss": 2.111, + "step": 12381 + }, + { + "epoch": 2.7, + "learning_rate": 3.652829364212366e-06, + "loss": 2.3912, + "step": 12382 + }, + { + "epoch": 2.7, + "learning_rate": 3.6527201223508852e-06, + "loss": 1.9621, + "step": 12383 + }, + { + "epoch": 2.7, + "learning_rate": 3.652610880489404e-06, + "loss": 1.0482, + "step": 12384 + }, + { + "epoch": 2.7, + "learning_rate": 3.652501638627923e-06, + "loss": 2.4655, + "step": 12385 + }, + { + "epoch": 2.7, + "learning_rate": 3.652392396766441e-06, + "loss": 2.4689, + "step": 12386 + }, + { + "epoch": 2.7, + "learning_rate": 3.6522831549049597e-06, + "loss": 3.25, + "step": 12387 + }, + { + "epoch": 2.7, + "learning_rate": 3.6521739130434787e-06, + "loss": 2.8421, + "step": 12388 + }, + { + "epoch": 2.7, + "learning_rate": 3.6520646711819973e-06, + "loss": 3.2749, + "step": 12389 + }, + { + "epoch": 2.7, + "learning_rate": 3.6519554293205163e-06, + "loss": 2.8297, + "step": 12390 + }, + { + "epoch": 2.7, + "learning_rate": 3.6518461874590345e-06, + "loss": 2.6406, + "step": 12391 + }, + { + "epoch": 2.7, + "learning_rate": 3.651736945597553e-06, + "loss": 3.4258, + "step": 12392 + }, + { + "epoch": 2.7, + "learning_rate": 3.651627703736072e-06, + "loss": 2.7342, + "step": 12393 + }, + { + "epoch": 2.7, + "learning_rate": 3.651518461874591e-06, + "loss": 2.6596, + "step": 12394 + }, + { + "epoch": 2.71, + "learning_rate": 3.651409220013109e-06, + "loss": 1.7584, + "step": 12395 + }, + { + "epoch": 2.71, + "learning_rate": 3.651299978151628e-06, + "loss": 2.0002, + "step": 12396 + }, + { + "epoch": 2.71, + "learning_rate": 3.6511907362901466e-06, + "loss": 2.8329, + "step": 12397 + }, + { + "epoch": 2.71, + "learning_rate": 3.6510814944286652e-06, + "loss": 1.9678, + "step": 12398 + }, + { + "epoch": 2.71, + "learning_rate": 3.6509722525671843e-06, + "loss": 2.8423, + "step": 12399 + }, + { + "epoch": 2.71, + "learning_rate": 3.6508630107057025e-06, + "loss": 3.4736, + "step": 12400 + }, + { + "epoch": 2.71, + "learning_rate": 3.650753768844221e-06, + "loss": 2.0322, + "step": 12401 + }, + { + "epoch": 2.71, + "learning_rate": 3.65064452698274e-06, + "loss": 2.8447, + "step": 12402 + }, + { + "epoch": 2.71, + "learning_rate": 3.6505352851212587e-06, + "loss": 1.8423, + "step": 12403 + }, + { + "epoch": 2.71, + "learning_rate": 3.6504260432597777e-06, + "loss": 1.9502, + "step": 12404 + }, + { + "epoch": 2.71, + "learning_rate": 3.650316801398296e-06, + "loss": 3.5382, + "step": 12405 + }, + { + "epoch": 2.71, + "learning_rate": 3.6502075595368145e-06, + "loss": 1.8909, + "step": 12406 + }, + { + "epoch": 2.71, + "learning_rate": 3.6500983176753336e-06, + "loss": 0.9945, + "step": 12407 + }, + { + "epoch": 2.71, + "learning_rate": 3.649989075813852e-06, + "loss": 1.5569, + "step": 12408 + }, + { + "epoch": 2.71, + "learning_rate": 3.6498798339523712e-06, + "loss": 2.0736, + "step": 12409 + }, + { + "epoch": 2.71, + "learning_rate": 3.6497705920908894e-06, + "loss": 2.5458, + "step": 12410 + }, + { + "epoch": 2.71, + "learning_rate": 3.649661350229408e-06, + "loss": 1.2294, + "step": 12411 + }, + { + "epoch": 2.71, + "learning_rate": 3.649552108367927e-06, + "loss": 3.6773, + "step": 12412 + }, + { + "epoch": 2.71, + "learning_rate": 3.6494428665064457e-06, + "loss": 2.3046, + "step": 12413 + }, + { + "epoch": 2.71, + "learning_rate": 3.649333624644964e-06, + "loss": 2.2, + "step": 12414 + }, + { + "epoch": 2.71, + "learning_rate": 3.649224382783483e-06, + "loss": 1.5293, + "step": 12415 + }, + { + "epoch": 2.71, + "learning_rate": 3.6491151409220015e-06, + "loss": 1.862, + "step": 12416 + }, + { + "epoch": 2.71, + "learning_rate": 3.6490058990605205e-06, + "loss": 2.9078, + "step": 12417 + }, + { + "epoch": 2.71, + "learning_rate": 3.648896657199039e-06, + "loss": 2.1143, + "step": 12418 + }, + { + "epoch": 2.71, + "learning_rate": 3.6487874153375573e-06, + "loss": 2.5281, + "step": 12419 + }, + { + "epoch": 2.71, + "learning_rate": 3.6486781734760764e-06, + "loss": 2.3285, + "step": 12420 + }, + { + "epoch": 2.71, + "learning_rate": 3.648568931614595e-06, + "loss": 1.1823, + "step": 12421 + }, + { + "epoch": 2.71, + "learning_rate": 3.648459689753114e-06, + "loss": 1.574, + "step": 12422 + }, + { + "epoch": 2.71, + "learning_rate": 3.6483504478916326e-06, + "loss": 1.7844, + "step": 12423 + }, + { + "epoch": 2.71, + "learning_rate": 3.648241206030151e-06, + "loss": 0.9702, + "step": 12424 + }, + { + "epoch": 2.71, + "learning_rate": 3.64813196416867e-06, + "loss": 2.0809, + "step": 12425 + }, + { + "epoch": 2.71, + "learning_rate": 3.6480227223071885e-06, + "loss": 2.2685, + "step": 12426 + }, + { + "epoch": 2.71, + "learning_rate": 3.647913480445707e-06, + "loss": 1.838, + "step": 12427 + }, + { + "epoch": 2.71, + "learning_rate": 3.6478042385842257e-06, + "loss": 1.9863, + "step": 12428 + }, + { + "epoch": 2.71, + "learning_rate": 3.6476949967227443e-06, + "loss": 3.5769, + "step": 12429 + }, + { + "epoch": 2.71, + "learning_rate": 3.647585754861263e-06, + "loss": 1.4251, + "step": 12430 + }, + { + "epoch": 2.71, + "learning_rate": 3.647476512999782e-06, + "loss": 2.0188, + "step": 12431 + }, + { + "epoch": 2.71, + "learning_rate": 3.6473672711383006e-06, + "loss": 2.3651, + "step": 12432 + }, + { + "epoch": 2.71, + "learning_rate": 3.6472580292768187e-06, + "loss": 3.1878, + "step": 12433 + }, + { + "epoch": 2.71, + "learning_rate": 3.6471487874153378e-06, + "loss": 2.8369, + "step": 12434 + }, + { + "epoch": 2.71, + "learning_rate": 3.6470395455538564e-06, + "loss": 1.5821, + "step": 12435 + }, + { + "epoch": 2.71, + "learning_rate": 3.6469303036923754e-06, + "loss": 2.8519, + "step": 12436 + }, + { + "epoch": 2.71, + "learning_rate": 3.646821061830894e-06, + "loss": 2.3488, + "step": 12437 + }, + { + "epoch": 2.71, + "learning_rate": 3.6467118199694122e-06, + "loss": 1.843, + "step": 12438 + }, + { + "epoch": 2.71, + "learning_rate": 3.6466025781079313e-06, + "loss": 2.1307, + "step": 12439 + }, + { + "epoch": 2.71, + "learning_rate": 3.64649333624645e-06, + "loss": 3.424, + "step": 12440 + }, + { + "epoch": 2.72, + "learning_rate": 3.646384094384969e-06, + "loss": 1.3529, + "step": 12441 + }, + { + "epoch": 2.72, + "learning_rate": 3.6462748525234875e-06, + "loss": 1.3303, + "step": 12442 + }, + { + "epoch": 2.72, + "learning_rate": 3.6461656106620057e-06, + "loss": 3.018, + "step": 12443 + }, + { + "epoch": 2.72, + "learning_rate": 3.6460563688005247e-06, + "loss": 3.7232, + "step": 12444 + }, + { + "epoch": 2.72, + "learning_rate": 3.6459471269390433e-06, + "loss": 2.086, + "step": 12445 + }, + { + "epoch": 2.72, + "learning_rate": 3.6458378850775624e-06, + "loss": 2.0175, + "step": 12446 + }, + { + "epoch": 2.72, + "learning_rate": 3.6457286432160806e-06, + "loss": 2.5344, + "step": 12447 + }, + { + "epoch": 2.72, + "learning_rate": 3.645619401354599e-06, + "loss": 2.5914, + "step": 12448 + }, + { + "epoch": 2.72, + "learning_rate": 3.645510159493118e-06, + "loss": 3.074, + "step": 12449 + }, + { + "epoch": 2.72, + "learning_rate": 3.645400917631637e-06, + "loss": 1.532, + "step": 12450 + }, + { + "epoch": 2.72, + "learning_rate": 3.6452916757701554e-06, + "loss": 2.4067, + "step": 12451 + }, + { + "epoch": 2.72, + "learning_rate": 3.645182433908674e-06, + "loss": 2.4554, + "step": 12452 + }, + { + "epoch": 2.72, + "learning_rate": 3.6450731920471927e-06, + "loss": 2.2907, + "step": 12453 + }, + { + "epoch": 2.72, + "learning_rate": 3.6449639501857113e-06, + "loss": 0.8005, + "step": 12454 + }, + { + "epoch": 2.72, + "learning_rate": 3.6448547083242303e-06, + "loss": 1.1828, + "step": 12455 + }, + { + "epoch": 2.72, + "learning_rate": 3.644745466462749e-06, + "loss": 3.059, + "step": 12456 + }, + { + "epoch": 2.72, + "learning_rate": 3.6446362246012675e-06, + "loss": 2.1032, + "step": 12457 + }, + { + "epoch": 2.72, + "learning_rate": 3.644526982739786e-06, + "loss": 2.0547, + "step": 12458 + }, + { + "epoch": 2.72, + "learning_rate": 3.6444177408783047e-06, + "loss": 2.9804, + "step": 12459 + }, + { + "epoch": 2.72, + "learning_rate": 3.6443084990168238e-06, + "loss": 2.0655, + "step": 12460 + }, + { + "epoch": 2.72, + "learning_rate": 3.644199257155342e-06, + "loss": 2.4782, + "step": 12461 + }, + { + "epoch": 2.72, + "learning_rate": 3.6440900152938606e-06, + "loss": 2.9615, + "step": 12462 + }, + { + "epoch": 2.72, + "learning_rate": 3.6439807734323796e-06, + "loss": 2.2646, + "step": 12463 + }, + { + "epoch": 2.72, + "learning_rate": 3.6438715315708982e-06, + "loss": 1.603, + "step": 12464 + }, + { + "epoch": 2.72, + "learning_rate": 3.6437622897094173e-06, + "loss": 1.8643, + "step": 12465 + }, + { + "epoch": 2.72, + "learning_rate": 3.6436530478479354e-06, + "loss": 1.4383, + "step": 12466 + }, + { + "epoch": 2.72, + "learning_rate": 3.643543805986454e-06, + "loss": 2.2031, + "step": 12467 + }, + { + "epoch": 2.72, + "learning_rate": 3.643434564124973e-06, + "loss": 2.3955, + "step": 12468 + }, + { + "epoch": 2.72, + "learning_rate": 3.6433253222634917e-06, + "loss": 1.5566, + "step": 12469 + }, + { + "epoch": 2.72, + "learning_rate": 3.6432160804020107e-06, + "loss": 1.161, + "step": 12470 + }, + { + "epoch": 2.72, + "learning_rate": 3.643106838540529e-06, + "loss": 1.8833, + "step": 12471 + }, + { + "epoch": 2.72, + "learning_rate": 3.6429975966790475e-06, + "loss": 1.4209, + "step": 12472 + }, + { + "epoch": 2.72, + "learning_rate": 3.6428883548175666e-06, + "loss": 1.313, + "step": 12473 + }, + { + "epoch": 2.72, + "learning_rate": 3.642779112956085e-06, + "loss": 3.6308, + "step": 12474 + }, + { + "epoch": 2.72, + "learning_rate": 3.6426698710946042e-06, + "loss": 1.7947, + "step": 12475 + }, + { + "epoch": 2.72, + "learning_rate": 3.6425606292331224e-06, + "loss": 1.5025, + "step": 12476 + }, + { + "epoch": 2.72, + "learning_rate": 3.642451387371641e-06, + "loss": 2.0805, + "step": 12477 + }, + { + "epoch": 2.72, + "learning_rate": 3.64234214551016e-06, + "loss": 2.3381, + "step": 12478 + }, + { + "epoch": 2.72, + "learning_rate": 3.6422329036486787e-06, + "loss": 1.4276, + "step": 12479 + }, + { + "epoch": 2.72, + "learning_rate": 3.642123661787197e-06, + "loss": 1.2733, + "step": 12480 + }, + { + "epoch": 2.72, + "learning_rate": 3.642014419925716e-06, + "loss": 2.0465, + "step": 12481 + }, + { + "epoch": 2.72, + "learning_rate": 3.6419051780642345e-06, + "loss": 2.8947, + "step": 12482 + }, + { + "epoch": 2.72, + "learning_rate": 3.641795936202753e-06, + "loss": 2.3985, + "step": 12483 + }, + { + "epoch": 2.72, + "learning_rate": 3.641686694341272e-06, + "loss": 1.485, + "step": 12484 + }, + { + "epoch": 2.72, + "learning_rate": 3.6415774524797903e-06, + "loss": 2.3229, + "step": 12485 + }, + { + "epoch": 2.73, + "learning_rate": 3.641468210618309e-06, + "loss": 3.2685, + "step": 12486 + }, + { + "epoch": 2.73, + "learning_rate": 3.641358968756828e-06, + "loss": 2.4692, + "step": 12487 + }, + { + "epoch": 2.73, + "learning_rate": 3.6412497268953466e-06, + "loss": 2.2415, + "step": 12488 + }, + { + "epoch": 2.73, + "learning_rate": 3.6411404850338656e-06, + "loss": 1.7846, + "step": 12489 + }, + { + "epoch": 2.73, + "learning_rate": 3.641031243172384e-06, + "loss": 2.5017, + "step": 12490 + }, + { + "epoch": 2.73, + "learning_rate": 3.6409220013109024e-06, + "loss": 2.1333, + "step": 12491 + }, + { + "epoch": 2.73, + "learning_rate": 3.6408127594494214e-06, + "loss": 1.4146, + "step": 12492 + }, + { + "epoch": 2.73, + "learning_rate": 3.64070351758794e-06, + "loss": 2.6896, + "step": 12493 + }, + { + "epoch": 2.73, + "learning_rate": 3.6405942757264582e-06, + "loss": 2.1447, + "step": 12494 + }, + { + "epoch": 2.73, + "learning_rate": 3.6404850338649773e-06, + "loss": 2.6744, + "step": 12495 + }, + { + "epoch": 2.73, + "learning_rate": 3.640375792003496e-06, + "loss": 2.0941, + "step": 12496 + }, + { + "epoch": 2.73, + "learning_rate": 3.640266550142015e-06, + "loss": 1.335, + "step": 12497 + }, + { + "epoch": 2.73, + "learning_rate": 3.6401573082805335e-06, + "loss": 2.114, + "step": 12498 + }, + { + "epoch": 2.73, + "learning_rate": 3.6400480664190517e-06, + "loss": 2.6894, + "step": 12499 + }, + { + "epoch": 2.73, + "learning_rate": 3.6399388245575708e-06, + "loss": 2.304, + "step": 12500 + }, + { + "epoch": 2.73, + "learning_rate": 3.6398295826960894e-06, + "loss": 2.5437, + "step": 12501 + }, + { + "epoch": 2.73, + "learning_rate": 3.6397203408346084e-06, + "loss": 1.3459, + "step": 12502 + }, + { + "epoch": 2.73, + "learning_rate": 3.639611098973127e-06, + "loss": 1.882, + "step": 12503 + }, + { + "epoch": 2.73, + "learning_rate": 3.639501857111645e-06, + "loss": 2.3969, + "step": 12504 + }, + { + "epoch": 2.73, + "learning_rate": 3.6393926152501642e-06, + "loss": 1.9883, + "step": 12505 + }, + { + "epoch": 2.73, + "learning_rate": 3.639283373388683e-06, + "loss": 2.466, + "step": 12506 + }, + { + "epoch": 2.73, + "learning_rate": 3.639174131527202e-06, + "loss": 2.8526, + "step": 12507 + }, + { + "epoch": 2.73, + "learning_rate": 3.6390648896657205e-06, + "loss": 3.5323, + "step": 12508 + }, + { + "epoch": 2.73, + "learning_rate": 3.6389556478042387e-06, + "loss": 2.4431, + "step": 12509 + }, + { + "epoch": 2.73, + "learning_rate": 3.6388464059427577e-06, + "loss": 2.3903, + "step": 12510 + }, + { + "epoch": 2.73, + "learning_rate": 3.6387371640812763e-06, + "loss": 1.4004, + "step": 12511 + }, + { + "epoch": 2.73, + "learning_rate": 3.638627922219795e-06, + "loss": 2.0331, + "step": 12512 + }, + { + "epoch": 2.73, + "learning_rate": 3.6385186803583135e-06, + "loss": 3.252, + "step": 12513 + }, + { + "epoch": 2.73, + "learning_rate": 3.638409438496832e-06, + "loss": 2.8144, + "step": 12514 + }, + { + "epoch": 2.73, + "learning_rate": 3.6383001966353508e-06, + "loss": 2.5102, + "step": 12515 + }, + { + "epoch": 2.73, + "learning_rate": 3.63819095477387e-06, + "loss": 3.0703, + "step": 12516 + }, + { + "epoch": 2.73, + "learning_rate": 3.6380817129123884e-06, + "loss": 3.4391, + "step": 12517 + }, + { + "epoch": 2.73, + "learning_rate": 3.6379724710509066e-06, + "loss": 1.5948, + "step": 12518 + }, + { + "epoch": 2.73, + "learning_rate": 3.6378632291894256e-06, + "loss": 3.6512, + "step": 12519 + }, + { + "epoch": 2.73, + "learning_rate": 3.6377539873279442e-06, + "loss": 3.5917, + "step": 12520 + }, + { + "epoch": 2.73, + "learning_rate": 3.6376447454664633e-06, + "loss": 1.826, + "step": 12521 + }, + { + "epoch": 2.73, + "learning_rate": 3.637535503604982e-06, + "loss": 1.438, + "step": 12522 + }, + { + "epoch": 2.73, + "learning_rate": 3.6374262617435e-06, + "loss": 2.5297, + "step": 12523 + }, + { + "epoch": 2.73, + "learning_rate": 3.637317019882019e-06, + "loss": 3.374, + "step": 12524 + }, + { + "epoch": 2.73, + "learning_rate": 3.6372077780205377e-06, + "loss": 1.4058, + "step": 12525 + }, + { + "epoch": 2.73, + "learning_rate": 3.6370985361590568e-06, + "loss": 2.597, + "step": 12526 + }, + { + "epoch": 2.73, + "learning_rate": 3.636989294297575e-06, + "loss": 1.5682, + "step": 12527 + }, + { + "epoch": 2.73, + "learning_rate": 3.6368800524360936e-06, + "loss": 2.8488, + "step": 12528 + }, + { + "epoch": 2.73, + "learning_rate": 3.6367708105746126e-06, + "loss": 1.2536, + "step": 12529 + }, + { + "epoch": 2.73, + "learning_rate": 3.636661568713131e-06, + "loss": 3.5994, + "step": 12530 + }, + { + "epoch": 2.73, + "learning_rate": 3.6365523268516502e-06, + "loss": 1.5767, + "step": 12531 + }, + { + "epoch": 2.74, + "learning_rate": 3.6364430849901684e-06, + "loss": 1.4197, + "step": 12532 + }, + { + "epoch": 2.74, + "learning_rate": 3.636333843128687e-06, + "loss": 1.4196, + "step": 12533 + }, + { + "epoch": 2.74, + "learning_rate": 3.636224601267206e-06, + "loss": 3.1802, + "step": 12534 + }, + { + "epoch": 2.74, + "learning_rate": 3.6361153594057247e-06, + "loss": 1.9189, + "step": 12535 + }, + { + "epoch": 2.74, + "learning_rate": 3.6360061175442433e-06, + "loss": 3.1508, + "step": 12536 + }, + { + "epoch": 2.74, + "learning_rate": 3.635896875682762e-06, + "loss": 3.9321, + "step": 12537 + }, + { + "epoch": 2.74, + "learning_rate": 3.6357876338212805e-06, + "loss": 2.329, + "step": 12538 + }, + { + "epoch": 2.74, + "learning_rate": 3.635678391959799e-06, + "loss": 2.5351, + "step": 12539 + }, + { + "epoch": 2.74, + "learning_rate": 3.635569150098318e-06, + "loss": 2.0389, + "step": 12540 + }, + { + "epoch": 2.74, + "learning_rate": 3.6354599082368368e-06, + "loss": 2.257, + "step": 12541 + }, + { + "epoch": 2.74, + "learning_rate": 3.635350666375355e-06, + "loss": 2.2512, + "step": 12542 + }, + { + "epoch": 2.74, + "learning_rate": 3.635241424513874e-06, + "loss": 2.317, + "step": 12543 + }, + { + "epoch": 2.74, + "learning_rate": 3.6351321826523926e-06, + "loss": 0.9943, + "step": 12544 + }, + { + "epoch": 2.74, + "learning_rate": 3.6350229407909116e-06, + "loss": 1.548, + "step": 12545 + }, + { + "epoch": 2.74, + "learning_rate": 3.63491369892943e-06, + "loss": 2.0654, + "step": 12546 + }, + { + "epoch": 2.74, + "learning_rate": 3.6348044570679484e-06, + "loss": 2.4746, + "step": 12547 + }, + { + "epoch": 2.74, + "learning_rate": 3.6346952152064675e-06, + "loss": 1.9288, + "step": 12548 + }, + { + "epoch": 2.74, + "learning_rate": 3.634585973344986e-06, + "loss": 2.564, + "step": 12549 + }, + { + "epoch": 2.74, + "learning_rate": 3.634476731483505e-06, + "loss": 2.9064, + "step": 12550 + }, + { + "epoch": 2.74, + "learning_rate": 3.6343674896220233e-06, + "loss": 1.74, + "step": 12551 + }, + { + "epoch": 2.74, + "learning_rate": 3.634258247760542e-06, + "loss": 2.6516, + "step": 12552 + }, + { + "epoch": 2.74, + "learning_rate": 3.634149005899061e-06, + "loss": 2.1719, + "step": 12553 + }, + { + "epoch": 2.74, + "learning_rate": 3.6340397640375796e-06, + "loss": 1.9553, + "step": 12554 + }, + { + "epoch": 2.74, + "learning_rate": 3.6339305221760986e-06, + "loss": 1.7294, + "step": 12555 + }, + { + "epoch": 2.74, + "learning_rate": 3.6338212803146168e-06, + "loss": 3.2493, + "step": 12556 + }, + { + "epoch": 2.74, + "learning_rate": 3.6337120384531354e-06, + "loss": 1.6703, + "step": 12557 + }, + { + "epoch": 2.74, + "learning_rate": 3.6336027965916544e-06, + "loss": 1.5599, + "step": 12558 + }, + { + "epoch": 2.74, + "learning_rate": 3.633493554730173e-06, + "loss": 3.4413, + "step": 12559 + }, + { + "epoch": 2.74, + "learning_rate": 3.6333843128686912e-06, + "loss": 3.1467, + "step": 12560 + }, + { + "epoch": 2.74, + "learning_rate": 3.6332750710072103e-06, + "loss": 0.9679, + "step": 12561 + }, + { + "epoch": 2.74, + "learning_rate": 3.633165829145729e-06, + "loss": 1.456, + "step": 12562 + }, + { + "epoch": 2.74, + "learning_rate": 3.633056587284248e-06, + "loss": 2.501, + "step": 12563 + }, + { + "epoch": 2.74, + "learning_rate": 3.6329473454227665e-06, + "loss": 1.592, + "step": 12564 + }, + { + "epoch": 2.74, + "learning_rate": 3.6328381035612847e-06, + "loss": 2.8278, + "step": 12565 + }, + { + "epoch": 2.74, + "learning_rate": 3.6327288616998037e-06, + "loss": 2.6137, + "step": 12566 + }, + { + "epoch": 2.74, + "learning_rate": 3.6326196198383224e-06, + "loss": 1.8939, + "step": 12567 + }, + { + "epoch": 2.74, + "learning_rate": 3.632510377976841e-06, + "loss": 1.3838, + "step": 12568 + }, + { + "epoch": 2.74, + "learning_rate": 3.63240113611536e-06, + "loss": 2.1364, + "step": 12569 + }, + { + "epoch": 2.74, + "learning_rate": 3.632291894253878e-06, + "loss": 1.9923, + "step": 12570 + }, + { + "epoch": 2.74, + "learning_rate": 3.632182652392397e-06, + "loss": 2.2058, + "step": 12571 + }, + { + "epoch": 2.74, + "learning_rate": 3.632073410530916e-06, + "loss": 1.6732, + "step": 12572 + }, + { + "epoch": 2.74, + "learning_rate": 3.6319641686694344e-06, + "loss": 1.6506, + "step": 12573 + }, + { + "epoch": 2.74, + "learning_rate": 3.6318549268079535e-06, + "loss": 2.5945, + "step": 12574 + }, + { + "epoch": 2.74, + "learning_rate": 3.6317456849464717e-06, + "loss": 1.803, + "step": 12575 + }, + { + "epoch": 2.74, + "learning_rate": 3.6316364430849903e-06, + "loss": 2.4456, + "step": 12576 + }, + { + "epoch": 2.74, + "learning_rate": 3.6315272012235093e-06, + "loss": 1.3825, + "step": 12577 + }, + { + "epoch": 2.75, + "learning_rate": 3.631417959362028e-06, + "loss": 1.561, + "step": 12578 + }, + { + "epoch": 2.75, + "learning_rate": 3.631308717500546e-06, + "loss": 2.0709, + "step": 12579 + }, + { + "epoch": 2.75, + "learning_rate": 3.631199475639065e-06, + "loss": 2.045, + "step": 12580 + }, + { + "epoch": 2.75, + "learning_rate": 3.6310902337775838e-06, + "loss": 2.2666, + "step": 12581 + }, + { + "epoch": 2.75, + "learning_rate": 3.6309809919161028e-06, + "loss": 2.6613, + "step": 12582 + }, + { + "epoch": 2.75, + "learning_rate": 3.6308717500546214e-06, + "loss": 3.1159, + "step": 12583 + }, + { + "epoch": 2.75, + "learning_rate": 3.6307625081931396e-06, + "loss": 2.7804, + "step": 12584 + }, + { + "epoch": 2.75, + "learning_rate": 3.6306532663316586e-06, + "loss": 2.6535, + "step": 12585 + }, + { + "epoch": 2.75, + "learning_rate": 3.6305440244701772e-06, + "loss": 2.4321, + "step": 12586 + }, + { + "epoch": 2.75, + "learning_rate": 3.6304347826086963e-06, + "loss": 2.3949, + "step": 12587 + }, + { + "epoch": 2.75, + "learning_rate": 3.630325540747215e-06, + "loss": 2.3522, + "step": 12588 + }, + { + "epoch": 2.75, + "learning_rate": 3.630216298885733e-06, + "loss": 2.7378, + "step": 12589 + }, + { + "epoch": 2.75, + "learning_rate": 3.630107057024252e-06, + "loss": 2.8294, + "step": 12590 + }, + { + "epoch": 2.75, + "learning_rate": 3.6299978151627707e-06, + "loss": 1.4631, + "step": 12591 + }, + { + "epoch": 2.75, + "learning_rate": 3.6298885733012893e-06, + "loss": 1.3071, + "step": 12592 + }, + { + "epoch": 2.75, + "learning_rate": 3.629779331439808e-06, + "loss": 3.9015, + "step": 12593 + }, + { + "epoch": 2.75, + "learning_rate": 3.6296700895783265e-06, + "loss": 2.2577, + "step": 12594 + }, + { + "epoch": 2.75, + "learning_rate": 3.629560847716845e-06, + "loss": 1.3199, + "step": 12595 + }, + { + "epoch": 2.75, + "learning_rate": 3.629451605855364e-06, + "loss": 2.0253, + "step": 12596 + }, + { + "epoch": 2.75, + "learning_rate": 3.629342363993883e-06, + "loss": 1.3118, + "step": 12597 + }, + { + "epoch": 2.75, + "learning_rate": 3.6292331221324014e-06, + "loss": 2.0007, + "step": 12598 + }, + { + "epoch": 2.75, + "learning_rate": 3.62912388027092e-06, + "loss": 1.4835, + "step": 12599 + }, + { + "epoch": 2.75, + "learning_rate": 3.6290146384094386e-06, + "loss": 1.8809, + "step": 12600 + }, + { + "epoch": 2.75, + "learning_rate": 3.6289053965479577e-06, + "loss": 2.1927, + "step": 12601 + }, + { + "epoch": 2.75, + "learning_rate": 3.6287961546864763e-06, + "loss": 1.373, + "step": 12602 + }, + { + "epoch": 2.75, + "learning_rate": 3.6286869128249945e-06, + "loss": 2.8347, + "step": 12603 + }, + { + "epoch": 2.75, + "learning_rate": 3.6285776709635135e-06, + "loss": 2.1748, + "step": 12604 + }, + { + "epoch": 2.75, + "learning_rate": 3.628468429102032e-06, + "loss": 1.5978, + "step": 12605 + }, + { + "epoch": 2.75, + "learning_rate": 3.628359187240551e-06, + "loss": 3.5668, + "step": 12606 + }, + { + "epoch": 2.75, + "learning_rate": 3.6282499453790698e-06, + "loss": 2.1609, + "step": 12607 + }, + { + "epoch": 2.75, + "learning_rate": 3.628140703517588e-06, + "loss": 1.932, + "step": 12608 + }, + { + "epoch": 2.75, + "learning_rate": 3.628031461656107e-06, + "loss": 3.5486, + "step": 12609 + }, + { + "epoch": 2.75, + "learning_rate": 3.6279222197946256e-06, + "loss": 1.4713, + "step": 12610 + }, + { + "epoch": 2.75, + "learning_rate": 3.6278129779331446e-06, + "loss": 2.6132, + "step": 12611 + }, + { + "epoch": 2.75, + "learning_rate": 3.627703736071663e-06, + "loss": 2.3377, + "step": 12612 + }, + { + "epoch": 2.75, + "learning_rate": 3.6275944942101814e-06, + "loss": 1.3983, + "step": 12613 + }, + { + "epoch": 2.75, + "learning_rate": 3.6274852523487005e-06, + "loss": 2.8677, + "step": 12614 + }, + { + "epoch": 2.75, + "learning_rate": 3.627376010487219e-06, + "loss": 1.8262, + "step": 12615 + }, + { + "epoch": 2.75, + "learning_rate": 3.627266768625738e-06, + "loss": 2.2161, + "step": 12616 + }, + { + "epoch": 2.75, + "learning_rate": 3.6271575267642563e-06, + "loss": 1.8988, + "step": 12617 + }, + { + "epoch": 2.75, + "learning_rate": 3.627048284902775e-06, + "loss": 1.829, + "step": 12618 + }, + { + "epoch": 2.75, + "learning_rate": 3.626939043041294e-06, + "loss": 3.1335, + "step": 12619 + }, + { + "epoch": 2.75, + "learning_rate": 3.6268298011798125e-06, + "loss": 1.8281, + "step": 12620 + }, + { + "epoch": 2.75, + "learning_rate": 3.626720559318331e-06, + "loss": 1.9588, + "step": 12621 + }, + { + "epoch": 2.75, + "learning_rate": 3.6266113174568498e-06, + "loss": 1.8001, + "step": 12622 + }, + { + "epoch": 2.75, + "learning_rate": 3.6265020755953684e-06, + "loss": 3.1934, + "step": 12623 + }, + { + "epoch": 2.76, + "learning_rate": 3.626392833733887e-06, + "loss": 1.1932, + "step": 12624 + }, + { + "epoch": 2.76, + "learning_rate": 3.626283591872406e-06, + "loss": 2.467, + "step": 12625 + }, + { + "epoch": 2.76, + "learning_rate": 3.6261743500109242e-06, + "loss": 2.2978, + "step": 12626 + }, + { + "epoch": 2.76, + "learning_rate": 3.626065108149443e-06, + "loss": 1.9414, + "step": 12627 + }, + { + "epoch": 2.76, + "learning_rate": 3.625955866287962e-06, + "loss": 1.3291, + "step": 12628 + }, + { + "epoch": 2.76, + "learning_rate": 3.6258466244264805e-06, + "loss": 2.9702, + "step": 12629 + }, + { + "epoch": 2.76, + "learning_rate": 3.6257373825649995e-06, + "loss": 1.083, + "step": 12630 + }, + { + "epoch": 2.76, + "learning_rate": 3.6256281407035177e-06, + "loss": 1.3578, + "step": 12631 + }, + { + "epoch": 2.76, + "learning_rate": 3.6255188988420363e-06, + "loss": 2.762, + "step": 12632 + }, + { + "epoch": 2.76, + "learning_rate": 3.6254096569805553e-06, + "loss": 2.3497, + "step": 12633 + }, + { + "epoch": 2.76, + "learning_rate": 3.625300415119074e-06, + "loss": 2.7762, + "step": 12634 + }, + { + "epoch": 2.76, + "learning_rate": 3.625191173257593e-06, + "loss": 2.9503, + "step": 12635 + }, + { + "epoch": 2.76, + "learning_rate": 3.625081931396111e-06, + "loss": 2.2823, + "step": 12636 + }, + { + "epoch": 2.76, + "learning_rate": 3.6249726895346298e-06, + "loss": 3.2146, + "step": 12637 + }, + { + "epoch": 2.76, + "learning_rate": 3.624863447673149e-06, + "loss": 1.4484, + "step": 12638 + }, + { + "epoch": 2.76, + "learning_rate": 3.6247542058116674e-06, + "loss": 0.8053, + "step": 12639 + }, + { + "epoch": 2.76, + "learning_rate": 3.6246449639501865e-06, + "loss": 1.8516, + "step": 12640 + }, + { + "epoch": 2.76, + "learning_rate": 3.6245357220887046e-06, + "loss": 1.1834, + "step": 12641 + }, + { + "epoch": 2.76, + "learning_rate": 3.6244264802272233e-06, + "loss": 1.9652, + "step": 12642 + }, + { + "epoch": 2.76, + "learning_rate": 3.6243172383657423e-06, + "loss": 1.4565, + "step": 12643 + }, + { + "epoch": 2.76, + "learning_rate": 3.624207996504261e-06, + "loss": 2.4734, + "step": 12644 + }, + { + "epoch": 2.76, + "learning_rate": 3.624098754642779e-06, + "loss": 2.9764, + "step": 12645 + }, + { + "epoch": 2.76, + "learning_rate": 3.623989512781298e-06, + "loss": 1.6729, + "step": 12646 + }, + { + "epoch": 2.76, + "learning_rate": 3.6238802709198167e-06, + "loss": 2.0579, + "step": 12647 + }, + { + "epoch": 2.76, + "learning_rate": 3.6237710290583353e-06, + "loss": 1.5252, + "step": 12648 + }, + { + "epoch": 2.76, + "learning_rate": 3.6236617871968544e-06, + "loss": 2.572, + "step": 12649 + }, + { + "epoch": 2.76, + "learning_rate": 3.6235525453353726e-06, + "loss": 1.84, + "step": 12650 + }, + { + "epoch": 2.76, + "learning_rate": 3.6234433034738916e-06, + "loss": 1.3278, + "step": 12651 + }, + { + "epoch": 2.76, + "learning_rate": 3.6233340616124102e-06, + "loss": 2.8461, + "step": 12652 + }, + { + "epoch": 2.76, + "learning_rate": 3.623224819750929e-06, + "loss": 2.6664, + "step": 12653 + }, + { + "epoch": 2.76, + "learning_rate": 3.623115577889448e-06, + "loss": 2.9805, + "step": 12654 + }, + { + "epoch": 2.76, + "learning_rate": 3.623006336027966e-06, + "loss": 1.8418, + "step": 12655 + }, + { + "epoch": 2.76, + "learning_rate": 3.6228970941664847e-06, + "loss": 2.0265, + "step": 12656 + }, + { + "epoch": 2.76, + "learning_rate": 3.6227878523050037e-06, + "loss": 2.6976, + "step": 12657 + }, + { + "epoch": 2.76, + "learning_rate": 3.6226786104435223e-06, + "loss": 1.4287, + "step": 12658 + }, + { + "epoch": 2.76, + "learning_rate": 3.6225693685820405e-06, + "loss": 0.9643, + "step": 12659 + }, + { + "epoch": 2.76, + "learning_rate": 3.6224601267205595e-06, + "loss": 3.2164, + "step": 12660 + }, + { + "epoch": 2.76, + "learning_rate": 3.622350884859078e-06, + "loss": 2.6298, + "step": 12661 + }, + { + "epoch": 2.76, + "learning_rate": 3.622241642997597e-06, + "loss": 3.2196, + "step": 12662 + }, + { + "epoch": 2.76, + "learning_rate": 3.6221324011361158e-06, + "loss": 1.2181, + "step": 12663 + }, + { + "epoch": 2.76, + "learning_rate": 3.622023159274634e-06, + "loss": 2.1565, + "step": 12664 + }, + { + "epoch": 2.76, + "learning_rate": 3.621913917413153e-06, + "loss": 2.0307, + "step": 12665 + }, + { + "epoch": 2.76, + "learning_rate": 3.6218046755516716e-06, + "loss": 1.9814, + "step": 12666 + }, + { + "epoch": 2.76, + "learning_rate": 3.6216954336901907e-06, + "loss": 1.9734, + "step": 12667 + }, + { + "epoch": 2.76, + "learning_rate": 3.6215861918287093e-06, + "loss": 1.3481, + "step": 12668 + }, + { + "epoch": 2.76, + "learning_rate": 3.6214769499672274e-06, + "loss": 1.2999, + "step": 12669 + }, + { + "epoch": 2.77, + "learning_rate": 3.6213677081057465e-06, + "loss": 2.8377, + "step": 12670 + }, + { + "epoch": 2.77, + "learning_rate": 3.621258466244265e-06, + "loss": 1.2823, + "step": 12671 + }, + { + "epoch": 2.77, + "learning_rate": 3.621149224382784e-06, + "loss": 2.0433, + "step": 12672 + }, + { + "epoch": 2.77, + "learning_rate": 3.6210399825213027e-06, + "loss": 3.1634, + "step": 12673 + }, + { + "epoch": 2.77, + "learning_rate": 3.620930740659821e-06, + "loss": 2.0594, + "step": 12674 + }, + { + "epoch": 2.77, + "learning_rate": 3.62082149879834e-06, + "loss": 1.9226, + "step": 12675 + }, + { + "epoch": 2.77, + "learning_rate": 3.6207122569368586e-06, + "loss": 1.4075, + "step": 12676 + }, + { + "epoch": 2.77, + "learning_rate": 3.620603015075377e-06, + "loss": 2.5766, + "step": 12677 + }, + { + "epoch": 2.77, + "learning_rate": 3.620493773213896e-06, + "loss": 3.3647, + "step": 12678 + }, + { + "epoch": 2.77, + "learning_rate": 3.6203845313524144e-06, + "loss": 0.8871, + "step": 12679 + }, + { + "epoch": 2.77, + "learning_rate": 3.620275289490933e-06, + "loss": 1.857, + "step": 12680 + }, + { + "epoch": 2.77, + "learning_rate": 3.620166047629452e-06, + "loss": 1.7242, + "step": 12681 + }, + { + "epoch": 2.77, + "learning_rate": 3.6200568057679707e-06, + "loss": 1.6841, + "step": 12682 + }, + { + "epoch": 2.77, + "learning_rate": 3.619947563906489e-06, + "loss": 1.8823, + "step": 12683 + }, + { + "epoch": 2.77, + "learning_rate": 3.619838322045008e-06, + "loss": 1.6598, + "step": 12684 + }, + { + "epoch": 2.77, + "learning_rate": 3.6197290801835265e-06, + "loss": 2.2161, + "step": 12685 + }, + { + "epoch": 2.77, + "learning_rate": 3.6196198383220455e-06, + "loss": 2.5662, + "step": 12686 + }, + { + "epoch": 2.77, + "learning_rate": 3.619510596460564e-06, + "loss": 1.9304, + "step": 12687 + }, + { + "epoch": 2.77, + "learning_rate": 3.6194013545990823e-06, + "loss": 2.6414, + "step": 12688 + }, + { + "epoch": 2.77, + "learning_rate": 3.6192921127376014e-06, + "loss": 2.2298, + "step": 12689 + }, + { + "epoch": 2.77, + "learning_rate": 3.61918287087612e-06, + "loss": 1.5603, + "step": 12690 + }, + { + "epoch": 2.77, + "learning_rate": 3.619073629014639e-06, + "loss": 2.5713, + "step": 12691 + }, + { + "epoch": 2.77, + "learning_rate": 3.618964387153157e-06, + "loss": 1.7671, + "step": 12692 + }, + { + "epoch": 2.77, + "learning_rate": 3.618855145291676e-06, + "loss": 1.3137, + "step": 12693 + }, + { + "epoch": 2.77, + "learning_rate": 3.618745903430195e-06, + "loss": 1.6323, + "step": 12694 + }, + { + "epoch": 2.77, + "learning_rate": 3.6186366615687135e-06, + "loss": 2.499, + "step": 12695 + }, + { + "epoch": 2.77, + "learning_rate": 3.6185274197072325e-06, + "loss": 1.6288, + "step": 12696 + }, + { + "epoch": 2.77, + "learning_rate": 3.6184181778457507e-06, + "loss": 2.5776, + "step": 12697 + }, + { + "epoch": 2.77, + "learning_rate": 3.6183089359842693e-06, + "loss": 1.0732, + "step": 12698 + }, + { + "epoch": 2.77, + "learning_rate": 3.6181996941227883e-06, + "loss": 1.53, + "step": 12699 + }, + { + "epoch": 2.77, + "learning_rate": 3.618090452261307e-06, + "loss": 2.4278, + "step": 12700 + }, + { + "epoch": 2.77, + "learning_rate": 3.6179812103998255e-06, + "loss": 2.0929, + "step": 12701 + }, + { + "epoch": 2.77, + "learning_rate": 3.617871968538344e-06, + "loss": 1.324, + "step": 12702 + }, + { + "epoch": 2.77, + "learning_rate": 3.6177627266768628e-06, + "loss": 2.2553, + "step": 12703 + }, + { + "epoch": 2.77, + "learning_rate": 3.617653484815382e-06, + "loss": 2.8643, + "step": 12704 + }, + { + "epoch": 2.77, + "learning_rate": 3.6175442429539004e-06, + "loss": 2.8625, + "step": 12705 + }, + { + "epoch": 2.77, + "learning_rate": 3.617435001092419e-06, + "loss": 1.9977, + "step": 12706 + }, + { + "epoch": 2.77, + "learning_rate": 3.6173257592309376e-06, + "loss": 2.7114, + "step": 12707 + }, + { + "epoch": 2.77, + "learning_rate": 3.6172165173694562e-06, + "loss": 1.1216, + "step": 12708 + }, + { + "epoch": 2.77, + "learning_rate": 3.617107275507975e-06, + "loss": 2.4071, + "step": 12709 + }, + { + "epoch": 2.77, + "learning_rate": 3.616998033646494e-06, + "loss": 2.2249, + "step": 12710 + }, + { + "epoch": 2.77, + "learning_rate": 3.616888791785012e-06, + "loss": 1.5283, + "step": 12711 + }, + { + "epoch": 2.77, + "learning_rate": 3.6167795499235307e-06, + "loss": 2.9935, + "step": 12712 + }, + { + "epoch": 2.77, + "learning_rate": 3.6166703080620497e-06, + "loss": 1.869, + "step": 12713 + }, + { + "epoch": 2.77, + "learning_rate": 3.6165610662005683e-06, + "loss": 2.2718, + "step": 12714 + }, + { + "epoch": 2.77, + "learning_rate": 3.6164518243390874e-06, + "loss": 2.2185, + "step": 12715 + }, + { + "epoch": 2.78, + "learning_rate": 3.6163425824776056e-06, + "loss": 0.9995, + "step": 12716 + }, + { + "epoch": 2.78, + "learning_rate": 3.616233340616124e-06, + "loss": 4.4586, + "step": 12717 + }, + { + "epoch": 2.78, + "learning_rate": 3.616124098754643e-06, + "loss": 1.8104, + "step": 12718 + }, + { + "epoch": 2.78, + "learning_rate": 3.616014856893162e-06, + "loss": 2.7052, + "step": 12719 + }, + { + "epoch": 2.78, + "learning_rate": 3.615905615031681e-06, + "loss": 2.6688, + "step": 12720 + }, + { + "epoch": 2.78, + "learning_rate": 3.615796373170199e-06, + "loss": 2.1156, + "step": 12721 + }, + { + "epoch": 2.78, + "learning_rate": 3.6156871313087176e-06, + "loss": 3.6928, + "step": 12722 + }, + { + "epoch": 2.78, + "learning_rate": 3.6155778894472367e-06, + "loss": 2.4932, + "step": 12723 + }, + { + "epoch": 2.78, + "learning_rate": 3.6154686475857553e-06, + "loss": 2.4515, + "step": 12724 + }, + { + "epoch": 2.78, + "learning_rate": 3.6153594057242735e-06, + "loss": 2.3465, + "step": 12725 + }, + { + "epoch": 2.78, + "learning_rate": 3.6152501638627925e-06, + "loss": 1.8793, + "step": 12726 + }, + { + "epoch": 2.78, + "learning_rate": 3.615140922001311e-06, + "loss": 3.3294, + "step": 12727 + }, + { + "epoch": 2.78, + "learning_rate": 3.61503168013983e-06, + "loss": 2.0346, + "step": 12728 + }, + { + "epoch": 2.78, + "learning_rate": 3.6149224382783488e-06, + "loss": 2.733, + "step": 12729 + }, + { + "epoch": 2.78, + "learning_rate": 3.614813196416867e-06, + "loss": 1.8066, + "step": 12730 + }, + { + "epoch": 2.78, + "learning_rate": 3.614703954555386e-06, + "loss": 2.7933, + "step": 12731 + }, + { + "epoch": 2.78, + "learning_rate": 3.6145947126939046e-06, + "loss": 2.7317, + "step": 12732 + }, + { + "epoch": 2.78, + "learning_rate": 3.614485470832423e-06, + "loss": 3.1546, + "step": 12733 + }, + { + "epoch": 2.78, + "learning_rate": 3.6143762289709422e-06, + "loss": 1.9242, + "step": 12734 + }, + { + "epoch": 2.78, + "learning_rate": 3.6142669871094604e-06, + "loss": 3.0639, + "step": 12735 + }, + { + "epoch": 2.78, + "learning_rate": 3.614157745247979e-06, + "loss": 1.295, + "step": 12736 + }, + { + "epoch": 2.78, + "learning_rate": 3.614048503386498e-06, + "loss": 2.094, + "step": 12737 + }, + { + "epoch": 2.78, + "learning_rate": 3.6139392615250167e-06, + "loss": 1.9485, + "step": 12738 + }, + { + "epoch": 2.78, + "learning_rate": 3.6138300196635357e-06, + "loss": 1.9149, + "step": 12739 + }, + { + "epoch": 2.78, + "learning_rate": 3.613720777802054e-06, + "loss": 2.2903, + "step": 12740 + }, + { + "epoch": 2.78, + "learning_rate": 3.6136115359405725e-06, + "loss": 2.4436, + "step": 12741 + }, + { + "epoch": 2.78, + "learning_rate": 3.6135022940790916e-06, + "loss": 1.9151, + "step": 12742 + }, + { + "epoch": 2.78, + "learning_rate": 3.61339305221761e-06, + "loss": 1.874, + "step": 12743 + }, + { + "epoch": 2.78, + "learning_rate": 3.6132838103561284e-06, + "loss": 2.2361, + "step": 12744 + }, + { + "epoch": 2.78, + "learning_rate": 3.6131745684946474e-06, + "loss": 1.9332, + "step": 12745 + }, + { + "epoch": 2.78, + "learning_rate": 3.613065326633166e-06, + "loss": 3.9272, + "step": 12746 + }, + { + "epoch": 2.78, + "learning_rate": 3.612956084771685e-06, + "loss": 2.5278, + "step": 12747 + }, + { + "epoch": 2.78, + "learning_rate": 3.6128468429102036e-06, + "loss": 0.9621, + "step": 12748 + }, + { + "epoch": 2.78, + "learning_rate": 3.612737601048722e-06, + "loss": 2.1611, + "step": 12749 + }, + { + "epoch": 2.78, + "learning_rate": 3.612628359187241e-06, + "loss": 1.8757, + "step": 12750 + }, + { + "epoch": 2.78, + "learning_rate": 3.6125191173257595e-06, + "loss": 2.9082, + "step": 12751 + }, + { + "epoch": 2.78, + "learning_rate": 3.6124098754642785e-06, + "loss": 1.39, + "step": 12752 + }, + { + "epoch": 2.78, + "learning_rate": 3.612300633602797e-06, + "loss": 3.4804, + "step": 12753 + }, + { + "epoch": 2.78, + "learning_rate": 3.6121913917413153e-06, + "loss": 1.6537, + "step": 12754 + }, + { + "epoch": 2.78, + "learning_rate": 3.6120821498798343e-06, + "loss": 2.3348, + "step": 12755 + }, + { + "epoch": 2.78, + "learning_rate": 3.611972908018353e-06, + "loss": 2.5524, + "step": 12756 + }, + { + "epoch": 2.78, + "learning_rate": 3.611863666156872e-06, + "loss": 2.4793, + "step": 12757 + }, + { + "epoch": 2.78, + "learning_rate": 3.61175442429539e-06, + "loss": 2.7752, + "step": 12758 + }, + { + "epoch": 2.78, + "learning_rate": 3.611645182433909e-06, + "loss": 2.7098, + "step": 12759 + }, + { + "epoch": 2.78, + "learning_rate": 3.611535940572428e-06, + "loss": 1.7713, + "step": 12760 + }, + { + "epoch": 2.79, + "learning_rate": 3.6114266987109464e-06, + "loss": 2.5499, + "step": 12761 + }, + { + "epoch": 2.79, + "learning_rate": 3.611317456849465e-06, + "loss": 1.8371, + "step": 12762 + }, + { + "epoch": 2.79, + "learning_rate": 3.6112082149879837e-06, + "loss": 1.3549, + "step": 12763 + }, + { + "epoch": 2.79, + "learning_rate": 3.6110989731265023e-06, + "loss": 1.1693, + "step": 12764 + }, + { + "epoch": 2.79, + "learning_rate": 3.610989731265021e-06, + "loss": 2.3508, + "step": 12765 + }, + { + "epoch": 2.79, + "learning_rate": 3.61088048940354e-06, + "loss": 2.7425, + "step": 12766 + }, + { + "epoch": 2.79, + "learning_rate": 3.6107712475420585e-06, + "loss": 2.2677, + "step": 12767 + }, + { + "epoch": 2.79, + "learning_rate": 3.6106620056805767e-06, + "loss": 2.4816, + "step": 12768 + }, + { + "epoch": 2.79, + "learning_rate": 3.6105527638190957e-06, + "loss": 2.9851, + "step": 12769 + }, + { + "epoch": 2.79, + "learning_rate": 3.6104435219576144e-06, + "loss": 1.1831, + "step": 12770 + }, + { + "epoch": 2.79, + "learning_rate": 3.6103342800961334e-06, + "loss": 1.8735, + "step": 12771 + }, + { + "epoch": 2.79, + "learning_rate": 3.610225038234652e-06, + "loss": 1.2316, + "step": 12772 + }, + { + "epoch": 2.79, + "learning_rate": 3.61011579637317e-06, + "loss": 2.5251, + "step": 12773 + }, + { + "epoch": 2.79, + "learning_rate": 3.6100065545116892e-06, + "loss": 2.4069, + "step": 12774 + }, + { + "epoch": 2.79, + "learning_rate": 3.609897312650208e-06, + "loss": 2.9693, + "step": 12775 + }, + { + "epoch": 2.79, + "learning_rate": 3.609788070788727e-06, + "loss": 1.7476, + "step": 12776 + }, + { + "epoch": 2.79, + "learning_rate": 3.609678828927245e-06, + "loss": 2.023, + "step": 12777 + }, + { + "epoch": 2.79, + "learning_rate": 3.6095695870657637e-06, + "loss": 1.59, + "step": 12778 + }, + { + "epoch": 2.79, + "learning_rate": 3.6094603452042827e-06, + "loss": 2.2844, + "step": 12779 + }, + { + "epoch": 2.79, + "learning_rate": 3.6093511033428013e-06, + "loss": 2.6082, + "step": 12780 + }, + { + "epoch": 2.79, + "learning_rate": 3.6092418614813204e-06, + "loss": 1.7976, + "step": 12781 + }, + { + "epoch": 2.79, + "learning_rate": 3.6091326196198385e-06, + "loss": 2.8234, + "step": 12782 + }, + { + "epoch": 2.79, + "learning_rate": 3.609023377758357e-06, + "loss": 2.1156, + "step": 12783 + }, + { + "epoch": 2.79, + "learning_rate": 3.608914135896876e-06, + "loss": 2.8902, + "step": 12784 + }, + { + "epoch": 2.79, + "learning_rate": 3.608804894035395e-06, + "loss": 2.6926, + "step": 12785 + }, + { + "epoch": 2.79, + "learning_rate": 3.6086956521739134e-06, + "loss": 1.1638, + "step": 12786 + }, + { + "epoch": 2.79, + "learning_rate": 3.608586410312432e-06, + "loss": 1.193, + "step": 12787 + }, + { + "epoch": 2.79, + "learning_rate": 3.6084771684509506e-06, + "loss": 0.9912, + "step": 12788 + }, + { + "epoch": 2.79, + "learning_rate": 3.6083679265894692e-06, + "loss": 0.7038, + "step": 12789 + }, + { + "epoch": 2.79, + "learning_rate": 3.6082586847279883e-06, + "loss": 2.1893, + "step": 12790 + }, + { + "epoch": 2.79, + "learning_rate": 3.6081494428665065e-06, + "loss": 2.2479, + "step": 12791 + }, + { + "epoch": 2.79, + "learning_rate": 3.608040201005025e-06, + "loss": 2.0543, + "step": 12792 + }, + { + "epoch": 2.79, + "learning_rate": 3.607930959143544e-06, + "loss": 2.5216, + "step": 12793 + }, + { + "epoch": 2.79, + "learning_rate": 3.6078217172820627e-06, + "loss": 1.9755, + "step": 12794 + }, + { + "epoch": 2.79, + "learning_rate": 3.6077124754205818e-06, + "loss": 2.5656, + "step": 12795 + }, + { + "epoch": 2.79, + "learning_rate": 3.6076032335591e-06, + "loss": 1.6433, + "step": 12796 + }, + { + "epoch": 2.79, + "learning_rate": 3.6074939916976185e-06, + "loss": 1.5803, + "step": 12797 + }, + { + "epoch": 2.79, + "learning_rate": 3.6073847498361376e-06, + "loss": 1.1434, + "step": 12798 + }, + { + "epoch": 2.79, + "learning_rate": 3.607275507974656e-06, + "loss": 1.9901, + "step": 12799 + }, + { + "epoch": 2.79, + "learning_rate": 3.6071662661131752e-06, + "loss": 2.484, + "step": 12800 + }, + { + "epoch": 2.79, + "learning_rate": 3.6070570242516934e-06, + "loss": 1.7581, + "step": 12801 + }, + { + "epoch": 2.79, + "learning_rate": 3.606947782390212e-06, + "loss": 1.5054, + "step": 12802 + }, + { + "epoch": 2.79, + "learning_rate": 3.606838540528731e-06, + "loss": 1.1127, + "step": 12803 + }, + { + "epoch": 2.79, + "learning_rate": 3.6067292986672497e-06, + "loss": 3.0674, + "step": 12804 + }, + { + "epoch": 2.79, + "learning_rate": 3.606620056805768e-06, + "loss": 2.1254, + "step": 12805 + }, + { + "epoch": 2.79, + "learning_rate": 3.606510814944287e-06, + "loss": 1.5481, + "step": 12806 + }, + { + "epoch": 2.8, + "learning_rate": 3.6064015730828055e-06, + "loss": 2.6203, + "step": 12807 + }, + { + "epoch": 2.8, + "learning_rate": 3.6062923312213245e-06, + "loss": 2.2777, + "step": 12808 + }, + { + "epoch": 2.8, + "learning_rate": 3.606183089359843e-06, + "loss": 3.1623, + "step": 12809 + }, + { + "epoch": 2.8, + "learning_rate": 3.6060738474983613e-06, + "loss": 1.9963, + "step": 12810 + }, + { + "epoch": 2.8, + "learning_rate": 3.6059646056368804e-06, + "loss": 1.6796, + "step": 12811 + }, + { + "epoch": 2.8, + "learning_rate": 3.605855363775399e-06, + "loss": 1.7633, + "step": 12812 + }, + { + "epoch": 2.8, + "learning_rate": 3.605746121913918e-06, + "loss": 1.83, + "step": 12813 + }, + { + "epoch": 2.8, + "learning_rate": 3.6056368800524366e-06, + "loss": 1.9635, + "step": 12814 + }, + { + "epoch": 2.8, + "learning_rate": 3.605527638190955e-06, + "loss": 2.0862, + "step": 12815 + }, + { + "epoch": 2.8, + "learning_rate": 3.605418396329474e-06, + "loss": 2.3919, + "step": 12816 + }, + { + "epoch": 2.8, + "learning_rate": 3.6053091544679925e-06, + "loss": 2.513, + "step": 12817 + }, + { + "epoch": 2.8, + "learning_rate": 3.605199912606511e-06, + "loss": 2.5812, + "step": 12818 + }, + { + "epoch": 2.8, + "learning_rate": 3.60509067074503e-06, + "loss": 1.9091, + "step": 12819 + }, + { + "epoch": 2.8, + "learning_rate": 3.6049814288835483e-06, + "loss": 1.8106, + "step": 12820 + }, + { + "epoch": 2.8, + "learning_rate": 3.604872187022067e-06, + "loss": 3.4425, + "step": 12821 + }, + { + "epoch": 2.8, + "learning_rate": 3.604762945160586e-06, + "loss": 2.2311, + "step": 12822 + }, + { + "epoch": 2.8, + "learning_rate": 3.6046537032991046e-06, + "loss": 1.6271, + "step": 12823 + }, + { + "epoch": 2.8, + "learning_rate": 3.6045444614376227e-06, + "loss": 3.1372, + "step": 12824 + }, + { + "epoch": 2.8, + "learning_rate": 3.6044352195761418e-06, + "loss": 2.6788, + "step": 12825 + }, + { + "epoch": 2.8, + "learning_rate": 3.6043259777146604e-06, + "loss": 2.439, + "step": 12826 + }, + { + "epoch": 2.8, + "learning_rate": 3.6042167358531794e-06, + "loss": 3.7917, + "step": 12827 + }, + { + "epoch": 2.8, + "learning_rate": 3.604107493991698e-06, + "loss": 4.3899, + "step": 12828 + }, + { + "epoch": 2.8, + "learning_rate": 3.6039982521302162e-06, + "loss": 3.4578, + "step": 12829 + }, + { + "epoch": 2.8, + "learning_rate": 3.6038890102687353e-06, + "loss": 1.7652, + "step": 12830 + }, + { + "epoch": 2.8, + "learning_rate": 3.603779768407254e-06, + "loss": 2.0054, + "step": 12831 + }, + { + "epoch": 2.8, + "learning_rate": 3.603670526545773e-06, + "loss": 1.1516, + "step": 12832 + }, + { + "epoch": 2.8, + "learning_rate": 3.6035612846842915e-06, + "loss": 2.0734, + "step": 12833 + }, + { + "epoch": 2.8, + "learning_rate": 3.6034520428228097e-06, + "loss": 1.5498, + "step": 12834 + }, + { + "epoch": 2.8, + "learning_rate": 3.6033428009613287e-06, + "loss": 1.098, + "step": 12835 + }, + { + "epoch": 2.8, + "learning_rate": 3.6032335590998473e-06, + "loss": 2.869, + "step": 12836 + }, + { + "epoch": 2.8, + "learning_rate": 3.6031243172383664e-06, + "loss": 2.1254, + "step": 12837 + }, + { + "epoch": 2.8, + "learning_rate": 3.6030150753768846e-06, + "loss": 2.55, + "step": 12838 + }, + { + "epoch": 2.8, + "learning_rate": 3.602905833515403e-06, + "loss": 2.4939, + "step": 12839 + }, + { + "epoch": 2.8, + "learning_rate": 3.602796591653922e-06, + "loss": 2.1681, + "step": 12840 + }, + { + "epoch": 2.8, + "learning_rate": 3.602687349792441e-06, + "loss": 2.8638, + "step": 12841 + }, + { + "epoch": 2.8, + "learning_rate": 3.6025781079309594e-06, + "loss": 2.4586, + "step": 12842 + }, + { + "epoch": 2.8, + "learning_rate": 3.602468866069478e-06, + "loss": 2.7839, + "step": 12843 + }, + { + "epoch": 2.8, + "learning_rate": 3.6023596242079967e-06, + "loss": 2.4722, + "step": 12844 + }, + { + "epoch": 2.8, + "learning_rate": 3.6022503823465153e-06, + "loss": 1.9965, + "step": 12845 + }, + { + "epoch": 2.8, + "learning_rate": 3.6021411404850343e-06, + "loss": 1.4944, + "step": 12846 + }, + { + "epoch": 2.8, + "learning_rate": 3.602031898623553e-06, + "loss": 2.5332, + "step": 12847 + }, + { + "epoch": 2.8, + "learning_rate": 3.6019226567620715e-06, + "loss": 2.0157, + "step": 12848 + }, + { + "epoch": 2.8, + "learning_rate": 3.60181341490059e-06, + "loss": 2.1164, + "step": 12849 + }, + { + "epoch": 2.8, + "learning_rate": 3.6017041730391087e-06, + "loss": 2.1079, + "step": 12850 + }, + { + "epoch": 2.8, + "learning_rate": 3.6015949311776278e-06, + "loss": 2.1286, + "step": 12851 + }, + { + "epoch": 2.8, + "learning_rate": 3.6014856893161464e-06, + "loss": 1.8297, + "step": 12852 + }, + { + "epoch": 2.81, + "learning_rate": 3.6013764474546646e-06, + "loss": 2.5239, + "step": 12853 + }, + { + "epoch": 2.81, + "learning_rate": 3.6012672055931836e-06, + "loss": 1.835, + "step": 12854 + }, + { + "epoch": 2.81, + "learning_rate": 3.6011579637317022e-06, + "loss": 2.7819, + "step": 12855 + }, + { + "epoch": 2.81, + "learning_rate": 3.6010487218702213e-06, + "loss": 3.0089, + "step": 12856 + }, + { + "epoch": 2.81, + "learning_rate": 3.6009394800087394e-06, + "loss": 2.6285, + "step": 12857 + }, + { + "epoch": 2.81, + "learning_rate": 3.600830238147258e-06, + "loss": 2.751, + "step": 12858 + }, + { + "epoch": 2.81, + "learning_rate": 3.600720996285777e-06, + "loss": 1.4119, + "step": 12859 + }, + { + "epoch": 2.81, + "learning_rate": 3.6006117544242957e-06, + "loss": 3.2884, + "step": 12860 + }, + { + "epoch": 2.81, + "learning_rate": 3.6005025125628147e-06, + "loss": 2.2106, + "step": 12861 + }, + { + "epoch": 2.81, + "learning_rate": 3.600393270701333e-06, + "loss": 2.2847, + "step": 12862 + }, + { + "epoch": 2.81, + "learning_rate": 3.6002840288398515e-06, + "loss": 2.5993, + "step": 12863 + }, + { + "epoch": 2.81, + "learning_rate": 3.6001747869783706e-06, + "loss": 1.6057, + "step": 12864 + }, + { + "epoch": 2.81, + "learning_rate": 3.600065545116889e-06, + "loss": 1.4056, + "step": 12865 + }, + { + "epoch": 2.81, + "learning_rate": 3.5999563032554082e-06, + "loss": 1.1642, + "step": 12866 + }, + { + "epoch": 2.81, + "learning_rate": 3.5998470613939264e-06, + "loss": 1.35, + "step": 12867 + }, + { + "epoch": 2.81, + "learning_rate": 3.599737819532445e-06, + "loss": 2.1421, + "step": 12868 + }, + { + "epoch": 2.81, + "learning_rate": 3.599628577670964e-06, + "loss": 1.6597, + "step": 12869 + }, + { + "epoch": 2.81, + "learning_rate": 3.5995193358094827e-06, + "loss": 3.0696, + "step": 12870 + }, + { + "epoch": 2.81, + "learning_rate": 3.599410093948001e-06, + "loss": 2.4424, + "step": 12871 + }, + { + "epoch": 2.81, + "learning_rate": 3.59930085208652e-06, + "loss": 1.5632, + "step": 12872 + }, + { + "epoch": 2.81, + "learning_rate": 3.5991916102250385e-06, + "loss": 1.2245, + "step": 12873 + }, + { + "epoch": 2.81, + "learning_rate": 3.599082368363557e-06, + "loss": 2.3161, + "step": 12874 + }, + { + "epoch": 2.81, + "learning_rate": 3.598973126502076e-06, + "loss": 3.249, + "step": 12875 + }, + { + "epoch": 2.81, + "learning_rate": 3.5988638846405943e-06, + "loss": 2.0578, + "step": 12876 + }, + { + "epoch": 2.81, + "learning_rate": 3.598754642779113e-06, + "loss": 2.9562, + "step": 12877 + }, + { + "epoch": 2.81, + "learning_rate": 3.598645400917632e-06, + "loss": 1.5853, + "step": 12878 + }, + { + "epoch": 2.81, + "learning_rate": 3.5985361590561506e-06, + "loss": 2.8071, + "step": 12879 + }, + { + "epoch": 2.81, + "learning_rate": 3.5984269171946696e-06, + "loss": 1.678, + "step": 12880 + }, + { + "epoch": 2.81, + "learning_rate": 3.598317675333188e-06, + "loss": 2.9214, + "step": 12881 + }, + { + "epoch": 2.81, + "learning_rate": 3.5982084334717064e-06, + "loss": 4.0957, + "step": 12882 + }, + { + "epoch": 2.81, + "learning_rate": 3.5980991916102254e-06, + "loss": 2.5149, + "step": 12883 + }, + { + "epoch": 2.81, + "learning_rate": 3.597989949748744e-06, + "loss": 3.0481, + "step": 12884 + }, + { + "epoch": 2.81, + "learning_rate": 3.597880707887263e-06, + "loss": 1.1061, + "step": 12885 + }, + { + "epoch": 2.81, + "learning_rate": 3.5977714660257813e-06, + "loss": 2.2702, + "step": 12886 + }, + { + "epoch": 2.81, + "learning_rate": 3.5976622241643e-06, + "loss": 3.404, + "step": 12887 + }, + { + "epoch": 2.81, + "learning_rate": 3.597552982302819e-06, + "loss": 2.8149, + "step": 12888 + }, + { + "epoch": 2.81, + "learning_rate": 3.5974437404413375e-06, + "loss": 1.9286, + "step": 12889 + }, + { + "epoch": 2.81, + "learning_rate": 3.5973344985798557e-06, + "loss": 2.7575, + "step": 12890 + }, + { + "epoch": 2.81, + "learning_rate": 3.5972252567183748e-06, + "loss": 2.3835, + "step": 12891 + }, + { + "epoch": 2.81, + "learning_rate": 3.5971160148568934e-06, + "loss": 2.2679, + "step": 12892 + }, + { + "epoch": 2.81, + "learning_rate": 3.5970067729954124e-06, + "loss": 2.2619, + "step": 12893 + }, + { + "epoch": 2.81, + "learning_rate": 3.596897531133931e-06, + "loss": 2.419, + "step": 12894 + }, + { + "epoch": 2.81, + "learning_rate": 3.596788289272449e-06, + "loss": 2.492, + "step": 12895 + }, + { + "epoch": 2.81, + "learning_rate": 3.5966790474109682e-06, + "loss": 1.3311, + "step": 12896 + }, + { + "epoch": 2.81, + "learning_rate": 3.596569805549487e-06, + "loss": 1.3944, + "step": 12897 + }, + { + "epoch": 2.81, + "learning_rate": 3.5964605636880055e-06, + "loss": 1.2308, + "step": 12898 + }, + { + "epoch": 2.82, + "learning_rate": 3.5963513218265245e-06, + "loss": 2.2454, + "step": 12899 + }, + { + "epoch": 2.82, + "learning_rate": 3.5962420799650427e-06, + "loss": 1.9985, + "step": 12900 + }, + { + "epoch": 2.82, + "learning_rate": 3.5961328381035617e-06, + "loss": 1.2386, + "step": 12901 + }, + { + "epoch": 2.82, + "learning_rate": 3.5960235962420803e-06, + "loss": 0.8501, + "step": 12902 + }, + { + "epoch": 2.82, + "learning_rate": 3.595914354380599e-06, + "loss": 0.7984, + "step": 12903 + }, + { + "epoch": 2.82, + "learning_rate": 3.5958051125191175e-06, + "loss": 2.3399, + "step": 12904 + }, + { + "epoch": 2.82, + "learning_rate": 3.595695870657636e-06, + "loss": 1.6238, + "step": 12905 + }, + { + "epoch": 2.82, + "learning_rate": 3.5955866287961548e-06, + "loss": 2.1794, + "step": 12906 + }, + { + "epoch": 2.82, + "learning_rate": 3.595477386934674e-06, + "loss": 1.767, + "step": 12907 + }, + { + "epoch": 2.82, + "learning_rate": 3.5953681450731924e-06, + "loss": 0.7084, + "step": 12908 + }, + { + "epoch": 2.82, + "learning_rate": 3.5952589032117106e-06, + "loss": 1.7304, + "step": 12909 + }, + { + "epoch": 2.82, + "learning_rate": 3.5951496613502296e-06, + "loss": 1.8618, + "step": 12910 + }, + { + "epoch": 2.82, + "learning_rate": 3.5950404194887482e-06, + "loss": 2.1894, + "step": 12911 + }, + { + "epoch": 2.82, + "learning_rate": 3.5949311776272673e-06, + "loss": 1.8708, + "step": 12912 + }, + { + "epoch": 2.82, + "learning_rate": 3.594821935765786e-06, + "loss": 1.7002, + "step": 12913 + }, + { + "epoch": 2.82, + "learning_rate": 3.594712693904304e-06, + "loss": 1.7156, + "step": 12914 + }, + { + "epoch": 2.82, + "learning_rate": 3.594603452042823e-06, + "loss": 2.0569, + "step": 12915 + }, + { + "epoch": 2.82, + "learning_rate": 3.5944942101813417e-06, + "loss": 2.398, + "step": 12916 + }, + { + "epoch": 2.82, + "learning_rate": 3.5943849683198608e-06, + "loss": 1.7174, + "step": 12917 + }, + { + "epoch": 2.82, + "learning_rate": 3.5942757264583794e-06, + "loss": 2.5367, + "step": 12918 + }, + { + "epoch": 2.82, + "learning_rate": 3.5941664845968976e-06, + "loss": 2.4605, + "step": 12919 + }, + { + "epoch": 2.82, + "learning_rate": 3.5940572427354166e-06, + "loss": 1.5946, + "step": 12920 + }, + { + "epoch": 2.82, + "learning_rate": 3.593948000873935e-06, + "loss": 1.4479, + "step": 12921 + }, + { + "epoch": 2.82, + "learning_rate": 3.5938387590124542e-06, + "loss": 1.8255, + "step": 12922 + }, + { + "epoch": 2.82, + "learning_rate": 3.5937295171509724e-06, + "loss": 2.4517, + "step": 12923 + }, + { + "epoch": 2.82, + "learning_rate": 3.593620275289491e-06, + "loss": 2.0189, + "step": 12924 + }, + { + "epoch": 2.82, + "learning_rate": 3.59351103342801e-06, + "loss": 1.8108, + "step": 12925 + }, + { + "epoch": 2.82, + "learning_rate": 3.5934017915665287e-06, + "loss": 2.0497, + "step": 12926 + }, + { + "epoch": 2.82, + "learning_rate": 3.5932925497050473e-06, + "loss": 2.1865, + "step": 12927 + }, + { + "epoch": 2.82, + "learning_rate": 3.593183307843566e-06, + "loss": 3.5628, + "step": 12928 + }, + { + "epoch": 2.82, + "learning_rate": 3.5930740659820845e-06, + "loss": 2.3047, + "step": 12929 + }, + { + "epoch": 2.82, + "learning_rate": 3.592964824120603e-06, + "loss": 2.4468, + "step": 12930 + }, + { + "epoch": 2.82, + "learning_rate": 3.592855582259122e-06, + "loss": 0.5874, + "step": 12931 + }, + { + "epoch": 2.82, + "learning_rate": 3.5927463403976408e-06, + "loss": 2.4201, + "step": 12932 + }, + { + "epoch": 2.82, + "learning_rate": 3.592637098536159e-06, + "loss": 2.9443, + "step": 12933 + }, + { + "epoch": 2.82, + "learning_rate": 3.592527856674678e-06, + "loss": 1.6623, + "step": 12934 + }, + { + "epoch": 2.82, + "learning_rate": 3.5924186148131966e-06, + "loss": 2.1907, + "step": 12935 + }, + { + "epoch": 2.82, + "learning_rate": 3.5923093729517156e-06, + "loss": 1.398, + "step": 12936 + }, + { + "epoch": 2.82, + "learning_rate": 3.592200131090234e-06, + "loss": 2.8779, + "step": 12937 + }, + { + "epoch": 2.82, + "learning_rate": 3.5920908892287524e-06, + "loss": 1.8355, + "step": 12938 + }, + { + "epoch": 2.82, + "learning_rate": 3.5919816473672715e-06, + "loss": 2.7487, + "step": 12939 + }, + { + "epoch": 2.82, + "learning_rate": 3.59187240550579e-06, + "loss": 1.4981, + "step": 12940 + }, + { + "epoch": 2.82, + "learning_rate": 3.591763163644309e-06, + "loss": 2.6975, + "step": 12941 + }, + { + "epoch": 2.82, + "learning_rate": 3.5916539217828273e-06, + "loss": 1.5582, + "step": 12942 + }, + { + "epoch": 2.82, + "learning_rate": 3.591544679921346e-06, + "loss": 3.6408, + "step": 12943 + }, + { + "epoch": 2.82, + "learning_rate": 3.591435438059865e-06, + "loss": 1.9713, + "step": 12944 + }, + { + "epoch": 2.83, + "learning_rate": 3.5913261961983836e-06, + "loss": 1.5771, + "step": 12945 + }, + { + "epoch": 2.83, + "learning_rate": 3.5912169543369026e-06, + "loss": 2.0722, + "step": 12946 + }, + { + "epoch": 2.83, + "learning_rate": 3.5911077124754208e-06, + "loss": 2.0206, + "step": 12947 + }, + { + "epoch": 2.83, + "learning_rate": 3.5909984706139394e-06, + "loss": 1.9651, + "step": 12948 + }, + { + "epoch": 2.83, + "learning_rate": 3.5908892287524584e-06, + "loss": 2.8838, + "step": 12949 + }, + { + "epoch": 2.83, + "learning_rate": 3.590779986890977e-06, + "loss": 1.6356, + "step": 12950 + }, + { + "epoch": 2.83, + "learning_rate": 3.5906707450294957e-06, + "loss": 3.0202, + "step": 12951 + }, + { + "epoch": 2.83, + "learning_rate": 3.5905615031680143e-06, + "loss": 2.5247, + "step": 12952 + }, + { + "epoch": 2.83, + "learning_rate": 3.590452261306533e-06, + "loss": 2.4024, + "step": 12953 + }, + { + "epoch": 2.83, + "learning_rate": 3.590343019445052e-06, + "loss": 2.3207, + "step": 12954 + }, + { + "epoch": 2.83, + "learning_rate": 3.5902337775835705e-06, + "loss": 2.9195, + "step": 12955 + }, + { + "epoch": 2.83, + "learning_rate": 3.5901245357220887e-06, + "loss": 2.1833, + "step": 12956 + }, + { + "epoch": 2.83, + "learning_rate": 3.5900152938606077e-06, + "loss": 2.017, + "step": 12957 + }, + { + "epoch": 2.83, + "learning_rate": 3.5899060519991264e-06, + "loss": 2.7429, + "step": 12958 + }, + { + "epoch": 2.83, + "learning_rate": 3.589796810137645e-06, + "loss": 1.6256, + "step": 12959 + }, + { + "epoch": 2.83, + "learning_rate": 3.589687568276164e-06, + "loss": 1.5003, + "step": 12960 + }, + { + "epoch": 2.83, + "learning_rate": 3.589578326414682e-06, + "loss": 1.9802, + "step": 12961 + }, + { + "epoch": 2.83, + "learning_rate": 3.589469084553201e-06, + "loss": 1.8106, + "step": 12962 + }, + { + "epoch": 2.83, + "learning_rate": 3.58935984269172e-06, + "loss": 2.0921, + "step": 12963 + }, + { + "epoch": 2.83, + "learning_rate": 3.5892506008302384e-06, + "loss": 3.1534, + "step": 12964 + }, + { + "epoch": 2.83, + "learning_rate": 3.5891413589687575e-06, + "loss": 3.0891, + "step": 12965 + }, + { + "epoch": 2.83, + "learning_rate": 3.5890321171072757e-06, + "loss": 3.1786, + "step": 12966 + }, + { + "epoch": 2.83, + "learning_rate": 3.5889228752457943e-06, + "loss": 2.806, + "step": 12967 + }, + { + "epoch": 2.83, + "learning_rate": 3.5888136333843133e-06, + "loss": 2.679, + "step": 12968 + }, + { + "epoch": 2.83, + "learning_rate": 3.588704391522832e-06, + "loss": 1.1788, + "step": 12969 + }, + { + "epoch": 2.83, + "learning_rate": 3.58859514966135e-06, + "loss": 1.287, + "step": 12970 + }, + { + "epoch": 2.83, + "learning_rate": 3.588485907799869e-06, + "loss": 1.9377, + "step": 12971 + }, + { + "epoch": 2.83, + "learning_rate": 3.5883766659383878e-06, + "loss": 2.7572, + "step": 12972 + }, + { + "epoch": 2.83, + "learning_rate": 3.588267424076907e-06, + "loss": 2.4929, + "step": 12973 + }, + { + "epoch": 2.83, + "learning_rate": 3.5881581822154254e-06, + "loss": 1.2981, + "step": 12974 + }, + { + "epoch": 2.83, + "learning_rate": 3.5880489403539436e-06, + "loss": 2.4454, + "step": 12975 + }, + { + "epoch": 2.83, + "learning_rate": 3.5879396984924626e-06, + "loss": 2.4384, + "step": 12976 + }, + { + "epoch": 2.83, + "learning_rate": 3.5878304566309812e-06, + "loss": 1.4548, + "step": 12977 + }, + { + "epoch": 2.83, + "learning_rate": 3.5877212147695003e-06, + "loss": 2.1626, + "step": 12978 + }, + { + "epoch": 2.83, + "learning_rate": 3.587611972908019e-06, + "loss": 0.9619, + "step": 12979 + }, + { + "epoch": 2.83, + "learning_rate": 3.587502731046537e-06, + "loss": 1.5891, + "step": 12980 + }, + { + "epoch": 2.83, + "learning_rate": 3.587393489185056e-06, + "loss": 3.6744, + "step": 12981 + }, + { + "epoch": 2.83, + "learning_rate": 3.5872842473235747e-06, + "loss": 1.398, + "step": 12982 + }, + { + "epoch": 2.83, + "learning_rate": 3.5871750054620933e-06, + "loss": 1.397, + "step": 12983 + }, + { + "epoch": 2.83, + "learning_rate": 3.5870657636006124e-06, + "loss": 2.6803, + "step": 12984 + }, + { + "epoch": 2.83, + "learning_rate": 3.5869565217391305e-06, + "loss": 2.2623, + "step": 12985 + }, + { + "epoch": 2.83, + "learning_rate": 3.586847279877649e-06, + "loss": 1.3439, + "step": 12986 + }, + { + "epoch": 2.83, + "learning_rate": 3.586738038016168e-06, + "loss": 1.5289, + "step": 12987 + }, + { + "epoch": 2.83, + "learning_rate": 3.586628796154687e-06, + "loss": 2.4856, + "step": 12988 + }, + { + "epoch": 2.83, + "learning_rate": 3.586519554293205e-06, + "loss": 2.9043, + "step": 12989 + }, + { + "epoch": 2.84, + "learning_rate": 3.586410312431724e-06, + "loss": 1.9253, + "step": 12990 + }, + { + "epoch": 2.84, + "learning_rate": 3.5863010705702426e-06, + "loss": 2.0808, + "step": 12991 + }, + { + "epoch": 2.84, + "learning_rate": 3.5861918287087617e-06, + "loss": 1.547, + "step": 12992 + }, + { + "epoch": 2.84, + "learning_rate": 3.5860825868472803e-06, + "loss": 2.5508, + "step": 12993 + }, + { + "epoch": 2.84, + "learning_rate": 3.5859733449857985e-06, + "loss": 1.4693, + "step": 12994 + }, + { + "epoch": 2.84, + "learning_rate": 3.5858641031243175e-06, + "loss": 2.1658, + "step": 12995 + }, + { + "epoch": 2.84, + "learning_rate": 3.585754861262836e-06, + "loss": 1.2551, + "step": 12996 + }, + { + "epoch": 2.84, + "learning_rate": 3.585645619401355e-06, + "loss": 2.4048, + "step": 12997 + }, + { + "epoch": 2.84, + "learning_rate": 3.5855363775398738e-06, + "loss": 3.3978, + "step": 12998 + }, + { + "epoch": 2.84, + "learning_rate": 3.585427135678392e-06, + "loss": 2.6934, + "step": 12999 + }, + { + "epoch": 2.84, + "learning_rate": 3.585317893816911e-06, + "loss": 1.6584, + "step": 13000 + }, + { + "epoch": 2.84, + "learning_rate": 3.5852086519554296e-06, + "loss": 3.5488, + "step": 13001 + }, + { + "epoch": 2.84, + "learning_rate": 3.5850994100939486e-06, + "loss": 2.4702, + "step": 13002 + }, + { + "epoch": 2.84, + "learning_rate": 3.584990168232467e-06, + "loss": 1.803, + "step": 13003 + }, + { + "epoch": 2.84, + "learning_rate": 3.5848809263709854e-06, + "loss": 2.0636, + "step": 13004 + }, + { + "epoch": 2.84, + "learning_rate": 3.5847716845095045e-06, + "loss": 2.9456, + "step": 13005 + }, + { + "epoch": 2.84, + "learning_rate": 3.584662442648023e-06, + "loss": 1.915, + "step": 13006 + }, + { + "epoch": 2.84, + "learning_rate": 3.584553200786542e-06, + "loss": 1.1295, + "step": 13007 + }, + { + "epoch": 2.84, + "learning_rate": 3.5844439589250603e-06, + "loss": 2.1962, + "step": 13008 + }, + { + "epoch": 2.84, + "learning_rate": 3.584334717063579e-06, + "loss": 1.9831, + "step": 13009 + }, + { + "epoch": 2.84, + "learning_rate": 3.584225475202098e-06, + "loss": 3.0142, + "step": 13010 + }, + { + "epoch": 2.84, + "learning_rate": 3.5841162333406165e-06, + "loss": 2.1714, + "step": 13011 + }, + { + "epoch": 2.84, + "learning_rate": 3.584006991479135e-06, + "loss": 1.7603, + "step": 13012 + }, + { + "epoch": 2.84, + "learning_rate": 3.5838977496176538e-06, + "loss": 2.4054, + "step": 13013 + }, + { + "epoch": 2.84, + "learning_rate": 3.5837885077561724e-06, + "loss": 1.2972, + "step": 13014 + }, + { + "epoch": 2.84, + "learning_rate": 3.583679265894691e-06, + "loss": 1.8346, + "step": 13015 + }, + { + "epoch": 2.84, + "learning_rate": 3.58357002403321e-06, + "loss": 2.8355, + "step": 13016 + }, + { + "epoch": 2.84, + "learning_rate": 3.5834607821717286e-06, + "loss": 2.9576, + "step": 13017 + }, + { + "epoch": 2.84, + "learning_rate": 3.583351540310247e-06, + "loss": 2.0342, + "step": 13018 + }, + { + "epoch": 2.84, + "learning_rate": 3.583242298448766e-06, + "loss": 1.9283, + "step": 13019 + }, + { + "epoch": 2.84, + "learning_rate": 3.5831330565872845e-06, + "loss": 1.6652, + "step": 13020 + }, + { + "epoch": 2.84, + "learning_rate": 3.5830238147258035e-06, + "loss": 2.6754, + "step": 13021 + }, + { + "epoch": 2.84, + "learning_rate": 3.5829145728643217e-06, + "loss": 1.3987, + "step": 13022 + }, + { + "epoch": 2.84, + "learning_rate": 3.5828053310028403e-06, + "loss": 2.3009, + "step": 13023 + }, + { + "epoch": 2.84, + "learning_rate": 3.5826960891413593e-06, + "loss": 3.2694, + "step": 13024 + }, + { + "epoch": 2.84, + "learning_rate": 3.582586847279878e-06, + "loss": 2.4123, + "step": 13025 + }, + { + "epoch": 2.84, + "learning_rate": 3.582477605418397e-06, + "loss": 2.697, + "step": 13026 + }, + { + "epoch": 2.84, + "learning_rate": 3.582368363556915e-06, + "loss": 2.7486, + "step": 13027 + }, + { + "epoch": 2.84, + "learning_rate": 3.5822591216954338e-06, + "loss": 2.8588, + "step": 13028 + }, + { + "epoch": 2.84, + "learning_rate": 3.582149879833953e-06, + "loss": 2.9261, + "step": 13029 + }, + { + "epoch": 2.84, + "learning_rate": 3.5820406379724714e-06, + "loss": 0.7532, + "step": 13030 + }, + { + "epoch": 2.84, + "learning_rate": 3.5819313961109905e-06, + "loss": 2.2014, + "step": 13031 + }, + { + "epoch": 2.84, + "learning_rate": 3.5818221542495086e-06, + "loss": 1.2075, + "step": 13032 + }, + { + "epoch": 2.84, + "learning_rate": 3.5817129123880273e-06, + "loss": 2.2304, + "step": 13033 + }, + { + "epoch": 2.84, + "learning_rate": 3.5816036705265463e-06, + "loss": 2.0737, + "step": 13034 + }, + { + "epoch": 2.84, + "learning_rate": 3.581494428665065e-06, + "loss": 1.6083, + "step": 13035 + }, + { + "epoch": 2.85, + "learning_rate": 3.581385186803583e-06, + "loss": 2.9899, + "step": 13036 + }, + { + "epoch": 2.85, + "learning_rate": 3.581275944942102e-06, + "loss": 2.4537, + "step": 13037 + }, + { + "epoch": 2.85, + "learning_rate": 3.5811667030806207e-06, + "loss": 1.3968, + "step": 13038 + }, + { + "epoch": 2.85, + "learning_rate": 3.5810574612191393e-06, + "loss": 2.4027, + "step": 13039 + }, + { + "epoch": 2.85, + "learning_rate": 3.5809482193576584e-06, + "loss": 2.7841, + "step": 13040 + }, + { + "epoch": 2.85, + "learning_rate": 3.5808389774961766e-06, + "loss": 2.7028, + "step": 13041 + }, + { + "epoch": 2.85, + "learning_rate": 3.580729735634695e-06, + "loss": 1.5973, + "step": 13042 + }, + { + "epoch": 2.85, + "learning_rate": 3.5806204937732142e-06, + "loss": 2.3092, + "step": 13043 + }, + { + "epoch": 2.85, + "learning_rate": 3.580511251911733e-06, + "loss": 1.7505, + "step": 13044 + }, + { + "epoch": 2.85, + "learning_rate": 3.580402010050252e-06, + "loss": 1.1844, + "step": 13045 + }, + { + "epoch": 2.85, + "learning_rate": 3.58029276818877e-06, + "loss": 2.4383, + "step": 13046 + }, + { + "epoch": 2.85, + "learning_rate": 3.5801835263272887e-06, + "loss": 3.2507, + "step": 13047 + }, + { + "epoch": 2.85, + "learning_rate": 3.5800742844658077e-06, + "loss": 1.6142, + "step": 13048 + }, + { + "epoch": 2.85, + "learning_rate": 3.5799650426043263e-06, + "loss": 2.06, + "step": 13049 + }, + { + "epoch": 2.85, + "learning_rate": 3.5798558007428453e-06, + "loss": 2.0174, + "step": 13050 + }, + { + "epoch": 2.85, + "learning_rate": 3.5797465588813635e-06, + "loss": 2.2179, + "step": 13051 + }, + { + "epoch": 2.85, + "learning_rate": 3.579637317019882e-06, + "loss": 1.5101, + "step": 13052 + }, + { + "epoch": 2.85, + "learning_rate": 3.579528075158401e-06, + "loss": 0.8836, + "step": 13053 + }, + { + "epoch": 2.85, + "learning_rate": 3.5794188332969198e-06, + "loss": 1.5915, + "step": 13054 + }, + { + "epoch": 2.85, + "learning_rate": 3.579309591435438e-06, + "loss": 1.7551, + "step": 13055 + }, + { + "epoch": 2.85, + "learning_rate": 3.579200349573957e-06, + "loss": 1.8816, + "step": 13056 + }, + { + "epoch": 2.85, + "learning_rate": 3.5790911077124756e-06, + "loss": 0.6986, + "step": 13057 + }, + { + "epoch": 2.85, + "learning_rate": 3.5789818658509947e-06, + "loss": 1.4986, + "step": 13058 + }, + { + "epoch": 2.85, + "learning_rate": 3.5788726239895133e-06, + "loss": 0.4834, + "step": 13059 + }, + { + "epoch": 2.85, + "learning_rate": 3.5787633821280315e-06, + "loss": 1.8216, + "step": 13060 + }, + { + "epoch": 2.85, + "learning_rate": 3.5786541402665505e-06, + "loss": 2.445, + "step": 13061 + }, + { + "epoch": 2.85, + "learning_rate": 3.578544898405069e-06, + "loss": 1.2622, + "step": 13062 + }, + { + "epoch": 2.85, + "learning_rate": 3.578435656543588e-06, + "loss": 0.9973, + "step": 13063 + }, + { + "epoch": 2.85, + "learning_rate": 3.5783264146821067e-06, + "loss": 1.8648, + "step": 13064 + }, + { + "epoch": 2.85, + "learning_rate": 3.578217172820625e-06, + "loss": 1.2695, + "step": 13065 + }, + { + "epoch": 2.85, + "learning_rate": 3.578107930959144e-06, + "loss": 3.3072, + "step": 13066 + }, + { + "epoch": 2.85, + "learning_rate": 3.5779986890976626e-06, + "loss": 1.0147, + "step": 13067 + }, + { + "epoch": 2.85, + "learning_rate": 3.577889447236181e-06, + "loss": 1.7857, + "step": 13068 + }, + { + "epoch": 2.85, + "learning_rate": 3.5777802053747e-06, + "loss": 2.5678, + "step": 13069 + }, + { + "epoch": 2.85, + "learning_rate": 3.5776709635132184e-06, + "loss": 2.429, + "step": 13070 + }, + { + "epoch": 2.85, + "learning_rate": 3.577561721651737e-06, + "loss": 2.6134, + "step": 13071 + }, + { + "epoch": 2.85, + "learning_rate": 3.577452479790256e-06, + "loss": 1.4488, + "step": 13072 + }, + { + "epoch": 2.85, + "learning_rate": 3.5773432379287747e-06, + "loss": 3.4652, + "step": 13073 + }, + { + "epoch": 2.85, + "learning_rate": 3.577233996067293e-06, + "loss": 1.6752, + "step": 13074 + }, + { + "epoch": 2.85, + "learning_rate": 3.577124754205812e-06, + "loss": 0.981, + "step": 13075 + }, + { + "epoch": 2.85, + "learning_rate": 3.5770155123443305e-06, + "loss": 2.0879, + "step": 13076 + }, + { + "epoch": 2.85, + "learning_rate": 3.5769062704828495e-06, + "loss": 3.0355, + "step": 13077 + }, + { + "epoch": 2.85, + "learning_rate": 3.576797028621368e-06, + "loss": 2.1895, + "step": 13078 + }, + { + "epoch": 2.85, + "learning_rate": 3.5766877867598863e-06, + "loss": 1.8431, + "step": 13079 + }, + { + "epoch": 2.85, + "learning_rate": 3.5765785448984054e-06, + "loss": 1.3917, + "step": 13080 + }, + { + "epoch": 2.85, + "learning_rate": 3.576469303036924e-06, + "loss": 2.2954, + "step": 13081 + }, + { + "epoch": 2.86, + "learning_rate": 3.576360061175443e-06, + "loss": 1.6603, + "step": 13082 + }, + { + "epoch": 2.86, + "learning_rate": 3.5762508193139616e-06, + "loss": 2.6049, + "step": 13083 + }, + { + "epoch": 2.86, + "learning_rate": 3.57614157745248e-06, + "loss": 1.9708, + "step": 13084 + }, + { + "epoch": 2.86, + "learning_rate": 3.576032335590999e-06, + "loss": 3.0254, + "step": 13085 + }, + { + "epoch": 2.86, + "learning_rate": 3.5759230937295175e-06, + "loss": 1.7738, + "step": 13086 + }, + { + "epoch": 2.86, + "learning_rate": 3.5758138518680365e-06, + "loss": 3.0956, + "step": 13087 + }, + { + "epoch": 2.86, + "learning_rate": 3.5757046100065547e-06, + "loss": 1.1078, + "step": 13088 + }, + { + "epoch": 2.86, + "learning_rate": 3.5755953681450733e-06, + "loss": 1.7657, + "step": 13089 + }, + { + "epoch": 2.86, + "learning_rate": 3.5754861262835923e-06, + "loss": 3.2133, + "step": 13090 + }, + { + "epoch": 2.86, + "learning_rate": 3.575376884422111e-06, + "loss": 0.9776, + "step": 13091 + }, + { + "epoch": 2.86, + "learning_rate": 3.5752676425606295e-06, + "loss": 2.5489, + "step": 13092 + }, + { + "epoch": 2.86, + "learning_rate": 3.575158400699148e-06, + "loss": 2.6894, + "step": 13093 + }, + { + "epoch": 2.86, + "learning_rate": 3.5750491588376668e-06, + "loss": 3.136, + "step": 13094 + }, + { + "epoch": 2.86, + "learning_rate": 3.5749399169761854e-06, + "loss": 3.0801, + "step": 13095 + }, + { + "epoch": 2.86, + "learning_rate": 3.5748306751147044e-06, + "loss": 2.2894, + "step": 13096 + }, + { + "epoch": 2.86, + "learning_rate": 3.574721433253223e-06, + "loss": 2.6907, + "step": 13097 + }, + { + "epoch": 2.86, + "learning_rate": 3.5746121913917416e-06, + "loss": 2.4806, + "step": 13098 + }, + { + "epoch": 2.86, + "learning_rate": 3.5745029495302602e-06, + "loss": 1.725, + "step": 13099 + }, + { + "epoch": 2.86, + "learning_rate": 3.574393707668779e-06, + "loss": 0.9299, + "step": 13100 + }, + { + "epoch": 2.86, + "learning_rate": 3.574284465807298e-06, + "loss": 2.8262, + "step": 13101 + }, + { + "epoch": 2.86, + "learning_rate": 3.574175223945816e-06, + "loss": 2.7237, + "step": 13102 + }, + { + "epoch": 2.86, + "learning_rate": 3.5740659820843347e-06, + "loss": 2.2386, + "step": 13103 + }, + { + "epoch": 2.86, + "learning_rate": 3.5739567402228537e-06, + "loss": 2.5706, + "step": 13104 + }, + { + "epoch": 2.86, + "learning_rate": 3.5738474983613723e-06, + "loss": 2.3549, + "step": 13105 + }, + { + "epoch": 2.86, + "learning_rate": 3.5737382564998914e-06, + "loss": 2.1833, + "step": 13106 + }, + { + "epoch": 2.86, + "learning_rate": 3.5736290146384096e-06, + "loss": 1.6496, + "step": 13107 + }, + { + "epoch": 2.86, + "learning_rate": 3.573519772776928e-06, + "loss": 2.8409, + "step": 13108 + }, + { + "epoch": 2.86, + "learning_rate": 3.573410530915447e-06, + "loss": 2.7267, + "step": 13109 + }, + { + "epoch": 2.86, + "learning_rate": 3.573301289053966e-06, + "loss": 2.9325, + "step": 13110 + }, + { + "epoch": 2.86, + "learning_rate": 3.573192047192485e-06, + "loss": 1.604, + "step": 13111 + }, + { + "epoch": 2.86, + "learning_rate": 3.573082805331003e-06, + "loss": 1.8038, + "step": 13112 + }, + { + "epoch": 2.86, + "learning_rate": 3.5729735634695216e-06, + "loss": 1.9675, + "step": 13113 + }, + { + "epoch": 2.86, + "learning_rate": 3.5728643216080407e-06, + "loss": 2.1621, + "step": 13114 + }, + { + "epoch": 2.86, + "learning_rate": 3.5727550797465593e-06, + "loss": 1.996, + "step": 13115 + }, + { + "epoch": 2.86, + "learning_rate": 3.5726458378850783e-06, + "loss": 2.2556, + "step": 13116 + }, + { + "epoch": 2.86, + "learning_rate": 3.5725365960235965e-06, + "loss": 1.9203, + "step": 13117 + }, + { + "epoch": 2.86, + "learning_rate": 3.572427354162115e-06, + "loss": 2.7033, + "step": 13118 + }, + { + "epoch": 2.86, + "learning_rate": 3.572318112300634e-06, + "loss": 1.9614, + "step": 13119 + }, + { + "epoch": 2.86, + "learning_rate": 3.5722088704391528e-06, + "loss": 2.3663, + "step": 13120 + }, + { + "epoch": 2.86, + "learning_rate": 3.572099628577671e-06, + "loss": 2.1602, + "step": 13121 + }, + { + "epoch": 2.86, + "learning_rate": 3.57199038671619e-06, + "loss": 2.2862, + "step": 13122 + }, + { + "epoch": 2.86, + "learning_rate": 3.5718811448547086e-06, + "loss": 3.1306, + "step": 13123 + }, + { + "epoch": 2.86, + "learning_rate": 3.5717719029932272e-06, + "loss": 2.2583, + "step": 13124 + }, + { + "epoch": 2.86, + "learning_rate": 3.5716626611317462e-06, + "loss": 1.3515, + "step": 13125 + }, + { + "epoch": 2.86, + "learning_rate": 3.5715534192702644e-06, + "loss": 2.9387, + "step": 13126 + }, + { + "epoch": 2.86, + "learning_rate": 3.571444177408783e-06, + "loss": 2.2808, + "step": 13127 + }, + { + "epoch": 2.87, + "learning_rate": 3.571334935547302e-06, + "loss": 2.1991, + "step": 13128 + }, + { + "epoch": 2.87, + "learning_rate": 3.5712256936858207e-06, + "loss": 2.6316, + "step": 13129 + }, + { + "epoch": 2.87, + "learning_rate": 3.5711164518243397e-06, + "loss": 1.6707, + "step": 13130 + }, + { + "epoch": 2.87, + "learning_rate": 3.571007209962858e-06, + "loss": 1.6232, + "step": 13131 + }, + { + "epoch": 2.87, + "learning_rate": 3.5708979681013765e-06, + "loss": 0.9996, + "step": 13132 + }, + { + "epoch": 2.87, + "learning_rate": 3.5707887262398956e-06, + "loss": 2.6396, + "step": 13133 + }, + { + "epoch": 2.87, + "learning_rate": 3.570679484378414e-06, + "loss": 2.1825, + "step": 13134 + }, + { + "epoch": 2.87, + "learning_rate": 3.5705702425169324e-06, + "loss": 0.9191, + "step": 13135 + }, + { + "epoch": 2.87, + "learning_rate": 3.5704610006554514e-06, + "loss": 1.6605, + "step": 13136 + }, + { + "epoch": 2.87, + "learning_rate": 3.57035175879397e-06, + "loss": 1.2845, + "step": 13137 + }, + { + "epoch": 2.87, + "learning_rate": 3.570242516932489e-06, + "loss": 2.8759, + "step": 13138 + }, + { + "epoch": 2.87, + "learning_rate": 3.5701332750710076e-06, + "loss": 1.22, + "step": 13139 + }, + { + "epoch": 2.87, + "learning_rate": 3.570024033209526e-06, + "loss": 2.0786, + "step": 13140 + }, + { + "epoch": 2.87, + "learning_rate": 3.569914791348045e-06, + "loss": 2.2355, + "step": 13141 + }, + { + "epoch": 2.87, + "learning_rate": 3.5698055494865635e-06, + "loss": 2.0169, + "step": 13142 + }, + { + "epoch": 2.87, + "learning_rate": 3.5696963076250825e-06, + "loss": 2.1633, + "step": 13143 + }, + { + "epoch": 2.87, + "learning_rate": 3.569587065763601e-06, + "loss": 2.2153, + "step": 13144 + }, + { + "epoch": 2.87, + "learning_rate": 3.5694778239021193e-06, + "loss": 1.8783, + "step": 13145 + }, + { + "epoch": 2.87, + "learning_rate": 3.5693685820406383e-06, + "loss": 1.2973, + "step": 13146 + }, + { + "epoch": 2.87, + "learning_rate": 3.569259340179157e-06, + "loss": 2.3966, + "step": 13147 + }, + { + "epoch": 2.87, + "learning_rate": 3.569150098317676e-06, + "loss": 3.3771, + "step": 13148 + }, + { + "epoch": 2.87, + "learning_rate": 3.5690408564561946e-06, + "loss": 2.4372, + "step": 13149 + }, + { + "epoch": 2.87, + "learning_rate": 3.568931614594713e-06, + "loss": 2.9919, + "step": 13150 + }, + { + "epoch": 2.87, + "learning_rate": 3.568822372733232e-06, + "loss": 2.5983, + "step": 13151 + }, + { + "epoch": 2.87, + "learning_rate": 3.5687131308717504e-06, + "loss": 1.7556, + "step": 13152 + }, + { + "epoch": 2.87, + "learning_rate": 3.568603889010269e-06, + "loss": 3.312, + "step": 13153 + }, + { + "epoch": 2.87, + "learning_rate": 3.5684946471487877e-06, + "loss": 1.4373, + "step": 13154 + }, + { + "epoch": 2.87, + "learning_rate": 3.5683854052873063e-06, + "loss": 1.228, + "step": 13155 + }, + { + "epoch": 2.87, + "learning_rate": 3.568276163425825e-06, + "loss": 2.3526, + "step": 13156 + }, + { + "epoch": 2.87, + "learning_rate": 3.568166921564344e-06, + "loss": 1.1814, + "step": 13157 + }, + { + "epoch": 2.87, + "learning_rate": 3.5680576797028625e-06, + "loss": 0.8803, + "step": 13158 + }, + { + "epoch": 2.87, + "learning_rate": 3.5679484378413807e-06, + "loss": 1.6393, + "step": 13159 + }, + { + "epoch": 2.87, + "learning_rate": 3.5678391959798997e-06, + "loss": 3.0834, + "step": 13160 + }, + { + "epoch": 2.87, + "learning_rate": 3.5677299541184184e-06, + "loss": 1.8806, + "step": 13161 + }, + { + "epoch": 2.87, + "learning_rate": 3.5676207122569374e-06, + "loss": 2.4769, + "step": 13162 + }, + { + "epoch": 2.87, + "learning_rate": 3.567511470395456e-06, + "loss": 2.463, + "step": 13163 + }, + { + "epoch": 2.87, + "learning_rate": 3.567402228533974e-06, + "loss": 2.7699, + "step": 13164 + }, + { + "epoch": 2.87, + "learning_rate": 3.5672929866724932e-06, + "loss": 2.8952, + "step": 13165 + }, + { + "epoch": 2.87, + "learning_rate": 3.567183744811012e-06, + "loss": 2.9593, + "step": 13166 + }, + { + "epoch": 2.87, + "learning_rate": 3.567074502949531e-06, + "loss": 1.8331, + "step": 13167 + }, + { + "epoch": 2.87, + "learning_rate": 3.566965261088049e-06, + "loss": 2.5001, + "step": 13168 + }, + { + "epoch": 2.87, + "learning_rate": 3.5668560192265677e-06, + "loss": 1.2905, + "step": 13169 + }, + { + "epoch": 2.87, + "learning_rate": 3.5667467773650867e-06, + "loss": 2.5692, + "step": 13170 + }, + { + "epoch": 2.87, + "learning_rate": 3.5666375355036053e-06, + "loss": 1.1496, + "step": 13171 + }, + { + "epoch": 2.87, + "learning_rate": 3.5665282936421244e-06, + "loss": 2.4288, + "step": 13172 + }, + { + "epoch": 2.87, + "learning_rate": 3.5664190517806425e-06, + "loss": 1.6685, + "step": 13173 + }, + { + "epoch": 2.88, + "learning_rate": 3.566309809919161e-06, + "loss": 1.9649, + "step": 13174 + }, + { + "epoch": 2.88, + "learning_rate": 3.56620056805768e-06, + "loss": 1.8898, + "step": 13175 + }, + { + "epoch": 2.88, + "learning_rate": 3.566091326196199e-06, + "loss": 2.4752, + "step": 13176 + }, + { + "epoch": 2.88, + "learning_rate": 3.5659820843347174e-06, + "loss": 1.659, + "step": 13177 + }, + { + "epoch": 2.88, + "learning_rate": 3.565872842473236e-06, + "loss": 2.6112, + "step": 13178 + }, + { + "epoch": 2.88, + "learning_rate": 3.5657636006117546e-06, + "loss": 0.8178, + "step": 13179 + }, + { + "epoch": 2.88, + "learning_rate": 3.5656543587502732e-06, + "loss": 1.5948, + "step": 13180 + }, + { + "epoch": 2.88, + "learning_rate": 3.5655451168887923e-06, + "loss": 2.4645, + "step": 13181 + }, + { + "epoch": 2.88, + "learning_rate": 3.565435875027311e-06, + "loss": 1.7807, + "step": 13182 + }, + { + "epoch": 2.88, + "learning_rate": 3.565326633165829e-06, + "loss": 2.7695, + "step": 13183 + }, + { + "epoch": 2.88, + "learning_rate": 3.565217391304348e-06, + "loss": 2.3014, + "step": 13184 + }, + { + "epoch": 2.88, + "learning_rate": 3.5651081494428667e-06, + "loss": 2.5624, + "step": 13185 + }, + { + "epoch": 2.88, + "learning_rate": 3.5649989075813858e-06, + "loss": 1.8744, + "step": 13186 + }, + { + "epoch": 2.88, + "learning_rate": 3.564889665719904e-06, + "loss": 3.0174, + "step": 13187 + }, + { + "epoch": 2.88, + "learning_rate": 3.5647804238584226e-06, + "loss": 2.2826, + "step": 13188 + }, + { + "epoch": 2.88, + "learning_rate": 3.5646711819969416e-06, + "loss": 2.0518, + "step": 13189 + }, + { + "epoch": 2.88, + "learning_rate": 3.56456194013546e-06, + "loss": 2.4696, + "step": 13190 + }, + { + "epoch": 2.88, + "learning_rate": 3.5644526982739792e-06, + "loss": 1.745, + "step": 13191 + }, + { + "epoch": 2.88, + "learning_rate": 3.5643434564124974e-06, + "loss": 1.5662, + "step": 13192 + }, + { + "epoch": 2.88, + "learning_rate": 3.564234214551016e-06, + "loss": 1.9929, + "step": 13193 + }, + { + "epoch": 2.88, + "learning_rate": 3.564124972689535e-06, + "loss": 2.9449, + "step": 13194 + }, + { + "epoch": 2.88, + "learning_rate": 3.5640157308280537e-06, + "loss": 2.2713, + "step": 13195 + }, + { + "epoch": 2.88, + "learning_rate": 3.5639064889665727e-06, + "loss": 2.6261, + "step": 13196 + }, + { + "epoch": 2.88, + "learning_rate": 3.563797247105091e-06, + "loss": 3.7882, + "step": 13197 + }, + { + "epoch": 2.88, + "learning_rate": 3.5636880052436095e-06, + "loss": 2.6731, + "step": 13198 + }, + { + "epoch": 2.88, + "learning_rate": 3.5635787633821285e-06, + "loss": 2.3621, + "step": 13199 + }, + { + "epoch": 2.88, + "learning_rate": 3.563469521520647e-06, + "loss": 1.9544, + "step": 13200 + }, + { + "epoch": 2.88, + "learning_rate": 3.5633602796591653e-06, + "loss": 2.9677, + "step": 13201 + }, + { + "epoch": 2.88, + "learning_rate": 3.5632510377976844e-06, + "loss": 2.2386, + "step": 13202 + }, + { + "epoch": 2.88, + "learning_rate": 3.563141795936203e-06, + "loss": 2.1901, + "step": 13203 + }, + { + "epoch": 2.88, + "learning_rate": 3.563032554074722e-06, + "loss": 2.0604, + "step": 13204 + }, + { + "epoch": 2.88, + "learning_rate": 3.5629233122132406e-06, + "loss": 1.4188, + "step": 13205 + }, + { + "epoch": 2.88, + "learning_rate": 3.562814070351759e-06, + "loss": 1.0463, + "step": 13206 + }, + { + "epoch": 2.88, + "learning_rate": 3.562704828490278e-06, + "loss": 1.4716, + "step": 13207 + }, + { + "epoch": 2.88, + "learning_rate": 3.5625955866287965e-06, + "loss": 2.1092, + "step": 13208 + }, + { + "epoch": 2.88, + "learning_rate": 3.562486344767315e-06, + "loss": 2.3386, + "step": 13209 + }, + { + "epoch": 2.88, + "learning_rate": 3.562377102905834e-06, + "loss": 2.4944, + "step": 13210 + }, + { + "epoch": 2.88, + "learning_rate": 3.5622678610443523e-06, + "loss": 2.1464, + "step": 13211 + }, + { + "epoch": 2.88, + "learning_rate": 3.562158619182871e-06, + "loss": 2.8457, + "step": 13212 + }, + { + "epoch": 2.88, + "learning_rate": 3.56204937732139e-06, + "loss": 2.3913, + "step": 13213 + }, + { + "epoch": 2.88, + "learning_rate": 3.5619401354599086e-06, + "loss": 2.55, + "step": 13214 + }, + { + "epoch": 2.88, + "learning_rate": 3.5618308935984276e-06, + "loss": 2.2121, + "step": 13215 + }, + { + "epoch": 2.88, + "learning_rate": 3.5617216517369458e-06, + "loss": 2.7368, + "step": 13216 + }, + { + "epoch": 2.88, + "learning_rate": 3.5616124098754644e-06, + "loss": 3.1061, + "step": 13217 + }, + { + "epoch": 2.88, + "learning_rate": 3.5615031680139834e-06, + "loss": 1.9008, + "step": 13218 + }, + { + "epoch": 2.88, + "learning_rate": 3.561393926152502e-06, + "loss": 2.8542, + "step": 13219 + }, + { + "epoch": 2.89, + "learning_rate": 3.5612846842910202e-06, + "loss": 2.1781, + "step": 13220 + }, + { + "epoch": 2.89, + "learning_rate": 3.5611754424295393e-06, + "loss": 2.6716, + "step": 13221 + }, + { + "epoch": 2.89, + "learning_rate": 3.561066200568058e-06, + "loss": 1.3514, + "step": 13222 + }, + { + "epoch": 2.89, + "learning_rate": 3.560956958706577e-06, + "loss": 3.0882, + "step": 13223 + }, + { + "epoch": 2.89, + "learning_rate": 3.5608477168450955e-06, + "loss": 2.596, + "step": 13224 + }, + { + "epoch": 2.89, + "learning_rate": 3.5607384749836137e-06, + "loss": 1.7417, + "step": 13225 + }, + { + "epoch": 2.89, + "learning_rate": 3.5606292331221327e-06, + "loss": 2.2539, + "step": 13226 + }, + { + "epoch": 2.89, + "learning_rate": 3.5605199912606513e-06, + "loss": 2.7719, + "step": 13227 + }, + { + "epoch": 2.89, + "learning_rate": 3.5604107493991704e-06, + "loss": 2.3331, + "step": 13228 + }, + { + "epoch": 2.89, + "learning_rate": 3.560301507537689e-06, + "loss": 2.3008, + "step": 13229 + }, + { + "epoch": 2.89, + "learning_rate": 3.560192265676207e-06, + "loss": 0.9852, + "step": 13230 + }, + { + "epoch": 2.89, + "learning_rate": 3.560083023814726e-06, + "loss": 3.0894, + "step": 13231 + }, + { + "epoch": 2.89, + "learning_rate": 3.559973781953245e-06, + "loss": 3.4408, + "step": 13232 + }, + { + "epoch": 2.89, + "learning_rate": 3.5598645400917634e-06, + "loss": 2.9333, + "step": 13233 + }, + { + "epoch": 2.89, + "learning_rate": 3.559755298230282e-06, + "loss": 2.3068, + "step": 13234 + }, + { + "epoch": 2.89, + "learning_rate": 3.5596460563688007e-06, + "loss": 2.6797, + "step": 13235 + }, + { + "epoch": 2.89, + "learning_rate": 3.5595368145073193e-06, + "loss": 2.7608, + "step": 13236 + }, + { + "epoch": 2.89, + "learning_rate": 3.5594275726458383e-06, + "loss": 2.4675, + "step": 13237 + }, + { + "epoch": 2.89, + "learning_rate": 3.559318330784357e-06, + "loss": 2.3045, + "step": 13238 + }, + { + "epoch": 2.89, + "learning_rate": 3.559209088922875e-06, + "loss": 2.8746, + "step": 13239 + }, + { + "epoch": 2.89, + "learning_rate": 3.559099847061394e-06, + "loss": 1.5769, + "step": 13240 + }, + { + "epoch": 2.89, + "learning_rate": 3.5589906051999127e-06, + "loss": 0.8652, + "step": 13241 + }, + { + "epoch": 2.89, + "learning_rate": 3.5588813633384318e-06, + "loss": 3.3751, + "step": 13242 + }, + { + "epoch": 2.89, + "learning_rate": 3.5587721214769504e-06, + "loss": 2.3962, + "step": 13243 + }, + { + "epoch": 2.89, + "learning_rate": 3.5586628796154686e-06, + "loss": 1.8818, + "step": 13244 + }, + { + "epoch": 2.89, + "learning_rate": 3.5585536377539876e-06, + "loss": 2.5279, + "step": 13245 + }, + { + "epoch": 2.89, + "learning_rate": 3.5584443958925062e-06, + "loss": 2.8058, + "step": 13246 + }, + { + "epoch": 2.89, + "learning_rate": 3.5583351540310253e-06, + "loss": 3.0504, + "step": 13247 + }, + { + "epoch": 2.89, + "learning_rate": 3.5582259121695434e-06, + "loss": 1.8273, + "step": 13248 + }, + { + "epoch": 2.89, + "learning_rate": 3.558116670308062e-06, + "loss": 1.8471, + "step": 13249 + }, + { + "epoch": 2.89, + "learning_rate": 3.558007428446581e-06, + "loss": 2.4892, + "step": 13250 + }, + { + "epoch": 2.89, + "learning_rate": 3.5578981865850997e-06, + "loss": 3.365, + "step": 13251 + }, + { + "epoch": 2.89, + "learning_rate": 3.5577889447236187e-06, + "loss": 2.4649, + "step": 13252 + }, + { + "epoch": 2.89, + "learning_rate": 3.557679702862137e-06, + "loss": 2.2227, + "step": 13253 + }, + { + "epoch": 2.89, + "learning_rate": 3.5575704610006555e-06, + "loss": 1.5349, + "step": 13254 + }, + { + "epoch": 2.89, + "learning_rate": 3.5574612191391746e-06, + "loss": 2.2308, + "step": 13255 + }, + { + "epoch": 2.89, + "learning_rate": 3.557351977277693e-06, + "loss": 1.7434, + "step": 13256 + }, + { + "epoch": 2.89, + "learning_rate": 3.5572427354162122e-06, + "loss": 2.4487, + "step": 13257 + }, + { + "epoch": 2.89, + "learning_rate": 3.5571334935547304e-06, + "loss": 2.2186, + "step": 13258 + }, + { + "epoch": 2.89, + "learning_rate": 3.557024251693249e-06, + "loss": 2.5699, + "step": 13259 + }, + { + "epoch": 2.89, + "learning_rate": 3.556915009831768e-06, + "loss": 2.667, + "step": 13260 + }, + { + "epoch": 2.89, + "learning_rate": 3.5568057679702867e-06, + "loss": 2.6966, + "step": 13261 + }, + { + "epoch": 2.89, + "learning_rate": 3.5566965261088053e-06, + "loss": 1.6018, + "step": 13262 + }, + { + "epoch": 2.89, + "learning_rate": 3.556587284247324e-06, + "loss": 2.5938, + "step": 13263 + }, + { + "epoch": 2.89, + "learning_rate": 3.5564780423858425e-06, + "loss": 2.2504, + "step": 13264 + }, + { + "epoch": 2.9, + "learning_rate": 3.556368800524361e-06, + "loss": 3.4869, + "step": 13265 + }, + { + "epoch": 2.9, + "learning_rate": 3.55625955866288e-06, + "loss": 2.169, + "step": 13266 + }, + { + "epoch": 2.9, + "learning_rate": 3.5561503168013983e-06, + "loss": 2.0431, + "step": 13267 + }, + { + "epoch": 2.9, + "learning_rate": 3.556041074939917e-06, + "loss": 0.8434, + "step": 13268 + }, + { + "epoch": 2.9, + "learning_rate": 3.555931833078436e-06, + "loss": 1.7817, + "step": 13269 + }, + { + "epoch": 2.9, + "learning_rate": 3.5558225912169546e-06, + "loss": 2.0524, + "step": 13270 + }, + { + "epoch": 2.9, + "learning_rate": 3.5557133493554736e-06, + "loss": 1.7865, + "step": 13271 + }, + { + "epoch": 2.9, + "learning_rate": 3.555604107493992e-06, + "loss": 2.4967, + "step": 13272 + }, + { + "epoch": 2.9, + "learning_rate": 3.5554948656325104e-06, + "loss": 1.673, + "step": 13273 + }, + { + "epoch": 2.9, + "learning_rate": 3.5553856237710294e-06, + "loss": 2.0662, + "step": 13274 + }, + { + "epoch": 2.9, + "learning_rate": 3.555276381909548e-06, + "loss": 1.4306, + "step": 13275 + }, + { + "epoch": 2.9, + "learning_rate": 3.555167140048067e-06, + "loss": 1.3685, + "step": 13276 + }, + { + "epoch": 2.9, + "learning_rate": 3.5550578981865853e-06, + "loss": 3.0942, + "step": 13277 + }, + { + "epoch": 2.9, + "learning_rate": 3.554948656325104e-06, + "loss": 1.613, + "step": 13278 + }, + { + "epoch": 2.9, + "learning_rate": 3.554839414463623e-06, + "loss": 3.5619, + "step": 13279 + }, + { + "epoch": 2.9, + "learning_rate": 3.5547301726021415e-06, + "loss": 1.7924, + "step": 13280 + }, + { + "epoch": 2.9, + "learning_rate": 3.5546209307406597e-06, + "loss": 2.1621, + "step": 13281 + }, + { + "epoch": 2.9, + "learning_rate": 3.5545116888791788e-06, + "loss": 2.5815, + "step": 13282 + }, + { + "epoch": 2.9, + "learning_rate": 3.5544024470176974e-06, + "loss": 2.9965, + "step": 13283 + }, + { + "epoch": 2.9, + "learning_rate": 3.5542932051562164e-06, + "loss": 1.6958, + "step": 13284 + }, + { + "epoch": 2.9, + "learning_rate": 3.554183963294735e-06, + "loss": 2.3851, + "step": 13285 + }, + { + "epoch": 2.9, + "learning_rate": 3.554074721433253e-06, + "loss": 1.8023, + "step": 13286 + }, + { + "epoch": 2.9, + "learning_rate": 3.5539654795717722e-06, + "loss": 2.3817, + "step": 13287 + }, + { + "epoch": 2.9, + "learning_rate": 3.553856237710291e-06, + "loss": 2.637, + "step": 13288 + }, + { + "epoch": 2.9, + "learning_rate": 3.5537469958488095e-06, + "loss": 1.6253, + "step": 13289 + }, + { + "epoch": 2.9, + "learning_rate": 3.5536377539873285e-06, + "loss": 2.965, + "step": 13290 + }, + { + "epoch": 2.9, + "learning_rate": 3.5535285121258467e-06, + "loss": 2.4526, + "step": 13291 + }, + { + "epoch": 2.9, + "learning_rate": 3.5534192702643657e-06, + "loss": 1.9897, + "step": 13292 + }, + { + "epoch": 2.9, + "learning_rate": 3.5533100284028843e-06, + "loss": 2.9202, + "step": 13293 + }, + { + "epoch": 2.9, + "learning_rate": 3.553200786541403e-06, + "loss": 1.5368, + "step": 13294 + }, + { + "epoch": 2.9, + "learning_rate": 3.553091544679922e-06, + "loss": 1.19, + "step": 13295 + }, + { + "epoch": 2.9, + "learning_rate": 3.55298230281844e-06, + "loss": 2.4682, + "step": 13296 + }, + { + "epoch": 2.9, + "learning_rate": 3.5528730609569588e-06, + "loss": 3.2868, + "step": 13297 + }, + { + "epoch": 2.9, + "learning_rate": 3.552763819095478e-06, + "loss": 1.7049, + "step": 13298 + }, + { + "epoch": 2.9, + "learning_rate": 3.5526545772339964e-06, + "loss": 1.1038, + "step": 13299 + }, + { + "epoch": 2.9, + "learning_rate": 3.5525453353725146e-06, + "loss": 2.9307, + "step": 13300 + }, + { + "epoch": 2.9, + "learning_rate": 3.5524360935110336e-06, + "loss": 1.5304, + "step": 13301 + }, + { + "epoch": 2.9, + "learning_rate": 3.5523268516495523e-06, + "loss": 1.6283, + "step": 13302 + }, + { + "epoch": 2.9, + "learning_rate": 3.5522176097880713e-06, + "loss": 1.8757, + "step": 13303 + }, + { + "epoch": 2.9, + "learning_rate": 3.55210836792659e-06, + "loss": 1.3202, + "step": 13304 + }, + { + "epoch": 2.9, + "learning_rate": 3.551999126065108e-06, + "loss": 1.411, + "step": 13305 + }, + { + "epoch": 2.9, + "learning_rate": 3.551889884203627e-06, + "loss": 2.7301, + "step": 13306 + }, + { + "epoch": 2.9, + "learning_rate": 3.5517806423421457e-06, + "loss": 2.2496, + "step": 13307 + }, + { + "epoch": 2.9, + "learning_rate": 3.5516714004806648e-06, + "loss": 2.0054, + "step": 13308 + }, + { + "epoch": 2.9, + "learning_rate": 3.5515621586191834e-06, + "loss": 1.6303, + "step": 13309 + }, + { + "epoch": 2.9, + "learning_rate": 3.5514529167577016e-06, + "loss": 1.2167, + "step": 13310 + }, + { + "epoch": 2.91, + "learning_rate": 3.5513436748962206e-06, + "loss": 2.0301, + "step": 13311 + }, + { + "epoch": 2.91, + "learning_rate": 3.551234433034739e-06, + "loss": 2.7874, + "step": 13312 + }, + { + "epoch": 2.91, + "learning_rate": 3.5511251911732582e-06, + "loss": 2.4814, + "step": 13313 + }, + { + "epoch": 2.91, + "learning_rate": 3.5510159493117764e-06, + "loss": 2.4758, + "step": 13314 + }, + { + "epoch": 2.91, + "learning_rate": 3.550906707450295e-06, + "loss": 1.7446, + "step": 13315 + }, + { + "epoch": 2.91, + "learning_rate": 3.550797465588814e-06, + "loss": 2.7527, + "step": 13316 + }, + { + "epoch": 2.91, + "learning_rate": 3.5506882237273327e-06, + "loss": 2.7981, + "step": 13317 + }, + { + "epoch": 2.91, + "learning_rate": 3.5505789818658513e-06, + "loss": 2.2662, + "step": 13318 + }, + { + "epoch": 2.91, + "learning_rate": 3.55046974000437e-06, + "loss": 1.7642, + "step": 13319 + }, + { + "epoch": 2.91, + "learning_rate": 3.5503604981428885e-06, + "loss": 1.0423, + "step": 13320 + }, + { + "epoch": 2.91, + "learning_rate": 3.550251256281407e-06, + "loss": 2.7856, + "step": 13321 + }, + { + "epoch": 2.91, + "learning_rate": 3.550142014419926e-06, + "loss": 2.5748, + "step": 13322 + }, + { + "epoch": 2.91, + "learning_rate": 3.5500327725584448e-06, + "loss": 2.0818, + "step": 13323 + }, + { + "epoch": 2.91, + "learning_rate": 3.549923530696963e-06, + "loss": 1.9075, + "step": 13324 + }, + { + "epoch": 2.91, + "learning_rate": 3.549814288835482e-06, + "loss": 2.1736, + "step": 13325 + }, + { + "epoch": 2.91, + "learning_rate": 3.5497050469740006e-06, + "loss": 3.0848, + "step": 13326 + }, + { + "epoch": 2.91, + "learning_rate": 3.5495958051125196e-06, + "loss": 1.8408, + "step": 13327 + }, + { + "epoch": 2.91, + "learning_rate": 3.5494865632510383e-06, + "loss": 1.3135, + "step": 13328 + }, + { + "epoch": 2.91, + "learning_rate": 3.5493773213895564e-06, + "loss": 1.0601, + "step": 13329 + }, + { + "epoch": 2.91, + "learning_rate": 3.5492680795280755e-06, + "loss": 2.8844, + "step": 13330 + }, + { + "epoch": 2.91, + "learning_rate": 3.549158837666594e-06, + "loss": 3.4256, + "step": 13331 + }, + { + "epoch": 2.91, + "learning_rate": 3.549049595805113e-06, + "loss": 1.9899, + "step": 13332 + }, + { + "epoch": 2.91, + "learning_rate": 3.5489403539436313e-06, + "loss": 1.7983, + "step": 13333 + }, + { + "epoch": 2.91, + "learning_rate": 3.54883111208215e-06, + "loss": 4.1221, + "step": 13334 + }, + { + "epoch": 2.91, + "learning_rate": 3.548721870220669e-06, + "loss": 1.7697, + "step": 13335 + }, + { + "epoch": 2.91, + "learning_rate": 3.5486126283591876e-06, + "loss": 2.7493, + "step": 13336 + }, + { + "epoch": 2.91, + "learning_rate": 3.5485033864977066e-06, + "loss": 1.5735, + "step": 13337 + }, + { + "epoch": 2.91, + "learning_rate": 3.5483941446362248e-06, + "loss": 1.483, + "step": 13338 + }, + { + "epoch": 2.91, + "learning_rate": 3.5482849027747434e-06, + "loss": 2.4538, + "step": 13339 + }, + { + "epoch": 2.91, + "learning_rate": 3.5481756609132624e-06, + "loss": 1.2531, + "step": 13340 + }, + { + "epoch": 2.91, + "learning_rate": 3.548066419051781e-06, + "loss": 3.6624, + "step": 13341 + }, + { + "epoch": 2.91, + "learning_rate": 3.5479571771902997e-06, + "loss": 2.4233, + "step": 13342 + }, + { + "epoch": 2.91, + "learning_rate": 3.5478479353288183e-06, + "loss": 3.4746, + "step": 13343 + }, + { + "epoch": 2.91, + "learning_rate": 3.547738693467337e-06, + "loss": 2.3912, + "step": 13344 + }, + { + "epoch": 2.91, + "learning_rate": 3.547629451605856e-06, + "loss": 2.4102, + "step": 13345 + }, + { + "epoch": 2.91, + "learning_rate": 3.5475202097443745e-06, + "loss": 2.4287, + "step": 13346 + }, + { + "epoch": 2.91, + "learning_rate": 3.5474109678828927e-06, + "loss": 1.2729, + "step": 13347 + }, + { + "epoch": 2.91, + "learning_rate": 3.5473017260214117e-06, + "loss": 2.3661, + "step": 13348 + }, + { + "epoch": 2.91, + "learning_rate": 3.5471924841599304e-06, + "loss": 0.8915, + "step": 13349 + }, + { + "epoch": 2.91, + "learning_rate": 3.547083242298449e-06, + "loss": 2.8476, + "step": 13350 + }, + { + "epoch": 2.91, + "learning_rate": 3.546974000436968e-06, + "loss": 1.957, + "step": 13351 + }, + { + "epoch": 2.91, + "learning_rate": 3.546864758575486e-06, + "loss": 2.5782, + "step": 13352 + }, + { + "epoch": 2.91, + "learning_rate": 3.546755516714005e-06, + "loss": 3.6715, + "step": 13353 + }, + { + "epoch": 2.91, + "learning_rate": 3.546646274852524e-06, + "loss": 2.7093, + "step": 13354 + }, + { + "epoch": 2.91, + "learning_rate": 3.5465370329910424e-06, + "loss": 1.6364, + "step": 13355 + }, + { + "epoch": 2.91, + "learning_rate": 3.5464277911295615e-06, + "loss": 2.918, + "step": 13356 + }, + { + "epoch": 2.92, + "learning_rate": 3.5463185492680797e-06, + "loss": 0.428, + "step": 13357 + }, + { + "epoch": 2.92, + "learning_rate": 3.5462093074065983e-06, + "loss": 1.7978, + "step": 13358 + }, + { + "epoch": 2.92, + "learning_rate": 3.5461000655451173e-06, + "loss": 1.8423, + "step": 13359 + }, + { + "epoch": 2.92, + "learning_rate": 3.545990823683636e-06, + "loss": 2.8969, + "step": 13360 + }, + { + "epoch": 2.92, + "learning_rate": 3.545881581822155e-06, + "loss": 1.1706, + "step": 13361 + }, + { + "epoch": 2.92, + "learning_rate": 3.545772339960673e-06, + "loss": 1.9901, + "step": 13362 + }, + { + "epoch": 2.92, + "learning_rate": 3.5456630980991918e-06, + "loss": 2.1611, + "step": 13363 + }, + { + "epoch": 2.92, + "learning_rate": 3.545553856237711e-06, + "loss": 1.9748, + "step": 13364 + }, + { + "epoch": 2.92, + "learning_rate": 3.5454446143762294e-06, + "loss": 3.4713, + "step": 13365 + }, + { + "epoch": 2.92, + "learning_rate": 3.5453353725147476e-06, + "loss": 1.4885, + "step": 13366 + }, + { + "epoch": 2.92, + "learning_rate": 3.5452261306532666e-06, + "loss": 2.4983, + "step": 13367 + }, + { + "epoch": 2.92, + "learning_rate": 3.5451168887917852e-06, + "loss": 1.51, + "step": 13368 + }, + { + "epoch": 2.92, + "learning_rate": 3.5450076469303043e-06, + "loss": 2.8191, + "step": 13369 + }, + { + "epoch": 2.92, + "learning_rate": 3.544898405068823e-06, + "loss": 2.3073, + "step": 13370 + }, + { + "epoch": 2.92, + "learning_rate": 3.544789163207341e-06, + "loss": 3.0093, + "step": 13371 + }, + { + "epoch": 2.92, + "learning_rate": 3.54467992134586e-06, + "loss": 2.1091, + "step": 13372 + }, + { + "epoch": 2.92, + "learning_rate": 3.5445706794843787e-06, + "loss": 2.0944, + "step": 13373 + }, + { + "epoch": 2.92, + "learning_rate": 3.5444614376228973e-06, + "loss": 2.1192, + "step": 13374 + }, + { + "epoch": 2.92, + "learning_rate": 3.5443521957614164e-06, + "loss": 2.3421, + "step": 13375 + }, + { + "epoch": 2.92, + "learning_rate": 3.5442429538999345e-06, + "loss": 1.4102, + "step": 13376 + }, + { + "epoch": 2.92, + "learning_rate": 3.544133712038453e-06, + "loss": 2.0085, + "step": 13377 + }, + { + "epoch": 2.92, + "learning_rate": 3.544024470176972e-06, + "loss": 2.5108, + "step": 13378 + }, + { + "epoch": 2.92, + "learning_rate": 3.543915228315491e-06, + "loss": 2.9568, + "step": 13379 + }, + { + "epoch": 2.92, + "learning_rate": 3.543805986454009e-06, + "loss": 1.8284, + "step": 13380 + }, + { + "epoch": 2.92, + "learning_rate": 3.543696744592528e-06, + "loss": 1.4317, + "step": 13381 + }, + { + "epoch": 2.92, + "learning_rate": 3.5435875027310466e-06, + "loss": 2.8865, + "step": 13382 + }, + { + "epoch": 2.92, + "learning_rate": 3.5434782608695657e-06, + "loss": 0.9573, + "step": 13383 + }, + { + "epoch": 2.92, + "learning_rate": 3.5433690190080843e-06, + "loss": 2.3845, + "step": 13384 + }, + { + "epoch": 2.92, + "learning_rate": 3.5432597771466025e-06, + "loss": 1.3899, + "step": 13385 + }, + { + "epoch": 2.92, + "learning_rate": 3.5431505352851215e-06, + "loss": 2.5446, + "step": 13386 + }, + { + "epoch": 2.92, + "learning_rate": 3.54304129342364e-06, + "loss": 2.5371, + "step": 13387 + }, + { + "epoch": 2.92, + "learning_rate": 3.542932051562159e-06, + "loss": 3.3854, + "step": 13388 + }, + { + "epoch": 2.92, + "learning_rate": 3.5428228097006778e-06, + "loss": 1.2304, + "step": 13389 + }, + { + "epoch": 2.92, + "learning_rate": 3.542713567839196e-06, + "loss": 2.2979, + "step": 13390 + }, + { + "epoch": 2.92, + "learning_rate": 3.542604325977715e-06, + "loss": 1.9955, + "step": 13391 + }, + { + "epoch": 2.92, + "learning_rate": 3.5424950841162336e-06, + "loss": 2.0657, + "step": 13392 + }, + { + "epoch": 2.92, + "learning_rate": 3.5423858422547526e-06, + "loss": 3.0853, + "step": 13393 + }, + { + "epoch": 2.92, + "learning_rate": 3.5422766003932712e-06, + "loss": 1.1549, + "step": 13394 + }, + { + "epoch": 2.92, + "learning_rate": 3.5421673585317894e-06, + "loss": 2.6979, + "step": 13395 + }, + { + "epoch": 2.92, + "learning_rate": 3.5420581166703085e-06, + "loss": 2.5433, + "step": 13396 + }, + { + "epoch": 2.92, + "learning_rate": 3.541948874808827e-06, + "loss": 1.2169, + "step": 13397 + }, + { + "epoch": 2.92, + "learning_rate": 3.541839632947346e-06, + "loss": 2.8408, + "step": 13398 + }, + { + "epoch": 2.92, + "learning_rate": 3.5417303910858643e-06, + "loss": 2.4212, + "step": 13399 + }, + { + "epoch": 2.92, + "learning_rate": 3.541621149224383e-06, + "loss": 2.182, + "step": 13400 + }, + { + "epoch": 2.92, + "learning_rate": 3.541511907362902e-06, + "loss": 1.626, + "step": 13401 + }, + { + "epoch": 2.92, + "learning_rate": 3.5414026655014205e-06, + "loss": 1.55, + "step": 13402 + }, + { + "epoch": 2.93, + "learning_rate": 3.541293423639939e-06, + "loss": 1.6866, + "step": 13403 + }, + { + "epoch": 2.93, + "learning_rate": 3.5411841817784578e-06, + "loss": 1.3685, + "step": 13404 + }, + { + "epoch": 2.93, + "learning_rate": 3.5410749399169764e-06, + "loss": 2.5857, + "step": 13405 + }, + { + "epoch": 2.93, + "learning_rate": 3.540965698055495e-06, + "loss": 1.4431, + "step": 13406 + }, + { + "epoch": 2.93, + "learning_rate": 3.540856456194014e-06, + "loss": 3.5067, + "step": 13407 + }, + { + "epoch": 2.93, + "learning_rate": 3.5407472143325326e-06, + "loss": 1.8196, + "step": 13408 + }, + { + "epoch": 2.93, + "learning_rate": 3.540637972471051e-06, + "loss": 0.8556, + "step": 13409 + }, + { + "epoch": 2.93, + "learning_rate": 3.54052873060957e-06, + "loss": 2.7867, + "step": 13410 + }, + { + "epoch": 2.93, + "learning_rate": 3.5404194887480885e-06, + "loss": 3.2572, + "step": 13411 + }, + { + "epoch": 2.93, + "learning_rate": 3.5403102468866075e-06, + "loss": 3.4703, + "step": 13412 + }, + { + "epoch": 2.93, + "learning_rate": 3.5402010050251257e-06, + "loss": 2.6507, + "step": 13413 + }, + { + "epoch": 2.93, + "learning_rate": 3.5400917631636443e-06, + "loss": 2.9112, + "step": 13414 + }, + { + "epoch": 2.93, + "learning_rate": 3.5399825213021633e-06, + "loss": 2.8561, + "step": 13415 + }, + { + "epoch": 2.93, + "learning_rate": 3.539873279440682e-06, + "loss": 2.4803, + "step": 13416 + }, + { + "epoch": 2.93, + "learning_rate": 3.539764037579201e-06, + "loss": 2.4185, + "step": 13417 + }, + { + "epoch": 2.93, + "learning_rate": 3.539654795717719e-06, + "loss": 2.6198, + "step": 13418 + }, + { + "epoch": 2.93, + "learning_rate": 3.5395455538562378e-06, + "loss": 1.721, + "step": 13419 + }, + { + "epoch": 2.93, + "learning_rate": 3.539436311994757e-06, + "loss": 2.8378, + "step": 13420 + }, + { + "epoch": 2.93, + "learning_rate": 3.5393270701332754e-06, + "loss": 1.8557, + "step": 13421 + }, + { + "epoch": 2.93, + "learning_rate": 3.5392178282717945e-06, + "loss": 2.5068, + "step": 13422 + }, + { + "epoch": 2.93, + "learning_rate": 3.5391085864103126e-06, + "loss": 2.6982, + "step": 13423 + }, + { + "epoch": 2.93, + "learning_rate": 3.5389993445488313e-06, + "loss": 2.8729, + "step": 13424 + }, + { + "epoch": 2.93, + "learning_rate": 3.5388901026873503e-06, + "loss": 0.9849, + "step": 13425 + }, + { + "epoch": 2.93, + "learning_rate": 3.538780860825869e-06, + "loss": 3.2729, + "step": 13426 + }, + { + "epoch": 2.93, + "learning_rate": 3.5386716189643875e-06, + "loss": 2.9049, + "step": 13427 + }, + { + "epoch": 2.93, + "learning_rate": 3.538562377102906e-06, + "loss": 2.5318, + "step": 13428 + }, + { + "epoch": 2.93, + "learning_rate": 3.5384531352414247e-06, + "loss": 2.1631, + "step": 13429 + }, + { + "epoch": 2.93, + "learning_rate": 3.5383438933799434e-06, + "loss": 1.9711, + "step": 13430 + }, + { + "epoch": 2.93, + "learning_rate": 3.5382346515184624e-06, + "loss": 1.9384, + "step": 13431 + }, + { + "epoch": 2.93, + "learning_rate": 3.5381254096569806e-06, + "loss": 1.8486, + "step": 13432 + }, + { + "epoch": 2.93, + "learning_rate": 3.538016167795499e-06, + "loss": 1.6873, + "step": 13433 + }, + { + "epoch": 2.93, + "learning_rate": 3.5379069259340182e-06, + "loss": 1.4282, + "step": 13434 + }, + { + "epoch": 2.93, + "learning_rate": 3.537797684072537e-06, + "loss": 2.4952, + "step": 13435 + }, + { + "epoch": 2.93, + "learning_rate": 3.537688442211056e-06, + "loss": 2.077, + "step": 13436 + }, + { + "epoch": 2.93, + "learning_rate": 3.537579200349574e-06, + "loss": 2.6066, + "step": 13437 + }, + { + "epoch": 2.93, + "learning_rate": 3.5374699584880927e-06, + "loss": 1.4957, + "step": 13438 + }, + { + "epoch": 2.93, + "learning_rate": 3.5373607166266117e-06, + "loss": 2.1042, + "step": 13439 + }, + { + "epoch": 2.93, + "learning_rate": 3.5372514747651303e-06, + "loss": 2.654, + "step": 13440 + }, + { + "epoch": 2.93, + "learning_rate": 3.5371422329036493e-06, + "loss": 2.7113, + "step": 13441 + }, + { + "epoch": 2.93, + "learning_rate": 3.5370329910421675e-06, + "loss": 2.1468, + "step": 13442 + }, + { + "epoch": 2.93, + "learning_rate": 3.536923749180686e-06, + "loss": 1.7233, + "step": 13443 + }, + { + "epoch": 2.93, + "learning_rate": 3.536814507319205e-06, + "loss": 1.4748, + "step": 13444 + }, + { + "epoch": 2.93, + "learning_rate": 3.5367052654577238e-06, + "loss": 3.7496, + "step": 13445 + }, + { + "epoch": 2.93, + "learning_rate": 3.536596023596242e-06, + "loss": 3.4381, + "step": 13446 + }, + { + "epoch": 2.93, + "learning_rate": 3.536486781734761e-06, + "loss": 2.8964, + "step": 13447 + }, + { + "epoch": 2.93, + "learning_rate": 3.5363775398732796e-06, + "loss": 2.6073, + "step": 13448 + }, + { + "epoch": 2.94, + "learning_rate": 3.5362682980117987e-06, + "loss": 2.409, + "step": 13449 + }, + { + "epoch": 2.94, + "learning_rate": 3.5361590561503173e-06, + "loss": 1.9333, + "step": 13450 + }, + { + "epoch": 2.94, + "learning_rate": 3.5360498142888355e-06, + "loss": 2.2867, + "step": 13451 + }, + { + "epoch": 2.94, + "learning_rate": 3.5359405724273545e-06, + "loss": 2.8468, + "step": 13452 + }, + { + "epoch": 2.94, + "learning_rate": 3.535831330565873e-06, + "loss": 2.3967, + "step": 13453 + }, + { + "epoch": 2.94, + "learning_rate": 3.535722088704392e-06, + "loss": 1.6137, + "step": 13454 + }, + { + "epoch": 2.94, + "learning_rate": 3.5356128468429107e-06, + "loss": 1.6661, + "step": 13455 + }, + { + "epoch": 2.94, + "learning_rate": 3.535503604981429e-06, + "loss": 1.9785, + "step": 13456 + }, + { + "epoch": 2.94, + "learning_rate": 3.535394363119948e-06, + "loss": 2.2686, + "step": 13457 + }, + { + "epoch": 2.94, + "learning_rate": 3.5352851212584666e-06, + "loss": 2.3297, + "step": 13458 + }, + { + "epoch": 2.94, + "learning_rate": 3.535175879396985e-06, + "loss": 1.5185, + "step": 13459 + }, + { + "epoch": 2.94, + "learning_rate": 3.5350666375355042e-06, + "loss": 2.2054, + "step": 13460 + }, + { + "epoch": 2.94, + "learning_rate": 3.5349573956740224e-06, + "loss": 3.4874, + "step": 13461 + }, + { + "epoch": 2.94, + "learning_rate": 3.534848153812541e-06, + "loss": 1.6864, + "step": 13462 + }, + { + "epoch": 2.94, + "learning_rate": 3.53473891195106e-06, + "loss": 3.0897, + "step": 13463 + }, + { + "epoch": 2.94, + "learning_rate": 3.5346296700895787e-06, + "loss": 1.4249, + "step": 13464 + }, + { + "epoch": 2.94, + "learning_rate": 3.534520428228097e-06, + "loss": 1.9146, + "step": 13465 + }, + { + "epoch": 2.94, + "learning_rate": 3.534411186366616e-06, + "loss": 1.796, + "step": 13466 + }, + { + "epoch": 2.94, + "learning_rate": 3.5343019445051345e-06, + "loss": 2.0672, + "step": 13467 + }, + { + "epoch": 2.94, + "learning_rate": 3.5341927026436535e-06, + "loss": 2.358, + "step": 13468 + }, + { + "epoch": 2.94, + "learning_rate": 3.534083460782172e-06, + "loss": 1.2798, + "step": 13469 + }, + { + "epoch": 2.94, + "learning_rate": 3.5339742189206903e-06, + "loss": 1.0528, + "step": 13470 + }, + { + "epoch": 2.94, + "learning_rate": 3.5338649770592094e-06, + "loss": 2.6191, + "step": 13471 + }, + { + "epoch": 2.94, + "learning_rate": 3.533755735197728e-06, + "loss": 1.8927, + "step": 13472 + }, + { + "epoch": 2.94, + "learning_rate": 3.533646493336247e-06, + "loss": 2.0794, + "step": 13473 + }, + { + "epoch": 2.94, + "learning_rate": 3.5335372514747656e-06, + "loss": 3.2042, + "step": 13474 + }, + { + "epoch": 2.94, + "learning_rate": 3.533428009613284e-06, + "loss": 2.0779, + "step": 13475 + }, + { + "epoch": 2.94, + "learning_rate": 3.533318767751803e-06, + "loss": 0.9468, + "step": 13476 + }, + { + "epoch": 2.94, + "learning_rate": 3.5332095258903215e-06, + "loss": 3.0861, + "step": 13477 + }, + { + "epoch": 2.94, + "learning_rate": 3.5331002840288405e-06, + "loss": 2.601, + "step": 13478 + }, + { + "epoch": 2.94, + "learning_rate": 3.5329910421673587e-06, + "loss": 2.8585, + "step": 13479 + }, + { + "epoch": 2.94, + "learning_rate": 3.5328818003058773e-06, + "loss": 2.6028, + "step": 13480 + }, + { + "epoch": 2.94, + "learning_rate": 3.5327725584443963e-06, + "loss": 1.2681, + "step": 13481 + }, + { + "epoch": 2.94, + "learning_rate": 3.532663316582915e-06, + "loss": 2.5189, + "step": 13482 + }, + { + "epoch": 2.94, + "learning_rate": 3.5325540747214335e-06, + "loss": 1.8811, + "step": 13483 + }, + { + "epoch": 2.94, + "learning_rate": 3.532444832859952e-06, + "loss": 1.6792, + "step": 13484 + }, + { + "epoch": 2.94, + "learning_rate": 3.5323355909984708e-06, + "loss": 1.7492, + "step": 13485 + }, + { + "epoch": 2.94, + "learning_rate": 3.5322263491369894e-06, + "loss": 2.0978, + "step": 13486 + }, + { + "epoch": 2.94, + "learning_rate": 3.5321171072755084e-06, + "loss": 3.2587, + "step": 13487 + }, + { + "epoch": 2.94, + "learning_rate": 3.532007865414027e-06, + "loss": 1.8682, + "step": 13488 + }, + { + "epoch": 2.94, + "learning_rate": 3.5318986235525456e-06, + "loss": 2.627, + "step": 13489 + }, + { + "epoch": 2.94, + "learning_rate": 3.5317893816910642e-06, + "loss": 2.1466, + "step": 13490 + }, + { + "epoch": 2.94, + "learning_rate": 3.531680139829583e-06, + "loss": 2.6934, + "step": 13491 + }, + { + "epoch": 2.94, + "learning_rate": 3.531570897968102e-06, + "loss": 2.1864, + "step": 13492 + }, + { + "epoch": 2.94, + "learning_rate": 3.5314616561066205e-06, + "loss": 3.1702, + "step": 13493 + }, + { + "epoch": 2.95, + "learning_rate": 3.5313524142451387e-06, + "loss": 3.0091, + "step": 13494 + }, + { + "epoch": 2.95, + "learning_rate": 3.5312431723836577e-06, + "loss": 2.0662, + "step": 13495 + }, + { + "epoch": 2.95, + "learning_rate": 3.5311339305221763e-06, + "loss": 2.5046, + "step": 13496 + }, + { + "epoch": 2.95, + "learning_rate": 3.5310246886606954e-06, + "loss": 2.4242, + "step": 13497 + }, + { + "epoch": 2.95, + "learning_rate": 3.5309154467992136e-06, + "loss": 2.6324, + "step": 13498 + }, + { + "epoch": 2.95, + "learning_rate": 3.530806204937732e-06, + "loss": 2.6786, + "step": 13499 + }, + { + "epoch": 2.95, + "learning_rate": 3.530696963076251e-06, + "loss": 2.1942, + "step": 13500 + }, + { + "epoch": 2.95, + "learning_rate": 3.53058772121477e-06, + "loss": 1.3117, + "step": 13501 + }, + { + "epoch": 2.95, + "learning_rate": 3.530478479353289e-06, + "loss": 1.5809, + "step": 13502 + }, + { + "epoch": 2.95, + "learning_rate": 3.530369237491807e-06, + "loss": 2.8175, + "step": 13503 + }, + { + "epoch": 2.95, + "learning_rate": 3.5302599956303256e-06, + "loss": 1.1358, + "step": 13504 + }, + { + "epoch": 2.95, + "learning_rate": 3.5301507537688447e-06, + "loss": 1.6835, + "step": 13505 + }, + { + "epoch": 2.95, + "learning_rate": 3.5300415119073633e-06, + "loss": 2.1484, + "step": 13506 + }, + { + "epoch": 2.95, + "learning_rate": 3.5299322700458823e-06, + "loss": 1.8553, + "step": 13507 + }, + { + "epoch": 2.95, + "learning_rate": 3.5298230281844005e-06, + "loss": 2.1182, + "step": 13508 + }, + { + "epoch": 2.95, + "learning_rate": 3.529713786322919e-06, + "loss": 2.9531, + "step": 13509 + }, + { + "epoch": 2.95, + "learning_rate": 3.529604544461438e-06, + "loss": 2.9985, + "step": 13510 + }, + { + "epoch": 2.95, + "learning_rate": 3.5294953025999568e-06, + "loss": 1.5768, + "step": 13511 + }, + { + "epoch": 2.95, + "learning_rate": 3.529386060738475e-06, + "loss": 2.4714, + "step": 13512 + }, + { + "epoch": 2.95, + "learning_rate": 3.529276818876994e-06, + "loss": 2.829, + "step": 13513 + }, + { + "epoch": 2.95, + "learning_rate": 3.5291675770155126e-06, + "loss": 0.9946, + "step": 13514 + }, + { + "epoch": 2.95, + "learning_rate": 3.5290583351540312e-06, + "loss": 2.6433, + "step": 13515 + }, + { + "epoch": 2.95, + "learning_rate": 3.5289490932925502e-06, + "loss": 2.116, + "step": 13516 + }, + { + "epoch": 2.95, + "learning_rate": 3.5288398514310684e-06, + "loss": 1.4588, + "step": 13517 + }, + { + "epoch": 2.95, + "learning_rate": 3.528730609569587e-06, + "loss": 3.6762, + "step": 13518 + }, + { + "epoch": 2.95, + "learning_rate": 3.528621367708106e-06, + "loss": 1.4109, + "step": 13519 + }, + { + "epoch": 2.95, + "learning_rate": 3.5285121258466247e-06, + "loss": 2.2304, + "step": 13520 + }, + { + "epoch": 2.95, + "learning_rate": 3.5284028839851437e-06, + "loss": 3.2978, + "step": 13521 + }, + { + "epoch": 2.95, + "learning_rate": 3.528293642123662e-06, + "loss": 2.149, + "step": 13522 + }, + { + "epoch": 2.95, + "learning_rate": 3.5281844002621805e-06, + "loss": 2.6071, + "step": 13523 + }, + { + "epoch": 2.95, + "learning_rate": 3.5280751584006996e-06, + "loss": 1.4304, + "step": 13524 + }, + { + "epoch": 2.95, + "learning_rate": 3.527965916539218e-06, + "loss": 3.3424, + "step": 13525 + }, + { + "epoch": 2.95, + "learning_rate": 3.527856674677737e-06, + "loss": 1.9716, + "step": 13526 + }, + { + "epoch": 2.95, + "learning_rate": 3.5277474328162554e-06, + "loss": 3.0674, + "step": 13527 + }, + { + "epoch": 2.95, + "learning_rate": 3.527638190954774e-06, + "loss": 1.7878, + "step": 13528 + }, + { + "epoch": 2.95, + "learning_rate": 3.527528949093293e-06, + "loss": 2.2307, + "step": 13529 + }, + { + "epoch": 2.95, + "learning_rate": 3.5274197072318116e-06, + "loss": 3.0637, + "step": 13530 + }, + { + "epoch": 2.95, + "learning_rate": 3.52731046537033e-06, + "loss": 3.1724, + "step": 13531 + }, + { + "epoch": 2.95, + "learning_rate": 3.527201223508849e-06, + "loss": 2.2264, + "step": 13532 + }, + { + "epoch": 2.95, + "learning_rate": 3.5270919816473675e-06, + "loss": 3.0393, + "step": 13533 + }, + { + "epoch": 2.95, + "learning_rate": 3.5269827397858865e-06, + "loss": 3.43, + "step": 13534 + }, + { + "epoch": 2.95, + "learning_rate": 3.526873497924405e-06, + "loss": 2.2869, + "step": 13535 + }, + { + "epoch": 2.95, + "learning_rate": 3.5267642560629233e-06, + "loss": 0.839, + "step": 13536 + }, + { + "epoch": 2.95, + "learning_rate": 3.5266550142014423e-06, + "loss": 2.4025, + "step": 13537 + }, + { + "epoch": 2.95, + "learning_rate": 3.526545772339961e-06, + "loss": 1.9696, + "step": 13538 + }, + { + "epoch": 2.95, + "learning_rate": 3.5264365304784796e-06, + "loss": 2.4191, + "step": 13539 + }, + { + "epoch": 2.96, + "learning_rate": 3.5263272886169986e-06, + "loss": 2.2504, + "step": 13540 + }, + { + "epoch": 2.96, + "learning_rate": 3.526218046755517e-06, + "loss": 2.2214, + "step": 13541 + }, + { + "epoch": 2.96, + "learning_rate": 3.526108804894036e-06, + "loss": 2.5975, + "step": 13542 + }, + { + "epoch": 2.96, + "learning_rate": 3.5259995630325544e-06, + "loss": 2.2432, + "step": 13543 + }, + { + "epoch": 2.96, + "learning_rate": 3.525890321171073e-06, + "loss": 1.1799, + "step": 13544 + }, + { + "epoch": 2.96, + "learning_rate": 3.5257810793095917e-06, + "loss": 2.3633, + "step": 13545 + }, + { + "epoch": 2.96, + "learning_rate": 3.5256718374481103e-06, + "loss": 2.6994, + "step": 13546 + }, + { + "epoch": 2.96, + "learning_rate": 3.525562595586629e-06, + "loss": 2.8125, + "step": 13547 + }, + { + "epoch": 2.96, + "learning_rate": 3.525453353725148e-06, + "loss": 3.3239, + "step": 13548 + }, + { + "epoch": 2.96, + "learning_rate": 3.5253441118636665e-06, + "loss": 1.5402, + "step": 13549 + }, + { + "epoch": 2.96, + "learning_rate": 3.5252348700021847e-06, + "loss": 2.6188, + "step": 13550 + }, + { + "epoch": 2.96, + "learning_rate": 3.5251256281407038e-06, + "loss": 1.8074, + "step": 13551 + }, + { + "epoch": 2.96, + "learning_rate": 3.5250163862792224e-06, + "loss": 2.2843, + "step": 13552 + }, + { + "epoch": 2.96, + "learning_rate": 3.5249071444177414e-06, + "loss": 1.4083, + "step": 13553 + }, + { + "epoch": 2.96, + "learning_rate": 3.52479790255626e-06, + "loss": 1.7759, + "step": 13554 + }, + { + "epoch": 2.96, + "learning_rate": 3.524688660694778e-06, + "loss": 1.8081, + "step": 13555 + }, + { + "epoch": 2.96, + "learning_rate": 3.5245794188332972e-06, + "loss": 1.4635, + "step": 13556 + }, + { + "epoch": 2.96, + "learning_rate": 3.524470176971816e-06, + "loss": 2.9968, + "step": 13557 + }, + { + "epoch": 2.96, + "learning_rate": 3.524360935110335e-06, + "loss": 2.7443, + "step": 13558 + }, + { + "epoch": 2.96, + "learning_rate": 3.5242516932488535e-06, + "loss": 1.1767, + "step": 13559 + }, + { + "epoch": 2.96, + "learning_rate": 3.5241424513873717e-06, + "loss": 2.8707, + "step": 13560 + }, + { + "epoch": 2.96, + "learning_rate": 3.5240332095258907e-06, + "loss": 2.0389, + "step": 13561 + }, + { + "epoch": 2.96, + "learning_rate": 3.5239239676644093e-06, + "loss": 2.3758, + "step": 13562 + }, + { + "epoch": 2.96, + "learning_rate": 3.5238147258029284e-06, + "loss": 2.7049, + "step": 13563 + }, + { + "epoch": 2.96, + "learning_rate": 3.5237054839414465e-06, + "loss": 1.4686, + "step": 13564 + }, + { + "epoch": 2.96, + "learning_rate": 3.523596242079965e-06, + "loss": 1.5736, + "step": 13565 + }, + { + "epoch": 2.96, + "learning_rate": 3.523487000218484e-06, + "loss": 2.6642, + "step": 13566 + }, + { + "epoch": 2.96, + "learning_rate": 3.523377758357003e-06, + "loss": 2.5273, + "step": 13567 + }, + { + "epoch": 2.96, + "learning_rate": 3.5232685164955214e-06, + "loss": 2.2495, + "step": 13568 + }, + { + "epoch": 2.96, + "learning_rate": 3.52315927463404e-06, + "loss": 1.4207, + "step": 13569 + }, + { + "epoch": 2.96, + "learning_rate": 3.5230500327725586e-06, + "loss": 2.2107, + "step": 13570 + }, + { + "epoch": 2.96, + "learning_rate": 3.5229407909110772e-06, + "loss": 1.8551, + "step": 13571 + }, + { + "epoch": 2.96, + "learning_rate": 3.5228315490495963e-06, + "loss": 1.1289, + "step": 13572 + }, + { + "epoch": 2.96, + "learning_rate": 3.522722307188115e-06, + "loss": 2.2105, + "step": 13573 + }, + { + "epoch": 2.96, + "learning_rate": 3.522613065326633e-06, + "loss": 1.9234, + "step": 13574 + }, + { + "epoch": 2.96, + "learning_rate": 3.522503823465152e-06, + "loss": 2.9524, + "step": 13575 + }, + { + "epoch": 2.96, + "learning_rate": 3.5223945816036707e-06, + "loss": 3.0691, + "step": 13576 + }, + { + "epoch": 2.96, + "learning_rate": 3.5222853397421898e-06, + "loss": 1.7762, + "step": 13577 + }, + { + "epoch": 2.96, + "learning_rate": 3.522176097880708e-06, + "loss": 2.139, + "step": 13578 + }, + { + "epoch": 2.96, + "learning_rate": 3.5220668560192266e-06, + "loss": 1.2163, + "step": 13579 + }, + { + "epoch": 2.96, + "learning_rate": 3.5219576141577456e-06, + "loss": 2.5679, + "step": 13580 + }, + { + "epoch": 2.96, + "learning_rate": 3.521848372296264e-06, + "loss": 2.4404, + "step": 13581 + }, + { + "epoch": 2.96, + "learning_rate": 3.5217391304347832e-06, + "loss": 2.1678, + "step": 13582 + }, + { + "epoch": 2.96, + "learning_rate": 3.5216298885733014e-06, + "loss": 1.4889, + "step": 13583 + }, + { + "epoch": 2.96, + "learning_rate": 3.52152064671182e-06, + "loss": 2.6205, + "step": 13584 + }, + { + "epoch": 2.96, + "learning_rate": 3.521411404850339e-06, + "loss": 1.4386, + "step": 13585 + }, + { + "epoch": 2.97, + "learning_rate": 3.5213021629888577e-06, + "loss": 1.5619, + "step": 13586 + }, + { + "epoch": 2.97, + "learning_rate": 3.5211929211273767e-06, + "loss": 1.376, + "step": 13587 + }, + { + "epoch": 2.97, + "learning_rate": 3.521083679265895e-06, + "loss": 1.5808, + "step": 13588 + }, + { + "epoch": 2.97, + "learning_rate": 3.5209744374044135e-06, + "loss": 2.8971, + "step": 13589 + }, + { + "epoch": 2.97, + "learning_rate": 3.5208651955429325e-06, + "loss": 2.1274, + "step": 13590 + }, + { + "epoch": 2.97, + "learning_rate": 3.520755953681451e-06, + "loss": 2.0473, + "step": 13591 + }, + { + "epoch": 2.97, + "learning_rate": 3.5206467118199698e-06, + "loss": 2.9741, + "step": 13592 + }, + { + "epoch": 2.97, + "learning_rate": 3.5205374699584884e-06, + "loss": 0.8059, + "step": 13593 + }, + { + "epoch": 2.97, + "learning_rate": 3.520428228097007e-06, + "loss": 2.0961, + "step": 13594 + }, + { + "epoch": 2.97, + "learning_rate": 3.520318986235526e-06, + "loss": 1.0728, + "step": 13595 + }, + { + "epoch": 2.97, + "learning_rate": 3.5202097443740446e-06, + "loss": 2.515, + "step": 13596 + }, + { + "epoch": 2.97, + "learning_rate": 3.520100502512563e-06, + "loss": 1.8913, + "step": 13597 + }, + { + "epoch": 2.97, + "learning_rate": 3.519991260651082e-06, + "loss": 3.3633, + "step": 13598 + }, + { + "epoch": 2.97, + "learning_rate": 3.5198820187896005e-06, + "loss": 1.2393, + "step": 13599 + }, + { + "epoch": 2.97, + "learning_rate": 3.519772776928119e-06, + "loss": 1.3255, + "step": 13600 + }, + { + "epoch": 2.97, + "learning_rate": 3.519663535066638e-06, + "loss": 2.8848, + "step": 13601 + }, + { + "epoch": 2.97, + "learning_rate": 3.5195542932051563e-06, + "loss": 2.2694, + "step": 13602 + }, + { + "epoch": 2.97, + "learning_rate": 3.519445051343675e-06, + "loss": 2.6952, + "step": 13603 + }, + { + "epoch": 2.97, + "learning_rate": 3.519335809482194e-06, + "loss": 2.6313, + "step": 13604 + }, + { + "epoch": 2.97, + "learning_rate": 3.5192265676207126e-06, + "loss": 1.1226, + "step": 13605 + }, + { + "epoch": 2.97, + "learning_rate": 3.5191173257592316e-06, + "loss": 1.8448, + "step": 13606 + }, + { + "epoch": 2.97, + "learning_rate": 3.5190080838977498e-06, + "loss": 2.0512, + "step": 13607 + }, + { + "epoch": 2.97, + "learning_rate": 3.5188988420362684e-06, + "loss": 2.0633, + "step": 13608 + }, + { + "epoch": 2.97, + "learning_rate": 3.5187896001747874e-06, + "loss": 2.9899, + "step": 13609 + }, + { + "epoch": 2.97, + "learning_rate": 3.518680358313306e-06, + "loss": 2.0351, + "step": 13610 + }, + { + "epoch": 2.97, + "learning_rate": 3.5185711164518242e-06, + "loss": 2.9346, + "step": 13611 + }, + { + "epoch": 2.97, + "learning_rate": 3.5184618745903433e-06, + "loss": 2.0508, + "step": 13612 + }, + { + "epoch": 2.97, + "learning_rate": 3.518352632728862e-06, + "loss": 2.3913, + "step": 13613 + }, + { + "epoch": 2.97, + "learning_rate": 3.518243390867381e-06, + "loss": 1.9537, + "step": 13614 + }, + { + "epoch": 2.97, + "learning_rate": 3.5181341490058995e-06, + "loss": 1.2082, + "step": 13615 + }, + { + "epoch": 2.97, + "learning_rate": 3.5180249071444177e-06, + "loss": 1.8588, + "step": 13616 + }, + { + "epoch": 2.97, + "learning_rate": 3.5179156652829367e-06, + "loss": 2.524, + "step": 13617 + }, + { + "epoch": 2.97, + "learning_rate": 3.5178064234214553e-06, + "loss": 1.0046, + "step": 13618 + }, + { + "epoch": 2.97, + "learning_rate": 3.5176971815599744e-06, + "loss": 3.1909, + "step": 13619 + }, + { + "epoch": 2.97, + "learning_rate": 3.517587939698493e-06, + "loss": 1.8567, + "step": 13620 + }, + { + "epoch": 2.97, + "learning_rate": 3.517478697837011e-06, + "loss": 3.6241, + "step": 13621 + }, + { + "epoch": 2.97, + "learning_rate": 3.5173694559755302e-06, + "loss": 1.4013, + "step": 13622 + }, + { + "epoch": 2.97, + "learning_rate": 3.517260214114049e-06, + "loss": 1.3509, + "step": 13623 + }, + { + "epoch": 2.97, + "learning_rate": 3.5171509722525674e-06, + "loss": 1.7917, + "step": 13624 + }, + { + "epoch": 2.97, + "learning_rate": 3.5170417303910865e-06, + "loss": 2.5202, + "step": 13625 + }, + { + "epoch": 2.97, + "learning_rate": 3.5169324885296047e-06, + "loss": 2.5313, + "step": 13626 + }, + { + "epoch": 2.97, + "learning_rate": 3.5168232466681233e-06, + "loss": 0.644, + "step": 13627 + }, + { + "epoch": 2.97, + "learning_rate": 3.5167140048066423e-06, + "loss": 2.4405, + "step": 13628 + }, + { + "epoch": 2.97, + "learning_rate": 3.516604762945161e-06, + "loss": 2.9629, + "step": 13629 + }, + { + "epoch": 2.97, + "learning_rate": 3.516495521083679e-06, + "loss": 2.6905, + "step": 13630 + }, + { + "epoch": 2.97, + "learning_rate": 3.516386279222198e-06, + "loss": 2.6534, + "step": 13631 + }, + { + "epoch": 2.98, + "learning_rate": 3.5162770373607167e-06, + "loss": 1.0927, + "step": 13632 + }, + { + "epoch": 2.98, + "learning_rate": 3.5161677954992358e-06, + "loss": 2.367, + "step": 13633 + }, + { + "epoch": 2.98, + "learning_rate": 3.5160585536377544e-06, + "loss": 2.1343, + "step": 13634 + }, + { + "epoch": 2.98, + "learning_rate": 3.5159493117762726e-06, + "loss": 1.86, + "step": 13635 + }, + { + "epoch": 2.98, + "learning_rate": 3.5158400699147916e-06, + "loss": 2.3096, + "step": 13636 + }, + { + "epoch": 2.98, + "learning_rate": 3.5157308280533102e-06, + "loss": 2.5646, + "step": 13637 + }, + { + "epoch": 2.98, + "learning_rate": 3.5156215861918293e-06, + "loss": 2.2862, + "step": 13638 + }, + { + "epoch": 2.98, + "learning_rate": 3.515512344330348e-06, + "loss": 1.5712, + "step": 13639 + }, + { + "epoch": 2.98, + "learning_rate": 3.515403102468866e-06, + "loss": 3.2072, + "step": 13640 + }, + { + "epoch": 2.98, + "learning_rate": 3.515293860607385e-06, + "loss": 4.0114, + "step": 13641 + }, + { + "epoch": 2.98, + "learning_rate": 3.5151846187459037e-06, + "loss": 2.6869, + "step": 13642 + }, + { + "epoch": 2.98, + "learning_rate": 3.5150753768844227e-06, + "loss": 1.0126, + "step": 13643 + }, + { + "epoch": 2.98, + "learning_rate": 3.514966135022941e-06, + "loss": 2.037, + "step": 13644 + }, + { + "epoch": 2.98, + "learning_rate": 3.5148568931614595e-06, + "loss": 1.3445, + "step": 13645 + }, + { + "epoch": 2.98, + "learning_rate": 3.5147476512999786e-06, + "loss": 1.7522, + "step": 13646 + }, + { + "epoch": 2.98, + "learning_rate": 3.514638409438497e-06, + "loss": 2.4648, + "step": 13647 + }, + { + "epoch": 2.98, + "learning_rate": 3.5145291675770162e-06, + "loss": 1.7004, + "step": 13648 + }, + { + "epoch": 2.98, + "learning_rate": 3.5144199257155344e-06, + "loss": 1.308, + "step": 13649 + }, + { + "epoch": 2.98, + "learning_rate": 3.514310683854053e-06, + "loss": 2.0722, + "step": 13650 + }, + { + "epoch": 2.98, + "learning_rate": 3.514201441992572e-06, + "loss": 1.8433, + "step": 13651 + }, + { + "epoch": 2.98, + "learning_rate": 3.5140922001310907e-06, + "loss": 2.7905, + "step": 13652 + }, + { + "epoch": 2.98, + "learning_rate": 3.5139829582696093e-06, + "loss": 2.7844, + "step": 13653 + }, + { + "epoch": 2.98, + "learning_rate": 3.513873716408128e-06, + "loss": 1.765, + "step": 13654 + }, + { + "epoch": 2.98, + "learning_rate": 3.5137644745466465e-06, + "loss": 1.415, + "step": 13655 + }, + { + "epoch": 2.98, + "learning_rate": 3.513655232685165e-06, + "loss": 2.3383, + "step": 13656 + }, + { + "epoch": 2.98, + "learning_rate": 3.513545990823684e-06, + "loss": 3.6578, + "step": 13657 + }, + { + "epoch": 2.98, + "learning_rate": 3.5134367489622023e-06, + "loss": 3.76, + "step": 13658 + }, + { + "epoch": 2.98, + "learning_rate": 3.513327507100721e-06, + "loss": 2.3415, + "step": 13659 + }, + { + "epoch": 2.98, + "learning_rate": 3.51321826523924e-06, + "loss": 2.6852, + "step": 13660 + }, + { + "epoch": 2.98, + "learning_rate": 3.5131090233777586e-06, + "loss": 1.3225, + "step": 13661 + }, + { + "epoch": 2.98, + "learning_rate": 3.5129997815162776e-06, + "loss": 2.9946, + "step": 13662 + }, + { + "epoch": 2.98, + "learning_rate": 3.512890539654796e-06, + "loss": 2.5184, + "step": 13663 + }, + { + "epoch": 2.98, + "learning_rate": 3.5127812977933144e-06, + "loss": 1.837, + "step": 13664 + }, + { + "epoch": 2.98, + "learning_rate": 3.5126720559318334e-06, + "loss": 1.8727, + "step": 13665 + }, + { + "epoch": 2.98, + "learning_rate": 3.512562814070352e-06, + "loss": 2.0711, + "step": 13666 + }, + { + "epoch": 2.98, + "learning_rate": 3.512453572208871e-06, + "loss": 2.0629, + "step": 13667 + }, + { + "epoch": 2.98, + "learning_rate": 3.5123443303473893e-06, + "loss": 3.0222, + "step": 13668 + }, + { + "epoch": 2.98, + "learning_rate": 3.512235088485908e-06, + "loss": 1.2162, + "step": 13669 + }, + { + "epoch": 2.98, + "learning_rate": 3.512125846624427e-06, + "loss": 1.7463, + "step": 13670 + }, + { + "epoch": 2.98, + "learning_rate": 3.5120166047629455e-06, + "loss": 2.7644, + "step": 13671 + }, + { + "epoch": 2.98, + "learning_rate": 3.5119073629014646e-06, + "loss": 0.9205, + "step": 13672 + }, + { + "epoch": 2.98, + "learning_rate": 3.5117981210399828e-06, + "loss": 3.1135, + "step": 13673 + }, + { + "epoch": 2.98, + "learning_rate": 3.5116888791785014e-06, + "loss": 3.2421, + "step": 13674 + }, + { + "epoch": 2.98, + "learning_rate": 3.5115796373170204e-06, + "loss": 2.9195, + "step": 13675 + }, + { + "epoch": 2.98, + "learning_rate": 3.511470395455539e-06, + "loss": 1.0348, + "step": 13676 + }, + { + "epoch": 2.98, + "learning_rate": 3.511361153594057e-06, + "loss": 2.2701, + "step": 13677 + }, + { + "epoch": 2.99, + "learning_rate": 3.5112519117325762e-06, + "loss": 1.8043, + "step": 13678 + }, + { + "epoch": 2.99, + "learning_rate": 3.511142669871095e-06, + "loss": 1.4129, + "step": 13679 + }, + { + "epoch": 2.99, + "learning_rate": 3.5110334280096135e-06, + "loss": 2.3267, + "step": 13680 + }, + { + "epoch": 2.99, + "learning_rate": 3.5109241861481325e-06, + "loss": 1.9668, + "step": 13681 + }, + { + "epoch": 2.99, + "learning_rate": 3.5108149442866507e-06, + "loss": 2.219, + "step": 13682 + }, + { + "epoch": 2.99, + "learning_rate": 3.5107057024251693e-06, + "loss": 2.3101, + "step": 13683 + }, + { + "epoch": 2.99, + "learning_rate": 3.5105964605636883e-06, + "loss": 2.4271, + "step": 13684 + }, + { + "epoch": 2.99, + "learning_rate": 3.510487218702207e-06, + "loss": 1.9123, + "step": 13685 + }, + { + "epoch": 2.99, + "learning_rate": 3.510377976840726e-06, + "loss": 4.1113, + "step": 13686 + }, + { + "epoch": 2.99, + "learning_rate": 3.510268734979244e-06, + "loss": 2.746, + "step": 13687 + }, + { + "epoch": 2.99, + "learning_rate": 3.5101594931177628e-06, + "loss": 2.3579, + "step": 13688 + }, + { + "epoch": 2.99, + "learning_rate": 3.510050251256282e-06, + "loss": 0.6584, + "step": 13689 + }, + { + "epoch": 2.99, + "learning_rate": 3.5099410093948004e-06, + "loss": 1.8506, + "step": 13690 + }, + { + "epoch": 2.99, + "learning_rate": 3.5098317675333186e-06, + "loss": 2.3877, + "step": 13691 + }, + { + "epoch": 2.99, + "learning_rate": 3.5097225256718376e-06, + "loss": 2.0883, + "step": 13692 + }, + { + "epoch": 2.99, + "learning_rate": 3.5096132838103563e-06, + "loss": 2.365, + "step": 13693 + }, + { + "epoch": 2.99, + "learning_rate": 3.5095040419488753e-06, + "loss": 2.0358, + "step": 13694 + }, + { + "epoch": 2.99, + "learning_rate": 3.509394800087394e-06, + "loss": 2.219, + "step": 13695 + }, + { + "epoch": 2.99, + "learning_rate": 3.509285558225912e-06, + "loss": 1.7972, + "step": 13696 + }, + { + "epoch": 2.99, + "learning_rate": 3.509176316364431e-06, + "loss": 2.2117, + "step": 13697 + }, + { + "epoch": 2.99, + "learning_rate": 3.5090670745029497e-06, + "loss": 2.0377, + "step": 13698 + }, + { + "epoch": 2.99, + "learning_rate": 3.5089578326414688e-06, + "loss": 2.1965, + "step": 13699 + }, + { + "epoch": 2.99, + "learning_rate": 3.5088485907799874e-06, + "loss": 2.6425, + "step": 13700 + }, + { + "epoch": 2.99, + "learning_rate": 3.5087393489185056e-06, + "loss": 2.6589, + "step": 13701 + }, + { + "epoch": 2.99, + "learning_rate": 3.5086301070570246e-06, + "loss": 1.9918, + "step": 13702 + }, + { + "epoch": 2.99, + "learning_rate": 3.508520865195543e-06, + "loss": 3.9439, + "step": 13703 + }, + { + "epoch": 2.99, + "learning_rate": 3.5084116233340622e-06, + "loss": 3.521, + "step": 13704 + }, + { + "epoch": 2.99, + "learning_rate": 3.508302381472581e-06, + "loss": 0.8521, + "step": 13705 + }, + { + "epoch": 2.99, + "learning_rate": 3.508193139611099e-06, + "loss": 2.0036, + "step": 13706 + }, + { + "epoch": 2.99, + "learning_rate": 3.508083897749618e-06, + "loss": 1.8661, + "step": 13707 + }, + { + "epoch": 2.99, + "learning_rate": 3.5079746558881367e-06, + "loss": 3.1752, + "step": 13708 + }, + { + "epoch": 2.99, + "learning_rate": 3.5078654140266553e-06, + "loss": 2.3449, + "step": 13709 + }, + { + "epoch": 2.99, + "learning_rate": 3.507756172165174e-06, + "loss": 2.3549, + "step": 13710 + }, + { + "epoch": 2.99, + "learning_rate": 3.5076469303036925e-06, + "loss": 2.6632, + "step": 13711 + }, + { + "epoch": 2.99, + "learning_rate": 3.507537688442211e-06, + "loss": 2.1471, + "step": 13712 + }, + { + "epoch": 2.99, + "learning_rate": 3.50742844658073e-06, + "loss": 1.9694, + "step": 13713 + }, + { + "epoch": 2.99, + "learning_rate": 3.5073192047192488e-06, + "loss": 1.9933, + "step": 13714 + }, + { + "epoch": 2.99, + "learning_rate": 3.507209962857767e-06, + "loss": 1.6928, + "step": 13715 + }, + { + "epoch": 2.99, + "learning_rate": 3.507100720996286e-06, + "loss": 2.7658, + "step": 13716 + }, + { + "epoch": 2.99, + "learning_rate": 3.5069914791348046e-06, + "loss": 1.4966, + "step": 13717 + }, + { + "epoch": 2.99, + "learning_rate": 3.5068822372733236e-06, + "loss": 2.3366, + "step": 13718 + }, + { + "epoch": 2.99, + "learning_rate": 3.5067729954118423e-06, + "loss": 3.4802, + "step": 13719 + }, + { + "epoch": 2.99, + "learning_rate": 3.5066637535503604e-06, + "loss": 2.9561, + "step": 13720 + }, + { + "epoch": 2.99, + "learning_rate": 3.5065545116888795e-06, + "loss": 2.781, + "step": 13721 + }, + { + "epoch": 2.99, + "learning_rate": 3.506445269827398e-06, + "loss": 2.5751, + "step": 13722 + }, + { + "epoch": 2.99, + "learning_rate": 3.506336027965917e-06, + "loss": 1.335, + "step": 13723 + }, + { + "epoch": 3.0, + "learning_rate": 3.5062267861044353e-06, + "loss": 2.5392, + "step": 13724 + }, + { + "epoch": 3.0, + "learning_rate": 3.506117544242954e-06, + "loss": 2.6193, + "step": 13725 + }, + { + "epoch": 3.0, + "learning_rate": 3.506008302381473e-06, + "loss": 1.3108, + "step": 13726 + }, + { + "epoch": 3.0, + "learning_rate": 3.5058990605199916e-06, + "loss": 2.0548, + "step": 13727 + }, + { + "epoch": 3.0, + "learning_rate": 3.5057898186585106e-06, + "loss": 2.1015, + "step": 13728 + }, + { + "epoch": 3.0, + "learning_rate": 3.5056805767970288e-06, + "loss": 1.4877, + "step": 13729 + }, + { + "epoch": 3.0, + "learning_rate": 3.5055713349355474e-06, + "loss": 1.2659, + "step": 13730 + }, + { + "epoch": 3.0, + "learning_rate": 3.5054620930740664e-06, + "loss": 2.1324, + "step": 13731 + }, + { + "epoch": 3.0, + "learning_rate": 3.505352851212585e-06, + "loss": 1.5238, + "step": 13732 + }, + { + "epoch": 3.0, + "learning_rate": 3.5052436093511037e-06, + "loss": 1.554, + "step": 13733 + }, + { + "epoch": 3.0, + "learning_rate": 3.5051343674896223e-06, + "loss": 1.7335, + "step": 13734 + }, + { + "epoch": 3.0, + "learning_rate": 3.505025125628141e-06, + "loss": 1.687, + "step": 13735 + }, + { + "epoch": 3.0, + "learning_rate": 3.5049158837666595e-06, + "loss": 1.9152, + "step": 13736 + }, + { + "epoch": 3.0, + "learning_rate": 3.5048066419051785e-06, + "loss": 0.8026, + "step": 13737 + }, + { + "epoch": 3.0, + "learning_rate": 3.504697400043697e-06, + "loss": 1.8119, + "step": 13738 + }, + { + "epoch": 3.0, + "learning_rate": 3.5045881581822157e-06, + "loss": 2.0043, + "step": 13739 + }, + { + "epoch": 3.0, + "learning_rate": 3.5044789163207344e-06, + "loss": 1.6297, + "step": 13740 + }, + { + "epoch": 3.0, + "learning_rate": 3.504369674459253e-06, + "loss": 3.0301, + "step": 13741 + }, + { + "epoch": 3.0, + "learning_rate": 3.504260432597772e-06, + "loss": 2.6065, + "step": 13742 + }, + { + "epoch": 3.0, + "learning_rate": 3.50415119073629e-06, + "loss": 1.4565, + "step": 13743 + }, + { + "epoch": 3.0, + "learning_rate": 3.504041948874809e-06, + "loss": 2.5315, + "step": 13744 + }, + { + "epoch": 3.0, + "learning_rate": 3.503932707013328e-06, + "loss": 2.1339, + "step": 13745 + }, + { + "epoch": 3.0, + "learning_rate": 3.5038234651518464e-06, + "loss": 2.6564, + "step": 13746 + }, + { + "epoch": 3.0, + "eval_accuracy": 0.47806873977086745, + "eval_loss": 2.014918804168701, + "eval_runtime": 332.0122, + "eval_samples_per_second": 18.403, + "eval_steps_per_second": 4.602, + "step": 13746 + }, + { + "epoch": 3.0, + "learning_rate": 3.5037142232903655e-06, + "loss": 2.6364, + "step": 13747 + }, + { + "epoch": 3.0, + "learning_rate": 3.5036049814288837e-06, + "loss": 2.0441, + "step": 13748 + }, + { + "epoch": 3.0, + "learning_rate": 3.5034957395674023e-06, + "loss": 2.1014, + "step": 13749 + }, + { + "epoch": 3.0, + "learning_rate": 3.5033864977059213e-06, + "loss": 1.6434, + "step": 13750 + }, + { + "epoch": 3.0, + "learning_rate": 3.50327725584444e-06, + "loss": 2.2892, + "step": 13751 + }, + { + "epoch": 3.0, + "learning_rate": 3.503168013982959e-06, + "loss": 2.1947, + "step": 13752 + }, + { + "epoch": 3.0, + "learning_rate": 3.503058772121477e-06, + "loss": 3.0909, + "step": 13753 + }, + { + "epoch": 3.0, + "learning_rate": 3.5029495302599958e-06, + "loss": 2.4596, + "step": 13754 + }, + { + "epoch": 3.0, + "learning_rate": 3.502840288398515e-06, + "loss": 3.5023, + "step": 13755 + }, + { + "epoch": 3.0, + "learning_rate": 3.5027310465370334e-06, + "loss": 0.5155, + "step": 13756 + }, + { + "epoch": 3.0, + "learning_rate": 3.5026218046755516e-06, + "loss": 2.4618, + "step": 13757 + }, + { + "epoch": 3.0, + "learning_rate": 3.5025125628140706e-06, + "loss": 2.1786, + "step": 13758 + }, + { + "epoch": 3.0, + "learning_rate": 3.5024033209525892e-06, + "loss": 1.9006, + "step": 13759 + }, + { + "epoch": 3.0, + "learning_rate": 3.5022940790911083e-06, + "loss": 1.7772, + "step": 13760 + }, + { + "epoch": 3.0, + "learning_rate": 3.502184837229627e-06, + "loss": 2.9609, + "step": 13761 + }, + { + "epoch": 3.0, + "learning_rate": 3.502075595368145e-06, + "loss": 2.0508, + "step": 13762 + }, + { + "epoch": 3.0, + "learning_rate": 3.501966353506664e-06, + "loss": 1.0313, + "step": 13763 + }, + { + "epoch": 3.0, + "learning_rate": 3.5018571116451827e-06, + "loss": 1.4882, + "step": 13764 + }, + { + "epoch": 3.0, + "learning_rate": 3.5017478697837013e-06, + "loss": 0.9598, + "step": 13765 + }, + { + "epoch": 3.0, + "learning_rate": 3.5016386279222204e-06, + "loss": 1.5247, + "step": 13766 + }, + { + "epoch": 3.0, + "learning_rate": 3.5015293860607385e-06, + "loss": 2.3279, + "step": 13767 + }, + { + "epoch": 3.0, + "learning_rate": 3.501420144199257e-06, + "loss": 0.6133, + "step": 13768 + }, + { + "epoch": 3.01, + "learning_rate": 3.501310902337776e-06, + "loss": 1.6532, + "step": 13769 + }, + { + "epoch": 3.01, + "learning_rate": 3.501201660476295e-06, + "loss": 2.5542, + "step": 13770 + }, + { + "epoch": 3.01, + "learning_rate": 3.501092418614814e-06, + "loss": 1.5428, + "step": 13771 + }, + { + "epoch": 3.01, + "learning_rate": 3.500983176753332e-06, + "loss": 2.0086, + "step": 13772 + }, + { + "epoch": 3.01, + "learning_rate": 3.5008739348918506e-06, + "loss": 3.4462, + "step": 13773 + }, + { + "epoch": 3.01, + "learning_rate": 3.5007646930303697e-06, + "loss": 1.661, + "step": 13774 + }, + { + "epoch": 3.01, + "learning_rate": 3.5006554511688883e-06, + "loss": 1.2713, + "step": 13775 + }, + { + "epoch": 3.01, + "learning_rate": 3.5005462093074065e-06, + "loss": 1.7729, + "step": 13776 + }, + { + "epoch": 3.01, + "learning_rate": 3.5004369674459255e-06, + "loss": 1.291, + "step": 13777 + }, + { + "epoch": 3.01, + "learning_rate": 3.500327725584444e-06, + "loss": 2.4133, + "step": 13778 + }, + { + "epoch": 3.01, + "learning_rate": 3.500218483722963e-06, + "loss": 1.7784, + "step": 13779 + }, + { + "epoch": 3.01, + "learning_rate": 3.5001092418614818e-06, + "loss": 2.5278, + "step": 13780 + }, + { + "epoch": 3.01, + "learning_rate": 3.5e-06, + "loss": 3.3575, + "step": 13781 + }, + { + "epoch": 3.01, + "learning_rate": 3.499890758138519e-06, + "loss": 3.769, + "step": 13782 + }, + { + "epoch": 3.01, + "learning_rate": 3.4997815162770376e-06, + "loss": 2.5634, + "step": 13783 + }, + { + "epoch": 3.01, + "learning_rate": 3.4996722744155566e-06, + "loss": 3.1838, + "step": 13784 + }, + { + "epoch": 3.01, + "learning_rate": 3.4995630325540752e-06, + "loss": 1.5784, + "step": 13785 + }, + { + "epoch": 3.01, + "learning_rate": 3.4994537906925934e-06, + "loss": 1.8834, + "step": 13786 + }, + { + "epoch": 3.01, + "learning_rate": 3.4993445488311125e-06, + "loss": 1.6285, + "step": 13787 + }, + { + "epoch": 3.01, + "learning_rate": 3.499235306969631e-06, + "loss": 1.8645, + "step": 13788 + }, + { + "epoch": 3.01, + "learning_rate": 3.49912606510815e-06, + "loss": 1.8137, + "step": 13789 + }, + { + "epoch": 3.01, + "learning_rate": 3.4990168232466683e-06, + "loss": 2.0141, + "step": 13790 + }, + { + "epoch": 3.01, + "learning_rate": 3.498907581385187e-06, + "loss": 3.3112, + "step": 13791 + }, + { + "epoch": 3.01, + "learning_rate": 3.498798339523706e-06, + "loss": 2.5024, + "step": 13792 + }, + { + "epoch": 3.01, + "learning_rate": 3.4986890976622245e-06, + "loss": 1.4845, + "step": 13793 + }, + { + "epoch": 3.01, + "learning_rate": 3.498579855800743e-06, + "loss": 0.6987, + "step": 13794 + }, + { + "epoch": 3.01, + "learning_rate": 3.4984706139392618e-06, + "loss": 2.4536, + "step": 13795 + }, + { + "epoch": 3.01, + "learning_rate": 3.4983613720777804e-06, + "loss": 2.8834, + "step": 13796 + }, + { + "epoch": 3.01, + "learning_rate": 3.498252130216299e-06, + "loss": 1.2222, + "step": 13797 + }, + { + "epoch": 3.01, + "learning_rate": 3.498142888354818e-06, + "loss": 0.9092, + "step": 13798 + }, + { + "epoch": 3.01, + "learning_rate": 3.4980336464933366e-06, + "loss": 2.8428, + "step": 13799 + }, + { + "epoch": 3.01, + "learning_rate": 3.497924404631855e-06, + "loss": 2.3146, + "step": 13800 + }, + { + "epoch": 3.01, + "learning_rate": 3.497815162770374e-06, + "loss": 2.0107, + "step": 13801 + }, + { + "epoch": 3.01, + "learning_rate": 3.4977059209088925e-06, + "loss": 2.4116, + "step": 13802 + }, + { + "epoch": 3.01, + "learning_rate": 3.4975966790474115e-06, + "loss": 2.139, + "step": 13803 + }, + { + "epoch": 3.01, + "learning_rate": 3.49748743718593e-06, + "loss": 1.4871, + "step": 13804 + }, + { + "epoch": 3.01, + "learning_rate": 3.4973781953244483e-06, + "loss": 2.4902, + "step": 13805 + }, + { + "epoch": 3.01, + "learning_rate": 3.4972689534629673e-06, + "loss": 2.637, + "step": 13806 + }, + { + "epoch": 3.01, + "learning_rate": 3.497159711601486e-06, + "loss": 1.5016, + "step": 13807 + }, + { + "epoch": 3.01, + "learning_rate": 3.497050469740005e-06, + "loss": 1.75, + "step": 13808 + }, + { + "epoch": 3.01, + "learning_rate": 3.496941227878523e-06, + "loss": 2.5041, + "step": 13809 + }, + { + "epoch": 3.01, + "learning_rate": 3.4968319860170418e-06, + "loss": 2.0155, + "step": 13810 + }, + { + "epoch": 3.01, + "learning_rate": 3.496722744155561e-06, + "loss": 1.6389, + "step": 13811 + }, + { + "epoch": 3.01, + "learning_rate": 3.4966135022940794e-06, + "loss": 1.3282, + "step": 13812 + }, + { + "epoch": 3.01, + "learning_rate": 3.4965042604325985e-06, + "loss": 1.3707, + "step": 13813 + }, + { + "epoch": 3.01, + "learning_rate": 3.4963950185711167e-06, + "loss": 2.0085, + "step": 13814 + }, + { + "epoch": 3.02, + "learning_rate": 3.4962857767096353e-06, + "loss": 1.093, + "step": 13815 + }, + { + "epoch": 3.02, + "learning_rate": 3.4961765348481543e-06, + "loss": 2.1842, + "step": 13816 + }, + { + "epoch": 3.02, + "learning_rate": 3.496067292986673e-06, + "loss": 1.3709, + "step": 13817 + }, + { + "epoch": 3.02, + "learning_rate": 3.4959580511251915e-06, + "loss": 3.0469, + "step": 13818 + }, + { + "epoch": 3.02, + "learning_rate": 3.49584880926371e-06, + "loss": 2.2492, + "step": 13819 + }, + { + "epoch": 3.02, + "learning_rate": 3.4957395674022287e-06, + "loss": 2.1767, + "step": 13820 + }, + { + "epoch": 3.02, + "learning_rate": 3.4956303255407474e-06, + "loss": 2.0362, + "step": 13821 + }, + { + "epoch": 3.02, + "learning_rate": 3.4955210836792664e-06, + "loss": 2.0385, + "step": 13822 + }, + { + "epoch": 3.02, + "learning_rate": 3.4954118418177846e-06, + "loss": 2.136, + "step": 13823 + }, + { + "epoch": 3.02, + "learning_rate": 3.495302599956303e-06, + "loss": 1.9906, + "step": 13824 + }, + { + "epoch": 3.02, + "learning_rate": 3.4951933580948222e-06, + "loss": 1.1217, + "step": 13825 + }, + { + "epoch": 3.02, + "learning_rate": 3.495084116233341e-06, + "loss": 1.4084, + "step": 13826 + }, + { + "epoch": 3.02, + "learning_rate": 3.49497487437186e-06, + "loss": 2.4835, + "step": 13827 + }, + { + "epoch": 3.02, + "learning_rate": 3.494865632510378e-06, + "loss": 1.698, + "step": 13828 + }, + { + "epoch": 3.02, + "learning_rate": 3.4947563906488967e-06, + "loss": 1.8479, + "step": 13829 + }, + { + "epoch": 3.02, + "learning_rate": 3.4946471487874157e-06, + "loss": 0.665, + "step": 13830 + }, + { + "epoch": 3.02, + "learning_rate": 3.4945379069259343e-06, + "loss": 1.3351, + "step": 13831 + }, + { + "epoch": 3.02, + "learning_rate": 3.4944286650644533e-06, + "loss": 2.3812, + "step": 13832 + }, + { + "epoch": 3.02, + "learning_rate": 3.4943194232029715e-06, + "loss": 2.5705, + "step": 13833 + }, + { + "epoch": 3.02, + "learning_rate": 3.49421018134149e-06, + "loss": 1.6135, + "step": 13834 + }, + { + "epoch": 3.02, + "learning_rate": 3.494100939480009e-06, + "loss": 2.1897, + "step": 13835 + }, + { + "epoch": 3.02, + "learning_rate": 3.4939916976185278e-06, + "loss": 2.4067, + "step": 13836 + }, + { + "epoch": 3.02, + "learning_rate": 3.493882455757047e-06, + "loss": 3.1287, + "step": 13837 + }, + { + "epoch": 3.02, + "learning_rate": 3.493773213895565e-06, + "loss": 2.4861, + "step": 13838 + }, + { + "epoch": 3.02, + "learning_rate": 3.4936639720340836e-06, + "loss": 2.6856, + "step": 13839 + }, + { + "epoch": 3.02, + "learning_rate": 3.4935547301726027e-06, + "loss": 2.7004, + "step": 13840 + }, + { + "epoch": 3.02, + "learning_rate": 3.4934454883111213e-06, + "loss": 2.0431, + "step": 13841 + }, + { + "epoch": 3.02, + "learning_rate": 3.4933362464496395e-06, + "loss": 2.9126, + "step": 13842 + }, + { + "epoch": 3.02, + "learning_rate": 3.4932270045881585e-06, + "loss": 1.9052, + "step": 13843 + }, + { + "epoch": 3.02, + "learning_rate": 3.493117762726677e-06, + "loss": 2.2523, + "step": 13844 + }, + { + "epoch": 3.02, + "learning_rate": 3.493008520865196e-06, + "loss": 1.6013, + "step": 13845 + }, + { + "epoch": 3.02, + "learning_rate": 3.4928992790037147e-06, + "loss": 2.767, + "step": 13846 + }, + { + "epoch": 3.02, + "learning_rate": 3.492790037142233e-06, + "loss": 2.7009, + "step": 13847 + }, + { + "epoch": 3.02, + "learning_rate": 3.492680795280752e-06, + "loss": 2.7493, + "step": 13848 + }, + { + "epoch": 3.02, + "learning_rate": 3.4925715534192706e-06, + "loss": 1.6866, + "step": 13849 + }, + { + "epoch": 3.02, + "learning_rate": 3.492462311557789e-06, + "loss": 2.429, + "step": 13850 + }, + { + "epoch": 3.02, + "learning_rate": 3.4923530696963082e-06, + "loss": 1.5875, + "step": 13851 + }, + { + "epoch": 3.02, + "learning_rate": 3.4922438278348264e-06, + "loss": 2.6503, + "step": 13852 + }, + { + "epoch": 3.02, + "learning_rate": 3.492134585973345e-06, + "loss": 2.5482, + "step": 13853 + }, + { + "epoch": 3.02, + "learning_rate": 3.492025344111864e-06, + "loss": 2.92, + "step": 13854 + }, + { + "epoch": 3.02, + "learning_rate": 3.4919161022503827e-06, + "loss": 1.9809, + "step": 13855 + }, + { + "epoch": 3.02, + "learning_rate": 3.491806860388901e-06, + "loss": 1.9045, + "step": 13856 + }, + { + "epoch": 3.02, + "learning_rate": 3.49169761852742e-06, + "loss": 1.5722, + "step": 13857 + }, + { + "epoch": 3.02, + "learning_rate": 3.4915883766659385e-06, + "loss": 1.2704, + "step": 13858 + }, + { + "epoch": 3.02, + "learning_rate": 3.4914791348044575e-06, + "loss": 2.139, + "step": 13859 + }, + { + "epoch": 3.02, + "learning_rate": 3.491369892942976e-06, + "loss": 1.7429, + "step": 13860 + }, + { + "epoch": 3.03, + "learning_rate": 3.4912606510814943e-06, + "loss": 2.9529, + "step": 13861 + }, + { + "epoch": 3.03, + "learning_rate": 3.4911514092200134e-06, + "loss": 2.2536, + "step": 13862 + }, + { + "epoch": 3.03, + "learning_rate": 3.491042167358532e-06, + "loss": 2.0386, + "step": 13863 + }, + { + "epoch": 3.03, + "learning_rate": 3.490932925497051e-06, + "loss": 2.609, + "step": 13864 + }, + { + "epoch": 3.03, + "learning_rate": 3.4908236836355696e-06, + "loss": 2.2355, + "step": 13865 + }, + { + "epoch": 3.03, + "learning_rate": 3.490714441774088e-06, + "loss": 1.736, + "step": 13866 + }, + { + "epoch": 3.03, + "learning_rate": 3.490605199912607e-06, + "loss": 3.4424, + "step": 13867 + }, + { + "epoch": 3.03, + "learning_rate": 3.4904959580511255e-06, + "loss": 0.7944, + "step": 13868 + }, + { + "epoch": 3.03, + "learning_rate": 3.4903867161896445e-06, + "loss": 2.6819, + "step": 13869 + }, + { + "epoch": 3.03, + "learning_rate": 3.490277474328163e-06, + "loss": 1.1882, + "step": 13870 + }, + { + "epoch": 3.03, + "learning_rate": 3.4901682324666813e-06, + "loss": 1.6061, + "step": 13871 + }, + { + "epoch": 3.03, + "learning_rate": 3.4900589906052003e-06, + "loss": 1.3561, + "step": 13872 + }, + { + "epoch": 3.03, + "learning_rate": 3.489949748743719e-06, + "loss": 3.2101, + "step": 13873 + }, + { + "epoch": 3.03, + "learning_rate": 3.4898405068822375e-06, + "loss": 1.7587, + "step": 13874 + }, + { + "epoch": 3.03, + "learning_rate": 3.489731265020756e-06, + "loss": 2.126, + "step": 13875 + }, + { + "epoch": 3.03, + "learning_rate": 3.4896220231592748e-06, + "loss": 2.0911, + "step": 13876 + }, + { + "epoch": 3.03, + "learning_rate": 3.4895127812977934e-06, + "loss": 2.7213, + "step": 13877 + }, + { + "epoch": 3.03, + "learning_rate": 3.4894035394363124e-06, + "loss": 1.5831, + "step": 13878 + }, + { + "epoch": 3.03, + "learning_rate": 3.489294297574831e-06, + "loss": 1.5582, + "step": 13879 + }, + { + "epoch": 3.03, + "learning_rate": 3.489185055713349e-06, + "loss": 2.8073, + "step": 13880 + }, + { + "epoch": 3.03, + "learning_rate": 3.4890758138518682e-06, + "loss": 4.5295, + "step": 13881 + }, + { + "epoch": 3.03, + "learning_rate": 3.488966571990387e-06, + "loss": 1.0433, + "step": 13882 + }, + { + "epoch": 3.03, + "learning_rate": 3.488857330128906e-06, + "loss": 1.5948, + "step": 13883 + }, + { + "epoch": 3.03, + "learning_rate": 3.4887480882674245e-06, + "loss": 1.9375, + "step": 13884 + }, + { + "epoch": 3.03, + "learning_rate": 3.4886388464059427e-06, + "loss": 1.504, + "step": 13885 + }, + { + "epoch": 3.03, + "learning_rate": 3.4885296045444617e-06, + "loss": 2.632, + "step": 13886 + }, + { + "epoch": 3.03, + "learning_rate": 3.4884203626829803e-06, + "loss": 1.5592, + "step": 13887 + }, + { + "epoch": 3.03, + "learning_rate": 3.4883111208214994e-06, + "loss": 1.8251, + "step": 13888 + }, + { + "epoch": 3.03, + "learning_rate": 3.4882018789600176e-06, + "loss": 1.9966, + "step": 13889 + }, + { + "epoch": 3.03, + "learning_rate": 3.488092637098536e-06, + "loss": 2.7879, + "step": 13890 + }, + { + "epoch": 3.03, + "learning_rate": 3.487983395237055e-06, + "loss": 3.3331, + "step": 13891 + }, + { + "epoch": 3.03, + "learning_rate": 3.487874153375574e-06, + "loss": 2.6085, + "step": 13892 + }, + { + "epoch": 3.03, + "learning_rate": 3.487764911514093e-06, + "loss": 1.6236, + "step": 13893 + }, + { + "epoch": 3.03, + "learning_rate": 3.487655669652611e-06, + "loss": 1.0809, + "step": 13894 + }, + { + "epoch": 3.03, + "learning_rate": 3.4875464277911296e-06, + "loss": 1.4431, + "step": 13895 + }, + { + "epoch": 3.03, + "learning_rate": 3.4874371859296487e-06, + "loss": 2.7875, + "step": 13896 + }, + { + "epoch": 3.03, + "learning_rate": 3.4873279440681673e-06, + "loss": 2.2495, + "step": 13897 + }, + { + "epoch": 3.03, + "learning_rate": 3.4872187022066863e-06, + "loss": 2.9259, + "step": 13898 + }, + { + "epoch": 3.03, + "learning_rate": 3.4871094603452045e-06, + "loss": 1.8124, + "step": 13899 + }, + { + "epoch": 3.03, + "learning_rate": 3.487000218483723e-06, + "loss": 1.6156, + "step": 13900 + }, + { + "epoch": 3.03, + "learning_rate": 3.486890976622242e-06, + "loss": 1.9063, + "step": 13901 + }, + { + "epoch": 3.03, + "learning_rate": 3.4867817347607608e-06, + "loss": 1.7687, + "step": 13902 + }, + { + "epoch": 3.03, + "learning_rate": 3.4866724928992794e-06, + "loss": 1.4362, + "step": 13903 + }, + { + "epoch": 3.03, + "learning_rate": 3.486563251037798e-06, + "loss": 1.6289, + "step": 13904 + }, + { + "epoch": 3.03, + "learning_rate": 3.4864540091763166e-06, + "loss": 2.6948, + "step": 13905 + }, + { + "epoch": 3.03, + "learning_rate": 3.4863447673148352e-06, + "loss": 1.9117, + "step": 13906 + }, + { + "epoch": 3.04, + "learning_rate": 3.4862355254533542e-06, + "loss": 2.9387, + "step": 13907 + }, + { + "epoch": 3.04, + "learning_rate": 3.4861262835918724e-06, + "loss": 3.1103, + "step": 13908 + }, + { + "epoch": 3.04, + "learning_rate": 3.486017041730391e-06, + "loss": 2.702, + "step": 13909 + }, + { + "epoch": 3.04, + "learning_rate": 3.48590779986891e-06, + "loss": 2.9425, + "step": 13910 + }, + { + "epoch": 3.04, + "learning_rate": 3.4857985580074287e-06, + "loss": 2.056, + "step": 13911 + }, + { + "epoch": 3.04, + "learning_rate": 3.4856893161459477e-06, + "loss": 1.2758, + "step": 13912 + }, + { + "epoch": 3.04, + "learning_rate": 3.485580074284466e-06, + "loss": 4.3833, + "step": 13913 + }, + { + "epoch": 3.04, + "learning_rate": 3.4854708324229845e-06, + "loss": 3.2095, + "step": 13914 + }, + { + "epoch": 3.04, + "learning_rate": 3.4853615905615036e-06, + "loss": 1.5998, + "step": 13915 + }, + { + "epoch": 3.04, + "learning_rate": 3.485252348700022e-06, + "loss": 2.873, + "step": 13916 + }, + { + "epoch": 3.04, + "learning_rate": 3.485143106838541e-06, + "loss": 1.8009, + "step": 13917 + }, + { + "epoch": 3.04, + "learning_rate": 3.4850338649770594e-06, + "loss": 1.6203, + "step": 13918 + }, + { + "epoch": 3.04, + "learning_rate": 3.484924623115578e-06, + "loss": 1.8627, + "step": 13919 + }, + { + "epoch": 3.04, + "learning_rate": 3.484815381254097e-06, + "loss": 2.693, + "step": 13920 + }, + { + "epoch": 3.04, + "learning_rate": 3.4847061393926157e-06, + "loss": 1.2679, + "step": 13921 + }, + { + "epoch": 3.04, + "learning_rate": 3.484596897531134e-06, + "loss": 2.3169, + "step": 13922 + }, + { + "epoch": 3.04, + "learning_rate": 3.484487655669653e-06, + "loss": 1.6755, + "step": 13923 + }, + { + "epoch": 3.04, + "learning_rate": 3.4843784138081715e-06, + "loss": 1.7477, + "step": 13924 + }, + { + "epoch": 3.04, + "learning_rate": 3.4842691719466905e-06, + "loss": 2.1396, + "step": 13925 + }, + { + "epoch": 3.04, + "learning_rate": 3.484159930085209e-06, + "loss": 2.4046, + "step": 13926 + }, + { + "epoch": 3.04, + "learning_rate": 3.4840506882237273e-06, + "loss": 1.0118, + "step": 13927 + }, + { + "epoch": 3.04, + "learning_rate": 3.4839414463622464e-06, + "loss": 2.393, + "step": 13928 + }, + { + "epoch": 3.04, + "learning_rate": 3.483832204500765e-06, + "loss": 1.7631, + "step": 13929 + }, + { + "epoch": 3.04, + "learning_rate": 3.4837229626392836e-06, + "loss": 1.4177, + "step": 13930 + }, + { + "epoch": 3.04, + "learning_rate": 3.4836137207778026e-06, + "loss": 0.7393, + "step": 13931 + }, + { + "epoch": 3.04, + "learning_rate": 3.483504478916321e-06, + "loss": 1.8008, + "step": 13932 + }, + { + "epoch": 3.04, + "learning_rate": 3.48339523705484e-06, + "loss": 1.6557, + "step": 13933 + }, + { + "epoch": 3.04, + "learning_rate": 3.4832859951933584e-06, + "loss": 2.2116, + "step": 13934 + }, + { + "epoch": 3.04, + "learning_rate": 3.483176753331877e-06, + "loss": 1.9371, + "step": 13935 + }, + { + "epoch": 3.04, + "learning_rate": 3.483067511470396e-06, + "loss": 0.8481, + "step": 13936 + }, + { + "epoch": 3.04, + "learning_rate": 3.4829582696089143e-06, + "loss": 1.226, + "step": 13937 + }, + { + "epoch": 3.04, + "learning_rate": 3.482849027747433e-06, + "loss": 0.3476, + "step": 13938 + }, + { + "epoch": 3.04, + "learning_rate": 3.482739785885952e-06, + "loss": 1.9339, + "step": 13939 + }, + { + "epoch": 3.04, + "learning_rate": 3.4826305440244705e-06, + "loss": 1.8068, + "step": 13940 + }, + { + "epoch": 3.04, + "learning_rate": 3.4825213021629887e-06, + "loss": 1.0671, + "step": 13941 + }, + { + "epoch": 3.04, + "learning_rate": 3.4824120603015078e-06, + "loss": 2.6581, + "step": 13942 + }, + { + "epoch": 3.04, + "learning_rate": 3.4823028184400264e-06, + "loss": 2.1198, + "step": 13943 + }, + { + "epoch": 3.04, + "learning_rate": 3.4821935765785454e-06, + "loss": 2.4895, + "step": 13944 + }, + { + "epoch": 3.04, + "learning_rate": 3.482084334717064e-06, + "loss": 3.31, + "step": 13945 + }, + { + "epoch": 3.04, + "learning_rate": 3.481975092855582e-06, + "loss": 3.1561, + "step": 13946 + }, + { + "epoch": 3.04, + "learning_rate": 3.4818658509941012e-06, + "loss": 2.5998, + "step": 13947 + }, + { + "epoch": 3.04, + "learning_rate": 3.48175660913262e-06, + "loss": 2.4559, + "step": 13948 + }, + { + "epoch": 3.04, + "learning_rate": 3.481647367271139e-06, + "loss": 1.7203, + "step": 13949 + }, + { + "epoch": 3.04, + "learning_rate": 3.4815381254096575e-06, + "loss": 2.777, + "step": 13950 + }, + { + "epoch": 3.04, + "learning_rate": 3.4814288835481757e-06, + "loss": 1.8481, + "step": 13951 + }, + { + "epoch": 3.04, + "learning_rate": 3.4813196416866947e-06, + "loss": 3.3552, + "step": 13952 + }, + { + "epoch": 3.05, + "learning_rate": 3.4812103998252133e-06, + "loss": 2.9392, + "step": 13953 + }, + { + "epoch": 3.05, + "learning_rate": 3.4811011579637324e-06, + "loss": 1.5227, + "step": 13954 + }, + { + "epoch": 3.05, + "learning_rate": 3.4809919161022505e-06, + "loss": 1.4265, + "step": 13955 + }, + { + "epoch": 3.05, + "learning_rate": 3.480882674240769e-06, + "loss": 2.0675, + "step": 13956 + }, + { + "epoch": 3.05, + "learning_rate": 3.480773432379288e-06, + "loss": 1.3109, + "step": 13957 + }, + { + "epoch": 3.05, + "learning_rate": 3.480664190517807e-06, + "loss": 1.8828, + "step": 13958 + }, + { + "epoch": 3.05, + "learning_rate": 3.4805549486563254e-06, + "loss": 1.4699, + "step": 13959 + }, + { + "epoch": 3.05, + "learning_rate": 3.480445706794844e-06, + "loss": 1.3551, + "step": 13960 + }, + { + "epoch": 3.05, + "learning_rate": 3.4803364649333626e-06, + "loss": 1.8083, + "step": 13961 + }, + { + "epoch": 3.05, + "learning_rate": 3.4802272230718812e-06, + "loss": 1.4936, + "step": 13962 + }, + { + "epoch": 3.05, + "learning_rate": 3.4801179812104003e-06, + "loss": 1.9748, + "step": 13963 + }, + { + "epoch": 3.05, + "learning_rate": 3.480008739348919e-06, + "loss": 2.5979, + "step": 13964 + }, + { + "epoch": 3.05, + "learning_rate": 3.479899497487437e-06, + "loss": 1.8668, + "step": 13965 + }, + { + "epoch": 3.05, + "learning_rate": 3.479790255625956e-06, + "loss": 2.8527, + "step": 13966 + }, + { + "epoch": 3.05, + "learning_rate": 3.4796810137644747e-06, + "loss": 3.4407, + "step": 13967 + }, + { + "epoch": 3.05, + "learning_rate": 3.4795717719029938e-06, + "loss": 2.1589, + "step": 13968 + }, + { + "epoch": 3.05, + "learning_rate": 3.4794625300415124e-06, + "loss": 1.7664, + "step": 13969 + }, + { + "epoch": 3.05, + "learning_rate": 3.4793532881800306e-06, + "loss": 1.2449, + "step": 13970 + }, + { + "epoch": 3.05, + "learning_rate": 3.4792440463185496e-06, + "loss": 1.8299, + "step": 13971 + }, + { + "epoch": 3.05, + "learning_rate": 3.479134804457068e-06, + "loss": 2.1319, + "step": 13972 + }, + { + "epoch": 3.05, + "learning_rate": 3.4790255625955872e-06, + "loss": 3.5663, + "step": 13973 + }, + { + "epoch": 3.05, + "learning_rate": 3.4789163207341054e-06, + "loss": 2.8816, + "step": 13974 + }, + { + "epoch": 3.05, + "learning_rate": 3.478807078872624e-06, + "loss": 2.2341, + "step": 13975 + }, + { + "epoch": 3.05, + "learning_rate": 3.478697837011143e-06, + "loss": 1.6513, + "step": 13976 + }, + { + "epoch": 3.05, + "learning_rate": 3.4785885951496617e-06, + "loss": 2.53, + "step": 13977 + }, + { + "epoch": 3.05, + "learning_rate": 3.4784793532881807e-06, + "loss": 0.9595, + "step": 13978 + }, + { + "epoch": 3.05, + "learning_rate": 3.478370111426699e-06, + "loss": 4.3754, + "step": 13979 + }, + { + "epoch": 3.05, + "learning_rate": 3.4782608695652175e-06, + "loss": 2.2211, + "step": 13980 + }, + { + "epoch": 3.05, + "learning_rate": 3.4781516277037365e-06, + "loss": 1.857, + "step": 13981 + }, + { + "epoch": 3.05, + "learning_rate": 3.478042385842255e-06, + "loss": 2.4774, + "step": 13982 + }, + { + "epoch": 3.05, + "learning_rate": 3.4779331439807738e-06, + "loss": 2.4686, + "step": 13983 + }, + { + "epoch": 3.05, + "learning_rate": 3.4778239021192924e-06, + "loss": 1.6365, + "step": 13984 + }, + { + "epoch": 3.05, + "learning_rate": 3.477714660257811e-06, + "loss": 2.2015, + "step": 13985 + }, + { + "epoch": 3.05, + "learning_rate": 3.47760541839633e-06, + "loss": 1.7714, + "step": 13986 + }, + { + "epoch": 3.05, + "learning_rate": 3.4774961765348486e-06, + "loss": 2.7892, + "step": 13987 + }, + { + "epoch": 3.05, + "learning_rate": 3.477386934673367e-06, + "loss": 2.6469, + "step": 13988 + }, + { + "epoch": 3.05, + "learning_rate": 3.477277692811886e-06, + "loss": 1.6171, + "step": 13989 + }, + { + "epoch": 3.05, + "learning_rate": 3.4771684509504045e-06, + "loss": 2.0849, + "step": 13990 + }, + { + "epoch": 3.05, + "learning_rate": 3.477059209088923e-06, + "loss": 3.2086, + "step": 13991 + }, + { + "epoch": 3.05, + "learning_rate": 3.476949967227442e-06, + "loss": 1.1557, + "step": 13992 + }, + { + "epoch": 3.05, + "learning_rate": 3.4768407253659603e-06, + "loss": 1.3709, + "step": 13993 + }, + { + "epoch": 3.05, + "learning_rate": 3.476731483504479e-06, + "loss": 2.7867, + "step": 13994 + }, + { + "epoch": 3.05, + "learning_rate": 3.476622241642998e-06, + "loss": 1.7429, + "step": 13995 + }, + { + "epoch": 3.05, + "learning_rate": 3.4765129997815166e-06, + "loss": 3.1411, + "step": 13996 + }, + { + "epoch": 3.05, + "learning_rate": 3.4764037579200356e-06, + "loss": 1.5202, + "step": 13997 + }, + { + "epoch": 3.05, + "learning_rate": 3.4762945160585538e-06, + "loss": 1.0811, + "step": 13998 + }, + { + "epoch": 3.06, + "learning_rate": 3.4761852741970724e-06, + "loss": 1.1773, + "step": 13999 + }, + { + "epoch": 3.06, + "learning_rate": 3.4760760323355914e-06, + "loss": 1.9906, + "step": 14000 + }, + { + "epoch": 3.06, + "learning_rate": 3.47596679047411e-06, + "loss": 2.3331, + "step": 14001 + }, + { + "epoch": 3.06, + "learning_rate": 3.475857548612629e-06, + "loss": 2.3434, + "step": 14002 + }, + { + "epoch": 3.06, + "learning_rate": 3.4757483067511473e-06, + "loss": 1.0148, + "step": 14003 + }, + { + "epoch": 3.06, + "learning_rate": 3.475639064889666e-06, + "loss": 0.9212, + "step": 14004 + }, + { + "epoch": 3.06, + "learning_rate": 3.475529823028185e-06, + "loss": 1.9345, + "step": 14005 + }, + { + "epoch": 3.06, + "learning_rate": 3.4754205811667035e-06, + "loss": 1.8047, + "step": 14006 + }, + { + "epoch": 3.06, + "learning_rate": 3.4753113393052217e-06, + "loss": 2.3804, + "step": 14007 + }, + { + "epoch": 3.06, + "learning_rate": 3.4752020974437407e-06, + "loss": 2.845, + "step": 14008 + }, + { + "epoch": 3.06, + "learning_rate": 3.4750928555822593e-06, + "loss": 1.6861, + "step": 14009 + }, + { + "epoch": 3.06, + "learning_rate": 3.4749836137207784e-06, + "loss": 1.7873, + "step": 14010 + }, + { + "epoch": 3.06, + "learning_rate": 3.474874371859297e-06, + "loss": 1.4145, + "step": 14011 + }, + { + "epoch": 3.06, + "learning_rate": 3.474765129997815e-06, + "loss": 1.5597, + "step": 14012 + }, + { + "epoch": 3.06, + "learning_rate": 3.4746558881363342e-06, + "loss": 2.2227, + "step": 14013 + }, + { + "epoch": 3.06, + "learning_rate": 3.474546646274853e-06, + "loss": 2.8908, + "step": 14014 + }, + { + "epoch": 3.06, + "learning_rate": 3.4744374044133714e-06, + "loss": 0.8774, + "step": 14015 + }, + { + "epoch": 3.06, + "learning_rate": 3.4743281625518905e-06, + "loss": 2.0751, + "step": 14016 + }, + { + "epoch": 3.06, + "learning_rate": 3.4742189206904087e-06, + "loss": 0.9866, + "step": 14017 + }, + { + "epoch": 3.06, + "learning_rate": 3.4741096788289273e-06, + "loss": 1.7675, + "step": 14018 + }, + { + "epoch": 3.06, + "learning_rate": 3.4740004369674463e-06, + "loss": 1.8538, + "step": 14019 + }, + { + "epoch": 3.06, + "learning_rate": 3.473891195105965e-06, + "loss": 0.7093, + "step": 14020 + }, + { + "epoch": 3.06, + "learning_rate": 3.473781953244483e-06, + "loss": 1.7868, + "step": 14021 + }, + { + "epoch": 3.06, + "learning_rate": 3.473672711383002e-06, + "loss": 3.6943, + "step": 14022 + }, + { + "epoch": 3.06, + "learning_rate": 3.4735634695215207e-06, + "loss": 2.1768, + "step": 14023 + }, + { + "epoch": 3.06, + "learning_rate": 3.4734542276600398e-06, + "loss": 2.3533, + "step": 14024 + }, + { + "epoch": 3.06, + "learning_rate": 3.4733449857985584e-06, + "loss": 2.0158, + "step": 14025 + }, + { + "epoch": 3.06, + "learning_rate": 3.4732357439370766e-06, + "loss": 1.5446, + "step": 14026 + }, + { + "epoch": 3.06, + "learning_rate": 3.4731265020755956e-06, + "loss": 1.0486, + "step": 14027 + }, + { + "epoch": 3.06, + "learning_rate": 3.4730172602141142e-06, + "loss": 1.6004, + "step": 14028 + }, + { + "epoch": 3.06, + "learning_rate": 3.4729080183526333e-06, + "loss": 2.5764, + "step": 14029 + }, + { + "epoch": 3.06, + "learning_rate": 3.472798776491152e-06, + "loss": 0.9168, + "step": 14030 + }, + { + "epoch": 3.06, + "learning_rate": 3.47268953462967e-06, + "loss": 2.4616, + "step": 14031 + }, + { + "epoch": 3.06, + "learning_rate": 3.472580292768189e-06, + "loss": 2.5364, + "step": 14032 + }, + { + "epoch": 3.06, + "learning_rate": 3.4724710509067077e-06, + "loss": 0.5544, + "step": 14033 + }, + { + "epoch": 3.06, + "learning_rate": 3.4723618090452267e-06, + "loss": 2.0738, + "step": 14034 + }, + { + "epoch": 3.06, + "learning_rate": 3.4722525671837453e-06, + "loss": 1.067, + "step": 14035 + }, + { + "epoch": 3.06, + "learning_rate": 3.4721433253222635e-06, + "loss": 2.1095, + "step": 14036 + }, + { + "epoch": 3.06, + "learning_rate": 3.4720340834607826e-06, + "loss": 3.4136, + "step": 14037 + }, + { + "epoch": 3.06, + "learning_rate": 3.471924841599301e-06, + "loss": 3.4262, + "step": 14038 + }, + { + "epoch": 3.06, + "learning_rate": 3.4718155997378202e-06, + "loss": 1.3855, + "step": 14039 + }, + { + "epoch": 3.06, + "learning_rate": 3.4717063578763384e-06, + "loss": 2.2374, + "step": 14040 + }, + { + "epoch": 3.06, + "learning_rate": 3.471597116014857e-06, + "loss": 1.6287, + "step": 14041 + }, + { + "epoch": 3.06, + "learning_rate": 3.471487874153376e-06, + "loss": 2.7378, + "step": 14042 + }, + { + "epoch": 3.06, + "learning_rate": 3.4713786322918947e-06, + "loss": 2.7074, + "step": 14043 + }, + { + "epoch": 3.07, + "learning_rate": 3.4712693904304133e-06, + "loss": 1.896, + "step": 14044 + }, + { + "epoch": 3.07, + "learning_rate": 3.471160148568932e-06, + "loss": 1.9883, + "step": 14045 + }, + { + "epoch": 3.07, + "learning_rate": 3.4710509067074505e-06, + "loss": 1.4937, + "step": 14046 + }, + { + "epoch": 3.07, + "learning_rate": 3.470941664845969e-06, + "loss": 1.5094, + "step": 14047 + }, + { + "epoch": 3.07, + "learning_rate": 3.470832422984488e-06, + "loss": 1.4773, + "step": 14048 + }, + { + "epoch": 3.07, + "learning_rate": 3.4707231811230068e-06, + "loss": 2.5547, + "step": 14049 + }, + { + "epoch": 3.07, + "learning_rate": 3.470613939261525e-06, + "loss": 2.6, + "step": 14050 + }, + { + "epoch": 3.07, + "learning_rate": 3.470504697400044e-06, + "loss": 2.181, + "step": 14051 + }, + { + "epoch": 3.07, + "learning_rate": 3.4703954555385626e-06, + "loss": 2.0315, + "step": 14052 + }, + { + "epoch": 3.07, + "learning_rate": 3.4702862136770816e-06, + "loss": 2.8841, + "step": 14053 + }, + { + "epoch": 3.07, + "learning_rate": 3.4701769718156e-06, + "loss": 0.4796, + "step": 14054 + }, + { + "epoch": 3.07, + "learning_rate": 3.4700677299541184e-06, + "loss": 2.6046, + "step": 14055 + }, + { + "epoch": 3.07, + "learning_rate": 3.4699584880926375e-06, + "loss": 3.407, + "step": 14056 + }, + { + "epoch": 3.07, + "learning_rate": 3.469849246231156e-06, + "loss": 1.4123, + "step": 14057 + }, + { + "epoch": 3.07, + "learning_rate": 3.469740004369675e-06, + "loss": 1.4799, + "step": 14058 + }, + { + "epoch": 3.07, + "learning_rate": 3.4696307625081933e-06, + "loss": 2.4544, + "step": 14059 + }, + { + "epoch": 3.07, + "learning_rate": 3.469521520646712e-06, + "loss": 1.7172, + "step": 14060 + }, + { + "epoch": 3.07, + "learning_rate": 3.469412278785231e-06, + "loss": 1.8631, + "step": 14061 + }, + { + "epoch": 3.07, + "learning_rate": 3.4693030369237495e-06, + "loss": 1.6057, + "step": 14062 + }, + { + "epoch": 3.07, + "learning_rate": 3.4691937950622686e-06, + "loss": 3.3443, + "step": 14063 + }, + { + "epoch": 3.07, + "learning_rate": 3.4690845532007868e-06, + "loss": 3.6001, + "step": 14064 + }, + { + "epoch": 3.07, + "learning_rate": 3.4689753113393054e-06, + "loss": 1.9516, + "step": 14065 + }, + { + "epoch": 3.07, + "learning_rate": 3.4688660694778244e-06, + "loss": 2.9775, + "step": 14066 + }, + { + "epoch": 3.07, + "learning_rate": 3.468756827616343e-06, + "loss": 3.4864, + "step": 14067 + }, + { + "epoch": 3.07, + "learning_rate": 3.4686475857548616e-06, + "loss": 1.6015, + "step": 14068 + }, + { + "epoch": 3.07, + "learning_rate": 3.4685383438933802e-06, + "loss": 1.2939, + "step": 14069 + }, + { + "epoch": 3.07, + "learning_rate": 3.468429102031899e-06, + "loss": 3.4572, + "step": 14070 + }, + { + "epoch": 3.07, + "learning_rate": 3.4683198601704175e-06, + "loss": 2.459, + "step": 14071 + }, + { + "epoch": 3.07, + "learning_rate": 3.4682106183089365e-06, + "loss": 1.8349, + "step": 14072 + }, + { + "epoch": 3.07, + "learning_rate": 3.4681013764474547e-06, + "loss": 1.8475, + "step": 14073 + }, + { + "epoch": 3.07, + "learning_rate": 3.4679921345859733e-06, + "loss": 2.8903, + "step": 14074 + }, + { + "epoch": 3.07, + "learning_rate": 3.4678828927244923e-06, + "loss": 3.0193, + "step": 14075 + }, + { + "epoch": 3.07, + "learning_rate": 3.467773650863011e-06, + "loss": 2.1592, + "step": 14076 + }, + { + "epoch": 3.07, + "learning_rate": 3.46766440900153e-06, + "loss": 2.9854, + "step": 14077 + }, + { + "epoch": 3.07, + "learning_rate": 3.467555167140048e-06, + "loss": 3.5878, + "step": 14078 + }, + { + "epoch": 3.07, + "learning_rate": 3.4674459252785668e-06, + "loss": 1.1877, + "step": 14079 + }, + { + "epoch": 3.07, + "learning_rate": 3.467336683417086e-06, + "loss": 1.385, + "step": 14080 + }, + { + "epoch": 3.07, + "learning_rate": 3.4672274415556044e-06, + "loss": 2.3672, + "step": 14081 + }, + { + "epoch": 3.07, + "learning_rate": 3.4671181996941235e-06, + "loss": 2.2676, + "step": 14082 + }, + { + "epoch": 3.07, + "learning_rate": 3.4670089578326416e-06, + "loss": 0.8812, + "step": 14083 + }, + { + "epoch": 3.07, + "learning_rate": 3.4668997159711603e-06, + "loss": 2.6918, + "step": 14084 + }, + { + "epoch": 3.07, + "learning_rate": 3.4667904741096793e-06, + "loss": 2.8762, + "step": 14085 + }, + { + "epoch": 3.07, + "learning_rate": 3.466681232248198e-06, + "loss": 2.8961, + "step": 14086 + }, + { + "epoch": 3.07, + "learning_rate": 3.466571990386716e-06, + "loss": 3.1404, + "step": 14087 + }, + { + "epoch": 3.07, + "learning_rate": 3.466462748525235e-06, + "loss": 2.053, + "step": 14088 + }, + { + "epoch": 3.07, + "learning_rate": 3.4663535066637537e-06, + "loss": 1.5484, + "step": 14089 + }, + { + "epoch": 3.08, + "learning_rate": 3.4662442648022728e-06, + "loss": 2.6052, + "step": 14090 + }, + { + "epoch": 3.08, + "learning_rate": 3.4661350229407914e-06, + "loss": 3.6865, + "step": 14091 + }, + { + "epoch": 3.08, + "learning_rate": 3.4660257810793096e-06, + "loss": 2.6117, + "step": 14092 + }, + { + "epoch": 3.08, + "learning_rate": 3.4659165392178286e-06, + "loss": 3.1488, + "step": 14093 + }, + { + "epoch": 3.08, + "learning_rate": 3.465807297356347e-06, + "loss": 3.2582, + "step": 14094 + }, + { + "epoch": 3.08, + "learning_rate": 3.4656980554948662e-06, + "loss": 2.4936, + "step": 14095 + }, + { + "epoch": 3.08, + "learning_rate": 3.465588813633385e-06, + "loss": 2.5434, + "step": 14096 + }, + { + "epoch": 3.08, + "learning_rate": 3.465479571771903e-06, + "loss": 1.9525, + "step": 14097 + }, + { + "epoch": 3.08, + "learning_rate": 3.465370329910422e-06, + "loss": 2.2274, + "step": 14098 + }, + { + "epoch": 3.08, + "learning_rate": 3.4652610880489407e-06, + "loss": 1.9072, + "step": 14099 + }, + { + "epoch": 3.08, + "learning_rate": 3.4651518461874593e-06, + "loss": 2.8727, + "step": 14100 + }, + { + "epoch": 3.08, + "learning_rate": 3.465042604325978e-06, + "loss": 2.2184, + "step": 14101 + }, + { + "epoch": 3.08, + "learning_rate": 3.4649333624644965e-06, + "loss": 2.2457, + "step": 14102 + }, + { + "epoch": 3.08, + "learning_rate": 3.464824120603015e-06, + "loss": 2.7275, + "step": 14103 + }, + { + "epoch": 3.08, + "learning_rate": 3.464714878741534e-06, + "loss": 2.2057, + "step": 14104 + }, + { + "epoch": 3.08, + "learning_rate": 3.4646056368800528e-06, + "loss": 1.8866, + "step": 14105 + }, + { + "epoch": 3.08, + "learning_rate": 3.464496395018571e-06, + "loss": 2.4554, + "step": 14106 + }, + { + "epoch": 3.08, + "learning_rate": 3.46438715315709e-06, + "loss": 1.632, + "step": 14107 + }, + { + "epoch": 3.08, + "learning_rate": 3.4642779112956086e-06, + "loss": 1.5729, + "step": 14108 + }, + { + "epoch": 3.08, + "learning_rate": 3.4641686694341276e-06, + "loss": 1.8347, + "step": 14109 + }, + { + "epoch": 3.08, + "learning_rate": 3.4640594275726463e-06, + "loss": 1.5224, + "step": 14110 + }, + { + "epoch": 3.08, + "learning_rate": 3.4639501857111644e-06, + "loss": 2.2889, + "step": 14111 + }, + { + "epoch": 3.08, + "learning_rate": 3.4638409438496835e-06, + "loss": 2.3929, + "step": 14112 + }, + { + "epoch": 3.08, + "learning_rate": 3.463731701988202e-06, + "loss": 1.9428, + "step": 14113 + }, + { + "epoch": 3.08, + "learning_rate": 3.463622460126721e-06, + "loss": 1.4669, + "step": 14114 + }, + { + "epoch": 3.08, + "learning_rate": 3.4635132182652397e-06, + "loss": 1.6117, + "step": 14115 + }, + { + "epoch": 3.08, + "learning_rate": 3.463403976403758e-06, + "loss": 2.2301, + "step": 14116 + }, + { + "epoch": 3.08, + "learning_rate": 3.463294734542277e-06, + "loss": 1.2345, + "step": 14117 + }, + { + "epoch": 3.08, + "learning_rate": 3.4631854926807956e-06, + "loss": 1.8409, + "step": 14118 + }, + { + "epoch": 3.08, + "learning_rate": 3.4630762508193146e-06, + "loss": 1.1711, + "step": 14119 + }, + { + "epoch": 3.08, + "learning_rate": 3.462967008957833e-06, + "loss": 1.5241, + "step": 14120 + }, + { + "epoch": 3.08, + "learning_rate": 3.4628577670963514e-06, + "loss": 3.433, + "step": 14121 + }, + { + "epoch": 3.08, + "learning_rate": 3.4627485252348704e-06, + "loss": 1.9941, + "step": 14122 + }, + { + "epoch": 3.08, + "learning_rate": 3.462639283373389e-06, + "loss": 3.2233, + "step": 14123 + }, + { + "epoch": 3.08, + "learning_rate": 3.4625300415119077e-06, + "loss": 2.2087, + "step": 14124 + }, + { + "epoch": 3.08, + "learning_rate": 3.4624207996504263e-06, + "loss": 2.0114, + "step": 14125 + }, + { + "epoch": 3.08, + "learning_rate": 3.462311557788945e-06, + "loss": 1.6671, + "step": 14126 + }, + { + "epoch": 3.08, + "learning_rate": 3.4622023159274635e-06, + "loss": 2.3103, + "step": 14127 + }, + { + "epoch": 3.08, + "learning_rate": 3.4620930740659825e-06, + "loss": 2.4039, + "step": 14128 + }, + { + "epoch": 3.08, + "learning_rate": 3.461983832204501e-06, + "loss": 2.2864, + "step": 14129 + }, + { + "epoch": 3.08, + "learning_rate": 3.4618745903430197e-06, + "loss": 2.0202, + "step": 14130 + }, + { + "epoch": 3.08, + "learning_rate": 3.4617653484815384e-06, + "loss": 2.488, + "step": 14131 + }, + { + "epoch": 3.08, + "learning_rate": 3.461656106620057e-06, + "loss": 1.6244, + "step": 14132 + }, + { + "epoch": 3.08, + "learning_rate": 3.461546864758576e-06, + "loss": 1.5907, + "step": 14133 + }, + { + "epoch": 3.08, + "learning_rate": 3.461437622897094e-06, + "loss": 2.1388, + "step": 14134 + }, + { + "epoch": 3.08, + "learning_rate": 3.461328381035613e-06, + "loss": 3.1417, + "step": 14135 + }, + { + "epoch": 3.09, + "learning_rate": 3.461219139174132e-06, + "loss": 1.3012, + "step": 14136 + }, + { + "epoch": 3.09, + "learning_rate": 3.4611098973126504e-06, + "loss": 2.2267, + "step": 14137 + }, + { + "epoch": 3.09, + "learning_rate": 3.4610006554511695e-06, + "loss": 1.2568, + "step": 14138 + }, + { + "epoch": 3.09, + "learning_rate": 3.4608914135896877e-06, + "loss": 1.7512, + "step": 14139 + }, + { + "epoch": 3.09, + "learning_rate": 3.4607821717282063e-06, + "loss": 1.7781, + "step": 14140 + }, + { + "epoch": 3.09, + "learning_rate": 3.4606729298667253e-06, + "loss": 2.199, + "step": 14141 + }, + { + "epoch": 3.09, + "learning_rate": 3.460563688005244e-06, + "loss": 2.066, + "step": 14142 + }, + { + "epoch": 3.09, + "learning_rate": 3.460454446143763e-06, + "loss": 2.4841, + "step": 14143 + }, + { + "epoch": 3.09, + "learning_rate": 3.460345204282281e-06, + "loss": 1.1148, + "step": 14144 + }, + { + "epoch": 3.09, + "learning_rate": 3.4602359624207998e-06, + "loss": 1.0088, + "step": 14145 + }, + { + "epoch": 3.09, + "learning_rate": 3.460126720559319e-06, + "loss": 2.7314, + "step": 14146 + }, + { + "epoch": 3.09, + "learning_rate": 3.4600174786978374e-06, + "loss": 0.6649, + "step": 14147 + }, + { + "epoch": 3.09, + "learning_rate": 3.4599082368363564e-06, + "loss": 1.9226, + "step": 14148 + }, + { + "epoch": 3.09, + "learning_rate": 3.4597989949748746e-06, + "loss": 0.8739, + "step": 14149 + }, + { + "epoch": 3.09, + "learning_rate": 3.4596897531133932e-06, + "loss": 2.9876, + "step": 14150 + }, + { + "epoch": 3.09, + "learning_rate": 3.4595805112519123e-06, + "loss": 1.4089, + "step": 14151 + }, + { + "epoch": 3.09, + "learning_rate": 3.459471269390431e-06, + "loss": 1.1851, + "step": 14152 + }, + { + "epoch": 3.09, + "learning_rate": 3.459362027528949e-06, + "loss": 1.655, + "step": 14153 + }, + { + "epoch": 3.09, + "learning_rate": 3.459252785667468e-06, + "loss": 1.8775, + "step": 14154 + }, + { + "epoch": 3.09, + "learning_rate": 3.4591435438059867e-06, + "loss": 0.8805, + "step": 14155 + }, + { + "epoch": 3.09, + "learning_rate": 3.4590343019445053e-06, + "loss": 1.5887, + "step": 14156 + }, + { + "epoch": 3.09, + "learning_rate": 3.4589250600830244e-06, + "loss": 1.6422, + "step": 14157 + }, + { + "epoch": 3.09, + "learning_rate": 3.4588158182215425e-06, + "loss": 3.6939, + "step": 14158 + }, + { + "epoch": 3.09, + "learning_rate": 3.458706576360061e-06, + "loss": 0.9719, + "step": 14159 + }, + { + "epoch": 3.09, + "learning_rate": 3.45859733449858e-06, + "loss": 3.0246, + "step": 14160 + }, + { + "epoch": 3.09, + "learning_rate": 3.458488092637099e-06, + "loss": 1.4058, + "step": 14161 + }, + { + "epoch": 3.09, + "learning_rate": 3.458378850775618e-06, + "loss": 1.9473, + "step": 14162 + }, + { + "epoch": 3.09, + "learning_rate": 3.458269608914136e-06, + "loss": 1.8023, + "step": 14163 + }, + { + "epoch": 3.09, + "learning_rate": 3.4581603670526546e-06, + "loss": 2.023, + "step": 14164 + }, + { + "epoch": 3.09, + "learning_rate": 3.4580511251911737e-06, + "loss": 1.57, + "step": 14165 + }, + { + "epoch": 3.09, + "learning_rate": 3.4579418833296923e-06, + "loss": 2.3391, + "step": 14166 + }, + { + "epoch": 3.09, + "learning_rate": 3.4578326414682105e-06, + "loss": 2.0548, + "step": 14167 + }, + { + "epoch": 3.09, + "learning_rate": 3.4577233996067295e-06, + "loss": 2.1104, + "step": 14168 + }, + { + "epoch": 3.09, + "learning_rate": 3.457614157745248e-06, + "loss": 1.9875, + "step": 14169 + }, + { + "epoch": 3.09, + "learning_rate": 3.457504915883767e-06, + "loss": 1.2139, + "step": 14170 + }, + { + "epoch": 3.09, + "learning_rate": 3.4573956740222858e-06, + "loss": 2.1078, + "step": 14171 + }, + { + "epoch": 3.09, + "learning_rate": 3.457286432160804e-06, + "loss": 3.6625, + "step": 14172 + }, + { + "epoch": 3.09, + "learning_rate": 3.457177190299323e-06, + "loss": 2.2081, + "step": 14173 + }, + { + "epoch": 3.09, + "learning_rate": 3.4570679484378416e-06, + "loss": 2.7848, + "step": 14174 + }, + { + "epoch": 3.09, + "learning_rate": 3.4569587065763606e-06, + "loss": 2.0041, + "step": 14175 + }, + { + "epoch": 3.09, + "learning_rate": 3.4568494647148792e-06, + "loss": 1.5085, + "step": 14176 + }, + { + "epoch": 3.09, + "learning_rate": 3.4567402228533974e-06, + "loss": 1.5216, + "step": 14177 + }, + { + "epoch": 3.09, + "learning_rate": 3.4566309809919165e-06, + "loss": 0.7869, + "step": 14178 + }, + { + "epoch": 3.09, + "learning_rate": 3.456521739130435e-06, + "loss": 2.0533, + "step": 14179 + }, + { + "epoch": 3.09, + "learning_rate": 3.4564124972689537e-06, + "loss": 2.8886, + "step": 14180 + }, + { + "epoch": 3.09, + "learning_rate": 3.4563032554074727e-06, + "loss": 0.8868, + "step": 14181 + }, + { + "epoch": 3.1, + "learning_rate": 3.456194013545991e-06, + "loss": 2.0424, + "step": 14182 + }, + { + "epoch": 3.1, + "learning_rate": 3.45608477168451e-06, + "loss": 2.1517, + "step": 14183 + }, + { + "epoch": 3.1, + "learning_rate": 3.4559755298230286e-06, + "loss": 3.3037, + "step": 14184 + }, + { + "epoch": 3.1, + "learning_rate": 3.455866287961547e-06, + "loss": 1.686, + "step": 14185 + }, + { + "epoch": 3.1, + "learning_rate": 3.4557570461000658e-06, + "loss": 1.0761, + "step": 14186 + }, + { + "epoch": 3.1, + "learning_rate": 3.4556478042385844e-06, + "loss": 1.8376, + "step": 14187 + }, + { + "epoch": 3.1, + "learning_rate": 3.455538562377103e-06, + "loss": 2.8295, + "step": 14188 + }, + { + "epoch": 3.1, + "learning_rate": 3.455429320515622e-06, + "loss": 2.1382, + "step": 14189 + }, + { + "epoch": 3.1, + "learning_rate": 3.4553200786541406e-06, + "loss": 2.1171, + "step": 14190 + }, + { + "epoch": 3.1, + "learning_rate": 3.455210836792659e-06, + "loss": 2.7802, + "step": 14191 + }, + { + "epoch": 3.1, + "learning_rate": 3.455101594931178e-06, + "loss": 1.6605, + "step": 14192 + }, + { + "epoch": 3.1, + "learning_rate": 3.4549923530696965e-06, + "loss": 2.2642, + "step": 14193 + }, + { + "epoch": 3.1, + "learning_rate": 3.4548831112082155e-06, + "loss": 2.4284, + "step": 14194 + }, + { + "epoch": 3.1, + "learning_rate": 3.454773869346734e-06, + "loss": 2.6087, + "step": 14195 + }, + { + "epoch": 3.1, + "learning_rate": 3.4546646274852523e-06, + "loss": 2.1321, + "step": 14196 + }, + { + "epoch": 3.1, + "learning_rate": 3.4545553856237713e-06, + "loss": 1.9543, + "step": 14197 + }, + { + "epoch": 3.1, + "learning_rate": 3.45444614376229e-06, + "loss": 2.0809, + "step": 14198 + }, + { + "epoch": 3.1, + "learning_rate": 3.454336901900809e-06, + "loss": 0.8474, + "step": 14199 + }, + { + "epoch": 3.1, + "learning_rate": 3.454227660039327e-06, + "loss": 2.8628, + "step": 14200 + }, + { + "epoch": 3.1, + "learning_rate": 3.4541184181778458e-06, + "loss": 2.327, + "step": 14201 + }, + { + "epoch": 3.1, + "learning_rate": 3.454009176316365e-06, + "loss": 2.6109, + "step": 14202 + }, + { + "epoch": 3.1, + "learning_rate": 3.4538999344548834e-06, + "loss": 2.5249, + "step": 14203 + }, + { + "epoch": 3.1, + "learning_rate": 3.4537906925934025e-06, + "loss": 2.1585, + "step": 14204 + }, + { + "epoch": 3.1, + "learning_rate": 3.4536814507319207e-06, + "loss": 2.5231, + "step": 14205 + }, + { + "epoch": 3.1, + "learning_rate": 3.4535722088704393e-06, + "loss": 2.8027, + "step": 14206 + }, + { + "epoch": 3.1, + "learning_rate": 3.4534629670089583e-06, + "loss": 2.1669, + "step": 14207 + }, + { + "epoch": 3.1, + "learning_rate": 3.453353725147477e-06, + "loss": 1.8771, + "step": 14208 + }, + { + "epoch": 3.1, + "learning_rate": 3.4532444832859955e-06, + "loss": 3.2531, + "step": 14209 + }, + { + "epoch": 3.1, + "learning_rate": 3.453135241424514e-06, + "loss": 3.015, + "step": 14210 + }, + { + "epoch": 3.1, + "learning_rate": 3.4530259995630327e-06, + "loss": 2.2977, + "step": 14211 + }, + { + "epoch": 3.1, + "learning_rate": 3.4529167577015514e-06, + "loss": 2.0122, + "step": 14212 + }, + { + "epoch": 3.1, + "learning_rate": 3.4528075158400704e-06, + "loss": 1.3338, + "step": 14213 + }, + { + "epoch": 3.1, + "learning_rate": 3.452698273978589e-06, + "loss": 1.9746, + "step": 14214 + }, + { + "epoch": 3.1, + "learning_rate": 3.452589032117107e-06, + "loss": 2.715, + "step": 14215 + }, + { + "epoch": 3.1, + "learning_rate": 3.4524797902556262e-06, + "loss": 2.5513, + "step": 14216 + }, + { + "epoch": 3.1, + "learning_rate": 3.452370548394145e-06, + "loss": 1.7782, + "step": 14217 + }, + { + "epoch": 3.1, + "learning_rate": 3.452261306532664e-06, + "loss": 2.2697, + "step": 14218 + }, + { + "epoch": 3.1, + "learning_rate": 3.452152064671182e-06, + "loss": 1.7337, + "step": 14219 + }, + { + "epoch": 3.1, + "learning_rate": 3.4520428228097007e-06, + "loss": 1.5018, + "step": 14220 + }, + { + "epoch": 3.1, + "learning_rate": 3.4519335809482197e-06, + "loss": 2.0849, + "step": 14221 + }, + { + "epoch": 3.1, + "learning_rate": 3.4518243390867383e-06, + "loss": 1.8997, + "step": 14222 + }, + { + "epoch": 3.1, + "learning_rate": 3.4517150972252573e-06, + "loss": 1.3566, + "step": 14223 + }, + { + "epoch": 3.1, + "learning_rate": 3.4516058553637755e-06, + "loss": 2.3822, + "step": 14224 + }, + { + "epoch": 3.1, + "learning_rate": 3.451496613502294e-06, + "loss": 2.3692, + "step": 14225 + }, + { + "epoch": 3.1, + "learning_rate": 3.451387371640813e-06, + "loss": 0.6231, + "step": 14226 + }, + { + "epoch": 3.1, + "learning_rate": 3.451278129779332e-06, + "loss": 3.6844, + "step": 14227 + }, + { + "epoch": 3.11, + "learning_rate": 3.451168887917851e-06, + "loss": 2.3258, + "step": 14228 + }, + { + "epoch": 3.11, + "learning_rate": 3.451059646056369e-06, + "loss": 2.3372, + "step": 14229 + }, + { + "epoch": 3.11, + "learning_rate": 3.4509504041948876e-06, + "loss": 1.944, + "step": 14230 + }, + { + "epoch": 3.11, + "learning_rate": 3.4508411623334067e-06, + "loss": 1.6434, + "step": 14231 + }, + { + "epoch": 3.11, + "learning_rate": 3.4507319204719253e-06, + "loss": 2.572, + "step": 14232 + }, + { + "epoch": 3.11, + "learning_rate": 3.4506226786104435e-06, + "loss": 1.0157, + "step": 14233 + }, + { + "epoch": 3.11, + "learning_rate": 3.4505134367489625e-06, + "loss": 3.0184, + "step": 14234 + }, + { + "epoch": 3.11, + "learning_rate": 3.450404194887481e-06, + "loss": 2.2833, + "step": 14235 + }, + { + "epoch": 3.11, + "learning_rate": 3.450294953026e-06, + "loss": 2.2686, + "step": 14236 + }, + { + "epoch": 3.11, + "learning_rate": 3.4501857111645187e-06, + "loss": 1.7222, + "step": 14237 + }, + { + "epoch": 3.11, + "learning_rate": 3.450076469303037e-06, + "loss": 3.3102, + "step": 14238 + }, + { + "epoch": 3.11, + "learning_rate": 3.449967227441556e-06, + "loss": 1.8433, + "step": 14239 + }, + { + "epoch": 3.11, + "learning_rate": 3.4498579855800746e-06, + "loss": 1.4535, + "step": 14240 + }, + { + "epoch": 3.11, + "learning_rate": 3.449748743718593e-06, + "loss": 1.1343, + "step": 14241 + }, + { + "epoch": 3.11, + "learning_rate": 3.4496395018571122e-06, + "loss": 0.8795, + "step": 14242 + }, + { + "epoch": 3.11, + "learning_rate": 3.4495302599956304e-06, + "loss": 1.8641, + "step": 14243 + }, + { + "epoch": 3.11, + "learning_rate": 3.449421018134149e-06, + "loss": 2.7318, + "step": 14244 + }, + { + "epoch": 3.11, + "learning_rate": 3.449311776272668e-06, + "loss": 2.254, + "step": 14245 + }, + { + "epoch": 3.11, + "learning_rate": 3.4492025344111867e-06, + "loss": 1.2492, + "step": 14246 + }, + { + "epoch": 3.11, + "learning_rate": 3.4490932925497057e-06, + "loss": 1.8379, + "step": 14247 + }, + { + "epoch": 3.11, + "learning_rate": 3.448984050688224e-06, + "loss": 2.2997, + "step": 14248 + }, + { + "epoch": 3.11, + "learning_rate": 3.4488748088267425e-06, + "loss": 2.4195, + "step": 14249 + }, + { + "epoch": 3.11, + "learning_rate": 3.4487655669652615e-06, + "loss": 3.6759, + "step": 14250 + }, + { + "epoch": 3.11, + "learning_rate": 3.44865632510378e-06, + "loss": 1.2692, + "step": 14251 + }, + { + "epoch": 3.11, + "learning_rate": 3.4485470832422983e-06, + "loss": 2.8718, + "step": 14252 + }, + { + "epoch": 3.11, + "learning_rate": 3.4484378413808174e-06, + "loss": 2.552, + "step": 14253 + }, + { + "epoch": 3.11, + "learning_rate": 3.448328599519336e-06, + "loss": 1.3203, + "step": 14254 + }, + { + "epoch": 3.11, + "learning_rate": 3.448219357657855e-06, + "loss": 2.3812, + "step": 14255 + }, + { + "epoch": 3.11, + "learning_rate": 3.4481101157963736e-06, + "loss": 3.4726, + "step": 14256 + }, + { + "epoch": 3.11, + "learning_rate": 3.448000873934892e-06, + "loss": 2.1426, + "step": 14257 + }, + { + "epoch": 3.11, + "learning_rate": 3.447891632073411e-06, + "loss": 1.0249, + "step": 14258 + }, + { + "epoch": 3.11, + "learning_rate": 3.4477823902119295e-06, + "loss": 2.7063, + "step": 14259 + }, + { + "epoch": 3.11, + "learning_rate": 3.4476731483504485e-06, + "loss": 2.3094, + "step": 14260 + }, + { + "epoch": 3.11, + "learning_rate": 3.447563906488967e-06, + "loss": 3.2478, + "step": 14261 + }, + { + "epoch": 3.11, + "learning_rate": 3.4474546646274853e-06, + "loss": 2.9786, + "step": 14262 + }, + { + "epoch": 3.11, + "learning_rate": 3.4473454227660043e-06, + "loss": 1.0481, + "step": 14263 + }, + { + "epoch": 3.11, + "learning_rate": 3.447236180904523e-06, + "loss": 0.3462, + "step": 14264 + }, + { + "epoch": 3.11, + "learning_rate": 3.4471269390430415e-06, + "loss": 1.5791, + "step": 14265 + }, + { + "epoch": 3.11, + "learning_rate": 3.44701769718156e-06, + "loss": 0.8962, + "step": 14266 + }, + { + "epoch": 3.11, + "learning_rate": 3.4469084553200788e-06, + "loss": 1.6622, + "step": 14267 + }, + { + "epoch": 3.11, + "learning_rate": 3.4467992134585974e-06, + "loss": 2.5155, + "step": 14268 + }, + { + "epoch": 3.11, + "learning_rate": 3.4466899715971164e-06, + "loss": 2.3169, + "step": 14269 + }, + { + "epoch": 3.11, + "learning_rate": 3.446580729735635e-06, + "loss": 3.0751, + "step": 14270 + }, + { + "epoch": 3.11, + "learning_rate": 3.4464714878741532e-06, + "loss": 2.9408, + "step": 14271 + }, + { + "epoch": 3.11, + "learning_rate": 3.4463622460126722e-06, + "loss": 2.4415, + "step": 14272 + }, + { + "epoch": 3.12, + "learning_rate": 3.446253004151191e-06, + "loss": 3.0329, + "step": 14273 + }, + { + "epoch": 3.12, + "learning_rate": 3.44614376228971e-06, + "loss": 1.6278, + "step": 14274 + }, + { + "epoch": 3.12, + "learning_rate": 3.4460345204282285e-06, + "loss": 2.043, + "step": 14275 + }, + { + "epoch": 3.12, + "learning_rate": 3.4459252785667467e-06, + "loss": 1.1384, + "step": 14276 + }, + { + "epoch": 3.12, + "learning_rate": 3.4458160367052657e-06, + "loss": 3.0023, + "step": 14277 + }, + { + "epoch": 3.12, + "learning_rate": 3.4457067948437843e-06, + "loss": 1.1122, + "step": 14278 + }, + { + "epoch": 3.12, + "learning_rate": 3.4455975529823034e-06, + "loss": 3.7091, + "step": 14279 + }, + { + "epoch": 3.12, + "learning_rate": 3.445488311120822e-06, + "loss": 2.7331, + "step": 14280 + }, + { + "epoch": 3.12, + "learning_rate": 3.44537906925934e-06, + "loss": 2.5775, + "step": 14281 + }, + { + "epoch": 3.12, + "learning_rate": 3.445269827397859e-06, + "loss": 1.6561, + "step": 14282 + }, + { + "epoch": 3.12, + "learning_rate": 3.445160585536378e-06, + "loss": 3.4539, + "step": 14283 + }, + { + "epoch": 3.12, + "learning_rate": 3.445051343674897e-06, + "loss": 1.6595, + "step": 14284 + }, + { + "epoch": 3.12, + "learning_rate": 3.444942101813415e-06, + "loss": 2.1129, + "step": 14285 + }, + { + "epoch": 3.12, + "learning_rate": 3.4448328599519336e-06, + "loss": 2.0122, + "step": 14286 + }, + { + "epoch": 3.12, + "learning_rate": 3.4447236180904527e-06, + "loss": 2.3056, + "step": 14287 + }, + { + "epoch": 3.12, + "learning_rate": 3.4446143762289713e-06, + "loss": 2.2279, + "step": 14288 + }, + { + "epoch": 3.12, + "learning_rate": 3.4445051343674903e-06, + "loss": 3.767, + "step": 14289 + }, + { + "epoch": 3.12, + "learning_rate": 3.4443958925060085e-06, + "loss": 1.2644, + "step": 14290 + }, + { + "epoch": 3.12, + "learning_rate": 3.444286650644527e-06, + "loss": 4.3311, + "step": 14291 + }, + { + "epoch": 3.12, + "learning_rate": 3.444177408783046e-06, + "loss": 1.5147, + "step": 14292 + }, + { + "epoch": 3.12, + "learning_rate": 3.4440681669215648e-06, + "loss": 1.6826, + "step": 14293 + }, + { + "epoch": 3.12, + "learning_rate": 3.4439589250600834e-06, + "loss": 1.2676, + "step": 14294 + }, + { + "epoch": 3.12, + "learning_rate": 3.443849683198602e-06, + "loss": 2.1869, + "step": 14295 + }, + { + "epoch": 3.12, + "learning_rate": 3.4437404413371206e-06, + "loss": 1.0805, + "step": 14296 + }, + { + "epoch": 3.12, + "learning_rate": 3.4436311994756392e-06, + "loss": 2.7961, + "step": 14297 + }, + { + "epoch": 3.12, + "learning_rate": 3.4435219576141583e-06, + "loss": 3.462, + "step": 14298 + }, + { + "epoch": 3.12, + "learning_rate": 3.4434127157526764e-06, + "loss": 2.6853, + "step": 14299 + }, + { + "epoch": 3.12, + "learning_rate": 3.443303473891195e-06, + "loss": 2.0304, + "step": 14300 + }, + { + "epoch": 3.12, + "learning_rate": 3.443194232029714e-06, + "loss": 2.2536, + "step": 14301 + }, + { + "epoch": 3.12, + "learning_rate": 3.4430849901682327e-06, + "loss": 3.2905, + "step": 14302 + }, + { + "epoch": 3.12, + "learning_rate": 3.4429757483067517e-06, + "loss": 2.3798, + "step": 14303 + }, + { + "epoch": 3.12, + "learning_rate": 3.44286650644527e-06, + "loss": 2.8744, + "step": 14304 + }, + { + "epoch": 3.12, + "learning_rate": 3.4427572645837885e-06, + "loss": 2.2109, + "step": 14305 + }, + { + "epoch": 3.12, + "learning_rate": 3.4426480227223076e-06, + "loss": 1.8641, + "step": 14306 + }, + { + "epoch": 3.12, + "learning_rate": 3.442538780860826e-06, + "loss": 1.565, + "step": 14307 + }, + { + "epoch": 3.12, + "learning_rate": 3.442429538999345e-06, + "loss": 1.9862, + "step": 14308 + }, + { + "epoch": 3.12, + "learning_rate": 3.4423202971378634e-06, + "loss": 2.1073, + "step": 14309 + }, + { + "epoch": 3.12, + "learning_rate": 3.442211055276382e-06, + "loss": 2.6246, + "step": 14310 + }, + { + "epoch": 3.12, + "learning_rate": 3.442101813414901e-06, + "loss": 2.7488, + "step": 14311 + }, + { + "epoch": 3.12, + "learning_rate": 3.4419925715534197e-06, + "loss": 1.9788, + "step": 14312 + }, + { + "epoch": 3.12, + "learning_rate": 3.4418833296919387e-06, + "loss": 2.1089, + "step": 14313 + }, + { + "epoch": 3.12, + "learning_rate": 3.441774087830457e-06, + "loss": 1.5146, + "step": 14314 + }, + { + "epoch": 3.12, + "learning_rate": 3.4416648459689755e-06, + "loss": 2.7632, + "step": 14315 + }, + { + "epoch": 3.12, + "learning_rate": 3.4415556041074945e-06, + "loss": 2.1391, + "step": 14316 + }, + { + "epoch": 3.12, + "learning_rate": 3.441446362246013e-06, + "loss": 2.4235, + "step": 14317 + }, + { + "epoch": 3.12, + "learning_rate": 3.4413371203845313e-06, + "loss": 3.0019, + "step": 14318 + }, + { + "epoch": 3.13, + "learning_rate": 3.4412278785230504e-06, + "loss": 2.0437, + "step": 14319 + }, + { + "epoch": 3.13, + "learning_rate": 3.441118636661569e-06, + "loss": 2.0785, + "step": 14320 + }, + { + "epoch": 3.13, + "learning_rate": 3.4410093948000876e-06, + "loss": 2.9491, + "step": 14321 + }, + { + "epoch": 3.13, + "learning_rate": 3.4409001529386066e-06, + "loss": 2.6723, + "step": 14322 + }, + { + "epoch": 3.13, + "learning_rate": 3.440790911077125e-06, + "loss": 2.1826, + "step": 14323 + }, + { + "epoch": 3.13, + "learning_rate": 3.4406816692156434e-06, + "loss": 0.8072, + "step": 14324 + }, + { + "epoch": 3.13, + "learning_rate": 3.4405724273541624e-06, + "loss": 2.1067, + "step": 14325 + }, + { + "epoch": 3.13, + "learning_rate": 3.440463185492681e-06, + "loss": 3.2666, + "step": 14326 + }, + { + "epoch": 3.13, + "learning_rate": 3.4403539436312e-06, + "loss": 1.893, + "step": 14327 + }, + { + "epoch": 3.13, + "learning_rate": 3.4402447017697183e-06, + "loss": 2.3433, + "step": 14328 + }, + { + "epoch": 3.13, + "learning_rate": 3.440135459908237e-06, + "loss": 1.0509, + "step": 14329 + }, + { + "epoch": 3.13, + "learning_rate": 3.440026218046756e-06, + "loss": 2.1176, + "step": 14330 + }, + { + "epoch": 3.13, + "learning_rate": 3.4399169761852745e-06, + "loss": 2.1378, + "step": 14331 + }, + { + "epoch": 3.13, + "learning_rate": 3.4398077343237927e-06, + "loss": 2.3697, + "step": 14332 + }, + { + "epoch": 3.13, + "learning_rate": 3.4396984924623118e-06, + "loss": 1.9612, + "step": 14333 + }, + { + "epoch": 3.13, + "learning_rate": 3.4395892506008304e-06, + "loss": 1.6902, + "step": 14334 + }, + { + "epoch": 3.13, + "learning_rate": 3.4394800087393494e-06, + "loss": 1.6233, + "step": 14335 + }, + { + "epoch": 3.13, + "learning_rate": 3.439370766877868e-06, + "loss": 2.2487, + "step": 14336 + }, + { + "epoch": 3.13, + "learning_rate": 3.439261525016386e-06, + "loss": 1.8658, + "step": 14337 + }, + { + "epoch": 3.13, + "learning_rate": 3.4391522831549052e-06, + "loss": 2.6759, + "step": 14338 + }, + { + "epoch": 3.13, + "learning_rate": 3.439043041293424e-06, + "loss": 2.0914, + "step": 14339 + }, + { + "epoch": 3.13, + "learning_rate": 3.438933799431943e-06, + "loss": 2.9139, + "step": 14340 + }, + { + "epoch": 3.13, + "learning_rate": 3.4388245575704615e-06, + "loss": 1.392, + "step": 14341 + }, + { + "epoch": 3.13, + "learning_rate": 3.4387153157089797e-06, + "loss": 1.7252, + "step": 14342 + }, + { + "epoch": 3.13, + "learning_rate": 3.4386060738474987e-06, + "loss": 2.4149, + "step": 14343 + }, + { + "epoch": 3.13, + "learning_rate": 3.4384968319860173e-06, + "loss": 2.7177, + "step": 14344 + }, + { + "epoch": 3.13, + "learning_rate": 3.4383875901245364e-06, + "loss": 2.6929, + "step": 14345 + }, + { + "epoch": 3.13, + "learning_rate": 3.438278348263055e-06, + "loss": 2.2522, + "step": 14346 + }, + { + "epoch": 3.13, + "learning_rate": 3.438169106401573e-06, + "loss": 2.5355, + "step": 14347 + }, + { + "epoch": 3.13, + "learning_rate": 3.438059864540092e-06, + "loss": 2.007, + "step": 14348 + }, + { + "epoch": 3.13, + "learning_rate": 3.437950622678611e-06, + "loss": 1.1023, + "step": 14349 + }, + { + "epoch": 3.13, + "learning_rate": 3.4378413808171294e-06, + "loss": 2.1969, + "step": 14350 + }, + { + "epoch": 3.13, + "learning_rate": 3.437732138955648e-06, + "loss": 3.1763, + "step": 14351 + }, + { + "epoch": 3.13, + "learning_rate": 3.4376228970941666e-06, + "loss": 1.8822, + "step": 14352 + }, + { + "epoch": 3.13, + "learning_rate": 3.4375136552326852e-06, + "loss": 2.3723, + "step": 14353 + }, + { + "epoch": 3.13, + "learning_rate": 3.4374044133712043e-06, + "loss": 2.8442, + "step": 14354 + }, + { + "epoch": 3.13, + "learning_rate": 3.437295171509723e-06, + "loss": 2.6419, + "step": 14355 + }, + { + "epoch": 3.13, + "learning_rate": 3.437185929648241e-06, + "loss": 2.2283, + "step": 14356 + }, + { + "epoch": 3.13, + "learning_rate": 3.43707668778676e-06, + "loss": 2.1285, + "step": 14357 + }, + { + "epoch": 3.13, + "learning_rate": 3.4369674459252787e-06, + "loss": 3.135, + "step": 14358 + }, + { + "epoch": 3.13, + "learning_rate": 3.4368582040637978e-06, + "loss": 2.1575, + "step": 14359 + }, + { + "epoch": 3.13, + "learning_rate": 3.4367489622023164e-06, + "loss": 2.5587, + "step": 14360 + }, + { + "epoch": 3.13, + "learning_rate": 3.4366397203408346e-06, + "loss": 3.264, + "step": 14361 + }, + { + "epoch": 3.13, + "learning_rate": 3.4365304784793536e-06, + "loss": 2.7691, + "step": 14362 + }, + { + "epoch": 3.13, + "learning_rate": 3.436421236617872e-06, + "loss": 2.0328, + "step": 14363 + }, + { + "epoch": 3.13, + "learning_rate": 3.4363119947563912e-06, + "loss": 1.5309, + "step": 14364 + }, + { + "epoch": 3.14, + "learning_rate": 3.4362027528949094e-06, + "loss": 2.3045, + "step": 14365 + }, + { + "epoch": 3.14, + "learning_rate": 3.436093511033428e-06, + "loss": 2.3009, + "step": 14366 + }, + { + "epoch": 3.14, + "learning_rate": 3.435984269171947e-06, + "loss": 2.2038, + "step": 14367 + }, + { + "epoch": 3.14, + "learning_rate": 3.4358750273104657e-06, + "loss": 3.253, + "step": 14368 + }, + { + "epoch": 3.14, + "learning_rate": 3.4357657854489847e-06, + "loss": 1.2333, + "step": 14369 + }, + { + "epoch": 3.14, + "learning_rate": 3.435656543587503e-06, + "loss": 2.2988, + "step": 14370 + }, + { + "epoch": 3.14, + "learning_rate": 3.4355473017260215e-06, + "loss": 3.1314, + "step": 14371 + }, + { + "epoch": 3.14, + "learning_rate": 3.4354380598645405e-06, + "loss": 2.2019, + "step": 14372 + }, + { + "epoch": 3.14, + "learning_rate": 3.435328818003059e-06, + "loss": 2.3981, + "step": 14373 + }, + { + "epoch": 3.14, + "learning_rate": 3.4352195761415778e-06, + "loss": 1.4339, + "step": 14374 + }, + { + "epoch": 3.14, + "learning_rate": 3.4351103342800964e-06, + "loss": 0.9256, + "step": 14375 + }, + { + "epoch": 3.14, + "learning_rate": 3.435001092418615e-06, + "loss": 2.1436, + "step": 14376 + }, + { + "epoch": 3.14, + "learning_rate": 3.4348918505571336e-06, + "loss": 1.8033, + "step": 14377 + }, + { + "epoch": 3.14, + "learning_rate": 3.4347826086956526e-06, + "loss": 2.5799, + "step": 14378 + }, + { + "epoch": 3.14, + "learning_rate": 3.4346733668341712e-06, + "loss": 0.8728, + "step": 14379 + }, + { + "epoch": 3.14, + "learning_rate": 3.43456412497269e-06, + "loss": 3.8086, + "step": 14380 + }, + { + "epoch": 3.14, + "learning_rate": 3.4344548831112085e-06, + "loss": 1.4437, + "step": 14381 + }, + { + "epoch": 3.14, + "learning_rate": 3.434345641249727e-06, + "loss": 1.6821, + "step": 14382 + }, + { + "epoch": 3.14, + "learning_rate": 3.434236399388246e-06, + "loss": 3.0676, + "step": 14383 + }, + { + "epoch": 3.14, + "learning_rate": 3.4341271575267643e-06, + "loss": 1.2612, + "step": 14384 + }, + { + "epoch": 3.14, + "learning_rate": 3.434017915665283e-06, + "loss": 1.5406, + "step": 14385 + }, + { + "epoch": 3.14, + "learning_rate": 3.433908673803802e-06, + "loss": 1.5921, + "step": 14386 + }, + { + "epoch": 3.14, + "learning_rate": 3.4337994319423206e-06, + "loss": 2.9995, + "step": 14387 + }, + { + "epoch": 3.14, + "learning_rate": 3.4336901900808396e-06, + "loss": 2.364, + "step": 14388 + }, + { + "epoch": 3.14, + "learning_rate": 3.4335809482193578e-06, + "loss": 2.684, + "step": 14389 + }, + { + "epoch": 3.14, + "learning_rate": 3.4334717063578764e-06, + "loss": 2.1737, + "step": 14390 + }, + { + "epoch": 3.14, + "learning_rate": 3.4333624644963954e-06, + "loss": 1.6278, + "step": 14391 + }, + { + "epoch": 3.14, + "learning_rate": 3.433253222634914e-06, + "loss": 1.4996, + "step": 14392 + }, + { + "epoch": 3.14, + "learning_rate": 3.433143980773433e-06, + "loss": 1.6136, + "step": 14393 + }, + { + "epoch": 3.14, + "learning_rate": 3.4330347389119513e-06, + "loss": 3.5561, + "step": 14394 + }, + { + "epoch": 3.14, + "learning_rate": 3.43292549705047e-06, + "loss": 3.6213, + "step": 14395 + }, + { + "epoch": 3.14, + "learning_rate": 3.432816255188989e-06, + "loss": 1.809, + "step": 14396 + }, + { + "epoch": 3.14, + "learning_rate": 3.4327070133275075e-06, + "loss": 3.6678, + "step": 14397 + }, + { + "epoch": 3.14, + "learning_rate": 3.4325977714660257e-06, + "loss": 2.9128, + "step": 14398 + }, + { + "epoch": 3.14, + "learning_rate": 3.4324885296045447e-06, + "loss": 2.8282, + "step": 14399 + }, + { + "epoch": 3.14, + "learning_rate": 3.4323792877430633e-06, + "loss": 1.8733, + "step": 14400 + }, + { + "epoch": 3.14, + "learning_rate": 3.4322700458815824e-06, + "loss": 1.3261, + "step": 14401 + }, + { + "epoch": 3.14, + "learning_rate": 3.432160804020101e-06, + "loss": 1.5329, + "step": 14402 + }, + { + "epoch": 3.14, + "learning_rate": 3.432051562158619e-06, + "loss": 2.9488, + "step": 14403 + }, + { + "epoch": 3.14, + "learning_rate": 3.4319423202971382e-06, + "loss": 2.2712, + "step": 14404 + }, + { + "epoch": 3.14, + "learning_rate": 3.431833078435657e-06, + "loss": 2.3658, + "step": 14405 + }, + { + "epoch": 3.14, + "learning_rate": 3.4317238365741754e-06, + "loss": 2.0426, + "step": 14406 + }, + { + "epoch": 3.14, + "learning_rate": 3.4316145947126945e-06, + "loss": 1.3044, + "step": 14407 + }, + { + "epoch": 3.14, + "learning_rate": 3.4315053528512127e-06, + "loss": 1.8035, + "step": 14408 + }, + { + "epoch": 3.14, + "learning_rate": 3.4313961109897313e-06, + "loss": 1.5698, + "step": 14409 + }, + { + "epoch": 3.14, + "learning_rate": 3.4312868691282503e-06, + "loss": 3.2038, + "step": 14410 + }, + { + "epoch": 3.15, + "learning_rate": 3.431177627266769e-06, + "loss": 1.9528, + "step": 14411 + }, + { + "epoch": 3.15, + "learning_rate": 3.431068385405288e-06, + "loss": 1.0593, + "step": 14412 + }, + { + "epoch": 3.15, + "learning_rate": 3.430959143543806e-06, + "loss": 2.0526, + "step": 14413 + }, + { + "epoch": 3.15, + "learning_rate": 3.4308499016823247e-06, + "loss": 1.5506, + "step": 14414 + }, + { + "epoch": 3.15, + "learning_rate": 3.4307406598208438e-06, + "loss": 1.8545, + "step": 14415 + }, + { + "epoch": 3.15, + "learning_rate": 3.4306314179593624e-06, + "loss": 1.6372, + "step": 14416 + }, + { + "epoch": 3.15, + "learning_rate": 3.4305221760978806e-06, + "loss": 1.6005, + "step": 14417 + }, + { + "epoch": 3.15, + "learning_rate": 3.4304129342363996e-06, + "loss": 2.5985, + "step": 14418 + }, + { + "epoch": 3.15, + "learning_rate": 3.4303036923749182e-06, + "loss": 1.4962, + "step": 14419 + }, + { + "epoch": 3.15, + "learning_rate": 3.4301944505134373e-06, + "loss": 2.4227, + "step": 14420 + }, + { + "epoch": 3.15, + "learning_rate": 3.430085208651956e-06, + "loss": 1.2588, + "step": 14421 + }, + { + "epoch": 3.15, + "learning_rate": 3.429975966790474e-06, + "loss": 1.0654, + "step": 14422 + }, + { + "epoch": 3.15, + "learning_rate": 3.429866724928993e-06, + "loss": 2.2735, + "step": 14423 + }, + { + "epoch": 3.15, + "learning_rate": 3.4297574830675117e-06, + "loss": 2.7229, + "step": 14424 + }, + { + "epoch": 3.15, + "learning_rate": 3.4296482412060307e-06, + "loss": 2.4535, + "step": 14425 + }, + { + "epoch": 3.15, + "learning_rate": 3.4295389993445494e-06, + "loss": 1.8512, + "step": 14426 + }, + { + "epoch": 3.15, + "learning_rate": 3.4294297574830675e-06, + "loss": 2.4256, + "step": 14427 + }, + { + "epoch": 3.15, + "learning_rate": 3.4293205156215866e-06, + "loss": 1.9148, + "step": 14428 + }, + { + "epoch": 3.15, + "learning_rate": 3.429211273760105e-06, + "loss": 2.7045, + "step": 14429 + }, + { + "epoch": 3.15, + "learning_rate": 3.429102031898624e-06, + "loss": 2.2713, + "step": 14430 + }, + { + "epoch": 3.15, + "learning_rate": 3.4289927900371424e-06, + "loss": 2.3829, + "step": 14431 + }, + { + "epoch": 3.15, + "learning_rate": 3.428883548175661e-06, + "loss": 2.4664, + "step": 14432 + }, + { + "epoch": 3.15, + "learning_rate": 3.42877430631418e-06, + "loss": 1.8818, + "step": 14433 + }, + { + "epoch": 3.15, + "learning_rate": 3.4286650644526987e-06, + "loss": 1.6619, + "step": 14434 + }, + { + "epoch": 3.15, + "learning_rate": 3.4285558225912173e-06, + "loss": 0.715, + "step": 14435 + }, + { + "epoch": 3.15, + "learning_rate": 3.428446580729736e-06, + "loss": 1.4638, + "step": 14436 + }, + { + "epoch": 3.15, + "learning_rate": 3.4283373388682545e-06, + "loss": 2.2389, + "step": 14437 + }, + { + "epoch": 3.15, + "learning_rate": 3.428228097006773e-06, + "loss": 3.0434, + "step": 14438 + }, + { + "epoch": 3.15, + "learning_rate": 3.428118855145292e-06, + "loss": 1.5904, + "step": 14439 + }, + { + "epoch": 3.15, + "learning_rate": 3.4280096132838108e-06, + "loss": 1.9516, + "step": 14440 + }, + { + "epoch": 3.15, + "learning_rate": 3.427900371422329e-06, + "loss": 1.877, + "step": 14441 + }, + { + "epoch": 3.15, + "learning_rate": 3.427791129560848e-06, + "loss": 0.7875, + "step": 14442 + }, + { + "epoch": 3.15, + "learning_rate": 3.4276818876993666e-06, + "loss": 2.083, + "step": 14443 + }, + { + "epoch": 3.15, + "learning_rate": 3.4275726458378856e-06, + "loss": 2.3059, + "step": 14444 + }, + { + "epoch": 3.15, + "learning_rate": 3.4274634039764042e-06, + "loss": 2.4684, + "step": 14445 + }, + { + "epoch": 3.15, + "learning_rate": 3.4273541621149224e-06, + "loss": 1.8682, + "step": 14446 + }, + { + "epoch": 3.15, + "learning_rate": 3.4272449202534415e-06, + "loss": 2.24, + "step": 14447 + }, + { + "epoch": 3.15, + "learning_rate": 3.42713567839196e-06, + "loss": 2.8726, + "step": 14448 + }, + { + "epoch": 3.15, + "learning_rate": 3.427026436530479e-06, + "loss": 1.5032, + "step": 14449 + }, + { + "epoch": 3.15, + "learning_rate": 3.4269171946689973e-06, + "loss": 2.2957, + "step": 14450 + }, + { + "epoch": 3.15, + "learning_rate": 3.426807952807516e-06, + "loss": 1.4283, + "step": 14451 + }, + { + "epoch": 3.15, + "learning_rate": 3.426698710946035e-06, + "loss": 2.41, + "step": 14452 + }, + { + "epoch": 3.15, + "learning_rate": 3.4265894690845535e-06, + "loss": 2.2325, + "step": 14453 + }, + { + "epoch": 3.15, + "learning_rate": 3.4264802272230726e-06, + "loss": 2.197, + "step": 14454 + }, + { + "epoch": 3.15, + "learning_rate": 3.4263709853615908e-06, + "loss": 1.6287, + "step": 14455 + }, + { + "epoch": 3.15, + "learning_rate": 3.4262617435001094e-06, + "loss": 2.1098, + "step": 14456 + }, + { + "epoch": 3.16, + "learning_rate": 3.4261525016386284e-06, + "loss": 2.8536, + "step": 14457 + }, + { + "epoch": 3.16, + "learning_rate": 3.426043259777147e-06, + "loss": 1.154, + "step": 14458 + }, + { + "epoch": 3.16, + "learning_rate": 3.4259340179156656e-06, + "loss": 2.5045, + "step": 14459 + }, + { + "epoch": 3.16, + "learning_rate": 3.4258247760541842e-06, + "loss": 3.1923, + "step": 14460 + }, + { + "epoch": 3.16, + "learning_rate": 3.425715534192703e-06, + "loss": 1.4637, + "step": 14461 + }, + { + "epoch": 3.16, + "learning_rate": 3.4256062923312215e-06, + "loss": 2.839, + "step": 14462 + }, + { + "epoch": 3.16, + "learning_rate": 3.4254970504697405e-06, + "loss": 2.0093, + "step": 14463 + }, + { + "epoch": 3.16, + "learning_rate": 3.4253878086082587e-06, + "loss": 1.1576, + "step": 14464 + }, + { + "epoch": 3.16, + "learning_rate": 3.4252785667467773e-06, + "loss": 1.3447, + "step": 14465 + }, + { + "epoch": 3.16, + "learning_rate": 3.4251693248852963e-06, + "loss": 0.5308, + "step": 14466 + }, + { + "epoch": 3.16, + "learning_rate": 3.425060083023815e-06, + "loss": 1.0448, + "step": 14467 + }, + { + "epoch": 3.16, + "learning_rate": 3.424950841162334e-06, + "loss": 1.8759, + "step": 14468 + }, + { + "epoch": 3.16, + "learning_rate": 3.424841599300852e-06, + "loss": 1.8557, + "step": 14469 + }, + { + "epoch": 3.16, + "learning_rate": 3.4247323574393708e-06, + "loss": 1.1707, + "step": 14470 + }, + { + "epoch": 3.16, + "learning_rate": 3.42462311557789e-06, + "loss": 2.5802, + "step": 14471 + }, + { + "epoch": 3.16, + "learning_rate": 3.4245138737164084e-06, + "loss": 2.0544, + "step": 14472 + }, + { + "epoch": 3.16, + "learning_rate": 3.4244046318549275e-06, + "loss": 0.8039, + "step": 14473 + }, + { + "epoch": 3.16, + "learning_rate": 3.4242953899934456e-06, + "loss": 2.7219, + "step": 14474 + }, + { + "epoch": 3.16, + "learning_rate": 3.4241861481319643e-06, + "loss": 2.1291, + "step": 14475 + }, + { + "epoch": 3.16, + "learning_rate": 3.4240769062704833e-06, + "loss": 2.3779, + "step": 14476 + }, + { + "epoch": 3.16, + "learning_rate": 3.423967664409002e-06, + "loss": 1.686, + "step": 14477 + }, + { + "epoch": 3.16, + "learning_rate": 3.423858422547521e-06, + "loss": 2.0108, + "step": 14478 + }, + { + "epoch": 3.16, + "learning_rate": 3.423749180686039e-06, + "loss": 1.7859, + "step": 14479 + }, + { + "epoch": 3.16, + "learning_rate": 3.4236399388245577e-06, + "loss": 2.9473, + "step": 14480 + }, + { + "epoch": 3.16, + "learning_rate": 3.4235306969630768e-06, + "loss": 1.6928, + "step": 14481 + }, + { + "epoch": 3.16, + "learning_rate": 3.4234214551015954e-06, + "loss": 1.1533, + "step": 14482 + }, + { + "epoch": 3.16, + "learning_rate": 3.4233122132401136e-06, + "loss": 2.4419, + "step": 14483 + }, + { + "epoch": 3.16, + "learning_rate": 3.4232029713786326e-06, + "loss": 2.0367, + "step": 14484 + }, + { + "epoch": 3.16, + "learning_rate": 3.423093729517151e-06, + "loss": 2.7925, + "step": 14485 + }, + { + "epoch": 3.16, + "learning_rate": 3.4229844876556702e-06, + "loss": 2.1271, + "step": 14486 + }, + { + "epoch": 3.16, + "learning_rate": 3.422875245794189e-06, + "loss": 1.6152, + "step": 14487 + }, + { + "epoch": 3.16, + "learning_rate": 3.422766003932707e-06, + "loss": 1.8266, + "step": 14488 + }, + { + "epoch": 3.16, + "learning_rate": 3.422656762071226e-06, + "loss": 1.673, + "step": 14489 + }, + { + "epoch": 3.16, + "learning_rate": 3.4225475202097447e-06, + "loss": 2.1355, + "step": 14490 + }, + { + "epoch": 3.16, + "learning_rate": 3.4224382783482633e-06, + "loss": 1.4666, + "step": 14491 + }, + { + "epoch": 3.16, + "learning_rate": 3.4223290364867823e-06, + "loss": 1.5211, + "step": 14492 + }, + { + "epoch": 3.16, + "learning_rate": 3.4222197946253005e-06, + "loss": 1.5524, + "step": 14493 + }, + { + "epoch": 3.16, + "learning_rate": 3.422110552763819e-06, + "loss": 3.1823, + "step": 14494 + }, + { + "epoch": 3.16, + "learning_rate": 3.422001310902338e-06, + "loss": 2.3571, + "step": 14495 + }, + { + "epoch": 3.16, + "learning_rate": 3.4218920690408568e-06, + "loss": 1.0922, + "step": 14496 + }, + { + "epoch": 3.16, + "learning_rate": 3.421782827179375e-06, + "loss": 2.5218, + "step": 14497 + }, + { + "epoch": 3.16, + "learning_rate": 3.421673585317894e-06, + "loss": 3.195, + "step": 14498 + }, + { + "epoch": 3.16, + "learning_rate": 3.4215643434564126e-06, + "loss": 2.1125, + "step": 14499 + }, + { + "epoch": 3.16, + "learning_rate": 3.4214551015949316e-06, + "loss": 1.7183, + "step": 14500 + }, + { + "epoch": 3.16, + "learning_rate": 3.4213458597334503e-06, + "loss": 1.9991, + "step": 14501 + }, + { + "epoch": 3.16, + "learning_rate": 3.4212366178719684e-06, + "loss": 2.8527, + "step": 14502 + }, + { + "epoch": 3.17, + "learning_rate": 3.4211273760104875e-06, + "loss": 2.6443, + "step": 14503 + }, + { + "epoch": 3.17, + "learning_rate": 3.421018134149006e-06, + "loss": 2.563, + "step": 14504 + }, + { + "epoch": 3.17, + "learning_rate": 3.420908892287525e-06, + "loss": 2.7107, + "step": 14505 + }, + { + "epoch": 3.17, + "learning_rate": 3.4207996504260437e-06, + "loss": 1.032, + "step": 14506 + }, + { + "epoch": 3.17, + "learning_rate": 3.420690408564562e-06, + "loss": 1.6534, + "step": 14507 + }, + { + "epoch": 3.17, + "learning_rate": 3.420581166703081e-06, + "loss": 2.4885, + "step": 14508 + }, + { + "epoch": 3.17, + "learning_rate": 3.4204719248415996e-06, + "loss": 2.6801, + "step": 14509 + }, + { + "epoch": 3.17, + "learning_rate": 3.4203626829801186e-06, + "loss": 1.121, + "step": 14510 + }, + { + "epoch": 3.17, + "learning_rate": 3.4202534411186372e-06, + "loss": 2.4174, + "step": 14511 + }, + { + "epoch": 3.17, + "learning_rate": 3.4201441992571554e-06, + "loss": 2.1842, + "step": 14512 + }, + { + "epoch": 3.17, + "learning_rate": 3.4200349573956744e-06, + "loss": 1.506, + "step": 14513 + }, + { + "epoch": 3.17, + "learning_rate": 3.419925715534193e-06, + "loss": 2.5708, + "step": 14514 + }, + { + "epoch": 3.17, + "learning_rate": 3.4198164736727117e-06, + "loss": 2.8642, + "step": 14515 + }, + { + "epoch": 3.17, + "learning_rate": 3.4197072318112303e-06, + "loss": 1.7077, + "step": 14516 + }, + { + "epoch": 3.17, + "learning_rate": 3.419597989949749e-06, + "loss": 3.6922, + "step": 14517 + }, + { + "epoch": 3.17, + "learning_rate": 3.4194887480882675e-06, + "loss": 2.4625, + "step": 14518 + }, + { + "epoch": 3.17, + "learning_rate": 3.4193795062267865e-06, + "loss": 3.0, + "step": 14519 + }, + { + "epoch": 3.17, + "learning_rate": 3.419270264365305e-06, + "loss": 2.4305, + "step": 14520 + }, + { + "epoch": 3.17, + "learning_rate": 3.4191610225038233e-06, + "loss": 2.5813, + "step": 14521 + }, + { + "epoch": 3.17, + "learning_rate": 3.4190517806423424e-06, + "loss": 2.2905, + "step": 14522 + }, + { + "epoch": 3.17, + "learning_rate": 3.418942538780861e-06, + "loss": 1.4127, + "step": 14523 + }, + { + "epoch": 3.17, + "learning_rate": 3.41883329691938e-06, + "loss": 1.6507, + "step": 14524 + }, + { + "epoch": 3.17, + "learning_rate": 3.4187240550578986e-06, + "loss": 1.7948, + "step": 14525 + }, + { + "epoch": 3.17, + "learning_rate": 3.418614813196417e-06, + "loss": 2.4135, + "step": 14526 + }, + { + "epoch": 3.17, + "learning_rate": 3.418505571334936e-06, + "loss": 2.7392, + "step": 14527 + }, + { + "epoch": 3.17, + "learning_rate": 3.4183963294734544e-06, + "loss": 2.2477, + "step": 14528 + }, + { + "epoch": 3.17, + "learning_rate": 3.4182870876119735e-06, + "loss": 1.8216, + "step": 14529 + }, + { + "epoch": 3.17, + "learning_rate": 3.4181778457504917e-06, + "loss": 1.6889, + "step": 14530 + }, + { + "epoch": 3.17, + "learning_rate": 3.4180686038890103e-06, + "loss": 2.2342, + "step": 14531 + }, + { + "epoch": 3.17, + "learning_rate": 3.4179593620275293e-06, + "loss": 1.3909, + "step": 14532 + }, + { + "epoch": 3.17, + "learning_rate": 3.417850120166048e-06, + "loss": 1.9071, + "step": 14533 + }, + { + "epoch": 3.17, + "learning_rate": 3.417740878304567e-06, + "loss": 2.5424, + "step": 14534 + }, + { + "epoch": 3.17, + "learning_rate": 3.417631636443085e-06, + "loss": 1.9142, + "step": 14535 + }, + { + "epoch": 3.17, + "learning_rate": 3.4175223945816038e-06, + "loss": 2.0819, + "step": 14536 + }, + { + "epoch": 3.17, + "learning_rate": 3.417413152720123e-06, + "loss": 1.3529, + "step": 14537 + }, + { + "epoch": 3.17, + "learning_rate": 3.4173039108586414e-06, + "loss": 1.39, + "step": 14538 + }, + { + "epoch": 3.17, + "learning_rate": 3.4171946689971604e-06, + "loss": 1.2607, + "step": 14539 + }, + { + "epoch": 3.17, + "learning_rate": 3.4170854271356786e-06, + "loss": 2.0353, + "step": 14540 + }, + { + "epoch": 3.17, + "learning_rate": 3.4169761852741972e-06, + "loss": 1.8194, + "step": 14541 + }, + { + "epoch": 3.17, + "learning_rate": 3.4168669434127163e-06, + "loss": 1.3637, + "step": 14542 + }, + { + "epoch": 3.17, + "learning_rate": 3.416757701551235e-06, + "loss": 3.2416, + "step": 14543 + }, + { + "epoch": 3.17, + "learning_rate": 3.416648459689753e-06, + "loss": 1.5368, + "step": 14544 + }, + { + "epoch": 3.17, + "learning_rate": 3.416539217828272e-06, + "loss": 0.9867, + "step": 14545 + }, + { + "epoch": 3.17, + "learning_rate": 3.4164299759667907e-06, + "loss": 1.7785, + "step": 14546 + }, + { + "epoch": 3.17, + "learning_rate": 3.4163207341053093e-06, + "loss": 1.9688, + "step": 14547 + }, + { + "epoch": 3.18, + "learning_rate": 3.4162114922438284e-06, + "loss": 2.9767, + "step": 14548 + }, + { + "epoch": 3.18, + "learning_rate": 3.4161022503823465e-06, + "loss": 1.4848, + "step": 14549 + }, + { + "epoch": 3.18, + "learning_rate": 3.415993008520865e-06, + "loss": 0.7645, + "step": 14550 + }, + { + "epoch": 3.18, + "learning_rate": 3.415883766659384e-06, + "loss": 2.0882, + "step": 14551 + }, + { + "epoch": 3.18, + "learning_rate": 3.415774524797903e-06, + "loss": 1.742, + "step": 14552 + }, + { + "epoch": 3.18, + "learning_rate": 3.415665282936422e-06, + "loss": 2.4114, + "step": 14553 + }, + { + "epoch": 3.18, + "learning_rate": 3.41555604107494e-06, + "loss": 3.5178, + "step": 14554 + }, + { + "epoch": 3.18, + "learning_rate": 3.4154467992134586e-06, + "loss": 2.0308, + "step": 14555 + }, + { + "epoch": 3.18, + "learning_rate": 3.4153375573519777e-06, + "loss": 2.8202, + "step": 14556 + }, + { + "epoch": 3.18, + "learning_rate": 3.4152283154904963e-06, + "loss": 2.887, + "step": 14557 + }, + { + "epoch": 3.18, + "learning_rate": 3.4151190736290153e-06, + "loss": 2.3217, + "step": 14558 + }, + { + "epoch": 3.18, + "learning_rate": 3.4150098317675335e-06, + "loss": 1.6144, + "step": 14559 + }, + { + "epoch": 3.18, + "learning_rate": 3.414900589906052e-06, + "loss": 2.1734, + "step": 14560 + }, + { + "epoch": 3.18, + "learning_rate": 3.414791348044571e-06, + "loss": 2.0983, + "step": 14561 + }, + { + "epoch": 3.18, + "learning_rate": 3.4146821061830898e-06, + "loss": 3.3902, + "step": 14562 + }, + { + "epoch": 3.18, + "learning_rate": 3.414572864321608e-06, + "loss": 1.9023, + "step": 14563 + }, + { + "epoch": 3.18, + "learning_rate": 3.414463622460127e-06, + "loss": 2.2934, + "step": 14564 + }, + { + "epoch": 3.18, + "learning_rate": 3.4143543805986456e-06, + "loss": 1.9207, + "step": 14565 + }, + { + "epoch": 3.18, + "learning_rate": 3.4142451387371646e-06, + "loss": 1.867, + "step": 14566 + }, + { + "epoch": 3.18, + "learning_rate": 3.4141358968756832e-06, + "loss": 3.3053, + "step": 14567 + }, + { + "epoch": 3.18, + "learning_rate": 3.4140266550142014e-06, + "loss": 2.3064, + "step": 14568 + }, + { + "epoch": 3.18, + "learning_rate": 3.4139174131527205e-06, + "loss": 2.7687, + "step": 14569 + }, + { + "epoch": 3.18, + "learning_rate": 3.413808171291239e-06, + "loss": 1.79, + "step": 14570 + }, + { + "epoch": 3.18, + "learning_rate": 3.4136989294297577e-06, + "loss": 2.3214, + "step": 14571 + }, + { + "epoch": 3.18, + "learning_rate": 3.4135896875682767e-06, + "loss": 3.1327, + "step": 14572 + }, + { + "epoch": 3.18, + "learning_rate": 3.413480445706795e-06, + "loss": 2.1089, + "step": 14573 + }, + { + "epoch": 3.18, + "learning_rate": 3.413371203845314e-06, + "loss": 3.0067, + "step": 14574 + }, + { + "epoch": 3.18, + "learning_rate": 3.4132619619838326e-06, + "loss": 2.3712, + "step": 14575 + }, + { + "epoch": 3.18, + "learning_rate": 3.413152720122351e-06, + "loss": 1.7475, + "step": 14576 + }, + { + "epoch": 3.18, + "learning_rate": 3.4130434782608698e-06, + "loss": 1.3859, + "step": 14577 + }, + { + "epoch": 3.18, + "learning_rate": 3.4129342363993884e-06, + "loss": 2.4291, + "step": 14578 + }, + { + "epoch": 3.18, + "learning_rate": 3.412824994537907e-06, + "loss": 2.4159, + "step": 14579 + }, + { + "epoch": 3.18, + "learning_rate": 3.412715752676426e-06, + "loss": 2.3719, + "step": 14580 + }, + { + "epoch": 3.18, + "learning_rate": 3.4126065108149446e-06, + "loss": 2.6795, + "step": 14581 + }, + { + "epoch": 3.18, + "learning_rate": 3.412497268953463e-06, + "loss": 1.9545, + "step": 14582 + }, + { + "epoch": 3.18, + "learning_rate": 3.412388027091982e-06, + "loss": 1.7037, + "step": 14583 + }, + { + "epoch": 3.18, + "learning_rate": 3.4122787852305005e-06, + "loss": 2.3889, + "step": 14584 + }, + { + "epoch": 3.18, + "learning_rate": 3.4121695433690195e-06, + "loss": 2.3846, + "step": 14585 + }, + { + "epoch": 3.18, + "learning_rate": 3.412060301507538e-06, + "loss": 2.0619, + "step": 14586 + }, + { + "epoch": 3.18, + "learning_rate": 3.4119510596460563e-06, + "loss": 0.446, + "step": 14587 + }, + { + "epoch": 3.18, + "learning_rate": 3.4118418177845753e-06, + "loss": 2.154, + "step": 14588 + }, + { + "epoch": 3.18, + "learning_rate": 3.411732575923094e-06, + "loss": 1.3516, + "step": 14589 + }, + { + "epoch": 3.18, + "learning_rate": 3.411623334061613e-06, + "loss": 2.9273, + "step": 14590 + }, + { + "epoch": 3.18, + "learning_rate": 3.4115140922001316e-06, + "loss": 1.2003, + "step": 14591 + }, + { + "epoch": 3.18, + "learning_rate": 3.4114048503386498e-06, + "loss": 1.8341, + "step": 14592 + }, + { + "epoch": 3.18, + "learning_rate": 3.411295608477169e-06, + "loss": 2.7472, + "step": 14593 + }, + { + "epoch": 3.19, + "learning_rate": 3.4111863666156874e-06, + "loss": 3.258, + "step": 14594 + }, + { + "epoch": 3.19, + "learning_rate": 3.4110771247542065e-06, + "loss": 1.2529, + "step": 14595 + }, + { + "epoch": 3.19, + "learning_rate": 3.4109678828927247e-06, + "loss": 1.7312, + "step": 14596 + }, + { + "epoch": 3.19, + "learning_rate": 3.4108586410312433e-06, + "loss": 0.665, + "step": 14597 + }, + { + "epoch": 3.19, + "learning_rate": 3.4107493991697623e-06, + "loss": 1.0523, + "step": 14598 + }, + { + "epoch": 3.19, + "learning_rate": 3.410640157308281e-06, + "loss": 1.0959, + "step": 14599 + }, + { + "epoch": 3.19, + "learning_rate": 3.4105309154467995e-06, + "loss": 1.5348, + "step": 14600 + }, + { + "epoch": 3.19, + "learning_rate": 3.410421673585318e-06, + "loss": 1.8086, + "step": 14601 + }, + { + "epoch": 3.19, + "learning_rate": 3.4103124317238367e-06, + "loss": 1.6206, + "step": 14602 + }, + { + "epoch": 3.19, + "learning_rate": 3.4102031898623554e-06, + "loss": 1.8133, + "step": 14603 + }, + { + "epoch": 3.19, + "learning_rate": 3.4100939480008744e-06, + "loss": 2.3125, + "step": 14604 + }, + { + "epoch": 3.19, + "learning_rate": 3.409984706139393e-06, + "loss": 1.3504, + "step": 14605 + }, + { + "epoch": 3.19, + "learning_rate": 3.409875464277911e-06, + "loss": 1.8058, + "step": 14606 + }, + { + "epoch": 3.19, + "learning_rate": 3.4097662224164302e-06, + "loss": 2.5337, + "step": 14607 + }, + { + "epoch": 3.19, + "learning_rate": 3.409656980554949e-06, + "loss": 2.3629, + "step": 14608 + }, + { + "epoch": 3.19, + "learning_rate": 3.409547738693468e-06, + "loss": 3.2496, + "step": 14609 + }, + { + "epoch": 3.19, + "learning_rate": 3.409438496831986e-06, + "loss": 3.1934, + "step": 14610 + }, + { + "epoch": 3.19, + "learning_rate": 3.4093292549705047e-06, + "loss": 2.6203, + "step": 14611 + }, + { + "epoch": 3.19, + "learning_rate": 3.4092200131090237e-06, + "loss": 2.6797, + "step": 14612 + }, + { + "epoch": 3.19, + "learning_rate": 3.4091107712475423e-06, + "loss": 2.196, + "step": 14613 + }, + { + "epoch": 3.19, + "learning_rate": 3.4090015293860613e-06, + "loss": 2.5279, + "step": 14614 + }, + { + "epoch": 3.19, + "learning_rate": 3.4088922875245795e-06, + "loss": 2.2945, + "step": 14615 + }, + { + "epoch": 3.19, + "learning_rate": 3.408783045663098e-06, + "loss": 2.2668, + "step": 14616 + }, + { + "epoch": 3.19, + "learning_rate": 3.408673803801617e-06, + "loss": 2.4452, + "step": 14617 + }, + { + "epoch": 3.19, + "learning_rate": 3.408564561940136e-06, + "loss": 0.9859, + "step": 14618 + }, + { + "epoch": 3.19, + "learning_rate": 3.408455320078655e-06, + "loss": 1.6725, + "step": 14619 + }, + { + "epoch": 3.19, + "learning_rate": 3.408346078217173e-06, + "loss": 1.5022, + "step": 14620 + }, + { + "epoch": 3.19, + "learning_rate": 3.4082368363556916e-06, + "loss": 1.7791, + "step": 14621 + }, + { + "epoch": 3.19, + "learning_rate": 3.4081275944942107e-06, + "loss": 1.6928, + "step": 14622 + }, + { + "epoch": 3.19, + "learning_rate": 3.4080183526327293e-06, + "loss": 0.6965, + "step": 14623 + }, + { + "epoch": 3.19, + "learning_rate": 3.407909110771248e-06, + "loss": 3.1657, + "step": 14624 + }, + { + "epoch": 3.19, + "learning_rate": 3.4077998689097665e-06, + "loss": 3.033, + "step": 14625 + }, + { + "epoch": 3.19, + "learning_rate": 3.407690627048285e-06, + "loss": 3.0971, + "step": 14626 + }, + { + "epoch": 3.19, + "learning_rate": 3.407581385186804e-06, + "loss": 2.1551, + "step": 14627 + }, + { + "epoch": 3.19, + "learning_rate": 3.4074721433253227e-06, + "loss": 1.6644, + "step": 14628 + }, + { + "epoch": 3.19, + "learning_rate": 3.407362901463841e-06, + "loss": 1.5489, + "step": 14629 + }, + { + "epoch": 3.19, + "learning_rate": 3.40725365960236e-06, + "loss": 1.7462, + "step": 14630 + }, + { + "epoch": 3.19, + "learning_rate": 3.4071444177408786e-06, + "loss": 2.4545, + "step": 14631 + }, + { + "epoch": 3.19, + "learning_rate": 3.407035175879397e-06, + "loss": 1.2395, + "step": 14632 + }, + { + "epoch": 3.19, + "learning_rate": 3.4069259340179162e-06, + "loss": 1.7354, + "step": 14633 + }, + { + "epoch": 3.19, + "learning_rate": 3.4068166921564344e-06, + "loss": 2.0193, + "step": 14634 + }, + { + "epoch": 3.19, + "learning_rate": 3.406707450294953e-06, + "loss": 2.4664, + "step": 14635 + }, + { + "epoch": 3.19, + "learning_rate": 3.406598208433472e-06, + "loss": 1.5175, + "step": 14636 + }, + { + "epoch": 3.19, + "learning_rate": 3.4064889665719907e-06, + "loss": 2.2502, + "step": 14637 + }, + { + "epoch": 3.19, + "learning_rate": 3.4063797247105097e-06, + "loss": 2.3122, + "step": 14638 + }, + { + "epoch": 3.19, + "learning_rate": 3.406270482849028e-06, + "loss": 0.4718, + "step": 14639 + }, + { + "epoch": 3.2, + "learning_rate": 3.4061612409875465e-06, + "loss": 1.5965, + "step": 14640 + }, + { + "epoch": 3.2, + "learning_rate": 3.4060519991260655e-06, + "loss": 1.6471, + "step": 14641 + }, + { + "epoch": 3.2, + "learning_rate": 3.405942757264584e-06, + "loss": 2.3183, + "step": 14642 + }, + { + "epoch": 3.2, + "learning_rate": 3.4058335154031023e-06, + "loss": 2.1341, + "step": 14643 + }, + { + "epoch": 3.2, + "learning_rate": 3.4057242735416214e-06, + "loss": 0.7684, + "step": 14644 + }, + { + "epoch": 3.2, + "learning_rate": 3.40561503168014e-06, + "loss": 1.9154, + "step": 14645 + }, + { + "epoch": 3.2, + "learning_rate": 3.405505789818659e-06, + "loss": 2.2126, + "step": 14646 + }, + { + "epoch": 3.2, + "learning_rate": 3.4053965479571776e-06, + "loss": 2.1085, + "step": 14647 + }, + { + "epoch": 3.2, + "learning_rate": 3.405287306095696e-06, + "loss": 2.7711, + "step": 14648 + }, + { + "epoch": 3.2, + "learning_rate": 3.405178064234215e-06, + "loss": 1.4075, + "step": 14649 + }, + { + "epoch": 3.2, + "learning_rate": 3.4050688223727335e-06, + "loss": 3.5396, + "step": 14650 + }, + { + "epoch": 3.2, + "learning_rate": 3.4049595805112525e-06, + "loss": 0.9747, + "step": 14651 + }, + { + "epoch": 3.2, + "learning_rate": 3.404850338649771e-06, + "loss": 1.3563, + "step": 14652 + }, + { + "epoch": 3.2, + "learning_rate": 3.4047410967882893e-06, + "loss": 1.7985, + "step": 14653 + }, + { + "epoch": 3.2, + "learning_rate": 3.4046318549268083e-06, + "loss": 1.0724, + "step": 14654 + }, + { + "epoch": 3.2, + "learning_rate": 3.404522613065327e-06, + "loss": 0.7143, + "step": 14655 + }, + { + "epoch": 3.2, + "learning_rate": 3.4044133712038455e-06, + "loss": 1.7545, + "step": 14656 + }, + { + "epoch": 3.2, + "learning_rate": 3.4043041293423646e-06, + "loss": 1.345, + "step": 14657 + }, + { + "epoch": 3.2, + "learning_rate": 3.4041948874808828e-06, + "loss": 1.3608, + "step": 14658 + }, + { + "epoch": 3.2, + "learning_rate": 3.4040856456194014e-06, + "loss": 2.7645, + "step": 14659 + }, + { + "epoch": 3.2, + "learning_rate": 3.4039764037579204e-06, + "loss": 3.094, + "step": 14660 + }, + { + "epoch": 3.2, + "learning_rate": 3.403867161896439e-06, + "loss": 1.2407, + "step": 14661 + }, + { + "epoch": 3.2, + "learning_rate": 3.4037579200349572e-06, + "loss": 2.6183, + "step": 14662 + }, + { + "epoch": 3.2, + "learning_rate": 3.4036486781734762e-06, + "loss": 1.3258, + "step": 14663 + }, + { + "epoch": 3.2, + "learning_rate": 3.403539436311995e-06, + "loss": 2.044, + "step": 14664 + }, + { + "epoch": 3.2, + "learning_rate": 3.403430194450514e-06, + "loss": 2.2908, + "step": 14665 + }, + { + "epoch": 3.2, + "learning_rate": 3.4033209525890325e-06, + "loss": 3.2768, + "step": 14666 + }, + { + "epoch": 3.2, + "learning_rate": 3.4032117107275507e-06, + "loss": 2.6068, + "step": 14667 + }, + { + "epoch": 3.2, + "learning_rate": 3.4031024688660697e-06, + "loss": 1.2916, + "step": 14668 + }, + { + "epoch": 3.2, + "learning_rate": 3.4029932270045883e-06, + "loss": 2.1581, + "step": 14669 + }, + { + "epoch": 3.2, + "learning_rate": 3.4028839851431074e-06, + "loss": 1.5626, + "step": 14670 + }, + { + "epoch": 3.2, + "learning_rate": 3.402774743281626e-06, + "loss": 1.8775, + "step": 14671 + }, + { + "epoch": 3.2, + "learning_rate": 3.402665501420144e-06, + "loss": 2.5879, + "step": 14672 + }, + { + "epoch": 3.2, + "learning_rate": 3.402556259558663e-06, + "loss": 2.0294, + "step": 14673 + }, + { + "epoch": 3.2, + "learning_rate": 3.402447017697182e-06, + "loss": 2.9476, + "step": 14674 + }, + { + "epoch": 3.2, + "learning_rate": 3.402337775835701e-06, + "loss": 0.9734, + "step": 14675 + }, + { + "epoch": 3.2, + "learning_rate": 3.402228533974219e-06, + "loss": 1.9796, + "step": 14676 + }, + { + "epoch": 3.2, + "learning_rate": 3.4021192921127376e-06, + "loss": 1.3524, + "step": 14677 + }, + { + "epoch": 3.2, + "learning_rate": 3.4020100502512567e-06, + "loss": 2.3527, + "step": 14678 + }, + { + "epoch": 3.2, + "learning_rate": 3.4019008083897753e-06, + "loss": 3.0028, + "step": 14679 + }, + { + "epoch": 3.2, + "learning_rate": 3.4017915665282943e-06, + "loss": 2.1136, + "step": 14680 + }, + { + "epoch": 3.2, + "learning_rate": 3.4016823246668125e-06, + "loss": 2.7019, + "step": 14681 + }, + { + "epoch": 3.2, + "learning_rate": 3.401573082805331e-06, + "loss": 0.895, + "step": 14682 + }, + { + "epoch": 3.2, + "learning_rate": 3.40146384094385e-06, + "loss": 2.6426, + "step": 14683 + }, + { + "epoch": 3.2, + "learning_rate": 3.4013545990823688e-06, + "loss": 1.9969, + "step": 14684 + }, + { + "epoch": 3.2, + "learning_rate": 3.4012453572208874e-06, + "loss": 2.2912, + "step": 14685 + }, + { + "epoch": 3.21, + "learning_rate": 3.401136115359406e-06, + "loss": 1.9812, + "step": 14686 + }, + { + "epoch": 3.21, + "learning_rate": 3.4010268734979246e-06, + "loss": 1.9164, + "step": 14687 + }, + { + "epoch": 3.21, + "learning_rate": 3.4009176316364432e-06, + "loss": 1.3062, + "step": 14688 + }, + { + "epoch": 3.21, + "learning_rate": 3.4008083897749623e-06, + "loss": 3.4943, + "step": 14689 + }, + { + "epoch": 3.21, + "learning_rate": 3.400699147913481e-06, + "loss": 3.0098, + "step": 14690 + }, + { + "epoch": 3.21, + "learning_rate": 3.400589906051999e-06, + "loss": 2.7807, + "step": 14691 + }, + { + "epoch": 3.21, + "learning_rate": 3.400480664190518e-06, + "loss": 0.7328, + "step": 14692 + }, + { + "epoch": 3.21, + "learning_rate": 3.4003714223290367e-06, + "loss": 1.2189, + "step": 14693 + }, + { + "epoch": 3.21, + "learning_rate": 3.4002621804675557e-06, + "loss": 1.1336, + "step": 14694 + }, + { + "epoch": 3.21, + "learning_rate": 3.400152938606074e-06, + "loss": 1.4119, + "step": 14695 + }, + { + "epoch": 3.21, + "learning_rate": 3.4000436967445925e-06, + "loss": 2.4211, + "step": 14696 + }, + { + "epoch": 3.21, + "learning_rate": 3.3999344548831116e-06, + "loss": 1.2822, + "step": 14697 + }, + { + "epoch": 3.21, + "learning_rate": 3.39982521302163e-06, + "loss": 2.8128, + "step": 14698 + }, + { + "epoch": 3.21, + "learning_rate": 3.399715971160149e-06, + "loss": 2.1416, + "step": 14699 + }, + { + "epoch": 3.21, + "learning_rate": 3.3996067292986674e-06, + "loss": 1.5138, + "step": 14700 + }, + { + "epoch": 3.21, + "learning_rate": 3.399497487437186e-06, + "loss": 1.9724, + "step": 14701 + }, + { + "epoch": 3.21, + "learning_rate": 3.399388245575705e-06, + "loss": 2.3641, + "step": 14702 + }, + { + "epoch": 3.21, + "learning_rate": 3.3992790037142237e-06, + "loss": 1.8857, + "step": 14703 + }, + { + "epoch": 3.21, + "learning_rate": 3.3991697618527427e-06, + "loss": 3.2709, + "step": 14704 + }, + { + "epoch": 3.21, + "learning_rate": 3.399060519991261e-06, + "loss": 1.9091, + "step": 14705 + }, + { + "epoch": 3.21, + "learning_rate": 3.3989512781297795e-06, + "loss": 3.0801, + "step": 14706 + }, + { + "epoch": 3.21, + "learning_rate": 3.3988420362682985e-06, + "loss": 2.4951, + "step": 14707 + }, + { + "epoch": 3.21, + "learning_rate": 3.398732794406817e-06, + "loss": 3.0372, + "step": 14708 + }, + { + "epoch": 3.21, + "learning_rate": 3.3986235525453353e-06, + "loss": 3.0984, + "step": 14709 + }, + { + "epoch": 3.21, + "learning_rate": 3.3985143106838544e-06, + "loss": 2.6468, + "step": 14710 + }, + { + "epoch": 3.21, + "learning_rate": 3.398405068822373e-06, + "loss": 1.4145, + "step": 14711 + }, + { + "epoch": 3.21, + "learning_rate": 3.3982958269608916e-06, + "loss": 1.8696, + "step": 14712 + }, + { + "epoch": 3.21, + "learning_rate": 3.3981865850994106e-06, + "loss": 2.4275, + "step": 14713 + }, + { + "epoch": 3.21, + "learning_rate": 3.398077343237929e-06, + "loss": 2.4269, + "step": 14714 + }, + { + "epoch": 3.21, + "learning_rate": 3.3979681013764474e-06, + "loss": 1.6524, + "step": 14715 + }, + { + "epoch": 3.21, + "learning_rate": 3.3978588595149664e-06, + "loss": 0.8936, + "step": 14716 + }, + { + "epoch": 3.21, + "learning_rate": 3.397749617653485e-06, + "loss": 2.4434, + "step": 14717 + }, + { + "epoch": 3.21, + "learning_rate": 3.397640375792004e-06, + "loss": 2.5918, + "step": 14718 + }, + { + "epoch": 3.21, + "learning_rate": 3.3975311339305223e-06, + "loss": 1.6025, + "step": 14719 + }, + { + "epoch": 3.21, + "learning_rate": 3.397421892069041e-06, + "loss": 3.0603, + "step": 14720 + }, + { + "epoch": 3.21, + "learning_rate": 3.39731265020756e-06, + "loss": 3.3184, + "step": 14721 + }, + { + "epoch": 3.21, + "learning_rate": 3.3972034083460785e-06, + "loss": 2.8331, + "step": 14722 + }, + { + "epoch": 3.21, + "learning_rate": 3.3970941664845976e-06, + "loss": 1.4719, + "step": 14723 + }, + { + "epoch": 3.21, + "learning_rate": 3.3969849246231158e-06, + "loss": 2.0224, + "step": 14724 + }, + { + "epoch": 3.21, + "learning_rate": 3.3968756827616344e-06, + "loss": 2.7535, + "step": 14725 + }, + { + "epoch": 3.21, + "learning_rate": 3.3967664409001534e-06, + "loss": 1.6055, + "step": 14726 + }, + { + "epoch": 3.21, + "learning_rate": 3.396657199038672e-06, + "loss": 2.0465, + "step": 14727 + }, + { + "epoch": 3.21, + "learning_rate": 3.39654795717719e-06, + "loss": 2.9726, + "step": 14728 + }, + { + "epoch": 3.21, + "learning_rate": 3.3964387153157092e-06, + "loss": 1.9419, + "step": 14729 + }, + { + "epoch": 3.21, + "learning_rate": 3.396329473454228e-06, + "loss": 2.4509, + "step": 14730 + }, + { + "epoch": 3.21, + "learning_rate": 3.396220231592747e-06, + "loss": 1.9748, + "step": 14731 + }, + { + "epoch": 3.22, + "learning_rate": 3.3961109897312655e-06, + "loss": 2.2437, + "step": 14732 + }, + { + "epoch": 3.22, + "learning_rate": 3.3960017478697837e-06, + "loss": 2.0603, + "step": 14733 + }, + { + "epoch": 3.22, + "learning_rate": 3.3958925060083027e-06, + "loss": 2.3869, + "step": 14734 + }, + { + "epoch": 3.22, + "learning_rate": 3.3957832641468213e-06, + "loss": 2.277, + "step": 14735 + }, + { + "epoch": 3.22, + "learning_rate": 3.3956740222853404e-06, + "loss": 2.3514, + "step": 14736 + }, + { + "epoch": 3.22, + "learning_rate": 3.395564780423859e-06, + "loss": 1.5799, + "step": 14737 + }, + { + "epoch": 3.22, + "learning_rate": 3.395455538562377e-06, + "loss": 2.3636, + "step": 14738 + }, + { + "epoch": 3.22, + "learning_rate": 3.395346296700896e-06, + "loss": 1.7939, + "step": 14739 + }, + { + "epoch": 3.22, + "learning_rate": 3.395237054839415e-06, + "loss": 1.7295, + "step": 14740 + }, + { + "epoch": 3.22, + "learning_rate": 3.3951278129779334e-06, + "loss": 2.6282, + "step": 14741 + }, + { + "epoch": 3.22, + "learning_rate": 3.395018571116452e-06, + "loss": 1.9259, + "step": 14742 + }, + { + "epoch": 3.22, + "learning_rate": 3.3949093292549706e-06, + "loss": 2.5297, + "step": 14743 + }, + { + "epoch": 3.22, + "learning_rate": 3.3948000873934892e-06, + "loss": 1.7711, + "step": 14744 + }, + { + "epoch": 3.22, + "learning_rate": 3.3946908455320083e-06, + "loss": 0.9359, + "step": 14745 + }, + { + "epoch": 3.22, + "learning_rate": 3.394581603670527e-06, + "loss": 1.6907, + "step": 14746 + }, + { + "epoch": 3.22, + "learning_rate": 3.394472361809045e-06, + "loss": 1.6894, + "step": 14747 + }, + { + "epoch": 3.22, + "learning_rate": 3.394363119947564e-06, + "loss": 1.8103, + "step": 14748 + }, + { + "epoch": 3.22, + "learning_rate": 3.3942538780860827e-06, + "loss": 1.4099, + "step": 14749 + }, + { + "epoch": 3.22, + "learning_rate": 3.3941446362246018e-06, + "loss": 2.8459, + "step": 14750 + }, + { + "epoch": 3.22, + "learning_rate": 3.3940353943631204e-06, + "loss": 2.2907, + "step": 14751 + }, + { + "epoch": 3.22, + "learning_rate": 3.3939261525016386e-06, + "loss": 1.8851, + "step": 14752 + }, + { + "epoch": 3.22, + "learning_rate": 3.3938169106401576e-06, + "loss": 1.3747, + "step": 14753 + }, + { + "epoch": 3.22, + "learning_rate": 3.393707668778676e-06, + "loss": 2.1611, + "step": 14754 + }, + { + "epoch": 3.22, + "learning_rate": 3.3935984269171952e-06, + "loss": 1.6058, + "step": 14755 + }, + { + "epoch": 3.22, + "learning_rate": 3.393489185055714e-06, + "loss": 1.3573, + "step": 14756 + }, + { + "epoch": 3.22, + "learning_rate": 3.393379943194232e-06, + "loss": 2.9661, + "step": 14757 + }, + { + "epoch": 3.22, + "learning_rate": 3.393270701332751e-06, + "loss": 2.6523, + "step": 14758 + }, + { + "epoch": 3.22, + "learning_rate": 3.3931614594712697e-06, + "loss": 1.56, + "step": 14759 + }, + { + "epoch": 3.22, + "learning_rate": 3.3930522176097887e-06, + "loss": 1.8788, + "step": 14760 + }, + { + "epoch": 3.22, + "learning_rate": 3.392942975748307e-06, + "loss": 1.1671, + "step": 14761 + }, + { + "epoch": 3.22, + "learning_rate": 3.3928337338868255e-06, + "loss": 1.8153, + "step": 14762 + }, + { + "epoch": 3.22, + "learning_rate": 3.3927244920253445e-06, + "loss": 2.4471, + "step": 14763 + }, + { + "epoch": 3.22, + "learning_rate": 3.392615250163863e-06, + "loss": 2.1602, + "step": 14764 + }, + { + "epoch": 3.22, + "learning_rate": 3.3925060083023818e-06, + "loss": 2.5299, + "step": 14765 + }, + { + "epoch": 3.22, + "learning_rate": 3.3923967664409004e-06, + "loss": 1.2974, + "step": 14766 + }, + { + "epoch": 3.22, + "learning_rate": 3.392287524579419e-06, + "loss": 2.3797, + "step": 14767 + }, + { + "epoch": 3.22, + "learning_rate": 3.3921782827179376e-06, + "loss": 2.9182, + "step": 14768 + }, + { + "epoch": 3.22, + "learning_rate": 3.3920690408564566e-06, + "loss": 1.8444, + "step": 14769 + }, + { + "epoch": 3.22, + "learning_rate": 3.3919597989949752e-06, + "loss": 1.3676, + "step": 14770 + }, + { + "epoch": 3.22, + "learning_rate": 3.391850557133494e-06, + "loss": 2.5227, + "step": 14771 + }, + { + "epoch": 3.22, + "learning_rate": 3.3917413152720125e-06, + "loss": 1.2165, + "step": 14772 + }, + { + "epoch": 3.22, + "learning_rate": 3.391632073410531e-06, + "loss": 2.9378, + "step": 14773 + }, + { + "epoch": 3.22, + "learning_rate": 3.39152283154905e-06, + "loss": 1.5618, + "step": 14774 + }, + { + "epoch": 3.22, + "learning_rate": 3.3914135896875683e-06, + "loss": 1.402, + "step": 14775 + }, + { + "epoch": 3.22, + "learning_rate": 3.391304347826087e-06, + "loss": 2.0061, + "step": 14776 + }, + { + "epoch": 3.23, + "learning_rate": 3.391195105964606e-06, + "loss": 2.6747, + "step": 14777 + }, + { + "epoch": 3.23, + "learning_rate": 3.3910858641031246e-06, + "loss": 1.3436, + "step": 14778 + }, + { + "epoch": 3.23, + "learning_rate": 3.3909766222416436e-06, + "loss": 1.8703, + "step": 14779 + }, + { + "epoch": 3.23, + "learning_rate": 3.3908673803801618e-06, + "loss": 2.3871, + "step": 14780 + }, + { + "epoch": 3.23, + "learning_rate": 3.3907581385186804e-06, + "loss": 1.3694, + "step": 14781 + }, + { + "epoch": 3.23, + "learning_rate": 3.3906488966571994e-06, + "loss": 1.5317, + "step": 14782 + }, + { + "epoch": 3.23, + "learning_rate": 3.390539654795718e-06, + "loss": 1.4696, + "step": 14783 + }, + { + "epoch": 3.23, + "learning_rate": 3.390430412934237e-06, + "loss": 1.3991, + "step": 14784 + }, + { + "epoch": 3.23, + "learning_rate": 3.3903211710727553e-06, + "loss": 2.2761, + "step": 14785 + }, + { + "epoch": 3.23, + "learning_rate": 3.390211929211274e-06, + "loss": 2.3767, + "step": 14786 + }, + { + "epoch": 3.23, + "learning_rate": 3.390102687349793e-06, + "loss": 2.9662, + "step": 14787 + }, + { + "epoch": 3.23, + "learning_rate": 3.3899934454883115e-06, + "loss": 2.5369, + "step": 14788 + }, + { + "epoch": 3.23, + "learning_rate": 3.3898842036268306e-06, + "loss": 2.7504, + "step": 14789 + }, + { + "epoch": 3.23, + "learning_rate": 3.3897749617653487e-06, + "loss": 2.4903, + "step": 14790 + }, + { + "epoch": 3.23, + "learning_rate": 3.3896657199038673e-06, + "loss": 2.6266, + "step": 14791 + }, + { + "epoch": 3.23, + "learning_rate": 3.3895564780423864e-06, + "loss": 1.352, + "step": 14792 + }, + { + "epoch": 3.23, + "learning_rate": 3.389447236180905e-06, + "loss": 1.7834, + "step": 14793 + }, + { + "epoch": 3.23, + "learning_rate": 3.389337994319423e-06, + "loss": 2.1737, + "step": 14794 + }, + { + "epoch": 3.23, + "learning_rate": 3.3892287524579422e-06, + "loss": 2.2612, + "step": 14795 + }, + { + "epoch": 3.23, + "learning_rate": 3.389119510596461e-06, + "loss": 3.0747, + "step": 14796 + }, + { + "epoch": 3.23, + "learning_rate": 3.3890102687349794e-06, + "loss": 2.8174, + "step": 14797 + }, + { + "epoch": 3.23, + "learning_rate": 3.3889010268734985e-06, + "loss": 1.1462, + "step": 14798 + }, + { + "epoch": 3.23, + "learning_rate": 3.3887917850120167e-06, + "loss": 2.0526, + "step": 14799 + }, + { + "epoch": 3.23, + "learning_rate": 3.3886825431505353e-06, + "loss": 1.8285, + "step": 14800 + }, + { + "epoch": 3.23, + "learning_rate": 3.3885733012890543e-06, + "loss": 0.8617, + "step": 14801 + }, + { + "epoch": 3.23, + "learning_rate": 3.388464059427573e-06, + "loss": 3.3558, + "step": 14802 + }, + { + "epoch": 3.23, + "learning_rate": 3.388354817566092e-06, + "loss": 2.4057, + "step": 14803 + }, + { + "epoch": 3.23, + "learning_rate": 3.38824557570461e-06, + "loss": 1.4104, + "step": 14804 + }, + { + "epoch": 3.23, + "learning_rate": 3.3881363338431287e-06, + "loss": 2.3485, + "step": 14805 + }, + { + "epoch": 3.23, + "learning_rate": 3.3880270919816478e-06, + "loss": 1.8739, + "step": 14806 + }, + { + "epoch": 3.23, + "learning_rate": 3.3879178501201664e-06, + "loss": 2.379, + "step": 14807 + }, + { + "epoch": 3.23, + "learning_rate": 3.3878086082586846e-06, + "loss": 3.0778, + "step": 14808 + }, + { + "epoch": 3.23, + "learning_rate": 3.3876993663972036e-06, + "loss": 1.5553, + "step": 14809 + }, + { + "epoch": 3.23, + "learning_rate": 3.3875901245357222e-06, + "loss": 1.4283, + "step": 14810 + }, + { + "epoch": 3.23, + "learning_rate": 3.3874808826742413e-06, + "loss": 1.9112, + "step": 14811 + }, + { + "epoch": 3.23, + "learning_rate": 3.38737164081276e-06, + "loss": 2.0434, + "step": 14812 + }, + { + "epoch": 3.23, + "learning_rate": 3.387262398951278e-06, + "loss": 2.1797, + "step": 14813 + }, + { + "epoch": 3.23, + "learning_rate": 3.387153157089797e-06, + "loss": 3.1602, + "step": 14814 + }, + { + "epoch": 3.23, + "learning_rate": 3.3870439152283157e-06, + "loss": 2.8058, + "step": 14815 + }, + { + "epoch": 3.23, + "learning_rate": 3.3869346733668347e-06, + "loss": 1.6084, + "step": 14816 + }, + { + "epoch": 3.23, + "learning_rate": 3.3868254315053534e-06, + "loss": 1.3898, + "step": 14817 + }, + { + "epoch": 3.23, + "learning_rate": 3.3867161896438715e-06, + "loss": 1.5542, + "step": 14818 + }, + { + "epoch": 3.23, + "learning_rate": 3.3866069477823906e-06, + "loss": 2.0749, + "step": 14819 + }, + { + "epoch": 3.23, + "learning_rate": 3.386497705920909e-06, + "loss": 2.8268, + "step": 14820 + }, + { + "epoch": 3.23, + "learning_rate": 3.386388464059428e-06, + "loss": 2.0891, + "step": 14821 + }, + { + "epoch": 3.23, + "learning_rate": 3.386279222197947e-06, + "loss": 0.1343, + "step": 14822 + }, + { + "epoch": 3.24, + "learning_rate": 3.386169980336465e-06, + "loss": 1.1586, + "step": 14823 + }, + { + "epoch": 3.24, + "learning_rate": 3.386060738474984e-06, + "loss": 3.0207, + "step": 14824 + }, + { + "epoch": 3.24, + "learning_rate": 3.3859514966135027e-06, + "loss": 2.0719, + "step": 14825 + }, + { + "epoch": 3.24, + "learning_rate": 3.3858422547520213e-06, + "loss": 1.0617, + "step": 14826 + }, + { + "epoch": 3.24, + "learning_rate": 3.38573301289054e-06, + "loss": 2.0028, + "step": 14827 + }, + { + "epoch": 3.24, + "learning_rate": 3.3856237710290585e-06, + "loss": 1.9815, + "step": 14828 + }, + { + "epoch": 3.24, + "learning_rate": 3.385514529167577e-06, + "loss": 2.8585, + "step": 14829 + }, + { + "epoch": 3.24, + "learning_rate": 3.385405287306096e-06, + "loss": 2.7812, + "step": 14830 + }, + { + "epoch": 3.24, + "learning_rate": 3.3852960454446148e-06, + "loss": 4.0122, + "step": 14831 + }, + { + "epoch": 3.24, + "learning_rate": 3.385186803583133e-06, + "loss": 2.9371, + "step": 14832 + }, + { + "epoch": 3.24, + "learning_rate": 3.385077561721652e-06, + "loss": 1.9831, + "step": 14833 + }, + { + "epoch": 3.24, + "learning_rate": 3.3849683198601706e-06, + "loss": 1.7099, + "step": 14834 + }, + { + "epoch": 3.24, + "learning_rate": 3.3848590779986896e-06, + "loss": 1.7752, + "step": 14835 + }, + { + "epoch": 3.24, + "learning_rate": 3.3847498361372082e-06, + "loss": 1.4539, + "step": 14836 + }, + { + "epoch": 3.24, + "learning_rate": 3.3846405942757264e-06, + "loss": 1.2561, + "step": 14837 + }, + { + "epoch": 3.24, + "learning_rate": 3.3845313524142455e-06, + "loss": 2.3916, + "step": 14838 + }, + { + "epoch": 3.24, + "learning_rate": 3.384422110552764e-06, + "loss": 2.8966, + "step": 14839 + }, + { + "epoch": 3.24, + "learning_rate": 3.384312868691283e-06, + "loss": 2.5941, + "step": 14840 + }, + { + "epoch": 3.24, + "learning_rate": 3.3842036268298013e-06, + "loss": 1.9385, + "step": 14841 + }, + { + "epoch": 3.24, + "learning_rate": 3.38409438496832e-06, + "loss": 2.2976, + "step": 14842 + }, + { + "epoch": 3.24, + "learning_rate": 3.383985143106839e-06, + "loss": 2.621, + "step": 14843 + }, + { + "epoch": 3.24, + "learning_rate": 3.3838759012453575e-06, + "loss": 2.4133, + "step": 14844 + }, + { + "epoch": 3.24, + "learning_rate": 3.3837666593838766e-06, + "loss": 1.972, + "step": 14845 + }, + { + "epoch": 3.24, + "learning_rate": 3.3836574175223948e-06, + "loss": 2.5575, + "step": 14846 + }, + { + "epoch": 3.24, + "learning_rate": 3.3835481756609134e-06, + "loss": 1.6077, + "step": 14847 + }, + { + "epoch": 3.24, + "learning_rate": 3.3834389337994324e-06, + "loss": 2.0074, + "step": 14848 + }, + { + "epoch": 3.24, + "learning_rate": 3.383329691937951e-06, + "loss": 2.3376, + "step": 14849 + }, + { + "epoch": 3.24, + "learning_rate": 3.3832204500764696e-06, + "loss": 1.3677, + "step": 14850 + }, + { + "epoch": 3.24, + "learning_rate": 3.3831112082149882e-06, + "loss": 1.3083, + "step": 14851 + }, + { + "epoch": 3.24, + "learning_rate": 3.383001966353507e-06, + "loss": 1.5845, + "step": 14852 + }, + { + "epoch": 3.24, + "learning_rate": 3.3828927244920255e-06, + "loss": 3.2863, + "step": 14853 + }, + { + "epoch": 3.24, + "learning_rate": 3.3827834826305445e-06, + "loss": 2.8464, + "step": 14854 + }, + { + "epoch": 3.24, + "learning_rate": 3.382674240769063e-06, + "loss": 2.7181, + "step": 14855 + }, + { + "epoch": 3.24, + "learning_rate": 3.3825649989075813e-06, + "loss": 2.1172, + "step": 14856 + }, + { + "epoch": 3.24, + "learning_rate": 3.3824557570461003e-06, + "loss": 1.9613, + "step": 14857 + }, + { + "epoch": 3.24, + "learning_rate": 3.382346515184619e-06, + "loss": 2.607, + "step": 14858 + }, + { + "epoch": 3.24, + "learning_rate": 3.382237273323138e-06, + "loss": 2.3764, + "step": 14859 + }, + { + "epoch": 3.24, + "learning_rate": 3.382128031461656e-06, + "loss": 1.8489, + "step": 14860 + }, + { + "epoch": 3.24, + "learning_rate": 3.3820187896001748e-06, + "loss": 0.8759, + "step": 14861 + }, + { + "epoch": 3.24, + "learning_rate": 3.381909547738694e-06, + "loss": 1.2105, + "step": 14862 + }, + { + "epoch": 3.24, + "learning_rate": 3.3818003058772124e-06, + "loss": 1.4158, + "step": 14863 + }, + { + "epoch": 3.24, + "learning_rate": 3.3816910640157315e-06, + "loss": 1.715, + "step": 14864 + }, + { + "epoch": 3.24, + "learning_rate": 3.3815818221542496e-06, + "loss": 3.8524, + "step": 14865 + }, + { + "epoch": 3.24, + "learning_rate": 3.3814725802927683e-06, + "loss": 2.279, + "step": 14866 + }, + { + "epoch": 3.24, + "learning_rate": 3.3813633384312873e-06, + "loss": 1.8071, + "step": 14867 + }, + { + "epoch": 3.24, + "learning_rate": 3.381254096569806e-06, + "loss": 3.0331, + "step": 14868 + }, + { + "epoch": 3.25, + "learning_rate": 3.381144854708325e-06, + "loss": 1.9636, + "step": 14869 + }, + { + "epoch": 3.25, + "learning_rate": 3.381035612846843e-06, + "loss": 2.4693, + "step": 14870 + }, + { + "epoch": 3.25, + "learning_rate": 3.3809263709853617e-06, + "loss": 0.7373, + "step": 14871 + }, + { + "epoch": 3.25, + "learning_rate": 3.3808171291238808e-06, + "loss": 1.9775, + "step": 14872 + }, + { + "epoch": 3.25, + "learning_rate": 3.3807078872623994e-06, + "loss": 2.0933, + "step": 14873 + }, + { + "epoch": 3.25, + "learning_rate": 3.3805986454009176e-06, + "loss": 1.6327, + "step": 14874 + }, + { + "epoch": 3.25, + "learning_rate": 3.3804894035394366e-06, + "loss": 2.1573, + "step": 14875 + }, + { + "epoch": 3.25, + "learning_rate": 3.3803801616779552e-06, + "loss": 1.4503, + "step": 14876 + }, + { + "epoch": 3.25, + "learning_rate": 3.3802709198164742e-06, + "loss": 0.8054, + "step": 14877 + }, + { + "epoch": 3.25, + "learning_rate": 3.380161677954993e-06, + "loss": 2.4156, + "step": 14878 + }, + { + "epoch": 3.25, + "learning_rate": 3.380052436093511e-06, + "loss": 1.5676, + "step": 14879 + }, + { + "epoch": 3.25, + "learning_rate": 3.37994319423203e-06, + "loss": 1.3532, + "step": 14880 + }, + { + "epoch": 3.25, + "learning_rate": 3.3798339523705487e-06, + "loss": 2.3129, + "step": 14881 + }, + { + "epoch": 3.25, + "learning_rate": 3.3797247105090673e-06, + "loss": 2.1776, + "step": 14882 + }, + { + "epoch": 3.25, + "learning_rate": 3.3796154686475863e-06, + "loss": 0.8487, + "step": 14883 + }, + { + "epoch": 3.25, + "learning_rate": 3.3795062267861045e-06, + "loss": 2.3456, + "step": 14884 + }, + { + "epoch": 3.25, + "learning_rate": 3.379396984924623e-06, + "loss": 2.6527, + "step": 14885 + }, + { + "epoch": 3.25, + "learning_rate": 3.379287743063142e-06, + "loss": 2.3909, + "step": 14886 + }, + { + "epoch": 3.25, + "learning_rate": 3.3791785012016608e-06, + "loss": 2.9909, + "step": 14887 + }, + { + "epoch": 3.25, + "learning_rate": 3.37906925934018e-06, + "loss": 2.3841, + "step": 14888 + }, + { + "epoch": 3.25, + "learning_rate": 3.378960017478698e-06, + "loss": 1.6109, + "step": 14889 + }, + { + "epoch": 3.25, + "learning_rate": 3.3788507756172166e-06, + "loss": 2.3397, + "step": 14890 + }, + { + "epoch": 3.25, + "learning_rate": 3.3787415337557356e-06, + "loss": 2.1266, + "step": 14891 + }, + { + "epoch": 3.25, + "learning_rate": 3.3786322918942543e-06, + "loss": 1.66, + "step": 14892 + }, + { + "epoch": 3.25, + "learning_rate": 3.3785230500327724e-06, + "loss": 2.8716, + "step": 14893 + }, + { + "epoch": 3.25, + "learning_rate": 3.3784138081712915e-06, + "loss": 2.5007, + "step": 14894 + }, + { + "epoch": 3.25, + "learning_rate": 3.37830456630981e-06, + "loss": 1.7798, + "step": 14895 + }, + { + "epoch": 3.25, + "learning_rate": 3.378195324448329e-06, + "loss": 1.716, + "step": 14896 + }, + { + "epoch": 3.25, + "learning_rate": 3.3780860825868477e-06, + "loss": 2.7255, + "step": 14897 + }, + { + "epoch": 3.25, + "learning_rate": 3.377976840725366e-06, + "loss": 2.7784, + "step": 14898 + }, + { + "epoch": 3.25, + "learning_rate": 3.377867598863885e-06, + "loss": 2.1303, + "step": 14899 + }, + { + "epoch": 3.25, + "learning_rate": 3.3777583570024036e-06, + "loss": 3.3917, + "step": 14900 + }, + { + "epoch": 3.25, + "learning_rate": 3.3776491151409226e-06, + "loss": 2.0763, + "step": 14901 + }, + { + "epoch": 3.25, + "learning_rate": 3.3775398732794412e-06, + "loss": 2.6275, + "step": 14902 + }, + { + "epoch": 3.25, + "learning_rate": 3.3774306314179594e-06, + "loss": 4.0642, + "step": 14903 + }, + { + "epoch": 3.25, + "learning_rate": 3.3773213895564784e-06, + "loss": 2.1991, + "step": 14904 + }, + { + "epoch": 3.25, + "learning_rate": 3.377212147694997e-06, + "loss": 2.3006, + "step": 14905 + }, + { + "epoch": 3.25, + "learning_rate": 3.3771029058335157e-06, + "loss": 1.6484, + "step": 14906 + }, + { + "epoch": 3.25, + "learning_rate": 3.3769936639720343e-06, + "loss": 2.4544, + "step": 14907 + }, + { + "epoch": 3.25, + "learning_rate": 3.376884422110553e-06, + "loss": 3.3046, + "step": 14908 + }, + { + "epoch": 3.25, + "learning_rate": 3.3767751802490715e-06, + "loss": 2.6806, + "step": 14909 + }, + { + "epoch": 3.25, + "learning_rate": 3.3766659383875905e-06, + "loss": 1.2551, + "step": 14910 + }, + { + "epoch": 3.25, + "learning_rate": 3.376556696526109e-06, + "loss": 1.5903, + "step": 14911 + }, + { + "epoch": 3.25, + "learning_rate": 3.3764474546646273e-06, + "loss": 1.6788, + "step": 14912 + }, + { + "epoch": 3.25, + "learning_rate": 3.3763382128031464e-06, + "loss": 3.0533, + "step": 14913 + }, + { + "epoch": 3.25, + "learning_rate": 3.376228970941665e-06, + "loss": 3.2441, + "step": 14914 + }, + { + "epoch": 3.26, + "learning_rate": 3.376119729080184e-06, + "loss": 2.808, + "step": 14915 + }, + { + "epoch": 3.26, + "learning_rate": 3.3760104872187026e-06, + "loss": 2.0237, + "step": 14916 + }, + { + "epoch": 3.26, + "learning_rate": 3.375901245357221e-06, + "loss": 1.5266, + "step": 14917 + }, + { + "epoch": 3.26, + "learning_rate": 3.37579200349574e-06, + "loss": 3.1164, + "step": 14918 + }, + { + "epoch": 3.26, + "learning_rate": 3.3756827616342584e-06, + "loss": 1.9647, + "step": 14919 + }, + { + "epoch": 3.26, + "learning_rate": 3.3755735197727775e-06, + "loss": 1.7985, + "step": 14920 + }, + { + "epoch": 3.26, + "learning_rate": 3.375464277911296e-06, + "loss": 2.2023, + "step": 14921 + }, + { + "epoch": 3.26, + "learning_rate": 3.3753550360498143e-06, + "loss": 2.6156, + "step": 14922 + }, + { + "epoch": 3.26, + "learning_rate": 3.3752457941883333e-06, + "loss": 2.1878, + "step": 14923 + }, + { + "epoch": 3.26, + "learning_rate": 3.375136552326852e-06, + "loss": 1.4086, + "step": 14924 + }, + { + "epoch": 3.26, + "learning_rate": 3.375027310465371e-06, + "loss": 2.6398, + "step": 14925 + }, + { + "epoch": 3.26, + "learning_rate": 3.374918068603889e-06, + "loss": 1.9285, + "step": 14926 + }, + { + "epoch": 3.26, + "learning_rate": 3.3748088267424078e-06, + "loss": 2.565, + "step": 14927 + }, + { + "epoch": 3.26, + "learning_rate": 3.374699584880927e-06, + "loss": 0.9979, + "step": 14928 + }, + { + "epoch": 3.26, + "learning_rate": 3.3745903430194454e-06, + "loss": 1.4209, + "step": 14929 + }, + { + "epoch": 3.26, + "learning_rate": 3.3744811011579644e-06, + "loss": 1.4234, + "step": 14930 + }, + { + "epoch": 3.26, + "learning_rate": 3.3743718592964826e-06, + "loss": 1.8013, + "step": 14931 + }, + { + "epoch": 3.26, + "learning_rate": 3.3742626174350012e-06, + "loss": 2.1599, + "step": 14932 + }, + { + "epoch": 3.26, + "learning_rate": 3.3741533755735203e-06, + "loss": 1.5969, + "step": 14933 + }, + { + "epoch": 3.26, + "learning_rate": 3.374044133712039e-06, + "loss": 2.2415, + "step": 14934 + }, + { + "epoch": 3.26, + "learning_rate": 3.3739348918505575e-06, + "loss": 0.8295, + "step": 14935 + }, + { + "epoch": 3.26, + "learning_rate": 3.373825649989076e-06, + "loss": 2.2153, + "step": 14936 + }, + { + "epoch": 3.26, + "learning_rate": 3.3737164081275947e-06, + "loss": 3.3033, + "step": 14937 + }, + { + "epoch": 3.26, + "learning_rate": 3.3736071662661133e-06, + "loss": 1.5781, + "step": 14938 + }, + { + "epoch": 3.26, + "learning_rate": 3.3734979244046324e-06, + "loss": 1.8033, + "step": 14939 + }, + { + "epoch": 3.26, + "learning_rate": 3.3733886825431506e-06, + "loss": 1.3878, + "step": 14940 + }, + { + "epoch": 3.26, + "learning_rate": 3.373279440681669e-06, + "loss": 1.3812, + "step": 14941 + }, + { + "epoch": 3.26, + "learning_rate": 3.373170198820188e-06, + "loss": 1.2028, + "step": 14942 + }, + { + "epoch": 3.26, + "learning_rate": 3.373060956958707e-06, + "loss": 4.0371, + "step": 14943 + }, + { + "epoch": 3.26, + "learning_rate": 3.372951715097226e-06, + "loss": 0.7811, + "step": 14944 + }, + { + "epoch": 3.26, + "learning_rate": 3.372842473235744e-06, + "loss": 1.6599, + "step": 14945 + }, + { + "epoch": 3.26, + "learning_rate": 3.3727332313742626e-06, + "loss": 1.1924, + "step": 14946 + }, + { + "epoch": 3.26, + "learning_rate": 3.3726239895127817e-06, + "loss": 2.5798, + "step": 14947 + }, + { + "epoch": 3.26, + "learning_rate": 3.3725147476513003e-06, + "loss": 2.7066, + "step": 14948 + }, + { + "epoch": 3.26, + "learning_rate": 3.3724055057898193e-06, + "loss": 3.5199, + "step": 14949 + }, + { + "epoch": 3.26, + "learning_rate": 3.3722962639283375e-06, + "loss": 2.8096, + "step": 14950 + }, + { + "epoch": 3.26, + "learning_rate": 3.372187022066856e-06, + "loss": 2.0348, + "step": 14951 + }, + { + "epoch": 3.26, + "learning_rate": 3.372077780205375e-06, + "loss": 2.0808, + "step": 14952 + }, + { + "epoch": 3.26, + "learning_rate": 3.3719685383438938e-06, + "loss": 3.1683, + "step": 14953 + }, + { + "epoch": 3.26, + "learning_rate": 3.371859296482412e-06, + "loss": 3.9055, + "step": 14954 + }, + { + "epoch": 3.26, + "learning_rate": 3.371750054620931e-06, + "loss": 1.862, + "step": 14955 + }, + { + "epoch": 3.26, + "learning_rate": 3.3716408127594496e-06, + "loss": 2.2351, + "step": 14956 + }, + { + "epoch": 3.26, + "learning_rate": 3.3715315708979686e-06, + "loss": 0.4926, + "step": 14957 + }, + { + "epoch": 3.26, + "learning_rate": 3.3714223290364872e-06, + "loss": 2.7532, + "step": 14958 + }, + { + "epoch": 3.26, + "learning_rate": 3.3713130871750054e-06, + "loss": 1.2809, + "step": 14959 + }, + { + "epoch": 3.26, + "learning_rate": 3.3712038453135245e-06, + "loss": 1.7067, + "step": 14960 + }, + { + "epoch": 3.27, + "learning_rate": 3.371094603452043e-06, + "loss": 1.3163, + "step": 14961 + }, + { + "epoch": 3.27, + "learning_rate": 3.3709853615905617e-06, + "loss": 3.0093, + "step": 14962 + }, + { + "epoch": 3.27, + "learning_rate": 3.3708761197290807e-06, + "loss": 1.0732, + "step": 14963 + }, + { + "epoch": 3.27, + "learning_rate": 3.370766877867599e-06, + "loss": 1.6208, + "step": 14964 + }, + { + "epoch": 3.27, + "learning_rate": 3.3706576360061175e-06, + "loss": 1.4242, + "step": 14965 + }, + { + "epoch": 3.27, + "learning_rate": 3.3705483941446366e-06, + "loss": 2.4508, + "step": 14966 + }, + { + "epoch": 3.27, + "learning_rate": 3.370439152283155e-06, + "loss": 1.858, + "step": 14967 + }, + { + "epoch": 3.27, + "learning_rate": 3.370329910421674e-06, + "loss": 4.0321, + "step": 14968 + }, + { + "epoch": 3.27, + "learning_rate": 3.3702206685601924e-06, + "loss": 1.9439, + "step": 14969 + }, + { + "epoch": 3.27, + "learning_rate": 3.370111426698711e-06, + "loss": 1.6444, + "step": 14970 + }, + { + "epoch": 3.27, + "learning_rate": 3.37000218483723e-06, + "loss": 2.4022, + "step": 14971 + }, + { + "epoch": 3.27, + "learning_rate": 3.3698929429757486e-06, + "loss": 2.2544, + "step": 14972 + }, + { + "epoch": 3.27, + "learning_rate": 3.369783701114267e-06, + "loss": 1.7249, + "step": 14973 + }, + { + "epoch": 3.27, + "learning_rate": 3.369674459252786e-06, + "loss": 2.1373, + "step": 14974 + }, + { + "epoch": 3.27, + "learning_rate": 3.3695652173913045e-06, + "loss": 2.5896, + "step": 14975 + }, + { + "epoch": 3.27, + "learning_rate": 3.3694559755298235e-06, + "loss": 2.5502, + "step": 14976 + }, + { + "epoch": 3.27, + "learning_rate": 3.369346733668342e-06, + "loss": 2.2076, + "step": 14977 + }, + { + "epoch": 3.27, + "learning_rate": 3.3692374918068603e-06, + "loss": 2.8308, + "step": 14978 + }, + { + "epoch": 3.27, + "learning_rate": 3.3691282499453793e-06, + "loss": 2.7041, + "step": 14979 + }, + { + "epoch": 3.27, + "learning_rate": 3.369019008083898e-06, + "loss": 0.7315, + "step": 14980 + }, + { + "epoch": 3.27, + "learning_rate": 3.368909766222417e-06, + "loss": 2.5024, + "step": 14981 + }, + { + "epoch": 3.27, + "learning_rate": 3.3688005243609356e-06, + "loss": 2.1582, + "step": 14982 + }, + { + "epoch": 3.27, + "learning_rate": 3.3686912824994538e-06, + "loss": 2.9303, + "step": 14983 + }, + { + "epoch": 3.27, + "learning_rate": 3.368582040637973e-06, + "loss": 1.8127, + "step": 14984 + }, + { + "epoch": 3.27, + "learning_rate": 3.3684727987764914e-06, + "loss": 1.9633, + "step": 14985 + }, + { + "epoch": 3.27, + "learning_rate": 3.3683635569150105e-06, + "loss": 2.1981, + "step": 14986 + }, + { + "epoch": 3.27, + "learning_rate": 3.3682543150535287e-06, + "loss": 2.1885, + "step": 14987 + }, + { + "epoch": 3.27, + "learning_rate": 3.3681450731920473e-06, + "loss": 1.8153, + "step": 14988 + }, + { + "epoch": 3.27, + "learning_rate": 3.3680358313305663e-06, + "loss": 1.7004, + "step": 14989 + }, + { + "epoch": 3.27, + "learning_rate": 3.367926589469085e-06, + "loss": 2.0979, + "step": 14990 + }, + { + "epoch": 3.27, + "learning_rate": 3.3678173476076035e-06, + "loss": 1.556, + "step": 14991 + }, + { + "epoch": 3.27, + "learning_rate": 3.367708105746122e-06, + "loss": 1.2644, + "step": 14992 + }, + { + "epoch": 3.27, + "learning_rate": 3.3675988638846407e-06, + "loss": 1.4961, + "step": 14993 + }, + { + "epoch": 3.27, + "learning_rate": 3.3674896220231594e-06, + "loss": 2.1829, + "step": 14994 + }, + { + "epoch": 3.27, + "learning_rate": 3.3673803801616784e-06, + "loss": 2.4772, + "step": 14995 + }, + { + "epoch": 3.27, + "learning_rate": 3.367271138300197e-06, + "loss": 1.9657, + "step": 14996 + }, + { + "epoch": 3.27, + "learning_rate": 3.367161896438715e-06, + "loss": 1.0694, + "step": 14997 + }, + { + "epoch": 3.27, + "learning_rate": 3.3670526545772342e-06, + "loss": 2.0738, + "step": 14998 + }, + { + "epoch": 3.27, + "learning_rate": 3.366943412715753e-06, + "loss": 2.7746, + "step": 14999 + }, + { + "epoch": 3.27, + "learning_rate": 3.366834170854272e-06, + "loss": 2.5447, + "step": 15000 + }, + { + "epoch": 3.27, + "learning_rate": 3.3667249289927905e-06, + "loss": 0.8341, + "step": 15001 + }, + { + "epoch": 3.27, + "learning_rate": 3.3666156871313087e-06, + "loss": 2.7376, + "step": 15002 + }, + { + "epoch": 3.27, + "learning_rate": 3.3665064452698277e-06, + "loss": 0.4534, + "step": 15003 + }, + { + "epoch": 3.27, + "learning_rate": 3.3663972034083463e-06, + "loss": 1.6252, + "step": 15004 + }, + { + "epoch": 3.27, + "learning_rate": 3.3662879615468653e-06, + "loss": 2.184, + "step": 15005 + }, + { + "epoch": 3.27, + "learning_rate": 3.3661787196853835e-06, + "loss": 1.7071, + "step": 15006 + }, + { + "epoch": 3.28, + "learning_rate": 3.366069477823902e-06, + "loss": 1.0488, + "step": 15007 + }, + { + "epoch": 3.28, + "learning_rate": 3.365960235962421e-06, + "loss": 2.7966, + "step": 15008 + }, + { + "epoch": 3.28, + "learning_rate": 3.36585099410094e-06, + "loss": 2.0814, + "step": 15009 + }, + { + "epoch": 3.28, + "learning_rate": 3.365741752239459e-06, + "loss": 1.693, + "step": 15010 + }, + { + "epoch": 3.28, + "learning_rate": 3.365632510377977e-06, + "loss": 1.9515, + "step": 15011 + }, + { + "epoch": 3.28, + "learning_rate": 3.3655232685164956e-06, + "loss": 1.4904, + "step": 15012 + }, + { + "epoch": 3.28, + "learning_rate": 3.3654140266550147e-06, + "loss": 2.3223, + "step": 15013 + }, + { + "epoch": 3.28, + "learning_rate": 3.3653047847935333e-06, + "loss": 2.2461, + "step": 15014 + }, + { + "epoch": 3.28, + "learning_rate": 3.365195542932052e-06, + "loss": 0.7717, + "step": 15015 + }, + { + "epoch": 3.28, + "learning_rate": 3.3650863010705705e-06, + "loss": 1.2474, + "step": 15016 + }, + { + "epoch": 3.28, + "learning_rate": 3.364977059209089e-06, + "loss": 3.1269, + "step": 15017 + }, + { + "epoch": 3.28, + "learning_rate": 3.3648678173476077e-06, + "loss": 0.6717, + "step": 15018 + }, + { + "epoch": 3.28, + "learning_rate": 3.3647585754861267e-06, + "loss": 2.7257, + "step": 15019 + }, + { + "epoch": 3.28, + "learning_rate": 3.364649333624645e-06, + "loss": 2.1517, + "step": 15020 + }, + { + "epoch": 3.28, + "learning_rate": 3.364540091763164e-06, + "loss": 3.237, + "step": 15021 + }, + { + "epoch": 3.28, + "learning_rate": 3.3644308499016826e-06, + "loss": 2.7835, + "step": 15022 + }, + { + "epoch": 3.28, + "learning_rate": 3.364321608040201e-06, + "loss": 1.1287, + "step": 15023 + }, + { + "epoch": 3.28, + "learning_rate": 3.3642123661787202e-06, + "loss": 2.5012, + "step": 15024 + }, + { + "epoch": 3.28, + "learning_rate": 3.3641031243172384e-06, + "loss": 1.6458, + "step": 15025 + }, + { + "epoch": 3.28, + "learning_rate": 3.363993882455757e-06, + "loss": 2.5064, + "step": 15026 + }, + { + "epoch": 3.28, + "learning_rate": 3.363884640594276e-06, + "loss": 2.3563, + "step": 15027 + }, + { + "epoch": 3.28, + "learning_rate": 3.3637753987327947e-06, + "loss": 2.4929, + "step": 15028 + }, + { + "epoch": 3.28, + "learning_rate": 3.3636661568713137e-06, + "loss": 1.2919, + "step": 15029 + }, + { + "epoch": 3.28, + "learning_rate": 3.363556915009832e-06, + "loss": 2.9367, + "step": 15030 + }, + { + "epoch": 3.28, + "learning_rate": 3.3634476731483505e-06, + "loss": 2.969, + "step": 15031 + }, + { + "epoch": 3.28, + "learning_rate": 3.3633384312868695e-06, + "loss": 2.8617, + "step": 15032 + }, + { + "epoch": 3.28, + "learning_rate": 3.363229189425388e-06, + "loss": 2.7697, + "step": 15033 + }, + { + "epoch": 3.28, + "learning_rate": 3.363119947563907e-06, + "loss": 2.2991, + "step": 15034 + }, + { + "epoch": 3.28, + "learning_rate": 3.3630107057024254e-06, + "loss": 3.3048, + "step": 15035 + }, + { + "epoch": 3.28, + "learning_rate": 3.362901463840944e-06, + "loss": 2.0083, + "step": 15036 + }, + { + "epoch": 3.28, + "learning_rate": 3.362792221979463e-06, + "loss": 2.733, + "step": 15037 + }, + { + "epoch": 3.28, + "learning_rate": 3.3626829801179816e-06, + "loss": 1.8123, + "step": 15038 + }, + { + "epoch": 3.28, + "learning_rate": 3.3625737382565e-06, + "loss": 0.851, + "step": 15039 + }, + { + "epoch": 3.28, + "learning_rate": 3.362464496395019e-06, + "loss": 2.1685, + "step": 15040 + }, + { + "epoch": 3.28, + "learning_rate": 3.3623552545335375e-06, + "loss": 1.4714, + "step": 15041 + }, + { + "epoch": 3.28, + "learning_rate": 3.3622460126720565e-06, + "loss": 1.7659, + "step": 15042 + }, + { + "epoch": 3.28, + "learning_rate": 3.362136770810575e-06, + "loss": 2.2789, + "step": 15043 + }, + { + "epoch": 3.28, + "learning_rate": 3.3620275289490933e-06, + "loss": 2.0131, + "step": 15044 + }, + { + "epoch": 3.28, + "learning_rate": 3.3619182870876123e-06, + "loss": 1.6205, + "step": 15045 + }, + { + "epoch": 3.28, + "learning_rate": 3.361809045226131e-06, + "loss": 2.6192, + "step": 15046 + }, + { + "epoch": 3.28, + "learning_rate": 3.3616998033646495e-06, + "loss": 2.9503, + "step": 15047 + }, + { + "epoch": 3.28, + "learning_rate": 3.3615905615031686e-06, + "loss": 0.4457, + "step": 15048 + }, + { + "epoch": 3.28, + "learning_rate": 3.3614813196416868e-06, + "loss": 1.7117, + "step": 15049 + }, + { + "epoch": 3.28, + "learning_rate": 3.3613720777802054e-06, + "loss": 2.543, + "step": 15050 + }, + { + "epoch": 3.28, + "learning_rate": 3.3612628359187244e-06, + "loss": 0.9574, + "step": 15051 + }, + { + "epoch": 3.29, + "learning_rate": 3.361153594057243e-06, + "loss": 1.2728, + "step": 15052 + }, + { + "epoch": 3.29, + "learning_rate": 3.3610443521957612e-06, + "loss": 2.1217, + "step": 15053 + }, + { + "epoch": 3.29, + "learning_rate": 3.3609351103342802e-06, + "loss": 2.0382, + "step": 15054 + }, + { + "epoch": 3.29, + "learning_rate": 3.360825868472799e-06, + "loss": 1.6865, + "step": 15055 + }, + { + "epoch": 3.29, + "learning_rate": 3.360716626611318e-06, + "loss": 2.4379, + "step": 15056 + }, + { + "epoch": 3.29, + "learning_rate": 3.3606073847498365e-06, + "loss": 1.0522, + "step": 15057 + }, + { + "epoch": 3.29, + "learning_rate": 3.3604981428883547e-06, + "loss": 1.4875, + "step": 15058 + }, + { + "epoch": 3.29, + "learning_rate": 3.3603889010268737e-06, + "loss": 1.5158, + "step": 15059 + }, + { + "epoch": 3.29, + "learning_rate": 3.3602796591653923e-06, + "loss": 2.788, + "step": 15060 + }, + { + "epoch": 3.29, + "learning_rate": 3.3601704173039114e-06, + "loss": 2.5237, + "step": 15061 + }, + { + "epoch": 3.29, + "learning_rate": 3.36006117544243e-06, + "loss": 1.7766, + "step": 15062 + }, + { + "epoch": 3.29, + "learning_rate": 3.359951933580948e-06, + "loss": 2.6352, + "step": 15063 + }, + { + "epoch": 3.29, + "learning_rate": 3.359842691719467e-06, + "loss": 2.503, + "step": 15064 + }, + { + "epoch": 3.29, + "learning_rate": 3.359733449857986e-06, + "loss": 2.4844, + "step": 15065 + }, + { + "epoch": 3.29, + "learning_rate": 3.359624207996505e-06, + "loss": 1.5195, + "step": 15066 + }, + { + "epoch": 3.29, + "learning_rate": 3.3595149661350235e-06, + "loss": 1.6589, + "step": 15067 + }, + { + "epoch": 3.29, + "learning_rate": 3.3594057242735417e-06, + "loss": 2.7492, + "step": 15068 + }, + { + "epoch": 3.29, + "learning_rate": 3.3592964824120607e-06, + "loss": 1.96, + "step": 15069 + }, + { + "epoch": 3.29, + "learning_rate": 3.3591872405505793e-06, + "loss": 1.8011, + "step": 15070 + }, + { + "epoch": 3.29, + "learning_rate": 3.359077998689098e-06, + "loss": 2.2616, + "step": 15071 + }, + { + "epoch": 3.29, + "learning_rate": 3.3589687568276165e-06, + "loss": 1.0453, + "step": 15072 + }, + { + "epoch": 3.29, + "learning_rate": 3.358859514966135e-06, + "loss": 1.917, + "step": 15073 + }, + { + "epoch": 3.29, + "learning_rate": 3.358750273104654e-06, + "loss": 1.5354, + "step": 15074 + }, + { + "epoch": 3.29, + "learning_rate": 3.3586410312431728e-06, + "loss": 1.7862, + "step": 15075 + }, + { + "epoch": 3.29, + "learning_rate": 3.3585317893816914e-06, + "loss": 2.0228, + "step": 15076 + }, + { + "epoch": 3.29, + "learning_rate": 3.35842254752021e-06, + "loss": 1.3811, + "step": 15077 + }, + { + "epoch": 3.29, + "learning_rate": 3.3583133056587286e-06, + "loss": 0.7624, + "step": 15078 + }, + { + "epoch": 3.29, + "learning_rate": 3.3582040637972472e-06, + "loss": 2.4432, + "step": 15079 + }, + { + "epoch": 3.29, + "learning_rate": 3.3580948219357663e-06, + "loss": 3.0022, + "step": 15080 + }, + { + "epoch": 3.29, + "learning_rate": 3.357985580074285e-06, + "loss": 3.7338, + "step": 15081 + }, + { + "epoch": 3.29, + "learning_rate": 3.357876338212803e-06, + "loss": 2.5039, + "step": 15082 + }, + { + "epoch": 3.29, + "learning_rate": 3.357767096351322e-06, + "loss": 1.77, + "step": 15083 + }, + { + "epoch": 3.29, + "learning_rate": 3.3576578544898407e-06, + "loss": 1.3082, + "step": 15084 + }, + { + "epoch": 3.29, + "learning_rate": 3.3575486126283597e-06, + "loss": 2.0849, + "step": 15085 + }, + { + "epoch": 3.29, + "learning_rate": 3.357439370766878e-06, + "loss": 2.1313, + "step": 15086 + }, + { + "epoch": 3.29, + "learning_rate": 3.3573301289053965e-06, + "loss": 2.2155, + "step": 15087 + }, + { + "epoch": 3.29, + "learning_rate": 3.3572208870439156e-06, + "loss": 2.7888, + "step": 15088 + }, + { + "epoch": 3.29, + "learning_rate": 3.357111645182434e-06, + "loss": 2.3089, + "step": 15089 + }, + { + "epoch": 3.29, + "learning_rate": 3.357002403320953e-06, + "loss": 1.9779, + "step": 15090 + }, + { + "epoch": 3.29, + "learning_rate": 3.3568931614594714e-06, + "loss": 3.3215, + "step": 15091 + }, + { + "epoch": 3.29, + "learning_rate": 3.35678391959799e-06, + "loss": 1.7247, + "step": 15092 + }, + { + "epoch": 3.29, + "learning_rate": 3.356674677736509e-06, + "loss": 1.9596, + "step": 15093 + }, + { + "epoch": 3.29, + "learning_rate": 3.3565654358750277e-06, + "loss": 2.1621, + "step": 15094 + }, + { + "epoch": 3.29, + "learning_rate": 3.3564561940135467e-06, + "loss": 2.345, + "step": 15095 + }, + { + "epoch": 3.29, + "learning_rate": 3.356346952152065e-06, + "loss": 2.3959, + "step": 15096 + }, + { + "epoch": 3.29, + "learning_rate": 3.3562377102905835e-06, + "loss": 1.8515, + "step": 15097 + }, + { + "epoch": 3.3, + "learning_rate": 3.3561284684291025e-06, + "loss": 2.1742, + "step": 15098 + }, + { + "epoch": 3.3, + "learning_rate": 3.356019226567621e-06, + "loss": 3.8594, + "step": 15099 + }, + { + "epoch": 3.3, + "learning_rate": 3.3559099847061397e-06, + "loss": 2.2058, + "step": 15100 + }, + { + "epoch": 3.3, + "learning_rate": 3.3558007428446584e-06, + "loss": 2.3597, + "step": 15101 + }, + { + "epoch": 3.3, + "learning_rate": 3.355691500983177e-06, + "loss": 3.9261, + "step": 15102 + }, + { + "epoch": 3.3, + "learning_rate": 3.3555822591216956e-06, + "loss": 3.2149, + "step": 15103 + }, + { + "epoch": 3.3, + "learning_rate": 3.3554730172602146e-06, + "loss": 0.9324, + "step": 15104 + }, + { + "epoch": 3.3, + "learning_rate": 3.355363775398733e-06, + "loss": 2.4561, + "step": 15105 + }, + { + "epoch": 3.3, + "learning_rate": 3.3552545335372514e-06, + "loss": 2.7588, + "step": 15106 + }, + { + "epoch": 3.3, + "learning_rate": 3.3551452916757704e-06, + "loss": 1.3062, + "step": 15107 + }, + { + "epoch": 3.3, + "learning_rate": 3.355036049814289e-06, + "loss": 1.9653, + "step": 15108 + }, + { + "epoch": 3.3, + "learning_rate": 3.354926807952808e-06, + "loss": 1.9809, + "step": 15109 + }, + { + "epoch": 3.3, + "learning_rate": 3.3548175660913263e-06, + "loss": 1.1725, + "step": 15110 + }, + { + "epoch": 3.3, + "learning_rate": 3.354708324229845e-06, + "loss": 1.7093, + "step": 15111 + }, + { + "epoch": 3.3, + "learning_rate": 3.354599082368364e-06, + "loss": 1.747, + "step": 15112 + }, + { + "epoch": 3.3, + "learning_rate": 3.3544898405068825e-06, + "loss": 2.2833, + "step": 15113 + }, + { + "epoch": 3.3, + "learning_rate": 3.3543805986454016e-06, + "loss": 2.3206, + "step": 15114 + }, + { + "epoch": 3.3, + "learning_rate": 3.3542713567839198e-06, + "loss": 2.5405, + "step": 15115 + }, + { + "epoch": 3.3, + "learning_rate": 3.3541621149224384e-06, + "loss": 2.3437, + "step": 15116 + }, + { + "epoch": 3.3, + "learning_rate": 3.3540528730609574e-06, + "loss": 1.8835, + "step": 15117 + }, + { + "epoch": 3.3, + "learning_rate": 3.353943631199476e-06, + "loss": 1.0616, + "step": 15118 + }, + { + "epoch": 3.3, + "learning_rate": 3.353834389337994e-06, + "loss": 1.4238, + "step": 15119 + }, + { + "epoch": 3.3, + "learning_rate": 3.3537251474765132e-06, + "loss": 1.4116, + "step": 15120 + }, + { + "epoch": 3.3, + "learning_rate": 3.353615905615032e-06, + "loss": 2.822, + "step": 15121 + }, + { + "epoch": 3.3, + "learning_rate": 3.353506663753551e-06, + "loss": 3.137, + "step": 15122 + }, + { + "epoch": 3.3, + "learning_rate": 3.3533974218920695e-06, + "loss": 1.2408, + "step": 15123 + }, + { + "epoch": 3.3, + "learning_rate": 3.3532881800305877e-06, + "loss": 1.697, + "step": 15124 + }, + { + "epoch": 3.3, + "learning_rate": 3.3531789381691067e-06, + "loss": 1.8767, + "step": 15125 + }, + { + "epoch": 3.3, + "learning_rate": 3.3530696963076253e-06, + "loss": 3.1258, + "step": 15126 + }, + { + "epoch": 3.3, + "learning_rate": 3.3529604544461444e-06, + "loss": 1.9653, + "step": 15127 + }, + { + "epoch": 3.3, + "learning_rate": 3.352851212584663e-06, + "loss": 1.7841, + "step": 15128 + }, + { + "epoch": 3.3, + "learning_rate": 3.352741970723181e-06, + "loss": 1.6792, + "step": 15129 + }, + { + "epoch": 3.3, + "learning_rate": 3.3526327288617e-06, + "loss": 2.1456, + "step": 15130 + }, + { + "epoch": 3.3, + "learning_rate": 3.352523487000219e-06, + "loss": 1.3689, + "step": 15131 + }, + { + "epoch": 3.3, + "learning_rate": 3.3524142451387374e-06, + "loss": 1.8706, + "step": 15132 + }, + { + "epoch": 3.3, + "learning_rate": 3.3523050032772564e-06, + "loss": 2.77, + "step": 15133 + }, + { + "epoch": 3.3, + "learning_rate": 3.3521957614157746e-06, + "loss": 2.8853, + "step": 15134 + }, + { + "epoch": 3.3, + "learning_rate": 3.3520865195542932e-06, + "loss": 2.3612, + "step": 15135 + }, + { + "epoch": 3.3, + "learning_rate": 3.3519772776928123e-06, + "loss": 2.7204, + "step": 15136 + }, + { + "epoch": 3.3, + "learning_rate": 3.351868035831331e-06, + "loss": 2.009, + "step": 15137 + }, + { + "epoch": 3.3, + "learning_rate": 3.351758793969849e-06, + "loss": 1.1844, + "step": 15138 + }, + { + "epoch": 3.3, + "learning_rate": 3.351649552108368e-06, + "loss": 1.5232, + "step": 15139 + }, + { + "epoch": 3.3, + "learning_rate": 3.3515403102468867e-06, + "loss": 0.9882, + "step": 15140 + }, + { + "epoch": 3.3, + "learning_rate": 3.3514310683854058e-06, + "loss": 2.5731, + "step": 15141 + }, + { + "epoch": 3.3, + "learning_rate": 3.3513218265239244e-06, + "loss": 2.0085, + "step": 15142 + }, + { + "epoch": 3.3, + "learning_rate": 3.3512125846624426e-06, + "loss": 1.1359, + "step": 15143 + }, + { + "epoch": 3.31, + "learning_rate": 3.3511033428009616e-06, + "loss": 1.6521, + "step": 15144 + }, + { + "epoch": 3.31, + "learning_rate": 3.35099410093948e-06, + "loss": 3.264, + "step": 15145 + }, + { + "epoch": 3.31, + "learning_rate": 3.3508848590779992e-06, + "loss": 1.8477, + "step": 15146 + }, + { + "epoch": 3.31, + "learning_rate": 3.350775617216518e-06, + "loss": 2.9089, + "step": 15147 + }, + { + "epoch": 3.31, + "learning_rate": 3.350666375355036e-06, + "loss": 1.1261, + "step": 15148 + }, + { + "epoch": 3.31, + "learning_rate": 3.350557133493555e-06, + "loss": 1.6146, + "step": 15149 + }, + { + "epoch": 3.31, + "learning_rate": 3.3504478916320737e-06, + "loss": 1.6215, + "step": 15150 + }, + { + "epoch": 3.31, + "learning_rate": 3.3503386497705927e-06, + "loss": 2.5109, + "step": 15151 + }, + { + "epoch": 3.31, + "learning_rate": 3.350229407909111e-06, + "loss": 0.9896, + "step": 15152 + }, + { + "epoch": 3.31, + "learning_rate": 3.3501201660476295e-06, + "loss": 2.6234, + "step": 15153 + }, + { + "epoch": 3.31, + "learning_rate": 3.3500109241861485e-06, + "loss": 3.5776, + "step": 15154 + }, + { + "epoch": 3.31, + "learning_rate": 3.349901682324667e-06, + "loss": 1.611, + "step": 15155 + }, + { + "epoch": 3.31, + "learning_rate": 3.3497924404631858e-06, + "loss": 2.0753, + "step": 15156 + }, + { + "epoch": 3.31, + "learning_rate": 3.3496831986017044e-06, + "loss": 2.3114, + "step": 15157 + }, + { + "epoch": 3.31, + "learning_rate": 3.349573956740223e-06, + "loss": 1.9264, + "step": 15158 + }, + { + "epoch": 3.31, + "learning_rate": 3.3494647148787416e-06, + "loss": 1.6702, + "step": 15159 + }, + { + "epoch": 3.31, + "learning_rate": 3.3493554730172606e-06, + "loss": 1.8251, + "step": 15160 + }, + { + "epoch": 3.31, + "learning_rate": 3.3492462311557792e-06, + "loss": 3.3453, + "step": 15161 + }, + { + "epoch": 3.31, + "learning_rate": 3.3491369892942974e-06, + "loss": 1.7926, + "step": 15162 + }, + { + "epoch": 3.31, + "learning_rate": 3.3490277474328165e-06, + "loss": 1.6619, + "step": 15163 + }, + { + "epoch": 3.31, + "learning_rate": 3.348918505571335e-06, + "loss": 2.2302, + "step": 15164 + }, + { + "epoch": 3.31, + "learning_rate": 3.348809263709854e-06, + "loss": 1.5669, + "step": 15165 + }, + { + "epoch": 3.31, + "learning_rate": 3.3487000218483727e-06, + "loss": 1.349, + "step": 15166 + }, + { + "epoch": 3.31, + "learning_rate": 3.348590779986891e-06, + "loss": 1.638, + "step": 15167 + }, + { + "epoch": 3.31, + "learning_rate": 3.34848153812541e-06, + "loss": 2.8534, + "step": 15168 + }, + { + "epoch": 3.31, + "learning_rate": 3.3483722962639286e-06, + "loss": 2.7626, + "step": 15169 + }, + { + "epoch": 3.31, + "learning_rate": 3.3482630544024476e-06, + "loss": 0.8952, + "step": 15170 + }, + { + "epoch": 3.31, + "learning_rate": 3.3481538125409658e-06, + "loss": 1.1473, + "step": 15171 + }, + { + "epoch": 3.31, + "learning_rate": 3.3480445706794844e-06, + "loss": 2.0065, + "step": 15172 + }, + { + "epoch": 3.31, + "learning_rate": 3.3479353288180034e-06, + "loss": 2.0646, + "step": 15173 + }, + { + "epoch": 3.31, + "learning_rate": 3.347826086956522e-06, + "loss": 2.3684, + "step": 15174 + }, + { + "epoch": 3.31, + "learning_rate": 3.347716845095041e-06, + "loss": 1.9017, + "step": 15175 + }, + { + "epoch": 3.31, + "learning_rate": 3.3476076032335593e-06, + "loss": 1.8878, + "step": 15176 + }, + { + "epoch": 3.31, + "learning_rate": 3.347498361372078e-06, + "loss": 3.241, + "step": 15177 + }, + { + "epoch": 3.31, + "learning_rate": 3.347389119510597e-06, + "loss": 2.0126, + "step": 15178 + }, + { + "epoch": 3.31, + "learning_rate": 3.3472798776491155e-06, + "loss": 2.2949, + "step": 15179 + }, + { + "epoch": 3.31, + "learning_rate": 3.3471706357876346e-06, + "loss": 1.1883, + "step": 15180 + }, + { + "epoch": 3.31, + "learning_rate": 3.3470613939261527e-06, + "loss": 2.2709, + "step": 15181 + }, + { + "epoch": 3.31, + "learning_rate": 3.3469521520646714e-06, + "loss": 1.5292, + "step": 15182 + }, + { + "epoch": 3.31, + "learning_rate": 3.3468429102031904e-06, + "loss": 2.0861, + "step": 15183 + }, + { + "epoch": 3.31, + "learning_rate": 3.346733668341709e-06, + "loss": 2.0435, + "step": 15184 + }, + { + "epoch": 3.31, + "learning_rate": 3.346624426480227e-06, + "loss": 1.7141, + "step": 15185 + }, + { + "epoch": 3.31, + "learning_rate": 3.3465151846187462e-06, + "loss": 2.229, + "step": 15186 + }, + { + "epoch": 3.31, + "learning_rate": 3.346405942757265e-06, + "loss": 1.3486, + "step": 15187 + }, + { + "epoch": 3.31, + "learning_rate": 3.3462967008957834e-06, + "loss": 1.9993, + "step": 15188 + }, + { + "epoch": 3.31, + "learning_rate": 3.3461874590343025e-06, + "loss": 2.7747, + "step": 15189 + }, + { + "epoch": 3.32, + "learning_rate": 3.3460782171728207e-06, + "loss": 1.9733, + "step": 15190 + }, + { + "epoch": 3.32, + "learning_rate": 3.3459689753113393e-06, + "loss": 1.2697, + "step": 15191 + }, + { + "epoch": 3.32, + "learning_rate": 3.3458597334498583e-06, + "loss": 1.8638, + "step": 15192 + }, + { + "epoch": 3.32, + "learning_rate": 3.345750491588377e-06, + "loss": 0.934, + "step": 15193 + }, + { + "epoch": 3.32, + "learning_rate": 3.345641249726896e-06, + "loss": 1.4786, + "step": 15194 + }, + { + "epoch": 3.32, + "learning_rate": 3.345532007865414e-06, + "loss": 1.6672, + "step": 15195 + }, + { + "epoch": 3.32, + "learning_rate": 3.3454227660039328e-06, + "loss": 3.4365, + "step": 15196 + }, + { + "epoch": 3.32, + "learning_rate": 3.3453135241424518e-06, + "loss": 2.7325, + "step": 15197 + }, + { + "epoch": 3.32, + "learning_rate": 3.3452042822809704e-06, + "loss": 3.4111, + "step": 15198 + }, + { + "epoch": 3.32, + "learning_rate": 3.3450950404194894e-06, + "loss": 1.401, + "step": 15199 + }, + { + "epoch": 3.32, + "learning_rate": 3.3449857985580076e-06, + "loss": 2.0075, + "step": 15200 + }, + { + "epoch": 3.32, + "learning_rate": 3.3448765566965262e-06, + "loss": 2.0135, + "step": 15201 + }, + { + "epoch": 3.32, + "learning_rate": 3.3447673148350453e-06, + "loss": 1.3545, + "step": 15202 + }, + { + "epoch": 3.32, + "learning_rate": 3.344658072973564e-06, + "loss": 2.5505, + "step": 15203 + }, + { + "epoch": 3.32, + "learning_rate": 3.344548831112082e-06, + "loss": 1.8425, + "step": 15204 + }, + { + "epoch": 3.32, + "learning_rate": 3.344439589250601e-06, + "loss": 2.8359, + "step": 15205 + }, + { + "epoch": 3.32, + "learning_rate": 3.3443303473891197e-06, + "loss": 1.9995, + "step": 15206 + }, + { + "epoch": 3.32, + "learning_rate": 3.3442211055276387e-06, + "loss": 1.5397, + "step": 15207 + }, + { + "epoch": 3.32, + "learning_rate": 3.3441118636661574e-06, + "loss": 1.1071, + "step": 15208 + }, + { + "epoch": 3.32, + "learning_rate": 3.3440026218046755e-06, + "loss": 0.9367, + "step": 15209 + }, + { + "epoch": 3.32, + "learning_rate": 3.3438933799431946e-06, + "loss": 2.626, + "step": 15210 + }, + { + "epoch": 3.32, + "learning_rate": 3.343784138081713e-06, + "loss": 1.8088, + "step": 15211 + }, + { + "epoch": 3.32, + "learning_rate": 3.343674896220232e-06, + "loss": 2.1199, + "step": 15212 + }, + { + "epoch": 3.32, + "learning_rate": 3.343565654358751e-06, + "loss": 2.499, + "step": 15213 + }, + { + "epoch": 3.32, + "learning_rate": 3.343456412497269e-06, + "loss": 3.0495, + "step": 15214 + }, + { + "epoch": 3.32, + "learning_rate": 3.3433471706357876e-06, + "loss": 2.1287, + "step": 15215 + }, + { + "epoch": 3.32, + "learning_rate": 3.3432379287743067e-06, + "loss": 1.468, + "step": 15216 + }, + { + "epoch": 3.32, + "learning_rate": 3.3431286869128253e-06, + "loss": 1.2718, + "step": 15217 + }, + { + "epoch": 3.32, + "learning_rate": 3.343019445051344e-06, + "loss": 2.8677, + "step": 15218 + }, + { + "epoch": 3.32, + "learning_rate": 3.3429102031898625e-06, + "loss": 2.7186, + "step": 15219 + }, + { + "epoch": 3.32, + "learning_rate": 3.342800961328381e-06, + "loss": 1.042, + "step": 15220 + }, + { + "epoch": 3.32, + "learning_rate": 3.3426917194669e-06, + "loss": 2.0339, + "step": 15221 + }, + { + "epoch": 3.32, + "learning_rate": 3.3425824776054188e-06, + "loss": 0.9757, + "step": 15222 + }, + { + "epoch": 3.32, + "learning_rate": 3.342473235743937e-06, + "loss": 1.378, + "step": 15223 + }, + { + "epoch": 3.32, + "learning_rate": 3.342363993882456e-06, + "loss": 3.1433, + "step": 15224 + }, + { + "epoch": 3.32, + "learning_rate": 3.3422547520209746e-06, + "loss": 2.6827, + "step": 15225 + }, + { + "epoch": 3.32, + "learning_rate": 3.3421455101594936e-06, + "loss": 2.2698, + "step": 15226 + }, + { + "epoch": 3.32, + "learning_rate": 3.3420362682980122e-06, + "loss": 0.6683, + "step": 15227 + }, + { + "epoch": 3.32, + "learning_rate": 3.3419270264365304e-06, + "loss": 1.7718, + "step": 15228 + }, + { + "epoch": 3.32, + "learning_rate": 3.3418177845750495e-06, + "loss": 1.7143, + "step": 15229 + }, + { + "epoch": 3.32, + "learning_rate": 3.341708542713568e-06, + "loss": 1.7487, + "step": 15230 + }, + { + "epoch": 3.32, + "learning_rate": 3.341599300852087e-06, + "loss": 1.974, + "step": 15231 + }, + { + "epoch": 3.32, + "learning_rate": 3.3414900589906057e-06, + "loss": 1.2901, + "step": 15232 + }, + { + "epoch": 3.32, + "learning_rate": 3.341380817129124e-06, + "loss": 2.4214, + "step": 15233 + }, + { + "epoch": 3.32, + "learning_rate": 3.341271575267643e-06, + "loss": 1.2881, + "step": 15234 + }, + { + "epoch": 3.32, + "learning_rate": 3.3411623334061615e-06, + "loss": 2.5119, + "step": 15235 + }, + { + "epoch": 3.33, + "learning_rate": 3.3410530915446806e-06, + "loss": 1.6365, + "step": 15236 + }, + { + "epoch": 3.33, + "learning_rate": 3.3409438496831988e-06, + "loss": 2.2998, + "step": 15237 + }, + { + "epoch": 3.33, + "learning_rate": 3.3408346078217174e-06, + "loss": 3.5402, + "step": 15238 + }, + { + "epoch": 3.33, + "learning_rate": 3.3407253659602364e-06, + "loss": 1.0662, + "step": 15239 + }, + { + "epoch": 3.33, + "learning_rate": 3.340616124098755e-06, + "loss": 2.1249, + "step": 15240 + }, + { + "epoch": 3.33, + "learning_rate": 3.3405068822372736e-06, + "loss": 1.5175, + "step": 15241 + }, + { + "epoch": 3.33, + "learning_rate": 3.3403976403757922e-06, + "loss": 1.4008, + "step": 15242 + }, + { + "epoch": 3.33, + "learning_rate": 3.340288398514311e-06, + "loss": 1.9434, + "step": 15243 + }, + { + "epoch": 3.33, + "learning_rate": 3.3401791566528295e-06, + "loss": 1.076, + "step": 15244 + }, + { + "epoch": 3.33, + "learning_rate": 3.3400699147913485e-06, + "loss": 0.7803, + "step": 15245 + }, + { + "epoch": 3.33, + "learning_rate": 3.339960672929867e-06, + "loss": 2.2865, + "step": 15246 + }, + { + "epoch": 3.33, + "learning_rate": 3.3398514310683853e-06, + "loss": 1.8838, + "step": 15247 + }, + { + "epoch": 3.33, + "learning_rate": 3.3397421892069043e-06, + "loss": 2.7764, + "step": 15248 + }, + { + "epoch": 3.33, + "learning_rate": 3.339632947345423e-06, + "loss": 1.1847, + "step": 15249 + }, + { + "epoch": 3.33, + "learning_rate": 3.339523705483942e-06, + "loss": 2.5288, + "step": 15250 + }, + { + "epoch": 3.33, + "learning_rate": 3.33941446362246e-06, + "loss": 2.309, + "step": 15251 + }, + { + "epoch": 3.33, + "learning_rate": 3.3393052217609788e-06, + "loss": 1.9206, + "step": 15252 + }, + { + "epoch": 3.33, + "learning_rate": 3.339195979899498e-06, + "loss": 1.1712, + "step": 15253 + }, + { + "epoch": 3.33, + "learning_rate": 3.3390867380380164e-06, + "loss": 2.0991, + "step": 15254 + }, + { + "epoch": 3.33, + "learning_rate": 3.3389774961765355e-06, + "loss": 2.0376, + "step": 15255 + }, + { + "epoch": 3.33, + "learning_rate": 3.3388682543150536e-06, + "loss": 2.1533, + "step": 15256 + }, + { + "epoch": 3.33, + "learning_rate": 3.3387590124535723e-06, + "loss": 2.0794, + "step": 15257 + }, + { + "epoch": 3.33, + "learning_rate": 3.3386497705920913e-06, + "loss": 2.3586, + "step": 15258 + }, + { + "epoch": 3.33, + "learning_rate": 3.33854052873061e-06, + "loss": 1.7323, + "step": 15259 + }, + { + "epoch": 3.33, + "learning_rate": 3.338431286869129e-06, + "loss": 1.6355, + "step": 15260 + }, + { + "epoch": 3.33, + "learning_rate": 3.338322045007647e-06, + "loss": 2.9978, + "step": 15261 + }, + { + "epoch": 3.33, + "learning_rate": 3.3382128031461657e-06, + "loss": 1.408, + "step": 15262 + }, + { + "epoch": 3.33, + "learning_rate": 3.3381035612846848e-06, + "loss": 1.4517, + "step": 15263 + }, + { + "epoch": 3.33, + "learning_rate": 3.3379943194232034e-06, + "loss": 2.2824, + "step": 15264 + }, + { + "epoch": 3.33, + "learning_rate": 3.337885077561722e-06, + "loss": 3.312, + "step": 15265 + }, + { + "epoch": 3.33, + "learning_rate": 3.3377758357002406e-06, + "loss": 2.249, + "step": 15266 + }, + { + "epoch": 3.33, + "learning_rate": 3.3376665938387592e-06, + "loss": 3.7954, + "step": 15267 + }, + { + "epoch": 3.33, + "learning_rate": 3.3375573519772782e-06, + "loss": 1.894, + "step": 15268 + }, + { + "epoch": 3.33, + "learning_rate": 3.337448110115797e-06, + "loss": 1.9335, + "step": 15269 + }, + { + "epoch": 3.33, + "learning_rate": 3.337338868254315e-06, + "loss": 2.1891, + "step": 15270 + }, + { + "epoch": 3.33, + "learning_rate": 3.337229626392834e-06, + "loss": 1.8295, + "step": 15271 + }, + { + "epoch": 3.33, + "learning_rate": 3.3371203845313527e-06, + "loss": 3.3686, + "step": 15272 + }, + { + "epoch": 3.33, + "learning_rate": 3.3370111426698713e-06, + "loss": 1.7603, + "step": 15273 + }, + { + "epoch": 3.33, + "learning_rate": 3.3369019008083903e-06, + "loss": 2.5109, + "step": 15274 + }, + { + "epoch": 3.33, + "learning_rate": 3.3367926589469085e-06, + "loss": 1.3048, + "step": 15275 + }, + { + "epoch": 3.33, + "learning_rate": 3.336683417085427e-06, + "loss": 3.1141, + "step": 15276 + }, + { + "epoch": 3.33, + "learning_rate": 3.336574175223946e-06, + "loss": 1.8676, + "step": 15277 + }, + { + "epoch": 3.33, + "learning_rate": 3.3364649333624648e-06, + "loss": 2.1375, + "step": 15278 + }, + { + "epoch": 3.33, + "learning_rate": 3.336355691500984e-06, + "loss": 2.4392, + "step": 15279 + }, + { + "epoch": 3.33, + "learning_rate": 3.336246449639502e-06, + "loss": 1.2704, + "step": 15280 + }, + { + "epoch": 3.34, + "learning_rate": 3.3361372077780206e-06, + "loss": 1.516, + "step": 15281 + }, + { + "epoch": 3.34, + "learning_rate": 3.3360279659165396e-06, + "loss": 1.9918, + "step": 15282 + }, + { + "epoch": 3.34, + "learning_rate": 3.3359187240550583e-06, + "loss": 0.6349, + "step": 15283 + }, + { + "epoch": 3.34, + "learning_rate": 3.3358094821935764e-06, + "loss": 0.6678, + "step": 15284 + }, + { + "epoch": 3.34, + "learning_rate": 3.3357002403320955e-06, + "loss": 1.7021, + "step": 15285 + }, + { + "epoch": 3.34, + "learning_rate": 3.335590998470614e-06, + "loss": 1.7891, + "step": 15286 + }, + { + "epoch": 3.34, + "learning_rate": 3.335481756609133e-06, + "loss": 1.4162, + "step": 15287 + }, + { + "epoch": 3.34, + "learning_rate": 3.3353725147476517e-06, + "loss": 3.3898, + "step": 15288 + }, + { + "epoch": 3.34, + "learning_rate": 3.33526327288617e-06, + "loss": 2.2105, + "step": 15289 + }, + { + "epoch": 3.34, + "learning_rate": 3.335154031024689e-06, + "loss": 2.1481, + "step": 15290 + }, + { + "epoch": 3.34, + "learning_rate": 3.3350447891632076e-06, + "loss": 2.3249, + "step": 15291 + }, + { + "epoch": 3.34, + "learning_rate": 3.3349355473017266e-06, + "loss": 1.5901, + "step": 15292 + }, + { + "epoch": 3.34, + "learning_rate": 3.3348263054402452e-06, + "loss": 0.9125, + "step": 15293 + }, + { + "epoch": 3.34, + "learning_rate": 3.3347170635787634e-06, + "loss": 2.2328, + "step": 15294 + }, + { + "epoch": 3.34, + "learning_rate": 3.3346078217172824e-06, + "loss": 2.1598, + "step": 15295 + }, + { + "epoch": 3.34, + "learning_rate": 3.334498579855801e-06, + "loss": 2.8448, + "step": 15296 + }, + { + "epoch": 3.34, + "learning_rate": 3.3343893379943197e-06, + "loss": 3.1816, + "step": 15297 + }, + { + "epoch": 3.34, + "learning_rate": 3.3342800961328387e-06, + "loss": 3.0464, + "step": 15298 + }, + { + "epoch": 3.34, + "learning_rate": 3.334170854271357e-06, + "loss": 2.2345, + "step": 15299 + }, + { + "epoch": 3.34, + "learning_rate": 3.3340616124098755e-06, + "loss": 1.6087, + "step": 15300 + }, + { + "epoch": 3.34, + "learning_rate": 3.3339523705483945e-06, + "loss": 1.8638, + "step": 15301 + }, + { + "epoch": 3.34, + "learning_rate": 3.333843128686913e-06, + "loss": 1.2788, + "step": 15302 + }, + { + "epoch": 3.34, + "learning_rate": 3.3337338868254313e-06, + "loss": 1.0717, + "step": 15303 + }, + { + "epoch": 3.34, + "learning_rate": 3.3336246449639504e-06, + "loss": 3.6776, + "step": 15304 + }, + { + "epoch": 3.34, + "learning_rate": 3.333515403102469e-06, + "loss": 0.6474, + "step": 15305 + }, + { + "epoch": 3.34, + "learning_rate": 3.333406161240988e-06, + "loss": 2.2142, + "step": 15306 + }, + { + "epoch": 3.34, + "learning_rate": 3.3332969193795066e-06, + "loss": 2.0854, + "step": 15307 + }, + { + "epoch": 3.34, + "learning_rate": 3.333187677518025e-06, + "loss": 1.647, + "step": 15308 + }, + { + "epoch": 3.34, + "learning_rate": 3.333078435656544e-06, + "loss": 1.3235, + "step": 15309 + }, + { + "epoch": 3.34, + "learning_rate": 3.3329691937950625e-06, + "loss": 2.1596, + "step": 15310 + }, + { + "epoch": 3.34, + "learning_rate": 3.3328599519335815e-06, + "loss": 1.2882, + "step": 15311 + }, + { + "epoch": 3.34, + "learning_rate": 3.3327507100721e-06, + "loss": 1.4752, + "step": 15312 + }, + { + "epoch": 3.34, + "learning_rate": 3.3326414682106183e-06, + "loss": 1.9755, + "step": 15313 + }, + { + "epoch": 3.34, + "learning_rate": 3.3325322263491373e-06, + "loss": 0.7239, + "step": 15314 + }, + { + "epoch": 3.34, + "learning_rate": 3.332422984487656e-06, + "loss": 1.2094, + "step": 15315 + }, + { + "epoch": 3.34, + "learning_rate": 3.332313742626175e-06, + "loss": 1.23, + "step": 15316 + }, + { + "epoch": 3.34, + "learning_rate": 3.332204500764693e-06, + "loss": 1.0706, + "step": 15317 + }, + { + "epoch": 3.34, + "learning_rate": 3.3320952589032118e-06, + "loss": 1.3194, + "step": 15318 + }, + { + "epoch": 3.34, + "learning_rate": 3.331986017041731e-06, + "loss": 2.4926, + "step": 15319 + }, + { + "epoch": 3.34, + "learning_rate": 3.3318767751802494e-06, + "loss": 1.3461, + "step": 15320 + }, + { + "epoch": 3.34, + "learning_rate": 3.3317675333187684e-06, + "loss": 2.6759, + "step": 15321 + }, + { + "epoch": 3.34, + "learning_rate": 3.3316582914572866e-06, + "loss": 2.3008, + "step": 15322 + }, + { + "epoch": 3.34, + "learning_rate": 3.3315490495958052e-06, + "loss": 2.2895, + "step": 15323 + }, + { + "epoch": 3.34, + "learning_rate": 3.3314398077343243e-06, + "loss": 1.1022, + "step": 15324 + }, + { + "epoch": 3.34, + "learning_rate": 3.331330565872843e-06, + "loss": 3.6956, + "step": 15325 + }, + { + "epoch": 3.34, + "learning_rate": 3.3312213240113615e-06, + "loss": 0.9455, + "step": 15326 + }, + { + "epoch": 3.35, + "learning_rate": 3.33111208214988e-06, + "loss": 2.0291, + "step": 15327 + }, + { + "epoch": 3.35, + "learning_rate": 3.3310028402883987e-06, + "loss": 0.9455, + "step": 15328 + }, + { + "epoch": 3.35, + "learning_rate": 3.3308935984269173e-06, + "loss": 2.0659, + "step": 15329 + }, + { + "epoch": 3.35, + "learning_rate": 3.3307843565654364e-06, + "loss": 2.2386, + "step": 15330 + }, + { + "epoch": 3.35, + "learning_rate": 3.330675114703955e-06, + "loss": 2.4065, + "step": 15331 + }, + { + "epoch": 3.35, + "learning_rate": 3.330565872842473e-06, + "loss": 2.0508, + "step": 15332 + }, + { + "epoch": 3.35, + "learning_rate": 3.330456630980992e-06, + "loss": 2.375, + "step": 15333 + }, + { + "epoch": 3.35, + "learning_rate": 3.330347389119511e-06, + "loss": 1.3957, + "step": 15334 + }, + { + "epoch": 3.35, + "learning_rate": 3.33023814725803e-06, + "loss": 1.883, + "step": 15335 + }, + { + "epoch": 3.35, + "learning_rate": 3.330128905396548e-06, + "loss": 0.7988, + "step": 15336 + }, + { + "epoch": 3.35, + "learning_rate": 3.3300196635350666e-06, + "loss": 1.4712, + "step": 15337 + }, + { + "epoch": 3.35, + "learning_rate": 3.3299104216735857e-06, + "loss": 4.228, + "step": 15338 + }, + { + "epoch": 3.35, + "learning_rate": 3.3298011798121043e-06, + "loss": 1.3969, + "step": 15339 + }, + { + "epoch": 3.35, + "learning_rate": 3.3296919379506233e-06, + "loss": 3.036, + "step": 15340 + }, + { + "epoch": 3.35, + "learning_rate": 3.3295826960891415e-06, + "loss": 1.5645, + "step": 15341 + }, + { + "epoch": 3.35, + "learning_rate": 3.32947345422766e-06, + "loss": 0.8552, + "step": 15342 + }, + { + "epoch": 3.35, + "learning_rate": 3.329364212366179e-06, + "loss": 1.7964, + "step": 15343 + }, + { + "epoch": 3.35, + "learning_rate": 3.3292549705046978e-06, + "loss": 1.8225, + "step": 15344 + }, + { + "epoch": 3.35, + "learning_rate": 3.329145728643217e-06, + "loss": 1.8369, + "step": 15345 + }, + { + "epoch": 3.35, + "learning_rate": 3.329036486781735e-06, + "loss": 2.5002, + "step": 15346 + }, + { + "epoch": 3.35, + "learning_rate": 3.3289272449202536e-06, + "loss": 0.7048, + "step": 15347 + }, + { + "epoch": 3.35, + "learning_rate": 3.3288180030587726e-06, + "loss": 3.0713, + "step": 15348 + }, + { + "epoch": 3.35, + "learning_rate": 3.3287087611972912e-06, + "loss": 1.6417, + "step": 15349 + }, + { + "epoch": 3.35, + "learning_rate": 3.3285995193358094e-06, + "loss": 3.2044, + "step": 15350 + }, + { + "epoch": 3.35, + "learning_rate": 3.3284902774743285e-06, + "loss": 1.003, + "step": 15351 + }, + { + "epoch": 3.35, + "learning_rate": 3.328381035612847e-06, + "loss": 1.463, + "step": 15352 + }, + { + "epoch": 3.35, + "learning_rate": 3.3282717937513657e-06, + "loss": 2.1365, + "step": 15353 + }, + { + "epoch": 3.35, + "learning_rate": 3.3281625518898847e-06, + "loss": 0.1726, + "step": 15354 + }, + { + "epoch": 3.35, + "learning_rate": 3.328053310028403e-06, + "loss": 0.7354, + "step": 15355 + }, + { + "epoch": 3.35, + "learning_rate": 3.3279440681669215e-06, + "loss": 2.2056, + "step": 15356 + }, + { + "epoch": 3.35, + "learning_rate": 3.3278348263054406e-06, + "loss": 1.6328, + "step": 15357 + }, + { + "epoch": 3.35, + "learning_rate": 3.327725584443959e-06, + "loss": 1.9183, + "step": 15358 + }, + { + "epoch": 3.35, + "learning_rate": 3.327616342582478e-06, + "loss": 1.9518, + "step": 15359 + }, + { + "epoch": 3.35, + "learning_rate": 3.3275071007209964e-06, + "loss": 1.0704, + "step": 15360 + }, + { + "epoch": 3.35, + "learning_rate": 3.327397858859515e-06, + "loss": 2.101, + "step": 15361 + }, + { + "epoch": 3.35, + "learning_rate": 3.327288616998034e-06, + "loss": 2.2503, + "step": 15362 + }, + { + "epoch": 3.35, + "learning_rate": 3.3271793751365526e-06, + "loss": 2.0097, + "step": 15363 + }, + { + "epoch": 3.35, + "learning_rate": 3.3270701332750717e-06, + "loss": 1.7184, + "step": 15364 + }, + { + "epoch": 3.35, + "learning_rate": 3.32696089141359e-06, + "loss": 1.1407, + "step": 15365 + }, + { + "epoch": 3.35, + "learning_rate": 3.3268516495521085e-06, + "loss": 2.1717, + "step": 15366 + }, + { + "epoch": 3.35, + "learning_rate": 3.3267424076906275e-06, + "loss": 0.9653, + "step": 15367 + }, + { + "epoch": 3.35, + "learning_rate": 3.326633165829146e-06, + "loss": 2.8032, + "step": 15368 + }, + { + "epoch": 3.35, + "learning_rate": 3.3265239239676643e-06, + "loss": 1.0315, + "step": 15369 + }, + { + "epoch": 3.35, + "learning_rate": 3.3264146821061833e-06, + "loss": 2.3218, + "step": 15370 + }, + { + "epoch": 3.35, + "learning_rate": 3.326305440244702e-06, + "loss": 1.3877, + "step": 15371 + }, + { + "epoch": 3.35, + "learning_rate": 3.326196198383221e-06, + "loss": 2.945, + "step": 15372 + }, + { + "epoch": 3.36, + "learning_rate": 3.3260869565217396e-06, + "loss": 2.3248, + "step": 15373 + }, + { + "epoch": 3.36, + "learning_rate": 3.325977714660258e-06, + "loss": 0.6211, + "step": 15374 + }, + { + "epoch": 3.36, + "learning_rate": 3.325868472798777e-06, + "loss": 1.8511, + "step": 15375 + }, + { + "epoch": 3.36, + "learning_rate": 3.3257592309372954e-06, + "loss": 2.6633, + "step": 15376 + }, + { + "epoch": 3.36, + "learning_rate": 3.3256499890758145e-06, + "loss": 2.3223, + "step": 15377 + }, + { + "epoch": 3.36, + "learning_rate": 3.325540747214333e-06, + "loss": 2.6894, + "step": 15378 + }, + { + "epoch": 3.36, + "learning_rate": 3.3254315053528513e-06, + "loss": 2.2663, + "step": 15379 + }, + { + "epoch": 3.36, + "learning_rate": 3.3253222634913703e-06, + "loss": 2.5504, + "step": 15380 + }, + { + "epoch": 3.36, + "learning_rate": 3.325213021629889e-06, + "loss": 2.6481, + "step": 15381 + }, + { + "epoch": 3.36, + "learning_rate": 3.3251037797684075e-06, + "loss": 0.7591, + "step": 15382 + }, + { + "epoch": 3.36, + "learning_rate": 3.324994537906926e-06, + "loss": 2.7781, + "step": 15383 + }, + { + "epoch": 3.36, + "learning_rate": 3.3248852960454447e-06, + "loss": 1.9343, + "step": 15384 + }, + { + "epoch": 3.36, + "learning_rate": 3.3247760541839634e-06, + "loss": 1.7497, + "step": 15385 + }, + { + "epoch": 3.36, + "learning_rate": 3.3246668123224824e-06, + "loss": 4.3089, + "step": 15386 + }, + { + "epoch": 3.36, + "learning_rate": 3.324557570461001e-06, + "loss": 1.9599, + "step": 15387 + }, + { + "epoch": 3.36, + "learning_rate": 3.324448328599519e-06, + "loss": 1.3795, + "step": 15388 + }, + { + "epoch": 3.36, + "learning_rate": 3.3243390867380382e-06, + "loss": 1.2065, + "step": 15389 + }, + { + "epoch": 3.36, + "learning_rate": 3.324229844876557e-06, + "loss": 2.7304, + "step": 15390 + }, + { + "epoch": 3.36, + "learning_rate": 3.324120603015076e-06, + "loss": 2.0344, + "step": 15391 + }, + { + "epoch": 3.36, + "learning_rate": 3.3240113611535945e-06, + "loss": 1.435, + "step": 15392 + }, + { + "epoch": 3.36, + "learning_rate": 3.3239021192921127e-06, + "loss": 1.8003, + "step": 15393 + }, + { + "epoch": 3.36, + "learning_rate": 3.3237928774306317e-06, + "loss": 2.5331, + "step": 15394 + }, + { + "epoch": 3.36, + "learning_rate": 3.3236836355691503e-06, + "loss": 3.5005, + "step": 15395 + }, + { + "epoch": 3.36, + "learning_rate": 3.3235743937076693e-06, + "loss": 2.7834, + "step": 15396 + }, + { + "epoch": 3.36, + "learning_rate": 3.3234651518461875e-06, + "loss": 1.5401, + "step": 15397 + }, + { + "epoch": 3.36, + "learning_rate": 3.323355909984706e-06, + "loss": 2.0555, + "step": 15398 + }, + { + "epoch": 3.36, + "learning_rate": 3.323246668123225e-06, + "loss": 1.5021, + "step": 15399 + }, + { + "epoch": 3.36, + "learning_rate": 3.323137426261744e-06, + "loss": 2.8523, + "step": 15400 + }, + { + "epoch": 3.36, + "learning_rate": 3.323028184400263e-06, + "loss": 2.4098, + "step": 15401 + }, + { + "epoch": 3.36, + "learning_rate": 3.322918942538781e-06, + "loss": 3.0388, + "step": 15402 + }, + { + "epoch": 3.36, + "learning_rate": 3.3228097006772996e-06, + "loss": 1.238, + "step": 15403 + }, + { + "epoch": 3.36, + "learning_rate": 3.3227004588158187e-06, + "loss": 1.2931, + "step": 15404 + }, + { + "epoch": 3.36, + "learning_rate": 3.3225912169543373e-06, + "loss": 2.1333, + "step": 15405 + }, + { + "epoch": 3.36, + "learning_rate": 3.322481975092856e-06, + "loss": 2.2385, + "step": 15406 + }, + { + "epoch": 3.36, + "learning_rate": 3.3223727332313745e-06, + "loss": 1.9776, + "step": 15407 + }, + { + "epoch": 3.36, + "learning_rate": 3.322263491369893e-06, + "loss": 1.5024, + "step": 15408 + }, + { + "epoch": 3.36, + "learning_rate": 3.3221542495084117e-06, + "loss": 1.9846, + "step": 15409 + }, + { + "epoch": 3.36, + "learning_rate": 3.3220450076469307e-06, + "loss": 1.536, + "step": 15410 + }, + { + "epoch": 3.36, + "learning_rate": 3.3219357657854494e-06, + "loss": 2.4122, + "step": 15411 + }, + { + "epoch": 3.36, + "learning_rate": 3.321826523923968e-06, + "loss": 2.1889, + "step": 15412 + }, + { + "epoch": 3.36, + "learning_rate": 3.3217172820624866e-06, + "loss": 2.1425, + "step": 15413 + }, + { + "epoch": 3.36, + "learning_rate": 3.321608040201005e-06, + "loss": 0.9621, + "step": 15414 + }, + { + "epoch": 3.36, + "learning_rate": 3.3214987983395242e-06, + "loss": 2.7782, + "step": 15415 + }, + { + "epoch": 3.36, + "learning_rate": 3.3213895564780424e-06, + "loss": 2.6112, + "step": 15416 + }, + { + "epoch": 3.36, + "learning_rate": 3.321280314616561e-06, + "loss": 1.6133, + "step": 15417 + }, + { + "epoch": 3.36, + "learning_rate": 3.32117107275508e-06, + "loss": 1.293, + "step": 15418 + }, + { + "epoch": 3.37, + "learning_rate": 3.3210618308935987e-06, + "loss": 1.5473, + "step": 15419 + }, + { + "epoch": 3.37, + "learning_rate": 3.3209525890321177e-06, + "loss": 2.3695, + "step": 15420 + }, + { + "epoch": 3.37, + "learning_rate": 3.320843347170636e-06, + "loss": 2.4504, + "step": 15421 + }, + { + "epoch": 3.37, + "learning_rate": 3.3207341053091545e-06, + "loss": 2.6188, + "step": 15422 + }, + { + "epoch": 3.37, + "learning_rate": 3.3206248634476735e-06, + "loss": 2.0664, + "step": 15423 + }, + { + "epoch": 3.37, + "learning_rate": 3.320515621586192e-06, + "loss": 1.4428, + "step": 15424 + }, + { + "epoch": 3.37, + "learning_rate": 3.320406379724711e-06, + "loss": 2.4745, + "step": 15425 + }, + { + "epoch": 3.37, + "learning_rate": 3.3202971378632294e-06, + "loss": 1.3312, + "step": 15426 + }, + { + "epoch": 3.37, + "learning_rate": 3.320187896001748e-06, + "loss": 2.8832, + "step": 15427 + }, + { + "epoch": 3.37, + "learning_rate": 3.320078654140267e-06, + "loss": 1.7147, + "step": 15428 + }, + { + "epoch": 3.37, + "learning_rate": 3.3199694122787856e-06, + "loss": 2.1405, + "step": 15429 + }, + { + "epoch": 3.37, + "learning_rate": 3.319860170417304e-06, + "loss": 2.3172, + "step": 15430 + }, + { + "epoch": 3.37, + "learning_rate": 3.319750928555823e-06, + "loss": 1.0351, + "step": 15431 + }, + { + "epoch": 3.37, + "learning_rate": 3.3196416866943415e-06, + "loss": 1.9351, + "step": 15432 + }, + { + "epoch": 3.37, + "learning_rate": 3.3195324448328605e-06, + "loss": 1.7721, + "step": 15433 + }, + { + "epoch": 3.37, + "learning_rate": 3.319423202971379e-06, + "loss": 2.3146, + "step": 15434 + }, + { + "epoch": 3.37, + "learning_rate": 3.3193139611098973e-06, + "loss": 3.6248, + "step": 15435 + }, + { + "epoch": 3.37, + "learning_rate": 3.3192047192484163e-06, + "loss": 2.0416, + "step": 15436 + }, + { + "epoch": 3.37, + "learning_rate": 3.319095477386935e-06, + "loss": 2.2731, + "step": 15437 + }, + { + "epoch": 3.37, + "learning_rate": 3.3189862355254536e-06, + "loss": 0.7436, + "step": 15438 + }, + { + "epoch": 3.37, + "learning_rate": 3.3188769936639726e-06, + "loss": 0.885, + "step": 15439 + }, + { + "epoch": 3.37, + "learning_rate": 3.3187677518024908e-06, + "loss": 1.4999, + "step": 15440 + }, + { + "epoch": 3.37, + "learning_rate": 3.3186585099410094e-06, + "loss": 2.244, + "step": 15441 + }, + { + "epoch": 3.37, + "learning_rate": 3.3185492680795284e-06, + "loss": 1.8187, + "step": 15442 + }, + { + "epoch": 3.37, + "learning_rate": 3.318440026218047e-06, + "loss": 2.3857, + "step": 15443 + }, + { + "epoch": 3.37, + "learning_rate": 3.318330784356566e-06, + "loss": 2.6068, + "step": 15444 + }, + { + "epoch": 3.37, + "learning_rate": 3.3182215424950843e-06, + "loss": 1.6723, + "step": 15445 + }, + { + "epoch": 3.37, + "learning_rate": 3.318112300633603e-06, + "loss": 2.5212, + "step": 15446 + }, + { + "epoch": 3.37, + "learning_rate": 3.318003058772122e-06, + "loss": 1.5495, + "step": 15447 + }, + { + "epoch": 3.37, + "learning_rate": 3.3178938169106405e-06, + "loss": 2.1263, + "step": 15448 + }, + { + "epoch": 3.37, + "learning_rate": 3.3177845750491587e-06, + "loss": 1.5035, + "step": 15449 + }, + { + "epoch": 3.37, + "learning_rate": 3.3176753331876777e-06, + "loss": 3.7722, + "step": 15450 + }, + { + "epoch": 3.37, + "learning_rate": 3.3175660913261963e-06, + "loss": 1.9823, + "step": 15451 + }, + { + "epoch": 3.37, + "learning_rate": 3.3174568494647154e-06, + "loss": 2.9737, + "step": 15452 + }, + { + "epoch": 3.37, + "learning_rate": 3.317347607603234e-06, + "loss": 2.625, + "step": 15453 + }, + { + "epoch": 3.37, + "learning_rate": 3.317238365741752e-06, + "loss": 2.1452, + "step": 15454 + }, + { + "epoch": 3.37, + "learning_rate": 3.317129123880271e-06, + "loss": 1.2792, + "step": 15455 + }, + { + "epoch": 3.37, + "learning_rate": 3.31701988201879e-06, + "loss": 1.5021, + "step": 15456 + }, + { + "epoch": 3.37, + "learning_rate": 3.316910640157309e-06, + "loss": 1.6238, + "step": 15457 + }, + { + "epoch": 3.37, + "learning_rate": 3.3168013982958275e-06, + "loss": 2.8822, + "step": 15458 + }, + { + "epoch": 3.37, + "learning_rate": 3.3166921564343457e-06, + "loss": 1.939, + "step": 15459 + }, + { + "epoch": 3.37, + "learning_rate": 3.3165829145728647e-06, + "loss": 1.4529, + "step": 15460 + }, + { + "epoch": 3.37, + "learning_rate": 3.3164736727113833e-06, + "loss": 2.4944, + "step": 15461 + }, + { + "epoch": 3.37, + "learning_rate": 3.316364430849902e-06, + "loss": 1.1257, + "step": 15462 + }, + { + "epoch": 3.37, + "learning_rate": 3.3162551889884205e-06, + "loss": 2.7113, + "step": 15463 + }, + { + "epoch": 3.37, + "learning_rate": 3.316145947126939e-06, + "loss": 1.1789, + "step": 15464 + }, + { + "epoch": 3.38, + "learning_rate": 3.316036705265458e-06, + "loss": 3.3385, + "step": 15465 + }, + { + "epoch": 3.38, + "learning_rate": 3.3159274634039768e-06, + "loss": 1.5694, + "step": 15466 + }, + { + "epoch": 3.38, + "learning_rate": 3.3158182215424954e-06, + "loss": 1.5635, + "step": 15467 + }, + { + "epoch": 3.38, + "learning_rate": 3.315708979681014e-06, + "loss": 1.8893, + "step": 15468 + }, + { + "epoch": 3.38, + "learning_rate": 3.3155997378195326e-06, + "loss": 1.0618, + "step": 15469 + }, + { + "epoch": 3.38, + "learning_rate": 3.3154904959580512e-06, + "loss": 2.3549, + "step": 15470 + }, + { + "epoch": 3.38, + "learning_rate": 3.3153812540965703e-06, + "loss": 2.3534, + "step": 15471 + }, + { + "epoch": 3.38, + "learning_rate": 3.315272012235089e-06, + "loss": 0.7965, + "step": 15472 + }, + { + "epoch": 3.38, + "learning_rate": 3.315162770373607e-06, + "loss": 1.8504, + "step": 15473 + }, + { + "epoch": 3.38, + "learning_rate": 3.315053528512126e-06, + "loss": 2.0667, + "step": 15474 + }, + { + "epoch": 3.38, + "learning_rate": 3.3149442866506447e-06, + "loss": 2.4862, + "step": 15475 + }, + { + "epoch": 3.38, + "learning_rate": 3.3148350447891637e-06, + "loss": 2.3527, + "step": 15476 + }, + { + "epoch": 3.38, + "learning_rate": 3.3147258029276823e-06, + "loss": 1.7852, + "step": 15477 + }, + { + "epoch": 3.38, + "learning_rate": 3.3146165610662005e-06, + "loss": 2.7056, + "step": 15478 + }, + { + "epoch": 3.38, + "learning_rate": 3.3145073192047196e-06, + "loss": 2.653, + "step": 15479 + }, + { + "epoch": 3.38, + "learning_rate": 3.314398077343238e-06, + "loss": 1.8072, + "step": 15480 + }, + { + "epoch": 3.38, + "learning_rate": 3.314288835481757e-06, + "loss": 1.3624, + "step": 15481 + }, + { + "epoch": 3.38, + "learning_rate": 3.3141795936202754e-06, + "loss": 1.444, + "step": 15482 + }, + { + "epoch": 3.38, + "learning_rate": 3.314070351758794e-06, + "loss": 1.3739, + "step": 15483 + }, + { + "epoch": 3.38, + "learning_rate": 3.313961109897313e-06, + "loss": 2.1566, + "step": 15484 + }, + { + "epoch": 3.38, + "learning_rate": 3.3138518680358317e-06, + "loss": 0.8545, + "step": 15485 + }, + { + "epoch": 3.38, + "learning_rate": 3.3137426261743507e-06, + "loss": 2.1865, + "step": 15486 + }, + { + "epoch": 3.38, + "learning_rate": 3.313633384312869e-06, + "loss": 3.2145, + "step": 15487 + }, + { + "epoch": 3.38, + "learning_rate": 3.3135241424513875e-06, + "loss": 1.7111, + "step": 15488 + }, + { + "epoch": 3.38, + "learning_rate": 3.3134149005899065e-06, + "loss": 3.7846, + "step": 15489 + }, + { + "epoch": 3.38, + "learning_rate": 3.313305658728425e-06, + "loss": 1.8284, + "step": 15490 + }, + { + "epoch": 3.38, + "learning_rate": 3.3131964168669437e-06, + "loss": 1.8428, + "step": 15491 + }, + { + "epoch": 3.38, + "learning_rate": 3.3130871750054624e-06, + "loss": 2.4558, + "step": 15492 + }, + { + "epoch": 3.38, + "learning_rate": 3.312977933143981e-06, + "loss": 2.2592, + "step": 15493 + }, + { + "epoch": 3.38, + "learning_rate": 3.3128686912824996e-06, + "loss": 1.857, + "step": 15494 + }, + { + "epoch": 3.38, + "learning_rate": 3.3127594494210186e-06, + "loss": 2.0848, + "step": 15495 + }, + { + "epoch": 3.38, + "learning_rate": 3.312650207559537e-06, + "loss": 1.7455, + "step": 15496 + }, + { + "epoch": 3.38, + "learning_rate": 3.3125409656980554e-06, + "loss": 1.6382, + "step": 15497 + }, + { + "epoch": 3.38, + "learning_rate": 3.3124317238365744e-06, + "loss": 0.7805, + "step": 15498 + }, + { + "epoch": 3.38, + "learning_rate": 3.312322481975093e-06, + "loss": 1.5779, + "step": 15499 + }, + { + "epoch": 3.38, + "learning_rate": 3.312213240113612e-06, + "loss": 2.0492, + "step": 15500 + }, + { + "epoch": 3.38, + "learning_rate": 3.3121039982521303e-06, + "loss": 3.8605, + "step": 15501 + }, + { + "epoch": 3.38, + "learning_rate": 3.311994756390649e-06, + "loss": 3.3892, + "step": 15502 + }, + { + "epoch": 3.38, + "learning_rate": 3.311885514529168e-06, + "loss": 2.0218, + "step": 15503 + }, + { + "epoch": 3.38, + "learning_rate": 3.3117762726676865e-06, + "loss": 3.5155, + "step": 15504 + }, + { + "epoch": 3.38, + "learning_rate": 3.3116670308062056e-06, + "loss": 3.1043, + "step": 15505 + }, + { + "epoch": 3.38, + "learning_rate": 3.3115577889447238e-06, + "loss": 1.9465, + "step": 15506 + }, + { + "epoch": 3.38, + "learning_rate": 3.3114485470832424e-06, + "loss": 2.2651, + "step": 15507 + }, + { + "epoch": 3.38, + "learning_rate": 3.3113393052217614e-06, + "loss": 2.1102, + "step": 15508 + }, + { + "epoch": 3.38, + "learning_rate": 3.31123006336028e-06, + "loss": 0.6967, + "step": 15509 + }, + { + "epoch": 3.38, + "learning_rate": 3.311120821498799e-06, + "loss": 1.7527, + "step": 15510 + }, + { + "epoch": 3.39, + "learning_rate": 3.3110115796373172e-06, + "loss": 1.9548, + "step": 15511 + }, + { + "epoch": 3.39, + "learning_rate": 3.310902337775836e-06, + "loss": 2.2863, + "step": 15512 + }, + { + "epoch": 3.39, + "learning_rate": 3.310793095914355e-06, + "loss": 2.3608, + "step": 15513 + }, + { + "epoch": 3.39, + "learning_rate": 3.3106838540528735e-06, + "loss": 1.7955, + "step": 15514 + }, + { + "epoch": 3.39, + "learning_rate": 3.3105746121913917e-06, + "loss": 1.8802, + "step": 15515 + }, + { + "epoch": 3.39, + "learning_rate": 3.3104653703299107e-06, + "loss": 3.0035, + "step": 15516 + }, + { + "epoch": 3.39, + "learning_rate": 3.3103561284684293e-06, + "loss": 3.7212, + "step": 15517 + }, + { + "epoch": 3.39, + "learning_rate": 3.3102468866069484e-06, + "loss": 2.7135, + "step": 15518 + }, + { + "epoch": 3.39, + "learning_rate": 3.310137644745467e-06, + "loss": 1.9075, + "step": 15519 + }, + { + "epoch": 3.39, + "learning_rate": 3.310028402883985e-06, + "loss": 2.519, + "step": 15520 + }, + { + "epoch": 3.39, + "learning_rate": 3.309919161022504e-06, + "loss": 2.2536, + "step": 15521 + }, + { + "epoch": 3.39, + "learning_rate": 3.309809919161023e-06, + "loss": 1.8391, + "step": 15522 + }, + { + "epoch": 3.39, + "learning_rate": 3.3097006772995414e-06, + "loss": 2.0374, + "step": 15523 + }, + { + "epoch": 3.39, + "learning_rate": 3.3095914354380604e-06, + "loss": 1.3458, + "step": 15524 + }, + { + "epoch": 3.39, + "learning_rate": 3.3094821935765786e-06, + "loss": 2.5317, + "step": 15525 + }, + { + "epoch": 3.39, + "learning_rate": 3.3093729517150972e-06, + "loss": 2.521, + "step": 15526 + }, + { + "epoch": 3.39, + "learning_rate": 3.3092637098536163e-06, + "loss": 2.2427, + "step": 15527 + }, + { + "epoch": 3.39, + "learning_rate": 3.309154467992135e-06, + "loss": 1.5515, + "step": 15528 + }, + { + "epoch": 3.39, + "learning_rate": 3.309045226130653e-06, + "loss": 2.5067, + "step": 15529 + }, + { + "epoch": 3.39, + "learning_rate": 3.308935984269172e-06, + "loss": 1.6108, + "step": 15530 + }, + { + "epoch": 3.39, + "learning_rate": 3.3088267424076907e-06, + "loss": 1.7076, + "step": 15531 + }, + { + "epoch": 3.39, + "learning_rate": 3.3087175005462098e-06, + "loss": 0.982, + "step": 15532 + }, + { + "epoch": 3.39, + "learning_rate": 3.3086082586847284e-06, + "loss": 2.1724, + "step": 15533 + }, + { + "epoch": 3.39, + "learning_rate": 3.3084990168232466e-06, + "loss": 2.3949, + "step": 15534 + }, + { + "epoch": 3.39, + "learning_rate": 3.3083897749617656e-06, + "loss": 2.0277, + "step": 15535 + }, + { + "epoch": 3.39, + "learning_rate": 3.308280533100284e-06, + "loss": 1.7087, + "step": 15536 + }, + { + "epoch": 3.39, + "learning_rate": 3.3081712912388032e-06, + "loss": 3.4052, + "step": 15537 + }, + { + "epoch": 3.39, + "learning_rate": 3.308062049377322e-06, + "loss": 3.6105, + "step": 15538 + }, + { + "epoch": 3.39, + "learning_rate": 3.30795280751584e-06, + "loss": 1.7664, + "step": 15539 + }, + { + "epoch": 3.39, + "learning_rate": 3.307843565654359e-06, + "loss": 2.7317, + "step": 15540 + }, + { + "epoch": 3.39, + "learning_rate": 3.3077343237928777e-06, + "loss": 2.3505, + "step": 15541 + }, + { + "epoch": 3.39, + "learning_rate": 3.3076250819313967e-06, + "loss": 2.8729, + "step": 15542 + }, + { + "epoch": 3.39, + "learning_rate": 3.3075158400699153e-06, + "loss": 2.2168, + "step": 15543 + }, + { + "epoch": 3.39, + "learning_rate": 3.3074065982084335e-06, + "loss": 1.2883, + "step": 15544 + }, + { + "epoch": 3.39, + "learning_rate": 3.3072973563469525e-06, + "loss": 2.4868, + "step": 15545 + }, + { + "epoch": 3.39, + "learning_rate": 3.307188114485471e-06, + "loss": 1.2361, + "step": 15546 + }, + { + "epoch": 3.39, + "learning_rate": 3.3070788726239898e-06, + "loss": 0.8451, + "step": 15547 + }, + { + "epoch": 3.39, + "learning_rate": 3.3069696307625084e-06, + "loss": 2.2789, + "step": 15548 + }, + { + "epoch": 3.39, + "learning_rate": 3.306860388901027e-06, + "loss": 1.1157, + "step": 15549 + }, + { + "epoch": 3.39, + "learning_rate": 3.3067511470395456e-06, + "loss": 2.3256, + "step": 15550 + }, + { + "epoch": 3.39, + "learning_rate": 3.3066419051780646e-06, + "loss": 2.5826, + "step": 15551 + }, + { + "epoch": 3.39, + "learning_rate": 3.3065326633165833e-06, + "loss": 0.5684, + "step": 15552 + }, + { + "epoch": 3.39, + "learning_rate": 3.3064234214551014e-06, + "loss": 1.8499, + "step": 15553 + }, + { + "epoch": 3.39, + "learning_rate": 3.3063141795936205e-06, + "loss": 1.2613, + "step": 15554 + }, + { + "epoch": 3.39, + "learning_rate": 3.306204937732139e-06, + "loss": 1.9168, + "step": 15555 + }, + { + "epoch": 3.4, + "learning_rate": 3.306095695870658e-06, + "loss": 1.7743, + "step": 15556 + }, + { + "epoch": 3.4, + "learning_rate": 3.3059864540091767e-06, + "loss": 1.5132, + "step": 15557 + }, + { + "epoch": 3.4, + "learning_rate": 3.305877212147695e-06, + "loss": 2.6241, + "step": 15558 + }, + { + "epoch": 3.4, + "learning_rate": 3.305767970286214e-06, + "loss": 1.9469, + "step": 15559 + }, + { + "epoch": 3.4, + "learning_rate": 3.3056587284247326e-06, + "loss": 1.8356, + "step": 15560 + }, + { + "epoch": 3.4, + "learning_rate": 3.3055494865632516e-06, + "loss": 1.1111, + "step": 15561 + }, + { + "epoch": 3.4, + "learning_rate": 3.3054402447017698e-06, + "loss": 1.5886, + "step": 15562 + }, + { + "epoch": 3.4, + "learning_rate": 3.3053310028402884e-06, + "loss": 0.5801, + "step": 15563 + }, + { + "epoch": 3.4, + "learning_rate": 3.3052217609788074e-06, + "loss": 1.6343, + "step": 15564 + }, + { + "epoch": 3.4, + "learning_rate": 3.305112519117326e-06, + "loss": 1.4833, + "step": 15565 + }, + { + "epoch": 3.4, + "learning_rate": 3.305003277255845e-06, + "loss": 1.3792, + "step": 15566 + }, + { + "epoch": 3.4, + "learning_rate": 3.3048940353943633e-06, + "loss": 5.1115, + "step": 15567 + }, + { + "epoch": 3.4, + "learning_rate": 3.304784793532882e-06, + "loss": 2.1457, + "step": 15568 + }, + { + "epoch": 3.4, + "learning_rate": 3.304675551671401e-06, + "loss": 1.991, + "step": 15569 + }, + { + "epoch": 3.4, + "learning_rate": 3.3045663098099195e-06, + "loss": 2.1054, + "step": 15570 + }, + { + "epoch": 3.4, + "learning_rate": 3.3044570679484386e-06, + "loss": 2.2691, + "step": 15571 + }, + { + "epoch": 3.4, + "learning_rate": 3.3043478260869567e-06, + "loss": 1.5081, + "step": 15572 + }, + { + "epoch": 3.4, + "learning_rate": 3.3042385842254754e-06, + "loss": 1.7125, + "step": 15573 + }, + { + "epoch": 3.4, + "learning_rate": 3.3041293423639944e-06, + "loss": 1.0956, + "step": 15574 + }, + { + "epoch": 3.4, + "learning_rate": 3.304020100502513e-06, + "loss": 2.3512, + "step": 15575 + }, + { + "epoch": 3.4, + "learning_rate": 3.3039108586410316e-06, + "loss": 2.1778, + "step": 15576 + }, + { + "epoch": 3.4, + "learning_rate": 3.3038016167795502e-06, + "loss": 1.6273, + "step": 15577 + }, + { + "epoch": 3.4, + "learning_rate": 3.303692374918069e-06, + "loss": 1.8165, + "step": 15578 + }, + { + "epoch": 3.4, + "learning_rate": 3.3035831330565874e-06, + "loss": 1.3823, + "step": 15579 + }, + { + "epoch": 3.4, + "learning_rate": 3.3034738911951065e-06, + "loss": 0.7255, + "step": 15580 + }, + { + "epoch": 3.4, + "learning_rate": 3.3033646493336247e-06, + "loss": 2.6038, + "step": 15581 + }, + { + "epoch": 3.4, + "learning_rate": 3.3032554074721433e-06, + "loss": 1.1016, + "step": 15582 + }, + { + "epoch": 3.4, + "learning_rate": 3.3031461656106623e-06, + "loss": 0.8506, + "step": 15583 + }, + { + "epoch": 3.4, + "learning_rate": 3.303036923749181e-06, + "loss": 2.9032, + "step": 15584 + }, + { + "epoch": 3.4, + "learning_rate": 3.3029276818877e-06, + "loss": 1.3484, + "step": 15585 + }, + { + "epoch": 3.4, + "learning_rate": 3.302818440026218e-06, + "loss": 2.206, + "step": 15586 + }, + { + "epoch": 3.4, + "learning_rate": 3.3027091981647368e-06, + "loss": 2.6108, + "step": 15587 + }, + { + "epoch": 3.4, + "learning_rate": 3.3025999563032558e-06, + "loss": 1.2508, + "step": 15588 + }, + { + "epoch": 3.4, + "learning_rate": 3.3024907144417744e-06, + "loss": 1.0206, + "step": 15589 + }, + { + "epoch": 3.4, + "learning_rate": 3.3023814725802934e-06, + "loss": 2.8277, + "step": 15590 + }, + { + "epoch": 3.4, + "learning_rate": 3.3022722307188116e-06, + "loss": 0.8552, + "step": 15591 + }, + { + "epoch": 3.4, + "learning_rate": 3.3021629888573302e-06, + "loss": 2.2085, + "step": 15592 + }, + { + "epoch": 3.4, + "learning_rate": 3.3020537469958493e-06, + "loss": 1.3041, + "step": 15593 + }, + { + "epoch": 3.4, + "learning_rate": 3.301944505134368e-06, + "loss": 3.3347, + "step": 15594 + }, + { + "epoch": 3.4, + "learning_rate": 3.301835263272886e-06, + "loss": 0.8209, + "step": 15595 + }, + { + "epoch": 3.4, + "learning_rate": 3.301726021411405e-06, + "loss": 2.9755, + "step": 15596 + }, + { + "epoch": 3.4, + "learning_rate": 3.3016167795499237e-06, + "loss": 2.611, + "step": 15597 + }, + { + "epoch": 3.4, + "learning_rate": 3.3015075376884427e-06, + "loss": 1.6657, + "step": 15598 + }, + { + "epoch": 3.4, + "learning_rate": 3.3013982958269614e-06, + "loss": 1.8848, + "step": 15599 + }, + { + "epoch": 3.4, + "learning_rate": 3.3012890539654795e-06, + "loss": 3.0162, + "step": 15600 + }, + { + "epoch": 3.4, + "learning_rate": 3.3011798121039986e-06, + "loss": 3.0048, + "step": 15601 + }, + { + "epoch": 3.41, + "learning_rate": 3.301070570242517e-06, + "loss": 2.8833, + "step": 15602 + }, + { + "epoch": 3.41, + "learning_rate": 3.300961328381036e-06, + "loss": 2.142, + "step": 15603 + }, + { + "epoch": 3.41, + "learning_rate": 3.300852086519555e-06, + "loss": 1.8586, + "step": 15604 + }, + { + "epoch": 3.41, + "learning_rate": 3.300742844658073e-06, + "loss": 2.1378, + "step": 15605 + }, + { + "epoch": 3.41, + "learning_rate": 3.3006336027965916e-06, + "loss": 1.744, + "step": 15606 + }, + { + "epoch": 3.41, + "learning_rate": 3.3005243609351107e-06, + "loss": 0.6506, + "step": 15607 + }, + { + "epoch": 3.41, + "learning_rate": 3.3004151190736293e-06, + "loss": 4.2465, + "step": 15608 + }, + { + "epoch": 3.41, + "learning_rate": 3.3003058772121483e-06, + "loss": 1.9103, + "step": 15609 + }, + { + "epoch": 3.41, + "learning_rate": 3.3001966353506665e-06, + "loss": 2.0826, + "step": 15610 + }, + { + "epoch": 3.41, + "learning_rate": 3.300087393489185e-06, + "loss": 2.3673, + "step": 15611 + }, + { + "epoch": 3.41, + "learning_rate": 3.299978151627704e-06, + "loss": 1.4571, + "step": 15612 + }, + { + "epoch": 3.41, + "learning_rate": 3.2998689097662228e-06, + "loss": 2.2004, + "step": 15613 + }, + { + "epoch": 3.41, + "learning_rate": 3.299759667904741e-06, + "loss": 2.4665, + "step": 15614 + }, + { + "epoch": 3.41, + "learning_rate": 3.29965042604326e-06, + "loss": 2.1107, + "step": 15615 + }, + { + "epoch": 3.41, + "learning_rate": 3.2995411841817786e-06, + "loss": 2.2635, + "step": 15616 + }, + { + "epoch": 3.41, + "learning_rate": 3.2994319423202976e-06, + "loss": 1.8669, + "step": 15617 + }, + { + "epoch": 3.41, + "learning_rate": 3.2993227004588162e-06, + "loss": 2.2941, + "step": 15618 + }, + { + "epoch": 3.41, + "learning_rate": 3.2992134585973344e-06, + "loss": 1.9716, + "step": 15619 + }, + { + "epoch": 3.41, + "learning_rate": 3.2991042167358535e-06, + "loss": 2.7684, + "step": 15620 + }, + { + "epoch": 3.41, + "learning_rate": 3.298994974874372e-06, + "loss": 1.7354, + "step": 15621 + }, + { + "epoch": 3.41, + "learning_rate": 3.298885733012891e-06, + "loss": 2.268, + "step": 15622 + }, + { + "epoch": 3.41, + "learning_rate": 3.2987764911514097e-06, + "loss": 2.7944, + "step": 15623 + }, + { + "epoch": 3.41, + "learning_rate": 3.298667249289928e-06, + "loss": 2.4664, + "step": 15624 + }, + { + "epoch": 3.41, + "learning_rate": 3.298558007428447e-06, + "loss": 2.4262, + "step": 15625 + }, + { + "epoch": 3.41, + "learning_rate": 3.2984487655669655e-06, + "loss": 1.8314, + "step": 15626 + }, + { + "epoch": 3.41, + "learning_rate": 3.2983395237054846e-06, + "loss": 2.1173, + "step": 15627 + }, + { + "epoch": 3.41, + "learning_rate": 3.2982302818440028e-06, + "loss": 2.3273, + "step": 15628 + }, + { + "epoch": 3.41, + "learning_rate": 3.2981210399825214e-06, + "loss": 2.0015, + "step": 15629 + }, + { + "epoch": 3.41, + "learning_rate": 3.2980117981210404e-06, + "loss": 1.3034, + "step": 15630 + }, + { + "epoch": 3.41, + "learning_rate": 3.297902556259559e-06, + "loss": 0.5212, + "step": 15631 + }, + { + "epoch": 3.41, + "learning_rate": 3.2977933143980776e-06, + "loss": 2.2341, + "step": 15632 + }, + { + "epoch": 3.41, + "learning_rate": 3.2976840725365962e-06, + "loss": 1.6869, + "step": 15633 + }, + { + "epoch": 3.41, + "learning_rate": 3.297574830675115e-06, + "loss": 2.0617, + "step": 15634 + }, + { + "epoch": 3.41, + "learning_rate": 3.2974655888136335e-06, + "loss": 0.6285, + "step": 15635 + }, + { + "epoch": 3.41, + "learning_rate": 3.2973563469521525e-06, + "loss": 2.4718, + "step": 15636 + }, + { + "epoch": 3.41, + "learning_rate": 3.297247105090671e-06, + "loss": 3.6561, + "step": 15637 + }, + { + "epoch": 3.41, + "learning_rate": 3.2971378632291893e-06, + "loss": 1.9046, + "step": 15638 + }, + { + "epoch": 3.41, + "learning_rate": 3.2970286213677083e-06, + "loss": 3.1256, + "step": 15639 + }, + { + "epoch": 3.41, + "learning_rate": 3.296919379506227e-06, + "loss": 1.9154, + "step": 15640 + }, + { + "epoch": 3.41, + "learning_rate": 3.296810137644746e-06, + "loss": 1.6901, + "step": 15641 + }, + { + "epoch": 3.41, + "learning_rate": 3.2967008957832646e-06, + "loss": 2.7333, + "step": 15642 + }, + { + "epoch": 3.41, + "learning_rate": 3.2965916539217828e-06, + "loss": 3.0127, + "step": 15643 + }, + { + "epoch": 3.41, + "learning_rate": 3.296482412060302e-06, + "loss": 1.6545, + "step": 15644 + }, + { + "epoch": 3.41, + "learning_rate": 3.2963731701988204e-06, + "loss": 1.9863, + "step": 15645 + }, + { + "epoch": 3.41, + "learning_rate": 3.2962639283373395e-06, + "loss": 1.7326, + "step": 15646 + }, + { + "epoch": 3.41, + "learning_rate": 3.2961546864758576e-06, + "loss": 1.5103, + "step": 15647 + }, + { + "epoch": 3.42, + "learning_rate": 3.2960454446143763e-06, + "loss": 3.2161, + "step": 15648 + }, + { + "epoch": 3.42, + "learning_rate": 3.2959362027528953e-06, + "loss": 1.8569, + "step": 15649 + }, + { + "epoch": 3.42, + "learning_rate": 3.295826960891414e-06, + "loss": 1.4437, + "step": 15650 + }, + { + "epoch": 3.42, + "learning_rate": 3.295717719029933e-06, + "loss": 2.003, + "step": 15651 + }, + { + "epoch": 3.42, + "learning_rate": 3.295608477168451e-06, + "loss": 1.4872, + "step": 15652 + }, + { + "epoch": 3.42, + "learning_rate": 3.2954992353069697e-06, + "loss": 1.1223, + "step": 15653 + }, + { + "epoch": 3.42, + "learning_rate": 3.2953899934454888e-06, + "loss": 1.434, + "step": 15654 + }, + { + "epoch": 3.42, + "learning_rate": 3.2952807515840074e-06, + "loss": 2.0014, + "step": 15655 + }, + { + "epoch": 3.42, + "learning_rate": 3.295171509722526e-06, + "loss": 1.2703, + "step": 15656 + }, + { + "epoch": 3.42, + "learning_rate": 3.2950622678610446e-06, + "loss": 2.3602, + "step": 15657 + }, + { + "epoch": 3.42, + "learning_rate": 3.2949530259995632e-06, + "loss": 1.968, + "step": 15658 + }, + { + "epoch": 3.42, + "learning_rate": 3.294843784138082e-06, + "loss": 2.4933, + "step": 15659 + }, + { + "epoch": 3.42, + "learning_rate": 3.294734542276601e-06, + "loss": 1.6868, + "step": 15660 + }, + { + "epoch": 3.42, + "learning_rate": 3.294625300415119e-06, + "loss": 2.7131, + "step": 15661 + }, + { + "epoch": 3.42, + "learning_rate": 3.294516058553638e-06, + "loss": 1.8381, + "step": 15662 + }, + { + "epoch": 3.42, + "learning_rate": 3.2944068166921567e-06, + "loss": 1.8975, + "step": 15663 + }, + { + "epoch": 3.42, + "learning_rate": 3.2942975748306753e-06, + "loss": 1.8415, + "step": 15664 + }, + { + "epoch": 3.42, + "learning_rate": 3.2941883329691943e-06, + "loss": 2.0642, + "step": 15665 + }, + { + "epoch": 3.42, + "learning_rate": 3.2940790911077125e-06, + "loss": 1.648, + "step": 15666 + }, + { + "epoch": 3.42, + "learning_rate": 3.293969849246231e-06, + "loss": 1.5727, + "step": 15667 + }, + { + "epoch": 3.42, + "learning_rate": 3.29386060738475e-06, + "loss": 2.3378, + "step": 15668 + }, + { + "epoch": 3.42, + "learning_rate": 3.2937513655232688e-06, + "loss": 2.1997, + "step": 15669 + }, + { + "epoch": 3.42, + "learning_rate": 3.293642123661788e-06, + "loss": 1.4204, + "step": 15670 + }, + { + "epoch": 3.42, + "learning_rate": 3.293532881800306e-06, + "loss": 2.0456, + "step": 15671 + }, + { + "epoch": 3.42, + "learning_rate": 3.2934236399388246e-06, + "loss": 1.8324, + "step": 15672 + }, + { + "epoch": 3.42, + "learning_rate": 3.2933143980773437e-06, + "loss": 1.8449, + "step": 15673 + }, + { + "epoch": 3.42, + "learning_rate": 3.2932051562158623e-06, + "loss": 2.2844, + "step": 15674 + }, + { + "epoch": 3.42, + "learning_rate": 3.2930959143543813e-06, + "loss": 2.0194, + "step": 15675 + }, + { + "epoch": 3.42, + "learning_rate": 3.2929866724928995e-06, + "loss": 2.2269, + "step": 15676 + }, + { + "epoch": 3.42, + "learning_rate": 3.292877430631418e-06, + "loss": 1.3505, + "step": 15677 + }, + { + "epoch": 3.42, + "learning_rate": 3.292768188769937e-06, + "loss": 1.8788, + "step": 15678 + }, + { + "epoch": 3.42, + "learning_rate": 3.2926589469084557e-06, + "loss": 1.9094, + "step": 15679 + }, + { + "epoch": 3.42, + "learning_rate": 3.292549705046974e-06, + "loss": 2.698, + "step": 15680 + }, + { + "epoch": 3.42, + "learning_rate": 3.292440463185493e-06, + "loss": 1.5135, + "step": 15681 + }, + { + "epoch": 3.42, + "learning_rate": 3.2923312213240116e-06, + "loss": 1.7303, + "step": 15682 + }, + { + "epoch": 3.42, + "learning_rate": 3.2922219794625306e-06, + "loss": 2.8147, + "step": 15683 + }, + { + "epoch": 3.42, + "learning_rate": 3.2921127376010492e-06, + "loss": 1.2569, + "step": 15684 + }, + { + "epoch": 3.42, + "learning_rate": 3.2920034957395674e-06, + "loss": 1.4769, + "step": 15685 + }, + { + "epoch": 3.42, + "learning_rate": 3.2918942538780864e-06, + "loss": 2.3941, + "step": 15686 + }, + { + "epoch": 3.42, + "learning_rate": 3.291785012016605e-06, + "loss": 1.5365, + "step": 15687 + }, + { + "epoch": 3.42, + "learning_rate": 3.2916757701551237e-06, + "loss": 1.9774, + "step": 15688 + }, + { + "epoch": 3.42, + "learning_rate": 3.2915665282936427e-06, + "loss": 1.7379, + "step": 15689 + }, + { + "epoch": 3.42, + "learning_rate": 3.291457286432161e-06, + "loss": 2.5356, + "step": 15690 + }, + { + "epoch": 3.42, + "learning_rate": 3.2913480445706795e-06, + "loss": 2.6474, + "step": 15691 + }, + { + "epoch": 3.42, + "learning_rate": 3.2912388027091985e-06, + "loss": 2.826, + "step": 15692 + }, + { + "epoch": 3.42, + "learning_rate": 3.291129560847717e-06, + "loss": 2.0115, + "step": 15693 + }, + { + "epoch": 3.43, + "learning_rate": 3.2910203189862353e-06, + "loss": 2.8486, + "step": 15694 + }, + { + "epoch": 3.43, + "learning_rate": 3.2909110771247544e-06, + "loss": 2.1367, + "step": 15695 + }, + { + "epoch": 3.43, + "learning_rate": 3.290801835263273e-06, + "loss": 2.5274, + "step": 15696 + }, + { + "epoch": 3.43, + "learning_rate": 3.290692593401792e-06, + "loss": 1.8452, + "step": 15697 + }, + { + "epoch": 3.43, + "learning_rate": 3.2905833515403106e-06, + "loss": 2.1332, + "step": 15698 + }, + { + "epoch": 3.43, + "learning_rate": 3.290474109678829e-06, + "loss": 2.8269, + "step": 15699 + }, + { + "epoch": 3.43, + "learning_rate": 3.290364867817348e-06, + "loss": 0.6104, + "step": 15700 + }, + { + "epoch": 3.43, + "learning_rate": 3.2902556259558665e-06, + "loss": 2.0733, + "step": 15701 + }, + { + "epoch": 3.43, + "learning_rate": 3.2901463840943855e-06, + "loss": 2.4249, + "step": 15702 + }, + { + "epoch": 3.43, + "learning_rate": 3.290037142232904e-06, + "loss": 2.0868, + "step": 15703 + }, + { + "epoch": 3.43, + "learning_rate": 3.2899279003714223e-06, + "loss": 1.4057, + "step": 15704 + }, + { + "epoch": 3.43, + "learning_rate": 3.2898186585099413e-06, + "loss": 3.1919, + "step": 15705 + }, + { + "epoch": 3.43, + "learning_rate": 3.28970941664846e-06, + "loss": 2.1623, + "step": 15706 + }, + { + "epoch": 3.43, + "learning_rate": 3.289600174786979e-06, + "loss": 1.9591, + "step": 15707 + }, + { + "epoch": 3.43, + "learning_rate": 3.2894909329254976e-06, + "loss": 0.7441, + "step": 15708 + }, + { + "epoch": 3.43, + "learning_rate": 3.2893816910640158e-06, + "loss": 2.5562, + "step": 15709 + }, + { + "epoch": 3.43, + "learning_rate": 3.289272449202535e-06, + "loss": 1.6775, + "step": 15710 + }, + { + "epoch": 3.43, + "learning_rate": 3.2891632073410534e-06, + "loss": 2.3655, + "step": 15711 + }, + { + "epoch": 3.43, + "learning_rate": 3.289053965479572e-06, + "loss": 2.3345, + "step": 15712 + }, + { + "epoch": 3.43, + "learning_rate": 3.2889447236180906e-06, + "loss": 2.5353, + "step": 15713 + }, + { + "epoch": 3.43, + "learning_rate": 3.2888354817566092e-06, + "loss": 1.6864, + "step": 15714 + }, + { + "epoch": 3.43, + "learning_rate": 3.2887262398951283e-06, + "loss": 1.9309, + "step": 15715 + }, + { + "epoch": 3.43, + "learning_rate": 3.288616998033647e-06, + "loss": 1.7765, + "step": 15716 + }, + { + "epoch": 3.43, + "learning_rate": 3.2885077561721655e-06, + "loss": 1.6998, + "step": 15717 + }, + { + "epoch": 3.43, + "learning_rate": 3.288398514310684e-06, + "loss": 2.7359, + "step": 15718 + }, + { + "epoch": 3.43, + "learning_rate": 3.2882892724492027e-06, + "loss": 1.9472, + "step": 15719 + }, + { + "epoch": 3.43, + "learning_rate": 3.2881800305877213e-06, + "loss": 2.8433, + "step": 15720 + }, + { + "epoch": 3.43, + "learning_rate": 3.2880707887262404e-06, + "loss": 0.9482, + "step": 15721 + }, + { + "epoch": 3.43, + "learning_rate": 3.287961546864759e-06, + "loss": 1.3863, + "step": 15722 + }, + { + "epoch": 3.43, + "learning_rate": 3.287852305003277e-06, + "loss": 0.8597, + "step": 15723 + }, + { + "epoch": 3.43, + "learning_rate": 3.287743063141796e-06, + "loss": 1.3769, + "step": 15724 + }, + { + "epoch": 3.43, + "learning_rate": 3.287633821280315e-06, + "loss": 2.4394, + "step": 15725 + }, + { + "epoch": 3.43, + "learning_rate": 3.287524579418834e-06, + "loss": 1.6306, + "step": 15726 + }, + { + "epoch": 3.43, + "learning_rate": 3.287415337557352e-06, + "loss": 1.0614, + "step": 15727 + }, + { + "epoch": 3.43, + "learning_rate": 3.2873060956958706e-06, + "loss": 2.1632, + "step": 15728 + }, + { + "epoch": 3.43, + "learning_rate": 3.2871968538343897e-06, + "loss": 0.8738, + "step": 15729 + }, + { + "epoch": 3.43, + "learning_rate": 3.2870876119729083e-06, + "loss": 2.1498, + "step": 15730 + }, + { + "epoch": 3.43, + "learning_rate": 3.2869783701114273e-06, + "loss": 0.892, + "step": 15731 + }, + { + "epoch": 3.43, + "learning_rate": 3.2868691282499455e-06, + "loss": 2.0818, + "step": 15732 + }, + { + "epoch": 3.43, + "learning_rate": 3.286759886388464e-06, + "loss": 2.6452, + "step": 15733 + }, + { + "epoch": 3.43, + "learning_rate": 3.286650644526983e-06, + "loss": 2.2365, + "step": 15734 + }, + { + "epoch": 3.43, + "learning_rate": 3.2865414026655018e-06, + "loss": 1.3337, + "step": 15735 + }, + { + "epoch": 3.43, + "learning_rate": 3.286432160804021e-06, + "loss": 2.3487, + "step": 15736 + }, + { + "epoch": 3.43, + "learning_rate": 3.286322918942539e-06, + "loss": 1.5856, + "step": 15737 + }, + { + "epoch": 3.43, + "learning_rate": 3.2862136770810576e-06, + "loss": 1.0342, + "step": 15738 + }, + { + "epoch": 3.43, + "learning_rate": 3.2861044352195766e-06, + "loss": 2.6169, + "step": 15739 + }, + { + "epoch": 3.44, + "learning_rate": 3.2859951933580952e-06, + "loss": 1.5743, + "step": 15740 + }, + { + "epoch": 3.44, + "learning_rate": 3.285885951496614e-06, + "loss": 1.5263, + "step": 15741 + }, + { + "epoch": 3.44, + "learning_rate": 3.2857767096351325e-06, + "loss": 2.8323, + "step": 15742 + }, + { + "epoch": 3.44, + "learning_rate": 3.285667467773651e-06, + "loss": 1.7057, + "step": 15743 + }, + { + "epoch": 3.44, + "learning_rate": 3.2855582259121697e-06, + "loss": 2.542, + "step": 15744 + }, + { + "epoch": 3.44, + "learning_rate": 3.2854489840506887e-06, + "loss": 1.4833, + "step": 15745 + }, + { + "epoch": 3.44, + "learning_rate": 3.285339742189207e-06, + "loss": 1.9457, + "step": 15746 + }, + { + "epoch": 3.44, + "learning_rate": 3.2852305003277255e-06, + "loss": 1.0565, + "step": 15747 + }, + { + "epoch": 3.44, + "learning_rate": 3.2851212584662446e-06, + "loss": 0.9481, + "step": 15748 + }, + { + "epoch": 3.44, + "learning_rate": 3.285012016604763e-06, + "loss": 1.9635, + "step": 15749 + }, + { + "epoch": 3.44, + "learning_rate": 3.284902774743282e-06, + "loss": 1.076, + "step": 15750 + }, + { + "epoch": 3.44, + "learning_rate": 3.2847935328818004e-06, + "loss": 1.6871, + "step": 15751 + }, + { + "epoch": 3.44, + "learning_rate": 3.284684291020319e-06, + "loss": 2.4116, + "step": 15752 + }, + { + "epoch": 3.44, + "learning_rate": 3.284575049158838e-06, + "loss": 2.5928, + "step": 15753 + }, + { + "epoch": 3.44, + "learning_rate": 3.2844658072973566e-06, + "loss": 4.4022, + "step": 15754 + }, + { + "epoch": 3.44, + "learning_rate": 3.2843565654358757e-06, + "loss": 2.4537, + "step": 15755 + }, + { + "epoch": 3.44, + "learning_rate": 3.284247323574394e-06, + "loss": 0.9339, + "step": 15756 + }, + { + "epoch": 3.44, + "learning_rate": 3.2841380817129125e-06, + "loss": 2.5625, + "step": 15757 + }, + { + "epoch": 3.44, + "learning_rate": 3.2840288398514315e-06, + "loss": 1.021, + "step": 15758 + }, + { + "epoch": 3.44, + "learning_rate": 3.28391959798995e-06, + "loss": 2.3695, + "step": 15759 + }, + { + "epoch": 3.44, + "learning_rate": 3.2838103561284683e-06, + "loss": 3.1094, + "step": 15760 + }, + { + "epoch": 3.44, + "learning_rate": 3.2837011142669873e-06, + "loss": 1.5524, + "step": 15761 + }, + { + "epoch": 3.44, + "learning_rate": 3.283591872405506e-06, + "loss": 2.023, + "step": 15762 + }, + { + "epoch": 3.44, + "learning_rate": 3.283482630544025e-06, + "loss": 1.8853, + "step": 15763 + }, + { + "epoch": 3.44, + "learning_rate": 3.2833733886825436e-06, + "loss": 3.0613, + "step": 15764 + }, + { + "epoch": 3.44, + "learning_rate": 3.283264146821062e-06, + "loss": 3.2988, + "step": 15765 + }, + { + "epoch": 3.44, + "learning_rate": 3.283154904959581e-06, + "loss": 1.6791, + "step": 15766 + }, + { + "epoch": 3.44, + "learning_rate": 3.2830456630980994e-06, + "loss": 1.3195, + "step": 15767 + }, + { + "epoch": 3.44, + "learning_rate": 3.2829364212366185e-06, + "loss": 1.6302, + "step": 15768 + }, + { + "epoch": 3.44, + "learning_rate": 3.282827179375137e-06, + "loss": 3.0572, + "step": 15769 + }, + { + "epoch": 3.44, + "learning_rate": 3.2827179375136553e-06, + "loss": 2.251, + "step": 15770 + }, + { + "epoch": 3.44, + "learning_rate": 3.2826086956521743e-06, + "loss": 3.0377, + "step": 15771 + }, + { + "epoch": 3.44, + "learning_rate": 3.282499453790693e-06, + "loss": 1.7481, + "step": 15772 + }, + { + "epoch": 3.44, + "learning_rate": 3.2823902119292115e-06, + "loss": 1.9224, + "step": 15773 + }, + { + "epoch": 3.44, + "learning_rate": 3.2822809700677306e-06, + "loss": 2.9997, + "step": 15774 + }, + { + "epoch": 3.44, + "learning_rate": 3.2821717282062487e-06, + "loss": 2.9651, + "step": 15775 + }, + { + "epoch": 3.44, + "learning_rate": 3.2820624863447674e-06, + "loss": 2.1886, + "step": 15776 + }, + { + "epoch": 3.44, + "learning_rate": 3.2819532444832864e-06, + "loss": 1.4422, + "step": 15777 + }, + { + "epoch": 3.44, + "learning_rate": 3.281844002621805e-06, + "loss": 3.2167, + "step": 15778 + }, + { + "epoch": 3.44, + "learning_rate": 3.281734760760323e-06, + "loss": 2.9743, + "step": 15779 + }, + { + "epoch": 3.44, + "learning_rate": 3.2816255188988422e-06, + "loss": 1.2105, + "step": 15780 + }, + { + "epoch": 3.44, + "learning_rate": 3.281516277037361e-06, + "loss": 2.8178, + "step": 15781 + }, + { + "epoch": 3.44, + "learning_rate": 3.28140703517588e-06, + "loss": 3.0138, + "step": 15782 + }, + { + "epoch": 3.44, + "learning_rate": 3.2812977933143985e-06, + "loss": 1.3949, + "step": 15783 + }, + { + "epoch": 3.44, + "learning_rate": 3.2811885514529167e-06, + "loss": 2.0941, + "step": 15784 + }, + { + "epoch": 3.45, + "learning_rate": 3.2810793095914357e-06, + "loss": 1.5149, + "step": 15785 + }, + { + "epoch": 3.45, + "learning_rate": 3.2809700677299543e-06, + "loss": 2.494, + "step": 15786 + }, + { + "epoch": 3.45, + "learning_rate": 3.2808608258684733e-06, + "loss": 0.6869, + "step": 15787 + }, + { + "epoch": 3.45, + "learning_rate": 3.280751584006992e-06, + "loss": 1.169, + "step": 15788 + }, + { + "epoch": 3.45, + "learning_rate": 3.28064234214551e-06, + "loss": 1.6889, + "step": 15789 + }, + { + "epoch": 3.45, + "learning_rate": 3.280533100284029e-06, + "loss": 1.5726, + "step": 15790 + }, + { + "epoch": 3.45, + "learning_rate": 3.280423858422548e-06, + "loss": 2.4306, + "step": 15791 + }, + { + "epoch": 3.45, + "learning_rate": 3.280314616561067e-06, + "loss": 0.9212, + "step": 15792 + }, + { + "epoch": 3.45, + "learning_rate": 3.280205374699585e-06, + "loss": 2.0721, + "step": 15793 + }, + { + "epoch": 3.45, + "learning_rate": 3.2800961328381036e-06, + "loss": 1.1271, + "step": 15794 + }, + { + "epoch": 3.45, + "learning_rate": 3.2799868909766227e-06, + "loss": 1.6822, + "step": 15795 + }, + { + "epoch": 3.45, + "learning_rate": 3.2798776491151413e-06, + "loss": 1.9338, + "step": 15796 + }, + { + "epoch": 3.45, + "learning_rate": 3.27976840725366e-06, + "loss": 1.4225, + "step": 15797 + }, + { + "epoch": 3.45, + "learning_rate": 3.2796591653921785e-06, + "loss": 2.4837, + "step": 15798 + }, + { + "epoch": 3.45, + "learning_rate": 3.279549923530697e-06, + "loss": 1.015, + "step": 15799 + }, + { + "epoch": 3.45, + "learning_rate": 3.2794406816692157e-06, + "loss": 1.9141, + "step": 15800 + }, + { + "epoch": 3.45, + "learning_rate": 3.2793314398077348e-06, + "loss": 2.9229, + "step": 15801 + }, + { + "epoch": 3.45, + "learning_rate": 3.2792221979462534e-06, + "loss": 1.2296, + "step": 15802 + }, + { + "epoch": 3.45, + "learning_rate": 3.2791129560847715e-06, + "loss": 2.3088, + "step": 15803 + }, + { + "epoch": 3.45, + "learning_rate": 3.2790037142232906e-06, + "loss": 0.9025, + "step": 15804 + }, + { + "epoch": 3.45, + "learning_rate": 3.278894472361809e-06, + "loss": 1.6388, + "step": 15805 + }, + { + "epoch": 3.45, + "learning_rate": 3.2787852305003282e-06, + "loss": 0.6209, + "step": 15806 + }, + { + "epoch": 3.45, + "learning_rate": 3.278675988638847e-06, + "loss": 2.0361, + "step": 15807 + }, + { + "epoch": 3.45, + "learning_rate": 3.278566746777365e-06, + "loss": 1.1598, + "step": 15808 + }, + { + "epoch": 3.45, + "learning_rate": 3.278457504915884e-06, + "loss": 2.4298, + "step": 15809 + }, + { + "epoch": 3.45, + "learning_rate": 3.2783482630544027e-06, + "loss": 1.482, + "step": 15810 + }, + { + "epoch": 3.45, + "learning_rate": 3.2782390211929217e-06, + "loss": 2.9148, + "step": 15811 + }, + { + "epoch": 3.45, + "learning_rate": 3.27812977933144e-06, + "loss": 3.1557, + "step": 15812 + }, + { + "epoch": 3.45, + "learning_rate": 3.2780205374699585e-06, + "loss": 2.1257, + "step": 15813 + }, + { + "epoch": 3.45, + "learning_rate": 3.2779112956084775e-06, + "loss": 2.7996, + "step": 15814 + }, + { + "epoch": 3.45, + "learning_rate": 3.277802053746996e-06, + "loss": 2.6004, + "step": 15815 + }, + { + "epoch": 3.45, + "learning_rate": 3.277692811885515e-06, + "loss": 2.3018, + "step": 15816 + }, + { + "epoch": 3.45, + "learning_rate": 3.2775835700240334e-06, + "loss": 1.4291, + "step": 15817 + }, + { + "epoch": 3.45, + "learning_rate": 3.277474328162552e-06, + "loss": 1.2765, + "step": 15818 + }, + { + "epoch": 3.45, + "learning_rate": 3.277365086301071e-06, + "loss": 1.6963, + "step": 15819 + }, + { + "epoch": 3.45, + "learning_rate": 3.2772558444395896e-06, + "loss": 2.1418, + "step": 15820 + }, + { + "epoch": 3.45, + "learning_rate": 3.2771466025781087e-06, + "loss": 0.9525, + "step": 15821 + }, + { + "epoch": 3.45, + "learning_rate": 3.277037360716627e-06, + "loss": 1.3451, + "step": 15822 + }, + { + "epoch": 3.45, + "learning_rate": 3.2769281188551455e-06, + "loss": 3.4258, + "step": 15823 + }, + { + "epoch": 3.45, + "learning_rate": 3.2768188769936645e-06, + "loss": 2.4256, + "step": 15824 + }, + { + "epoch": 3.45, + "learning_rate": 3.276709635132183e-06, + "loss": 1.0211, + "step": 15825 + }, + { + "epoch": 3.45, + "learning_rate": 3.2766003932707013e-06, + "loss": 1.9713, + "step": 15826 + }, + { + "epoch": 3.45, + "learning_rate": 3.2764911514092203e-06, + "loss": 2.3303, + "step": 15827 + }, + { + "epoch": 3.45, + "learning_rate": 3.276381909547739e-06, + "loss": 1.7952, + "step": 15828 + }, + { + "epoch": 3.45, + "learning_rate": 3.2762726676862576e-06, + "loss": 1.6722, + "step": 15829 + }, + { + "epoch": 3.45, + "learning_rate": 3.2761634258247766e-06, + "loss": 1.8826, + "step": 15830 + }, + { + "epoch": 3.46, + "learning_rate": 3.2760541839632948e-06, + "loss": 3.1552, + "step": 15831 + }, + { + "epoch": 3.46, + "learning_rate": 3.2759449421018134e-06, + "loss": 2.5856, + "step": 15832 + }, + { + "epoch": 3.46, + "learning_rate": 3.2758357002403324e-06, + "loss": 1.0758, + "step": 15833 + }, + { + "epoch": 3.46, + "learning_rate": 3.275726458378851e-06, + "loss": 0.8596, + "step": 15834 + }, + { + "epoch": 3.46, + "learning_rate": 3.27561721651737e-06, + "loss": 1.7095, + "step": 15835 + }, + { + "epoch": 3.46, + "learning_rate": 3.2755079746558883e-06, + "loss": 1.6952, + "step": 15836 + }, + { + "epoch": 3.46, + "learning_rate": 3.275398732794407e-06, + "loss": 2.9298, + "step": 15837 + }, + { + "epoch": 3.46, + "learning_rate": 3.275289490932926e-06, + "loss": 1.4787, + "step": 15838 + }, + { + "epoch": 3.46, + "learning_rate": 3.2751802490714445e-06, + "loss": 1.0682, + "step": 15839 + }, + { + "epoch": 3.46, + "learning_rate": 3.2750710072099627e-06, + "loss": 1.6552, + "step": 15840 + }, + { + "epoch": 3.46, + "learning_rate": 3.2749617653484817e-06, + "loss": 1.2637, + "step": 15841 + }, + { + "epoch": 3.46, + "learning_rate": 3.2748525234870003e-06, + "loss": 2.2261, + "step": 15842 + }, + { + "epoch": 3.46, + "learning_rate": 3.2747432816255194e-06, + "loss": 2.7279, + "step": 15843 + }, + { + "epoch": 3.46, + "learning_rate": 3.274634039764038e-06, + "loss": 1.3889, + "step": 15844 + }, + { + "epoch": 3.46, + "learning_rate": 3.274524797902556e-06, + "loss": 0.6081, + "step": 15845 + }, + { + "epoch": 3.46, + "learning_rate": 3.274415556041075e-06, + "loss": 2.8001, + "step": 15846 + }, + { + "epoch": 3.46, + "learning_rate": 3.274306314179594e-06, + "loss": 1.9806, + "step": 15847 + }, + { + "epoch": 3.46, + "learning_rate": 3.274197072318113e-06, + "loss": 2.6799, + "step": 15848 + }, + { + "epoch": 3.46, + "learning_rate": 3.2740878304566315e-06, + "loss": 2.4964, + "step": 15849 + }, + { + "epoch": 3.46, + "learning_rate": 3.2739785885951497e-06, + "loss": 1.4524, + "step": 15850 + }, + { + "epoch": 3.46, + "learning_rate": 3.2738693467336687e-06, + "loss": 1.8325, + "step": 15851 + }, + { + "epoch": 3.46, + "learning_rate": 3.2737601048721873e-06, + "loss": 2.0498, + "step": 15852 + }, + { + "epoch": 3.46, + "learning_rate": 3.273650863010706e-06, + "loss": 2.3658, + "step": 15853 + }, + { + "epoch": 3.46, + "learning_rate": 3.273541621149225e-06, + "loss": 1.7664, + "step": 15854 + }, + { + "epoch": 3.46, + "learning_rate": 3.273432379287743e-06, + "loss": 1.8042, + "step": 15855 + }, + { + "epoch": 3.46, + "learning_rate": 3.2733231374262617e-06, + "loss": 2.9838, + "step": 15856 + }, + { + "epoch": 3.46, + "learning_rate": 3.2732138955647808e-06, + "loss": 1.4091, + "step": 15857 + }, + { + "epoch": 3.46, + "learning_rate": 3.2731046537032994e-06, + "loss": 1.0904, + "step": 15858 + }, + { + "epoch": 3.46, + "learning_rate": 3.272995411841818e-06, + "loss": 1.808, + "step": 15859 + }, + { + "epoch": 3.46, + "learning_rate": 3.2728861699803366e-06, + "loss": 3.3426, + "step": 15860 + }, + { + "epoch": 3.46, + "learning_rate": 3.2727769281188552e-06, + "loss": 2.5427, + "step": 15861 + }, + { + "epoch": 3.46, + "learning_rate": 3.2726676862573743e-06, + "loss": 1.3602, + "step": 15862 + }, + { + "epoch": 3.46, + "learning_rate": 3.272558444395893e-06, + "loss": 2.1533, + "step": 15863 + }, + { + "epoch": 3.46, + "learning_rate": 3.272449202534411e-06, + "loss": 2.7585, + "step": 15864 + }, + { + "epoch": 3.46, + "learning_rate": 3.27233996067293e-06, + "loss": 2.2854, + "step": 15865 + }, + { + "epoch": 3.46, + "learning_rate": 3.2722307188114487e-06, + "loss": 1.5287, + "step": 15866 + }, + { + "epoch": 3.46, + "learning_rate": 3.2721214769499677e-06, + "loss": 2.608, + "step": 15867 + }, + { + "epoch": 3.46, + "learning_rate": 3.2720122350884863e-06, + "loss": 1.3102, + "step": 15868 + }, + { + "epoch": 3.46, + "learning_rate": 3.2719029932270045e-06, + "loss": 1.7723, + "step": 15869 + }, + { + "epoch": 3.46, + "learning_rate": 3.2717937513655236e-06, + "loss": 1.3913, + "step": 15870 + }, + { + "epoch": 3.46, + "learning_rate": 3.271684509504042e-06, + "loss": 2.269, + "step": 15871 + }, + { + "epoch": 3.46, + "learning_rate": 3.2715752676425612e-06, + "loss": 1.8627, + "step": 15872 + }, + { + "epoch": 3.46, + "learning_rate": 3.2714660257810794e-06, + "loss": 2.7744, + "step": 15873 + }, + { + "epoch": 3.46, + "learning_rate": 3.271356783919598e-06, + "loss": 2.7441, + "step": 15874 + }, + { + "epoch": 3.46, + "learning_rate": 3.271247542058117e-06, + "loss": 2.2313, + "step": 15875 + }, + { + "epoch": 3.46, + "learning_rate": 3.2711383001966357e-06, + "loss": 1.6005, + "step": 15876 + }, + { + "epoch": 3.47, + "learning_rate": 3.2710290583351547e-06, + "loss": 2.446, + "step": 15877 + }, + { + "epoch": 3.47, + "learning_rate": 3.270919816473673e-06, + "loss": 1.8551, + "step": 15878 + }, + { + "epoch": 3.47, + "learning_rate": 3.2708105746121915e-06, + "loss": 2.4001, + "step": 15879 + }, + { + "epoch": 3.47, + "learning_rate": 3.2707013327507105e-06, + "loss": 2.5464, + "step": 15880 + }, + { + "epoch": 3.47, + "learning_rate": 3.270592090889229e-06, + "loss": 1.7489, + "step": 15881 + }, + { + "epoch": 3.47, + "learning_rate": 3.2704828490277477e-06, + "loss": 1.7333, + "step": 15882 + }, + { + "epoch": 3.47, + "learning_rate": 3.2703736071662664e-06, + "loss": 2.3921, + "step": 15883 + }, + { + "epoch": 3.47, + "learning_rate": 3.270264365304785e-06, + "loss": 2.1191, + "step": 15884 + }, + { + "epoch": 3.47, + "learning_rate": 3.2701551234433036e-06, + "loss": 0.9442, + "step": 15885 + }, + { + "epoch": 3.47, + "learning_rate": 3.2700458815818226e-06, + "loss": 2.1947, + "step": 15886 + }, + { + "epoch": 3.47, + "learning_rate": 3.2699366397203412e-06, + "loss": 1.8445, + "step": 15887 + }, + { + "epoch": 3.47, + "learning_rate": 3.2698273978588594e-06, + "loss": 0.8509, + "step": 15888 + }, + { + "epoch": 3.47, + "learning_rate": 3.2697181559973784e-06, + "loss": 1.2798, + "step": 15889 + }, + { + "epoch": 3.47, + "learning_rate": 3.269608914135897e-06, + "loss": 1.7094, + "step": 15890 + }, + { + "epoch": 3.47, + "learning_rate": 3.269499672274416e-06, + "loss": 1.6249, + "step": 15891 + }, + { + "epoch": 3.47, + "learning_rate": 3.2693904304129343e-06, + "loss": 2.2176, + "step": 15892 + }, + { + "epoch": 3.47, + "learning_rate": 3.269281188551453e-06, + "loss": 3.5715, + "step": 15893 + }, + { + "epoch": 3.47, + "learning_rate": 3.269171946689972e-06, + "loss": 0.8544, + "step": 15894 + }, + { + "epoch": 3.47, + "learning_rate": 3.2690627048284905e-06, + "loss": 2.3577, + "step": 15895 + }, + { + "epoch": 3.47, + "learning_rate": 3.2689534629670096e-06, + "loss": 2.0076, + "step": 15896 + }, + { + "epoch": 3.47, + "learning_rate": 3.2688442211055278e-06, + "loss": 1.5386, + "step": 15897 + }, + { + "epoch": 3.47, + "learning_rate": 3.2687349792440464e-06, + "loss": 0.9289, + "step": 15898 + }, + { + "epoch": 3.47, + "learning_rate": 3.2686257373825654e-06, + "loss": 1.9673, + "step": 15899 + }, + { + "epoch": 3.47, + "learning_rate": 3.268516495521084e-06, + "loss": 2.4454, + "step": 15900 + }, + { + "epoch": 3.47, + "learning_rate": 3.268407253659603e-06, + "loss": 1.766, + "step": 15901 + }, + { + "epoch": 3.47, + "learning_rate": 3.2682980117981212e-06, + "loss": 0.2569, + "step": 15902 + }, + { + "epoch": 3.47, + "learning_rate": 3.26818876993664e-06, + "loss": 1.9559, + "step": 15903 + }, + { + "epoch": 3.47, + "learning_rate": 3.268079528075159e-06, + "loss": 2.1487, + "step": 15904 + }, + { + "epoch": 3.47, + "learning_rate": 3.2679702862136775e-06, + "loss": 1.6174, + "step": 15905 + }, + { + "epoch": 3.47, + "learning_rate": 3.2678610443521957e-06, + "loss": 2.9603, + "step": 15906 + }, + { + "epoch": 3.47, + "learning_rate": 3.2677518024907147e-06, + "loss": 2.3954, + "step": 15907 + }, + { + "epoch": 3.47, + "learning_rate": 3.2676425606292333e-06, + "loss": 1.4141, + "step": 15908 + }, + { + "epoch": 3.47, + "learning_rate": 3.2675333187677524e-06, + "loss": 2.5281, + "step": 15909 + }, + { + "epoch": 3.47, + "learning_rate": 3.267424076906271e-06, + "loss": 1.5117, + "step": 15910 + }, + { + "epoch": 3.47, + "learning_rate": 3.267314835044789e-06, + "loss": 2.1709, + "step": 15911 + }, + { + "epoch": 3.47, + "learning_rate": 3.267205593183308e-06, + "loss": 1.7809, + "step": 15912 + }, + { + "epoch": 3.47, + "learning_rate": 3.267096351321827e-06, + "loss": 0.9945, + "step": 15913 + }, + { + "epoch": 3.47, + "learning_rate": 3.2669871094603454e-06, + "loss": 2.4341, + "step": 15914 + }, + { + "epoch": 3.47, + "learning_rate": 3.2668778675988644e-06, + "loss": 2.8461, + "step": 15915 + }, + { + "epoch": 3.47, + "learning_rate": 3.2667686257373826e-06, + "loss": 1.5955, + "step": 15916 + }, + { + "epoch": 3.47, + "learning_rate": 3.2666593838759012e-06, + "loss": 2.0479, + "step": 15917 + }, + { + "epoch": 3.47, + "learning_rate": 3.2665501420144203e-06, + "loss": 1.2569, + "step": 15918 + }, + { + "epoch": 3.47, + "learning_rate": 3.266440900152939e-06, + "loss": 1.7888, + "step": 15919 + }, + { + "epoch": 3.47, + "learning_rate": 3.266331658291458e-06, + "loss": 0.8917, + "step": 15920 + }, + { + "epoch": 3.47, + "learning_rate": 3.266222416429976e-06, + "loss": 2.1956, + "step": 15921 + }, + { + "epoch": 3.47, + "learning_rate": 3.2661131745684947e-06, + "loss": 2.9817, + "step": 15922 + }, + { + "epoch": 3.48, + "learning_rate": 3.2660039327070138e-06, + "loss": 2.3809, + "step": 15923 + }, + { + "epoch": 3.48, + "learning_rate": 3.2658946908455324e-06, + "loss": 1.7117, + "step": 15924 + }, + { + "epoch": 3.48, + "learning_rate": 3.2657854489840506e-06, + "loss": 3.3026, + "step": 15925 + }, + { + "epoch": 3.48, + "learning_rate": 3.2656762071225696e-06, + "loss": 1.6328, + "step": 15926 + }, + { + "epoch": 3.48, + "learning_rate": 3.265566965261088e-06, + "loss": 2.0597, + "step": 15927 + }, + { + "epoch": 3.48, + "learning_rate": 3.2654577233996072e-06, + "loss": 1.9456, + "step": 15928 + }, + { + "epoch": 3.48, + "learning_rate": 3.265348481538126e-06, + "loss": 1.909, + "step": 15929 + }, + { + "epoch": 3.48, + "learning_rate": 3.265239239676644e-06, + "loss": 2.0832, + "step": 15930 + }, + { + "epoch": 3.48, + "learning_rate": 3.265129997815163e-06, + "loss": 3.4919, + "step": 15931 + }, + { + "epoch": 3.48, + "learning_rate": 3.2650207559536817e-06, + "loss": 1.3759, + "step": 15932 + }, + { + "epoch": 3.48, + "learning_rate": 3.2649115140922007e-06, + "loss": 1.9316, + "step": 15933 + }, + { + "epoch": 3.48, + "learning_rate": 3.2648022722307193e-06, + "loss": 2.6855, + "step": 15934 + }, + { + "epoch": 3.48, + "learning_rate": 3.2646930303692375e-06, + "loss": 1.9531, + "step": 15935 + }, + { + "epoch": 3.48, + "learning_rate": 3.2645837885077566e-06, + "loss": 2.7316, + "step": 15936 + }, + { + "epoch": 3.48, + "learning_rate": 3.264474546646275e-06, + "loss": 0.7214, + "step": 15937 + }, + { + "epoch": 3.48, + "learning_rate": 3.2643653047847938e-06, + "loss": 2.085, + "step": 15938 + }, + { + "epoch": 3.48, + "learning_rate": 3.2642560629233124e-06, + "loss": 2.4047, + "step": 15939 + }, + { + "epoch": 3.48, + "learning_rate": 3.264146821061831e-06, + "loss": 1.3529, + "step": 15940 + }, + { + "epoch": 3.48, + "learning_rate": 3.2640375792003496e-06, + "loss": 2.4107, + "step": 15941 + }, + { + "epoch": 3.48, + "learning_rate": 3.2639283373388686e-06, + "loss": 1.6634, + "step": 15942 + }, + { + "epoch": 3.48, + "learning_rate": 3.2638190954773873e-06, + "loss": 1.9299, + "step": 15943 + }, + { + "epoch": 3.48, + "learning_rate": 3.2637098536159054e-06, + "loss": 0.6145, + "step": 15944 + }, + { + "epoch": 3.48, + "learning_rate": 3.2636006117544245e-06, + "loss": 2.0154, + "step": 15945 + }, + { + "epoch": 3.48, + "learning_rate": 3.263491369892943e-06, + "loss": 2.5596, + "step": 15946 + }, + { + "epoch": 3.48, + "learning_rate": 3.263382128031462e-06, + "loss": 1.1527, + "step": 15947 + }, + { + "epoch": 3.48, + "learning_rate": 3.2632728861699807e-06, + "loss": 2.1892, + "step": 15948 + }, + { + "epoch": 3.48, + "learning_rate": 3.263163644308499e-06, + "loss": 1.6146, + "step": 15949 + }, + { + "epoch": 3.48, + "learning_rate": 3.263054402447018e-06, + "loss": 2.4564, + "step": 15950 + }, + { + "epoch": 3.48, + "learning_rate": 3.2629451605855366e-06, + "loss": 1.5798, + "step": 15951 + }, + { + "epoch": 3.48, + "learning_rate": 3.2628359187240556e-06, + "loss": 1.6824, + "step": 15952 + }, + { + "epoch": 3.48, + "learning_rate": 3.262726676862574e-06, + "loss": 0.7874, + "step": 15953 + }, + { + "epoch": 3.48, + "learning_rate": 3.2626174350010924e-06, + "loss": 1.5446, + "step": 15954 + }, + { + "epoch": 3.48, + "learning_rate": 3.2625081931396114e-06, + "loss": 0.9893, + "step": 15955 + }, + { + "epoch": 3.48, + "learning_rate": 3.26239895127813e-06, + "loss": 2.7138, + "step": 15956 + }, + { + "epoch": 3.48, + "learning_rate": 3.262289709416649e-06, + "loss": 1.1917, + "step": 15957 + }, + { + "epoch": 3.48, + "learning_rate": 3.2621804675551673e-06, + "loss": 1.1907, + "step": 15958 + }, + { + "epoch": 3.48, + "learning_rate": 3.262071225693686e-06, + "loss": 1.4003, + "step": 15959 + }, + { + "epoch": 3.48, + "learning_rate": 3.261961983832205e-06, + "loss": 2.2278, + "step": 15960 + }, + { + "epoch": 3.48, + "learning_rate": 3.2618527419707235e-06, + "loss": 1.6573, + "step": 15961 + }, + { + "epoch": 3.48, + "learning_rate": 3.2617435001092426e-06, + "loss": 2.0401, + "step": 15962 + }, + { + "epoch": 3.48, + "learning_rate": 3.2616342582477607e-06, + "loss": 2.603, + "step": 15963 + }, + { + "epoch": 3.48, + "learning_rate": 3.2615250163862794e-06, + "loss": 1.9642, + "step": 15964 + }, + { + "epoch": 3.48, + "learning_rate": 3.2614157745247984e-06, + "loss": 1.8244, + "step": 15965 + }, + { + "epoch": 3.48, + "learning_rate": 3.261306532663317e-06, + "loss": 2.3265, + "step": 15966 + }, + { + "epoch": 3.48, + "learning_rate": 3.2611972908018356e-06, + "loss": 2.5913, + "step": 15967 + }, + { + "epoch": 3.48, + "learning_rate": 3.2610880489403542e-06, + "loss": 3.4899, + "step": 15968 + }, + { + "epoch": 3.49, + "learning_rate": 3.260978807078873e-06, + "loss": 2.1043, + "step": 15969 + }, + { + "epoch": 3.49, + "learning_rate": 3.2608695652173914e-06, + "loss": 0.8148, + "step": 15970 + }, + { + "epoch": 3.49, + "learning_rate": 3.2607603233559105e-06, + "loss": 1.5696, + "step": 15971 + }, + { + "epoch": 3.49, + "learning_rate": 3.2606510814944287e-06, + "loss": 1.124, + "step": 15972 + }, + { + "epoch": 3.49, + "learning_rate": 3.2605418396329473e-06, + "loss": 1.8097, + "step": 15973 + }, + { + "epoch": 3.49, + "learning_rate": 3.2604325977714663e-06, + "loss": 3.0708, + "step": 15974 + }, + { + "epoch": 3.49, + "learning_rate": 3.260323355909985e-06, + "loss": 2.1878, + "step": 15975 + }, + { + "epoch": 3.49, + "learning_rate": 3.260214114048504e-06, + "loss": 1.6578, + "step": 15976 + }, + { + "epoch": 3.49, + "learning_rate": 3.260104872187022e-06, + "loss": 2.4933, + "step": 15977 + }, + { + "epoch": 3.49, + "learning_rate": 3.2599956303255408e-06, + "loss": 2.0071, + "step": 15978 + }, + { + "epoch": 3.49, + "learning_rate": 3.2598863884640598e-06, + "loss": 1.9371, + "step": 15979 + }, + { + "epoch": 3.49, + "learning_rate": 3.2597771466025784e-06, + "loss": 2.1236, + "step": 15980 + }, + { + "epoch": 3.49, + "learning_rate": 3.2596679047410974e-06, + "loss": 2.0159, + "step": 15981 + }, + { + "epoch": 3.49, + "learning_rate": 3.2595586628796156e-06, + "loss": 1.2861, + "step": 15982 + }, + { + "epoch": 3.49, + "learning_rate": 3.2594494210181342e-06, + "loss": 2.8942, + "step": 15983 + }, + { + "epoch": 3.49, + "learning_rate": 3.2593401791566533e-06, + "loss": 1.2446, + "step": 15984 + }, + { + "epoch": 3.49, + "learning_rate": 3.259230937295172e-06, + "loss": 2.0522, + "step": 15985 + }, + { + "epoch": 3.49, + "learning_rate": 3.259121695433691e-06, + "loss": 1.3362, + "step": 15986 + }, + { + "epoch": 3.49, + "learning_rate": 3.259012453572209e-06, + "loss": 1.7887, + "step": 15987 + }, + { + "epoch": 3.49, + "learning_rate": 3.2589032117107277e-06, + "loss": 2.1601, + "step": 15988 + }, + { + "epoch": 3.49, + "learning_rate": 3.2587939698492467e-06, + "loss": 1.6078, + "step": 15989 + }, + { + "epoch": 3.49, + "learning_rate": 3.2586847279877654e-06, + "loss": 2.1713, + "step": 15990 + }, + { + "epoch": 3.49, + "learning_rate": 3.2585754861262835e-06, + "loss": 1.1831, + "step": 15991 + }, + { + "epoch": 3.49, + "learning_rate": 3.2584662442648026e-06, + "loss": 1.1235, + "step": 15992 + }, + { + "epoch": 3.49, + "learning_rate": 3.258357002403321e-06, + "loss": 1.9773, + "step": 15993 + }, + { + "epoch": 3.49, + "learning_rate": 3.25824776054184e-06, + "loss": 1.3018, + "step": 15994 + }, + { + "epoch": 3.49, + "learning_rate": 3.258138518680359e-06, + "loss": 2.4599, + "step": 15995 + }, + { + "epoch": 3.49, + "learning_rate": 3.258029276818877e-06, + "loss": 0.698, + "step": 15996 + }, + { + "epoch": 3.49, + "learning_rate": 3.2579200349573956e-06, + "loss": 2.2006, + "step": 15997 + }, + { + "epoch": 3.49, + "learning_rate": 3.2578107930959147e-06, + "loss": 1.2022, + "step": 15998 + }, + { + "epoch": 3.49, + "learning_rate": 3.2577015512344333e-06, + "loss": 2.3743, + "step": 15999 + }, + { + "epoch": 3.49, + "learning_rate": 3.2575923093729523e-06, + "loss": 1.265, + "step": 16000 + }, + { + "epoch": 3.49, + "learning_rate": 3.2574830675114705e-06, + "loss": 1.7393, + "step": 16001 + }, + { + "epoch": 3.49, + "learning_rate": 3.257373825649989e-06, + "loss": 2.3338, + "step": 16002 + }, + { + "epoch": 3.49, + "learning_rate": 3.257264583788508e-06, + "loss": 1.8244, + "step": 16003 + }, + { + "epoch": 3.49, + "learning_rate": 3.2571553419270268e-06, + "loss": 1.0441, + "step": 16004 + }, + { + "epoch": 3.49, + "learning_rate": 3.257046100065545e-06, + "loss": 2.4309, + "step": 16005 + }, + { + "epoch": 3.49, + "learning_rate": 3.256936858204064e-06, + "loss": 2.1817, + "step": 16006 + }, + { + "epoch": 3.49, + "learning_rate": 3.2568276163425826e-06, + "loss": 1.241, + "step": 16007 + }, + { + "epoch": 3.49, + "learning_rate": 3.2567183744811016e-06, + "loss": 1.6429, + "step": 16008 + }, + { + "epoch": 3.49, + "learning_rate": 3.2566091326196202e-06, + "loss": 3.4724, + "step": 16009 + }, + { + "epoch": 3.49, + "learning_rate": 3.2564998907581384e-06, + "loss": 2.941, + "step": 16010 + }, + { + "epoch": 3.49, + "learning_rate": 3.2563906488966575e-06, + "loss": 2.9386, + "step": 16011 + }, + { + "epoch": 3.49, + "learning_rate": 3.256281407035176e-06, + "loss": 2.0512, + "step": 16012 + }, + { + "epoch": 3.49, + "learning_rate": 3.256172165173695e-06, + "loss": 2.1974, + "step": 16013 + }, + { + "epoch": 3.49, + "learning_rate": 3.2560629233122137e-06, + "loss": 1.4036, + "step": 16014 + }, + { + "epoch": 3.5, + "learning_rate": 3.255953681450732e-06, + "loss": 1.4743, + "step": 16015 + }, + { + "epoch": 3.5, + "learning_rate": 3.255844439589251e-06, + "loss": 1.2357, + "step": 16016 + }, + { + "epoch": 3.5, + "learning_rate": 3.2557351977277695e-06, + "loss": 1.7487, + "step": 16017 + }, + { + "epoch": 3.5, + "learning_rate": 3.2556259558662886e-06, + "loss": 1.4443, + "step": 16018 + }, + { + "epoch": 3.5, + "learning_rate": 3.255516714004807e-06, + "loss": 2.3384, + "step": 16019 + }, + { + "epoch": 3.5, + "learning_rate": 3.2554074721433254e-06, + "loss": 3.0344, + "step": 16020 + }, + { + "epoch": 3.5, + "learning_rate": 3.2552982302818444e-06, + "loss": 1.6274, + "step": 16021 + }, + { + "epoch": 3.5, + "learning_rate": 3.255188988420363e-06, + "loss": 2.7204, + "step": 16022 + }, + { + "epoch": 3.5, + "learning_rate": 3.2550797465588816e-06, + "loss": 1.6781, + "step": 16023 + }, + { + "epoch": 3.5, + "learning_rate": 3.2549705046974002e-06, + "loss": 2.0228, + "step": 16024 + }, + { + "epoch": 3.5, + "learning_rate": 3.254861262835919e-06, + "loss": 1.1207, + "step": 16025 + }, + { + "epoch": 3.5, + "learning_rate": 3.2547520209744375e-06, + "loss": 1.9553, + "step": 16026 + }, + { + "epoch": 3.5, + "learning_rate": 3.2546427791129565e-06, + "loss": 2.5142, + "step": 16027 + }, + { + "epoch": 3.5, + "learning_rate": 3.254533537251475e-06, + "loss": 3.0876, + "step": 16028 + }, + { + "epoch": 3.5, + "learning_rate": 3.2544242953899933e-06, + "loss": 1.24, + "step": 16029 + }, + { + "epoch": 3.5, + "learning_rate": 3.2543150535285123e-06, + "loss": 2.8274, + "step": 16030 + }, + { + "epoch": 3.5, + "learning_rate": 3.254205811667031e-06, + "loss": 0.9481, + "step": 16031 + }, + { + "epoch": 3.5, + "learning_rate": 3.25409656980555e-06, + "loss": 2.0703, + "step": 16032 + }, + { + "epoch": 3.5, + "learning_rate": 3.2539873279440686e-06, + "loss": 1.887, + "step": 16033 + }, + { + "epoch": 3.5, + "learning_rate": 3.2538780860825868e-06, + "loss": 0.4421, + "step": 16034 + }, + { + "epoch": 3.5, + "learning_rate": 3.253768844221106e-06, + "loss": 2.0397, + "step": 16035 + }, + { + "epoch": 3.5, + "learning_rate": 3.2536596023596244e-06, + "loss": 2.1943, + "step": 16036 + }, + { + "epoch": 3.5, + "learning_rate": 3.2535503604981435e-06, + "loss": 1.3141, + "step": 16037 + }, + { + "epoch": 3.5, + "learning_rate": 3.2534411186366616e-06, + "loss": 1.0137, + "step": 16038 + }, + { + "epoch": 3.5, + "learning_rate": 3.2533318767751803e-06, + "loss": 0.6568, + "step": 16039 + }, + { + "epoch": 3.5, + "learning_rate": 3.2532226349136993e-06, + "loss": 1.7032, + "step": 16040 + }, + { + "epoch": 3.5, + "learning_rate": 3.253113393052218e-06, + "loss": 1.8939, + "step": 16041 + }, + { + "epoch": 3.5, + "learning_rate": 3.253004151190737e-06, + "loss": 2.9273, + "step": 16042 + }, + { + "epoch": 3.5, + "learning_rate": 3.252894909329255e-06, + "loss": 2.1134, + "step": 16043 + }, + { + "epoch": 3.5, + "learning_rate": 3.2527856674677737e-06, + "loss": 2.6695, + "step": 16044 + }, + { + "epoch": 3.5, + "learning_rate": 3.2526764256062928e-06, + "loss": 1.5999, + "step": 16045 + }, + { + "epoch": 3.5, + "learning_rate": 3.2525671837448114e-06, + "loss": 0.594, + "step": 16046 + }, + { + "epoch": 3.5, + "learning_rate": 3.25245794188333e-06, + "loss": 2.8691, + "step": 16047 + }, + { + "epoch": 3.5, + "learning_rate": 3.2523487000218486e-06, + "loss": 1.9811, + "step": 16048 + }, + { + "epoch": 3.5, + "learning_rate": 3.2522394581603672e-06, + "loss": 2.5597, + "step": 16049 + }, + { + "epoch": 3.5, + "learning_rate": 3.252130216298886e-06, + "loss": 2.6232, + "step": 16050 + }, + { + "epoch": 3.5, + "learning_rate": 3.252020974437405e-06, + "loss": 1.7174, + "step": 16051 + }, + { + "epoch": 3.5, + "learning_rate": 3.2519117325759235e-06, + "loss": 1.2995, + "step": 16052 + }, + { + "epoch": 3.5, + "learning_rate": 3.251802490714442e-06, + "loss": 1.4452, + "step": 16053 + }, + { + "epoch": 3.5, + "learning_rate": 3.2516932488529607e-06, + "loss": 1.2982, + "step": 16054 + }, + { + "epoch": 3.5, + "learning_rate": 3.2515840069914793e-06, + "loss": 3.4842, + "step": 16055 + }, + { + "epoch": 3.5, + "learning_rate": 3.2514747651299983e-06, + "loss": 1.0551, + "step": 16056 + }, + { + "epoch": 3.5, + "learning_rate": 3.2513655232685165e-06, + "loss": 2.9042, + "step": 16057 + }, + { + "epoch": 3.5, + "learning_rate": 3.251256281407035e-06, + "loss": 0.8552, + "step": 16058 + }, + { + "epoch": 3.5, + "learning_rate": 3.251147039545554e-06, + "loss": 0.2998, + "step": 16059 + }, + { + "epoch": 3.51, + "learning_rate": 3.2510377976840728e-06, + "loss": 1.8419, + "step": 16060 + }, + { + "epoch": 3.51, + "learning_rate": 3.250928555822592e-06, + "loss": 1.6125, + "step": 16061 + }, + { + "epoch": 3.51, + "learning_rate": 3.25081931396111e-06, + "loss": 1.8358, + "step": 16062 + }, + { + "epoch": 3.51, + "learning_rate": 3.2507100720996286e-06, + "loss": 1.5589, + "step": 16063 + }, + { + "epoch": 3.51, + "learning_rate": 3.2506008302381477e-06, + "loss": 4.1894, + "step": 16064 + }, + { + "epoch": 3.51, + "learning_rate": 3.2504915883766663e-06, + "loss": 3.0632, + "step": 16065 + }, + { + "epoch": 3.51, + "learning_rate": 3.2503823465151853e-06, + "loss": 2.7875, + "step": 16066 + }, + { + "epoch": 3.51, + "learning_rate": 3.2502731046537035e-06, + "loss": 1.4451, + "step": 16067 + }, + { + "epoch": 3.51, + "learning_rate": 3.250163862792222e-06, + "loss": 2.1449, + "step": 16068 + }, + { + "epoch": 3.51, + "learning_rate": 3.250054620930741e-06, + "loss": 1.5876, + "step": 16069 + }, + { + "epoch": 3.51, + "learning_rate": 3.2499453790692597e-06, + "loss": 1.555, + "step": 16070 + }, + { + "epoch": 3.51, + "learning_rate": 3.249836137207778e-06, + "loss": 2.0536, + "step": 16071 + }, + { + "epoch": 3.51, + "learning_rate": 3.249726895346297e-06, + "loss": 2.0977, + "step": 16072 + }, + { + "epoch": 3.51, + "learning_rate": 3.2496176534848156e-06, + "loss": 1.9444, + "step": 16073 + }, + { + "epoch": 3.51, + "learning_rate": 3.2495084116233346e-06, + "loss": 2.8385, + "step": 16074 + }, + { + "epoch": 3.51, + "learning_rate": 3.2493991697618532e-06, + "loss": 2.3475, + "step": 16075 + }, + { + "epoch": 3.51, + "learning_rate": 3.2492899279003714e-06, + "loss": 1.4536, + "step": 16076 + }, + { + "epoch": 3.51, + "learning_rate": 3.2491806860388904e-06, + "loss": 3.4787, + "step": 16077 + }, + { + "epoch": 3.51, + "learning_rate": 3.249071444177409e-06, + "loss": 1.9871, + "step": 16078 + }, + { + "epoch": 3.51, + "learning_rate": 3.2489622023159277e-06, + "loss": 1.645, + "step": 16079 + }, + { + "epoch": 3.51, + "learning_rate": 3.2488529604544467e-06, + "loss": 2.6924, + "step": 16080 + }, + { + "epoch": 3.51, + "learning_rate": 3.248743718592965e-06, + "loss": 1.4166, + "step": 16081 + }, + { + "epoch": 3.51, + "learning_rate": 3.2486344767314835e-06, + "loss": 2.085, + "step": 16082 + }, + { + "epoch": 3.51, + "learning_rate": 3.2485252348700025e-06, + "loss": 0.8922, + "step": 16083 + }, + { + "epoch": 3.51, + "learning_rate": 3.248415993008521e-06, + "loss": 1.3142, + "step": 16084 + }, + { + "epoch": 3.51, + "learning_rate": 3.24830675114704e-06, + "loss": 0.4487, + "step": 16085 + }, + { + "epoch": 3.51, + "learning_rate": 3.2481975092855584e-06, + "loss": 1.6249, + "step": 16086 + }, + { + "epoch": 3.51, + "learning_rate": 3.248088267424077e-06, + "loss": 1.0452, + "step": 16087 + }, + { + "epoch": 3.51, + "learning_rate": 3.247979025562596e-06, + "loss": 1.7649, + "step": 16088 + }, + { + "epoch": 3.51, + "learning_rate": 3.2478697837011146e-06, + "loss": 3.0981, + "step": 16089 + }, + { + "epoch": 3.51, + "learning_rate": 3.247760541839633e-06, + "loss": 1.5273, + "step": 16090 + }, + { + "epoch": 3.51, + "learning_rate": 3.247651299978152e-06, + "loss": 1.7126, + "step": 16091 + }, + { + "epoch": 3.51, + "learning_rate": 3.2475420581166705e-06, + "loss": 0.8985, + "step": 16092 + }, + { + "epoch": 3.51, + "learning_rate": 3.2474328162551895e-06, + "loss": 2.8155, + "step": 16093 + }, + { + "epoch": 3.51, + "learning_rate": 3.247323574393708e-06, + "loss": 2.241, + "step": 16094 + }, + { + "epoch": 3.51, + "learning_rate": 3.2472143325322263e-06, + "loss": 2.7516, + "step": 16095 + }, + { + "epoch": 3.51, + "learning_rate": 3.2471050906707453e-06, + "loss": 2.0765, + "step": 16096 + }, + { + "epoch": 3.51, + "learning_rate": 3.246995848809264e-06, + "loss": 2.2951, + "step": 16097 + }, + { + "epoch": 3.51, + "learning_rate": 3.246886606947783e-06, + "loss": 2.2248, + "step": 16098 + }, + { + "epoch": 3.51, + "learning_rate": 3.2467773650863016e-06, + "loss": 1.4599, + "step": 16099 + }, + { + "epoch": 3.51, + "learning_rate": 3.2466681232248198e-06, + "loss": 1.7764, + "step": 16100 + }, + { + "epoch": 3.51, + "learning_rate": 3.246558881363339e-06, + "loss": 1.8611, + "step": 16101 + }, + { + "epoch": 3.51, + "learning_rate": 3.2464496395018574e-06, + "loss": 0.968, + "step": 16102 + }, + { + "epoch": 3.51, + "learning_rate": 3.246340397640376e-06, + "loss": 2.7344, + "step": 16103 + }, + { + "epoch": 3.51, + "learning_rate": 3.2462311557788946e-06, + "loss": 2.6066, + "step": 16104 + }, + { + "epoch": 3.51, + "learning_rate": 3.2461219139174132e-06, + "loss": 1.3785, + "step": 16105 + }, + { + "epoch": 3.52, + "learning_rate": 3.2460126720559323e-06, + "loss": 3.0598, + "step": 16106 + }, + { + "epoch": 3.52, + "learning_rate": 3.245903430194451e-06, + "loss": 1.7286, + "step": 16107 + }, + { + "epoch": 3.52, + "learning_rate": 3.2457941883329695e-06, + "loss": 1.3347, + "step": 16108 + }, + { + "epoch": 3.52, + "learning_rate": 3.245684946471488e-06, + "loss": 2.1207, + "step": 16109 + }, + { + "epoch": 3.52, + "learning_rate": 3.2455757046100067e-06, + "loss": 1.5912, + "step": 16110 + }, + { + "epoch": 3.52, + "learning_rate": 3.2454664627485253e-06, + "loss": 1.5929, + "step": 16111 + }, + { + "epoch": 3.52, + "learning_rate": 3.2453572208870444e-06, + "loss": 3.0431, + "step": 16112 + }, + { + "epoch": 3.52, + "learning_rate": 3.245247979025563e-06, + "loss": 0.4943, + "step": 16113 + }, + { + "epoch": 3.52, + "learning_rate": 3.245138737164081e-06, + "loss": 2.1275, + "step": 16114 + }, + { + "epoch": 3.52, + "learning_rate": 3.2450294953026e-06, + "loss": 3.1457, + "step": 16115 + }, + { + "epoch": 3.52, + "learning_rate": 3.244920253441119e-06, + "loss": 0.8333, + "step": 16116 + }, + { + "epoch": 3.52, + "learning_rate": 3.244811011579638e-06, + "loss": 1.4563, + "step": 16117 + }, + { + "epoch": 3.52, + "learning_rate": 3.2447017697181565e-06, + "loss": 2.4385, + "step": 16118 + }, + { + "epoch": 3.52, + "learning_rate": 3.2445925278566746e-06, + "loss": 2.284, + "step": 16119 + }, + { + "epoch": 3.52, + "learning_rate": 3.2444832859951937e-06, + "loss": 2.0423, + "step": 16120 + }, + { + "epoch": 3.52, + "learning_rate": 3.2443740441337123e-06, + "loss": 1.8037, + "step": 16121 + }, + { + "epoch": 3.52, + "learning_rate": 3.2442648022722313e-06, + "loss": 1.9208, + "step": 16122 + }, + { + "epoch": 3.52, + "learning_rate": 3.2441555604107495e-06, + "loss": 0.7963, + "step": 16123 + }, + { + "epoch": 3.52, + "learning_rate": 3.244046318549268e-06, + "loss": 2.5901, + "step": 16124 + }, + { + "epoch": 3.52, + "learning_rate": 3.243937076687787e-06, + "loss": 1.0719, + "step": 16125 + }, + { + "epoch": 3.52, + "learning_rate": 3.2438278348263058e-06, + "loss": 1.7764, + "step": 16126 + }, + { + "epoch": 3.52, + "learning_rate": 3.243718592964825e-06, + "loss": 2.1734, + "step": 16127 + }, + { + "epoch": 3.52, + "learning_rate": 3.243609351103343e-06, + "loss": 2.2189, + "step": 16128 + }, + { + "epoch": 3.52, + "learning_rate": 3.2435001092418616e-06, + "loss": 2.6934, + "step": 16129 + }, + { + "epoch": 3.52, + "learning_rate": 3.2433908673803806e-06, + "loss": 1.2861, + "step": 16130 + }, + { + "epoch": 3.52, + "learning_rate": 3.2432816255188992e-06, + "loss": 1.2179, + "step": 16131 + }, + { + "epoch": 3.52, + "learning_rate": 3.243172383657418e-06, + "loss": 2.7758, + "step": 16132 + }, + { + "epoch": 3.52, + "learning_rate": 3.2430631417959365e-06, + "loss": 2.5439, + "step": 16133 + }, + { + "epoch": 3.52, + "learning_rate": 3.242953899934455e-06, + "loss": 2.1718, + "step": 16134 + }, + { + "epoch": 3.52, + "learning_rate": 3.2428446580729737e-06, + "loss": 0.8788, + "step": 16135 + }, + { + "epoch": 3.52, + "learning_rate": 3.2427354162114927e-06, + "loss": 1.919, + "step": 16136 + }, + { + "epoch": 3.52, + "learning_rate": 3.242626174350011e-06, + "loss": 1.4237, + "step": 16137 + }, + { + "epoch": 3.52, + "learning_rate": 3.2425169324885295e-06, + "loss": 2.4176, + "step": 16138 + }, + { + "epoch": 3.52, + "learning_rate": 3.2424076906270486e-06, + "loss": 3.0625, + "step": 16139 + }, + { + "epoch": 3.52, + "learning_rate": 3.242298448765567e-06, + "loss": 0.9032, + "step": 16140 + }, + { + "epoch": 3.52, + "learning_rate": 3.242189206904086e-06, + "loss": 3.2385, + "step": 16141 + }, + { + "epoch": 3.52, + "learning_rate": 3.2420799650426044e-06, + "loss": 1.8553, + "step": 16142 + }, + { + "epoch": 3.52, + "learning_rate": 3.241970723181123e-06, + "loss": 1.7371, + "step": 16143 + }, + { + "epoch": 3.52, + "learning_rate": 3.241861481319642e-06, + "loss": 1.2874, + "step": 16144 + }, + { + "epoch": 3.52, + "learning_rate": 3.2417522394581606e-06, + "loss": 0.415, + "step": 16145 + }, + { + "epoch": 3.52, + "learning_rate": 3.2416429975966797e-06, + "loss": 2.0676, + "step": 16146 + }, + { + "epoch": 3.52, + "learning_rate": 3.241533755735198e-06, + "loss": 1.6475, + "step": 16147 + }, + { + "epoch": 3.52, + "learning_rate": 3.2414245138737165e-06, + "loss": 1.3834, + "step": 16148 + }, + { + "epoch": 3.52, + "learning_rate": 3.2413152720122355e-06, + "loss": 2.9254, + "step": 16149 + }, + { + "epoch": 3.52, + "learning_rate": 3.241206030150754e-06, + "loss": 2.8251, + "step": 16150 + }, + { + "epoch": 3.52, + "learning_rate": 3.241096788289273e-06, + "loss": 1.2995, + "step": 16151 + }, + { + "epoch": 3.53, + "learning_rate": 3.2409875464277913e-06, + "loss": 1.569, + "step": 16152 + }, + { + "epoch": 3.53, + "learning_rate": 3.24087830456631e-06, + "loss": 1.8124, + "step": 16153 + }, + { + "epoch": 3.53, + "learning_rate": 3.240769062704829e-06, + "loss": 1.7528, + "step": 16154 + }, + { + "epoch": 3.53, + "learning_rate": 3.2406598208433476e-06, + "loss": 2.0008, + "step": 16155 + }, + { + "epoch": 3.53, + "learning_rate": 3.240550578981866e-06, + "loss": 0.6519, + "step": 16156 + }, + { + "epoch": 3.53, + "learning_rate": 3.240441337120385e-06, + "loss": 2.2578, + "step": 16157 + }, + { + "epoch": 3.53, + "learning_rate": 3.2403320952589034e-06, + "loss": 2.4349, + "step": 16158 + }, + { + "epoch": 3.53, + "learning_rate": 3.2402228533974225e-06, + "loss": 0.3498, + "step": 16159 + }, + { + "epoch": 3.53, + "learning_rate": 3.240113611535941e-06, + "loss": 1.905, + "step": 16160 + }, + { + "epoch": 3.53, + "learning_rate": 3.2400043696744593e-06, + "loss": 2.6745, + "step": 16161 + }, + { + "epoch": 3.53, + "learning_rate": 3.2398951278129783e-06, + "loss": 2.1837, + "step": 16162 + }, + { + "epoch": 3.53, + "learning_rate": 3.239785885951497e-06, + "loss": 3.0309, + "step": 16163 + }, + { + "epoch": 3.53, + "learning_rate": 3.2396766440900155e-06, + "loss": 3.0447, + "step": 16164 + }, + { + "epoch": 3.53, + "learning_rate": 3.2395674022285346e-06, + "loss": 1.6495, + "step": 16165 + }, + { + "epoch": 3.53, + "learning_rate": 3.2394581603670527e-06, + "loss": 1.4289, + "step": 16166 + }, + { + "epoch": 3.53, + "learning_rate": 3.2393489185055714e-06, + "loss": 2.5356, + "step": 16167 + }, + { + "epoch": 3.53, + "learning_rate": 3.2392396766440904e-06, + "loss": 2.9398, + "step": 16168 + }, + { + "epoch": 3.53, + "learning_rate": 3.239130434782609e-06, + "loss": 1.866, + "step": 16169 + }, + { + "epoch": 3.53, + "learning_rate": 3.239021192921127e-06, + "loss": 1.9565, + "step": 16170 + }, + { + "epoch": 3.53, + "learning_rate": 3.2389119510596462e-06, + "loss": 1.4089, + "step": 16171 + }, + { + "epoch": 3.53, + "learning_rate": 3.238802709198165e-06, + "loss": 1.6537, + "step": 16172 + }, + { + "epoch": 3.53, + "learning_rate": 3.238693467336684e-06, + "loss": 0.7655, + "step": 16173 + }, + { + "epoch": 3.53, + "learning_rate": 3.2385842254752025e-06, + "loss": 1.8038, + "step": 16174 + }, + { + "epoch": 3.53, + "learning_rate": 3.2384749836137207e-06, + "loss": 2.2612, + "step": 16175 + }, + { + "epoch": 3.53, + "learning_rate": 3.2383657417522397e-06, + "loss": 1.2512, + "step": 16176 + }, + { + "epoch": 3.53, + "learning_rate": 3.2382564998907583e-06, + "loss": 2.1515, + "step": 16177 + }, + { + "epoch": 3.53, + "learning_rate": 3.2381472580292774e-06, + "loss": 1.4656, + "step": 16178 + }, + { + "epoch": 3.53, + "learning_rate": 3.238038016167796e-06, + "loss": 1.7155, + "step": 16179 + }, + { + "epoch": 3.53, + "learning_rate": 3.237928774306314e-06, + "loss": 2.7465, + "step": 16180 + }, + { + "epoch": 3.53, + "learning_rate": 3.237819532444833e-06, + "loss": 3.0536, + "step": 16181 + }, + { + "epoch": 3.53, + "learning_rate": 3.237710290583352e-06, + "loss": 2.6166, + "step": 16182 + }, + { + "epoch": 3.53, + "learning_rate": 3.237601048721871e-06, + "loss": 1.2522, + "step": 16183 + }, + { + "epoch": 3.53, + "learning_rate": 3.2374918068603894e-06, + "loss": 2.0365, + "step": 16184 + }, + { + "epoch": 3.53, + "learning_rate": 3.2373825649989076e-06, + "loss": 1.3621, + "step": 16185 + }, + { + "epoch": 3.53, + "learning_rate": 3.2372733231374267e-06, + "loss": 2.3416, + "step": 16186 + }, + { + "epoch": 3.53, + "learning_rate": 3.2371640812759453e-06, + "loss": 3.225, + "step": 16187 + }, + { + "epoch": 3.53, + "learning_rate": 3.237054839414464e-06, + "loss": 1.1768, + "step": 16188 + }, + { + "epoch": 3.53, + "learning_rate": 3.2369455975529825e-06, + "loss": 1.2692, + "step": 16189 + }, + { + "epoch": 3.53, + "learning_rate": 3.236836355691501e-06, + "loss": 2.9194, + "step": 16190 + }, + { + "epoch": 3.53, + "learning_rate": 3.2367271138300197e-06, + "loss": 2.1259, + "step": 16191 + }, + { + "epoch": 3.53, + "learning_rate": 3.2366178719685388e-06, + "loss": 2.348, + "step": 16192 + }, + { + "epoch": 3.53, + "learning_rate": 3.2365086301070574e-06, + "loss": 2.6722, + "step": 16193 + }, + { + "epoch": 3.53, + "learning_rate": 3.2363993882455756e-06, + "loss": 0.7184, + "step": 16194 + }, + { + "epoch": 3.53, + "learning_rate": 3.2362901463840946e-06, + "loss": 1.7706, + "step": 16195 + }, + { + "epoch": 3.53, + "learning_rate": 3.236180904522613e-06, + "loss": 1.5278, + "step": 16196 + }, + { + "epoch": 3.53, + "learning_rate": 3.2360716626611322e-06, + "loss": 1.5428, + "step": 16197 + }, + { + "epoch": 3.54, + "learning_rate": 3.235962420799651e-06, + "loss": 0.5929, + "step": 16198 + }, + { + "epoch": 3.54, + "learning_rate": 3.235853178938169e-06, + "loss": 1.0303, + "step": 16199 + }, + { + "epoch": 3.54, + "learning_rate": 3.235743937076688e-06, + "loss": 2.8198, + "step": 16200 + }, + { + "epoch": 3.54, + "learning_rate": 3.2356346952152067e-06, + "loss": 0.98, + "step": 16201 + }, + { + "epoch": 3.54, + "learning_rate": 3.2355254533537257e-06, + "loss": 3.3596, + "step": 16202 + }, + { + "epoch": 3.54, + "learning_rate": 3.235416211492244e-06, + "loss": 2.1223, + "step": 16203 + }, + { + "epoch": 3.54, + "learning_rate": 3.2353069696307625e-06, + "loss": 1.4105, + "step": 16204 + }, + { + "epoch": 3.54, + "learning_rate": 3.2351977277692815e-06, + "loss": 1.8356, + "step": 16205 + }, + { + "epoch": 3.54, + "learning_rate": 3.2350884859078e-06, + "loss": 1.2118, + "step": 16206 + }, + { + "epoch": 3.54, + "learning_rate": 3.234979244046319e-06, + "loss": 2.1685, + "step": 16207 + }, + { + "epoch": 3.54, + "learning_rate": 3.2348700021848374e-06, + "loss": 1.9082, + "step": 16208 + }, + { + "epoch": 3.54, + "learning_rate": 3.234760760323356e-06, + "loss": 1.2444, + "step": 16209 + }, + { + "epoch": 3.54, + "learning_rate": 3.234651518461875e-06, + "loss": 1.7302, + "step": 16210 + }, + { + "epoch": 3.54, + "learning_rate": 3.2345422766003936e-06, + "loss": 1.4533, + "step": 16211 + }, + { + "epoch": 3.54, + "learning_rate": 3.2344330347389127e-06, + "loss": 1.2306, + "step": 16212 + }, + { + "epoch": 3.54, + "learning_rate": 3.234323792877431e-06, + "loss": 0.903, + "step": 16213 + }, + { + "epoch": 3.54, + "learning_rate": 3.2342145510159495e-06, + "loss": 2.88, + "step": 16214 + }, + { + "epoch": 3.54, + "learning_rate": 3.2341053091544685e-06, + "loss": 2.4646, + "step": 16215 + }, + { + "epoch": 3.54, + "learning_rate": 3.233996067292987e-06, + "loss": 0.6529, + "step": 16216 + }, + { + "epoch": 3.54, + "learning_rate": 3.2338868254315057e-06, + "loss": 0.3321, + "step": 16217 + }, + { + "epoch": 3.54, + "learning_rate": 3.2337775835700243e-06, + "loss": 2.0202, + "step": 16218 + }, + { + "epoch": 3.54, + "learning_rate": 3.233668341708543e-06, + "loss": 2.1821, + "step": 16219 + }, + { + "epoch": 3.54, + "learning_rate": 3.2335590998470616e-06, + "loss": 1.8353, + "step": 16220 + }, + { + "epoch": 3.54, + "learning_rate": 3.2334498579855806e-06, + "loss": 3.153, + "step": 16221 + }, + { + "epoch": 3.54, + "learning_rate": 3.2333406161240988e-06, + "loss": 2.4453, + "step": 16222 + }, + { + "epoch": 3.54, + "learning_rate": 3.2332313742626174e-06, + "loss": 2.1454, + "step": 16223 + }, + { + "epoch": 3.54, + "learning_rate": 3.2331221324011364e-06, + "loss": 0.9306, + "step": 16224 + }, + { + "epoch": 3.54, + "learning_rate": 3.233012890539655e-06, + "loss": 1.958, + "step": 16225 + }, + { + "epoch": 3.54, + "learning_rate": 3.232903648678174e-06, + "loss": 2.2984, + "step": 16226 + }, + { + "epoch": 3.54, + "learning_rate": 3.2327944068166923e-06, + "loss": 3.4138, + "step": 16227 + }, + { + "epoch": 3.54, + "learning_rate": 3.232685164955211e-06, + "loss": 1.4207, + "step": 16228 + }, + { + "epoch": 3.54, + "learning_rate": 3.23257592309373e-06, + "loss": 2.193, + "step": 16229 + }, + { + "epoch": 3.54, + "learning_rate": 3.2324666812322485e-06, + "loss": 1.7172, + "step": 16230 + }, + { + "epoch": 3.54, + "learning_rate": 3.2323574393707675e-06, + "loss": 2.0506, + "step": 16231 + }, + { + "epoch": 3.54, + "learning_rate": 3.2322481975092857e-06, + "loss": 3.1226, + "step": 16232 + }, + { + "epoch": 3.54, + "learning_rate": 3.2321389556478043e-06, + "loss": 1.5416, + "step": 16233 + }, + { + "epoch": 3.54, + "learning_rate": 3.2320297137863234e-06, + "loss": 2.0543, + "step": 16234 + }, + { + "epoch": 3.54, + "learning_rate": 3.231920471924842e-06, + "loss": 2.4172, + "step": 16235 + }, + { + "epoch": 3.54, + "learning_rate": 3.23181123006336e-06, + "loss": 3.698, + "step": 16236 + }, + { + "epoch": 3.54, + "learning_rate": 3.231701988201879e-06, + "loss": 1.9297, + "step": 16237 + }, + { + "epoch": 3.54, + "learning_rate": 3.231592746340398e-06, + "loss": 1.3724, + "step": 16238 + }, + { + "epoch": 3.54, + "learning_rate": 3.231483504478917e-06, + "loss": 2.0925, + "step": 16239 + }, + { + "epoch": 3.54, + "learning_rate": 3.2313742626174355e-06, + "loss": 1.264, + "step": 16240 + }, + { + "epoch": 3.54, + "learning_rate": 3.2312650207559537e-06, + "loss": 0.9598, + "step": 16241 + }, + { + "epoch": 3.54, + "learning_rate": 3.2311557788944727e-06, + "loss": 3.1268, + "step": 16242 + }, + { + "epoch": 3.54, + "learning_rate": 3.2310465370329913e-06, + "loss": 2.1375, + "step": 16243 + }, + { + "epoch": 3.55, + "learning_rate": 3.23093729517151e-06, + "loss": 2.2535, + "step": 16244 + }, + { + "epoch": 3.55, + "learning_rate": 3.230828053310029e-06, + "loss": 2.0683, + "step": 16245 + }, + { + "epoch": 3.55, + "learning_rate": 3.230718811448547e-06, + "loss": 2.0639, + "step": 16246 + }, + { + "epoch": 3.55, + "learning_rate": 3.2306095695870657e-06, + "loss": 1.1023, + "step": 16247 + }, + { + "epoch": 3.55, + "learning_rate": 3.2305003277255848e-06, + "loss": 1.6645, + "step": 16248 + }, + { + "epoch": 3.55, + "learning_rate": 3.2303910858641034e-06, + "loss": 2.1115, + "step": 16249 + }, + { + "epoch": 3.55, + "learning_rate": 3.230281844002622e-06, + "loss": 2.1169, + "step": 16250 + }, + { + "epoch": 3.55, + "learning_rate": 3.2301726021411406e-06, + "loss": 2.5054, + "step": 16251 + }, + { + "epoch": 3.55, + "learning_rate": 3.2300633602796592e-06, + "loss": 3.1345, + "step": 16252 + }, + { + "epoch": 3.55, + "learning_rate": 3.2299541184181783e-06, + "loss": 2.5025, + "step": 16253 + }, + { + "epoch": 3.55, + "learning_rate": 3.229844876556697e-06, + "loss": 3.0117, + "step": 16254 + }, + { + "epoch": 3.55, + "learning_rate": 3.229735634695215e-06, + "loss": 2.3863, + "step": 16255 + }, + { + "epoch": 3.55, + "learning_rate": 3.229626392833734e-06, + "loss": 1.9775, + "step": 16256 + }, + { + "epoch": 3.55, + "learning_rate": 3.2295171509722527e-06, + "loss": 1.5629, + "step": 16257 + }, + { + "epoch": 3.55, + "learning_rate": 3.2294079091107717e-06, + "loss": 2.4559, + "step": 16258 + }, + { + "epoch": 3.55, + "learning_rate": 3.2292986672492903e-06, + "loss": 2.5162, + "step": 16259 + }, + { + "epoch": 3.55, + "learning_rate": 3.2291894253878085e-06, + "loss": 2.4547, + "step": 16260 + }, + { + "epoch": 3.55, + "learning_rate": 3.2290801835263276e-06, + "loss": 1.95, + "step": 16261 + }, + { + "epoch": 3.55, + "learning_rate": 3.228970941664846e-06, + "loss": 1.3902, + "step": 16262 + }, + { + "epoch": 3.55, + "learning_rate": 3.2288616998033652e-06, + "loss": 2.1354, + "step": 16263 + }, + { + "epoch": 3.55, + "learning_rate": 3.228752457941884e-06, + "loss": 2.0648, + "step": 16264 + }, + { + "epoch": 3.55, + "learning_rate": 3.228643216080402e-06, + "loss": 2.3936, + "step": 16265 + }, + { + "epoch": 3.55, + "learning_rate": 3.228533974218921e-06, + "loss": 1.8194, + "step": 16266 + }, + { + "epoch": 3.55, + "learning_rate": 3.2284247323574397e-06, + "loss": 2.3518, + "step": 16267 + }, + { + "epoch": 3.55, + "learning_rate": 3.2283154904959587e-06, + "loss": 1.6716, + "step": 16268 + }, + { + "epoch": 3.55, + "learning_rate": 3.228206248634477e-06, + "loss": 1.1362, + "step": 16269 + }, + { + "epoch": 3.55, + "learning_rate": 3.2280970067729955e-06, + "loss": 2.8257, + "step": 16270 + }, + { + "epoch": 3.55, + "learning_rate": 3.2279877649115145e-06, + "loss": 1.718, + "step": 16271 + }, + { + "epoch": 3.55, + "learning_rate": 3.227878523050033e-06, + "loss": 1.5478, + "step": 16272 + }, + { + "epoch": 3.55, + "learning_rate": 3.2277692811885517e-06, + "loss": 2.522, + "step": 16273 + }, + { + "epoch": 3.55, + "learning_rate": 3.2276600393270704e-06, + "loss": 1.2497, + "step": 16274 + }, + { + "epoch": 3.55, + "learning_rate": 3.227550797465589e-06, + "loss": 1.7098, + "step": 16275 + }, + { + "epoch": 3.55, + "learning_rate": 3.2274415556041076e-06, + "loss": 2.0961, + "step": 16276 + }, + { + "epoch": 3.55, + "learning_rate": 3.2273323137426266e-06, + "loss": 3.2634, + "step": 16277 + }, + { + "epoch": 3.55, + "learning_rate": 3.2272230718811452e-06, + "loss": 2.348, + "step": 16278 + }, + { + "epoch": 3.55, + "learning_rate": 3.2271138300196634e-06, + "loss": 1.7678, + "step": 16279 + }, + { + "epoch": 3.55, + "learning_rate": 3.2270045881581824e-06, + "loss": 2.6536, + "step": 16280 + }, + { + "epoch": 3.55, + "learning_rate": 3.226895346296701e-06, + "loss": 1.3643, + "step": 16281 + }, + { + "epoch": 3.55, + "learning_rate": 3.22678610443522e-06, + "loss": 2.5734, + "step": 16282 + }, + { + "epoch": 3.55, + "learning_rate": 3.2266768625737383e-06, + "loss": 3.0671, + "step": 16283 + }, + { + "epoch": 3.55, + "learning_rate": 3.226567620712257e-06, + "loss": 0.9193, + "step": 16284 + }, + { + "epoch": 3.55, + "learning_rate": 3.226458378850776e-06, + "loss": 2.6944, + "step": 16285 + }, + { + "epoch": 3.55, + "learning_rate": 3.2263491369892945e-06, + "loss": 2.3476, + "step": 16286 + }, + { + "epoch": 3.55, + "learning_rate": 3.2262398951278136e-06, + "loss": 0.933, + "step": 16287 + }, + { + "epoch": 3.55, + "learning_rate": 3.2261306532663318e-06, + "loss": 3.647, + "step": 16288 + }, + { + "epoch": 3.55, + "learning_rate": 3.2260214114048504e-06, + "loss": 0.834, + "step": 16289 + }, + { + "epoch": 3.56, + "learning_rate": 3.2259121695433694e-06, + "loss": 1.8583, + "step": 16290 + }, + { + "epoch": 3.56, + "learning_rate": 3.225802927681888e-06, + "loss": 1.5413, + "step": 16291 + }, + { + "epoch": 3.56, + "learning_rate": 3.225693685820407e-06, + "loss": 1.0665, + "step": 16292 + }, + { + "epoch": 3.56, + "learning_rate": 3.2255844439589252e-06, + "loss": 2.0066, + "step": 16293 + }, + { + "epoch": 3.56, + "learning_rate": 3.225475202097444e-06, + "loss": 1.1904, + "step": 16294 + }, + { + "epoch": 3.56, + "learning_rate": 3.225365960235963e-06, + "loss": 1.5536, + "step": 16295 + }, + { + "epoch": 3.56, + "learning_rate": 3.2252567183744815e-06, + "loss": 1.8414, + "step": 16296 + }, + { + "epoch": 3.56, + "learning_rate": 3.225147476513e-06, + "loss": 2.1428, + "step": 16297 + }, + { + "epoch": 3.56, + "learning_rate": 3.2250382346515187e-06, + "loss": 2.1767, + "step": 16298 + }, + { + "epoch": 3.56, + "learning_rate": 3.2249289927900373e-06, + "loss": 1.0355, + "step": 16299 + }, + { + "epoch": 3.56, + "learning_rate": 3.224819750928556e-06, + "loss": 2.7198, + "step": 16300 + }, + { + "epoch": 3.56, + "learning_rate": 3.224710509067075e-06, + "loss": 1.7491, + "step": 16301 + }, + { + "epoch": 3.56, + "learning_rate": 3.224601267205593e-06, + "loss": 1.2097, + "step": 16302 + }, + { + "epoch": 3.56, + "learning_rate": 3.224492025344112e-06, + "loss": 2.3487, + "step": 16303 + }, + { + "epoch": 3.56, + "learning_rate": 3.224382783482631e-06, + "loss": 0.5952, + "step": 16304 + }, + { + "epoch": 3.56, + "learning_rate": 3.2242735416211494e-06, + "loss": 1.7309, + "step": 16305 + }, + { + "epoch": 3.56, + "learning_rate": 3.2241642997596685e-06, + "loss": 2.9369, + "step": 16306 + }, + { + "epoch": 3.56, + "learning_rate": 3.2240550578981866e-06, + "loss": 1.7907, + "step": 16307 + }, + { + "epoch": 3.56, + "learning_rate": 3.2239458160367052e-06, + "loss": 2.6227, + "step": 16308 + }, + { + "epoch": 3.56, + "learning_rate": 3.2238365741752243e-06, + "loss": 2.0216, + "step": 16309 + }, + { + "epoch": 3.56, + "learning_rate": 3.223727332313743e-06, + "loss": 2.0937, + "step": 16310 + }, + { + "epoch": 3.56, + "learning_rate": 3.223618090452262e-06, + "loss": 0.9983, + "step": 16311 + }, + { + "epoch": 3.56, + "learning_rate": 3.22350884859078e-06, + "loss": 2.9603, + "step": 16312 + }, + { + "epoch": 3.56, + "learning_rate": 3.2233996067292987e-06, + "loss": 2.5146, + "step": 16313 + }, + { + "epoch": 3.56, + "learning_rate": 3.2232903648678178e-06, + "loss": 1.0606, + "step": 16314 + }, + { + "epoch": 3.56, + "learning_rate": 3.2231811230063364e-06, + "loss": 1.0572, + "step": 16315 + }, + { + "epoch": 3.56, + "learning_rate": 3.2230718811448546e-06, + "loss": 2.5527, + "step": 16316 + }, + { + "epoch": 3.56, + "learning_rate": 3.2229626392833736e-06, + "loss": 1.7725, + "step": 16317 + }, + { + "epoch": 3.56, + "learning_rate": 3.222853397421892e-06, + "loss": 1.2538, + "step": 16318 + }, + { + "epoch": 3.56, + "learning_rate": 3.2227441555604112e-06, + "loss": 2.4162, + "step": 16319 + }, + { + "epoch": 3.56, + "learning_rate": 3.22263491369893e-06, + "loss": 2.701, + "step": 16320 + }, + { + "epoch": 3.56, + "learning_rate": 3.222525671837448e-06, + "loss": 0.9609, + "step": 16321 + }, + { + "epoch": 3.56, + "learning_rate": 3.222416429975967e-06, + "loss": 1.9796, + "step": 16322 + }, + { + "epoch": 3.56, + "learning_rate": 3.2223071881144857e-06, + "loss": 0.8574, + "step": 16323 + }, + { + "epoch": 3.56, + "learning_rate": 3.2221979462530047e-06, + "loss": 1.623, + "step": 16324 + }, + { + "epoch": 3.56, + "learning_rate": 3.2220887043915233e-06, + "loss": 1.4714, + "step": 16325 + }, + { + "epoch": 3.56, + "learning_rate": 3.2219794625300415e-06, + "loss": 0.842, + "step": 16326 + }, + { + "epoch": 3.56, + "learning_rate": 3.2218702206685606e-06, + "loss": 2.0842, + "step": 16327 + }, + { + "epoch": 3.56, + "learning_rate": 3.221760978807079e-06, + "loss": 1.0758, + "step": 16328 + }, + { + "epoch": 3.56, + "learning_rate": 3.2216517369455978e-06, + "loss": 2.5538, + "step": 16329 + }, + { + "epoch": 3.56, + "learning_rate": 3.221542495084117e-06, + "loss": 1.2654, + "step": 16330 + }, + { + "epoch": 3.56, + "learning_rate": 3.221433253222635e-06, + "loss": 0.7651, + "step": 16331 + }, + { + "epoch": 3.56, + "learning_rate": 3.2213240113611536e-06, + "loss": 3.2418, + "step": 16332 + }, + { + "epoch": 3.56, + "learning_rate": 3.2212147694996726e-06, + "loss": 1.545, + "step": 16333 + }, + { + "epoch": 3.56, + "learning_rate": 3.2211055276381913e-06, + "loss": 2.3976, + "step": 16334 + }, + { + "epoch": 3.57, + "learning_rate": 3.2209962857767094e-06, + "loss": 1.2984, + "step": 16335 + }, + { + "epoch": 3.57, + "learning_rate": 3.2208870439152285e-06, + "loss": 0.6216, + "step": 16336 + }, + { + "epoch": 3.57, + "learning_rate": 3.220777802053747e-06, + "loss": 1.2197, + "step": 16337 + }, + { + "epoch": 3.57, + "learning_rate": 3.220668560192266e-06, + "loss": 1.5477, + "step": 16338 + }, + { + "epoch": 3.57, + "learning_rate": 3.2205593183307847e-06, + "loss": 1.7006, + "step": 16339 + }, + { + "epoch": 3.57, + "learning_rate": 3.220450076469303e-06, + "loss": 1.6023, + "step": 16340 + }, + { + "epoch": 3.57, + "learning_rate": 3.220340834607822e-06, + "loss": 1.3257, + "step": 16341 + }, + { + "epoch": 3.57, + "learning_rate": 3.2202315927463406e-06, + "loss": 1.3996, + "step": 16342 + }, + { + "epoch": 3.57, + "learning_rate": 3.2201223508848596e-06, + "loss": 2.9285, + "step": 16343 + }, + { + "epoch": 3.57, + "learning_rate": 3.220013109023378e-06, + "loss": 1.5204, + "step": 16344 + }, + { + "epoch": 3.57, + "learning_rate": 3.2199038671618964e-06, + "loss": 3.5009, + "step": 16345 + }, + { + "epoch": 3.57, + "learning_rate": 3.2197946253004154e-06, + "loss": 1.6442, + "step": 16346 + }, + { + "epoch": 3.57, + "learning_rate": 3.219685383438934e-06, + "loss": 1.2492, + "step": 16347 + }, + { + "epoch": 3.57, + "learning_rate": 3.219576141577453e-06, + "loss": 2.2659, + "step": 16348 + }, + { + "epoch": 3.57, + "learning_rate": 3.2194668997159713e-06, + "loss": 0.7542, + "step": 16349 + }, + { + "epoch": 3.57, + "learning_rate": 3.21935765785449e-06, + "loss": 2.361, + "step": 16350 + }, + { + "epoch": 3.57, + "learning_rate": 3.219248415993009e-06, + "loss": 1.3633, + "step": 16351 + }, + { + "epoch": 3.57, + "learning_rate": 3.2191391741315275e-06, + "loss": 1.8053, + "step": 16352 + }, + { + "epoch": 3.57, + "learning_rate": 3.219029932270046e-06, + "loss": 2.6375, + "step": 16353 + }, + { + "epoch": 3.57, + "learning_rate": 3.2189206904085647e-06, + "loss": 1.9818, + "step": 16354 + }, + { + "epoch": 3.57, + "learning_rate": 3.2188114485470834e-06, + "loss": 1.976, + "step": 16355 + }, + { + "epoch": 3.57, + "learning_rate": 3.2187022066856024e-06, + "loss": 2.3399, + "step": 16356 + }, + { + "epoch": 3.57, + "learning_rate": 3.218592964824121e-06, + "loss": 0.932, + "step": 16357 + }, + { + "epoch": 3.57, + "learning_rate": 3.2184837229626396e-06, + "loss": 2.5933, + "step": 16358 + }, + { + "epoch": 3.57, + "learning_rate": 3.2183744811011582e-06, + "loss": 2.6166, + "step": 16359 + }, + { + "epoch": 3.57, + "learning_rate": 3.218265239239677e-06, + "loss": 0.6141, + "step": 16360 + }, + { + "epoch": 3.57, + "learning_rate": 3.2181559973781954e-06, + "loss": 2.9513, + "step": 16361 + }, + { + "epoch": 3.57, + "learning_rate": 3.2180467555167145e-06, + "loss": 1.8473, + "step": 16362 + }, + { + "epoch": 3.57, + "learning_rate": 3.217937513655233e-06, + "loss": 1.8253, + "step": 16363 + }, + { + "epoch": 3.57, + "learning_rate": 3.2178282717937513e-06, + "loss": 1.703, + "step": 16364 + }, + { + "epoch": 3.57, + "learning_rate": 3.2177190299322703e-06, + "loss": 1.6176, + "step": 16365 + }, + { + "epoch": 3.57, + "learning_rate": 3.217609788070789e-06, + "loss": 1.8262, + "step": 16366 + }, + { + "epoch": 3.57, + "learning_rate": 3.217500546209308e-06, + "loss": 1.925, + "step": 16367 + }, + { + "epoch": 3.57, + "learning_rate": 3.217391304347826e-06, + "loss": 0.8528, + "step": 16368 + }, + { + "epoch": 3.57, + "learning_rate": 3.2172820624863448e-06, + "loss": 1.3165, + "step": 16369 + }, + { + "epoch": 3.57, + "learning_rate": 3.217172820624864e-06, + "loss": 2.5351, + "step": 16370 + }, + { + "epoch": 3.57, + "learning_rate": 3.2170635787633824e-06, + "loss": 1.3414, + "step": 16371 + }, + { + "epoch": 3.57, + "learning_rate": 3.2169543369019014e-06, + "loss": 2.032, + "step": 16372 + }, + { + "epoch": 3.57, + "learning_rate": 3.2168450950404196e-06, + "loss": 1.7156, + "step": 16373 + }, + { + "epoch": 3.57, + "learning_rate": 3.2167358531789382e-06, + "loss": 0.8237, + "step": 16374 + }, + { + "epoch": 3.57, + "learning_rate": 3.2166266113174573e-06, + "loss": 1.6696, + "step": 16375 + }, + { + "epoch": 3.57, + "learning_rate": 3.216517369455976e-06, + "loss": 1.4218, + "step": 16376 + }, + { + "epoch": 3.57, + "learning_rate": 3.216408127594495e-06, + "loss": 1.7519, + "step": 16377 + }, + { + "epoch": 3.57, + "learning_rate": 3.216298885733013e-06, + "loss": 2.4334, + "step": 16378 + }, + { + "epoch": 3.57, + "learning_rate": 3.2161896438715317e-06, + "loss": 1.7952, + "step": 16379 + }, + { + "epoch": 3.57, + "learning_rate": 3.2160804020100507e-06, + "loss": 3.3242, + "step": 16380 + }, + { + "epoch": 3.58, + "learning_rate": 3.2159711601485694e-06, + "loss": 1.9511, + "step": 16381 + }, + { + "epoch": 3.58, + "learning_rate": 3.2158619182870875e-06, + "loss": 1.8213, + "step": 16382 + }, + { + "epoch": 3.58, + "learning_rate": 3.2157526764256066e-06, + "loss": 1.3079, + "step": 16383 + }, + { + "epoch": 3.58, + "learning_rate": 3.215643434564125e-06, + "loss": 1.5769, + "step": 16384 + }, + { + "epoch": 3.58, + "learning_rate": 3.215534192702644e-06, + "loss": 1.1428, + "step": 16385 + }, + { + "epoch": 3.58, + "learning_rate": 3.215424950841163e-06, + "loss": 1.6368, + "step": 16386 + }, + { + "epoch": 3.58, + "learning_rate": 3.215315708979681e-06, + "loss": 1.6621, + "step": 16387 + }, + { + "epoch": 3.58, + "learning_rate": 3.2152064671181996e-06, + "loss": 1.474, + "step": 16388 + }, + { + "epoch": 3.58, + "learning_rate": 3.2150972252567187e-06, + "loss": 1.8834, + "step": 16389 + }, + { + "epoch": 3.58, + "learning_rate": 3.2149879833952373e-06, + "loss": 1.1207, + "step": 16390 + }, + { + "epoch": 3.58, + "learning_rate": 3.2148787415337563e-06, + "loss": 2.2485, + "step": 16391 + }, + { + "epoch": 3.58, + "learning_rate": 3.2147694996722745e-06, + "loss": 2.1266, + "step": 16392 + }, + { + "epoch": 3.58, + "learning_rate": 3.214660257810793e-06, + "loss": 0.9586, + "step": 16393 + }, + { + "epoch": 3.58, + "learning_rate": 3.214551015949312e-06, + "loss": 1.4381, + "step": 16394 + }, + { + "epoch": 3.58, + "learning_rate": 3.2144417740878308e-06, + "loss": 1.7974, + "step": 16395 + }, + { + "epoch": 3.58, + "learning_rate": 3.21433253222635e-06, + "loss": 2.5441, + "step": 16396 + }, + { + "epoch": 3.58, + "learning_rate": 3.214223290364868e-06, + "loss": 2.0959, + "step": 16397 + }, + { + "epoch": 3.58, + "learning_rate": 3.2141140485033866e-06, + "loss": 3.1568, + "step": 16398 + }, + { + "epoch": 3.58, + "learning_rate": 3.2140048066419056e-06, + "loss": 1.9362, + "step": 16399 + }, + { + "epoch": 3.58, + "learning_rate": 3.2138955647804242e-06, + "loss": 2.9721, + "step": 16400 + }, + { + "epoch": 3.58, + "learning_rate": 3.2137863229189424e-06, + "loss": 1.903, + "step": 16401 + }, + { + "epoch": 3.58, + "learning_rate": 3.2136770810574615e-06, + "loss": 1.3582, + "step": 16402 + }, + { + "epoch": 3.58, + "learning_rate": 3.21356783919598e-06, + "loss": 0.6695, + "step": 16403 + }, + { + "epoch": 3.58, + "learning_rate": 3.213458597334499e-06, + "loss": 2.3446, + "step": 16404 + }, + { + "epoch": 3.58, + "learning_rate": 3.2133493554730177e-06, + "loss": 2.5887, + "step": 16405 + }, + { + "epoch": 3.58, + "learning_rate": 3.213240113611536e-06, + "loss": 1.7188, + "step": 16406 + }, + { + "epoch": 3.58, + "learning_rate": 3.213130871750055e-06, + "loss": 1.8452, + "step": 16407 + }, + { + "epoch": 3.58, + "learning_rate": 3.2130216298885735e-06, + "loss": 4.4994, + "step": 16408 + }, + { + "epoch": 3.58, + "learning_rate": 3.2129123880270926e-06, + "loss": 1.9276, + "step": 16409 + }, + { + "epoch": 3.58, + "learning_rate": 3.212803146165611e-06, + "loss": 1.2671, + "step": 16410 + }, + { + "epoch": 3.58, + "learning_rate": 3.2126939043041294e-06, + "loss": 1.2918, + "step": 16411 + }, + { + "epoch": 3.58, + "learning_rate": 3.2125846624426484e-06, + "loss": 1.9559, + "step": 16412 + }, + { + "epoch": 3.58, + "learning_rate": 3.212475420581167e-06, + "loss": 1.6783, + "step": 16413 + }, + { + "epoch": 3.58, + "learning_rate": 3.2123661787196856e-06, + "loss": 2.526, + "step": 16414 + }, + { + "epoch": 3.58, + "learning_rate": 3.2122569368582042e-06, + "loss": 1.6486, + "step": 16415 + }, + { + "epoch": 3.58, + "learning_rate": 3.212147694996723e-06, + "loss": 3.0323, + "step": 16416 + }, + { + "epoch": 3.58, + "learning_rate": 3.2120384531352415e-06, + "loss": 0.6445, + "step": 16417 + }, + { + "epoch": 3.58, + "learning_rate": 3.2119292112737605e-06, + "loss": 1.753, + "step": 16418 + }, + { + "epoch": 3.58, + "learning_rate": 3.211819969412279e-06, + "loss": 1.7235, + "step": 16419 + }, + { + "epoch": 3.58, + "learning_rate": 3.2117107275507973e-06, + "loss": 1.0337, + "step": 16420 + }, + { + "epoch": 3.58, + "learning_rate": 3.2116014856893163e-06, + "loss": 1.5027, + "step": 16421 + }, + { + "epoch": 3.58, + "learning_rate": 3.211492243827835e-06, + "loss": 2.7062, + "step": 16422 + }, + { + "epoch": 3.58, + "learning_rate": 3.211383001966354e-06, + "loss": 1.079, + "step": 16423 + }, + { + "epoch": 3.58, + "learning_rate": 3.2112737601048726e-06, + "loss": 1.8079, + "step": 16424 + }, + { + "epoch": 3.58, + "learning_rate": 3.2111645182433908e-06, + "loss": 1.0357, + "step": 16425 + }, + { + "epoch": 3.58, + "learning_rate": 3.21105527638191e-06, + "loss": 2.0028, + "step": 16426 + }, + { + "epoch": 3.59, + "learning_rate": 3.2109460345204284e-06, + "loss": 1.7947, + "step": 16427 + }, + { + "epoch": 3.59, + "learning_rate": 3.2108367926589475e-06, + "loss": 0.9566, + "step": 16428 + }, + { + "epoch": 3.59, + "learning_rate": 3.210727550797466e-06, + "loss": 2.0278, + "step": 16429 + }, + { + "epoch": 3.59, + "learning_rate": 3.2106183089359843e-06, + "loss": 2.5791, + "step": 16430 + }, + { + "epoch": 3.59, + "learning_rate": 3.2105090670745033e-06, + "loss": 2.1253, + "step": 16431 + }, + { + "epoch": 3.59, + "learning_rate": 3.210399825213022e-06, + "loss": 1.453, + "step": 16432 + }, + { + "epoch": 3.59, + "learning_rate": 3.210290583351541e-06, + "loss": 1.0538, + "step": 16433 + }, + { + "epoch": 3.59, + "learning_rate": 3.210181341490059e-06, + "loss": 0.9165, + "step": 16434 + }, + { + "epoch": 3.59, + "learning_rate": 3.2100720996285777e-06, + "loss": 2.1235, + "step": 16435 + }, + { + "epoch": 3.59, + "learning_rate": 3.2099628577670968e-06, + "loss": 0.9217, + "step": 16436 + }, + { + "epoch": 3.59, + "learning_rate": 3.2098536159056154e-06, + "loss": 2.6184, + "step": 16437 + }, + { + "epoch": 3.59, + "learning_rate": 3.209744374044134e-06, + "loss": 2.2471, + "step": 16438 + }, + { + "epoch": 3.59, + "learning_rate": 3.2096351321826526e-06, + "loss": 1.6617, + "step": 16439 + }, + { + "epoch": 3.59, + "learning_rate": 3.2095258903211712e-06, + "loss": 1.7907, + "step": 16440 + }, + { + "epoch": 3.59, + "learning_rate": 3.20941664845969e-06, + "loss": 1.2772, + "step": 16441 + }, + { + "epoch": 3.59, + "learning_rate": 3.209307406598209e-06, + "loss": 1.4007, + "step": 16442 + }, + { + "epoch": 3.59, + "learning_rate": 3.2091981647367275e-06, + "loss": 1.4245, + "step": 16443 + }, + { + "epoch": 3.59, + "learning_rate": 3.2090889228752457e-06, + "loss": 2.2132, + "step": 16444 + }, + { + "epoch": 3.59, + "learning_rate": 3.2089796810137647e-06, + "loss": 0.2951, + "step": 16445 + }, + { + "epoch": 3.59, + "learning_rate": 3.2088704391522833e-06, + "loss": 2.6477, + "step": 16446 + }, + { + "epoch": 3.59, + "learning_rate": 3.2087611972908023e-06, + "loss": 1.7745, + "step": 16447 + }, + { + "epoch": 3.59, + "learning_rate": 3.2086519554293205e-06, + "loss": 3.606, + "step": 16448 + }, + { + "epoch": 3.59, + "learning_rate": 3.208542713567839e-06, + "loss": 1.1782, + "step": 16449 + }, + { + "epoch": 3.59, + "learning_rate": 3.208433471706358e-06, + "loss": 2.4322, + "step": 16450 + }, + { + "epoch": 3.59, + "learning_rate": 3.2083242298448768e-06, + "loss": 2.8148, + "step": 16451 + }, + { + "epoch": 3.59, + "learning_rate": 3.208214987983396e-06, + "loss": 1.5846, + "step": 16452 + }, + { + "epoch": 3.59, + "learning_rate": 3.208105746121914e-06, + "loss": 1.0851, + "step": 16453 + }, + { + "epoch": 3.59, + "learning_rate": 3.2079965042604326e-06, + "loss": 1.8258, + "step": 16454 + }, + { + "epoch": 3.59, + "learning_rate": 3.2078872623989517e-06, + "loss": 2.4793, + "step": 16455 + }, + { + "epoch": 3.59, + "learning_rate": 3.2077780205374703e-06, + "loss": 3.5609, + "step": 16456 + }, + { + "epoch": 3.59, + "learning_rate": 3.2076687786759893e-06, + "loss": 1.4092, + "step": 16457 + }, + { + "epoch": 3.59, + "learning_rate": 3.2075595368145075e-06, + "loss": 1.3772, + "step": 16458 + }, + { + "epoch": 3.59, + "learning_rate": 3.207450294953026e-06, + "loss": 1.2034, + "step": 16459 + }, + { + "epoch": 3.59, + "learning_rate": 3.207341053091545e-06, + "loss": 2.0197, + "step": 16460 + }, + { + "epoch": 3.59, + "learning_rate": 3.2072318112300637e-06, + "loss": 2.7861, + "step": 16461 + }, + { + "epoch": 3.59, + "learning_rate": 3.2071225693685828e-06, + "loss": 2.9671, + "step": 16462 + }, + { + "epoch": 3.59, + "learning_rate": 3.207013327507101e-06, + "loss": 2.1869, + "step": 16463 + }, + { + "epoch": 3.59, + "learning_rate": 3.2069040856456196e-06, + "loss": 1.5558, + "step": 16464 + }, + { + "epoch": 3.59, + "learning_rate": 3.2067948437841386e-06, + "loss": 1.8806, + "step": 16465 + }, + { + "epoch": 3.59, + "learning_rate": 3.2066856019226572e-06, + "loss": 2.0701, + "step": 16466 + }, + { + "epoch": 3.59, + "learning_rate": 3.2065763600611754e-06, + "loss": 1.6174, + "step": 16467 + }, + { + "epoch": 3.59, + "learning_rate": 3.2064671181996944e-06, + "loss": 1.8352, + "step": 16468 + }, + { + "epoch": 3.59, + "learning_rate": 3.206357876338213e-06, + "loss": 2.634, + "step": 16469 + }, + { + "epoch": 3.59, + "learning_rate": 3.2062486344767317e-06, + "loss": 1.4482, + "step": 16470 + }, + { + "epoch": 3.59, + "learning_rate": 3.2061393926152507e-06, + "loss": 1.7864, + "step": 16471 + }, + { + "epoch": 3.59, + "learning_rate": 3.206030150753769e-06, + "loss": 2.4846, + "step": 16472 + }, + { + "epoch": 3.6, + "learning_rate": 3.2059209088922875e-06, + "loss": 2.1618, + "step": 16473 + }, + { + "epoch": 3.6, + "learning_rate": 3.2058116670308065e-06, + "loss": 1.6582, + "step": 16474 + }, + { + "epoch": 3.6, + "learning_rate": 3.205702425169325e-06, + "loss": 1.4266, + "step": 16475 + }, + { + "epoch": 3.6, + "learning_rate": 3.205593183307844e-06, + "loss": 2.3991, + "step": 16476 + }, + { + "epoch": 3.6, + "learning_rate": 3.2054839414463624e-06, + "loss": 2.1818, + "step": 16477 + }, + { + "epoch": 3.6, + "learning_rate": 3.205374699584881e-06, + "loss": 3.0939, + "step": 16478 + }, + { + "epoch": 3.6, + "learning_rate": 3.2052654577234e-06, + "loss": 1.5261, + "step": 16479 + }, + { + "epoch": 3.6, + "learning_rate": 3.2051562158619186e-06, + "loss": 1.7297, + "step": 16480 + }, + { + "epoch": 3.6, + "learning_rate": 3.205046974000437e-06, + "loss": 1.6185, + "step": 16481 + }, + { + "epoch": 3.6, + "learning_rate": 3.204937732138956e-06, + "loss": 1.7405, + "step": 16482 + }, + { + "epoch": 3.6, + "learning_rate": 3.2048284902774745e-06, + "loss": 2.5676, + "step": 16483 + }, + { + "epoch": 3.6, + "learning_rate": 3.2047192484159935e-06, + "loss": 2.2178, + "step": 16484 + }, + { + "epoch": 3.6, + "learning_rate": 3.204610006554512e-06, + "loss": 1.6305, + "step": 16485 + }, + { + "epoch": 3.6, + "learning_rate": 3.2045007646930303e-06, + "loss": 1.6272, + "step": 16486 + }, + { + "epoch": 3.6, + "learning_rate": 3.2043915228315493e-06, + "loss": 2.896, + "step": 16487 + }, + { + "epoch": 3.6, + "learning_rate": 3.204282280970068e-06, + "loss": 1.6014, + "step": 16488 + }, + { + "epoch": 3.6, + "learning_rate": 3.204173039108587e-06, + "loss": 1.1147, + "step": 16489 + }, + { + "epoch": 3.6, + "learning_rate": 3.2040637972471056e-06, + "loss": 2.0868, + "step": 16490 + }, + { + "epoch": 3.6, + "learning_rate": 3.2039545553856238e-06, + "loss": 0.9289, + "step": 16491 + }, + { + "epoch": 3.6, + "learning_rate": 3.203845313524143e-06, + "loss": 1.4802, + "step": 16492 + }, + { + "epoch": 3.6, + "learning_rate": 3.2037360716626614e-06, + "loss": 2.3586, + "step": 16493 + }, + { + "epoch": 3.6, + "learning_rate": 3.20362682980118e-06, + "loss": 0.6342, + "step": 16494 + }, + { + "epoch": 3.6, + "learning_rate": 3.203517587939699e-06, + "loss": 0.9313, + "step": 16495 + }, + { + "epoch": 3.6, + "learning_rate": 3.2034083460782172e-06, + "loss": 1.7661, + "step": 16496 + }, + { + "epoch": 3.6, + "learning_rate": 3.203299104216736e-06, + "loss": 2.6092, + "step": 16497 + }, + { + "epoch": 3.6, + "learning_rate": 3.203189862355255e-06, + "loss": 1.2887, + "step": 16498 + }, + { + "epoch": 3.6, + "learning_rate": 3.2030806204937735e-06, + "loss": 3.4182, + "step": 16499 + }, + { + "epoch": 3.6, + "learning_rate": 3.202971378632292e-06, + "loss": 3.2556, + "step": 16500 + }, + { + "epoch": 3.6, + "learning_rate": 3.2028621367708107e-06, + "loss": 4.3717, + "step": 16501 + }, + { + "epoch": 3.6, + "learning_rate": 3.2027528949093293e-06, + "loss": 2.0205, + "step": 16502 + }, + { + "epoch": 3.6, + "learning_rate": 3.2026436530478484e-06, + "loss": 2.752, + "step": 16503 + }, + { + "epoch": 3.6, + "learning_rate": 3.202534411186367e-06, + "loss": 1.3144, + "step": 16504 + }, + { + "epoch": 3.6, + "learning_rate": 3.202425169324885e-06, + "loss": 1.0792, + "step": 16505 + }, + { + "epoch": 3.6, + "learning_rate": 3.202315927463404e-06, + "loss": 2.4335, + "step": 16506 + }, + { + "epoch": 3.6, + "learning_rate": 3.202206685601923e-06, + "loss": 0.7555, + "step": 16507 + }, + { + "epoch": 3.6, + "learning_rate": 3.202097443740442e-06, + "loss": 2.23, + "step": 16508 + }, + { + "epoch": 3.6, + "learning_rate": 3.2019882018789605e-06, + "loss": 2.6369, + "step": 16509 + }, + { + "epoch": 3.6, + "learning_rate": 3.2018789600174786e-06, + "loss": 2.7174, + "step": 16510 + }, + { + "epoch": 3.6, + "learning_rate": 3.2017697181559977e-06, + "loss": 1.1442, + "step": 16511 + }, + { + "epoch": 3.6, + "learning_rate": 3.2016604762945163e-06, + "loss": 1.526, + "step": 16512 + }, + { + "epoch": 3.6, + "learning_rate": 3.2015512344330353e-06, + "loss": 2.11, + "step": 16513 + }, + { + "epoch": 3.6, + "learning_rate": 3.2014419925715535e-06, + "loss": 1.5443, + "step": 16514 + }, + { + "epoch": 3.6, + "learning_rate": 3.201332750710072e-06, + "loss": 1.0246, + "step": 16515 + }, + { + "epoch": 3.6, + "learning_rate": 3.201223508848591e-06, + "loss": 2.176, + "step": 16516 + }, + { + "epoch": 3.6, + "learning_rate": 3.2011142669871098e-06, + "loss": 0.5253, + "step": 16517 + }, + { + "epoch": 3.6, + "learning_rate": 3.201005025125629e-06, + "loss": 3.0245, + "step": 16518 + }, + { + "epoch": 3.61, + "learning_rate": 3.200895783264147e-06, + "loss": 0.6669, + "step": 16519 + }, + { + "epoch": 3.61, + "learning_rate": 3.2007865414026656e-06, + "loss": 0.9211, + "step": 16520 + }, + { + "epoch": 3.61, + "learning_rate": 3.2006772995411846e-06, + "loss": 1.8263, + "step": 16521 + }, + { + "epoch": 3.61, + "learning_rate": 3.2005680576797032e-06, + "loss": 0.8292, + "step": 16522 + }, + { + "epoch": 3.61, + "learning_rate": 3.200458815818222e-06, + "loss": 1.4905, + "step": 16523 + }, + { + "epoch": 3.61, + "learning_rate": 3.2003495739567405e-06, + "loss": 1.2958, + "step": 16524 + }, + { + "epoch": 3.61, + "learning_rate": 3.200240332095259e-06, + "loss": 1.7248, + "step": 16525 + }, + { + "epoch": 3.61, + "learning_rate": 3.2001310902337777e-06, + "loss": 1.4645, + "step": 16526 + }, + { + "epoch": 3.61, + "learning_rate": 3.2000218483722967e-06, + "loss": 1.5502, + "step": 16527 + }, + { + "epoch": 3.61, + "learning_rate": 3.1999126065108153e-06, + "loss": 2.3942, + "step": 16528 + }, + { + "epoch": 3.61, + "learning_rate": 3.1998033646493335e-06, + "loss": 2.0485, + "step": 16529 + }, + { + "epoch": 3.61, + "learning_rate": 3.1996941227878526e-06, + "loss": 2.0365, + "step": 16530 + }, + { + "epoch": 3.61, + "learning_rate": 3.199584880926371e-06, + "loss": 1.7061, + "step": 16531 + }, + { + "epoch": 3.61, + "learning_rate": 3.19947563906489e-06, + "loss": 1.3176, + "step": 16532 + }, + { + "epoch": 3.61, + "learning_rate": 3.1993663972034084e-06, + "loss": 1.7872, + "step": 16533 + }, + { + "epoch": 3.61, + "learning_rate": 3.199257155341927e-06, + "loss": 2.1755, + "step": 16534 + }, + { + "epoch": 3.61, + "learning_rate": 3.199147913480446e-06, + "loss": 2.8831, + "step": 16535 + }, + { + "epoch": 3.61, + "learning_rate": 3.1990386716189646e-06, + "loss": 2.3133, + "step": 16536 + }, + { + "epoch": 3.61, + "learning_rate": 3.1989294297574837e-06, + "loss": 1.4351, + "step": 16537 + }, + { + "epoch": 3.61, + "learning_rate": 3.198820187896002e-06, + "loss": 1.9651, + "step": 16538 + }, + { + "epoch": 3.61, + "learning_rate": 3.1987109460345205e-06, + "loss": 1.4345, + "step": 16539 + }, + { + "epoch": 3.61, + "learning_rate": 3.1986017041730395e-06, + "loss": 1.6096, + "step": 16540 + }, + { + "epoch": 3.61, + "learning_rate": 3.198492462311558e-06, + "loss": 2.8161, + "step": 16541 + }, + { + "epoch": 3.61, + "learning_rate": 3.198383220450077e-06, + "loss": 2.19, + "step": 16542 + }, + { + "epoch": 3.61, + "learning_rate": 3.1982739785885953e-06, + "loss": 0.7676, + "step": 16543 + }, + { + "epoch": 3.61, + "learning_rate": 3.198164736727114e-06, + "loss": 1.3962, + "step": 16544 + }, + { + "epoch": 3.61, + "learning_rate": 3.198055494865633e-06, + "loss": 1.2852, + "step": 16545 + }, + { + "epoch": 3.61, + "learning_rate": 3.1979462530041516e-06, + "loss": 3.167, + "step": 16546 + }, + { + "epoch": 3.61, + "learning_rate": 3.19783701114267e-06, + "loss": 2.2054, + "step": 16547 + }, + { + "epoch": 3.61, + "learning_rate": 3.197727769281189e-06, + "loss": 2.3409, + "step": 16548 + }, + { + "epoch": 3.61, + "learning_rate": 3.1976185274197074e-06, + "loss": 0.9059, + "step": 16549 + }, + { + "epoch": 3.61, + "learning_rate": 3.1975092855582265e-06, + "loss": 1.7412, + "step": 16550 + }, + { + "epoch": 3.61, + "learning_rate": 3.197400043696745e-06, + "loss": 2.4334, + "step": 16551 + }, + { + "epoch": 3.61, + "learning_rate": 3.1972908018352633e-06, + "loss": 1.3573, + "step": 16552 + }, + { + "epoch": 3.61, + "learning_rate": 3.1971815599737823e-06, + "loss": 1.2784, + "step": 16553 + }, + { + "epoch": 3.61, + "learning_rate": 3.197072318112301e-06, + "loss": 1.3754, + "step": 16554 + }, + { + "epoch": 3.61, + "learning_rate": 3.1969630762508195e-06, + "loss": 2.2238, + "step": 16555 + }, + { + "epoch": 3.61, + "learning_rate": 3.1968538343893386e-06, + "loss": 1.8809, + "step": 16556 + }, + { + "epoch": 3.61, + "learning_rate": 3.1967445925278567e-06, + "loss": 2.221, + "step": 16557 + }, + { + "epoch": 3.61, + "learning_rate": 3.1966353506663754e-06, + "loss": 2.3051, + "step": 16558 + }, + { + "epoch": 3.61, + "learning_rate": 3.1965261088048944e-06, + "loss": 1.7393, + "step": 16559 + }, + { + "epoch": 3.61, + "learning_rate": 3.196416866943413e-06, + "loss": 2.7005, + "step": 16560 + }, + { + "epoch": 3.61, + "learning_rate": 3.196307625081932e-06, + "loss": 1.6932, + "step": 16561 + }, + { + "epoch": 3.61, + "learning_rate": 3.1961983832204502e-06, + "loss": 1.7426, + "step": 16562 + }, + { + "epoch": 3.61, + "learning_rate": 3.196089141358969e-06, + "loss": 2.2966, + "step": 16563 + }, + { + "epoch": 3.62, + "learning_rate": 3.195979899497488e-06, + "loss": 0.8691, + "step": 16564 + }, + { + "epoch": 3.62, + "learning_rate": 3.1958706576360065e-06, + "loss": 1.329, + "step": 16565 + }, + { + "epoch": 3.62, + "learning_rate": 3.1957614157745247e-06, + "loss": 2.2983, + "step": 16566 + }, + { + "epoch": 3.62, + "learning_rate": 3.1956521739130437e-06, + "loss": 0.9276, + "step": 16567 + }, + { + "epoch": 3.62, + "learning_rate": 3.1955429320515623e-06, + "loss": 1.9023, + "step": 16568 + }, + { + "epoch": 3.62, + "learning_rate": 3.1954336901900814e-06, + "loss": 1.5164, + "step": 16569 + }, + { + "epoch": 3.62, + "learning_rate": 3.1953244483286e-06, + "loss": 2.9025, + "step": 16570 + }, + { + "epoch": 3.62, + "learning_rate": 3.195215206467118e-06, + "loss": 1.0395, + "step": 16571 + }, + { + "epoch": 3.62, + "learning_rate": 3.195105964605637e-06, + "loss": 2.9038, + "step": 16572 + }, + { + "epoch": 3.62, + "learning_rate": 3.194996722744156e-06, + "loss": 2.8132, + "step": 16573 + }, + { + "epoch": 3.62, + "learning_rate": 3.194887480882675e-06, + "loss": 2.3338, + "step": 16574 + }, + { + "epoch": 3.62, + "learning_rate": 3.1947782390211934e-06, + "loss": 1.9834, + "step": 16575 + }, + { + "epoch": 3.62, + "learning_rate": 3.1946689971597116e-06, + "loss": 1.692, + "step": 16576 + }, + { + "epoch": 3.62, + "learning_rate": 3.1945597552982307e-06, + "loss": 1.5749, + "step": 16577 + }, + { + "epoch": 3.62, + "learning_rate": 3.1944505134367493e-06, + "loss": 2.8618, + "step": 16578 + }, + { + "epoch": 3.62, + "learning_rate": 3.194341271575268e-06, + "loss": 3.2598, + "step": 16579 + }, + { + "epoch": 3.62, + "learning_rate": 3.1942320297137865e-06, + "loss": 2.7469, + "step": 16580 + }, + { + "epoch": 3.62, + "learning_rate": 3.194122787852305e-06, + "loss": 2.151, + "step": 16581 + }, + { + "epoch": 3.62, + "learning_rate": 3.1940135459908237e-06, + "loss": 0.8429, + "step": 16582 + }, + { + "epoch": 3.62, + "learning_rate": 3.1939043041293428e-06, + "loss": 2.7619, + "step": 16583 + }, + { + "epoch": 3.62, + "learning_rate": 3.1937950622678614e-06, + "loss": 2.1143, + "step": 16584 + }, + { + "epoch": 3.62, + "learning_rate": 3.1936858204063796e-06, + "loss": 1.3398, + "step": 16585 + }, + { + "epoch": 3.62, + "learning_rate": 3.1935765785448986e-06, + "loss": 2.4901, + "step": 16586 + }, + { + "epoch": 3.62, + "learning_rate": 3.193467336683417e-06, + "loss": 2.5476, + "step": 16587 + }, + { + "epoch": 3.62, + "learning_rate": 3.1933580948219362e-06, + "loss": 2.3058, + "step": 16588 + }, + { + "epoch": 3.62, + "learning_rate": 3.193248852960455e-06, + "loss": 1.748, + "step": 16589 + }, + { + "epoch": 3.62, + "learning_rate": 3.193139611098973e-06, + "loss": 2.0777, + "step": 16590 + }, + { + "epoch": 3.62, + "learning_rate": 3.193030369237492e-06, + "loss": 2.189, + "step": 16591 + }, + { + "epoch": 3.62, + "learning_rate": 3.1929211273760107e-06, + "loss": 1.9805, + "step": 16592 + }, + { + "epoch": 3.62, + "learning_rate": 3.1928118855145297e-06, + "loss": 4.2361, + "step": 16593 + }, + { + "epoch": 3.62, + "learning_rate": 3.1927026436530483e-06, + "loss": 0.9168, + "step": 16594 + }, + { + "epoch": 3.62, + "learning_rate": 3.1925934017915665e-06, + "loss": 1.1115, + "step": 16595 + }, + { + "epoch": 3.62, + "learning_rate": 3.1924841599300855e-06, + "loss": 3.8759, + "step": 16596 + }, + { + "epoch": 3.62, + "learning_rate": 3.192374918068604e-06, + "loss": 2.4562, + "step": 16597 + }, + { + "epoch": 3.62, + "learning_rate": 3.192265676207123e-06, + "loss": 1.4717, + "step": 16598 + }, + { + "epoch": 3.62, + "learning_rate": 3.1921564343456414e-06, + "loss": 1.1342, + "step": 16599 + }, + { + "epoch": 3.62, + "learning_rate": 3.19204719248416e-06, + "loss": 1.5098, + "step": 16600 + }, + { + "epoch": 3.62, + "learning_rate": 3.191937950622679e-06, + "loss": 1.9791, + "step": 16601 + }, + { + "epoch": 3.62, + "learning_rate": 3.1918287087611976e-06, + "loss": 1.0382, + "step": 16602 + }, + { + "epoch": 3.62, + "learning_rate": 3.1917194668997167e-06, + "loss": 1.5245, + "step": 16603 + }, + { + "epoch": 3.62, + "learning_rate": 3.191610225038235e-06, + "loss": 2.7325, + "step": 16604 + }, + { + "epoch": 3.62, + "learning_rate": 3.1915009831767535e-06, + "loss": 1.5955, + "step": 16605 + }, + { + "epoch": 3.62, + "learning_rate": 3.1913917413152725e-06, + "loss": 2.8328, + "step": 16606 + }, + { + "epoch": 3.62, + "learning_rate": 3.191282499453791e-06, + "loss": 1.2521, + "step": 16607 + }, + { + "epoch": 3.62, + "learning_rate": 3.1911732575923097e-06, + "loss": 0.6556, + "step": 16608 + }, + { + "epoch": 3.62, + "learning_rate": 3.1910640157308283e-06, + "loss": 0.4799, + "step": 16609 + }, + { + "epoch": 3.63, + "learning_rate": 3.190954773869347e-06, + "loss": 1.7019, + "step": 16610 + }, + { + "epoch": 3.63, + "learning_rate": 3.1908455320078656e-06, + "loss": 0.8898, + "step": 16611 + }, + { + "epoch": 3.63, + "learning_rate": 3.1907362901463846e-06, + "loss": 2.3555, + "step": 16612 + }, + { + "epoch": 3.63, + "learning_rate": 3.1906270482849028e-06, + "loss": 1.564, + "step": 16613 + }, + { + "epoch": 3.63, + "learning_rate": 3.1905178064234214e-06, + "loss": 1.0344, + "step": 16614 + }, + { + "epoch": 3.63, + "learning_rate": 3.1904085645619404e-06, + "loss": 0.7941, + "step": 16615 + }, + { + "epoch": 3.63, + "learning_rate": 3.190299322700459e-06, + "loss": 3.2023, + "step": 16616 + }, + { + "epoch": 3.63, + "learning_rate": 3.190190080838978e-06, + "loss": 1.1793, + "step": 16617 + }, + { + "epoch": 3.63, + "learning_rate": 3.1900808389774963e-06, + "loss": 1.1463, + "step": 16618 + }, + { + "epoch": 3.63, + "learning_rate": 3.189971597116015e-06, + "loss": 2.6126, + "step": 16619 + }, + { + "epoch": 3.63, + "learning_rate": 3.189862355254534e-06, + "loss": 2.331, + "step": 16620 + }, + { + "epoch": 3.63, + "learning_rate": 3.1897531133930525e-06, + "loss": 1.5595, + "step": 16621 + }, + { + "epoch": 3.63, + "learning_rate": 3.1896438715315715e-06, + "loss": 2.4429, + "step": 16622 + }, + { + "epoch": 3.63, + "learning_rate": 3.1895346296700897e-06, + "loss": 1.6443, + "step": 16623 + }, + { + "epoch": 3.63, + "learning_rate": 3.1894253878086083e-06, + "loss": 2.5743, + "step": 16624 + }, + { + "epoch": 3.63, + "learning_rate": 3.1893161459471274e-06, + "loss": 2.8774, + "step": 16625 + }, + { + "epoch": 3.63, + "learning_rate": 3.189206904085646e-06, + "loss": 3.3599, + "step": 16626 + }, + { + "epoch": 3.63, + "learning_rate": 3.189097662224165e-06, + "loss": 0.5043, + "step": 16627 + }, + { + "epoch": 3.63, + "learning_rate": 3.1889884203626832e-06, + "loss": 2.0123, + "step": 16628 + }, + { + "epoch": 3.63, + "learning_rate": 3.188879178501202e-06, + "loss": 1.7062, + "step": 16629 + }, + { + "epoch": 3.63, + "learning_rate": 3.188769936639721e-06, + "loss": 1.6865, + "step": 16630 + }, + { + "epoch": 3.63, + "learning_rate": 3.1886606947782395e-06, + "loss": 2.093, + "step": 16631 + }, + { + "epoch": 3.63, + "learning_rate": 3.1885514529167577e-06, + "loss": 2.4137, + "step": 16632 + }, + { + "epoch": 3.63, + "learning_rate": 3.1884422110552767e-06, + "loss": 1.1152, + "step": 16633 + }, + { + "epoch": 3.63, + "learning_rate": 3.1883329691937953e-06, + "loss": 1.342, + "step": 16634 + }, + { + "epoch": 3.63, + "learning_rate": 3.188223727332314e-06, + "loss": 2.7767, + "step": 16635 + }, + { + "epoch": 3.63, + "learning_rate": 3.188114485470833e-06, + "loss": 2.3061, + "step": 16636 + }, + { + "epoch": 3.63, + "learning_rate": 3.188005243609351e-06, + "loss": 1.3251, + "step": 16637 + }, + { + "epoch": 3.63, + "learning_rate": 3.1878960017478697e-06, + "loss": 2.0754, + "step": 16638 + }, + { + "epoch": 3.63, + "learning_rate": 3.1877867598863888e-06, + "loss": 1.9976, + "step": 16639 + }, + { + "epoch": 3.63, + "learning_rate": 3.1876775180249074e-06, + "loss": 2.1501, + "step": 16640 + }, + { + "epoch": 3.63, + "learning_rate": 3.1875682761634264e-06, + "loss": 1.582, + "step": 16641 + }, + { + "epoch": 3.63, + "learning_rate": 3.1874590343019446e-06, + "loss": 2.8026, + "step": 16642 + }, + { + "epoch": 3.63, + "learning_rate": 3.1873497924404632e-06, + "loss": 1.7558, + "step": 16643 + }, + { + "epoch": 3.63, + "learning_rate": 3.1872405505789823e-06, + "loss": 1.6688, + "step": 16644 + }, + { + "epoch": 3.63, + "learning_rate": 3.187131308717501e-06, + "loss": 1.487, + "step": 16645 + }, + { + "epoch": 3.63, + "learning_rate": 3.187022066856019e-06, + "loss": 3.3591, + "step": 16646 + }, + { + "epoch": 3.63, + "learning_rate": 3.186912824994538e-06, + "loss": 2.6632, + "step": 16647 + }, + { + "epoch": 3.63, + "learning_rate": 3.1868035831330567e-06, + "loss": 2.864, + "step": 16648 + }, + { + "epoch": 3.63, + "learning_rate": 3.1866943412715757e-06, + "loss": 1.4332, + "step": 16649 + }, + { + "epoch": 3.63, + "learning_rate": 3.1865850994100943e-06, + "loss": 1.821, + "step": 16650 + }, + { + "epoch": 3.63, + "learning_rate": 3.1864758575486125e-06, + "loss": 1.276, + "step": 16651 + }, + { + "epoch": 3.63, + "learning_rate": 3.1863666156871316e-06, + "loss": 2.663, + "step": 16652 + }, + { + "epoch": 3.63, + "learning_rate": 3.18625737382565e-06, + "loss": 2.6244, + "step": 16653 + }, + { + "epoch": 3.63, + "learning_rate": 3.1861481319641692e-06, + "loss": 2.4489, + "step": 16654 + }, + { + "epoch": 3.63, + "learning_rate": 3.186038890102688e-06, + "loss": 0.8183, + "step": 16655 + }, + { + "epoch": 3.64, + "learning_rate": 3.185929648241206e-06, + "loss": 2.3197, + "step": 16656 + }, + { + "epoch": 3.64, + "learning_rate": 3.185820406379725e-06, + "loss": 1.1817, + "step": 16657 + }, + { + "epoch": 3.64, + "learning_rate": 3.1857111645182437e-06, + "loss": 1.6754, + "step": 16658 + }, + { + "epoch": 3.64, + "learning_rate": 3.1856019226567627e-06, + "loss": 2.1869, + "step": 16659 + }, + { + "epoch": 3.64, + "learning_rate": 3.1854926807952813e-06, + "loss": 1.2076, + "step": 16660 + }, + { + "epoch": 3.64, + "learning_rate": 3.1853834389337995e-06, + "loss": 1.9146, + "step": 16661 + }, + { + "epoch": 3.64, + "learning_rate": 3.1852741970723185e-06, + "loss": 1.6011, + "step": 16662 + }, + { + "epoch": 3.64, + "learning_rate": 3.185164955210837e-06, + "loss": 2.5565, + "step": 16663 + }, + { + "epoch": 3.64, + "learning_rate": 3.1850557133493557e-06, + "loss": 1.6366, + "step": 16664 + }, + { + "epoch": 3.64, + "learning_rate": 3.1849464714878744e-06, + "loss": 1.8395, + "step": 16665 + }, + { + "epoch": 3.64, + "learning_rate": 3.184837229626393e-06, + "loss": 2.4503, + "step": 16666 + }, + { + "epoch": 3.64, + "learning_rate": 3.1847279877649116e-06, + "loss": 3.1382, + "step": 16667 + }, + { + "epoch": 3.64, + "learning_rate": 3.1846187459034306e-06, + "loss": 3.3429, + "step": 16668 + }, + { + "epoch": 3.64, + "learning_rate": 3.1845095040419492e-06, + "loss": 1.5873, + "step": 16669 + }, + { + "epoch": 3.64, + "learning_rate": 3.1844002621804674e-06, + "loss": 1.7157, + "step": 16670 + }, + { + "epoch": 3.64, + "learning_rate": 3.1842910203189864e-06, + "loss": 1.7556, + "step": 16671 + }, + { + "epoch": 3.64, + "learning_rate": 3.184181778457505e-06, + "loss": 1.1444, + "step": 16672 + }, + { + "epoch": 3.64, + "learning_rate": 3.184072536596024e-06, + "loss": 2.7251, + "step": 16673 + }, + { + "epoch": 3.64, + "learning_rate": 3.1839632947345427e-06, + "loss": 2.827, + "step": 16674 + }, + { + "epoch": 3.64, + "learning_rate": 3.183854052873061e-06, + "loss": 0.8399, + "step": 16675 + }, + { + "epoch": 3.64, + "learning_rate": 3.18374481101158e-06, + "loss": 0.9909, + "step": 16676 + }, + { + "epoch": 3.64, + "learning_rate": 3.1836355691500985e-06, + "loss": 2.16, + "step": 16677 + }, + { + "epoch": 3.64, + "learning_rate": 3.1835263272886176e-06, + "loss": 1.4683, + "step": 16678 + }, + { + "epoch": 3.64, + "learning_rate": 3.1834170854271358e-06, + "loss": 1.2169, + "step": 16679 + }, + { + "epoch": 3.64, + "learning_rate": 3.1833078435656544e-06, + "loss": 1.0072, + "step": 16680 + }, + { + "epoch": 3.64, + "learning_rate": 3.1831986017041734e-06, + "loss": 1.8116, + "step": 16681 + }, + { + "epoch": 3.64, + "learning_rate": 3.183089359842692e-06, + "loss": 1.7022, + "step": 16682 + }, + { + "epoch": 3.64, + "learning_rate": 3.182980117981211e-06, + "loss": 1.4926, + "step": 16683 + }, + { + "epoch": 3.64, + "learning_rate": 3.1828708761197292e-06, + "loss": 1.2828, + "step": 16684 + }, + { + "epoch": 3.64, + "learning_rate": 3.182761634258248e-06, + "loss": 1.8123, + "step": 16685 + }, + { + "epoch": 3.64, + "learning_rate": 3.182652392396767e-06, + "loss": 2.9905, + "step": 16686 + }, + { + "epoch": 3.64, + "learning_rate": 3.1825431505352855e-06, + "loss": 2.7517, + "step": 16687 + }, + { + "epoch": 3.64, + "learning_rate": 3.182433908673804e-06, + "loss": 2.7033, + "step": 16688 + }, + { + "epoch": 3.64, + "learning_rate": 3.1823246668123227e-06, + "loss": 1.3105, + "step": 16689 + }, + { + "epoch": 3.64, + "learning_rate": 3.1822154249508413e-06, + "loss": 2.1026, + "step": 16690 + }, + { + "epoch": 3.64, + "learning_rate": 3.18210618308936e-06, + "loss": 1.229, + "step": 16691 + }, + { + "epoch": 3.64, + "learning_rate": 3.181996941227879e-06, + "loss": 1.8209, + "step": 16692 + }, + { + "epoch": 3.64, + "learning_rate": 3.181887699366397e-06, + "loss": 1.1879, + "step": 16693 + }, + { + "epoch": 3.64, + "learning_rate": 3.181778457504916e-06, + "loss": 1.0538, + "step": 16694 + }, + { + "epoch": 3.64, + "learning_rate": 3.181669215643435e-06, + "loss": 1.397, + "step": 16695 + }, + { + "epoch": 3.64, + "learning_rate": 3.1815599737819534e-06, + "loss": 1.8866, + "step": 16696 + }, + { + "epoch": 3.64, + "learning_rate": 3.1814507319204725e-06, + "loss": 1.3316, + "step": 16697 + }, + { + "epoch": 3.64, + "learning_rate": 3.1813414900589906e-06, + "loss": 2.501, + "step": 16698 + }, + { + "epoch": 3.64, + "learning_rate": 3.1812322481975093e-06, + "loss": 4.207, + "step": 16699 + }, + { + "epoch": 3.64, + "learning_rate": 3.1811230063360283e-06, + "loss": 2.52, + "step": 16700 + }, + { + "epoch": 3.64, + "learning_rate": 3.181013764474547e-06, + "loss": 2.1259, + "step": 16701 + }, + { + "epoch": 3.65, + "learning_rate": 3.180904522613066e-06, + "loss": 0.9271, + "step": 16702 + }, + { + "epoch": 3.65, + "learning_rate": 3.180795280751584e-06, + "loss": 1.3614, + "step": 16703 + }, + { + "epoch": 3.65, + "learning_rate": 3.1806860388901027e-06, + "loss": 1.2675, + "step": 16704 + }, + { + "epoch": 3.65, + "learning_rate": 3.1805767970286218e-06, + "loss": 0.3764, + "step": 16705 + }, + { + "epoch": 3.65, + "learning_rate": 3.1804675551671404e-06, + "loss": 2.3024, + "step": 16706 + }, + { + "epoch": 3.65, + "learning_rate": 3.1803583133056594e-06, + "loss": 1.1452, + "step": 16707 + }, + { + "epoch": 3.65, + "learning_rate": 3.1802490714441776e-06, + "loss": 2.231, + "step": 16708 + }, + { + "epoch": 3.65, + "learning_rate": 3.180139829582696e-06, + "loss": 1.5566, + "step": 16709 + }, + { + "epoch": 3.65, + "learning_rate": 3.1800305877212152e-06, + "loss": 2.299, + "step": 16710 + }, + { + "epoch": 3.65, + "learning_rate": 3.179921345859734e-06, + "loss": 1.805, + "step": 16711 + }, + { + "epoch": 3.65, + "learning_rate": 3.179812103998252e-06, + "loss": 0.748, + "step": 16712 + }, + { + "epoch": 3.65, + "learning_rate": 3.179702862136771e-06, + "loss": 2.5029, + "step": 16713 + }, + { + "epoch": 3.65, + "learning_rate": 3.1795936202752897e-06, + "loss": 1.8231, + "step": 16714 + }, + { + "epoch": 3.65, + "learning_rate": 3.1794843784138087e-06, + "loss": 2.7851, + "step": 16715 + }, + { + "epoch": 3.65, + "learning_rate": 3.1793751365523273e-06, + "loss": 1.7117, + "step": 16716 + }, + { + "epoch": 3.65, + "learning_rate": 3.1792658946908455e-06, + "loss": 1.8529, + "step": 16717 + }, + { + "epoch": 3.65, + "learning_rate": 3.1791566528293646e-06, + "loss": 2.1142, + "step": 16718 + }, + { + "epoch": 3.65, + "learning_rate": 3.179047410967883e-06, + "loss": 2.0508, + "step": 16719 + }, + { + "epoch": 3.65, + "learning_rate": 3.1789381691064018e-06, + "loss": 1.7453, + "step": 16720 + }, + { + "epoch": 3.65, + "learning_rate": 3.178828927244921e-06, + "loss": 2.2924, + "step": 16721 + }, + { + "epoch": 3.65, + "learning_rate": 3.178719685383439e-06, + "loss": 2.2273, + "step": 16722 + }, + { + "epoch": 3.65, + "learning_rate": 3.1786104435219576e-06, + "loss": 1.8665, + "step": 16723 + }, + { + "epoch": 3.65, + "learning_rate": 3.1785012016604766e-06, + "loss": 1.8918, + "step": 16724 + }, + { + "epoch": 3.65, + "learning_rate": 3.1783919597989953e-06, + "loss": 2.0277, + "step": 16725 + }, + { + "epoch": 3.65, + "learning_rate": 3.1782827179375134e-06, + "loss": 2.3266, + "step": 16726 + }, + { + "epoch": 3.65, + "learning_rate": 3.1781734760760325e-06, + "loss": 1.9734, + "step": 16727 + }, + { + "epoch": 3.65, + "learning_rate": 3.178064234214551e-06, + "loss": 2.5098, + "step": 16728 + }, + { + "epoch": 3.65, + "learning_rate": 3.17795499235307e-06, + "loss": 3.3499, + "step": 16729 + }, + { + "epoch": 3.65, + "learning_rate": 3.1778457504915887e-06, + "loss": 1.6318, + "step": 16730 + }, + { + "epoch": 3.65, + "learning_rate": 3.177736508630107e-06, + "loss": 2.1435, + "step": 16731 + }, + { + "epoch": 3.65, + "learning_rate": 3.177627266768626e-06, + "loss": 1.4072, + "step": 16732 + }, + { + "epoch": 3.65, + "learning_rate": 3.1775180249071446e-06, + "loss": 1.1389, + "step": 16733 + }, + { + "epoch": 3.65, + "learning_rate": 3.1774087830456636e-06, + "loss": 1.5627, + "step": 16734 + }, + { + "epoch": 3.65, + "learning_rate": 3.177299541184182e-06, + "loss": 2.7974, + "step": 16735 + }, + { + "epoch": 3.65, + "learning_rate": 3.1771902993227004e-06, + "loss": 1.643, + "step": 16736 + }, + { + "epoch": 3.65, + "learning_rate": 3.1770810574612194e-06, + "loss": 1.3936, + "step": 16737 + }, + { + "epoch": 3.65, + "learning_rate": 3.176971815599738e-06, + "loss": 1.8613, + "step": 16738 + }, + { + "epoch": 3.65, + "learning_rate": 3.176862573738257e-06, + "loss": 1.8498, + "step": 16739 + }, + { + "epoch": 3.65, + "learning_rate": 3.1767533318767757e-06, + "loss": 1.6207, + "step": 16740 + }, + { + "epoch": 3.65, + "learning_rate": 3.176644090015294e-06, + "loss": 2.3073, + "step": 16741 + }, + { + "epoch": 3.65, + "learning_rate": 3.176534848153813e-06, + "loss": 3.3153, + "step": 16742 + }, + { + "epoch": 3.65, + "learning_rate": 3.1764256062923315e-06, + "loss": 0.567, + "step": 16743 + }, + { + "epoch": 3.65, + "learning_rate": 3.17631636443085e-06, + "loss": 2.4407, + "step": 16744 + }, + { + "epoch": 3.65, + "learning_rate": 3.1762071225693687e-06, + "loss": 1.0419, + "step": 16745 + }, + { + "epoch": 3.65, + "learning_rate": 3.1760978807078874e-06, + "loss": 2.8267, + "step": 16746 + }, + { + "epoch": 3.65, + "learning_rate": 3.1759886388464064e-06, + "loss": 1.7221, + "step": 16747 + }, + { + "epoch": 3.66, + "learning_rate": 3.175879396984925e-06, + "loss": 1.5504, + "step": 16748 + }, + { + "epoch": 3.66, + "learning_rate": 3.1757701551234436e-06, + "loss": 1.1188, + "step": 16749 + }, + { + "epoch": 3.66, + "learning_rate": 3.1756609132619622e-06, + "loss": 1.8338, + "step": 16750 + }, + { + "epoch": 3.66, + "learning_rate": 3.175551671400481e-06, + "loss": 1.6667, + "step": 16751 + }, + { + "epoch": 3.66, + "learning_rate": 3.1754424295389994e-06, + "loss": 2.1233, + "step": 16752 + }, + { + "epoch": 3.66, + "learning_rate": 3.1753331876775185e-06, + "loss": 1.8571, + "step": 16753 + }, + { + "epoch": 3.66, + "learning_rate": 3.175223945816037e-06, + "loss": 2.1085, + "step": 16754 + }, + { + "epoch": 3.66, + "learning_rate": 3.1751147039545553e-06, + "loss": 1.4338, + "step": 16755 + }, + { + "epoch": 3.66, + "learning_rate": 3.1750054620930743e-06, + "loss": 1.6346, + "step": 16756 + }, + { + "epoch": 3.66, + "learning_rate": 3.174896220231593e-06, + "loss": 1.3844, + "step": 16757 + }, + { + "epoch": 3.66, + "learning_rate": 3.174786978370112e-06, + "loss": 1.8337, + "step": 16758 + }, + { + "epoch": 3.66, + "learning_rate": 3.17467773650863e-06, + "loss": 2.0746, + "step": 16759 + }, + { + "epoch": 3.66, + "learning_rate": 3.1745684946471488e-06, + "loss": 1.1006, + "step": 16760 + }, + { + "epoch": 3.66, + "learning_rate": 3.174459252785668e-06, + "loss": 2.6884, + "step": 16761 + }, + { + "epoch": 3.66, + "learning_rate": 3.1743500109241864e-06, + "loss": 2.8042, + "step": 16762 + }, + { + "epoch": 3.66, + "learning_rate": 3.1742407690627054e-06, + "loss": 1.6527, + "step": 16763 + }, + { + "epoch": 3.66, + "learning_rate": 3.1741315272012236e-06, + "loss": 1.4278, + "step": 16764 + }, + { + "epoch": 3.66, + "learning_rate": 3.1740222853397422e-06, + "loss": 2.3606, + "step": 16765 + }, + { + "epoch": 3.66, + "learning_rate": 3.1739130434782613e-06, + "loss": 2.1027, + "step": 16766 + }, + { + "epoch": 3.66, + "learning_rate": 3.17380380161678e-06, + "loss": 0.7008, + "step": 16767 + }, + { + "epoch": 3.66, + "learning_rate": 3.173694559755299e-06, + "loss": 1.7915, + "step": 16768 + }, + { + "epoch": 3.66, + "learning_rate": 3.173585317893817e-06, + "loss": 2.0562, + "step": 16769 + }, + { + "epoch": 3.66, + "learning_rate": 3.1734760760323357e-06, + "loss": 3.036, + "step": 16770 + }, + { + "epoch": 3.66, + "learning_rate": 3.1733668341708547e-06, + "loss": 1.6024, + "step": 16771 + }, + { + "epoch": 3.66, + "learning_rate": 3.1732575923093734e-06, + "loss": 2.2882, + "step": 16772 + }, + { + "epoch": 3.66, + "learning_rate": 3.173148350447892e-06, + "loss": 1.3995, + "step": 16773 + }, + { + "epoch": 3.66, + "learning_rate": 3.1730391085864106e-06, + "loss": 1.6235, + "step": 16774 + }, + { + "epoch": 3.66, + "learning_rate": 3.172929866724929e-06, + "loss": 1.713, + "step": 16775 + }, + { + "epoch": 3.66, + "learning_rate": 3.172820624863448e-06, + "loss": 1.7796, + "step": 16776 + }, + { + "epoch": 3.66, + "learning_rate": 3.172711383001967e-06, + "loss": 1.3727, + "step": 16777 + }, + { + "epoch": 3.66, + "learning_rate": 3.172602141140485e-06, + "loss": 2.7294, + "step": 16778 + }, + { + "epoch": 3.66, + "learning_rate": 3.1724928992790036e-06, + "loss": 1.4188, + "step": 16779 + }, + { + "epoch": 3.66, + "learning_rate": 3.1723836574175227e-06, + "loss": 1.1607, + "step": 16780 + }, + { + "epoch": 3.66, + "learning_rate": 3.1722744155560413e-06, + "loss": 1.2414, + "step": 16781 + }, + { + "epoch": 3.66, + "learning_rate": 3.1721651736945603e-06, + "loss": 0.8992, + "step": 16782 + }, + { + "epoch": 3.66, + "learning_rate": 3.1720559318330785e-06, + "loss": 1.8773, + "step": 16783 + }, + { + "epoch": 3.66, + "learning_rate": 3.171946689971597e-06, + "loss": 1.5768, + "step": 16784 + }, + { + "epoch": 3.66, + "learning_rate": 3.171837448110116e-06, + "loss": 3.9788, + "step": 16785 + }, + { + "epoch": 3.66, + "learning_rate": 3.1717282062486348e-06, + "loss": 2.1041, + "step": 16786 + }, + { + "epoch": 3.66, + "learning_rate": 3.171618964387154e-06, + "loss": 2.2487, + "step": 16787 + }, + { + "epoch": 3.66, + "learning_rate": 3.171509722525672e-06, + "loss": 1.2952, + "step": 16788 + }, + { + "epoch": 3.66, + "learning_rate": 3.1714004806641906e-06, + "loss": 1.4458, + "step": 16789 + }, + { + "epoch": 3.66, + "learning_rate": 3.1712912388027096e-06, + "loss": 1.1203, + "step": 16790 + }, + { + "epoch": 3.66, + "learning_rate": 3.1711819969412282e-06, + "loss": 1.4659, + "step": 16791 + }, + { + "epoch": 3.66, + "learning_rate": 3.1710727550797464e-06, + "loss": 3.8491, + "step": 16792 + }, + { + "epoch": 3.66, + "learning_rate": 3.1709635132182655e-06, + "loss": 2.1303, + "step": 16793 + }, + { + "epoch": 3.67, + "learning_rate": 3.170854271356784e-06, + "loss": 1.9854, + "step": 16794 + }, + { + "epoch": 3.67, + "learning_rate": 3.170745029495303e-06, + "loss": 0.981, + "step": 16795 + }, + { + "epoch": 3.67, + "learning_rate": 3.1706357876338217e-06, + "loss": 1.5925, + "step": 16796 + }, + { + "epoch": 3.67, + "learning_rate": 3.17052654577234e-06, + "loss": 2.741, + "step": 16797 + }, + { + "epoch": 3.67, + "learning_rate": 3.170417303910859e-06, + "loss": 1.762, + "step": 16798 + }, + { + "epoch": 3.67, + "learning_rate": 3.1703080620493775e-06, + "loss": 2.0694, + "step": 16799 + }, + { + "epoch": 3.67, + "learning_rate": 3.1701988201878966e-06, + "loss": 1.4003, + "step": 16800 + }, + { + "epoch": 3.67, + "learning_rate": 3.170089578326415e-06, + "loss": 1.4761, + "step": 16801 + }, + { + "epoch": 3.67, + "learning_rate": 3.1699803364649334e-06, + "loss": 3.0523, + "step": 16802 + }, + { + "epoch": 3.67, + "learning_rate": 3.1698710946034524e-06, + "loss": 0.5302, + "step": 16803 + }, + { + "epoch": 3.67, + "learning_rate": 3.169761852741971e-06, + "loss": 1.6754, + "step": 16804 + }, + { + "epoch": 3.67, + "learning_rate": 3.1696526108804896e-06, + "loss": 1.7572, + "step": 16805 + }, + { + "epoch": 3.67, + "learning_rate": 3.1695433690190087e-06, + "loss": 1.4157, + "step": 16806 + }, + { + "epoch": 3.67, + "learning_rate": 3.169434127157527e-06, + "loss": 2.7767, + "step": 16807 + }, + { + "epoch": 3.67, + "learning_rate": 3.1693248852960455e-06, + "loss": 0.7463, + "step": 16808 + }, + { + "epoch": 3.67, + "learning_rate": 3.1692156434345645e-06, + "loss": 0.8664, + "step": 16809 + }, + { + "epoch": 3.67, + "learning_rate": 3.169106401573083e-06, + "loss": 1.2462, + "step": 16810 + }, + { + "epoch": 3.67, + "learning_rate": 3.1689971597116013e-06, + "loss": 1.8124, + "step": 16811 + }, + { + "epoch": 3.67, + "learning_rate": 3.1688879178501203e-06, + "loss": 2.5738, + "step": 16812 + }, + { + "epoch": 3.67, + "learning_rate": 3.168778675988639e-06, + "loss": 1.9466, + "step": 16813 + }, + { + "epoch": 3.67, + "learning_rate": 3.168669434127158e-06, + "loss": 1.3893, + "step": 16814 + }, + { + "epoch": 3.67, + "learning_rate": 3.1685601922656766e-06, + "loss": 1.8791, + "step": 16815 + }, + { + "epoch": 3.67, + "learning_rate": 3.1684509504041948e-06, + "loss": 4.1622, + "step": 16816 + }, + { + "epoch": 3.67, + "learning_rate": 3.168341708542714e-06, + "loss": 1.1371, + "step": 16817 + }, + { + "epoch": 3.67, + "learning_rate": 3.1682324666812324e-06, + "loss": 3.3646, + "step": 16818 + }, + { + "epoch": 3.67, + "learning_rate": 3.1681232248197515e-06, + "loss": 1.5395, + "step": 16819 + }, + { + "epoch": 3.67, + "learning_rate": 3.16801398295827e-06, + "loss": 1.2725, + "step": 16820 + }, + { + "epoch": 3.67, + "learning_rate": 3.1679047410967883e-06, + "loss": 2.5135, + "step": 16821 + }, + { + "epoch": 3.67, + "learning_rate": 3.1677954992353073e-06, + "loss": 2.3649, + "step": 16822 + }, + { + "epoch": 3.67, + "learning_rate": 3.167686257373826e-06, + "loss": 1.0451, + "step": 16823 + }, + { + "epoch": 3.67, + "learning_rate": 3.167577015512345e-06, + "loss": 2.4889, + "step": 16824 + }, + { + "epoch": 3.67, + "learning_rate": 3.167467773650863e-06, + "loss": 1.3794, + "step": 16825 + }, + { + "epoch": 3.67, + "learning_rate": 3.1673585317893817e-06, + "loss": 2.2905, + "step": 16826 + }, + { + "epoch": 3.67, + "learning_rate": 3.1672492899279008e-06, + "loss": 1.7406, + "step": 16827 + }, + { + "epoch": 3.67, + "learning_rate": 3.1671400480664194e-06, + "loss": 1.2913, + "step": 16828 + }, + { + "epoch": 3.67, + "learning_rate": 3.167030806204938e-06, + "loss": 1.9582, + "step": 16829 + }, + { + "epoch": 3.67, + "learning_rate": 3.1669215643434566e-06, + "loss": 1.05, + "step": 16830 + }, + { + "epoch": 3.67, + "learning_rate": 3.1668123224819752e-06, + "loss": 2.5806, + "step": 16831 + }, + { + "epoch": 3.67, + "learning_rate": 3.166703080620494e-06, + "loss": 2.3807, + "step": 16832 + }, + { + "epoch": 3.67, + "learning_rate": 3.166593838759013e-06, + "loss": 2.6558, + "step": 16833 + }, + { + "epoch": 3.67, + "learning_rate": 3.1664845968975315e-06, + "loss": 2.3179, + "step": 16834 + }, + { + "epoch": 3.67, + "learning_rate": 3.1663753550360497e-06, + "loss": 1.5272, + "step": 16835 + }, + { + "epoch": 3.67, + "learning_rate": 3.1662661131745687e-06, + "loss": 1.1814, + "step": 16836 + }, + { + "epoch": 3.67, + "learning_rate": 3.1661568713130873e-06, + "loss": 1.9355, + "step": 16837 + }, + { + "epoch": 3.67, + "learning_rate": 3.1660476294516063e-06, + "loss": 2.0018, + "step": 16838 + }, + { + "epoch": 3.68, + "learning_rate": 3.165938387590125e-06, + "loss": 2.675, + "step": 16839 + }, + { + "epoch": 3.68, + "learning_rate": 3.165829145728643e-06, + "loss": 0.8588, + "step": 16840 + }, + { + "epoch": 3.68, + "learning_rate": 3.165719903867162e-06, + "loss": 0.9716, + "step": 16841 + }, + { + "epoch": 3.68, + "learning_rate": 3.1656106620056808e-06, + "loss": 1.9083, + "step": 16842 + }, + { + "epoch": 3.68, + "learning_rate": 3.1655014201442e-06, + "loss": 3.4129, + "step": 16843 + }, + { + "epoch": 3.68, + "learning_rate": 3.165392178282718e-06, + "loss": 1.6436, + "step": 16844 + }, + { + "epoch": 3.68, + "learning_rate": 3.1652829364212366e-06, + "loss": 2.4559, + "step": 16845 + }, + { + "epoch": 3.68, + "learning_rate": 3.1651736945597557e-06, + "loss": 2.2812, + "step": 16846 + }, + { + "epoch": 3.68, + "learning_rate": 3.1650644526982743e-06, + "loss": 1.6959, + "step": 16847 + }, + { + "epoch": 3.68, + "learning_rate": 3.1649552108367933e-06, + "loss": 1.3561, + "step": 16848 + }, + { + "epoch": 3.68, + "learning_rate": 3.1648459689753115e-06, + "loss": 2.7416, + "step": 16849 + }, + { + "epoch": 3.68, + "learning_rate": 3.16473672711383e-06, + "loss": 1.4208, + "step": 16850 + }, + { + "epoch": 3.68, + "learning_rate": 3.164627485252349e-06, + "loss": 1.8592, + "step": 16851 + }, + { + "epoch": 3.68, + "learning_rate": 3.1645182433908677e-06, + "loss": 1.6255, + "step": 16852 + }, + { + "epoch": 3.68, + "learning_rate": 3.1644090015293868e-06, + "loss": 1.8027, + "step": 16853 + }, + { + "epoch": 3.68, + "learning_rate": 3.164299759667905e-06, + "loss": 2.0514, + "step": 16854 + }, + { + "epoch": 3.68, + "learning_rate": 3.1641905178064236e-06, + "loss": 2.5897, + "step": 16855 + }, + { + "epoch": 3.68, + "learning_rate": 3.1640812759449426e-06, + "loss": 2.74, + "step": 16856 + }, + { + "epoch": 3.68, + "learning_rate": 3.1639720340834612e-06, + "loss": 2.6933, + "step": 16857 + }, + { + "epoch": 3.68, + "learning_rate": 3.1638627922219794e-06, + "loss": 2.4759, + "step": 16858 + }, + { + "epoch": 3.68, + "learning_rate": 3.1637535503604984e-06, + "loss": 0.7343, + "step": 16859 + }, + { + "epoch": 3.68, + "learning_rate": 3.163644308499017e-06, + "loss": 3.0291, + "step": 16860 + }, + { + "epoch": 3.68, + "learning_rate": 3.1635350666375357e-06, + "loss": 2.0597, + "step": 16861 + }, + { + "epoch": 3.68, + "learning_rate": 3.1634258247760547e-06, + "loss": 2.573, + "step": 16862 + }, + { + "epoch": 3.68, + "learning_rate": 3.163316582914573e-06, + "loss": 2.7007, + "step": 16863 + }, + { + "epoch": 3.68, + "learning_rate": 3.1632073410530915e-06, + "loss": 2.0276, + "step": 16864 + }, + { + "epoch": 3.68, + "learning_rate": 3.1630980991916105e-06, + "loss": 2.2372, + "step": 16865 + }, + { + "epoch": 3.68, + "learning_rate": 3.162988857330129e-06, + "loss": 1.7741, + "step": 16866 + }, + { + "epoch": 3.68, + "learning_rate": 3.162879615468648e-06, + "loss": 1.5074, + "step": 16867 + }, + { + "epoch": 3.68, + "learning_rate": 3.1627703736071664e-06, + "loss": 1.9608, + "step": 16868 + }, + { + "epoch": 3.68, + "learning_rate": 3.162661131745685e-06, + "loss": 2.405, + "step": 16869 + }, + { + "epoch": 3.68, + "learning_rate": 3.162551889884204e-06, + "loss": 2.3527, + "step": 16870 + }, + { + "epoch": 3.68, + "learning_rate": 3.1624426480227226e-06, + "loss": 2.3998, + "step": 16871 + }, + { + "epoch": 3.68, + "learning_rate": 3.1623334061612417e-06, + "loss": 1.9421, + "step": 16872 + }, + { + "epoch": 3.68, + "learning_rate": 3.16222416429976e-06, + "loss": 2.7144, + "step": 16873 + }, + { + "epoch": 3.68, + "learning_rate": 3.1621149224382785e-06, + "loss": 1.6018, + "step": 16874 + }, + { + "epoch": 3.68, + "learning_rate": 3.1620056805767975e-06, + "loss": 1.3501, + "step": 16875 + }, + { + "epoch": 3.68, + "learning_rate": 3.161896438715316e-06, + "loss": 1.6354, + "step": 16876 + }, + { + "epoch": 3.68, + "learning_rate": 3.1617871968538343e-06, + "loss": 1.909, + "step": 16877 + }, + { + "epoch": 3.68, + "learning_rate": 3.1616779549923533e-06, + "loss": 2.0106, + "step": 16878 + }, + { + "epoch": 3.68, + "learning_rate": 3.161568713130872e-06, + "loss": 0.7822, + "step": 16879 + }, + { + "epoch": 3.68, + "learning_rate": 3.161459471269391e-06, + "loss": 1.3499, + "step": 16880 + }, + { + "epoch": 3.68, + "learning_rate": 3.1613502294079096e-06, + "loss": 1.2036, + "step": 16881 + }, + { + "epoch": 3.68, + "learning_rate": 3.1612409875464278e-06, + "loss": 2.6625, + "step": 16882 + }, + { + "epoch": 3.68, + "learning_rate": 3.161131745684947e-06, + "loss": 1.1313, + "step": 16883 + }, + { + "epoch": 3.68, + "learning_rate": 3.1610225038234654e-06, + "loss": 1.8987, + "step": 16884 + }, + { + "epoch": 3.69, + "learning_rate": 3.160913261961984e-06, + "loss": 3.0322, + "step": 16885 + }, + { + "epoch": 3.69, + "learning_rate": 3.160804020100503e-06, + "loss": 1.4322, + "step": 16886 + }, + { + "epoch": 3.69, + "learning_rate": 3.1606947782390212e-06, + "loss": 1.933, + "step": 16887 + }, + { + "epoch": 3.69, + "learning_rate": 3.16058553637754e-06, + "loss": 1.2719, + "step": 16888 + }, + { + "epoch": 3.69, + "learning_rate": 3.160476294516059e-06, + "loss": 3.1004, + "step": 16889 + }, + { + "epoch": 3.69, + "learning_rate": 3.1603670526545775e-06, + "loss": 1.8774, + "step": 16890 + }, + { + "epoch": 3.69, + "learning_rate": 3.160257810793096e-06, + "loss": 3.0422, + "step": 16891 + }, + { + "epoch": 3.69, + "learning_rate": 3.1601485689316147e-06, + "loss": 1.3686, + "step": 16892 + }, + { + "epoch": 3.69, + "learning_rate": 3.1600393270701333e-06, + "loss": 2.3619, + "step": 16893 + }, + { + "epoch": 3.69, + "learning_rate": 3.1599300852086524e-06, + "loss": 2.1835, + "step": 16894 + }, + { + "epoch": 3.69, + "learning_rate": 3.159820843347171e-06, + "loss": 2.2718, + "step": 16895 + }, + { + "epoch": 3.69, + "learning_rate": 3.159711601485689e-06, + "loss": 1.6905, + "step": 16896 + }, + { + "epoch": 3.69, + "learning_rate": 3.159602359624208e-06, + "loss": 2.7959, + "step": 16897 + }, + { + "epoch": 3.69, + "learning_rate": 3.159493117762727e-06, + "loss": 1.8695, + "step": 16898 + }, + { + "epoch": 3.69, + "learning_rate": 3.159383875901246e-06, + "loss": 1.6797, + "step": 16899 + }, + { + "epoch": 3.69, + "learning_rate": 3.1592746340397645e-06, + "loss": 1.5112, + "step": 16900 + }, + { + "epoch": 3.69, + "learning_rate": 3.1591653921782826e-06, + "loss": 1.4278, + "step": 16901 + }, + { + "epoch": 3.69, + "learning_rate": 3.1590561503168017e-06, + "loss": 2.0731, + "step": 16902 + }, + { + "epoch": 3.69, + "learning_rate": 3.1589469084553203e-06, + "loss": 1.7152, + "step": 16903 + }, + { + "epoch": 3.69, + "learning_rate": 3.1588376665938393e-06, + "loss": 2.7482, + "step": 16904 + }, + { + "epoch": 3.69, + "learning_rate": 3.158728424732358e-06, + "loss": 1.1583, + "step": 16905 + }, + { + "epoch": 3.69, + "learning_rate": 3.158619182870876e-06, + "loss": 2.6765, + "step": 16906 + }, + { + "epoch": 3.69, + "learning_rate": 3.158509941009395e-06, + "loss": 2.489, + "step": 16907 + }, + { + "epoch": 3.69, + "learning_rate": 3.1584006991479138e-06, + "loss": 0.9946, + "step": 16908 + }, + { + "epoch": 3.69, + "learning_rate": 3.158291457286433e-06, + "loss": 1.7517, + "step": 16909 + }, + { + "epoch": 3.69, + "learning_rate": 3.158182215424951e-06, + "loss": 1.8045, + "step": 16910 + }, + { + "epoch": 3.69, + "learning_rate": 3.1580729735634696e-06, + "loss": 1.5837, + "step": 16911 + }, + { + "epoch": 3.69, + "learning_rate": 3.1579637317019886e-06, + "loss": 2.15, + "step": 16912 + }, + { + "epoch": 3.69, + "learning_rate": 3.1578544898405072e-06, + "loss": 1.2165, + "step": 16913 + }, + { + "epoch": 3.69, + "learning_rate": 3.157745247979026e-06, + "loss": 1.2139, + "step": 16914 + }, + { + "epoch": 3.69, + "learning_rate": 3.1576360061175445e-06, + "loss": 2.5786, + "step": 16915 + }, + { + "epoch": 3.69, + "learning_rate": 3.157526764256063e-06, + "loss": 2.5521, + "step": 16916 + }, + { + "epoch": 3.69, + "learning_rate": 3.1574175223945817e-06, + "loss": 1.4169, + "step": 16917 + }, + { + "epoch": 3.69, + "learning_rate": 3.1573082805331007e-06, + "loss": 1.1797, + "step": 16918 + }, + { + "epoch": 3.69, + "learning_rate": 3.1571990386716193e-06, + "loss": 1.2803, + "step": 16919 + }, + { + "epoch": 3.69, + "learning_rate": 3.1570897968101375e-06, + "loss": 2.1113, + "step": 16920 + }, + { + "epoch": 3.69, + "learning_rate": 3.1569805549486566e-06, + "loss": 1.8557, + "step": 16921 + }, + { + "epoch": 3.69, + "learning_rate": 3.156871313087175e-06, + "loss": 0.8461, + "step": 16922 + }, + { + "epoch": 3.69, + "learning_rate": 3.156762071225694e-06, + "loss": 0.6343, + "step": 16923 + }, + { + "epoch": 3.69, + "learning_rate": 3.1566528293642124e-06, + "loss": 2.0194, + "step": 16924 + }, + { + "epoch": 3.69, + "learning_rate": 3.156543587502731e-06, + "loss": 2.4684, + "step": 16925 + }, + { + "epoch": 3.69, + "learning_rate": 3.15643434564125e-06, + "loss": 3.4066, + "step": 16926 + }, + { + "epoch": 3.69, + "learning_rate": 3.1563251037797686e-06, + "loss": 2.3781, + "step": 16927 + }, + { + "epoch": 3.69, + "learning_rate": 3.1562158619182877e-06, + "loss": 2.8519, + "step": 16928 + }, + { + "epoch": 3.69, + "learning_rate": 3.156106620056806e-06, + "loss": 2.7711, + "step": 16929 + }, + { + "epoch": 3.69, + "learning_rate": 3.1559973781953245e-06, + "loss": 1.1116, + "step": 16930 + }, + { + "epoch": 3.7, + "learning_rate": 3.1558881363338435e-06, + "loss": 2.8426, + "step": 16931 + }, + { + "epoch": 3.7, + "learning_rate": 3.155778894472362e-06, + "loss": 1.7251, + "step": 16932 + }, + { + "epoch": 3.7, + "learning_rate": 3.155669652610881e-06, + "loss": 2.1855, + "step": 16933 + }, + { + "epoch": 3.7, + "learning_rate": 3.1555604107493994e-06, + "loss": 1.7437, + "step": 16934 + }, + { + "epoch": 3.7, + "learning_rate": 3.155451168887918e-06, + "loss": 2.7025, + "step": 16935 + }, + { + "epoch": 3.7, + "learning_rate": 3.155341927026437e-06, + "loss": 0.7355, + "step": 16936 + }, + { + "epoch": 3.7, + "learning_rate": 3.1552326851649556e-06, + "loss": 0.7058, + "step": 16937 + }, + { + "epoch": 3.7, + "learning_rate": 3.1551234433034742e-06, + "loss": 0.9275, + "step": 16938 + }, + { + "epoch": 3.7, + "learning_rate": 3.155014201441993e-06, + "loss": 3.0147, + "step": 16939 + }, + { + "epoch": 3.7, + "learning_rate": 3.1549049595805114e-06, + "loss": 1.4255, + "step": 16940 + }, + { + "epoch": 3.7, + "learning_rate": 3.15479571771903e-06, + "loss": 1.3206, + "step": 16941 + }, + { + "epoch": 3.7, + "learning_rate": 3.154686475857549e-06, + "loss": 1.9882, + "step": 16942 + }, + { + "epoch": 3.7, + "learning_rate": 3.1545772339960673e-06, + "loss": 1.3283, + "step": 16943 + }, + { + "epoch": 3.7, + "learning_rate": 3.1544679921345863e-06, + "loss": 2.2046, + "step": 16944 + }, + { + "epoch": 3.7, + "learning_rate": 3.154358750273105e-06, + "loss": 1.2475, + "step": 16945 + }, + { + "epoch": 3.7, + "learning_rate": 3.1542495084116235e-06, + "loss": 1.7634, + "step": 16946 + }, + { + "epoch": 3.7, + "learning_rate": 3.1541402665501426e-06, + "loss": 2.3273, + "step": 16947 + }, + { + "epoch": 3.7, + "learning_rate": 3.1540310246886608e-06, + "loss": 2.1342, + "step": 16948 + }, + { + "epoch": 3.7, + "learning_rate": 3.1539217828271794e-06, + "loss": 1.7628, + "step": 16949 + }, + { + "epoch": 3.7, + "learning_rate": 3.1538125409656984e-06, + "loss": 3.1184, + "step": 16950 + }, + { + "epoch": 3.7, + "learning_rate": 3.153703299104217e-06, + "loss": 1.876, + "step": 16951 + }, + { + "epoch": 3.7, + "learning_rate": 3.153594057242736e-06, + "loss": 2.0043, + "step": 16952 + }, + { + "epoch": 3.7, + "learning_rate": 3.1534848153812542e-06, + "loss": 1.0989, + "step": 16953 + }, + { + "epoch": 3.7, + "learning_rate": 3.153375573519773e-06, + "loss": 3.7018, + "step": 16954 + }, + { + "epoch": 3.7, + "learning_rate": 3.153266331658292e-06, + "loss": 1.6801, + "step": 16955 + }, + { + "epoch": 3.7, + "learning_rate": 3.1531570897968105e-06, + "loss": 2.5039, + "step": 16956 + }, + { + "epoch": 3.7, + "learning_rate": 3.1530478479353287e-06, + "loss": 1.9316, + "step": 16957 + }, + { + "epoch": 3.7, + "learning_rate": 3.1529386060738477e-06, + "loss": 1.9175, + "step": 16958 + }, + { + "epoch": 3.7, + "learning_rate": 3.1528293642123663e-06, + "loss": 1.0586, + "step": 16959 + }, + { + "epoch": 3.7, + "learning_rate": 3.1527201223508854e-06, + "loss": 1.1417, + "step": 16960 + }, + { + "epoch": 3.7, + "learning_rate": 3.152610880489404e-06, + "loss": 3.9546, + "step": 16961 + }, + { + "epoch": 3.7, + "learning_rate": 3.152501638627922e-06, + "loss": 1.8914, + "step": 16962 + }, + { + "epoch": 3.7, + "learning_rate": 3.152392396766441e-06, + "loss": 2.042, + "step": 16963 + }, + { + "epoch": 3.7, + "learning_rate": 3.15228315490496e-06, + "loss": 0.4231, + "step": 16964 + }, + { + "epoch": 3.7, + "learning_rate": 3.152173913043479e-06, + "loss": 2.333, + "step": 16965 + }, + { + "epoch": 3.7, + "learning_rate": 3.1520646711819974e-06, + "loss": 3.0339, + "step": 16966 + }, + { + "epoch": 3.7, + "learning_rate": 3.1519554293205156e-06, + "loss": 2.0442, + "step": 16967 + }, + { + "epoch": 3.7, + "learning_rate": 3.1518461874590347e-06, + "loss": 2.1775, + "step": 16968 + }, + { + "epoch": 3.7, + "learning_rate": 3.1517369455975533e-06, + "loss": 1.4124, + "step": 16969 + }, + { + "epoch": 3.7, + "learning_rate": 3.151627703736072e-06, + "loss": 3.4008, + "step": 16970 + }, + { + "epoch": 3.7, + "learning_rate": 3.151518461874591e-06, + "loss": 1.3989, + "step": 16971 + }, + { + "epoch": 3.7, + "learning_rate": 3.151409220013109e-06, + "loss": 2.1662, + "step": 16972 + }, + { + "epoch": 3.7, + "learning_rate": 3.1512999781516277e-06, + "loss": 2.7792, + "step": 16973 + }, + { + "epoch": 3.7, + "learning_rate": 3.1511907362901468e-06, + "loss": 1.6077, + "step": 16974 + }, + { + "epoch": 3.7, + "learning_rate": 3.1510814944286654e-06, + "loss": 1.2455, + "step": 16975 + }, + { + "epoch": 3.7, + "learning_rate": 3.1509722525671836e-06, + "loss": 2.1575, + "step": 16976 + }, + { + "epoch": 3.71, + "learning_rate": 3.1508630107057026e-06, + "loss": 2.7033, + "step": 16977 + }, + { + "epoch": 3.71, + "learning_rate": 3.150753768844221e-06, + "loss": 2.5951, + "step": 16978 + }, + { + "epoch": 3.71, + "learning_rate": 3.1506445269827402e-06, + "loss": 1.4026, + "step": 16979 + }, + { + "epoch": 3.71, + "learning_rate": 3.150535285121259e-06, + "loss": 3.4492, + "step": 16980 + }, + { + "epoch": 3.71, + "learning_rate": 3.150426043259777e-06, + "loss": 0.3554, + "step": 16981 + }, + { + "epoch": 3.71, + "learning_rate": 3.150316801398296e-06, + "loss": 2.2175, + "step": 16982 + }, + { + "epoch": 3.71, + "learning_rate": 3.1502075595368147e-06, + "loss": 4.179, + "step": 16983 + }, + { + "epoch": 3.71, + "learning_rate": 3.1500983176753337e-06, + "loss": 1.8503, + "step": 16984 + }, + { + "epoch": 3.71, + "learning_rate": 3.1499890758138523e-06, + "loss": 1.6928, + "step": 16985 + }, + { + "epoch": 3.71, + "learning_rate": 3.1498798339523705e-06, + "loss": 1.21, + "step": 16986 + }, + { + "epoch": 3.71, + "learning_rate": 3.1497705920908895e-06, + "loss": 0.8285, + "step": 16987 + }, + { + "epoch": 3.71, + "learning_rate": 3.149661350229408e-06, + "loss": 1.6094, + "step": 16988 + }, + { + "epoch": 3.71, + "learning_rate": 3.149552108367927e-06, + "loss": 0.7663, + "step": 16989 + }, + { + "epoch": 3.71, + "learning_rate": 3.1494428665064454e-06, + "loss": 2.1804, + "step": 16990 + }, + { + "epoch": 3.71, + "learning_rate": 3.149333624644964e-06, + "loss": 0.8157, + "step": 16991 + }, + { + "epoch": 3.71, + "learning_rate": 3.149224382783483e-06, + "loss": 1.0936, + "step": 16992 + }, + { + "epoch": 3.71, + "learning_rate": 3.1491151409220016e-06, + "loss": 2.8479, + "step": 16993 + }, + { + "epoch": 3.71, + "learning_rate": 3.1490058990605202e-06, + "loss": 2.39, + "step": 16994 + }, + { + "epoch": 3.71, + "learning_rate": 3.148896657199039e-06, + "loss": 1.7928, + "step": 16995 + }, + { + "epoch": 3.71, + "learning_rate": 3.1487874153375575e-06, + "loss": 2.0906, + "step": 16996 + }, + { + "epoch": 3.71, + "learning_rate": 3.1486781734760765e-06, + "loss": 2.2953, + "step": 16997 + }, + { + "epoch": 3.71, + "learning_rate": 3.148568931614595e-06, + "loss": 1.3282, + "step": 16998 + }, + { + "epoch": 3.71, + "learning_rate": 3.1484596897531137e-06, + "loss": 1.63, + "step": 16999 + }, + { + "epoch": 3.71, + "learning_rate": 3.1483504478916323e-06, + "loss": 0.807, + "step": 17000 + }, + { + "epoch": 3.71, + "learning_rate": 3.148241206030151e-06, + "loss": 2.7652, + "step": 17001 + }, + { + "epoch": 3.71, + "learning_rate": 3.1481319641686696e-06, + "loss": 1.6845, + "step": 17002 + }, + { + "epoch": 3.71, + "learning_rate": 3.1480227223071886e-06, + "loss": 3.2485, + "step": 17003 + }, + { + "epoch": 3.71, + "learning_rate": 3.147913480445707e-06, + "loss": 2.9107, + "step": 17004 + }, + { + "epoch": 3.71, + "learning_rate": 3.1478042385842254e-06, + "loss": 2.4059, + "step": 17005 + }, + { + "epoch": 3.71, + "learning_rate": 3.1476949967227444e-06, + "loss": 2.2499, + "step": 17006 + }, + { + "epoch": 3.71, + "learning_rate": 3.147585754861263e-06, + "loss": 2.4744, + "step": 17007 + }, + { + "epoch": 3.71, + "learning_rate": 3.147476512999782e-06, + "loss": 1.3196, + "step": 17008 + }, + { + "epoch": 3.71, + "learning_rate": 3.1473672711383003e-06, + "loss": 1.7309, + "step": 17009 + }, + { + "epoch": 3.71, + "learning_rate": 3.147258029276819e-06, + "loss": 1.9737, + "step": 17010 + }, + { + "epoch": 3.71, + "learning_rate": 3.147148787415338e-06, + "loss": 3.3801, + "step": 17011 + }, + { + "epoch": 3.71, + "learning_rate": 3.1470395455538565e-06, + "loss": 3.3592, + "step": 17012 + }, + { + "epoch": 3.71, + "learning_rate": 3.1469303036923755e-06, + "loss": 3.7565, + "step": 17013 + }, + { + "epoch": 3.71, + "learning_rate": 3.1468210618308937e-06, + "loss": 1.5432, + "step": 17014 + }, + { + "epoch": 3.71, + "learning_rate": 3.1467118199694123e-06, + "loss": 3.1028, + "step": 17015 + }, + { + "epoch": 3.71, + "learning_rate": 3.1466025781079314e-06, + "loss": 1.8544, + "step": 17016 + }, + { + "epoch": 3.71, + "learning_rate": 3.14649333624645e-06, + "loss": 3.5064, + "step": 17017 + }, + { + "epoch": 3.71, + "learning_rate": 3.146384094384969e-06, + "loss": 2.2935, + "step": 17018 + }, + { + "epoch": 3.71, + "learning_rate": 3.1462748525234872e-06, + "loss": 2.0449, + "step": 17019 + }, + { + "epoch": 3.71, + "learning_rate": 3.146165610662006e-06, + "loss": 1.6785, + "step": 17020 + }, + { + "epoch": 3.71, + "learning_rate": 3.146056368800525e-06, + "loss": 2.215, + "step": 17021 + }, + { + "epoch": 3.71, + "learning_rate": 3.1459471269390435e-06, + "loss": 2.2885, + "step": 17022 + }, + { + "epoch": 3.72, + "learning_rate": 3.1458378850775617e-06, + "loss": 2.1996, + "step": 17023 + }, + { + "epoch": 3.72, + "learning_rate": 3.1457286432160807e-06, + "loss": 1.3517, + "step": 17024 + }, + { + "epoch": 3.72, + "learning_rate": 3.1456194013545993e-06, + "loss": 2.6456, + "step": 17025 + }, + { + "epoch": 3.72, + "learning_rate": 3.145510159493118e-06, + "loss": 1.1792, + "step": 17026 + }, + { + "epoch": 3.72, + "learning_rate": 3.145400917631637e-06, + "loss": 2.2932, + "step": 17027 + }, + { + "epoch": 3.72, + "learning_rate": 3.145291675770155e-06, + "loss": 1.7336, + "step": 17028 + }, + { + "epoch": 3.72, + "learning_rate": 3.1451824339086737e-06, + "loss": 1.3257, + "step": 17029 + }, + { + "epoch": 3.72, + "learning_rate": 3.1450731920471928e-06, + "loss": 1.7423, + "step": 17030 + }, + { + "epoch": 3.72, + "learning_rate": 3.1449639501857114e-06, + "loss": 1.331, + "step": 17031 + }, + { + "epoch": 3.72, + "learning_rate": 3.1448547083242304e-06, + "loss": 1.6881, + "step": 17032 + }, + { + "epoch": 3.72, + "learning_rate": 3.1447454664627486e-06, + "loss": 2.107, + "step": 17033 + }, + { + "epoch": 3.72, + "learning_rate": 3.1446362246012672e-06, + "loss": 2.4928, + "step": 17034 + }, + { + "epoch": 3.72, + "learning_rate": 3.1445269827397863e-06, + "loss": 1.8922, + "step": 17035 + }, + { + "epoch": 3.72, + "learning_rate": 3.144417740878305e-06, + "loss": 2.418, + "step": 17036 + }, + { + "epoch": 3.72, + "learning_rate": 3.144308499016824e-06, + "loss": 1.8497, + "step": 17037 + }, + { + "epoch": 3.72, + "learning_rate": 3.144199257155342e-06, + "loss": 0.8425, + "step": 17038 + }, + { + "epoch": 3.72, + "learning_rate": 3.1440900152938607e-06, + "loss": 1.808, + "step": 17039 + }, + { + "epoch": 3.72, + "learning_rate": 3.1439807734323797e-06, + "loss": 1.8274, + "step": 17040 + }, + { + "epoch": 3.72, + "learning_rate": 3.1438715315708983e-06, + "loss": 1.5355, + "step": 17041 + }, + { + "epoch": 3.72, + "learning_rate": 3.1437622897094165e-06, + "loss": 2.0271, + "step": 17042 + }, + { + "epoch": 3.72, + "learning_rate": 3.1436530478479356e-06, + "loss": 0.9755, + "step": 17043 + }, + { + "epoch": 3.72, + "learning_rate": 3.143543805986454e-06, + "loss": 2.3731, + "step": 17044 + }, + { + "epoch": 3.72, + "learning_rate": 3.1434345641249732e-06, + "loss": 1.0098, + "step": 17045 + }, + { + "epoch": 3.72, + "learning_rate": 3.143325322263492e-06, + "loss": 2.5741, + "step": 17046 + }, + { + "epoch": 3.72, + "learning_rate": 3.14321608040201e-06, + "loss": 1.728, + "step": 17047 + }, + { + "epoch": 3.72, + "learning_rate": 3.143106838540529e-06, + "loss": 1.9483, + "step": 17048 + }, + { + "epoch": 3.72, + "learning_rate": 3.1429975966790477e-06, + "loss": 1.4704, + "step": 17049 + }, + { + "epoch": 3.72, + "learning_rate": 3.1428883548175667e-06, + "loss": 1.1665, + "step": 17050 + }, + { + "epoch": 3.72, + "learning_rate": 3.1427791129560853e-06, + "loss": 0.6857, + "step": 17051 + }, + { + "epoch": 3.72, + "learning_rate": 3.1426698710946035e-06, + "loss": 2.9273, + "step": 17052 + }, + { + "epoch": 3.72, + "learning_rate": 3.1425606292331225e-06, + "loss": 2.3612, + "step": 17053 + }, + { + "epoch": 3.72, + "learning_rate": 3.142451387371641e-06, + "loss": 2.17, + "step": 17054 + }, + { + "epoch": 3.72, + "learning_rate": 3.1423421455101597e-06, + "loss": 1.8028, + "step": 17055 + }, + { + "epoch": 3.72, + "learning_rate": 3.1422329036486784e-06, + "loss": 3.6235, + "step": 17056 + }, + { + "epoch": 3.72, + "learning_rate": 3.142123661787197e-06, + "loss": 1.4556, + "step": 17057 + }, + { + "epoch": 3.72, + "learning_rate": 3.1420144199257156e-06, + "loss": 2.5838, + "step": 17058 + }, + { + "epoch": 3.72, + "learning_rate": 3.1419051780642346e-06, + "loss": 1.3799, + "step": 17059 + }, + { + "epoch": 3.72, + "learning_rate": 3.1417959362027532e-06, + "loss": 1.2552, + "step": 17060 + }, + { + "epoch": 3.72, + "learning_rate": 3.1416866943412714e-06, + "loss": 0.6681, + "step": 17061 + }, + { + "epoch": 3.72, + "learning_rate": 3.1415774524797905e-06, + "loss": 1.7044, + "step": 17062 + }, + { + "epoch": 3.72, + "learning_rate": 3.141468210618309e-06, + "loss": 0.7319, + "step": 17063 + }, + { + "epoch": 3.72, + "learning_rate": 3.141358968756828e-06, + "loss": 2.0688, + "step": 17064 + }, + { + "epoch": 3.72, + "learning_rate": 3.1412497268953467e-06, + "loss": 1.251, + "step": 17065 + }, + { + "epoch": 3.72, + "learning_rate": 3.141140485033865e-06, + "loss": 2.4504, + "step": 17066 + }, + { + "epoch": 3.72, + "learning_rate": 3.141031243172384e-06, + "loss": 2.5264, + "step": 17067 + }, + { + "epoch": 3.73, + "learning_rate": 3.1409220013109025e-06, + "loss": 2.022, + "step": 17068 + }, + { + "epoch": 3.73, + "learning_rate": 3.1408127594494216e-06, + "loss": 1.4531, + "step": 17069 + }, + { + "epoch": 3.73, + "learning_rate": 3.14070351758794e-06, + "loss": 0.9543, + "step": 17070 + }, + { + "epoch": 3.73, + "learning_rate": 3.1405942757264584e-06, + "loss": 1.5941, + "step": 17071 + }, + { + "epoch": 3.73, + "learning_rate": 3.1404850338649774e-06, + "loss": 1.9977, + "step": 17072 + }, + { + "epoch": 3.73, + "learning_rate": 3.140375792003496e-06, + "loss": 0.7221, + "step": 17073 + }, + { + "epoch": 3.73, + "learning_rate": 3.140266550142015e-06, + "loss": 0.9192, + "step": 17074 + }, + { + "epoch": 3.73, + "learning_rate": 3.1401573082805332e-06, + "loss": 1.8043, + "step": 17075 + }, + { + "epoch": 3.73, + "learning_rate": 3.140048066419052e-06, + "loss": 1.1335, + "step": 17076 + }, + { + "epoch": 3.73, + "learning_rate": 3.139938824557571e-06, + "loss": 2.5466, + "step": 17077 + }, + { + "epoch": 3.73, + "learning_rate": 3.1398295826960895e-06, + "loss": 3.2212, + "step": 17078 + }, + { + "epoch": 3.73, + "learning_rate": 3.139720340834608e-06, + "loss": 2.6711, + "step": 17079 + }, + { + "epoch": 3.73, + "learning_rate": 3.1396110989731267e-06, + "loss": 1.2108, + "step": 17080 + }, + { + "epoch": 3.73, + "learning_rate": 3.1395018571116453e-06, + "loss": 1.4001, + "step": 17081 + }, + { + "epoch": 3.73, + "learning_rate": 3.139392615250164e-06, + "loss": 0.7982, + "step": 17082 + }, + { + "epoch": 3.73, + "learning_rate": 3.139283373388683e-06, + "loss": 1.844, + "step": 17083 + }, + { + "epoch": 3.73, + "learning_rate": 3.1391741315272016e-06, + "loss": 2.8102, + "step": 17084 + }, + { + "epoch": 3.73, + "learning_rate": 3.1390648896657198e-06, + "loss": 2.4973, + "step": 17085 + }, + { + "epoch": 3.73, + "learning_rate": 3.138955647804239e-06, + "loss": 1.286, + "step": 17086 + }, + { + "epoch": 3.73, + "learning_rate": 3.1388464059427574e-06, + "loss": 1.2842, + "step": 17087 + }, + { + "epoch": 3.73, + "learning_rate": 3.1387371640812765e-06, + "loss": 2.029, + "step": 17088 + }, + { + "epoch": 3.73, + "learning_rate": 3.1386279222197946e-06, + "loss": 0.8079, + "step": 17089 + }, + { + "epoch": 3.73, + "learning_rate": 3.1385186803583133e-06, + "loss": 2.9706, + "step": 17090 + }, + { + "epoch": 3.73, + "learning_rate": 3.1384094384968323e-06, + "loss": 2.2982, + "step": 17091 + }, + { + "epoch": 3.73, + "learning_rate": 3.138300196635351e-06, + "loss": 1.4868, + "step": 17092 + }, + { + "epoch": 3.73, + "learning_rate": 3.13819095477387e-06, + "loss": 1.1775, + "step": 17093 + }, + { + "epoch": 3.73, + "learning_rate": 3.138081712912388e-06, + "loss": 1.5496, + "step": 17094 + }, + { + "epoch": 3.73, + "learning_rate": 3.1379724710509067e-06, + "loss": 2.0892, + "step": 17095 + }, + { + "epoch": 3.73, + "learning_rate": 3.1378632291894258e-06, + "loss": 2.6185, + "step": 17096 + }, + { + "epoch": 3.73, + "learning_rate": 3.1377539873279444e-06, + "loss": 2.58, + "step": 17097 + }, + { + "epoch": 3.73, + "learning_rate": 3.1376447454664634e-06, + "loss": 0.8974, + "step": 17098 + }, + { + "epoch": 3.73, + "learning_rate": 3.1375355036049816e-06, + "loss": 1.9118, + "step": 17099 + }, + { + "epoch": 3.73, + "learning_rate": 3.1374262617435e-06, + "loss": 1.7845, + "step": 17100 + }, + { + "epoch": 3.73, + "learning_rate": 3.1373170198820192e-06, + "loss": 1.7035, + "step": 17101 + }, + { + "epoch": 3.73, + "learning_rate": 3.137207778020538e-06, + "loss": 2.4498, + "step": 17102 + }, + { + "epoch": 3.73, + "learning_rate": 3.137098536159057e-06, + "loss": 1.9109, + "step": 17103 + }, + { + "epoch": 3.73, + "learning_rate": 3.136989294297575e-06, + "loss": 1.1638, + "step": 17104 + }, + { + "epoch": 3.73, + "learning_rate": 3.1368800524360937e-06, + "loss": 1.6251, + "step": 17105 + }, + { + "epoch": 3.73, + "learning_rate": 3.1367708105746127e-06, + "loss": 1.7642, + "step": 17106 + }, + { + "epoch": 3.73, + "learning_rate": 3.1366615687131313e-06, + "loss": 2.7854, + "step": 17107 + }, + { + "epoch": 3.73, + "learning_rate": 3.1365523268516495e-06, + "loss": 1.6078, + "step": 17108 + }, + { + "epoch": 3.73, + "learning_rate": 3.1364430849901686e-06, + "loss": 1.3859, + "step": 17109 + }, + { + "epoch": 3.73, + "learning_rate": 3.136333843128687e-06, + "loss": 1.1702, + "step": 17110 + }, + { + "epoch": 3.73, + "learning_rate": 3.1362246012672058e-06, + "loss": 0.6959, + "step": 17111 + }, + { + "epoch": 3.73, + "learning_rate": 3.136115359405725e-06, + "loss": 1.1185, + "step": 17112 + }, + { + "epoch": 3.73, + "learning_rate": 3.136006117544243e-06, + "loss": 3.0666, + "step": 17113 + }, + { + "epoch": 3.74, + "learning_rate": 3.1358968756827616e-06, + "loss": 2.6734, + "step": 17114 + }, + { + "epoch": 3.74, + "learning_rate": 3.1357876338212806e-06, + "loss": 2.0461, + "step": 17115 + }, + { + "epoch": 3.74, + "learning_rate": 3.1356783919597993e-06, + "loss": 1.9067, + "step": 17116 + }, + { + "epoch": 3.74, + "learning_rate": 3.1355691500983183e-06, + "loss": 1.0073, + "step": 17117 + }, + { + "epoch": 3.74, + "learning_rate": 3.1354599082368365e-06, + "loss": 2.6659, + "step": 17118 + }, + { + "epoch": 3.74, + "learning_rate": 3.135350666375355e-06, + "loss": 1.7386, + "step": 17119 + }, + { + "epoch": 3.74, + "learning_rate": 3.135241424513874e-06, + "loss": 1.5255, + "step": 17120 + }, + { + "epoch": 3.74, + "learning_rate": 3.1351321826523927e-06, + "loss": 2.3645, + "step": 17121 + }, + { + "epoch": 3.74, + "learning_rate": 3.135022940790911e-06, + "loss": 1.165, + "step": 17122 + }, + { + "epoch": 3.74, + "learning_rate": 3.13491369892943e-06, + "loss": 3.017, + "step": 17123 + }, + { + "epoch": 3.74, + "learning_rate": 3.1348044570679486e-06, + "loss": 0.997, + "step": 17124 + }, + { + "epoch": 3.74, + "learning_rate": 3.1346952152064676e-06, + "loss": 1.5512, + "step": 17125 + }, + { + "epoch": 3.74, + "learning_rate": 3.1345859733449862e-06, + "loss": 2.1427, + "step": 17126 + }, + { + "epoch": 3.74, + "learning_rate": 3.1344767314835044e-06, + "loss": 1.9469, + "step": 17127 + }, + { + "epoch": 3.74, + "learning_rate": 3.1343674896220234e-06, + "loss": 1.2803, + "step": 17128 + }, + { + "epoch": 3.74, + "learning_rate": 3.134258247760542e-06, + "loss": 1.6015, + "step": 17129 + }, + { + "epoch": 3.74, + "learning_rate": 3.134149005899061e-06, + "loss": 3.2038, + "step": 17130 + }, + { + "epoch": 3.74, + "learning_rate": 3.1340397640375797e-06, + "loss": 2.1103, + "step": 17131 + }, + { + "epoch": 3.74, + "learning_rate": 3.133930522176098e-06, + "loss": 1.9135, + "step": 17132 + }, + { + "epoch": 3.74, + "learning_rate": 3.133821280314617e-06, + "loss": 2.2218, + "step": 17133 + }, + { + "epoch": 3.74, + "learning_rate": 3.1337120384531355e-06, + "loss": 1.7718, + "step": 17134 + }, + { + "epoch": 3.74, + "learning_rate": 3.133602796591654e-06, + "loss": 2.3699, + "step": 17135 + }, + { + "epoch": 3.74, + "learning_rate": 3.1334935547301727e-06, + "loss": 1.7812, + "step": 17136 + }, + { + "epoch": 3.74, + "learning_rate": 3.1333843128686914e-06, + "loss": 1.4066, + "step": 17137 + }, + { + "epoch": 3.74, + "learning_rate": 3.13327507100721e-06, + "loss": 1.9269, + "step": 17138 + }, + { + "epoch": 3.74, + "learning_rate": 3.133165829145729e-06, + "loss": 2.0307, + "step": 17139 + }, + { + "epoch": 3.74, + "learning_rate": 3.1330565872842476e-06, + "loss": 2.1011, + "step": 17140 + }, + { + "epoch": 3.74, + "learning_rate": 3.1329473454227662e-06, + "loss": 2.665, + "step": 17141 + }, + { + "epoch": 3.74, + "learning_rate": 3.132838103561285e-06, + "loss": 1.9092, + "step": 17142 + }, + { + "epoch": 3.74, + "learning_rate": 3.1327288616998034e-06, + "loss": 1.3155, + "step": 17143 + }, + { + "epoch": 3.74, + "learning_rate": 3.1326196198383225e-06, + "loss": 1.4775, + "step": 17144 + }, + { + "epoch": 3.74, + "learning_rate": 3.132510377976841e-06, + "loss": 2.6358, + "step": 17145 + }, + { + "epoch": 3.74, + "learning_rate": 3.1324011361153593e-06, + "loss": 1.9364, + "step": 17146 + }, + { + "epoch": 3.74, + "learning_rate": 3.1322918942538783e-06, + "loss": 3.9316, + "step": 17147 + }, + { + "epoch": 3.74, + "learning_rate": 3.132182652392397e-06, + "loss": 2.0248, + "step": 17148 + }, + { + "epoch": 3.74, + "learning_rate": 3.132073410530916e-06, + "loss": 2.2199, + "step": 17149 + }, + { + "epoch": 3.74, + "learning_rate": 3.1319641686694346e-06, + "loss": 1.2671, + "step": 17150 + }, + { + "epoch": 3.74, + "learning_rate": 3.1318549268079528e-06, + "loss": 1.8652, + "step": 17151 + }, + { + "epoch": 3.74, + "learning_rate": 3.131745684946472e-06, + "loss": 2.5995, + "step": 17152 + }, + { + "epoch": 3.74, + "learning_rate": 3.1316364430849904e-06, + "loss": 1.6074, + "step": 17153 + }, + { + "epoch": 3.74, + "learning_rate": 3.1315272012235094e-06, + "loss": 1.0594, + "step": 17154 + }, + { + "epoch": 3.74, + "learning_rate": 3.1314179593620276e-06, + "loss": 3.3899, + "step": 17155 + }, + { + "epoch": 3.74, + "learning_rate": 3.1313087175005462e-06, + "loss": 2.6882, + "step": 17156 + }, + { + "epoch": 3.74, + "learning_rate": 3.1311994756390653e-06, + "loss": 1.5155, + "step": 17157 + }, + { + "epoch": 3.74, + "learning_rate": 3.131090233777584e-06, + "loss": 1.3339, + "step": 17158 + }, + { + "epoch": 3.74, + "learning_rate": 3.130980991916103e-06, + "loss": 1.8792, + "step": 17159 + }, + { + "epoch": 3.75, + "learning_rate": 3.130871750054621e-06, + "loss": 2.5981, + "step": 17160 + }, + { + "epoch": 3.75, + "learning_rate": 3.1307625081931397e-06, + "loss": 2.0823, + "step": 17161 + }, + { + "epoch": 3.75, + "learning_rate": 3.1306532663316587e-06, + "loss": 2.7274, + "step": 17162 + }, + { + "epoch": 3.75, + "learning_rate": 3.1305440244701774e-06, + "loss": 2.6826, + "step": 17163 + }, + { + "epoch": 3.75, + "learning_rate": 3.130434782608696e-06, + "loss": 1.3333, + "step": 17164 + }, + { + "epoch": 3.75, + "learning_rate": 3.1303255407472146e-06, + "loss": 1.1725, + "step": 17165 + }, + { + "epoch": 3.75, + "learning_rate": 3.130216298885733e-06, + "loss": 2.4507, + "step": 17166 + }, + { + "epoch": 3.75, + "learning_rate": 3.130107057024252e-06, + "loss": 1.6454, + "step": 17167 + }, + { + "epoch": 3.75, + "learning_rate": 3.129997815162771e-06, + "loss": 1.9231, + "step": 17168 + }, + { + "epoch": 3.75, + "learning_rate": 3.129888573301289e-06, + "loss": 2.02, + "step": 17169 + }, + { + "epoch": 3.75, + "learning_rate": 3.1297793314398076e-06, + "loss": 3.1917, + "step": 17170 + }, + { + "epoch": 3.75, + "learning_rate": 3.1296700895783267e-06, + "loss": 2.4169, + "step": 17171 + }, + { + "epoch": 3.75, + "learning_rate": 3.1295608477168453e-06, + "loss": 1.7656, + "step": 17172 + }, + { + "epoch": 3.75, + "learning_rate": 3.1294516058553643e-06, + "loss": 3.7, + "step": 17173 + }, + { + "epoch": 3.75, + "learning_rate": 3.1293423639938825e-06, + "loss": 1.245, + "step": 17174 + }, + { + "epoch": 3.75, + "learning_rate": 3.129233122132401e-06, + "loss": 1.91, + "step": 17175 + }, + { + "epoch": 3.75, + "learning_rate": 3.12912388027092e-06, + "loss": 1.2699, + "step": 17176 + }, + { + "epoch": 3.75, + "learning_rate": 3.1290146384094388e-06, + "loss": 1.5722, + "step": 17177 + }, + { + "epoch": 3.75, + "learning_rate": 3.128905396547958e-06, + "loss": 2.1752, + "step": 17178 + }, + { + "epoch": 3.75, + "learning_rate": 3.128796154686476e-06, + "loss": 1.7778, + "step": 17179 + }, + { + "epoch": 3.75, + "learning_rate": 3.1286869128249946e-06, + "loss": 2.4841, + "step": 17180 + }, + { + "epoch": 3.75, + "learning_rate": 3.1285776709635136e-06, + "loss": 1.6278, + "step": 17181 + }, + { + "epoch": 3.75, + "learning_rate": 3.1284684291020322e-06, + "loss": 1.7029, + "step": 17182 + }, + { + "epoch": 3.75, + "learning_rate": 3.1283591872405513e-06, + "loss": 0.6969, + "step": 17183 + }, + { + "epoch": 3.75, + "learning_rate": 3.1282499453790695e-06, + "loss": 0.9671, + "step": 17184 + }, + { + "epoch": 3.75, + "learning_rate": 3.128140703517588e-06, + "loss": 2.3591, + "step": 17185 + }, + { + "epoch": 3.75, + "learning_rate": 3.128031461656107e-06, + "loss": 2.325, + "step": 17186 + }, + { + "epoch": 3.75, + "learning_rate": 3.1279222197946257e-06, + "loss": 1.6978, + "step": 17187 + }, + { + "epoch": 3.75, + "learning_rate": 3.127812977933144e-06, + "loss": 2.2943, + "step": 17188 + }, + { + "epoch": 3.75, + "learning_rate": 3.127703736071663e-06, + "loss": 0.5658, + "step": 17189 + }, + { + "epoch": 3.75, + "learning_rate": 3.1275944942101816e-06, + "loss": 1.9046, + "step": 17190 + }, + { + "epoch": 3.75, + "learning_rate": 3.1274852523487006e-06, + "loss": 2.1396, + "step": 17191 + }, + { + "epoch": 3.75, + "learning_rate": 3.127376010487219e-06, + "loss": 2.3892, + "step": 17192 + }, + { + "epoch": 3.75, + "learning_rate": 3.1272667686257374e-06, + "loss": 0.6176, + "step": 17193 + }, + { + "epoch": 3.75, + "learning_rate": 3.1271575267642564e-06, + "loss": 2.1615, + "step": 17194 + }, + { + "epoch": 3.75, + "learning_rate": 3.127048284902775e-06, + "loss": 2.3285, + "step": 17195 + }, + { + "epoch": 3.75, + "learning_rate": 3.1269390430412936e-06, + "loss": 1.8822, + "step": 17196 + }, + { + "epoch": 3.75, + "learning_rate": 3.1268298011798127e-06, + "loss": 0.8906, + "step": 17197 + }, + { + "epoch": 3.75, + "learning_rate": 3.126720559318331e-06, + "loss": 3.2228, + "step": 17198 + }, + { + "epoch": 3.75, + "learning_rate": 3.1266113174568495e-06, + "loss": 2.0159, + "step": 17199 + }, + { + "epoch": 3.75, + "learning_rate": 3.1265020755953685e-06, + "loss": 2.3062, + "step": 17200 + }, + { + "epoch": 3.75, + "learning_rate": 3.126392833733887e-06, + "loss": 1.5523, + "step": 17201 + }, + { + "epoch": 3.75, + "learning_rate": 3.1262835918724053e-06, + "loss": 2.4257, + "step": 17202 + }, + { + "epoch": 3.75, + "learning_rate": 3.1261743500109243e-06, + "loss": 2.9549, + "step": 17203 + }, + { + "epoch": 3.75, + "learning_rate": 3.126065108149443e-06, + "loss": 2.0919, + "step": 17204 + }, + { + "epoch": 3.75, + "learning_rate": 3.125955866287962e-06, + "loss": 1.0211, + "step": 17205 + }, + { + "epoch": 3.76, + "learning_rate": 3.1258466244264806e-06, + "loss": 1.3358, + "step": 17206 + }, + { + "epoch": 3.76, + "learning_rate": 3.1257373825649988e-06, + "loss": 1.0661, + "step": 17207 + }, + { + "epoch": 3.76, + "learning_rate": 3.125628140703518e-06, + "loss": 1.8061, + "step": 17208 + }, + { + "epoch": 3.76, + "learning_rate": 3.1255188988420364e-06, + "loss": 3.5273, + "step": 17209 + }, + { + "epoch": 3.76, + "learning_rate": 3.1254096569805555e-06, + "loss": 2.1632, + "step": 17210 + }, + { + "epoch": 3.76, + "learning_rate": 3.125300415119074e-06, + "loss": 2.5274, + "step": 17211 + }, + { + "epoch": 3.76, + "learning_rate": 3.1251911732575923e-06, + "loss": 1.7325, + "step": 17212 + }, + { + "epoch": 3.76, + "learning_rate": 3.1250819313961113e-06, + "loss": 1.0931, + "step": 17213 + }, + { + "epoch": 3.76, + "learning_rate": 3.12497268953463e-06, + "loss": 1.1888, + "step": 17214 + }, + { + "epoch": 3.76, + "learning_rate": 3.124863447673149e-06, + "loss": 2.6831, + "step": 17215 + }, + { + "epoch": 3.76, + "learning_rate": 3.1247542058116676e-06, + "loss": 3.2258, + "step": 17216 + }, + { + "epoch": 3.76, + "learning_rate": 3.1246449639501857e-06, + "loss": 3.4048, + "step": 17217 + }, + { + "epoch": 3.76, + "learning_rate": 3.1245357220887048e-06, + "loss": 1.5955, + "step": 17218 + }, + { + "epoch": 3.76, + "learning_rate": 3.1244264802272234e-06, + "loss": 2.3313, + "step": 17219 + }, + { + "epoch": 3.76, + "learning_rate": 3.124317238365742e-06, + "loss": 2.0646, + "step": 17220 + }, + { + "epoch": 3.76, + "learning_rate": 3.1242079965042606e-06, + "loss": 1.3569, + "step": 17221 + }, + { + "epoch": 3.76, + "learning_rate": 3.1240987546427792e-06, + "loss": 0.7242, + "step": 17222 + }, + { + "epoch": 3.76, + "learning_rate": 3.123989512781298e-06, + "loss": 1.4197, + "step": 17223 + }, + { + "epoch": 3.76, + "learning_rate": 3.123880270919817e-06, + "loss": 2.5962, + "step": 17224 + }, + { + "epoch": 3.76, + "learning_rate": 3.1237710290583355e-06, + "loss": 2.2434, + "step": 17225 + }, + { + "epoch": 3.76, + "learning_rate": 3.1236617871968537e-06, + "loss": 1.9723, + "step": 17226 + }, + { + "epoch": 3.76, + "learning_rate": 3.1235525453353727e-06, + "loss": 2.2445, + "step": 17227 + }, + { + "epoch": 3.76, + "learning_rate": 3.1234433034738913e-06, + "loss": 1.4594, + "step": 17228 + }, + { + "epoch": 3.76, + "learning_rate": 3.1233340616124103e-06, + "loss": 1.8066, + "step": 17229 + }, + { + "epoch": 3.76, + "learning_rate": 3.123224819750929e-06, + "loss": 2.4695, + "step": 17230 + }, + { + "epoch": 3.76, + "learning_rate": 3.123115577889447e-06, + "loss": 1.7594, + "step": 17231 + }, + { + "epoch": 3.76, + "learning_rate": 3.123006336027966e-06, + "loss": 1.9816, + "step": 17232 + }, + { + "epoch": 3.76, + "learning_rate": 3.1228970941664848e-06, + "loss": 2.7891, + "step": 17233 + }, + { + "epoch": 3.76, + "learning_rate": 3.122787852305004e-06, + "loss": 0.5671, + "step": 17234 + }, + { + "epoch": 3.76, + "learning_rate": 3.122678610443522e-06, + "loss": 1.4531, + "step": 17235 + }, + { + "epoch": 3.76, + "learning_rate": 3.1225693685820406e-06, + "loss": 1.7931, + "step": 17236 + }, + { + "epoch": 3.76, + "learning_rate": 3.1224601267205597e-06, + "loss": 2.3499, + "step": 17237 + }, + { + "epoch": 3.76, + "learning_rate": 3.1223508848590783e-06, + "loss": 1.691, + "step": 17238 + }, + { + "epoch": 3.76, + "learning_rate": 3.1222416429975973e-06, + "loss": 1.6544, + "step": 17239 + }, + { + "epoch": 3.76, + "learning_rate": 3.1221324011361155e-06, + "loss": 2.7144, + "step": 17240 + }, + { + "epoch": 3.76, + "learning_rate": 3.122023159274634e-06, + "loss": 0.4776, + "step": 17241 + }, + { + "epoch": 3.76, + "learning_rate": 3.121913917413153e-06, + "loss": 2.0388, + "step": 17242 + }, + { + "epoch": 3.76, + "learning_rate": 3.1218046755516717e-06, + "loss": 3.1466, + "step": 17243 + }, + { + "epoch": 3.76, + "learning_rate": 3.1216954336901908e-06, + "loss": 1.0507, + "step": 17244 + }, + { + "epoch": 3.76, + "learning_rate": 3.121586191828709e-06, + "loss": 1.7208, + "step": 17245 + }, + { + "epoch": 3.76, + "learning_rate": 3.1214769499672276e-06, + "loss": 2.7797, + "step": 17246 + }, + { + "epoch": 3.76, + "learning_rate": 3.1213677081057466e-06, + "loss": 1.3479, + "step": 17247 + }, + { + "epoch": 3.76, + "learning_rate": 3.1212584662442652e-06, + "loss": 1.2814, + "step": 17248 + }, + { + "epoch": 3.76, + "learning_rate": 3.121149224382784e-06, + "loss": 1.1188, + "step": 17249 + }, + { + "epoch": 3.76, + "learning_rate": 3.1210399825213024e-06, + "loss": 2.1695, + "step": 17250 + }, + { + "epoch": 3.76, + "learning_rate": 3.120930740659821e-06, + "loss": 2.4775, + "step": 17251 + }, + { + "epoch": 3.77, + "learning_rate": 3.1208214987983397e-06, + "loss": 2.1346, + "step": 17252 + }, + { + "epoch": 3.77, + "learning_rate": 3.1207122569368587e-06, + "loss": 1.8249, + "step": 17253 + }, + { + "epoch": 3.77, + "learning_rate": 3.120603015075377e-06, + "loss": 2.3405, + "step": 17254 + }, + { + "epoch": 3.77, + "learning_rate": 3.1204937732138955e-06, + "loss": 0.802, + "step": 17255 + }, + { + "epoch": 3.77, + "learning_rate": 3.1203845313524145e-06, + "loss": 2.0393, + "step": 17256 + }, + { + "epoch": 3.77, + "learning_rate": 3.120275289490933e-06, + "loss": 2.1177, + "step": 17257 + }, + { + "epoch": 3.77, + "learning_rate": 3.120166047629452e-06, + "loss": 1.5693, + "step": 17258 + }, + { + "epoch": 3.77, + "learning_rate": 3.1200568057679704e-06, + "loss": 1.2244, + "step": 17259 + }, + { + "epoch": 3.77, + "learning_rate": 3.119947563906489e-06, + "loss": 2.1197, + "step": 17260 + }, + { + "epoch": 3.77, + "learning_rate": 3.119838322045008e-06, + "loss": 1.9907, + "step": 17261 + }, + { + "epoch": 3.77, + "learning_rate": 3.1197290801835266e-06, + "loss": 1.751, + "step": 17262 + }, + { + "epoch": 3.77, + "learning_rate": 3.1196198383220457e-06, + "loss": 1.7086, + "step": 17263 + }, + { + "epoch": 3.77, + "learning_rate": 3.119510596460564e-06, + "loss": 2.0916, + "step": 17264 + }, + { + "epoch": 3.77, + "learning_rate": 3.1194013545990825e-06, + "loss": 1.7124, + "step": 17265 + }, + { + "epoch": 3.77, + "learning_rate": 3.1192921127376015e-06, + "loss": 2.1031, + "step": 17266 + }, + { + "epoch": 3.77, + "learning_rate": 3.11918287087612e-06, + "loss": 2.5296, + "step": 17267 + }, + { + "epoch": 3.77, + "learning_rate": 3.1190736290146383e-06, + "loss": 1.0484, + "step": 17268 + }, + { + "epoch": 3.77, + "learning_rate": 3.1189643871531573e-06, + "loss": 2.2033, + "step": 17269 + }, + { + "epoch": 3.77, + "learning_rate": 3.118855145291676e-06, + "loss": 3.0351, + "step": 17270 + }, + { + "epoch": 3.77, + "learning_rate": 3.118745903430195e-06, + "loss": 1.2974, + "step": 17271 + }, + { + "epoch": 3.77, + "learning_rate": 3.1186366615687136e-06, + "loss": 1.8848, + "step": 17272 + }, + { + "epoch": 3.77, + "learning_rate": 3.1185274197072318e-06, + "loss": 0.9145, + "step": 17273 + }, + { + "epoch": 3.77, + "learning_rate": 3.118418177845751e-06, + "loss": 2.8791, + "step": 17274 + }, + { + "epoch": 3.77, + "learning_rate": 3.1183089359842694e-06, + "loss": 2.6324, + "step": 17275 + }, + { + "epoch": 3.77, + "learning_rate": 3.118199694122788e-06, + "loss": 1.3092, + "step": 17276 + }, + { + "epoch": 3.77, + "learning_rate": 3.118090452261307e-06, + "loss": 2.9633, + "step": 17277 + }, + { + "epoch": 3.77, + "learning_rate": 3.1179812103998252e-06, + "loss": 0.9136, + "step": 17278 + }, + { + "epoch": 3.77, + "learning_rate": 3.117871968538344e-06, + "loss": 3.0315, + "step": 17279 + }, + { + "epoch": 3.77, + "learning_rate": 3.117762726676863e-06, + "loss": 3.4201, + "step": 17280 + }, + { + "epoch": 3.77, + "learning_rate": 3.1176534848153815e-06, + "loss": 2.8042, + "step": 17281 + }, + { + "epoch": 3.77, + "learning_rate": 3.1175442429539005e-06, + "loss": 2.7638, + "step": 17282 + }, + { + "epoch": 3.77, + "learning_rate": 3.1174350010924187e-06, + "loss": 2.2252, + "step": 17283 + }, + { + "epoch": 3.77, + "learning_rate": 3.1173257592309373e-06, + "loss": 1.8498, + "step": 17284 + }, + { + "epoch": 3.77, + "learning_rate": 3.1172165173694564e-06, + "loss": 3.7529, + "step": 17285 + }, + { + "epoch": 3.77, + "learning_rate": 3.117107275507975e-06, + "loss": 2.2378, + "step": 17286 + }, + { + "epoch": 3.77, + "learning_rate": 3.116998033646493e-06, + "loss": 3.2335, + "step": 17287 + }, + { + "epoch": 3.77, + "learning_rate": 3.116888791785012e-06, + "loss": 2.435, + "step": 17288 + }, + { + "epoch": 3.77, + "learning_rate": 3.116779549923531e-06, + "loss": 2.2307, + "step": 17289 + }, + { + "epoch": 3.77, + "learning_rate": 3.11667030806205e-06, + "loss": 0.9997, + "step": 17290 + }, + { + "epoch": 3.77, + "learning_rate": 3.1165610662005685e-06, + "loss": 2.3837, + "step": 17291 + }, + { + "epoch": 3.77, + "learning_rate": 3.1164518243390866e-06, + "loss": 1.7913, + "step": 17292 + }, + { + "epoch": 3.77, + "learning_rate": 3.1163425824776057e-06, + "loss": 1.351, + "step": 17293 + }, + { + "epoch": 3.77, + "learning_rate": 3.1162333406161243e-06, + "loss": 2.0942, + "step": 17294 + }, + { + "epoch": 3.77, + "learning_rate": 3.1161240987546433e-06, + "loss": 2.5868, + "step": 17295 + }, + { + "epoch": 3.77, + "learning_rate": 3.116014856893162e-06, + "loss": 1.622, + "step": 17296 + }, + { + "epoch": 3.77, + "learning_rate": 3.11590561503168e-06, + "loss": 2.7186, + "step": 17297 + }, + { + "epoch": 3.78, + "learning_rate": 3.115796373170199e-06, + "loss": 1.7467, + "step": 17298 + }, + { + "epoch": 3.78, + "learning_rate": 3.1156871313087178e-06, + "loss": 0.6315, + "step": 17299 + }, + { + "epoch": 3.78, + "learning_rate": 3.115577889447237e-06, + "loss": 2.4333, + "step": 17300 + }, + { + "epoch": 3.78, + "learning_rate": 3.115468647585755e-06, + "loss": 1.9748, + "step": 17301 + }, + { + "epoch": 3.78, + "learning_rate": 3.1153594057242736e-06, + "loss": 1.5195, + "step": 17302 + }, + { + "epoch": 3.78, + "learning_rate": 3.1152501638627926e-06, + "loss": 1.4295, + "step": 17303 + }, + { + "epoch": 3.78, + "learning_rate": 3.1151409220013113e-06, + "loss": 2.6531, + "step": 17304 + }, + { + "epoch": 3.78, + "learning_rate": 3.11503168013983e-06, + "loss": 2.1594, + "step": 17305 + }, + { + "epoch": 3.78, + "learning_rate": 3.1149224382783485e-06, + "loss": 1.9807, + "step": 17306 + }, + { + "epoch": 3.78, + "learning_rate": 3.114813196416867e-06, + "loss": 2.7379, + "step": 17307 + }, + { + "epoch": 3.78, + "learning_rate": 3.1147039545553857e-06, + "loss": 1.5754, + "step": 17308 + }, + { + "epoch": 3.78, + "learning_rate": 3.1145947126939047e-06, + "loss": 1.9161, + "step": 17309 + }, + { + "epoch": 3.78, + "learning_rate": 3.1144854708324233e-06, + "loss": 0.2587, + "step": 17310 + }, + { + "epoch": 3.78, + "learning_rate": 3.1143762289709415e-06, + "loss": 2.0376, + "step": 17311 + }, + { + "epoch": 3.78, + "learning_rate": 3.1142669871094606e-06, + "loss": 1.0727, + "step": 17312 + }, + { + "epoch": 3.78, + "learning_rate": 3.114157745247979e-06, + "loss": 2.1734, + "step": 17313 + }, + { + "epoch": 3.78, + "learning_rate": 3.114048503386498e-06, + "loss": 2.3029, + "step": 17314 + }, + { + "epoch": 3.78, + "learning_rate": 3.113939261525017e-06, + "loss": 1.8208, + "step": 17315 + }, + { + "epoch": 3.78, + "learning_rate": 3.113830019663535e-06, + "loss": 2.0973, + "step": 17316 + }, + { + "epoch": 3.78, + "learning_rate": 3.113720777802054e-06, + "loss": 1.5995, + "step": 17317 + }, + { + "epoch": 3.78, + "learning_rate": 3.1136115359405727e-06, + "loss": 1.4432, + "step": 17318 + }, + { + "epoch": 3.78, + "learning_rate": 3.1135022940790917e-06, + "loss": 3.1541, + "step": 17319 + }, + { + "epoch": 3.78, + "learning_rate": 3.11339305221761e-06, + "loss": 1.8138, + "step": 17320 + }, + { + "epoch": 3.78, + "learning_rate": 3.1132838103561285e-06, + "loss": 1.9577, + "step": 17321 + }, + { + "epoch": 3.78, + "learning_rate": 3.1131745684946475e-06, + "loss": 1.5807, + "step": 17322 + }, + { + "epoch": 3.78, + "learning_rate": 3.113065326633166e-06, + "loss": 0.9632, + "step": 17323 + }, + { + "epoch": 3.78, + "learning_rate": 3.112956084771685e-06, + "loss": 1.1277, + "step": 17324 + }, + { + "epoch": 3.78, + "learning_rate": 3.1128468429102034e-06, + "loss": 1.2536, + "step": 17325 + }, + { + "epoch": 3.78, + "learning_rate": 3.112737601048722e-06, + "loss": 1.7624, + "step": 17326 + }, + { + "epoch": 3.78, + "learning_rate": 3.112628359187241e-06, + "loss": 1.8052, + "step": 17327 + }, + { + "epoch": 3.78, + "learning_rate": 3.1125191173257596e-06, + "loss": 1.7168, + "step": 17328 + }, + { + "epoch": 3.78, + "learning_rate": 3.1124098754642782e-06, + "loss": 1.9426, + "step": 17329 + }, + { + "epoch": 3.78, + "learning_rate": 3.112300633602797e-06, + "loss": 0.8389, + "step": 17330 + }, + { + "epoch": 3.78, + "learning_rate": 3.1121913917413154e-06, + "loss": 1.9138, + "step": 17331 + }, + { + "epoch": 3.78, + "learning_rate": 3.112082149879834e-06, + "loss": 1.9006, + "step": 17332 + }, + { + "epoch": 3.78, + "learning_rate": 3.111972908018353e-06, + "loss": 2.3036, + "step": 17333 + }, + { + "epoch": 3.78, + "learning_rate": 3.1118636661568713e-06, + "loss": 1.5962, + "step": 17334 + }, + { + "epoch": 3.78, + "learning_rate": 3.1117544242953903e-06, + "loss": 1.4897, + "step": 17335 + }, + { + "epoch": 3.78, + "learning_rate": 3.111645182433909e-06, + "loss": 1.4408, + "step": 17336 + }, + { + "epoch": 3.78, + "learning_rate": 3.1115359405724275e-06, + "loss": 1.1069, + "step": 17337 + }, + { + "epoch": 3.78, + "learning_rate": 3.1114266987109466e-06, + "loss": 2.9872, + "step": 17338 + }, + { + "epoch": 3.78, + "learning_rate": 3.1113174568494648e-06, + "loss": 1.4605, + "step": 17339 + }, + { + "epoch": 3.78, + "learning_rate": 3.1112082149879834e-06, + "loss": 1.7564, + "step": 17340 + }, + { + "epoch": 3.78, + "learning_rate": 3.1110989731265024e-06, + "loss": 2.6021, + "step": 17341 + }, + { + "epoch": 3.78, + "learning_rate": 3.110989731265021e-06, + "loss": 1.1662, + "step": 17342 + }, + { + "epoch": 3.79, + "learning_rate": 3.11088048940354e-06, + "loss": 2.2558, + "step": 17343 + }, + { + "epoch": 3.79, + "learning_rate": 3.1107712475420582e-06, + "loss": 2.4143, + "step": 17344 + }, + { + "epoch": 3.79, + "learning_rate": 3.110662005680577e-06, + "loss": 2.1635, + "step": 17345 + }, + { + "epoch": 3.79, + "learning_rate": 3.110552763819096e-06, + "loss": 2.6186, + "step": 17346 + }, + { + "epoch": 3.79, + "learning_rate": 3.1104435219576145e-06, + "loss": 2.2773, + "step": 17347 + }, + { + "epoch": 3.79, + "learning_rate": 3.1103342800961335e-06, + "loss": 0.1956, + "step": 17348 + }, + { + "epoch": 3.79, + "learning_rate": 3.1102250382346517e-06, + "loss": 3.2247, + "step": 17349 + }, + { + "epoch": 3.79, + "learning_rate": 3.1101157963731703e-06, + "loss": 2.731, + "step": 17350 + }, + { + "epoch": 3.79, + "learning_rate": 3.1100065545116894e-06, + "loss": 1.9167, + "step": 17351 + }, + { + "epoch": 3.79, + "learning_rate": 3.109897312650208e-06, + "loss": 1.1116, + "step": 17352 + }, + { + "epoch": 3.79, + "learning_rate": 3.109788070788726e-06, + "loss": 1.906, + "step": 17353 + }, + { + "epoch": 3.79, + "learning_rate": 3.109678828927245e-06, + "loss": 2.3604, + "step": 17354 + }, + { + "epoch": 3.79, + "learning_rate": 3.109569587065764e-06, + "loss": 1.7369, + "step": 17355 + }, + { + "epoch": 3.79, + "learning_rate": 3.109460345204283e-06, + "loss": 1.5397, + "step": 17356 + }, + { + "epoch": 3.79, + "learning_rate": 3.1093511033428014e-06, + "loss": 2.4165, + "step": 17357 + }, + { + "epoch": 3.79, + "learning_rate": 3.1092418614813196e-06, + "loss": 2.8599, + "step": 17358 + }, + { + "epoch": 3.79, + "learning_rate": 3.1091326196198387e-06, + "loss": 2.2, + "step": 17359 + }, + { + "epoch": 3.79, + "learning_rate": 3.1090233777583573e-06, + "loss": 3.0837, + "step": 17360 + }, + { + "epoch": 3.79, + "learning_rate": 3.108914135896876e-06, + "loss": 1.34, + "step": 17361 + }, + { + "epoch": 3.79, + "learning_rate": 3.108804894035395e-06, + "loss": 1.1837, + "step": 17362 + }, + { + "epoch": 3.79, + "learning_rate": 3.108695652173913e-06, + "loss": 2.2794, + "step": 17363 + }, + { + "epoch": 3.79, + "learning_rate": 3.1085864103124317e-06, + "loss": 0.9634, + "step": 17364 + }, + { + "epoch": 3.79, + "learning_rate": 3.1084771684509508e-06, + "loss": 1.5473, + "step": 17365 + }, + { + "epoch": 3.79, + "learning_rate": 3.1083679265894694e-06, + "loss": 1.6982, + "step": 17366 + }, + { + "epoch": 3.79, + "learning_rate": 3.1082586847279876e-06, + "loss": 1.4593, + "step": 17367 + }, + { + "epoch": 3.79, + "learning_rate": 3.1081494428665066e-06, + "loss": 1.1703, + "step": 17368 + }, + { + "epoch": 3.79, + "learning_rate": 3.108040201005025e-06, + "loss": 2.4138, + "step": 17369 + }, + { + "epoch": 3.79, + "learning_rate": 3.1079309591435442e-06, + "loss": 2.1406, + "step": 17370 + }, + { + "epoch": 3.79, + "learning_rate": 3.107821717282063e-06, + "loss": 1.8233, + "step": 17371 + }, + { + "epoch": 3.79, + "learning_rate": 3.107712475420581e-06, + "loss": 1.6815, + "step": 17372 + }, + { + "epoch": 3.79, + "learning_rate": 3.1076032335591e-06, + "loss": 1.935, + "step": 17373 + }, + { + "epoch": 3.79, + "learning_rate": 3.1074939916976187e-06, + "loss": 1.6368, + "step": 17374 + }, + { + "epoch": 3.79, + "learning_rate": 3.1073847498361377e-06, + "loss": 3.4739, + "step": 17375 + }, + { + "epoch": 3.79, + "learning_rate": 3.1072755079746563e-06, + "loss": 1.1364, + "step": 17376 + }, + { + "epoch": 3.79, + "learning_rate": 3.1071662661131745e-06, + "loss": 2.7212, + "step": 17377 + }, + { + "epoch": 3.79, + "learning_rate": 3.1070570242516935e-06, + "loss": 2.4092, + "step": 17378 + }, + { + "epoch": 3.79, + "learning_rate": 3.106947782390212e-06, + "loss": 1.0988, + "step": 17379 + }, + { + "epoch": 3.79, + "learning_rate": 3.106838540528731e-06, + "loss": 1.0667, + "step": 17380 + }, + { + "epoch": 3.79, + "learning_rate": 3.10672929866725e-06, + "loss": 1.0109, + "step": 17381 + }, + { + "epoch": 3.79, + "learning_rate": 3.106620056805768e-06, + "loss": 2.1034, + "step": 17382 + }, + { + "epoch": 3.79, + "learning_rate": 3.106510814944287e-06, + "loss": 2.1142, + "step": 17383 + }, + { + "epoch": 3.79, + "learning_rate": 3.1064015730828056e-06, + "loss": 3.4409, + "step": 17384 + }, + { + "epoch": 3.79, + "learning_rate": 3.1062923312213242e-06, + "loss": 3.0666, + "step": 17385 + }, + { + "epoch": 3.79, + "learning_rate": 3.106183089359843e-06, + "loss": 2.8415, + "step": 17386 + }, + { + "epoch": 3.79, + "learning_rate": 3.1060738474983615e-06, + "loss": 2.1039, + "step": 17387 + }, + { + "epoch": 3.79, + "learning_rate": 3.1059646056368805e-06, + "loss": 1.3663, + "step": 17388 + }, + { + "epoch": 3.8, + "learning_rate": 3.105855363775399e-06, + "loss": 1.9165, + "step": 17389 + }, + { + "epoch": 3.8, + "learning_rate": 3.1057461219139177e-06, + "loss": 1.184, + "step": 17390 + }, + { + "epoch": 3.8, + "learning_rate": 3.1056368800524363e-06, + "loss": 1.6432, + "step": 17391 + }, + { + "epoch": 3.8, + "learning_rate": 3.105527638190955e-06, + "loss": 1.4796, + "step": 17392 + }, + { + "epoch": 3.8, + "learning_rate": 3.1054183963294736e-06, + "loss": 2.4539, + "step": 17393 + }, + { + "epoch": 3.8, + "learning_rate": 3.1053091544679926e-06, + "loss": 1.8813, + "step": 17394 + }, + { + "epoch": 3.8, + "learning_rate": 3.105199912606511e-06, + "loss": 1.172, + "step": 17395 + }, + { + "epoch": 3.8, + "learning_rate": 3.1050906707450294e-06, + "loss": 2.505, + "step": 17396 + }, + { + "epoch": 3.8, + "learning_rate": 3.1049814288835484e-06, + "loss": 2.0358, + "step": 17397 + }, + { + "epoch": 3.8, + "learning_rate": 3.104872187022067e-06, + "loss": 1.3477, + "step": 17398 + }, + { + "epoch": 3.8, + "learning_rate": 3.104762945160586e-06, + "loss": 0.6499, + "step": 17399 + }, + { + "epoch": 3.8, + "learning_rate": 3.1046537032991043e-06, + "loss": 1.814, + "step": 17400 + }, + { + "epoch": 3.8, + "learning_rate": 3.104544461437623e-06, + "loss": 0.9082, + "step": 17401 + }, + { + "epoch": 3.8, + "learning_rate": 3.104435219576142e-06, + "loss": 3.3734, + "step": 17402 + }, + { + "epoch": 3.8, + "learning_rate": 3.1043259777146605e-06, + "loss": 3.3159, + "step": 17403 + }, + { + "epoch": 3.8, + "learning_rate": 3.1042167358531795e-06, + "loss": 2.268, + "step": 17404 + }, + { + "epoch": 3.8, + "learning_rate": 3.1041074939916977e-06, + "loss": 1.8552, + "step": 17405 + }, + { + "epoch": 3.8, + "learning_rate": 3.1039982521302163e-06, + "loss": 1.9444, + "step": 17406 + }, + { + "epoch": 3.8, + "learning_rate": 3.1038890102687354e-06, + "loss": 3.1034, + "step": 17407 + }, + { + "epoch": 3.8, + "learning_rate": 3.103779768407254e-06, + "loss": 2.0136, + "step": 17408 + }, + { + "epoch": 3.8, + "learning_rate": 3.103670526545773e-06, + "loss": 1.9338, + "step": 17409 + }, + { + "epoch": 3.8, + "learning_rate": 3.1035612846842912e-06, + "loss": 2.303, + "step": 17410 + }, + { + "epoch": 3.8, + "learning_rate": 3.10345204282281e-06, + "loss": 1.7639, + "step": 17411 + }, + { + "epoch": 3.8, + "learning_rate": 3.103342800961329e-06, + "loss": 1.258, + "step": 17412 + }, + { + "epoch": 3.8, + "learning_rate": 3.1032335590998475e-06, + "loss": 1.9763, + "step": 17413 + }, + { + "epoch": 3.8, + "learning_rate": 3.103124317238366e-06, + "loss": 1.5662, + "step": 17414 + }, + { + "epoch": 3.8, + "learning_rate": 3.1030150753768847e-06, + "loss": 1.6153, + "step": 17415 + }, + { + "epoch": 3.8, + "learning_rate": 3.1029058335154033e-06, + "loss": 1.1401, + "step": 17416 + }, + { + "epoch": 3.8, + "learning_rate": 3.102796591653922e-06, + "loss": 1.683, + "step": 17417 + }, + { + "epoch": 3.8, + "learning_rate": 3.102687349792441e-06, + "loss": 2.1871, + "step": 17418 + }, + { + "epoch": 3.8, + "learning_rate": 3.102578107930959e-06, + "loss": 2.9059, + "step": 17419 + }, + { + "epoch": 3.8, + "learning_rate": 3.1024688660694777e-06, + "loss": 1.628, + "step": 17420 + }, + { + "epoch": 3.8, + "learning_rate": 3.1023596242079968e-06, + "loss": 2.2212, + "step": 17421 + }, + { + "epoch": 3.8, + "learning_rate": 3.1022503823465154e-06, + "loss": 1.734, + "step": 17422 + }, + { + "epoch": 3.8, + "learning_rate": 3.1021411404850344e-06, + "loss": 2.2621, + "step": 17423 + }, + { + "epoch": 3.8, + "learning_rate": 3.1020318986235526e-06, + "loss": 2.4618, + "step": 17424 + }, + { + "epoch": 3.8, + "learning_rate": 3.1019226567620712e-06, + "loss": 2.3932, + "step": 17425 + }, + { + "epoch": 3.8, + "learning_rate": 3.1018134149005903e-06, + "loss": 1.3839, + "step": 17426 + }, + { + "epoch": 3.8, + "learning_rate": 3.101704173039109e-06, + "loss": 1.3881, + "step": 17427 + }, + { + "epoch": 3.8, + "learning_rate": 3.101594931177628e-06, + "loss": 1.2846, + "step": 17428 + }, + { + "epoch": 3.8, + "learning_rate": 3.101485689316146e-06, + "loss": 3.8584, + "step": 17429 + }, + { + "epoch": 3.8, + "learning_rate": 3.1013764474546647e-06, + "loss": 2.9312, + "step": 17430 + }, + { + "epoch": 3.8, + "learning_rate": 3.1012672055931837e-06, + "loss": 1.7389, + "step": 17431 + }, + { + "epoch": 3.8, + "learning_rate": 3.1011579637317024e-06, + "loss": 1.2849, + "step": 17432 + }, + { + "epoch": 3.8, + "learning_rate": 3.1010487218702205e-06, + "loss": 1.3808, + "step": 17433 + }, + { + "epoch": 3.8, + "learning_rate": 3.1009394800087396e-06, + "loss": 2.0561, + "step": 17434 + }, + { + "epoch": 3.81, + "learning_rate": 3.100830238147258e-06, + "loss": 3.1964, + "step": 17435 + }, + { + "epoch": 3.81, + "learning_rate": 3.1007209962857772e-06, + "loss": 1.2505, + "step": 17436 + }, + { + "epoch": 3.81, + "learning_rate": 3.100611754424296e-06, + "loss": 1.7098, + "step": 17437 + }, + { + "epoch": 3.81, + "learning_rate": 3.100502512562814e-06, + "loss": 2.3458, + "step": 17438 + }, + { + "epoch": 3.81, + "learning_rate": 3.100393270701333e-06, + "loss": 2.6914, + "step": 17439 + }, + { + "epoch": 3.81, + "learning_rate": 3.1002840288398517e-06, + "loss": 2.1407, + "step": 17440 + }, + { + "epoch": 3.81, + "learning_rate": 3.1001747869783707e-06, + "loss": 1.7487, + "step": 17441 + }, + { + "epoch": 3.81, + "learning_rate": 3.1000655451168893e-06, + "loss": 2.4731, + "step": 17442 + }, + { + "epoch": 3.81, + "learning_rate": 3.0999563032554075e-06, + "loss": 1.2756, + "step": 17443 + }, + { + "epoch": 3.81, + "learning_rate": 3.0998470613939265e-06, + "loss": 1.2568, + "step": 17444 + }, + { + "epoch": 3.81, + "learning_rate": 3.099737819532445e-06, + "loss": 2.5623, + "step": 17445 + }, + { + "epoch": 3.81, + "learning_rate": 3.0996285776709638e-06, + "loss": 1.8774, + "step": 17446 + }, + { + "epoch": 3.81, + "learning_rate": 3.0995193358094828e-06, + "loss": 2.1262, + "step": 17447 + }, + { + "epoch": 3.81, + "learning_rate": 3.099410093948001e-06, + "loss": 1.584, + "step": 17448 + }, + { + "epoch": 3.81, + "learning_rate": 3.0993008520865196e-06, + "loss": 1.3733, + "step": 17449 + }, + { + "epoch": 3.81, + "learning_rate": 3.0991916102250386e-06, + "loss": 2.0604, + "step": 17450 + }, + { + "epoch": 3.81, + "learning_rate": 3.0990823683635572e-06, + "loss": 1.0583, + "step": 17451 + }, + { + "epoch": 3.81, + "learning_rate": 3.0989731265020754e-06, + "loss": 1.3972, + "step": 17452 + }, + { + "epoch": 3.81, + "learning_rate": 3.0988638846405945e-06, + "loss": 2.3315, + "step": 17453 + }, + { + "epoch": 3.81, + "learning_rate": 3.098754642779113e-06, + "loss": 0.9797, + "step": 17454 + }, + { + "epoch": 3.81, + "learning_rate": 3.098645400917632e-06, + "loss": 2.2224, + "step": 17455 + }, + { + "epoch": 3.81, + "learning_rate": 3.0985361590561507e-06, + "loss": 1.315, + "step": 17456 + }, + { + "epoch": 3.81, + "learning_rate": 3.098426917194669e-06, + "loss": 2.8243, + "step": 17457 + }, + { + "epoch": 3.81, + "learning_rate": 3.098317675333188e-06, + "loss": 1.7429, + "step": 17458 + }, + { + "epoch": 3.81, + "learning_rate": 3.0982084334717065e-06, + "loss": 1.6455, + "step": 17459 + }, + { + "epoch": 3.81, + "learning_rate": 3.0980991916102256e-06, + "loss": 1.8341, + "step": 17460 + }, + { + "epoch": 3.81, + "learning_rate": 3.097989949748744e-06, + "loss": 1.9742, + "step": 17461 + }, + { + "epoch": 3.81, + "learning_rate": 3.0978807078872624e-06, + "loss": 2.4227, + "step": 17462 + }, + { + "epoch": 3.81, + "learning_rate": 3.0977714660257814e-06, + "loss": 1.74, + "step": 17463 + }, + { + "epoch": 3.81, + "learning_rate": 3.0976622241643e-06, + "loss": 1.1661, + "step": 17464 + }, + { + "epoch": 3.81, + "learning_rate": 3.097552982302819e-06, + "loss": 1.3592, + "step": 17465 + }, + { + "epoch": 3.81, + "learning_rate": 3.0974437404413372e-06, + "loss": 2.2179, + "step": 17466 + }, + { + "epoch": 3.81, + "learning_rate": 3.097334498579856e-06, + "loss": 1.8468, + "step": 17467 + }, + { + "epoch": 3.81, + "learning_rate": 3.097225256718375e-06, + "loss": 3.0992, + "step": 17468 + }, + { + "epoch": 3.81, + "learning_rate": 3.0971160148568935e-06, + "loss": 2.9559, + "step": 17469 + }, + { + "epoch": 3.81, + "learning_rate": 3.097006772995412e-06, + "loss": 0.8258, + "step": 17470 + }, + { + "epoch": 3.81, + "learning_rate": 3.0968975311339307e-06, + "loss": 2.0104, + "step": 17471 + }, + { + "epoch": 3.81, + "learning_rate": 3.0967882892724493e-06, + "loss": 0.6855, + "step": 17472 + }, + { + "epoch": 3.81, + "learning_rate": 3.096679047410968e-06, + "loss": 2.0585, + "step": 17473 + }, + { + "epoch": 3.81, + "learning_rate": 3.096569805549487e-06, + "loss": 2.5313, + "step": 17474 + }, + { + "epoch": 3.81, + "learning_rate": 3.0964605636880056e-06, + "loss": 1.5273, + "step": 17475 + }, + { + "epoch": 3.81, + "learning_rate": 3.0963513218265238e-06, + "loss": 2.7288, + "step": 17476 + }, + { + "epoch": 3.81, + "learning_rate": 3.096242079965043e-06, + "loss": 1.8137, + "step": 17477 + }, + { + "epoch": 3.81, + "learning_rate": 3.0961328381035614e-06, + "loss": 0.4356, + "step": 17478 + }, + { + "epoch": 3.81, + "learning_rate": 3.0960235962420805e-06, + "loss": 2.1977, + "step": 17479 + }, + { + "epoch": 3.81, + "learning_rate": 3.095914354380599e-06, + "loss": 1.3328, + "step": 17480 + }, + { + "epoch": 3.82, + "learning_rate": 3.0958051125191173e-06, + "loss": 3.0678, + "step": 17481 + }, + { + "epoch": 3.82, + "learning_rate": 3.0956958706576363e-06, + "loss": 2.9301, + "step": 17482 + }, + { + "epoch": 3.82, + "learning_rate": 3.095586628796155e-06, + "loss": 1.9292, + "step": 17483 + }, + { + "epoch": 3.82, + "learning_rate": 3.095477386934674e-06, + "loss": 1.6001, + "step": 17484 + }, + { + "epoch": 3.82, + "learning_rate": 3.095368145073192e-06, + "loss": 0.2316, + "step": 17485 + }, + { + "epoch": 3.82, + "learning_rate": 3.0952589032117107e-06, + "loss": 3.4943, + "step": 17486 + }, + { + "epoch": 3.82, + "learning_rate": 3.0951496613502298e-06, + "loss": 3.6521, + "step": 17487 + }, + { + "epoch": 3.82, + "learning_rate": 3.0950404194887484e-06, + "loss": 1.7398, + "step": 17488 + }, + { + "epoch": 3.82, + "learning_rate": 3.0949311776272674e-06, + "loss": 0.5294, + "step": 17489 + }, + { + "epoch": 3.82, + "learning_rate": 3.0948219357657856e-06, + "loss": 2.3208, + "step": 17490 + }, + { + "epoch": 3.82, + "learning_rate": 3.094712693904304e-06, + "loss": 1.971, + "step": 17491 + }, + { + "epoch": 3.82, + "learning_rate": 3.0946034520428232e-06, + "loss": 0.8576, + "step": 17492 + }, + { + "epoch": 3.82, + "learning_rate": 3.094494210181342e-06, + "loss": 1.7591, + "step": 17493 + }, + { + "epoch": 3.82, + "learning_rate": 3.094384968319861e-06, + "loss": 1.9207, + "step": 17494 + }, + { + "epoch": 3.82, + "learning_rate": 3.094275726458379e-06, + "loss": 1.6345, + "step": 17495 + }, + { + "epoch": 3.82, + "learning_rate": 3.0941664845968977e-06, + "loss": 0.6233, + "step": 17496 + }, + { + "epoch": 3.82, + "learning_rate": 3.0940572427354167e-06, + "loss": 2.1177, + "step": 17497 + }, + { + "epoch": 3.82, + "learning_rate": 3.0939480008739353e-06, + "loss": 0.9779, + "step": 17498 + }, + { + "epoch": 3.82, + "learning_rate": 3.0938387590124535e-06, + "loss": 1.7636, + "step": 17499 + }, + { + "epoch": 3.82, + "learning_rate": 3.0937295171509726e-06, + "loss": 1.1925, + "step": 17500 + }, + { + "epoch": 3.82, + "learning_rate": 3.093620275289491e-06, + "loss": 2.5148, + "step": 17501 + }, + { + "epoch": 3.82, + "learning_rate": 3.0935110334280098e-06, + "loss": 2.7583, + "step": 17502 + }, + { + "epoch": 3.82, + "learning_rate": 3.093401791566529e-06, + "loss": 3.4577, + "step": 17503 + }, + { + "epoch": 3.82, + "learning_rate": 3.093292549705047e-06, + "loss": 1.823, + "step": 17504 + }, + { + "epoch": 3.82, + "learning_rate": 3.0931833078435656e-06, + "loss": 2.6049, + "step": 17505 + }, + { + "epoch": 3.82, + "learning_rate": 3.0930740659820846e-06, + "loss": 1.3287, + "step": 17506 + }, + { + "epoch": 3.82, + "learning_rate": 3.0929648241206033e-06, + "loss": 3.2024, + "step": 17507 + }, + { + "epoch": 3.82, + "learning_rate": 3.0928555822591223e-06, + "loss": 2.0089, + "step": 17508 + }, + { + "epoch": 3.82, + "learning_rate": 3.0927463403976405e-06, + "loss": 1.02, + "step": 17509 + }, + { + "epoch": 3.82, + "learning_rate": 3.092637098536159e-06, + "loss": 1.4594, + "step": 17510 + }, + { + "epoch": 3.82, + "learning_rate": 3.092527856674678e-06, + "loss": 2.1601, + "step": 17511 + }, + { + "epoch": 3.82, + "learning_rate": 3.0924186148131967e-06, + "loss": 2.2493, + "step": 17512 + }, + { + "epoch": 3.82, + "learning_rate": 3.0923093729517158e-06, + "loss": 1.9325, + "step": 17513 + }, + { + "epoch": 3.82, + "learning_rate": 3.092200131090234e-06, + "loss": 1.4882, + "step": 17514 + }, + { + "epoch": 3.82, + "learning_rate": 3.0920908892287526e-06, + "loss": 2.8949, + "step": 17515 + }, + { + "epoch": 3.82, + "learning_rate": 3.0919816473672716e-06, + "loss": 1.9344, + "step": 17516 + }, + { + "epoch": 3.82, + "learning_rate": 3.0918724055057902e-06, + "loss": 1.3399, + "step": 17517 + }, + { + "epoch": 3.82, + "learning_rate": 3.0917631636443084e-06, + "loss": 1.9352, + "step": 17518 + }, + { + "epoch": 3.82, + "learning_rate": 3.0916539217828274e-06, + "loss": 1.921, + "step": 17519 + }, + { + "epoch": 3.82, + "learning_rate": 3.091544679921346e-06, + "loss": 1.6955, + "step": 17520 + }, + { + "epoch": 3.82, + "learning_rate": 3.091435438059865e-06, + "loss": 2.807, + "step": 17521 + }, + { + "epoch": 3.82, + "learning_rate": 3.0913261961983837e-06, + "loss": 1.6251, + "step": 17522 + }, + { + "epoch": 3.82, + "learning_rate": 3.091216954336902e-06, + "loss": 3.919, + "step": 17523 + }, + { + "epoch": 3.82, + "learning_rate": 3.091107712475421e-06, + "loss": 2.4423, + "step": 17524 + }, + { + "epoch": 3.82, + "learning_rate": 3.0909984706139395e-06, + "loss": 4.0807, + "step": 17525 + }, + { + "epoch": 3.82, + "learning_rate": 3.090889228752458e-06, + "loss": 1.3432, + "step": 17526 + }, + { + "epoch": 3.83, + "learning_rate": 3.090779986890977e-06, + "loss": 1.8872, + "step": 17527 + }, + { + "epoch": 3.83, + "learning_rate": 3.0906707450294954e-06, + "loss": 1.8776, + "step": 17528 + }, + { + "epoch": 3.83, + "learning_rate": 3.090561503168014e-06, + "loss": 1.9356, + "step": 17529 + }, + { + "epoch": 3.83, + "learning_rate": 3.090452261306533e-06, + "loss": 2.2222, + "step": 17530 + }, + { + "epoch": 3.83, + "learning_rate": 3.0903430194450516e-06, + "loss": 1.411, + "step": 17531 + }, + { + "epoch": 3.83, + "learning_rate": 3.0902337775835702e-06, + "loss": 2.5175, + "step": 17532 + }, + { + "epoch": 3.83, + "learning_rate": 3.090124535722089e-06, + "loss": 1.0091, + "step": 17533 + }, + { + "epoch": 3.83, + "learning_rate": 3.0900152938606074e-06, + "loss": 1.9022, + "step": 17534 + }, + { + "epoch": 3.83, + "learning_rate": 3.0899060519991265e-06, + "loss": 2.7513, + "step": 17535 + }, + { + "epoch": 3.83, + "learning_rate": 3.089796810137645e-06, + "loss": 2.5779, + "step": 17536 + }, + { + "epoch": 3.83, + "learning_rate": 3.0896875682761633e-06, + "loss": 1.7022, + "step": 17537 + }, + { + "epoch": 3.83, + "learning_rate": 3.0895783264146823e-06, + "loss": 2.2857, + "step": 17538 + }, + { + "epoch": 3.83, + "learning_rate": 3.089469084553201e-06, + "loss": 3.0083, + "step": 17539 + }, + { + "epoch": 3.83, + "learning_rate": 3.08935984269172e-06, + "loss": 1.6349, + "step": 17540 + }, + { + "epoch": 3.83, + "learning_rate": 3.0892506008302386e-06, + "loss": 1.397, + "step": 17541 + }, + { + "epoch": 3.83, + "learning_rate": 3.0891413589687568e-06, + "loss": 3.1184, + "step": 17542 + }, + { + "epoch": 3.83, + "learning_rate": 3.089032117107276e-06, + "loss": 2.4441, + "step": 17543 + }, + { + "epoch": 3.83, + "learning_rate": 3.0889228752457944e-06, + "loss": 1.6994, + "step": 17544 + }, + { + "epoch": 3.83, + "learning_rate": 3.0888136333843134e-06, + "loss": 3.2368, + "step": 17545 + }, + { + "epoch": 3.83, + "learning_rate": 3.0887043915228316e-06, + "loss": 1.557, + "step": 17546 + }, + { + "epoch": 3.83, + "learning_rate": 3.0885951496613502e-06, + "loss": 2.1828, + "step": 17547 + }, + { + "epoch": 3.83, + "learning_rate": 3.0884859077998693e-06, + "loss": 1.5026, + "step": 17548 + }, + { + "epoch": 3.83, + "learning_rate": 3.088376665938388e-06, + "loss": 1.2838, + "step": 17549 + }, + { + "epoch": 3.83, + "learning_rate": 3.088267424076907e-06, + "loss": 1.7864, + "step": 17550 + }, + { + "epoch": 3.83, + "learning_rate": 3.088158182215425e-06, + "loss": 2.5666, + "step": 17551 + }, + { + "epoch": 3.83, + "learning_rate": 3.0880489403539437e-06, + "loss": 1.379, + "step": 17552 + }, + { + "epoch": 3.83, + "learning_rate": 3.0879396984924628e-06, + "loss": 2.9135, + "step": 17553 + }, + { + "epoch": 3.83, + "learning_rate": 3.0878304566309814e-06, + "loss": 1.5253, + "step": 17554 + }, + { + "epoch": 3.83, + "learning_rate": 3.0877212147695e-06, + "loss": 0.2411, + "step": 17555 + }, + { + "epoch": 3.83, + "learning_rate": 3.0876119729080186e-06, + "loss": 2.3209, + "step": 17556 + }, + { + "epoch": 3.83, + "learning_rate": 3.087502731046537e-06, + "loss": 2.5039, + "step": 17557 + }, + { + "epoch": 3.83, + "learning_rate": 3.087393489185056e-06, + "loss": 2.8679, + "step": 17558 + }, + { + "epoch": 3.83, + "learning_rate": 3.087284247323575e-06, + "loss": 2.6407, + "step": 17559 + }, + { + "epoch": 3.83, + "learning_rate": 3.0871750054620935e-06, + "loss": 2.9432, + "step": 17560 + }, + { + "epoch": 3.83, + "learning_rate": 3.0870657636006116e-06, + "loss": 1.8027, + "step": 17561 + }, + { + "epoch": 3.83, + "learning_rate": 3.0869565217391307e-06, + "loss": 1.643, + "step": 17562 + }, + { + "epoch": 3.83, + "learning_rate": 3.0868472798776493e-06, + "loss": 2.3746, + "step": 17563 + }, + { + "epoch": 3.83, + "learning_rate": 3.0867380380161683e-06, + "loss": 1.0395, + "step": 17564 + }, + { + "epoch": 3.83, + "learning_rate": 3.0866287961546865e-06, + "loss": 2.353, + "step": 17565 + }, + { + "epoch": 3.83, + "learning_rate": 3.086519554293205e-06, + "loss": 1.3484, + "step": 17566 + }, + { + "epoch": 3.83, + "learning_rate": 3.086410312431724e-06, + "loss": 0.8024, + "step": 17567 + }, + { + "epoch": 3.83, + "learning_rate": 3.0863010705702428e-06, + "loss": 1.8095, + "step": 17568 + }, + { + "epoch": 3.83, + "learning_rate": 3.086191828708762e-06, + "loss": 3.3173, + "step": 17569 + }, + { + "epoch": 3.83, + "learning_rate": 3.08608258684728e-06, + "loss": 1.8721, + "step": 17570 + }, + { + "epoch": 3.83, + "learning_rate": 3.0859733449857986e-06, + "loss": 1.4931, + "step": 17571 + }, + { + "epoch": 3.84, + "learning_rate": 3.0858641031243176e-06, + "loss": 1.7252, + "step": 17572 + }, + { + "epoch": 3.84, + "learning_rate": 3.0857548612628362e-06, + "loss": 0.8725, + "step": 17573 + }, + { + "epoch": 3.84, + "learning_rate": 3.0856456194013553e-06, + "loss": 1.5543, + "step": 17574 + }, + { + "epoch": 3.84, + "learning_rate": 3.0855363775398735e-06, + "loss": 2.2418, + "step": 17575 + }, + { + "epoch": 3.84, + "learning_rate": 3.085427135678392e-06, + "loss": 1.7079, + "step": 17576 + }, + { + "epoch": 3.84, + "learning_rate": 3.085317893816911e-06, + "loss": 0.4555, + "step": 17577 + }, + { + "epoch": 3.84, + "learning_rate": 3.0852086519554297e-06, + "loss": 2.4819, + "step": 17578 + }, + { + "epoch": 3.84, + "learning_rate": 3.085099410093948e-06, + "loss": 1.7441, + "step": 17579 + }, + { + "epoch": 3.84, + "learning_rate": 3.084990168232467e-06, + "loss": 2.7691, + "step": 17580 + }, + { + "epoch": 3.84, + "learning_rate": 3.0848809263709856e-06, + "loss": 1.6286, + "step": 17581 + }, + { + "epoch": 3.84, + "learning_rate": 3.084771684509504e-06, + "loss": 2.1236, + "step": 17582 + }, + { + "epoch": 3.84, + "learning_rate": 3.084662442648023e-06, + "loss": 1.6668, + "step": 17583 + }, + { + "epoch": 3.84, + "learning_rate": 3.0845532007865414e-06, + "loss": 2.4291, + "step": 17584 + }, + { + "epoch": 3.84, + "learning_rate": 3.0844439589250604e-06, + "loss": 1.7734, + "step": 17585 + }, + { + "epoch": 3.84, + "learning_rate": 3.084334717063579e-06, + "loss": 2.9349, + "step": 17586 + }, + { + "epoch": 3.84, + "learning_rate": 3.0842254752020976e-06, + "loss": 2.8902, + "step": 17587 + }, + { + "epoch": 3.84, + "learning_rate": 3.0841162333406167e-06, + "loss": 1.0867, + "step": 17588 + }, + { + "epoch": 3.84, + "learning_rate": 3.084006991479135e-06, + "loss": 1.8369, + "step": 17589 + }, + { + "epoch": 3.84, + "learning_rate": 3.0838977496176535e-06, + "loss": 1.1223, + "step": 17590 + }, + { + "epoch": 3.84, + "learning_rate": 3.0837885077561725e-06, + "loss": 1.2461, + "step": 17591 + }, + { + "epoch": 3.84, + "learning_rate": 3.083679265894691e-06, + "loss": 1.1526, + "step": 17592 + }, + { + "epoch": 3.84, + "learning_rate": 3.08357002403321e-06, + "loss": 2.034, + "step": 17593 + }, + { + "epoch": 3.84, + "learning_rate": 3.0834607821717283e-06, + "loss": 1.9182, + "step": 17594 + }, + { + "epoch": 3.84, + "learning_rate": 3.083351540310247e-06, + "loss": 2.9745, + "step": 17595 + }, + { + "epoch": 3.84, + "learning_rate": 3.083242298448766e-06, + "loss": 1.9803, + "step": 17596 + }, + { + "epoch": 3.84, + "learning_rate": 3.0831330565872846e-06, + "loss": 2.1123, + "step": 17597 + }, + { + "epoch": 3.84, + "learning_rate": 3.0830238147258028e-06, + "loss": 1.9292, + "step": 17598 + }, + { + "epoch": 3.84, + "learning_rate": 3.082914572864322e-06, + "loss": 2.1898, + "step": 17599 + }, + { + "epoch": 3.84, + "learning_rate": 3.0828053310028404e-06, + "loss": 1.7204, + "step": 17600 + }, + { + "epoch": 3.84, + "learning_rate": 3.0826960891413595e-06, + "loss": 1.2857, + "step": 17601 + }, + { + "epoch": 3.84, + "learning_rate": 3.082586847279878e-06, + "loss": 1.8391, + "step": 17602 + }, + { + "epoch": 3.84, + "learning_rate": 3.0824776054183963e-06, + "loss": 1.3566, + "step": 17603 + }, + { + "epoch": 3.84, + "learning_rate": 3.0823683635569153e-06, + "loss": 1.0938, + "step": 17604 + }, + { + "epoch": 3.84, + "learning_rate": 3.082259121695434e-06, + "loss": 3.4168, + "step": 17605 + }, + { + "epoch": 3.84, + "learning_rate": 3.082149879833953e-06, + "loss": 0.8607, + "step": 17606 + }, + { + "epoch": 3.84, + "learning_rate": 3.0820406379724716e-06, + "loss": 0.9458, + "step": 17607 + }, + { + "epoch": 3.84, + "learning_rate": 3.0819313961109897e-06, + "loss": 2.4634, + "step": 17608 + }, + { + "epoch": 3.84, + "learning_rate": 3.0818221542495088e-06, + "loss": 2.3514, + "step": 17609 + }, + { + "epoch": 3.84, + "learning_rate": 3.0817129123880274e-06, + "loss": 2.0939, + "step": 17610 + }, + { + "epoch": 3.84, + "learning_rate": 3.081603670526546e-06, + "loss": 2.4044, + "step": 17611 + }, + { + "epoch": 3.84, + "learning_rate": 3.0814944286650646e-06, + "loss": 0.9396, + "step": 17612 + }, + { + "epoch": 3.84, + "learning_rate": 3.0813851868035832e-06, + "loss": 1.8404, + "step": 17613 + }, + { + "epoch": 3.84, + "learning_rate": 3.081275944942102e-06, + "loss": 0.7585, + "step": 17614 + }, + { + "epoch": 3.84, + "learning_rate": 3.081166703080621e-06, + "loss": 0.6402, + "step": 17615 + }, + { + "epoch": 3.84, + "learning_rate": 3.0810574612191395e-06, + "loss": 2.3451, + "step": 17616 + }, + { + "epoch": 3.84, + "learning_rate": 3.0809482193576577e-06, + "loss": 2.14, + "step": 17617 + }, + { + "epoch": 3.85, + "learning_rate": 3.0808389774961767e-06, + "loss": 0.9785, + "step": 17618 + }, + { + "epoch": 3.85, + "learning_rate": 3.0807297356346953e-06, + "loss": 0.5327, + "step": 17619 + }, + { + "epoch": 3.85, + "learning_rate": 3.0806204937732143e-06, + "loss": 1.692, + "step": 17620 + }, + { + "epoch": 3.85, + "learning_rate": 3.080511251911733e-06, + "loss": 1.4479, + "step": 17621 + }, + { + "epoch": 3.85, + "learning_rate": 3.080402010050251e-06, + "loss": 3.204, + "step": 17622 + }, + { + "epoch": 3.85, + "learning_rate": 3.08029276818877e-06, + "loss": 3.0818, + "step": 17623 + }, + { + "epoch": 3.85, + "learning_rate": 3.080183526327289e-06, + "loss": 1.3202, + "step": 17624 + }, + { + "epoch": 3.85, + "learning_rate": 3.080074284465808e-06, + "loss": 2.3779, + "step": 17625 + }, + { + "epoch": 3.85, + "learning_rate": 3.0799650426043264e-06, + "loss": 2.876, + "step": 17626 + }, + { + "epoch": 3.85, + "learning_rate": 3.0798558007428446e-06, + "loss": 1.4368, + "step": 17627 + }, + { + "epoch": 3.85, + "learning_rate": 3.0797465588813637e-06, + "loss": 1.7034, + "step": 17628 + }, + { + "epoch": 3.85, + "learning_rate": 3.0796373170198823e-06, + "loss": 1.8734, + "step": 17629 + }, + { + "epoch": 3.85, + "learning_rate": 3.0795280751584013e-06, + "loss": 1.4458, + "step": 17630 + }, + { + "epoch": 3.85, + "learning_rate": 3.0794188332969195e-06, + "loss": 3.1725, + "step": 17631 + }, + { + "epoch": 3.85, + "learning_rate": 3.079309591435438e-06, + "loss": 2.5658, + "step": 17632 + }, + { + "epoch": 3.85, + "learning_rate": 3.079200349573957e-06, + "loss": 2.3702, + "step": 17633 + }, + { + "epoch": 3.85, + "learning_rate": 3.0790911077124757e-06, + "loss": 1.8141, + "step": 17634 + }, + { + "epoch": 3.85, + "learning_rate": 3.0789818658509944e-06, + "loss": 0.9134, + "step": 17635 + }, + { + "epoch": 3.85, + "learning_rate": 3.078872623989513e-06, + "loss": 1.0236, + "step": 17636 + }, + { + "epoch": 3.85, + "learning_rate": 3.0787633821280316e-06, + "loss": 3.269, + "step": 17637 + }, + { + "epoch": 3.85, + "learning_rate": 3.0786541402665506e-06, + "loss": 1.0754, + "step": 17638 + }, + { + "epoch": 3.85, + "learning_rate": 3.0785448984050692e-06, + "loss": 3.1364, + "step": 17639 + }, + { + "epoch": 3.85, + "learning_rate": 3.078435656543588e-06, + "loss": 1.0351, + "step": 17640 + }, + { + "epoch": 3.85, + "learning_rate": 3.0783264146821064e-06, + "loss": 2.2176, + "step": 17641 + }, + { + "epoch": 3.85, + "learning_rate": 3.078217172820625e-06, + "loss": 3.6864, + "step": 17642 + }, + { + "epoch": 3.85, + "learning_rate": 3.0781079309591437e-06, + "loss": 1.7688, + "step": 17643 + }, + { + "epoch": 3.85, + "learning_rate": 3.0779986890976627e-06, + "loss": 1.5148, + "step": 17644 + }, + { + "epoch": 3.85, + "learning_rate": 3.077889447236181e-06, + "loss": 1.4437, + "step": 17645 + }, + { + "epoch": 3.85, + "learning_rate": 3.0777802053746995e-06, + "loss": 3.009, + "step": 17646 + }, + { + "epoch": 3.85, + "learning_rate": 3.0776709635132185e-06, + "loss": 2.5286, + "step": 17647 + }, + { + "epoch": 3.85, + "learning_rate": 3.077561721651737e-06, + "loss": 0.9666, + "step": 17648 + }, + { + "epoch": 3.85, + "learning_rate": 3.077452479790256e-06, + "loss": 0.8971, + "step": 17649 + }, + { + "epoch": 3.85, + "learning_rate": 3.0773432379287744e-06, + "loss": 0.7185, + "step": 17650 + }, + { + "epoch": 3.85, + "learning_rate": 3.077233996067293e-06, + "loss": 1.4095, + "step": 17651 + }, + { + "epoch": 3.85, + "learning_rate": 3.077124754205812e-06, + "loss": 2.2815, + "step": 17652 + }, + { + "epoch": 3.85, + "learning_rate": 3.0770155123443306e-06, + "loss": 2.3524, + "step": 17653 + }, + { + "epoch": 3.85, + "learning_rate": 3.0769062704828497e-06, + "loss": 2.4949, + "step": 17654 + }, + { + "epoch": 3.85, + "learning_rate": 3.076797028621368e-06, + "loss": 2.9603, + "step": 17655 + }, + { + "epoch": 3.85, + "learning_rate": 3.0766877867598865e-06, + "loss": 1.3439, + "step": 17656 + }, + { + "epoch": 3.85, + "learning_rate": 3.0765785448984055e-06, + "loss": 1.1532, + "step": 17657 + }, + { + "epoch": 3.85, + "learning_rate": 3.076469303036924e-06, + "loss": 2.8341, + "step": 17658 + }, + { + "epoch": 3.85, + "learning_rate": 3.076360061175443e-06, + "loss": 1.7976, + "step": 17659 + }, + { + "epoch": 3.85, + "learning_rate": 3.0762508193139613e-06, + "loss": 1.944, + "step": 17660 + }, + { + "epoch": 3.85, + "learning_rate": 3.07614157745248e-06, + "loss": 2.8313, + "step": 17661 + }, + { + "epoch": 3.85, + "learning_rate": 3.076032335590999e-06, + "loss": 3.4446, + "step": 17662 + }, + { + "epoch": 3.85, + "learning_rate": 3.0759230937295176e-06, + "loss": 0.8879, + "step": 17663 + }, + { + "epoch": 3.86, + "learning_rate": 3.0758138518680358e-06, + "loss": 1.9252, + "step": 17664 + }, + { + "epoch": 3.86, + "learning_rate": 3.075704610006555e-06, + "loss": 2.1988, + "step": 17665 + }, + { + "epoch": 3.86, + "learning_rate": 3.0755953681450734e-06, + "loss": 1.3318, + "step": 17666 + }, + { + "epoch": 3.86, + "learning_rate": 3.075486126283592e-06, + "loss": 1.2258, + "step": 17667 + }, + { + "epoch": 3.86, + "learning_rate": 3.075376884422111e-06, + "loss": 2.1647, + "step": 17668 + }, + { + "epoch": 3.86, + "learning_rate": 3.0752676425606292e-06, + "loss": 1.8963, + "step": 17669 + }, + { + "epoch": 3.86, + "learning_rate": 3.075158400699148e-06, + "loss": 1.6937, + "step": 17670 + }, + { + "epoch": 3.86, + "learning_rate": 3.075049158837667e-06, + "loss": 2.438, + "step": 17671 + }, + { + "epoch": 3.86, + "learning_rate": 3.0749399169761855e-06, + "loss": 3.1636, + "step": 17672 + }, + { + "epoch": 3.86, + "learning_rate": 3.0748306751147045e-06, + "loss": 2.0377, + "step": 17673 + }, + { + "epoch": 3.86, + "learning_rate": 3.0747214332532227e-06, + "loss": 1.2334, + "step": 17674 + }, + { + "epoch": 3.86, + "learning_rate": 3.0746121913917413e-06, + "loss": 1.7258, + "step": 17675 + }, + { + "epoch": 3.86, + "learning_rate": 3.0745029495302604e-06, + "loss": 1.9373, + "step": 17676 + }, + { + "epoch": 3.86, + "learning_rate": 3.074393707668779e-06, + "loss": 1.517, + "step": 17677 + }, + { + "epoch": 3.86, + "learning_rate": 3.074284465807297e-06, + "loss": 0.5663, + "step": 17678 + }, + { + "epoch": 3.86, + "learning_rate": 3.074175223945816e-06, + "loss": 2.3108, + "step": 17679 + }, + { + "epoch": 3.86, + "learning_rate": 3.074065982084335e-06, + "loss": 1.5145, + "step": 17680 + }, + { + "epoch": 3.86, + "learning_rate": 3.073956740222854e-06, + "loss": 1.2037, + "step": 17681 + }, + { + "epoch": 3.86, + "learning_rate": 3.0738474983613725e-06, + "loss": 1.7602, + "step": 17682 + }, + { + "epoch": 3.86, + "learning_rate": 3.0737382564998906e-06, + "loss": 1.6868, + "step": 17683 + }, + { + "epoch": 3.86, + "learning_rate": 3.0736290146384097e-06, + "loss": 1.332, + "step": 17684 + }, + { + "epoch": 3.86, + "learning_rate": 3.0735197727769283e-06, + "loss": 1.9288, + "step": 17685 + }, + { + "epoch": 3.86, + "learning_rate": 3.0734105309154473e-06, + "loss": 1.6203, + "step": 17686 + }, + { + "epoch": 3.86, + "learning_rate": 3.073301289053966e-06, + "loss": 2.3045, + "step": 17687 + }, + { + "epoch": 3.86, + "learning_rate": 3.073192047192484e-06, + "loss": 1.6532, + "step": 17688 + }, + { + "epoch": 3.86, + "learning_rate": 3.073082805331003e-06, + "loss": 2.908, + "step": 17689 + }, + { + "epoch": 3.86, + "learning_rate": 3.0729735634695218e-06, + "loss": 2.2152, + "step": 17690 + }, + { + "epoch": 3.86, + "learning_rate": 3.072864321608041e-06, + "loss": 0.7015, + "step": 17691 + }, + { + "epoch": 3.86, + "learning_rate": 3.0727550797465594e-06, + "loss": 2.3562, + "step": 17692 + }, + { + "epoch": 3.86, + "learning_rate": 3.0726458378850776e-06, + "loss": 1.6591, + "step": 17693 + }, + { + "epoch": 3.86, + "learning_rate": 3.0725365960235966e-06, + "loss": 2.0675, + "step": 17694 + }, + { + "epoch": 3.86, + "learning_rate": 3.0724273541621153e-06, + "loss": 2.4542, + "step": 17695 + }, + { + "epoch": 3.86, + "learning_rate": 3.072318112300634e-06, + "loss": 3.5131, + "step": 17696 + }, + { + "epoch": 3.86, + "learning_rate": 3.0722088704391525e-06, + "loss": 2.3205, + "step": 17697 + }, + { + "epoch": 3.86, + "learning_rate": 3.072099628577671e-06, + "loss": 2.6152, + "step": 17698 + }, + { + "epoch": 3.86, + "learning_rate": 3.0719903867161897e-06, + "loss": 1.8376, + "step": 17699 + }, + { + "epoch": 3.86, + "learning_rate": 3.0718811448547087e-06, + "loss": 1.8396, + "step": 17700 + }, + { + "epoch": 3.86, + "learning_rate": 3.0717719029932273e-06, + "loss": 1.6174, + "step": 17701 + }, + { + "epoch": 3.86, + "learning_rate": 3.0716626611317455e-06, + "loss": 2.1203, + "step": 17702 + }, + { + "epoch": 3.86, + "learning_rate": 3.0715534192702646e-06, + "loss": 3.0692, + "step": 17703 + }, + { + "epoch": 3.86, + "learning_rate": 3.071444177408783e-06, + "loss": 2.5964, + "step": 17704 + }, + { + "epoch": 3.86, + "learning_rate": 3.071334935547302e-06, + "loss": 0.3075, + "step": 17705 + }, + { + "epoch": 3.86, + "learning_rate": 3.071225693685821e-06, + "loss": 2.042, + "step": 17706 + }, + { + "epoch": 3.86, + "learning_rate": 3.071116451824339e-06, + "loss": 1.7206, + "step": 17707 + }, + { + "epoch": 3.86, + "learning_rate": 3.071007209962858e-06, + "loss": 2.4091, + "step": 17708 + }, + { + "epoch": 3.86, + "learning_rate": 3.0708979681013767e-06, + "loss": 1.1818, + "step": 17709 + }, + { + "epoch": 3.87, + "learning_rate": 3.0707887262398957e-06, + "loss": 1.6885, + "step": 17710 + }, + { + "epoch": 3.87, + "learning_rate": 3.070679484378414e-06, + "loss": 2.9726, + "step": 17711 + }, + { + "epoch": 3.87, + "learning_rate": 3.0705702425169325e-06, + "loss": 1.4052, + "step": 17712 + }, + { + "epoch": 3.87, + "learning_rate": 3.0704610006554515e-06, + "loss": 1.3209, + "step": 17713 + }, + { + "epoch": 3.87, + "learning_rate": 3.07035175879397e-06, + "loss": 1.0951, + "step": 17714 + }, + { + "epoch": 3.87, + "learning_rate": 3.070242516932489e-06, + "loss": 1.7115, + "step": 17715 + }, + { + "epoch": 3.87, + "learning_rate": 3.0701332750710074e-06, + "loss": 1.8852, + "step": 17716 + }, + { + "epoch": 3.87, + "learning_rate": 3.070024033209526e-06, + "loss": 1.2818, + "step": 17717 + }, + { + "epoch": 3.87, + "learning_rate": 3.069914791348045e-06, + "loss": 1.7533, + "step": 17718 + }, + { + "epoch": 3.87, + "learning_rate": 3.0698055494865636e-06, + "loss": 1.4495, + "step": 17719 + }, + { + "epoch": 3.87, + "learning_rate": 3.0696963076250822e-06, + "loss": 2.3654, + "step": 17720 + }, + { + "epoch": 3.87, + "learning_rate": 3.069587065763601e-06, + "loss": 2.5868, + "step": 17721 + }, + { + "epoch": 3.87, + "learning_rate": 3.0694778239021194e-06, + "loss": 1.6662, + "step": 17722 + }, + { + "epoch": 3.87, + "learning_rate": 3.069368582040638e-06, + "loss": 1.6176, + "step": 17723 + }, + { + "epoch": 3.87, + "learning_rate": 3.069259340179157e-06, + "loss": 2.6967, + "step": 17724 + }, + { + "epoch": 3.87, + "learning_rate": 3.0691500983176757e-06, + "loss": 1.895, + "step": 17725 + }, + { + "epoch": 3.87, + "learning_rate": 3.069040856456194e-06, + "loss": 0.8581, + "step": 17726 + }, + { + "epoch": 3.87, + "learning_rate": 3.068931614594713e-06, + "loss": 1.6036, + "step": 17727 + }, + { + "epoch": 3.87, + "learning_rate": 3.0688223727332315e-06, + "loss": 2.0848, + "step": 17728 + }, + { + "epoch": 3.87, + "learning_rate": 3.0687131308717506e-06, + "loss": 1.6461, + "step": 17729 + }, + { + "epoch": 3.87, + "learning_rate": 3.0686038890102688e-06, + "loss": 3.3589, + "step": 17730 + }, + { + "epoch": 3.87, + "learning_rate": 3.0684946471487874e-06, + "loss": 1.5197, + "step": 17731 + }, + { + "epoch": 3.87, + "learning_rate": 3.0683854052873064e-06, + "loss": 1.2404, + "step": 17732 + }, + { + "epoch": 3.87, + "learning_rate": 3.068276163425825e-06, + "loss": 3.1445, + "step": 17733 + }, + { + "epoch": 3.87, + "learning_rate": 3.068166921564344e-06, + "loss": 1.012, + "step": 17734 + }, + { + "epoch": 3.87, + "learning_rate": 3.0680576797028622e-06, + "loss": 1.2747, + "step": 17735 + }, + { + "epoch": 3.87, + "learning_rate": 3.067948437841381e-06, + "loss": 2.4131, + "step": 17736 + }, + { + "epoch": 3.87, + "learning_rate": 3.0678391959799e-06, + "loss": 1.5994, + "step": 17737 + }, + { + "epoch": 3.87, + "learning_rate": 3.0677299541184185e-06, + "loss": 2.952, + "step": 17738 + }, + { + "epoch": 3.87, + "learning_rate": 3.0676207122569375e-06, + "loss": 3.5224, + "step": 17739 + }, + { + "epoch": 3.87, + "learning_rate": 3.0675114703954557e-06, + "loss": 2.384, + "step": 17740 + }, + { + "epoch": 3.87, + "learning_rate": 3.0674022285339743e-06, + "loss": 2.263, + "step": 17741 + }, + { + "epoch": 3.87, + "learning_rate": 3.0672929866724934e-06, + "loss": 1.1043, + "step": 17742 + }, + { + "epoch": 3.87, + "learning_rate": 3.067183744811012e-06, + "loss": 1.3261, + "step": 17743 + }, + { + "epoch": 3.87, + "learning_rate": 3.06707450294953e-06, + "loss": 2.9528, + "step": 17744 + }, + { + "epoch": 3.87, + "learning_rate": 3.066965261088049e-06, + "loss": 2.7168, + "step": 17745 + }, + { + "epoch": 3.87, + "learning_rate": 3.066856019226568e-06, + "loss": 2.2365, + "step": 17746 + }, + { + "epoch": 3.87, + "learning_rate": 3.066746777365087e-06, + "loss": 1.8393, + "step": 17747 + }, + { + "epoch": 3.87, + "learning_rate": 3.0666375355036054e-06, + "loss": 1.4581, + "step": 17748 + }, + { + "epoch": 3.87, + "learning_rate": 3.0665282936421236e-06, + "loss": 2.8965, + "step": 17749 + }, + { + "epoch": 3.87, + "learning_rate": 3.0664190517806427e-06, + "loss": 1.5477, + "step": 17750 + }, + { + "epoch": 3.87, + "learning_rate": 3.0663098099191613e-06, + "loss": 1.1352, + "step": 17751 + }, + { + "epoch": 3.87, + "learning_rate": 3.06620056805768e-06, + "loss": 1.9678, + "step": 17752 + }, + { + "epoch": 3.87, + "learning_rate": 3.066091326196199e-06, + "loss": 1.3115, + "step": 17753 + }, + { + "epoch": 3.87, + "learning_rate": 3.065982084334717e-06, + "loss": 2.1935, + "step": 17754 + }, + { + "epoch": 3.87, + "learning_rate": 3.0658728424732357e-06, + "loss": 0.2959, + "step": 17755 + }, + { + "epoch": 3.88, + "learning_rate": 3.0657636006117548e-06, + "loss": 1.477, + "step": 17756 + }, + { + "epoch": 3.88, + "learning_rate": 3.0656543587502734e-06, + "loss": 1.7429, + "step": 17757 + }, + { + "epoch": 3.88, + "learning_rate": 3.0655451168887924e-06, + "loss": 3.6174, + "step": 17758 + }, + { + "epoch": 3.88, + "learning_rate": 3.0654358750273106e-06, + "loss": 2.5576, + "step": 17759 + }, + { + "epoch": 3.88, + "learning_rate": 3.065326633165829e-06, + "loss": 2.4457, + "step": 17760 + }, + { + "epoch": 3.88, + "learning_rate": 3.0652173913043482e-06, + "loss": 1.5266, + "step": 17761 + }, + { + "epoch": 3.88, + "learning_rate": 3.065108149442867e-06, + "loss": 1.832, + "step": 17762 + }, + { + "epoch": 3.88, + "learning_rate": 3.064998907581385e-06, + "loss": 1.5354, + "step": 17763 + }, + { + "epoch": 3.88, + "learning_rate": 3.064889665719904e-06, + "loss": 2.6262, + "step": 17764 + }, + { + "epoch": 3.88, + "learning_rate": 3.0647804238584227e-06, + "loss": 2.1918, + "step": 17765 + }, + { + "epoch": 3.88, + "learning_rate": 3.0646711819969417e-06, + "loss": 3.0158, + "step": 17766 + }, + { + "epoch": 3.88, + "learning_rate": 3.0645619401354603e-06, + "loss": 1.2361, + "step": 17767 + }, + { + "epoch": 3.88, + "learning_rate": 3.0644526982739785e-06, + "loss": 2.0926, + "step": 17768 + }, + { + "epoch": 3.88, + "learning_rate": 3.0643434564124975e-06, + "loss": 1.4912, + "step": 17769 + }, + { + "epoch": 3.88, + "learning_rate": 3.064234214551016e-06, + "loss": 1.8243, + "step": 17770 + }, + { + "epoch": 3.88, + "learning_rate": 3.064124972689535e-06, + "loss": 1.7567, + "step": 17771 + }, + { + "epoch": 3.88, + "learning_rate": 3.064015730828054e-06, + "loss": 1.9989, + "step": 17772 + }, + { + "epoch": 3.88, + "learning_rate": 3.063906488966572e-06, + "loss": 1.7773, + "step": 17773 + }, + { + "epoch": 3.88, + "learning_rate": 3.063797247105091e-06, + "loss": 3.2749, + "step": 17774 + }, + { + "epoch": 3.88, + "learning_rate": 3.0636880052436096e-06, + "loss": 1.1513, + "step": 17775 + }, + { + "epoch": 3.88, + "learning_rate": 3.0635787633821282e-06, + "loss": 3.0946, + "step": 17776 + }, + { + "epoch": 3.88, + "learning_rate": 3.063469521520647e-06, + "loss": 2.3196, + "step": 17777 + }, + { + "epoch": 3.88, + "learning_rate": 3.0633602796591655e-06, + "loss": 2.3329, + "step": 17778 + }, + { + "epoch": 3.88, + "learning_rate": 3.063251037797684e-06, + "loss": 1.4535, + "step": 17779 + }, + { + "epoch": 3.88, + "learning_rate": 3.063141795936203e-06, + "loss": 1.6334, + "step": 17780 + }, + { + "epoch": 3.88, + "learning_rate": 3.0630325540747217e-06, + "loss": 1.7053, + "step": 17781 + }, + { + "epoch": 3.88, + "learning_rate": 3.0629233122132403e-06, + "loss": 2.2728, + "step": 17782 + }, + { + "epoch": 3.88, + "learning_rate": 3.062814070351759e-06, + "loss": 2.4767, + "step": 17783 + }, + { + "epoch": 3.88, + "learning_rate": 3.0627048284902776e-06, + "loss": 1.3045, + "step": 17784 + }, + { + "epoch": 3.88, + "learning_rate": 3.0625955866287966e-06, + "loss": 1.7115, + "step": 17785 + }, + { + "epoch": 3.88, + "learning_rate": 3.062486344767315e-06, + "loss": 1.0476, + "step": 17786 + }, + { + "epoch": 3.88, + "learning_rate": 3.0623771029058334e-06, + "loss": 1.6468, + "step": 17787 + }, + { + "epoch": 3.88, + "learning_rate": 3.0622678610443524e-06, + "loss": 1.1029, + "step": 17788 + }, + { + "epoch": 3.88, + "learning_rate": 3.062158619182871e-06, + "loss": 1.3675, + "step": 17789 + }, + { + "epoch": 3.88, + "learning_rate": 3.06204937732139e-06, + "loss": 1.2855, + "step": 17790 + }, + { + "epoch": 3.88, + "learning_rate": 3.0619401354599087e-06, + "loss": 0.5264, + "step": 17791 + }, + { + "epoch": 3.88, + "learning_rate": 3.061830893598427e-06, + "loss": 2.165, + "step": 17792 + }, + { + "epoch": 3.88, + "learning_rate": 3.061721651736946e-06, + "loss": 1.1514, + "step": 17793 + }, + { + "epoch": 3.88, + "learning_rate": 3.0616124098754645e-06, + "loss": 1.0142, + "step": 17794 + }, + { + "epoch": 3.88, + "learning_rate": 3.0615031680139835e-06, + "loss": 1.5972, + "step": 17795 + }, + { + "epoch": 3.88, + "learning_rate": 3.0613939261525017e-06, + "loss": 2.2684, + "step": 17796 + }, + { + "epoch": 3.88, + "learning_rate": 3.0612846842910203e-06, + "loss": 0.9986, + "step": 17797 + }, + { + "epoch": 3.88, + "learning_rate": 3.0611754424295394e-06, + "loss": 1.0185, + "step": 17798 + }, + { + "epoch": 3.88, + "learning_rate": 3.061066200568058e-06, + "loss": 1.6778, + "step": 17799 + }, + { + "epoch": 3.88, + "learning_rate": 3.060956958706577e-06, + "loss": 2.0795, + "step": 17800 + }, + { + "epoch": 3.88, + "learning_rate": 3.0608477168450952e-06, + "loss": 2.5615, + "step": 17801 + }, + { + "epoch": 3.89, + "learning_rate": 3.060738474983614e-06, + "loss": 1.2322, + "step": 17802 + }, + { + "epoch": 3.89, + "learning_rate": 3.060629233122133e-06, + "loss": 2.1424, + "step": 17803 + }, + { + "epoch": 3.89, + "learning_rate": 3.0605199912606515e-06, + "loss": 1.041, + "step": 17804 + }, + { + "epoch": 3.89, + "learning_rate": 3.06041074939917e-06, + "loss": 0.5818, + "step": 17805 + }, + { + "epoch": 3.89, + "learning_rate": 3.0603015075376887e-06, + "loss": 1.7276, + "step": 17806 + }, + { + "epoch": 3.89, + "learning_rate": 3.0601922656762073e-06, + "loss": 1.841, + "step": 17807 + }, + { + "epoch": 3.89, + "learning_rate": 3.060083023814726e-06, + "loss": 2.4071, + "step": 17808 + }, + { + "epoch": 3.89, + "learning_rate": 3.059973781953245e-06, + "loss": 2.5482, + "step": 17809 + }, + { + "epoch": 3.89, + "learning_rate": 3.059864540091763e-06, + "loss": 3.575, + "step": 17810 + }, + { + "epoch": 3.89, + "learning_rate": 3.0597552982302817e-06, + "loss": 1.1303, + "step": 17811 + }, + { + "epoch": 3.89, + "learning_rate": 3.0596460563688008e-06, + "loss": 1.6192, + "step": 17812 + }, + { + "epoch": 3.89, + "learning_rate": 3.0595368145073194e-06, + "loss": 3.2456, + "step": 17813 + }, + { + "epoch": 3.89, + "learning_rate": 3.0594275726458384e-06, + "loss": 2.8246, + "step": 17814 + }, + { + "epoch": 3.89, + "learning_rate": 3.0593183307843566e-06, + "loss": 1.1344, + "step": 17815 + }, + { + "epoch": 3.89, + "learning_rate": 3.0592090889228752e-06, + "loss": 2.1833, + "step": 17816 + }, + { + "epoch": 3.89, + "learning_rate": 3.0590998470613943e-06, + "loss": 1.5628, + "step": 17817 + }, + { + "epoch": 3.89, + "learning_rate": 3.058990605199913e-06, + "loss": 1.8804, + "step": 17818 + }, + { + "epoch": 3.89, + "learning_rate": 3.058881363338432e-06, + "loss": 1.6749, + "step": 17819 + }, + { + "epoch": 3.89, + "learning_rate": 3.05877212147695e-06, + "loss": 0.8418, + "step": 17820 + }, + { + "epoch": 3.89, + "learning_rate": 3.0586628796154687e-06, + "loss": 0.9697, + "step": 17821 + }, + { + "epoch": 3.89, + "learning_rate": 3.0585536377539877e-06, + "loss": 1.438, + "step": 17822 + }, + { + "epoch": 3.89, + "learning_rate": 3.0584443958925064e-06, + "loss": 1.1268, + "step": 17823 + }, + { + "epoch": 3.89, + "learning_rate": 3.0583351540310254e-06, + "loss": 2.0111, + "step": 17824 + }, + { + "epoch": 3.89, + "learning_rate": 3.0582259121695436e-06, + "loss": 3.5884, + "step": 17825 + }, + { + "epoch": 3.89, + "learning_rate": 3.058116670308062e-06, + "loss": 0.5582, + "step": 17826 + }, + { + "epoch": 3.89, + "learning_rate": 3.0580074284465812e-06, + "loss": 2.8456, + "step": 17827 + }, + { + "epoch": 3.89, + "learning_rate": 3.0578981865851e-06, + "loss": 1.4939, + "step": 17828 + }, + { + "epoch": 3.89, + "learning_rate": 3.057788944723618e-06, + "loss": 2.8564, + "step": 17829 + }, + { + "epoch": 3.89, + "learning_rate": 3.057679702862137e-06, + "loss": 1.1388, + "step": 17830 + }, + { + "epoch": 3.89, + "learning_rate": 3.0575704610006557e-06, + "loss": 0.7346, + "step": 17831 + }, + { + "epoch": 3.89, + "learning_rate": 3.0574612191391743e-06, + "loss": 1.4214, + "step": 17832 + }, + { + "epoch": 3.89, + "learning_rate": 3.0573519772776933e-06, + "loss": 1.7324, + "step": 17833 + }, + { + "epoch": 3.89, + "learning_rate": 3.0572427354162115e-06, + "loss": 1.8279, + "step": 17834 + }, + { + "epoch": 3.89, + "learning_rate": 3.0571334935547305e-06, + "loss": 1.0028, + "step": 17835 + }, + { + "epoch": 3.89, + "learning_rate": 3.057024251693249e-06, + "loss": 1.6206, + "step": 17836 + }, + { + "epoch": 3.89, + "learning_rate": 3.0569150098317678e-06, + "loss": 1.3373, + "step": 17837 + }, + { + "epoch": 3.89, + "learning_rate": 3.0568057679702868e-06, + "loss": 0.865, + "step": 17838 + }, + { + "epoch": 3.89, + "learning_rate": 3.056696526108805e-06, + "loss": 3.3405, + "step": 17839 + }, + { + "epoch": 3.89, + "learning_rate": 3.0565872842473236e-06, + "loss": 2.147, + "step": 17840 + }, + { + "epoch": 3.89, + "learning_rate": 3.0564780423858426e-06, + "loss": 1.3251, + "step": 17841 + }, + { + "epoch": 3.89, + "learning_rate": 3.0563688005243612e-06, + "loss": 1.5386, + "step": 17842 + }, + { + "epoch": 3.89, + "learning_rate": 3.0562595586628794e-06, + "loss": 2.2873, + "step": 17843 + }, + { + "epoch": 3.89, + "learning_rate": 3.0561503168013985e-06, + "loss": 1.9539, + "step": 17844 + }, + { + "epoch": 3.89, + "learning_rate": 3.056041074939917e-06, + "loss": 1.4618, + "step": 17845 + }, + { + "epoch": 3.89, + "learning_rate": 3.055931833078436e-06, + "loss": 1.53, + "step": 17846 + }, + { + "epoch": 3.9, + "learning_rate": 3.0558225912169547e-06, + "loss": 3.1601, + "step": 17847 + }, + { + "epoch": 3.9, + "learning_rate": 3.055713349355473e-06, + "loss": 1.6688, + "step": 17848 + }, + { + "epoch": 3.9, + "learning_rate": 3.055604107493992e-06, + "loss": 1.057, + "step": 17849 + }, + { + "epoch": 3.9, + "learning_rate": 3.0554948656325105e-06, + "loss": 1.5855, + "step": 17850 + }, + { + "epoch": 3.9, + "learning_rate": 3.0553856237710296e-06, + "loss": 2.7778, + "step": 17851 + }, + { + "epoch": 3.9, + "learning_rate": 3.055276381909548e-06, + "loss": 1.1382, + "step": 17852 + }, + { + "epoch": 3.9, + "learning_rate": 3.0551671400480664e-06, + "loss": 2.0591, + "step": 17853 + }, + { + "epoch": 3.9, + "learning_rate": 3.0550578981865854e-06, + "loss": 1.4338, + "step": 17854 + }, + { + "epoch": 3.9, + "learning_rate": 3.054948656325104e-06, + "loss": 1.9208, + "step": 17855 + }, + { + "epoch": 3.9, + "learning_rate": 3.054839414463623e-06, + "loss": 2.0424, + "step": 17856 + }, + { + "epoch": 3.9, + "learning_rate": 3.0547301726021417e-06, + "loss": 2.3201, + "step": 17857 + }, + { + "epoch": 3.9, + "learning_rate": 3.05462093074066e-06, + "loss": 1.4626, + "step": 17858 + }, + { + "epoch": 3.9, + "learning_rate": 3.054511688879179e-06, + "loss": 1.9323, + "step": 17859 + }, + { + "epoch": 3.9, + "learning_rate": 3.0544024470176975e-06, + "loss": 2.418, + "step": 17860 + }, + { + "epoch": 3.9, + "learning_rate": 3.054293205156216e-06, + "loss": 1.687, + "step": 17861 + }, + { + "epoch": 3.9, + "learning_rate": 3.0541839632947347e-06, + "loss": 0.5026, + "step": 17862 + }, + { + "epoch": 3.9, + "learning_rate": 3.0540747214332533e-06, + "loss": 1.3112, + "step": 17863 + }, + { + "epoch": 3.9, + "learning_rate": 3.053965479571772e-06, + "loss": 1.2157, + "step": 17864 + }, + { + "epoch": 3.9, + "learning_rate": 3.053856237710291e-06, + "loss": 0.9253, + "step": 17865 + }, + { + "epoch": 3.9, + "learning_rate": 3.0537469958488096e-06, + "loss": 2.033, + "step": 17866 + }, + { + "epoch": 3.9, + "learning_rate": 3.0536377539873278e-06, + "loss": 1.01, + "step": 17867 + }, + { + "epoch": 3.9, + "learning_rate": 3.053528512125847e-06, + "loss": 1.6261, + "step": 17868 + }, + { + "epoch": 3.9, + "learning_rate": 3.0534192702643654e-06, + "loss": 3.1055, + "step": 17869 + }, + { + "epoch": 3.9, + "learning_rate": 3.0533100284028845e-06, + "loss": 1.4179, + "step": 17870 + }, + { + "epoch": 3.9, + "learning_rate": 3.053200786541403e-06, + "loss": 0.8692, + "step": 17871 + }, + { + "epoch": 3.9, + "learning_rate": 3.0530915446799213e-06, + "loss": 2.1262, + "step": 17872 + }, + { + "epoch": 3.9, + "learning_rate": 3.0529823028184403e-06, + "loss": 1.1717, + "step": 17873 + }, + { + "epoch": 3.9, + "learning_rate": 3.052873060956959e-06, + "loss": 1.749, + "step": 17874 + }, + { + "epoch": 3.9, + "learning_rate": 3.052763819095478e-06, + "loss": 0.8251, + "step": 17875 + }, + { + "epoch": 3.9, + "learning_rate": 3.052654577233996e-06, + "loss": 2.0437, + "step": 17876 + }, + { + "epoch": 3.9, + "learning_rate": 3.0525453353725147e-06, + "loss": 1.451, + "step": 17877 + }, + { + "epoch": 3.9, + "learning_rate": 3.0524360935110338e-06, + "loss": 1.9627, + "step": 17878 + }, + { + "epoch": 3.9, + "learning_rate": 3.0523268516495524e-06, + "loss": 1.4904, + "step": 17879 + }, + { + "epoch": 3.9, + "learning_rate": 3.0522176097880714e-06, + "loss": 2.1007, + "step": 17880 + }, + { + "epoch": 3.9, + "learning_rate": 3.0521083679265896e-06, + "loss": 0.5963, + "step": 17881 + }, + { + "epoch": 3.9, + "learning_rate": 3.051999126065108e-06, + "loss": 1.8141, + "step": 17882 + }, + { + "epoch": 3.9, + "learning_rate": 3.0518898842036272e-06, + "loss": 1.8307, + "step": 17883 + }, + { + "epoch": 3.9, + "learning_rate": 3.051780642342146e-06, + "loss": 1.8931, + "step": 17884 + }, + { + "epoch": 3.9, + "learning_rate": 3.051671400480665e-06, + "loss": 1.7294, + "step": 17885 + }, + { + "epoch": 3.9, + "learning_rate": 3.051562158619183e-06, + "loss": 1.1165, + "step": 17886 + }, + { + "epoch": 3.9, + "learning_rate": 3.0514529167577017e-06, + "loss": 1.9574, + "step": 17887 + }, + { + "epoch": 3.9, + "learning_rate": 3.0513436748962207e-06, + "loss": 1.6342, + "step": 17888 + }, + { + "epoch": 3.9, + "learning_rate": 3.0512344330347393e-06, + "loss": 0.6354, + "step": 17889 + }, + { + "epoch": 3.9, + "learning_rate": 3.051125191173258e-06, + "loss": 3.0076, + "step": 17890 + }, + { + "epoch": 3.9, + "learning_rate": 3.0510159493117766e-06, + "loss": 1.6128, + "step": 17891 + }, + { + "epoch": 3.9, + "learning_rate": 3.050906707450295e-06, + "loss": 1.7409, + "step": 17892 + }, + { + "epoch": 3.91, + "learning_rate": 3.0507974655888138e-06, + "loss": 1.3614, + "step": 17893 + }, + { + "epoch": 3.91, + "learning_rate": 3.050688223727333e-06, + "loss": 1.9708, + "step": 17894 + }, + { + "epoch": 3.91, + "learning_rate": 3.050578981865851e-06, + "loss": 1.4899, + "step": 17895 + }, + { + "epoch": 3.91, + "learning_rate": 3.0504697400043696e-06, + "loss": 1.8052, + "step": 17896 + }, + { + "epoch": 3.91, + "learning_rate": 3.0503604981428886e-06, + "loss": 1.9444, + "step": 17897 + }, + { + "epoch": 3.91, + "learning_rate": 3.0502512562814073e-06, + "loss": 1.9537, + "step": 17898 + }, + { + "epoch": 3.91, + "learning_rate": 3.0501420144199263e-06, + "loss": 3.1263, + "step": 17899 + }, + { + "epoch": 3.91, + "learning_rate": 3.0500327725584445e-06, + "loss": 1.5405, + "step": 17900 + }, + { + "epoch": 3.91, + "learning_rate": 3.049923530696963e-06, + "loss": 3.3543, + "step": 17901 + }, + { + "epoch": 3.91, + "learning_rate": 3.049814288835482e-06, + "loss": 3.1209, + "step": 17902 + }, + { + "epoch": 3.91, + "learning_rate": 3.0497050469740007e-06, + "loss": 1.9443, + "step": 17903 + }, + { + "epoch": 3.91, + "learning_rate": 3.0495958051125198e-06, + "loss": 3.1827, + "step": 17904 + }, + { + "epoch": 3.91, + "learning_rate": 3.049486563251038e-06, + "loss": 3.1059, + "step": 17905 + }, + { + "epoch": 3.91, + "learning_rate": 3.0493773213895566e-06, + "loss": 1.0452, + "step": 17906 + }, + { + "epoch": 3.91, + "learning_rate": 3.0492680795280756e-06, + "loss": 1.0629, + "step": 17907 + }, + { + "epoch": 3.91, + "learning_rate": 3.0491588376665942e-06, + "loss": 1.3661, + "step": 17908 + }, + { + "epoch": 3.91, + "learning_rate": 3.0490495958051124e-06, + "loss": 0.508, + "step": 17909 + }, + { + "epoch": 3.91, + "learning_rate": 3.0489403539436314e-06, + "loss": 1.7272, + "step": 17910 + }, + { + "epoch": 3.91, + "learning_rate": 3.04883111208215e-06, + "loss": 3.2283, + "step": 17911 + }, + { + "epoch": 3.91, + "learning_rate": 3.048721870220669e-06, + "loss": 3.199, + "step": 17912 + }, + { + "epoch": 3.91, + "learning_rate": 3.0486126283591877e-06, + "loss": 1.1662, + "step": 17913 + }, + { + "epoch": 3.91, + "learning_rate": 3.048503386497706e-06, + "loss": 1.2348, + "step": 17914 + }, + { + "epoch": 3.91, + "learning_rate": 3.048394144636225e-06, + "loss": 0.9355, + "step": 17915 + }, + { + "epoch": 3.91, + "learning_rate": 3.0482849027747435e-06, + "loss": 0.6935, + "step": 17916 + }, + { + "epoch": 3.91, + "learning_rate": 3.048175660913262e-06, + "loss": 1.4607, + "step": 17917 + }, + { + "epoch": 3.91, + "learning_rate": 3.048066419051781e-06, + "loss": 0.7623, + "step": 17918 + }, + { + "epoch": 3.91, + "learning_rate": 3.0479571771902994e-06, + "loss": 0.8899, + "step": 17919 + }, + { + "epoch": 3.91, + "learning_rate": 3.047847935328818e-06, + "loss": 2.3935, + "step": 17920 + }, + { + "epoch": 3.91, + "learning_rate": 3.047738693467337e-06, + "loss": 2.4915, + "step": 17921 + }, + { + "epoch": 3.91, + "learning_rate": 3.0476294516058556e-06, + "loss": 3.0079, + "step": 17922 + }, + { + "epoch": 3.91, + "learning_rate": 3.0475202097443747e-06, + "loss": 2.2212, + "step": 17923 + }, + { + "epoch": 3.91, + "learning_rate": 3.047410967882893e-06, + "loss": 1.3363, + "step": 17924 + }, + { + "epoch": 3.91, + "learning_rate": 3.0473017260214114e-06, + "loss": 1.7102, + "step": 17925 + }, + { + "epoch": 3.91, + "learning_rate": 3.0471924841599305e-06, + "loss": 1.7304, + "step": 17926 + }, + { + "epoch": 3.91, + "learning_rate": 3.047083242298449e-06, + "loss": 1.2339, + "step": 17927 + }, + { + "epoch": 3.91, + "learning_rate": 3.0469740004369673e-06, + "loss": 2.9672, + "step": 17928 + }, + { + "epoch": 3.91, + "learning_rate": 3.0468647585754863e-06, + "loss": 2.0626, + "step": 17929 + }, + { + "epoch": 3.91, + "learning_rate": 3.046755516714005e-06, + "loss": 2.2105, + "step": 17930 + }, + { + "epoch": 3.91, + "learning_rate": 3.046646274852524e-06, + "loss": 2.5851, + "step": 17931 + }, + { + "epoch": 3.91, + "learning_rate": 3.0465370329910426e-06, + "loss": 2.6842, + "step": 17932 + }, + { + "epoch": 3.91, + "learning_rate": 3.0464277911295608e-06, + "loss": 2.0184, + "step": 17933 + }, + { + "epoch": 3.91, + "learning_rate": 3.04631854926808e-06, + "loss": 2.3303, + "step": 17934 + }, + { + "epoch": 3.91, + "learning_rate": 3.0462093074065984e-06, + "loss": 1.1438, + "step": 17935 + }, + { + "epoch": 3.91, + "learning_rate": 3.0461000655451174e-06, + "loss": 2.4658, + "step": 17936 + }, + { + "epoch": 3.91, + "learning_rate": 3.045990823683636e-06, + "loss": 0.6404, + "step": 17937 + }, + { + "epoch": 3.91, + "learning_rate": 3.0458815818221542e-06, + "loss": 0.5046, + "step": 17938 + }, + { + "epoch": 3.92, + "learning_rate": 3.0457723399606733e-06, + "loss": 2.7432, + "step": 17939 + }, + { + "epoch": 3.92, + "learning_rate": 3.045663098099192e-06, + "loss": 0.9966, + "step": 17940 + }, + { + "epoch": 3.92, + "learning_rate": 3.045553856237711e-06, + "loss": 2.0662, + "step": 17941 + }, + { + "epoch": 3.92, + "learning_rate": 3.045444614376229e-06, + "loss": 0.943, + "step": 17942 + }, + { + "epoch": 3.92, + "learning_rate": 3.0453353725147477e-06, + "loss": 1.9412, + "step": 17943 + }, + { + "epoch": 3.92, + "learning_rate": 3.0452261306532668e-06, + "loss": 1.2472, + "step": 17944 + }, + { + "epoch": 3.92, + "learning_rate": 3.0451168887917854e-06, + "loss": 1.7763, + "step": 17945 + }, + { + "epoch": 3.92, + "learning_rate": 3.045007646930304e-06, + "loss": 2.5022, + "step": 17946 + }, + { + "epoch": 3.92, + "learning_rate": 3.0448984050688226e-06, + "loss": 3.0228, + "step": 17947 + }, + { + "epoch": 3.92, + "learning_rate": 3.044789163207341e-06, + "loss": 4.0219, + "step": 17948 + }, + { + "epoch": 3.92, + "learning_rate": 3.04467992134586e-06, + "loss": 2.5487, + "step": 17949 + }, + { + "epoch": 3.92, + "learning_rate": 3.044570679484379e-06, + "loss": 1.6029, + "step": 17950 + }, + { + "epoch": 3.92, + "learning_rate": 3.0444614376228975e-06, + "loss": 1.1208, + "step": 17951 + }, + { + "epoch": 3.92, + "learning_rate": 3.0443521957614156e-06, + "loss": 0.4654, + "step": 17952 + }, + { + "epoch": 3.92, + "learning_rate": 3.0442429538999347e-06, + "loss": 2.6387, + "step": 17953 + }, + { + "epoch": 3.92, + "learning_rate": 3.0441337120384533e-06, + "loss": 2.3792, + "step": 17954 + }, + { + "epoch": 3.92, + "learning_rate": 3.0440244701769723e-06, + "loss": 2.996, + "step": 17955 + }, + { + "epoch": 3.92, + "learning_rate": 3.043915228315491e-06, + "loss": 2.7916, + "step": 17956 + }, + { + "epoch": 3.92, + "learning_rate": 3.043805986454009e-06, + "loss": 2.2144, + "step": 17957 + }, + { + "epoch": 3.92, + "learning_rate": 3.043696744592528e-06, + "loss": 0.9794, + "step": 17958 + }, + { + "epoch": 3.92, + "learning_rate": 3.0435875027310468e-06, + "loss": 3.028, + "step": 17959 + }, + { + "epoch": 3.92, + "learning_rate": 3.043478260869566e-06, + "loss": 1.943, + "step": 17960 + }, + { + "epoch": 3.92, + "learning_rate": 3.043369019008084e-06, + "loss": 1.1408, + "step": 17961 + }, + { + "epoch": 3.92, + "learning_rate": 3.0432597771466026e-06, + "loss": 0.8761, + "step": 17962 + }, + { + "epoch": 3.92, + "learning_rate": 3.0431505352851216e-06, + "loss": 0.5264, + "step": 17963 + }, + { + "epoch": 3.92, + "learning_rate": 3.0430412934236402e-06, + "loss": 2.5473, + "step": 17964 + }, + { + "epoch": 3.92, + "learning_rate": 3.0429320515621593e-06, + "loss": 1.6165, + "step": 17965 + }, + { + "epoch": 3.92, + "learning_rate": 3.0428228097006775e-06, + "loss": 2.2954, + "step": 17966 + }, + { + "epoch": 3.92, + "learning_rate": 3.042713567839196e-06, + "loss": 1.5434, + "step": 17967 + }, + { + "epoch": 3.92, + "learning_rate": 3.042604325977715e-06, + "loss": 1.4491, + "step": 17968 + }, + { + "epoch": 3.92, + "learning_rate": 3.0424950841162337e-06, + "loss": 2.6012, + "step": 17969 + }, + { + "epoch": 3.92, + "learning_rate": 3.0423858422547523e-06, + "loss": 1.01, + "step": 17970 + }, + { + "epoch": 3.92, + "learning_rate": 3.042276600393271e-06, + "loss": 2.2122, + "step": 17971 + }, + { + "epoch": 3.92, + "learning_rate": 3.0421673585317896e-06, + "loss": 1.0718, + "step": 17972 + }, + { + "epoch": 3.92, + "learning_rate": 3.042058116670308e-06, + "loss": 1.6803, + "step": 17973 + }, + { + "epoch": 3.92, + "learning_rate": 3.041948874808827e-06, + "loss": 2.804, + "step": 17974 + }, + { + "epoch": 3.92, + "learning_rate": 3.0418396329473454e-06, + "loss": 2.0647, + "step": 17975 + }, + { + "epoch": 3.92, + "learning_rate": 3.0417303910858644e-06, + "loss": 2.599, + "step": 17976 + }, + { + "epoch": 3.92, + "learning_rate": 3.041621149224383e-06, + "loss": 1.3593, + "step": 17977 + }, + { + "epoch": 3.92, + "learning_rate": 3.0415119073629016e-06, + "loss": 1.2976, + "step": 17978 + }, + { + "epoch": 3.92, + "learning_rate": 3.0414026655014207e-06, + "loss": 0.9021, + "step": 17979 + }, + { + "epoch": 3.92, + "learning_rate": 3.041293423639939e-06, + "loss": 2.6141, + "step": 17980 + }, + { + "epoch": 3.92, + "learning_rate": 3.0411841817784575e-06, + "loss": 0.7576, + "step": 17981 + }, + { + "epoch": 3.92, + "learning_rate": 3.0410749399169765e-06, + "loss": 0.6243, + "step": 17982 + }, + { + "epoch": 3.92, + "learning_rate": 3.040965698055495e-06, + "loss": 0.8429, + "step": 17983 + }, + { + "epoch": 3.92, + "learning_rate": 3.040856456194014e-06, + "loss": 2.6367, + "step": 17984 + }, + { + "epoch": 3.93, + "learning_rate": 3.0407472143325323e-06, + "loss": 1.6301, + "step": 17985 + }, + { + "epoch": 3.93, + "learning_rate": 3.040637972471051e-06, + "loss": 2.2542, + "step": 17986 + }, + { + "epoch": 3.93, + "learning_rate": 3.04052873060957e-06, + "loss": 0.9764, + "step": 17987 + }, + { + "epoch": 3.93, + "learning_rate": 3.0404194887480886e-06, + "loss": 2.4075, + "step": 17988 + }, + { + "epoch": 3.93, + "learning_rate": 3.0403102468866068e-06, + "loss": 3.1946, + "step": 17989 + }, + { + "epoch": 3.93, + "learning_rate": 3.040201005025126e-06, + "loss": 1.5788, + "step": 17990 + }, + { + "epoch": 3.93, + "learning_rate": 3.0400917631636444e-06, + "loss": 0.3672, + "step": 17991 + }, + { + "epoch": 3.93, + "learning_rate": 3.0399825213021635e-06, + "loss": 1.0853, + "step": 17992 + }, + { + "epoch": 3.93, + "learning_rate": 3.039873279440682e-06, + "loss": 1.1296, + "step": 17993 + }, + { + "epoch": 3.93, + "learning_rate": 3.0397640375792003e-06, + "loss": 3.0123, + "step": 17994 + }, + { + "epoch": 3.93, + "learning_rate": 3.0396547957177193e-06, + "loss": 1.4167, + "step": 17995 + }, + { + "epoch": 3.93, + "learning_rate": 3.039545553856238e-06, + "loss": 1.0182, + "step": 17996 + }, + { + "epoch": 3.93, + "learning_rate": 3.039436311994757e-06, + "loss": 1.9119, + "step": 17997 + }, + { + "epoch": 3.93, + "learning_rate": 3.0393270701332756e-06, + "loss": 1.4484, + "step": 17998 + }, + { + "epoch": 3.93, + "learning_rate": 3.0392178282717937e-06, + "loss": 1.942, + "step": 17999 + }, + { + "epoch": 3.93, + "learning_rate": 3.0391085864103128e-06, + "loss": 3.1314, + "step": 18000 + }, + { + "epoch": 3.93, + "learning_rate": 3.0389993445488314e-06, + "loss": 3.4296, + "step": 18001 + }, + { + "epoch": 3.93, + "learning_rate": 3.03889010268735e-06, + "loss": 0.8369, + "step": 18002 + }, + { + "epoch": 3.93, + "learning_rate": 3.038780860825869e-06, + "loss": 0.9622, + "step": 18003 + }, + { + "epoch": 3.93, + "learning_rate": 3.0386716189643872e-06, + "loss": 2.7425, + "step": 18004 + }, + { + "epoch": 3.93, + "learning_rate": 3.038562377102906e-06, + "loss": 0.7657, + "step": 18005 + }, + { + "epoch": 3.93, + "learning_rate": 3.038453135241425e-06, + "loss": 2.4188, + "step": 18006 + }, + { + "epoch": 3.93, + "learning_rate": 3.0383438933799435e-06, + "loss": 1.3365, + "step": 18007 + }, + { + "epoch": 3.93, + "learning_rate": 3.0382346515184617e-06, + "loss": 0.553, + "step": 18008 + }, + { + "epoch": 3.93, + "learning_rate": 3.0381254096569807e-06, + "loss": 2.2486, + "step": 18009 + }, + { + "epoch": 3.93, + "learning_rate": 3.0380161677954993e-06, + "loss": 3.2235, + "step": 18010 + }, + { + "epoch": 3.93, + "learning_rate": 3.0379069259340183e-06, + "loss": 0.6046, + "step": 18011 + }, + { + "epoch": 3.93, + "learning_rate": 3.037797684072537e-06, + "loss": 2.2176, + "step": 18012 + }, + { + "epoch": 3.93, + "learning_rate": 3.037688442211055e-06, + "loss": 1.7537, + "step": 18013 + }, + { + "epoch": 3.93, + "learning_rate": 3.037579200349574e-06, + "loss": 0.9557, + "step": 18014 + }, + { + "epoch": 3.93, + "learning_rate": 3.037469958488093e-06, + "loss": 1.621, + "step": 18015 + }, + { + "epoch": 3.93, + "learning_rate": 3.037360716626612e-06, + "loss": 1.408, + "step": 18016 + }, + { + "epoch": 3.93, + "learning_rate": 3.0372514747651304e-06, + "loss": 0.9445, + "step": 18017 + }, + { + "epoch": 3.93, + "learning_rate": 3.0371422329036486e-06, + "loss": 1.5967, + "step": 18018 + }, + { + "epoch": 3.93, + "learning_rate": 3.0370329910421677e-06, + "loss": 2.9447, + "step": 18019 + }, + { + "epoch": 3.93, + "learning_rate": 3.0369237491806863e-06, + "loss": 1.3446, + "step": 18020 + }, + { + "epoch": 3.93, + "learning_rate": 3.0368145073192053e-06, + "loss": 1.575, + "step": 18021 + }, + { + "epoch": 3.93, + "learning_rate": 3.0367052654577235e-06, + "loss": 1.2828, + "step": 18022 + }, + { + "epoch": 3.93, + "learning_rate": 3.036596023596242e-06, + "loss": 3.1768, + "step": 18023 + }, + { + "epoch": 3.93, + "learning_rate": 3.036486781734761e-06, + "loss": 1.0791, + "step": 18024 + }, + { + "epoch": 3.93, + "learning_rate": 3.0363775398732797e-06, + "loss": 1.7323, + "step": 18025 + }, + { + "epoch": 3.93, + "learning_rate": 3.0362682980117984e-06, + "loss": 1.1573, + "step": 18026 + }, + { + "epoch": 3.93, + "learning_rate": 3.036159056150317e-06, + "loss": 1.6505, + "step": 18027 + }, + { + "epoch": 3.93, + "learning_rate": 3.0360498142888356e-06, + "loss": 0.7337, + "step": 18028 + }, + { + "epoch": 3.93, + "learning_rate": 3.0359405724273546e-06, + "loss": 1.8147, + "step": 18029 + }, + { + "epoch": 3.93, + "learning_rate": 3.0358313305658732e-06, + "loss": 2.5743, + "step": 18030 + }, + { + "epoch": 3.94, + "learning_rate": 3.035722088704392e-06, + "loss": 2.7471, + "step": 18031 + }, + { + "epoch": 3.94, + "learning_rate": 3.0356128468429104e-06, + "loss": 1.4613, + "step": 18032 + }, + { + "epoch": 3.94, + "learning_rate": 3.035503604981429e-06, + "loss": 2.6059, + "step": 18033 + }, + { + "epoch": 3.94, + "learning_rate": 3.0353943631199477e-06, + "loss": 1.5275, + "step": 18034 + }, + { + "epoch": 3.94, + "learning_rate": 3.0352851212584667e-06, + "loss": 1.0039, + "step": 18035 + }, + { + "epoch": 3.94, + "learning_rate": 3.0351758793969853e-06, + "loss": 2.4915, + "step": 18036 + }, + { + "epoch": 3.94, + "learning_rate": 3.0350666375355035e-06, + "loss": 4.0415, + "step": 18037 + }, + { + "epoch": 3.94, + "learning_rate": 3.0349573956740225e-06, + "loss": 1.4693, + "step": 18038 + }, + { + "epoch": 3.94, + "learning_rate": 3.034848153812541e-06, + "loss": 2.4927, + "step": 18039 + }, + { + "epoch": 3.94, + "learning_rate": 3.03473891195106e-06, + "loss": 3.3152, + "step": 18040 + }, + { + "epoch": 3.94, + "learning_rate": 3.0346296700895784e-06, + "loss": 1.5368, + "step": 18041 + }, + { + "epoch": 3.94, + "learning_rate": 3.034520428228097e-06, + "loss": 1.3585, + "step": 18042 + }, + { + "epoch": 3.94, + "learning_rate": 3.034411186366616e-06, + "loss": 1.3986, + "step": 18043 + }, + { + "epoch": 3.94, + "learning_rate": 3.0343019445051346e-06, + "loss": 2.4407, + "step": 18044 + }, + { + "epoch": 3.94, + "learning_rate": 3.0341927026436537e-06, + "loss": 0.592, + "step": 18045 + }, + { + "epoch": 3.94, + "learning_rate": 3.034083460782172e-06, + "loss": 1.5732, + "step": 18046 + }, + { + "epoch": 3.94, + "learning_rate": 3.0339742189206905e-06, + "loss": 1.0738, + "step": 18047 + }, + { + "epoch": 3.94, + "learning_rate": 3.0338649770592095e-06, + "loss": 3.2116, + "step": 18048 + }, + { + "epoch": 3.94, + "learning_rate": 3.033755735197728e-06, + "loss": 2.3228, + "step": 18049 + }, + { + "epoch": 3.94, + "learning_rate": 3.033646493336247e-06, + "loss": 2.9729, + "step": 18050 + }, + { + "epoch": 3.94, + "learning_rate": 3.0335372514747653e-06, + "loss": 3.3596, + "step": 18051 + }, + { + "epoch": 3.94, + "learning_rate": 3.033428009613284e-06, + "loss": 1.2988, + "step": 18052 + }, + { + "epoch": 3.94, + "learning_rate": 3.033318767751803e-06, + "loss": 2.6711, + "step": 18053 + }, + { + "epoch": 3.94, + "learning_rate": 3.0332095258903216e-06, + "loss": 0.9755, + "step": 18054 + }, + { + "epoch": 3.94, + "learning_rate": 3.0331002840288398e-06, + "loss": 2.0992, + "step": 18055 + }, + { + "epoch": 3.94, + "learning_rate": 3.032991042167359e-06, + "loss": 1.6544, + "step": 18056 + }, + { + "epoch": 3.94, + "learning_rate": 3.0328818003058774e-06, + "loss": 2.292, + "step": 18057 + }, + { + "epoch": 3.94, + "learning_rate": 3.032772558444396e-06, + "loss": 1.2786, + "step": 18058 + }, + { + "epoch": 3.94, + "learning_rate": 3.032663316582915e-06, + "loss": 1.9313, + "step": 18059 + }, + { + "epoch": 3.94, + "learning_rate": 3.0325540747214332e-06, + "loss": 2.6028, + "step": 18060 + }, + { + "epoch": 3.94, + "learning_rate": 3.032444832859952e-06, + "loss": 1.283, + "step": 18061 + }, + { + "epoch": 3.94, + "learning_rate": 3.032335590998471e-06, + "loss": 1.7027, + "step": 18062 + }, + { + "epoch": 3.94, + "learning_rate": 3.0322263491369895e-06, + "loss": 3.3058, + "step": 18063 + }, + { + "epoch": 3.94, + "learning_rate": 3.0321171072755085e-06, + "loss": 1.9211, + "step": 18064 + }, + { + "epoch": 3.94, + "learning_rate": 3.0320078654140267e-06, + "loss": 2.1866, + "step": 18065 + }, + { + "epoch": 3.94, + "learning_rate": 3.0318986235525453e-06, + "loss": 2.1668, + "step": 18066 + }, + { + "epoch": 3.94, + "learning_rate": 3.0317893816910644e-06, + "loss": 2.2736, + "step": 18067 + }, + { + "epoch": 3.94, + "learning_rate": 3.031680139829583e-06, + "loss": 1.3033, + "step": 18068 + }, + { + "epoch": 3.94, + "learning_rate": 3.031570897968102e-06, + "loss": 1.9635, + "step": 18069 + }, + { + "epoch": 3.94, + "learning_rate": 3.03146165610662e-06, + "loss": 1.9034, + "step": 18070 + }, + { + "epoch": 3.94, + "learning_rate": 3.031352414245139e-06, + "loss": 0.8902, + "step": 18071 + }, + { + "epoch": 3.94, + "learning_rate": 3.031243172383658e-06, + "loss": 1.4891, + "step": 18072 + }, + { + "epoch": 3.94, + "learning_rate": 3.0311339305221765e-06, + "loss": 1.4228, + "step": 18073 + }, + { + "epoch": 3.94, + "learning_rate": 3.0310246886606947e-06, + "loss": 1.6468, + "step": 18074 + }, + { + "epoch": 3.94, + "learning_rate": 3.0309154467992137e-06, + "loss": 2.6799, + "step": 18075 + }, + { + "epoch": 3.95, + "learning_rate": 3.0308062049377323e-06, + "loss": 2.0633, + "step": 18076 + }, + { + "epoch": 3.95, + "learning_rate": 3.0306969630762513e-06, + "loss": 2.9292, + "step": 18077 + }, + { + "epoch": 3.95, + "learning_rate": 3.03058772121477e-06, + "loss": 1.7307, + "step": 18078 + }, + { + "epoch": 3.95, + "learning_rate": 3.030478479353288e-06, + "loss": 3.2693, + "step": 18079 + }, + { + "epoch": 3.95, + "learning_rate": 3.030369237491807e-06, + "loss": 0.5881, + "step": 18080 + }, + { + "epoch": 3.95, + "learning_rate": 3.0302599956303258e-06, + "loss": 2.5568, + "step": 18081 + }, + { + "epoch": 3.95, + "learning_rate": 3.030150753768845e-06, + "loss": 1.7184, + "step": 18082 + }, + { + "epoch": 3.95, + "learning_rate": 3.0300415119073634e-06, + "loss": 2.4157, + "step": 18083 + }, + { + "epoch": 3.95, + "learning_rate": 3.0299322700458816e-06, + "loss": 1.5634, + "step": 18084 + }, + { + "epoch": 3.95, + "learning_rate": 3.0298230281844006e-06, + "loss": 1.2466, + "step": 18085 + }, + { + "epoch": 3.95, + "learning_rate": 3.0297137863229193e-06, + "loss": 0.7235, + "step": 18086 + }, + { + "epoch": 3.95, + "learning_rate": 3.029604544461438e-06, + "loss": 1.744, + "step": 18087 + }, + { + "epoch": 3.95, + "learning_rate": 3.0294953025999565e-06, + "loss": 1.3902, + "step": 18088 + }, + { + "epoch": 3.95, + "learning_rate": 3.029386060738475e-06, + "loss": 1.1131, + "step": 18089 + }, + { + "epoch": 3.95, + "learning_rate": 3.0292768188769937e-06, + "loss": 1.7149, + "step": 18090 + }, + { + "epoch": 3.95, + "learning_rate": 3.0291675770155127e-06, + "loss": 1.222, + "step": 18091 + }, + { + "epoch": 3.95, + "learning_rate": 3.0290583351540313e-06, + "loss": 2.961, + "step": 18092 + }, + { + "epoch": 3.95, + "learning_rate": 3.0289490932925495e-06, + "loss": 1.4024, + "step": 18093 + }, + { + "epoch": 3.95, + "learning_rate": 3.0288398514310686e-06, + "loss": 0.6922, + "step": 18094 + }, + { + "epoch": 3.95, + "learning_rate": 3.028730609569587e-06, + "loss": 1.7563, + "step": 18095 + }, + { + "epoch": 3.95, + "learning_rate": 3.028621367708106e-06, + "loss": 1.783, + "step": 18096 + }, + { + "epoch": 3.95, + "learning_rate": 3.028512125846625e-06, + "loss": 1.9127, + "step": 18097 + }, + { + "epoch": 3.95, + "learning_rate": 3.028402883985143e-06, + "loss": 3.2413, + "step": 18098 + }, + { + "epoch": 3.95, + "learning_rate": 3.028293642123662e-06, + "loss": 2.6688, + "step": 18099 + }, + { + "epoch": 3.95, + "learning_rate": 3.0281844002621807e-06, + "loss": 2.7754, + "step": 18100 + }, + { + "epoch": 3.95, + "learning_rate": 3.0280751584006997e-06, + "loss": 1.707, + "step": 18101 + }, + { + "epoch": 3.95, + "learning_rate": 3.0279659165392183e-06, + "loss": 0.5793, + "step": 18102 + }, + { + "epoch": 3.95, + "learning_rate": 3.0278566746777365e-06, + "loss": 2.0279, + "step": 18103 + }, + { + "epoch": 3.95, + "learning_rate": 3.0277474328162555e-06, + "loss": 1.2525, + "step": 18104 + }, + { + "epoch": 3.95, + "learning_rate": 3.027638190954774e-06, + "loss": 1.5497, + "step": 18105 + }, + { + "epoch": 3.95, + "learning_rate": 3.027528949093293e-06, + "loss": 1.2311, + "step": 18106 + }, + { + "epoch": 3.95, + "learning_rate": 3.0274197072318114e-06, + "loss": 0.9786, + "step": 18107 + }, + { + "epoch": 3.95, + "learning_rate": 3.02731046537033e-06, + "loss": 2.3868, + "step": 18108 + }, + { + "epoch": 3.95, + "learning_rate": 3.027201223508849e-06, + "loss": 1.6356, + "step": 18109 + }, + { + "epoch": 3.95, + "learning_rate": 3.0270919816473676e-06, + "loss": 1.1321, + "step": 18110 + }, + { + "epoch": 3.95, + "learning_rate": 3.0269827397858862e-06, + "loss": 1.6022, + "step": 18111 + }, + { + "epoch": 3.95, + "learning_rate": 3.026873497924405e-06, + "loss": 2.5627, + "step": 18112 + }, + { + "epoch": 3.95, + "learning_rate": 3.0267642560629234e-06, + "loss": 2.2504, + "step": 18113 + }, + { + "epoch": 3.95, + "learning_rate": 3.026655014201442e-06, + "loss": 1.498, + "step": 18114 + }, + { + "epoch": 3.95, + "learning_rate": 3.026545772339961e-06, + "loss": 1.3821, + "step": 18115 + }, + { + "epoch": 3.95, + "learning_rate": 3.0264365304784797e-06, + "loss": 3.3476, + "step": 18116 + }, + { + "epoch": 3.95, + "learning_rate": 3.026327288616998e-06, + "loss": 1.0495, + "step": 18117 + }, + { + "epoch": 3.95, + "learning_rate": 3.026218046755517e-06, + "loss": 2.2403, + "step": 18118 + }, + { + "epoch": 3.95, + "learning_rate": 3.0261088048940355e-06, + "loss": 1.1951, + "step": 18119 + }, + { + "epoch": 3.95, + "learning_rate": 3.0259995630325546e-06, + "loss": 1.8135, + "step": 18120 + }, + { + "epoch": 3.95, + "learning_rate": 3.0258903211710728e-06, + "loss": 1.8318, + "step": 18121 + }, + { + "epoch": 3.96, + "learning_rate": 3.0257810793095914e-06, + "loss": 1.9949, + "step": 18122 + }, + { + "epoch": 3.96, + "learning_rate": 3.0256718374481104e-06, + "loss": 2.7644, + "step": 18123 + }, + { + "epoch": 3.96, + "learning_rate": 3.025562595586629e-06, + "loss": 1.6436, + "step": 18124 + }, + { + "epoch": 3.96, + "learning_rate": 3.025453353725148e-06, + "loss": 1.888, + "step": 18125 + }, + { + "epoch": 3.96, + "learning_rate": 3.0253441118636662e-06, + "loss": 1.3185, + "step": 18126 + }, + { + "epoch": 3.96, + "learning_rate": 3.025234870002185e-06, + "loss": 1.0296, + "step": 18127 + }, + { + "epoch": 3.96, + "learning_rate": 3.025125628140704e-06, + "loss": 2.1961, + "step": 18128 + }, + { + "epoch": 3.96, + "learning_rate": 3.0250163862792225e-06, + "loss": 2.2092, + "step": 18129 + }, + { + "epoch": 3.96, + "learning_rate": 3.0249071444177415e-06, + "loss": 1.9182, + "step": 18130 + }, + { + "epoch": 3.96, + "learning_rate": 3.0247979025562597e-06, + "loss": 2.3889, + "step": 18131 + }, + { + "epoch": 3.96, + "learning_rate": 3.0246886606947783e-06, + "loss": 1.0003, + "step": 18132 + }, + { + "epoch": 3.96, + "learning_rate": 3.0245794188332974e-06, + "loss": 2.1147, + "step": 18133 + }, + { + "epoch": 3.96, + "learning_rate": 3.024470176971816e-06, + "loss": 2.4469, + "step": 18134 + }, + { + "epoch": 3.96, + "learning_rate": 3.024360935110335e-06, + "loss": 1.0279, + "step": 18135 + }, + { + "epoch": 3.96, + "learning_rate": 3.024251693248853e-06, + "loss": 1.0612, + "step": 18136 + }, + { + "epoch": 3.96, + "learning_rate": 3.024142451387372e-06, + "loss": 1.6119, + "step": 18137 + }, + { + "epoch": 3.96, + "learning_rate": 3.024033209525891e-06, + "loss": 0.5736, + "step": 18138 + }, + { + "epoch": 3.96, + "learning_rate": 3.0239239676644094e-06, + "loss": 3.1027, + "step": 18139 + }, + { + "epoch": 3.96, + "learning_rate": 3.0238147258029276e-06, + "loss": 2.5866, + "step": 18140 + }, + { + "epoch": 3.96, + "learning_rate": 3.0237054839414467e-06, + "loss": 1.7075, + "step": 18141 + }, + { + "epoch": 3.96, + "learning_rate": 3.0235962420799653e-06, + "loss": 1.71, + "step": 18142 + }, + { + "epoch": 3.96, + "learning_rate": 3.023487000218484e-06, + "loss": 1.6989, + "step": 18143 + }, + { + "epoch": 3.96, + "learning_rate": 3.023377758357003e-06, + "loss": 0.9927, + "step": 18144 + }, + { + "epoch": 3.96, + "learning_rate": 3.023268516495521e-06, + "loss": 2.2432, + "step": 18145 + }, + { + "epoch": 3.96, + "learning_rate": 3.0231592746340397e-06, + "loss": 2.5784, + "step": 18146 + }, + { + "epoch": 3.96, + "learning_rate": 3.0230500327725588e-06, + "loss": 4.8073, + "step": 18147 + }, + { + "epoch": 3.96, + "learning_rate": 3.0229407909110774e-06, + "loss": 1.3653, + "step": 18148 + }, + { + "epoch": 3.96, + "learning_rate": 3.0228315490495964e-06, + "loss": 2.3781, + "step": 18149 + }, + { + "epoch": 3.96, + "learning_rate": 3.0227223071881146e-06, + "loss": 1.7566, + "step": 18150 + }, + { + "epoch": 3.96, + "learning_rate": 3.022613065326633e-06, + "loss": 1.7137, + "step": 18151 + }, + { + "epoch": 3.96, + "learning_rate": 3.0225038234651522e-06, + "loss": 1.4492, + "step": 18152 + }, + { + "epoch": 3.96, + "learning_rate": 3.022394581603671e-06, + "loss": 2.7073, + "step": 18153 + }, + { + "epoch": 3.96, + "learning_rate": 3.022285339742189e-06, + "loss": 2.7012, + "step": 18154 + }, + { + "epoch": 3.96, + "learning_rate": 3.022176097880708e-06, + "loss": 2.183, + "step": 18155 + }, + { + "epoch": 3.96, + "learning_rate": 3.0220668560192267e-06, + "loss": 1.7052, + "step": 18156 + }, + { + "epoch": 3.96, + "learning_rate": 3.0219576141577457e-06, + "loss": 2.8175, + "step": 18157 + }, + { + "epoch": 3.96, + "learning_rate": 3.0218483722962643e-06, + "loss": 1.7089, + "step": 18158 + }, + { + "epoch": 3.96, + "learning_rate": 3.0217391304347825e-06, + "loss": 2.1292, + "step": 18159 + }, + { + "epoch": 3.96, + "learning_rate": 3.0216298885733015e-06, + "loss": 2.3166, + "step": 18160 + }, + { + "epoch": 3.96, + "learning_rate": 3.02152064671182e-06, + "loss": 0.1869, + "step": 18161 + }, + { + "epoch": 3.96, + "learning_rate": 3.021411404850339e-06, + "loss": 1.6752, + "step": 18162 + }, + { + "epoch": 3.96, + "learning_rate": 3.021302162988858e-06, + "loss": 1.8157, + "step": 18163 + }, + { + "epoch": 3.96, + "learning_rate": 3.021192921127376e-06, + "loss": 2.0416, + "step": 18164 + }, + { + "epoch": 3.96, + "learning_rate": 3.021083679265895e-06, + "loss": 2.0409, + "step": 18165 + }, + { + "epoch": 3.96, + "learning_rate": 3.0209744374044136e-06, + "loss": 1.8069, + "step": 18166 + }, + { + "epoch": 3.96, + "learning_rate": 3.0208651955429322e-06, + "loss": 3.1805, + "step": 18167 + }, + { + "epoch": 3.97, + "learning_rate": 3.0207559536814513e-06, + "loss": 1.0037, + "step": 18168 + }, + { + "epoch": 3.97, + "learning_rate": 3.0206467118199695e-06, + "loss": 1.0197, + "step": 18169 + }, + { + "epoch": 3.97, + "learning_rate": 3.020537469958488e-06, + "loss": 2.7428, + "step": 18170 + }, + { + "epoch": 3.97, + "learning_rate": 3.020428228097007e-06, + "loss": 3.1744, + "step": 18171 + }, + { + "epoch": 3.97, + "learning_rate": 3.0203189862355257e-06, + "loss": 1.4133, + "step": 18172 + }, + { + "epoch": 3.97, + "learning_rate": 3.0202097443740443e-06, + "loss": 1.0748, + "step": 18173 + }, + { + "epoch": 3.97, + "learning_rate": 3.020100502512563e-06, + "loss": 3.0279, + "step": 18174 + }, + { + "epoch": 3.97, + "learning_rate": 3.0199912606510816e-06, + "loss": 3.5973, + "step": 18175 + }, + { + "epoch": 3.97, + "learning_rate": 3.0198820187896006e-06, + "loss": 2.1176, + "step": 18176 + }, + { + "epoch": 3.97, + "learning_rate": 3.019772776928119e-06, + "loss": 0.7504, + "step": 18177 + }, + { + "epoch": 3.97, + "learning_rate": 3.0196635350666374e-06, + "loss": 1.6461, + "step": 18178 + }, + { + "epoch": 3.97, + "learning_rate": 3.0195542932051564e-06, + "loss": 1.8338, + "step": 18179 + }, + { + "epoch": 3.97, + "learning_rate": 3.019445051343675e-06, + "loss": 2.7219, + "step": 18180 + }, + { + "epoch": 3.97, + "learning_rate": 3.019335809482194e-06, + "loss": 0.8642, + "step": 18181 + }, + { + "epoch": 3.97, + "learning_rate": 3.0192265676207127e-06, + "loss": 1.4351, + "step": 18182 + }, + { + "epoch": 3.97, + "learning_rate": 3.019117325759231e-06, + "loss": 1.6428, + "step": 18183 + }, + { + "epoch": 3.97, + "learning_rate": 3.01900808389775e-06, + "loss": 3.0929, + "step": 18184 + }, + { + "epoch": 3.97, + "learning_rate": 3.0188988420362685e-06, + "loss": 1.303, + "step": 18185 + }, + { + "epoch": 3.97, + "learning_rate": 3.0187896001747876e-06, + "loss": 1.329, + "step": 18186 + }, + { + "epoch": 3.97, + "learning_rate": 3.0186803583133057e-06, + "loss": 0.5372, + "step": 18187 + }, + { + "epoch": 3.97, + "learning_rate": 3.0185711164518243e-06, + "loss": 0.4429, + "step": 18188 + }, + { + "epoch": 3.97, + "learning_rate": 3.0184618745903434e-06, + "loss": 0.8785, + "step": 18189 + }, + { + "epoch": 3.97, + "learning_rate": 3.018352632728862e-06, + "loss": 3.6264, + "step": 18190 + }, + { + "epoch": 3.97, + "learning_rate": 3.018243390867381e-06, + "loss": 1.8328, + "step": 18191 + }, + { + "epoch": 3.97, + "learning_rate": 3.0181341490058992e-06, + "loss": 1.2922, + "step": 18192 + }, + { + "epoch": 3.97, + "learning_rate": 3.018024907144418e-06, + "loss": 1.6377, + "step": 18193 + }, + { + "epoch": 3.97, + "learning_rate": 3.017915665282937e-06, + "loss": 1.1605, + "step": 18194 + }, + { + "epoch": 3.97, + "learning_rate": 3.0178064234214555e-06, + "loss": 1.126, + "step": 18195 + }, + { + "epoch": 3.97, + "learning_rate": 3.017697181559974e-06, + "loss": 1.311, + "step": 18196 + }, + { + "epoch": 3.97, + "learning_rate": 3.0175879396984927e-06, + "loss": 0.8406, + "step": 18197 + }, + { + "epoch": 3.97, + "learning_rate": 3.0174786978370113e-06, + "loss": 1.1847, + "step": 18198 + }, + { + "epoch": 3.97, + "learning_rate": 3.01736945597553e-06, + "loss": 2.5089, + "step": 18199 + }, + { + "epoch": 3.97, + "learning_rate": 3.017260214114049e-06, + "loss": 2.3914, + "step": 18200 + }, + { + "epoch": 3.97, + "learning_rate": 3.0171509722525676e-06, + "loss": 0.3691, + "step": 18201 + }, + { + "epoch": 3.97, + "learning_rate": 3.0170417303910858e-06, + "loss": 2.9266, + "step": 18202 + }, + { + "epoch": 3.97, + "learning_rate": 3.0169324885296048e-06, + "loss": 0.6033, + "step": 18203 + }, + { + "epoch": 3.97, + "learning_rate": 3.0168232466681234e-06, + "loss": 2.0029, + "step": 18204 + }, + { + "epoch": 3.97, + "learning_rate": 3.0167140048066424e-06, + "loss": 2.9576, + "step": 18205 + }, + { + "epoch": 3.97, + "learning_rate": 3.0166047629451606e-06, + "loss": 1.3439, + "step": 18206 + }, + { + "epoch": 3.97, + "learning_rate": 3.0164955210836792e-06, + "loss": 4.3552, + "step": 18207 + }, + { + "epoch": 3.97, + "learning_rate": 3.0163862792221983e-06, + "loss": 1.249, + "step": 18208 + }, + { + "epoch": 3.97, + "learning_rate": 3.016277037360717e-06, + "loss": 2.673, + "step": 18209 + }, + { + "epoch": 3.97, + "learning_rate": 3.016167795499236e-06, + "loss": 1.7978, + "step": 18210 + }, + { + "epoch": 3.97, + "learning_rate": 3.016058553637754e-06, + "loss": 0.3178, + "step": 18211 + }, + { + "epoch": 3.97, + "learning_rate": 3.0159493117762727e-06, + "loss": 1.2308, + "step": 18212 + }, + { + "epoch": 3.97, + "learning_rate": 3.0158400699147917e-06, + "loss": 0.8433, + "step": 18213 + }, + { + "epoch": 3.98, + "learning_rate": 3.0157308280533104e-06, + "loss": 3.1149, + "step": 18214 + }, + { + "epoch": 3.98, + "learning_rate": 3.0156215861918294e-06, + "loss": 0.7434, + "step": 18215 + }, + { + "epoch": 3.98, + "learning_rate": 3.0155123443303476e-06, + "loss": 2.0315, + "step": 18216 + }, + { + "epoch": 3.98, + "learning_rate": 3.015403102468866e-06, + "loss": 1.4892, + "step": 18217 + }, + { + "epoch": 3.98, + "learning_rate": 3.0152938606073852e-06, + "loss": 2.447, + "step": 18218 + }, + { + "epoch": 3.98, + "learning_rate": 3.015184618745904e-06, + "loss": 1.6752, + "step": 18219 + }, + { + "epoch": 3.98, + "learning_rate": 3.015075376884422e-06, + "loss": 3.2673, + "step": 18220 + }, + { + "epoch": 3.98, + "learning_rate": 3.014966135022941e-06, + "loss": 2.5972, + "step": 18221 + }, + { + "epoch": 3.98, + "learning_rate": 3.0148568931614597e-06, + "loss": 3.6046, + "step": 18222 + }, + { + "epoch": 3.98, + "learning_rate": 3.0147476512999783e-06, + "loss": 2.2896, + "step": 18223 + }, + { + "epoch": 3.98, + "learning_rate": 3.0146384094384973e-06, + "loss": 1.4658, + "step": 18224 + }, + { + "epoch": 3.98, + "learning_rate": 3.0145291675770155e-06, + "loss": 1.6283, + "step": 18225 + }, + { + "epoch": 3.98, + "learning_rate": 3.0144199257155345e-06, + "loss": 0.4514, + "step": 18226 + }, + { + "epoch": 3.98, + "learning_rate": 3.014310683854053e-06, + "loss": 2.8653, + "step": 18227 + }, + { + "epoch": 3.98, + "learning_rate": 3.0142014419925718e-06, + "loss": 1.122, + "step": 18228 + }, + { + "epoch": 3.98, + "learning_rate": 3.014092200131091e-06, + "loss": 1.7454, + "step": 18229 + }, + { + "epoch": 3.98, + "learning_rate": 3.013982958269609e-06, + "loss": 2.3905, + "step": 18230 + }, + { + "epoch": 3.98, + "learning_rate": 3.0138737164081276e-06, + "loss": 2.532, + "step": 18231 + }, + { + "epoch": 3.98, + "learning_rate": 3.0137644745466466e-06, + "loss": 1.9268, + "step": 18232 + }, + { + "epoch": 3.98, + "learning_rate": 3.0136552326851652e-06, + "loss": 1.2194, + "step": 18233 + }, + { + "epoch": 3.98, + "learning_rate": 3.0135459908236843e-06, + "loss": 2.8917, + "step": 18234 + }, + { + "epoch": 3.98, + "learning_rate": 3.0134367489622025e-06, + "loss": 1.8672, + "step": 18235 + }, + { + "epoch": 3.98, + "learning_rate": 3.013327507100721e-06, + "loss": 0.9865, + "step": 18236 + }, + { + "epoch": 3.98, + "learning_rate": 3.01321826523924e-06, + "loss": 2.1948, + "step": 18237 + }, + { + "epoch": 3.98, + "learning_rate": 3.0131090233777587e-06, + "loss": 1.7288, + "step": 18238 + }, + { + "epoch": 3.98, + "learning_rate": 3.012999781516277e-06, + "loss": 2.1222, + "step": 18239 + }, + { + "epoch": 3.98, + "learning_rate": 3.012890539654796e-06, + "loss": 1.9155, + "step": 18240 + }, + { + "epoch": 3.98, + "learning_rate": 3.0127812977933145e-06, + "loss": 0.9992, + "step": 18241 + }, + { + "epoch": 3.98, + "learning_rate": 3.0126720559318336e-06, + "loss": 1.7662, + "step": 18242 + }, + { + "epoch": 3.98, + "learning_rate": 3.012562814070352e-06, + "loss": 1.8366, + "step": 18243 + }, + { + "epoch": 3.98, + "learning_rate": 3.0124535722088704e-06, + "loss": 0.8692, + "step": 18244 + }, + { + "epoch": 3.98, + "learning_rate": 3.0123443303473894e-06, + "loss": 1.3235, + "step": 18245 + }, + { + "epoch": 3.98, + "learning_rate": 3.012235088485908e-06, + "loss": 1.7117, + "step": 18246 + }, + { + "epoch": 3.98, + "learning_rate": 3.012125846624427e-06, + "loss": 1.237, + "step": 18247 + }, + { + "epoch": 3.98, + "learning_rate": 3.0120166047629457e-06, + "loss": 1.8646, + "step": 18248 + }, + { + "epoch": 3.98, + "learning_rate": 3.011907362901464e-06, + "loss": 1.297, + "step": 18249 + }, + { + "epoch": 3.98, + "learning_rate": 3.011798121039983e-06, + "loss": 1.3191, + "step": 18250 + }, + { + "epoch": 3.98, + "learning_rate": 3.0116888791785015e-06, + "loss": 1.2857, + "step": 18251 + }, + { + "epoch": 3.98, + "learning_rate": 3.01157963731702e-06, + "loss": 2.2043, + "step": 18252 + }, + { + "epoch": 3.98, + "learning_rate": 3.0114703954555387e-06, + "loss": 1.5635, + "step": 18253 + }, + { + "epoch": 3.98, + "learning_rate": 3.0113611535940573e-06, + "loss": 1.9097, + "step": 18254 + }, + { + "epoch": 3.98, + "learning_rate": 3.011251911732576e-06, + "loss": 2.1766, + "step": 18255 + }, + { + "epoch": 3.98, + "learning_rate": 3.011142669871095e-06, + "loss": 1.6012, + "step": 18256 + }, + { + "epoch": 3.98, + "learning_rate": 3.0110334280096136e-06, + "loss": 1.7501, + "step": 18257 + }, + { + "epoch": 3.98, + "learning_rate": 3.0109241861481318e-06, + "loss": 0.7106, + "step": 18258 + }, + { + "epoch": 3.98, + "learning_rate": 3.010814944286651e-06, + "loss": 1.7974, + "step": 18259 + }, + { + "epoch": 3.99, + "learning_rate": 3.0107057024251694e-06, + "loss": 1.3964, + "step": 18260 + }, + { + "epoch": 3.99, + "learning_rate": 3.0105964605636885e-06, + "loss": 3.094, + "step": 18261 + }, + { + "epoch": 3.99, + "learning_rate": 3.010487218702207e-06, + "loss": 0.5881, + "step": 18262 + }, + { + "epoch": 3.99, + "learning_rate": 3.0103779768407253e-06, + "loss": 1.7851, + "step": 18263 + }, + { + "epoch": 3.99, + "learning_rate": 3.0102687349792443e-06, + "loss": 1.4488, + "step": 18264 + }, + { + "epoch": 3.99, + "learning_rate": 3.010159493117763e-06, + "loss": 0.7876, + "step": 18265 + }, + { + "epoch": 3.99, + "learning_rate": 3.010050251256282e-06, + "loss": 1.2185, + "step": 18266 + }, + { + "epoch": 3.99, + "learning_rate": 3.0099410093948005e-06, + "loss": 2.5609, + "step": 18267 + }, + { + "epoch": 3.99, + "learning_rate": 3.0098317675333187e-06, + "loss": 1.8093, + "step": 18268 + }, + { + "epoch": 3.99, + "learning_rate": 3.0097225256718378e-06, + "loss": 1.5122, + "step": 18269 + }, + { + "epoch": 3.99, + "learning_rate": 3.0096132838103564e-06, + "loss": 2.5116, + "step": 18270 + }, + { + "epoch": 3.99, + "learning_rate": 3.0095040419488754e-06, + "loss": 1.7535, + "step": 18271 + }, + { + "epoch": 3.99, + "learning_rate": 3.0093948000873936e-06, + "loss": 3.1321, + "step": 18272 + }, + { + "epoch": 3.99, + "learning_rate": 3.0092855582259122e-06, + "loss": 2.5108, + "step": 18273 + }, + { + "epoch": 3.99, + "learning_rate": 3.0091763163644312e-06, + "loss": 2.8912, + "step": 18274 + }, + { + "epoch": 3.99, + "learning_rate": 3.00906707450295e-06, + "loss": 0.7679, + "step": 18275 + }, + { + "epoch": 3.99, + "learning_rate": 3.0089578326414685e-06, + "loss": 2.3217, + "step": 18276 + }, + { + "epoch": 3.99, + "learning_rate": 3.008848590779987e-06, + "loss": 1.3798, + "step": 18277 + }, + { + "epoch": 3.99, + "learning_rate": 3.0087393489185057e-06, + "loss": 1.5105, + "step": 18278 + }, + { + "epoch": 3.99, + "learning_rate": 3.0086301070570247e-06, + "loss": 1.6313, + "step": 18279 + }, + { + "epoch": 3.99, + "learning_rate": 3.0085208651955433e-06, + "loss": 1.3787, + "step": 18280 + }, + { + "epoch": 3.99, + "learning_rate": 3.008411623334062e-06, + "loss": 2.183, + "step": 18281 + }, + { + "epoch": 3.99, + "learning_rate": 3.0083023814725806e-06, + "loss": 3.1822, + "step": 18282 + }, + { + "epoch": 3.99, + "learning_rate": 3.008193139611099e-06, + "loss": 2.1434, + "step": 18283 + }, + { + "epoch": 3.99, + "learning_rate": 3.0080838977496178e-06, + "loss": 2.6429, + "step": 18284 + }, + { + "epoch": 3.99, + "learning_rate": 3.007974655888137e-06, + "loss": 1.6809, + "step": 18285 + }, + { + "epoch": 3.99, + "learning_rate": 3.007865414026655e-06, + "loss": 0.9088, + "step": 18286 + }, + { + "epoch": 3.99, + "learning_rate": 3.0077561721651736e-06, + "loss": 1.385, + "step": 18287 + }, + { + "epoch": 3.99, + "learning_rate": 3.0076469303036926e-06, + "loss": 1.9719, + "step": 18288 + }, + { + "epoch": 3.99, + "learning_rate": 3.0075376884422113e-06, + "loss": 2.9102, + "step": 18289 + }, + { + "epoch": 3.99, + "learning_rate": 3.0074284465807303e-06, + "loss": 1.0697, + "step": 18290 + }, + { + "epoch": 3.99, + "learning_rate": 3.0073192047192485e-06, + "loss": 1.4216, + "step": 18291 + }, + { + "epoch": 3.99, + "learning_rate": 3.007209962857767e-06, + "loss": 1.4279, + "step": 18292 + }, + { + "epoch": 3.99, + "learning_rate": 3.007100720996286e-06, + "loss": 2.4545, + "step": 18293 + }, + { + "epoch": 3.99, + "learning_rate": 3.0069914791348047e-06, + "loss": 2.913, + "step": 18294 + }, + { + "epoch": 3.99, + "learning_rate": 3.0068822372733238e-06, + "loss": 1.9393, + "step": 18295 + }, + { + "epoch": 3.99, + "learning_rate": 3.006772995411842e-06, + "loss": 1.8478, + "step": 18296 + }, + { + "epoch": 3.99, + "learning_rate": 3.0066637535503606e-06, + "loss": 1.696, + "step": 18297 + }, + { + "epoch": 3.99, + "learning_rate": 3.0065545116888796e-06, + "loss": 1.4472, + "step": 18298 + }, + { + "epoch": 3.99, + "learning_rate": 3.0064452698273982e-06, + "loss": 1.8465, + "step": 18299 + }, + { + "epoch": 3.99, + "learning_rate": 3.0063360279659173e-06, + "loss": 2.9454, + "step": 18300 + }, + { + "epoch": 3.99, + "learning_rate": 3.0062267861044354e-06, + "loss": 1.515, + "step": 18301 + }, + { + "epoch": 3.99, + "learning_rate": 3.006117544242954e-06, + "loss": 1.8419, + "step": 18302 + }, + { + "epoch": 3.99, + "learning_rate": 3.006008302381473e-06, + "loss": 0.5381, + "step": 18303 + }, + { + "epoch": 3.99, + "learning_rate": 3.0058990605199917e-06, + "loss": 1.072, + "step": 18304 + }, + { + "epoch": 3.99, + "learning_rate": 3.00578981865851e-06, + "loss": 2.0185, + "step": 18305 + }, + { + "epoch": 4.0, + "learning_rate": 3.005680576797029e-06, + "loss": 1.4611, + "step": 18306 + }, + { + "epoch": 4.0, + "learning_rate": 3.0055713349355475e-06, + "loss": 1.5617, + "step": 18307 + }, + { + "epoch": 4.0, + "learning_rate": 3.005462093074066e-06, + "loss": 1.2808, + "step": 18308 + }, + { + "epoch": 4.0, + "learning_rate": 3.005352851212585e-06, + "loss": 0.9747, + "step": 18309 + }, + { + "epoch": 4.0, + "learning_rate": 3.0052436093511034e-06, + "loss": 2.8333, + "step": 18310 + }, + { + "epoch": 4.0, + "learning_rate": 3.005134367489622e-06, + "loss": 1.0861, + "step": 18311 + }, + { + "epoch": 4.0, + "learning_rate": 3.005025125628141e-06, + "loss": 2.595, + "step": 18312 + }, + { + "epoch": 4.0, + "learning_rate": 3.0049158837666596e-06, + "loss": 1.6592, + "step": 18313 + }, + { + "epoch": 4.0, + "learning_rate": 3.0048066419051787e-06, + "loss": 2.8368, + "step": 18314 + }, + { + "epoch": 4.0, + "learning_rate": 3.004697400043697e-06, + "loss": 4.3548, + "step": 18315 + }, + { + "epoch": 4.0, + "learning_rate": 3.0045881581822154e-06, + "loss": 2.0099, + "step": 18316 + }, + { + "epoch": 4.0, + "learning_rate": 3.0044789163207345e-06, + "loss": 3.4031, + "step": 18317 + }, + { + "epoch": 4.0, + "learning_rate": 3.004369674459253e-06, + "loss": 1.0319, + "step": 18318 + }, + { + "epoch": 4.0, + "learning_rate": 3.0042604325977713e-06, + "loss": 4.0856, + "step": 18319 + }, + { + "epoch": 4.0, + "learning_rate": 3.0041511907362903e-06, + "loss": 1.1313, + "step": 18320 + }, + { + "epoch": 4.0, + "learning_rate": 3.004041948874809e-06, + "loss": 1.8256, + "step": 18321 + }, + { + "epoch": 4.0, + "learning_rate": 3.003932707013328e-06, + "loss": 1.265, + "step": 18322 + }, + { + "epoch": 4.0, + "learning_rate": 3.0038234651518466e-06, + "loss": 0.8742, + "step": 18323 + }, + { + "epoch": 4.0, + "learning_rate": 3.0037142232903648e-06, + "loss": 2.6434, + "step": 18324 + }, + { + "epoch": 4.0, + "learning_rate": 3.003604981428884e-06, + "loss": 2.5735, + "step": 18325 + }, + { + "epoch": 4.0, + "learning_rate": 3.0034957395674024e-06, + "loss": 0.4768, + "step": 18326 + }, + { + "epoch": 4.0, + "learning_rate": 3.0033864977059214e-06, + "loss": 2.3667, + "step": 18327 + }, + { + "epoch": 4.0, + "learning_rate": 3.00327725584444e-06, + "loss": 2.2075, + "step": 18328 + }, + { + "epoch": 4.0, + "eval_accuracy": 0.5594108019639934, + "eval_loss": 1.8017919063568115, + "eval_runtime": 335.5063, + "eval_samples_per_second": 18.211, + "eval_steps_per_second": 4.554, + "step": 18328 + }, + { + "epoch": 4.0, + "learning_rate": 3.0031680139829582e-06, + "loss": 0.9875, + "step": 18329 + }, + { + "epoch": 4.0, + "learning_rate": 3.0030587721214773e-06, + "loss": 1.3781, + "step": 18330 + }, + { + "epoch": 4.0, + "learning_rate": 3.002949530259996e-06, + "loss": 2.0532, + "step": 18331 + }, + { + "epoch": 4.0, + "learning_rate": 3.002840288398515e-06, + "loss": 3.7395, + "step": 18332 + }, + { + "epoch": 4.0, + "learning_rate": 3.0027310465370335e-06, + "loss": 1.6207, + "step": 18333 + }, + { + "epoch": 4.0, + "learning_rate": 3.0026218046755517e-06, + "loss": 1.299, + "step": 18334 + }, + { + "epoch": 4.0, + "learning_rate": 3.0025125628140708e-06, + "loss": 0.5626, + "step": 18335 + }, + { + "epoch": 4.0, + "learning_rate": 3.0024033209525894e-06, + "loss": 0.9988, + "step": 18336 + }, + { + "epoch": 4.0, + "learning_rate": 3.002294079091108e-06, + "loss": 2.3779, + "step": 18337 + }, + { + "epoch": 4.0, + "learning_rate": 3.0021848372296266e-06, + "loss": 2.2103, + "step": 18338 + }, + { + "epoch": 4.0, + "learning_rate": 3.002075595368145e-06, + "loss": 0.9663, + "step": 18339 + }, + { + "epoch": 4.0, + "learning_rate": 3.001966353506664e-06, + "loss": 1.7399, + "step": 18340 + }, + { + "epoch": 4.0, + "learning_rate": 3.001857111645183e-06, + "loss": 2.618, + "step": 18341 + }, + { + "epoch": 4.0, + "learning_rate": 3.0017478697837015e-06, + "loss": 1.7028, + "step": 18342 + }, + { + "epoch": 4.0, + "learning_rate": 3.0016386279222196e-06, + "loss": 1.1588, + "step": 18343 + }, + { + "epoch": 4.0, + "learning_rate": 3.0015293860607387e-06, + "loss": 2.1752, + "step": 18344 + }, + { + "epoch": 4.0, + "learning_rate": 3.0014201441992573e-06, + "loss": 3.0477, + "step": 18345 + }, + { + "epoch": 4.0, + "learning_rate": 3.0013109023377763e-06, + "loss": 1.725, + "step": 18346 + }, + { + "epoch": 4.0, + "learning_rate": 3.001201660476295e-06, + "loss": 1.2971, + "step": 18347 + }, + { + "epoch": 4.0, + "learning_rate": 3.001092418614813e-06, + "loss": 2.3239, + "step": 18348 + }, + { + "epoch": 4.0, + "learning_rate": 3.000983176753332e-06, + "loss": 1.4088, + "step": 18349 + }, + { + "epoch": 4.0, + "learning_rate": 3.0008739348918508e-06, + "loss": 1.8489, + "step": 18350 + }, + { + "epoch": 4.01, + "learning_rate": 3.00076469303037e-06, + "loss": 1.7129, + "step": 18351 + }, + { + "epoch": 4.01, + "learning_rate": 3.000655451168888e-06, + "loss": 0.6461, + "step": 18352 + }, + { + "epoch": 4.01, + "learning_rate": 3.0005462093074066e-06, + "loss": 2.2315, + "step": 18353 + }, + { + "epoch": 4.01, + "learning_rate": 3.0004369674459256e-06, + "loss": 1.6139, + "step": 18354 + }, + { + "epoch": 4.01, + "learning_rate": 3.0003277255844442e-06, + "loss": 1.8263, + "step": 18355 + }, + { + "epoch": 4.01, + "learning_rate": 3.0002184837229633e-06, + "loss": 2.3551, + "step": 18356 + }, + { + "epoch": 4.01, + "learning_rate": 3.0001092418614815e-06, + "loss": 1.8575, + "step": 18357 + }, + { + "epoch": 4.01, + "learning_rate": 3e-06, + "loss": 1.4054, + "step": 18358 + }, + { + "epoch": 4.01, + "learning_rate": 2.999890758138519e-06, + "loss": 2.5894, + "step": 18359 + }, + { + "epoch": 4.01, + "learning_rate": 2.9997815162770377e-06, + "loss": 2.1709, + "step": 18360 + }, + { + "epoch": 4.01, + "learning_rate": 2.9996722744155563e-06, + "loss": 1.6981, + "step": 18361 + }, + { + "epoch": 4.01, + "learning_rate": 2.999563032554075e-06, + "loss": 2.6666, + "step": 18362 + }, + { + "epoch": 4.01, + "learning_rate": 2.9994537906925936e-06, + "loss": 0.9486, + "step": 18363 + }, + { + "epoch": 4.01, + "learning_rate": 2.999344548831112e-06, + "loss": 2.3673, + "step": 18364 + }, + { + "epoch": 4.01, + "learning_rate": 2.999235306969631e-06, + "loss": 3.0342, + "step": 18365 + }, + { + "epoch": 4.01, + "learning_rate": 2.99912606510815e-06, + "loss": 1.7855, + "step": 18366 + }, + { + "epoch": 4.01, + "learning_rate": 2.999016823246668e-06, + "loss": 0.2547, + "step": 18367 + }, + { + "epoch": 4.01, + "learning_rate": 2.998907581385187e-06, + "loss": 1.4936, + "step": 18368 + }, + { + "epoch": 4.01, + "learning_rate": 2.9987983395237056e-06, + "loss": 2.0034, + "step": 18369 + }, + { + "epoch": 4.01, + "learning_rate": 2.9986890976622247e-06, + "loss": 1.5383, + "step": 18370 + }, + { + "epoch": 4.01, + "learning_rate": 2.998579855800743e-06, + "loss": 4.1409, + "step": 18371 + }, + { + "epoch": 4.01, + "learning_rate": 2.9984706139392615e-06, + "loss": 2.049, + "step": 18372 + }, + { + "epoch": 4.01, + "learning_rate": 2.9983613720777805e-06, + "loss": 1.3788, + "step": 18373 + }, + { + "epoch": 4.01, + "learning_rate": 2.998252130216299e-06, + "loss": 1.7207, + "step": 18374 + }, + { + "epoch": 4.01, + "learning_rate": 2.998142888354818e-06, + "loss": 1.8245, + "step": 18375 + }, + { + "epoch": 4.01, + "learning_rate": 2.9980336464933363e-06, + "loss": 1.7425, + "step": 18376 + }, + { + "epoch": 4.01, + "learning_rate": 2.997924404631855e-06, + "loss": 1.4349, + "step": 18377 + }, + { + "epoch": 4.01, + "learning_rate": 2.997815162770374e-06, + "loss": 0.8551, + "step": 18378 + }, + { + "epoch": 4.01, + "learning_rate": 2.9977059209088926e-06, + "loss": 3.2574, + "step": 18379 + }, + { + "epoch": 4.01, + "learning_rate": 2.9975966790474116e-06, + "loss": 2.0065, + "step": 18380 + }, + { + "epoch": 4.01, + "learning_rate": 2.99748743718593e-06, + "loss": 1.8585, + "step": 18381 + }, + { + "epoch": 4.01, + "learning_rate": 2.9973781953244484e-06, + "loss": 2.0231, + "step": 18382 + }, + { + "epoch": 4.01, + "learning_rate": 2.9972689534629675e-06, + "loss": 0.729, + "step": 18383 + }, + { + "epoch": 4.01, + "learning_rate": 2.997159711601486e-06, + "loss": 1.1816, + "step": 18384 + }, + { + "epoch": 4.01, + "learning_rate": 2.9970504697400043e-06, + "loss": 1.3542, + "step": 18385 + }, + { + "epoch": 4.01, + "learning_rate": 2.9969412278785233e-06, + "loss": 1.5238, + "step": 18386 + }, + { + "epoch": 4.01, + "learning_rate": 2.996831986017042e-06, + "loss": 0.7754, + "step": 18387 + }, + { + "epoch": 4.01, + "learning_rate": 2.996722744155561e-06, + "loss": 2.2209, + "step": 18388 + }, + { + "epoch": 4.01, + "learning_rate": 2.9966135022940796e-06, + "loss": 1.3803, + "step": 18389 + }, + { + "epoch": 4.01, + "learning_rate": 2.9965042604325977e-06, + "loss": 2.3524, + "step": 18390 + }, + { + "epoch": 4.01, + "learning_rate": 2.9963950185711168e-06, + "loss": 1.474, + "step": 18391 + }, + { + "epoch": 4.01, + "learning_rate": 2.9962857767096354e-06, + "loss": 1.687, + "step": 18392 + }, + { + "epoch": 4.01, + "learning_rate": 2.996176534848154e-06, + "loss": 0.7371, + "step": 18393 + }, + { + "epoch": 4.01, + "learning_rate": 2.996067292986673e-06, + "loss": 1.8173, + "step": 18394 + }, + { + "epoch": 4.01, + "learning_rate": 2.9959580511251912e-06, + "loss": 3.2532, + "step": 18395 + }, + { + "epoch": 4.01, + "learning_rate": 2.99584880926371e-06, + "loss": 1.831, + "step": 18396 + }, + { + "epoch": 4.02, + "learning_rate": 2.995739567402229e-06, + "loss": 1.354, + "step": 18397 + }, + { + "epoch": 4.02, + "learning_rate": 2.9956303255407475e-06, + "loss": 2.1176, + "step": 18398 + }, + { + "epoch": 4.02, + "learning_rate": 2.9955210836792665e-06, + "loss": 0.9905, + "step": 18399 + }, + { + "epoch": 4.02, + "learning_rate": 2.9954118418177847e-06, + "loss": 3.0389, + "step": 18400 + }, + { + "epoch": 4.02, + "learning_rate": 2.9953025999563033e-06, + "loss": 2.0485, + "step": 18401 + }, + { + "epoch": 4.02, + "learning_rate": 2.9951933580948223e-06, + "loss": 0.6939, + "step": 18402 + }, + { + "epoch": 4.02, + "learning_rate": 2.995084116233341e-06, + "loss": 1.3221, + "step": 18403 + }, + { + "epoch": 4.02, + "learning_rate": 2.994974874371859e-06, + "loss": 3.1207, + "step": 18404 + }, + { + "epoch": 4.02, + "learning_rate": 2.994865632510378e-06, + "loss": 1.3892, + "step": 18405 + }, + { + "epoch": 4.02, + "learning_rate": 2.994756390648897e-06, + "loss": 3.0434, + "step": 18406 + }, + { + "epoch": 4.02, + "learning_rate": 2.994647148787416e-06, + "loss": 2.6999, + "step": 18407 + }, + { + "epoch": 4.02, + "learning_rate": 2.9945379069259344e-06, + "loss": 1.7372, + "step": 18408 + }, + { + "epoch": 4.02, + "learning_rate": 2.9944286650644526e-06, + "loss": 3.0986, + "step": 18409 + }, + { + "epoch": 4.02, + "learning_rate": 2.9943194232029717e-06, + "loss": 2.0082, + "step": 18410 + }, + { + "epoch": 4.02, + "learning_rate": 2.9942101813414903e-06, + "loss": 2.9283, + "step": 18411 + }, + { + "epoch": 4.02, + "learning_rate": 2.9941009394800093e-06, + "loss": 0.8694, + "step": 18412 + }, + { + "epoch": 4.02, + "learning_rate": 2.993991697618528e-06, + "loss": 0.4782, + "step": 18413 + }, + { + "epoch": 4.02, + "learning_rate": 2.993882455757046e-06, + "loss": 1.4323, + "step": 18414 + }, + { + "epoch": 4.02, + "learning_rate": 2.993773213895565e-06, + "loss": 1.1672, + "step": 18415 + }, + { + "epoch": 4.02, + "learning_rate": 2.9936639720340837e-06, + "loss": 1.7839, + "step": 18416 + }, + { + "epoch": 4.02, + "learning_rate": 2.9935547301726024e-06, + "loss": 1.732, + "step": 18417 + }, + { + "epoch": 4.02, + "learning_rate": 2.993445488311121e-06, + "loss": 3.2042, + "step": 18418 + }, + { + "epoch": 4.02, + "learning_rate": 2.9933362464496396e-06, + "loss": 0.747, + "step": 18419 + }, + { + "epoch": 4.02, + "learning_rate": 2.993227004588158e-06, + "loss": 2.5517, + "step": 18420 + }, + { + "epoch": 4.02, + "learning_rate": 2.9931177627266772e-06, + "loss": 1.4145, + "step": 18421 + }, + { + "epoch": 4.02, + "learning_rate": 2.993008520865196e-06, + "loss": 1.5548, + "step": 18422 + }, + { + "epoch": 4.02, + "learning_rate": 2.9928992790037144e-06, + "loss": 1.6627, + "step": 18423 + }, + { + "epoch": 4.02, + "learning_rate": 2.992790037142233e-06, + "loss": 1.379, + "step": 18424 + }, + { + "epoch": 4.02, + "learning_rate": 2.9926807952807517e-06, + "loss": 1.8732, + "step": 18425 + }, + { + "epoch": 4.02, + "learning_rate": 2.9925715534192707e-06, + "loss": 2.2615, + "step": 18426 + }, + { + "epoch": 4.02, + "learning_rate": 2.9924623115577893e-06, + "loss": 2.1367, + "step": 18427 + }, + { + "epoch": 4.02, + "learning_rate": 2.9923530696963075e-06, + "loss": 2.7834, + "step": 18428 + }, + { + "epoch": 4.02, + "learning_rate": 2.9922438278348265e-06, + "loss": 1.6808, + "step": 18429 + }, + { + "epoch": 4.02, + "learning_rate": 2.992134585973345e-06, + "loss": 1.4169, + "step": 18430 + }, + { + "epoch": 4.02, + "learning_rate": 2.992025344111864e-06, + "loss": 0.8893, + "step": 18431 + }, + { + "epoch": 4.02, + "learning_rate": 2.9919161022503824e-06, + "loss": 2.0378, + "step": 18432 + }, + { + "epoch": 4.02, + "learning_rate": 2.991806860388901e-06, + "loss": 1.7479, + "step": 18433 + }, + { + "epoch": 4.02, + "learning_rate": 2.99169761852742e-06, + "loss": 2.3438, + "step": 18434 + }, + { + "epoch": 4.02, + "learning_rate": 2.9915883766659386e-06, + "loss": 1.7958, + "step": 18435 + }, + { + "epoch": 4.02, + "learning_rate": 2.9914791348044577e-06, + "loss": 3.9694, + "step": 18436 + }, + { + "epoch": 4.02, + "learning_rate": 2.991369892942976e-06, + "loss": 2.9168, + "step": 18437 + }, + { + "epoch": 4.02, + "learning_rate": 2.9912606510814945e-06, + "loss": 1.0521, + "step": 18438 + }, + { + "epoch": 4.02, + "learning_rate": 2.9911514092200135e-06, + "loss": 0.9766, + "step": 18439 + }, + { + "epoch": 4.02, + "learning_rate": 2.991042167358532e-06, + "loss": 1.6252, + "step": 18440 + }, + { + "epoch": 4.02, + "learning_rate": 2.990932925497051e-06, + "loss": 1.9349, + "step": 18441 + }, + { + "epoch": 4.02, + "learning_rate": 2.9908236836355693e-06, + "loss": 1.4011, + "step": 18442 + }, + { + "epoch": 4.03, + "learning_rate": 2.990714441774088e-06, + "loss": 2.0649, + "step": 18443 + }, + { + "epoch": 4.03, + "learning_rate": 2.990605199912607e-06, + "loss": 1.6708, + "step": 18444 + }, + { + "epoch": 4.03, + "learning_rate": 2.9904959580511256e-06, + "loss": 1.8116, + "step": 18445 + }, + { + "epoch": 4.03, + "learning_rate": 2.990386716189644e-06, + "loss": 3.1599, + "step": 18446 + }, + { + "epoch": 4.03, + "learning_rate": 2.990277474328163e-06, + "loss": 1.9979, + "step": 18447 + }, + { + "epoch": 4.03, + "learning_rate": 2.9901682324666814e-06, + "loss": 1.753, + "step": 18448 + }, + { + "epoch": 4.03, + "learning_rate": 2.9900589906052e-06, + "loss": 2.5377, + "step": 18449 + }, + { + "epoch": 4.03, + "learning_rate": 2.989949748743719e-06, + "loss": 0.3041, + "step": 18450 + }, + { + "epoch": 4.03, + "learning_rate": 2.9898405068822373e-06, + "loss": 2.7288, + "step": 18451 + }, + { + "epoch": 4.03, + "learning_rate": 2.989731265020756e-06, + "loss": 1.0738, + "step": 18452 + }, + { + "epoch": 4.03, + "learning_rate": 2.989622023159275e-06, + "loss": 0.8288, + "step": 18453 + }, + { + "epoch": 4.03, + "learning_rate": 2.9895127812977935e-06, + "loss": 1.614, + "step": 18454 + }, + { + "epoch": 4.03, + "learning_rate": 2.9894035394363125e-06, + "loss": 3.0745, + "step": 18455 + }, + { + "epoch": 4.03, + "learning_rate": 2.9892942975748307e-06, + "loss": 1.4384, + "step": 18456 + }, + { + "epoch": 4.03, + "learning_rate": 2.9891850557133493e-06, + "loss": 1.6474, + "step": 18457 + }, + { + "epoch": 4.03, + "learning_rate": 2.9890758138518684e-06, + "loss": 2.4082, + "step": 18458 + }, + { + "epoch": 4.03, + "learning_rate": 2.988966571990387e-06, + "loss": 1.7494, + "step": 18459 + }, + { + "epoch": 4.03, + "learning_rate": 2.988857330128906e-06, + "loss": 0.702, + "step": 18460 + }, + { + "epoch": 4.03, + "learning_rate": 2.988748088267424e-06, + "loss": 1.3061, + "step": 18461 + }, + { + "epoch": 4.03, + "learning_rate": 2.988638846405943e-06, + "loss": 1.0331, + "step": 18462 + }, + { + "epoch": 4.03, + "learning_rate": 2.988529604544462e-06, + "loss": 1.9837, + "step": 18463 + }, + { + "epoch": 4.03, + "learning_rate": 2.9884203626829805e-06, + "loss": 1.4735, + "step": 18464 + }, + { + "epoch": 4.03, + "learning_rate": 2.9883111208214987e-06, + "loss": 1.9208, + "step": 18465 + }, + { + "epoch": 4.03, + "learning_rate": 2.9882018789600177e-06, + "loss": 0.2703, + "step": 18466 + }, + { + "epoch": 4.03, + "learning_rate": 2.9880926370985363e-06, + "loss": 2.0717, + "step": 18467 + }, + { + "epoch": 4.03, + "learning_rate": 2.9879833952370553e-06, + "loss": 2.0843, + "step": 18468 + }, + { + "epoch": 4.03, + "learning_rate": 2.987874153375574e-06, + "loss": 1.8231, + "step": 18469 + }, + { + "epoch": 4.03, + "learning_rate": 2.987764911514092e-06, + "loss": 1.7635, + "step": 18470 + }, + { + "epoch": 4.03, + "learning_rate": 2.987655669652611e-06, + "loss": 0.7446, + "step": 18471 + }, + { + "epoch": 4.03, + "learning_rate": 2.9875464277911298e-06, + "loss": 1.6251, + "step": 18472 + }, + { + "epoch": 4.03, + "learning_rate": 2.9874371859296484e-06, + "loss": 0.8954, + "step": 18473 + }, + { + "epoch": 4.03, + "learning_rate": 2.9873279440681674e-06, + "loss": 2.0213, + "step": 18474 + }, + { + "epoch": 4.03, + "learning_rate": 2.9872187022066856e-06, + "loss": 1.1124, + "step": 18475 + }, + { + "epoch": 4.03, + "learning_rate": 2.9871094603452046e-06, + "loss": 2.165, + "step": 18476 + }, + { + "epoch": 4.03, + "learning_rate": 2.9870002184837233e-06, + "loss": 2.1608, + "step": 18477 + }, + { + "epoch": 4.03, + "learning_rate": 2.986890976622242e-06, + "loss": 1.9728, + "step": 18478 + }, + { + "epoch": 4.03, + "learning_rate": 2.986781734760761e-06, + "loss": 1.7001, + "step": 18479 + }, + { + "epoch": 4.03, + "learning_rate": 2.986672492899279e-06, + "loss": 1.1285, + "step": 18480 + }, + { + "epoch": 4.03, + "learning_rate": 2.9865632510377977e-06, + "loss": 2.3659, + "step": 18481 + }, + { + "epoch": 4.03, + "learning_rate": 2.9864540091763167e-06, + "loss": 2.4438, + "step": 18482 + }, + { + "epoch": 4.03, + "learning_rate": 2.9863447673148353e-06, + "loss": 2.0704, + "step": 18483 + }, + { + "epoch": 4.03, + "learning_rate": 2.9862355254533535e-06, + "loss": 2.2714, + "step": 18484 + }, + { + "epoch": 4.03, + "learning_rate": 2.9861262835918726e-06, + "loss": 1.2584, + "step": 18485 + }, + { + "epoch": 4.03, + "learning_rate": 2.986017041730391e-06, + "loss": 1.0492, + "step": 18486 + }, + { + "epoch": 4.03, + "learning_rate": 2.98590779986891e-06, + "loss": 1.4667, + "step": 18487 + }, + { + "epoch": 4.03, + "learning_rate": 2.985798558007429e-06, + "loss": 1.9561, + "step": 18488 + }, + { + "epoch": 4.04, + "learning_rate": 2.985689316145947e-06, + "loss": 1.7773, + "step": 18489 + }, + { + "epoch": 4.04, + "learning_rate": 2.985580074284466e-06, + "loss": 1.5576, + "step": 18490 + }, + { + "epoch": 4.04, + "learning_rate": 2.9854708324229847e-06, + "loss": 1.368, + "step": 18491 + }, + { + "epoch": 4.04, + "learning_rate": 2.9853615905615037e-06, + "loss": 2.6382, + "step": 18492 + }, + { + "epoch": 4.04, + "learning_rate": 2.9852523487000223e-06, + "loss": 2.2972, + "step": 18493 + }, + { + "epoch": 4.04, + "learning_rate": 2.9851431068385405e-06, + "loss": 1.9326, + "step": 18494 + }, + { + "epoch": 4.04, + "learning_rate": 2.9850338649770595e-06, + "loss": 1.548, + "step": 18495 + }, + { + "epoch": 4.04, + "learning_rate": 2.984924623115578e-06, + "loss": 1.7718, + "step": 18496 + }, + { + "epoch": 4.04, + "learning_rate": 2.984815381254097e-06, + "loss": 1.5009, + "step": 18497 + }, + { + "epoch": 4.04, + "learning_rate": 2.9847061393926154e-06, + "loss": 2.5087, + "step": 18498 + }, + { + "epoch": 4.04, + "learning_rate": 2.984596897531134e-06, + "loss": 1.5358, + "step": 18499 + }, + { + "epoch": 4.04, + "learning_rate": 2.984487655669653e-06, + "loss": 1.7427, + "step": 18500 + }, + { + "epoch": 4.04, + "learning_rate": 2.9843784138081716e-06, + "loss": 2.2655, + "step": 18501 + }, + { + "epoch": 4.04, + "learning_rate": 2.9842691719466902e-06, + "loss": 1.4212, + "step": 18502 + }, + { + "epoch": 4.04, + "learning_rate": 2.984159930085209e-06, + "loss": 2.7166, + "step": 18503 + }, + { + "epoch": 4.04, + "learning_rate": 2.9840506882237274e-06, + "loss": 1.7877, + "step": 18504 + }, + { + "epoch": 4.04, + "learning_rate": 2.983941446362246e-06, + "loss": 2.6656, + "step": 18505 + }, + { + "epoch": 4.04, + "learning_rate": 2.983832204500765e-06, + "loss": 1.3904, + "step": 18506 + }, + { + "epoch": 4.04, + "learning_rate": 2.9837229626392837e-06, + "loss": 2.2828, + "step": 18507 + }, + { + "epoch": 4.04, + "learning_rate": 2.983613720777802e-06, + "loss": 2.4811, + "step": 18508 + }, + { + "epoch": 4.04, + "learning_rate": 2.983504478916321e-06, + "loss": 0.6649, + "step": 18509 + }, + { + "epoch": 4.04, + "learning_rate": 2.9833952370548395e-06, + "loss": 0.552, + "step": 18510 + }, + { + "epoch": 4.04, + "learning_rate": 2.9832859951933586e-06, + "loss": 3.9469, + "step": 18511 + }, + { + "epoch": 4.04, + "learning_rate": 2.983176753331877e-06, + "loss": 2.6831, + "step": 18512 + }, + { + "epoch": 4.04, + "learning_rate": 2.9830675114703954e-06, + "loss": 1.9275, + "step": 18513 + }, + { + "epoch": 4.04, + "learning_rate": 2.9829582696089144e-06, + "loss": 2.2581, + "step": 18514 + }, + { + "epoch": 4.04, + "learning_rate": 2.982849027747433e-06, + "loss": 1.4255, + "step": 18515 + }, + { + "epoch": 4.04, + "learning_rate": 2.982739785885952e-06, + "loss": 2.1172, + "step": 18516 + }, + { + "epoch": 4.04, + "learning_rate": 2.9826305440244702e-06, + "loss": 2.9442, + "step": 18517 + }, + { + "epoch": 4.04, + "learning_rate": 2.982521302162989e-06, + "loss": 0.6984, + "step": 18518 + }, + { + "epoch": 4.04, + "learning_rate": 2.982412060301508e-06, + "loss": 1.9563, + "step": 18519 + }, + { + "epoch": 4.04, + "learning_rate": 2.9823028184400265e-06, + "loss": 0.8356, + "step": 18520 + }, + { + "epoch": 4.04, + "learning_rate": 2.9821935765785455e-06, + "loss": 1.2973, + "step": 18521 + }, + { + "epoch": 4.04, + "learning_rate": 2.9820843347170637e-06, + "loss": 2.5924, + "step": 18522 + }, + { + "epoch": 4.04, + "learning_rate": 2.9819750928555823e-06, + "loss": 1.7383, + "step": 18523 + }, + { + "epoch": 4.04, + "learning_rate": 2.9818658509941014e-06, + "loss": 1.5718, + "step": 18524 + }, + { + "epoch": 4.04, + "learning_rate": 2.98175660913262e-06, + "loss": 1.5231, + "step": 18525 + }, + { + "epoch": 4.04, + "learning_rate": 2.981647367271139e-06, + "loss": 2.2519, + "step": 18526 + }, + { + "epoch": 4.04, + "learning_rate": 2.981538125409657e-06, + "loss": 2.6603, + "step": 18527 + }, + { + "epoch": 4.04, + "learning_rate": 2.981428883548176e-06, + "loss": 1.9501, + "step": 18528 + }, + { + "epoch": 4.04, + "learning_rate": 2.981319641686695e-06, + "loss": 0.772, + "step": 18529 + }, + { + "epoch": 4.04, + "learning_rate": 2.9812103998252134e-06, + "loss": 1.8419, + "step": 18530 + }, + { + "epoch": 4.04, + "learning_rate": 2.9811011579637316e-06, + "loss": 1.8796, + "step": 18531 + }, + { + "epoch": 4.04, + "learning_rate": 2.9809919161022507e-06, + "loss": 0.4665, + "step": 18532 + }, + { + "epoch": 4.04, + "learning_rate": 2.9808826742407693e-06, + "loss": 2.7775, + "step": 18533 + }, + { + "epoch": 4.04, + "learning_rate": 2.980773432379288e-06, + "loss": 1.5381, + "step": 18534 + }, + { + "epoch": 4.05, + "learning_rate": 2.980664190517807e-06, + "loss": 0.6945, + "step": 18535 + }, + { + "epoch": 4.05, + "learning_rate": 2.980554948656325e-06, + "loss": 1.5974, + "step": 18536 + }, + { + "epoch": 4.05, + "learning_rate": 2.9804457067948437e-06, + "loss": 1.4979, + "step": 18537 + }, + { + "epoch": 4.05, + "learning_rate": 2.9803364649333628e-06, + "loss": 1.5428, + "step": 18538 + }, + { + "epoch": 4.05, + "learning_rate": 2.9802272230718814e-06, + "loss": 1.0581, + "step": 18539 + }, + { + "epoch": 4.05, + "learning_rate": 2.9801179812104004e-06, + "loss": 1.2533, + "step": 18540 + }, + { + "epoch": 4.05, + "learning_rate": 2.9800087393489186e-06, + "loss": 0.9599, + "step": 18541 + }, + { + "epoch": 4.05, + "learning_rate": 2.979899497487437e-06, + "loss": 2.2752, + "step": 18542 + }, + { + "epoch": 4.05, + "learning_rate": 2.9797902556259562e-06, + "loss": 1.0866, + "step": 18543 + }, + { + "epoch": 4.05, + "learning_rate": 2.979681013764475e-06, + "loss": 0.7421, + "step": 18544 + }, + { + "epoch": 4.05, + "learning_rate": 2.979571771902994e-06, + "loss": 1.9964, + "step": 18545 + }, + { + "epoch": 4.05, + "learning_rate": 2.979462530041512e-06, + "loss": 1.861, + "step": 18546 + }, + { + "epoch": 4.05, + "learning_rate": 2.9793532881800307e-06, + "loss": 2.1587, + "step": 18547 + }, + { + "epoch": 4.05, + "learning_rate": 2.9792440463185497e-06, + "loss": 2.8868, + "step": 18548 + }, + { + "epoch": 4.05, + "learning_rate": 2.9791348044570683e-06, + "loss": 1.7682, + "step": 18549 + }, + { + "epoch": 4.05, + "learning_rate": 2.9790255625955865e-06, + "loss": 1.7731, + "step": 18550 + }, + { + "epoch": 4.05, + "learning_rate": 2.9789163207341055e-06, + "loss": 0.8608, + "step": 18551 + }, + { + "epoch": 4.05, + "learning_rate": 2.978807078872624e-06, + "loss": 1.1181, + "step": 18552 + }, + { + "epoch": 4.05, + "learning_rate": 2.978697837011143e-06, + "loss": 1.809, + "step": 18553 + }, + { + "epoch": 4.05, + "learning_rate": 2.978588595149662e-06, + "loss": 1.0908, + "step": 18554 + }, + { + "epoch": 4.05, + "learning_rate": 2.97847935328818e-06, + "loss": 1.2195, + "step": 18555 + }, + { + "epoch": 4.05, + "learning_rate": 2.978370111426699e-06, + "loss": 4.3483, + "step": 18556 + }, + { + "epoch": 4.05, + "learning_rate": 2.9782608695652176e-06, + "loss": 0.9443, + "step": 18557 + }, + { + "epoch": 4.05, + "learning_rate": 2.9781516277037362e-06, + "loss": 2.139, + "step": 18558 + }, + { + "epoch": 4.05, + "learning_rate": 2.9780423858422553e-06, + "loss": 2.1756, + "step": 18559 + }, + { + "epoch": 4.05, + "learning_rate": 2.9779331439807735e-06, + "loss": 2.83, + "step": 18560 + }, + { + "epoch": 4.05, + "learning_rate": 2.977823902119292e-06, + "loss": 0.7409, + "step": 18561 + }, + { + "epoch": 4.05, + "learning_rate": 2.977714660257811e-06, + "loss": 1.5093, + "step": 18562 + }, + { + "epoch": 4.05, + "learning_rate": 2.9776054183963297e-06, + "loss": 1.7138, + "step": 18563 + }, + { + "epoch": 4.05, + "learning_rate": 2.977496176534848e-06, + "loss": 1.6102, + "step": 18564 + }, + { + "epoch": 4.05, + "learning_rate": 2.977386934673367e-06, + "loss": 1.3717, + "step": 18565 + }, + { + "epoch": 4.05, + "learning_rate": 2.9772776928118856e-06, + "loss": 1.1945, + "step": 18566 + }, + { + "epoch": 4.05, + "learning_rate": 2.9771684509504046e-06, + "loss": 0.8123, + "step": 18567 + }, + { + "epoch": 4.05, + "learning_rate": 2.977059209088923e-06, + "loss": 2.2047, + "step": 18568 + }, + { + "epoch": 4.05, + "learning_rate": 2.9769499672274414e-06, + "loss": 1.1037, + "step": 18569 + }, + { + "epoch": 4.05, + "learning_rate": 2.9768407253659604e-06, + "loss": 1.4005, + "step": 18570 + }, + { + "epoch": 4.05, + "learning_rate": 2.976731483504479e-06, + "loss": 2.4019, + "step": 18571 + }, + { + "epoch": 4.05, + "learning_rate": 2.976622241642998e-06, + "loss": 2.3003, + "step": 18572 + }, + { + "epoch": 4.05, + "learning_rate": 2.9765129997815167e-06, + "loss": 2.5977, + "step": 18573 + }, + { + "epoch": 4.05, + "learning_rate": 2.976403757920035e-06, + "loss": 1.6097, + "step": 18574 + }, + { + "epoch": 4.05, + "learning_rate": 2.976294516058554e-06, + "loss": 1.2092, + "step": 18575 + }, + { + "epoch": 4.05, + "learning_rate": 2.9761852741970725e-06, + "loss": 1.2837, + "step": 18576 + }, + { + "epoch": 4.05, + "learning_rate": 2.9760760323355916e-06, + "loss": 4.3406, + "step": 18577 + }, + { + "epoch": 4.05, + "learning_rate": 2.97596679047411e-06, + "loss": 2.4572, + "step": 18578 + }, + { + "epoch": 4.05, + "learning_rate": 2.9758575486126284e-06, + "loss": 1.9987, + "step": 18579 + }, + { + "epoch": 4.05, + "learning_rate": 2.9757483067511474e-06, + "loss": 1.6463, + "step": 18580 + }, + { + "epoch": 4.06, + "learning_rate": 2.975639064889666e-06, + "loss": 0.5402, + "step": 18581 + }, + { + "epoch": 4.06, + "learning_rate": 2.975529823028185e-06, + "loss": 2.0425, + "step": 18582 + }, + { + "epoch": 4.06, + "learning_rate": 2.9754205811667032e-06, + "loss": 1.4559, + "step": 18583 + }, + { + "epoch": 4.06, + "learning_rate": 2.975311339305222e-06, + "loss": 0.7355, + "step": 18584 + }, + { + "epoch": 4.06, + "learning_rate": 2.975202097443741e-06, + "loss": 1.7866, + "step": 18585 + }, + { + "epoch": 4.06, + "learning_rate": 2.9750928555822595e-06, + "loss": 1.35, + "step": 18586 + }, + { + "epoch": 4.06, + "learning_rate": 2.974983613720778e-06, + "loss": 0.4247, + "step": 18587 + }, + { + "epoch": 4.06, + "learning_rate": 2.9748743718592967e-06, + "loss": 2.1708, + "step": 18588 + }, + { + "epoch": 4.06, + "learning_rate": 2.9747651299978153e-06, + "loss": 1.3465, + "step": 18589 + }, + { + "epoch": 4.06, + "learning_rate": 2.974655888136334e-06, + "loss": 0.7594, + "step": 18590 + }, + { + "epoch": 4.06, + "learning_rate": 2.974546646274853e-06, + "loss": 1.1584, + "step": 18591 + }, + { + "epoch": 4.06, + "learning_rate": 2.9744374044133716e-06, + "loss": 0.8651, + "step": 18592 + }, + { + "epoch": 4.06, + "learning_rate": 2.9743281625518898e-06, + "loss": 1.6289, + "step": 18593 + }, + { + "epoch": 4.06, + "learning_rate": 2.9742189206904088e-06, + "loss": 2.2674, + "step": 18594 + }, + { + "epoch": 4.06, + "learning_rate": 2.9741096788289274e-06, + "loss": 1.0404, + "step": 18595 + }, + { + "epoch": 4.06, + "learning_rate": 2.9740004369674464e-06, + "loss": 0.6663, + "step": 18596 + }, + { + "epoch": 4.06, + "learning_rate": 2.9738911951059646e-06, + "loss": 2.176, + "step": 18597 + }, + { + "epoch": 4.06, + "learning_rate": 2.9737819532444832e-06, + "loss": 2.0548, + "step": 18598 + }, + { + "epoch": 4.06, + "learning_rate": 2.9736727113830023e-06, + "loss": 1.0665, + "step": 18599 + }, + { + "epoch": 4.06, + "learning_rate": 2.973563469521521e-06, + "loss": 2.9007, + "step": 18600 + }, + { + "epoch": 4.06, + "learning_rate": 2.97345422766004e-06, + "loss": 1.499, + "step": 18601 + }, + { + "epoch": 4.06, + "learning_rate": 2.973344985798558e-06, + "loss": 1.7351, + "step": 18602 + }, + { + "epoch": 4.06, + "learning_rate": 2.9732357439370767e-06, + "loss": 2.0318, + "step": 18603 + }, + { + "epoch": 4.06, + "learning_rate": 2.9731265020755957e-06, + "loss": 3.4223, + "step": 18604 + }, + { + "epoch": 4.06, + "learning_rate": 2.9730172602141144e-06, + "loss": 1.7429, + "step": 18605 + }, + { + "epoch": 4.06, + "learning_rate": 2.9729080183526334e-06, + "loss": 1.5823, + "step": 18606 + }, + { + "epoch": 4.06, + "learning_rate": 2.9727987764911516e-06, + "loss": 3.9352, + "step": 18607 + }, + { + "epoch": 4.06, + "learning_rate": 2.97268953462967e-06, + "loss": 2.0579, + "step": 18608 + }, + { + "epoch": 4.06, + "learning_rate": 2.9725802927681892e-06, + "loss": 1.5762, + "step": 18609 + }, + { + "epoch": 4.06, + "learning_rate": 2.972471050906708e-06, + "loss": 2.2268, + "step": 18610 + }, + { + "epoch": 4.06, + "learning_rate": 2.9723618090452264e-06, + "loss": 1.9635, + "step": 18611 + }, + { + "epoch": 4.06, + "learning_rate": 2.972252567183745e-06, + "loss": 0.4944, + "step": 18612 + }, + { + "epoch": 4.06, + "learning_rate": 2.9721433253222637e-06, + "loss": 0.8435, + "step": 18613 + }, + { + "epoch": 4.06, + "learning_rate": 2.9720340834607823e-06, + "loss": 1.1715, + "step": 18614 + }, + { + "epoch": 4.06, + "learning_rate": 2.9719248415993013e-06, + "loss": 2.665, + "step": 18615 + }, + { + "epoch": 4.06, + "learning_rate": 2.9718155997378195e-06, + "loss": 1.6098, + "step": 18616 + }, + { + "epoch": 4.06, + "learning_rate": 2.9717063578763385e-06, + "loss": 0.8484, + "step": 18617 + }, + { + "epoch": 4.06, + "learning_rate": 2.971597116014857e-06, + "loss": 1.9094, + "step": 18618 + }, + { + "epoch": 4.06, + "learning_rate": 2.9714878741533758e-06, + "loss": 1.2179, + "step": 18619 + }, + { + "epoch": 4.06, + "learning_rate": 2.971378632291895e-06, + "loss": 1.0578, + "step": 18620 + }, + { + "epoch": 4.06, + "learning_rate": 2.971269390430413e-06, + "loss": 1.1835, + "step": 18621 + }, + { + "epoch": 4.06, + "learning_rate": 2.9711601485689316e-06, + "loss": 1.8083, + "step": 18622 + }, + { + "epoch": 4.06, + "learning_rate": 2.9710509067074506e-06, + "loss": 0.5861, + "step": 18623 + }, + { + "epoch": 4.06, + "learning_rate": 2.9709416648459692e-06, + "loss": 1.9751, + "step": 18624 + }, + { + "epoch": 4.06, + "learning_rate": 2.9708324229844883e-06, + "loss": 2.523, + "step": 18625 + }, + { + "epoch": 4.07, + "learning_rate": 2.9707231811230065e-06, + "loss": 1.3018, + "step": 18626 + }, + { + "epoch": 4.07, + "learning_rate": 2.970613939261525e-06, + "loss": 1.4051, + "step": 18627 + }, + { + "epoch": 4.07, + "learning_rate": 2.970504697400044e-06, + "loss": 1.1631, + "step": 18628 + }, + { + "epoch": 4.07, + "learning_rate": 2.9703954555385627e-06, + "loss": 2.2708, + "step": 18629 + }, + { + "epoch": 4.07, + "learning_rate": 2.970286213677081e-06, + "loss": 0.83, + "step": 18630 + }, + { + "epoch": 4.07, + "learning_rate": 2.9701769718156e-06, + "loss": 0.8011, + "step": 18631 + }, + { + "epoch": 4.07, + "learning_rate": 2.9700677299541185e-06, + "loss": 2.9093, + "step": 18632 + }, + { + "epoch": 4.07, + "learning_rate": 2.9699584880926376e-06, + "loss": 1.2732, + "step": 18633 + }, + { + "epoch": 4.07, + "learning_rate": 2.969849246231156e-06, + "loss": 0.8905, + "step": 18634 + }, + { + "epoch": 4.07, + "learning_rate": 2.9697400043696744e-06, + "loss": 1.1465, + "step": 18635 + }, + { + "epoch": 4.07, + "learning_rate": 2.9696307625081934e-06, + "loss": 1.8804, + "step": 18636 + }, + { + "epoch": 4.07, + "learning_rate": 2.969521520646712e-06, + "loss": 1.7762, + "step": 18637 + }, + { + "epoch": 4.07, + "learning_rate": 2.969412278785231e-06, + "loss": 0.1571, + "step": 18638 + }, + { + "epoch": 4.07, + "learning_rate": 2.9693030369237497e-06, + "loss": 2.9747, + "step": 18639 + }, + { + "epoch": 4.07, + "learning_rate": 2.969193795062268e-06, + "loss": 1.8974, + "step": 18640 + }, + { + "epoch": 4.07, + "learning_rate": 2.969084553200787e-06, + "loss": 2.586, + "step": 18641 + }, + { + "epoch": 4.07, + "learning_rate": 2.9689753113393055e-06, + "loss": 3.0374, + "step": 18642 + }, + { + "epoch": 4.07, + "learning_rate": 2.968866069477824e-06, + "loss": 1.5971, + "step": 18643 + }, + { + "epoch": 4.07, + "learning_rate": 2.968756827616343e-06, + "loss": 2.2401, + "step": 18644 + }, + { + "epoch": 4.07, + "learning_rate": 2.9686475857548613e-06, + "loss": 0.852, + "step": 18645 + }, + { + "epoch": 4.07, + "learning_rate": 2.96853834389338e-06, + "loss": 0.9952, + "step": 18646 + }, + { + "epoch": 4.07, + "learning_rate": 2.968429102031899e-06, + "loss": 1.9257, + "step": 18647 + }, + { + "epoch": 4.07, + "learning_rate": 2.9683198601704176e-06, + "loss": 1.4466, + "step": 18648 + }, + { + "epoch": 4.07, + "learning_rate": 2.9682106183089358e-06, + "loss": 2.5936, + "step": 18649 + }, + { + "epoch": 4.07, + "learning_rate": 2.968101376447455e-06, + "loss": 2.3342, + "step": 18650 + }, + { + "epoch": 4.07, + "learning_rate": 2.9679921345859734e-06, + "loss": 2.1609, + "step": 18651 + }, + { + "epoch": 4.07, + "learning_rate": 2.9678828927244925e-06, + "loss": 2.1084, + "step": 18652 + }, + { + "epoch": 4.07, + "learning_rate": 2.967773650863011e-06, + "loss": 2.2855, + "step": 18653 + }, + { + "epoch": 4.07, + "learning_rate": 2.9676644090015293e-06, + "loss": 0.9326, + "step": 18654 + }, + { + "epoch": 4.07, + "learning_rate": 2.9675551671400483e-06, + "loss": 0.8674, + "step": 18655 + }, + { + "epoch": 4.07, + "learning_rate": 2.967445925278567e-06, + "loss": 2.3313, + "step": 18656 + }, + { + "epoch": 4.07, + "learning_rate": 2.967336683417086e-06, + "loss": 1.042, + "step": 18657 + }, + { + "epoch": 4.07, + "learning_rate": 2.9672274415556045e-06, + "loss": 1.784, + "step": 18658 + }, + { + "epoch": 4.07, + "learning_rate": 2.9671181996941227e-06, + "loss": 2.1299, + "step": 18659 + }, + { + "epoch": 4.07, + "learning_rate": 2.9670089578326418e-06, + "loss": 1.2588, + "step": 18660 + }, + { + "epoch": 4.07, + "learning_rate": 2.9668997159711604e-06, + "loss": 0.9013, + "step": 18661 + }, + { + "epoch": 4.07, + "learning_rate": 2.9667904741096794e-06, + "loss": 3.6196, + "step": 18662 + }, + { + "epoch": 4.07, + "learning_rate": 2.9666812322481976e-06, + "loss": 2.3821, + "step": 18663 + }, + { + "epoch": 4.07, + "learning_rate": 2.9665719903867162e-06, + "loss": 2.2108, + "step": 18664 + }, + { + "epoch": 4.07, + "learning_rate": 2.9664627485252352e-06, + "loss": 1.9164, + "step": 18665 + }, + { + "epoch": 4.07, + "learning_rate": 2.966353506663754e-06, + "loss": 0.479, + "step": 18666 + }, + { + "epoch": 4.07, + "learning_rate": 2.9662442648022725e-06, + "loss": 0.9449, + "step": 18667 + }, + { + "epoch": 4.07, + "learning_rate": 2.966135022940791e-06, + "loss": 2.168, + "step": 18668 + }, + { + "epoch": 4.07, + "learning_rate": 2.9660257810793097e-06, + "loss": 1.3001, + "step": 18669 + }, + { + "epoch": 4.07, + "learning_rate": 2.9659165392178287e-06, + "loss": 0.9096, + "step": 18670 + }, + { + "epoch": 4.07, + "learning_rate": 2.9658072973563473e-06, + "loss": 2.5048, + "step": 18671 + }, + { + "epoch": 4.08, + "learning_rate": 2.965698055494866e-06, + "loss": 0.9926, + "step": 18672 + }, + { + "epoch": 4.08, + "learning_rate": 2.9655888136333846e-06, + "loss": 1.5868, + "step": 18673 + }, + { + "epoch": 4.08, + "learning_rate": 2.965479571771903e-06, + "loss": 1.4241, + "step": 18674 + }, + { + "epoch": 4.08, + "learning_rate": 2.9653703299104218e-06, + "loss": 2.0678, + "step": 18675 + }, + { + "epoch": 4.08, + "learning_rate": 2.965261088048941e-06, + "loss": 1.9777, + "step": 18676 + }, + { + "epoch": 4.08, + "learning_rate": 2.9651518461874594e-06, + "loss": 2.0672, + "step": 18677 + }, + { + "epoch": 4.08, + "learning_rate": 2.9650426043259776e-06, + "loss": 2.0694, + "step": 18678 + }, + { + "epoch": 4.08, + "learning_rate": 2.9649333624644966e-06, + "loss": 0.9404, + "step": 18679 + }, + { + "epoch": 4.08, + "learning_rate": 2.9648241206030153e-06, + "loss": 1.9638, + "step": 18680 + }, + { + "epoch": 4.08, + "learning_rate": 2.9647148787415343e-06, + "loss": 2.1668, + "step": 18681 + }, + { + "epoch": 4.08, + "learning_rate": 2.9646056368800525e-06, + "loss": 1.8083, + "step": 18682 + }, + { + "epoch": 4.08, + "learning_rate": 2.964496395018571e-06, + "loss": 3.1334, + "step": 18683 + }, + { + "epoch": 4.08, + "learning_rate": 2.96438715315709e-06, + "loss": 1.1456, + "step": 18684 + }, + { + "epoch": 4.08, + "learning_rate": 2.9642779112956087e-06, + "loss": 1.3043, + "step": 18685 + }, + { + "epoch": 4.08, + "learning_rate": 2.9641686694341278e-06, + "loss": 2.6807, + "step": 18686 + }, + { + "epoch": 4.08, + "learning_rate": 2.964059427572646e-06, + "loss": 2.4981, + "step": 18687 + }, + { + "epoch": 4.08, + "learning_rate": 2.9639501857111646e-06, + "loss": 1.7556, + "step": 18688 + }, + { + "epoch": 4.08, + "learning_rate": 2.9638409438496836e-06, + "loss": 1.5584, + "step": 18689 + }, + { + "epoch": 4.08, + "learning_rate": 2.9637317019882022e-06, + "loss": 1.0561, + "step": 18690 + }, + { + "epoch": 4.08, + "learning_rate": 2.9636224601267213e-06, + "loss": 2.6896, + "step": 18691 + }, + { + "epoch": 4.08, + "learning_rate": 2.9635132182652394e-06, + "loss": 0.7113, + "step": 18692 + }, + { + "epoch": 4.08, + "learning_rate": 2.963403976403758e-06, + "loss": 1.2659, + "step": 18693 + }, + { + "epoch": 4.08, + "learning_rate": 2.963294734542277e-06, + "loss": 2.3172, + "step": 18694 + }, + { + "epoch": 4.08, + "learning_rate": 2.9631854926807957e-06, + "loss": 1.7876, + "step": 18695 + }, + { + "epoch": 4.08, + "learning_rate": 2.963076250819314e-06, + "loss": 1.1998, + "step": 18696 + }, + { + "epoch": 4.08, + "learning_rate": 2.962967008957833e-06, + "loss": 1.5577, + "step": 18697 + }, + { + "epoch": 4.08, + "learning_rate": 2.9628577670963515e-06, + "loss": 2.2471, + "step": 18698 + }, + { + "epoch": 4.08, + "learning_rate": 2.96274852523487e-06, + "loss": 2.7455, + "step": 18699 + }, + { + "epoch": 4.08, + "learning_rate": 2.962639283373389e-06, + "loss": 0.6203, + "step": 18700 + }, + { + "epoch": 4.08, + "learning_rate": 2.9625300415119074e-06, + "loss": 1.8342, + "step": 18701 + }, + { + "epoch": 4.08, + "learning_rate": 2.962420799650426e-06, + "loss": 1.5594, + "step": 18702 + }, + { + "epoch": 4.08, + "learning_rate": 2.962311557788945e-06, + "loss": 1.4092, + "step": 18703 + }, + { + "epoch": 4.08, + "learning_rate": 2.9622023159274636e-06, + "loss": 1.8104, + "step": 18704 + }, + { + "epoch": 4.08, + "learning_rate": 2.9620930740659827e-06, + "loss": 2.2336, + "step": 18705 + }, + { + "epoch": 4.08, + "learning_rate": 2.961983832204501e-06, + "loss": 0.9734, + "step": 18706 + }, + { + "epoch": 4.08, + "learning_rate": 2.9618745903430195e-06, + "loss": 1.8285, + "step": 18707 + }, + { + "epoch": 4.08, + "learning_rate": 2.9617653484815385e-06, + "loss": 1.6246, + "step": 18708 + }, + { + "epoch": 4.08, + "learning_rate": 2.961656106620057e-06, + "loss": 1.3414, + "step": 18709 + }, + { + "epoch": 4.08, + "learning_rate": 2.961546864758576e-06, + "loss": 2.6269, + "step": 18710 + }, + { + "epoch": 4.08, + "learning_rate": 2.9614376228970943e-06, + "loss": 2.0505, + "step": 18711 + }, + { + "epoch": 4.08, + "learning_rate": 2.961328381035613e-06, + "loss": 2.9389, + "step": 18712 + }, + { + "epoch": 4.08, + "learning_rate": 2.961219139174132e-06, + "loss": 1.5448, + "step": 18713 + }, + { + "epoch": 4.08, + "learning_rate": 2.9611098973126506e-06, + "loss": 0.5013, + "step": 18714 + }, + { + "epoch": 4.08, + "learning_rate": 2.9610006554511688e-06, + "loss": 0.9971, + "step": 18715 + }, + { + "epoch": 4.08, + "learning_rate": 2.960891413589688e-06, + "loss": 2.878, + "step": 18716 + }, + { + "epoch": 4.08, + "learning_rate": 2.9607821717282064e-06, + "loss": 3.0631, + "step": 18717 + }, + { + "epoch": 4.09, + "learning_rate": 2.9606729298667254e-06, + "loss": 1.548, + "step": 18718 + }, + { + "epoch": 4.09, + "learning_rate": 2.960563688005244e-06, + "loss": 1.6737, + "step": 18719 + }, + { + "epoch": 4.09, + "learning_rate": 2.9604544461437622e-06, + "loss": 2.169, + "step": 18720 + }, + { + "epoch": 4.09, + "learning_rate": 2.9603452042822813e-06, + "loss": 1.154, + "step": 18721 + }, + { + "epoch": 4.09, + "learning_rate": 2.9602359624208e-06, + "loss": 1.7867, + "step": 18722 + }, + { + "epoch": 4.09, + "learning_rate": 2.960126720559319e-06, + "loss": 0.3237, + "step": 18723 + }, + { + "epoch": 4.09, + "learning_rate": 2.9600174786978375e-06, + "loss": 3.185, + "step": 18724 + }, + { + "epoch": 4.09, + "learning_rate": 2.9599082368363557e-06, + "loss": 2.9238, + "step": 18725 + }, + { + "epoch": 4.09, + "learning_rate": 2.9597989949748748e-06, + "loss": 2.1476, + "step": 18726 + }, + { + "epoch": 4.09, + "learning_rate": 2.9596897531133934e-06, + "loss": 2.0657, + "step": 18727 + }, + { + "epoch": 4.09, + "learning_rate": 2.959580511251912e-06, + "loss": 1.3594, + "step": 18728 + }, + { + "epoch": 4.09, + "learning_rate": 2.9594712693904306e-06, + "loss": 2.5237, + "step": 18729 + }, + { + "epoch": 4.09, + "learning_rate": 2.959362027528949e-06, + "loss": 0.5476, + "step": 18730 + }, + { + "epoch": 4.09, + "learning_rate": 2.959252785667468e-06, + "loss": 1.5915, + "step": 18731 + }, + { + "epoch": 4.09, + "learning_rate": 2.959143543805987e-06, + "loss": 1.5941, + "step": 18732 + }, + { + "epoch": 4.09, + "learning_rate": 2.9590343019445055e-06, + "loss": 1.6549, + "step": 18733 + }, + { + "epoch": 4.09, + "learning_rate": 2.9589250600830236e-06, + "loss": 1.6514, + "step": 18734 + }, + { + "epoch": 4.09, + "learning_rate": 2.9588158182215427e-06, + "loss": 2.209, + "step": 18735 + }, + { + "epoch": 4.09, + "learning_rate": 2.9587065763600613e-06, + "loss": 2.1755, + "step": 18736 + }, + { + "epoch": 4.09, + "learning_rate": 2.9585973344985803e-06, + "loss": 0.4705, + "step": 18737 + }, + { + "epoch": 4.09, + "learning_rate": 2.958488092637099e-06, + "loss": 0.7165, + "step": 18738 + }, + { + "epoch": 4.09, + "learning_rate": 2.958378850775617e-06, + "loss": 1.5185, + "step": 18739 + }, + { + "epoch": 4.09, + "learning_rate": 2.958269608914136e-06, + "loss": 2.6468, + "step": 18740 + }, + { + "epoch": 4.09, + "learning_rate": 2.9581603670526548e-06, + "loss": 0.8268, + "step": 18741 + }, + { + "epoch": 4.09, + "learning_rate": 2.958051125191174e-06, + "loss": 2.306, + "step": 18742 + }, + { + "epoch": 4.09, + "learning_rate": 2.9579418833296924e-06, + "loss": 2.1862, + "step": 18743 + }, + { + "epoch": 4.09, + "learning_rate": 2.9578326414682106e-06, + "loss": 1.7941, + "step": 18744 + }, + { + "epoch": 4.09, + "learning_rate": 2.9577233996067296e-06, + "loss": 0.9545, + "step": 18745 + }, + { + "epoch": 4.09, + "learning_rate": 2.9576141577452482e-06, + "loss": 0.7576, + "step": 18746 + }, + { + "epoch": 4.09, + "learning_rate": 2.9575049158837673e-06, + "loss": 2.1976, + "step": 18747 + }, + { + "epoch": 4.09, + "learning_rate": 2.9573956740222855e-06, + "loss": 1.8851, + "step": 18748 + }, + { + "epoch": 4.09, + "learning_rate": 2.957286432160804e-06, + "loss": 1.9589, + "step": 18749 + }, + { + "epoch": 4.09, + "learning_rate": 2.957177190299323e-06, + "loss": 2.3478, + "step": 18750 + }, + { + "epoch": 4.09, + "learning_rate": 2.9570679484378417e-06, + "loss": 1.4904, + "step": 18751 + }, + { + "epoch": 4.09, + "learning_rate": 2.9569587065763603e-06, + "loss": 1.2174, + "step": 18752 + }, + { + "epoch": 4.09, + "learning_rate": 2.956849464714879e-06, + "loss": 2.687, + "step": 18753 + }, + { + "epoch": 4.09, + "learning_rate": 2.9567402228533976e-06, + "loss": 1.4201, + "step": 18754 + }, + { + "epoch": 4.09, + "learning_rate": 2.956630980991916e-06, + "loss": 1.4979, + "step": 18755 + }, + { + "epoch": 4.09, + "learning_rate": 2.956521739130435e-06, + "loss": 0.9098, + "step": 18756 + }, + { + "epoch": 4.09, + "learning_rate": 2.956412497268954e-06, + "loss": 2.8608, + "step": 18757 + }, + { + "epoch": 4.09, + "learning_rate": 2.956303255407472e-06, + "loss": 1.1175, + "step": 18758 + }, + { + "epoch": 4.09, + "learning_rate": 2.956194013545991e-06, + "loss": 2.7282, + "step": 18759 + }, + { + "epoch": 4.09, + "learning_rate": 2.9560847716845096e-06, + "loss": 2.5006, + "step": 18760 + }, + { + "epoch": 4.09, + "learning_rate": 2.9559755298230287e-06, + "loss": 1.7135, + "step": 18761 + }, + { + "epoch": 4.09, + "learning_rate": 2.955866287961547e-06, + "loss": 0.9706, + "step": 18762 + }, + { + "epoch": 4.09, + "learning_rate": 2.9557570461000655e-06, + "loss": 1.1201, + "step": 18763 + }, + { + "epoch": 4.1, + "learning_rate": 2.9556478042385845e-06, + "loss": 1.0653, + "step": 18764 + }, + { + "epoch": 4.1, + "learning_rate": 2.955538562377103e-06, + "loss": 2.6905, + "step": 18765 + }, + { + "epoch": 4.1, + "learning_rate": 2.955429320515622e-06, + "loss": 2.4208, + "step": 18766 + }, + { + "epoch": 4.1, + "learning_rate": 2.9553200786541403e-06, + "loss": 1.9262, + "step": 18767 + }, + { + "epoch": 4.1, + "learning_rate": 2.955210836792659e-06, + "loss": 2.3863, + "step": 18768 + }, + { + "epoch": 4.1, + "learning_rate": 2.955101594931178e-06, + "loss": 1.807, + "step": 18769 + }, + { + "epoch": 4.1, + "learning_rate": 2.9549923530696966e-06, + "loss": 2.6743, + "step": 18770 + }, + { + "epoch": 4.1, + "learning_rate": 2.9548831112082156e-06, + "loss": 1.604, + "step": 18771 + }, + { + "epoch": 4.1, + "learning_rate": 2.954773869346734e-06, + "loss": 2.7556, + "step": 18772 + }, + { + "epoch": 4.1, + "learning_rate": 2.9546646274852524e-06, + "loss": 2.9754, + "step": 18773 + }, + { + "epoch": 4.1, + "learning_rate": 2.9545553856237715e-06, + "loss": 2.1127, + "step": 18774 + }, + { + "epoch": 4.1, + "learning_rate": 2.95444614376229e-06, + "loss": 1.2992, + "step": 18775 + }, + { + "epoch": 4.1, + "learning_rate": 2.954336901900809e-06, + "loss": 1.2449, + "step": 18776 + }, + { + "epoch": 4.1, + "learning_rate": 2.9542276600393273e-06, + "loss": 2.2396, + "step": 18777 + }, + { + "epoch": 4.1, + "learning_rate": 2.954118418177846e-06, + "loss": 1.2144, + "step": 18778 + }, + { + "epoch": 4.1, + "learning_rate": 2.954009176316365e-06, + "loss": 0.8351, + "step": 18779 + }, + { + "epoch": 4.1, + "learning_rate": 2.9538999344548836e-06, + "loss": 0.8756, + "step": 18780 + }, + { + "epoch": 4.1, + "learning_rate": 2.9537906925934017e-06, + "loss": 2.2659, + "step": 18781 + }, + { + "epoch": 4.1, + "learning_rate": 2.9536814507319208e-06, + "loss": 1.4663, + "step": 18782 + }, + { + "epoch": 4.1, + "learning_rate": 2.9535722088704394e-06, + "loss": 1.7255, + "step": 18783 + }, + { + "epoch": 4.1, + "learning_rate": 2.953462967008958e-06, + "loss": 1.2423, + "step": 18784 + }, + { + "epoch": 4.1, + "learning_rate": 2.953353725147477e-06, + "loss": 1.1977, + "step": 18785 + }, + { + "epoch": 4.1, + "learning_rate": 2.9532444832859952e-06, + "loss": 1.4472, + "step": 18786 + }, + { + "epoch": 4.1, + "learning_rate": 2.953135241424514e-06, + "loss": 2.5303, + "step": 18787 + }, + { + "epoch": 4.1, + "learning_rate": 2.953025999563033e-06, + "loss": 2.2687, + "step": 18788 + }, + { + "epoch": 4.1, + "learning_rate": 2.9529167577015515e-06, + "loss": 2.9302, + "step": 18789 + }, + { + "epoch": 4.1, + "learning_rate": 2.9528075158400705e-06, + "loss": 1.3689, + "step": 18790 + }, + { + "epoch": 4.1, + "learning_rate": 2.9526982739785887e-06, + "loss": 1.5818, + "step": 18791 + }, + { + "epoch": 4.1, + "learning_rate": 2.9525890321171073e-06, + "loss": 2.7525, + "step": 18792 + }, + { + "epoch": 4.1, + "learning_rate": 2.9524797902556263e-06, + "loss": 3.5252, + "step": 18793 + }, + { + "epoch": 4.1, + "learning_rate": 2.952370548394145e-06, + "loss": 2.8885, + "step": 18794 + }, + { + "epoch": 4.1, + "learning_rate": 2.952261306532663e-06, + "loss": 2.943, + "step": 18795 + }, + { + "epoch": 4.1, + "learning_rate": 2.952152064671182e-06, + "loss": 1.0475, + "step": 18796 + }, + { + "epoch": 4.1, + "learning_rate": 2.952042822809701e-06, + "loss": 0.6237, + "step": 18797 + }, + { + "epoch": 4.1, + "learning_rate": 2.95193358094822e-06, + "loss": 2.7049, + "step": 18798 + }, + { + "epoch": 4.1, + "learning_rate": 2.9518243390867384e-06, + "loss": 1.6692, + "step": 18799 + }, + { + "epoch": 4.1, + "learning_rate": 2.9517150972252566e-06, + "loss": 2.1641, + "step": 18800 + }, + { + "epoch": 4.1, + "learning_rate": 2.9516058553637757e-06, + "loss": 2.9989, + "step": 18801 + }, + { + "epoch": 4.1, + "learning_rate": 2.9514966135022943e-06, + "loss": 2.3674, + "step": 18802 + }, + { + "epoch": 4.1, + "learning_rate": 2.9513873716408133e-06, + "loss": 1.5207, + "step": 18803 + }, + { + "epoch": 4.1, + "learning_rate": 2.951278129779332e-06, + "loss": 1.3663, + "step": 18804 + }, + { + "epoch": 4.1, + "learning_rate": 2.95116888791785e-06, + "loss": 2.4617, + "step": 18805 + }, + { + "epoch": 4.1, + "learning_rate": 2.951059646056369e-06, + "loss": 2.4578, + "step": 18806 + }, + { + "epoch": 4.1, + "learning_rate": 2.9509504041948877e-06, + "loss": 1.3462, + "step": 18807 + }, + { + "epoch": 4.1, + "learning_rate": 2.9508411623334064e-06, + "loss": 1.8565, + "step": 18808 + }, + { + "epoch": 4.1, + "learning_rate": 2.9507319204719254e-06, + "loss": 2.0502, + "step": 18809 + }, + { + "epoch": 4.11, + "learning_rate": 2.9506226786104436e-06, + "loss": 2.4857, + "step": 18810 + }, + { + "epoch": 4.11, + "learning_rate": 2.950513436748962e-06, + "loss": 0.8117, + "step": 18811 + }, + { + "epoch": 4.11, + "learning_rate": 2.9504041948874812e-06, + "loss": 2.1461, + "step": 18812 + }, + { + "epoch": 4.11, + "learning_rate": 2.950294953026e-06, + "loss": 2.2253, + "step": 18813 + }, + { + "epoch": 4.11, + "learning_rate": 2.9501857111645185e-06, + "loss": 1.2219, + "step": 18814 + }, + { + "epoch": 4.11, + "learning_rate": 2.950076469303037e-06, + "loss": 0.2741, + "step": 18815 + }, + { + "epoch": 4.11, + "learning_rate": 2.9499672274415557e-06, + "loss": 2.0872, + "step": 18816 + }, + { + "epoch": 4.11, + "learning_rate": 2.9498579855800747e-06, + "loss": 1.0908, + "step": 18817 + }, + { + "epoch": 4.11, + "learning_rate": 2.9497487437185933e-06, + "loss": 0.7335, + "step": 18818 + }, + { + "epoch": 4.11, + "learning_rate": 2.9496395018571115e-06, + "loss": 1.214, + "step": 18819 + }, + { + "epoch": 4.11, + "learning_rate": 2.9495302599956305e-06, + "loss": 2.0727, + "step": 18820 + }, + { + "epoch": 4.11, + "learning_rate": 2.949421018134149e-06, + "loss": 2.5413, + "step": 18821 + }, + { + "epoch": 4.11, + "learning_rate": 2.949311776272668e-06, + "loss": 2.3389, + "step": 18822 + }, + { + "epoch": 4.11, + "learning_rate": 2.949202534411187e-06, + "loss": 2.4597, + "step": 18823 + }, + { + "epoch": 4.11, + "learning_rate": 2.949093292549705e-06, + "loss": 1.8986, + "step": 18824 + }, + { + "epoch": 4.11, + "learning_rate": 2.948984050688224e-06, + "loss": 2.8301, + "step": 18825 + }, + { + "epoch": 4.11, + "learning_rate": 2.9488748088267426e-06, + "loss": 1.8332, + "step": 18826 + }, + { + "epoch": 4.11, + "learning_rate": 2.9487655669652617e-06, + "loss": 1.9919, + "step": 18827 + }, + { + "epoch": 4.11, + "learning_rate": 2.94865632510378e-06, + "loss": 1.5388, + "step": 18828 + }, + { + "epoch": 4.11, + "learning_rate": 2.9485470832422985e-06, + "loss": 1.1783, + "step": 18829 + }, + { + "epoch": 4.11, + "learning_rate": 2.9484378413808175e-06, + "loss": 1.654, + "step": 18830 + }, + { + "epoch": 4.11, + "learning_rate": 2.948328599519336e-06, + "loss": 1.5972, + "step": 18831 + }, + { + "epoch": 4.11, + "learning_rate": 2.948219357657855e-06, + "loss": 2.4519, + "step": 18832 + }, + { + "epoch": 4.11, + "learning_rate": 2.9481101157963733e-06, + "loss": 2.093, + "step": 18833 + }, + { + "epoch": 4.11, + "learning_rate": 2.948000873934892e-06, + "loss": 2.4486, + "step": 18834 + }, + { + "epoch": 4.11, + "learning_rate": 2.947891632073411e-06, + "loss": 1.0243, + "step": 18835 + }, + { + "epoch": 4.11, + "learning_rate": 2.9477823902119296e-06, + "loss": 0.8195, + "step": 18836 + }, + { + "epoch": 4.11, + "learning_rate": 2.947673148350448e-06, + "loss": 2.967, + "step": 18837 + }, + { + "epoch": 4.11, + "learning_rate": 2.947563906488967e-06, + "loss": 1.6987, + "step": 18838 + }, + { + "epoch": 4.11, + "learning_rate": 2.9474546646274854e-06, + "loss": 1.6624, + "step": 18839 + }, + { + "epoch": 4.11, + "learning_rate": 2.947345422766004e-06, + "loss": 0.8852, + "step": 18840 + }, + { + "epoch": 4.11, + "learning_rate": 2.947236180904523e-06, + "loss": 2.2249, + "step": 18841 + }, + { + "epoch": 4.11, + "learning_rate": 2.9471269390430413e-06, + "loss": 1.9639, + "step": 18842 + }, + { + "epoch": 4.11, + "learning_rate": 2.94701769718156e-06, + "loss": 1.3827, + "step": 18843 + }, + { + "epoch": 4.11, + "learning_rate": 2.946908455320079e-06, + "loss": 1.9844, + "step": 18844 + }, + { + "epoch": 4.11, + "learning_rate": 2.9467992134585975e-06, + "loss": 0.3844, + "step": 18845 + }, + { + "epoch": 4.11, + "learning_rate": 2.9466899715971165e-06, + "loss": 2.5298, + "step": 18846 + }, + { + "epoch": 4.11, + "learning_rate": 2.9465807297356347e-06, + "loss": 2.0011, + "step": 18847 + }, + { + "epoch": 4.11, + "learning_rate": 2.9464714878741533e-06, + "loss": 1.7386, + "step": 18848 + }, + { + "epoch": 4.11, + "learning_rate": 2.9463622460126724e-06, + "loss": 1.5155, + "step": 18849 + }, + { + "epoch": 4.11, + "learning_rate": 2.946253004151191e-06, + "loss": 2.3052, + "step": 18850 + }, + { + "epoch": 4.11, + "learning_rate": 2.94614376228971e-06, + "loss": 0.7105, + "step": 18851 + }, + { + "epoch": 4.11, + "learning_rate": 2.946034520428228e-06, + "loss": 1.3357, + "step": 18852 + }, + { + "epoch": 4.11, + "learning_rate": 2.945925278566747e-06, + "loss": 2.2245, + "step": 18853 + }, + { + "epoch": 4.11, + "learning_rate": 2.945816036705266e-06, + "loss": 0.9734, + "step": 18854 + }, + { + "epoch": 4.12, + "learning_rate": 2.9457067948437845e-06, + "loss": 0.7224, + "step": 18855 + }, + { + "epoch": 4.12, + "learning_rate": 2.9455975529823035e-06, + "loss": 2.0986, + "step": 18856 + }, + { + "epoch": 4.12, + "learning_rate": 2.9454883111208217e-06, + "loss": 1.6661, + "step": 18857 + }, + { + "epoch": 4.12, + "learning_rate": 2.9453790692593403e-06, + "loss": 0.7154, + "step": 18858 + }, + { + "epoch": 4.12, + "learning_rate": 2.9452698273978593e-06, + "loss": 1.4091, + "step": 18859 + }, + { + "epoch": 4.12, + "learning_rate": 2.945160585536378e-06, + "loss": 1.4726, + "step": 18860 + }, + { + "epoch": 4.12, + "learning_rate": 2.945051343674896e-06, + "loss": 3.136, + "step": 18861 + }, + { + "epoch": 4.12, + "learning_rate": 2.944942101813415e-06, + "loss": 1.3036, + "step": 18862 + }, + { + "epoch": 4.12, + "learning_rate": 2.9448328599519338e-06, + "loss": 1.3387, + "step": 18863 + }, + { + "epoch": 4.12, + "learning_rate": 2.9447236180904524e-06, + "loss": 2.1469, + "step": 18864 + }, + { + "epoch": 4.12, + "learning_rate": 2.9446143762289714e-06, + "loss": 1.283, + "step": 18865 + }, + { + "epoch": 4.12, + "learning_rate": 2.9445051343674896e-06, + "loss": 1.8418, + "step": 18866 + }, + { + "epoch": 4.12, + "learning_rate": 2.9443958925060086e-06, + "loss": 2.4472, + "step": 18867 + }, + { + "epoch": 4.12, + "learning_rate": 2.9442866506445273e-06, + "loss": 3.2507, + "step": 18868 + }, + { + "epoch": 4.12, + "learning_rate": 2.944177408783046e-06, + "loss": 2.768, + "step": 18869 + }, + { + "epoch": 4.12, + "learning_rate": 2.944068166921565e-06, + "loss": 1.7077, + "step": 18870 + }, + { + "epoch": 4.12, + "learning_rate": 2.943958925060083e-06, + "loss": 2.7693, + "step": 18871 + }, + { + "epoch": 4.12, + "learning_rate": 2.9438496831986017e-06, + "loss": 2.4354, + "step": 18872 + }, + { + "epoch": 4.12, + "learning_rate": 2.9437404413371207e-06, + "loss": 1.2322, + "step": 18873 + }, + { + "epoch": 4.12, + "learning_rate": 2.9436311994756393e-06, + "loss": 1.5113, + "step": 18874 + }, + { + "epoch": 4.12, + "learning_rate": 2.9435219576141575e-06, + "loss": 1.908, + "step": 18875 + }, + { + "epoch": 4.12, + "learning_rate": 2.9434127157526766e-06, + "loss": 1.7368, + "step": 18876 + }, + { + "epoch": 4.12, + "learning_rate": 2.943303473891195e-06, + "loss": 2.981, + "step": 18877 + }, + { + "epoch": 4.12, + "learning_rate": 2.9431942320297142e-06, + "loss": 1.5706, + "step": 18878 + }, + { + "epoch": 4.12, + "learning_rate": 2.943084990168233e-06, + "loss": 1.6759, + "step": 18879 + }, + { + "epoch": 4.12, + "learning_rate": 2.942975748306751e-06, + "loss": 3.2169, + "step": 18880 + }, + { + "epoch": 4.12, + "learning_rate": 2.94286650644527e-06, + "loss": 1.4762, + "step": 18881 + }, + { + "epoch": 4.12, + "learning_rate": 2.9427572645837887e-06, + "loss": 3.4149, + "step": 18882 + }, + { + "epoch": 4.12, + "learning_rate": 2.9426480227223077e-06, + "loss": 3.1611, + "step": 18883 + }, + { + "epoch": 4.12, + "learning_rate": 2.9425387808608263e-06, + "loss": 2.1762, + "step": 18884 + }, + { + "epoch": 4.12, + "learning_rate": 2.9424295389993445e-06, + "loss": 2.738, + "step": 18885 + }, + { + "epoch": 4.12, + "learning_rate": 2.9423202971378635e-06, + "loss": 3.2221, + "step": 18886 + }, + { + "epoch": 4.12, + "learning_rate": 2.942211055276382e-06, + "loss": 1.1072, + "step": 18887 + }, + { + "epoch": 4.12, + "learning_rate": 2.942101813414901e-06, + "loss": 1.9523, + "step": 18888 + }, + { + "epoch": 4.12, + "learning_rate": 2.9419925715534198e-06, + "loss": 1.6925, + "step": 18889 + }, + { + "epoch": 4.12, + "learning_rate": 2.941883329691938e-06, + "loss": 1.8123, + "step": 18890 + }, + { + "epoch": 4.12, + "learning_rate": 2.941774087830457e-06, + "loss": 1.5843, + "step": 18891 + }, + { + "epoch": 4.12, + "learning_rate": 2.9416648459689756e-06, + "loss": 2.9882, + "step": 18892 + }, + { + "epoch": 4.12, + "learning_rate": 2.9415556041074942e-06, + "loss": 2.3118, + "step": 18893 + }, + { + "epoch": 4.12, + "learning_rate": 2.941446362246013e-06, + "loss": 1.6759, + "step": 18894 + }, + { + "epoch": 4.12, + "learning_rate": 2.9413371203845314e-06, + "loss": 0.8087, + "step": 18895 + }, + { + "epoch": 4.12, + "learning_rate": 2.94122787852305e-06, + "loss": 0.6903, + "step": 18896 + }, + { + "epoch": 4.12, + "learning_rate": 2.941118636661569e-06, + "loss": 1.023, + "step": 18897 + }, + { + "epoch": 4.12, + "learning_rate": 2.9410093948000877e-06, + "loss": 1.4103, + "step": 18898 + }, + { + "epoch": 4.12, + "learning_rate": 2.940900152938606e-06, + "loss": 2.136, + "step": 18899 + }, + { + "epoch": 4.12, + "learning_rate": 2.940790911077125e-06, + "loss": 1.842, + "step": 18900 + }, + { + "epoch": 4.13, + "learning_rate": 2.9406816692156435e-06, + "loss": 1.1511, + "step": 18901 + }, + { + "epoch": 4.13, + "learning_rate": 2.9405724273541626e-06, + "loss": 2.345, + "step": 18902 + }, + { + "epoch": 4.13, + "learning_rate": 2.940463185492681e-06, + "loss": 2.1732, + "step": 18903 + }, + { + "epoch": 4.13, + "learning_rate": 2.9403539436311994e-06, + "loss": 1.487, + "step": 18904 + }, + { + "epoch": 4.13, + "learning_rate": 2.9402447017697184e-06, + "loss": 2.4975, + "step": 18905 + }, + { + "epoch": 4.13, + "learning_rate": 2.940135459908237e-06, + "loss": 1.3927, + "step": 18906 + }, + { + "epoch": 4.13, + "learning_rate": 2.940026218046756e-06, + "loss": 1.8432, + "step": 18907 + }, + { + "epoch": 4.13, + "learning_rate": 2.9399169761852742e-06, + "loss": 2.1749, + "step": 18908 + }, + { + "epoch": 4.13, + "learning_rate": 2.939807734323793e-06, + "loss": 2.4835, + "step": 18909 + }, + { + "epoch": 4.13, + "learning_rate": 2.939698492462312e-06, + "loss": 1.3395, + "step": 18910 + }, + { + "epoch": 4.13, + "learning_rate": 2.9395892506008305e-06, + "loss": 1.664, + "step": 18911 + }, + { + "epoch": 4.13, + "learning_rate": 2.9394800087393495e-06, + "loss": 1.9872, + "step": 18912 + }, + { + "epoch": 4.13, + "learning_rate": 2.9393707668778677e-06, + "loss": 1.9999, + "step": 18913 + }, + { + "epoch": 4.13, + "learning_rate": 2.9392615250163863e-06, + "loss": 1.4783, + "step": 18914 + }, + { + "epoch": 4.13, + "learning_rate": 2.9391522831549054e-06, + "loss": 1.8021, + "step": 18915 + }, + { + "epoch": 4.13, + "learning_rate": 2.939043041293424e-06, + "loss": 2.1386, + "step": 18916 + }, + { + "epoch": 4.13, + "learning_rate": 2.9389337994319426e-06, + "loss": 1.8112, + "step": 18917 + }, + { + "epoch": 4.13, + "learning_rate": 2.938824557570461e-06, + "loss": 0.2156, + "step": 18918 + }, + { + "epoch": 4.13, + "learning_rate": 2.93871531570898e-06, + "loss": 1.0764, + "step": 18919 + }, + { + "epoch": 4.13, + "learning_rate": 2.938606073847499e-06, + "loss": 1.4189, + "step": 18920 + }, + { + "epoch": 4.13, + "learning_rate": 2.9384968319860174e-06, + "loss": 1.2854, + "step": 18921 + }, + { + "epoch": 4.13, + "learning_rate": 2.938387590124536e-06, + "loss": 1.3677, + "step": 18922 + }, + { + "epoch": 4.13, + "learning_rate": 2.9382783482630547e-06, + "loss": 1.6887, + "step": 18923 + }, + { + "epoch": 4.13, + "learning_rate": 2.9381691064015733e-06, + "loss": 1.767, + "step": 18924 + }, + { + "epoch": 4.13, + "learning_rate": 2.938059864540092e-06, + "loss": 0.4668, + "step": 18925 + }, + { + "epoch": 4.13, + "learning_rate": 2.937950622678611e-06, + "loss": 0.8465, + "step": 18926 + }, + { + "epoch": 4.13, + "learning_rate": 2.937841380817129e-06, + "loss": 2.5098, + "step": 18927 + }, + { + "epoch": 4.13, + "learning_rate": 2.9377321389556477e-06, + "loss": 4.2068, + "step": 18928 + }, + { + "epoch": 4.13, + "learning_rate": 2.9376228970941668e-06, + "loss": 1.7194, + "step": 18929 + }, + { + "epoch": 4.13, + "learning_rate": 2.9375136552326854e-06, + "loss": 1.3718, + "step": 18930 + }, + { + "epoch": 4.13, + "learning_rate": 2.9374044133712044e-06, + "loss": 0.9155, + "step": 18931 + }, + { + "epoch": 4.13, + "learning_rate": 2.9372951715097226e-06, + "loss": 2.2737, + "step": 18932 + }, + { + "epoch": 4.13, + "learning_rate": 2.937185929648241e-06, + "loss": 2.2026, + "step": 18933 + }, + { + "epoch": 4.13, + "learning_rate": 2.9370766877867602e-06, + "loss": 3.0646, + "step": 18934 + }, + { + "epoch": 4.13, + "learning_rate": 2.936967445925279e-06, + "loss": 2.2718, + "step": 18935 + }, + { + "epoch": 4.13, + "learning_rate": 2.936858204063798e-06, + "loss": 0.8113, + "step": 18936 + }, + { + "epoch": 4.13, + "learning_rate": 2.936748962202316e-06, + "loss": 0.8694, + "step": 18937 + }, + { + "epoch": 4.13, + "learning_rate": 2.9366397203408347e-06, + "loss": 3.0935, + "step": 18938 + }, + { + "epoch": 4.13, + "learning_rate": 2.9365304784793537e-06, + "loss": 2.811, + "step": 18939 + }, + { + "epoch": 4.13, + "learning_rate": 2.9364212366178723e-06, + "loss": 2.5107, + "step": 18940 + }, + { + "epoch": 4.13, + "learning_rate": 2.9363119947563905e-06, + "loss": 1.4527, + "step": 18941 + }, + { + "epoch": 4.13, + "learning_rate": 2.9362027528949096e-06, + "loss": 3.3747, + "step": 18942 + }, + { + "epoch": 4.13, + "learning_rate": 2.936093511033428e-06, + "loss": 0.2676, + "step": 18943 + }, + { + "epoch": 4.13, + "learning_rate": 2.935984269171947e-06, + "loss": 1.5921, + "step": 18944 + }, + { + "epoch": 4.13, + "learning_rate": 2.935875027310466e-06, + "loss": 1.7972, + "step": 18945 + }, + { + "epoch": 4.13, + "learning_rate": 2.935765785448984e-06, + "loss": 2.8992, + "step": 18946 + }, + { + "epoch": 4.14, + "learning_rate": 2.935656543587503e-06, + "loss": 0.8526, + "step": 18947 + }, + { + "epoch": 4.14, + "learning_rate": 2.9355473017260216e-06, + "loss": 2.2538, + "step": 18948 + }, + { + "epoch": 4.14, + "learning_rate": 2.9354380598645403e-06, + "loss": 2.3651, + "step": 18949 + }, + { + "epoch": 4.14, + "learning_rate": 2.9353288180030593e-06, + "loss": 1.686, + "step": 18950 + }, + { + "epoch": 4.14, + "learning_rate": 2.9352195761415775e-06, + "loss": 1.7657, + "step": 18951 + }, + { + "epoch": 4.14, + "learning_rate": 2.935110334280096e-06, + "loss": 1.4643, + "step": 18952 + }, + { + "epoch": 4.14, + "learning_rate": 2.935001092418615e-06, + "loss": 1.3641, + "step": 18953 + }, + { + "epoch": 4.14, + "learning_rate": 2.9348918505571337e-06, + "loss": 0.9344, + "step": 18954 + }, + { + "epoch": 4.14, + "learning_rate": 2.9347826086956528e-06, + "loss": 1.9037, + "step": 18955 + }, + { + "epoch": 4.14, + "learning_rate": 2.934673366834171e-06, + "loss": 1.961, + "step": 18956 + }, + { + "epoch": 4.14, + "learning_rate": 2.9345641249726896e-06, + "loss": 3.3534, + "step": 18957 + }, + { + "epoch": 4.14, + "learning_rate": 2.9344548831112086e-06, + "loss": 1.2201, + "step": 18958 + }, + { + "epoch": 4.14, + "learning_rate": 2.934345641249727e-06, + "loss": 2.5907, + "step": 18959 + }, + { + "epoch": 4.14, + "learning_rate": 2.9342363993882454e-06, + "loss": 1.4824, + "step": 18960 + }, + { + "epoch": 4.14, + "learning_rate": 2.9341271575267644e-06, + "loss": 1.3281, + "step": 18961 + }, + { + "epoch": 4.14, + "learning_rate": 2.934017915665283e-06, + "loss": 2.5896, + "step": 18962 + }, + { + "epoch": 4.14, + "learning_rate": 2.933908673803802e-06, + "loss": 2.9047, + "step": 18963 + }, + { + "epoch": 4.14, + "learning_rate": 2.9337994319423207e-06, + "loss": 2.2918, + "step": 18964 + }, + { + "epoch": 4.14, + "learning_rate": 2.933690190080839e-06, + "loss": 1.4388, + "step": 18965 + }, + { + "epoch": 4.14, + "learning_rate": 2.933580948219358e-06, + "loss": 2.0217, + "step": 18966 + }, + { + "epoch": 4.14, + "learning_rate": 2.9334717063578765e-06, + "loss": 3.503, + "step": 18967 + }, + { + "epoch": 4.14, + "learning_rate": 2.9333624644963956e-06, + "loss": 2.4608, + "step": 18968 + }, + { + "epoch": 4.14, + "learning_rate": 2.933253222634914e-06, + "loss": 2.2046, + "step": 18969 + }, + { + "epoch": 4.14, + "learning_rate": 2.9331439807734324e-06, + "loss": 0.7602, + "step": 18970 + }, + { + "epoch": 4.14, + "learning_rate": 2.9330347389119514e-06, + "loss": 1.7053, + "step": 18971 + }, + { + "epoch": 4.14, + "learning_rate": 2.93292549705047e-06, + "loss": 2.203, + "step": 18972 + }, + { + "epoch": 4.14, + "learning_rate": 2.932816255188989e-06, + "loss": 1.0321, + "step": 18973 + }, + { + "epoch": 4.14, + "learning_rate": 2.9327070133275072e-06, + "loss": 2.106, + "step": 18974 + }, + { + "epoch": 4.14, + "learning_rate": 2.932597771466026e-06, + "loss": 1.0555, + "step": 18975 + }, + { + "epoch": 4.14, + "learning_rate": 2.932488529604545e-06, + "loss": 0.5591, + "step": 18976 + }, + { + "epoch": 4.14, + "learning_rate": 2.9323792877430635e-06, + "loss": 1.3829, + "step": 18977 + }, + { + "epoch": 4.14, + "learning_rate": 2.932270045881582e-06, + "loss": 1.9851, + "step": 18978 + }, + { + "epoch": 4.14, + "learning_rate": 2.9321608040201007e-06, + "loss": 1.3997, + "step": 18979 + }, + { + "epoch": 4.14, + "learning_rate": 2.9320515621586193e-06, + "loss": 2.5056, + "step": 18980 + }, + { + "epoch": 4.14, + "learning_rate": 2.931942320297138e-06, + "loss": 1.7522, + "step": 18981 + }, + { + "epoch": 4.14, + "learning_rate": 2.931833078435657e-06, + "loss": 1.1779, + "step": 18982 + }, + { + "epoch": 4.14, + "learning_rate": 2.9317238365741756e-06, + "loss": 1.067, + "step": 18983 + }, + { + "epoch": 4.14, + "learning_rate": 2.9316145947126938e-06, + "loss": 1.5766, + "step": 18984 + }, + { + "epoch": 4.14, + "learning_rate": 2.9315053528512128e-06, + "loss": 2.8627, + "step": 18985 + }, + { + "epoch": 4.14, + "learning_rate": 2.9313961109897314e-06, + "loss": 1.7826, + "step": 18986 + }, + { + "epoch": 4.14, + "learning_rate": 2.9312868691282504e-06, + "loss": 0.9824, + "step": 18987 + }, + { + "epoch": 4.14, + "learning_rate": 2.931177627266769e-06, + "loss": 2.6817, + "step": 18988 + }, + { + "epoch": 4.14, + "learning_rate": 2.9310683854052872e-06, + "loss": 1.5563, + "step": 18989 + }, + { + "epoch": 4.14, + "learning_rate": 2.9309591435438063e-06, + "loss": 2.3645, + "step": 18990 + }, + { + "epoch": 4.14, + "learning_rate": 2.930849901682325e-06, + "loss": 1.4857, + "step": 18991 + }, + { + "epoch": 4.14, + "learning_rate": 2.930740659820844e-06, + "loss": 1.0179, + "step": 18992 + }, + { + "epoch": 4.15, + "learning_rate": 2.930631417959362e-06, + "loss": 0.9713, + "step": 18993 + }, + { + "epoch": 4.15, + "learning_rate": 2.9305221760978807e-06, + "loss": 2.1137, + "step": 18994 + }, + { + "epoch": 4.15, + "learning_rate": 2.9304129342363997e-06, + "loss": 1.2687, + "step": 18995 + }, + { + "epoch": 4.15, + "learning_rate": 2.9303036923749184e-06, + "loss": 1.6514, + "step": 18996 + }, + { + "epoch": 4.15, + "learning_rate": 2.9301944505134374e-06, + "loss": 0.9434, + "step": 18997 + }, + { + "epoch": 4.15, + "learning_rate": 2.9300852086519556e-06, + "loss": 1.7321, + "step": 18998 + }, + { + "epoch": 4.15, + "learning_rate": 2.929975966790474e-06, + "loss": 0.7014, + "step": 18999 + }, + { + "epoch": 4.15, + "learning_rate": 2.9298667249289932e-06, + "loss": 0.5845, + "step": 19000 + }, + { + "epoch": 4.15, + "learning_rate": 2.929757483067512e-06, + "loss": 2.4631, + "step": 19001 + }, + { + "epoch": 4.15, + "learning_rate": 2.9296482412060304e-06, + "loss": 1.8941, + "step": 19002 + }, + { + "epoch": 4.15, + "learning_rate": 2.929538999344549e-06, + "loss": 2.0721, + "step": 19003 + }, + { + "epoch": 4.15, + "learning_rate": 2.9294297574830677e-06, + "loss": 0.8047, + "step": 19004 + }, + { + "epoch": 4.15, + "learning_rate": 2.9293205156215863e-06, + "loss": 2.16, + "step": 19005 + }, + { + "epoch": 4.15, + "learning_rate": 2.9292112737601053e-06, + "loss": 2.4017, + "step": 19006 + }, + { + "epoch": 4.15, + "learning_rate": 2.9291020318986235e-06, + "loss": 1.9341, + "step": 19007 + }, + { + "epoch": 4.15, + "learning_rate": 2.928992790037142e-06, + "loss": 0.7614, + "step": 19008 + }, + { + "epoch": 4.15, + "learning_rate": 2.928883548175661e-06, + "loss": 0.6713, + "step": 19009 + }, + { + "epoch": 4.15, + "learning_rate": 2.9287743063141798e-06, + "loss": 1.0015, + "step": 19010 + }, + { + "epoch": 4.15, + "learning_rate": 2.928665064452699e-06, + "loss": 1.6455, + "step": 19011 + }, + { + "epoch": 4.15, + "learning_rate": 2.928555822591217e-06, + "loss": 0.3957, + "step": 19012 + }, + { + "epoch": 4.15, + "learning_rate": 2.9284465807297356e-06, + "loss": 0.8346, + "step": 19013 + }, + { + "epoch": 4.15, + "learning_rate": 2.9283373388682546e-06, + "loss": 2.0785, + "step": 19014 + }, + { + "epoch": 4.15, + "learning_rate": 2.9282280970067732e-06, + "loss": 0.9785, + "step": 19015 + }, + { + "epoch": 4.15, + "learning_rate": 2.9281188551452923e-06, + "loss": 0.588, + "step": 19016 + }, + { + "epoch": 4.15, + "learning_rate": 2.9280096132838105e-06, + "loss": 2.4887, + "step": 19017 + }, + { + "epoch": 4.15, + "learning_rate": 2.927900371422329e-06, + "loss": 1.4171, + "step": 19018 + }, + { + "epoch": 4.15, + "learning_rate": 2.927791129560848e-06, + "loss": 1.1321, + "step": 19019 + }, + { + "epoch": 4.15, + "learning_rate": 2.9276818876993667e-06, + "loss": 3.166, + "step": 19020 + }, + { + "epoch": 4.15, + "learning_rate": 2.9275726458378857e-06, + "loss": 1.1239, + "step": 19021 + }, + { + "epoch": 4.15, + "learning_rate": 2.927463403976404e-06, + "loss": 1.5408, + "step": 19022 + }, + { + "epoch": 4.15, + "learning_rate": 2.9273541621149225e-06, + "loss": 1.5364, + "step": 19023 + }, + { + "epoch": 4.15, + "learning_rate": 2.9272449202534416e-06, + "loss": 2.8231, + "step": 19024 + }, + { + "epoch": 4.15, + "learning_rate": 2.92713567839196e-06, + "loss": 2.2265, + "step": 19025 + }, + { + "epoch": 4.15, + "learning_rate": 2.9270264365304784e-06, + "loss": 0.7546, + "step": 19026 + }, + { + "epoch": 4.15, + "learning_rate": 2.9269171946689974e-06, + "loss": 1.0583, + "step": 19027 + }, + { + "epoch": 4.15, + "learning_rate": 2.926807952807516e-06, + "loss": 1.2706, + "step": 19028 + }, + { + "epoch": 4.15, + "learning_rate": 2.926698710946035e-06, + "loss": 1.5582, + "step": 19029 + }, + { + "epoch": 4.15, + "learning_rate": 2.9265894690845537e-06, + "loss": 1.6718, + "step": 19030 + }, + { + "epoch": 4.15, + "learning_rate": 2.926480227223072e-06, + "loss": 2.8528, + "step": 19031 + }, + { + "epoch": 4.15, + "learning_rate": 2.926370985361591e-06, + "loss": 1.8675, + "step": 19032 + }, + { + "epoch": 4.15, + "learning_rate": 2.9262617435001095e-06, + "loss": 1.6603, + "step": 19033 + }, + { + "epoch": 4.15, + "learning_rate": 2.926152501638628e-06, + "loss": 2.481, + "step": 19034 + }, + { + "epoch": 4.15, + "learning_rate": 2.926043259777147e-06, + "loss": 2.9045, + "step": 19035 + }, + { + "epoch": 4.15, + "learning_rate": 2.9259340179156653e-06, + "loss": 1.4487, + "step": 19036 + }, + { + "epoch": 4.15, + "learning_rate": 2.925824776054184e-06, + "loss": 2.1865, + "step": 19037 + }, + { + "epoch": 4.15, + "learning_rate": 2.925715534192703e-06, + "loss": 1.7321, + "step": 19038 + }, + { + "epoch": 4.16, + "learning_rate": 2.9256062923312216e-06, + "loss": 1.7266, + "step": 19039 + }, + { + "epoch": 4.16, + "learning_rate": 2.9254970504697398e-06, + "loss": 1.0582, + "step": 19040 + }, + { + "epoch": 4.16, + "learning_rate": 2.925387808608259e-06, + "loss": 1.2391, + "step": 19041 + }, + { + "epoch": 4.16, + "learning_rate": 2.9252785667467774e-06, + "loss": 1.6357, + "step": 19042 + }, + { + "epoch": 4.16, + "learning_rate": 2.9251693248852965e-06, + "loss": 1.8398, + "step": 19043 + }, + { + "epoch": 4.16, + "learning_rate": 2.925060083023815e-06, + "loss": 1.8611, + "step": 19044 + }, + { + "epoch": 4.16, + "learning_rate": 2.9249508411623333e-06, + "loss": 1.2913, + "step": 19045 + }, + { + "epoch": 4.16, + "learning_rate": 2.9248415993008523e-06, + "loss": 1.2271, + "step": 19046 + }, + { + "epoch": 4.16, + "learning_rate": 2.924732357439371e-06, + "loss": 0.8338, + "step": 19047 + }, + { + "epoch": 4.16, + "learning_rate": 2.92462311557789e-06, + "loss": 1.5003, + "step": 19048 + }, + { + "epoch": 4.16, + "learning_rate": 2.9245138737164085e-06, + "loss": 0.7608, + "step": 19049 + }, + { + "epoch": 4.16, + "learning_rate": 2.9244046318549267e-06, + "loss": 2.6729, + "step": 19050 + }, + { + "epoch": 4.16, + "learning_rate": 2.9242953899934458e-06, + "loss": 0.6276, + "step": 19051 + }, + { + "epoch": 4.16, + "learning_rate": 2.9241861481319644e-06, + "loss": 2.2421, + "step": 19052 + }, + { + "epoch": 4.16, + "learning_rate": 2.9240769062704834e-06, + "loss": 1.7133, + "step": 19053 + }, + { + "epoch": 4.16, + "learning_rate": 2.923967664409002e-06, + "loss": 0.9454, + "step": 19054 + }, + { + "epoch": 4.16, + "learning_rate": 2.9238584225475202e-06, + "loss": 0.844, + "step": 19055 + }, + { + "epoch": 4.16, + "learning_rate": 2.9237491806860392e-06, + "loss": 1.551, + "step": 19056 + }, + { + "epoch": 4.16, + "learning_rate": 2.923639938824558e-06, + "loss": 1.6953, + "step": 19057 + }, + { + "epoch": 4.16, + "learning_rate": 2.9235306969630765e-06, + "loss": 0.8573, + "step": 19058 + }, + { + "epoch": 4.16, + "learning_rate": 2.923421455101595e-06, + "loss": 2.5603, + "step": 19059 + }, + { + "epoch": 4.16, + "learning_rate": 2.9233122132401137e-06, + "loss": 2.0735, + "step": 19060 + }, + { + "epoch": 4.16, + "learning_rate": 2.9232029713786323e-06, + "loss": 2.0763, + "step": 19061 + }, + { + "epoch": 4.16, + "learning_rate": 2.9230937295171513e-06, + "loss": 1.7775, + "step": 19062 + }, + { + "epoch": 4.16, + "learning_rate": 2.92298448765567e-06, + "loss": 1.4688, + "step": 19063 + }, + { + "epoch": 4.16, + "learning_rate": 2.9228752457941886e-06, + "loss": 0.6927, + "step": 19064 + }, + { + "epoch": 4.16, + "learning_rate": 2.922766003932707e-06, + "loss": 0.9562, + "step": 19065 + }, + { + "epoch": 4.16, + "learning_rate": 2.9226567620712258e-06, + "loss": 2.5928, + "step": 19066 + }, + { + "epoch": 4.16, + "learning_rate": 2.922547520209745e-06, + "loss": 1.494, + "step": 19067 + }, + { + "epoch": 4.16, + "learning_rate": 2.9224382783482634e-06, + "loss": 1.3712, + "step": 19068 + }, + { + "epoch": 4.16, + "learning_rate": 2.9223290364867816e-06, + "loss": 1.3903, + "step": 19069 + }, + { + "epoch": 4.16, + "learning_rate": 2.9222197946253007e-06, + "loss": 2.8034, + "step": 19070 + }, + { + "epoch": 4.16, + "learning_rate": 2.9221105527638193e-06, + "loss": 1.099, + "step": 19071 + }, + { + "epoch": 4.16, + "learning_rate": 2.9220013109023383e-06, + "loss": 1.1274, + "step": 19072 + }, + { + "epoch": 4.16, + "learning_rate": 2.9218920690408565e-06, + "loss": 1.1171, + "step": 19073 + }, + { + "epoch": 4.16, + "learning_rate": 2.921782827179375e-06, + "loss": 1.5246, + "step": 19074 + }, + { + "epoch": 4.16, + "learning_rate": 2.921673585317894e-06, + "loss": 3.5212, + "step": 19075 + }, + { + "epoch": 4.16, + "learning_rate": 2.9215643434564127e-06, + "loss": 1.0581, + "step": 19076 + }, + { + "epoch": 4.16, + "learning_rate": 2.9214551015949318e-06, + "loss": 1.7756, + "step": 19077 + }, + { + "epoch": 4.16, + "learning_rate": 2.92134585973345e-06, + "loss": 1.3726, + "step": 19078 + }, + { + "epoch": 4.16, + "learning_rate": 2.9212366178719686e-06, + "loss": 3.6008, + "step": 19079 + }, + { + "epoch": 4.16, + "learning_rate": 2.9211273760104876e-06, + "loss": 1.4878, + "step": 19080 + }, + { + "epoch": 4.16, + "learning_rate": 2.9210181341490062e-06, + "loss": 1.6739, + "step": 19081 + }, + { + "epoch": 4.16, + "learning_rate": 2.9209088922875253e-06, + "loss": 1.3415, + "step": 19082 + }, + { + "epoch": 4.16, + "learning_rate": 2.9207996504260434e-06, + "loss": 1.4533, + "step": 19083 + }, + { + "epoch": 4.16, + "learning_rate": 2.920690408564562e-06, + "loss": 1.8331, + "step": 19084 + }, + { + "epoch": 4.17, + "learning_rate": 2.920581166703081e-06, + "loss": 1.6713, + "step": 19085 + }, + { + "epoch": 4.17, + "learning_rate": 2.9204719248415997e-06, + "loss": 1.5171, + "step": 19086 + }, + { + "epoch": 4.17, + "learning_rate": 2.9203626829801183e-06, + "loss": 1.4162, + "step": 19087 + }, + { + "epoch": 4.17, + "learning_rate": 2.920253441118637e-06, + "loss": 3.0246, + "step": 19088 + }, + { + "epoch": 4.17, + "learning_rate": 2.9201441992571555e-06, + "loss": 1.3384, + "step": 19089 + }, + { + "epoch": 4.17, + "learning_rate": 2.920034957395674e-06, + "loss": 1.3698, + "step": 19090 + }, + { + "epoch": 4.17, + "learning_rate": 2.919925715534193e-06, + "loss": 0.9855, + "step": 19091 + }, + { + "epoch": 4.17, + "learning_rate": 2.9198164736727114e-06, + "loss": 1.9639, + "step": 19092 + }, + { + "epoch": 4.17, + "learning_rate": 2.91970723181123e-06, + "loss": 2.0161, + "step": 19093 + }, + { + "epoch": 4.17, + "learning_rate": 2.919597989949749e-06, + "loss": 2.5022, + "step": 19094 + }, + { + "epoch": 4.17, + "learning_rate": 2.9194887480882676e-06, + "loss": 1.2378, + "step": 19095 + }, + { + "epoch": 4.17, + "learning_rate": 2.9193795062267867e-06, + "loss": 1.5122, + "step": 19096 + }, + { + "epoch": 4.17, + "learning_rate": 2.919270264365305e-06, + "loss": 2.0975, + "step": 19097 + }, + { + "epoch": 4.17, + "learning_rate": 2.9191610225038235e-06, + "loss": 1.372, + "step": 19098 + }, + { + "epoch": 4.17, + "learning_rate": 2.9190517806423425e-06, + "loss": 2.0254, + "step": 19099 + }, + { + "epoch": 4.17, + "learning_rate": 2.918942538780861e-06, + "loss": 1.3616, + "step": 19100 + }, + { + "epoch": 4.17, + "learning_rate": 2.91883329691938e-06, + "loss": 1.6791, + "step": 19101 + }, + { + "epoch": 4.17, + "learning_rate": 2.9187240550578983e-06, + "loss": 1.8044, + "step": 19102 + }, + { + "epoch": 4.17, + "learning_rate": 2.918614813196417e-06, + "loss": 1.6126, + "step": 19103 + }, + { + "epoch": 4.17, + "learning_rate": 2.918505571334936e-06, + "loss": 2.5468, + "step": 19104 + }, + { + "epoch": 4.17, + "learning_rate": 2.9183963294734546e-06, + "loss": 1.2262, + "step": 19105 + }, + { + "epoch": 4.17, + "learning_rate": 2.9182870876119728e-06, + "loss": 1.3815, + "step": 19106 + }, + { + "epoch": 4.17, + "learning_rate": 2.918177845750492e-06, + "loss": 1.936, + "step": 19107 + }, + { + "epoch": 4.17, + "learning_rate": 2.9180686038890104e-06, + "loss": 3.265, + "step": 19108 + }, + { + "epoch": 4.17, + "learning_rate": 2.9179593620275294e-06, + "loss": 0.8922, + "step": 19109 + }, + { + "epoch": 4.17, + "learning_rate": 2.917850120166048e-06, + "loss": 1.6791, + "step": 19110 + }, + { + "epoch": 4.17, + "learning_rate": 2.9177408783045662e-06, + "loss": 1.7775, + "step": 19111 + }, + { + "epoch": 4.17, + "learning_rate": 2.9176316364430853e-06, + "loss": 1.0674, + "step": 19112 + }, + { + "epoch": 4.17, + "learning_rate": 2.917522394581604e-06, + "loss": 1.4721, + "step": 19113 + }, + { + "epoch": 4.17, + "learning_rate": 2.9174131527201225e-06, + "loss": 1.5802, + "step": 19114 + }, + { + "epoch": 4.17, + "learning_rate": 2.9173039108586415e-06, + "loss": 1.2318, + "step": 19115 + }, + { + "epoch": 4.17, + "learning_rate": 2.9171946689971597e-06, + "loss": 1.272, + "step": 19116 + }, + { + "epoch": 4.17, + "learning_rate": 2.9170854271356788e-06, + "loss": 2.219, + "step": 19117 + }, + { + "epoch": 4.17, + "learning_rate": 2.9169761852741974e-06, + "loss": 1.1755, + "step": 19118 + }, + { + "epoch": 4.17, + "learning_rate": 2.916866943412716e-06, + "loss": 2.986, + "step": 19119 + }, + { + "epoch": 4.17, + "learning_rate": 2.916757701551235e-06, + "loss": 1.4747, + "step": 19120 + }, + { + "epoch": 4.17, + "learning_rate": 2.916648459689753e-06, + "loss": 1.4592, + "step": 19121 + }, + { + "epoch": 4.17, + "learning_rate": 2.916539217828272e-06, + "loss": 2.0467, + "step": 19122 + }, + { + "epoch": 4.17, + "learning_rate": 2.916429975966791e-06, + "loss": 2.0233, + "step": 19123 + }, + { + "epoch": 4.17, + "learning_rate": 2.9163207341053095e-06, + "loss": 0.6211, + "step": 19124 + }, + { + "epoch": 4.17, + "learning_rate": 2.9162114922438276e-06, + "loss": 1.6353, + "step": 19125 + }, + { + "epoch": 4.17, + "learning_rate": 2.9161022503823467e-06, + "loss": 1.0449, + "step": 19126 + }, + { + "epoch": 4.17, + "learning_rate": 2.9159930085208653e-06, + "loss": 0.394, + "step": 19127 + }, + { + "epoch": 4.17, + "learning_rate": 2.9158837666593843e-06, + "loss": 2.4592, + "step": 19128 + }, + { + "epoch": 4.17, + "learning_rate": 2.915774524797903e-06, + "loss": 1.1414, + "step": 19129 + }, + { + "epoch": 4.18, + "learning_rate": 2.915665282936421e-06, + "loss": 1.2067, + "step": 19130 + }, + { + "epoch": 4.18, + "learning_rate": 2.91555604107494e-06, + "loss": 1.9267, + "step": 19131 + }, + { + "epoch": 4.18, + "learning_rate": 2.9154467992134588e-06, + "loss": 1.6844, + "step": 19132 + }, + { + "epoch": 4.18, + "learning_rate": 2.915337557351978e-06, + "loss": 2.0037, + "step": 19133 + }, + { + "epoch": 4.18, + "learning_rate": 2.9152283154904964e-06, + "loss": 1.9104, + "step": 19134 + }, + { + "epoch": 4.18, + "learning_rate": 2.9151190736290146e-06, + "loss": 0.9625, + "step": 19135 + }, + { + "epoch": 4.18, + "learning_rate": 2.9150098317675336e-06, + "loss": 1.9637, + "step": 19136 + }, + { + "epoch": 4.18, + "learning_rate": 2.9149005899060522e-06, + "loss": 1.5003, + "step": 19137 + }, + { + "epoch": 4.18, + "learning_rate": 2.9147913480445713e-06, + "loss": 1.8934, + "step": 19138 + }, + { + "epoch": 4.18, + "learning_rate": 2.9146821061830895e-06, + "loss": 0.9013, + "step": 19139 + }, + { + "epoch": 4.18, + "learning_rate": 2.914572864321608e-06, + "loss": 1.8853, + "step": 19140 + }, + { + "epoch": 4.18, + "learning_rate": 2.914463622460127e-06, + "loss": 1.7012, + "step": 19141 + }, + { + "epoch": 4.18, + "learning_rate": 2.9143543805986457e-06, + "loss": 1.8493, + "step": 19142 + }, + { + "epoch": 4.18, + "learning_rate": 2.9142451387371643e-06, + "loss": 1.738, + "step": 19143 + }, + { + "epoch": 4.18, + "learning_rate": 2.914135896875683e-06, + "loss": 2.6288, + "step": 19144 + }, + { + "epoch": 4.18, + "learning_rate": 2.9140266550142016e-06, + "loss": 0.4325, + "step": 19145 + }, + { + "epoch": 4.18, + "learning_rate": 2.91391741315272e-06, + "loss": 2.4115, + "step": 19146 + }, + { + "epoch": 4.18, + "learning_rate": 2.913808171291239e-06, + "loss": 1.4178, + "step": 19147 + }, + { + "epoch": 4.18, + "learning_rate": 2.913698929429758e-06, + "loss": 1.9564, + "step": 19148 + }, + { + "epoch": 4.18, + "learning_rate": 2.913589687568276e-06, + "loss": 1.1524, + "step": 19149 + }, + { + "epoch": 4.18, + "learning_rate": 2.913480445706795e-06, + "loss": 0.9966, + "step": 19150 + }, + { + "epoch": 4.18, + "learning_rate": 2.9133712038453136e-06, + "loss": 1.6177, + "step": 19151 + }, + { + "epoch": 4.18, + "learning_rate": 2.9132619619838327e-06, + "loss": 1.5364, + "step": 19152 + }, + { + "epoch": 4.18, + "learning_rate": 2.9131527201223513e-06, + "loss": 1.0914, + "step": 19153 + }, + { + "epoch": 4.18, + "learning_rate": 2.9130434782608695e-06, + "loss": 0.5264, + "step": 19154 + }, + { + "epoch": 4.18, + "learning_rate": 2.9129342363993885e-06, + "loss": 2.0014, + "step": 19155 + }, + { + "epoch": 4.18, + "learning_rate": 2.912824994537907e-06, + "loss": 1.5249, + "step": 19156 + }, + { + "epoch": 4.18, + "learning_rate": 2.912715752676426e-06, + "loss": 1.3367, + "step": 19157 + }, + { + "epoch": 4.18, + "learning_rate": 2.9126065108149443e-06, + "loss": 1.7099, + "step": 19158 + }, + { + "epoch": 4.18, + "learning_rate": 2.912497268953463e-06, + "loss": 0.3512, + "step": 19159 + }, + { + "epoch": 4.18, + "learning_rate": 2.912388027091982e-06, + "loss": 1.4468, + "step": 19160 + }, + { + "epoch": 4.18, + "learning_rate": 2.9122787852305006e-06, + "loss": 1.7289, + "step": 19161 + }, + { + "epoch": 4.18, + "learning_rate": 2.9121695433690196e-06, + "loss": 1.2817, + "step": 19162 + }, + { + "epoch": 4.18, + "learning_rate": 2.912060301507538e-06, + "loss": 1.0005, + "step": 19163 + }, + { + "epoch": 4.18, + "learning_rate": 2.9119510596460564e-06, + "loss": 1.1717, + "step": 19164 + }, + { + "epoch": 4.18, + "learning_rate": 2.9118418177845755e-06, + "loss": 0.7881, + "step": 19165 + }, + { + "epoch": 4.18, + "learning_rate": 2.911732575923094e-06, + "loss": 1.1408, + "step": 19166 + }, + { + "epoch": 4.18, + "learning_rate": 2.911623334061613e-06, + "loss": 3.7075, + "step": 19167 + }, + { + "epoch": 4.18, + "learning_rate": 2.9115140922001313e-06, + "loss": 1.4231, + "step": 19168 + }, + { + "epoch": 4.18, + "learning_rate": 2.91140485033865e-06, + "loss": 1.4381, + "step": 19169 + }, + { + "epoch": 4.18, + "learning_rate": 2.911295608477169e-06, + "loss": 2.4241, + "step": 19170 + }, + { + "epoch": 4.18, + "learning_rate": 2.9111863666156876e-06, + "loss": 1.5325, + "step": 19171 + }, + { + "epoch": 4.18, + "learning_rate": 2.9110771247542057e-06, + "loss": 1.503, + "step": 19172 + }, + { + "epoch": 4.18, + "learning_rate": 2.9109678828927248e-06, + "loss": 2.4328, + "step": 19173 + }, + { + "epoch": 4.18, + "learning_rate": 2.9108586410312434e-06, + "loss": 3.1336, + "step": 19174 + }, + { + "epoch": 4.18, + "learning_rate": 2.910749399169762e-06, + "loss": 1.0522, + "step": 19175 + }, + { + "epoch": 4.19, + "learning_rate": 2.910640157308281e-06, + "loss": 0.7281, + "step": 19176 + }, + { + "epoch": 4.19, + "learning_rate": 2.9105309154467992e-06, + "loss": 1.4484, + "step": 19177 + }, + { + "epoch": 4.19, + "learning_rate": 2.910421673585318e-06, + "loss": 0.9343, + "step": 19178 + }, + { + "epoch": 4.19, + "learning_rate": 2.910312431723837e-06, + "loss": 1.0996, + "step": 19179 + }, + { + "epoch": 4.19, + "learning_rate": 2.9102031898623555e-06, + "loss": 0.3107, + "step": 19180 + }, + { + "epoch": 4.19, + "learning_rate": 2.9100939480008745e-06, + "loss": 1.4051, + "step": 19181 + }, + { + "epoch": 4.19, + "learning_rate": 2.9099847061393927e-06, + "loss": 1.303, + "step": 19182 + }, + { + "epoch": 4.19, + "learning_rate": 2.9098754642779113e-06, + "loss": 0.2341, + "step": 19183 + }, + { + "epoch": 4.19, + "learning_rate": 2.9097662224164304e-06, + "loss": 1.5643, + "step": 19184 + }, + { + "epoch": 4.19, + "learning_rate": 2.909656980554949e-06, + "loss": 2.8272, + "step": 19185 + }, + { + "epoch": 4.19, + "learning_rate": 2.909547738693468e-06, + "loss": 3.7292, + "step": 19186 + }, + { + "epoch": 4.19, + "learning_rate": 2.909438496831986e-06, + "loss": 2.0058, + "step": 19187 + }, + { + "epoch": 4.19, + "learning_rate": 2.909329254970505e-06, + "loss": 1.3877, + "step": 19188 + }, + { + "epoch": 4.19, + "learning_rate": 2.909220013109024e-06, + "loss": 1.9858, + "step": 19189 + }, + { + "epoch": 4.19, + "learning_rate": 2.9091107712475424e-06, + "loss": 0.6897, + "step": 19190 + }, + { + "epoch": 4.19, + "learning_rate": 2.9090015293860606e-06, + "loss": 1.929, + "step": 19191 + }, + { + "epoch": 4.19, + "learning_rate": 2.9088922875245797e-06, + "loss": 2.127, + "step": 19192 + }, + { + "epoch": 4.19, + "learning_rate": 2.9087830456630983e-06, + "loss": 1.0591, + "step": 19193 + }, + { + "epoch": 4.19, + "learning_rate": 2.9086738038016173e-06, + "loss": 2.102, + "step": 19194 + }, + { + "epoch": 4.19, + "learning_rate": 2.908564561940136e-06, + "loss": 1.8991, + "step": 19195 + }, + { + "epoch": 4.19, + "learning_rate": 2.908455320078654e-06, + "loss": 2.7276, + "step": 19196 + }, + { + "epoch": 4.19, + "learning_rate": 2.908346078217173e-06, + "loss": 2.0397, + "step": 19197 + }, + { + "epoch": 4.19, + "learning_rate": 2.9082368363556918e-06, + "loss": 1.3677, + "step": 19198 + }, + { + "epoch": 4.19, + "learning_rate": 2.9081275944942104e-06, + "loss": 0.965, + "step": 19199 + }, + { + "epoch": 4.19, + "learning_rate": 2.9080183526327294e-06, + "loss": 2.3261, + "step": 19200 + }, + { + "epoch": 4.19, + "learning_rate": 2.9079091107712476e-06, + "loss": 1.3297, + "step": 19201 + }, + { + "epoch": 4.19, + "learning_rate": 2.907799868909766e-06, + "loss": 1.3912, + "step": 19202 + }, + { + "epoch": 4.19, + "learning_rate": 2.9076906270482852e-06, + "loss": 3.0701, + "step": 19203 + }, + { + "epoch": 4.19, + "learning_rate": 2.907581385186804e-06, + "loss": 1.204, + "step": 19204 + }, + { + "epoch": 4.19, + "learning_rate": 2.907472143325322e-06, + "loss": 2.3703, + "step": 19205 + }, + { + "epoch": 4.19, + "learning_rate": 2.907362901463841e-06, + "loss": 2.7027, + "step": 19206 + }, + { + "epoch": 4.19, + "learning_rate": 2.9072536596023597e-06, + "loss": 1.0981, + "step": 19207 + }, + { + "epoch": 4.19, + "learning_rate": 2.9071444177408787e-06, + "loss": 1.1662, + "step": 19208 + }, + { + "epoch": 4.19, + "learning_rate": 2.9070351758793973e-06, + "loss": 1.5327, + "step": 19209 + }, + { + "epoch": 4.19, + "learning_rate": 2.9069259340179155e-06, + "loss": 1.3961, + "step": 19210 + }, + { + "epoch": 4.19, + "learning_rate": 2.9068166921564345e-06, + "loss": 2.8153, + "step": 19211 + }, + { + "epoch": 4.19, + "learning_rate": 2.906707450294953e-06, + "loss": 3.0795, + "step": 19212 + }, + { + "epoch": 4.19, + "learning_rate": 2.906598208433472e-06, + "loss": 0.4903, + "step": 19213 + }, + { + "epoch": 4.19, + "learning_rate": 2.906488966571991e-06, + "loss": 2.8167, + "step": 19214 + }, + { + "epoch": 4.19, + "learning_rate": 2.906379724710509e-06, + "loss": 1.0712, + "step": 19215 + }, + { + "epoch": 4.19, + "learning_rate": 2.906270482849028e-06, + "loss": 2.018, + "step": 19216 + }, + { + "epoch": 4.19, + "learning_rate": 2.9061612409875466e-06, + "loss": 1.7034, + "step": 19217 + }, + { + "epoch": 4.19, + "learning_rate": 2.9060519991260657e-06, + "loss": 0.7839, + "step": 19218 + }, + { + "epoch": 4.19, + "learning_rate": 2.9059427572645843e-06, + "loss": 2.3634, + "step": 19219 + }, + { + "epoch": 4.19, + "learning_rate": 2.9058335154031025e-06, + "loss": 1.2839, + "step": 19220 + }, + { + "epoch": 4.19, + "learning_rate": 2.9057242735416215e-06, + "loss": 3.2386, + "step": 19221 + }, + { + "epoch": 4.2, + "learning_rate": 2.90561503168014e-06, + "loss": 1.291, + "step": 19222 + }, + { + "epoch": 4.2, + "learning_rate": 2.905505789818659e-06, + "loss": 1.9022, + "step": 19223 + }, + { + "epoch": 4.2, + "learning_rate": 2.9053965479571773e-06, + "loss": 1.4735, + "step": 19224 + }, + { + "epoch": 4.2, + "learning_rate": 2.905287306095696e-06, + "loss": 1.6187, + "step": 19225 + }, + { + "epoch": 4.2, + "learning_rate": 2.905178064234215e-06, + "loss": 1.8794, + "step": 19226 + }, + { + "epoch": 4.2, + "learning_rate": 2.9050688223727336e-06, + "loss": 2.1483, + "step": 19227 + }, + { + "epoch": 4.2, + "learning_rate": 2.904959580511252e-06, + "loss": 1.377, + "step": 19228 + }, + { + "epoch": 4.2, + "learning_rate": 2.904850338649771e-06, + "loss": 1.3128, + "step": 19229 + }, + { + "epoch": 4.2, + "learning_rate": 2.9047410967882894e-06, + "loss": 2.0981, + "step": 19230 + }, + { + "epoch": 4.2, + "learning_rate": 2.904631854926808e-06, + "loss": 0.8086, + "step": 19231 + }, + { + "epoch": 4.2, + "learning_rate": 2.904522613065327e-06, + "loss": 2.7083, + "step": 19232 + }, + { + "epoch": 4.2, + "learning_rate": 2.9044133712038457e-06, + "loss": 2.3683, + "step": 19233 + }, + { + "epoch": 4.2, + "learning_rate": 2.904304129342364e-06, + "loss": 1.8201, + "step": 19234 + }, + { + "epoch": 4.2, + "learning_rate": 2.904194887480883e-06, + "loss": 2.7308, + "step": 19235 + }, + { + "epoch": 4.2, + "learning_rate": 2.9040856456194015e-06, + "loss": 1.3557, + "step": 19236 + }, + { + "epoch": 4.2, + "learning_rate": 2.9039764037579205e-06, + "loss": 2.5687, + "step": 19237 + }, + { + "epoch": 4.2, + "learning_rate": 2.9038671618964387e-06, + "loss": 2.1689, + "step": 19238 + }, + { + "epoch": 4.2, + "learning_rate": 2.9037579200349573e-06, + "loss": 2.1495, + "step": 19239 + }, + { + "epoch": 4.2, + "learning_rate": 2.9036486781734764e-06, + "loss": 1.3082, + "step": 19240 + }, + { + "epoch": 4.2, + "learning_rate": 2.903539436311995e-06, + "loss": 1.5826, + "step": 19241 + }, + { + "epoch": 4.2, + "learning_rate": 2.903430194450514e-06, + "loss": 2.013, + "step": 19242 + }, + { + "epoch": 4.2, + "learning_rate": 2.903320952589032e-06, + "loss": 1.1025, + "step": 19243 + }, + { + "epoch": 4.2, + "learning_rate": 2.903211710727551e-06, + "loss": 1.4551, + "step": 19244 + }, + { + "epoch": 4.2, + "learning_rate": 2.90310246886607e-06, + "loss": 0.9939, + "step": 19245 + }, + { + "epoch": 4.2, + "learning_rate": 2.9029932270045885e-06, + "loss": 2.0151, + "step": 19246 + }, + { + "epoch": 4.2, + "learning_rate": 2.9028839851431075e-06, + "loss": 1.6108, + "step": 19247 + }, + { + "epoch": 4.2, + "learning_rate": 2.9027747432816257e-06, + "loss": 2.4456, + "step": 19248 + }, + { + "epoch": 4.2, + "learning_rate": 2.9026655014201443e-06, + "loss": 1.9024, + "step": 19249 + }, + { + "epoch": 4.2, + "learning_rate": 2.9025562595586633e-06, + "loss": 1.4278, + "step": 19250 + }, + { + "epoch": 4.2, + "learning_rate": 2.902447017697182e-06, + "loss": 2.6648, + "step": 19251 + }, + { + "epoch": 4.2, + "learning_rate": 2.9023377758357006e-06, + "loss": 1.5711, + "step": 19252 + }, + { + "epoch": 4.2, + "learning_rate": 2.902228533974219e-06, + "loss": 1.4864, + "step": 19253 + }, + { + "epoch": 4.2, + "learning_rate": 2.9021192921127378e-06, + "loss": 0.6482, + "step": 19254 + }, + { + "epoch": 4.2, + "learning_rate": 2.9020100502512564e-06, + "loss": 2.1967, + "step": 19255 + }, + { + "epoch": 4.2, + "learning_rate": 2.9019008083897754e-06, + "loss": 2.1079, + "step": 19256 + }, + { + "epoch": 4.2, + "learning_rate": 2.9017915665282936e-06, + "loss": 2.2854, + "step": 19257 + }, + { + "epoch": 4.2, + "learning_rate": 2.9016823246668122e-06, + "loss": 2.6134, + "step": 19258 + }, + { + "epoch": 4.2, + "learning_rate": 2.9015730828053313e-06, + "loss": 2.1454, + "step": 19259 + }, + { + "epoch": 4.2, + "learning_rate": 2.90146384094385e-06, + "loss": 2.3095, + "step": 19260 + }, + { + "epoch": 4.2, + "learning_rate": 2.901354599082369e-06, + "loss": 2.2322, + "step": 19261 + }, + { + "epoch": 4.2, + "learning_rate": 2.901245357220887e-06, + "loss": 0.9326, + "step": 19262 + }, + { + "epoch": 4.2, + "learning_rate": 2.9011361153594057e-06, + "loss": 2.77, + "step": 19263 + }, + { + "epoch": 4.2, + "learning_rate": 2.9010268734979247e-06, + "loss": 1.7841, + "step": 19264 + }, + { + "epoch": 4.2, + "learning_rate": 2.9009176316364433e-06, + "loss": 2.7252, + "step": 19265 + }, + { + "epoch": 4.2, + "learning_rate": 2.9008083897749624e-06, + "loss": 0.8989, + "step": 19266 + }, + { + "epoch": 4.2, + "learning_rate": 2.9006991479134806e-06, + "loss": 3.5715, + "step": 19267 + }, + { + "epoch": 4.21, + "learning_rate": 2.900589906051999e-06, + "loss": 1.5312, + "step": 19268 + }, + { + "epoch": 4.21, + "learning_rate": 2.9004806641905182e-06, + "loss": 0.7744, + "step": 19269 + }, + { + "epoch": 4.21, + "learning_rate": 2.900371422329037e-06, + "loss": 2.0432, + "step": 19270 + }, + { + "epoch": 4.21, + "learning_rate": 2.900262180467555e-06, + "loss": 1.406, + "step": 19271 + }, + { + "epoch": 4.21, + "learning_rate": 2.900152938606074e-06, + "loss": 2.1799, + "step": 19272 + }, + { + "epoch": 4.21, + "learning_rate": 2.9000436967445927e-06, + "loss": 2.4321, + "step": 19273 + }, + { + "epoch": 4.21, + "learning_rate": 2.8999344548831117e-06, + "loss": 1.1787, + "step": 19274 + }, + { + "epoch": 4.21, + "learning_rate": 2.8998252130216303e-06, + "loss": 1.8093, + "step": 19275 + }, + { + "epoch": 4.21, + "learning_rate": 2.8997159711601485e-06, + "loss": 2.2416, + "step": 19276 + }, + { + "epoch": 4.21, + "learning_rate": 2.8996067292986675e-06, + "loss": 1.0283, + "step": 19277 + }, + { + "epoch": 4.21, + "learning_rate": 2.899497487437186e-06, + "loss": 1.0265, + "step": 19278 + }, + { + "epoch": 4.21, + "learning_rate": 2.899388245575705e-06, + "loss": 1.2808, + "step": 19279 + }, + { + "epoch": 4.21, + "learning_rate": 2.8992790037142238e-06, + "loss": 2.5754, + "step": 19280 + }, + { + "epoch": 4.21, + "learning_rate": 2.899169761852742e-06, + "loss": 1.7828, + "step": 19281 + }, + { + "epoch": 4.21, + "learning_rate": 2.899060519991261e-06, + "loss": 1.706, + "step": 19282 + }, + { + "epoch": 4.21, + "learning_rate": 2.8989512781297796e-06, + "loss": 1.4424, + "step": 19283 + }, + { + "epoch": 4.21, + "learning_rate": 2.8988420362682982e-06, + "loss": 1.0808, + "step": 19284 + }, + { + "epoch": 4.21, + "learning_rate": 2.898732794406817e-06, + "loss": 1.0982, + "step": 19285 + }, + { + "epoch": 4.21, + "learning_rate": 2.8986235525453354e-06, + "loss": 3.2763, + "step": 19286 + }, + { + "epoch": 4.21, + "learning_rate": 2.898514310683854e-06, + "loss": 3.7984, + "step": 19287 + }, + { + "epoch": 4.21, + "learning_rate": 2.898405068822373e-06, + "loss": 1.298, + "step": 19288 + }, + { + "epoch": 4.21, + "learning_rate": 2.8982958269608917e-06, + "loss": 1.055, + "step": 19289 + }, + { + "epoch": 4.21, + "learning_rate": 2.89818658509941e-06, + "loss": 0.644, + "step": 19290 + }, + { + "epoch": 4.21, + "learning_rate": 2.898077343237929e-06, + "loss": 3.2903, + "step": 19291 + }, + { + "epoch": 4.21, + "learning_rate": 2.8979681013764475e-06, + "loss": 2.2191, + "step": 19292 + }, + { + "epoch": 4.21, + "learning_rate": 2.8978588595149666e-06, + "loss": 2.7516, + "step": 19293 + }, + { + "epoch": 4.21, + "learning_rate": 2.897749617653485e-06, + "loss": 1.605, + "step": 19294 + }, + { + "epoch": 4.21, + "learning_rate": 2.8976403757920034e-06, + "loss": 2.0305, + "step": 19295 + }, + { + "epoch": 4.21, + "learning_rate": 2.8975311339305224e-06, + "loss": 3.0229, + "step": 19296 + }, + { + "epoch": 4.21, + "learning_rate": 2.897421892069041e-06, + "loss": 1.0703, + "step": 19297 + }, + { + "epoch": 4.21, + "learning_rate": 2.89731265020756e-06, + "loss": 0.878, + "step": 19298 + }, + { + "epoch": 4.21, + "learning_rate": 2.8972034083460787e-06, + "loss": 2.3333, + "step": 19299 + }, + { + "epoch": 4.21, + "learning_rate": 2.897094166484597e-06, + "loss": 1.7682, + "step": 19300 + }, + { + "epoch": 4.21, + "learning_rate": 2.896984924623116e-06, + "loss": 2.4906, + "step": 19301 + }, + { + "epoch": 4.21, + "learning_rate": 2.8968756827616345e-06, + "loss": 1.3867, + "step": 19302 + }, + { + "epoch": 4.21, + "learning_rate": 2.8967664409001535e-06, + "loss": 0.5934, + "step": 19303 + }, + { + "epoch": 4.21, + "learning_rate": 2.8966571990386717e-06, + "loss": 2.4099, + "step": 19304 + }, + { + "epoch": 4.21, + "learning_rate": 2.8965479571771903e-06, + "loss": 0.8027, + "step": 19305 + }, + { + "epoch": 4.21, + "learning_rate": 2.8964387153157094e-06, + "loss": 2.6694, + "step": 19306 + }, + { + "epoch": 4.21, + "learning_rate": 2.896329473454228e-06, + "loss": 1.6093, + "step": 19307 + }, + { + "epoch": 4.21, + "learning_rate": 2.8962202315927466e-06, + "loss": 1.7854, + "step": 19308 + }, + { + "epoch": 4.21, + "learning_rate": 2.896110989731265e-06, + "loss": 2.0117, + "step": 19309 + }, + { + "epoch": 4.21, + "learning_rate": 2.896001747869784e-06, + "loss": 1.414, + "step": 19310 + }, + { + "epoch": 4.21, + "learning_rate": 2.895892506008303e-06, + "loss": 3.4348, + "step": 19311 + }, + { + "epoch": 4.21, + "learning_rate": 2.8957832641468215e-06, + "loss": 0.8212, + "step": 19312 + }, + { + "epoch": 4.21, + "learning_rate": 2.89567402228534e-06, + "loss": 1.808, + "step": 19313 + }, + { + "epoch": 4.22, + "learning_rate": 2.8955647804238587e-06, + "loss": 1.9451, + "step": 19314 + }, + { + "epoch": 4.22, + "learning_rate": 2.8954555385623773e-06, + "loss": 1.8544, + "step": 19315 + }, + { + "epoch": 4.22, + "learning_rate": 2.895346296700896e-06, + "loss": 1.4494, + "step": 19316 + }, + { + "epoch": 4.22, + "learning_rate": 2.895237054839415e-06, + "loss": 0.7828, + "step": 19317 + }, + { + "epoch": 4.22, + "learning_rate": 2.895127812977933e-06, + "loss": 1.5771, + "step": 19318 + }, + { + "epoch": 4.22, + "learning_rate": 2.8950185711164517e-06, + "loss": 2.9786, + "step": 19319 + }, + { + "epoch": 4.22, + "learning_rate": 2.8949093292549708e-06, + "loss": 1.5154, + "step": 19320 + }, + { + "epoch": 4.22, + "learning_rate": 2.8948000873934894e-06, + "loss": 1.2519, + "step": 19321 + }, + { + "epoch": 4.22, + "learning_rate": 2.8946908455320084e-06, + "loss": 1.3777, + "step": 19322 + }, + { + "epoch": 4.22, + "learning_rate": 2.8945816036705266e-06, + "loss": 1.8207, + "step": 19323 + }, + { + "epoch": 4.22, + "learning_rate": 2.894472361809045e-06, + "loss": 2.7424, + "step": 19324 + }, + { + "epoch": 4.22, + "learning_rate": 2.8943631199475642e-06, + "loss": 0.9879, + "step": 19325 + }, + { + "epoch": 4.22, + "learning_rate": 2.894253878086083e-06, + "loss": 1.1432, + "step": 19326 + }, + { + "epoch": 4.22, + "learning_rate": 2.894144636224602e-06, + "loss": 0.7712, + "step": 19327 + }, + { + "epoch": 4.22, + "learning_rate": 2.89403539436312e-06, + "loss": 1.6727, + "step": 19328 + }, + { + "epoch": 4.22, + "learning_rate": 2.8939261525016387e-06, + "loss": 0.7878, + "step": 19329 + }, + { + "epoch": 4.22, + "learning_rate": 2.8938169106401577e-06, + "loss": 0.9578, + "step": 19330 + }, + { + "epoch": 4.22, + "learning_rate": 2.8937076687786763e-06, + "loss": 1.4941, + "step": 19331 + }, + { + "epoch": 4.22, + "learning_rate": 2.8935984269171954e-06, + "loss": 1.8251, + "step": 19332 + }, + { + "epoch": 4.22, + "learning_rate": 2.8934891850557136e-06, + "loss": 0.4076, + "step": 19333 + }, + { + "epoch": 4.22, + "learning_rate": 2.893379943194232e-06, + "loss": 1.4392, + "step": 19334 + }, + { + "epoch": 4.22, + "learning_rate": 2.893270701332751e-06, + "loss": 1.3319, + "step": 19335 + }, + { + "epoch": 4.22, + "learning_rate": 2.89316145947127e-06, + "loss": 2.8914, + "step": 19336 + }, + { + "epoch": 4.22, + "learning_rate": 2.893052217609788e-06, + "loss": 1.2213, + "step": 19337 + }, + { + "epoch": 4.22, + "learning_rate": 2.892942975748307e-06, + "loss": 1.4721, + "step": 19338 + }, + { + "epoch": 4.22, + "learning_rate": 2.8928337338868256e-06, + "loss": 2.6151, + "step": 19339 + }, + { + "epoch": 4.22, + "learning_rate": 2.8927244920253443e-06, + "loss": 4.4704, + "step": 19340 + }, + { + "epoch": 4.22, + "learning_rate": 2.8926152501638633e-06, + "loss": 1.362, + "step": 19341 + }, + { + "epoch": 4.22, + "learning_rate": 2.8925060083023815e-06, + "loss": 2.679, + "step": 19342 + }, + { + "epoch": 4.22, + "learning_rate": 2.8923967664409e-06, + "loss": 1.8935, + "step": 19343 + }, + { + "epoch": 4.22, + "learning_rate": 2.892287524579419e-06, + "loss": 1.2789, + "step": 19344 + }, + { + "epoch": 4.22, + "learning_rate": 2.8921782827179377e-06, + "loss": 0.8973, + "step": 19345 + }, + { + "epoch": 4.22, + "learning_rate": 2.8920690408564568e-06, + "loss": 1.0279, + "step": 19346 + }, + { + "epoch": 4.22, + "learning_rate": 2.891959798994975e-06, + "loss": 1.183, + "step": 19347 + }, + { + "epoch": 4.22, + "learning_rate": 2.8918505571334936e-06, + "loss": 3.2187, + "step": 19348 + }, + { + "epoch": 4.22, + "learning_rate": 2.8917413152720126e-06, + "loss": 1.9869, + "step": 19349 + }, + { + "epoch": 4.22, + "learning_rate": 2.891632073410531e-06, + "loss": 1.1679, + "step": 19350 + }, + { + "epoch": 4.22, + "learning_rate": 2.8915228315490494e-06, + "loss": 2.4642, + "step": 19351 + }, + { + "epoch": 4.22, + "learning_rate": 2.8914135896875684e-06, + "loss": 1.6644, + "step": 19352 + }, + { + "epoch": 4.22, + "learning_rate": 2.891304347826087e-06, + "loss": 1.9712, + "step": 19353 + }, + { + "epoch": 4.22, + "learning_rate": 2.891195105964606e-06, + "loss": 2.033, + "step": 19354 + }, + { + "epoch": 4.22, + "learning_rate": 2.8910858641031247e-06, + "loss": 2.2025, + "step": 19355 + }, + { + "epoch": 4.22, + "learning_rate": 2.890976622241643e-06, + "loss": 2.0573, + "step": 19356 + }, + { + "epoch": 4.22, + "learning_rate": 2.890867380380162e-06, + "loss": 2.2865, + "step": 19357 + }, + { + "epoch": 4.22, + "learning_rate": 2.8907581385186805e-06, + "loss": 1.4129, + "step": 19358 + }, + { + "epoch": 4.23, + "learning_rate": 2.8906488966571996e-06, + "loss": 0.7009, + "step": 19359 + }, + { + "epoch": 4.23, + "learning_rate": 2.890539654795718e-06, + "loss": 2.1982, + "step": 19360 + }, + { + "epoch": 4.23, + "learning_rate": 2.8904304129342364e-06, + "loss": 1.3322, + "step": 19361 + }, + { + "epoch": 4.23, + "learning_rate": 2.8903211710727554e-06, + "loss": 2.2618, + "step": 19362 + }, + { + "epoch": 4.23, + "learning_rate": 2.890211929211274e-06, + "loss": 1.6725, + "step": 19363 + }, + { + "epoch": 4.23, + "learning_rate": 2.890102687349793e-06, + "loss": 1.6741, + "step": 19364 + }, + { + "epoch": 4.23, + "learning_rate": 2.8899934454883116e-06, + "loss": 1.9378, + "step": 19365 + }, + { + "epoch": 4.23, + "learning_rate": 2.88988420362683e-06, + "loss": 2.4359, + "step": 19366 + }, + { + "epoch": 4.23, + "learning_rate": 2.889774961765349e-06, + "loss": 1.3913, + "step": 19367 + }, + { + "epoch": 4.23, + "learning_rate": 2.8896657199038675e-06, + "loss": 1.2565, + "step": 19368 + }, + { + "epoch": 4.23, + "learning_rate": 2.889556478042386e-06, + "loss": 0.7615, + "step": 19369 + }, + { + "epoch": 4.23, + "learning_rate": 2.8894472361809047e-06, + "loss": 2.4682, + "step": 19370 + }, + { + "epoch": 4.23, + "learning_rate": 2.8893379943194233e-06, + "loss": 2.1028, + "step": 19371 + }, + { + "epoch": 4.23, + "learning_rate": 2.889228752457942e-06, + "loss": 1.6304, + "step": 19372 + }, + { + "epoch": 4.23, + "learning_rate": 2.889119510596461e-06, + "loss": 1.6256, + "step": 19373 + }, + { + "epoch": 4.23, + "learning_rate": 2.8890102687349796e-06, + "loss": 0.944, + "step": 19374 + }, + { + "epoch": 4.23, + "learning_rate": 2.8889010268734978e-06, + "loss": 2.5507, + "step": 19375 + }, + { + "epoch": 4.23, + "learning_rate": 2.888791785012017e-06, + "loss": 2.1116, + "step": 19376 + }, + { + "epoch": 4.23, + "learning_rate": 2.8886825431505354e-06, + "loss": 2.436, + "step": 19377 + }, + { + "epoch": 4.23, + "learning_rate": 2.8885733012890544e-06, + "loss": 0.8526, + "step": 19378 + }, + { + "epoch": 4.23, + "learning_rate": 2.888464059427573e-06, + "loss": 1.0433, + "step": 19379 + }, + { + "epoch": 4.23, + "learning_rate": 2.8883548175660912e-06, + "loss": 2.2782, + "step": 19380 + }, + { + "epoch": 4.23, + "learning_rate": 2.8882455757046103e-06, + "loss": 1.8237, + "step": 19381 + }, + { + "epoch": 4.23, + "learning_rate": 2.888136333843129e-06, + "loss": 1.0892, + "step": 19382 + }, + { + "epoch": 4.23, + "learning_rate": 2.888027091981648e-06, + "loss": 1.3499, + "step": 19383 + }, + { + "epoch": 4.23, + "learning_rate": 2.887917850120166e-06, + "loss": 1.5679, + "step": 19384 + }, + { + "epoch": 4.23, + "learning_rate": 2.8878086082586847e-06, + "loss": 1.4781, + "step": 19385 + }, + { + "epoch": 4.23, + "learning_rate": 2.8876993663972037e-06, + "loss": 0.6174, + "step": 19386 + }, + { + "epoch": 4.23, + "learning_rate": 2.8875901245357224e-06, + "loss": 2.4396, + "step": 19387 + }, + { + "epoch": 4.23, + "learning_rate": 2.8874808826742414e-06, + "loss": 1.7693, + "step": 19388 + }, + { + "epoch": 4.23, + "learning_rate": 2.8873716408127596e-06, + "loss": 2.2916, + "step": 19389 + }, + { + "epoch": 4.23, + "learning_rate": 2.887262398951278e-06, + "loss": 1.2847, + "step": 19390 + }, + { + "epoch": 4.23, + "learning_rate": 2.8871531570897972e-06, + "loss": 2.428, + "step": 19391 + }, + { + "epoch": 4.23, + "learning_rate": 2.887043915228316e-06, + "loss": 0.698, + "step": 19392 + }, + { + "epoch": 4.23, + "learning_rate": 2.8869346733668344e-06, + "loss": 1.7807, + "step": 19393 + }, + { + "epoch": 4.23, + "learning_rate": 2.886825431505353e-06, + "loss": 2.0497, + "step": 19394 + }, + { + "epoch": 4.23, + "learning_rate": 2.8867161896438717e-06, + "loss": 2.9004, + "step": 19395 + }, + { + "epoch": 4.23, + "learning_rate": 2.8866069477823903e-06, + "loss": 1.8609, + "step": 19396 + }, + { + "epoch": 4.23, + "learning_rate": 2.8864977059209093e-06, + "loss": 1.6566, + "step": 19397 + }, + { + "epoch": 4.23, + "learning_rate": 2.886388464059428e-06, + "loss": 0.8031, + "step": 19398 + }, + { + "epoch": 4.23, + "learning_rate": 2.886279222197946e-06, + "loss": 1.0494, + "step": 19399 + }, + { + "epoch": 4.23, + "learning_rate": 2.886169980336465e-06, + "loss": 0.7751, + "step": 19400 + }, + { + "epoch": 4.23, + "learning_rate": 2.8860607384749838e-06, + "loss": 1.2499, + "step": 19401 + }, + { + "epoch": 4.23, + "learning_rate": 2.885951496613503e-06, + "loss": 1.0343, + "step": 19402 + }, + { + "epoch": 4.23, + "learning_rate": 2.885842254752021e-06, + "loss": 0.3949, + "step": 19403 + }, + { + "epoch": 4.23, + "learning_rate": 2.8857330128905396e-06, + "loss": 0.6774, + "step": 19404 + }, + { + "epoch": 4.24, + "learning_rate": 2.8856237710290586e-06, + "loss": 1.4904, + "step": 19405 + }, + { + "epoch": 4.24, + "learning_rate": 2.8855145291675772e-06, + "loss": 1.1986, + "step": 19406 + }, + { + "epoch": 4.24, + "learning_rate": 2.8854052873060963e-06, + "loss": 1.8333, + "step": 19407 + }, + { + "epoch": 4.24, + "learning_rate": 2.8852960454446145e-06, + "loss": 0.8815, + "step": 19408 + }, + { + "epoch": 4.24, + "learning_rate": 2.885186803583133e-06, + "loss": 2.4666, + "step": 19409 + }, + { + "epoch": 4.24, + "learning_rate": 2.885077561721652e-06, + "loss": 2.0931, + "step": 19410 + }, + { + "epoch": 4.24, + "learning_rate": 2.8849683198601707e-06, + "loss": 0.3896, + "step": 19411 + }, + { + "epoch": 4.24, + "learning_rate": 2.8848590779986897e-06, + "loss": 1.1565, + "step": 19412 + }, + { + "epoch": 4.24, + "learning_rate": 2.884749836137208e-06, + "loss": 1.9613, + "step": 19413 + }, + { + "epoch": 4.24, + "learning_rate": 2.8846405942757265e-06, + "loss": 1.7483, + "step": 19414 + }, + { + "epoch": 4.24, + "learning_rate": 2.8845313524142456e-06, + "loss": 1.6172, + "step": 19415 + }, + { + "epoch": 4.24, + "learning_rate": 2.884422110552764e-06, + "loss": 3.1268, + "step": 19416 + }, + { + "epoch": 4.24, + "learning_rate": 2.8843128686912824e-06, + "loss": 1.2607, + "step": 19417 + }, + { + "epoch": 4.24, + "learning_rate": 2.8842036268298014e-06, + "loss": 2.049, + "step": 19418 + }, + { + "epoch": 4.24, + "learning_rate": 2.88409438496832e-06, + "loss": 2.4544, + "step": 19419 + }, + { + "epoch": 4.24, + "learning_rate": 2.883985143106839e-06, + "loss": 2.0303, + "step": 19420 + }, + { + "epoch": 4.24, + "learning_rate": 2.8838759012453577e-06, + "loss": 1.3029, + "step": 19421 + }, + { + "epoch": 4.24, + "learning_rate": 2.883766659383876e-06, + "loss": 1.981, + "step": 19422 + }, + { + "epoch": 4.24, + "learning_rate": 2.883657417522395e-06, + "loss": 1.3295, + "step": 19423 + }, + { + "epoch": 4.24, + "learning_rate": 2.8835481756609135e-06, + "loss": 0.7707, + "step": 19424 + }, + { + "epoch": 4.24, + "learning_rate": 2.883438933799432e-06, + "loss": 1.898, + "step": 19425 + }, + { + "epoch": 4.24, + "learning_rate": 2.883329691937951e-06, + "loss": 1.9834, + "step": 19426 + }, + { + "epoch": 4.24, + "learning_rate": 2.8832204500764693e-06, + "loss": 0.4639, + "step": 19427 + }, + { + "epoch": 4.24, + "learning_rate": 2.883111208214988e-06, + "loss": 2.7666, + "step": 19428 + }, + { + "epoch": 4.24, + "learning_rate": 2.883001966353507e-06, + "loss": 2.527, + "step": 19429 + }, + { + "epoch": 4.24, + "learning_rate": 2.8828927244920256e-06, + "loss": 0.9854, + "step": 19430 + }, + { + "epoch": 4.24, + "learning_rate": 2.8827834826305446e-06, + "loss": 2.412, + "step": 19431 + }, + { + "epoch": 4.24, + "learning_rate": 2.882674240769063e-06, + "loss": 0.9179, + "step": 19432 + }, + { + "epoch": 4.24, + "learning_rate": 2.8825649989075814e-06, + "loss": 0.7212, + "step": 19433 + }, + { + "epoch": 4.24, + "learning_rate": 2.8824557570461005e-06, + "loss": 2.5123, + "step": 19434 + }, + { + "epoch": 4.24, + "learning_rate": 2.882346515184619e-06, + "loss": 1.4002, + "step": 19435 + }, + { + "epoch": 4.24, + "learning_rate": 2.8822372733231373e-06, + "loss": 1.8292, + "step": 19436 + }, + { + "epoch": 4.24, + "learning_rate": 2.8821280314616563e-06, + "loss": 1.4741, + "step": 19437 + }, + { + "epoch": 4.24, + "learning_rate": 2.882018789600175e-06, + "loss": 1.6732, + "step": 19438 + }, + { + "epoch": 4.24, + "learning_rate": 2.881909547738694e-06, + "loss": 2.1658, + "step": 19439 + }, + { + "epoch": 4.24, + "learning_rate": 2.8818003058772126e-06, + "loss": 1.4966, + "step": 19440 + }, + { + "epoch": 4.24, + "learning_rate": 2.8816910640157307e-06, + "loss": 1.8033, + "step": 19441 + }, + { + "epoch": 4.24, + "learning_rate": 2.8815818221542498e-06, + "loss": 1.7778, + "step": 19442 + }, + { + "epoch": 4.24, + "learning_rate": 2.8814725802927684e-06, + "loss": 2.7661, + "step": 19443 + }, + { + "epoch": 4.24, + "learning_rate": 2.8813633384312874e-06, + "loss": 2.3694, + "step": 19444 + }, + { + "epoch": 4.24, + "learning_rate": 2.881254096569806e-06, + "loss": 1.9345, + "step": 19445 + }, + { + "epoch": 4.24, + "learning_rate": 2.8811448547083242e-06, + "loss": 0.2165, + "step": 19446 + }, + { + "epoch": 4.24, + "learning_rate": 2.8810356128468433e-06, + "loss": 1.4989, + "step": 19447 + }, + { + "epoch": 4.24, + "learning_rate": 2.880926370985362e-06, + "loss": 2.1929, + "step": 19448 + }, + { + "epoch": 4.24, + "learning_rate": 2.8808171291238805e-06, + "loss": 2.0917, + "step": 19449 + }, + { + "epoch": 4.24, + "learning_rate": 2.880707887262399e-06, + "loss": 1.7615, + "step": 19450 + }, + { + "epoch": 4.25, + "learning_rate": 2.8805986454009177e-06, + "loss": 1.2008, + "step": 19451 + }, + { + "epoch": 4.25, + "learning_rate": 2.8804894035394363e-06, + "loss": 1.6278, + "step": 19452 + }, + { + "epoch": 4.25, + "learning_rate": 2.8803801616779553e-06, + "loss": 0.9458, + "step": 19453 + }, + { + "epoch": 4.25, + "learning_rate": 2.880270919816474e-06, + "loss": 2.0289, + "step": 19454 + }, + { + "epoch": 4.25, + "learning_rate": 2.8801616779549926e-06, + "loss": 1.9091, + "step": 19455 + }, + { + "epoch": 4.25, + "learning_rate": 2.880052436093511e-06, + "loss": 2.4055, + "step": 19456 + }, + { + "epoch": 4.25, + "learning_rate": 2.8799431942320298e-06, + "loss": 1.3581, + "step": 19457 + }, + { + "epoch": 4.25, + "learning_rate": 2.879833952370549e-06, + "loss": 0.9526, + "step": 19458 + }, + { + "epoch": 4.25, + "learning_rate": 2.8797247105090674e-06, + "loss": 2.0738, + "step": 19459 + }, + { + "epoch": 4.25, + "learning_rate": 2.8796154686475856e-06, + "loss": 2.5878, + "step": 19460 + }, + { + "epoch": 4.25, + "learning_rate": 2.8795062267861047e-06, + "loss": 1.4511, + "step": 19461 + }, + { + "epoch": 4.25, + "learning_rate": 2.8793969849246233e-06, + "loss": 3.3175, + "step": 19462 + }, + { + "epoch": 4.25, + "learning_rate": 2.8792877430631423e-06, + "loss": 1.9735, + "step": 19463 + }, + { + "epoch": 4.25, + "learning_rate": 2.879178501201661e-06, + "loss": 1.1143, + "step": 19464 + }, + { + "epoch": 4.25, + "learning_rate": 2.879069259340179e-06, + "loss": 2.6706, + "step": 19465 + }, + { + "epoch": 4.25, + "learning_rate": 2.878960017478698e-06, + "loss": 3.6879, + "step": 19466 + }, + { + "epoch": 4.25, + "learning_rate": 2.8788507756172167e-06, + "loss": 1.4768, + "step": 19467 + }, + { + "epoch": 4.25, + "learning_rate": 2.8787415337557358e-06, + "loss": 2.678, + "step": 19468 + }, + { + "epoch": 4.25, + "learning_rate": 2.878632291894254e-06, + "loss": 2.207, + "step": 19469 + }, + { + "epoch": 4.25, + "learning_rate": 2.8785230500327726e-06, + "loss": 2.2071, + "step": 19470 + }, + { + "epoch": 4.25, + "learning_rate": 2.8784138081712916e-06, + "loss": 2.1977, + "step": 19471 + }, + { + "epoch": 4.25, + "learning_rate": 2.8783045663098102e-06, + "loss": 1.3151, + "step": 19472 + }, + { + "epoch": 4.25, + "learning_rate": 2.8781953244483293e-06, + "loss": 1.6901, + "step": 19473 + }, + { + "epoch": 4.25, + "learning_rate": 2.8780860825868474e-06, + "loss": 0.9891, + "step": 19474 + }, + { + "epoch": 4.25, + "learning_rate": 2.877976840725366e-06, + "loss": 1.3833, + "step": 19475 + }, + { + "epoch": 4.25, + "learning_rate": 2.877867598863885e-06, + "loss": 1.6315, + "step": 19476 + }, + { + "epoch": 4.25, + "learning_rate": 2.8777583570024037e-06, + "loss": 2.1711, + "step": 19477 + }, + { + "epoch": 4.25, + "learning_rate": 2.8776491151409223e-06, + "loss": 3.1191, + "step": 19478 + }, + { + "epoch": 4.25, + "learning_rate": 2.877539873279441e-06, + "loss": 1.2672, + "step": 19479 + }, + { + "epoch": 4.25, + "learning_rate": 2.8774306314179595e-06, + "loss": 2.0618, + "step": 19480 + }, + { + "epoch": 4.25, + "learning_rate": 2.877321389556478e-06, + "loss": 0.7482, + "step": 19481 + }, + { + "epoch": 4.25, + "learning_rate": 2.877212147694997e-06, + "loss": 3.0379, + "step": 19482 + }, + { + "epoch": 4.25, + "learning_rate": 2.8771029058335154e-06, + "loss": 0.5368, + "step": 19483 + }, + { + "epoch": 4.25, + "learning_rate": 2.876993663972034e-06, + "loss": 2.1752, + "step": 19484 + }, + { + "epoch": 4.25, + "learning_rate": 2.876884422110553e-06, + "loss": 1.7179, + "step": 19485 + }, + { + "epoch": 4.25, + "learning_rate": 2.8767751802490716e-06, + "loss": 3.004, + "step": 19486 + }, + { + "epoch": 4.25, + "learning_rate": 2.8766659383875907e-06, + "loss": 2.0837, + "step": 19487 + }, + { + "epoch": 4.25, + "learning_rate": 2.876556696526109e-06, + "loss": 1.2672, + "step": 19488 + }, + { + "epoch": 4.25, + "learning_rate": 2.8764474546646275e-06, + "loss": 1.2504, + "step": 19489 + }, + { + "epoch": 4.25, + "learning_rate": 2.8763382128031465e-06, + "loss": 1.8166, + "step": 19490 + }, + { + "epoch": 4.25, + "learning_rate": 2.876228970941665e-06, + "loss": 1.6775, + "step": 19491 + }, + { + "epoch": 4.25, + "learning_rate": 2.876119729080184e-06, + "loss": 0.541, + "step": 19492 + }, + { + "epoch": 4.25, + "learning_rate": 2.8760104872187023e-06, + "loss": 4.3256, + "step": 19493 + }, + { + "epoch": 4.25, + "learning_rate": 2.875901245357221e-06, + "loss": 1.1319, + "step": 19494 + }, + { + "epoch": 4.25, + "learning_rate": 2.87579200349574e-06, + "loss": 0.4612, + "step": 19495 + }, + { + "epoch": 4.25, + "learning_rate": 2.8756827616342586e-06, + "loss": 1.71, + "step": 19496 + }, + { + "epoch": 4.26, + "learning_rate": 2.8755735197727776e-06, + "loss": 1.2267, + "step": 19497 + }, + { + "epoch": 4.26, + "learning_rate": 2.875464277911296e-06, + "loss": 0.4836, + "step": 19498 + }, + { + "epoch": 4.26, + "learning_rate": 2.8753550360498144e-06, + "loss": 1.5483, + "step": 19499 + }, + { + "epoch": 4.26, + "learning_rate": 2.8752457941883334e-06, + "loss": 1.1035, + "step": 19500 + }, + { + "epoch": 4.26, + "learning_rate": 2.875136552326852e-06, + "loss": 1.5912, + "step": 19501 + }, + { + "epoch": 4.26, + "learning_rate": 2.8750273104653702e-06, + "loss": 1.897, + "step": 19502 + }, + { + "epoch": 4.26, + "learning_rate": 2.8749180686038893e-06, + "loss": 2.466, + "step": 19503 + }, + { + "epoch": 4.26, + "learning_rate": 2.874808826742408e-06, + "loss": 3.3691, + "step": 19504 + }, + { + "epoch": 4.26, + "learning_rate": 2.8746995848809265e-06, + "loss": 1.3794, + "step": 19505 + }, + { + "epoch": 4.26, + "learning_rate": 2.8745903430194455e-06, + "loss": 2.5028, + "step": 19506 + }, + { + "epoch": 4.26, + "learning_rate": 2.8744811011579637e-06, + "loss": 2.2609, + "step": 19507 + }, + { + "epoch": 4.26, + "learning_rate": 2.8743718592964828e-06, + "loss": 2.2323, + "step": 19508 + }, + { + "epoch": 4.26, + "learning_rate": 2.8742626174350014e-06, + "loss": 3.0448, + "step": 19509 + }, + { + "epoch": 4.26, + "learning_rate": 2.87415337557352e-06, + "loss": 2.7485, + "step": 19510 + }, + { + "epoch": 4.26, + "learning_rate": 2.874044133712039e-06, + "loss": 1.4717, + "step": 19511 + }, + { + "epoch": 4.26, + "learning_rate": 2.873934891850557e-06, + "loss": 2.5161, + "step": 19512 + }, + { + "epoch": 4.26, + "learning_rate": 2.873825649989076e-06, + "loss": 2.9221, + "step": 19513 + }, + { + "epoch": 4.26, + "learning_rate": 2.873716408127595e-06, + "loss": 1.0414, + "step": 19514 + }, + { + "epoch": 4.26, + "learning_rate": 2.8736071662661135e-06, + "loss": 1.9821, + "step": 19515 + }, + { + "epoch": 4.26, + "learning_rate": 2.8734979244046316e-06, + "loss": 2.3224, + "step": 19516 + }, + { + "epoch": 4.26, + "learning_rate": 2.8733886825431507e-06, + "loss": 1.045, + "step": 19517 + }, + { + "epoch": 4.26, + "learning_rate": 2.8732794406816693e-06, + "loss": 2.8719, + "step": 19518 + }, + { + "epoch": 4.26, + "learning_rate": 2.8731701988201883e-06, + "loss": 1.5753, + "step": 19519 + }, + { + "epoch": 4.26, + "learning_rate": 2.873060956958707e-06, + "loss": 2.9548, + "step": 19520 + }, + { + "epoch": 4.26, + "learning_rate": 2.872951715097225e-06, + "loss": 2.7794, + "step": 19521 + }, + { + "epoch": 4.26, + "learning_rate": 2.872842473235744e-06, + "loss": 1.9231, + "step": 19522 + }, + { + "epoch": 4.26, + "learning_rate": 2.8727332313742628e-06, + "loss": 1.2856, + "step": 19523 + }, + { + "epoch": 4.26, + "learning_rate": 2.872623989512782e-06, + "loss": 0.4386, + "step": 19524 + }, + { + "epoch": 4.26, + "learning_rate": 2.8725147476513004e-06, + "loss": 1.6359, + "step": 19525 + }, + { + "epoch": 4.26, + "learning_rate": 2.8724055057898186e-06, + "loss": 0.9926, + "step": 19526 + }, + { + "epoch": 4.26, + "learning_rate": 2.8722962639283376e-06, + "loss": 2.7373, + "step": 19527 + }, + { + "epoch": 4.26, + "learning_rate": 2.8721870220668562e-06, + "loss": 1.0212, + "step": 19528 + }, + { + "epoch": 4.26, + "learning_rate": 2.8720777802053753e-06, + "loss": 3.2729, + "step": 19529 + }, + { + "epoch": 4.26, + "learning_rate": 2.871968538343894e-06, + "loss": 1.4049, + "step": 19530 + }, + { + "epoch": 4.26, + "learning_rate": 2.871859296482412e-06, + "loss": 2.7403, + "step": 19531 + }, + { + "epoch": 4.26, + "learning_rate": 2.871750054620931e-06, + "loss": 1.6139, + "step": 19532 + }, + { + "epoch": 4.26, + "learning_rate": 2.8716408127594497e-06, + "loss": 1.9592, + "step": 19533 + }, + { + "epoch": 4.26, + "learning_rate": 2.8715315708979683e-06, + "loss": 0.6417, + "step": 19534 + }, + { + "epoch": 4.26, + "learning_rate": 2.871422329036487e-06, + "loss": 1.4275, + "step": 19535 + }, + { + "epoch": 4.26, + "learning_rate": 2.8713130871750056e-06, + "loss": 1.6114, + "step": 19536 + }, + { + "epoch": 4.26, + "learning_rate": 2.871203845313524e-06, + "loss": 2.1957, + "step": 19537 + }, + { + "epoch": 4.26, + "learning_rate": 2.871094603452043e-06, + "loss": 1.4748, + "step": 19538 + }, + { + "epoch": 4.26, + "learning_rate": 2.870985361590562e-06, + "loss": 1.1996, + "step": 19539 + }, + { + "epoch": 4.26, + "learning_rate": 2.87087611972908e-06, + "loss": 1.9787, + "step": 19540 + }, + { + "epoch": 4.26, + "learning_rate": 2.870766877867599e-06, + "loss": 0.9119, + "step": 19541 + }, + { + "epoch": 4.26, + "learning_rate": 2.8706576360061176e-06, + "loss": 2.0073, + "step": 19542 + }, + { + "epoch": 4.27, + "learning_rate": 2.8705483941446367e-06, + "loss": 2.5516, + "step": 19543 + }, + { + "epoch": 4.27, + "learning_rate": 2.8704391522831553e-06, + "loss": 2.1768, + "step": 19544 + }, + { + "epoch": 4.27, + "learning_rate": 2.8703299104216735e-06, + "loss": 1.3581, + "step": 19545 + }, + { + "epoch": 4.27, + "learning_rate": 2.8702206685601925e-06, + "loss": 1.9418, + "step": 19546 + }, + { + "epoch": 4.27, + "learning_rate": 2.870111426698711e-06, + "loss": 2.7457, + "step": 19547 + }, + { + "epoch": 4.27, + "learning_rate": 2.87000218483723e-06, + "loss": 2.5572, + "step": 19548 + }, + { + "epoch": 4.27, + "learning_rate": 2.8698929429757483e-06, + "loss": 2.2484, + "step": 19549 + }, + { + "epoch": 4.27, + "learning_rate": 2.869783701114267e-06, + "loss": 0.8045, + "step": 19550 + }, + { + "epoch": 4.27, + "learning_rate": 2.869674459252786e-06, + "loss": 2.6168, + "step": 19551 + }, + { + "epoch": 4.27, + "learning_rate": 2.8695652173913046e-06, + "loss": 2.1922, + "step": 19552 + }, + { + "epoch": 4.27, + "learning_rate": 2.8694559755298236e-06, + "loss": 1.5826, + "step": 19553 + }, + { + "epoch": 4.27, + "learning_rate": 2.869346733668342e-06, + "loss": 1.1049, + "step": 19554 + }, + { + "epoch": 4.27, + "learning_rate": 2.8692374918068604e-06, + "loss": 2.6662, + "step": 19555 + }, + { + "epoch": 4.27, + "learning_rate": 2.8691282499453795e-06, + "loss": 2.521, + "step": 19556 + }, + { + "epoch": 4.27, + "learning_rate": 2.869019008083898e-06, + "loss": 2.2359, + "step": 19557 + }, + { + "epoch": 4.27, + "learning_rate": 2.8689097662224167e-06, + "loss": 1.6127, + "step": 19558 + }, + { + "epoch": 4.27, + "learning_rate": 2.8688005243609353e-06, + "loss": 1.2526, + "step": 19559 + }, + { + "epoch": 4.27, + "learning_rate": 2.868691282499454e-06, + "loss": 1.1271, + "step": 19560 + }, + { + "epoch": 4.27, + "learning_rate": 2.868582040637973e-06, + "loss": 1.6145, + "step": 19561 + }, + { + "epoch": 4.27, + "learning_rate": 2.8684727987764916e-06, + "loss": 1.793, + "step": 19562 + }, + { + "epoch": 4.27, + "learning_rate": 2.86836355691501e-06, + "loss": 1.0075, + "step": 19563 + }, + { + "epoch": 4.27, + "learning_rate": 2.8682543150535288e-06, + "loss": 1.0823, + "step": 19564 + }, + { + "epoch": 4.27, + "learning_rate": 2.8681450731920474e-06, + "loss": 3.239, + "step": 19565 + }, + { + "epoch": 4.27, + "learning_rate": 2.868035831330566e-06, + "loss": 1.2852, + "step": 19566 + }, + { + "epoch": 4.27, + "learning_rate": 2.867926589469085e-06, + "loss": 0.7895, + "step": 19567 + }, + { + "epoch": 4.27, + "learning_rate": 2.8678173476076032e-06, + "loss": 2.3555, + "step": 19568 + }, + { + "epoch": 4.27, + "learning_rate": 2.867708105746122e-06, + "loss": 2.5503, + "step": 19569 + }, + { + "epoch": 4.27, + "learning_rate": 2.867598863884641e-06, + "loss": 1.9767, + "step": 19570 + }, + { + "epoch": 4.27, + "learning_rate": 2.8674896220231595e-06, + "loss": 1.2711, + "step": 19571 + }, + { + "epoch": 4.27, + "learning_rate": 2.8673803801616785e-06, + "loss": 1.9012, + "step": 19572 + }, + { + "epoch": 4.27, + "learning_rate": 2.8672711383001967e-06, + "loss": 1.443, + "step": 19573 + }, + { + "epoch": 4.27, + "learning_rate": 2.8671618964387153e-06, + "loss": 1.0063, + "step": 19574 + }, + { + "epoch": 4.27, + "learning_rate": 2.8670526545772344e-06, + "loss": 3.2048, + "step": 19575 + }, + { + "epoch": 4.27, + "learning_rate": 2.866943412715753e-06, + "loss": 1.9561, + "step": 19576 + }, + { + "epoch": 4.27, + "learning_rate": 2.866834170854272e-06, + "loss": 0.8267, + "step": 19577 + }, + { + "epoch": 4.27, + "learning_rate": 2.86672492899279e-06, + "loss": 1.765, + "step": 19578 + }, + { + "epoch": 4.27, + "learning_rate": 2.866615687131309e-06, + "loss": 2.9657, + "step": 19579 + }, + { + "epoch": 4.27, + "learning_rate": 2.866506445269828e-06, + "loss": 2.158, + "step": 19580 + }, + { + "epoch": 4.27, + "learning_rate": 2.8663972034083464e-06, + "loss": 1.8904, + "step": 19581 + }, + { + "epoch": 4.27, + "learning_rate": 2.8662879615468646e-06, + "loss": 1.8298, + "step": 19582 + }, + { + "epoch": 4.27, + "learning_rate": 2.8661787196853837e-06, + "loss": 1.6554, + "step": 19583 + }, + { + "epoch": 4.27, + "learning_rate": 2.8660694778239023e-06, + "loss": 1.6494, + "step": 19584 + }, + { + "epoch": 4.27, + "learning_rate": 2.8659602359624213e-06, + "loss": 1.2398, + "step": 19585 + }, + { + "epoch": 4.27, + "learning_rate": 2.86585099410094e-06, + "loss": 1.8432, + "step": 19586 + }, + { + "epoch": 4.27, + "learning_rate": 2.865741752239458e-06, + "loss": 2.3131, + "step": 19587 + }, + { + "epoch": 4.27, + "learning_rate": 2.865632510377977e-06, + "loss": 2.3465, + "step": 19588 + }, + { + "epoch": 4.28, + "learning_rate": 2.8655232685164958e-06, + "loss": 0.7066, + "step": 19589 + }, + { + "epoch": 4.28, + "learning_rate": 2.8654140266550144e-06, + "loss": 1.3941, + "step": 19590 + }, + { + "epoch": 4.28, + "learning_rate": 2.8653047847935334e-06, + "loss": 2.3652, + "step": 19591 + }, + { + "epoch": 4.28, + "learning_rate": 2.8651955429320516e-06, + "loss": 0.7424, + "step": 19592 + }, + { + "epoch": 4.28, + "learning_rate": 2.86508630107057e-06, + "loss": 2.0283, + "step": 19593 + }, + { + "epoch": 4.28, + "learning_rate": 2.8649770592090892e-06, + "loss": 0.8995, + "step": 19594 + }, + { + "epoch": 4.28, + "learning_rate": 2.864867817347608e-06, + "loss": 2.1381, + "step": 19595 + }, + { + "epoch": 4.28, + "learning_rate": 2.864758575486127e-06, + "loss": 2.2282, + "step": 19596 + }, + { + "epoch": 4.28, + "learning_rate": 2.864649333624645e-06, + "loss": 0.8108, + "step": 19597 + }, + { + "epoch": 4.28, + "learning_rate": 2.8645400917631637e-06, + "loss": 2.4297, + "step": 19598 + }, + { + "epoch": 4.28, + "learning_rate": 2.8644308499016827e-06, + "loss": 2.3207, + "step": 19599 + }, + { + "epoch": 4.28, + "learning_rate": 2.8643216080402013e-06, + "loss": 0.8101, + "step": 19600 + }, + { + "epoch": 4.28, + "learning_rate": 2.8642123661787195e-06, + "loss": 0.8222, + "step": 19601 + }, + { + "epoch": 4.28, + "learning_rate": 2.8641031243172385e-06, + "loss": 1.9076, + "step": 19602 + }, + { + "epoch": 4.28, + "learning_rate": 2.863993882455757e-06, + "loss": 3.0575, + "step": 19603 + }, + { + "epoch": 4.28, + "learning_rate": 2.863884640594276e-06, + "loss": 1.6289, + "step": 19604 + }, + { + "epoch": 4.28, + "learning_rate": 2.863775398732795e-06, + "loss": 0.6866, + "step": 19605 + }, + { + "epoch": 4.28, + "learning_rate": 2.863666156871313e-06, + "loss": 1.0653, + "step": 19606 + }, + { + "epoch": 4.28, + "learning_rate": 2.863556915009832e-06, + "loss": 1.5263, + "step": 19607 + }, + { + "epoch": 4.28, + "learning_rate": 2.8634476731483506e-06, + "loss": 1.7237, + "step": 19608 + }, + { + "epoch": 4.28, + "learning_rate": 2.8633384312868697e-06, + "loss": 1.2093, + "step": 19609 + }, + { + "epoch": 4.28, + "learning_rate": 2.8632291894253883e-06, + "loss": 2.5312, + "step": 19610 + }, + { + "epoch": 4.28, + "learning_rate": 2.8631199475639065e-06, + "loss": 0.9646, + "step": 19611 + }, + { + "epoch": 4.28, + "learning_rate": 2.8630107057024255e-06, + "loss": 1.5244, + "step": 19612 + }, + { + "epoch": 4.28, + "learning_rate": 2.862901463840944e-06, + "loss": 1.2397, + "step": 19613 + }, + { + "epoch": 4.28, + "learning_rate": 2.862792221979463e-06, + "loss": 1.7414, + "step": 19614 + }, + { + "epoch": 4.28, + "learning_rate": 2.8626829801179813e-06, + "loss": 1.8953, + "step": 19615 + }, + { + "epoch": 4.28, + "learning_rate": 2.8625737382565e-06, + "loss": 2.5377, + "step": 19616 + }, + { + "epoch": 4.28, + "learning_rate": 2.862464496395019e-06, + "loss": 1.9399, + "step": 19617 + }, + { + "epoch": 4.28, + "learning_rate": 2.8623552545335376e-06, + "loss": 2.4958, + "step": 19618 + }, + { + "epoch": 4.28, + "learning_rate": 2.862246012672056e-06, + "loss": 0.9538, + "step": 19619 + }, + { + "epoch": 4.28, + "learning_rate": 2.862136770810575e-06, + "loss": 1.6222, + "step": 19620 + }, + { + "epoch": 4.28, + "learning_rate": 2.8620275289490934e-06, + "loss": 1.1119, + "step": 19621 + }, + { + "epoch": 4.28, + "learning_rate": 2.861918287087612e-06, + "loss": 0.6859, + "step": 19622 + }, + { + "epoch": 4.28, + "learning_rate": 2.861809045226131e-06, + "loss": 3.0562, + "step": 19623 + }, + { + "epoch": 4.28, + "learning_rate": 2.8616998033646497e-06, + "loss": 1.6208, + "step": 19624 + }, + { + "epoch": 4.28, + "learning_rate": 2.861590561503168e-06, + "loss": 2.21, + "step": 19625 + }, + { + "epoch": 4.28, + "learning_rate": 2.861481319641687e-06, + "loss": 2.9161, + "step": 19626 + }, + { + "epoch": 4.28, + "learning_rate": 2.8613720777802055e-06, + "loss": 1.2471, + "step": 19627 + }, + { + "epoch": 4.28, + "learning_rate": 2.8612628359187245e-06, + "loss": 2.2686, + "step": 19628 + }, + { + "epoch": 4.28, + "learning_rate": 2.861153594057243e-06, + "loss": 2.3659, + "step": 19629 + }, + { + "epoch": 4.28, + "learning_rate": 2.8610443521957613e-06, + "loss": 2.8136, + "step": 19630 + }, + { + "epoch": 4.28, + "learning_rate": 2.8609351103342804e-06, + "loss": 1.2331, + "step": 19631 + }, + { + "epoch": 4.28, + "learning_rate": 2.860825868472799e-06, + "loss": 3.1207, + "step": 19632 + }, + { + "epoch": 4.28, + "learning_rate": 2.860716626611318e-06, + "loss": 0.5625, + "step": 19633 + }, + { + "epoch": 4.29, + "learning_rate": 2.860607384749836e-06, + "loss": 2.7317, + "step": 19634 + }, + { + "epoch": 4.29, + "learning_rate": 2.860498142888355e-06, + "loss": 2.4461, + "step": 19635 + }, + { + "epoch": 4.29, + "learning_rate": 2.860388901026874e-06, + "loss": 1.6127, + "step": 19636 + }, + { + "epoch": 4.29, + "learning_rate": 2.8602796591653925e-06, + "loss": 0.9383, + "step": 19637 + }, + { + "epoch": 4.29, + "learning_rate": 2.8601704173039115e-06, + "loss": 1.3072, + "step": 19638 + }, + { + "epoch": 4.29, + "learning_rate": 2.8600611754424297e-06, + "loss": 2.6261, + "step": 19639 + }, + { + "epoch": 4.29, + "learning_rate": 2.8599519335809483e-06, + "loss": 2.749, + "step": 19640 + }, + { + "epoch": 4.29, + "learning_rate": 2.8598426917194673e-06, + "loss": 2.5313, + "step": 19641 + }, + { + "epoch": 4.29, + "learning_rate": 2.859733449857986e-06, + "loss": 1.6556, + "step": 19642 + }, + { + "epoch": 4.29, + "learning_rate": 2.8596242079965046e-06, + "loss": 1.6756, + "step": 19643 + }, + { + "epoch": 4.29, + "learning_rate": 2.859514966135023e-06, + "loss": 1.9193, + "step": 19644 + }, + { + "epoch": 4.29, + "learning_rate": 2.8594057242735418e-06, + "loss": 0.9508, + "step": 19645 + }, + { + "epoch": 4.29, + "learning_rate": 2.8592964824120604e-06, + "loss": 0.7327, + "step": 19646 + }, + { + "epoch": 4.29, + "learning_rate": 2.8591872405505794e-06, + "loss": 2.2377, + "step": 19647 + }, + { + "epoch": 4.29, + "learning_rate": 2.8590779986890976e-06, + "loss": 1.7286, + "step": 19648 + }, + { + "epoch": 4.29, + "learning_rate": 2.8589687568276162e-06, + "loss": 1.2041, + "step": 19649 + }, + { + "epoch": 4.29, + "learning_rate": 2.8588595149661353e-06, + "loss": 1.2915, + "step": 19650 + }, + { + "epoch": 4.29, + "learning_rate": 2.858750273104654e-06, + "loss": 1.3828, + "step": 19651 + }, + { + "epoch": 4.29, + "learning_rate": 2.858641031243173e-06, + "loss": 2.1626, + "step": 19652 + }, + { + "epoch": 4.29, + "learning_rate": 2.858531789381691e-06, + "loss": 1.6698, + "step": 19653 + }, + { + "epoch": 4.29, + "learning_rate": 2.8584225475202097e-06, + "loss": 2.795, + "step": 19654 + }, + { + "epoch": 4.29, + "learning_rate": 2.8583133056587287e-06, + "loss": 2.2347, + "step": 19655 + }, + { + "epoch": 4.29, + "learning_rate": 2.8582040637972473e-06, + "loss": 1.4348, + "step": 19656 + }, + { + "epoch": 4.29, + "learning_rate": 2.8580948219357664e-06, + "loss": 0.6511, + "step": 19657 + }, + { + "epoch": 4.29, + "learning_rate": 2.8579855800742846e-06, + "loss": 0.9312, + "step": 19658 + }, + { + "epoch": 4.29, + "learning_rate": 2.857876338212803e-06, + "loss": 0.7884, + "step": 19659 + }, + { + "epoch": 4.29, + "learning_rate": 2.8577670963513222e-06, + "loss": 2.7999, + "step": 19660 + }, + { + "epoch": 4.29, + "learning_rate": 2.857657854489841e-06, + "loss": 1.4898, + "step": 19661 + }, + { + "epoch": 4.29, + "learning_rate": 2.85754861262836e-06, + "loss": 2.4176, + "step": 19662 + }, + { + "epoch": 4.29, + "learning_rate": 2.857439370766878e-06, + "loss": 1.4217, + "step": 19663 + }, + { + "epoch": 4.29, + "learning_rate": 2.8573301289053967e-06, + "loss": 0.6048, + "step": 19664 + }, + { + "epoch": 4.29, + "learning_rate": 2.8572208870439157e-06, + "loss": 0.7297, + "step": 19665 + }, + { + "epoch": 4.29, + "learning_rate": 2.8571116451824343e-06, + "loss": 3.6929, + "step": 19666 + }, + { + "epoch": 4.29, + "learning_rate": 2.8570024033209525e-06, + "loss": 1.6872, + "step": 19667 + }, + { + "epoch": 4.29, + "learning_rate": 2.8568931614594715e-06, + "loss": 2.9073, + "step": 19668 + }, + { + "epoch": 4.29, + "learning_rate": 2.85678391959799e-06, + "loss": 1.1809, + "step": 19669 + }, + { + "epoch": 4.29, + "learning_rate": 2.856674677736509e-06, + "loss": 1.5505, + "step": 19670 + }, + { + "epoch": 4.29, + "learning_rate": 2.8565654358750278e-06, + "loss": 1.1666, + "step": 19671 + }, + { + "epoch": 4.29, + "learning_rate": 2.856456194013546e-06, + "loss": 2.1079, + "step": 19672 + }, + { + "epoch": 4.29, + "learning_rate": 2.856346952152065e-06, + "loss": 1.5738, + "step": 19673 + }, + { + "epoch": 4.29, + "learning_rate": 2.8562377102905836e-06, + "loss": 0.8767, + "step": 19674 + }, + { + "epoch": 4.29, + "learning_rate": 2.8561284684291022e-06, + "loss": 1.9723, + "step": 19675 + }, + { + "epoch": 4.29, + "learning_rate": 2.8560192265676213e-06, + "loss": 2.2563, + "step": 19676 + }, + { + "epoch": 4.29, + "learning_rate": 2.8559099847061394e-06, + "loss": 0.5979, + "step": 19677 + }, + { + "epoch": 4.29, + "learning_rate": 2.855800742844658e-06, + "loss": 2.9502, + "step": 19678 + }, + { + "epoch": 4.29, + "learning_rate": 2.855691500983177e-06, + "loss": 1.2581, + "step": 19679 + }, + { + "epoch": 4.3, + "learning_rate": 2.8555822591216957e-06, + "loss": 1.6777, + "step": 19680 + }, + { + "epoch": 4.3, + "learning_rate": 2.855473017260214e-06, + "loss": 2.0648, + "step": 19681 + }, + { + "epoch": 4.3, + "learning_rate": 2.855363775398733e-06, + "loss": 1.3592, + "step": 19682 + }, + { + "epoch": 4.3, + "learning_rate": 2.8552545335372515e-06, + "loss": 1.0741, + "step": 19683 + }, + { + "epoch": 4.3, + "learning_rate": 2.8551452916757706e-06, + "loss": 0.6678, + "step": 19684 + }, + { + "epoch": 4.3, + "learning_rate": 2.855036049814289e-06, + "loss": 2.1278, + "step": 19685 + }, + { + "epoch": 4.3, + "learning_rate": 2.8549268079528074e-06, + "loss": 0.7512, + "step": 19686 + }, + { + "epoch": 4.3, + "learning_rate": 2.8548175660913264e-06, + "loss": 0.8434, + "step": 19687 + }, + { + "epoch": 4.3, + "learning_rate": 2.854708324229845e-06, + "loss": 2.1172, + "step": 19688 + }, + { + "epoch": 4.3, + "learning_rate": 2.854599082368364e-06, + "loss": 1.7415, + "step": 19689 + }, + { + "epoch": 4.3, + "learning_rate": 2.8544898405068827e-06, + "loss": 0.5026, + "step": 19690 + }, + { + "epoch": 4.3, + "learning_rate": 2.854380598645401e-06, + "loss": 2.6279, + "step": 19691 + }, + { + "epoch": 4.3, + "learning_rate": 2.85427135678392e-06, + "loss": 1.6352, + "step": 19692 + }, + { + "epoch": 4.3, + "learning_rate": 2.8541621149224385e-06, + "loss": 2.4434, + "step": 19693 + }, + { + "epoch": 4.3, + "learning_rate": 2.8540528730609575e-06, + "loss": 2.7512, + "step": 19694 + }, + { + "epoch": 4.3, + "learning_rate": 2.853943631199476e-06, + "loss": 1.0426, + "step": 19695 + }, + { + "epoch": 4.3, + "learning_rate": 2.8538343893379943e-06, + "loss": 2.4826, + "step": 19696 + }, + { + "epoch": 4.3, + "learning_rate": 2.8537251474765134e-06, + "loss": 0.7332, + "step": 19697 + }, + { + "epoch": 4.3, + "learning_rate": 2.853615905615032e-06, + "loss": 2.8315, + "step": 19698 + }, + { + "epoch": 4.3, + "learning_rate": 2.8535066637535506e-06, + "loss": 1.9087, + "step": 19699 + }, + { + "epoch": 4.3, + "learning_rate": 2.853397421892069e-06, + "loss": 2.2177, + "step": 19700 + }, + { + "epoch": 4.3, + "learning_rate": 2.853288180030588e-06, + "loss": 1.6326, + "step": 19701 + }, + { + "epoch": 4.3, + "learning_rate": 2.8531789381691064e-06, + "loss": 2.0003, + "step": 19702 + }, + { + "epoch": 4.3, + "learning_rate": 2.8530696963076255e-06, + "loss": 1.6627, + "step": 19703 + }, + { + "epoch": 4.3, + "learning_rate": 2.852960454446144e-06, + "loss": 0.7358, + "step": 19704 + }, + { + "epoch": 4.3, + "learning_rate": 2.8528512125846627e-06, + "loss": 1.269, + "step": 19705 + }, + { + "epoch": 4.3, + "learning_rate": 2.8527419707231813e-06, + "loss": 2.5026, + "step": 19706 + }, + { + "epoch": 4.3, + "learning_rate": 2.8526327288617e-06, + "loss": 1.441, + "step": 19707 + }, + { + "epoch": 4.3, + "learning_rate": 2.852523487000219e-06, + "loss": 2.1225, + "step": 19708 + }, + { + "epoch": 4.3, + "learning_rate": 2.8524142451387375e-06, + "loss": 1.3581, + "step": 19709 + }, + { + "epoch": 4.3, + "learning_rate": 2.8523050032772557e-06, + "loss": 3.3055, + "step": 19710 + }, + { + "epoch": 4.3, + "learning_rate": 2.8521957614157748e-06, + "loss": 1.5973, + "step": 19711 + }, + { + "epoch": 4.3, + "learning_rate": 2.8520865195542934e-06, + "loss": 1.1564, + "step": 19712 + }, + { + "epoch": 4.3, + "learning_rate": 2.8519772776928124e-06, + "loss": 1.3962, + "step": 19713 + }, + { + "epoch": 4.3, + "learning_rate": 2.8518680358313306e-06, + "loss": 2.1751, + "step": 19714 + }, + { + "epoch": 4.3, + "learning_rate": 2.851758793969849e-06, + "loss": 1.2864, + "step": 19715 + }, + { + "epoch": 4.3, + "learning_rate": 2.8516495521083682e-06, + "loss": 1.3621, + "step": 19716 + }, + { + "epoch": 4.3, + "learning_rate": 2.851540310246887e-06, + "loss": 1.1258, + "step": 19717 + }, + { + "epoch": 4.3, + "learning_rate": 2.851431068385406e-06, + "loss": 2.3784, + "step": 19718 + }, + { + "epoch": 4.3, + "learning_rate": 2.851321826523924e-06, + "loss": 2.0059, + "step": 19719 + }, + { + "epoch": 4.3, + "learning_rate": 2.8512125846624427e-06, + "loss": 1.5679, + "step": 19720 + }, + { + "epoch": 4.3, + "learning_rate": 2.8511033428009617e-06, + "loss": 0.7675, + "step": 19721 + }, + { + "epoch": 4.3, + "learning_rate": 2.8509941009394803e-06, + "loss": 1.2454, + "step": 19722 + }, + { + "epoch": 4.3, + "learning_rate": 2.8508848590779994e-06, + "loss": 1.1535, + "step": 19723 + }, + { + "epoch": 4.3, + "learning_rate": 2.8507756172165176e-06, + "loss": 1.2909, + "step": 19724 + }, + { + "epoch": 4.3, + "learning_rate": 2.850666375355036e-06, + "loss": 1.7175, + "step": 19725 + }, + { + "epoch": 4.31, + "learning_rate": 2.850557133493555e-06, + "loss": 1.7529, + "step": 19726 + }, + { + "epoch": 4.31, + "learning_rate": 2.850447891632074e-06, + "loss": 1.5444, + "step": 19727 + }, + { + "epoch": 4.31, + "learning_rate": 2.850338649770592e-06, + "loss": 2.865, + "step": 19728 + }, + { + "epoch": 4.31, + "learning_rate": 2.850229407909111e-06, + "loss": 2.1714, + "step": 19729 + }, + { + "epoch": 4.31, + "learning_rate": 2.8501201660476296e-06, + "loss": 1.0573, + "step": 19730 + }, + { + "epoch": 4.31, + "learning_rate": 2.8500109241861483e-06, + "loss": 1.4879, + "step": 19731 + }, + { + "epoch": 4.31, + "learning_rate": 2.8499016823246673e-06, + "loss": 2.6148, + "step": 19732 + }, + { + "epoch": 4.31, + "learning_rate": 2.8497924404631855e-06, + "loss": 1.8871, + "step": 19733 + }, + { + "epoch": 4.31, + "learning_rate": 2.849683198601704e-06, + "loss": 1.1059, + "step": 19734 + }, + { + "epoch": 4.31, + "learning_rate": 2.849573956740223e-06, + "loss": 0.8149, + "step": 19735 + }, + { + "epoch": 4.31, + "learning_rate": 2.8494647148787417e-06, + "loss": 1.3058, + "step": 19736 + }, + { + "epoch": 4.31, + "learning_rate": 2.8493554730172608e-06, + "loss": 1.8841, + "step": 19737 + }, + { + "epoch": 4.31, + "learning_rate": 2.849246231155779e-06, + "loss": 2.8055, + "step": 19738 + }, + { + "epoch": 4.31, + "learning_rate": 2.8491369892942976e-06, + "loss": 1.9272, + "step": 19739 + }, + { + "epoch": 4.31, + "learning_rate": 2.8490277474328166e-06, + "loss": 0.3159, + "step": 19740 + }, + { + "epoch": 4.31, + "learning_rate": 2.848918505571335e-06, + "loss": 2.4432, + "step": 19741 + }, + { + "epoch": 4.31, + "learning_rate": 2.8488092637098542e-06, + "loss": 1.4005, + "step": 19742 + }, + { + "epoch": 4.31, + "learning_rate": 2.8487000218483724e-06, + "loss": 1.0929, + "step": 19743 + }, + { + "epoch": 4.31, + "learning_rate": 2.848590779986891e-06, + "loss": 3.3732, + "step": 19744 + }, + { + "epoch": 4.31, + "learning_rate": 2.84848153812541e-06, + "loss": 1.0568, + "step": 19745 + }, + { + "epoch": 4.31, + "learning_rate": 2.8483722962639287e-06, + "loss": 1.7704, + "step": 19746 + }, + { + "epoch": 4.31, + "learning_rate": 2.848263054402447e-06, + "loss": 2.6759, + "step": 19747 + }, + { + "epoch": 4.31, + "learning_rate": 2.848153812540966e-06, + "loss": 0.5815, + "step": 19748 + }, + { + "epoch": 4.31, + "learning_rate": 2.8480445706794845e-06, + "loss": 1.5904, + "step": 19749 + }, + { + "epoch": 4.31, + "learning_rate": 2.8479353288180036e-06, + "loss": 1.943, + "step": 19750 + }, + { + "epoch": 4.31, + "learning_rate": 2.847826086956522e-06, + "loss": 0.7658, + "step": 19751 + }, + { + "epoch": 4.31, + "learning_rate": 2.8477168450950404e-06, + "loss": 1.3341, + "step": 19752 + }, + { + "epoch": 4.31, + "learning_rate": 2.8476076032335594e-06, + "loss": 2.1253, + "step": 19753 + }, + { + "epoch": 4.31, + "learning_rate": 2.847498361372078e-06, + "loss": 0.8025, + "step": 19754 + }, + { + "epoch": 4.31, + "learning_rate": 2.8473891195105966e-06, + "loss": 0.7288, + "step": 19755 + }, + { + "epoch": 4.31, + "learning_rate": 2.8472798776491156e-06, + "loss": 2.3154, + "step": 19756 + }, + { + "epoch": 4.31, + "learning_rate": 2.847170635787634e-06, + "loss": 2.1902, + "step": 19757 + }, + { + "epoch": 4.31, + "learning_rate": 2.847061393926153e-06, + "loss": 1.0799, + "step": 19758 + }, + { + "epoch": 4.31, + "learning_rate": 2.8469521520646715e-06, + "loss": 2.2893, + "step": 19759 + }, + { + "epoch": 4.31, + "learning_rate": 2.84684291020319e-06, + "loss": 1.5319, + "step": 19760 + }, + { + "epoch": 4.31, + "learning_rate": 2.8467336683417087e-06, + "loss": 2.6208, + "step": 19761 + }, + { + "epoch": 4.31, + "learning_rate": 2.8466244264802273e-06, + "loss": 0.9582, + "step": 19762 + }, + { + "epoch": 4.31, + "learning_rate": 2.846515184618746e-06, + "loss": 1.3158, + "step": 19763 + }, + { + "epoch": 4.31, + "learning_rate": 2.846405942757265e-06, + "loss": 1.5265, + "step": 19764 + }, + { + "epoch": 4.31, + "learning_rate": 2.8462967008957836e-06, + "loss": 1.4413, + "step": 19765 + }, + { + "epoch": 4.31, + "learning_rate": 2.8461874590343018e-06, + "loss": 0.7232, + "step": 19766 + }, + { + "epoch": 4.31, + "learning_rate": 2.846078217172821e-06, + "loss": 1.2697, + "step": 19767 + }, + { + "epoch": 4.31, + "learning_rate": 2.8459689753113394e-06, + "loss": 1.0392, + "step": 19768 + }, + { + "epoch": 4.31, + "learning_rate": 2.8458597334498584e-06, + "loss": 1.9208, + "step": 19769 + }, + { + "epoch": 4.31, + "learning_rate": 2.845750491588377e-06, + "loss": 1.51, + "step": 19770 + }, + { + "epoch": 4.31, + "learning_rate": 2.8456412497268952e-06, + "loss": 1.5063, + "step": 19771 + }, + { + "epoch": 4.32, + "learning_rate": 2.8455320078654143e-06, + "loss": 2.1582, + "step": 19772 + }, + { + "epoch": 4.32, + "learning_rate": 2.845422766003933e-06, + "loss": 1.7884, + "step": 19773 + }, + { + "epoch": 4.32, + "learning_rate": 2.845313524142452e-06, + "loss": 1.2608, + "step": 19774 + }, + { + "epoch": 4.32, + "learning_rate": 2.8452042822809705e-06, + "loss": 1.6663, + "step": 19775 + }, + { + "epoch": 4.32, + "learning_rate": 2.8450950404194887e-06, + "loss": 0.7466, + "step": 19776 + }, + { + "epoch": 4.32, + "learning_rate": 2.8449857985580077e-06, + "loss": 1.4308, + "step": 19777 + }, + { + "epoch": 4.32, + "learning_rate": 2.8448765566965264e-06, + "loss": 2.0323, + "step": 19778 + }, + { + "epoch": 4.32, + "learning_rate": 2.8447673148350454e-06, + "loss": 1.7837, + "step": 19779 + }, + { + "epoch": 4.32, + "learning_rate": 2.8446580729735636e-06, + "loss": 1.9561, + "step": 19780 + }, + { + "epoch": 4.32, + "learning_rate": 2.844548831112082e-06, + "loss": 0.8981, + "step": 19781 + }, + { + "epoch": 4.32, + "learning_rate": 2.8444395892506012e-06, + "loss": 0.8244, + "step": 19782 + }, + { + "epoch": 4.32, + "learning_rate": 2.84433034738912e-06, + "loss": 1.6233, + "step": 19783 + }, + { + "epoch": 4.32, + "learning_rate": 2.8442211055276384e-06, + "loss": 2.3806, + "step": 19784 + }, + { + "epoch": 4.32, + "learning_rate": 2.844111863666157e-06, + "loss": 1.9941, + "step": 19785 + }, + { + "epoch": 4.32, + "learning_rate": 2.8440026218046757e-06, + "loss": 1.9366, + "step": 19786 + }, + { + "epoch": 4.32, + "learning_rate": 2.8438933799431943e-06, + "loss": 3.0579, + "step": 19787 + }, + { + "epoch": 4.32, + "learning_rate": 2.8437841380817133e-06, + "loss": 1.1696, + "step": 19788 + }, + { + "epoch": 4.32, + "learning_rate": 2.843674896220232e-06, + "loss": 1.2373, + "step": 19789 + }, + { + "epoch": 4.32, + "learning_rate": 2.84356565435875e-06, + "loss": 1.6985, + "step": 19790 + }, + { + "epoch": 4.32, + "learning_rate": 2.843456412497269e-06, + "loss": 2.4103, + "step": 19791 + }, + { + "epoch": 4.32, + "learning_rate": 2.8433471706357878e-06, + "loss": 1.2776, + "step": 19792 + }, + { + "epoch": 4.32, + "learning_rate": 2.843237928774307e-06, + "loss": 1.5012, + "step": 19793 + }, + { + "epoch": 4.32, + "learning_rate": 2.843128686912825e-06, + "loss": 1.6925, + "step": 19794 + }, + { + "epoch": 4.32, + "learning_rate": 2.8430194450513436e-06, + "loss": 1.2378, + "step": 19795 + }, + { + "epoch": 4.32, + "learning_rate": 2.8429102031898626e-06, + "loss": 2.1339, + "step": 19796 + }, + { + "epoch": 4.32, + "learning_rate": 2.8428009613283812e-06, + "loss": 0.7716, + "step": 19797 + }, + { + "epoch": 4.32, + "learning_rate": 2.8426917194669003e-06, + "loss": 2.3128, + "step": 19798 + }, + { + "epoch": 4.32, + "learning_rate": 2.8425824776054185e-06, + "loss": 3.055, + "step": 19799 + }, + { + "epoch": 4.32, + "learning_rate": 2.842473235743937e-06, + "loss": 1.0112, + "step": 19800 + }, + { + "epoch": 4.32, + "learning_rate": 2.842363993882456e-06, + "loss": 2.7324, + "step": 19801 + }, + { + "epoch": 4.32, + "learning_rate": 2.8422547520209747e-06, + "loss": 1.8334, + "step": 19802 + }, + { + "epoch": 4.32, + "learning_rate": 2.8421455101594938e-06, + "loss": 2.4905, + "step": 19803 + }, + { + "epoch": 4.32, + "learning_rate": 2.842036268298012e-06, + "loss": 2.0309, + "step": 19804 + }, + { + "epoch": 4.32, + "learning_rate": 2.8419270264365305e-06, + "loss": 1.7757, + "step": 19805 + }, + { + "epoch": 4.32, + "learning_rate": 2.8418177845750496e-06, + "loss": 2.4848, + "step": 19806 + }, + { + "epoch": 4.32, + "learning_rate": 2.841708542713568e-06, + "loss": 1.3985, + "step": 19807 + }, + { + "epoch": 4.32, + "learning_rate": 2.8415993008520872e-06, + "loss": 1.6245, + "step": 19808 + }, + { + "epoch": 4.32, + "learning_rate": 2.8414900589906054e-06, + "loss": 2.2779, + "step": 19809 + }, + { + "epoch": 4.32, + "learning_rate": 2.841380817129124e-06, + "loss": 0.3336, + "step": 19810 + }, + { + "epoch": 4.32, + "learning_rate": 2.841271575267643e-06, + "loss": 1.6644, + "step": 19811 + }, + { + "epoch": 4.32, + "learning_rate": 2.8411623334061617e-06, + "loss": 2.0547, + "step": 19812 + }, + { + "epoch": 4.32, + "learning_rate": 2.84105309154468e-06, + "loss": 2.072, + "step": 19813 + }, + { + "epoch": 4.32, + "learning_rate": 2.840943849683199e-06, + "loss": 2.3651, + "step": 19814 + }, + { + "epoch": 4.32, + "learning_rate": 2.8408346078217175e-06, + "loss": 0.9797, + "step": 19815 + }, + { + "epoch": 4.32, + "learning_rate": 2.840725365960236e-06, + "loss": 0.9106, + "step": 19816 + }, + { + "epoch": 4.32, + "learning_rate": 2.840616124098755e-06, + "loss": 1.5659, + "step": 19817 + }, + { + "epoch": 4.33, + "learning_rate": 2.8405068822372733e-06, + "loss": 1.4704, + "step": 19818 + }, + { + "epoch": 4.33, + "learning_rate": 2.840397640375792e-06, + "loss": 0.9452, + "step": 19819 + }, + { + "epoch": 4.33, + "learning_rate": 2.840288398514311e-06, + "loss": 1.5007, + "step": 19820 + }, + { + "epoch": 4.33, + "learning_rate": 2.8401791566528296e-06, + "loss": 1.5933, + "step": 19821 + }, + { + "epoch": 4.33, + "learning_rate": 2.8400699147913486e-06, + "loss": 1.0479, + "step": 19822 + }, + { + "epoch": 4.33, + "learning_rate": 2.839960672929867e-06, + "loss": 2.8906, + "step": 19823 + }, + { + "epoch": 4.33, + "learning_rate": 2.8398514310683854e-06, + "loss": 2.4791, + "step": 19824 + }, + { + "epoch": 4.33, + "learning_rate": 2.8397421892069045e-06, + "loss": 1.1116, + "step": 19825 + }, + { + "epoch": 4.33, + "learning_rate": 2.839632947345423e-06, + "loss": 1.2763, + "step": 19826 + }, + { + "epoch": 4.33, + "learning_rate": 2.8395237054839413e-06, + "loss": 1.2411, + "step": 19827 + }, + { + "epoch": 4.33, + "learning_rate": 2.8394144636224603e-06, + "loss": 1.7506, + "step": 19828 + }, + { + "epoch": 4.33, + "learning_rate": 2.839305221760979e-06, + "loss": 1.1466, + "step": 19829 + }, + { + "epoch": 4.33, + "learning_rate": 2.839195979899498e-06, + "loss": 2.861, + "step": 19830 + }, + { + "epoch": 4.33, + "learning_rate": 2.8390867380380166e-06, + "loss": 1.616, + "step": 19831 + }, + { + "epoch": 4.33, + "learning_rate": 2.8389774961765347e-06, + "loss": 3.0108, + "step": 19832 + }, + { + "epoch": 4.33, + "learning_rate": 2.8388682543150538e-06, + "loss": 2.0879, + "step": 19833 + }, + { + "epoch": 4.33, + "learning_rate": 2.8387590124535724e-06, + "loss": 0.7648, + "step": 19834 + }, + { + "epoch": 4.33, + "learning_rate": 2.8386497705920914e-06, + "loss": 1.3687, + "step": 19835 + }, + { + "epoch": 4.33, + "learning_rate": 2.83854052873061e-06, + "loss": 1.4255, + "step": 19836 + }, + { + "epoch": 4.33, + "learning_rate": 2.8384312868691282e-06, + "loss": 2.2282, + "step": 19837 + }, + { + "epoch": 4.33, + "learning_rate": 2.8383220450076473e-06, + "loss": 2.1966, + "step": 19838 + }, + { + "epoch": 4.33, + "learning_rate": 2.838212803146166e-06, + "loss": 1.2721, + "step": 19839 + }, + { + "epoch": 4.33, + "learning_rate": 2.8381035612846845e-06, + "loss": 3.0128, + "step": 19840 + }, + { + "epoch": 4.33, + "learning_rate": 2.8379943194232035e-06, + "loss": 1.9936, + "step": 19841 + }, + { + "epoch": 4.33, + "learning_rate": 2.8378850775617217e-06, + "loss": 3.0353, + "step": 19842 + }, + { + "epoch": 4.33, + "learning_rate": 2.8377758357002403e-06, + "loss": 1.1461, + "step": 19843 + }, + { + "epoch": 4.33, + "learning_rate": 2.8376665938387593e-06, + "loss": 1.9884, + "step": 19844 + }, + { + "epoch": 4.33, + "learning_rate": 2.837557351977278e-06, + "loss": 0.9476, + "step": 19845 + }, + { + "epoch": 4.33, + "learning_rate": 2.837448110115796e-06, + "loss": 2.4678, + "step": 19846 + }, + { + "epoch": 4.33, + "learning_rate": 2.837338868254315e-06, + "loss": 1.1313, + "step": 19847 + }, + { + "epoch": 4.33, + "learning_rate": 2.8372296263928338e-06, + "loss": 1.6167, + "step": 19848 + }, + { + "epoch": 4.33, + "learning_rate": 2.837120384531353e-06, + "loss": 1.8685, + "step": 19849 + }, + { + "epoch": 4.33, + "learning_rate": 2.8370111426698714e-06, + "loss": 1.885, + "step": 19850 + }, + { + "epoch": 4.33, + "learning_rate": 2.8369019008083896e-06, + "loss": 2.6629, + "step": 19851 + }, + { + "epoch": 4.33, + "learning_rate": 2.8367926589469087e-06, + "loss": 2.2577, + "step": 19852 + }, + { + "epoch": 4.33, + "learning_rate": 2.8366834170854273e-06, + "loss": 2.6934, + "step": 19853 + }, + { + "epoch": 4.33, + "learning_rate": 2.8365741752239463e-06, + "loss": 1.6228, + "step": 19854 + }, + { + "epoch": 4.33, + "learning_rate": 2.836464933362465e-06, + "loss": 1.9287, + "step": 19855 + }, + { + "epoch": 4.33, + "learning_rate": 2.836355691500983e-06, + "loss": 1.0625, + "step": 19856 + }, + { + "epoch": 4.33, + "learning_rate": 2.836246449639502e-06, + "loss": 2.2751, + "step": 19857 + }, + { + "epoch": 4.33, + "learning_rate": 2.8361372077780207e-06, + "loss": 3.3602, + "step": 19858 + }, + { + "epoch": 4.33, + "learning_rate": 2.8360279659165398e-06, + "loss": 2.1179, + "step": 19859 + }, + { + "epoch": 4.33, + "learning_rate": 2.835918724055058e-06, + "loss": 1.5935, + "step": 19860 + }, + { + "epoch": 4.33, + "learning_rate": 2.8358094821935766e-06, + "loss": 1.8744, + "step": 19861 + }, + { + "epoch": 4.33, + "learning_rate": 2.8357002403320956e-06, + "loss": 0.4579, + "step": 19862 + }, + { + "epoch": 4.34, + "learning_rate": 2.8355909984706142e-06, + "loss": 2.4599, + "step": 19863 + }, + { + "epoch": 4.34, + "learning_rate": 2.8354817566091333e-06, + "loss": 1.5864, + "step": 19864 + }, + { + "epoch": 4.34, + "learning_rate": 2.8353725147476514e-06, + "loss": 1.7008, + "step": 19865 + }, + { + "epoch": 4.34, + "learning_rate": 2.83526327288617e-06, + "loss": 1.7687, + "step": 19866 + }, + { + "epoch": 4.34, + "learning_rate": 2.835154031024689e-06, + "loss": 1.3659, + "step": 19867 + }, + { + "epoch": 4.34, + "learning_rate": 2.8350447891632077e-06, + "loss": 0.9297, + "step": 19868 + }, + { + "epoch": 4.34, + "learning_rate": 2.8349355473017263e-06, + "loss": 1.7428, + "step": 19869 + }, + { + "epoch": 4.34, + "learning_rate": 2.834826305440245e-06, + "loss": 0.9437, + "step": 19870 + }, + { + "epoch": 4.34, + "learning_rate": 2.8347170635787635e-06, + "loss": 1.9678, + "step": 19871 + }, + { + "epoch": 4.34, + "learning_rate": 2.834607821717282e-06, + "loss": 2.1375, + "step": 19872 + }, + { + "epoch": 4.34, + "learning_rate": 2.834498579855801e-06, + "loss": 1.919, + "step": 19873 + }, + { + "epoch": 4.34, + "learning_rate": 2.83438933799432e-06, + "loss": 2.6207, + "step": 19874 + }, + { + "epoch": 4.34, + "learning_rate": 2.834280096132838e-06, + "loss": 1.4168, + "step": 19875 + }, + { + "epoch": 4.34, + "learning_rate": 2.834170854271357e-06, + "loss": 3.0337, + "step": 19876 + }, + { + "epoch": 4.34, + "learning_rate": 2.8340616124098756e-06, + "loss": 3.2216, + "step": 19877 + }, + { + "epoch": 4.34, + "learning_rate": 2.8339523705483947e-06, + "loss": 1.3482, + "step": 19878 + }, + { + "epoch": 4.34, + "learning_rate": 2.833843128686913e-06, + "loss": 0.6179, + "step": 19879 + }, + { + "epoch": 4.34, + "learning_rate": 2.8337338868254315e-06, + "loss": 3.2122, + "step": 19880 + }, + { + "epoch": 4.34, + "learning_rate": 2.8336246449639505e-06, + "loss": 2.9643, + "step": 19881 + }, + { + "epoch": 4.34, + "learning_rate": 2.833515403102469e-06, + "loss": 1.9384, + "step": 19882 + }, + { + "epoch": 4.34, + "learning_rate": 2.833406161240988e-06, + "loss": 0.5776, + "step": 19883 + }, + { + "epoch": 4.34, + "learning_rate": 2.8332969193795063e-06, + "loss": 2.7112, + "step": 19884 + }, + { + "epoch": 4.34, + "learning_rate": 2.833187677518025e-06, + "loss": 1.7604, + "step": 19885 + }, + { + "epoch": 4.34, + "learning_rate": 2.833078435656544e-06, + "loss": 2.5825, + "step": 19886 + }, + { + "epoch": 4.34, + "learning_rate": 2.8329691937950626e-06, + "loss": 1.6708, + "step": 19887 + }, + { + "epoch": 4.34, + "learning_rate": 2.8328599519335816e-06, + "loss": 1.3548, + "step": 19888 + }, + { + "epoch": 4.34, + "learning_rate": 2.8327507100721e-06, + "loss": 2.2913, + "step": 19889 + }, + { + "epoch": 4.34, + "learning_rate": 2.8326414682106184e-06, + "loss": 2.4382, + "step": 19890 + }, + { + "epoch": 4.34, + "learning_rate": 2.8325322263491374e-06, + "loss": 2.7504, + "step": 19891 + }, + { + "epoch": 4.34, + "learning_rate": 2.832422984487656e-06, + "loss": 2.5172, + "step": 19892 + }, + { + "epoch": 4.34, + "learning_rate": 2.8323137426261742e-06, + "loss": 1.2556, + "step": 19893 + }, + { + "epoch": 4.34, + "learning_rate": 2.8322045007646933e-06, + "loss": 2.8224, + "step": 19894 + }, + { + "epoch": 4.34, + "learning_rate": 2.832095258903212e-06, + "loss": 1.2156, + "step": 19895 + }, + { + "epoch": 4.34, + "learning_rate": 2.8319860170417305e-06, + "loss": 1.5096, + "step": 19896 + }, + { + "epoch": 4.34, + "learning_rate": 2.8318767751802495e-06, + "loss": 3.3901, + "step": 19897 + }, + { + "epoch": 4.34, + "learning_rate": 2.8317675333187677e-06, + "loss": 1.1522, + "step": 19898 + }, + { + "epoch": 4.34, + "learning_rate": 2.8316582914572863e-06, + "loss": 1.012, + "step": 19899 + }, + { + "epoch": 4.34, + "learning_rate": 2.8315490495958054e-06, + "loss": 2.5864, + "step": 19900 + }, + { + "epoch": 4.34, + "learning_rate": 2.831439807734324e-06, + "loss": 0.8564, + "step": 19901 + }, + { + "epoch": 4.34, + "learning_rate": 2.831330565872843e-06, + "loss": 1.6713, + "step": 19902 + }, + { + "epoch": 4.34, + "learning_rate": 2.831221324011361e-06, + "loss": 1.2351, + "step": 19903 + }, + { + "epoch": 4.34, + "learning_rate": 2.83111208214988e-06, + "loss": 0.3416, + "step": 19904 + }, + { + "epoch": 4.34, + "learning_rate": 2.831002840288399e-06, + "loss": 1.0052, + "step": 19905 + }, + { + "epoch": 4.34, + "learning_rate": 2.8308935984269175e-06, + "loss": 1.2229, + "step": 19906 + }, + { + "epoch": 4.34, + "learning_rate": 2.8307843565654365e-06, + "loss": 1.7319, + "step": 19907 + }, + { + "epoch": 4.34, + "learning_rate": 2.8306751147039547e-06, + "loss": 1.7625, + "step": 19908 + }, + { + "epoch": 4.35, + "learning_rate": 2.8305658728424733e-06, + "loss": 2.6316, + "step": 19909 + }, + { + "epoch": 4.35, + "learning_rate": 2.8304566309809923e-06, + "loss": 2.1613, + "step": 19910 + }, + { + "epoch": 4.35, + "learning_rate": 2.830347389119511e-06, + "loss": 1.8336, + "step": 19911 + }, + { + "epoch": 4.35, + "learning_rate": 2.830238147258029e-06, + "loss": 1.6829, + "step": 19912 + }, + { + "epoch": 4.35, + "learning_rate": 2.830128905396548e-06, + "loss": 1.275, + "step": 19913 + }, + { + "epoch": 4.35, + "learning_rate": 2.8300196635350668e-06, + "loss": 3.0889, + "step": 19914 + }, + { + "epoch": 4.35, + "learning_rate": 2.829910421673586e-06, + "loss": 1.1982, + "step": 19915 + }, + { + "epoch": 4.35, + "learning_rate": 2.8298011798121044e-06, + "loss": 0.4609, + "step": 19916 + }, + { + "epoch": 4.35, + "learning_rate": 2.8296919379506226e-06, + "loss": 2.1307, + "step": 19917 + }, + { + "epoch": 4.35, + "learning_rate": 2.8295826960891416e-06, + "loss": 0.4947, + "step": 19918 + }, + { + "epoch": 4.35, + "learning_rate": 2.8294734542276602e-06, + "loss": 1.5069, + "step": 19919 + }, + { + "epoch": 4.35, + "learning_rate": 2.8293642123661793e-06, + "loss": 2.0132, + "step": 19920 + }, + { + "epoch": 4.35, + "learning_rate": 2.829254970504698e-06, + "loss": 1.2958, + "step": 19921 + }, + { + "epoch": 4.35, + "learning_rate": 2.829145728643216e-06, + "loss": 1.6713, + "step": 19922 + }, + { + "epoch": 4.35, + "learning_rate": 2.829036486781735e-06, + "loss": 1.3462, + "step": 19923 + }, + { + "epoch": 4.35, + "learning_rate": 2.8289272449202537e-06, + "loss": 2.2253, + "step": 19924 + }, + { + "epoch": 4.35, + "learning_rate": 2.8288180030587723e-06, + "loss": 1.4709, + "step": 19925 + }, + { + "epoch": 4.35, + "learning_rate": 2.828708761197291e-06, + "loss": 1.7465, + "step": 19926 + }, + { + "epoch": 4.35, + "learning_rate": 2.8285995193358096e-06, + "loss": 2.1307, + "step": 19927 + }, + { + "epoch": 4.35, + "learning_rate": 2.828490277474328e-06, + "loss": 1.6228, + "step": 19928 + }, + { + "epoch": 4.35, + "learning_rate": 2.828381035612847e-06, + "loss": 1.2816, + "step": 19929 + }, + { + "epoch": 4.35, + "learning_rate": 2.828271793751366e-06, + "loss": 1.2206, + "step": 19930 + }, + { + "epoch": 4.35, + "learning_rate": 2.828162551889884e-06, + "loss": 1.7796, + "step": 19931 + }, + { + "epoch": 4.35, + "learning_rate": 2.828053310028403e-06, + "loss": 1.266, + "step": 19932 + }, + { + "epoch": 4.35, + "learning_rate": 2.8279440681669216e-06, + "loss": 1.4382, + "step": 19933 + }, + { + "epoch": 4.35, + "learning_rate": 2.8278348263054407e-06, + "loss": 1.7444, + "step": 19934 + }, + { + "epoch": 4.35, + "learning_rate": 2.8277255844439593e-06, + "loss": 2.0495, + "step": 19935 + }, + { + "epoch": 4.35, + "learning_rate": 2.8276163425824775e-06, + "loss": 1.8849, + "step": 19936 + }, + { + "epoch": 4.35, + "learning_rate": 2.8275071007209965e-06, + "loss": 2.6391, + "step": 19937 + }, + { + "epoch": 4.35, + "learning_rate": 2.827397858859515e-06, + "loss": 2.5392, + "step": 19938 + }, + { + "epoch": 4.35, + "learning_rate": 2.827288616998034e-06, + "loss": 1.5169, + "step": 19939 + }, + { + "epoch": 4.35, + "learning_rate": 2.8271793751365528e-06, + "loss": 1.9666, + "step": 19940 + }, + { + "epoch": 4.35, + "learning_rate": 2.827070133275071e-06, + "loss": 1.1019, + "step": 19941 + }, + { + "epoch": 4.35, + "learning_rate": 2.82696089141359e-06, + "loss": 1.3767, + "step": 19942 + }, + { + "epoch": 4.35, + "learning_rate": 2.8268516495521086e-06, + "loss": 1.5193, + "step": 19943 + }, + { + "epoch": 4.35, + "learning_rate": 2.8267424076906276e-06, + "loss": 1.0214, + "step": 19944 + }, + { + "epoch": 4.35, + "learning_rate": 2.826633165829146e-06, + "loss": 1.0356, + "step": 19945 + }, + { + "epoch": 4.35, + "learning_rate": 2.8265239239676644e-06, + "loss": 2.273, + "step": 19946 + }, + { + "epoch": 4.35, + "learning_rate": 2.8264146821061835e-06, + "loss": 0.3349, + "step": 19947 + }, + { + "epoch": 4.35, + "learning_rate": 2.826305440244702e-06, + "loss": 2.3463, + "step": 19948 + }, + { + "epoch": 4.35, + "learning_rate": 2.8261961983832207e-06, + "loss": 1.4935, + "step": 19949 + }, + { + "epoch": 4.35, + "learning_rate": 2.8260869565217393e-06, + "loss": 1.5031, + "step": 19950 + }, + { + "epoch": 4.35, + "learning_rate": 2.825977714660258e-06, + "loss": 1.2659, + "step": 19951 + }, + { + "epoch": 4.35, + "learning_rate": 2.825868472798777e-06, + "loss": 3.0486, + "step": 19952 + }, + { + "epoch": 4.35, + "learning_rate": 2.8257592309372956e-06, + "loss": 1.7865, + "step": 19953 + }, + { + "epoch": 4.35, + "learning_rate": 2.825649989075814e-06, + "loss": 2.2729, + "step": 19954 + }, + { + "epoch": 4.36, + "learning_rate": 2.8255407472143328e-06, + "loss": 2.4843, + "step": 19955 + }, + { + "epoch": 4.36, + "learning_rate": 2.8254315053528514e-06, + "loss": 3.5189, + "step": 19956 + }, + { + "epoch": 4.36, + "learning_rate": 2.82532226349137e-06, + "loss": 2.2087, + "step": 19957 + }, + { + "epoch": 4.36, + "learning_rate": 2.825213021629889e-06, + "loss": 1.8147, + "step": 19958 + }, + { + "epoch": 4.36, + "learning_rate": 2.8251037797684072e-06, + "loss": 1.1737, + "step": 19959 + }, + { + "epoch": 4.36, + "learning_rate": 2.824994537906926e-06, + "loss": 1.6913, + "step": 19960 + }, + { + "epoch": 4.36, + "learning_rate": 2.824885296045445e-06, + "loss": 2.3171, + "step": 19961 + }, + { + "epoch": 4.36, + "learning_rate": 2.8247760541839635e-06, + "loss": 1.5228, + "step": 19962 + }, + { + "epoch": 4.36, + "learning_rate": 2.8246668123224825e-06, + "loss": 2.1918, + "step": 19963 + }, + { + "epoch": 4.36, + "learning_rate": 2.8245575704610007e-06, + "loss": 0.9591, + "step": 19964 + }, + { + "epoch": 4.36, + "learning_rate": 2.8244483285995193e-06, + "loss": 1.6779, + "step": 19965 + }, + { + "epoch": 4.36, + "learning_rate": 2.8243390867380384e-06, + "loss": 2.0337, + "step": 19966 + }, + { + "epoch": 4.36, + "learning_rate": 2.824229844876557e-06, + "loss": 1.5259, + "step": 19967 + }, + { + "epoch": 4.36, + "learning_rate": 2.824120603015076e-06, + "loss": 1.1459, + "step": 19968 + }, + { + "epoch": 4.36, + "learning_rate": 2.824011361153594e-06, + "loss": 2.1527, + "step": 19969 + }, + { + "epoch": 4.36, + "learning_rate": 2.823902119292113e-06, + "loss": 2.6494, + "step": 19970 + }, + { + "epoch": 4.36, + "learning_rate": 2.823792877430632e-06, + "loss": 1.7371, + "step": 19971 + }, + { + "epoch": 4.36, + "learning_rate": 2.8236836355691504e-06, + "loss": 1.5324, + "step": 19972 + }, + { + "epoch": 4.36, + "learning_rate": 2.8235743937076695e-06, + "loss": 1.8725, + "step": 19973 + }, + { + "epoch": 4.36, + "learning_rate": 2.8234651518461877e-06, + "loss": 1.5385, + "step": 19974 + }, + { + "epoch": 4.36, + "learning_rate": 2.8233559099847063e-06, + "loss": 1.0982, + "step": 19975 + }, + { + "epoch": 4.36, + "learning_rate": 2.8232466681232253e-06, + "loss": 2.1896, + "step": 19976 + }, + { + "epoch": 4.36, + "learning_rate": 2.823137426261744e-06, + "loss": 1.9607, + "step": 19977 + }, + { + "epoch": 4.36, + "learning_rate": 2.823028184400262e-06, + "loss": 1.2929, + "step": 19978 + }, + { + "epoch": 4.36, + "learning_rate": 2.822918942538781e-06, + "loss": 1.186, + "step": 19979 + }, + { + "epoch": 4.36, + "learning_rate": 2.8228097006772998e-06, + "loss": 1.3325, + "step": 19980 + }, + { + "epoch": 4.36, + "learning_rate": 2.8227004588158184e-06, + "loss": 2.8987, + "step": 19981 + }, + { + "epoch": 4.36, + "learning_rate": 2.8225912169543374e-06, + "loss": 2.1134, + "step": 19982 + }, + { + "epoch": 4.36, + "learning_rate": 2.8224819750928556e-06, + "loss": 2.307, + "step": 19983 + }, + { + "epoch": 4.36, + "learning_rate": 2.822372733231374e-06, + "loss": 1.7377, + "step": 19984 + }, + { + "epoch": 4.36, + "learning_rate": 2.8222634913698932e-06, + "loss": 1.2688, + "step": 19985 + }, + { + "epoch": 4.36, + "learning_rate": 2.822154249508412e-06, + "loss": 2.0972, + "step": 19986 + }, + { + "epoch": 4.36, + "learning_rate": 2.822045007646931e-06, + "loss": 1.8707, + "step": 19987 + }, + { + "epoch": 4.36, + "learning_rate": 2.821935765785449e-06, + "loss": 1.5981, + "step": 19988 + }, + { + "epoch": 4.36, + "learning_rate": 2.8218265239239677e-06, + "loss": 0.7723, + "step": 19989 + }, + { + "epoch": 4.36, + "learning_rate": 2.8217172820624867e-06, + "loss": 1.944, + "step": 19990 + }, + { + "epoch": 4.36, + "learning_rate": 2.8216080402010053e-06, + "loss": 1.7854, + "step": 19991 + }, + { + "epoch": 4.36, + "learning_rate": 2.8214987983395235e-06, + "loss": 2.9611, + "step": 19992 + }, + { + "epoch": 4.36, + "learning_rate": 2.8213895564780425e-06, + "loss": 2.3855, + "step": 19993 + }, + { + "epoch": 4.36, + "learning_rate": 2.821280314616561e-06, + "loss": 1.6496, + "step": 19994 + }, + { + "epoch": 4.36, + "learning_rate": 2.82117107275508e-06, + "loss": 1.697, + "step": 19995 + }, + { + "epoch": 4.36, + "learning_rate": 2.821061830893599e-06, + "loss": 2.6605, + "step": 19996 + }, + { + "epoch": 4.36, + "learning_rate": 2.820952589032117e-06, + "loss": 1.1512, + "step": 19997 + }, + { + "epoch": 4.36, + "learning_rate": 2.820843347170636e-06, + "loss": 2.2005, + "step": 19998 + }, + { + "epoch": 4.36, + "learning_rate": 2.8207341053091546e-06, + "loss": 2.8322, + "step": 19999 + }, + { + "epoch": 4.36, + "learning_rate": 2.8206248634476737e-06, + "loss": 1.2719, + "step": 20000 + }, + { + "epoch": 4.37, + "learning_rate": 2.8205156215861923e-06, + "loss": 1.1528, + "step": 20001 + }, + { + "epoch": 4.37, + "learning_rate": 2.8204063797247105e-06, + "loss": 1.2176, + "step": 20002 + }, + { + "epoch": 4.37, + "learning_rate": 2.8202971378632295e-06, + "loss": 1.9945, + "step": 20003 + }, + { + "epoch": 4.37, + "learning_rate": 2.820187896001748e-06, + "loss": 2.5688, + "step": 20004 + }, + { + "epoch": 4.37, + "learning_rate": 2.820078654140267e-06, + "loss": 1.5007, + "step": 20005 + }, + { + "epoch": 4.37, + "learning_rate": 2.8199694122787858e-06, + "loss": 1.1622, + "step": 20006 + }, + { + "epoch": 4.37, + "learning_rate": 2.819860170417304e-06, + "loss": 1.317, + "step": 20007 + }, + { + "epoch": 4.37, + "learning_rate": 2.819750928555823e-06, + "loss": 0.8803, + "step": 20008 + }, + { + "epoch": 4.37, + "learning_rate": 2.8196416866943416e-06, + "loss": 1.1974, + "step": 20009 + }, + { + "epoch": 4.37, + "learning_rate": 2.81953244483286e-06, + "loss": 1.0792, + "step": 20010 + }, + { + "epoch": 4.37, + "learning_rate": 2.819423202971379e-06, + "loss": 1.7956, + "step": 20011 + }, + { + "epoch": 4.37, + "learning_rate": 2.8193139611098974e-06, + "loss": 2.0743, + "step": 20012 + }, + { + "epoch": 4.37, + "learning_rate": 2.819204719248416e-06, + "loss": 0.8183, + "step": 20013 + }, + { + "epoch": 4.37, + "learning_rate": 2.819095477386935e-06, + "loss": 0.5803, + "step": 20014 + }, + { + "epoch": 4.37, + "learning_rate": 2.8189862355254537e-06, + "loss": 0.8919, + "step": 20015 + }, + { + "epoch": 4.37, + "learning_rate": 2.818876993663972e-06, + "loss": 1.4378, + "step": 20016 + }, + { + "epoch": 4.37, + "learning_rate": 2.818767751802491e-06, + "loss": 2.1628, + "step": 20017 + }, + { + "epoch": 4.37, + "learning_rate": 2.8186585099410095e-06, + "loss": 2.1048, + "step": 20018 + }, + { + "epoch": 4.37, + "learning_rate": 2.8185492680795285e-06, + "loss": 2.1748, + "step": 20019 + }, + { + "epoch": 4.37, + "learning_rate": 2.818440026218047e-06, + "loss": 1.7992, + "step": 20020 + }, + { + "epoch": 4.37, + "learning_rate": 2.8183307843565653e-06, + "loss": 2.8206, + "step": 20021 + }, + { + "epoch": 4.37, + "learning_rate": 2.8182215424950844e-06, + "loss": 0.1626, + "step": 20022 + }, + { + "epoch": 4.37, + "learning_rate": 2.818112300633603e-06, + "loss": 2.5187, + "step": 20023 + }, + { + "epoch": 4.37, + "learning_rate": 2.818003058772122e-06, + "loss": 1.3797, + "step": 20024 + }, + { + "epoch": 4.37, + "learning_rate": 2.8178938169106402e-06, + "loss": 1.7942, + "step": 20025 + }, + { + "epoch": 4.37, + "learning_rate": 2.817784575049159e-06, + "loss": 1.2159, + "step": 20026 + }, + { + "epoch": 4.37, + "learning_rate": 2.817675333187678e-06, + "loss": 1.3778, + "step": 20027 + }, + { + "epoch": 4.37, + "learning_rate": 2.8175660913261965e-06, + "loss": 1.8418, + "step": 20028 + }, + { + "epoch": 4.37, + "learning_rate": 2.8174568494647155e-06, + "loss": 2.1348, + "step": 20029 + }, + { + "epoch": 4.37, + "learning_rate": 2.8173476076032337e-06, + "loss": 0.8257, + "step": 20030 + }, + { + "epoch": 4.37, + "learning_rate": 2.8172383657417523e-06, + "loss": 1.042, + "step": 20031 + }, + { + "epoch": 4.37, + "learning_rate": 2.8171291238802713e-06, + "loss": 1.5673, + "step": 20032 + }, + { + "epoch": 4.37, + "learning_rate": 2.81701988201879e-06, + "loss": 2.1058, + "step": 20033 + }, + { + "epoch": 4.37, + "learning_rate": 2.8169106401573086e-06, + "loss": 2.4025, + "step": 20034 + }, + { + "epoch": 4.37, + "learning_rate": 2.816801398295827e-06, + "loss": 0.6053, + "step": 20035 + }, + { + "epoch": 4.37, + "learning_rate": 2.8166921564343458e-06, + "loss": 1.5712, + "step": 20036 + }, + { + "epoch": 4.37, + "learning_rate": 2.8165829145728644e-06, + "loss": 0.8202, + "step": 20037 + }, + { + "epoch": 4.37, + "learning_rate": 2.8164736727113834e-06, + "loss": 2.0101, + "step": 20038 + }, + { + "epoch": 4.37, + "learning_rate": 2.816364430849902e-06, + "loss": 2.241, + "step": 20039 + }, + { + "epoch": 4.37, + "learning_rate": 2.8162551889884202e-06, + "loss": 1.8933, + "step": 20040 + }, + { + "epoch": 4.37, + "learning_rate": 2.8161459471269393e-06, + "loss": 2.4095, + "step": 20041 + }, + { + "epoch": 4.37, + "learning_rate": 2.816036705265458e-06, + "loss": 0.5309, + "step": 20042 + }, + { + "epoch": 4.37, + "learning_rate": 2.815927463403977e-06, + "loss": 1.166, + "step": 20043 + }, + { + "epoch": 4.37, + "learning_rate": 2.815818221542495e-06, + "loss": 1.6698, + "step": 20044 + }, + { + "epoch": 4.37, + "learning_rate": 2.8157089796810137e-06, + "loss": 4.2154, + "step": 20045 + }, + { + "epoch": 4.37, + "learning_rate": 2.8155997378195327e-06, + "loss": 1.798, + "step": 20046 + }, + { + "epoch": 4.38, + "learning_rate": 2.8154904959580513e-06, + "loss": 2.8641, + "step": 20047 + }, + { + "epoch": 4.38, + "learning_rate": 2.8153812540965704e-06, + "loss": 2.7688, + "step": 20048 + }, + { + "epoch": 4.38, + "learning_rate": 2.8152720122350886e-06, + "loss": 2.1599, + "step": 20049 + }, + { + "epoch": 4.38, + "learning_rate": 2.815162770373607e-06, + "loss": 0.7857, + "step": 20050 + }, + { + "epoch": 4.38, + "learning_rate": 2.8150535285121262e-06, + "loss": 2.6794, + "step": 20051 + }, + { + "epoch": 4.38, + "learning_rate": 2.814944286650645e-06, + "loss": 0.7872, + "step": 20052 + }, + { + "epoch": 4.38, + "learning_rate": 2.814835044789164e-06, + "loss": 0.3775, + "step": 20053 + }, + { + "epoch": 4.38, + "learning_rate": 2.814725802927682e-06, + "loss": 2.1721, + "step": 20054 + }, + { + "epoch": 4.38, + "learning_rate": 2.8146165610662007e-06, + "loss": 1.2288, + "step": 20055 + }, + { + "epoch": 4.38, + "learning_rate": 2.8145073192047197e-06, + "loss": 2.2237, + "step": 20056 + }, + { + "epoch": 4.38, + "learning_rate": 2.8143980773432383e-06, + "loss": 1.1585, + "step": 20057 + }, + { + "epoch": 4.38, + "learning_rate": 2.8142888354817565e-06, + "loss": 1.5153, + "step": 20058 + }, + { + "epoch": 4.38, + "learning_rate": 2.8141795936202755e-06, + "loss": 1.9599, + "step": 20059 + }, + { + "epoch": 4.38, + "learning_rate": 2.814070351758794e-06, + "loss": 0.9662, + "step": 20060 + }, + { + "epoch": 4.38, + "learning_rate": 2.813961109897313e-06, + "loss": 1.7496, + "step": 20061 + }, + { + "epoch": 4.38, + "learning_rate": 2.8138518680358318e-06, + "loss": 2.7887, + "step": 20062 + }, + { + "epoch": 4.38, + "learning_rate": 2.81374262617435e-06, + "loss": 0.263, + "step": 20063 + }, + { + "epoch": 4.38, + "learning_rate": 2.813633384312869e-06, + "loss": 1.5224, + "step": 20064 + }, + { + "epoch": 4.38, + "learning_rate": 2.8135241424513876e-06, + "loss": 1.7956, + "step": 20065 + }, + { + "epoch": 4.38, + "learning_rate": 2.8134149005899062e-06, + "loss": 3.5152, + "step": 20066 + }, + { + "epoch": 4.38, + "learning_rate": 2.8133056587284253e-06, + "loss": 2.198, + "step": 20067 + }, + { + "epoch": 4.38, + "learning_rate": 2.8131964168669434e-06, + "loss": 1.8704, + "step": 20068 + }, + { + "epoch": 4.38, + "learning_rate": 2.813087175005462e-06, + "loss": 0.8111, + "step": 20069 + }, + { + "epoch": 4.38, + "learning_rate": 2.812977933143981e-06, + "loss": 2.1943, + "step": 20070 + }, + { + "epoch": 4.38, + "learning_rate": 2.8128686912824997e-06, + "loss": 1.4658, + "step": 20071 + }, + { + "epoch": 4.38, + "learning_rate": 2.8127594494210187e-06, + "loss": 0.2673, + "step": 20072 + }, + { + "epoch": 4.38, + "learning_rate": 2.812650207559537e-06, + "loss": 1.0423, + "step": 20073 + }, + { + "epoch": 4.38, + "learning_rate": 2.8125409656980555e-06, + "loss": 2.9862, + "step": 20074 + }, + { + "epoch": 4.38, + "learning_rate": 2.8124317238365746e-06, + "loss": 1.6448, + "step": 20075 + }, + { + "epoch": 4.38, + "learning_rate": 2.812322481975093e-06, + "loss": 1.9971, + "step": 20076 + }, + { + "epoch": 4.38, + "learning_rate": 2.8122132401136114e-06, + "loss": 2.3708, + "step": 20077 + }, + { + "epoch": 4.38, + "learning_rate": 2.8121039982521304e-06, + "loss": 2.4656, + "step": 20078 + }, + { + "epoch": 4.38, + "learning_rate": 2.811994756390649e-06, + "loss": 1.0216, + "step": 20079 + }, + { + "epoch": 4.38, + "learning_rate": 2.811885514529168e-06, + "loss": 2.745, + "step": 20080 + }, + { + "epoch": 4.38, + "learning_rate": 2.8117762726676867e-06, + "loss": 2.2134, + "step": 20081 + }, + { + "epoch": 4.38, + "learning_rate": 2.811667030806205e-06, + "loss": 1.5268, + "step": 20082 + }, + { + "epoch": 4.38, + "learning_rate": 2.811557788944724e-06, + "loss": 2.3112, + "step": 20083 + }, + { + "epoch": 4.38, + "learning_rate": 2.8114485470832425e-06, + "loss": 3.3448, + "step": 20084 + }, + { + "epoch": 4.38, + "learning_rate": 2.8113393052217615e-06, + "loss": 1.3614, + "step": 20085 + }, + { + "epoch": 4.38, + "learning_rate": 2.81123006336028e-06, + "loss": 2.4192, + "step": 20086 + }, + { + "epoch": 4.38, + "learning_rate": 2.8111208214987983e-06, + "loss": 1.9069, + "step": 20087 + }, + { + "epoch": 4.38, + "learning_rate": 2.8110115796373174e-06, + "loss": 2.2057, + "step": 20088 + }, + { + "epoch": 4.38, + "learning_rate": 2.810902337775836e-06, + "loss": 1.408, + "step": 20089 + }, + { + "epoch": 4.38, + "learning_rate": 2.8107930959143546e-06, + "loss": 2.4551, + "step": 20090 + }, + { + "epoch": 4.38, + "learning_rate": 2.810683854052873e-06, + "loss": 1.9339, + "step": 20091 + }, + { + "epoch": 4.38, + "learning_rate": 2.810574612191392e-06, + "loss": 1.8965, + "step": 20092 + }, + { + "epoch": 4.39, + "learning_rate": 2.8104653703299104e-06, + "loss": 3.6367, + "step": 20093 + }, + { + "epoch": 4.39, + "learning_rate": 2.8103561284684295e-06, + "loss": 2.5243, + "step": 20094 + }, + { + "epoch": 4.39, + "learning_rate": 2.810246886606948e-06, + "loss": 2.1486, + "step": 20095 + }, + { + "epoch": 4.39, + "learning_rate": 2.8101376447454667e-06, + "loss": 2.2366, + "step": 20096 + }, + { + "epoch": 4.39, + "learning_rate": 2.8100284028839853e-06, + "loss": 1.4873, + "step": 20097 + }, + { + "epoch": 4.39, + "learning_rate": 2.809919161022504e-06, + "loss": 2.5016, + "step": 20098 + }, + { + "epoch": 4.39, + "learning_rate": 2.809809919161023e-06, + "loss": 0.6134, + "step": 20099 + }, + { + "epoch": 4.39, + "learning_rate": 2.8097006772995415e-06, + "loss": 2.1322, + "step": 20100 + }, + { + "epoch": 4.39, + "learning_rate": 2.8095914354380597e-06, + "loss": 0.8251, + "step": 20101 + }, + { + "epoch": 4.39, + "learning_rate": 2.8094821935765788e-06, + "loss": 1.26, + "step": 20102 + }, + { + "epoch": 4.39, + "learning_rate": 2.8093729517150974e-06, + "loss": 1.4241, + "step": 20103 + }, + { + "epoch": 4.39, + "learning_rate": 2.8092637098536164e-06, + "loss": 2.0477, + "step": 20104 + }, + { + "epoch": 4.39, + "learning_rate": 2.809154467992135e-06, + "loss": 1.3727, + "step": 20105 + }, + { + "epoch": 4.39, + "learning_rate": 2.809045226130653e-06, + "loss": 1.2167, + "step": 20106 + }, + { + "epoch": 4.39, + "learning_rate": 2.8089359842691722e-06, + "loss": 1.8794, + "step": 20107 + }, + { + "epoch": 4.39, + "learning_rate": 2.808826742407691e-06, + "loss": 2.0042, + "step": 20108 + }, + { + "epoch": 4.39, + "learning_rate": 2.80871750054621e-06, + "loss": 1.0484, + "step": 20109 + }, + { + "epoch": 4.39, + "learning_rate": 2.808608258684728e-06, + "loss": 2.025, + "step": 20110 + }, + { + "epoch": 4.39, + "learning_rate": 2.8084990168232467e-06, + "loss": 1.5146, + "step": 20111 + }, + { + "epoch": 4.39, + "learning_rate": 2.8083897749617657e-06, + "loss": 2.4637, + "step": 20112 + }, + { + "epoch": 4.39, + "learning_rate": 2.8082805331002843e-06, + "loss": 1.8206, + "step": 20113 + }, + { + "epoch": 4.39, + "learning_rate": 2.8081712912388034e-06, + "loss": 2.1549, + "step": 20114 + }, + { + "epoch": 4.39, + "learning_rate": 2.8080620493773216e-06, + "loss": 1.9318, + "step": 20115 + }, + { + "epoch": 4.39, + "learning_rate": 2.80795280751584e-06, + "loss": 1.0078, + "step": 20116 + }, + { + "epoch": 4.39, + "learning_rate": 2.807843565654359e-06, + "loss": 1.958, + "step": 20117 + }, + { + "epoch": 4.39, + "learning_rate": 2.807734323792878e-06, + "loss": 1.8605, + "step": 20118 + }, + { + "epoch": 4.39, + "learning_rate": 2.8076250819313964e-06, + "loss": 1.5441, + "step": 20119 + }, + { + "epoch": 4.39, + "learning_rate": 2.807515840069915e-06, + "loss": 2.0681, + "step": 20120 + }, + { + "epoch": 4.39, + "learning_rate": 2.8074065982084336e-06, + "loss": 2.9187, + "step": 20121 + }, + { + "epoch": 4.39, + "learning_rate": 2.8072973563469523e-06, + "loss": 1.7174, + "step": 20122 + }, + { + "epoch": 4.39, + "learning_rate": 2.8071881144854713e-06, + "loss": 2.0008, + "step": 20123 + }, + { + "epoch": 4.39, + "learning_rate": 2.8070788726239895e-06, + "loss": 3.4398, + "step": 20124 + }, + { + "epoch": 4.39, + "learning_rate": 2.806969630762508e-06, + "loss": 4.5032, + "step": 20125 + }, + { + "epoch": 4.39, + "learning_rate": 2.806860388901027e-06, + "loss": 2.0261, + "step": 20126 + }, + { + "epoch": 4.39, + "learning_rate": 2.8067511470395457e-06, + "loss": 1.9145, + "step": 20127 + }, + { + "epoch": 4.39, + "learning_rate": 2.8066419051780648e-06, + "loss": 1.2119, + "step": 20128 + }, + { + "epoch": 4.39, + "learning_rate": 2.806532663316583e-06, + "loss": 2.0793, + "step": 20129 + }, + { + "epoch": 4.39, + "learning_rate": 2.8064234214551016e-06, + "loss": 2.6864, + "step": 20130 + }, + { + "epoch": 4.39, + "learning_rate": 2.8063141795936206e-06, + "loss": 3.7692, + "step": 20131 + }, + { + "epoch": 4.39, + "learning_rate": 2.806204937732139e-06, + "loss": 2.9698, + "step": 20132 + }, + { + "epoch": 4.39, + "learning_rate": 2.8060956958706582e-06, + "loss": 2.2224, + "step": 20133 + }, + { + "epoch": 4.39, + "learning_rate": 2.8059864540091764e-06, + "loss": 1.0643, + "step": 20134 + }, + { + "epoch": 4.39, + "learning_rate": 2.805877212147695e-06, + "loss": 2.2391, + "step": 20135 + }, + { + "epoch": 4.39, + "learning_rate": 2.805767970286214e-06, + "loss": 2.1284, + "step": 20136 + }, + { + "epoch": 4.39, + "learning_rate": 2.8056587284247327e-06, + "loss": 0.4599, + "step": 20137 + }, + { + "epoch": 4.4, + "learning_rate": 2.805549486563251e-06, + "loss": 0.4495, + "step": 20138 + }, + { + "epoch": 4.4, + "learning_rate": 2.80544024470177e-06, + "loss": 2.3455, + "step": 20139 + }, + { + "epoch": 4.4, + "learning_rate": 2.8053310028402885e-06, + "loss": 2.8383, + "step": 20140 + }, + { + "epoch": 4.4, + "learning_rate": 2.8052217609788076e-06, + "loss": 2.3973, + "step": 20141 + }, + { + "epoch": 4.4, + "learning_rate": 2.805112519117326e-06, + "loss": 0.6246, + "step": 20142 + }, + { + "epoch": 4.4, + "learning_rate": 2.8050032772558444e-06, + "loss": 1.3362, + "step": 20143 + }, + { + "epoch": 4.4, + "learning_rate": 2.8048940353943634e-06, + "loss": 2.5322, + "step": 20144 + }, + { + "epoch": 4.4, + "learning_rate": 2.804784793532882e-06, + "loss": 0.6024, + "step": 20145 + }, + { + "epoch": 4.4, + "learning_rate": 2.8046755516714006e-06, + "loss": 2.5391, + "step": 20146 + }, + { + "epoch": 4.4, + "learning_rate": 2.8045663098099196e-06, + "loss": 1.4873, + "step": 20147 + }, + { + "epoch": 4.4, + "learning_rate": 2.804457067948438e-06, + "loss": 2.6696, + "step": 20148 + }, + { + "epoch": 4.4, + "learning_rate": 2.804347826086957e-06, + "loss": 1.691, + "step": 20149 + }, + { + "epoch": 4.4, + "learning_rate": 2.8042385842254755e-06, + "loss": 0.9176, + "step": 20150 + }, + { + "epoch": 4.4, + "learning_rate": 2.804129342363994e-06, + "loss": 1.5618, + "step": 20151 + }, + { + "epoch": 4.4, + "learning_rate": 2.804020100502513e-06, + "loss": 1.2682, + "step": 20152 + }, + { + "epoch": 4.4, + "learning_rate": 2.8039108586410313e-06, + "loss": 2.7185, + "step": 20153 + }, + { + "epoch": 4.4, + "learning_rate": 2.80380161677955e-06, + "loss": 2.0371, + "step": 20154 + }, + { + "epoch": 4.4, + "learning_rate": 2.803692374918069e-06, + "loss": 1.5177, + "step": 20155 + }, + { + "epoch": 4.4, + "learning_rate": 2.8035831330565876e-06, + "loss": 0.862, + "step": 20156 + }, + { + "epoch": 4.4, + "learning_rate": 2.8034738911951058e-06, + "loss": 1.0993, + "step": 20157 + }, + { + "epoch": 4.4, + "learning_rate": 2.803364649333625e-06, + "loss": 1.9188, + "step": 20158 + }, + { + "epoch": 4.4, + "learning_rate": 2.8032554074721434e-06, + "loss": 3.0811, + "step": 20159 + }, + { + "epoch": 4.4, + "learning_rate": 2.8031461656106624e-06, + "loss": 2.6752, + "step": 20160 + }, + { + "epoch": 4.4, + "learning_rate": 2.803036923749181e-06, + "loss": 2.4122, + "step": 20161 + }, + { + "epoch": 4.4, + "learning_rate": 2.8029276818876992e-06, + "loss": 2.1259, + "step": 20162 + }, + { + "epoch": 4.4, + "learning_rate": 2.8028184400262183e-06, + "loss": 2.3702, + "step": 20163 + }, + { + "epoch": 4.4, + "learning_rate": 2.802709198164737e-06, + "loss": 2.1122, + "step": 20164 + }, + { + "epoch": 4.4, + "learning_rate": 2.802599956303256e-06, + "loss": 0.9389, + "step": 20165 + }, + { + "epoch": 4.4, + "learning_rate": 2.8024907144417745e-06, + "loss": 0.6365, + "step": 20166 + }, + { + "epoch": 4.4, + "learning_rate": 2.8023814725802927e-06, + "loss": 1.8029, + "step": 20167 + }, + { + "epoch": 4.4, + "learning_rate": 2.8022722307188117e-06, + "loss": 1.6904, + "step": 20168 + }, + { + "epoch": 4.4, + "learning_rate": 2.8021629888573304e-06, + "loss": 0.6803, + "step": 20169 + }, + { + "epoch": 4.4, + "learning_rate": 2.8020537469958494e-06, + "loss": 3.0457, + "step": 20170 + }, + { + "epoch": 4.4, + "learning_rate": 2.8019445051343676e-06, + "loss": 2.8872, + "step": 20171 + }, + { + "epoch": 4.4, + "learning_rate": 2.801835263272886e-06, + "loss": 0.472, + "step": 20172 + }, + { + "epoch": 4.4, + "learning_rate": 2.8017260214114052e-06, + "loss": 1.4046, + "step": 20173 + }, + { + "epoch": 4.4, + "learning_rate": 2.801616779549924e-06, + "loss": 1.1658, + "step": 20174 + }, + { + "epoch": 4.4, + "learning_rate": 2.8015075376884424e-06, + "loss": 1.8928, + "step": 20175 + }, + { + "epoch": 4.4, + "learning_rate": 2.801398295826961e-06, + "loss": 0.8893, + "step": 20176 + }, + { + "epoch": 4.4, + "learning_rate": 2.8012890539654797e-06, + "loss": 1.1909, + "step": 20177 + }, + { + "epoch": 4.4, + "learning_rate": 2.8011798121039983e-06, + "loss": 0.6507, + "step": 20178 + }, + { + "epoch": 4.4, + "learning_rate": 2.8010705702425173e-06, + "loss": 0.9533, + "step": 20179 + }, + { + "epoch": 4.4, + "learning_rate": 2.800961328381036e-06, + "loss": 2.344, + "step": 20180 + }, + { + "epoch": 4.4, + "learning_rate": 2.800852086519554e-06, + "loss": 1.8532, + "step": 20181 + }, + { + "epoch": 4.4, + "learning_rate": 2.800742844658073e-06, + "loss": 1.8454, + "step": 20182 + }, + { + "epoch": 4.4, + "learning_rate": 2.8006336027965918e-06, + "loss": 1.85, + "step": 20183 + }, + { + "epoch": 4.41, + "learning_rate": 2.800524360935111e-06, + "loss": 0.9603, + "step": 20184 + }, + { + "epoch": 4.41, + "learning_rate": 2.8004151190736294e-06, + "loss": 2.1042, + "step": 20185 + }, + { + "epoch": 4.41, + "learning_rate": 2.8003058772121476e-06, + "loss": 1.0643, + "step": 20186 + }, + { + "epoch": 4.41, + "learning_rate": 2.8001966353506666e-06, + "loss": 0.9627, + "step": 20187 + }, + { + "epoch": 4.41, + "learning_rate": 2.8000873934891852e-06, + "loss": 1.7617, + "step": 20188 + }, + { + "epoch": 4.41, + "learning_rate": 2.7999781516277043e-06, + "loss": 1.7212, + "step": 20189 + }, + { + "epoch": 4.41, + "learning_rate": 2.7998689097662225e-06, + "loss": 3.1919, + "step": 20190 + }, + { + "epoch": 4.41, + "learning_rate": 2.799759667904741e-06, + "loss": 2.9401, + "step": 20191 + }, + { + "epoch": 4.41, + "learning_rate": 2.79965042604326e-06, + "loss": 1.8057, + "step": 20192 + }, + { + "epoch": 4.41, + "learning_rate": 2.7995411841817787e-06, + "loss": 2.6128, + "step": 20193 + }, + { + "epoch": 4.41, + "learning_rate": 2.7994319423202978e-06, + "loss": 0.971, + "step": 20194 + }, + { + "epoch": 4.41, + "learning_rate": 2.799322700458816e-06, + "loss": 1.2859, + "step": 20195 + }, + { + "epoch": 4.41, + "learning_rate": 2.7992134585973346e-06, + "loss": 1.6069, + "step": 20196 + }, + { + "epoch": 4.41, + "learning_rate": 2.7991042167358536e-06, + "loss": 1.5847, + "step": 20197 + }, + { + "epoch": 4.41, + "learning_rate": 2.798994974874372e-06, + "loss": 1.2913, + "step": 20198 + }, + { + "epoch": 4.41, + "learning_rate": 2.798885733012891e-06, + "loss": 2.1189, + "step": 20199 + }, + { + "epoch": 4.41, + "learning_rate": 2.7987764911514094e-06, + "loss": 1.078, + "step": 20200 + }, + { + "epoch": 4.41, + "learning_rate": 2.798667249289928e-06, + "loss": 3.0906, + "step": 20201 + }, + { + "epoch": 4.41, + "learning_rate": 2.798558007428447e-06, + "loss": 1.7267, + "step": 20202 + }, + { + "epoch": 4.41, + "learning_rate": 2.7984487655669657e-06, + "loss": 1.8316, + "step": 20203 + }, + { + "epoch": 4.41, + "learning_rate": 2.798339523705484e-06, + "loss": 0.9188, + "step": 20204 + }, + { + "epoch": 4.41, + "learning_rate": 2.798230281844003e-06, + "loss": 0.6225, + "step": 20205 + }, + { + "epoch": 4.41, + "learning_rate": 2.7981210399825215e-06, + "loss": 0.5414, + "step": 20206 + }, + { + "epoch": 4.41, + "learning_rate": 2.79801179812104e-06, + "loss": 2.7324, + "step": 20207 + }, + { + "epoch": 4.41, + "learning_rate": 2.797902556259559e-06, + "loss": 2.526, + "step": 20208 + }, + { + "epoch": 4.41, + "learning_rate": 2.7977933143980773e-06, + "loss": 0.9695, + "step": 20209 + }, + { + "epoch": 4.41, + "learning_rate": 2.797684072536596e-06, + "loss": 2.1679, + "step": 20210 + }, + { + "epoch": 4.41, + "learning_rate": 2.797574830675115e-06, + "loss": 2.1617, + "step": 20211 + }, + { + "epoch": 4.41, + "learning_rate": 2.7974655888136336e-06, + "loss": 2.2768, + "step": 20212 + }, + { + "epoch": 4.41, + "learning_rate": 2.7973563469521526e-06, + "loss": 0.8937, + "step": 20213 + }, + { + "epoch": 4.41, + "learning_rate": 2.797247105090671e-06, + "loss": 1.3718, + "step": 20214 + }, + { + "epoch": 4.41, + "learning_rate": 2.7971378632291894e-06, + "loss": 0.8317, + "step": 20215 + }, + { + "epoch": 4.41, + "learning_rate": 2.7970286213677085e-06, + "loss": 2.5685, + "step": 20216 + }, + { + "epoch": 4.41, + "learning_rate": 2.796919379506227e-06, + "loss": 0.6344, + "step": 20217 + }, + { + "epoch": 4.41, + "learning_rate": 2.796810137644746e-06, + "loss": 1.6156, + "step": 20218 + }, + { + "epoch": 4.41, + "learning_rate": 2.7967008957832643e-06, + "loss": 1.3225, + "step": 20219 + }, + { + "epoch": 4.41, + "learning_rate": 2.796591653921783e-06, + "loss": 2.6002, + "step": 20220 + }, + { + "epoch": 4.41, + "learning_rate": 2.796482412060302e-06, + "loss": 2.4164, + "step": 20221 + }, + { + "epoch": 4.41, + "learning_rate": 2.7963731701988206e-06, + "loss": 2.3945, + "step": 20222 + }, + { + "epoch": 4.41, + "learning_rate": 2.7962639283373387e-06, + "loss": 2.4695, + "step": 20223 + }, + { + "epoch": 4.41, + "learning_rate": 2.7961546864758578e-06, + "loss": 3.2452, + "step": 20224 + }, + { + "epoch": 4.41, + "learning_rate": 2.7960454446143764e-06, + "loss": 0.4939, + "step": 20225 + }, + { + "epoch": 4.41, + "learning_rate": 2.7959362027528954e-06, + "loss": 2.9984, + "step": 20226 + }, + { + "epoch": 4.41, + "learning_rate": 2.795826960891414e-06, + "loss": 0.7496, + "step": 20227 + }, + { + "epoch": 4.41, + "learning_rate": 2.7957177190299322e-06, + "loss": 2.4875, + "step": 20228 + }, + { + "epoch": 4.41, + "learning_rate": 2.7956084771684513e-06, + "loss": 1.3555, + "step": 20229 + }, + { + "epoch": 4.42, + "learning_rate": 2.79549923530697e-06, + "loss": 2.6931, + "step": 20230 + }, + { + "epoch": 4.42, + "learning_rate": 2.7953899934454885e-06, + "loss": 0.7244, + "step": 20231 + }, + { + "epoch": 4.42, + "learning_rate": 2.7952807515840075e-06, + "loss": 1.4185, + "step": 20232 + }, + { + "epoch": 4.42, + "learning_rate": 2.7951715097225257e-06, + "loss": 3.3066, + "step": 20233 + }, + { + "epoch": 4.42, + "learning_rate": 2.7950622678610443e-06, + "loss": 1.4176, + "step": 20234 + }, + { + "epoch": 4.42, + "learning_rate": 2.7949530259995633e-06, + "loss": 1.6903, + "step": 20235 + }, + { + "epoch": 4.42, + "learning_rate": 2.794843784138082e-06, + "loss": 1.2203, + "step": 20236 + }, + { + "epoch": 4.42, + "learning_rate": 2.7947345422766e-06, + "loss": 0.6206, + "step": 20237 + }, + { + "epoch": 4.42, + "learning_rate": 2.794625300415119e-06, + "loss": 1.6262, + "step": 20238 + }, + { + "epoch": 4.42, + "learning_rate": 2.7945160585536378e-06, + "loss": 1.4378, + "step": 20239 + }, + { + "epoch": 4.42, + "learning_rate": 2.794406816692157e-06, + "loss": 2.382, + "step": 20240 + }, + { + "epoch": 4.42, + "learning_rate": 2.7942975748306754e-06, + "loss": 1.563, + "step": 20241 + }, + { + "epoch": 4.42, + "learning_rate": 2.7941883329691936e-06, + "loss": 0.8232, + "step": 20242 + }, + { + "epoch": 4.42, + "learning_rate": 2.7940790911077127e-06, + "loss": 2.0738, + "step": 20243 + }, + { + "epoch": 4.42, + "learning_rate": 2.7939698492462313e-06, + "loss": 3.0718, + "step": 20244 + }, + { + "epoch": 4.42, + "learning_rate": 2.7938606073847503e-06, + "loss": 2.5358, + "step": 20245 + }, + { + "epoch": 4.42, + "learning_rate": 2.793751365523269e-06, + "loss": 1.5371, + "step": 20246 + }, + { + "epoch": 4.42, + "learning_rate": 2.793642123661787e-06, + "loss": 1.4603, + "step": 20247 + }, + { + "epoch": 4.42, + "learning_rate": 2.793532881800306e-06, + "loss": 2.7392, + "step": 20248 + }, + { + "epoch": 4.42, + "learning_rate": 2.7934236399388247e-06, + "loss": 2.5426, + "step": 20249 + }, + { + "epoch": 4.42, + "learning_rate": 2.7933143980773438e-06, + "loss": 1.6089, + "step": 20250 + }, + { + "epoch": 4.42, + "learning_rate": 2.7932051562158624e-06, + "loss": 2.2803, + "step": 20251 + }, + { + "epoch": 4.42, + "learning_rate": 2.7930959143543806e-06, + "loss": 1.4585, + "step": 20252 + }, + { + "epoch": 4.42, + "learning_rate": 2.7929866724928996e-06, + "loss": 1.2948, + "step": 20253 + }, + { + "epoch": 4.42, + "learning_rate": 2.7928774306314182e-06, + "loss": 1.6217, + "step": 20254 + }, + { + "epoch": 4.42, + "learning_rate": 2.7927681887699373e-06, + "loss": 1.2902, + "step": 20255 + }, + { + "epoch": 4.42, + "learning_rate": 2.7926589469084554e-06, + "loss": 1.9412, + "step": 20256 + }, + { + "epoch": 4.42, + "learning_rate": 2.792549705046974e-06, + "loss": 1.328, + "step": 20257 + }, + { + "epoch": 4.42, + "learning_rate": 2.792440463185493e-06, + "loss": 1.4023, + "step": 20258 + }, + { + "epoch": 4.42, + "learning_rate": 2.7923312213240117e-06, + "loss": 1.7104, + "step": 20259 + }, + { + "epoch": 4.42, + "learning_rate": 2.7922219794625303e-06, + "loss": 2.3282, + "step": 20260 + }, + { + "epoch": 4.42, + "learning_rate": 2.792112737601049e-06, + "loss": 1.2229, + "step": 20261 + }, + { + "epoch": 4.42, + "learning_rate": 2.7920034957395675e-06, + "loss": 1.0479, + "step": 20262 + }, + { + "epoch": 4.42, + "learning_rate": 2.791894253878086e-06, + "loss": 2.5364, + "step": 20263 + }, + { + "epoch": 4.42, + "learning_rate": 2.791785012016605e-06, + "loss": 0.8275, + "step": 20264 + }, + { + "epoch": 4.42, + "learning_rate": 2.791675770155124e-06, + "loss": 3.3271, + "step": 20265 + }, + { + "epoch": 4.42, + "learning_rate": 2.791566528293642e-06, + "loss": 0.9915, + "step": 20266 + }, + { + "epoch": 4.42, + "learning_rate": 2.791457286432161e-06, + "loss": 0.3785, + "step": 20267 + }, + { + "epoch": 4.42, + "learning_rate": 2.7913480445706796e-06, + "loss": 1.2639, + "step": 20268 + }, + { + "epoch": 4.42, + "learning_rate": 2.7912388027091987e-06, + "loss": 2.3654, + "step": 20269 + }, + { + "epoch": 4.42, + "learning_rate": 2.791129560847717e-06, + "loss": 1.5908, + "step": 20270 + }, + { + "epoch": 4.42, + "learning_rate": 2.7910203189862355e-06, + "loss": 1.911, + "step": 20271 + }, + { + "epoch": 4.42, + "learning_rate": 2.7909110771247545e-06, + "loss": 1.8359, + "step": 20272 + }, + { + "epoch": 4.42, + "learning_rate": 2.790801835263273e-06, + "loss": 1.4539, + "step": 20273 + }, + { + "epoch": 4.42, + "learning_rate": 2.790692593401792e-06, + "loss": 0.9336, + "step": 20274 + }, + { + "epoch": 4.42, + "learning_rate": 2.7905833515403103e-06, + "loss": 2.6325, + "step": 20275 + }, + { + "epoch": 4.43, + "learning_rate": 2.790474109678829e-06, + "loss": 0.7197, + "step": 20276 + }, + { + "epoch": 4.43, + "learning_rate": 2.790364867817348e-06, + "loss": 0.617, + "step": 20277 + }, + { + "epoch": 4.43, + "learning_rate": 2.7902556259558666e-06, + "loss": 3.0017, + "step": 20278 + }, + { + "epoch": 4.43, + "learning_rate": 2.7901463840943856e-06, + "loss": 1.4586, + "step": 20279 + }, + { + "epoch": 4.43, + "learning_rate": 2.790037142232904e-06, + "loss": 2.2063, + "step": 20280 + }, + { + "epoch": 4.43, + "learning_rate": 2.7899279003714224e-06, + "loss": 1.2913, + "step": 20281 + }, + { + "epoch": 4.43, + "learning_rate": 2.7898186585099414e-06, + "loss": 2.9434, + "step": 20282 + }, + { + "epoch": 4.43, + "learning_rate": 2.78970941664846e-06, + "loss": 1.7134, + "step": 20283 + }, + { + "epoch": 4.43, + "learning_rate": 2.7896001747869787e-06, + "loss": 2.2046, + "step": 20284 + }, + { + "epoch": 4.43, + "learning_rate": 2.7894909329254973e-06, + "loss": 1.1705, + "step": 20285 + }, + { + "epoch": 4.43, + "learning_rate": 2.789381691064016e-06, + "loss": 1.2533, + "step": 20286 + }, + { + "epoch": 4.43, + "learning_rate": 2.7892724492025345e-06, + "loss": 0.7977, + "step": 20287 + }, + { + "epoch": 4.43, + "learning_rate": 2.7891632073410535e-06, + "loss": 1.1172, + "step": 20288 + }, + { + "epoch": 4.43, + "learning_rate": 2.7890539654795717e-06, + "loss": 2.7519, + "step": 20289 + }, + { + "epoch": 4.43, + "learning_rate": 2.7889447236180903e-06, + "loss": 1.1782, + "step": 20290 + }, + { + "epoch": 4.43, + "learning_rate": 2.7888354817566094e-06, + "loss": 1.2912, + "step": 20291 + }, + { + "epoch": 4.43, + "learning_rate": 2.788726239895128e-06, + "loss": 0.7316, + "step": 20292 + }, + { + "epoch": 4.43, + "learning_rate": 2.788616998033647e-06, + "loss": 2.4411, + "step": 20293 + }, + { + "epoch": 4.43, + "learning_rate": 2.788507756172165e-06, + "loss": 0.8506, + "step": 20294 + }, + { + "epoch": 4.43, + "learning_rate": 2.788398514310684e-06, + "loss": 1.6983, + "step": 20295 + }, + { + "epoch": 4.43, + "learning_rate": 2.788289272449203e-06, + "loss": 2.3899, + "step": 20296 + }, + { + "epoch": 4.43, + "learning_rate": 2.7881800305877215e-06, + "loss": 0.9953, + "step": 20297 + }, + { + "epoch": 4.43, + "learning_rate": 2.7880707887262405e-06, + "loss": 0.5758, + "step": 20298 + }, + { + "epoch": 4.43, + "learning_rate": 2.7879615468647587e-06, + "loss": 1.2033, + "step": 20299 + }, + { + "epoch": 4.43, + "learning_rate": 2.7878523050032773e-06, + "loss": 1.9109, + "step": 20300 + }, + { + "epoch": 4.43, + "learning_rate": 2.7877430631417963e-06, + "loss": 1.4363, + "step": 20301 + }, + { + "epoch": 4.43, + "learning_rate": 2.787633821280315e-06, + "loss": 1.5365, + "step": 20302 + }, + { + "epoch": 4.43, + "learning_rate": 2.787524579418833e-06, + "loss": 2.9002, + "step": 20303 + }, + { + "epoch": 4.43, + "learning_rate": 2.787415337557352e-06, + "loss": 3.2784, + "step": 20304 + }, + { + "epoch": 4.43, + "learning_rate": 2.7873060956958708e-06, + "loss": 1.8901, + "step": 20305 + }, + { + "epoch": 4.43, + "learning_rate": 2.78719685383439e-06, + "loss": 1.087, + "step": 20306 + }, + { + "epoch": 4.43, + "learning_rate": 2.7870876119729084e-06, + "loss": 2.1199, + "step": 20307 + }, + { + "epoch": 4.43, + "learning_rate": 2.7869783701114266e-06, + "loss": 1.0344, + "step": 20308 + }, + { + "epoch": 4.43, + "learning_rate": 2.7868691282499456e-06, + "loss": 1.6095, + "step": 20309 + }, + { + "epoch": 4.43, + "learning_rate": 2.7867598863884642e-06, + "loss": 2.3748, + "step": 20310 + }, + { + "epoch": 4.43, + "learning_rate": 2.7866506445269833e-06, + "loss": 2.5203, + "step": 20311 + }, + { + "epoch": 4.43, + "learning_rate": 2.786541402665502e-06, + "loss": 0.9064, + "step": 20312 + }, + { + "epoch": 4.43, + "learning_rate": 2.78643216080402e-06, + "loss": 1.2783, + "step": 20313 + }, + { + "epoch": 4.43, + "learning_rate": 2.786322918942539e-06, + "loss": 1.4062, + "step": 20314 + }, + { + "epoch": 4.43, + "learning_rate": 2.7862136770810577e-06, + "loss": 1.0997, + "step": 20315 + }, + { + "epoch": 4.43, + "learning_rate": 2.7861044352195763e-06, + "loss": 2.2581, + "step": 20316 + }, + { + "epoch": 4.43, + "learning_rate": 2.7859951933580954e-06, + "loss": 1.4066, + "step": 20317 + }, + { + "epoch": 4.43, + "learning_rate": 2.7858859514966136e-06, + "loss": 1.5256, + "step": 20318 + }, + { + "epoch": 4.43, + "learning_rate": 2.785776709635132e-06, + "loss": 1.9076, + "step": 20319 + }, + { + "epoch": 4.43, + "learning_rate": 2.785667467773651e-06, + "loss": 1.9908, + "step": 20320 + }, + { + "epoch": 4.43, + "learning_rate": 2.78555822591217e-06, + "loss": 1.8653, + "step": 20321 + }, + { + "epoch": 4.44, + "learning_rate": 2.785448984050688e-06, + "loss": 1.4792, + "step": 20322 + }, + { + "epoch": 4.44, + "learning_rate": 2.785339742189207e-06, + "loss": 2.3395, + "step": 20323 + }, + { + "epoch": 4.44, + "learning_rate": 2.7852305003277257e-06, + "loss": 0.8316, + "step": 20324 + }, + { + "epoch": 4.44, + "learning_rate": 2.7851212584662447e-06, + "loss": 2.8081, + "step": 20325 + }, + { + "epoch": 4.44, + "learning_rate": 2.7850120166047633e-06, + "loss": 2.1937, + "step": 20326 + }, + { + "epoch": 4.44, + "learning_rate": 2.7849027747432815e-06, + "loss": 1.3097, + "step": 20327 + }, + { + "epoch": 4.44, + "learning_rate": 2.7847935328818005e-06, + "loss": 2.8474, + "step": 20328 + }, + { + "epoch": 4.44, + "learning_rate": 2.784684291020319e-06, + "loss": 0.6094, + "step": 20329 + }, + { + "epoch": 4.44, + "learning_rate": 2.784575049158838e-06, + "loss": 1.9378, + "step": 20330 + }, + { + "epoch": 4.44, + "learning_rate": 2.7844658072973568e-06, + "loss": 1.3161, + "step": 20331 + }, + { + "epoch": 4.44, + "learning_rate": 2.784356565435875e-06, + "loss": 1.4171, + "step": 20332 + }, + { + "epoch": 4.44, + "learning_rate": 2.784247323574394e-06, + "loss": 2.1007, + "step": 20333 + }, + { + "epoch": 4.44, + "learning_rate": 2.7841380817129126e-06, + "loss": 0.9353, + "step": 20334 + }, + { + "epoch": 4.44, + "learning_rate": 2.7840288398514316e-06, + "loss": 0.8858, + "step": 20335 + }, + { + "epoch": 4.44, + "learning_rate": 2.78391959798995e-06, + "loss": 0.7414, + "step": 20336 + }, + { + "epoch": 4.44, + "learning_rate": 2.7838103561284684e-06, + "loss": 0.5432, + "step": 20337 + }, + { + "epoch": 4.44, + "learning_rate": 2.7837011142669875e-06, + "loss": 1.0299, + "step": 20338 + }, + { + "epoch": 4.44, + "learning_rate": 2.783591872405506e-06, + "loss": 1.2479, + "step": 20339 + }, + { + "epoch": 4.44, + "learning_rate": 2.7834826305440247e-06, + "loss": 2.7067, + "step": 20340 + }, + { + "epoch": 4.44, + "learning_rate": 2.7833733886825433e-06, + "loss": 2.4647, + "step": 20341 + }, + { + "epoch": 4.44, + "learning_rate": 2.783264146821062e-06, + "loss": 0.7591, + "step": 20342 + }, + { + "epoch": 4.44, + "learning_rate": 2.7831549049595805e-06, + "loss": 1.3248, + "step": 20343 + }, + { + "epoch": 4.44, + "learning_rate": 2.7830456630980996e-06, + "loss": 0.7058, + "step": 20344 + }, + { + "epoch": 4.44, + "learning_rate": 2.782936421236618e-06, + "loss": 1.4788, + "step": 20345 + }, + { + "epoch": 4.44, + "learning_rate": 2.7828271793751368e-06, + "loss": 1.3394, + "step": 20346 + }, + { + "epoch": 4.44, + "learning_rate": 2.7827179375136554e-06, + "loss": 1.4042, + "step": 20347 + }, + { + "epoch": 4.44, + "learning_rate": 2.782608695652174e-06, + "loss": 2.9493, + "step": 20348 + }, + { + "epoch": 4.44, + "learning_rate": 2.782499453790693e-06, + "loss": 1.5946, + "step": 20349 + }, + { + "epoch": 4.44, + "learning_rate": 2.7823902119292117e-06, + "loss": 3.1179, + "step": 20350 + }, + { + "epoch": 4.44, + "learning_rate": 2.78228097006773e-06, + "loss": 1.1599, + "step": 20351 + }, + { + "epoch": 4.44, + "learning_rate": 2.782171728206249e-06, + "loss": 3.4861, + "step": 20352 + }, + { + "epoch": 4.44, + "learning_rate": 2.7820624863447675e-06, + "loss": 1.3058, + "step": 20353 + }, + { + "epoch": 4.44, + "learning_rate": 2.7819532444832865e-06, + "loss": 1.6781, + "step": 20354 + }, + { + "epoch": 4.44, + "learning_rate": 2.7818440026218047e-06, + "loss": 1.6104, + "step": 20355 + }, + { + "epoch": 4.44, + "learning_rate": 2.7817347607603233e-06, + "loss": 1.1466, + "step": 20356 + }, + { + "epoch": 4.44, + "learning_rate": 2.7816255188988424e-06, + "loss": 1.677, + "step": 20357 + }, + { + "epoch": 4.44, + "learning_rate": 2.781516277037361e-06, + "loss": 1.8852, + "step": 20358 + }, + { + "epoch": 4.44, + "learning_rate": 2.78140703517588e-06, + "loss": 1.3009, + "step": 20359 + }, + { + "epoch": 4.44, + "learning_rate": 2.781297793314398e-06, + "loss": 1.6807, + "step": 20360 + }, + { + "epoch": 4.44, + "learning_rate": 2.781188551452917e-06, + "loss": 1.6397, + "step": 20361 + }, + { + "epoch": 4.44, + "learning_rate": 2.781079309591436e-06, + "loss": 0.991, + "step": 20362 + }, + { + "epoch": 4.44, + "learning_rate": 2.7809700677299544e-06, + "loss": 1.4261, + "step": 20363 + }, + { + "epoch": 4.44, + "learning_rate": 2.7808608258684735e-06, + "loss": 1.7528, + "step": 20364 + }, + { + "epoch": 4.44, + "learning_rate": 2.7807515840069917e-06, + "loss": 1.3221, + "step": 20365 + }, + { + "epoch": 4.44, + "learning_rate": 2.7806423421455103e-06, + "loss": 0.1976, + "step": 20366 + }, + { + "epoch": 4.45, + "learning_rate": 2.7805331002840293e-06, + "loss": 1.4019, + "step": 20367 + }, + { + "epoch": 4.45, + "learning_rate": 2.780423858422548e-06, + "loss": 2.4609, + "step": 20368 + }, + { + "epoch": 4.45, + "learning_rate": 2.780314616561066e-06, + "loss": 2.166, + "step": 20369 + }, + { + "epoch": 4.45, + "learning_rate": 2.780205374699585e-06, + "loss": 2.5146, + "step": 20370 + }, + { + "epoch": 4.45, + "learning_rate": 2.7800961328381038e-06, + "loss": 1.6364, + "step": 20371 + }, + { + "epoch": 4.45, + "learning_rate": 2.7799868909766224e-06, + "loss": 2.2457, + "step": 20372 + }, + { + "epoch": 4.45, + "learning_rate": 2.7798776491151414e-06, + "loss": 0.8294, + "step": 20373 + }, + { + "epoch": 4.45, + "learning_rate": 2.7797684072536596e-06, + "loss": 2.2165, + "step": 20374 + }, + { + "epoch": 4.45, + "learning_rate": 2.779659165392178e-06, + "loss": 1.5663, + "step": 20375 + }, + { + "epoch": 4.45, + "learning_rate": 2.7795499235306972e-06, + "loss": 1.9671, + "step": 20376 + }, + { + "epoch": 4.45, + "learning_rate": 2.779440681669216e-06, + "loss": 1.8476, + "step": 20377 + }, + { + "epoch": 4.45, + "learning_rate": 2.779331439807735e-06, + "loss": 2.7322, + "step": 20378 + }, + { + "epoch": 4.45, + "learning_rate": 2.779222197946253e-06, + "loss": 2.0782, + "step": 20379 + }, + { + "epoch": 4.45, + "learning_rate": 2.7791129560847717e-06, + "loss": 4.6392, + "step": 20380 + }, + { + "epoch": 4.45, + "learning_rate": 2.7790037142232907e-06, + "loss": 1.336, + "step": 20381 + }, + { + "epoch": 4.45, + "learning_rate": 2.7788944723618093e-06, + "loss": 2.7695, + "step": 20382 + }, + { + "epoch": 4.45, + "learning_rate": 2.7787852305003284e-06, + "loss": 1.1668, + "step": 20383 + }, + { + "epoch": 4.45, + "learning_rate": 2.7786759886388465e-06, + "loss": 0.999, + "step": 20384 + }, + { + "epoch": 4.45, + "learning_rate": 2.778566746777365e-06, + "loss": 3.4428, + "step": 20385 + }, + { + "epoch": 4.45, + "learning_rate": 2.778457504915884e-06, + "loss": 1.2274, + "step": 20386 + }, + { + "epoch": 4.45, + "learning_rate": 2.778348263054403e-06, + "loss": 1.5772, + "step": 20387 + }, + { + "epoch": 4.45, + "learning_rate": 2.778239021192921e-06, + "loss": 1.954, + "step": 20388 + }, + { + "epoch": 4.45, + "learning_rate": 2.77812977933144e-06, + "loss": 0.653, + "step": 20389 + }, + { + "epoch": 4.45, + "learning_rate": 2.7780205374699586e-06, + "loss": 0.4316, + "step": 20390 + }, + { + "epoch": 4.45, + "learning_rate": 2.7779112956084777e-06, + "loss": 1.4592, + "step": 20391 + }, + { + "epoch": 4.45, + "learning_rate": 2.7778020537469963e-06, + "loss": 1.3035, + "step": 20392 + }, + { + "epoch": 4.45, + "learning_rate": 2.7776928118855145e-06, + "loss": 1.2094, + "step": 20393 + }, + { + "epoch": 4.45, + "learning_rate": 2.7775835700240335e-06, + "loss": 2.7146, + "step": 20394 + }, + { + "epoch": 4.45, + "learning_rate": 2.777474328162552e-06, + "loss": 1.5207, + "step": 20395 + }, + { + "epoch": 4.45, + "learning_rate": 2.7773650863010707e-06, + "loss": 1.8818, + "step": 20396 + }, + { + "epoch": 4.45, + "learning_rate": 2.7772558444395898e-06, + "loss": 1.5863, + "step": 20397 + }, + { + "epoch": 4.45, + "learning_rate": 2.777146602578108e-06, + "loss": 1.3171, + "step": 20398 + }, + { + "epoch": 4.45, + "learning_rate": 2.777037360716627e-06, + "loss": 1.3589, + "step": 20399 + }, + { + "epoch": 4.45, + "learning_rate": 2.7769281188551456e-06, + "loss": 1.0639, + "step": 20400 + }, + { + "epoch": 4.45, + "learning_rate": 2.776818876993664e-06, + "loss": 3.4149, + "step": 20401 + }, + { + "epoch": 4.45, + "learning_rate": 2.776709635132183e-06, + "loss": 1.6698, + "step": 20402 + }, + { + "epoch": 4.45, + "learning_rate": 2.7766003932707014e-06, + "loss": 1.8989, + "step": 20403 + }, + { + "epoch": 4.45, + "learning_rate": 2.77649115140922e-06, + "loss": 1.3592, + "step": 20404 + }, + { + "epoch": 4.45, + "learning_rate": 2.776381909547739e-06, + "loss": 1.132, + "step": 20405 + }, + { + "epoch": 4.45, + "learning_rate": 2.7762726676862577e-06, + "loss": 2.7352, + "step": 20406 + }, + { + "epoch": 4.45, + "learning_rate": 2.776163425824776e-06, + "loss": 0.9865, + "step": 20407 + }, + { + "epoch": 4.45, + "learning_rate": 2.776054183963295e-06, + "loss": 2.577, + "step": 20408 + }, + { + "epoch": 4.45, + "learning_rate": 2.7759449421018135e-06, + "loss": 0.8601, + "step": 20409 + }, + { + "epoch": 4.45, + "learning_rate": 2.7758357002403325e-06, + "loss": 1.1259, + "step": 20410 + }, + { + "epoch": 4.45, + "learning_rate": 2.775726458378851e-06, + "loss": 1.4792, + "step": 20411 + }, + { + "epoch": 4.45, + "learning_rate": 2.7756172165173693e-06, + "loss": 2.7257, + "step": 20412 + }, + { + "epoch": 4.46, + "learning_rate": 2.7755079746558884e-06, + "loss": 1.81, + "step": 20413 + }, + { + "epoch": 4.46, + "learning_rate": 2.775398732794407e-06, + "loss": 0.9569, + "step": 20414 + }, + { + "epoch": 4.46, + "learning_rate": 2.775289490932926e-06, + "loss": 1.4759, + "step": 20415 + }, + { + "epoch": 4.46, + "learning_rate": 2.7751802490714446e-06, + "loss": 0.6438, + "step": 20416 + }, + { + "epoch": 4.46, + "learning_rate": 2.775071007209963e-06, + "loss": 1.3507, + "step": 20417 + }, + { + "epoch": 4.46, + "learning_rate": 2.774961765348482e-06, + "loss": 1.6153, + "step": 20418 + }, + { + "epoch": 4.46, + "learning_rate": 2.7748525234870005e-06, + "loss": 1.1392, + "step": 20419 + }, + { + "epoch": 4.46, + "learning_rate": 2.7747432816255195e-06, + "loss": 1.2932, + "step": 20420 + }, + { + "epoch": 4.46, + "learning_rate": 2.7746340397640377e-06, + "loss": 1.5724, + "step": 20421 + }, + { + "epoch": 4.46, + "learning_rate": 2.7745247979025563e-06, + "loss": 2.2458, + "step": 20422 + }, + { + "epoch": 4.46, + "learning_rate": 2.7744155560410753e-06, + "loss": 1.5263, + "step": 20423 + }, + { + "epoch": 4.46, + "learning_rate": 2.774306314179594e-06, + "loss": 1.8038, + "step": 20424 + }, + { + "epoch": 4.46, + "learning_rate": 2.7741970723181126e-06, + "loss": 2.2735, + "step": 20425 + }, + { + "epoch": 4.46, + "learning_rate": 2.774087830456631e-06, + "loss": 0.8037, + "step": 20426 + }, + { + "epoch": 4.46, + "learning_rate": 2.7739785885951498e-06, + "loss": 1.0466, + "step": 20427 + }, + { + "epoch": 4.46, + "learning_rate": 2.7738693467336684e-06, + "loss": 2.6854, + "step": 20428 + }, + { + "epoch": 4.46, + "learning_rate": 2.7737601048721874e-06, + "loss": 2.5986, + "step": 20429 + }, + { + "epoch": 4.46, + "learning_rate": 2.773650863010706e-06, + "loss": 1.5011, + "step": 20430 + }, + { + "epoch": 4.46, + "learning_rate": 2.7735416211492242e-06, + "loss": 0.9804, + "step": 20431 + }, + { + "epoch": 4.46, + "learning_rate": 2.7734323792877433e-06, + "loss": 1.4849, + "step": 20432 + }, + { + "epoch": 4.46, + "learning_rate": 2.773323137426262e-06, + "loss": 0.7451, + "step": 20433 + }, + { + "epoch": 4.46, + "learning_rate": 2.773213895564781e-06, + "loss": 1.4633, + "step": 20434 + }, + { + "epoch": 4.46, + "learning_rate": 2.773104653703299e-06, + "loss": 2.0338, + "step": 20435 + }, + { + "epoch": 4.46, + "learning_rate": 2.7729954118418177e-06, + "loss": 1.6732, + "step": 20436 + }, + { + "epoch": 4.46, + "learning_rate": 2.7728861699803367e-06, + "loss": 2.9272, + "step": 20437 + }, + { + "epoch": 4.46, + "learning_rate": 2.7727769281188553e-06, + "loss": 2.6839, + "step": 20438 + }, + { + "epoch": 4.46, + "learning_rate": 2.7726676862573744e-06, + "loss": 1.4196, + "step": 20439 + }, + { + "epoch": 4.46, + "learning_rate": 2.7725584443958926e-06, + "loss": 2.8366, + "step": 20440 + }, + { + "epoch": 4.46, + "learning_rate": 2.772449202534411e-06, + "loss": 1.1985, + "step": 20441 + }, + { + "epoch": 4.46, + "learning_rate": 2.7723399606729302e-06, + "loss": 2.2576, + "step": 20442 + }, + { + "epoch": 4.46, + "learning_rate": 2.772230718811449e-06, + "loss": 2.6463, + "step": 20443 + }, + { + "epoch": 4.46, + "learning_rate": 2.772121476949968e-06, + "loss": 0.5171, + "step": 20444 + }, + { + "epoch": 4.46, + "learning_rate": 2.772012235088486e-06, + "loss": 1.2673, + "step": 20445 + }, + { + "epoch": 4.46, + "learning_rate": 2.7719029932270047e-06, + "loss": 2.2406, + "step": 20446 + }, + { + "epoch": 4.46, + "learning_rate": 2.7717937513655237e-06, + "loss": 1.7963, + "step": 20447 + }, + { + "epoch": 4.46, + "learning_rate": 2.7716845095040423e-06, + "loss": 1.66, + "step": 20448 + }, + { + "epoch": 4.46, + "learning_rate": 2.771575267642561e-06, + "loss": 1.9745, + "step": 20449 + }, + { + "epoch": 4.46, + "learning_rate": 2.7714660257810795e-06, + "loss": 1.4466, + "step": 20450 + }, + { + "epoch": 4.46, + "learning_rate": 2.771356783919598e-06, + "loss": 1.3575, + "step": 20451 + }, + { + "epoch": 4.46, + "learning_rate": 2.771247542058117e-06, + "loss": 1.8937, + "step": 20452 + }, + { + "epoch": 4.46, + "learning_rate": 2.7711383001966358e-06, + "loss": 1.1802, + "step": 20453 + }, + { + "epoch": 4.46, + "learning_rate": 2.771029058335154e-06, + "loss": 1.7288, + "step": 20454 + }, + { + "epoch": 4.46, + "learning_rate": 2.770919816473673e-06, + "loss": 1.7282, + "step": 20455 + }, + { + "epoch": 4.46, + "learning_rate": 2.7708105746121916e-06, + "loss": 1.6032, + "step": 20456 + }, + { + "epoch": 4.46, + "learning_rate": 2.7707013327507102e-06, + "loss": 1.3716, + "step": 20457 + }, + { + "epoch": 4.46, + "learning_rate": 2.7705920908892293e-06, + "loss": 2.3079, + "step": 20458 + }, + { + "epoch": 4.47, + "learning_rate": 2.7704828490277475e-06, + "loss": 0.8271, + "step": 20459 + }, + { + "epoch": 4.47, + "learning_rate": 2.770373607166266e-06, + "loss": 2.0038, + "step": 20460 + }, + { + "epoch": 4.47, + "learning_rate": 2.770264365304785e-06, + "loss": 2.4562, + "step": 20461 + }, + { + "epoch": 4.47, + "learning_rate": 2.7701551234433037e-06, + "loss": 2.663, + "step": 20462 + }, + { + "epoch": 4.47, + "learning_rate": 2.7700458815818227e-06, + "loss": 1.8594, + "step": 20463 + }, + { + "epoch": 4.47, + "learning_rate": 2.769936639720341e-06, + "loss": 1.9773, + "step": 20464 + }, + { + "epoch": 4.47, + "learning_rate": 2.7698273978588595e-06, + "loss": 0.9607, + "step": 20465 + }, + { + "epoch": 4.47, + "learning_rate": 2.7697181559973786e-06, + "loss": 2.2484, + "step": 20466 + }, + { + "epoch": 4.47, + "learning_rate": 2.769608914135897e-06, + "loss": 2.3059, + "step": 20467 + }, + { + "epoch": 4.47, + "learning_rate": 2.7694996722744154e-06, + "loss": 1.8411, + "step": 20468 + }, + { + "epoch": 4.47, + "learning_rate": 2.7693904304129344e-06, + "loss": 1.2687, + "step": 20469 + }, + { + "epoch": 4.47, + "learning_rate": 2.769281188551453e-06, + "loss": 0.6487, + "step": 20470 + }, + { + "epoch": 4.47, + "learning_rate": 2.769171946689972e-06, + "loss": 2.0802, + "step": 20471 + }, + { + "epoch": 4.47, + "learning_rate": 2.7690627048284907e-06, + "loss": 1.2306, + "step": 20472 + }, + { + "epoch": 4.47, + "learning_rate": 2.768953462967009e-06, + "loss": 1.7971, + "step": 20473 + }, + { + "epoch": 4.47, + "learning_rate": 2.768844221105528e-06, + "loss": 1.0196, + "step": 20474 + }, + { + "epoch": 4.47, + "learning_rate": 2.7687349792440465e-06, + "loss": 1.7844, + "step": 20475 + }, + { + "epoch": 4.47, + "learning_rate": 2.7686257373825655e-06, + "loss": 2.9728, + "step": 20476 + }, + { + "epoch": 4.47, + "learning_rate": 2.768516495521084e-06, + "loss": 4.4917, + "step": 20477 + }, + { + "epoch": 4.47, + "learning_rate": 2.7684072536596023e-06, + "loss": 2.4054, + "step": 20478 + }, + { + "epoch": 4.47, + "learning_rate": 2.7682980117981214e-06, + "loss": 0.629, + "step": 20479 + }, + { + "epoch": 4.47, + "learning_rate": 2.76818876993664e-06, + "loss": 1.6687, + "step": 20480 + }, + { + "epoch": 4.47, + "learning_rate": 2.7680795280751586e-06, + "loss": 1.62, + "step": 20481 + }, + { + "epoch": 4.47, + "learning_rate": 2.7679702862136776e-06, + "loss": 2.9905, + "step": 20482 + }, + { + "epoch": 4.47, + "learning_rate": 2.767861044352196e-06, + "loss": 0.705, + "step": 20483 + }, + { + "epoch": 4.47, + "learning_rate": 2.7677518024907144e-06, + "loss": 1.3178, + "step": 20484 + }, + { + "epoch": 4.47, + "learning_rate": 2.7676425606292335e-06, + "loss": 0.773, + "step": 20485 + }, + { + "epoch": 4.47, + "learning_rate": 2.767533318767752e-06, + "loss": 2.4218, + "step": 20486 + }, + { + "epoch": 4.47, + "learning_rate": 2.7674240769062703e-06, + "loss": 2.4406, + "step": 20487 + }, + { + "epoch": 4.47, + "learning_rate": 2.7673148350447893e-06, + "loss": 2.1323, + "step": 20488 + }, + { + "epoch": 4.47, + "learning_rate": 2.767205593183308e-06, + "loss": 1.1228, + "step": 20489 + }, + { + "epoch": 4.47, + "learning_rate": 2.767096351321827e-06, + "loss": 1.566, + "step": 20490 + }, + { + "epoch": 4.47, + "learning_rate": 2.7669871094603455e-06, + "loss": 1.2088, + "step": 20491 + }, + { + "epoch": 4.47, + "learning_rate": 2.7668778675988637e-06, + "loss": 1.7394, + "step": 20492 + }, + { + "epoch": 4.47, + "learning_rate": 2.7667686257373828e-06, + "loss": 1.5341, + "step": 20493 + }, + { + "epoch": 4.47, + "learning_rate": 2.7666593838759014e-06, + "loss": 2.5799, + "step": 20494 + }, + { + "epoch": 4.47, + "learning_rate": 2.7665501420144204e-06, + "loss": 1.8102, + "step": 20495 + }, + { + "epoch": 4.47, + "learning_rate": 2.766440900152939e-06, + "loss": 1.9786, + "step": 20496 + }, + { + "epoch": 4.47, + "learning_rate": 2.766331658291457e-06, + "loss": 1.9048, + "step": 20497 + }, + { + "epoch": 4.47, + "learning_rate": 2.7662224164299762e-06, + "loss": 1.0012, + "step": 20498 + }, + { + "epoch": 4.47, + "learning_rate": 2.766113174568495e-06, + "loss": 1.1143, + "step": 20499 + }, + { + "epoch": 4.47, + "learning_rate": 2.766003932707014e-06, + "loss": 1.4561, + "step": 20500 + }, + { + "epoch": 4.47, + "learning_rate": 2.765894690845532e-06, + "loss": 1.4676, + "step": 20501 + }, + { + "epoch": 4.47, + "learning_rate": 2.7657854489840507e-06, + "loss": 1.7779, + "step": 20502 + }, + { + "epoch": 4.47, + "learning_rate": 2.7656762071225697e-06, + "loss": 2.058, + "step": 20503 + }, + { + "epoch": 4.47, + "learning_rate": 2.7655669652610883e-06, + "loss": 1.3966, + "step": 20504 + }, + { + "epoch": 4.48, + "learning_rate": 2.7654577233996074e-06, + "loss": 1.134, + "step": 20505 + }, + { + "epoch": 4.48, + "learning_rate": 2.7653484815381256e-06, + "loss": 1.7091, + "step": 20506 + }, + { + "epoch": 4.48, + "learning_rate": 2.765239239676644e-06, + "loss": 1.2829, + "step": 20507 + }, + { + "epoch": 4.48, + "learning_rate": 2.765129997815163e-06, + "loss": 1.0962, + "step": 20508 + }, + { + "epoch": 4.48, + "learning_rate": 2.765020755953682e-06, + "loss": 1.3036, + "step": 20509 + }, + { + "epoch": 4.48, + "learning_rate": 2.7649115140922004e-06, + "loss": 3.0485, + "step": 20510 + }, + { + "epoch": 4.48, + "learning_rate": 2.764802272230719e-06, + "loss": 1.2516, + "step": 20511 + }, + { + "epoch": 4.48, + "learning_rate": 2.7646930303692376e-06, + "loss": 2.3268, + "step": 20512 + }, + { + "epoch": 4.48, + "learning_rate": 2.7645837885077563e-06, + "loss": 0.6363, + "step": 20513 + }, + { + "epoch": 4.48, + "learning_rate": 2.7644745466462753e-06, + "loss": 1.3277, + "step": 20514 + }, + { + "epoch": 4.48, + "learning_rate": 2.764365304784794e-06, + "loss": 1.5351, + "step": 20515 + }, + { + "epoch": 4.48, + "learning_rate": 2.764256062923312e-06, + "loss": 1.7243, + "step": 20516 + }, + { + "epoch": 4.48, + "learning_rate": 2.764146821061831e-06, + "loss": 1.9254, + "step": 20517 + }, + { + "epoch": 4.48, + "learning_rate": 2.7640375792003497e-06, + "loss": 1.7135, + "step": 20518 + }, + { + "epoch": 4.48, + "learning_rate": 2.7639283373388688e-06, + "loss": 1.6235, + "step": 20519 + }, + { + "epoch": 4.48, + "learning_rate": 2.763819095477387e-06, + "loss": 0.2851, + "step": 20520 + }, + { + "epoch": 4.48, + "learning_rate": 2.7637098536159056e-06, + "loss": 2.2427, + "step": 20521 + }, + { + "epoch": 4.48, + "learning_rate": 2.7636006117544246e-06, + "loss": 0.6445, + "step": 20522 + }, + { + "epoch": 4.48, + "learning_rate": 2.7634913698929432e-06, + "loss": 1.8023, + "step": 20523 + }, + { + "epoch": 4.48, + "learning_rate": 2.7633821280314622e-06, + "loss": 1.4134, + "step": 20524 + }, + { + "epoch": 4.48, + "learning_rate": 2.7632728861699804e-06, + "loss": 4.2994, + "step": 20525 + }, + { + "epoch": 4.48, + "learning_rate": 2.763163644308499e-06, + "loss": 2.6167, + "step": 20526 + }, + { + "epoch": 4.48, + "learning_rate": 2.763054402447018e-06, + "loss": 0.7219, + "step": 20527 + }, + { + "epoch": 4.48, + "learning_rate": 2.7629451605855367e-06, + "loss": 1.0661, + "step": 20528 + }, + { + "epoch": 4.48, + "learning_rate": 2.7628359187240557e-06, + "loss": 2.5678, + "step": 20529 + }, + { + "epoch": 4.48, + "learning_rate": 2.762726676862574e-06, + "loss": 1.2848, + "step": 20530 + }, + { + "epoch": 4.48, + "learning_rate": 2.7626174350010925e-06, + "loss": 2.2234, + "step": 20531 + }, + { + "epoch": 4.48, + "learning_rate": 2.7625081931396116e-06, + "loss": 1.5041, + "step": 20532 + }, + { + "epoch": 4.48, + "learning_rate": 2.76239895127813e-06, + "loss": 3.2465, + "step": 20533 + }, + { + "epoch": 4.48, + "learning_rate": 2.7622897094166484e-06, + "loss": 1.488, + "step": 20534 + }, + { + "epoch": 4.48, + "learning_rate": 2.7621804675551674e-06, + "loss": 1.04, + "step": 20535 + }, + { + "epoch": 4.48, + "learning_rate": 2.762071225693686e-06, + "loss": 3.3096, + "step": 20536 + }, + { + "epoch": 4.48, + "learning_rate": 2.7619619838322046e-06, + "loss": 1.2431, + "step": 20537 + }, + { + "epoch": 4.48, + "learning_rate": 2.7618527419707236e-06, + "loss": 1.8551, + "step": 20538 + }, + { + "epoch": 4.48, + "learning_rate": 2.761743500109242e-06, + "loss": 1.6724, + "step": 20539 + }, + { + "epoch": 4.48, + "learning_rate": 2.7616342582477604e-06, + "loss": 3.0227, + "step": 20540 + }, + { + "epoch": 4.48, + "learning_rate": 2.7615250163862795e-06, + "loss": 3.1049, + "step": 20541 + }, + { + "epoch": 4.48, + "learning_rate": 2.761415774524798e-06, + "loss": 0.4866, + "step": 20542 + }, + { + "epoch": 4.48, + "learning_rate": 2.761306532663317e-06, + "loss": 2.8003, + "step": 20543 + }, + { + "epoch": 4.48, + "learning_rate": 2.7611972908018353e-06, + "loss": 2.633, + "step": 20544 + }, + { + "epoch": 4.48, + "learning_rate": 2.761088048940354e-06, + "loss": 1.2446, + "step": 20545 + }, + { + "epoch": 4.48, + "learning_rate": 2.760978807078873e-06, + "loss": 1.5387, + "step": 20546 + }, + { + "epoch": 4.48, + "learning_rate": 2.7608695652173916e-06, + "loss": 1.2004, + "step": 20547 + }, + { + "epoch": 4.48, + "learning_rate": 2.7607603233559106e-06, + "loss": 2.618, + "step": 20548 + }, + { + "epoch": 4.48, + "learning_rate": 2.760651081494429e-06, + "loss": 1.3548, + "step": 20549 + }, + { + "epoch": 4.48, + "learning_rate": 2.7605418396329474e-06, + "loss": 1.6411, + "step": 20550 + }, + { + "epoch": 4.49, + "learning_rate": 2.7604325977714664e-06, + "loss": 1.7214, + "step": 20551 + }, + { + "epoch": 4.49, + "learning_rate": 2.760323355909985e-06, + "loss": 0.5839, + "step": 20552 + }, + { + "epoch": 4.49, + "learning_rate": 2.7602141140485032e-06, + "loss": 1.7328, + "step": 20553 + }, + { + "epoch": 4.49, + "learning_rate": 2.7601048721870223e-06, + "loss": 0.9959, + "step": 20554 + }, + { + "epoch": 4.49, + "learning_rate": 2.759995630325541e-06, + "loss": 2.8988, + "step": 20555 + }, + { + "epoch": 4.49, + "learning_rate": 2.75988638846406e-06, + "loss": 1.5059, + "step": 20556 + }, + { + "epoch": 4.49, + "learning_rate": 2.7597771466025785e-06, + "loss": 3.5004, + "step": 20557 + }, + { + "epoch": 4.49, + "learning_rate": 2.7596679047410967e-06, + "loss": 2.4676, + "step": 20558 + }, + { + "epoch": 4.49, + "learning_rate": 2.7595586628796157e-06, + "loss": 2.5347, + "step": 20559 + }, + { + "epoch": 4.49, + "learning_rate": 2.7594494210181344e-06, + "loss": 1.7904, + "step": 20560 + }, + { + "epoch": 4.49, + "learning_rate": 2.7593401791566534e-06, + "loss": 3.428, + "step": 20561 + }, + { + "epoch": 4.49, + "learning_rate": 2.759230937295172e-06, + "loss": 2.2428, + "step": 20562 + }, + { + "epoch": 4.49, + "learning_rate": 2.75912169543369e-06, + "loss": 1.215, + "step": 20563 + }, + { + "epoch": 4.49, + "learning_rate": 2.7590124535722092e-06, + "loss": 2.976, + "step": 20564 + }, + { + "epoch": 4.49, + "learning_rate": 2.758903211710728e-06, + "loss": 1.6942, + "step": 20565 + }, + { + "epoch": 4.49, + "learning_rate": 2.7587939698492465e-06, + "loss": 1.1158, + "step": 20566 + }, + { + "epoch": 4.49, + "learning_rate": 2.758684727987765e-06, + "loss": 1.7068, + "step": 20567 + }, + { + "epoch": 4.49, + "learning_rate": 2.7585754861262837e-06, + "loss": 2.1658, + "step": 20568 + }, + { + "epoch": 4.49, + "learning_rate": 2.7584662442648023e-06, + "loss": 2.0999, + "step": 20569 + }, + { + "epoch": 4.49, + "learning_rate": 2.7583570024033213e-06, + "loss": 2.2226, + "step": 20570 + }, + { + "epoch": 4.49, + "learning_rate": 2.75824776054184e-06, + "loss": 0.4365, + "step": 20571 + }, + { + "epoch": 4.49, + "learning_rate": 2.758138518680358e-06, + "loss": 1.687, + "step": 20572 + }, + { + "epoch": 4.49, + "learning_rate": 2.758029276818877e-06, + "loss": 1.7264, + "step": 20573 + }, + { + "epoch": 4.49, + "learning_rate": 2.7579200349573958e-06, + "loss": 1.4306, + "step": 20574 + }, + { + "epoch": 4.49, + "learning_rate": 2.757810793095915e-06, + "loss": 2.5827, + "step": 20575 + }, + { + "epoch": 4.49, + "learning_rate": 2.7577015512344334e-06, + "loss": 0.4804, + "step": 20576 + }, + { + "epoch": 4.49, + "learning_rate": 2.7575923093729516e-06, + "loss": 1.4814, + "step": 20577 + }, + { + "epoch": 4.49, + "learning_rate": 2.7574830675114706e-06, + "loss": 2.0831, + "step": 20578 + }, + { + "epoch": 4.49, + "learning_rate": 2.7573738256499892e-06, + "loss": 1.0424, + "step": 20579 + }, + { + "epoch": 4.49, + "learning_rate": 2.7572645837885083e-06, + "loss": 2.8426, + "step": 20580 + }, + { + "epoch": 4.49, + "learning_rate": 2.7571553419270265e-06, + "loss": 2.64, + "step": 20581 + }, + { + "epoch": 4.49, + "learning_rate": 2.757046100065545e-06, + "loss": 2.5132, + "step": 20582 + }, + { + "epoch": 4.49, + "learning_rate": 2.756936858204064e-06, + "loss": 1.2065, + "step": 20583 + }, + { + "epoch": 4.49, + "learning_rate": 2.7568276163425827e-06, + "loss": 0.7051, + "step": 20584 + }, + { + "epoch": 4.49, + "learning_rate": 2.7567183744811018e-06, + "loss": 2.7202, + "step": 20585 + }, + { + "epoch": 4.49, + "learning_rate": 2.75660913261962e-06, + "loss": 1.6707, + "step": 20586 + }, + { + "epoch": 4.49, + "learning_rate": 2.7564998907581386e-06, + "loss": 0.514, + "step": 20587 + }, + { + "epoch": 4.49, + "learning_rate": 2.7563906488966576e-06, + "loss": 0.9493, + "step": 20588 + }, + { + "epoch": 4.49, + "learning_rate": 2.756281407035176e-06, + "loss": 1.1938, + "step": 20589 + }, + { + "epoch": 4.49, + "learning_rate": 2.756172165173695e-06, + "loss": 0.4348, + "step": 20590 + }, + { + "epoch": 4.49, + "learning_rate": 2.7560629233122134e-06, + "loss": 3.2036, + "step": 20591 + }, + { + "epoch": 4.49, + "learning_rate": 2.755953681450732e-06, + "loss": 1.4761, + "step": 20592 + }, + { + "epoch": 4.49, + "learning_rate": 2.755844439589251e-06, + "loss": 0.9921, + "step": 20593 + }, + { + "epoch": 4.49, + "learning_rate": 2.7557351977277697e-06, + "loss": 0.7489, + "step": 20594 + }, + { + "epoch": 4.49, + "learning_rate": 2.7556259558662883e-06, + "loss": 2.8787, + "step": 20595 + }, + { + "epoch": 4.49, + "learning_rate": 2.755516714004807e-06, + "loss": 1.2589, + "step": 20596 + }, + { + "epoch": 4.5, + "learning_rate": 2.7554074721433255e-06, + "loss": 2.823, + "step": 20597 + }, + { + "epoch": 4.5, + "learning_rate": 2.755298230281844e-06, + "loss": 2.5518, + "step": 20598 + }, + { + "epoch": 4.5, + "learning_rate": 2.755188988420363e-06, + "loss": 2.1628, + "step": 20599 + }, + { + "epoch": 4.5, + "learning_rate": 2.7550797465588813e-06, + "loss": 2.5023, + "step": 20600 + }, + { + "epoch": 4.5, + "learning_rate": 2.7549705046974e-06, + "loss": 1.4901, + "step": 20601 + }, + { + "epoch": 4.5, + "learning_rate": 2.754861262835919e-06, + "loss": 1.3273, + "step": 20602 + }, + { + "epoch": 4.5, + "learning_rate": 2.7547520209744376e-06, + "loss": 1.6987, + "step": 20603 + }, + { + "epoch": 4.5, + "learning_rate": 2.7546427791129566e-06, + "loss": 1.7329, + "step": 20604 + }, + { + "epoch": 4.5, + "learning_rate": 2.754533537251475e-06, + "loss": 2.8498, + "step": 20605 + }, + { + "epoch": 4.5, + "learning_rate": 2.7544242953899934e-06, + "loss": 1.3318, + "step": 20606 + }, + { + "epoch": 4.5, + "learning_rate": 2.7543150535285125e-06, + "loss": 1.5311, + "step": 20607 + }, + { + "epoch": 4.5, + "learning_rate": 2.754205811667031e-06, + "loss": 1.1141, + "step": 20608 + }, + { + "epoch": 4.5, + "learning_rate": 2.75409656980555e-06, + "loss": 0.721, + "step": 20609 + }, + { + "epoch": 4.5, + "learning_rate": 2.7539873279440683e-06, + "loss": 3.4319, + "step": 20610 + }, + { + "epoch": 4.5, + "learning_rate": 2.753878086082587e-06, + "loss": 0.8859, + "step": 20611 + }, + { + "epoch": 4.5, + "learning_rate": 2.753768844221106e-06, + "loss": 1.5647, + "step": 20612 + }, + { + "epoch": 4.5, + "learning_rate": 2.7536596023596246e-06, + "loss": 0.8506, + "step": 20613 + }, + { + "epoch": 4.5, + "learning_rate": 2.7535503604981427e-06, + "loss": 1.2681, + "step": 20614 + }, + { + "epoch": 4.5, + "learning_rate": 2.7534411186366618e-06, + "loss": 3.3071, + "step": 20615 + }, + { + "epoch": 4.5, + "learning_rate": 2.7533318767751804e-06, + "loss": 3.1298, + "step": 20616 + }, + { + "epoch": 4.5, + "learning_rate": 2.7532226349136994e-06, + "loss": 1.9599, + "step": 20617 + }, + { + "epoch": 4.5, + "learning_rate": 2.753113393052218e-06, + "loss": 1.428, + "step": 20618 + }, + { + "epoch": 4.5, + "learning_rate": 2.7530041511907362e-06, + "loss": 2.1269, + "step": 20619 + }, + { + "epoch": 4.5, + "learning_rate": 2.7528949093292553e-06, + "loss": 1.54, + "step": 20620 + }, + { + "epoch": 4.5, + "learning_rate": 2.752785667467774e-06, + "loss": 1.783, + "step": 20621 + }, + { + "epoch": 4.5, + "learning_rate": 2.7526764256062925e-06, + "loss": 0.9743, + "step": 20622 + }, + { + "epoch": 4.5, + "learning_rate": 2.7525671837448115e-06, + "loss": 3.3677, + "step": 20623 + }, + { + "epoch": 4.5, + "learning_rate": 2.7524579418833297e-06, + "loss": 3.0636, + "step": 20624 + }, + { + "epoch": 4.5, + "learning_rate": 2.7523487000218483e-06, + "loss": 1.0038, + "step": 20625 + }, + { + "epoch": 4.5, + "learning_rate": 2.7522394581603673e-06, + "loss": 1.3711, + "step": 20626 + }, + { + "epoch": 4.5, + "learning_rate": 2.752130216298886e-06, + "loss": 1.4452, + "step": 20627 + }, + { + "epoch": 4.5, + "learning_rate": 2.752020974437405e-06, + "loss": 0.9563, + "step": 20628 + }, + { + "epoch": 4.5, + "learning_rate": 2.751911732575923e-06, + "loss": 0.3875, + "step": 20629 + }, + { + "epoch": 4.5, + "learning_rate": 2.751802490714442e-06, + "loss": 1.733, + "step": 20630 + }, + { + "epoch": 4.5, + "learning_rate": 2.751693248852961e-06, + "loss": 1.4213, + "step": 20631 + }, + { + "epoch": 4.5, + "learning_rate": 2.7515840069914794e-06, + "loss": 2.111, + "step": 20632 + }, + { + "epoch": 4.5, + "learning_rate": 2.7514747651299976e-06, + "loss": 2.265, + "step": 20633 + }, + { + "epoch": 4.5, + "learning_rate": 2.7513655232685167e-06, + "loss": 0.8292, + "step": 20634 + }, + { + "epoch": 4.5, + "learning_rate": 2.7512562814070353e-06, + "loss": 1.3382, + "step": 20635 + }, + { + "epoch": 4.5, + "learning_rate": 2.7511470395455543e-06, + "loss": 1.6927, + "step": 20636 + }, + { + "epoch": 4.5, + "learning_rate": 2.751037797684073e-06, + "loss": 0.9808, + "step": 20637 + }, + { + "epoch": 4.5, + "learning_rate": 2.750928555822591e-06, + "loss": 3.1231, + "step": 20638 + }, + { + "epoch": 4.5, + "learning_rate": 2.75081931396111e-06, + "loss": 1.2537, + "step": 20639 + }, + { + "epoch": 4.5, + "learning_rate": 2.7507100720996287e-06, + "loss": 1.9064, + "step": 20640 + }, + { + "epoch": 4.5, + "learning_rate": 2.7506008302381478e-06, + "loss": 2.7533, + "step": 20641 + }, + { + "epoch": 4.51, + "learning_rate": 2.7504915883766664e-06, + "loss": 1.992, + "step": 20642 + }, + { + "epoch": 4.51, + "learning_rate": 2.7503823465151846e-06, + "loss": 1.4352, + "step": 20643 + }, + { + "epoch": 4.51, + "learning_rate": 2.7502731046537036e-06, + "loss": 0.9031, + "step": 20644 + }, + { + "epoch": 4.51, + "learning_rate": 2.7501638627922222e-06, + "loss": 2.6613, + "step": 20645 + }, + { + "epoch": 4.51, + "learning_rate": 2.7500546209307413e-06, + "loss": 1.0139, + "step": 20646 + }, + { + "epoch": 4.51, + "learning_rate": 2.7499453790692594e-06, + "loss": 2.3985, + "step": 20647 + }, + { + "epoch": 4.51, + "learning_rate": 2.749836137207778e-06, + "loss": 2.1204, + "step": 20648 + }, + { + "epoch": 4.51, + "learning_rate": 2.749726895346297e-06, + "loss": 0.9716, + "step": 20649 + }, + { + "epoch": 4.51, + "learning_rate": 2.7496176534848157e-06, + "loss": 1.228, + "step": 20650 + }, + { + "epoch": 4.51, + "learning_rate": 2.7495084116233343e-06, + "loss": 1.2648, + "step": 20651 + }, + { + "epoch": 4.51, + "learning_rate": 2.749399169761853e-06, + "loss": 2.0264, + "step": 20652 + }, + { + "epoch": 4.51, + "learning_rate": 2.7492899279003715e-06, + "loss": 1.3826, + "step": 20653 + }, + { + "epoch": 4.51, + "learning_rate": 2.74918068603889e-06, + "loss": 0.998, + "step": 20654 + }, + { + "epoch": 4.51, + "learning_rate": 2.749071444177409e-06, + "loss": 1.9936, + "step": 20655 + }, + { + "epoch": 4.51, + "learning_rate": 2.748962202315928e-06, + "loss": 2.0753, + "step": 20656 + }, + { + "epoch": 4.51, + "learning_rate": 2.748852960454446e-06, + "loss": 0.9201, + "step": 20657 + }, + { + "epoch": 4.51, + "learning_rate": 2.748743718592965e-06, + "loss": 1.7049, + "step": 20658 + }, + { + "epoch": 4.51, + "learning_rate": 2.7486344767314836e-06, + "loss": 1.6122, + "step": 20659 + }, + { + "epoch": 4.51, + "learning_rate": 2.7485252348700027e-06, + "loss": 0.2904, + "step": 20660 + }, + { + "epoch": 4.51, + "learning_rate": 2.7484159930085213e-06, + "loss": 0.9621, + "step": 20661 + }, + { + "epoch": 4.51, + "learning_rate": 2.7483067511470395e-06, + "loss": 1.8723, + "step": 20662 + }, + { + "epoch": 4.51, + "learning_rate": 2.7481975092855585e-06, + "loss": 1.1266, + "step": 20663 + }, + { + "epoch": 4.51, + "learning_rate": 2.748088267424077e-06, + "loss": 2.1308, + "step": 20664 + }, + { + "epoch": 4.51, + "learning_rate": 2.747979025562596e-06, + "loss": 1.0954, + "step": 20665 + }, + { + "epoch": 4.51, + "learning_rate": 2.7478697837011143e-06, + "loss": 1.8194, + "step": 20666 + }, + { + "epoch": 4.51, + "learning_rate": 2.747760541839633e-06, + "loss": 0.9353, + "step": 20667 + }, + { + "epoch": 4.51, + "learning_rate": 2.747651299978152e-06, + "loss": 0.7348, + "step": 20668 + }, + { + "epoch": 4.51, + "learning_rate": 2.7475420581166706e-06, + "loss": 1.2271, + "step": 20669 + }, + { + "epoch": 4.51, + "learning_rate": 2.7474328162551896e-06, + "loss": 2.5446, + "step": 20670 + }, + { + "epoch": 4.51, + "learning_rate": 2.747323574393708e-06, + "loss": 3.8584, + "step": 20671 + }, + { + "epoch": 4.51, + "learning_rate": 2.7472143325322264e-06, + "loss": 1.8507, + "step": 20672 + }, + { + "epoch": 4.51, + "learning_rate": 2.7471050906707454e-06, + "loss": 2.363, + "step": 20673 + }, + { + "epoch": 4.51, + "learning_rate": 2.746995848809264e-06, + "loss": 1.6582, + "step": 20674 + }, + { + "epoch": 4.51, + "learning_rate": 2.7468866069477827e-06, + "loss": 0.884, + "step": 20675 + }, + { + "epoch": 4.51, + "learning_rate": 2.7467773650863013e-06, + "loss": 1.5654, + "step": 20676 + }, + { + "epoch": 4.51, + "learning_rate": 2.74666812322482e-06, + "loss": 2.3127, + "step": 20677 + }, + { + "epoch": 4.51, + "learning_rate": 2.7465588813633385e-06, + "loss": 1.8966, + "step": 20678 + }, + { + "epoch": 4.51, + "learning_rate": 2.7464496395018575e-06, + "loss": 1.4529, + "step": 20679 + }, + { + "epoch": 4.51, + "learning_rate": 2.7463403976403757e-06, + "loss": 1.54, + "step": 20680 + }, + { + "epoch": 4.51, + "learning_rate": 2.7462311557788943e-06, + "loss": 2.0722, + "step": 20681 + }, + { + "epoch": 4.51, + "learning_rate": 2.7461219139174134e-06, + "loss": 1.2378, + "step": 20682 + }, + { + "epoch": 4.51, + "learning_rate": 2.746012672055932e-06, + "loss": 1.6405, + "step": 20683 + }, + { + "epoch": 4.51, + "learning_rate": 2.745903430194451e-06, + "loss": 2.7997, + "step": 20684 + }, + { + "epoch": 4.51, + "learning_rate": 2.745794188332969e-06, + "loss": 1.7926, + "step": 20685 + }, + { + "epoch": 4.51, + "learning_rate": 2.745684946471488e-06, + "loss": 0.671, + "step": 20686 + }, + { + "epoch": 4.51, + "learning_rate": 2.745575704610007e-06, + "loss": 0.7386, + "step": 20687 + }, + { + "epoch": 4.52, + "learning_rate": 2.7454664627485255e-06, + "loss": 0.9487, + "step": 20688 + }, + { + "epoch": 4.52, + "learning_rate": 2.7453572208870445e-06, + "loss": 1.0435, + "step": 20689 + }, + { + "epoch": 4.52, + "learning_rate": 2.7452479790255627e-06, + "loss": 3.032, + "step": 20690 + }, + { + "epoch": 4.52, + "learning_rate": 2.7451387371640813e-06, + "loss": 1.782, + "step": 20691 + }, + { + "epoch": 4.52, + "learning_rate": 2.7450294953026003e-06, + "loss": 2.0195, + "step": 20692 + }, + { + "epoch": 4.52, + "learning_rate": 2.744920253441119e-06, + "loss": 1.4971, + "step": 20693 + }, + { + "epoch": 4.52, + "learning_rate": 2.744811011579638e-06, + "loss": 0.8405, + "step": 20694 + }, + { + "epoch": 4.52, + "learning_rate": 2.744701769718156e-06, + "loss": 2.4851, + "step": 20695 + }, + { + "epoch": 4.52, + "learning_rate": 2.7445925278566748e-06, + "loss": 1.8743, + "step": 20696 + }, + { + "epoch": 4.52, + "learning_rate": 2.744483285995194e-06, + "loss": 1.0299, + "step": 20697 + }, + { + "epoch": 4.52, + "learning_rate": 2.7443740441337124e-06, + "loss": 2.6933, + "step": 20698 + }, + { + "epoch": 4.52, + "learning_rate": 2.7442648022722306e-06, + "loss": 2.3964, + "step": 20699 + }, + { + "epoch": 4.52, + "learning_rate": 2.7441555604107496e-06, + "loss": 2.1883, + "step": 20700 + }, + { + "epoch": 4.52, + "learning_rate": 2.7440463185492683e-06, + "loss": 1.7339, + "step": 20701 + }, + { + "epoch": 4.52, + "learning_rate": 2.7439370766877873e-06, + "loss": 1.5379, + "step": 20702 + }, + { + "epoch": 4.52, + "learning_rate": 2.743827834826306e-06, + "loss": 1.5694, + "step": 20703 + }, + { + "epoch": 4.52, + "learning_rate": 2.743718592964824e-06, + "loss": 2.6237, + "step": 20704 + }, + { + "epoch": 4.52, + "learning_rate": 2.743609351103343e-06, + "loss": 1.8381, + "step": 20705 + }, + { + "epoch": 4.52, + "learning_rate": 2.7435001092418617e-06, + "loss": 1.2435, + "step": 20706 + }, + { + "epoch": 4.52, + "learning_rate": 2.7433908673803803e-06, + "loss": 1.0122, + "step": 20707 + }, + { + "epoch": 4.52, + "learning_rate": 2.7432816255188994e-06, + "loss": 1.3967, + "step": 20708 + }, + { + "epoch": 4.52, + "learning_rate": 2.7431723836574176e-06, + "loss": 1.0671, + "step": 20709 + }, + { + "epoch": 4.52, + "learning_rate": 2.743063141795936e-06, + "loss": 0.121, + "step": 20710 + }, + { + "epoch": 4.52, + "learning_rate": 2.742953899934455e-06, + "loss": 2.2721, + "step": 20711 + }, + { + "epoch": 4.52, + "learning_rate": 2.742844658072974e-06, + "loss": 1.9506, + "step": 20712 + }, + { + "epoch": 4.52, + "learning_rate": 2.742735416211492e-06, + "loss": 2.3519, + "step": 20713 + }, + { + "epoch": 4.52, + "learning_rate": 2.742626174350011e-06, + "loss": 1.1655, + "step": 20714 + }, + { + "epoch": 4.52, + "learning_rate": 2.7425169324885297e-06, + "loss": 1.5391, + "step": 20715 + }, + { + "epoch": 4.52, + "learning_rate": 2.7424076906270487e-06, + "loss": 0.4578, + "step": 20716 + }, + { + "epoch": 4.52, + "learning_rate": 2.7422984487655673e-06, + "loss": 2.032, + "step": 20717 + }, + { + "epoch": 4.52, + "learning_rate": 2.7421892069040855e-06, + "loss": 0.8849, + "step": 20718 + }, + { + "epoch": 4.52, + "learning_rate": 2.7420799650426045e-06, + "loss": 1.419, + "step": 20719 + }, + { + "epoch": 4.52, + "learning_rate": 2.741970723181123e-06, + "loss": 0.9809, + "step": 20720 + }, + { + "epoch": 4.52, + "learning_rate": 2.741861481319642e-06, + "loss": 0.7245, + "step": 20721 + }, + { + "epoch": 4.52, + "learning_rate": 2.7417522394581608e-06, + "loss": 1.7328, + "step": 20722 + }, + { + "epoch": 4.52, + "learning_rate": 2.741642997596679e-06, + "loss": 1.5661, + "step": 20723 + }, + { + "epoch": 4.52, + "learning_rate": 2.741533755735198e-06, + "loss": 2.8611, + "step": 20724 + }, + { + "epoch": 4.52, + "learning_rate": 2.7414245138737166e-06, + "loss": 2.4213, + "step": 20725 + }, + { + "epoch": 4.52, + "learning_rate": 2.7413152720122356e-06, + "loss": 2.9219, + "step": 20726 + }, + { + "epoch": 4.52, + "learning_rate": 2.7412060301507543e-06, + "loss": 1.9107, + "step": 20727 + }, + { + "epoch": 4.52, + "learning_rate": 2.7410967882892724e-06, + "loss": 0.7178, + "step": 20728 + }, + { + "epoch": 4.52, + "learning_rate": 2.7409875464277915e-06, + "loss": 1.3645, + "step": 20729 + }, + { + "epoch": 4.52, + "learning_rate": 2.74087830456631e-06, + "loss": 1.2889, + "step": 20730 + }, + { + "epoch": 4.52, + "learning_rate": 2.7407690627048287e-06, + "loss": 1.7094, + "step": 20731 + }, + { + "epoch": 4.52, + "learning_rate": 2.7406598208433473e-06, + "loss": 1.3689, + "step": 20732 + }, + { + "epoch": 4.52, + "learning_rate": 2.740550578981866e-06, + "loss": 1.2352, + "step": 20733 + }, + { + "epoch": 4.53, + "learning_rate": 2.7404413371203845e-06, + "loss": 1.8613, + "step": 20734 + }, + { + "epoch": 4.53, + "learning_rate": 2.7403320952589036e-06, + "loss": 1.7793, + "step": 20735 + }, + { + "epoch": 4.53, + "learning_rate": 2.740222853397422e-06, + "loss": 1.6215, + "step": 20736 + }, + { + "epoch": 4.53, + "learning_rate": 2.7401136115359408e-06, + "loss": 1.802, + "step": 20737 + }, + { + "epoch": 4.53, + "learning_rate": 2.7400043696744594e-06, + "loss": 2.578, + "step": 20738 + }, + { + "epoch": 4.53, + "learning_rate": 2.739895127812978e-06, + "loss": 2.8583, + "step": 20739 + }, + { + "epoch": 4.53, + "learning_rate": 2.739785885951497e-06, + "loss": 1.6914, + "step": 20740 + }, + { + "epoch": 4.53, + "learning_rate": 2.7396766440900157e-06, + "loss": 0.8309, + "step": 20741 + }, + { + "epoch": 4.53, + "learning_rate": 2.739567402228534e-06, + "loss": 1.6881, + "step": 20742 + }, + { + "epoch": 4.53, + "learning_rate": 2.739458160367053e-06, + "loss": 1.6226, + "step": 20743 + }, + { + "epoch": 4.53, + "learning_rate": 2.7393489185055715e-06, + "loss": 0.4999, + "step": 20744 + }, + { + "epoch": 4.53, + "learning_rate": 2.7392396766440905e-06, + "loss": 0.507, + "step": 20745 + }, + { + "epoch": 4.53, + "learning_rate": 2.7391304347826087e-06, + "loss": 1.9293, + "step": 20746 + }, + { + "epoch": 4.53, + "learning_rate": 2.7390211929211273e-06, + "loss": 0.7668, + "step": 20747 + }, + { + "epoch": 4.53, + "learning_rate": 2.7389119510596464e-06, + "loss": 1.6365, + "step": 20748 + }, + { + "epoch": 4.53, + "learning_rate": 2.738802709198165e-06, + "loss": 2.1399, + "step": 20749 + }, + { + "epoch": 4.53, + "learning_rate": 2.738693467336684e-06, + "loss": 0.688, + "step": 20750 + }, + { + "epoch": 4.53, + "learning_rate": 2.738584225475202e-06, + "loss": 1.5702, + "step": 20751 + }, + { + "epoch": 4.53, + "learning_rate": 2.738474983613721e-06, + "loss": 1.4018, + "step": 20752 + }, + { + "epoch": 4.53, + "learning_rate": 2.73836574175224e-06, + "loss": 0.8496, + "step": 20753 + }, + { + "epoch": 4.53, + "learning_rate": 2.7382564998907584e-06, + "loss": 2.3905, + "step": 20754 + }, + { + "epoch": 4.53, + "learning_rate": 2.7381472580292775e-06, + "loss": 0.9132, + "step": 20755 + }, + { + "epoch": 4.53, + "learning_rate": 2.7380380161677957e-06, + "loss": 1.849, + "step": 20756 + }, + { + "epoch": 4.53, + "learning_rate": 2.7379287743063143e-06, + "loss": 2.4528, + "step": 20757 + }, + { + "epoch": 4.53, + "learning_rate": 2.7378195324448333e-06, + "loss": 1.6745, + "step": 20758 + }, + { + "epoch": 4.53, + "learning_rate": 2.737710290583352e-06, + "loss": 2.4752, + "step": 20759 + }, + { + "epoch": 4.53, + "learning_rate": 2.7376010487218705e-06, + "loss": 1.995, + "step": 20760 + }, + { + "epoch": 4.53, + "learning_rate": 2.737491806860389e-06, + "loss": 1.1758, + "step": 20761 + }, + { + "epoch": 4.53, + "learning_rate": 2.7373825649989078e-06, + "loss": 1.2751, + "step": 20762 + }, + { + "epoch": 4.53, + "learning_rate": 2.7372733231374264e-06, + "loss": 2.3471, + "step": 20763 + }, + { + "epoch": 4.53, + "learning_rate": 2.7371640812759454e-06, + "loss": 0.9867, + "step": 20764 + }, + { + "epoch": 4.53, + "learning_rate": 2.7370548394144636e-06, + "loss": 1.6973, + "step": 20765 + }, + { + "epoch": 4.53, + "learning_rate": 2.736945597552982e-06, + "loss": 0.7631, + "step": 20766 + }, + { + "epoch": 4.53, + "learning_rate": 2.7368363556915012e-06, + "loss": 1.8367, + "step": 20767 + }, + { + "epoch": 4.53, + "learning_rate": 2.73672711383002e-06, + "loss": 2.0558, + "step": 20768 + }, + { + "epoch": 4.53, + "learning_rate": 2.736617871968539e-06, + "loss": 1.9846, + "step": 20769 + }, + { + "epoch": 4.53, + "learning_rate": 2.736508630107057e-06, + "loss": 2.4965, + "step": 20770 + }, + { + "epoch": 4.53, + "learning_rate": 2.7363993882455757e-06, + "loss": 2.337, + "step": 20771 + }, + { + "epoch": 4.53, + "learning_rate": 2.7362901463840947e-06, + "loss": 2.4883, + "step": 20772 + }, + { + "epoch": 4.53, + "learning_rate": 2.7361809045226133e-06, + "loss": 1.5606, + "step": 20773 + }, + { + "epoch": 4.53, + "learning_rate": 2.7360716626611324e-06, + "loss": 1.0887, + "step": 20774 + }, + { + "epoch": 4.53, + "learning_rate": 2.7359624207996505e-06, + "loss": 1.5031, + "step": 20775 + }, + { + "epoch": 4.53, + "learning_rate": 2.735853178938169e-06, + "loss": 2.0639, + "step": 20776 + }, + { + "epoch": 4.53, + "learning_rate": 2.735743937076688e-06, + "loss": 3.398, + "step": 20777 + }, + { + "epoch": 4.53, + "learning_rate": 2.735634695215207e-06, + "loss": 1.868, + "step": 20778 + }, + { + "epoch": 4.53, + "learning_rate": 2.735525453353725e-06, + "loss": 1.4564, + "step": 20779 + }, + { + "epoch": 4.54, + "learning_rate": 2.735416211492244e-06, + "loss": 1.235, + "step": 20780 + }, + { + "epoch": 4.54, + "learning_rate": 2.7353069696307626e-06, + "loss": 2.6061, + "step": 20781 + }, + { + "epoch": 4.54, + "learning_rate": 2.7351977277692817e-06, + "loss": 0.2134, + "step": 20782 + }, + { + "epoch": 4.54, + "learning_rate": 2.7350884859078003e-06, + "loss": 0.5148, + "step": 20783 + }, + { + "epoch": 4.54, + "learning_rate": 2.7349792440463185e-06, + "loss": 0.4903, + "step": 20784 + }, + { + "epoch": 4.54, + "learning_rate": 2.7348700021848375e-06, + "loss": 1.6969, + "step": 20785 + }, + { + "epoch": 4.54, + "learning_rate": 2.734760760323356e-06, + "loss": 2.0054, + "step": 20786 + }, + { + "epoch": 4.54, + "learning_rate": 2.7346515184618747e-06, + "loss": 1.1144, + "step": 20787 + }, + { + "epoch": 4.54, + "learning_rate": 2.7345422766003938e-06, + "loss": 0.7753, + "step": 20788 + }, + { + "epoch": 4.54, + "learning_rate": 2.734433034738912e-06, + "loss": 1.9438, + "step": 20789 + }, + { + "epoch": 4.54, + "learning_rate": 2.734323792877431e-06, + "loss": 1.7343, + "step": 20790 + }, + { + "epoch": 4.54, + "learning_rate": 2.7342145510159496e-06, + "loss": 1.1531, + "step": 20791 + }, + { + "epoch": 4.54, + "learning_rate": 2.734105309154468e-06, + "loss": 2.3364, + "step": 20792 + }, + { + "epoch": 4.54, + "learning_rate": 2.7339960672929872e-06, + "loss": 1.8222, + "step": 20793 + }, + { + "epoch": 4.54, + "learning_rate": 2.7338868254315054e-06, + "loss": 2.867, + "step": 20794 + }, + { + "epoch": 4.54, + "learning_rate": 2.733777583570024e-06, + "loss": 2.3372, + "step": 20795 + }, + { + "epoch": 4.54, + "learning_rate": 2.733668341708543e-06, + "loss": 1.9483, + "step": 20796 + }, + { + "epoch": 4.54, + "learning_rate": 2.7335590998470617e-06, + "loss": 2.3937, + "step": 20797 + }, + { + "epoch": 4.54, + "learning_rate": 2.73344985798558e-06, + "loss": 2.3975, + "step": 20798 + }, + { + "epoch": 4.54, + "learning_rate": 2.733340616124099e-06, + "loss": 1.4484, + "step": 20799 + }, + { + "epoch": 4.54, + "learning_rate": 2.7332313742626175e-06, + "loss": 1.5589, + "step": 20800 + }, + { + "epoch": 4.54, + "learning_rate": 2.7331221324011365e-06, + "loss": 2.1464, + "step": 20801 + }, + { + "epoch": 4.54, + "learning_rate": 2.733012890539655e-06, + "loss": 2.165, + "step": 20802 + }, + { + "epoch": 4.54, + "learning_rate": 2.7329036486781733e-06, + "loss": 1.1843, + "step": 20803 + }, + { + "epoch": 4.54, + "learning_rate": 2.7327944068166924e-06, + "loss": 1.3025, + "step": 20804 + }, + { + "epoch": 4.54, + "learning_rate": 2.732685164955211e-06, + "loss": 2.7723, + "step": 20805 + }, + { + "epoch": 4.54, + "learning_rate": 2.73257592309373e-06, + "loss": 2.1164, + "step": 20806 + }, + { + "epoch": 4.54, + "learning_rate": 2.7324666812322486e-06, + "loss": 1.4518, + "step": 20807 + }, + { + "epoch": 4.54, + "learning_rate": 2.732357439370767e-06, + "loss": 1.803, + "step": 20808 + }, + { + "epoch": 4.54, + "learning_rate": 2.732248197509286e-06, + "loss": 1.882, + "step": 20809 + }, + { + "epoch": 4.54, + "learning_rate": 2.7321389556478045e-06, + "loss": 1.1137, + "step": 20810 + }, + { + "epoch": 4.54, + "learning_rate": 2.7320297137863235e-06, + "loss": 0.8576, + "step": 20811 + }, + { + "epoch": 4.54, + "learning_rate": 2.7319204719248417e-06, + "loss": 2.4677, + "step": 20812 + }, + { + "epoch": 4.54, + "learning_rate": 2.7318112300633603e-06, + "loss": 1.3117, + "step": 20813 + }, + { + "epoch": 4.54, + "learning_rate": 2.7317019882018793e-06, + "loss": 1.5798, + "step": 20814 + }, + { + "epoch": 4.54, + "learning_rate": 2.731592746340398e-06, + "loss": 2.0991, + "step": 20815 + }, + { + "epoch": 4.54, + "learning_rate": 2.7314835044789166e-06, + "loss": 3.2234, + "step": 20816 + }, + { + "epoch": 4.54, + "learning_rate": 2.731374262617435e-06, + "loss": 1.0432, + "step": 20817 + }, + { + "epoch": 4.54, + "learning_rate": 2.7312650207559538e-06, + "loss": 2.047, + "step": 20818 + }, + { + "epoch": 4.54, + "learning_rate": 2.7311557788944724e-06, + "loss": 2.7916, + "step": 20819 + }, + { + "epoch": 4.54, + "learning_rate": 2.7310465370329914e-06, + "loss": 1.7346, + "step": 20820 + }, + { + "epoch": 4.54, + "learning_rate": 2.73093729517151e-06, + "loss": 2.0121, + "step": 20821 + }, + { + "epoch": 4.54, + "learning_rate": 2.7308280533100282e-06, + "loss": 1.1798, + "step": 20822 + }, + { + "epoch": 4.54, + "learning_rate": 2.7307188114485473e-06, + "loss": 1.2398, + "step": 20823 + }, + { + "epoch": 4.54, + "learning_rate": 2.730609569587066e-06, + "loss": 1.9103, + "step": 20824 + }, + { + "epoch": 4.54, + "learning_rate": 2.730500327725585e-06, + "loss": 2.2173, + "step": 20825 + }, + { + "epoch": 4.55, + "learning_rate": 2.7303910858641035e-06, + "loss": 2.3319, + "step": 20826 + }, + { + "epoch": 4.55, + "learning_rate": 2.7302818440026217e-06, + "loss": 0.9866, + "step": 20827 + }, + { + "epoch": 4.55, + "learning_rate": 2.7301726021411407e-06, + "loss": 1.0564, + "step": 20828 + }, + { + "epoch": 4.55, + "learning_rate": 2.7300633602796594e-06, + "loss": 0.9331, + "step": 20829 + }, + { + "epoch": 4.55, + "learning_rate": 2.7299541184181784e-06, + "loss": 2.8947, + "step": 20830 + }, + { + "epoch": 4.55, + "learning_rate": 2.7298448765566966e-06, + "loss": 2.3312, + "step": 20831 + }, + { + "epoch": 4.55, + "learning_rate": 2.729735634695215e-06, + "loss": 2.496, + "step": 20832 + }, + { + "epoch": 4.55, + "learning_rate": 2.7296263928337342e-06, + "loss": 1.5756, + "step": 20833 + }, + { + "epoch": 4.55, + "learning_rate": 2.729517150972253e-06, + "loss": 1.2852, + "step": 20834 + }, + { + "epoch": 4.55, + "learning_rate": 2.729407909110772e-06, + "loss": 2.3661, + "step": 20835 + }, + { + "epoch": 4.55, + "learning_rate": 2.72929866724929e-06, + "loss": 1.6822, + "step": 20836 + }, + { + "epoch": 4.55, + "learning_rate": 2.7291894253878087e-06, + "loss": 2.7468, + "step": 20837 + }, + { + "epoch": 4.55, + "learning_rate": 2.7290801835263277e-06, + "loss": 1.7511, + "step": 20838 + }, + { + "epoch": 4.55, + "learning_rate": 2.7289709416648463e-06, + "loss": 2.4103, + "step": 20839 + }, + { + "epoch": 4.55, + "learning_rate": 2.728861699803365e-06, + "loss": 1.2752, + "step": 20840 + }, + { + "epoch": 4.55, + "learning_rate": 2.7287524579418835e-06, + "loss": 1.3943, + "step": 20841 + }, + { + "epoch": 4.55, + "learning_rate": 2.728643216080402e-06, + "loss": 1.6243, + "step": 20842 + }, + { + "epoch": 4.55, + "learning_rate": 2.728533974218921e-06, + "loss": 1.5017, + "step": 20843 + }, + { + "epoch": 4.55, + "learning_rate": 2.7284247323574398e-06, + "loss": 0.4543, + "step": 20844 + }, + { + "epoch": 4.55, + "learning_rate": 2.728315490495958e-06, + "loss": 4.3627, + "step": 20845 + }, + { + "epoch": 4.55, + "learning_rate": 2.728206248634477e-06, + "loss": 2.7816, + "step": 20846 + }, + { + "epoch": 4.55, + "learning_rate": 2.7280970067729956e-06, + "loss": 0.9767, + "step": 20847 + }, + { + "epoch": 4.55, + "learning_rate": 2.7279877649115142e-06, + "loss": 2.9286, + "step": 20848 + }, + { + "epoch": 4.55, + "learning_rate": 2.7278785230500333e-06, + "loss": 1.5885, + "step": 20849 + }, + { + "epoch": 4.55, + "learning_rate": 2.7277692811885515e-06, + "loss": 1.5606, + "step": 20850 + }, + { + "epoch": 4.55, + "learning_rate": 2.72766003932707e-06, + "loss": 2.7978, + "step": 20851 + }, + { + "epoch": 4.55, + "learning_rate": 2.727550797465589e-06, + "loss": 1.0205, + "step": 20852 + }, + { + "epoch": 4.55, + "learning_rate": 2.7274415556041077e-06, + "loss": 1.5653, + "step": 20853 + }, + { + "epoch": 4.55, + "learning_rate": 2.7273323137426267e-06, + "loss": 2.4449, + "step": 20854 + }, + { + "epoch": 4.55, + "learning_rate": 2.727223071881145e-06, + "loss": 3.5339, + "step": 20855 + }, + { + "epoch": 4.55, + "learning_rate": 2.7271138300196635e-06, + "loss": 1.4353, + "step": 20856 + }, + { + "epoch": 4.55, + "learning_rate": 2.7270045881581826e-06, + "loss": 1.5548, + "step": 20857 + }, + { + "epoch": 4.55, + "learning_rate": 2.726895346296701e-06, + "loss": 0.4095, + "step": 20858 + }, + { + "epoch": 4.55, + "learning_rate": 2.7267861044352202e-06, + "loss": 2.0335, + "step": 20859 + }, + { + "epoch": 4.55, + "learning_rate": 2.7266768625737384e-06, + "loss": 1.4953, + "step": 20860 + }, + { + "epoch": 4.55, + "learning_rate": 2.726567620712257e-06, + "loss": 1.9412, + "step": 20861 + }, + { + "epoch": 4.55, + "learning_rate": 2.726458378850776e-06, + "loss": 1.2731, + "step": 20862 + }, + { + "epoch": 4.55, + "learning_rate": 2.7263491369892947e-06, + "loss": 1.8296, + "step": 20863 + }, + { + "epoch": 4.55, + "learning_rate": 2.726239895127813e-06, + "loss": 1.7874, + "step": 20864 + }, + { + "epoch": 4.55, + "learning_rate": 2.726130653266332e-06, + "loss": 0.6736, + "step": 20865 + }, + { + "epoch": 4.55, + "learning_rate": 2.7260214114048505e-06, + "loss": 1.3495, + "step": 20866 + }, + { + "epoch": 4.55, + "learning_rate": 2.7259121695433695e-06, + "loss": 1.9836, + "step": 20867 + }, + { + "epoch": 4.55, + "learning_rate": 2.725802927681888e-06, + "loss": 2.6616, + "step": 20868 + }, + { + "epoch": 4.55, + "learning_rate": 2.7256936858204063e-06, + "loss": 1.812, + "step": 20869 + }, + { + "epoch": 4.55, + "learning_rate": 2.7255844439589254e-06, + "loss": 2.9049, + "step": 20870 + }, + { + "epoch": 4.55, + "learning_rate": 2.725475202097444e-06, + "loss": 1.6755, + "step": 20871 + }, + { + "epoch": 4.56, + "learning_rate": 2.7253659602359626e-06, + "loss": 2.0819, + "step": 20872 + }, + { + "epoch": 4.56, + "learning_rate": 2.7252567183744816e-06, + "loss": 0.5706, + "step": 20873 + }, + { + "epoch": 4.56, + "learning_rate": 2.725147476513e-06, + "loss": 2.5494, + "step": 20874 + }, + { + "epoch": 4.56, + "learning_rate": 2.7250382346515184e-06, + "loss": 0.7196, + "step": 20875 + }, + { + "epoch": 4.56, + "learning_rate": 2.7249289927900375e-06, + "loss": 2.9328, + "step": 20876 + }, + { + "epoch": 4.56, + "learning_rate": 2.724819750928556e-06, + "loss": 1.3737, + "step": 20877 + }, + { + "epoch": 4.56, + "learning_rate": 2.7247105090670743e-06, + "loss": 1.9652, + "step": 20878 + }, + { + "epoch": 4.56, + "learning_rate": 2.7246012672055933e-06, + "loss": 1.0442, + "step": 20879 + }, + { + "epoch": 4.56, + "learning_rate": 2.724492025344112e-06, + "loss": 0.6188, + "step": 20880 + }, + { + "epoch": 4.56, + "learning_rate": 2.724382783482631e-06, + "loss": 2.5161, + "step": 20881 + }, + { + "epoch": 4.56, + "learning_rate": 2.7242735416211495e-06, + "loss": 1.6386, + "step": 20882 + }, + { + "epoch": 4.56, + "learning_rate": 2.7241642997596677e-06, + "loss": 2.0195, + "step": 20883 + }, + { + "epoch": 4.56, + "learning_rate": 2.7240550578981868e-06, + "loss": 3.0008, + "step": 20884 + }, + { + "epoch": 4.56, + "learning_rate": 2.7239458160367054e-06, + "loss": 1.3554, + "step": 20885 + }, + { + "epoch": 4.56, + "learning_rate": 2.7238365741752244e-06, + "loss": 1.7582, + "step": 20886 + }, + { + "epoch": 4.56, + "learning_rate": 2.723727332313743e-06, + "loss": 1.1183, + "step": 20887 + }, + { + "epoch": 4.56, + "learning_rate": 2.723618090452261e-06, + "loss": 2.1024, + "step": 20888 + }, + { + "epoch": 4.56, + "learning_rate": 2.7235088485907802e-06, + "loss": 2.8025, + "step": 20889 + }, + { + "epoch": 4.56, + "learning_rate": 2.723399606729299e-06, + "loss": 0.5822, + "step": 20890 + }, + { + "epoch": 4.56, + "learning_rate": 2.723290364867818e-06, + "loss": 1.6941, + "step": 20891 + }, + { + "epoch": 4.56, + "learning_rate": 2.7231811230063365e-06, + "loss": 2.0817, + "step": 20892 + }, + { + "epoch": 4.56, + "learning_rate": 2.7230718811448547e-06, + "loss": 1.2363, + "step": 20893 + }, + { + "epoch": 4.56, + "learning_rate": 2.7229626392833737e-06, + "loss": 1.2863, + "step": 20894 + }, + { + "epoch": 4.56, + "learning_rate": 2.7228533974218923e-06, + "loss": 1.969, + "step": 20895 + }, + { + "epoch": 4.56, + "learning_rate": 2.7227441555604114e-06, + "loss": 1.8864, + "step": 20896 + }, + { + "epoch": 4.56, + "learning_rate": 2.7226349136989296e-06, + "loss": 2.1941, + "step": 20897 + }, + { + "epoch": 4.56, + "learning_rate": 2.722525671837448e-06, + "loss": 0.6933, + "step": 20898 + }, + { + "epoch": 4.56, + "learning_rate": 2.722416429975967e-06, + "loss": 3.1606, + "step": 20899 + }, + { + "epoch": 4.56, + "learning_rate": 2.722307188114486e-06, + "loss": 2.2715, + "step": 20900 + }, + { + "epoch": 4.56, + "learning_rate": 2.7221979462530044e-06, + "loss": 0.8002, + "step": 20901 + }, + { + "epoch": 4.56, + "learning_rate": 2.722088704391523e-06, + "loss": 2.2775, + "step": 20902 + }, + { + "epoch": 4.56, + "learning_rate": 2.7219794625300416e-06, + "loss": 1.9664, + "step": 20903 + }, + { + "epoch": 4.56, + "learning_rate": 2.7218702206685603e-06, + "loss": 1.9874, + "step": 20904 + }, + { + "epoch": 4.56, + "learning_rate": 2.7217609788070793e-06, + "loss": 0.9274, + "step": 20905 + }, + { + "epoch": 4.56, + "learning_rate": 2.721651736945598e-06, + "loss": 2.1973, + "step": 20906 + }, + { + "epoch": 4.56, + "learning_rate": 2.721542495084116e-06, + "loss": 1.2635, + "step": 20907 + }, + { + "epoch": 4.56, + "learning_rate": 2.721433253222635e-06, + "loss": 2.2152, + "step": 20908 + }, + { + "epoch": 4.56, + "learning_rate": 2.7213240113611537e-06, + "loss": 2.5515, + "step": 20909 + }, + { + "epoch": 4.56, + "learning_rate": 2.7212147694996728e-06, + "loss": 2.3441, + "step": 20910 + }, + { + "epoch": 4.56, + "learning_rate": 2.721105527638191e-06, + "loss": 1.5721, + "step": 20911 + }, + { + "epoch": 4.56, + "learning_rate": 2.7209962857767096e-06, + "loss": 1.9165, + "step": 20912 + }, + { + "epoch": 4.56, + "learning_rate": 2.7208870439152286e-06, + "loss": 1.394, + "step": 20913 + }, + { + "epoch": 4.56, + "learning_rate": 2.7207778020537472e-06, + "loss": 1.8193, + "step": 20914 + }, + { + "epoch": 4.56, + "learning_rate": 2.7206685601922662e-06, + "loss": 0.4054, + "step": 20915 + }, + { + "epoch": 4.56, + "learning_rate": 2.7205593183307844e-06, + "loss": 0.9545, + "step": 20916 + }, + { + "epoch": 4.57, + "learning_rate": 2.720450076469303e-06, + "loss": 0.8401, + "step": 20917 + }, + { + "epoch": 4.57, + "learning_rate": 2.720340834607822e-06, + "loss": 1.1121, + "step": 20918 + }, + { + "epoch": 4.57, + "learning_rate": 2.7202315927463407e-06, + "loss": 1.297, + "step": 20919 + }, + { + "epoch": 4.57, + "learning_rate": 2.7201223508848597e-06, + "loss": 1.2361, + "step": 20920 + }, + { + "epoch": 4.57, + "learning_rate": 2.720013109023378e-06, + "loss": 1.6727, + "step": 20921 + }, + { + "epoch": 4.57, + "learning_rate": 2.7199038671618965e-06, + "loss": 1.8205, + "step": 20922 + }, + { + "epoch": 4.57, + "learning_rate": 2.7197946253004156e-06, + "loss": 2.0665, + "step": 20923 + }, + { + "epoch": 4.57, + "learning_rate": 2.719685383438934e-06, + "loss": 2.5581, + "step": 20924 + }, + { + "epoch": 4.57, + "learning_rate": 2.7195761415774528e-06, + "loss": 1.2842, + "step": 20925 + }, + { + "epoch": 4.57, + "learning_rate": 2.7194668997159714e-06, + "loss": 2.7483, + "step": 20926 + }, + { + "epoch": 4.57, + "learning_rate": 2.71935765785449e-06, + "loss": 1.3514, + "step": 20927 + }, + { + "epoch": 4.57, + "learning_rate": 2.7192484159930086e-06, + "loss": 1.3789, + "step": 20928 + }, + { + "epoch": 4.57, + "learning_rate": 2.7191391741315276e-06, + "loss": 2.1737, + "step": 20929 + }, + { + "epoch": 4.57, + "learning_rate": 2.719029932270046e-06, + "loss": 1.8816, + "step": 20930 + }, + { + "epoch": 4.57, + "learning_rate": 2.7189206904085644e-06, + "loss": 2.8771, + "step": 20931 + }, + { + "epoch": 4.57, + "learning_rate": 2.7188114485470835e-06, + "loss": 1.9302, + "step": 20932 + }, + { + "epoch": 4.57, + "learning_rate": 2.718702206685602e-06, + "loss": 1.5527, + "step": 20933 + }, + { + "epoch": 4.57, + "learning_rate": 2.718592964824121e-06, + "loss": 1.207, + "step": 20934 + }, + { + "epoch": 4.57, + "learning_rate": 2.7184837229626393e-06, + "loss": 1.1498, + "step": 20935 + }, + { + "epoch": 4.57, + "learning_rate": 2.718374481101158e-06, + "loss": 2.3632, + "step": 20936 + }, + { + "epoch": 4.57, + "learning_rate": 2.718265239239677e-06, + "loss": 1.5543, + "step": 20937 + }, + { + "epoch": 4.57, + "learning_rate": 2.7181559973781956e-06, + "loss": 2.8731, + "step": 20938 + }, + { + "epoch": 4.57, + "learning_rate": 2.7180467555167146e-06, + "loss": 3.5358, + "step": 20939 + }, + { + "epoch": 4.57, + "learning_rate": 2.717937513655233e-06, + "loss": 2.2645, + "step": 20940 + }, + { + "epoch": 4.57, + "learning_rate": 2.7178282717937514e-06, + "loss": 1.5952, + "step": 20941 + }, + { + "epoch": 4.57, + "learning_rate": 2.7177190299322704e-06, + "loss": 2.0327, + "step": 20942 + }, + { + "epoch": 4.57, + "learning_rate": 2.717609788070789e-06, + "loss": 1.0588, + "step": 20943 + }, + { + "epoch": 4.57, + "learning_rate": 2.7175005462093072e-06, + "loss": 1.5239, + "step": 20944 + }, + { + "epoch": 4.57, + "learning_rate": 2.7173913043478263e-06, + "loss": 0.7199, + "step": 20945 + }, + { + "epoch": 4.57, + "learning_rate": 2.717282062486345e-06, + "loss": 1.3451, + "step": 20946 + }, + { + "epoch": 4.57, + "learning_rate": 2.717172820624864e-06, + "loss": 3.3842, + "step": 20947 + }, + { + "epoch": 4.57, + "learning_rate": 2.7170635787633825e-06, + "loss": 1.2954, + "step": 20948 + }, + { + "epoch": 4.57, + "learning_rate": 2.7169543369019007e-06, + "loss": 1.9256, + "step": 20949 + }, + { + "epoch": 4.57, + "learning_rate": 2.7168450950404198e-06, + "loss": 1.5902, + "step": 20950 + }, + { + "epoch": 4.57, + "learning_rate": 2.7167358531789384e-06, + "loss": 2.1601, + "step": 20951 + }, + { + "epoch": 4.57, + "learning_rate": 2.7166266113174574e-06, + "loss": 1.4097, + "step": 20952 + }, + { + "epoch": 4.57, + "learning_rate": 2.716517369455976e-06, + "loss": 1.0844, + "step": 20953 + }, + { + "epoch": 4.57, + "learning_rate": 2.716408127594494e-06, + "loss": 1.3379, + "step": 20954 + }, + { + "epoch": 4.57, + "learning_rate": 2.7162988857330132e-06, + "loss": 1.8496, + "step": 20955 + }, + { + "epoch": 4.57, + "learning_rate": 2.716189643871532e-06, + "loss": 0.7887, + "step": 20956 + }, + { + "epoch": 4.57, + "learning_rate": 2.7160804020100505e-06, + "loss": 1.7353, + "step": 20957 + }, + { + "epoch": 4.57, + "learning_rate": 2.7159711601485695e-06, + "loss": 1.64, + "step": 20958 + }, + { + "epoch": 4.57, + "learning_rate": 2.7158619182870877e-06, + "loss": 2.8586, + "step": 20959 + }, + { + "epoch": 4.57, + "learning_rate": 2.7157526764256063e-06, + "loss": 1.4045, + "step": 20960 + }, + { + "epoch": 4.57, + "learning_rate": 2.7156434345641253e-06, + "loss": 2.6915, + "step": 20961 + }, + { + "epoch": 4.57, + "learning_rate": 2.715534192702644e-06, + "loss": 2.0664, + "step": 20962 + }, + { + "epoch": 4.58, + "learning_rate": 2.715424950841162e-06, + "loss": 2.5726, + "step": 20963 + }, + { + "epoch": 4.58, + "learning_rate": 2.715315708979681e-06, + "loss": 1.3033, + "step": 20964 + }, + { + "epoch": 4.58, + "learning_rate": 2.7152064671181998e-06, + "loss": 1.2832, + "step": 20965 + }, + { + "epoch": 4.58, + "learning_rate": 2.715097225256719e-06, + "loss": 2.4278, + "step": 20966 + }, + { + "epoch": 4.58, + "learning_rate": 2.7149879833952374e-06, + "loss": 1.1895, + "step": 20967 + }, + { + "epoch": 4.58, + "learning_rate": 2.7148787415337556e-06, + "loss": 1.6839, + "step": 20968 + }, + { + "epoch": 4.58, + "learning_rate": 2.7147694996722746e-06, + "loss": 1.1897, + "step": 20969 + }, + { + "epoch": 4.58, + "learning_rate": 2.7146602578107932e-06, + "loss": 2.2117, + "step": 20970 + }, + { + "epoch": 4.58, + "learning_rate": 2.7145510159493123e-06, + "loss": 1.0065, + "step": 20971 + }, + { + "epoch": 4.58, + "learning_rate": 2.714441774087831e-06, + "loss": 2.3311, + "step": 20972 + }, + { + "epoch": 4.58, + "learning_rate": 2.714332532226349e-06, + "loss": 0.7248, + "step": 20973 + }, + { + "epoch": 4.58, + "learning_rate": 2.714223290364868e-06, + "loss": 1.5819, + "step": 20974 + }, + { + "epoch": 4.58, + "learning_rate": 2.7141140485033867e-06, + "loss": 1.1315, + "step": 20975 + }, + { + "epoch": 4.58, + "learning_rate": 2.7140048066419058e-06, + "loss": 1.8836, + "step": 20976 + }, + { + "epoch": 4.58, + "learning_rate": 2.713895564780424e-06, + "loss": 2.5617, + "step": 20977 + }, + { + "epoch": 4.58, + "learning_rate": 2.7137863229189426e-06, + "loss": 2.7992, + "step": 20978 + }, + { + "epoch": 4.58, + "learning_rate": 2.7136770810574616e-06, + "loss": 1.5948, + "step": 20979 + }, + { + "epoch": 4.58, + "learning_rate": 2.71356783919598e-06, + "loss": 2.0718, + "step": 20980 + }, + { + "epoch": 4.58, + "learning_rate": 2.713458597334499e-06, + "loss": 1.2924, + "step": 20981 + }, + { + "epoch": 4.58, + "learning_rate": 2.7133493554730174e-06, + "loss": 2.3043, + "step": 20982 + }, + { + "epoch": 4.58, + "learning_rate": 2.713240113611536e-06, + "loss": 2.0966, + "step": 20983 + }, + { + "epoch": 4.58, + "learning_rate": 2.7131308717500546e-06, + "loss": 2.0489, + "step": 20984 + }, + { + "epoch": 4.58, + "learning_rate": 2.7130216298885737e-06, + "loss": 2.9729, + "step": 20985 + }, + { + "epoch": 4.58, + "learning_rate": 2.7129123880270923e-06, + "loss": 2.4786, + "step": 20986 + }, + { + "epoch": 4.58, + "learning_rate": 2.712803146165611e-06, + "loss": 2.5046, + "step": 20987 + }, + { + "epoch": 4.58, + "learning_rate": 2.7126939043041295e-06, + "loss": 0.9611, + "step": 20988 + }, + { + "epoch": 4.58, + "learning_rate": 2.712584662442648e-06, + "loss": 1.7809, + "step": 20989 + }, + { + "epoch": 4.58, + "learning_rate": 2.712475420581167e-06, + "loss": 1.1047, + "step": 20990 + }, + { + "epoch": 4.58, + "learning_rate": 2.7123661787196858e-06, + "loss": 4.2479, + "step": 20991 + }, + { + "epoch": 4.58, + "learning_rate": 2.712256936858204e-06, + "loss": 2.8833, + "step": 20992 + }, + { + "epoch": 4.58, + "learning_rate": 2.712147694996723e-06, + "loss": 1.0541, + "step": 20993 + }, + { + "epoch": 4.58, + "learning_rate": 2.7120384531352416e-06, + "loss": 0.6848, + "step": 20994 + }, + { + "epoch": 4.58, + "learning_rate": 2.7119292112737606e-06, + "loss": 2.5475, + "step": 20995 + }, + { + "epoch": 4.58, + "learning_rate": 2.711819969412279e-06, + "loss": 1.485, + "step": 20996 + }, + { + "epoch": 4.58, + "learning_rate": 2.7117107275507974e-06, + "loss": 1.7621, + "step": 20997 + }, + { + "epoch": 4.58, + "learning_rate": 2.7116014856893165e-06, + "loss": 1.7359, + "step": 20998 + }, + { + "epoch": 4.58, + "learning_rate": 2.711492243827835e-06, + "loss": 1.9236, + "step": 20999 + }, + { + "epoch": 4.58, + "learning_rate": 2.711383001966354e-06, + "loss": 1.3678, + "step": 21000 + }, + { + "epoch": 4.58, + "learning_rate": 2.7112737601048723e-06, + "loss": 1.1045, + "step": 21001 + }, + { + "epoch": 4.58, + "learning_rate": 2.711164518243391e-06, + "loss": 0.1121, + "step": 21002 + }, + { + "epoch": 4.58, + "learning_rate": 2.71105527638191e-06, + "loss": 1.3878, + "step": 21003 + }, + { + "epoch": 4.58, + "learning_rate": 2.7109460345204286e-06, + "loss": 1.0162, + "step": 21004 + }, + { + "epoch": 4.58, + "learning_rate": 2.7108367926589476e-06, + "loss": 1.7444, + "step": 21005 + }, + { + "epoch": 4.58, + "learning_rate": 2.7107275507974658e-06, + "loss": 1.7118, + "step": 21006 + }, + { + "epoch": 4.58, + "learning_rate": 2.7106183089359844e-06, + "loss": 1.254, + "step": 21007 + }, + { + "epoch": 4.58, + "learning_rate": 2.7105090670745034e-06, + "loss": 2.4959, + "step": 21008 + }, + { + "epoch": 4.59, + "learning_rate": 2.710399825213022e-06, + "loss": 3.3116, + "step": 21009 + }, + { + "epoch": 4.59, + "learning_rate": 2.7102905833515402e-06, + "loss": 2.0714, + "step": 21010 + }, + { + "epoch": 4.59, + "learning_rate": 2.7101813414900593e-06, + "loss": 1.3267, + "step": 21011 + }, + { + "epoch": 4.59, + "learning_rate": 2.710072099628578e-06, + "loss": 0.5574, + "step": 21012 + }, + { + "epoch": 4.59, + "learning_rate": 2.7099628577670965e-06, + "loss": 0.4292, + "step": 21013 + }, + { + "epoch": 4.59, + "learning_rate": 2.7098536159056155e-06, + "loss": 1.7612, + "step": 21014 + }, + { + "epoch": 4.59, + "learning_rate": 2.7097443740441337e-06, + "loss": 2.4238, + "step": 21015 + }, + { + "epoch": 4.59, + "learning_rate": 2.7096351321826523e-06, + "loss": 2.0235, + "step": 21016 + }, + { + "epoch": 4.59, + "learning_rate": 2.7095258903211713e-06, + "loss": 2.368, + "step": 21017 + }, + { + "epoch": 4.59, + "learning_rate": 2.70941664845969e-06, + "loss": 1.7518, + "step": 21018 + }, + { + "epoch": 4.59, + "learning_rate": 2.709307406598209e-06, + "loss": 1.7599, + "step": 21019 + }, + { + "epoch": 4.59, + "learning_rate": 2.709198164736727e-06, + "loss": 0.7863, + "step": 21020 + }, + { + "epoch": 4.59, + "learning_rate": 2.709088922875246e-06, + "loss": 3.5569, + "step": 21021 + }, + { + "epoch": 4.59, + "learning_rate": 2.708979681013765e-06, + "loss": 0.7697, + "step": 21022 + }, + { + "epoch": 4.59, + "learning_rate": 2.7088704391522834e-06, + "loss": 0.7839, + "step": 21023 + }, + { + "epoch": 4.59, + "learning_rate": 2.7087611972908016e-06, + "loss": 1.8195, + "step": 21024 + }, + { + "epoch": 4.59, + "learning_rate": 2.7086519554293207e-06, + "loss": 0.7919, + "step": 21025 + }, + { + "epoch": 4.59, + "learning_rate": 2.7085427135678393e-06, + "loss": 2.46, + "step": 21026 + }, + { + "epoch": 4.59, + "learning_rate": 2.7084334717063583e-06, + "loss": 2.0449, + "step": 21027 + }, + { + "epoch": 4.59, + "learning_rate": 2.708324229844877e-06, + "loss": 3.4301, + "step": 21028 + }, + { + "epoch": 4.59, + "learning_rate": 2.708214987983395e-06, + "loss": 1.3158, + "step": 21029 + }, + { + "epoch": 4.59, + "learning_rate": 2.708105746121914e-06, + "loss": 1.8362, + "step": 21030 + }, + { + "epoch": 4.59, + "learning_rate": 2.7079965042604327e-06, + "loss": 1.9649, + "step": 21031 + }, + { + "epoch": 4.59, + "learning_rate": 2.7078872623989518e-06, + "loss": 3.2961, + "step": 21032 + }, + { + "epoch": 4.59, + "learning_rate": 2.7077780205374704e-06, + "loss": 1.3901, + "step": 21033 + }, + { + "epoch": 4.59, + "learning_rate": 2.7076687786759886e-06, + "loss": 2.6816, + "step": 21034 + }, + { + "epoch": 4.59, + "learning_rate": 2.7075595368145076e-06, + "loss": 0.6557, + "step": 21035 + }, + { + "epoch": 4.59, + "learning_rate": 2.7074502949530262e-06, + "loss": 1.3205, + "step": 21036 + }, + { + "epoch": 4.59, + "learning_rate": 2.707341053091545e-06, + "loss": 2.0695, + "step": 21037 + }, + { + "epoch": 4.59, + "learning_rate": 2.707231811230064e-06, + "loss": 1.826, + "step": 21038 + }, + { + "epoch": 4.59, + "learning_rate": 2.707122569368582e-06, + "loss": 0.5657, + "step": 21039 + }, + { + "epoch": 4.59, + "learning_rate": 2.707013327507101e-06, + "loss": 2.8799, + "step": 21040 + }, + { + "epoch": 4.59, + "learning_rate": 2.7069040856456197e-06, + "loss": 1.9098, + "step": 21041 + }, + { + "epoch": 4.59, + "learning_rate": 2.7067948437841383e-06, + "loss": 0.7809, + "step": 21042 + }, + { + "epoch": 4.59, + "learning_rate": 2.706685601922657e-06, + "loss": 3.1395, + "step": 21043 + }, + { + "epoch": 4.59, + "learning_rate": 2.7065763600611755e-06, + "loss": 0.9378, + "step": 21044 + }, + { + "epoch": 4.59, + "learning_rate": 2.706467118199694e-06, + "loss": 2.6284, + "step": 21045 + }, + { + "epoch": 4.59, + "learning_rate": 2.706357876338213e-06, + "loss": 1.4754, + "step": 21046 + }, + { + "epoch": 4.59, + "learning_rate": 2.706248634476732e-06, + "loss": 1.6823, + "step": 21047 + }, + { + "epoch": 4.59, + "learning_rate": 2.70613939261525e-06, + "loss": 0.9901, + "step": 21048 + }, + { + "epoch": 4.59, + "learning_rate": 2.706030150753769e-06, + "loss": 1.8369, + "step": 21049 + }, + { + "epoch": 4.59, + "learning_rate": 2.7059209088922876e-06, + "loss": 1.1212, + "step": 21050 + }, + { + "epoch": 4.59, + "learning_rate": 2.7058116670308067e-06, + "loss": 1.8932, + "step": 21051 + }, + { + "epoch": 4.59, + "learning_rate": 2.7057024251693253e-06, + "loss": 1.8195, + "step": 21052 + }, + { + "epoch": 4.59, + "learning_rate": 2.7055931833078435e-06, + "loss": 1.9687, + "step": 21053 + }, + { + "epoch": 4.59, + "learning_rate": 2.7054839414463625e-06, + "loss": 2.1109, + "step": 21054 + }, + { + "epoch": 4.6, + "learning_rate": 2.705374699584881e-06, + "loss": 1.0492, + "step": 21055 + }, + { + "epoch": 4.6, + "learning_rate": 2.7052654577234e-06, + "loss": 1.8294, + "step": 21056 + }, + { + "epoch": 4.6, + "learning_rate": 2.7051562158619183e-06, + "loss": 1.1476, + "step": 21057 + }, + { + "epoch": 4.6, + "learning_rate": 2.705046974000437e-06, + "loss": 3.7322, + "step": 21058 + }, + { + "epoch": 4.6, + "learning_rate": 2.704937732138956e-06, + "loss": 0.9957, + "step": 21059 + }, + { + "epoch": 4.6, + "learning_rate": 2.7048284902774746e-06, + "loss": 0.7532, + "step": 21060 + }, + { + "epoch": 4.6, + "learning_rate": 2.7047192484159936e-06, + "loss": 1.5368, + "step": 21061 + }, + { + "epoch": 4.6, + "learning_rate": 2.704610006554512e-06, + "loss": 1.628, + "step": 21062 + }, + { + "epoch": 4.6, + "learning_rate": 2.7045007646930304e-06, + "loss": 0.1844, + "step": 21063 + }, + { + "epoch": 4.6, + "learning_rate": 2.7043915228315495e-06, + "loss": 1.4353, + "step": 21064 + }, + { + "epoch": 4.6, + "learning_rate": 2.704282280970068e-06, + "loss": 2.9361, + "step": 21065 + }, + { + "epoch": 4.6, + "learning_rate": 2.7041730391085867e-06, + "loss": 1.8388, + "step": 21066 + }, + { + "epoch": 4.6, + "learning_rate": 2.7040637972471053e-06, + "loss": 2.4758, + "step": 21067 + }, + { + "epoch": 4.6, + "learning_rate": 2.703954555385624e-06, + "loss": 1.5171, + "step": 21068 + }, + { + "epoch": 4.6, + "learning_rate": 2.7038453135241425e-06, + "loss": 1.0868, + "step": 21069 + }, + { + "epoch": 4.6, + "learning_rate": 2.7037360716626615e-06, + "loss": 1.1623, + "step": 21070 + }, + { + "epoch": 4.6, + "learning_rate": 2.70362682980118e-06, + "loss": 2.1357, + "step": 21071 + }, + { + "epoch": 4.6, + "learning_rate": 2.7035175879396983e-06, + "loss": 1.3008, + "step": 21072 + }, + { + "epoch": 4.6, + "learning_rate": 2.7034083460782174e-06, + "loss": 1.9684, + "step": 21073 + }, + { + "epoch": 4.6, + "learning_rate": 2.703299104216736e-06, + "loss": 2.3722, + "step": 21074 + }, + { + "epoch": 4.6, + "learning_rate": 2.703189862355255e-06, + "loss": 1.017, + "step": 21075 + }, + { + "epoch": 4.6, + "learning_rate": 2.703080620493773e-06, + "loss": 1.3938, + "step": 21076 + }, + { + "epoch": 4.6, + "learning_rate": 2.702971378632292e-06, + "loss": 2.6408, + "step": 21077 + }, + { + "epoch": 4.6, + "learning_rate": 2.702862136770811e-06, + "loss": 1.4651, + "step": 21078 + }, + { + "epoch": 4.6, + "learning_rate": 2.7027528949093295e-06, + "loss": 2.4697, + "step": 21079 + }, + { + "epoch": 4.6, + "learning_rate": 2.7026436530478485e-06, + "loss": 2.7261, + "step": 21080 + }, + { + "epoch": 4.6, + "learning_rate": 2.7025344111863667e-06, + "loss": 0.9701, + "step": 21081 + }, + { + "epoch": 4.6, + "learning_rate": 2.7024251693248853e-06, + "loss": 2.2181, + "step": 21082 + }, + { + "epoch": 4.6, + "learning_rate": 2.7023159274634043e-06, + "loss": 1.5239, + "step": 21083 + }, + { + "epoch": 4.6, + "learning_rate": 2.702206685601923e-06, + "loss": 0.5595, + "step": 21084 + }, + { + "epoch": 4.6, + "learning_rate": 2.702097443740442e-06, + "loss": 1.5008, + "step": 21085 + }, + { + "epoch": 4.6, + "learning_rate": 2.70198820187896e-06, + "loss": 0.6062, + "step": 21086 + }, + { + "epoch": 4.6, + "learning_rate": 2.7018789600174788e-06, + "loss": 2.9635, + "step": 21087 + }, + { + "epoch": 4.6, + "learning_rate": 2.701769718155998e-06, + "loss": 1.503, + "step": 21088 + }, + { + "epoch": 4.6, + "learning_rate": 2.7016604762945164e-06, + "loss": 1.274, + "step": 21089 + }, + { + "epoch": 4.6, + "learning_rate": 2.7015512344330346e-06, + "loss": 1.7305, + "step": 21090 + }, + { + "epoch": 4.6, + "learning_rate": 2.7014419925715536e-06, + "loss": 0.573, + "step": 21091 + }, + { + "epoch": 4.6, + "learning_rate": 2.7013327507100723e-06, + "loss": 2.1662, + "step": 21092 + }, + { + "epoch": 4.6, + "learning_rate": 2.7012235088485913e-06, + "loss": 1.2731, + "step": 21093 + }, + { + "epoch": 4.6, + "learning_rate": 2.70111426698711e-06, + "loss": 1.4498, + "step": 21094 + }, + { + "epoch": 4.6, + "learning_rate": 2.701005025125628e-06, + "loss": 2.6474, + "step": 21095 + }, + { + "epoch": 4.6, + "learning_rate": 2.700895783264147e-06, + "loss": 0.8664, + "step": 21096 + }, + { + "epoch": 4.6, + "learning_rate": 2.7007865414026657e-06, + "loss": 0.8058, + "step": 21097 + }, + { + "epoch": 4.6, + "learning_rate": 2.7006772995411843e-06, + "loss": 0.3582, + "step": 21098 + }, + { + "epoch": 4.6, + "learning_rate": 2.7005680576797034e-06, + "loss": 0.9402, + "step": 21099 + }, + { + "epoch": 4.6, + "learning_rate": 2.7004588158182216e-06, + "loss": 3.3009, + "step": 21100 + }, + { + "epoch": 4.61, + "learning_rate": 2.70034957395674e-06, + "loss": 0.5874, + "step": 21101 + }, + { + "epoch": 4.61, + "learning_rate": 2.700240332095259e-06, + "loss": 1.4697, + "step": 21102 + }, + { + "epoch": 4.61, + "learning_rate": 2.700131090233778e-06, + "loss": 2.8639, + "step": 21103 + }, + { + "epoch": 4.61, + "learning_rate": 2.700021848372297e-06, + "loss": 2.3867, + "step": 21104 + }, + { + "epoch": 4.61, + "learning_rate": 2.699912606510815e-06, + "loss": 1.6116, + "step": 21105 + }, + { + "epoch": 4.61, + "learning_rate": 2.6998033646493337e-06, + "loss": 1.7992, + "step": 21106 + }, + { + "epoch": 4.61, + "learning_rate": 2.6996941227878527e-06, + "loss": 1.6006, + "step": 21107 + }, + { + "epoch": 4.61, + "learning_rate": 2.6995848809263713e-06, + "loss": 1.4823, + "step": 21108 + }, + { + "epoch": 4.61, + "learning_rate": 2.6994756390648895e-06, + "loss": 2.1213, + "step": 21109 + }, + { + "epoch": 4.61, + "learning_rate": 2.6993663972034085e-06, + "loss": 2.2491, + "step": 21110 + }, + { + "epoch": 4.61, + "learning_rate": 2.699257155341927e-06, + "loss": 2.2104, + "step": 21111 + }, + { + "epoch": 4.61, + "learning_rate": 2.699147913480446e-06, + "loss": 2.2916, + "step": 21112 + }, + { + "epoch": 4.61, + "learning_rate": 2.6990386716189648e-06, + "loss": 2.0763, + "step": 21113 + }, + { + "epoch": 4.61, + "learning_rate": 2.698929429757483e-06, + "loss": 1.4047, + "step": 21114 + }, + { + "epoch": 4.61, + "learning_rate": 2.698820187896002e-06, + "loss": 0.311, + "step": 21115 + }, + { + "epoch": 4.61, + "learning_rate": 2.6987109460345206e-06, + "loss": 1.1581, + "step": 21116 + }, + { + "epoch": 4.61, + "learning_rate": 2.6986017041730396e-06, + "loss": 2.3371, + "step": 21117 + }, + { + "epoch": 4.61, + "learning_rate": 2.6984924623115583e-06, + "loss": 1.6852, + "step": 21118 + }, + { + "epoch": 4.61, + "learning_rate": 2.6983832204500764e-06, + "loss": 1.252, + "step": 21119 + }, + { + "epoch": 4.61, + "learning_rate": 2.6982739785885955e-06, + "loss": 2.4341, + "step": 21120 + }, + { + "epoch": 4.61, + "learning_rate": 2.698164736727114e-06, + "loss": 1.6388, + "step": 21121 + }, + { + "epoch": 4.61, + "learning_rate": 2.6980554948656327e-06, + "loss": 1.43, + "step": 21122 + }, + { + "epoch": 4.61, + "learning_rate": 2.6979462530041513e-06, + "loss": 2.3075, + "step": 21123 + }, + { + "epoch": 4.61, + "learning_rate": 2.69783701114267e-06, + "loss": 0.8365, + "step": 21124 + }, + { + "epoch": 4.61, + "learning_rate": 2.6977277692811885e-06, + "loss": 1.1338, + "step": 21125 + }, + { + "epoch": 4.61, + "learning_rate": 2.6976185274197076e-06, + "loss": 1.1681, + "step": 21126 + }, + { + "epoch": 4.61, + "learning_rate": 2.697509285558226e-06, + "loss": 0.9511, + "step": 21127 + }, + { + "epoch": 4.61, + "learning_rate": 2.6974000436967444e-06, + "loss": 0.4428, + "step": 21128 + }, + { + "epoch": 4.61, + "learning_rate": 2.6972908018352634e-06, + "loss": 0.7176, + "step": 21129 + }, + { + "epoch": 4.61, + "learning_rate": 2.697181559973782e-06, + "loss": 2.4318, + "step": 21130 + }, + { + "epoch": 4.61, + "learning_rate": 2.697072318112301e-06, + "loss": 1.6613, + "step": 21131 + }, + { + "epoch": 4.61, + "learning_rate": 2.6969630762508197e-06, + "loss": 1.7475, + "step": 21132 + }, + { + "epoch": 4.61, + "learning_rate": 2.696853834389338e-06, + "loss": 2.9592, + "step": 21133 + }, + { + "epoch": 4.61, + "learning_rate": 2.696744592527857e-06, + "loss": 2.1055, + "step": 21134 + }, + { + "epoch": 4.61, + "learning_rate": 2.6966353506663755e-06, + "loss": 2.0022, + "step": 21135 + }, + { + "epoch": 4.61, + "learning_rate": 2.6965261088048945e-06, + "loss": 0.2937, + "step": 21136 + }, + { + "epoch": 4.61, + "learning_rate": 2.696416866943413e-06, + "loss": 2.7147, + "step": 21137 + }, + { + "epoch": 4.61, + "learning_rate": 2.6963076250819313e-06, + "loss": 2.5923, + "step": 21138 + }, + { + "epoch": 4.61, + "learning_rate": 2.6961983832204504e-06, + "loss": 2.2202, + "step": 21139 + }, + { + "epoch": 4.61, + "learning_rate": 2.696089141358969e-06, + "loss": 1.2314, + "step": 21140 + }, + { + "epoch": 4.61, + "learning_rate": 2.695979899497488e-06, + "loss": 0.743, + "step": 21141 + }, + { + "epoch": 4.61, + "learning_rate": 2.695870657636006e-06, + "loss": 2.4328, + "step": 21142 + }, + { + "epoch": 4.61, + "learning_rate": 2.695761415774525e-06, + "loss": 0.8757, + "step": 21143 + }, + { + "epoch": 4.61, + "learning_rate": 2.695652173913044e-06, + "loss": 1.4839, + "step": 21144 + }, + { + "epoch": 4.61, + "learning_rate": 2.6955429320515624e-06, + "loss": 1.5972, + "step": 21145 + }, + { + "epoch": 4.62, + "learning_rate": 2.6954336901900815e-06, + "loss": 0.3947, + "step": 21146 + }, + { + "epoch": 4.62, + "learning_rate": 2.6953244483285997e-06, + "loss": 1.7594, + "step": 21147 + }, + { + "epoch": 4.62, + "learning_rate": 2.6952152064671183e-06, + "loss": 1.1968, + "step": 21148 + }, + { + "epoch": 4.62, + "learning_rate": 2.6951059646056373e-06, + "loss": 2.2713, + "step": 21149 + }, + { + "epoch": 4.62, + "learning_rate": 2.694996722744156e-06, + "loss": 2.6516, + "step": 21150 + }, + { + "epoch": 4.62, + "learning_rate": 2.6948874808826745e-06, + "loss": 1.337, + "step": 21151 + }, + { + "epoch": 4.62, + "learning_rate": 2.694778239021193e-06, + "loss": 1.5917, + "step": 21152 + }, + { + "epoch": 4.62, + "learning_rate": 2.6946689971597118e-06, + "loss": 1.4101, + "step": 21153 + }, + { + "epoch": 4.62, + "learning_rate": 2.6945597552982304e-06, + "loss": 2.5278, + "step": 21154 + }, + { + "epoch": 4.62, + "learning_rate": 2.6944505134367494e-06, + "loss": 2.3856, + "step": 21155 + }, + { + "epoch": 4.62, + "learning_rate": 2.6943412715752676e-06, + "loss": 1.5916, + "step": 21156 + }, + { + "epoch": 4.62, + "learning_rate": 2.694232029713786e-06, + "loss": 0.8766, + "step": 21157 + }, + { + "epoch": 4.62, + "learning_rate": 2.6941227878523052e-06, + "loss": 1.4485, + "step": 21158 + }, + { + "epoch": 4.62, + "learning_rate": 2.694013545990824e-06, + "loss": 1.8584, + "step": 21159 + }, + { + "epoch": 4.62, + "learning_rate": 2.693904304129343e-06, + "loss": 1.0353, + "step": 21160 + }, + { + "epoch": 4.62, + "learning_rate": 2.693795062267861e-06, + "loss": 1.8257, + "step": 21161 + }, + { + "epoch": 4.62, + "learning_rate": 2.6936858204063797e-06, + "loss": 1.3531, + "step": 21162 + }, + { + "epoch": 4.62, + "learning_rate": 2.6935765785448987e-06, + "loss": 1.1229, + "step": 21163 + }, + { + "epoch": 4.62, + "learning_rate": 2.6934673366834173e-06, + "loss": 0.3185, + "step": 21164 + }, + { + "epoch": 4.62, + "learning_rate": 2.6933580948219364e-06, + "loss": 1.3672, + "step": 21165 + }, + { + "epoch": 4.62, + "learning_rate": 2.6932488529604545e-06, + "loss": 3.5863, + "step": 21166 + }, + { + "epoch": 4.62, + "learning_rate": 2.693139611098973e-06, + "loss": 0.8533, + "step": 21167 + }, + { + "epoch": 4.62, + "learning_rate": 2.693030369237492e-06, + "loss": 1.1768, + "step": 21168 + }, + { + "epoch": 4.62, + "learning_rate": 2.692921127376011e-06, + "loss": 2.1054, + "step": 21169 + }, + { + "epoch": 4.62, + "learning_rate": 2.69281188551453e-06, + "loss": 0.7648, + "step": 21170 + }, + { + "epoch": 4.62, + "learning_rate": 2.692702643653048e-06, + "loss": 2.6983, + "step": 21171 + }, + { + "epoch": 4.62, + "learning_rate": 2.6925934017915666e-06, + "loss": 2.7397, + "step": 21172 + }, + { + "epoch": 4.62, + "learning_rate": 2.6924841599300857e-06, + "loss": 1.7331, + "step": 21173 + }, + { + "epoch": 4.62, + "learning_rate": 2.6923749180686043e-06, + "loss": 0.8777, + "step": 21174 + }, + { + "epoch": 4.62, + "learning_rate": 2.6922656762071225e-06, + "loss": 1.6759, + "step": 21175 + }, + { + "epoch": 4.62, + "learning_rate": 2.6921564343456415e-06, + "loss": 3.8716, + "step": 21176 + }, + { + "epoch": 4.62, + "learning_rate": 2.69204719248416e-06, + "loss": 3.1598, + "step": 21177 + }, + { + "epoch": 4.62, + "learning_rate": 2.6919379506226787e-06, + "loss": 2.9738, + "step": 21178 + }, + { + "epoch": 4.62, + "learning_rate": 2.6918287087611978e-06, + "loss": 1.1871, + "step": 21179 + }, + { + "epoch": 4.62, + "learning_rate": 2.691719466899716e-06, + "loss": 1.7257, + "step": 21180 + }, + { + "epoch": 4.62, + "learning_rate": 2.6916102250382346e-06, + "loss": 1.3592, + "step": 21181 + }, + { + "epoch": 4.62, + "learning_rate": 2.6915009831767536e-06, + "loss": 1.9744, + "step": 21182 + }, + { + "epoch": 4.62, + "learning_rate": 2.691391741315272e-06, + "loss": 2.5754, + "step": 21183 + }, + { + "epoch": 4.62, + "learning_rate": 2.6912824994537912e-06, + "loss": 1.9852, + "step": 21184 + }, + { + "epoch": 4.62, + "learning_rate": 2.6911732575923094e-06, + "loss": 0.7034, + "step": 21185 + }, + { + "epoch": 4.62, + "learning_rate": 2.691064015730828e-06, + "loss": 2.74, + "step": 21186 + }, + { + "epoch": 4.62, + "learning_rate": 2.690954773869347e-06, + "loss": 1.9195, + "step": 21187 + }, + { + "epoch": 4.62, + "learning_rate": 2.6908455320078657e-06, + "loss": 1.0884, + "step": 21188 + }, + { + "epoch": 4.62, + "learning_rate": 2.690736290146384e-06, + "loss": 1.5413, + "step": 21189 + }, + { + "epoch": 4.62, + "learning_rate": 2.690627048284903e-06, + "loss": 1.7728, + "step": 21190 + }, + { + "epoch": 4.62, + "learning_rate": 2.6905178064234215e-06, + "loss": 1.2347, + "step": 21191 + }, + { + "epoch": 4.63, + "learning_rate": 2.6904085645619406e-06, + "loss": 1.7783, + "step": 21192 + }, + { + "epoch": 4.63, + "learning_rate": 2.690299322700459e-06, + "loss": 3.4748, + "step": 21193 + }, + { + "epoch": 4.63, + "learning_rate": 2.6901900808389773e-06, + "loss": 2.2517, + "step": 21194 + }, + { + "epoch": 4.63, + "learning_rate": 2.6900808389774964e-06, + "loss": 1.5858, + "step": 21195 + }, + { + "epoch": 4.63, + "learning_rate": 2.689971597116015e-06, + "loss": 2.2066, + "step": 21196 + }, + { + "epoch": 4.63, + "learning_rate": 2.689862355254534e-06, + "loss": 2.1326, + "step": 21197 + }, + { + "epoch": 4.63, + "learning_rate": 2.6897531133930526e-06, + "loss": 2.4582, + "step": 21198 + }, + { + "epoch": 4.63, + "learning_rate": 2.689643871531571e-06, + "loss": 0.7521, + "step": 21199 + }, + { + "epoch": 4.63, + "learning_rate": 2.68953462967009e-06, + "loss": 2.4603, + "step": 21200 + }, + { + "epoch": 4.63, + "learning_rate": 2.6894253878086085e-06, + "loss": 1.1549, + "step": 21201 + }, + { + "epoch": 4.63, + "learning_rate": 2.6893161459471275e-06, + "loss": 1.2462, + "step": 21202 + }, + { + "epoch": 4.63, + "learning_rate": 2.689206904085646e-06, + "loss": 1.1181, + "step": 21203 + }, + { + "epoch": 4.63, + "learning_rate": 2.6890976622241643e-06, + "loss": 1.3484, + "step": 21204 + }, + { + "epoch": 4.63, + "learning_rate": 2.6889884203626833e-06, + "loss": 1.1062, + "step": 21205 + }, + { + "epoch": 4.63, + "learning_rate": 2.688879178501202e-06, + "loss": 1.7621, + "step": 21206 + }, + { + "epoch": 4.63, + "learning_rate": 2.6887699366397206e-06, + "loss": 0.584, + "step": 21207 + }, + { + "epoch": 4.63, + "learning_rate": 2.688660694778239e-06, + "loss": 2.268, + "step": 21208 + }, + { + "epoch": 4.63, + "learning_rate": 2.6885514529167578e-06, + "loss": 1.9442, + "step": 21209 + }, + { + "epoch": 4.63, + "learning_rate": 2.6884422110552764e-06, + "loss": 1.796, + "step": 21210 + }, + { + "epoch": 4.63, + "learning_rate": 2.6883329691937954e-06, + "loss": 0.7523, + "step": 21211 + }, + { + "epoch": 4.63, + "learning_rate": 2.688223727332314e-06, + "loss": 1.9403, + "step": 21212 + }, + { + "epoch": 4.63, + "learning_rate": 2.6881144854708322e-06, + "loss": 1.3469, + "step": 21213 + }, + { + "epoch": 4.63, + "learning_rate": 2.6880052436093513e-06, + "loss": 1.9757, + "step": 21214 + }, + { + "epoch": 4.63, + "learning_rate": 2.68789600174787e-06, + "loss": 2.0546, + "step": 21215 + }, + { + "epoch": 4.63, + "learning_rate": 2.687786759886389e-06, + "loss": 3.0512, + "step": 21216 + }, + { + "epoch": 4.63, + "learning_rate": 2.6876775180249075e-06, + "loss": 2.7872, + "step": 21217 + }, + { + "epoch": 4.63, + "learning_rate": 2.6875682761634257e-06, + "loss": 1.337, + "step": 21218 + }, + { + "epoch": 4.63, + "learning_rate": 2.6874590343019447e-06, + "loss": 0.8349, + "step": 21219 + }, + { + "epoch": 4.63, + "learning_rate": 2.6873497924404634e-06, + "loss": 1.5532, + "step": 21220 + }, + { + "epoch": 4.63, + "learning_rate": 2.6872405505789824e-06, + "loss": 2.2614, + "step": 21221 + }, + { + "epoch": 4.63, + "learning_rate": 2.6871313087175006e-06, + "loss": 1.8681, + "step": 21222 + }, + { + "epoch": 4.63, + "learning_rate": 2.687022066856019e-06, + "loss": 2.7242, + "step": 21223 + }, + { + "epoch": 4.63, + "learning_rate": 2.6869128249945382e-06, + "loss": 1.3939, + "step": 21224 + }, + { + "epoch": 4.63, + "learning_rate": 2.686803583133057e-06, + "loss": 0.4889, + "step": 21225 + }, + { + "epoch": 4.63, + "learning_rate": 2.686694341271576e-06, + "loss": 2.1601, + "step": 21226 + }, + { + "epoch": 4.63, + "learning_rate": 2.686585099410094e-06, + "loss": 2.0864, + "step": 21227 + }, + { + "epoch": 4.63, + "learning_rate": 2.6864758575486127e-06, + "loss": 2.2611, + "step": 21228 + }, + { + "epoch": 4.63, + "learning_rate": 2.6863666156871317e-06, + "loss": 1.2894, + "step": 21229 + }, + { + "epoch": 4.63, + "learning_rate": 2.6862573738256503e-06, + "loss": 0.509, + "step": 21230 + }, + { + "epoch": 4.63, + "learning_rate": 2.686148131964169e-06, + "loss": 1.6928, + "step": 21231 + }, + { + "epoch": 4.63, + "learning_rate": 2.6860388901026875e-06, + "loss": 1.9631, + "step": 21232 + }, + { + "epoch": 4.63, + "learning_rate": 2.685929648241206e-06, + "loss": 1.2508, + "step": 21233 + }, + { + "epoch": 4.63, + "learning_rate": 2.6858204063797248e-06, + "loss": 1.0229, + "step": 21234 + }, + { + "epoch": 4.63, + "learning_rate": 2.6857111645182438e-06, + "loss": 1.448, + "step": 21235 + }, + { + "epoch": 4.63, + "learning_rate": 2.6856019226567624e-06, + "loss": 0.1869, + "step": 21236 + }, + { + "epoch": 4.63, + "learning_rate": 2.685492680795281e-06, + "loss": 1.9167, + "step": 21237 + }, + { + "epoch": 4.64, + "learning_rate": 2.6853834389337996e-06, + "loss": 1.6941, + "step": 21238 + }, + { + "epoch": 4.64, + "learning_rate": 2.6852741970723182e-06, + "loss": 1.9308, + "step": 21239 + }, + { + "epoch": 4.64, + "learning_rate": 2.6851649552108373e-06, + "loss": 1.8216, + "step": 21240 + }, + { + "epoch": 4.64, + "learning_rate": 2.6850557133493555e-06, + "loss": 1.2124, + "step": 21241 + }, + { + "epoch": 4.64, + "learning_rate": 2.684946471487874e-06, + "loss": 2.2683, + "step": 21242 + }, + { + "epoch": 4.64, + "learning_rate": 2.684837229626393e-06, + "loss": 2.322, + "step": 21243 + }, + { + "epoch": 4.64, + "learning_rate": 2.6847279877649117e-06, + "loss": 2.0898, + "step": 21244 + }, + { + "epoch": 4.64, + "learning_rate": 2.6846187459034307e-06, + "loss": 0.7499, + "step": 21245 + }, + { + "epoch": 4.64, + "learning_rate": 2.684509504041949e-06, + "loss": 1.45, + "step": 21246 + }, + { + "epoch": 4.64, + "learning_rate": 2.6844002621804675e-06, + "loss": 2.5572, + "step": 21247 + }, + { + "epoch": 4.64, + "learning_rate": 2.6842910203189866e-06, + "loss": 1.6336, + "step": 21248 + }, + { + "epoch": 4.64, + "learning_rate": 2.684181778457505e-06, + "loss": 1.6666, + "step": 21249 + }, + { + "epoch": 4.64, + "learning_rate": 2.6840725365960242e-06, + "loss": 0.183, + "step": 21250 + }, + { + "epoch": 4.64, + "learning_rate": 2.6839632947345424e-06, + "loss": 2.4573, + "step": 21251 + }, + { + "epoch": 4.64, + "learning_rate": 2.683854052873061e-06, + "loss": 0.0795, + "step": 21252 + }, + { + "epoch": 4.64, + "learning_rate": 2.68374481101158e-06, + "loss": 1.2535, + "step": 21253 + }, + { + "epoch": 4.64, + "learning_rate": 2.6836355691500987e-06, + "loss": 2.6993, + "step": 21254 + }, + { + "epoch": 4.64, + "learning_rate": 2.683526327288617e-06, + "loss": 0.5733, + "step": 21255 + }, + { + "epoch": 4.64, + "learning_rate": 2.683417085427136e-06, + "loss": 1.5086, + "step": 21256 + }, + { + "epoch": 4.64, + "learning_rate": 2.6833078435656545e-06, + "loss": 1.7491, + "step": 21257 + }, + { + "epoch": 4.64, + "learning_rate": 2.6831986017041735e-06, + "loss": 1.2454, + "step": 21258 + }, + { + "epoch": 4.64, + "learning_rate": 2.683089359842692e-06, + "loss": 1.641, + "step": 21259 + }, + { + "epoch": 4.64, + "learning_rate": 2.6829801179812103e-06, + "loss": 0.7414, + "step": 21260 + }, + { + "epoch": 4.64, + "learning_rate": 2.6828708761197294e-06, + "loss": 0.3568, + "step": 21261 + }, + { + "epoch": 4.64, + "learning_rate": 2.682761634258248e-06, + "loss": 1.2165, + "step": 21262 + }, + { + "epoch": 4.64, + "learning_rate": 2.6826523923967666e-06, + "loss": 0.8908, + "step": 21263 + }, + { + "epoch": 4.64, + "learning_rate": 2.6825431505352856e-06, + "loss": 2.6752, + "step": 21264 + }, + { + "epoch": 4.64, + "learning_rate": 2.682433908673804e-06, + "loss": 2.1969, + "step": 21265 + }, + { + "epoch": 4.64, + "learning_rate": 2.6823246668123224e-06, + "loss": 0.9178, + "step": 21266 + }, + { + "epoch": 4.64, + "learning_rate": 2.6822154249508415e-06, + "loss": 0.6165, + "step": 21267 + }, + { + "epoch": 4.64, + "learning_rate": 2.68210618308936e-06, + "loss": 0.8557, + "step": 21268 + }, + { + "epoch": 4.64, + "learning_rate": 2.681996941227879e-06, + "loss": 1.1764, + "step": 21269 + }, + { + "epoch": 4.64, + "learning_rate": 2.6818876993663973e-06, + "loss": 1.5445, + "step": 21270 + }, + { + "epoch": 4.64, + "learning_rate": 2.681778457504916e-06, + "loss": 3.2172, + "step": 21271 + }, + { + "epoch": 4.64, + "learning_rate": 2.681669215643435e-06, + "loss": 1.2399, + "step": 21272 + }, + { + "epoch": 4.64, + "learning_rate": 2.6815599737819535e-06, + "loss": 1.0509, + "step": 21273 + }, + { + "epoch": 4.64, + "learning_rate": 2.6814507319204717e-06, + "loss": 2.7505, + "step": 21274 + }, + { + "epoch": 4.64, + "learning_rate": 2.6813414900589908e-06, + "loss": 2.0036, + "step": 21275 + }, + { + "epoch": 4.64, + "learning_rate": 2.6812322481975094e-06, + "loss": 1.542, + "step": 21276 + }, + { + "epoch": 4.64, + "learning_rate": 2.6811230063360284e-06, + "loss": 0.7247, + "step": 21277 + }, + { + "epoch": 4.64, + "learning_rate": 2.681013764474547e-06, + "loss": 0.3839, + "step": 21278 + }, + { + "epoch": 4.64, + "learning_rate": 2.6809045226130652e-06, + "loss": 1.3024, + "step": 21279 + }, + { + "epoch": 4.64, + "learning_rate": 2.6807952807515842e-06, + "loss": 1.0881, + "step": 21280 + }, + { + "epoch": 4.64, + "learning_rate": 2.680686038890103e-06, + "loss": 1.3952, + "step": 21281 + }, + { + "epoch": 4.64, + "learning_rate": 2.680576797028622e-06, + "loss": 0.6665, + "step": 21282 + }, + { + "epoch": 4.64, + "learning_rate": 2.6804675551671405e-06, + "loss": 0.7349, + "step": 21283 + }, + { + "epoch": 4.65, + "learning_rate": 2.6803583133056587e-06, + "loss": 2.252, + "step": 21284 + }, + { + "epoch": 4.65, + "learning_rate": 2.6802490714441777e-06, + "loss": 1.2339, + "step": 21285 + }, + { + "epoch": 4.65, + "learning_rate": 2.6801398295826963e-06, + "loss": 0.8692, + "step": 21286 + }, + { + "epoch": 4.65, + "learning_rate": 2.6800305877212154e-06, + "loss": 1.0458, + "step": 21287 + }, + { + "epoch": 4.65, + "learning_rate": 2.6799213458597336e-06, + "loss": 2.0039, + "step": 21288 + }, + { + "epoch": 4.65, + "learning_rate": 2.679812103998252e-06, + "loss": 2.1627, + "step": 21289 + }, + { + "epoch": 4.65, + "learning_rate": 2.679702862136771e-06, + "loss": 1.1836, + "step": 21290 + }, + { + "epoch": 4.65, + "learning_rate": 2.67959362027529e-06, + "loss": 0.9305, + "step": 21291 + }, + { + "epoch": 4.65, + "learning_rate": 2.6794843784138084e-06, + "loss": 1.5385, + "step": 21292 + }, + { + "epoch": 4.65, + "learning_rate": 2.679375136552327e-06, + "loss": 0.9482, + "step": 21293 + }, + { + "epoch": 4.65, + "learning_rate": 2.6792658946908456e-06, + "loss": 1.634, + "step": 21294 + }, + { + "epoch": 4.65, + "learning_rate": 2.6791566528293643e-06, + "loss": 0.8735, + "step": 21295 + }, + { + "epoch": 4.65, + "learning_rate": 2.6790474109678833e-06, + "loss": 1.2301, + "step": 21296 + }, + { + "epoch": 4.65, + "learning_rate": 2.678938169106402e-06, + "loss": 0.8651, + "step": 21297 + }, + { + "epoch": 4.65, + "learning_rate": 2.67882892724492e-06, + "loss": 1.3949, + "step": 21298 + }, + { + "epoch": 4.65, + "learning_rate": 2.678719685383439e-06, + "loss": 1.11, + "step": 21299 + }, + { + "epoch": 4.65, + "learning_rate": 2.6786104435219577e-06, + "loss": 1.3643, + "step": 21300 + }, + { + "epoch": 4.65, + "learning_rate": 2.6785012016604768e-06, + "loss": 0.8153, + "step": 21301 + }, + { + "epoch": 4.65, + "learning_rate": 2.6783919597989954e-06, + "loss": 1.7883, + "step": 21302 + }, + { + "epoch": 4.65, + "learning_rate": 2.6782827179375136e-06, + "loss": 2.4451, + "step": 21303 + }, + { + "epoch": 4.65, + "learning_rate": 2.6781734760760326e-06, + "loss": 1.6031, + "step": 21304 + }, + { + "epoch": 4.65, + "learning_rate": 2.6780642342145512e-06, + "loss": 1.0805, + "step": 21305 + }, + { + "epoch": 4.65, + "learning_rate": 2.6779549923530703e-06, + "loss": 1.1791, + "step": 21306 + }, + { + "epoch": 4.65, + "learning_rate": 2.6778457504915884e-06, + "loss": 1.1182, + "step": 21307 + }, + { + "epoch": 4.65, + "learning_rate": 2.677736508630107e-06, + "loss": 1.5275, + "step": 21308 + }, + { + "epoch": 4.65, + "learning_rate": 2.677627266768626e-06, + "loss": 2.603, + "step": 21309 + }, + { + "epoch": 4.65, + "learning_rate": 2.6775180249071447e-06, + "loss": 1.421, + "step": 21310 + }, + { + "epoch": 4.65, + "learning_rate": 2.6774087830456637e-06, + "loss": 1.5415, + "step": 21311 + }, + { + "epoch": 4.65, + "learning_rate": 2.677299541184182e-06, + "loss": 1.4762, + "step": 21312 + }, + { + "epoch": 4.65, + "learning_rate": 2.6771902993227005e-06, + "loss": 2.2991, + "step": 21313 + }, + { + "epoch": 4.65, + "learning_rate": 2.6770810574612196e-06, + "loss": 1.4261, + "step": 21314 + }, + { + "epoch": 4.65, + "learning_rate": 2.676971815599738e-06, + "loss": 1.2378, + "step": 21315 + }, + { + "epoch": 4.65, + "learning_rate": 2.6768625737382568e-06, + "loss": 0.3006, + "step": 21316 + }, + { + "epoch": 4.65, + "learning_rate": 2.6767533318767754e-06, + "loss": 1.5516, + "step": 21317 + }, + { + "epoch": 4.65, + "learning_rate": 2.676644090015294e-06, + "loss": 1.3609, + "step": 21318 + }, + { + "epoch": 4.65, + "learning_rate": 2.6765348481538126e-06, + "loss": 0.7072, + "step": 21319 + }, + { + "epoch": 4.65, + "learning_rate": 2.6764256062923317e-06, + "loss": 1.1499, + "step": 21320 + }, + { + "epoch": 4.65, + "learning_rate": 2.67631636443085e-06, + "loss": 2.2877, + "step": 21321 + }, + { + "epoch": 4.65, + "learning_rate": 2.6762071225693684e-06, + "loss": 1.5833, + "step": 21322 + }, + { + "epoch": 4.65, + "learning_rate": 2.6760978807078875e-06, + "loss": 1.861, + "step": 21323 + }, + { + "epoch": 4.65, + "learning_rate": 2.675988638846406e-06, + "loss": 2.1598, + "step": 21324 + }, + { + "epoch": 4.65, + "learning_rate": 2.675879396984925e-06, + "loss": 2.5425, + "step": 21325 + }, + { + "epoch": 4.65, + "learning_rate": 2.6757701551234433e-06, + "loss": 1.4944, + "step": 21326 + }, + { + "epoch": 4.65, + "learning_rate": 2.675660913261962e-06, + "loss": 1.5249, + "step": 21327 + }, + { + "epoch": 4.65, + "learning_rate": 2.675551671400481e-06, + "loss": 0.8561, + "step": 21328 + }, + { + "epoch": 4.65, + "learning_rate": 2.6754424295389996e-06, + "loss": 0.7489, + "step": 21329 + }, + { + "epoch": 4.66, + "learning_rate": 2.6753331876775186e-06, + "loss": 1.0374, + "step": 21330 + }, + { + "epoch": 4.66, + "learning_rate": 2.675223945816037e-06, + "loss": 2.5382, + "step": 21331 + }, + { + "epoch": 4.66, + "learning_rate": 2.6751147039545554e-06, + "loss": 1.2942, + "step": 21332 + }, + { + "epoch": 4.66, + "learning_rate": 2.6750054620930744e-06, + "loss": 3.3625, + "step": 21333 + }, + { + "epoch": 4.66, + "learning_rate": 2.674896220231593e-06, + "loss": 1.6413, + "step": 21334 + }, + { + "epoch": 4.66, + "learning_rate": 2.674786978370112e-06, + "loss": 0.2243, + "step": 21335 + }, + { + "epoch": 4.66, + "learning_rate": 2.6746777365086303e-06, + "loss": 0.6486, + "step": 21336 + }, + { + "epoch": 4.66, + "learning_rate": 2.674568494647149e-06, + "loss": 1.2116, + "step": 21337 + }, + { + "epoch": 4.66, + "learning_rate": 2.674459252785668e-06, + "loss": 0.6679, + "step": 21338 + }, + { + "epoch": 4.66, + "learning_rate": 2.6743500109241865e-06, + "loss": 2.2034, + "step": 21339 + }, + { + "epoch": 4.66, + "learning_rate": 2.6742407690627047e-06, + "loss": 0.9727, + "step": 21340 + }, + { + "epoch": 4.66, + "learning_rate": 2.6741315272012238e-06, + "loss": 0.4233, + "step": 21341 + }, + { + "epoch": 4.66, + "learning_rate": 2.6740222853397424e-06, + "loss": 1.0451, + "step": 21342 + }, + { + "epoch": 4.66, + "learning_rate": 2.6739130434782614e-06, + "loss": 2.401, + "step": 21343 + }, + { + "epoch": 4.66, + "learning_rate": 2.67380380161678e-06, + "loss": 1.5193, + "step": 21344 + }, + { + "epoch": 4.66, + "learning_rate": 2.673694559755298e-06, + "loss": 1.9378, + "step": 21345 + }, + { + "epoch": 4.66, + "learning_rate": 2.6735853178938172e-06, + "loss": 1.7319, + "step": 21346 + }, + { + "epoch": 4.66, + "learning_rate": 2.673476076032336e-06, + "loss": 2.1531, + "step": 21347 + }, + { + "epoch": 4.66, + "learning_rate": 2.6733668341708545e-06, + "loss": 2.0266, + "step": 21348 + }, + { + "epoch": 4.66, + "learning_rate": 2.6732575923093735e-06, + "loss": 1.8979, + "step": 21349 + }, + { + "epoch": 4.66, + "learning_rate": 2.6731483504478917e-06, + "loss": 1.9225, + "step": 21350 + }, + { + "epoch": 4.66, + "learning_rate": 2.6730391085864103e-06, + "loss": 1.1533, + "step": 21351 + }, + { + "epoch": 4.66, + "learning_rate": 2.6729298667249293e-06, + "loss": 1.7185, + "step": 21352 + }, + { + "epoch": 4.66, + "learning_rate": 2.672820624863448e-06, + "loss": 2.2141, + "step": 21353 + }, + { + "epoch": 4.66, + "learning_rate": 2.672711383001966e-06, + "loss": 1.6009, + "step": 21354 + }, + { + "epoch": 4.66, + "learning_rate": 2.672602141140485e-06, + "loss": 1.4528, + "step": 21355 + }, + { + "epoch": 4.66, + "learning_rate": 2.6724928992790038e-06, + "loss": 1.4269, + "step": 21356 + }, + { + "epoch": 4.66, + "learning_rate": 2.672383657417523e-06, + "loss": 1.7296, + "step": 21357 + }, + { + "epoch": 4.66, + "learning_rate": 2.6722744155560414e-06, + "loss": 1.7574, + "step": 21358 + }, + { + "epoch": 4.66, + "learning_rate": 2.6721651736945596e-06, + "loss": 2.3403, + "step": 21359 + }, + { + "epoch": 4.66, + "learning_rate": 2.6720559318330786e-06, + "loss": 1.4258, + "step": 21360 + }, + { + "epoch": 4.66, + "learning_rate": 2.6719466899715972e-06, + "loss": 2.0768, + "step": 21361 + }, + { + "epoch": 4.66, + "learning_rate": 2.6718374481101163e-06, + "loss": 3.8012, + "step": 21362 + }, + { + "epoch": 4.66, + "learning_rate": 2.671728206248635e-06, + "loss": 2.2161, + "step": 21363 + }, + { + "epoch": 4.66, + "learning_rate": 2.671618964387153e-06, + "loss": 1.7931, + "step": 21364 + }, + { + "epoch": 4.66, + "learning_rate": 2.671509722525672e-06, + "loss": 0.9347, + "step": 21365 + }, + { + "epoch": 4.66, + "learning_rate": 2.6714004806641907e-06, + "loss": 1.5755, + "step": 21366 + }, + { + "epoch": 4.66, + "learning_rate": 2.6712912388027098e-06, + "loss": 1.6622, + "step": 21367 + }, + { + "epoch": 4.66, + "learning_rate": 2.6711819969412284e-06, + "loss": 2.1099, + "step": 21368 + }, + { + "epoch": 4.66, + "learning_rate": 2.6710727550797466e-06, + "loss": 0.8387, + "step": 21369 + }, + { + "epoch": 4.66, + "learning_rate": 2.6709635132182656e-06, + "loss": 1.1376, + "step": 21370 + }, + { + "epoch": 4.66, + "learning_rate": 2.670854271356784e-06, + "loss": 2.1269, + "step": 21371 + }, + { + "epoch": 4.66, + "learning_rate": 2.670745029495303e-06, + "loss": 1.1988, + "step": 21372 + }, + { + "epoch": 4.66, + "learning_rate": 2.6706357876338214e-06, + "loss": 2.9659, + "step": 21373 + }, + { + "epoch": 4.66, + "learning_rate": 2.67052654577234e-06, + "loss": 0.3573, + "step": 21374 + }, + { + "epoch": 4.66, + "learning_rate": 2.6704173039108586e-06, + "loss": 1.268, + "step": 21375 + }, + { + "epoch": 4.67, + "learning_rate": 2.6703080620493777e-06, + "loss": 3.1998, + "step": 21376 + }, + { + "epoch": 4.67, + "learning_rate": 2.6701988201878963e-06, + "loss": 1.9156, + "step": 21377 + }, + { + "epoch": 4.67, + "learning_rate": 2.670089578326415e-06, + "loss": 2.2324, + "step": 21378 + }, + { + "epoch": 4.67, + "learning_rate": 2.6699803364649335e-06, + "loss": 1.5817, + "step": 21379 + }, + { + "epoch": 4.67, + "learning_rate": 2.669871094603452e-06, + "loss": 1.4922, + "step": 21380 + }, + { + "epoch": 4.67, + "learning_rate": 2.669761852741971e-06, + "loss": 1.1818, + "step": 21381 + }, + { + "epoch": 4.67, + "learning_rate": 2.6696526108804898e-06, + "loss": 2.1251, + "step": 21382 + }, + { + "epoch": 4.67, + "learning_rate": 2.669543369019008e-06, + "loss": 1.8437, + "step": 21383 + }, + { + "epoch": 4.67, + "learning_rate": 2.669434127157527e-06, + "loss": 2.1468, + "step": 21384 + }, + { + "epoch": 4.67, + "learning_rate": 2.6693248852960456e-06, + "loss": 0.8232, + "step": 21385 + }, + { + "epoch": 4.67, + "learning_rate": 2.6692156434345646e-06, + "loss": 2.2069, + "step": 21386 + }, + { + "epoch": 4.67, + "learning_rate": 2.669106401573083e-06, + "loss": 0.5645, + "step": 21387 + }, + { + "epoch": 4.67, + "learning_rate": 2.6689971597116014e-06, + "loss": 0.7946, + "step": 21388 + }, + { + "epoch": 4.67, + "learning_rate": 2.6688879178501205e-06, + "loss": 0.9835, + "step": 21389 + }, + { + "epoch": 4.67, + "learning_rate": 2.668778675988639e-06, + "loss": 2.8758, + "step": 21390 + }, + { + "epoch": 4.67, + "learning_rate": 2.668669434127158e-06, + "loss": 1.7857, + "step": 21391 + }, + { + "epoch": 4.67, + "learning_rate": 2.6685601922656763e-06, + "loss": 1.1589, + "step": 21392 + }, + { + "epoch": 4.67, + "learning_rate": 2.668450950404195e-06, + "loss": 1.4821, + "step": 21393 + }, + { + "epoch": 4.67, + "learning_rate": 2.668341708542714e-06, + "loss": 4.0138, + "step": 21394 + }, + { + "epoch": 4.67, + "learning_rate": 2.6682324666812326e-06, + "loss": 0.4176, + "step": 21395 + }, + { + "epoch": 4.67, + "learning_rate": 2.6681232248197516e-06, + "loss": 0.8193, + "step": 21396 + }, + { + "epoch": 4.67, + "learning_rate": 2.6680139829582698e-06, + "loss": 2.2058, + "step": 21397 + }, + { + "epoch": 4.67, + "learning_rate": 2.6679047410967884e-06, + "loss": 0.6409, + "step": 21398 + }, + { + "epoch": 4.67, + "learning_rate": 2.6677954992353074e-06, + "loss": 1.9729, + "step": 21399 + }, + { + "epoch": 4.67, + "learning_rate": 2.667686257373826e-06, + "loss": 1.8078, + "step": 21400 + }, + { + "epoch": 4.67, + "learning_rate": 2.6675770155123446e-06, + "loss": 0.986, + "step": 21401 + }, + { + "epoch": 4.67, + "learning_rate": 2.6674677736508633e-06, + "loss": 1.9022, + "step": 21402 + }, + { + "epoch": 4.67, + "learning_rate": 2.667358531789382e-06, + "loss": 2.2933, + "step": 21403 + }, + { + "epoch": 4.67, + "learning_rate": 2.6672492899279005e-06, + "loss": 0.9769, + "step": 21404 + }, + { + "epoch": 4.67, + "learning_rate": 2.6671400480664195e-06, + "loss": 1.4852, + "step": 21405 + }, + { + "epoch": 4.67, + "learning_rate": 2.6670308062049377e-06, + "loss": 0.5746, + "step": 21406 + }, + { + "epoch": 4.67, + "learning_rate": 2.6669215643434563e-06, + "loss": 0.6831, + "step": 21407 + }, + { + "epoch": 4.67, + "learning_rate": 2.6668123224819753e-06, + "loss": 0.7151, + "step": 21408 + }, + { + "epoch": 4.67, + "learning_rate": 2.666703080620494e-06, + "loss": 1.2407, + "step": 21409 + }, + { + "epoch": 4.67, + "learning_rate": 2.666593838759013e-06, + "loss": 0.7943, + "step": 21410 + }, + { + "epoch": 4.67, + "learning_rate": 2.666484596897531e-06, + "loss": 2.1544, + "step": 21411 + }, + { + "epoch": 4.67, + "learning_rate": 2.66637535503605e-06, + "loss": 2.0186, + "step": 21412 + }, + { + "epoch": 4.67, + "learning_rate": 2.666266113174569e-06, + "loss": 2.9143, + "step": 21413 + }, + { + "epoch": 4.67, + "learning_rate": 2.6661568713130874e-06, + "loss": 1.8346, + "step": 21414 + }, + { + "epoch": 4.67, + "learning_rate": 2.6660476294516065e-06, + "loss": 1.2861, + "step": 21415 + }, + { + "epoch": 4.67, + "learning_rate": 2.6659383875901247e-06, + "loss": 1.8712, + "step": 21416 + }, + { + "epoch": 4.67, + "learning_rate": 2.6658291457286433e-06, + "loss": 1.5474, + "step": 21417 + }, + { + "epoch": 4.67, + "learning_rate": 2.6657199038671623e-06, + "loss": 1.7454, + "step": 21418 + }, + { + "epoch": 4.67, + "learning_rate": 2.665610662005681e-06, + "loss": 0.9473, + "step": 21419 + }, + { + "epoch": 4.67, + "learning_rate": 2.665501420144199e-06, + "loss": 0.8752, + "step": 21420 + }, + { + "epoch": 4.68, + "learning_rate": 2.665392178282718e-06, + "loss": 0.8111, + "step": 21421 + }, + { + "epoch": 4.68, + "learning_rate": 2.6652829364212367e-06, + "loss": 1.9999, + "step": 21422 + }, + { + "epoch": 4.68, + "learning_rate": 2.6651736945597558e-06, + "loss": 0.8841, + "step": 21423 + }, + { + "epoch": 4.68, + "learning_rate": 2.6650644526982744e-06, + "loss": 1.7346, + "step": 21424 + }, + { + "epoch": 4.68, + "learning_rate": 2.6649552108367926e-06, + "loss": 1.2502, + "step": 21425 + }, + { + "epoch": 4.68, + "learning_rate": 2.6648459689753116e-06, + "loss": 2.288, + "step": 21426 + }, + { + "epoch": 4.68, + "learning_rate": 2.6647367271138302e-06, + "loss": 1.0529, + "step": 21427 + }, + { + "epoch": 4.68, + "learning_rate": 2.664627485252349e-06, + "loss": 1.4334, + "step": 21428 + }, + { + "epoch": 4.68, + "learning_rate": 2.664518243390868e-06, + "loss": 1.3687, + "step": 21429 + }, + { + "epoch": 4.68, + "learning_rate": 2.664409001529386e-06, + "loss": 2.2264, + "step": 21430 + }, + { + "epoch": 4.68, + "learning_rate": 2.664299759667905e-06, + "loss": 2.3828, + "step": 21431 + }, + { + "epoch": 4.68, + "learning_rate": 2.6641905178064237e-06, + "loss": 1.9363, + "step": 21432 + }, + { + "epoch": 4.68, + "learning_rate": 2.6640812759449423e-06, + "loss": 2.3271, + "step": 21433 + }, + { + "epoch": 4.68, + "learning_rate": 2.663972034083461e-06, + "loss": 2.1668, + "step": 21434 + }, + { + "epoch": 4.68, + "learning_rate": 2.6638627922219795e-06, + "loss": 2.5273, + "step": 21435 + }, + { + "epoch": 4.68, + "learning_rate": 2.663753550360498e-06, + "loss": 1.7969, + "step": 21436 + }, + { + "epoch": 4.68, + "learning_rate": 2.663644308499017e-06, + "loss": 0.8439, + "step": 21437 + }, + { + "epoch": 4.68, + "learning_rate": 2.663535066637536e-06, + "loss": 0.8565, + "step": 21438 + }, + { + "epoch": 4.68, + "learning_rate": 2.663425824776054e-06, + "loss": 0.9229, + "step": 21439 + }, + { + "epoch": 4.68, + "learning_rate": 2.663316582914573e-06, + "loss": 1.4335, + "step": 21440 + }, + { + "epoch": 4.68, + "learning_rate": 2.6632073410530916e-06, + "loss": 1.6354, + "step": 21441 + }, + { + "epoch": 4.68, + "learning_rate": 2.6630980991916107e-06, + "loss": 4.391, + "step": 21442 + }, + { + "epoch": 4.68, + "learning_rate": 2.6629888573301293e-06, + "loss": 1.2257, + "step": 21443 + }, + { + "epoch": 4.68, + "learning_rate": 2.6628796154686475e-06, + "loss": 2.6817, + "step": 21444 + }, + { + "epoch": 4.68, + "learning_rate": 2.6627703736071665e-06, + "loss": 1.2924, + "step": 21445 + }, + { + "epoch": 4.68, + "learning_rate": 2.662661131745685e-06, + "loss": 0.9599, + "step": 21446 + }, + { + "epoch": 4.68, + "learning_rate": 2.662551889884204e-06, + "loss": 2.2028, + "step": 21447 + }, + { + "epoch": 4.68, + "learning_rate": 2.6624426480227228e-06, + "loss": 1.222, + "step": 21448 + }, + { + "epoch": 4.68, + "learning_rate": 2.662333406161241e-06, + "loss": 2.1746, + "step": 21449 + }, + { + "epoch": 4.68, + "learning_rate": 2.66222416429976e-06, + "loss": 0.9554, + "step": 21450 + }, + { + "epoch": 4.68, + "learning_rate": 2.6621149224382786e-06, + "loss": 2.2477, + "step": 21451 + }, + { + "epoch": 4.68, + "learning_rate": 2.6620056805767976e-06, + "loss": 1.2288, + "step": 21452 + }, + { + "epoch": 4.68, + "learning_rate": 2.661896438715316e-06, + "loss": 2.0267, + "step": 21453 + }, + { + "epoch": 4.68, + "learning_rate": 2.6617871968538344e-06, + "loss": 0.3994, + "step": 21454 + }, + { + "epoch": 4.68, + "learning_rate": 2.6616779549923535e-06, + "loss": 1.5708, + "step": 21455 + }, + { + "epoch": 4.68, + "learning_rate": 2.661568713130872e-06, + "loss": 2.2763, + "step": 21456 + }, + { + "epoch": 4.68, + "learning_rate": 2.6614594712693907e-06, + "loss": 2.2896, + "step": 21457 + }, + { + "epoch": 4.68, + "learning_rate": 2.6613502294079093e-06, + "loss": 3.0385, + "step": 21458 + }, + { + "epoch": 4.68, + "learning_rate": 2.661240987546428e-06, + "loss": 1.573, + "step": 21459 + }, + { + "epoch": 4.68, + "learning_rate": 2.6611317456849465e-06, + "loss": 2.2739, + "step": 21460 + }, + { + "epoch": 4.68, + "learning_rate": 2.6610225038234655e-06, + "loss": 2.9755, + "step": 21461 + }, + { + "epoch": 4.68, + "learning_rate": 2.660913261961984e-06, + "loss": 2.1278, + "step": 21462 + }, + { + "epoch": 4.68, + "learning_rate": 2.6608040201005023e-06, + "loss": 1.2465, + "step": 21463 + }, + { + "epoch": 4.68, + "learning_rate": 2.6606947782390214e-06, + "loss": 3.5496, + "step": 21464 + }, + { + "epoch": 4.68, + "learning_rate": 2.66058553637754e-06, + "loss": 0.9661, + "step": 21465 + }, + { + "epoch": 4.68, + "learning_rate": 2.660476294516059e-06, + "loss": 1.6132, + "step": 21466 + }, + { + "epoch": 4.69, + "learning_rate": 2.660367052654577e-06, + "loss": 0.8627, + "step": 21467 + }, + { + "epoch": 4.69, + "learning_rate": 2.660257810793096e-06, + "loss": 1.4333, + "step": 21468 + }, + { + "epoch": 4.69, + "learning_rate": 2.660148568931615e-06, + "loss": 1.8346, + "step": 21469 + }, + { + "epoch": 4.69, + "learning_rate": 2.6600393270701335e-06, + "loss": 1.6475, + "step": 21470 + }, + { + "epoch": 4.69, + "learning_rate": 2.6599300852086525e-06, + "loss": 2.7218, + "step": 21471 + }, + { + "epoch": 4.69, + "learning_rate": 2.6598208433471707e-06, + "loss": 2.245, + "step": 21472 + }, + { + "epoch": 4.69, + "learning_rate": 2.6597116014856893e-06, + "loss": 1.8494, + "step": 21473 + }, + { + "epoch": 4.69, + "learning_rate": 2.6596023596242083e-06, + "loss": 2.0872, + "step": 21474 + }, + { + "epoch": 4.69, + "learning_rate": 2.659493117762727e-06, + "loss": 0.6369, + "step": 21475 + }, + { + "epoch": 4.69, + "learning_rate": 2.659383875901246e-06, + "loss": 0.8913, + "step": 21476 + }, + { + "epoch": 4.69, + "learning_rate": 2.659274634039764e-06, + "loss": 2.0086, + "step": 21477 + }, + { + "epoch": 4.69, + "learning_rate": 2.6591653921782828e-06, + "loss": 1.956, + "step": 21478 + }, + { + "epoch": 4.69, + "learning_rate": 2.659056150316802e-06, + "loss": 1.5018, + "step": 21479 + }, + { + "epoch": 4.69, + "learning_rate": 2.6589469084553204e-06, + "loss": 1.6971, + "step": 21480 + }, + { + "epoch": 4.69, + "learning_rate": 2.658837666593839e-06, + "loss": 1.0825, + "step": 21481 + }, + { + "epoch": 4.69, + "learning_rate": 2.6587284247323576e-06, + "loss": 1.8797, + "step": 21482 + }, + { + "epoch": 4.69, + "learning_rate": 2.6586191828708763e-06, + "loss": 1.9161, + "step": 21483 + }, + { + "epoch": 4.69, + "learning_rate": 2.6585099410093953e-06, + "loss": 2.9548, + "step": 21484 + }, + { + "epoch": 4.69, + "learning_rate": 2.658400699147914e-06, + "loss": 1.2638, + "step": 21485 + }, + { + "epoch": 4.69, + "learning_rate": 2.658291457286432e-06, + "loss": 2.0988, + "step": 21486 + }, + { + "epoch": 4.69, + "learning_rate": 2.658182215424951e-06, + "loss": 0.9596, + "step": 21487 + }, + { + "epoch": 4.69, + "learning_rate": 2.6580729735634697e-06, + "loss": 2.7703, + "step": 21488 + }, + { + "epoch": 4.69, + "learning_rate": 2.6579637317019883e-06, + "loss": 1.3057, + "step": 21489 + }, + { + "epoch": 4.69, + "learning_rate": 2.6578544898405074e-06, + "loss": 0.9889, + "step": 21490 + }, + { + "epoch": 4.69, + "learning_rate": 2.6577452479790256e-06, + "loss": 1.9513, + "step": 21491 + }, + { + "epoch": 4.69, + "learning_rate": 2.657636006117544e-06, + "loss": 1.0409, + "step": 21492 + }, + { + "epoch": 4.69, + "learning_rate": 2.657526764256063e-06, + "loss": 2.7, + "step": 21493 + }, + { + "epoch": 4.69, + "learning_rate": 2.657417522394582e-06, + "loss": 2.4171, + "step": 21494 + }, + { + "epoch": 4.69, + "learning_rate": 2.657308280533101e-06, + "loss": 2.6197, + "step": 21495 + }, + { + "epoch": 4.69, + "learning_rate": 2.657199038671619e-06, + "loss": 0.6598, + "step": 21496 + }, + { + "epoch": 4.69, + "learning_rate": 2.6570897968101377e-06, + "loss": 2.7114, + "step": 21497 + }, + { + "epoch": 4.69, + "learning_rate": 2.6569805549486567e-06, + "loss": 0.4682, + "step": 21498 + }, + { + "epoch": 4.69, + "learning_rate": 2.6568713130871753e-06, + "loss": 2.5379, + "step": 21499 + }, + { + "epoch": 4.69, + "learning_rate": 2.6567620712256935e-06, + "loss": 1.3657, + "step": 21500 + }, + { + "epoch": 4.69, + "learning_rate": 2.6566528293642125e-06, + "loss": 1.5496, + "step": 21501 + }, + { + "epoch": 4.69, + "learning_rate": 2.656543587502731e-06, + "loss": 1.8455, + "step": 21502 + }, + { + "epoch": 4.69, + "learning_rate": 2.65643434564125e-06, + "loss": 2.8401, + "step": 21503 + }, + { + "epoch": 4.69, + "learning_rate": 2.6563251037797688e-06, + "loss": 0.8592, + "step": 21504 + }, + { + "epoch": 4.69, + "learning_rate": 2.656215861918287e-06, + "loss": 0.8657, + "step": 21505 + }, + { + "epoch": 4.69, + "learning_rate": 2.656106620056806e-06, + "loss": 1.4264, + "step": 21506 + }, + { + "epoch": 4.69, + "learning_rate": 2.6559973781953246e-06, + "loss": 2.3279, + "step": 21507 + }, + { + "epoch": 4.69, + "learning_rate": 2.6558881363338436e-06, + "loss": 0.6679, + "step": 21508 + }, + { + "epoch": 4.69, + "learning_rate": 2.6557788944723623e-06, + "loss": 1.2394, + "step": 21509 + }, + { + "epoch": 4.69, + "learning_rate": 2.6556696526108804e-06, + "loss": 0.6863, + "step": 21510 + }, + { + "epoch": 4.69, + "learning_rate": 2.6555604107493995e-06, + "loss": 1.383, + "step": 21511 + }, + { + "epoch": 4.69, + "learning_rate": 2.655451168887918e-06, + "loss": 0.5174, + "step": 21512 + }, + { + "epoch": 4.7, + "learning_rate": 2.6553419270264367e-06, + "loss": 1.3016, + "step": 21513 + }, + { + "epoch": 4.7, + "learning_rate": 2.6552326851649557e-06, + "loss": 1.1887, + "step": 21514 + }, + { + "epoch": 4.7, + "learning_rate": 2.655123443303474e-06, + "loss": 1.0977, + "step": 21515 + }, + { + "epoch": 4.7, + "learning_rate": 2.6550142014419925e-06, + "loss": 1.1191, + "step": 21516 + }, + { + "epoch": 4.7, + "learning_rate": 2.6549049595805116e-06, + "loss": 2.0357, + "step": 21517 + }, + { + "epoch": 4.7, + "learning_rate": 2.65479571771903e-06, + "loss": 2.4395, + "step": 21518 + }, + { + "epoch": 4.7, + "learning_rate": 2.6546864758575484e-06, + "loss": 0.679, + "step": 21519 + }, + { + "epoch": 4.7, + "learning_rate": 2.6545772339960674e-06, + "loss": 2.0171, + "step": 21520 + }, + { + "epoch": 4.7, + "learning_rate": 2.654467992134586e-06, + "loss": 2.5001, + "step": 21521 + }, + { + "epoch": 4.7, + "learning_rate": 2.654358750273105e-06, + "loss": 1.6131, + "step": 21522 + }, + { + "epoch": 4.7, + "learning_rate": 2.6542495084116237e-06, + "loss": 1.6892, + "step": 21523 + }, + { + "epoch": 4.7, + "learning_rate": 2.654140266550142e-06, + "loss": 1.7881, + "step": 21524 + }, + { + "epoch": 4.7, + "learning_rate": 2.654031024688661e-06, + "loss": 2.7903, + "step": 21525 + }, + { + "epoch": 4.7, + "learning_rate": 2.6539217828271795e-06, + "loss": 1.0957, + "step": 21526 + }, + { + "epoch": 4.7, + "learning_rate": 2.6538125409656985e-06, + "loss": 1.5103, + "step": 21527 + }, + { + "epoch": 4.7, + "learning_rate": 2.653703299104217e-06, + "loss": 2.1187, + "step": 21528 + }, + { + "epoch": 4.7, + "learning_rate": 2.6535940572427353e-06, + "loss": 0.8384, + "step": 21529 + }, + { + "epoch": 4.7, + "learning_rate": 2.6534848153812544e-06, + "loss": 1.8839, + "step": 21530 + }, + { + "epoch": 4.7, + "learning_rate": 2.653375573519773e-06, + "loss": 2.0243, + "step": 21531 + }, + { + "epoch": 4.7, + "learning_rate": 2.653266331658292e-06, + "loss": 1.0367, + "step": 21532 + }, + { + "epoch": 4.7, + "learning_rate": 2.65315708979681e-06, + "loss": 2.2743, + "step": 21533 + }, + { + "epoch": 4.7, + "learning_rate": 2.653047847935329e-06, + "loss": 1.094, + "step": 21534 + }, + { + "epoch": 4.7, + "learning_rate": 2.652938606073848e-06, + "loss": 1.7681, + "step": 21535 + }, + { + "epoch": 4.7, + "learning_rate": 2.6528293642123664e-06, + "loss": 0.5723, + "step": 21536 + }, + { + "epoch": 4.7, + "learning_rate": 2.6527201223508855e-06, + "loss": 1.2458, + "step": 21537 + }, + { + "epoch": 4.7, + "learning_rate": 2.6526108804894037e-06, + "loss": 2.8098, + "step": 21538 + }, + { + "epoch": 4.7, + "learning_rate": 2.6525016386279223e-06, + "loss": 0.8863, + "step": 21539 + }, + { + "epoch": 4.7, + "learning_rate": 2.6523923967664413e-06, + "loss": 2.5799, + "step": 21540 + }, + { + "epoch": 4.7, + "learning_rate": 2.65228315490496e-06, + "loss": 2.475, + "step": 21541 + }, + { + "epoch": 4.7, + "learning_rate": 2.6521739130434785e-06, + "loss": 1.7539, + "step": 21542 + }, + { + "epoch": 4.7, + "learning_rate": 2.652064671181997e-06, + "loss": 1.484, + "step": 21543 + }, + { + "epoch": 4.7, + "learning_rate": 2.6519554293205158e-06, + "loss": 1.2117, + "step": 21544 + }, + { + "epoch": 4.7, + "learning_rate": 2.6518461874590344e-06, + "loss": 2.3894, + "step": 21545 + }, + { + "epoch": 4.7, + "learning_rate": 2.6517369455975534e-06, + "loss": 3.388, + "step": 21546 + }, + { + "epoch": 4.7, + "learning_rate": 2.651627703736072e-06, + "loss": 1.0197, + "step": 21547 + }, + { + "epoch": 4.7, + "learning_rate": 2.65151846187459e-06, + "loss": 1.07, + "step": 21548 + }, + { + "epoch": 4.7, + "learning_rate": 2.6514092200131092e-06, + "loss": 2.107, + "step": 21549 + }, + { + "epoch": 4.7, + "learning_rate": 2.651299978151628e-06, + "loss": 0.954, + "step": 21550 + }, + { + "epoch": 4.7, + "learning_rate": 2.651190736290147e-06, + "loss": 2.1468, + "step": 21551 + }, + { + "epoch": 4.7, + "learning_rate": 2.651081494428665e-06, + "loss": 1.6265, + "step": 21552 + }, + { + "epoch": 4.7, + "learning_rate": 2.6509722525671837e-06, + "loss": 3.1788, + "step": 21553 + }, + { + "epoch": 4.7, + "learning_rate": 2.6508630107057027e-06, + "loss": 2.3102, + "step": 21554 + }, + { + "epoch": 4.7, + "learning_rate": 2.6507537688442213e-06, + "loss": 2.1497, + "step": 21555 + }, + { + "epoch": 4.7, + "learning_rate": 2.6506445269827404e-06, + "loss": 2.8385, + "step": 21556 + }, + { + "epoch": 4.7, + "learning_rate": 2.6505352851212585e-06, + "loss": 2.7249, + "step": 21557 + }, + { + "epoch": 4.7, + "learning_rate": 2.650426043259777e-06, + "loss": 0.3737, + "step": 21558 + }, + { + "epoch": 4.71, + "learning_rate": 2.650316801398296e-06, + "loss": 0.8283, + "step": 21559 + }, + { + "epoch": 4.71, + "learning_rate": 2.650207559536815e-06, + "loss": 0.9395, + "step": 21560 + }, + { + "epoch": 4.71, + "learning_rate": 2.650098317675334e-06, + "loss": 1.9846, + "step": 21561 + }, + { + "epoch": 4.71, + "learning_rate": 2.649989075813852e-06, + "loss": 1.9535, + "step": 21562 + }, + { + "epoch": 4.71, + "learning_rate": 2.6498798339523706e-06, + "loss": 1.3355, + "step": 21563 + }, + { + "epoch": 4.71, + "learning_rate": 2.6497705920908897e-06, + "loss": 1.148, + "step": 21564 + }, + { + "epoch": 4.71, + "learning_rate": 2.6496613502294083e-06, + "loss": 1.1601, + "step": 21565 + }, + { + "epoch": 4.71, + "learning_rate": 2.6495521083679265e-06, + "loss": 2.8038, + "step": 21566 + }, + { + "epoch": 4.71, + "learning_rate": 2.6494428665064455e-06, + "loss": 1.224, + "step": 21567 + }, + { + "epoch": 4.71, + "learning_rate": 2.649333624644964e-06, + "loss": 1.9762, + "step": 21568 + }, + { + "epoch": 4.71, + "learning_rate": 2.6492243827834827e-06, + "loss": 1.5526, + "step": 21569 + }, + { + "epoch": 4.71, + "learning_rate": 2.6491151409220018e-06, + "loss": 1.2972, + "step": 21570 + }, + { + "epoch": 4.71, + "learning_rate": 2.64900589906052e-06, + "loss": 1.2781, + "step": 21571 + }, + { + "epoch": 4.71, + "learning_rate": 2.6488966571990386e-06, + "loss": 0.9109, + "step": 21572 + }, + { + "epoch": 4.71, + "learning_rate": 2.6487874153375576e-06, + "loss": 2.0543, + "step": 21573 + }, + { + "epoch": 4.71, + "learning_rate": 2.648678173476076e-06, + "loss": 2.3583, + "step": 21574 + }, + { + "epoch": 4.71, + "learning_rate": 2.6485689316145952e-06, + "loss": 1.7646, + "step": 21575 + }, + { + "epoch": 4.71, + "learning_rate": 2.6484596897531134e-06, + "loss": 2.0445, + "step": 21576 + }, + { + "epoch": 4.71, + "learning_rate": 2.648350447891632e-06, + "loss": 2.1269, + "step": 21577 + }, + { + "epoch": 4.71, + "learning_rate": 2.648241206030151e-06, + "loss": 2.2733, + "step": 21578 + }, + { + "epoch": 4.71, + "learning_rate": 2.6481319641686697e-06, + "loss": 1.4521, + "step": 21579 + }, + { + "epoch": 4.71, + "learning_rate": 2.6480227223071887e-06, + "loss": 1.8956, + "step": 21580 + }, + { + "epoch": 4.71, + "learning_rate": 2.647913480445707e-06, + "loss": 2.2614, + "step": 21581 + }, + { + "epoch": 4.71, + "learning_rate": 2.6478042385842255e-06, + "loss": 2.4575, + "step": 21582 + }, + { + "epoch": 4.71, + "learning_rate": 2.6476949967227446e-06, + "loss": 0.6372, + "step": 21583 + }, + { + "epoch": 4.71, + "learning_rate": 2.647585754861263e-06, + "loss": 1.4624, + "step": 21584 + }, + { + "epoch": 4.71, + "learning_rate": 2.6474765129997814e-06, + "loss": 1.7632, + "step": 21585 + }, + { + "epoch": 4.71, + "learning_rate": 2.6473672711383004e-06, + "loss": 1.0646, + "step": 21586 + }, + { + "epoch": 4.71, + "learning_rate": 2.647258029276819e-06, + "loss": 0.959, + "step": 21587 + }, + { + "epoch": 4.71, + "learning_rate": 2.647148787415338e-06, + "loss": 0.8555, + "step": 21588 + }, + { + "epoch": 4.71, + "learning_rate": 2.6470395455538566e-06, + "loss": 0.6115, + "step": 21589 + }, + { + "epoch": 4.71, + "learning_rate": 2.646930303692375e-06, + "loss": 1.8372, + "step": 21590 + }, + { + "epoch": 4.71, + "learning_rate": 2.646821061830894e-06, + "loss": 2.2542, + "step": 21591 + }, + { + "epoch": 4.71, + "learning_rate": 2.6467118199694125e-06, + "loss": 3.2282, + "step": 21592 + }, + { + "epoch": 4.71, + "learning_rate": 2.6466025781079315e-06, + "loss": 0.6428, + "step": 21593 + }, + { + "epoch": 4.71, + "learning_rate": 2.64649333624645e-06, + "loss": 3.1525, + "step": 21594 + }, + { + "epoch": 4.71, + "learning_rate": 2.6463840943849683e-06, + "loss": 2.8942, + "step": 21595 + }, + { + "epoch": 4.71, + "learning_rate": 2.6462748525234873e-06, + "loss": 2.8626, + "step": 21596 + }, + { + "epoch": 4.71, + "learning_rate": 2.646165610662006e-06, + "loss": 2.1413, + "step": 21597 + }, + { + "epoch": 4.71, + "learning_rate": 2.6460563688005246e-06, + "loss": 1.1994, + "step": 21598 + }, + { + "epoch": 4.71, + "learning_rate": 2.645947126939043e-06, + "loss": 1.4896, + "step": 21599 + }, + { + "epoch": 4.71, + "learning_rate": 2.6458378850775618e-06, + "loss": 0.7399, + "step": 21600 + }, + { + "epoch": 4.71, + "learning_rate": 2.6457286432160804e-06, + "loss": 1.7888, + "step": 21601 + }, + { + "epoch": 4.71, + "learning_rate": 2.6456194013545994e-06, + "loss": 0.7181, + "step": 21602 + }, + { + "epoch": 4.71, + "learning_rate": 2.645510159493118e-06, + "loss": 1.7786, + "step": 21603 + }, + { + "epoch": 4.71, + "learning_rate": 2.6454009176316362e-06, + "loss": 2.1267, + "step": 21604 + }, + { + "epoch": 4.72, + "learning_rate": 2.6452916757701553e-06, + "loss": 1.0349, + "step": 21605 + }, + { + "epoch": 4.72, + "learning_rate": 2.645182433908674e-06, + "loss": 1.3963, + "step": 21606 + }, + { + "epoch": 4.72, + "learning_rate": 2.645073192047193e-06, + "loss": 2.271, + "step": 21607 + }, + { + "epoch": 4.72, + "learning_rate": 2.6449639501857115e-06, + "loss": 3.3682, + "step": 21608 + }, + { + "epoch": 4.72, + "learning_rate": 2.6448547083242297e-06, + "loss": 0.9754, + "step": 21609 + }, + { + "epoch": 4.72, + "learning_rate": 2.6447454664627487e-06, + "loss": 1.7092, + "step": 21610 + }, + { + "epoch": 4.72, + "learning_rate": 2.6446362246012674e-06, + "loss": 1.1667, + "step": 21611 + }, + { + "epoch": 4.72, + "learning_rate": 2.6445269827397864e-06, + "loss": 1.8077, + "step": 21612 + }, + { + "epoch": 4.72, + "learning_rate": 2.644417740878305e-06, + "loss": 2.4322, + "step": 21613 + }, + { + "epoch": 4.72, + "learning_rate": 2.644308499016823e-06, + "loss": 1.353, + "step": 21614 + }, + { + "epoch": 4.72, + "learning_rate": 2.6441992571553422e-06, + "loss": 1.5902, + "step": 21615 + }, + { + "epoch": 4.72, + "learning_rate": 2.644090015293861e-06, + "loss": 2.0386, + "step": 21616 + }, + { + "epoch": 4.72, + "learning_rate": 2.64398077343238e-06, + "loss": 2.0329, + "step": 21617 + }, + { + "epoch": 4.72, + "learning_rate": 2.643871531570898e-06, + "loss": 1.4102, + "step": 21618 + }, + { + "epoch": 4.72, + "learning_rate": 2.6437622897094167e-06, + "loss": 0.7036, + "step": 21619 + }, + { + "epoch": 4.72, + "learning_rate": 2.6436530478479357e-06, + "loss": 0.8121, + "step": 21620 + }, + { + "epoch": 4.72, + "learning_rate": 2.6435438059864543e-06, + "loss": 2.1152, + "step": 21621 + }, + { + "epoch": 4.72, + "learning_rate": 2.643434564124973e-06, + "loss": 1.7353, + "step": 21622 + }, + { + "epoch": 4.72, + "learning_rate": 2.6433253222634915e-06, + "loss": 1.5666, + "step": 21623 + }, + { + "epoch": 4.72, + "learning_rate": 2.64321608040201e-06, + "loss": 1.2762, + "step": 21624 + }, + { + "epoch": 4.72, + "learning_rate": 2.6431068385405288e-06, + "loss": 1.4744, + "step": 21625 + }, + { + "epoch": 4.72, + "learning_rate": 2.642997596679048e-06, + "loss": 1.446, + "step": 21626 + }, + { + "epoch": 4.72, + "learning_rate": 2.6428883548175664e-06, + "loss": 2.1527, + "step": 21627 + }, + { + "epoch": 4.72, + "learning_rate": 2.642779112956085e-06, + "loss": 1.0705, + "step": 21628 + }, + { + "epoch": 4.72, + "learning_rate": 2.6426698710946036e-06, + "loss": 1.6433, + "step": 21629 + }, + { + "epoch": 4.72, + "learning_rate": 2.6425606292331222e-06, + "loss": 1.3836, + "step": 21630 + }, + { + "epoch": 4.72, + "learning_rate": 2.6424513873716413e-06, + "loss": 1.4117, + "step": 21631 + }, + { + "epoch": 4.72, + "learning_rate": 2.6423421455101595e-06, + "loss": 2.2108, + "step": 21632 + }, + { + "epoch": 4.72, + "learning_rate": 2.642232903648678e-06, + "loss": 1.9328, + "step": 21633 + }, + { + "epoch": 4.72, + "learning_rate": 2.642123661787197e-06, + "loss": 1.0864, + "step": 21634 + }, + { + "epoch": 4.72, + "learning_rate": 2.6420144199257157e-06, + "loss": 0.4579, + "step": 21635 + }, + { + "epoch": 4.72, + "learning_rate": 2.6419051780642347e-06, + "loss": 2.621, + "step": 21636 + }, + { + "epoch": 4.72, + "learning_rate": 2.641795936202753e-06, + "loss": 1.7283, + "step": 21637 + }, + { + "epoch": 4.72, + "learning_rate": 2.6416866943412715e-06, + "loss": 2.2049, + "step": 21638 + }, + { + "epoch": 4.72, + "learning_rate": 2.6415774524797906e-06, + "loss": 3.5636, + "step": 21639 + }, + { + "epoch": 4.72, + "learning_rate": 2.641468210618309e-06, + "loss": 0.7494, + "step": 21640 + }, + { + "epoch": 4.72, + "learning_rate": 2.6413589687568282e-06, + "loss": 2.6391, + "step": 21641 + }, + { + "epoch": 4.72, + "learning_rate": 2.6412497268953464e-06, + "loss": 0.8892, + "step": 21642 + }, + { + "epoch": 4.72, + "learning_rate": 2.641140485033865e-06, + "loss": 1.9124, + "step": 21643 + }, + { + "epoch": 4.72, + "learning_rate": 2.641031243172384e-06, + "loss": 1.475, + "step": 21644 + }, + { + "epoch": 4.72, + "learning_rate": 2.6409220013109027e-06, + "loss": 2.6234, + "step": 21645 + }, + { + "epoch": 4.72, + "learning_rate": 2.6408127594494217e-06, + "loss": 0.1539, + "step": 21646 + }, + { + "epoch": 4.72, + "learning_rate": 2.64070351758794e-06, + "loss": 2.8854, + "step": 21647 + }, + { + "epoch": 4.72, + "learning_rate": 2.6405942757264585e-06, + "loss": 2.6668, + "step": 21648 + }, + { + "epoch": 4.72, + "learning_rate": 2.6404850338649775e-06, + "loss": 2.6886, + "step": 21649 + }, + { + "epoch": 4.73, + "learning_rate": 2.640375792003496e-06, + "loss": 1.9366, + "step": 21650 + }, + { + "epoch": 4.73, + "learning_rate": 2.6402665501420143e-06, + "loss": 1.1797, + "step": 21651 + }, + { + "epoch": 4.73, + "learning_rate": 2.6401573082805334e-06, + "loss": 2.0749, + "step": 21652 + }, + { + "epoch": 4.73, + "learning_rate": 2.640048066419052e-06, + "loss": 2.5768, + "step": 21653 + }, + { + "epoch": 4.73, + "learning_rate": 2.6399388245575706e-06, + "loss": 1.8037, + "step": 21654 + }, + { + "epoch": 4.73, + "learning_rate": 2.6398295826960896e-06, + "loss": 1.4422, + "step": 21655 + }, + { + "epoch": 4.73, + "learning_rate": 2.639720340834608e-06, + "loss": 2.9847, + "step": 21656 + }, + { + "epoch": 4.73, + "learning_rate": 2.6396110989731264e-06, + "loss": 2.421, + "step": 21657 + }, + { + "epoch": 4.73, + "learning_rate": 2.6395018571116455e-06, + "loss": 2.209, + "step": 21658 + }, + { + "epoch": 4.73, + "learning_rate": 2.639392615250164e-06, + "loss": 2.4836, + "step": 21659 + }, + { + "epoch": 4.73, + "learning_rate": 2.639283373388683e-06, + "loss": 1.9184, + "step": 21660 + }, + { + "epoch": 4.73, + "learning_rate": 2.6391741315272013e-06, + "loss": 1.8026, + "step": 21661 + }, + { + "epoch": 4.73, + "learning_rate": 2.63906488966572e-06, + "loss": 3.2817, + "step": 21662 + }, + { + "epoch": 4.73, + "learning_rate": 2.638955647804239e-06, + "loss": 0.5176, + "step": 21663 + }, + { + "epoch": 4.73, + "learning_rate": 2.6388464059427575e-06, + "loss": 2.418, + "step": 21664 + }, + { + "epoch": 4.73, + "learning_rate": 2.6387371640812757e-06, + "loss": 1.2215, + "step": 21665 + }, + { + "epoch": 4.73, + "learning_rate": 2.6386279222197948e-06, + "loss": 2.5189, + "step": 21666 + }, + { + "epoch": 4.73, + "learning_rate": 2.6385186803583134e-06, + "loss": 0.3921, + "step": 21667 + }, + { + "epoch": 4.73, + "learning_rate": 2.6384094384968324e-06, + "loss": 2.2501, + "step": 21668 + }, + { + "epoch": 4.73, + "learning_rate": 2.638300196635351e-06, + "loss": 1.7974, + "step": 21669 + }, + { + "epoch": 4.73, + "learning_rate": 2.6381909547738692e-06, + "loss": 1.3711, + "step": 21670 + }, + { + "epoch": 4.73, + "learning_rate": 2.6380817129123882e-06, + "loss": 2.7623, + "step": 21671 + }, + { + "epoch": 4.73, + "learning_rate": 2.637972471050907e-06, + "loss": 1.6619, + "step": 21672 + }, + { + "epoch": 4.73, + "learning_rate": 2.637863229189426e-06, + "loss": 1.3939, + "step": 21673 + }, + { + "epoch": 4.73, + "learning_rate": 2.6377539873279445e-06, + "loss": 1.1321, + "step": 21674 + }, + { + "epoch": 4.73, + "learning_rate": 2.6376447454664627e-06, + "loss": 0.8607, + "step": 21675 + }, + { + "epoch": 4.73, + "learning_rate": 2.6375355036049817e-06, + "loss": 1.4552, + "step": 21676 + }, + { + "epoch": 4.73, + "learning_rate": 2.6374262617435003e-06, + "loss": 2.0283, + "step": 21677 + }, + { + "epoch": 4.73, + "learning_rate": 2.637317019882019e-06, + "loss": 0.3254, + "step": 21678 + }, + { + "epoch": 4.73, + "learning_rate": 2.637207778020538e-06, + "loss": 1.6556, + "step": 21679 + }, + { + "epoch": 4.73, + "learning_rate": 2.637098536159056e-06, + "loss": 2.4701, + "step": 21680 + }, + { + "epoch": 4.73, + "learning_rate": 2.636989294297575e-06, + "loss": 1.2139, + "step": 21681 + }, + { + "epoch": 4.73, + "learning_rate": 2.636880052436094e-06, + "loss": 1.1626, + "step": 21682 + }, + { + "epoch": 4.73, + "learning_rate": 2.6367708105746124e-06, + "loss": 0.1542, + "step": 21683 + }, + { + "epoch": 4.73, + "learning_rate": 2.636661568713131e-06, + "loss": 2.1763, + "step": 21684 + }, + { + "epoch": 4.73, + "learning_rate": 2.6365523268516496e-06, + "loss": 0.8079, + "step": 21685 + }, + { + "epoch": 4.73, + "learning_rate": 2.6364430849901683e-06, + "loss": 0.9422, + "step": 21686 + }, + { + "epoch": 4.73, + "learning_rate": 2.6363338431286873e-06, + "loss": 2.6447, + "step": 21687 + }, + { + "epoch": 4.73, + "learning_rate": 2.636224601267206e-06, + "loss": 2.4277, + "step": 21688 + }, + { + "epoch": 4.73, + "learning_rate": 2.636115359405724e-06, + "loss": 2.1618, + "step": 21689 + }, + { + "epoch": 4.73, + "learning_rate": 2.636006117544243e-06, + "loss": 1.2475, + "step": 21690 + }, + { + "epoch": 4.73, + "learning_rate": 2.6358968756827617e-06, + "loss": 1.6782, + "step": 21691 + }, + { + "epoch": 4.73, + "learning_rate": 2.6357876338212808e-06, + "loss": 1.7999, + "step": 21692 + }, + { + "epoch": 4.73, + "learning_rate": 2.6356783919597994e-06, + "loss": 1.8442, + "step": 21693 + }, + { + "epoch": 4.73, + "learning_rate": 2.6355691500983176e-06, + "loss": 1.5373, + "step": 21694 + }, + { + "epoch": 4.73, + "learning_rate": 2.6354599082368366e-06, + "loss": 1.191, + "step": 21695 + }, + { + "epoch": 4.74, + "learning_rate": 2.6353506663753552e-06, + "loss": 0.5648, + "step": 21696 + }, + { + "epoch": 4.74, + "learning_rate": 2.6352414245138743e-06, + "loss": 1.9743, + "step": 21697 + }, + { + "epoch": 4.74, + "learning_rate": 2.6351321826523924e-06, + "loss": 2.6729, + "step": 21698 + }, + { + "epoch": 4.74, + "learning_rate": 2.635022940790911e-06, + "loss": 1.2102, + "step": 21699 + }, + { + "epoch": 4.74, + "learning_rate": 2.63491369892943e-06, + "loss": 1.0946, + "step": 21700 + }, + { + "epoch": 4.74, + "learning_rate": 2.6348044570679487e-06, + "loss": 1.1863, + "step": 21701 + }, + { + "epoch": 4.74, + "learning_rate": 2.6346952152064677e-06, + "loss": 2.7268, + "step": 21702 + }, + { + "epoch": 4.74, + "learning_rate": 2.634585973344986e-06, + "loss": 2.129, + "step": 21703 + }, + { + "epoch": 4.74, + "learning_rate": 2.6344767314835045e-06, + "loss": 1.2316, + "step": 21704 + }, + { + "epoch": 4.74, + "learning_rate": 2.6343674896220236e-06, + "loss": 2.2163, + "step": 21705 + }, + { + "epoch": 4.74, + "learning_rate": 2.634258247760542e-06, + "loss": 0.9739, + "step": 21706 + }, + { + "epoch": 4.74, + "learning_rate": 2.6341490058990608e-06, + "loss": 2.159, + "step": 21707 + }, + { + "epoch": 4.74, + "learning_rate": 2.6340397640375794e-06, + "loss": 0.4986, + "step": 21708 + }, + { + "epoch": 4.74, + "learning_rate": 2.633930522176098e-06, + "loss": 2.7435, + "step": 21709 + }, + { + "epoch": 4.74, + "learning_rate": 2.6338212803146166e-06, + "loss": 1.8303, + "step": 21710 + }, + { + "epoch": 4.74, + "learning_rate": 2.6337120384531357e-06, + "loss": 1.4865, + "step": 21711 + }, + { + "epoch": 4.74, + "learning_rate": 2.6336027965916543e-06, + "loss": 0.2467, + "step": 21712 + }, + { + "epoch": 4.74, + "learning_rate": 2.6334935547301725e-06, + "loss": 2.2815, + "step": 21713 + }, + { + "epoch": 4.74, + "learning_rate": 2.6333843128686915e-06, + "loss": 1.4559, + "step": 21714 + }, + { + "epoch": 4.74, + "learning_rate": 2.63327507100721e-06, + "loss": 1.0184, + "step": 21715 + }, + { + "epoch": 4.74, + "learning_rate": 2.633165829145729e-06, + "loss": 1.4421, + "step": 21716 + }, + { + "epoch": 4.74, + "learning_rate": 2.6330565872842473e-06, + "loss": 0.9312, + "step": 21717 + }, + { + "epoch": 4.74, + "learning_rate": 2.632947345422766e-06, + "loss": 0.973, + "step": 21718 + }, + { + "epoch": 4.74, + "learning_rate": 2.632838103561285e-06, + "loss": 3.6094, + "step": 21719 + }, + { + "epoch": 4.74, + "learning_rate": 2.6327288616998036e-06, + "loss": 2.064, + "step": 21720 + }, + { + "epoch": 4.74, + "learning_rate": 2.6326196198383226e-06, + "loss": 3.1524, + "step": 21721 + }, + { + "epoch": 4.74, + "learning_rate": 2.632510377976841e-06, + "loss": 2.2372, + "step": 21722 + }, + { + "epoch": 4.74, + "learning_rate": 2.6324011361153594e-06, + "loss": 1.6459, + "step": 21723 + }, + { + "epoch": 4.74, + "learning_rate": 2.6322918942538784e-06, + "loss": 1.4753, + "step": 21724 + }, + { + "epoch": 4.74, + "learning_rate": 2.632182652392397e-06, + "loss": 3.0444, + "step": 21725 + }, + { + "epoch": 4.74, + "learning_rate": 2.632073410530916e-06, + "loss": 1.1891, + "step": 21726 + }, + { + "epoch": 4.74, + "learning_rate": 2.6319641686694343e-06, + "loss": 1.6326, + "step": 21727 + }, + { + "epoch": 4.74, + "learning_rate": 2.631854926807953e-06, + "loss": 2.4312, + "step": 21728 + }, + { + "epoch": 4.74, + "learning_rate": 2.631745684946472e-06, + "loss": 2.0991, + "step": 21729 + }, + { + "epoch": 4.74, + "learning_rate": 2.6316364430849905e-06, + "loss": 0.79, + "step": 21730 + }, + { + "epoch": 4.74, + "learning_rate": 2.6315272012235087e-06, + "loss": 1.8206, + "step": 21731 + }, + { + "epoch": 4.74, + "learning_rate": 2.6314179593620278e-06, + "loss": 3.4312, + "step": 21732 + }, + { + "epoch": 4.74, + "learning_rate": 2.6313087175005464e-06, + "loss": 0.3619, + "step": 21733 + }, + { + "epoch": 4.74, + "learning_rate": 2.6311994756390654e-06, + "loss": 1.0288, + "step": 21734 + }, + { + "epoch": 4.74, + "learning_rate": 2.631090233777584e-06, + "loss": 1.3306, + "step": 21735 + }, + { + "epoch": 4.74, + "learning_rate": 2.630980991916102e-06, + "loss": 2.2545, + "step": 21736 + }, + { + "epoch": 4.74, + "learning_rate": 2.6308717500546212e-06, + "loss": 0.7961, + "step": 21737 + }, + { + "epoch": 4.74, + "learning_rate": 2.63076250819314e-06, + "loss": 2.556, + "step": 21738 + }, + { + "epoch": 4.74, + "learning_rate": 2.6306532663316585e-06, + "loss": 1.6645, + "step": 21739 + }, + { + "epoch": 4.74, + "learning_rate": 2.6305440244701775e-06, + "loss": 1.3298, + "step": 21740 + }, + { + "epoch": 4.74, + "learning_rate": 2.6304347826086957e-06, + "loss": 3.8052, + "step": 21741 + }, + { + "epoch": 4.75, + "learning_rate": 2.6303255407472143e-06, + "loss": 2.2945, + "step": 21742 + }, + { + "epoch": 4.75, + "learning_rate": 2.6302162988857333e-06, + "loss": 1.9324, + "step": 21743 + }, + { + "epoch": 4.75, + "learning_rate": 2.630107057024252e-06, + "loss": 1.7704, + "step": 21744 + }, + { + "epoch": 4.75, + "learning_rate": 2.629997815162771e-06, + "loss": 1.1147, + "step": 21745 + }, + { + "epoch": 4.75, + "learning_rate": 2.629888573301289e-06, + "loss": 1.6662, + "step": 21746 + }, + { + "epoch": 4.75, + "learning_rate": 2.6297793314398078e-06, + "loss": 1.7827, + "step": 21747 + }, + { + "epoch": 4.75, + "learning_rate": 2.629670089578327e-06, + "loss": 1.2487, + "step": 21748 + }, + { + "epoch": 4.75, + "learning_rate": 2.6295608477168454e-06, + "loss": 2.9244, + "step": 21749 + }, + { + "epoch": 4.75, + "learning_rate": 2.6294516058553636e-06, + "loss": 1.3323, + "step": 21750 + }, + { + "epoch": 4.75, + "learning_rate": 2.6293423639938826e-06, + "loss": 2.1987, + "step": 21751 + }, + { + "epoch": 4.75, + "learning_rate": 2.6292331221324012e-06, + "loss": 0.8436, + "step": 21752 + }, + { + "epoch": 4.75, + "learning_rate": 2.6291238802709203e-06, + "loss": 0.3777, + "step": 21753 + }, + { + "epoch": 4.75, + "learning_rate": 2.629014638409439e-06, + "loss": 0.895, + "step": 21754 + }, + { + "epoch": 4.75, + "learning_rate": 2.628905396547957e-06, + "loss": 1.2804, + "step": 21755 + }, + { + "epoch": 4.75, + "learning_rate": 2.628796154686476e-06, + "loss": 2.3773, + "step": 21756 + }, + { + "epoch": 4.75, + "learning_rate": 2.6286869128249947e-06, + "loss": 1.2736, + "step": 21757 + }, + { + "epoch": 4.75, + "learning_rate": 2.6285776709635138e-06, + "loss": 1.5294, + "step": 21758 + }, + { + "epoch": 4.75, + "learning_rate": 2.6284684291020324e-06, + "loss": 1.5209, + "step": 21759 + }, + { + "epoch": 4.75, + "learning_rate": 2.6283591872405506e-06, + "loss": 0.8419, + "step": 21760 + }, + { + "epoch": 4.75, + "learning_rate": 2.6282499453790696e-06, + "loss": 1.5951, + "step": 21761 + }, + { + "epoch": 4.75, + "learning_rate": 2.628140703517588e-06, + "loss": 2.3461, + "step": 21762 + }, + { + "epoch": 4.75, + "learning_rate": 2.628031461656107e-06, + "loss": 1.2521, + "step": 21763 + }, + { + "epoch": 4.75, + "learning_rate": 2.6279222197946254e-06, + "loss": 0.9806, + "step": 21764 + }, + { + "epoch": 4.75, + "learning_rate": 2.627812977933144e-06, + "loss": 2.2378, + "step": 21765 + }, + { + "epoch": 4.75, + "learning_rate": 2.6277037360716626e-06, + "loss": 0.6428, + "step": 21766 + }, + { + "epoch": 4.75, + "learning_rate": 2.6275944942101817e-06, + "loss": 3.2154, + "step": 21767 + }, + { + "epoch": 4.75, + "learning_rate": 2.6274852523487003e-06, + "loss": 1.7542, + "step": 21768 + }, + { + "epoch": 4.75, + "learning_rate": 2.6273760104872185e-06, + "loss": 1.624, + "step": 21769 + }, + { + "epoch": 4.75, + "learning_rate": 2.6272667686257375e-06, + "loss": 0.9139, + "step": 21770 + }, + { + "epoch": 4.75, + "learning_rate": 2.627157526764256e-06, + "loss": 2.1921, + "step": 21771 + }, + { + "epoch": 4.75, + "learning_rate": 2.627048284902775e-06, + "loss": 1.8504, + "step": 21772 + }, + { + "epoch": 4.75, + "learning_rate": 2.6269390430412938e-06, + "loss": 1.9242, + "step": 21773 + }, + { + "epoch": 4.75, + "learning_rate": 2.626829801179812e-06, + "loss": 1.936, + "step": 21774 + }, + { + "epoch": 4.75, + "learning_rate": 2.626720559318331e-06, + "loss": 1.7313, + "step": 21775 + }, + { + "epoch": 4.75, + "learning_rate": 2.6266113174568496e-06, + "loss": 1.22, + "step": 21776 + }, + { + "epoch": 4.75, + "learning_rate": 2.6265020755953686e-06, + "loss": 1.9329, + "step": 21777 + }, + { + "epoch": 4.75, + "learning_rate": 2.6263928337338872e-06, + "loss": 1.5773, + "step": 21778 + }, + { + "epoch": 4.75, + "learning_rate": 2.6262835918724054e-06, + "loss": 1.8366, + "step": 21779 + }, + { + "epoch": 4.75, + "learning_rate": 2.6261743500109245e-06, + "loss": 2.2105, + "step": 21780 + }, + { + "epoch": 4.75, + "learning_rate": 2.626065108149443e-06, + "loss": 1.7436, + "step": 21781 + }, + { + "epoch": 4.75, + "learning_rate": 2.625955866287962e-06, + "loss": 1.8747, + "step": 21782 + }, + { + "epoch": 4.75, + "learning_rate": 2.6258466244264803e-06, + "loss": 0.8285, + "step": 21783 + }, + { + "epoch": 4.75, + "learning_rate": 2.625737382564999e-06, + "loss": 1.2369, + "step": 21784 + }, + { + "epoch": 4.75, + "learning_rate": 2.625628140703518e-06, + "loss": 1.1033, + "step": 21785 + }, + { + "epoch": 4.75, + "learning_rate": 2.6255188988420366e-06, + "loss": 0.9279, + "step": 21786 + }, + { + "epoch": 4.75, + "learning_rate": 2.6254096569805556e-06, + "loss": 1.982, + "step": 21787 + }, + { + "epoch": 4.76, + "learning_rate": 2.6253004151190738e-06, + "loss": 1.0219, + "step": 21788 + }, + { + "epoch": 4.76, + "learning_rate": 2.6251911732575924e-06, + "loss": 1.6433, + "step": 21789 + }, + { + "epoch": 4.76, + "learning_rate": 2.6250819313961114e-06, + "loss": 0.8742, + "step": 21790 + }, + { + "epoch": 4.76, + "learning_rate": 2.62497268953463e-06, + "loss": 0.7322, + "step": 21791 + }, + { + "epoch": 4.76, + "learning_rate": 2.6248634476731486e-06, + "loss": 1.5251, + "step": 21792 + }, + { + "epoch": 4.76, + "learning_rate": 2.6247542058116673e-06, + "loss": 1.8955, + "step": 21793 + }, + { + "epoch": 4.76, + "learning_rate": 2.624644963950186e-06, + "loss": 2.1124, + "step": 21794 + }, + { + "epoch": 4.76, + "learning_rate": 2.6245357220887045e-06, + "loss": 0.9474, + "step": 21795 + }, + { + "epoch": 4.76, + "learning_rate": 2.6244264802272235e-06, + "loss": 2.584, + "step": 21796 + }, + { + "epoch": 4.76, + "learning_rate": 2.6243172383657417e-06, + "loss": 2.2625, + "step": 21797 + }, + { + "epoch": 4.76, + "learning_rate": 2.6242079965042603e-06, + "loss": 1.566, + "step": 21798 + }, + { + "epoch": 4.76, + "learning_rate": 2.6240987546427793e-06, + "loss": 2.0011, + "step": 21799 + }, + { + "epoch": 4.76, + "learning_rate": 2.623989512781298e-06, + "loss": 0.8325, + "step": 21800 + }, + { + "epoch": 4.76, + "learning_rate": 2.623880270919817e-06, + "loss": 1.0765, + "step": 21801 + }, + { + "epoch": 4.76, + "learning_rate": 2.623771029058335e-06, + "loss": 1.8849, + "step": 21802 + }, + { + "epoch": 4.76, + "learning_rate": 2.623661787196854e-06, + "loss": 1.425, + "step": 21803 + }, + { + "epoch": 4.76, + "learning_rate": 2.623552545335373e-06, + "loss": 1.3359, + "step": 21804 + }, + { + "epoch": 4.76, + "learning_rate": 2.6234433034738914e-06, + "loss": 1.4599, + "step": 21805 + }, + { + "epoch": 4.76, + "learning_rate": 2.6233340616124105e-06, + "loss": 0.4776, + "step": 21806 + }, + { + "epoch": 4.76, + "learning_rate": 2.6232248197509287e-06, + "loss": 1.3982, + "step": 21807 + }, + { + "epoch": 4.76, + "learning_rate": 2.6231155778894473e-06, + "loss": 2.8177, + "step": 21808 + }, + { + "epoch": 4.76, + "learning_rate": 2.6230063360279663e-06, + "loss": 1.5322, + "step": 21809 + }, + { + "epoch": 4.76, + "learning_rate": 2.622897094166485e-06, + "loss": 0.4745, + "step": 21810 + }, + { + "epoch": 4.76, + "learning_rate": 2.622787852305004e-06, + "loss": 1.5471, + "step": 21811 + }, + { + "epoch": 4.76, + "learning_rate": 2.622678610443522e-06, + "loss": 2.4574, + "step": 21812 + }, + { + "epoch": 4.76, + "learning_rate": 2.6225693685820407e-06, + "loss": 2.7438, + "step": 21813 + }, + { + "epoch": 4.76, + "learning_rate": 2.6224601267205598e-06, + "loss": 0.491, + "step": 21814 + }, + { + "epoch": 4.76, + "learning_rate": 2.6223508848590784e-06, + "loss": 2.0076, + "step": 21815 + }, + { + "epoch": 4.76, + "learning_rate": 2.6222416429975966e-06, + "loss": 1.0481, + "step": 21816 + }, + { + "epoch": 4.76, + "learning_rate": 2.6221324011361156e-06, + "loss": 1.3887, + "step": 21817 + }, + { + "epoch": 4.76, + "learning_rate": 2.6220231592746342e-06, + "loss": 0.9071, + "step": 21818 + }, + { + "epoch": 4.76, + "learning_rate": 2.621913917413153e-06, + "loss": 2.0588, + "step": 21819 + }, + { + "epoch": 4.76, + "learning_rate": 2.621804675551672e-06, + "loss": 1.5799, + "step": 21820 + }, + { + "epoch": 4.76, + "learning_rate": 2.62169543369019e-06, + "loss": 2.0237, + "step": 21821 + }, + { + "epoch": 4.76, + "learning_rate": 2.6215861918287087e-06, + "loss": 0.302, + "step": 21822 + }, + { + "epoch": 4.76, + "learning_rate": 2.6214769499672277e-06, + "loss": 0.5186, + "step": 21823 + }, + { + "epoch": 4.76, + "learning_rate": 2.6213677081057463e-06, + "loss": 2.3304, + "step": 21824 + }, + { + "epoch": 4.76, + "learning_rate": 2.6212584662442654e-06, + "loss": 2.5655, + "step": 21825 + }, + { + "epoch": 4.76, + "learning_rate": 2.6211492243827835e-06, + "loss": 0.2936, + "step": 21826 + }, + { + "epoch": 4.76, + "learning_rate": 2.621039982521302e-06, + "loss": 2.7519, + "step": 21827 + }, + { + "epoch": 4.76, + "learning_rate": 2.620930740659821e-06, + "loss": 2.083, + "step": 21828 + }, + { + "epoch": 4.76, + "learning_rate": 2.62082149879834e-06, + "loss": 1.7005, + "step": 21829 + }, + { + "epoch": 4.76, + "learning_rate": 2.620712256936858e-06, + "loss": 1.4497, + "step": 21830 + }, + { + "epoch": 4.76, + "learning_rate": 2.620603015075377e-06, + "loss": 1.6945, + "step": 21831 + }, + { + "epoch": 4.76, + "learning_rate": 2.6204937732138956e-06, + "loss": 0.691, + "step": 21832 + }, + { + "epoch": 4.76, + "learning_rate": 2.6203845313524147e-06, + "loss": 0.7981, + "step": 21833 + }, + { + "epoch": 4.77, + "learning_rate": 2.6202752894909333e-06, + "loss": 1.8856, + "step": 21834 + }, + { + "epoch": 4.77, + "learning_rate": 2.6201660476294515e-06, + "loss": 1.8853, + "step": 21835 + }, + { + "epoch": 4.77, + "learning_rate": 2.6200568057679705e-06, + "loss": 1.8165, + "step": 21836 + }, + { + "epoch": 4.77, + "learning_rate": 2.619947563906489e-06, + "loss": 1.733, + "step": 21837 + }, + { + "epoch": 4.77, + "learning_rate": 2.619838322045008e-06, + "loss": 1.8355, + "step": 21838 + }, + { + "epoch": 4.77, + "learning_rate": 2.6197290801835268e-06, + "loss": 1.2672, + "step": 21839 + }, + { + "epoch": 4.77, + "learning_rate": 2.619619838322045e-06, + "loss": 1.3541, + "step": 21840 + }, + { + "epoch": 4.77, + "learning_rate": 2.619510596460564e-06, + "loss": 3.5223, + "step": 21841 + }, + { + "epoch": 4.77, + "learning_rate": 2.6194013545990826e-06, + "loss": 2.5915, + "step": 21842 + }, + { + "epoch": 4.77, + "learning_rate": 2.6192921127376016e-06, + "loss": 0.2164, + "step": 21843 + }, + { + "epoch": 4.77, + "learning_rate": 2.6191828708761202e-06, + "loss": 0.9039, + "step": 21844 + }, + { + "epoch": 4.77, + "learning_rate": 2.6190736290146384e-06, + "loss": 1.4003, + "step": 21845 + }, + { + "epoch": 4.77, + "learning_rate": 2.6189643871531575e-06, + "loss": 1.1146, + "step": 21846 + }, + { + "epoch": 4.77, + "learning_rate": 2.618855145291676e-06, + "loss": 1.7843, + "step": 21847 + }, + { + "epoch": 4.77, + "learning_rate": 2.6187459034301947e-06, + "loss": 2.4565, + "step": 21848 + }, + { + "epoch": 4.77, + "learning_rate": 2.6186366615687133e-06, + "loss": 1.6367, + "step": 21849 + }, + { + "epoch": 4.77, + "learning_rate": 2.618527419707232e-06, + "loss": 2.4668, + "step": 21850 + }, + { + "epoch": 4.77, + "learning_rate": 2.6184181778457505e-06, + "loss": 0.2821, + "step": 21851 + }, + { + "epoch": 4.77, + "learning_rate": 2.6183089359842695e-06, + "loss": 1.3674, + "step": 21852 + }, + { + "epoch": 4.77, + "learning_rate": 2.618199694122788e-06, + "loss": 2.3033, + "step": 21853 + }, + { + "epoch": 4.77, + "learning_rate": 2.6180904522613063e-06, + "loss": 1.9119, + "step": 21854 + }, + { + "epoch": 4.77, + "learning_rate": 2.6179812103998254e-06, + "loss": 0.6245, + "step": 21855 + }, + { + "epoch": 4.77, + "learning_rate": 2.617871968538344e-06, + "loss": 0.6596, + "step": 21856 + }, + { + "epoch": 4.77, + "learning_rate": 2.617762726676863e-06, + "loss": 0.9183, + "step": 21857 + }, + { + "epoch": 4.77, + "learning_rate": 2.6176534848153816e-06, + "loss": 3.486, + "step": 21858 + }, + { + "epoch": 4.77, + "learning_rate": 2.6175442429539e-06, + "loss": 1.4104, + "step": 21859 + }, + { + "epoch": 4.77, + "learning_rate": 2.617435001092419e-06, + "loss": 1.1767, + "step": 21860 + }, + { + "epoch": 4.77, + "learning_rate": 2.6173257592309375e-06, + "loss": 2.0669, + "step": 21861 + }, + { + "epoch": 4.77, + "learning_rate": 2.6172165173694565e-06, + "loss": 2.3799, + "step": 21862 + }, + { + "epoch": 4.77, + "learning_rate": 2.6171072755079747e-06, + "loss": 1.5659, + "step": 21863 + }, + { + "epoch": 4.77, + "learning_rate": 2.6169980336464933e-06, + "loss": 1.65, + "step": 21864 + }, + { + "epoch": 4.77, + "learning_rate": 2.6168887917850123e-06, + "loss": 1.2624, + "step": 21865 + }, + { + "epoch": 4.77, + "learning_rate": 2.616779549923531e-06, + "loss": 1.5952, + "step": 21866 + }, + { + "epoch": 4.77, + "learning_rate": 2.61667030806205e-06, + "loss": 1.9098, + "step": 21867 + }, + { + "epoch": 4.77, + "learning_rate": 2.616561066200568e-06, + "loss": 0.8396, + "step": 21868 + }, + { + "epoch": 4.77, + "learning_rate": 2.6164518243390868e-06, + "loss": 2.8031, + "step": 21869 + }, + { + "epoch": 4.77, + "learning_rate": 2.616342582477606e-06, + "loss": 1.9574, + "step": 21870 + }, + { + "epoch": 4.77, + "learning_rate": 2.6162333406161244e-06, + "loss": 1.7837, + "step": 21871 + }, + { + "epoch": 4.77, + "learning_rate": 2.616124098754643e-06, + "loss": 1.2292, + "step": 21872 + }, + { + "epoch": 4.77, + "learning_rate": 2.6160148568931616e-06, + "loss": 1.5028, + "step": 21873 + }, + { + "epoch": 4.77, + "learning_rate": 2.6159056150316803e-06, + "loss": 1.8904, + "step": 21874 + }, + { + "epoch": 4.77, + "learning_rate": 2.615796373170199e-06, + "loss": 0.494, + "step": 21875 + }, + { + "epoch": 4.77, + "learning_rate": 2.615687131308718e-06, + "loss": 1.4474, + "step": 21876 + }, + { + "epoch": 4.77, + "learning_rate": 2.615577889447236e-06, + "loss": 2.1112, + "step": 21877 + }, + { + "epoch": 4.77, + "learning_rate": 2.615468647585755e-06, + "loss": 1.8646, + "step": 21878 + }, + { + "epoch": 4.77, + "learning_rate": 2.6153594057242737e-06, + "loss": 1.4317, + "step": 21879 + }, + { + "epoch": 4.78, + "learning_rate": 2.6152501638627923e-06, + "loss": 2.0, + "step": 21880 + }, + { + "epoch": 4.78, + "learning_rate": 2.6151409220013114e-06, + "loss": 1.6847, + "step": 21881 + }, + { + "epoch": 4.78, + "learning_rate": 2.6150316801398296e-06, + "loss": 1.85, + "step": 21882 + }, + { + "epoch": 4.78, + "learning_rate": 2.614922438278348e-06, + "loss": 2.1866, + "step": 21883 + }, + { + "epoch": 4.78, + "learning_rate": 2.614813196416867e-06, + "loss": 0.7318, + "step": 21884 + }, + { + "epoch": 4.78, + "learning_rate": 2.614703954555386e-06, + "loss": 2.7164, + "step": 21885 + }, + { + "epoch": 4.78, + "learning_rate": 2.614594712693905e-06, + "loss": 1.4426, + "step": 21886 + }, + { + "epoch": 4.78, + "learning_rate": 2.614485470832423e-06, + "loss": 1.9507, + "step": 21887 + }, + { + "epoch": 4.78, + "learning_rate": 2.6143762289709417e-06, + "loss": 1.462, + "step": 21888 + }, + { + "epoch": 4.78, + "learning_rate": 2.6142669871094607e-06, + "loss": 1.9168, + "step": 21889 + }, + { + "epoch": 4.78, + "learning_rate": 2.6141577452479793e-06, + "loss": 1.116, + "step": 21890 + }, + { + "epoch": 4.78, + "learning_rate": 2.6140485033864983e-06, + "loss": 0.7207, + "step": 21891 + }, + { + "epoch": 4.78, + "learning_rate": 2.6139392615250165e-06, + "loss": 1.4665, + "step": 21892 + }, + { + "epoch": 4.78, + "learning_rate": 2.613830019663535e-06, + "loss": 3.3372, + "step": 21893 + }, + { + "epoch": 4.78, + "learning_rate": 2.613720777802054e-06, + "loss": 2.8395, + "step": 21894 + }, + { + "epoch": 4.78, + "learning_rate": 2.6136115359405728e-06, + "loss": 1.0492, + "step": 21895 + }, + { + "epoch": 4.78, + "learning_rate": 2.613502294079091e-06, + "loss": 2.3059, + "step": 21896 + }, + { + "epoch": 4.78, + "learning_rate": 2.61339305221761e-06, + "loss": 1.9808, + "step": 21897 + }, + { + "epoch": 4.78, + "learning_rate": 2.6132838103561286e-06, + "loss": 0.8187, + "step": 21898 + }, + { + "epoch": 4.78, + "learning_rate": 2.6131745684946476e-06, + "loss": 1.7043, + "step": 21899 + }, + { + "epoch": 4.78, + "learning_rate": 2.6130653266331663e-06, + "loss": 0.4747, + "step": 21900 + }, + { + "epoch": 4.78, + "learning_rate": 2.6129560847716844e-06, + "loss": 1.6506, + "step": 21901 + }, + { + "epoch": 4.78, + "learning_rate": 2.6128468429102035e-06, + "loss": 2.5502, + "step": 21902 + }, + { + "epoch": 4.78, + "learning_rate": 2.612737601048722e-06, + "loss": 2.3277, + "step": 21903 + }, + { + "epoch": 4.78, + "learning_rate": 2.6126283591872407e-06, + "loss": 1.8175, + "step": 21904 + }, + { + "epoch": 4.78, + "learning_rate": 2.6125191173257597e-06, + "loss": 0.766, + "step": 21905 + }, + { + "epoch": 4.78, + "learning_rate": 2.612409875464278e-06, + "loss": 1.6788, + "step": 21906 + }, + { + "epoch": 4.78, + "learning_rate": 2.6123006336027965e-06, + "loss": 1.7426, + "step": 21907 + }, + { + "epoch": 4.78, + "learning_rate": 2.6121913917413156e-06, + "loss": 1.6106, + "step": 21908 + }, + { + "epoch": 4.78, + "learning_rate": 2.612082149879834e-06, + "loss": 1.2778, + "step": 21909 + }, + { + "epoch": 4.78, + "learning_rate": 2.6119729080183524e-06, + "loss": 1.2595, + "step": 21910 + }, + { + "epoch": 4.78, + "learning_rate": 2.6118636661568714e-06, + "loss": 1.1541, + "step": 21911 + }, + { + "epoch": 4.78, + "learning_rate": 2.61175442429539e-06, + "loss": 1.4057, + "step": 21912 + }, + { + "epoch": 4.78, + "learning_rate": 2.611645182433909e-06, + "loss": 1.1966, + "step": 21913 + }, + { + "epoch": 4.78, + "learning_rate": 2.6115359405724277e-06, + "loss": 1.2965, + "step": 21914 + }, + { + "epoch": 4.78, + "learning_rate": 2.611426698710946e-06, + "loss": 1.8476, + "step": 21915 + }, + { + "epoch": 4.78, + "learning_rate": 2.611317456849465e-06, + "loss": 2.3158, + "step": 21916 + }, + { + "epoch": 4.78, + "learning_rate": 2.6112082149879835e-06, + "loss": 0.3915, + "step": 21917 + }, + { + "epoch": 4.78, + "learning_rate": 2.6110989731265025e-06, + "loss": 1.5832, + "step": 21918 + }, + { + "epoch": 4.78, + "learning_rate": 2.610989731265021e-06, + "loss": 0.9931, + "step": 21919 + }, + { + "epoch": 4.78, + "learning_rate": 2.6108804894035393e-06, + "loss": 1.6942, + "step": 21920 + }, + { + "epoch": 4.78, + "learning_rate": 2.6107712475420584e-06, + "loss": 1.1551, + "step": 21921 + }, + { + "epoch": 4.78, + "learning_rate": 2.610662005680577e-06, + "loss": 1.5099, + "step": 21922 + }, + { + "epoch": 4.78, + "learning_rate": 2.610552763819096e-06, + "loss": 3.038, + "step": 21923 + }, + { + "epoch": 4.78, + "learning_rate": 2.6104435219576146e-06, + "loss": 1.6806, + "step": 21924 + }, + { + "epoch": 4.79, + "learning_rate": 2.610334280096133e-06, + "loss": 1.3005, + "step": 21925 + }, + { + "epoch": 4.79, + "learning_rate": 2.610225038234652e-06, + "loss": 1.1007, + "step": 21926 + }, + { + "epoch": 4.79, + "learning_rate": 2.6101157963731704e-06, + "loss": 1.6281, + "step": 21927 + }, + { + "epoch": 4.79, + "learning_rate": 2.6100065545116895e-06, + "loss": 1.4129, + "step": 21928 + }, + { + "epoch": 4.79, + "learning_rate": 2.6098973126502077e-06, + "loss": 0.8133, + "step": 21929 + }, + { + "epoch": 4.79, + "learning_rate": 2.6097880707887263e-06, + "loss": 0.3338, + "step": 21930 + }, + { + "epoch": 4.79, + "learning_rate": 2.6096788289272453e-06, + "loss": 0.8602, + "step": 21931 + }, + { + "epoch": 4.79, + "learning_rate": 2.609569587065764e-06, + "loss": 0.924, + "step": 21932 + }, + { + "epoch": 4.79, + "learning_rate": 2.6094603452042825e-06, + "loss": 1.3037, + "step": 21933 + }, + { + "epoch": 4.79, + "learning_rate": 2.609351103342801e-06, + "loss": 1.0902, + "step": 21934 + }, + { + "epoch": 4.79, + "learning_rate": 2.6092418614813198e-06, + "loss": 0.9389, + "step": 21935 + }, + { + "epoch": 4.79, + "learning_rate": 2.6091326196198384e-06, + "loss": 0.6406, + "step": 21936 + }, + { + "epoch": 4.79, + "learning_rate": 2.6090233777583574e-06, + "loss": 0.6946, + "step": 21937 + }, + { + "epoch": 4.79, + "learning_rate": 2.608914135896876e-06, + "loss": 1.2671, + "step": 21938 + }, + { + "epoch": 4.79, + "learning_rate": 2.608804894035394e-06, + "loss": 0.8431, + "step": 21939 + }, + { + "epoch": 4.79, + "learning_rate": 2.6086956521739132e-06, + "loss": 0.7752, + "step": 21940 + }, + { + "epoch": 4.79, + "learning_rate": 2.608586410312432e-06, + "loss": 1.5187, + "step": 21941 + }, + { + "epoch": 4.79, + "learning_rate": 2.608477168450951e-06, + "loss": 2.441, + "step": 21942 + }, + { + "epoch": 4.79, + "learning_rate": 2.608367926589469e-06, + "loss": 1.7914, + "step": 21943 + }, + { + "epoch": 4.79, + "learning_rate": 2.6082586847279877e-06, + "loss": 1.1543, + "step": 21944 + }, + { + "epoch": 4.79, + "learning_rate": 2.6081494428665067e-06, + "loss": 1.4788, + "step": 21945 + }, + { + "epoch": 4.79, + "learning_rate": 2.6080402010050253e-06, + "loss": 0.4372, + "step": 21946 + }, + { + "epoch": 4.79, + "learning_rate": 2.6079309591435444e-06, + "loss": 1.0123, + "step": 21947 + }, + { + "epoch": 4.79, + "learning_rate": 2.6078217172820625e-06, + "loss": 1.2761, + "step": 21948 + }, + { + "epoch": 4.79, + "learning_rate": 2.607712475420581e-06, + "loss": 1.7201, + "step": 21949 + }, + { + "epoch": 4.79, + "learning_rate": 2.6076032335591e-06, + "loss": 1.7898, + "step": 21950 + }, + { + "epoch": 4.79, + "learning_rate": 2.607493991697619e-06, + "loss": 2.6577, + "step": 21951 + }, + { + "epoch": 4.79, + "learning_rate": 2.607384749836138e-06, + "loss": 2.9372, + "step": 21952 + }, + { + "epoch": 4.79, + "learning_rate": 2.607275507974656e-06, + "loss": 1.102, + "step": 21953 + }, + { + "epoch": 4.79, + "learning_rate": 2.6071662661131746e-06, + "loss": 3.0112, + "step": 21954 + }, + { + "epoch": 4.79, + "learning_rate": 2.6070570242516937e-06, + "loss": 1.4562, + "step": 21955 + }, + { + "epoch": 4.79, + "learning_rate": 2.6069477823902123e-06, + "loss": 1.8455, + "step": 21956 + }, + { + "epoch": 4.79, + "learning_rate": 2.606838540528731e-06, + "loss": 1.1285, + "step": 21957 + }, + { + "epoch": 4.79, + "learning_rate": 2.6067292986672495e-06, + "loss": 1.5943, + "step": 21958 + }, + { + "epoch": 4.79, + "learning_rate": 2.606620056805768e-06, + "loss": 1.9128, + "step": 21959 + }, + { + "epoch": 4.79, + "learning_rate": 2.6065108149442867e-06, + "loss": 1.1187, + "step": 21960 + }, + { + "epoch": 4.79, + "learning_rate": 2.6064015730828058e-06, + "loss": 1.3553, + "step": 21961 + }, + { + "epoch": 4.79, + "learning_rate": 2.606292331221324e-06, + "loss": 2.8574, + "step": 21962 + }, + { + "epoch": 4.79, + "learning_rate": 2.6061830893598426e-06, + "loss": 1.3295, + "step": 21963 + }, + { + "epoch": 4.79, + "learning_rate": 2.6060738474983616e-06, + "loss": 2.0038, + "step": 21964 + }, + { + "epoch": 4.79, + "learning_rate": 2.60596460563688e-06, + "loss": 0.5947, + "step": 21965 + }, + { + "epoch": 4.79, + "learning_rate": 2.6058553637753992e-06, + "loss": 0.9837, + "step": 21966 + }, + { + "epoch": 4.79, + "learning_rate": 2.6057461219139174e-06, + "loss": 0.7165, + "step": 21967 + }, + { + "epoch": 4.79, + "learning_rate": 2.605636880052436e-06, + "loss": 1.0334, + "step": 21968 + }, + { + "epoch": 4.79, + "learning_rate": 2.605527638190955e-06, + "loss": 1.1592, + "step": 21969 + }, + { + "epoch": 4.79, + "learning_rate": 2.6054183963294737e-06, + "loss": 0.8096, + "step": 21970 + }, + { + "epoch": 4.8, + "learning_rate": 2.6053091544679927e-06, + "loss": 2.6925, + "step": 21971 + }, + { + "epoch": 4.8, + "learning_rate": 2.605199912606511e-06, + "loss": 1.1517, + "step": 21972 + }, + { + "epoch": 4.8, + "learning_rate": 2.6050906707450295e-06, + "loss": 1.2495, + "step": 21973 + }, + { + "epoch": 4.8, + "learning_rate": 2.6049814288835486e-06, + "loss": 1.822, + "step": 21974 + }, + { + "epoch": 4.8, + "learning_rate": 2.604872187022067e-06, + "loss": 2.4795, + "step": 21975 + }, + { + "epoch": 4.8, + "learning_rate": 2.6047629451605854e-06, + "loss": 3.0999, + "step": 21976 + }, + { + "epoch": 4.8, + "learning_rate": 2.6046537032991044e-06, + "loss": 1.4148, + "step": 21977 + }, + { + "epoch": 4.8, + "learning_rate": 2.604544461437623e-06, + "loss": 1.7952, + "step": 21978 + }, + { + "epoch": 4.8, + "learning_rate": 2.604435219576142e-06, + "loss": 1.0759, + "step": 21979 + }, + { + "epoch": 4.8, + "learning_rate": 2.6043259777146606e-06, + "loss": 1.2542, + "step": 21980 + }, + { + "epoch": 4.8, + "learning_rate": 2.604216735853179e-06, + "loss": 0.6957, + "step": 21981 + }, + { + "epoch": 4.8, + "learning_rate": 2.604107493991698e-06, + "loss": 3.9825, + "step": 21982 + }, + { + "epoch": 4.8, + "learning_rate": 2.6039982521302165e-06, + "loss": 1.0844, + "step": 21983 + }, + { + "epoch": 4.8, + "learning_rate": 2.6038890102687355e-06, + "loss": 2.4349, + "step": 21984 + }, + { + "epoch": 4.8, + "learning_rate": 2.603779768407254e-06, + "loss": 0.3904, + "step": 21985 + }, + { + "epoch": 4.8, + "learning_rate": 2.6036705265457723e-06, + "loss": 1.0354, + "step": 21986 + }, + { + "epoch": 4.8, + "learning_rate": 2.6035612846842913e-06, + "loss": 1.4779, + "step": 21987 + }, + { + "epoch": 4.8, + "learning_rate": 2.60345204282281e-06, + "loss": 1.9181, + "step": 21988 + }, + { + "epoch": 4.8, + "learning_rate": 2.6033428009613286e-06, + "loss": 1.3489, + "step": 21989 + }, + { + "epoch": 4.8, + "learning_rate": 2.6032335590998476e-06, + "loss": 0.9459, + "step": 21990 + }, + { + "epoch": 4.8, + "learning_rate": 2.6031243172383658e-06, + "loss": 1.7159, + "step": 21991 + }, + { + "epoch": 4.8, + "learning_rate": 2.6030150753768844e-06, + "loss": 1.7761, + "step": 21992 + }, + { + "epoch": 4.8, + "learning_rate": 2.6029058335154034e-06, + "loss": 1.2715, + "step": 21993 + }, + { + "epoch": 4.8, + "learning_rate": 2.602796591653922e-06, + "loss": 1.0966, + "step": 21994 + }, + { + "epoch": 4.8, + "learning_rate": 2.6026873497924402e-06, + "loss": 1.0719, + "step": 21995 + }, + { + "epoch": 4.8, + "learning_rate": 2.6025781079309593e-06, + "loss": 2.7518, + "step": 21996 + }, + { + "epoch": 4.8, + "learning_rate": 2.602468866069478e-06, + "loss": 1.097, + "step": 21997 + }, + { + "epoch": 4.8, + "learning_rate": 2.602359624207997e-06, + "loss": 1.1903, + "step": 21998 + }, + { + "epoch": 4.8, + "learning_rate": 2.6022503823465155e-06, + "loss": 1.6364, + "step": 21999 + }, + { + "epoch": 4.8, + "learning_rate": 2.6021411404850337e-06, + "loss": 1.7724, + "step": 22000 + }, + { + "epoch": 4.8, + "learning_rate": 2.6020318986235527e-06, + "loss": 1.609, + "step": 22001 + }, + { + "epoch": 4.8, + "learning_rate": 2.6019226567620714e-06, + "loss": 2.1694, + "step": 22002 + }, + { + "epoch": 4.8, + "learning_rate": 2.6018134149005904e-06, + "loss": 0.3658, + "step": 22003 + }, + { + "epoch": 4.8, + "learning_rate": 2.601704173039109e-06, + "loss": 1.1442, + "step": 22004 + }, + { + "epoch": 4.8, + "learning_rate": 2.601594931177627e-06, + "loss": 2.3716, + "step": 22005 + }, + { + "epoch": 4.8, + "learning_rate": 2.6014856893161462e-06, + "loss": 0.5422, + "step": 22006 + }, + { + "epoch": 4.8, + "learning_rate": 2.601376447454665e-06, + "loss": 0.9464, + "step": 22007 + }, + { + "epoch": 4.8, + "learning_rate": 2.601267205593184e-06, + "loss": 0.7967, + "step": 22008 + }, + { + "epoch": 4.8, + "learning_rate": 2.601157963731702e-06, + "loss": 1.0703, + "step": 22009 + }, + { + "epoch": 4.8, + "learning_rate": 2.6010487218702207e-06, + "loss": 1.2735, + "step": 22010 + }, + { + "epoch": 4.8, + "learning_rate": 2.6009394800087397e-06, + "loss": 0.6979, + "step": 22011 + }, + { + "epoch": 4.8, + "learning_rate": 2.6008302381472583e-06, + "loss": 0.7264, + "step": 22012 + }, + { + "epoch": 4.8, + "learning_rate": 2.600720996285777e-06, + "loss": 2.7016, + "step": 22013 + }, + { + "epoch": 4.8, + "learning_rate": 2.6006117544242955e-06, + "loss": 1.7077, + "step": 22014 + }, + { + "epoch": 4.8, + "learning_rate": 2.600502512562814e-06, + "loss": 1.0307, + "step": 22015 + }, + { + "epoch": 4.8, + "learning_rate": 2.6003932707013328e-06, + "loss": 2.2247, + "step": 22016 + }, + { + "epoch": 4.81, + "learning_rate": 2.600284028839852e-06, + "loss": 1.5368, + "step": 22017 + }, + { + "epoch": 4.81, + "learning_rate": 2.6001747869783704e-06, + "loss": 0.3764, + "step": 22018 + }, + { + "epoch": 4.81, + "learning_rate": 2.600065545116889e-06, + "loss": 2.1375, + "step": 22019 + }, + { + "epoch": 4.81, + "learning_rate": 2.5999563032554076e-06, + "loss": 0.0838, + "step": 22020 + }, + { + "epoch": 4.81, + "learning_rate": 2.5998470613939262e-06, + "loss": 1.0594, + "step": 22021 + }, + { + "epoch": 4.81, + "learning_rate": 2.5997378195324453e-06, + "loss": 1.9256, + "step": 22022 + }, + { + "epoch": 4.81, + "learning_rate": 2.599628577670964e-06, + "loss": 1.8306, + "step": 22023 + }, + { + "epoch": 4.81, + "learning_rate": 2.599519335809482e-06, + "loss": 0.7578, + "step": 22024 + }, + { + "epoch": 4.81, + "learning_rate": 2.599410093948001e-06, + "loss": 1.201, + "step": 22025 + }, + { + "epoch": 4.81, + "learning_rate": 2.5993008520865197e-06, + "loss": 2.1875, + "step": 22026 + }, + { + "epoch": 4.81, + "learning_rate": 2.5991916102250387e-06, + "loss": 2.0352, + "step": 22027 + }, + { + "epoch": 4.81, + "learning_rate": 2.599082368363557e-06, + "loss": 1.0381, + "step": 22028 + }, + { + "epoch": 4.81, + "learning_rate": 2.5989731265020755e-06, + "loss": 0.9358, + "step": 22029 + }, + { + "epoch": 4.81, + "learning_rate": 2.5988638846405946e-06, + "loss": 2.531, + "step": 22030 + }, + { + "epoch": 4.81, + "learning_rate": 2.598754642779113e-06, + "loss": 0.5851, + "step": 22031 + }, + { + "epoch": 4.81, + "learning_rate": 2.5986454009176322e-06, + "loss": 1.5531, + "step": 22032 + }, + { + "epoch": 4.81, + "learning_rate": 2.5985361590561504e-06, + "loss": 0.9251, + "step": 22033 + }, + { + "epoch": 4.81, + "learning_rate": 2.598426917194669e-06, + "loss": 1.9442, + "step": 22034 + }, + { + "epoch": 4.81, + "learning_rate": 2.598317675333188e-06, + "loss": 0.7694, + "step": 22035 + }, + { + "epoch": 4.81, + "learning_rate": 2.5982084334717067e-06, + "loss": 1.8163, + "step": 22036 + }, + { + "epoch": 4.81, + "learning_rate": 2.5980991916102257e-06, + "loss": 0.7262, + "step": 22037 + }, + { + "epoch": 4.81, + "learning_rate": 2.597989949748744e-06, + "loss": 2.5156, + "step": 22038 + }, + { + "epoch": 4.81, + "learning_rate": 2.5978807078872625e-06, + "loss": 1.0924, + "step": 22039 + }, + { + "epoch": 4.81, + "learning_rate": 2.5977714660257815e-06, + "loss": 0.9902, + "step": 22040 + }, + { + "epoch": 4.81, + "learning_rate": 2.5976622241643e-06, + "loss": 2.3106, + "step": 22041 + }, + { + "epoch": 4.81, + "learning_rate": 2.5975529823028183e-06, + "loss": 1.8847, + "step": 22042 + }, + { + "epoch": 4.81, + "learning_rate": 2.5974437404413374e-06, + "loss": 1.3355, + "step": 22043 + }, + { + "epoch": 4.81, + "learning_rate": 2.597334498579856e-06, + "loss": 1.0566, + "step": 22044 + }, + { + "epoch": 4.81, + "learning_rate": 2.5972252567183746e-06, + "loss": 0.3545, + "step": 22045 + }, + { + "epoch": 4.81, + "learning_rate": 2.5971160148568936e-06, + "loss": 0.5565, + "step": 22046 + }, + { + "epoch": 4.81, + "learning_rate": 2.597006772995412e-06, + "loss": 1.3061, + "step": 22047 + }, + { + "epoch": 4.81, + "learning_rate": 2.5968975311339304e-06, + "loss": 0.9956, + "step": 22048 + }, + { + "epoch": 4.81, + "learning_rate": 2.5967882892724495e-06, + "loss": 0.6515, + "step": 22049 + }, + { + "epoch": 4.81, + "learning_rate": 2.596679047410968e-06, + "loss": 2.1888, + "step": 22050 + }, + { + "epoch": 4.81, + "learning_rate": 2.596569805549487e-06, + "loss": 1.4592, + "step": 22051 + }, + { + "epoch": 4.81, + "learning_rate": 2.5964605636880053e-06, + "loss": 1.6844, + "step": 22052 + }, + { + "epoch": 4.81, + "learning_rate": 2.596351321826524e-06, + "loss": 1.4257, + "step": 22053 + }, + { + "epoch": 4.81, + "learning_rate": 2.596242079965043e-06, + "loss": 2.3765, + "step": 22054 + }, + { + "epoch": 4.81, + "learning_rate": 2.5961328381035615e-06, + "loss": 0.8462, + "step": 22055 + }, + { + "epoch": 4.81, + "learning_rate": 2.5960235962420806e-06, + "loss": 1.7483, + "step": 22056 + }, + { + "epoch": 4.81, + "learning_rate": 2.5959143543805988e-06, + "loss": 1.7277, + "step": 22057 + }, + { + "epoch": 4.81, + "learning_rate": 2.5958051125191174e-06, + "loss": 1.3776, + "step": 22058 + }, + { + "epoch": 4.81, + "learning_rate": 2.5956958706576364e-06, + "loss": 2.072, + "step": 22059 + }, + { + "epoch": 4.81, + "learning_rate": 2.595586628796155e-06, + "loss": 1.8224, + "step": 22060 + }, + { + "epoch": 4.81, + "learning_rate": 2.5954773869346732e-06, + "loss": 1.4387, + "step": 22061 + }, + { + "epoch": 4.81, + "learning_rate": 2.5953681450731922e-06, + "loss": 1.421, + "step": 22062 + }, + { + "epoch": 4.82, + "learning_rate": 2.595258903211711e-06, + "loss": 1.955, + "step": 22063 + }, + { + "epoch": 4.82, + "learning_rate": 2.59514966135023e-06, + "loss": 1.7155, + "step": 22064 + }, + { + "epoch": 4.82, + "learning_rate": 2.5950404194887485e-06, + "loss": 0.7423, + "step": 22065 + }, + { + "epoch": 4.82, + "learning_rate": 2.5949311776272667e-06, + "loss": 1.7459, + "step": 22066 + }, + { + "epoch": 4.82, + "learning_rate": 2.5948219357657857e-06, + "loss": 2.3656, + "step": 22067 + }, + { + "epoch": 4.82, + "learning_rate": 2.5947126939043043e-06, + "loss": 1.3862, + "step": 22068 + }, + { + "epoch": 4.82, + "learning_rate": 2.594603452042823e-06, + "loss": 1.7207, + "step": 22069 + }, + { + "epoch": 4.82, + "learning_rate": 2.594494210181342e-06, + "loss": 1.2395, + "step": 22070 + }, + { + "epoch": 4.82, + "learning_rate": 2.59438496831986e-06, + "loss": 0.8747, + "step": 22071 + }, + { + "epoch": 4.82, + "learning_rate": 2.594275726458379e-06, + "loss": 1.9136, + "step": 22072 + }, + { + "epoch": 4.82, + "learning_rate": 2.594166484596898e-06, + "loss": 0.4458, + "step": 22073 + }, + { + "epoch": 4.82, + "learning_rate": 2.5940572427354164e-06, + "loss": 1.9761, + "step": 22074 + }, + { + "epoch": 4.82, + "learning_rate": 2.593948000873935e-06, + "loss": 0.6149, + "step": 22075 + }, + { + "epoch": 4.82, + "learning_rate": 2.5938387590124537e-06, + "loss": 2.4524, + "step": 22076 + }, + { + "epoch": 4.82, + "learning_rate": 2.5937295171509723e-06, + "loss": 1.6082, + "step": 22077 + }, + { + "epoch": 4.82, + "learning_rate": 2.5936202752894913e-06, + "loss": 1.6924, + "step": 22078 + }, + { + "epoch": 4.82, + "learning_rate": 2.59351103342801e-06, + "loss": 1.4936, + "step": 22079 + }, + { + "epoch": 4.82, + "learning_rate": 2.593401791566528e-06, + "loss": 2.9887, + "step": 22080 + }, + { + "epoch": 4.82, + "learning_rate": 2.593292549705047e-06, + "loss": 1.4781, + "step": 22081 + }, + { + "epoch": 4.82, + "learning_rate": 2.5931833078435657e-06, + "loss": 0.8637, + "step": 22082 + }, + { + "epoch": 4.82, + "learning_rate": 2.5930740659820848e-06, + "loss": 0.9389, + "step": 22083 + }, + { + "epoch": 4.82, + "learning_rate": 2.5929648241206034e-06, + "loss": 1.2488, + "step": 22084 + }, + { + "epoch": 4.82, + "learning_rate": 2.5928555822591216e-06, + "loss": 1.471, + "step": 22085 + }, + { + "epoch": 4.82, + "learning_rate": 2.5927463403976406e-06, + "loss": 2.741, + "step": 22086 + }, + { + "epoch": 4.82, + "learning_rate": 2.5926370985361592e-06, + "loss": 1.3935, + "step": 22087 + }, + { + "epoch": 4.82, + "learning_rate": 2.5925278566746783e-06, + "loss": 0.3829, + "step": 22088 + }, + { + "epoch": 4.82, + "learning_rate": 2.592418614813197e-06, + "loss": 0.6059, + "step": 22089 + }, + { + "epoch": 4.82, + "learning_rate": 2.592309372951715e-06, + "loss": 1.383, + "step": 22090 + }, + { + "epoch": 4.82, + "learning_rate": 2.592200131090234e-06, + "loss": 1.249, + "step": 22091 + }, + { + "epoch": 4.82, + "learning_rate": 2.5920908892287527e-06, + "loss": 1.3826, + "step": 22092 + }, + { + "epoch": 4.82, + "learning_rate": 2.5919816473672717e-06, + "loss": 3.0106, + "step": 22093 + }, + { + "epoch": 4.82, + "learning_rate": 2.59187240550579e-06, + "loss": 1.7519, + "step": 22094 + }, + { + "epoch": 4.82, + "learning_rate": 2.5917631636443085e-06, + "loss": 1.5083, + "step": 22095 + }, + { + "epoch": 4.82, + "learning_rate": 2.5916539217828276e-06, + "loss": 1.0513, + "step": 22096 + }, + { + "epoch": 4.82, + "learning_rate": 2.591544679921346e-06, + "loss": 1.3738, + "step": 22097 + }, + { + "epoch": 4.82, + "learning_rate": 2.5914354380598648e-06, + "loss": 1.6035, + "step": 22098 + }, + { + "epoch": 4.82, + "learning_rate": 2.5913261961983834e-06, + "loss": 0.6433, + "step": 22099 + }, + { + "epoch": 4.82, + "learning_rate": 2.591216954336902e-06, + "loss": 2.3546, + "step": 22100 + }, + { + "epoch": 4.82, + "learning_rate": 2.5911077124754206e-06, + "loss": 0.6789, + "step": 22101 + }, + { + "epoch": 4.82, + "learning_rate": 2.5909984706139397e-06, + "loss": 1.4079, + "step": 22102 + }, + { + "epoch": 4.82, + "learning_rate": 2.5908892287524583e-06, + "loss": 1.7346, + "step": 22103 + }, + { + "epoch": 4.82, + "learning_rate": 2.5907799868909765e-06, + "loss": 2.5751, + "step": 22104 + }, + { + "epoch": 4.82, + "learning_rate": 2.5906707450294955e-06, + "loss": 0.4379, + "step": 22105 + }, + { + "epoch": 4.82, + "learning_rate": 2.590561503168014e-06, + "loss": 0.8116, + "step": 22106 + }, + { + "epoch": 4.82, + "learning_rate": 2.590452261306533e-06, + "loss": 1.6591, + "step": 22107 + }, + { + "epoch": 4.82, + "learning_rate": 2.5903430194450513e-06, + "loss": 2.0047, + "step": 22108 + }, + { + "epoch": 4.83, + "learning_rate": 2.59023377758357e-06, + "loss": 1.9619, + "step": 22109 + }, + { + "epoch": 4.83, + "learning_rate": 2.590124535722089e-06, + "loss": 1.5179, + "step": 22110 + }, + { + "epoch": 4.83, + "learning_rate": 2.5900152938606076e-06, + "loss": 2.2882, + "step": 22111 + }, + { + "epoch": 4.83, + "learning_rate": 2.5899060519991266e-06, + "loss": 1.8419, + "step": 22112 + }, + { + "epoch": 4.83, + "learning_rate": 2.589796810137645e-06, + "loss": 1.8773, + "step": 22113 + }, + { + "epoch": 4.83, + "learning_rate": 2.5896875682761634e-06, + "loss": 1.1091, + "step": 22114 + }, + { + "epoch": 4.83, + "learning_rate": 2.5895783264146824e-06, + "loss": 1.8904, + "step": 22115 + }, + { + "epoch": 4.83, + "learning_rate": 2.589469084553201e-06, + "loss": 1.001, + "step": 22116 + }, + { + "epoch": 4.83, + "learning_rate": 2.58935984269172e-06, + "loss": 2.0885, + "step": 22117 + }, + { + "epoch": 4.83, + "learning_rate": 2.5892506008302383e-06, + "loss": 1.1796, + "step": 22118 + }, + { + "epoch": 4.83, + "learning_rate": 2.589141358968757e-06, + "loss": 2.6304, + "step": 22119 + }, + { + "epoch": 4.83, + "learning_rate": 2.589032117107276e-06, + "loss": 1.1806, + "step": 22120 + }, + { + "epoch": 4.83, + "learning_rate": 2.5889228752457945e-06, + "loss": 1.3254, + "step": 22121 + }, + { + "epoch": 4.83, + "learning_rate": 2.588813633384313e-06, + "loss": 1.865, + "step": 22122 + }, + { + "epoch": 4.83, + "learning_rate": 2.5887043915228318e-06, + "loss": 0.732, + "step": 22123 + }, + { + "epoch": 4.83, + "learning_rate": 2.5885951496613504e-06, + "loss": 1.5266, + "step": 22124 + }, + { + "epoch": 4.83, + "learning_rate": 2.5884859077998694e-06, + "loss": 1.7551, + "step": 22125 + }, + { + "epoch": 4.83, + "learning_rate": 2.588376665938388e-06, + "loss": 0.2917, + "step": 22126 + }, + { + "epoch": 4.83, + "learning_rate": 2.588267424076906e-06, + "loss": 1.4863, + "step": 22127 + }, + { + "epoch": 4.83, + "learning_rate": 2.5881581822154252e-06, + "loss": 1.0714, + "step": 22128 + }, + { + "epoch": 4.83, + "learning_rate": 2.588048940353944e-06, + "loss": 0.857, + "step": 22129 + }, + { + "epoch": 4.83, + "learning_rate": 2.5879396984924625e-06, + "loss": 2.1546, + "step": 22130 + }, + { + "epoch": 4.83, + "learning_rate": 2.5878304566309815e-06, + "loss": 1.3469, + "step": 22131 + }, + { + "epoch": 4.83, + "learning_rate": 2.5877212147694997e-06, + "loss": 1.9715, + "step": 22132 + }, + { + "epoch": 4.83, + "learning_rate": 2.5876119729080183e-06, + "loss": 1.6468, + "step": 22133 + }, + { + "epoch": 4.83, + "learning_rate": 2.5875027310465373e-06, + "loss": 1.7085, + "step": 22134 + }, + { + "epoch": 4.83, + "learning_rate": 2.587393489185056e-06, + "loss": 1.3416, + "step": 22135 + }, + { + "epoch": 4.83, + "learning_rate": 2.587284247323575e-06, + "loss": 1.7365, + "step": 22136 + }, + { + "epoch": 4.83, + "learning_rate": 2.587175005462093e-06, + "loss": 1.8938, + "step": 22137 + }, + { + "epoch": 4.83, + "learning_rate": 2.5870657636006118e-06, + "loss": 2.3719, + "step": 22138 + }, + { + "epoch": 4.83, + "learning_rate": 2.586956521739131e-06, + "loss": 0.5302, + "step": 22139 + }, + { + "epoch": 4.83, + "learning_rate": 2.5868472798776494e-06, + "loss": 0.5253, + "step": 22140 + }, + { + "epoch": 4.83, + "learning_rate": 2.5867380380161676e-06, + "loss": 0.6432, + "step": 22141 + }, + { + "epoch": 4.83, + "learning_rate": 2.5866287961546866e-06, + "loss": 1.2527, + "step": 22142 + }, + { + "epoch": 4.83, + "learning_rate": 2.5865195542932052e-06, + "loss": 1.7442, + "step": 22143 + }, + { + "epoch": 4.83, + "learning_rate": 2.5864103124317243e-06, + "loss": 2.8767, + "step": 22144 + }, + { + "epoch": 4.83, + "learning_rate": 2.586301070570243e-06, + "loss": 1.9062, + "step": 22145 + }, + { + "epoch": 4.83, + "learning_rate": 2.586191828708761e-06, + "loss": 1.4272, + "step": 22146 + }, + { + "epoch": 4.83, + "learning_rate": 2.58608258684728e-06, + "loss": 1.3014, + "step": 22147 + }, + { + "epoch": 4.83, + "learning_rate": 2.5859733449857987e-06, + "loss": 2.0739, + "step": 22148 + }, + { + "epoch": 4.83, + "learning_rate": 2.5858641031243178e-06, + "loss": 1.3402, + "step": 22149 + }, + { + "epoch": 4.83, + "learning_rate": 2.5857548612628364e-06, + "loss": 0.5255, + "step": 22150 + }, + { + "epoch": 4.83, + "learning_rate": 2.5856456194013546e-06, + "loss": 3.7334, + "step": 22151 + }, + { + "epoch": 4.83, + "learning_rate": 2.5855363775398736e-06, + "loss": 0.6004, + "step": 22152 + }, + { + "epoch": 4.83, + "learning_rate": 2.585427135678392e-06, + "loss": 0.6517, + "step": 22153 + }, + { + "epoch": 4.84, + "learning_rate": 2.585317893816911e-06, + "loss": 2.2221, + "step": 22154 + }, + { + "epoch": 4.84, + "learning_rate": 2.58520865195543e-06, + "loss": 1.0951, + "step": 22155 + }, + { + "epoch": 4.84, + "learning_rate": 2.585099410093948e-06, + "loss": 2.3657, + "step": 22156 + }, + { + "epoch": 4.84, + "learning_rate": 2.5849901682324666e-06, + "loss": 3.4605, + "step": 22157 + }, + { + "epoch": 4.84, + "learning_rate": 2.5848809263709857e-06, + "loss": 1.3565, + "step": 22158 + }, + { + "epoch": 4.84, + "learning_rate": 2.5847716845095043e-06, + "loss": 1.3136, + "step": 22159 + }, + { + "epoch": 4.84, + "learning_rate": 2.5846624426480225e-06, + "loss": 2.2465, + "step": 22160 + }, + { + "epoch": 4.84, + "learning_rate": 2.5845532007865415e-06, + "loss": 2.7258, + "step": 22161 + }, + { + "epoch": 4.84, + "learning_rate": 2.58444395892506e-06, + "loss": 2.0017, + "step": 22162 + }, + { + "epoch": 4.84, + "learning_rate": 2.584334717063579e-06, + "loss": 2.139, + "step": 22163 + }, + { + "epoch": 4.84, + "learning_rate": 2.5842254752020978e-06, + "loss": 1.2511, + "step": 22164 + }, + { + "epoch": 4.84, + "learning_rate": 2.584116233340616e-06, + "loss": 1.7421, + "step": 22165 + }, + { + "epoch": 4.84, + "learning_rate": 2.584006991479135e-06, + "loss": 1.7007, + "step": 22166 + }, + { + "epoch": 4.84, + "learning_rate": 2.5838977496176536e-06, + "loss": 2.1258, + "step": 22167 + }, + { + "epoch": 4.84, + "learning_rate": 2.5837885077561726e-06, + "loss": 3.2621, + "step": 22168 + }, + { + "epoch": 4.84, + "learning_rate": 2.5836792658946912e-06, + "loss": 1.2184, + "step": 22169 + }, + { + "epoch": 4.84, + "learning_rate": 2.5835700240332094e-06, + "loss": 1.8742, + "step": 22170 + }, + { + "epoch": 4.84, + "learning_rate": 2.5834607821717285e-06, + "loss": 1.621, + "step": 22171 + }, + { + "epoch": 4.84, + "learning_rate": 2.583351540310247e-06, + "loss": 1.3242, + "step": 22172 + }, + { + "epoch": 4.84, + "learning_rate": 2.583242298448766e-06, + "loss": 1.8814, + "step": 22173 + }, + { + "epoch": 4.84, + "learning_rate": 2.5831330565872843e-06, + "loss": 1.491, + "step": 22174 + }, + { + "epoch": 4.84, + "learning_rate": 2.583023814725803e-06, + "loss": 1.8225, + "step": 22175 + }, + { + "epoch": 4.84, + "learning_rate": 2.582914572864322e-06, + "loss": 1.5071, + "step": 22176 + }, + { + "epoch": 4.84, + "learning_rate": 2.5828053310028406e-06, + "loss": 1.4855, + "step": 22177 + }, + { + "epoch": 4.84, + "learning_rate": 2.5826960891413596e-06, + "loss": 1.8383, + "step": 22178 + }, + { + "epoch": 4.84, + "learning_rate": 2.5825868472798778e-06, + "loss": 1.6808, + "step": 22179 + }, + { + "epoch": 4.84, + "learning_rate": 2.5824776054183964e-06, + "loss": 0.7881, + "step": 22180 + }, + { + "epoch": 4.84, + "learning_rate": 2.5823683635569154e-06, + "loss": 1.5093, + "step": 22181 + }, + { + "epoch": 4.84, + "learning_rate": 2.582259121695434e-06, + "loss": 1.4059, + "step": 22182 + }, + { + "epoch": 4.84, + "learning_rate": 2.5821498798339526e-06, + "loss": 2.8981, + "step": 22183 + }, + { + "epoch": 4.84, + "learning_rate": 2.5820406379724713e-06, + "loss": 1.8167, + "step": 22184 + }, + { + "epoch": 4.84, + "learning_rate": 2.58193139611099e-06, + "loss": 1.5198, + "step": 22185 + }, + { + "epoch": 4.84, + "learning_rate": 2.5818221542495085e-06, + "loss": 0.9675, + "step": 22186 + }, + { + "epoch": 4.84, + "learning_rate": 2.5817129123880275e-06, + "loss": 0.8922, + "step": 22187 + }, + { + "epoch": 4.84, + "learning_rate": 2.581603670526546e-06, + "loss": 1.7852, + "step": 22188 + }, + { + "epoch": 4.84, + "learning_rate": 2.5814944286650643e-06, + "loss": 2.7663, + "step": 22189 + }, + { + "epoch": 4.84, + "learning_rate": 2.5813851868035833e-06, + "loss": 1.2711, + "step": 22190 + }, + { + "epoch": 4.84, + "learning_rate": 2.581275944942102e-06, + "loss": 1.5824, + "step": 22191 + }, + { + "epoch": 4.84, + "learning_rate": 2.581166703080621e-06, + "loss": 1.2258, + "step": 22192 + }, + { + "epoch": 4.84, + "learning_rate": 2.581057461219139e-06, + "loss": 3.3478, + "step": 22193 + }, + { + "epoch": 4.84, + "learning_rate": 2.580948219357658e-06, + "loss": 0.9039, + "step": 22194 + }, + { + "epoch": 4.84, + "learning_rate": 2.580838977496177e-06, + "loss": 0.9559, + "step": 22195 + }, + { + "epoch": 4.84, + "learning_rate": 2.5807297356346954e-06, + "loss": 1.0717, + "step": 22196 + }, + { + "epoch": 4.84, + "learning_rate": 2.5806204937732145e-06, + "loss": 1.3416, + "step": 22197 + }, + { + "epoch": 4.84, + "learning_rate": 2.5805112519117327e-06, + "loss": 0.8673, + "step": 22198 + }, + { + "epoch": 4.84, + "learning_rate": 2.5804020100502513e-06, + "loss": 1.5269, + "step": 22199 + }, + { + "epoch": 4.85, + "learning_rate": 2.5802927681887703e-06, + "loss": 2.069, + "step": 22200 + }, + { + "epoch": 4.85, + "learning_rate": 2.580183526327289e-06, + "loss": 0.9717, + "step": 22201 + }, + { + "epoch": 4.85, + "learning_rate": 2.580074284465808e-06, + "loss": 3.0597, + "step": 22202 + }, + { + "epoch": 4.85, + "learning_rate": 2.579965042604326e-06, + "loss": 1.7546, + "step": 22203 + }, + { + "epoch": 4.85, + "learning_rate": 2.5798558007428448e-06, + "loss": 2.9646, + "step": 22204 + }, + { + "epoch": 4.85, + "learning_rate": 2.5797465588813638e-06, + "loss": 1.2419, + "step": 22205 + }, + { + "epoch": 4.85, + "learning_rate": 2.5796373170198824e-06, + "loss": 0.4983, + "step": 22206 + }, + { + "epoch": 4.85, + "learning_rate": 2.5795280751584006e-06, + "loss": 0.7121, + "step": 22207 + }, + { + "epoch": 4.85, + "learning_rate": 2.5794188332969196e-06, + "loss": 1.1375, + "step": 22208 + }, + { + "epoch": 4.85, + "learning_rate": 2.5793095914354382e-06, + "loss": 1.2665, + "step": 22209 + }, + { + "epoch": 4.85, + "learning_rate": 2.579200349573957e-06, + "loss": 2.216, + "step": 22210 + }, + { + "epoch": 4.85, + "learning_rate": 2.579091107712476e-06, + "loss": 1.0684, + "step": 22211 + }, + { + "epoch": 4.85, + "learning_rate": 2.578981865850994e-06, + "loss": 1.288, + "step": 22212 + }, + { + "epoch": 4.85, + "learning_rate": 2.5788726239895127e-06, + "loss": 1.3809, + "step": 22213 + }, + { + "epoch": 4.85, + "learning_rate": 2.5787633821280317e-06, + "loss": 2.7478, + "step": 22214 + }, + { + "epoch": 4.85, + "learning_rate": 2.5786541402665503e-06, + "loss": 1.7513, + "step": 22215 + }, + { + "epoch": 4.85, + "learning_rate": 2.5785448984050694e-06, + "loss": 1.3769, + "step": 22216 + }, + { + "epoch": 4.85, + "learning_rate": 2.5784356565435875e-06, + "loss": 2.2119, + "step": 22217 + }, + { + "epoch": 4.85, + "learning_rate": 2.578326414682106e-06, + "loss": 1.2807, + "step": 22218 + }, + { + "epoch": 4.85, + "learning_rate": 2.578217172820625e-06, + "loss": 1.0754, + "step": 22219 + }, + { + "epoch": 4.85, + "learning_rate": 2.578107930959144e-06, + "loss": 1.0101, + "step": 22220 + }, + { + "epoch": 4.85, + "learning_rate": 2.577998689097663e-06, + "loss": 0.8264, + "step": 22221 + }, + { + "epoch": 4.85, + "learning_rate": 2.577889447236181e-06, + "loss": 2.4997, + "step": 22222 + }, + { + "epoch": 4.85, + "learning_rate": 2.5777802053746996e-06, + "loss": 1.8028, + "step": 22223 + }, + { + "epoch": 4.85, + "learning_rate": 2.5776709635132187e-06, + "loss": 1.234, + "step": 22224 + }, + { + "epoch": 4.85, + "learning_rate": 2.5775617216517373e-06, + "loss": 0.7697, + "step": 22225 + }, + { + "epoch": 4.85, + "learning_rate": 2.5774524797902555e-06, + "loss": 1.1305, + "step": 22226 + }, + { + "epoch": 4.85, + "learning_rate": 2.5773432379287745e-06, + "loss": 1.6401, + "step": 22227 + }, + { + "epoch": 4.85, + "learning_rate": 2.577233996067293e-06, + "loss": 3.035, + "step": 22228 + }, + { + "epoch": 4.85, + "learning_rate": 2.577124754205812e-06, + "loss": 0.1492, + "step": 22229 + }, + { + "epoch": 4.85, + "learning_rate": 2.5770155123443308e-06, + "loss": 1.3405, + "step": 22230 + }, + { + "epoch": 4.85, + "learning_rate": 2.576906270482849e-06, + "loss": 2.6842, + "step": 22231 + }, + { + "epoch": 4.85, + "learning_rate": 2.576797028621368e-06, + "loss": 2.198, + "step": 22232 + }, + { + "epoch": 4.85, + "learning_rate": 2.5766877867598866e-06, + "loss": 2.9564, + "step": 22233 + }, + { + "epoch": 4.85, + "learning_rate": 2.5765785448984056e-06, + "loss": 1.2575, + "step": 22234 + }, + { + "epoch": 4.85, + "learning_rate": 2.5764693030369242e-06, + "loss": 1.9254, + "step": 22235 + }, + { + "epoch": 4.85, + "learning_rate": 2.5763600611754424e-06, + "loss": 2.1157, + "step": 22236 + }, + { + "epoch": 4.85, + "learning_rate": 2.5762508193139615e-06, + "loss": 2.3493, + "step": 22237 + }, + { + "epoch": 4.85, + "learning_rate": 2.57614157745248e-06, + "loss": 0.6897, + "step": 22238 + }, + { + "epoch": 4.85, + "learning_rate": 2.5760323355909987e-06, + "loss": 1.059, + "step": 22239 + }, + { + "epoch": 4.85, + "learning_rate": 2.5759230937295173e-06, + "loss": 1.5584, + "step": 22240 + }, + { + "epoch": 4.85, + "learning_rate": 2.575813851868036e-06, + "loss": 0.9631, + "step": 22241 + }, + { + "epoch": 4.85, + "learning_rate": 2.5757046100065545e-06, + "loss": 0.9617, + "step": 22242 + }, + { + "epoch": 4.85, + "learning_rate": 2.5755953681450735e-06, + "loss": 1.4977, + "step": 22243 + }, + { + "epoch": 4.85, + "learning_rate": 2.575486126283592e-06, + "loss": 1.348, + "step": 22244 + }, + { + "epoch": 4.85, + "learning_rate": 2.5753768844221103e-06, + "loss": 2.1622, + "step": 22245 + }, + { + "epoch": 4.86, + "learning_rate": 2.5752676425606294e-06, + "loss": 0.5758, + "step": 22246 + }, + { + "epoch": 4.86, + "learning_rate": 2.575158400699148e-06, + "loss": 0.9959, + "step": 22247 + }, + { + "epoch": 4.86, + "learning_rate": 2.575049158837667e-06, + "loss": 1.3927, + "step": 22248 + }, + { + "epoch": 4.86, + "learning_rate": 2.5749399169761856e-06, + "loss": 1.3309, + "step": 22249 + }, + { + "epoch": 4.86, + "learning_rate": 2.574830675114704e-06, + "loss": 0.7014, + "step": 22250 + }, + { + "epoch": 4.86, + "learning_rate": 2.574721433253223e-06, + "loss": 1.1892, + "step": 22251 + }, + { + "epoch": 4.86, + "learning_rate": 2.5746121913917415e-06, + "loss": 2.9716, + "step": 22252 + }, + { + "epoch": 4.86, + "learning_rate": 2.5745029495302605e-06, + "loss": 1.7549, + "step": 22253 + }, + { + "epoch": 4.86, + "learning_rate": 2.574393707668779e-06, + "loss": 2.0618, + "step": 22254 + }, + { + "epoch": 4.86, + "learning_rate": 2.5742844658072973e-06, + "loss": 1.1424, + "step": 22255 + }, + { + "epoch": 4.86, + "learning_rate": 2.5741752239458163e-06, + "loss": 1.5959, + "step": 22256 + }, + { + "epoch": 4.86, + "learning_rate": 2.574065982084335e-06, + "loss": 1.1751, + "step": 22257 + }, + { + "epoch": 4.86, + "learning_rate": 2.573956740222854e-06, + "loss": 1.7809, + "step": 22258 + }, + { + "epoch": 4.86, + "learning_rate": 2.573847498361372e-06, + "loss": 2.2199, + "step": 22259 + }, + { + "epoch": 4.86, + "learning_rate": 2.5737382564998908e-06, + "loss": 1.6074, + "step": 22260 + }, + { + "epoch": 4.86, + "learning_rate": 2.57362901463841e-06, + "loss": 1.4302, + "step": 22261 + }, + { + "epoch": 4.86, + "learning_rate": 2.5735197727769284e-06, + "loss": 2.1233, + "step": 22262 + }, + { + "epoch": 4.86, + "learning_rate": 2.573410530915447e-06, + "loss": 2.3999, + "step": 22263 + }, + { + "epoch": 4.86, + "learning_rate": 2.5733012890539656e-06, + "loss": 1.9876, + "step": 22264 + }, + { + "epoch": 4.86, + "learning_rate": 2.5731920471924843e-06, + "loss": 1.2937, + "step": 22265 + }, + { + "epoch": 4.86, + "learning_rate": 2.573082805331003e-06, + "loss": 0.9384, + "step": 22266 + }, + { + "epoch": 4.86, + "learning_rate": 2.572973563469522e-06, + "loss": 2.279, + "step": 22267 + }, + { + "epoch": 4.86, + "learning_rate": 2.5728643216080405e-06, + "loss": 2.5495, + "step": 22268 + }, + { + "epoch": 4.86, + "learning_rate": 2.572755079746559e-06, + "loss": 1.9507, + "step": 22269 + }, + { + "epoch": 4.86, + "learning_rate": 2.5726458378850777e-06, + "loss": 2.8498, + "step": 22270 + }, + { + "epoch": 4.86, + "learning_rate": 2.5725365960235963e-06, + "loss": 1.0866, + "step": 22271 + }, + { + "epoch": 4.86, + "learning_rate": 2.5724273541621154e-06, + "loss": 0.2984, + "step": 22272 + }, + { + "epoch": 4.86, + "learning_rate": 2.5723181123006336e-06, + "loss": 0.7372, + "step": 22273 + }, + { + "epoch": 4.86, + "learning_rate": 2.572208870439152e-06, + "loss": 1.9056, + "step": 22274 + }, + { + "epoch": 4.86, + "learning_rate": 2.5720996285776712e-06, + "loss": 2.5501, + "step": 22275 + }, + { + "epoch": 4.86, + "learning_rate": 2.57199038671619e-06, + "loss": 1.6723, + "step": 22276 + }, + { + "epoch": 4.86, + "learning_rate": 2.571881144854709e-06, + "loss": 1.9019, + "step": 22277 + }, + { + "epoch": 4.86, + "learning_rate": 2.571771902993227e-06, + "loss": 1.7256, + "step": 22278 + }, + { + "epoch": 4.86, + "learning_rate": 2.5716626611317457e-06, + "loss": 1.3939, + "step": 22279 + }, + { + "epoch": 4.86, + "learning_rate": 2.5715534192702647e-06, + "loss": 1.856, + "step": 22280 + }, + { + "epoch": 4.86, + "learning_rate": 2.5714441774087833e-06, + "loss": 1.8005, + "step": 22281 + }, + { + "epoch": 4.86, + "learning_rate": 2.5713349355473023e-06, + "loss": 1.2123, + "step": 22282 + }, + { + "epoch": 4.86, + "learning_rate": 2.5712256936858205e-06, + "loss": 1.6379, + "step": 22283 + }, + { + "epoch": 4.86, + "learning_rate": 2.571116451824339e-06, + "loss": 1.2425, + "step": 22284 + }, + { + "epoch": 4.86, + "learning_rate": 2.571007209962858e-06, + "loss": 4.5147, + "step": 22285 + }, + { + "epoch": 4.86, + "learning_rate": 2.5708979681013768e-06, + "loss": 1.1598, + "step": 22286 + }, + { + "epoch": 4.86, + "learning_rate": 2.570788726239896e-06, + "loss": 0.7504, + "step": 22287 + }, + { + "epoch": 4.86, + "learning_rate": 2.570679484378414e-06, + "loss": 0.8384, + "step": 22288 + }, + { + "epoch": 4.86, + "learning_rate": 2.5705702425169326e-06, + "loss": 0.3131, + "step": 22289 + }, + { + "epoch": 4.86, + "learning_rate": 2.5704610006554516e-06, + "loss": 1.1116, + "step": 22290 + }, + { + "epoch": 4.86, + "learning_rate": 2.5703517587939703e-06, + "loss": 1.297, + "step": 22291 + }, + { + "epoch": 4.87, + "learning_rate": 2.5702425169324884e-06, + "loss": 2.2126, + "step": 22292 + }, + { + "epoch": 4.87, + "learning_rate": 2.5701332750710075e-06, + "loss": 0.3247, + "step": 22293 + }, + { + "epoch": 4.87, + "learning_rate": 2.570024033209526e-06, + "loss": 0.6415, + "step": 22294 + }, + { + "epoch": 4.87, + "learning_rate": 2.5699147913480447e-06, + "loss": 1.5797, + "step": 22295 + }, + { + "epoch": 4.87, + "learning_rate": 2.5698055494865637e-06, + "loss": 1.0221, + "step": 22296 + }, + { + "epoch": 4.87, + "learning_rate": 2.569696307625082e-06, + "loss": 1.9558, + "step": 22297 + }, + { + "epoch": 4.87, + "learning_rate": 2.5695870657636005e-06, + "loss": 1.9256, + "step": 22298 + }, + { + "epoch": 4.87, + "learning_rate": 2.5694778239021196e-06, + "loss": 3.171, + "step": 22299 + }, + { + "epoch": 4.87, + "learning_rate": 2.569368582040638e-06, + "loss": 1.3692, + "step": 22300 + }, + { + "epoch": 4.87, + "learning_rate": 2.5692593401791572e-06, + "loss": 1.1142, + "step": 22301 + }, + { + "epoch": 4.87, + "learning_rate": 2.5691500983176754e-06, + "loss": 1.7992, + "step": 22302 + }, + { + "epoch": 4.87, + "learning_rate": 2.569040856456194e-06, + "loss": 2.2101, + "step": 22303 + }, + { + "epoch": 4.87, + "learning_rate": 2.568931614594713e-06, + "loss": 2.5633, + "step": 22304 + }, + { + "epoch": 4.87, + "learning_rate": 2.5688223727332317e-06, + "loss": 1.6565, + "step": 22305 + }, + { + "epoch": 4.87, + "learning_rate": 2.56871313087175e-06, + "loss": 2.3781, + "step": 22306 + }, + { + "epoch": 4.87, + "learning_rate": 2.568603889010269e-06, + "loss": 1.5609, + "step": 22307 + }, + { + "epoch": 4.87, + "learning_rate": 2.5684946471487875e-06, + "loss": 0.5378, + "step": 22308 + }, + { + "epoch": 4.87, + "learning_rate": 2.5683854052873065e-06, + "loss": 1.2641, + "step": 22309 + }, + { + "epoch": 4.87, + "learning_rate": 2.568276163425825e-06, + "loss": 1.9628, + "step": 22310 + }, + { + "epoch": 4.87, + "learning_rate": 2.5681669215643433e-06, + "loss": 1.467, + "step": 22311 + }, + { + "epoch": 4.87, + "learning_rate": 2.5680576797028624e-06, + "loss": 0.1358, + "step": 22312 + }, + { + "epoch": 4.87, + "learning_rate": 2.567948437841381e-06, + "loss": 1.6336, + "step": 22313 + }, + { + "epoch": 4.87, + "learning_rate": 2.5678391959799e-06, + "loss": 2.0998, + "step": 22314 + }, + { + "epoch": 4.87, + "learning_rate": 2.5677299541184186e-06, + "loss": 2.5045, + "step": 22315 + }, + { + "epoch": 4.87, + "learning_rate": 2.567620712256937e-06, + "loss": 1.3129, + "step": 22316 + }, + { + "epoch": 4.87, + "learning_rate": 2.567511470395456e-06, + "loss": 1.4517, + "step": 22317 + }, + { + "epoch": 4.87, + "learning_rate": 2.5674022285339744e-06, + "loss": 0.4367, + "step": 22318 + }, + { + "epoch": 4.87, + "learning_rate": 2.567292986672493e-06, + "loss": 1.892, + "step": 22319 + }, + { + "epoch": 4.87, + "learning_rate": 2.5671837448110117e-06, + "loss": 0.8546, + "step": 22320 + }, + { + "epoch": 4.87, + "learning_rate": 2.5670745029495303e-06, + "loss": 1.283, + "step": 22321 + }, + { + "epoch": 4.87, + "learning_rate": 2.5669652610880493e-06, + "loss": 1.3702, + "step": 22322 + }, + { + "epoch": 4.87, + "learning_rate": 2.566856019226568e-06, + "loss": 2.1841, + "step": 22323 + }, + { + "epoch": 4.87, + "learning_rate": 2.5667467773650865e-06, + "loss": 0.7229, + "step": 22324 + }, + { + "epoch": 4.87, + "learning_rate": 2.566637535503605e-06, + "loss": 0.9502, + "step": 22325 + }, + { + "epoch": 4.87, + "learning_rate": 2.5665282936421238e-06, + "loss": 1.0972, + "step": 22326 + }, + { + "epoch": 4.87, + "learning_rate": 2.5664190517806424e-06, + "loss": 1.5764, + "step": 22327 + }, + { + "epoch": 4.87, + "learning_rate": 2.5663098099191614e-06, + "loss": 2.1627, + "step": 22328 + }, + { + "epoch": 4.87, + "learning_rate": 2.56620056805768e-06, + "loss": 1.0497, + "step": 22329 + }, + { + "epoch": 4.87, + "learning_rate": 2.566091326196198e-06, + "loss": 1.4698, + "step": 22330 + }, + { + "epoch": 4.87, + "learning_rate": 2.5659820843347172e-06, + "loss": 1.1461, + "step": 22331 + }, + { + "epoch": 4.87, + "learning_rate": 2.565872842473236e-06, + "loss": 2.091, + "step": 22332 + }, + { + "epoch": 4.87, + "learning_rate": 2.565763600611755e-06, + "loss": 3.1237, + "step": 22333 + }, + { + "epoch": 4.87, + "learning_rate": 2.5656543587502735e-06, + "loss": 1.0438, + "step": 22334 + }, + { + "epoch": 4.87, + "learning_rate": 2.5655451168887917e-06, + "loss": 2.2474, + "step": 22335 + }, + { + "epoch": 4.87, + "learning_rate": 2.5654358750273107e-06, + "loss": 2.0376, + "step": 22336 + }, + { + "epoch": 4.87, + "learning_rate": 2.5653266331658293e-06, + "loss": 2.2515, + "step": 22337 + }, + { + "epoch": 4.88, + "learning_rate": 2.5652173913043484e-06, + "loss": 1.7895, + "step": 22338 + }, + { + "epoch": 4.88, + "learning_rate": 2.5651081494428666e-06, + "loss": 1.1337, + "step": 22339 + }, + { + "epoch": 4.88, + "learning_rate": 2.564998907581385e-06, + "loss": 1.7569, + "step": 22340 + }, + { + "epoch": 4.88, + "learning_rate": 2.564889665719904e-06, + "loss": 1.9069, + "step": 22341 + }, + { + "epoch": 4.88, + "learning_rate": 2.564780423858423e-06, + "loss": 0.3986, + "step": 22342 + }, + { + "epoch": 4.88, + "learning_rate": 2.564671181996942e-06, + "loss": 0.6625, + "step": 22343 + }, + { + "epoch": 4.88, + "learning_rate": 2.56456194013546e-06, + "loss": 0.9161, + "step": 22344 + }, + { + "epoch": 4.88, + "learning_rate": 2.5644526982739786e-06, + "loss": 2.37, + "step": 22345 + }, + { + "epoch": 4.88, + "learning_rate": 2.5643434564124977e-06, + "loss": 1.3914, + "step": 22346 + }, + { + "epoch": 4.88, + "learning_rate": 2.5642342145510163e-06, + "loss": 1.8008, + "step": 22347 + }, + { + "epoch": 4.88, + "learning_rate": 2.564124972689535e-06, + "loss": 0.4601, + "step": 22348 + }, + { + "epoch": 4.88, + "learning_rate": 2.5640157308280535e-06, + "loss": 1.3219, + "step": 22349 + }, + { + "epoch": 4.88, + "learning_rate": 2.563906488966572e-06, + "loss": 2.2828, + "step": 22350 + }, + { + "epoch": 4.88, + "learning_rate": 2.5637972471050907e-06, + "loss": 1.3961, + "step": 22351 + }, + { + "epoch": 4.88, + "learning_rate": 2.5636880052436098e-06, + "loss": 1.0696, + "step": 22352 + }, + { + "epoch": 4.88, + "learning_rate": 2.563578763382128e-06, + "loss": 0.6325, + "step": 22353 + }, + { + "epoch": 4.88, + "learning_rate": 2.5634695215206466e-06, + "loss": 1.8248, + "step": 22354 + }, + { + "epoch": 4.88, + "learning_rate": 2.5633602796591656e-06, + "loss": 1.1849, + "step": 22355 + }, + { + "epoch": 4.88, + "learning_rate": 2.563251037797684e-06, + "loss": 2.3991, + "step": 22356 + }, + { + "epoch": 4.88, + "learning_rate": 2.5631417959362032e-06, + "loss": 0.9497, + "step": 22357 + }, + { + "epoch": 4.88, + "learning_rate": 2.5630325540747214e-06, + "loss": 2.6253, + "step": 22358 + }, + { + "epoch": 4.88, + "learning_rate": 2.56292331221324e-06, + "loss": 1.9683, + "step": 22359 + }, + { + "epoch": 4.88, + "learning_rate": 2.562814070351759e-06, + "loss": 2.1138, + "step": 22360 + }, + { + "epoch": 4.88, + "learning_rate": 2.5627048284902777e-06, + "loss": 1.1839, + "step": 22361 + }, + { + "epoch": 4.88, + "learning_rate": 2.5625955866287967e-06, + "loss": 1.0397, + "step": 22362 + }, + { + "epoch": 4.88, + "learning_rate": 2.562486344767315e-06, + "loss": 0.6882, + "step": 22363 + }, + { + "epoch": 4.88, + "learning_rate": 2.5623771029058335e-06, + "loss": 1.4905, + "step": 22364 + }, + { + "epoch": 4.88, + "learning_rate": 2.5622678610443526e-06, + "loss": 1.3268, + "step": 22365 + }, + { + "epoch": 4.88, + "learning_rate": 2.562158619182871e-06, + "loss": 1.8724, + "step": 22366 + }, + { + "epoch": 4.88, + "learning_rate": 2.56204937732139e-06, + "loss": 2.0248, + "step": 22367 + }, + { + "epoch": 4.88, + "learning_rate": 2.5619401354599084e-06, + "loss": 0.6603, + "step": 22368 + }, + { + "epoch": 4.88, + "learning_rate": 2.561830893598427e-06, + "loss": 1.2608, + "step": 22369 + }, + { + "epoch": 4.88, + "learning_rate": 2.561721651736946e-06, + "loss": 3.5535, + "step": 22370 + }, + { + "epoch": 4.88, + "learning_rate": 2.5616124098754646e-06, + "loss": 1.6528, + "step": 22371 + }, + { + "epoch": 4.88, + "learning_rate": 2.561503168013983e-06, + "loss": 0.7606, + "step": 22372 + }, + { + "epoch": 4.88, + "learning_rate": 2.561393926152502e-06, + "loss": 1.2766, + "step": 22373 + }, + { + "epoch": 4.88, + "learning_rate": 2.5612846842910205e-06, + "loss": 1.9229, + "step": 22374 + }, + { + "epoch": 4.88, + "learning_rate": 2.5611754424295395e-06, + "loss": 1.3953, + "step": 22375 + }, + { + "epoch": 4.88, + "learning_rate": 2.561066200568058e-06, + "loss": 1.6973, + "step": 22376 + }, + { + "epoch": 4.88, + "learning_rate": 2.5609569587065763e-06, + "loss": 3.5463, + "step": 22377 + }, + { + "epoch": 4.88, + "learning_rate": 2.5608477168450953e-06, + "loss": 2.2824, + "step": 22378 + }, + { + "epoch": 4.88, + "learning_rate": 2.560738474983614e-06, + "loss": 0.7639, + "step": 22379 + }, + { + "epoch": 4.88, + "learning_rate": 2.5606292331221326e-06, + "loss": 1.1248, + "step": 22380 + }, + { + "epoch": 4.88, + "learning_rate": 2.5605199912606516e-06, + "loss": 2.2533, + "step": 22381 + }, + { + "epoch": 4.88, + "learning_rate": 2.56041074939917e-06, + "loss": 2.2731, + "step": 22382 + }, + { + "epoch": 4.88, + "learning_rate": 2.5603015075376884e-06, + "loss": 1.6807, + "step": 22383 + }, + { + "epoch": 4.89, + "learning_rate": 2.5601922656762074e-06, + "loss": 0.6722, + "step": 22384 + }, + { + "epoch": 4.89, + "learning_rate": 2.560083023814726e-06, + "loss": 0.7067, + "step": 22385 + }, + { + "epoch": 4.89, + "learning_rate": 2.5599737819532442e-06, + "loss": 0.5843, + "step": 22386 + }, + { + "epoch": 4.89, + "learning_rate": 2.5598645400917633e-06, + "loss": 1.9842, + "step": 22387 + }, + { + "epoch": 4.89, + "learning_rate": 2.559755298230282e-06, + "loss": 0.5234, + "step": 22388 + }, + { + "epoch": 4.89, + "learning_rate": 2.559646056368801e-06, + "loss": 1.5501, + "step": 22389 + }, + { + "epoch": 4.89, + "learning_rate": 2.5595368145073195e-06, + "loss": 2.4225, + "step": 22390 + }, + { + "epoch": 4.89, + "learning_rate": 2.5594275726458377e-06, + "loss": 1.1837, + "step": 22391 + }, + { + "epoch": 4.89, + "learning_rate": 2.5593183307843567e-06, + "loss": 1.9618, + "step": 22392 + }, + { + "epoch": 4.89, + "learning_rate": 2.5592090889228754e-06, + "loss": 1.2221, + "step": 22393 + }, + { + "epoch": 4.89, + "learning_rate": 2.5590998470613944e-06, + "loss": 3.4692, + "step": 22394 + }, + { + "epoch": 4.89, + "learning_rate": 2.558990605199913e-06, + "loss": 1.2836, + "step": 22395 + }, + { + "epoch": 4.89, + "learning_rate": 2.558881363338431e-06, + "loss": 2.0758, + "step": 22396 + }, + { + "epoch": 4.89, + "learning_rate": 2.5587721214769502e-06, + "loss": 0.8692, + "step": 22397 + }, + { + "epoch": 4.89, + "learning_rate": 2.558662879615469e-06, + "loss": 2.7571, + "step": 22398 + }, + { + "epoch": 4.89, + "learning_rate": 2.558553637753988e-06, + "loss": 1.4985, + "step": 22399 + }, + { + "epoch": 4.89, + "learning_rate": 2.5584443958925065e-06, + "loss": 2.4659, + "step": 22400 + }, + { + "epoch": 4.89, + "learning_rate": 2.5583351540310247e-06, + "loss": 1.8607, + "step": 22401 + }, + { + "epoch": 4.89, + "learning_rate": 2.5582259121695437e-06, + "loss": 1.7827, + "step": 22402 + }, + { + "epoch": 4.89, + "learning_rate": 2.5581166703080623e-06, + "loss": 1.5002, + "step": 22403 + }, + { + "epoch": 4.89, + "learning_rate": 2.558007428446581e-06, + "loss": 1.4527, + "step": 22404 + }, + { + "epoch": 4.89, + "learning_rate": 2.5578981865850995e-06, + "loss": 1.5688, + "step": 22405 + }, + { + "epoch": 4.89, + "learning_rate": 2.557788944723618e-06, + "loss": 0.5874, + "step": 22406 + }, + { + "epoch": 4.89, + "learning_rate": 2.5576797028621368e-06, + "loss": 1.1465, + "step": 22407 + }, + { + "epoch": 4.89, + "learning_rate": 2.557570461000656e-06, + "loss": 0.523, + "step": 22408 + }, + { + "epoch": 4.89, + "learning_rate": 2.5574612191391744e-06, + "loss": 0.9659, + "step": 22409 + }, + { + "epoch": 4.89, + "learning_rate": 2.5573519772776926e-06, + "loss": 1.2898, + "step": 22410 + }, + { + "epoch": 4.89, + "learning_rate": 2.5572427354162116e-06, + "loss": 2.7215, + "step": 22411 + }, + { + "epoch": 4.89, + "learning_rate": 2.5571334935547302e-06, + "loss": 1.2394, + "step": 22412 + }, + { + "epoch": 4.89, + "learning_rate": 2.5570242516932493e-06, + "loss": 1.3923, + "step": 22413 + }, + { + "epoch": 4.89, + "learning_rate": 2.556915009831768e-06, + "loss": 0.495, + "step": 22414 + }, + { + "epoch": 4.89, + "learning_rate": 2.556805767970286e-06, + "loss": 2.0901, + "step": 22415 + }, + { + "epoch": 4.89, + "learning_rate": 2.556696526108805e-06, + "loss": 1.3163, + "step": 22416 + }, + { + "epoch": 4.89, + "learning_rate": 2.5565872842473237e-06, + "loss": 1.9382, + "step": 22417 + }, + { + "epoch": 4.89, + "learning_rate": 2.5564780423858427e-06, + "loss": 2.0664, + "step": 22418 + }, + { + "epoch": 4.89, + "learning_rate": 2.556368800524361e-06, + "loss": 1.4797, + "step": 22419 + }, + { + "epoch": 4.89, + "learning_rate": 2.5562595586628795e-06, + "loss": 1.703, + "step": 22420 + }, + { + "epoch": 4.89, + "learning_rate": 2.5561503168013986e-06, + "loss": 1.6708, + "step": 22421 + }, + { + "epoch": 4.89, + "learning_rate": 2.556041074939917e-06, + "loss": 0.5075, + "step": 22422 + }, + { + "epoch": 4.89, + "learning_rate": 2.5559318330784362e-06, + "loss": 2.272, + "step": 22423 + }, + { + "epoch": 4.89, + "learning_rate": 2.5558225912169544e-06, + "loss": 2.5021, + "step": 22424 + }, + { + "epoch": 4.89, + "learning_rate": 2.555713349355473e-06, + "loss": 1.1827, + "step": 22425 + }, + { + "epoch": 4.89, + "learning_rate": 2.555604107493992e-06, + "loss": 0.3004, + "step": 22426 + }, + { + "epoch": 4.89, + "learning_rate": 2.5554948656325107e-06, + "loss": 1.6133, + "step": 22427 + }, + { + "epoch": 4.89, + "learning_rate": 2.5553856237710297e-06, + "loss": 1.9152, + "step": 22428 + }, + { + "epoch": 4.9, + "learning_rate": 2.555276381909548e-06, + "loss": 2.9995, + "step": 22429 + }, + { + "epoch": 4.9, + "learning_rate": 2.5551671400480665e-06, + "loss": 1.6902, + "step": 22430 + }, + { + "epoch": 4.9, + "learning_rate": 2.5550578981865855e-06, + "loss": 1.5958, + "step": 22431 + }, + { + "epoch": 4.9, + "learning_rate": 2.554948656325104e-06, + "loss": 1.0566, + "step": 22432 + }, + { + "epoch": 4.9, + "learning_rate": 2.5548394144636228e-06, + "loss": 1.3874, + "step": 22433 + }, + { + "epoch": 4.9, + "learning_rate": 2.5547301726021414e-06, + "loss": 0.9641, + "step": 22434 + }, + { + "epoch": 4.9, + "learning_rate": 2.55462093074066e-06, + "loss": 1.7087, + "step": 22435 + }, + { + "epoch": 4.9, + "learning_rate": 2.5545116888791786e-06, + "loss": 1.1964, + "step": 22436 + }, + { + "epoch": 4.9, + "learning_rate": 2.5544024470176976e-06, + "loss": 1.2062, + "step": 22437 + }, + { + "epoch": 4.9, + "learning_rate": 2.554293205156216e-06, + "loss": 0.9938, + "step": 22438 + }, + { + "epoch": 4.9, + "learning_rate": 2.5541839632947344e-06, + "loss": 0.4266, + "step": 22439 + }, + { + "epoch": 4.9, + "learning_rate": 2.5540747214332535e-06, + "loss": 1.7046, + "step": 22440 + }, + { + "epoch": 4.9, + "learning_rate": 2.553965479571772e-06, + "loss": 3.0228, + "step": 22441 + }, + { + "epoch": 4.9, + "learning_rate": 2.553856237710291e-06, + "loss": 0.7743, + "step": 22442 + }, + { + "epoch": 4.9, + "learning_rate": 2.5537469958488093e-06, + "loss": 1.0518, + "step": 22443 + }, + { + "epoch": 4.9, + "learning_rate": 2.553637753987328e-06, + "loss": 1.0607, + "step": 22444 + }, + { + "epoch": 4.9, + "learning_rate": 2.553528512125847e-06, + "loss": 1.2308, + "step": 22445 + }, + { + "epoch": 4.9, + "learning_rate": 2.5534192702643656e-06, + "loss": 0.9527, + "step": 22446 + }, + { + "epoch": 4.9, + "learning_rate": 2.5533100284028846e-06, + "loss": 0.6106, + "step": 22447 + }, + { + "epoch": 4.9, + "learning_rate": 2.5532007865414028e-06, + "loss": 1.0069, + "step": 22448 + }, + { + "epoch": 4.9, + "learning_rate": 2.5530915446799214e-06, + "loss": 1.1233, + "step": 22449 + }, + { + "epoch": 4.9, + "learning_rate": 2.5529823028184404e-06, + "loss": 1.0038, + "step": 22450 + }, + { + "epoch": 4.9, + "learning_rate": 2.552873060956959e-06, + "loss": 2.206, + "step": 22451 + }, + { + "epoch": 4.9, + "learning_rate": 2.5527638190954772e-06, + "loss": 2.9227, + "step": 22452 + }, + { + "epoch": 4.9, + "learning_rate": 2.5526545772339963e-06, + "loss": 1.0454, + "step": 22453 + }, + { + "epoch": 4.9, + "learning_rate": 2.552545335372515e-06, + "loss": 1.5262, + "step": 22454 + }, + { + "epoch": 4.9, + "learning_rate": 2.552436093511034e-06, + "loss": 0.5271, + "step": 22455 + }, + { + "epoch": 4.9, + "learning_rate": 2.5523268516495525e-06, + "loss": 2.9514, + "step": 22456 + }, + { + "epoch": 4.9, + "learning_rate": 2.5522176097880707e-06, + "loss": 1.2079, + "step": 22457 + }, + { + "epoch": 4.9, + "learning_rate": 2.5521083679265897e-06, + "loss": 2.7206, + "step": 22458 + }, + { + "epoch": 4.9, + "learning_rate": 2.5519991260651083e-06, + "loss": 1.3321, + "step": 22459 + }, + { + "epoch": 4.9, + "learning_rate": 2.551889884203627e-06, + "loss": 1.4411, + "step": 22460 + }, + { + "epoch": 4.9, + "learning_rate": 2.551780642342146e-06, + "loss": 1.4257, + "step": 22461 + }, + { + "epoch": 4.9, + "learning_rate": 2.551671400480664e-06, + "loss": 3.3759, + "step": 22462 + }, + { + "epoch": 4.9, + "learning_rate": 2.5515621586191828e-06, + "loss": 2.3433, + "step": 22463 + }, + { + "epoch": 4.9, + "learning_rate": 2.551452916757702e-06, + "loss": 4.4618, + "step": 22464 + }, + { + "epoch": 4.9, + "learning_rate": 2.5513436748962204e-06, + "loss": 1.9005, + "step": 22465 + }, + { + "epoch": 4.9, + "learning_rate": 2.5512344330347395e-06, + "loss": 1.0783, + "step": 22466 + }, + { + "epoch": 4.9, + "learning_rate": 2.5511251911732577e-06, + "loss": 2.6981, + "step": 22467 + }, + { + "epoch": 4.9, + "learning_rate": 2.5510159493117763e-06, + "loss": 1.8192, + "step": 22468 + }, + { + "epoch": 4.9, + "learning_rate": 2.5509067074502953e-06, + "loss": 2.4882, + "step": 22469 + }, + { + "epoch": 4.9, + "learning_rate": 2.550797465588814e-06, + "loss": 1.1409, + "step": 22470 + }, + { + "epoch": 4.9, + "learning_rate": 2.550688223727332e-06, + "loss": 1.2647, + "step": 22471 + }, + { + "epoch": 4.9, + "learning_rate": 2.550578981865851e-06, + "loss": 1.4858, + "step": 22472 + }, + { + "epoch": 4.9, + "learning_rate": 2.5504697400043697e-06, + "loss": 2.3932, + "step": 22473 + }, + { + "epoch": 4.9, + "learning_rate": 2.5503604981428888e-06, + "loss": 2.0585, + "step": 22474 + }, + { + "epoch": 4.91, + "learning_rate": 2.5502512562814074e-06, + "loss": 0.7558, + "step": 22475 + }, + { + "epoch": 4.91, + "learning_rate": 2.5501420144199256e-06, + "loss": 1.5182, + "step": 22476 + }, + { + "epoch": 4.91, + "learning_rate": 2.5500327725584446e-06, + "loss": 0.4145, + "step": 22477 + }, + { + "epoch": 4.91, + "learning_rate": 2.5499235306969632e-06, + "loss": 1.7451, + "step": 22478 + }, + { + "epoch": 4.91, + "learning_rate": 2.5498142888354823e-06, + "loss": 2.2265, + "step": 22479 + }, + { + "epoch": 4.91, + "learning_rate": 2.549705046974001e-06, + "loss": 2.5197, + "step": 22480 + }, + { + "epoch": 4.91, + "learning_rate": 2.549595805112519e-06, + "loss": 1.5401, + "step": 22481 + }, + { + "epoch": 4.91, + "learning_rate": 2.549486563251038e-06, + "loss": 3.3355, + "step": 22482 + }, + { + "epoch": 4.91, + "learning_rate": 2.5493773213895567e-06, + "loss": 2.273, + "step": 22483 + }, + { + "epoch": 4.91, + "learning_rate": 2.5492680795280757e-06, + "loss": 0.7532, + "step": 22484 + }, + { + "epoch": 4.91, + "learning_rate": 2.549158837666594e-06, + "loss": 1.1961, + "step": 22485 + }, + { + "epoch": 4.91, + "learning_rate": 2.5490495958051125e-06, + "loss": 2.2575, + "step": 22486 + }, + { + "epoch": 4.91, + "learning_rate": 2.5489403539436316e-06, + "loss": 1.4285, + "step": 22487 + }, + { + "epoch": 4.91, + "learning_rate": 2.54883111208215e-06, + "loss": 0.9038, + "step": 22488 + }, + { + "epoch": 4.91, + "learning_rate": 2.5487218702206688e-06, + "loss": 0.7163, + "step": 22489 + }, + { + "epoch": 4.91, + "learning_rate": 2.5486126283591874e-06, + "loss": 1.2396, + "step": 22490 + }, + { + "epoch": 4.91, + "learning_rate": 2.548503386497706e-06, + "loss": 1.6131, + "step": 22491 + }, + { + "epoch": 4.91, + "learning_rate": 2.5483941446362246e-06, + "loss": 1.2367, + "step": 22492 + }, + { + "epoch": 4.91, + "learning_rate": 2.5482849027747437e-06, + "loss": 1.6393, + "step": 22493 + }, + { + "epoch": 4.91, + "learning_rate": 2.5481756609132623e-06, + "loss": 0.3254, + "step": 22494 + }, + { + "epoch": 4.91, + "learning_rate": 2.5480664190517805e-06, + "loss": 1.4688, + "step": 22495 + }, + { + "epoch": 4.91, + "learning_rate": 2.5479571771902995e-06, + "loss": 1.3804, + "step": 22496 + }, + { + "epoch": 4.91, + "learning_rate": 2.547847935328818e-06, + "loss": 1.6249, + "step": 22497 + }, + { + "epoch": 4.91, + "learning_rate": 2.547738693467337e-06, + "loss": 0.326, + "step": 22498 + }, + { + "epoch": 4.91, + "learning_rate": 2.5476294516058557e-06, + "loss": 1.9303, + "step": 22499 + }, + { + "epoch": 4.91, + "learning_rate": 2.547520209744374e-06, + "loss": 2.6582, + "step": 22500 + }, + { + "epoch": 4.91, + "learning_rate": 2.547410967882893e-06, + "loss": 1.4757, + "step": 22501 + }, + { + "epoch": 4.91, + "learning_rate": 2.5473017260214116e-06, + "loss": 2.9843, + "step": 22502 + }, + { + "epoch": 4.91, + "learning_rate": 2.5471924841599306e-06, + "loss": 0.9359, + "step": 22503 + }, + { + "epoch": 4.91, + "learning_rate": 2.547083242298449e-06, + "loss": 2.1578, + "step": 22504 + }, + { + "epoch": 4.91, + "learning_rate": 2.5469740004369674e-06, + "loss": 1.174, + "step": 22505 + }, + { + "epoch": 4.91, + "learning_rate": 2.5468647585754864e-06, + "loss": 0.5622, + "step": 22506 + }, + { + "epoch": 4.91, + "learning_rate": 2.546755516714005e-06, + "loss": 2.8588, + "step": 22507 + }, + { + "epoch": 4.91, + "learning_rate": 2.546646274852524e-06, + "loss": 1.9824, + "step": 22508 + }, + { + "epoch": 4.91, + "learning_rate": 2.5465370329910423e-06, + "loss": 1.5543, + "step": 22509 + }, + { + "epoch": 4.91, + "learning_rate": 2.546427791129561e-06, + "loss": 0.7151, + "step": 22510 + }, + { + "epoch": 4.91, + "learning_rate": 2.54631854926808e-06, + "loss": 2.9484, + "step": 22511 + }, + { + "epoch": 4.91, + "learning_rate": 2.5462093074065985e-06, + "loss": 1.6962, + "step": 22512 + }, + { + "epoch": 4.91, + "learning_rate": 2.546100065545117e-06, + "loss": 1.6851, + "step": 22513 + }, + { + "epoch": 4.91, + "learning_rate": 2.5459908236836358e-06, + "loss": 0.4721, + "step": 22514 + }, + { + "epoch": 4.91, + "learning_rate": 2.5458815818221544e-06, + "loss": 1.1077, + "step": 22515 + }, + { + "epoch": 4.91, + "learning_rate": 2.545772339960673e-06, + "loss": 1.3197, + "step": 22516 + }, + { + "epoch": 4.91, + "learning_rate": 2.545663098099192e-06, + "loss": 1.3842, + "step": 22517 + }, + { + "epoch": 4.91, + "learning_rate": 2.54555385623771e-06, + "loss": 1.8681, + "step": 22518 + }, + { + "epoch": 4.91, + "learning_rate": 2.5454446143762292e-06, + "loss": 0.2914, + "step": 22519 + }, + { + "epoch": 4.91, + "learning_rate": 2.545335372514748e-06, + "loss": 1.7861, + "step": 22520 + }, + { + "epoch": 4.92, + "learning_rate": 2.5452261306532665e-06, + "loss": 3.3534, + "step": 22521 + }, + { + "epoch": 4.92, + "learning_rate": 2.5451168887917855e-06, + "loss": 1.0145, + "step": 22522 + }, + { + "epoch": 4.92, + "learning_rate": 2.5450076469303037e-06, + "loss": 1.398, + "step": 22523 + }, + { + "epoch": 4.92, + "learning_rate": 2.5448984050688223e-06, + "loss": 2.1691, + "step": 22524 + }, + { + "epoch": 4.92, + "learning_rate": 2.5447891632073413e-06, + "loss": 1.7871, + "step": 22525 + }, + { + "epoch": 4.92, + "learning_rate": 2.54467992134586e-06, + "loss": 1.5888, + "step": 22526 + }, + { + "epoch": 4.92, + "learning_rate": 2.544570679484379e-06, + "loss": 1.371, + "step": 22527 + }, + { + "epoch": 4.92, + "learning_rate": 2.544461437622897e-06, + "loss": 1.3637, + "step": 22528 + }, + { + "epoch": 4.92, + "learning_rate": 2.5443521957614158e-06, + "loss": 2.0722, + "step": 22529 + }, + { + "epoch": 4.92, + "learning_rate": 2.544242953899935e-06, + "loss": 1.7221, + "step": 22530 + }, + { + "epoch": 4.92, + "learning_rate": 2.5441337120384534e-06, + "loss": 2.3492, + "step": 22531 + }, + { + "epoch": 4.92, + "learning_rate": 2.5440244701769724e-06, + "loss": 2.2543, + "step": 22532 + }, + { + "epoch": 4.92, + "learning_rate": 2.5439152283154906e-06, + "loss": 1.2161, + "step": 22533 + }, + { + "epoch": 4.92, + "learning_rate": 2.5438059864540092e-06, + "loss": 3.0283, + "step": 22534 + }, + { + "epoch": 4.92, + "learning_rate": 2.5436967445925283e-06, + "loss": 2.4214, + "step": 22535 + }, + { + "epoch": 4.92, + "learning_rate": 2.543587502731047e-06, + "loss": 1.5383, + "step": 22536 + }, + { + "epoch": 4.92, + "learning_rate": 2.543478260869565e-06, + "loss": 0.7278, + "step": 22537 + }, + { + "epoch": 4.92, + "learning_rate": 2.543369019008084e-06, + "loss": 1.1433, + "step": 22538 + }, + { + "epoch": 4.92, + "learning_rate": 2.5432597771466027e-06, + "loss": 1.9251, + "step": 22539 + }, + { + "epoch": 4.92, + "learning_rate": 2.5431505352851218e-06, + "loss": 1.9158, + "step": 22540 + }, + { + "epoch": 4.92, + "learning_rate": 2.5430412934236404e-06, + "loss": 1.2597, + "step": 22541 + }, + { + "epoch": 4.92, + "learning_rate": 2.5429320515621586e-06, + "loss": 1.5946, + "step": 22542 + }, + { + "epoch": 4.92, + "learning_rate": 2.5428228097006776e-06, + "loss": 1.7246, + "step": 22543 + }, + { + "epoch": 4.92, + "learning_rate": 2.542713567839196e-06, + "loss": 0.8665, + "step": 22544 + }, + { + "epoch": 4.92, + "learning_rate": 2.542604325977715e-06, + "loss": 3.7901, + "step": 22545 + }, + { + "epoch": 4.92, + "learning_rate": 2.542495084116234e-06, + "loss": 2.2778, + "step": 22546 + }, + { + "epoch": 4.92, + "learning_rate": 2.542385842254752e-06, + "loss": 2.0428, + "step": 22547 + }, + { + "epoch": 4.92, + "learning_rate": 2.5422766003932706e-06, + "loss": 1.0028, + "step": 22548 + }, + { + "epoch": 4.92, + "learning_rate": 2.5421673585317897e-06, + "loss": 2.6586, + "step": 22549 + }, + { + "epoch": 4.92, + "learning_rate": 2.5420581166703083e-06, + "loss": 2.2939, + "step": 22550 + }, + { + "epoch": 4.92, + "learning_rate": 2.5419488748088265e-06, + "loss": 0.9836, + "step": 22551 + }, + { + "epoch": 4.92, + "learning_rate": 2.5418396329473455e-06, + "loss": 1.6002, + "step": 22552 + }, + { + "epoch": 4.92, + "learning_rate": 2.541730391085864e-06, + "loss": 2.5825, + "step": 22553 + }, + { + "epoch": 4.92, + "learning_rate": 2.541621149224383e-06, + "loss": 2.1105, + "step": 22554 + }, + { + "epoch": 4.92, + "learning_rate": 2.5415119073629018e-06, + "loss": 1.3753, + "step": 22555 + }, + { + "epoch": 4.92, + "learning_rate": 2.54140266550142e-06, + "loss": 1.7898, + "step": 22556 + }, + { + "epoch": 4.92, + "learning_rate": 2.541293423639939e-06, + "loss": 1.557, + "step": 22557 + }, + { + "epoch": 4.92, + "learning_rate": 2.5411841817784576e-06, + "loss": 2.5775, + "step": 22558 + }, + { + "epoch": 4.92, + "learning_rate": 2.5410749399169766e-06, + "loss": 1.165, + "step": 22559 + }, + { + "epoch": 4.92, + "learning_rate": 2.5409656980554952e-06, + "loss": 1.1135, + "step": 22560 + }, + { + "epoch": 4.92, + "learning_rate": 2.5408564561940134e-06, + "loss": 2.5183, + "step": 22561 + }, + { + "epoch": 4.92, + "learning_rate": 2.5407472143325325e-06, + "loss": 1.2811, + "step": 22562 + }, + { + "epoch": 4.92, + "learning_rate": 2.540637972471051e-06, + "loss": 2.0131, + "step": 22563 + }, + { + "epoch": 4.92, + "learning_rate": 2.54052873060957e-06, + "loss": 2.3173, + "step": 22564 + }, + { + "epoch": 4.92, + "learning_rate": 2.5404194887480887e-06, + "loss": 1.9766, + "step": 22565 + }, + { + "epoch": 4.92, + "learning_rate": 2.540310246886607e-06, + "loss": 2.5854, + "step": 22566 + }, + { + "epoch": 4.93, + "learning_rate": 2.540201005025126e-06, + "loss": 2.241, + "step": 22567 + }, + { + "epoch": 4.93, + "learning_rate": 2.5400917631636446e-06, + "loss": 1.4692, + "step": 22568 + }, + { + "epoch": 4.93, + "learning_rate": 2.5399825213021636e-06, + "loss": 1.2744, + "step": 22569 + }, + { + "epoch": 4.93, + "learning_rate": 2.5398732794406818e-06, + "loss": 2.8541, + "step": 22570 + }, + { + "epoch": 4.93, + "learning_rate": 2.5397640375792004e-06, + "loss": 2.3703, + "step": 22571 + }, + { + "epoch": 4.93, + "learning_rate": 2.5396547957177194e-06, + "loss": 1.2673, + "step": 22572 + }, + { + "epoch": 4.93, + "learning_rate": 2.539545553856238e-06, + "loss": 2.1781, + "step": 22573 + }, + { + "epoch": 4.93, + "learning_rate": 2.5394363119947567e-06, + "loss": 2.7243, + "step": 22574 + }, + { + "epoch": 4.93, + "learning_rate": 2.5393270701332753e-06, + "loss": 2.0718, + "step": 22575 + }, + { + "epoch": 4.93, + "learning_rate": 2.539217828271794e-06, + "loss": 0.8578, + "step": 22576 + }, + { + "epoch": 4.93, + "learning_rate": 2.5391085864103125e-06, + "loss": 1.1959, + "step": 22577 + }, + { + "epoch": 4.93, + "learning_rate": 2.5389993445488315e-06, + "loss": 0.9976, + "step": 22578 + }, + { + "epoch": 4.93, + "learning_rate": 2.53889010268735e-06, + "loss": 1.7804, + "step": 22579 + }, + { + "epoch": 4.93, + "learning_rate": 2.5387808608258683e-06, + "loss": 1.357, + "step": 22580 + }, + { + "epoch": 4.93, + "learning_rate": 2.5386716189643874e-06, + "loss": 2.0958, + "step": 22581 + }, + { + "epoch": 4.93, + "learning_rate": 2.538562377102906e-06, + "loss": 3.0522, + "step": 22582 + }, + { + "epoch": 4.93, + "learning_rate": 2.538453135241425e-06, + "loss": 2.0087, + "step": 22583 + }, + { + "epoch": 4.93, + "learning_rate": 2.538343893379943e-06, + "loss": 1.5251, + "step": 22584 + }, + { + "epoch": 4.93, + "learning_rate": 2.538234651518462e-06, + "loss": 1.4296, + "step": 22585 + }, + { + "epoch": 4.93, + "learning_rate": 2.538125409656981e-06, + "loss": 1.3092, + "step": 22586 + }, + { + "epoch": 4.93, + "learning_rate": 2.5380161677954994e-06, + "loss": 1.8671, + "step": 22587 + }, + { + "epoch": 4.93, + "learning_rate": 2.5379069259340185e-06, + "loss": 0.842, + "step": 22588 + }, + { + "epoch": 4.93, + "learning_rate": 2.5377976840725367e-06, + "loss": 1.0463, + "step": 22589 + }, + { + "epoch": 4.93, + "learning_rate": 2.5376884422110553e-06, + "loss": 2.8135, + "step": 22590 + }, + { + "epoch": 4.93, + "learning_rate": 2.5375792003495743e-06, + "loss": 1.3256, + "step": 22591 + }, + { + "epoch": 4.93, + "learning_rate": 2.537469958488093e-06, + "loss": 1.9692, + "step": 22592 + }, + { + "epoch": 4.93, + "learning_rate": 2.537360716626612e-06, + "loss": 2.2188, + "step": 22593 + }, + { + "epoch": 4.93, + "learning_rate": 2.53725147476513e-06, + "loss": 1.9011, + "step": 22594 + }, + { + "epoch": 4.93, + "learning_rate": 2.5371422329036488e-06, + "loss": 1.8878, + "step": 22595 + }, + { + "epoch": 4.93, + "learning_rate": 2.5370329910421678e-06, + "loss": 0.8635, + "step": 22596 + }, + { + "epoch": 4.93, + "learning_rate": 2.5369237491806864e-06, + "loss": 1.6681, + "step": 22597 + }, + { + "epoch": 4.93, + "learning_rate": 2.536814507319205e-06, + "loss": 0.7175, + "step": 22598 + }, + { + "epoch": 4.93, + "learning_rate": 2.5367052654577236e-06, + "loss": 3.1288, + "step": 22599 + }, + { + "epoch": 4.93, + "learning_rate": 2.5365960235962422e-06, + "loss": 2.4038, + "step": 22600 + }, + { + "epoch": 4.93, + "learning_rate": 2.536486781734761e-06, + "loss": 2.3701, + "step": 22601 + }, + { + "epoch": 4.93, + "learning_rate": 2.53637753987328e-06, + "loss": 2.628, + "step": 22602 + }, + { + "epoch": 4.93, + "learning_rate": 2.536268298011798e-06, + "loss": 1.2665, + "step": 22603 + }, + { + "epoch": 4.93, + "learning_rate": 2.5361590561503167e-06, + "loss": 1.5323, + "step": 22604 + }, + { + "epoch": 4.93, + "learning_rate": 2.5360498142888357e-06, + "loss": 2.0147, + "step": 22605 + }, + { + "epoch": 4.93, + "learning_rate": 2.5359405724273543e-06, + "loss": 2.0471, + "step": 22606 + }, + { + "epoch": 4.93, + "learning_rate": 2.5358313305658734e-06, + "loss": 0.9642, + "step": 22607 + }, + { + "epoch": 4.93, + "learning_rate": 2.5357220887043915e-06, + "loss": 1.8125, + "step": 22608 + }, + { + "epoch": 4.93, + "learning_rate": 2.53561284684291e-06, + "loss": 2.155, + "step": 22609 + }, + { + "epoch": 4.93, + "learning_rate": 2.535503604981429e-06, + "loss": 1.5943, + "step": 22610 + }, + { + "epoch": 4.93, + "learning_rate": 2.535394363119948e-06, + "loss": 1.6989, + "step": 22611 + }, + { + "epoch": 4.93, + "learning_rate": 2.535285121258467e-06, + "loss": 3.0534, + "step": 22612 + }, + { + "epoch": 4.94, + "learning_rate": 2.535175879396985e-06, + "loss": 1.0676, + "step": 22613 + }, + { + "epoch": 4.94, + "learning_rate": 2.5350666375355036e-06, + "loss": 3.0331, + "step": 22614 + }, + { + "epoch": 4.94, + "learning_rate": 2.5349573956740227e-06, + "loss": 1.527, + "step": 22615 + }, + { + "epoch": 4.94, + "learning_rate": 2.5348481538125413e-06, + "loss": 1.1907, + "step": 22616 + }, + { + "epoch": 4.94, + "learning_rate": 2.5347389119510595e-06, + "loss": 2.4898, + "step": 22617 + }, + { + "epoch": 4.94, + "learning_rate": 2.5346296700895785e-06, + "loss": 2.2743, + "step": 22618 + }, + { + "epoch": 4.94, + "learning_rate": 2.534520428228097e-06, + "loss": 1.9294, + "step": 22619 + }, + { + "epoch": 4.94, + "learning_rate": 2.534411186366616e-06, + "loss": 1.1427, + "step": 22620 + }, + { + "epoch": 4.94, + "learning_rate": 2.5343019445051348e-06, + "loss": 2.102, + "step": 22621 + }, + { + "epoch": 4.94, + "learning_rate": 2.534192702643653e-06, + "loss": 1.5446, + "step": 22622 + }, + { + "epoch": 4.94, + "learning_rate": 2.534083460782172e-06, + "loss": 2.6143, + "step": 22623 + }, + { + "epoch": 4.94, + "learning_rate": 2.5339742189206906e-06, + "loss": 1.5908, + "step": 22624 + }, + { + "epoch": 4.94, + "learning_rate": 2.5338649770592096e-06, + "loss": 2.017, + "step": 22625 + }, + { + "epoch": 4.94, + "learning_rate": 2.5337557351977282e-06, + "loss": 3.0052, + "step": 22626 + }, + { + "epoch": 4.94, + "learning_rate": 2.5336464933362464e-06, + "loss": 1.8301, + "step": 22627 + }, + { + "epoch": 4.94, + "learning_rate": 2.5335372514747655e-06, + "loss": 0.6032, + "step": 22628 + }, + { + "epoch": 4.94, + "learning_rate": 2.533428009613284e-06, + "loss": 1.722, + "step": 22629 + }, + { + "epoch": 4.94, + "learning_rate": 2.5333187677518027e-06, + "loss": 2.173, + "step": 22630 + }, + { + "epoch": 4.94, + "learning_rate": 2.5332095258903217e-06, + "loss": 1.3656, + "step": 22631 + }, + { + "epoch": 4.94, + "learning_rate": 2.53310028402884e-06, + "loss": 1.5855, + "step": 22632 + }, + { + "epoch": 4.94, + "learning_rate": 2.5329910421673585e-06, + "loss": 1.4366, + "step": 22633 + }, + { + "epoch": 4.94, + "learning_rate": 2.5328818003058775e-06, + "loss": 3.2711, + "step": 22634 + }, + { + "epoch": 4.94, + "learning_rate": 2.532772558444396e-06, + "loss": 0.8101, + "step": 22635 + }, + { + "epoch": 4.94, + "learning_rate": 2.5326633165829143e-06, + "loss": 0.4369, + "step": 22636 + }, + { + "epoch": 4.94, + "learning_rate": 2.5325540747214334e-06, + "loss": 2.2689, + "step": 22637 + }, + { + "epoch": 4.94, + "learning_rate": 2.532444832859952e-06, + "loss": 2.4795, + "step": 22638 + }, + { + "epoch": 4.94, + "learning_rate": 2.532335590998471e-06, + "loss": 2.3497, + "step": 22639 + }, + { + "epoch": 4.94, + "learning_rate": 2.5322263491369896e-06, + "loss": 3.69, + "step": 22640 + }, + { + "epoch": 4.94, + "learning_rate": 2.532117107275508e-06, + "loss": 1.5035, + "step": 22641 + }, + { + "epoch": 4.94, + "learning_rate": 2.532007865414027e-06, + "loss": 3.2338, + "step": 22642 + }, + { + "epoch": 4.94, + "learning_rate": 2.5318986235525455e-06, + "loss": 1.9992, + "step": 22643 + }, + { + "epoch": 4.94, + "learning_rate": 2.5317893816910645e-06, + "loss": 2.277, + "step": 22644 + }, + { + "epoch": 4.94, + "learning_rate": 2.531680139829583e-06, + "loss": 0.8219, + "step": 22645 + }, + { + "epoch": 4.94, + "learning_rate": 2.5315708979681013e-06, + "loss": 2.3678, + "step": 22646 + }, + { + "epoch": 4.94, + "learning_rate": 2.5314616561066203e-06, + "loss": 1.77, + "step": 22647 + }, + { + "epoch": 4.94, + "learning_rate": 2.531352414245139e-06, + "loss": 3.0352, + "step": 22648 + }, + { + "epoch": 4.94, + "learning_rate": 2.531243172383658e-06, + "loss": 0.7698, + "step": 22649 + }, + { + "epoch": 4.94, + "learning_rate": 2.531133930522176e-06, + "loss": 2.8131, + "step": 22650 + }, + { + "epoch": 4.94, + "learning_rate": 2.5310246886606948e-06, + "loss": 2.171, + "step": 22651 + }, + { + "epoch": 4.94, + "learning_rate": 2.530915446799214e-06, + "loss": 1.1941, + "step": 22652 + }, + { + "epoch": 4.94, + "learning_rate": 2.5308062049377324e-06, + "loss": 0.2117, + "step": 22653 + }, + { + "epoch": 4.94, + "learning_rate": 2.530696963076251e-06, + "loss": 2.4147, + "step": 22654 + }, + { + "epoch": 4.94, + "learning_rate": 2.5305877212147696e-06, + "loss": 2.5812, + "step": 22655 + }, + { + "epoch": 4.94, + "learning_rate": 2.5304784793532883e-06, + "loss": 1.7655, + "step": 22656 + }, + { + "epoch": 4.94, + "learning_rate": 2.530369237491807e-06, + "loss": 0.9794, + "step": 22657 + }, + { + "epoch": 4.95, + "learning_rate": 2.530259995630326e-06, + "loss": 1.5552, + "step": 22658 + }, + { + "epoch": 4.95, + "learning_rate": 2.5301507537688445e-06, + "loss": 0.8479, + "step": 22659 + }, + { + "epoch": 4.95, + "learning_rate": 2.5300415119073627e-06, + "loss": 0.7839, + "step": 22660 + }, + { + "epoch": 4.95, + "learning_rate": 2.5299322700458817e-06, + "loss": 0.5759, + "step": 22661 + }, + { + "epoch": 4.95, + "learning_rate": 2.5298230281844003e-06, + "loss": 1.077, + "step": 22662 + }, + { + "epoch": 4.95, + "learning_rate": 2.5297137863229194e-06, + "loss": 2.4828, + "step": 22663 + }, + { + "epoch": 4.95, + "learning_rate": 2.529604544461438e-06, + "loss": 0.6945, + "step": 22664 + }, + { + "epoch": 4.95, + "learning_rate": 2.529495302599956e-06, + "loss": 1.8807, + "step": 22665 + }, + { + "epoch": 4.95, + "learning_rate": 2.5293860607384752e-06, + "loss": 0.7256, + "step": 22666 + }, + { + "epoch": 4.95, + "learning_rate": 2.529276818876994e-06, + "loss": 0.6641, + "step": 22667 + }, + { + "epoch": 4.95, + "learning_rate": 2.529167577015513e-06, + "loss": 1.1174, + "step": 22668 + }, + { + "epoch": 4.95, + "learning_rate": 2.529058335154031e-06, + "loss": 1.7934, + "step": 22669 + }, + { + "epoch": 4.95, + "learning_rate": 2.5289490932925497e-06, + "loss": 1.0598, + "step": 22670 + }, + { + "epoch": 4.95, + "learning_rate": 2.5288398514310687e-06, + "loss": 0.8504, + "step": 22671 + }, + { + "epoch": 4.95, + "learning_rate": 2.5287306095695873e-06, + "loss": 0.7866, + "step": 22672 + }, + { + "epoch": 4.95, + "learning_rate": 2.5286213677081063e-06, + "loss": 2.4182, + "step": 22673 + }, + { + "epoch": 4.95, + "learning_rate": 2.5285121258466245e-06, + "loss": 1.2178, + "step": 22674 + }, + { + "epoch": 4.95, + "learning_rate": 2.528402883985143e-06, + "loss": 1.1865, + "step": 22675 + }, + { + "epoch": 4.95, + "learning_rate": 2.528293642123662e-06, + "loss": 1.2, + "step": 22676 + }, + { + "epoch": 4.95, + "learning_rate": 2.5281844002621808e-06, + "loss": 2.2201, + "step": 22677 + }, + { + "epoch": 4.95, + "learning_rate": 2.5280751584007e-06, + "loss": 1.0543, + "step": 22678 + }, + { + "epoch": 4.95, + "learning_rate": 2.527965916539218e-06, + "loss": 1.4064, + "step": 22679 + }, + { + "epoch": 4.95, + "learning_rate": 2.5278566746777366e-06, + "loss": 1.6903, + "step": 22680 + }, + { + "epoch": 4.95, + "learning_rate": 2.5277474328162556e-06, + "loss": 1.0086, + "step": 22681 + }, + { + "epoch": 4.95, + "learning_rate": 2.5276381909547743e-06, + "loss": 1.2528, + "step": 22682 + }, + { + "epoch": 4.95, + "learning_rate": 2.5275289490932924e-06, + "loss": 0.8266, + "step": 22683 + }, + { + "epoch": 4.95, + "learning_rate": 2.5274197072318115e-06, + "loss": 2.2023, + "step": 22684 + }, + { + "epoch": 4.95, + "learning_rate": 2.52731046537033e-06, + "loss": 1.8634, + "step": 22685 + }, + { + "epoch": 4.95, + "learning_rate": 2.5272012235088487e-06, + "loss": 1.1189, + "step": 22686 + }, + { + "epoch": 4.95, + "learning_rate": 2.5270919816473677e-06, + "loss": 2.2217, + "step": 22687 + }, + { + "epoch": 4.95, + "learning_rate": 2.526982739785886e-06, + "loss": 0.9637, + "step": 22688 + }, + { + "epoch": 4.95, + "learning_rate": 2.5268734979244045e-06, + "loss": 0.438, + "step": 22689 + }, + { + "epoch": 4.95, + "learning_rate": 2.5267642560629236e-06, + "loss": 1.8721, + "step": 22690 + }, + { + "epoch": 4.95, + "learning_rate": 2.526655014201442e-06, + "loss": 1.1461, + "step": 22691 + }, + { + "epoch": 4.95, + "learning_rate": 2.5265457723399612e-06, + "loss": 2.4205, + "step": 22692 + }, + { + "epoch": 4.95, + "learning_rate": 2.5264365304784794e-06, + "loss": 0.8627, + "step": 22693 + }, + { + "epoch": 4.95, + "learning_rate": 2.526327288616998e-06, + "loss": 1.711, + "step": 22694 + }, + { + "epoch": 4.95, + "learning_rate": 2.526218046755517e-06, + "loss": 0.6687, + "step": 22695 + }, + { + "epoch": 4.95, + "learning_rate": 2.5261088048940357e-06, + "loss": 1.5554, + "step": 22696 + }, + { + "epoch": 4.95, + "learning_rate": 2.5259995630325547e-06, + "loss": 1.1572, + "step": 22697 + }, + { + "epoch": 4.95, + "learning_rate": 2.525890321171073e-06, + "loss": 1.2878, + "step": 22698 + }, + { + "epoch": 4.95, + "learning_rate": 2.5257810793095915e-06, + "loss": 1.4111, + "step": 22699 + }, + { + "epoch": 4.95, + "learning_rate": 2.5256718374481105e-06, + "loss": 2.1823, + "step": 22700 + }, + { + "epoch": 4.95, + "learning_rate": 2.525562595586629e-06, + "loss": 1.9014, + "step": 22701 + }, + { + "epoch": 4.95, + "learning_rate": 2.5254533537251473e-06, + "loss": 0.4911, + "step": 22702 + }, + { + "epoch": 4.95, + "learning_rate": 2.5253441118636664e-06, + "loss": 0.6296, + "step": 22703 + }, + { + "epoch": 4.96, + "learning_rate": 2.525234870002185e-06, + "loss": 1.6084, + "step": 22704 + }, + { + "epoch": 4.96, + "learning_rate": 2.525125628140704e-06, + "loss": 1.4582, + "step": 22705 + }, + { + "epoch": 4.96, + "learning_rate": 2.5250163862792226e-06, + "loss": 1.3975, + "step": 22706 + }, + { + "epoch": 4.96, + "learning_rate": 2.524907144417741e-06, + "loss": 2.6717, + "step": 22707 + }, + { + "epoch": 4.96, + "learning_rate": 2.52479790255626e-06, + "loss": 1.0387, + "step": 22708 + }, + { + "epoch": 4.96, + "learning_rate": 2.5246886606947785e-06, + "loss": 2.3074, + "step": 22709 + }, + { + "epoch": 4.96, + "learning_rate": 2.524579418833297e-06, + "loss": 1.8385, + "step": 22710 + }, + { + "epoch": 4.96, + "learning_rate": 2.524470176971816e-06, + "loss": 1.7194, + "step": 22711 + }, + { + "epoch": 4.96, + "learning_rate": 2.5243609351103343e-06, + "loss": 1.9288, + "step": 22712 + }, + { + "epoch": 4.96, + "learning_rate": 2.5242516932488533e-06, + "loss": 1.4595, + "step": 22713 + }, + { + "epoch": 4.96, + "learning_rate": 2.524142451387372e-06, + "loss": 0.7826, + "step": 22714 + }, + { + "epoch": 4.96, + "learning_rate": 2.5240332095258905e-06, + "loss": 2.7904, + "step": 22715 + }, + { + "epoch": 4.96, + "learning_rate": 2.523923967664409e-06, + "loss": 3.2244, + "step": 22716 + }, + { + "epoch": 4.96, + "learning_rate": 2.5238147258029278e-06, + "loss": 1.3054, + "step": 22717 + }, + { + "epoch": 4.96, + "learning_rate": 2.5237054839414464e-06, + "loss": 0.7642, + "step": 22718 + }, + { + "epoch": 4.96, + "learning_rate": 2.5235962420799654e-06, + "loss": 2.7353, + "step": 22719 + }, + { + "epoch": 4.96, + "learning_rate": 2.523487000218484e-06, + "loss": 1.384, + "step": 22720 + }, + { + "epoch": 4.96, + "learning_rate": 2.523377758357002e-06, + "loss": 1.239, + "step": 22721 + }, + { + "epoch": 4.96, + "learning_rate": 2.5232685164955212e-06, + "loss": 1.1859, + "step": 22722 + }, + { + "epoch": 4.96, + "learning_rate": 2.52315927463404e-06, + "loss": 2.709, + "step": 22723 + }, + { + "epoch": 4.96, + "learning_rate": 2.523050032772559e-06, + "loss": 2.2641, + "step": 22724 + }, + { + "epoch": 4.96, + "learning_rate": 2.5229407909110775e-06, + "loss": 0.6586, + "step": 22725 + }, + { + "epoch": 4.96, + "learning_rate": 2.5228315490495957e-06, + "loss": 0.867, + "step": 22726 + }, + { + "epoch": 4.96, + "learning_rate": 2.5227223071881147e-06, + "loss": 2.6297, + "step": 22727 + }, + { + "epoch": 4.96, + "learning_rate": 2.5226130653266333e-06, + "loss": 1.6081, + "step": 22728 + }, + { + "epoch": 4.96, + "learning_rate": 2.5225038234651524e-06, + "loss": 2.3901, + "step": 22729 + }, + { + "epoch": 4.96, + "learning_rate": 2.5223945816036706e-06, + "loss": 0.9619, + "step": 22730 + }, + { + "epoch": 4.96, + "learning_rate": 2.522285339742189e-06, + "loss": 0.9148, + "step": 22731 + }, + { + "epoch": 4.96, + "learning_rate": 2.522176097880708e-06, + "loss": 0.8805, + "step": 22732 + }, + { + "epoch": 4.96, + "learning_rate": 2.522066856019227e-06, + "loss": 1.0362, + "step": 22733 + }, + { + "epoch": 4.96, + "learning_rate": 2.521957614157746e-06, + "loss": 1.6826, + "step": 22734 + }, + { + "epoch": 4.96, + "learning_rate": 2.521848372296264e-06, + "loss": 0.7607, + "step": 22735 + }, + { + "epoch": 4.96, + "learning_rate": 2.5217391304347826e-06, + "loss": 0.9151, + "step": 22736 + }, + { + "epoch": 4.96, + "learning_rate": 2.5216298885733017e-06, + "loss": 1.684, + "step": 22737 + }, + { + "epoch": 4.96, + "learning_rate": 2.5215206467118203e-06, + "loss": 1.0811, + "step": 22738 + }, + { + "epoch": 4.96, + "learning_rate": 2.521411404850339e-06, + "loss": 0.816, + "step": 22739 + }, + { + "epoch": 4.96, + "learning_rate": 2.5213021629888575e-06, + "loss": 2.3753, + "step": 22740 + }, + { + "epoch": 4.96, + "learning_rate": 2.521192921127376e-06, + "loss": 2.1071, + "step": 22741 + }, + { + "epoch": 4.96, + "learning_rate": 2.5210836792658947e-06, + "loss": 2.5793, + "step": 22742 + }, + { + "epoch": 4.96, + "learning_rate": 2.5209744374044138e-06, + "loss": 0.9332, + "step": 22743 + }, + { + "epoch": 4.96, + "learning_rate": 2.5208651955429324e-06, + "loss": 2.1832, + "step": 22744 + }, + { + "epoch": 4.96, + "learning_rate": 2.5207559536814506e-06, + "loss": 0.3319, + "step": 22745 + }, + { + "epoch": 4.96, + "learning_rate": 2.5206467118199696e-06, + "loss": 1.1997, + "step": 22746 + }, + { + "epoch": 4.96, + "learning_rate": 2.520537469958488e-06, + "loss": 1.0785, + "step": 22747 + }, + { + "epoch": 4.96, + "learning_rate": 2.5204282280970072e-06, + "loss": 1.0489, + "step": 22748 + }, + { + "epoch": 4.96, + "learning_rate": 2.5203189862355254e-06, + "loss": 1.3227, + "step": 22749 + }, + { + "epoch": 4.97, + "learning_rate": 2.520209744374044e-06, + "loss": 1.7204, + "step": 22750 + }, + { + "epoch": 4.97, + "learning_rate": 2.520100502512563e-06, + "loss": 0.7121, + "step": 22751 + }, + { + "epoch": 4.97, + "learning_rate": 2.5199912606510817e-06, + "loss": 1.5857, + "step": 22752 + }, + { + "epoch": 4.97, + "learning_rate": 2.5198820187896007e-06, + "loss": 2.1762, + "step": 22753 + }, + { + "epoch": 4.97, + "learning_rate": 2.519772776928119e-06, + "loss": 1.1261, + "step": 22754 + }, + { + "epoch": 4.97, + "learning_rate": 2.5196635350666375e-06, + "loss": 1.2212, + "step": 22755 + }, + { + "epoch": 4.97, + "learning_rate": 2.5195542932051566e-06, + "loss": 1.0881, + "step": 22756 + }, + { + "epoch": 4.97, + "learning_rate": 2.519445051343675e-06, + "loss": 2.7283, + "step": 22757 + }, + { + "epoch": 4.97, + "learning_rate": 2.519335809482194e-06, + "loss": 1.0169, + "step": 22758 + }, + { + "epoch": 4.97, + "learning_rate": 2.5192265676207124e-06, + "loss": 0.3995, + "step": 22759 + }, + { + "epoch": 4.97, + "learning_rate": 2.519117325759231e-06, + "loss": 1.637, + "step": 22760 + }, + { + "epoch": 4.97, + "learning_rate": 2.51900808389775e-06, + "loss": 0.2739, + "step": 22761 + }, + { + "epoch": 4.97, + "learning_rate": 2.5188988420362686e-06, + "loss": 1.7073, + "step": 22762 + }, + { + "epoch": 4.97, + "learning_rate": 2.518789600174787e-06, + "loss": 0.9271, + "step": 22763 + }, + { + "epoch": 4.97, + "learning_rate": 2.518680358313306e-06, + "loss": 1.7938, + "step": 22764 + }, + { + "epoch": 4.97, + "learning_rate": 2.5185711164518245e-06, + "loss": 1.7439, + "step": 22765 + }, + { + "epoch": 4.97, + "learning_rate": 2.5184618745903435e-06, + "loss": 2.0612, + "step": 22766 + }, + { + "epoch": 4.97, + "learning_rate": 2.518352632728862e-06, + "loss": 1.5654, + "step": 22767 + }, + { + "epoch": 4.97, + "learning_rate": 2.5182433908673803e-06, + "loss": 1.6053, + "step": 22768 + }, + { + "epoch": 4.97, + "learning_rate": 2.5181341490058993e-06, + "loss": 1.0551, + "step": 22769 + }, + { + "epoch": 4.97, + "learning_rate": 2.518024907144418e-06, + "loss": 1.9236, + "step": 22770 + }, + { + "epoch": 4.97, + "learning_rate": 2.5179156652829366e-06, + "loss": 2.3404, + "step": 22771 + }, + { + "epoch": 4.97, + "learning_rate": 2.5178064234214556e-06, + "loss": 1.5998, + "step": 22772 + }, + { + "epoch": 4.97, + "learning_rate": 2.517697181559974e-06, + "loss": 2.613, + "step": 22773 + }, + { + "epoch": 4.97, + "learning_rate": 2.5175879396984924e-06, + "loss": 1.3607, + "step": 22774 + }, + { + "epoch": 4.97, + "learning_rate": 2.5174786978370114e-06, + "loss": 1.9223, + "step": 22775 + }, + { + "epoch": 4.97, + "learning_rate": 2.51736945597553e-06, + "loss": 1.8768, + "step": 22776 + }, + { + "epoch": 4.97, + "learning_rate": 2.517260214114049e-06, + "loss": 0.8034, + "step": 22777 + }, + { + "epoch": 4.97, + "learning_rate": 2.5171509722525673e-06, + "loss": 0.8833, + "step": 22778 + }, + { + "epoch": 4.97, + "learning_rate": 2.517041730391086e-06, + "loss": 2.2637, + "step": 22779 + }, + { + "epoch": 4.97, + "learning_rate": 2.516932488529605e-06, + "loss": 2.3066, + "step": 22780 + }, + { + "epoch": 4.97, + "learning_rate": 2.5168232466681235e-06, + "loss": 0.4617, + "step": 22781 + }, + { + "epoch": 4.97, + "learning_rate": 2.5167140048066417e-06, + "loss": 1.6131, + "step": 22782 + }, + { + "epoch": 4.97, + "learning_rate": 2.5166047629451607e-06, + "loss": 1.4567, + "step": 22783 + }, + { + "epoch": 4.97, + "learning_rate": 2.5164955210836794e-06, + "loss": 0.7243, + "step": 22784 + }, + { + "epoch": 4.97, + "learning_rate": 2.5163862792221984e-06, + "loss": 1.7776, + "step": 22785 + }, + { + "epoch": 4.97, + "learning_rate": 2.516277037360717e-06, + "loss": 1.5891, + "step": 22786 + }, + { + "epoch": 4.97, + "learning_rate": 2.516167795499235e-06, + "loss": 1.1685, + "step": 22787 + }, + { + "epoch": 4.97, + "learning_rate": 2.5160585536377542e-06, + "loss": 1.2109, + "step": 22788 + }, + { + "epoch": 4.97, + "learning_rate": 2.515949311776273e-06, + "loss": 1.5661, + "step": 22789 + }, + { + "epoch": 4.97, + "learning_rate": 2.515840069914792e-06, + "loss": 1.7593, + "step": 22790 + }, + { + "epoch": 4.97, + "learning_rate": 2.5157308280533105e-06, + "loss": 1.2129, + "step": 22791 + }, + { + "epoch": 4.97, + "learning_rate": 2.5156215861918287e-06, + "loss": 1.2964, + "step": 22792 + }, + { + "epoch": 4.97, + "learning_rate": 2.5155123443303477e-06, + "loss": 1.8962, + "step": 22793 + }, + { + "epoch": 4.97, + "learning_rate": 2.5154031024688663e-06, + "loss": 1.4385, + "step": 22794 + }, + { + "epoch": 4.97, + "learning_rate": 2.515293860607385e-06, + "loss": 1.6306, + "step": 22795 + }, + { + "epoch": 4.98, + "learning_rate": 2.5151846187459035e-06, + "loss": 2.2678, + "step": 22796 + }, + { + "epoch": 4.98, + "learning_rate": 2.515075376884422e-06, + "loss": 0.4612, + "step": 22797 + }, + { + "epoch": 4.98, + "learning_rate": 2.5149661350229408e-06, + "loss": 1.9795, + "step": 22798 + }, + { + "epoch": 4.98, + "learning_rate": 2.51485689316146e-06, + "loss": 1.3959, + "step": 22799 + }, + { + "epoch": 4.98, + "learning_rate": 2.5147476512999784e-06, + "loss": 1.7467, + "step": 22800 + }, + { + "epoch": 4.98, + "learning_rate": 2.5146384094384966e-06, + "loss": 1.5343, + "step": 22801 + }, + { + "epoch": 4.98, + "learning_rate": 2.5145291675770156e-06, + "loss": 2.6798, + "step": 22802 + }, + { + "epoch": 4.98, + "learning_rate": 2.5144199257155342e-06, + "loss": 2.2548, + "step": 22803 + }, + { + "epoch": 4.98, + "learning_rate": 2.5143106838540533e-06, + "loss": 2.0241, + "step": 22804 + }, + { + "epoch": 4.98, + "learning_rate": 2.514201441992572e-06, + "loss": 1.8609, + "step": 22805 + }, + { + "epoch": 4.98, + "learning_rate": 2.51409220013109e-06, + "loss": 1.6227, + "step": 22806 + }, + { + "epoch": 4.98, + "learning_rate": 2.513982958269609e-06, + "loss": 3.5462, + "step": 22807 + }, + { + "epoch": 4.98, + "learning_rate": 2.5138737164081277e-06, + "loss": 0.7211, + "step": 22808 + }, + { + "epoch": 4.98, + "learning_rate": 2.5137644745466467e-06, + "loss": 2.1325, + "step": 22809 + }, + { + "epoch": 4.98, + "learning_rate": 2.5136552326851654e-06, + "loss": 2.5843, + "step": 22810 + }, + { + "epoch": 4.98, + "learning_rate": 2.5135459908236835e-06, + "loss": 1.5172, + "step": 22811 + }, + { + "epoch": 4.98, + "learning_rate": 2.5134367489622026e-06, + "loss": 1.3627, + "step": 22812 + }, + { + "epoch": 4.98, + "learning_rate": 2.513327507100721e-06, + "loss": 2.3909, + "step": 22813 + }, + { + "epoch": 4.98, + "learning_rate": 2.5132182652392402e-06, + "loss": 2.2724, + "step": 22814 + }, + { + "epoch": 4.98, + "learning_rate": 2.5131090233777584e-06, + "loss": 1.2079, + "step": 22815 + }, + { + "epoch": 4.98, + "learning_rate": 2.512999781516277e-06, + "loss": 2.2701, + "step": 22816 + }, + { + "epoch": 4.98, + "learning_rate": 2.512890539654796e-06, + "loss": 0.7174, + "step": 22817 + }, + { + "epoch": 4.98, + "learning_rate": 2.5127812977933147e-06, + "loss": 2.5271, + "step": 22818 + }, + { + "epoch": 4.98, + "learning_rate": 2.5126720559318337e-06, + "loss": 1.6277, + "step": 22819 + }, + { + "epoch": 4.98, + "learning_rate": 2.512562814070352e-06, + "loss": 0.5012, + "step": 22820 + }, + { + "epoch": 4.98, + "learning_rate": 2.5124535722088705e-06, + "loss": 0.9416, + "step": 22821 + }, + { + "epoch": 4.98, + "learning_rate": 2.5123443303473895e-06, + "loss": 1.3326, + "step": 22822 + }, + { + "epoch": 4.98, + "learning_rate": 2.512235088485908e-06, + "loss": 1.6131, + "step": 22823 + }, + { + "epoch": 4.98, + "learning_rate": 2.5121258466244268e-06, + "loss": 1.6305, + "step": 22824 + }, + { + "epoch": 4.98, + "learning_rate": 2.5120166047629454e-06, + "loss": 3.1655, + "step": 22825 + }, + { + "epoch": 4.98, + "learning_rate": 2.511907362901464e-06, + "loss": 0.5576, + "step": 22826 + }, + { + "epoch": 4.98, + "learning_rate": 2.5117981210399826e-06, + "loss": 1.2586, + "step": 22827 + }, + { + "epoch": 4.98, + "learning_rate": 2.5116888791785016e-06, + "loss": 1.9021, + "step": 22828 + }, + { + "epoch": 4.98, + "learning_rate": 2.51157963731702e-06, + "loss": 1.5981, + "step": 22829 + }, + { + "epoch": 4.98, + "learning_rate": 2.5114703954555384e-06, + "loss": 2.4198, + "step": 22830 + }, + { + "epoch": 4.98, + "learning_rate": 2.5113611535940575e-06, + "loss": 0.9946, + "step": 22831 + }, + { + "epoch": 4.98, + "learning_rate": 2.511251911732576e-06, + "loss": 3.3713, + "step": 22832 + }, + { + "epoch": 4.98, + "learning_rate": 2.511142669871095e-06, + "loss": 1.0101, + "step": 22833 + }, + { + "epoch": 4.98, + "learning_rate": 2.5110334280096133e-06, + "loss": 1.0219, + "step": 22834 + }, + { + "epoch": 4.98, + "learning_rate": 2.510924186148132e-06, + "loss": 2.0575, + "step": 22835 + }, + { + "epoch": 4.98, + "learning_rate": 2.510814944286651e-06, + "loss": 1.2263, + "step": 22836 + }, + { + "epoch": 4.98, + "learning_rate": 2.5107057024251696e-06, + "loss": 1.8018, + "step": 22837 + }, + { + "epoch": 4.98, + "learning_rate": 2.5105964605636886e-06, + "loss": 0.6392, + "step": 22838 + }, + { + "epoch": 4.98, + "learning_rate": 2.5104872187022068e-06, + "loss": 1.0113, + "step": 22839 + }, + { + "epoch": 4.98, + "learning_rate": 2.5103779768407254e-06, + "loss": 1.3236, + "step": 22840 + }, + { + "epoch": 4.98, + "learning_rate": 2.5102687349792444e-06, + "loss": 2.3584, + "step": 22841 + }, + { + "epoch": 4.99, + "learning_rate": 2.510159493117763e-06, + "loss": 2.8613, + "step": 22842 + }, + { + "epoch": 4.99, + "learning_rate": 2.510050251256282e-06, + "loss": 0.8018, + "step": 22843 + }, + { + "epoch": 4.99, + "learning_rate": 2.5099410093948003e-06, + "loss": 1.8075, + "step": 22844 + }, + { + "epoch": 4.99, + "learning_rate": 2.509831767533319e-06, + "loss": 1.1705, + "step": 22845 + }, + { + "epoch": 4.99, + "learning_rate": 2.509722525671838e-06, + "loss": 0.6978, + "step": 22846 + }, + { + "epoch": 4.99, + "learning_rate": 2.5096132838103565e-06, + "loss": 1.51, + "step": 22847 + }, + { + "epoch": 4.99, + "learning_rate": 2.5095040419488747e-06, + "loss": 1.8302, + "step": 22848 + }, + { + "epoch": 4.99, + "learning_rate": 2.5093948000873937e-06, + "loss": 1.547, + "step": 22849 + }, + { + "epoch": 4.99, + "learning_rate": 2.5092855582259123e-06, + "loss": 1.9691, + "step": 22850 + }, + { + "epoch": 4.99, + "learning_rate": 2.509176316364431e-06, + "loss": 2.7652, + "step": 22851 + }, + { + "epoch": 4.99, + "learning_rate": 2.50906707450295e-06, + "loss": 0.8087, + "step": 22852 + }, + { + "epoch": 4.99, + "learning_rate": 2.508957832641468e-06, + "loss": 2.9821, + "step": 22853 + }, + { + "epoch": 4.99, + "learning_rate": 2.5088485907799868e-06, + "loss": 1.8435, + "step": 22854 + }, + { + "epoch": 4.99, + "learning_rate": 2.508739348918506e-06, + "loss": 1.4298, + "step": 22855 + }, + { + "epoch": 4.99, + "learning_rate": 2.5086301070570244e-06, + "loss": 0.9442, + "step": 22856 + }, + { + "epoch": 4.99, + "learning_rate": 2.5085208651955435e-06, + "loss": 0.9944, + "step": 22857 + }, + { + "epoch": 4.99, + "learning_rate": 2.5084116233340617e-06, + "loss": 1.4923, + "step": 22858 + }, + { + "epoch": 4.99, + "learning_rate": 2.5083023814725803e-06, + "loss": 1.6413, + "step": 22859 + }, + { + "epoch": 4.99, + "learning_rate": 2.5081931396110993e-06, + "loss": 1.1318, + "step": 22860 + }, + { + "epoch": 4.99, + "learning_rate": 2.508083897749618e-06, + "loss": 0.4507, + "step": 22861 + }, + { + "epoch": 4.99, + "learning_rate": 2.507974655888136e-06, + "loss": 0.1426, + "step": 22862 + }, + { + "epoch": 4.99, + "learning_rate": 2.507865414026655e-06, + "loss": 1.0726, + "step": 22863 + }, + { + "epoch": 4.99, + "learning_rate": 2.5077561721651737e-06, + "loss": 2.3619, + "step": 22864 + }, + { + "epoch": 4.99, + "learning_rate": 2.5076469303036928e-06, + "loss": 3.1041, + "step": 22865 + }, + { + "epoch": 4.99, + "learning_rate": 2.5075376884422114e-06, + "loss": 2.157, + "step": 22866 + }, + { + "epoch": 4.99, + "learning_rate": 2.5074284465807296e-06, + "loss": 1.0491, + "step": 22867 + }, + { + "epoch": 4.99, + "learning_rate": 2.5073192047192486e-06, + "loss": 2.1144, + "step": 22868 + }, + { + "epoch": 4.99, + "learning_rate": 2.5072099628577672e-06, + "loss": 0.5202, + "step": 22869 + }, + { + "epoch": 4.99, + "learning_rate": 2.5071007209962863e-06, + "loss": 0.686, + "step": 22870 + }, + { + "epoch": 4.99, + "learning_rate": 2.506991479134805e-06, + "loss": 1.5706, + "step": 22871 + }, + { + "epoch": 4.99, + "learning_rate": 2.506882237273323e-06, + "loss": 2.2863, + "step": 22872 + }, + { + "epoch": 4.99, + "learning_rate": 2.506772995411842e-06, + "loss": 2.1078, + "step": 22873 + }, + { + "epoch": 4.99, + "learning_rate": 2.5066637535503607e-06, + "loss": 3.0767, + "step": 22874 + }, + { + "epoch": 4.99, + "learning_rate": 2.5065545116888797e-06, + "loss": 1.6834, + "step": 22875 + }, + { + "epoch": 4.99, + "learning_rate": 2.5064452698273983e-06, + "loss": 2.0247, + "step": 22876 + }, + { + "epoch": 4.99, + "learning_rate": 2.5063360279659165e-06, + "loss": 2.178, + "step": 22877 + }, + { + "epoch": 4.99, + "learning_rate": 2.5062267861044356e-06, + "loss": 1.7503, + "step": 22878 + }, + { + "epoch": 4.99, + "learning_rate": 2.506117544242954e-06, + "loss": 2.1425, + "step": 22879 + }, + { + "epoch": 4.99, + "learning_rate": 2.506008302381473e-06, + "loss": 1.8775, + "step": 22880 + }, + { + "epoch": 4.99, + "learning_rate": 2.5058990605199914e-06, + "loss": 1.3953, + "step": 22881 + }, + { + "epoch": 4.99, + "learning_rate": 2.50578981865851e-06, + "loss": 1.7753, + "step": 22882 + }, + { + "epoch": 4.99, + "learning_rate": 2.5056805767970286e-06, + "loss": 3.3219, + "step": 22883 + }, + { + "epoch": 4.99, + "learning_rate": 2.5055713349355477e-06, + "loss": 1.924, + "step": 22884 + }, + { + "epoch": 4.99, + "learning_rate": 2.5054620930740663e-06, + "loss": 1.5228, + "step": 22885 + }, + { + "epoch": 4.99, + "learning_rate": 2.5053528512125845e-06, + "loss": 0.5499, + "step": 22886 + }, + { + "epoch": 4.99, + "learning_rate": 2.5052436093511035e-06, + "loss": 1.1558, + "step": 22887 + }, + { + "epoch": 5.0, + "learning_rate": 2.505134367489622e-06, + "loss": 1.9135, + "step": 22888 + }, + { + "epoch": 5.0, + "learning_rate": 2.505025125628141e-06, + "loss": 1.9171, + "step": 22889 + }, + { + "epoch": 5.0, + "learning_rate": 2.5049158837666597e-06, + "loss": 2.5007, + "step": 22890 + }, + { + "epoch": 5.0, + "learning_rate": 2.504806641905178e-06, + "loss": 1.0876, + "step": 22891 + }, + { + "epoch": 5.0, + "learning_rate": 2.504697400043697e-06, + "loss": 0.2303, + "step": 22892 + }, + { + "epoch": 5.0, + "learning_rate": 2.5045881581822156e-06, + "loss": 2.8821, + "step": 22893 + }, + { + "epoch": 5.0, + "learning_rate": 2.5044789163207346e-06, + "loss": 2.5304, + "step": 22894 + }, + { + "epoch": 5.0, + "learning_rate": 2.504369674459253e-06, + "loss": 2.5343, + "step": 22895 + }, + { + "epoch": 5.0, + "learning_rate": 2.5042604325977714e-06, + "loss": 1.9256, + "step": 22896 + }, + { + "epoch": 5.0, + "learning_rate": 2.5041511907362904e-06, + "loss": 1.8795, + "step": 22897 + }, + { + "epoch": 5.0, + "learning_rate": 2.504041948874809e-06, + "loss": 2.5412, + "step": 22898 + }, + { + "epoch": 5.0, + "learning_rate": 2.503932707013328e-06, + "loss": 1.244, + "step": 22899 + }, + { + "epoch": 5.0, + "learning_rate": 2.5038234651518463e-06, + "loss": 1.707, + "step": 22900 + }, + { + "epoch": 5.0, + "learning_rate": 2.503714223290365e-06, + "loss": 3.6331, + "step": 22901 + }, + { + "epoch": 5.0, + "learning_rate": 2.503604981428884e-06, + "loss": 1.2396, + "step": 22902 + }, + { + "epoch": 5.0, + "learning_rate": 2.5034957395674025e-06, + "loss": 1.1673, + "step": 22903 + }, + { + "epoch": 5.0, + "learning_rate": 2.503386497705921e-06, + "loss": 1.6849, + "step": 22904 + }, + { + "epoch": 5.0, + "learning_rate": 2.5032772558444398e-06, + "loss": 1.2936, + "step": 22905 + }, + { + "epoch": 5.0, + "learning_rate": 2.5031680139829584e-06, + "loss": 1.7036, + "step": 22906 + }, + { + "epoch": 5.0, + "learning_rate": 2.503058772121477e-06, + "loss": 1.7556, + "step": 22907 + }, + { + "epoch": 5.0, + "learning_rate": 2.502949530259996e-06, + "loss": 2.2248, + "step": 22908 + }, + { + "epoch": 5.0, + "learning_rate": 2.5028402883985146e-06, + "loss": 1.6356, + "step": 22909 + }, + { + "epoch": 5.0, + "learning_rate": 2.5027310465370332e-06, + "loss": 0.9259, + "step": 22910 + }, + { + "epoch": 5.0, + "eval_accuracy": 0.6091653027823241, + "eval_loss": 1.558190941810608, + "eval_runtime": 326.8889, + "eval_samples_per_second": 18.691, + "eval_steps_per_second": 4.674, + "step": 22910 + }, + { + "epoch": 5.0, + "learning_rate": 2.502621804675552e-06, + "loss": 1.253, + "step": 22911 + }, + { + "epoch": 5.0, + "learning_rate": 2.5025125628140705e-06, + "loss": 1.659, + "step": 22912 + }, + { + "epoch": 5.0, + "learning_rate": 2.5024033209525895e-06, + "loss": 2.1833, + "step": 22913 + }, + { + "epoch": 5.0, + "learning_rate": 2.5022940790911077e-06, + "loss": 1.418, + "step": 22914 + }, + { + "epoch": 5.0, + "learning_rate": 2.5021848372296263e-06, + "loss": 1.5618, + "step": 22915 + }, + { + "epoch": 5.0, + "learning_rate": 2.5020755953681453e-06, + "loss": 2.3402, + "step": 22916 + }, + { + "epoch": 5.0, + "learning_rate": 2.501966353506664e-06, + "loss": 0.8837, + "step": 22917 + }, + { + "epoch": 5.0, + "learning_rate": 2.501857111645183e-06, + "loss": 1.8855, + "step": 22918 + }, + { + "epoch": 5.0, + "learning_rate": 2.501747869783701e-06, + "loss": 1.7426, + "step": 22919 + }, + { + "epoch": 5.0, + "learning_rate": 2.5016386279222198e-06, + "loss": 0.8231, + "step": 22920 + }, + { + "epoch": 5.0, + "learning_rate": 2.501529386060739e-06, + "loss": 2.107, + "step": 22921 + }, + { + "epoch": 5.0, + "learning_rate": 2.5014201441992574e-06, + "loss": 1.3929, + "step": 22922 + }, + { + "epoch": 5.0, + "learning_rate": 2.5013109023377764e-06, + "loss": 1.401, + "step": 22923 + }, + { + "epoch": 5.0, + "learning_rate": 2.5012016604762946e-06, + "loss": 0.3905, + "step": 22924 + }, + { + "epoch": 5.0, + "learning_rate": 2.5010924186148132e-06, + "loss": 1.4302, + "step": 22925 + }, + { + "epoch": 5.0, + "learning_rate": 2.5009831767533323e-06, + "loss": 1.8224, + "step": 22926 + }, + { + "epoch": 5.0, + "learning_rate": 2.500873934891851e-06, + "loss": 1.7947, + "step": 22927 + }, + { + "epoch": 5.0, + "learning_rate": 2.500764693030369e-06, + "loss": 2.1428, + "step": 22928 + }, + { + "epoch": 5.0, + "learning_rate": 2.500655451168888e-06, + "loss": 1.6865, + "step": 22929 + }, + { + "epoch": 5.0, + "learning_rate": 2.5005462093074067e-06, + "loss": 0.4413, + "step": 22930 + }, + { + "epoch": 5.0, + "learning_rate": 2.5004369674459258e-06, + "loss": 1.1779, + "step": 22931 + }, + { + "epoch": 5.0, + "learning_rate": 2.5003277255844444e-06, + "loss": 1.5709, + "step": 22932 + }, + { + "epoch": 5.01, + "learning_rate": 2.5002184837229626e-06, + "loss": 1.5572, + "step": 22933 + }, + { + "epoch": 5.01, + "learning_rate": 2.5001092418614816e-06, + "loss": 0.7099, + "step": 22934 + }, + { + "epoch": 5.01, + "learning_rate": 2.5e-06, + "loss": 1.2208, + "step": 22935 + }, + { + "epoch": 5.01, + "learning_rate": 2.499890758138519e-06, + "loss": 0.2305, + "step": 22936 + }, + { + "epoch": 5.01, + "learning_rate": 2.4997815162770374e-06, + "loss": 1.0945, + "step": 22937 + }, + { + "epoch": 5.01, + "learning_rate": 2.4996722744155565e-06, + "loss": 1.9102, + "step": 22938 + }, + { + "epoch": 5.01, + "learning_rate": 2.4995630325540746e-06, + "loss": 1.6647, + "step": 22939 + }, + { + "epoch": 5.01, + "learning_rate": 2.4994537906925937e-06, + "loss": 0.4255, + "step": 22940 + }, + { + "epoch": 5.01, + "learning_rate": 2.4993445488311123e-06, + "loss": 1.3078, + "step": 22941 + }, + { + "epoch": 5.01, + "learning_rate": 2.499235306969631e-06, + "loss": 1.0317, + "step": 22942 + }, + { + "epoch": 5.01, + "learning_rate": 2.49912606510815e-06, + "loss": 1.5818, + "step": 22943 + }, + { + "epoch": 5.01, + "learning_rate": 2.499016823246668e-06, + "loss": 1.9975, + "step": 22944 + }, + { + "epoch": 5.01, + "learning_rate": 2.498907581385187e-06, + "loss": 1.0629, + "step": 22945 + }, + { + "epoch": 5.01, + "learning_rate": 2.4987983395237058e-06, + "loss": 2.1717, + "step": 22946 + }, + { + "epoch": 5.01, + "learning_rate": 2.4986890976622244e-06, + "loss": 1.7138, + "step": 22947 + }, + { + "epoch": 5.01, + "learning_rate": 2.498579855800743e-06, + "loss": 1.6217, + "step": 22948 + }, + { + "epoch": 5.01, + "learning_rate": 2.4984706139392616e-06, + "loss": 1.3398, + "step": 22949 + }, + { + "epoch": 5.01, + "learning_rate": 2.4983613720777806e-06, + "loss": 1.3675, + "step": 22950 + }, + { + "epoch": 5.01, + "learning_rate": 2.498252130216299e-06, + "loss": 1.596, + "step": 22951 + }, + { + "epoch": 5.01, + "learning_rate": 2.498142888354818e-06, + "loss": 0.9023, + "step": 22952 + }, + { + "epoch": 5.01, + "learning_rate": 2.4980336464933365e-06, + "loss": 1.1772, + "step": 22953 + }, + { + "epoch": 5.01, + "learning_rate": 2.497924404631855e-06, + "loss": 1.2677, + "step": 22954 + }, + { + "epoch": 5.01, + "learning_rate": 2.497815162770374e-06, + "loss": 0.4733, + "step": 22955 + }, + { + "epoch": 5.01, + "learning_rate": 2.4977059209088923e-06, + "loss": 2.8596, + "step": 22956 + }, + { + "epoch": 5.01, + "learning_rate": 2.4975966790474113e-06, + "loss": 2.1632, + "step": 22957 + }, + { + "epoch": 5.01, + "learning_rate": 2.49748743718593e-06, + "loss": 0.8107, + "step": 22958 + }, + { + "epoch": 5.01, + "learning_rate": 2.4973781953244486e-06, + "loss": 1.3278, + "step": 22959 + }, + { + "epoch": 5.01, + "learning_rate": 2.497268953462967e-06, + "loss": 1.4864, + "step": 22960 + }, + { + "epoch": 5.01, + "learning_rate": 2.4971597116014858e-06, + "loss": 1.2617, + "step": 22961 + }, + { + "epoch": 5.01, + "learning_rate": 2.497050469740005e-06, + "loss": 0.8668, + "step": 22962 + }, + { + "epoch": 5.01, + "learning_rate": 2.4969412278785234e-06, + "loss": 1.1291, + "step": 22963 + }, + { + "epoch": 5.01, + "learning_rate": 2.496831986017042e-06, + "loss": 1.2879, + "step": 22964 + }, + { + "epoch": 5.01, + "learning_rate": 2.4967227441555607e-06, + "loss": 1.879, + "step": 22965 + }, + { + "epoch": 5.01, + "learning_rate": 2.4966135022940793e-06, + "loss": 1.628, + "step": 22966 + }, + { + "epoch": 5.01, + "learning_rate": 2.496504260432598e-06, + "loss": 2.5049, + "step": 22967 + }, + { + "epoch": 5.01, + "learning_rate": 2.4963950185711165e-06, + "loss": 1.0664, + "step": 22968 + }, + { + "epoch": 5.01, + "learning_rate": 2.4962857767096355e-06, + "loss": 0.3132, + "step": 22969 + }, + { + "epoch": 5.01, + "learning_rate": 2.496176534848154e-06, + "loss": 1.8571, + "step": 22970 + }, + { + "epoch": 5.01, + "learning_rate": 2.4960672929866727e-06, + "loss": 1.7207, + "step": 22971 + }, + { + "epoch": 5.01, + "learning_rate": 2.4959580511251914e-06, + "loss": 1.4857, + "step": 22972 + }, + { + "epoch": 5.01, + "learning_rate": 2.49584880926371e-06, + "loss": 1.6502, + "step": 22973 + }, + { + "epoch": 5.01, + "learning_rate": 2.4957395674022286e-06, + "loss": 2.1711, + "step": 22974 + }, + { + "epoch": 5.01, + "learning_rate": 2.4956303255407476e-06, + "loss": 0.415, + "step": 22975 + }, + { + "epoch": 5.01, + "learning_rate": 2.4955210836792662e-06, + "loss": 1.4296, + "step": 22976 + }, + { + "epoch": 5.01, + "learning_rate": 2.495411841817785e-06, + "loss": 0.6591, + "step": 22977 + }, + { + "epoch": 5.01, + "learning_rate": 2.4953025999563034e-06, + "loss": 1.4321, + "step": 22978 + }, + { + "epoch": 5.02, + "learning_rate": 2.495193358094822e-06, + "loss": 1.9943, + "step": 22979 + }, + { + "epoch": 5.02, + "learning_rate": 2.4950841162333407e-06, + "loss": 2.2921, + "step": 22980 + }, + { + "epoch": 5.02, + "learning_rate": 2.4949748743718593e-06, + "loss": 1.4466, + "step": 22981 + }, + { + "epoch": 5.02, + "learning_rate": 2.4948656325103783e-06, + "loss": 1.1549, + "step": 22982 + }, + { + "epoch": 5.02, + "learning_rate": 2.494756390648897e-06, + "loss": 1.5257, + "step": 22983 + }, + { + "epoch": 5.02, + "learning_rate": 2.4946471487874155e-06, + "loss": 2.2212, + "step": 22984 + }, + { + "epoch": 5.02, + "learning_rate": 2.494537906925934e-06, + "loss": 0.7908, + "step": 22985 + }, + { + "epoch": 5.02, + "learning_rate": 2.4944286650644528e-06, + "loss": 1.8741, + "step": 22986 + }, + { + "epoch": 5.02, + "learning_rate": 2.4943194232029718e-06, + "loss": 2.9014, + "step": 22987 + }, + { + "epoch": 5.02, + "learning_rate": 2.4942101813414904e-06, + "loss": 1.648, + "step": 22988 + }, + { + "epoch": 5.02, + "learning_rate": 2.494100939480009e-06, + "loss": 1.8176, + "step": 22989 + }, + { + "epoch": 5.02, + "learning_rate": 2.4939916976185276e-06, + "loss": 2.9372, + "step": 22990 + }, + { + "epoch": 5.02, + "learning_rate": 2.4938824557570462e-06, + "loss": 1.5478, + "step": 22991 + }, + { + "epoch": 5.02, + "learning_rate": 2.493773213895565e-06, + "loss": 2.7425, + "step": 22992 + }, + { + "epoch": 5.02, + "learning_rate": 2.4936639720340835e-06, + "loss": 0.9003, + "step": 22993 + }, + { + "epoch": 5.02, + "learning_rate": 2.4935547301726025e-06, + "loss": 1.1155, + "step": 22994 + }, + { + "epoch": 5.02, + "learning_rate": 2.493445488311121e-06, + "loss": 2.6871, + "step": 22995 + }, + { + "epoch": 5.02, + "learning_rate": 2.4933362464496397e-06, + "loss": 0.37, + "step": 22996 + }, + { + "epoch": 5.02, + "learning_rate": 2.4932270045881583e-06, + "loss": 1.526, + "step": 22997 + }, + { + "epoch": 5.02, + "learning_rate": 2.493117762726677e-06, + "loss": 1.9239, + "step": 22998 + }, + { + "epoch": 5.02, + "learning_rate": 2.493008520865196e-06, + "loss": 1.9774, + "step": 22999 + }, + { + "epoch": 5.02, + "learning_rate": 2.492899279003714e-06, + "loss": 0.8035, + "step": 23000 + }, + { + "epoch": 5.02, + "learning_rate": 2.492790037142233e-06, + "loss": 2.367, + "step": 23001 + }, + { + "epoch": 5.02, + "learning_rate": 2.492680795280752e-06, + "loss": 2.1315, + "step": 23002 + }, + { + "epoch": 5.02, + "learning_rate": 2.4925715534192704e-06, + "loss": 1.2878, + "step": 23003 + }, + { + "epoch": 5.02, + "learning_rate": 2.492462311557789e-06, + "loss": 2.3417, + "step": 23004 + }, + { + "epoch": 5.02, + "learning_rate": 2.4923530696963076e-06, + "loss": 3.2477, + "step": 23005 + }, + { + "epoch": 5.02, + "learning_rate": 2.4922438278348267e-06, + "loss": 1.4174, + "step": 23006 + }, + { + "epoch": 5.02, + "learning_rate": 2.4921345859733453e-06, + "loss": 1.1674, + "step": 23007 + }, + { + "epoch": 5.02, + "learning_rate": 2.492025344111864e-06, + "loss": 1.7174, + "step": 23008 + }, + { + "epoch": 5.02, + "learning_rate": 2.4919161022503825e-06, + "loss": 0.8149, + "step": 23009 + }, + { + "epoch": 5.02, + "learning_rate": 2.491806860388901e-06, + "loss": 3.2643, + "step": 23010 + }, + { + "epoch": 5.02, + "learning_rate": 2.49169761852742e-06, + "loss": 0.4447, + "step": 23011 + }, + { + "epoch": 5.02, + "learning_rate": 2.4915883766659383e-06, + "loss": 2.4238, + "step": 23012 + }, + { + "epoch": 5.02, + "learning_rate": 2.4914791348044574e-06, + "loss": 1.0591, + "step": 23013 + }, + { + "epoch": 5.02, + "learning_rate": 2.491369892942976e-06, + "loss": 2.0027, + "step": 23014 + }, + { + "epoch": 5.02, + "learning_rate": 2.4912606510814946e-06, + "loss": 0.9558, + "step": 23015 + }, + { + "epoch": 5.02, + "learning_rate": 2.4911514092200136e-06, + "loss": 0.7685, + "step": 23016 + }, + { + "epoch": 5.02, + "learning_rate": 2.491042167358532e-06, + "loss": 2.457, + "step": 23017 + }, + { + "epoch": 5.02, + "learning_rate": 2.490932925497051e-06, + "loss": 1.297, + "step": 23018 + }, + { + "epoch": 5.02, + "learning_rate": 2.4908236836355695e-06, + "loss": 1.281, + "step": 23019 + }, + { + "epoch": 5.02, + "learning_rate": 2.490714441774088e-06, + "loss": 0.8716, + "step": 23020 + }, + { + "epoch": 5.02, + "learning_rate": 2.4906051999126067e-06, + "loss": 2.3009, + "step": 23021 + }, + { + "epoch": 5.02, + "learning_rate": 2.4904959580511253e-06, + "loss": 1.2217, + "step": 23022 + }, + { + "epoch": 5.02, + "learning_rate": 2.4903867161896443e-06, + "loss": 0.9414, + "step": 23023 + }, + { + "epoch": 5.02, + "learning_rate": 2.4902774743281625e-06, + "loss": 0.9464, + "step": 23024 + }, + { + "epoch": 5.03, + "learning_rate": 2.4901682324666815e-06, + "loss": 0.9218, + "step": 23025 + }, + { + "epoch": 5.03, + "learning_rate": 2.4900589906052e-06, + "loss": 1.9572, + "step": 23026 + }, + { + "epoch": 5.03, + "learning_rate": 2.4899497487437188e-06, + "loss": 0.7744, + "step": 23027 + }, + { + "epoch": 5.03, + "learning_rate": 2.489840506882238e-06, + "loss": 2.9743, + "step": 23028 + }, + { + "epoch": 5.03, + "learning_rate": 2.489731265020756e-06, + "loss": 1.0696, + "step": 23029 + }, + { + "epoch": 5.03, + "learning_rate": 2.489622023159275e-06, + "loss": 0.7398, + "step": 23030 + }, + { + "epoch": 5.03, + "learning_rate": 2.4895127812977936e-06, + "loss": 1.4968, + "step": 23031 + }, + { + "epoch": 5.03, + "learning_rate": 2.4894035394363122e-06, + "loss": 1.9666, + "step": 23032 + }, + { + "epoch": 5.03, + "learning_rate": 2.489294297574831e-06, + "loss": 1.5457, + "step": 23033 + }, + { + "epoch": 5.03, + "learning_rate": 2.4891850557133495e-06, + "loss": 2.7563, + "step": 23034 + }, + { + "epoch": 5.03, + "learning_rate": 2.4890758138518685e-06, + "loss": 0.668, + "step": 23035 + }, + { + "epoch": 5.03, + "learning_rate": 2.4889665719903867e-06, + "loss": 0.6701, + "step": 23036 + }, + { + "epoch": 5.03, + "learning_rate": 2.4888573301289057e-06, + "loss": 1.3183, + "step": 23037 + }, + { + "epoch": 5.03, + "learning_rate": 2.4887480882674243e-06, + "loss": 2.4414, + "step": 23038 + }, + { + "epoch": 5.03, + "learning_rate": 2.488638846405943e-06, + "loss": 2.4009, + "step": 23039 + }, + { + "epoch": 5.03, + "learning_rate": 2.4885296045444616e-06, + "loss": 1.7708, + "step": 23040 + }, + { + "epoch": 5.03, + "learning_rate": 2.48842036268298e-06, + "loss": 0.5428, + "step": 23041 + }, + { + "epoch": 5.03, + "learning_rate": 2.488311120821499e-06, + "loss": 1.847, + "step": 23042 + }, + { + "epoch": 5.03, + "learning_rate": 2.488201878960018e-06, + "loss": 0.9523, + "step": 23043 + }, + { + "epoch": 5.03, + "learning_rate": 2.4880926370985364e-06, + "loss": 0.9851, + "step": 23044 + }, + { + "epoch": 5.03, + "learning_rate": 2.487983395237055e-06, + "loss": 1.325, + "step": 23045 + }, + { + "epoch": 5.03, + "learning_rate": 2.4878741533755736e-06, + "loss": 2.8848, + "step": 23046 + }, + { + "epoch": 5.03, + "learning_rate": 2.4877649115140923e-06, + "loss": 1.5323, + "step": 23047 + }, + { + "epoch": 5.03, + "learning_rate": 2.487655669652611e-06, + "loss": 1.1189, + "step": 23048 + }, + { + "epoch": 5.03, + "learning_rate": 2.48754642779113e-06, + "loss": 1.8639, + "step": 23049 + }, + { + "epoch": 5.03, + "learning_rate": 2.4874371859296485e-06, + "loss": 2.4233, + "step": 23050 + }, + { + "epoch": 5.03, + "learning_rate": 2.487327944068167e-06, + "loss": 1.0965, + "step": 23051 + }, + { + "epoch": 5.03, + "learning_rate": 2.4872187022066857e-06, + "loss": 2.0273, + "step": 23052 + }, + { + "epoch": 5.03, + "learning_rate": 2.4871094603452043e-06, + "loss": 2.4468, + "step": 23053 + }, + { + "epoch": 5.03, + "learning_rate": 2.487000218483723e-06, + "loss": 1.2344, + "step": 23054 + }, + { + "epoch": 5.03, + "learning_rate": 2.486890976622242e-06, + "loss": 0.9316, + "step": 23055 + }, + { + "epoch": 5.03, + "learning_rate": 2.4867817347607606e-06, + "loss": 2.3633, + "step": 23056 + }, + { + "epoch": 5.03, + "learning_rate": 2.4866724928992792e-06, + "loss": 1.5848, + "step": 23057 + }, + { + "epoch": 5.03, + "learning_rate": 2.486563251037798e-06, + "loss": 2.5019, + "step": 23058 + }, + { + "epoch": 5.03, + "learning_rate": 2.4864540091763164e-06, + "loss": 1.377, + "step": 23059 + }, + { + "epoch": 5.03, + "learning_rate": 2.4863447673148355e-06, + "loss": 1.6595, + "step": 23060 + }, + { + "epoch": 5.03, + "learning_rate": 2.486235525453354e-06, + "loss": 1.7404, + "step": 23061 + }, + { + "epoch": 5.03, + "learning_rate": 2.4861262835918727e-06, + "loss": 1.1974, + "step": 23062 + }, + { + "epoch": 5.03, + "learning_rate": 2.4860170417303913e-06, + "loss": 1.1998, + "step": 23063 + }, + { + "epoch": 5.03, + "learning_rate": 2.48590779986891e-06, + "loss": 1.3726, + "step": 23064 + }, + { + "epoch": 5.03, + "learning_rate": 2.4857985580074285e-06, + "loss": 1.6504, + "step": 23065 + }, + { + "epoch": 5.03, + "learning_rate": 2.485689316145947e-06, + "loss": 1.2122, + "step": 23066 + }, + { + "epoch": 5.03, + "learning_rate": 2.485580074284466e-06, + "loss": 0.8166, + "step": 23067 + }, + { + "epoch": 5.03, + "learning_rate": 2.4854708324229848e-06, + "loss": 2.4771, + "step": 23068 + }, + { + "epoch": 5.03, + "learning_rate": 2.4853615905615034e-06, + "loss": 2.4089, + "step": 23069 + }, + { + "epoch": 5.03, + "learning_rate": 2.485252348700022e-06, + "loss": 2.833, + "step": 23070 + }, + { + "epoch": 5.04, + "learning_rate": 2.4851431068385406e-06, + "loss": 1.4641, + "step": 23071 + }, + { + "epoch": 5.04, + "learning_rate": 2.4850338649770597e-06, + "loss": 0.2941, + "step": 23072 + }, + { + "epoch": 5.04, + "learning_rate": 2.484924623115578e-06, + "loss": 1.0347, + "step": 23073 + }, + { + "epoch": 5.04, + "learning_rate": 2.484815381254097e-06, + "loss": 0.2814, + "step": 23074 + }, + { + "epoch": 5.04, + "learning_rate": 2.4847061393926155e-06, + "loss": 1.214, + "step": 23075 + }, + { + "epoch": 5.04, + "learning_rate": 2.484596897531134e-06, + "loss": 2.3051, + "step": 23076 + }, + { + "epoch": 5.04, + "learning_rate": 2.4844876556696527e-06, + "loss": 1.0094, + "step": 23077 + }, + { + "epoch": 5.04, + "learning_rate": 2.4843784138081713e-06, + "loss": 2.6208, + "step": 23078 + }, + { + "epoch": 5.04, + "learning_rate": 2.4842691719466904e-06, + "loss": 1.9, + "step": 23079 + }, + { + "epoch": 5.04, + "learning_rate": 2.4841599300852085e-06, + "loss": 0.1845, + "step": 23080 + }, + { + "epoch": 5.04, + "learning_rate": 2.4840506882237276e-06, + "loss": 1.6586, + "step": 23081 + }, + { + "epoch": 5.04, + "learning_rate": 2.483941446362246e-06, + "loss": 0.6827, + "step": 23082 + }, + { + "epoch": 5.04, + "learning_rate": 2.483832204500765e-06, + "loss": 0.3766, + "step": 23083 + }, + { + "epoch": 5.04, + "learning_rate": 2.483722962639284e-06, + "loss": 1.1643, + "step": 23084 + }, + { + "epoch": 5.04, + "learning_rate": 2.483613720777802e-06, + "loss": 1.8437, + "step": 23085 + }, + { + "epoch": 5.04, + "learning_rate": 2.483504478916321e-06, + "loss": 1.5759, + "step": 23086 + }, + { + "epoch": 5.04, + "learning_rate": 2.4833952370548397e-06, + "loss": 1.0891, + "step": 23087 + }, + { + "epoch": 5.04, + "learning_rate": 2.4832859951933583e-06, + "loss": 0.4665, + "step": 23088 + }, + { + "epoch": 5.04, + "learning_rate": 2.483176753331877e-06, + "loss": 1.8973, + "step": 23089 + }, + { + "epoch": 5.04, + "learning_rate": 2.4830675114703955e-06, + "loss": 0.7283, + "step": 23090 + }, + { + "epoch": 5.04, + "learning_rate": 2.4829582696089145e-06, + "loss": 0.9764, + "step": 23091 + }, + { + "epoch": 5.04, + "learning_rate": 2.4828490277474327e-06, + "loss": 2.6232, + "step": 23092 + }, + { + "epoch": 5.04, + "learning_rate": 2.4827397858859518e-06, + "loss": 1.5014, + "step": 23093 + }, + { + "epoch": 5.04, + "learning_rate": 2.4826305440244704e-06, + "loss": 0.7354, + "step": 23094 + }, + { + "epoch": 5.04, + "learning_rate": 2.482521302162989e-06, + "loss": 0.9913, + "step": 23095 + }, + { + "epoch": 5.04, + "learning_rate": 2.482412060301508e-06, + "loss": 0.7592, + "step": 23096 + }, + { + "epoch": 5.04, + "learning_rate": 2.482302818440026e-06, + "loss": 2.7198, + "step": 23097 + }, + { + "epoch": 5.04, + "learning_rate": 2.4821935765785452e-06, + "loss": 1.3853, + "step": 23098 + }, + { + "epoch": 5.04, + "learning_rate": 2.482084334717064e-06, + "loss": 1.7964, + "step": 23099 + }, + { + "epoch": 5.04, + "learning_rate": 2.4819750928555825e-06, + "loss": 0.2885, + "step": 23100 + }, + { + "epoch": 5.04, + "learning_rate": 2.481865850994101e-06, + "loss": 1.5202, + "step": 23101 + }, + { + "epoch": 5.04, + "learning_rate": 2.4817566091326197e-06, + "loss": 1.9538, + "step": 23102 + }, + { + "epoch": 5.04, + "learning_rate": 2.4816473672711387e-06, + "loss": 1.1572, + "step": 23103 + }, + { + "epoch": 5.04, + "learning_rate": 2.481538125409657e-06, + "loss": 1.4571, + "step": 23104 + }, + { + "epoch": 5.04, + "learning_rate": 2.481428883548176e-06, + "loss": 0.6376, + "step": 23105 + }, + { + "epoch": 5.04, + "learning_rate": 2.4813196416866945e-06, + "loss": 0.4804, + "step": 23106 + }, + { + "epoch": 5.04, + "learning_rate": 2.481210399825213e-06, + "loss": 1.7447, + "step": 23107 + }, + { + "epoch": 5.04, + "learning_rate": 2.481101157963732e-06, + "loss": 2.6966, + "step": 23108 + }, + { + "epoch": 5.04, + "learning_rate": 2.4809919161022504e-06, + "loss": 1.0223, + "step": 23109 + }, + { + "epoch": 5.04, + "learning_rate": 2.4808826742407694e-06, + "loss": 1.5363, + "step": 23110 + }, + { + "epoch": 5.04, + "learning_rate": 2.480773432379288e-06, + "loss": 2.055, + "step": 23111 + }, + { + "epoch": 5.04, + "learning_rate": 2.4806641905178066e-06, + "loss": 2.248, + "step": 23112 + }, + { + "epoch": 5.04, + "learning_rate": 2.4805549486563252e-06, + "loss": 2.6756, + "step": 23113 + }, + { + "epoch": 5.04, + "learning_rate": 2.480445706794844e-06, + "loss": 1.9719, + "step": 23114 + }, + { + "epoch": 5.04, + "learning_rate": 2.480336464933363e-06, + "loss": 3.7822, + "step": 23115 + }, + { + "epoch": 5.04, + "learning_rate": 2.4802272230718815e-06, + "loss": 1.2833, + "step": 23116 + }, + { + "epoch": 5.05, + "learning_rate": 2.4801179812104e-06, + "loss": 0.4931, + "step": 23117 + }, + { + "epoch": 5.05, + "learning_rate": 2.4800087393489187e-06, + "loss": 0.9107, + "step": 23118 + }, + { + "epoch": 5.05, + "learning_rate": 2.4798994974874373e-06, + "loss": 1.757, + "step": 23119 + }, + { + "epoch": 5.05, + "learning_rate": 2.479790255625956e-06, + "loss": 1.9371, + "step": 23120 + }, + { + "epoch": 5.05, + "learning_rate": 2.4796810137644746e-06, + "loss": 0.466, + "step": 23121 + }, + { + "epoch": 5.05, + "learning_rate": 2.4795717719029936e-06, + "loss": 0.8364, + "step": 23122 + }, + { + "epoch": 5.05, + "learning_rate": 2.479462530041512e-06, + "loss": 1.7576, + "step": 23123 + }, + { + "epoch": 5.05, + "learning_rate": 2.479353288180031e-06, + "loss": 0.902, + "step": 23124 + }, + { + "epoch": 5.05, + "learning_rate": 2.4792440463185494e-06, + "loss": 1.9553, + "step": 23125 + }, + { + "epoch": 5.05, + "learning_rate": 2.479134804457068e-06, + "loss": 2.7389, + "step": 23126 + }, + { + "epoch": 5.05, + "learning_rate": 2.479025562595587e-06, + "loss": 3.0846, + "step": 23127 + }, + { + "epoch": 5.05, + "learning_rate": 2.4789163207341057e-06, + "loss": 2.4551, + "step": 23128 + }, + { + "epoch": 5.05, + "learning_rate": 2.4788070788726243e-06, + "loss": 2.4239, + "step": 23129 + }, + { + "epoch": 5.05, + "learning_rate": 2.478697837011143e-06, + "loss": 1.4544, + "step": 23130 + }, + { + "epoch": 5.05, + "learning_rate": 2.4785885951496615e-06, + "loss": 0.2857, + "step": 23131 + }, + { + "epoch": 5.05, + "learning_rate": 2.47847935328818e-06, + "loss": 1.5342, + "step": 23132 + }, + { + "epoch": 5.05, + "learning_rate": 2.4783701114266987e-06, + "loss": 1.2429, + "step": 23133 + }, + { + "epoch": 5.05, + "learning_rate": 2.4782608695652178e-06, + "loss": 1.2003, + "step": 23134 + }, + { + "epoch": 5.05, + "learning_rate": 2.4781516277037364e-06, + "loss": 1.1751, + "step": 23135 + }, + { + "epoch": 5.05, + "learning_rate": 2.478042385842255e-06, + "loss": 1.3053, + "step": 23136 + }, + { + "epoch": 5.05, + "learning_rate": 2.4779331439807736e-06, + "loss": 0.3657, + "step": 23137 + }, + { + "epoch": 5.05, + "learning_rate": 2.477823902119292e-06, + "loss": 1.0405, + "step": 23138 + }, + { + "epoch": 5.05, + "learning_rate": 2.477714660257811e-06, + "loss": 1.7252, + "step": 23139 + }, + { + "epoch": 5.05, + "learning_rate": 2.47760541839633e-06, + "loss": 1.5135, + "step": 23140 + }, + { + "epoch": 5.05, + "learning_rate": 2.4774961765348485e-06, + "loss": 3.3271, + "step": 23141 + }, + { + "epoch": 5.05, + "learning_rate": 2.477386934673367e-06, + "loss": 2.0894, + "step": 23142 + }, + { + "epoch": 5.05, + "learning_rate": 2.4772776928118857e-06, + "loss": 2.4666, + "step": 23143 + }, + { + "epoch": 5.05, + "learning_rate": 2.4771684509504043e-06, + "loss": 1.0811, + "step": 23144 + }, + { + "epoch": 5.05, + "learning_rate": 2.477059209088923e-06, + "loss": 1.5873, + "step": 23145 + }, + { + "epoch": 5.05, + "learning_rate": 2.4769499672274415e-06, + "loss": 1.2094, + "step": 23146 + }, + { + "epoch": 5.05, + "learning_rate": 2.4768407253659606e-06, + "loss": 0.7214, + "step": 23147 + }, + { + "epoch": 5.05, + "learning_rate": 2.476731483504479e-06, + "loss": 1.851, + "step": 23148 + }, + { + "epoch": 5.05, + "learning_rate": 2.4766222416429978e-06, + "loss": 0.7975, + "step": 23149 + }, + { + "epoch": 5.05, + "learning_rate": 2.4765129997815164e-06, + "loss": 0.5387, + "step": 23150 + }, + { + "epoch": 5.05, + "learning_rate": 2.476403757920035e-06, + "loss": 1.2884, + "step": 23151 + }, + { + "epoch": 5.05, + "learning_rate": 2.476294516058554e-06, + "loss": 1.8229, + "step": 23152 + }, + { + "epoch": 5.05, + "learning_rate": 2.4761852741970722e-06, + "loss": 3.4667, + "step": 23153 + }, + { + "epoch": 5.05, + "learning_rate": 2.4760760323355913e-06, + "loss": 1.2484, + "step": 23154 + }, + { + "epoch": 5.05, + "learning_rate": 2.47596679047411e-06, + "loss": 1.5074, + "step": 23155 + }, + { + "epoch": 5.05, + "learning_rate": 2.4758575486126285e-06, + "loss": 2.9567, + "step": 23156 + }, + { + "epoch": 5.05, + "learning_rate": 2.475748306751147e-06, + "loss": 0.3178, + "step": 23157 + }, + { + "epoch": 5.05, + "learning_rate": 2.4756390648896657e-06, + "loss": 0.9661, + "step": 23158 + }, + { + "epoch": 5.05, + "learning_rate": 2.4755298230281847e-06, + "loss": 1.0832, + "step": 23159 + }, + { + "epoch": 5.05, + "learning_rate": 2.4754205811667033e-06, + "loss": 1.1687, + "step": 23160 + }, + { + "epoch": 5.05, + "learning_rate": 2.475311339305222e-06, + "loss": 0.3887, + "step": 23161 + }, + { + "epoch": 5.05, + "learning_rate": 2.4752020974437406e-06, + "loss": 1.5323, + "step": 23162 + }, + { + "epoch": 5.06, + "learning_rate": 2.475092855582259e-06, + "loss": 3.3015, + "step": 23163 + }, + { + "epoch": 5.06, + "learning_rate": 2.4749836137207782e-06, + "loss": 1.6745, + "step": 23164 + }, + { + "epoch": 5.06, + "learning_rate": 2.4748743718592964e-06, + "loss": 1.9482, + "step": 23165 + }, + { + "epoch": 5.06, + "learning_rate": 2.4747651299978154e-06, + "loss": 2.5561, + "step": 23166 + }, + { + "epoch": 5.06, + "learning_rate": 2.474655888136334e-06, + "loss": 3.066, + "step": 23167 + }, + { + "epoch": 5.06, + "learning_rate": 2.4745466462748527e-06, + "loss": 1.8125, + "step": 23168 + }, + { + "epoch": 5.06, + "learning_rate": 2.4744374044133717e-06, + "loss": 2.7587, + "step": 23169 + }, + { + "epoch": 5.06, + "learning_rate": 2.47432816255189e-06, + "loss": 0.5591, + "step": 23170 + }, + { + "epoch": 5.06, + "learning_rate": 2.474218920690409e-06, + "loss": 1.5083, + "step": 23171 + }, + { + "epoch": 5.06, + "learning_rate": 2.4741096788289275e-06, + "loss": 2.0107, + "step": 23172 + }, + { + "epoch": 5.06, + "learning_rate": 2.474000436967446e-06, + "loss": 0.7357, + "step": 23173 + }, + { + "epoch": 5.06, + "learning_rate": 2.4738911951059647e-06, + "loss": 2.1549, + "step": 23174 + }, + { + "epoch": 5.06, + "learning_rate": 2.4737819532444834e-06, + "loss": 2.0912, + "step": 23175 + }, + { + "epoch": 5.06, + "learning_rate": 2.4736727113830024e-06, + "loss": 0.557, + "step": 23176 + }, + { + "epoch": 5.06, + "learning_rate": 2.4735634695215206e-06, + "loss": 0.4323, + "step": 23177 + }, + { + "epoch": 5.06, + "learning_rate": 2.4734542276600396e-06, + "loss": 1.3984, + "step": 23178 + }, + { + "epoch": 5.06, + "learning_rate": 2.4733449857985582e-06, + "loss": 1.4281, + "step": 23179 + }, + { + "epoch": 5.06, + "learning_rate": 2.473235743937077e-06, + "loss": 3.4959, + "step": 23180 + }, + { + "epoch": 5.06, + "learning_rate": 2.473126502075596e-06, + "loss": 0.8636, + "step": 23181 + }, + { + "epoch": 5.06, + "learning_rate": 2.473017260214114e-06, + "loss": 0.9451, + "step": 23182 + }, + { + "epoch": 5.06, + "learning_rate": 2.472908018352633e-06, + "loss": 1.8487, + "step": 23183 + }, + { + "epoch": 5.06, + "learning_rate": 2.4727987764911517e-06, + "loss": 1.6218, + "step": 23184 + }, + { + "epoch": 5.06, + "learning_rate": 2.4726895346296703e-06, + "loss": 0.4785, + "step": 23185 + }, + { + "epoch": 5.06, + "learning_rate": 2.472580292768189e-06, + "loss": 0.588, + "step": 23186 + }, + { + "epoch": 5.06, + "learning_rate": 2.4724710509067075e-06, + "loss": 0.975, + "step": 23187 + }, + { + "epoch": 5.06, + "learning_rate": 2.4723618090452266e-06, + "loss": 1.5308, + "step": 23188 + }, + { + "epoch": 5.06, + "learning_rate": 2.4722525671837448e-06, + "loss": 0.6008, + "step": 23189 + }, + { + "epoch": 5.06, + "learning_rate": 2.472143325322264e-06, + "loss": 0.6166, + "step": 23190 + }, + { + "epoch": 5.06, + "learning_rate": 2.4720340834607824e-06, + "loss": 0.8953, + "step": 23191 + }, + { + "epoch": 5.06, + "learning_rate": 2.471924841599301e-06, + "loss": 1.0518, + "step": 23192 + }, + { + "epoch": 5.06, + "learning_rate": 2.47181559973782e-06, + "loss": 2.4685, + "step": 23193 + }, + { + "epoch": 5.06, + "learning_rate": 2.4717063578763382e-06, + "loss": 0.8908, + "step": 23194 + }, + { + "epoch": 5.06, + "learning_rate": 2.4715971160148573e-06, + "loss": 1.2058, + "step": 23195 + }, + { + "epoch": 5.06, + "learning_rate": 2.471487874153376e-06, + "loss": 1.3482, + "step": 23196 + }, + { + "epoch": 5.06, + "learning_rate": 2.4713786322918945e-06, + "loss": 1.4358, + "step": 23197 + }, + { + "epoch": 5.06, + "learning_rate": 2.471269390430413e-06, + "loss": 1.2583, + "step": 23198 + }, + { + "epoch": 5.06, + "learning_rate": 2.4711601485689317e-06, + "loss": 1.5008, + "step": 23199 + }, + { + "epoch": 5.06, + "learning_rate": 2.4710509067074508e-06, + "loss": 1.5906, + "step": 23200 + }, + { + "epoch": 5.06, + "learning_rate": 2.470941664845969e-06, + "loss": 1.8534, + "step": 23201 + }, + { + "epoch": 5.06, + "learning_rate": 2.470832422984488e-06, + "loss": 1.7487, + "step": 23202 + }, + { + "epoch": 5.06, + "learning_rate": 2.4707231811230066e-06, + "loss": 1.1161, + "step": 23203 + }, + { + "epoch": 5.06, + "learning_rate": 2.470613939261525e-06, + "loss": 1.4687, + "step": 23204 + }, + { + "epoch": 5.06, + "learning_rate": 2.470504697400044e-06, + "loss": 2.0332, + "step": 23205 + }, + { + "epoch": 5.06, + "learning_rate": 2.4703954555385624e-06, + "loss": 0.7091, + "step": 23206 + }, + { + "epoch": 5.06, + "learning_rate": 2.4702862136770815e-06, + "loss": 2.0295, + "step": 23207 + }, + { + "epoch": 5.07, + "learning_rate": 2.4701769718156e-06, + "loss": 0.1207, + "step": 23208 + }, + { + "epoch": 5.07, + "learning_rate": 2.4700677299541187e-06, + "loss": 0.9507, + "step": 23209 + }, + { + "epoch": 5.07, + "learning_rate": 2.4699584880926373e-06, + "loss": 1.9775, + "step": 23210 + }, + { + "epoch": 5.07, + "learning_rate": 2.469849246231156e-06, + "loss": 0.4925, + "step": 23211 + }, + { + "epoch": 5.07, + "learning_rate": 2.4697400043696745e-06, + "loss": 1.537, + "step": 23212 + }, + { + "epoch": 5.07, + "learning_rate": 2.4696307625081935e-06, + "loss": 1.3773, + "step": 23213 + }, + { + "epoch": 5.07, + "learning_rate": 2.469521520646712e-06, + "loss": 1.86, + "step": 23214 + }, + { + "epoch": 5.07, + "learning_rate": 2.4694122787852308e-06, + "loss": 2.3796, + "step": 23215 + }, + { + "epoch": 5.07, + "learning_rate": 2.4693030369237494e-06, + "loss": 2.1123, + "step": 23216 + }, + { + "epoch": 5.07, + "learning_rate": 2.469193795062268e-06, + "loss": 1.6349, + "step": 23217 + }, + { + "epoch": 5.07, + "learning_rate": 2.4690845532007866e-06, + "loss": 0.236, + "step": 23218 + }, + { + "epoch": 5.07, + "learning_rate": 2.468975311339305e-06, + "loss": 1.8887, + "step": 23219 + }, + { + "epoch": 5.07, + "learning_rate": 2.4688660694778242e-06, + "loss": 0.7871, + "step": 23220 + }, + { + "epoch": 5.07, + "learning_rate": 2.468756827616343e-06, + "loss": 0.7741, + "step": 23221 + }, + { + "epoch": 5.07, + "learning_rate": 2.4686475857548615e-06, + "loss": 0.6944, + "step": 23222 + }, + { + "epoch": 5.07, + "learning_rate": 2.46853834389338e-06, + "loss": 1.6836, + "step": 23223 + }, + { + "epoch": 5.07, + "learning_rate": 2.4684291020318987e-06, + "loss": 1.4394, + "step": 23224 + }, + { + "epoch": 5.07, + "learning_rate": 2.4683198601704177e-06, + "loss": 1.2137, + "step": 23225 + }, + { + "epoch": 5.07, + "learning_rate": 2.4682106183089363e-06, + "loss": 1.5173, + "step": 23226 + }, + { + "epoch": 5.07, + "learning_rate": 2.468101376447455e-06, + "loss": 1.8232, + "step": 23227 + }, + { + "epoch": 5.07, + "learning_rate": 2.4679921345859736e-06, + "loss": 1.9866, + "step": 23228 + }, + { + "epoch": 5.07, + "learning_rate": 2.467882892724492e-06, + "loss": 2.7193, + "step": 23229 + }, + { + "epoch": 5.07, + "learning_rate": 2.4677736508630108e-06, + "loss": 2.0369, + "step": 23230 + }, + { + "epoch": 5.07, + "learning_rate": 2.4676644090015294e-06, + "loss": 0.1546, + "step": 23231 + }, + { + "epoch": 5.07, + "learning_rate": 2.4675551671400484e-06, + "loss": 1.0903, + "step": 23232 + }, + { + "epoch": 5.07, + "learning_rate": 2.467445925278567e-06, + "loss": 1.0191, + "step": 23233 + }, + { + "epoch": 5.07, + "learning_rate": 2.4673366834170856e-06, + "loss": 1.6786, + "step": 23234 + }, + { + "epoch": 5.07, + "learning_rate": 2.4672274415556043e-06, + "loss": 2.4853, + "step": 23235 + }, + { + "epoch": 5.07, + "learning_rate": 2.467118199694123e-06, + "loss": 3.2052, + "step": 23236 + }, + { + "epoch": 5.07, + "learning_rate": 2.467008957832642e-06, + "loss": 1.786, + "step": 23237 + }, + { + "epoch": 5.07, + "learning_rate": 2.46689971597116e-06, + "loss": 1.6858, + "step": 23238 + }, + { + "epoch": 5.07, + "learning_rate": 2.466790474109679e-06, + "loss": 1.0887, + "step": 23239 + }, + { + "epoch": 5.07, + "learning_rate": 2.4666812322481977e-06, + "loss": 1.8718, + "step": 23240 + }, + { + "epoch": 5.07, + "learning_rate": 2.4665719903867163e-06, + "loss": 0.7193, + "step": 23241 + }, + { + "epoch": 5.07, + "learning_rate": 2.466462748525235e-06, + "loss": 1.6743, + "step": 23242 + }, + { + "epoch": 5.07, + "learning_rate": 2.4663535066637536e-06, + "loss": 2.5347, + "step": 23243 + }, + { + "epoch": 5.07, + "learning_rate": 2.4662442648022726e-06, + "loss": 0.8677, + "step": 23244 + }, + { + "epoch": 5.07, + "learning_rate": 2.4661350229407908e-06, + "loss": 1.4413, + "step": 23245 + }, + { + "epoch": 5.07, + "learning_rate": 2.46602578107931e-06, + "loss": 0.7221, + "step": 23246 + }, + { + "epoch": 5.07, + "learning_rate": 2.4659165392178284e-06, + "loss": 2.0853, + "step": 23247 + }, + { + "epoch": 5.07, + "learning_rate": 2.465807297356347e-06, + "loss": 1.1188, + "step": 23248 + }, + { + "epoch": 5.07, + "learning_rate": 2.465698055494866e-06, + "loss": 0.5162, + "step": 23249 + }, + { + "epoch": 5.07, + "learning_rate": 2.4655888136333843e-06, + "loss": 0.1487, + "step": 23250 + }, + { + "epoch": 5.07, + "learning_rate": 2.4654795717719033e-06, + "loss": 1.4906, + "step": 23251 + }, + { + "epoch": 5.07, + "learning_rate": 2.465370329910422e-06, + "loss": 1.1822, + "step": 23252 + }, + { + "epoch": 5.07, + "learning_rate": 2.4652610880489405e-06, + "loss": 1.7341, + "step": 23253 + }, + { + "epoch": 5.08, + "learning_rate": 2.465151846187459e-06, + "loss": 1.6767, + "step": 23254 + }, + { + "epoch": 5.08, + "learning_rate": 2.4650426043259777e-06, + "loss": 0.7925, + "step": 23255 + }, + { + "epoch": 5.08, + "learning_rate": 2.4649333624644968e-06, + "loss": 2.1855, + "step": 23256 + }, + { + "epoch": 5.08, + "learning_rate": 2.4648241206030154e-06, + "loss": 1.3105, + "step": 23257 + }, + { + "epoch": 5.08, + "learning_rate": 2.464714878741534e-06, + "loss": 1.1198, + "step": 23258 + }, + { + "epoch": 5.08, + "learning_rate": 2.4646056368800526e-06, + "loss": 1.0804, + "step": 23259 + }, + { + "epoch": 5.08, + "learning_rate": 2.4644963950185712e-06, + "loss": 1.8885, + "step": 23260 + }, + { + "epoch": 5.08, + "learning_rate": 2.4643871531570903e-06, + "loss": 1.2743, + "step": 23261 + }, + { + "epoch": 5.08, + "learning_rate": 2.4642779112956084e-06, + "loss": 1.4221, + "step": 23262 + }, + { + "epoch": 5.08, + "learning_rate": 2.4641686694341275e-06, + "loss": 2.1448, + "step": 23263 + }, + { + "epoch": 5.08, + "learning_rate": 2.464059427572646e-06, + "loss": 2.6319, + "step": 23264 + }, + { + "epoch": 5.08, + "learning_rate": 2.4639501857111647e-06, + "loss": 0.5542, + "step": 23265 + }, + { + "epoch": 5.08, + "learning_rate": 2.4638409438496837e-06, + "loss": 0.3477, + "step": 23266 + }, + { + "epoch": 5.08, + "learning_rate": 2.463731701988202e-06, + "loss": 2.1519, + "step": 23267 + }, + { + "epoch": 5.08, + "learning_rate": 2.463622460126721e-06, + "loss": 1.4263, + "step": 23268 + }, + { + "epoch": 5.08, + "learning_rate": 2.4635132182652396e-06, + "loss": 2.4246, + "step": 23269 + }, + { + "epoch": 5.08, + "learning_rate": 2.463403976403758e-06, + "loss": 0.746, + "step": 23270 + }, + { + "epoch": 5.08, + "learning_rate": 2.463294734542277e-06, + "loss": 1.7447, + "step": 23271 + }, + { + "epoch": 5.08, + "learning_rate": 2.4631854926807954e-06, + "loss": 3.4113, + "step": 23272 + }, + { + "epoch": 5.08, + "learning_rate": 2.4630762508193144e-06, + "loss": 0.9437, + "step": 23273 + }, + { + "epoch": 5.08, + "learning_rate": 2.4629670089578326e-06, + "loss": 0.5994, + "step": 23274 + }, + { + "epoch": 5.08, + "learning_rate": 2.4628577670963517e-06, + "loss": 2.1136, + "step": 23275 + }, + { + "epoch": 5.08, + "learning_rate": 2.4627485252348703e-06, + "loss": 3.1685, + "step": 23276 + }, + { + "epoch": 5.08, + "learning_rate": 2.462639283373389e-06, + "loss": 0.9477, + "step": 23277 + }, + { + "epoch": 5.08, + "learning_rate": 2.4625300415119075e-06, + "loss": 1.2701, + "step": 23278 + }, + { + "epoch": 5.08, + "learning_rate": 2.462420799650426e-06, + "loss": 1.4134, + "step": 23279 + }, + { + "epoch": 5.08, + "learning_rate": 2.462311557788945e-06, + "loss": 1.7406, + "step": 23280 + }, + { + "epoch": 5.08, + "learning_rate": 2.4622023159274637e-06, + "loss": 0.8671, + "step": 23281 + }, + { + "epoch": 5.08, + "learning_rate": 2.4620930740659824e-06, + "loss": 1.6605, + "step": 23282 + }, + { + "epoch": 5.08, + "learning_rate": 2.461983832204501e-06, + "loss": 0.9583, + "step": 23283 + }, + { + "epoch": 5.08, + "learning_rate": 2.4618745903430196e-06, + "loss": 2.6468, + "step": 23284 + }, + { + "epoch": 5.08, + "learning_rate": 2.461765348481538e-06, + "loss": 1.5648, + "step": 23285 + }, + { + "epoch": 5.08, + "learning_rate": 2.461656106620057e-06, + "loss": 1.1747, + "step": 23286 + }, + { + "epoch": 5.08, + "learning_rate": 2.461546864758576e-06, + "loss": 1.1857, + "step": 23287 + }, + { + "epoch": 5.08, + "learning_rate": 2.4614376228970944e-06, + "loss": 1.2479, + "step": 23288 + }, + { + "epoch": 5.08, + "learning_rate": 2.461328381035613e-06, + "loss": 1.6036, + "step": 23289 + }, + { + "epoch": 5.08, + "learning_rate": 2.4612191391741317e-06, + "loss": 1.4911, + "step": 23290 + }, + { + "epoch": 5.08, + "learning_rate": 2.4611098973126503e-06, + "loss": 1.8804, + "step": 23291 + }, + { + "epoch": 5.08, + "learning_rate": 2.461000655451169e-06, + "loss": 1.9849, + "step": 23292 + }, + { + "epoch": 5.08, + "learning_rate": 2.460891413589688e-06, + "loss": 1.2519, + "step": 23293 + }, + { + "epoch": 5.08, + "learning_rate": 2.4607821717282065e-06, + "loss": 1.1408, + "step": 23294 + }, + { + "epoch": 5.08, + "learning_rate": 2.460672929866725e-06, + "loss": 1.0742, + "step": 23295 + }, + { + "epoch": 5.08, + "learning_rate": 2.4605636880052438e-06, + "loss": 0.5724, + "step": 23296 + }, + { + "epoch": 5.08, + "learning_rate": 2.4604544461437624e-06, + "loss": 1.8088, + "step": 23297 + }, + { + "epoch": 5.08, + "learning_rate": 2.460345204282281e-06, + "loss": 1.6331, + "step": 23298 + }, + { + "epoch": 5.08, + "learning_rate": 2.4602359624208e-06, + "loss": 0.9767, + "step": 23299 + }, + { + "epoch": 5.09, + "learning_rate": 2.4601267205593186e-06, + "loss": 0.6567, + "step": 23300 + }, + { + "epoch": 5.09, + "learning_rate": 2.4600174786978372e-06, + "loss": 1.2981, + "step": 23301 + }, + { + "epoch": 5.09, + "learning_rate": 2.459908236836356e-06, + "loss": 0.9248, + "step": 23302 + }, + { + "epoch": 5.09, + "learning_rate": 2.4597989949748745e-06, + "loss": 0.6583, + "step": 23303 + }, + { + "epoch": 5.09, + "learning_rate": 2.459689753113393e-06, + "loss": 1.3344, + "step": 23304 + }, + { + "epoch": 5.09, + "learning_rate": 2.459580511251912e-06, + "loss": 0.423, + "step": 23305 + }, + { + "epoch": 5.09, + "learning_rate": 2.4594712693904307e-06, + "loss": 4.3625, + "step": 23306 + }, + { + "epoch": 5.09, + "learning_rate": 2.4593620275289493e-06, + "loss": 2.672, + "step": 23307 + }, + { + "epoch": 5.09, + "learning_rate": 2.459252785667468e-06, + "loss": 1.1635, + "step": 23308 + }, + { + "epoch": 5.09, + "learning_rate": 2.4591435438059865e-06, + "loss": 1.5682, + "step": 23309 + }, + { + "epoch": 5.09, + "learning_rate": 2.4590343019445056e-06, + "loss": 2.0059, + "step": 23310 + }, + { + "epoch": 5.09, + "learning_rate": 2.4589250600830238e-06, + "loss": 1.7598, + "step": 23311 + }, + { + "epoch": 5.09, + "learning_rate": 2.458815818221543e-06, + "loss": 0.9328, + "step": 23312 + }, + { + "epoch": 5.09, + "learning_rate": 2.4587065763600614e-06, + "loss": 0.768, + "step": 23313 + }, + { + "epoch": 5.09, + "learning_rate": 2.45859733449858e-06, + "loss": 0.4414, + "step": 23314 + }, + { + "epoch": 5.09, + "learning_rate": 2.4584880926370986e-06, + "loss": 0.8668, + "step": 23315 + }, + { + "epoch": 5.09, + "learning_rate": 2.4583788507756172e-06, + "loss": 1.4733, + "step": 23316 + }, + { + "epoch": 5.09, + "learning_rate": 2.4582696089141363e-06, + "loss": 1.252, + "step": 23317 + }, + { + "epoch": 5.09, + "learning_rate": 2.4581603670526545e-06, + "loss": 0.6305, + "step": 23318 + }, + { + "epoch": 5.09, + "learning_rate": 2.4580511251911735e-06, + "loss": 2.9973, + "step": 23319 + }, + { + "epoch": 5.09, + "learning_rate": 2.457941883329692e-06, + "loss": 2.1099, + "step": 23320 + }, + { + "epoch": 5.09, + "learning_rate": 2.4578326414682107e-06, + "loss": 2.1689, + "step": 23321 + }, + { + "epoch": 5.09, + "learning_rate": 2.4577233996067298e-06, + "loss": 0.8268, + "step": 23322 + }, + { + "epoch": 5.09, + "learning_rate": 2.457614157745248e-06, + "loss": 1.513, + "step": 23323 + }, + { + "epoch": 5.09, + "learning_rate": 2.457504915883767e-06, + "loss": 1.1802, + "step": 23324 + }, + { + "epoch": 5.09, + "learning_rate": 2.4573956740222856e-06, + "loss": 3.3481, + "step": 23325 + }, + { + "epoch": 5.09, + "learning_rate": 2.457286432160804e-06, + "loss": 0.9012, + "step": 23326 + }, + { + "epoch": 5.09, + "learning_rate": 2.457177190299323e-06, + "loss": 2.476, + "step": 23327 + }, + { + "epoch": 5.09, + "learning_rate": 2.4570679484378414e-06, + "loss": 1.9449, + "step": 23328 + }, + { + "epoch": 5.09, + "learning_rate": 2.4569587065763605e-06, + "loss": 0.8748, + "step": 23329 + }, + { + "epoch": 5.09, + "learning_rate": 2.4568494647148786e-06, + "loss": 1.7431, + "step": 23330 + }, + { + "epoch": 5.09, + "learning_rate": 2.4567402228533977e-06, + "loss": 1.6945, + "step": 23331 + }, + { + "epoch": 5.09, + "learning_rate": 2.4566309809919163e-06, + "loss": 2.2397, + "step": 23332 + }, + { + "epoch": 5.09, + "learning_rate": 2.456521739130435e-06, + "loss": 1.7346, + "step": 23333 + }, + { + "epoch": 5.09, + "learning_rate": 2.456412497268954e-06, + "loss": 1.035, + "step": 23334 + }, + { + "epoch": 5.09, + "learning_rate": 2.456303255407472e-06, + "loss": 0.9767, + "step": 23335 + }, + { + "epoch": 5.09, + "learning_rate": 2.456194013545991e-06, + "loss": 2.9235, + "step": 23336 + }, + { + "epoch": 5.09, + "learning_rate": 2.4560847716845098e-06, + "loss": 1.0639, + "step": 23337 + }, + { + "epoch": 5.09, + "learning_rate": 2.4559755298230284e-06, + "loss": 2.2144, + "step": 23338 + }, + { + "epoch": 5.09, + "learning_rate": 2.455866287961547e-06, + "loss": 1.4182, + "step": 23339 + }, + { + "epoch": 5.09, + "learning_rate": 2.4557570461000656e-06, + "loss": 1.353, + "step": 23340 + }, + { + "epoch": 5.09, + "learning_rate": 2.4556478042385846e-06, + "loss": 2.0296, + "step": 23341 + }, + { + "epoch": 5.09, + "learning_rate": 2.455538562377103e-06, + "loss": 1.7741, + "step": 23342 + }, + { + "epoch": 5.09, + "learning_rate": 2.455429320515622e-06, + "loss": 1.4842, + "step": 23343 + }, + { + "epoch": 5.09, + "learning_rate": 2.4553200786541405e-06, + "loss": 2.2258, + "step": 23344 + }, + { + "epoch": 5.09, + "learning_rate": 2.455210836792659e-06, + "loss": 3.3055, + "step": 23345 + }, + { + "epoch": 5.1, + "learning_rate": 2.455101594931178e-06, + "loss": 0.9533, + "step": 23346 + }, + { + "epoch": 5.1, + "learning_rate": 2.4549923530696963e-06, + "loss": 1.1272, + "step": 23347 + }, + { + "epoch": 5.1, + "learning_rate": 2.4548831112082153e-06, + "loss": 2.2142, + "step": 23348 + }, + { + "epoch": 5.1, + "learning_rate": 2.454773869346734e-06, + "loss": 3.1593, + "step": 23349 + }, + { + "epoch": 5.1, + "learning_rate": 2.4546646274852526e-06, + "loss": 0.9691, + "step": 23350 + }, + { + "epoch": 5.1, + "learning_rate": 2.454555385623771e-06, + "loss": 0.6765, + "step": 23351 + }, + { + "epoch": 5.1, + "learning_rate": 2.4544461437622898e-06, + "loss": 1.0378, + "step": 23352 + }, + { + "epoch": 5.1, + "learning_rate": 2.454336901900809e-06, + "loss": 0.9614, + "step": 23353 + }, + { + "epoch": 5.1, + "learning_rate": 2.4542276600393274e-06, + "loss": 1.7693, + "step": 23354 + }, + { + "epoch": 5.1, + "learning_rate": 2.454118418177846e-06, + "loss": 1.543, + "step": 23355 + }, + { + "epoch": 5.1, + "learning_rate": 2.4540091763163647e-06, + "loss": 1.8733, + "step": 23356 + }, + { + "epoch": 5.1, + "learning_rate": 2.4538999344548833e-06, + "loss": 1.7756, + "step": 23357 + }, + { + "epoch": 5.1, + "learning_rate": 2.453790692593402e-06, + "loss": 2.1819, + "step": 23358 + }, + { + "epoch": 5.1, + "learning_rate": 2.4536814507319205e-06, + "loss": 1.3016, + "step": 23359 + }, + { + "epoch": 5.1, + "learning_rate": 2.4535722088704395e-06, + "loss": 2.658, + "step": 23360 + }, + { + "epoch": 5.1, + "learning_rate": 2.453462967008958e-06, + "loss": 1.2859, + "step": 23361 + }, + { + "epoch": 5.1, + "learning_rate": 2.4533537251474767e-06, + "loss": 1.5117, + "step": 23362 + }, + { + "epoch": 5.1, + "learning_rate": 2.4532444832859954e-06, + "loss": 1.1735, + "step": 23363 + }, + { + "epoch": 5.1, + "learning_rate": 2.453135241424514e-06, + "loss": 0.8169, + "step": 23364 + }, + { + "epoch": 5.1, + "learning_rate": 2.453025999563033e-06, + "loss": 0.9234, + "step": 23365 + }, + { + "epoch": 5.1, + "learning_rate": 2.4529167577015516e-06, + "loss": 1.2832, + "step": 23366 + }, + { + "epoch": 5.1, + "learning_rate": 2.4528075158400702e-06, + "loss": 0.7824, + "step": 23367 + }, + { + "epoch": 5.1, + "learning_rate": 2.452698273978589e-06, + "loss": 1.5101, + "step": 23368 + }, + { + "epoch": 5.1, + "learning_rate": 2.4525890321171074e-06, + "loss": 2.0976, + "step": 23369 + }, + { + "epoch": 5.1, + "learning_rate": 2.452479790255626e-06, + "loss": 1.0574, + "step": 23370 + }, + { + "epoch": 5.1, + "learning_rate": 2.4523705483941447e-06, + "loss": 2.3299, + "step": 23371 + }, + { + "epoch": 5.1, + "learning_rate": 2.4522613065326637e-06, + "loss": 3.5636, + "step": 23372 + }, + { + "epoch": 5.1, + "learning_rate": 2.4521520646711823e-06, + "loss": 1.8187, + "step": 23373 + }, + { + "epoch": 5.1, + "learning_rate": 2.452042822809701e-06, + "loss": 3.1367, + "step": 23374 + }, + { + "epoch": 5.1, + "learning_rate": 2.4519335809482195e-06, + "loss": 1.9694, + "step": 23375 + }, + { + "epoch": 5.1, + "learning_rate": 2.451824339086738e-06, + "loss": 1.9694, + "step": 23376 + }, + { + "epoch": 5.1, + "learning_rate": 2.4517150972252568e-06, + "loss": 0.9847, + "step": 23377 + }, + { + "epoch": 5.1, + "learning_rate": 2.451605855363776e-06, + "loss": 1.7241, + "step": 23378 + }, + { + "epoch": 5.1, + "learning_rate": 2.4514966135022944e-06, + "loss": 1.3475, + "step": 23379 + }, + { + "epoch": 5.1, + "learning_rate": 2.451387371640813e-06, + "loss": 3.297, + "step": 23380 + }, + { + "epoch": 5.1, + "learning_rate": 2.4512781297793316e-06, + "loss": 1.2224, + "step": 23381 + }, + { + "epoch": 5.1, + "learning_rate": 2.4511688879178502e-06, + "loss": 1.8885, + "step": 23382 + }, + { + "epoch": 5.1, + "learning_rate": 2.451059646056369e-06, + "loss": 2.2358, + "step": 23383 + }, + { + "epoch": 5.1, + "learning_rate": 2.4509504041948875e-06, + "loss": 2.207, + "step": 23384 + }, + { + "epoch": 5.1, + "learning_rate": 2.4508411623334065e-06, + "loss": 1.9289, + "step": 23385 + }, + { + "epoch": 5.1, + "learning_rate": 2.450731920471925e-06, + "loss": 0.7048, + "step": 23386 + }, + { + "epoch": 5.1, + "learning_rate": 2.4506226786104437e-06, + "loss": 2.0857, + "step": 23387 + }, + { + "epoch": 5.1, + "learning_rate": 2.4505134367489623e-06, + "loss": 1.1462, + "step": 23388 + }, + { + "epoch": 5.1, + "learning_rate": 2.450404194887481e-06, + "loss": 0.8426, + "step": 23389 + }, + { + "epoch": 5.1, + "learning_rate": 2.450294953026e-06, + "loss": 1.1928, + "step": 23390 + }, + { + "epoch": 5.1, + "learning_rate": 2.450185711164518e-06, + "loss": 0.4671, + "step": 23391 + }, + { + "epoch": 5.11, + "learning_rate": 2.450076469303037e-06, + "loss": 1.0386, + "step": 23392 + }, + { + "epoch": 5.11, + "learning_rate": 2.449967227441556e-06, + "loss": 1.3974, + "step": 23393 + }, + { + "epoch": 5.11, + "learning_rate": 2.4498579855800744e-06, + "loss": 0.2486, + "step": 23394 + }, + { + "epoch": 5.11, + "learning_rate": 2.449748743718593e-06, + "loss": 2.0898, + "step": 23395 + }, + { + "epoch": 5.11, + "learning_rate": 2.4496395018571116e-06, + "loss": 1.7194, + "step": 23396 + }, + { + "epoch": 5.11, + "learning_rate": 2.4495302599956307e-06, + "loss": 2.8005, + "step": 23397 + }, + { + "epoch": 5.11, + "learning_rate": 2.4494210181341493e-06, + "loss": 0.3282, + "step": 23398 + }, + { + "epoch": 5.11, + "learning_rate": 2.449311776272668e-06, + "loss": 0.379, + "step": 23399 + }, + { + "epoch": 5.11, + "learning_rate": 2.4492025344111865e-06, + "loss": 1.6549, + "step": 23400 + }, + { + "epoch": 5.11, + "learning_rate": 2.449093292549705e-06, + "loss": 1.0038, + "step": 23401 + }, + { + "epoch": 5.11, + "learning_rate": 2.448984050688224e-06, + "loss": 1.0925, + "step": 23402 + }, + { + "epoch": 5.11, + "learning_rate": 2.4488748088267423e-06, + "loss": 1.179, + "step": 23403 + }, + { + "epoch": 5.11, + "learning_rate": 2.4487655669652614e-06, + "loss": 1.4208, + "step": 23404 + }, + { + "epoch": 5.11, + "learning_rate": 2.44865632510378e-06, + "loss": 1.0195, + "step": 23405 + }, + { + "epoch": 5.11, + "learning_rate": 2.4485470832422986e-06, + "loss": 0.6687, + "step": 23406 + }, + { + "epoch": 5.11, + "learning_rate": 2.4484378413808176e-06, + "loss": 1.5527, + "step": 23407 + }, + { + "epoch": 5.11, + "learning_rate": 2.448328599519336e-06, + "loss": 1.5602, + "step": 23408 + }, + { + "epoch": 5.11, + "learning_rate": 2.448219357657855e-06, + "loss": 2.4476, + "step": 23409 + }, + { + "epoch": 5.11, + "learning_rate": 2.4481101157963735e-06, + "loss": 0.9993, + "step": 23410 + }, + { + "epoch": 5.11, + "learning_rate": 2.448000873934892e-06, + "loss": 1.8174, + "step": 23411 + }, + { + "epoch": 5.11, + "learning_rate": 2.4478916320734107e-06, + "loss": 1.8843, + "step": 23412 + }, + { + "epoch": 5.11, + "learning_rate": 2.4477823902119293e-06, + "loss": 1.7809, + "step": 23413 + }, + { + "epoch": 5.11, + "learning_rate": 2.4476731483504483e-06, + "loss": 1.9212, + "step": 23414 + }, + { + "epoch": 5.11, + "learning_rate": 2.4475639064889665e-06, + "loss": 1.509, + "step": 23415 + }, + { + "epoch": 5.11, + "learning_rate": 2.4474546646274855e-06, + "loss": 1.3137, + "step": 23416 + }, + { + "epoch": 5.11, + "learning_rate": 2.447345422766004e-06, + "loss": 0.9656, + "step": 23417 + }, + { + "epoch": 5.11, + "learning_rate": 2.4472361809045228e-06, + "loss": 0.5001, + "step": 23418 + }, + { + "epoch": 5.11, + "learning_rate": 2.447126939043042e-06, + "loss": 1.2565, + "step": 23419 + }, + { + "epoch": 5.11, + "learning_rate": 2.44701769718156e-06, + "loss": 1.3366, + "step": 23420 + }, + { + "epoch": 5.11, + "learning_rate": 2.446908455320079e-06, + "loss": 0.9483, + "step": 23421 + }, + { + "epoch": 5.11, + "learning_rate": 2.4467992134585976e-06, + "loss": 1.1029, + "step": 23422 + }, + { + "epoch": 5.11, + "learning_rate": 2.4466899715971162e-06, + "loss": 0.4172, + "step": 23423 + }, + { + "epoch": 5.11, + "learning_rate": 2.446580729735635e-06, + "loss": 1.1469, + "step": 23424 + }, + { + "epoch": 5.11, + "learning_rate": 2.4464714878741535e-06, + "loss": 1.4243, + "step": 23425 + }, + { + "epoch": 5.11, + "learning_rate": 2.4463622460126725e-06, + "loss": 1.3363, + "step": 23426 + }, + { + "epoch": 5.11, + "learning_rate": 2.4462530041511907e-06, + "loss": 0.7697, + "step": 23427 + }, + { + "epoch": 5.11, + "learning_rate": 2.4461437622897097e-06, + "loss": 0.9491, + "step": 23428 + }, + { + "epoch": 5.11, + "learning_rate": 2.4460345204282283e-06, + "loss": 0.7326, + "step": 23429 + }, + { + "epoch": 5.11, + "learning_rate": 2.445925278566747e-06, + "loss": 1.6824, + "step": 23430 + }, + { + "epoch": 5.11, + "learning_rate": 2.445816036705266e-06, + "loss": 1.5923, + "step": 23431 + }, + { + "epoch": 5.11, + "learning_rate": 2.445706794843784e-06, + "loss": 1.8721, + "step": 23432 + }, + { + "epoch": 5.11, + "learning_rate": 2.445597552982303e-06, + "loss": 0.7728, + "step": 23433 + }, + { + "epoch": 5.11, + "learning_rate": 2.445488311120822e-06, + "loss": 1.4609, + "step": 23434 + }, + { + "epoch": 5.11, + "learning_rate": 2.4453790692593404e-06, + "loss": 1.5689, + "step": 23435 + }, + { + "epoch": 5.11, + "learning_rate": 2.445269827397859e-06, + "loss": 3.0677, + "step": 23436 + }, + { + "epoch": 5.12, + "learning_rate": 2.4451605855363776e-06, + "loss": 1.53, + "step": 23437 + }, + { + "epoch": 5.12, + "learning_rate": 2.4450513436748967e-06, + "loss": 1.43, + "step": 23438 + }, + { + "epoch": 5.12, + "learning_rate": 2.444942101813415e-06, + "loss": 0.567, + "step": 23439 + }, + { + "epoch": 5.12, + "learning_rate": 2.444832859951934e-06, + "loss": 1.2848, + "step": 23440 + }, + { + "epoch": 5.12, + "learning_rate": 2.4447236180904525e-06, + "loss": 1.1356, + "step": 23441 + }, + { + "epoch": 5.12, + "learning_rate": 2.444614376228971e-06, + "loss": 2.4939, + "step": 23442 + }, + { + "epoch": 5.12, + "learning_rate": 2.4445051343674897e-06, + "loss": 0.6119, + "step": 23443 + }, + { + "epoch": 5.12, + "learning_rate": 2.4443958925060083e-06, + "loss": 0.9608, + "step": 23444 + }, + { + "epoch": 5.12, + "learning_rate": 2.4442866506445274e-06, + "loss": 1.409, + "step": 23445 + }, + { + "epoch": 5.12, + "learning_rate": 2.444177408783046e-06, + "loss": 0.7849, + "step": 23446 + }, + { + "epoch": 5.12, + "learning_rate": 2.4440681669215646e-06, + "loss": 0.8993, + "step": 23447 + }, + { + "epoch": 5.12, + "learning_rate": 2.4439589250600832e-06, + "loss": 1.5014, + "step": 23448 + }, + { + "epoch": 5.12, + "learning_rate": 2.443849683198602e-06, + "loss": 1.1159, + "step": 23449 + }, + { + "epoch": 5.12, + "learning_rate": 2.4437404413371204e-06, + "loss": 1.2659, + "step": 23450 + }, + { + "epoch": 5.12, + "learning_rate": 2.443631199475639e-06, + "loss": 2.0877, + "step": 23451 + }, + { + "epoch": 5.12, + "learning_rate": 2.443521957614158e-06, + "loss": 2.514, + "step": 23452 + }, + { + "epoch": 5.12, + "learning_rate": 2.4434127157526767e-06, + "loss": 2.6755, + "step": 23453 + }, + { + "epoch": 5.12, + "learning_rate": 2.4433034738911953e-06, + "loss": 1.5822, + "step": 23454 + }, + { + "epoch": 5.12, + "learning_rate": 2.443194232029714e-06, + "loss": 0.7785, + "step": 23455 + }, + { + "epoch": 5.12, + "learning_rate": 2.4430849901682325e-06, + "loss": 1.5154, + "step": 23456 + }, + { + "epoch": 5.12, + "learning_rate": 2.442975748306751e-06, + "loss": 1.4444, + "step": 23457 + }, + { + "epoch": 5.12, + "learning_rate": 2.44286650644527e-06, + "loss": 3.1414, + "step": 23458 + }, + { + "epoch": 5.12, + "learning_rate": 2.4427572645837888e-06, + "loss": 1.9708, + "step": 23459 + }, + { + "epoch": 5.12, + "learning_rate": 2.4426480227223074e-06, + "loss": 1.7377, + "step": 23460 + }, + { + "epoch": 5.12, + "learning_rate": 2.442538780860826e-06, + "loss": 3.1424, + "step": 23461 + }, + { + "epoch": 5.12, + "learning_rate": 2.4424295389993446e-06, + "loss": 1.8602, + "step": 23462 + }, + { + "epoch": 5.12, + "learning_rate": 2.4423202971378637e-06, + "loss": 1.2752, + "step": 23463 + }, + { + "epoch": 5.12, + "learning_rate": 2.4422110552763823e-06, + "loss": 2.8931, + "step": 23464 + }, + { + "epoch": 5.12, + "learning_rate": 2.442101813414901e-06, + "loss": 1.562, + "step": 23465 + }, + { + "epoch": 5.12, + "learning_rate": 2.4419925715534195e-06, + "loss": 1.4014, + "step": 23466 + }, + { + "epoch": 5.12, + "learning_rate": 2.441883329691938e-06, + "loss": 0.5338, + "step": 23467 + }, + { + "epoch": 5.12, + "learning_rate": 2.4417740878304567e-06, + "loss": 1.7157, + "step": 23468 + }, + { + "epoch": 5.12, + "learning_rate": 2.4416648459689753e-06, + "loss": 1.3958, + "step": 23469 + }, + { + "epoch": 5.12, + "learning_rate": 2.4415556041074944e-06, + "loss": 1.328, + "step": 23470 + }, + { + "epoch": 5.12, + "learning_rate": 2.441446362246013e-06, + "loss": 1.1412, + "step": 23471 + }, + { + "epoch": 5.12, + "learning_rate": 2.4413371203845316e-06, + "loss": 1.4977, + "step": 23472 + }, + { + "epoch": 5.12, + "learning_rate": 2.44122787852305e-06, + "loss": 3.5689, + "step": 23473 + }, + { + "epoch": 5.12, + "learning_rate": 2.441118636661569e-06, + "loss": 1.9176, + "step": 23474 + }, + { + "epoch": 5.12, + "learning_rate": 2.441009394800088e-06, + "loss": 2.1753, + "step": 23475 + }, + { + "epoch": 5.12, + "learning_rate": 2.440900152938606e-06, + "loss": 2.2196, + "step": 23476 + }, + { + "epoch": 5.12, + "learning_rate": 2.440790911077125e-06, + "loss": 1.6442, + "step": 23477 + }, + { + "epoch": 5.12, + "learning_rate": 2.4406816692156437e-06, + "loss": 2.5848, + "step": 23478 + }, + { + "epoch": 5.12, + "learning_rate": 2.4405724273541623e-06, + "loss": 1.4676, + "step": 23479 + }, + { + "epoch": 5.12, + "learning_rate": 2.440463185492681e-06, + "loss": 1.0721, + "step": 23480 + }, + { + "epoch": 5.12, + "learning_rate": 2.4403539436311995e-06, + "loss": 3.2936, + "step": 23481 + }, + { + "epoch": 5.12, + "learning_rate": 2.4402447017697185e-06, + "loss": 1.431, + "step": 23482 + }, + { + "epoch": 5.13, + "learning_rate": 2.4401354599082367e-06, + "loss": 1.28, + "step": 23483 + }, + { + "epoch": 5.13, + "learning_rate": 2.4400262180467558e-06, + "loss": 1.004, + "step": 23484 + }, + { + "epoch": 5.13, + "learning_rate": 2.4399169761852744e-06, + "loss": 2.006, + "step": 23485 + }, + { + "epoch": 5.13, + "learning_rate": 2.439807734323793e-06, + "loss": 0.8879, + "step": 23486 + }, + { + "epoch": 5.13, + "learning_rate": 2.439698492462312e-06, + "loss": 1.4802, + "step": 23487 + }, + { + "epoch": 5.13, + "learning_rate": 2.43958925060083e-06, + "loss": 2.6403, + "step": 23488 + }, + { + "epoch": 5.13, + "learning_rate": 2.4394800087393492e-06, + "loss": 3.3947, + "step": 23489 + }, + { + "epoch": 5.13, + "learning_rate": 2.439370766877868e-06, + "loss": 1.8636, + "step": 23490 + }, + { + "epoch": 5.13, + "learning_rate": 2.4392615250163865e-06, + "loss": 2.0129, + "step": 23491 + }, + { + "epoch": 5.13, + "learning_rate": 2.439152283154905e-06, + "loss": 1.1942, + "step": 23492 + }, + { + "epoch": 5.13, + "learning_rate": 2.4390430412934237e-06, + "loss": 1.1029, + "step": 23493 + }, + { + "epoch": 5.13, + "learning_rate": 2.4389337994319427e-06, + "loss": 1.3708, + "step": 23494 + }, + { + "epoch": 5.13, + "learning_rate": 2.438824557570461e-06, + "loss": 0.9473, + "step": 23495 + }, + { + "epoch": 5.13, + "learning_rate": 2.43871531570898e-06, + "loss": 1.7595, + "step": 23496 + }, + { + "epoch": 5.13, + "learning_rate": 2.4386060738474985e-06, + "loss": 1.7242, + "step": 23497 + }, + { + "epoch": 5.13, + "learning_rate": 2.438496831986017e-06, + "loss": 0.8025, + "step": 23498 + }, + { + "epoch": 5.13, + "learning_rate": 2.438387590124536e-06, + "loss": 1.7276, + "step": 23499 + }, + { + "epoch": 5.13, + "learning_rate": 2.4382783482630544e-06, + "loss": 0.889, + "step": 23500 + }, + { + "epoch": 5.13, + "learning_rate": 2.4381691064015734e-06, + "loss": 2.451, + "step": 23501 + }, + { + "epoch": 5.13, + "learning_rate": 2.438059864540092e-06, + "loss": 1.0813, + "step": 23502 + }, + { + "epoch": 5.13, + "learning_rate": 2.4379506226786106e-06, + "loss": 1.4597, + "step": 23503 + }, + { + "epoch": 5.13, + "learning_rate": 2.4378413808171292e-06, + "loss": 2.9633, + "step": 23504 + }, + { + "epoch": 5.13, + "learning_rate": 2.437732138955648e-06, + "loss": 2.0722, + "step": 23505 + }, + { + "epoch": 5.13, + "learning_rate": 2.437622897094167e-06, + "loss": 1.358, + "step": 23506 + }, + { + "epoch": 5.13, + "learning_rate": 2.4375136552326855e-06, + "loss": 1.0889, + "step": 23507 + }, + { + "epoch": 5.13, + "learning_rate": 2.437404413371204e-06, + "loss": 0.486, + "step": 23508 + }, + { + "epoch": 5.13, + "learning_rate": 2.4372951715097227e-06, + "loss": 0.5104, + "step": 23509 + }, + { + "epoch": 5.13, + "learning_rate": 2.4371859296482413e-06, + "loss": 0.789, + "step": 23510 + }, + { + "epoch": 5.13, + "learning_rate": 2.4370766877867604e-06, + "loss": 1.1425, + "step": 23511 + }, + { + "epoch": 5.13, + "learning_rate": 2.4369674459252786e-06, + "loss": 2.913, + "step": 23512 + }, + { + "epoch": 5.13, + "learning_rate": 2.4368582040637976e-06, + "loss": 1.2299, + "step": 23513 + }, + { + "epoch": 5.13, + "learning_rate": 2.436748962202316e-06, + "loss": 1.7598, + "step": 23514 + }, + { + "epoch": 5.13, + "learning_rate": 2.436639720340835e-06, + "loss": 0.8429, + "step": 23515 + }, + { + "epoch": 5.13, + "learning_rate": 2.4365304784793534e-06, + "loss": 1.8434, + "step": 23516 + }, + { + "epoch": 5.13, + "learning_rate": 2.436421236617872e-06, + "loss": 0.9312, + "step": 23517 + }, + { + "epoch": 5.13, + "learning_rate": 2.436311994756391e-06, + "loss": 1.3357, + "step": 23518 + }, + { + "epoch": 5.13, + "learning_rate": 2.4362027528949097e-06, + "loss": 0.8915, + "step": 23519 + }, + { + "epoch": 5.13, + "learning_rate": 2.4360935110334283e-06, + "loss": 3.3156, + "step": 23520 + }, + { + "epoch": 5.13, + "learning_rate": 2.435984269171947e-06, + "loss": 1.3083, + "step": 23521 + }, + { + "epoch": 5.13, + "learning_rate": 2.4358750273104655e-06, + "loss": 1.806, + "step": 23522 + }, + { + "epoch": 5.13, + "learning_rate": 2.435765785448984e-06, + "loss": 1.3966, + "step": 23523 + }, + { + "epoch": 5.13, + "learning_rate": 2.4356565435875027e-06, + "loss": 1.5257, + "step": 23524 + }, + { + "epoch": 5.13, + "learning_rate": 2.4355473017260218e-06, + "loss": 0.3541, + "step": 23525 + }, + { + "epoch": 5.13, + "learning_rate": 2.4354380598645404e-06, + "loss": 1.6462, + "step": 23526 + }, + { + "epoch": 5.13, + "learning_rate": 2.435328818003059e-06, + "loss": 1.0228, + "step": 23527 + }, + { + "epoch": 5.13, + "learning_rate": 2.4352195761415776e-06, + "loss": 2.8179, + "step": 23528 + }, + { + "epoch": 5.14, + "learning_rate": 2.4351103342800962e-06, + "loss": 2.1351, + "step": 23529 + }, + { + "epoch": 5.14, + "learning_rate": 2.435001092418615e-06, + "loss": 1.2749, + "step": 23530 + }, + { + "epoch": 5.14, + "learning_rate": 2.434891850557134e-06, + "loss": 1.3427, + "step": 23531 + }, + { + "epoch": 5.14, + "learning_rate": 2.4347826086956525e-06, + "loss": 1.2616, + "step": 23532 + }, + { + "epoch": 5.14, + "learning_rate": 2.434673366834171e-06, + "loss": 2.1537, + "step": 23533 + }, + { + "epoch": 5.14, + "learning_rate": 2.4345641249726897e-06, + "loss": 1.6451, + "step": 23534 + }, + { + "epoch": 5.14, + "learning_rate": 2.4344548831112083e-06, + "loss": 0.5227, + "step": 23535 + }, + { + "epoch": 5.14, + "learning_rate": 2.434345641249727e-06, + "loss": 0.2511, + "step": 23536 + }, + { + "epoch": 5.14, + "learning_rate": 2.434236399388246e-06, + "loss": 1.9412, + "step": 23537 + }, + { + "epoch": 5.14, + "learning_rate": 2.4341271575267646e-06, + "loss": 1.6544, + "step": 23538 + }, + { + "epoch": 5.14, + "learning_rate": 2.434017915665283e-06, + "loss": 1.1509, + "step": 23539 + }, + { + "epoch": 5.14, + "learning_rate": 2.4339086738038018e-06, + "loss": 0.9631, + "step": 23540 + }, + { + "epoch": 5.14, + "learning_rate": 2.4337994319423204e-06, + "loss": 0.2518, + "step": 23541 + }, + { + "epoch": 5.14, + "learning_rate": 2.433690190080839e-06, + "loss": 1.8776, + "step": 23542 + }, + { + "epoch": 5.14, + "learning_rate": 2.433580948219358e-06, + "loss": 1.1307, + "step": 23543 + }, + { + "epoch": 5.14, + "learning_rate": 2.4334717063578766e-06, + "loss": 0.8897, + "step": 23544 + }, + { + "epoch": 5.14, + "learning_rate": 2.4333624644963953e-06, + "loss": 2.3824, + "step": 23545 + }, + { + "epoch": 5.14, + "learning_rate": 2.433253222634914e-06, + "loss": 1.4661, + "step": 23546 + }, + { + "epoch": 5.14, + "learning_rate": 2.4331439807734325e-06, + "loss": 0.9864, + "step": 23547 + }, + { + "epoch": 5.14, + "learning_rate": 2.433034738911951e-06, + "loss": 2.5477, + "step": 23548 + }, + { + "epoch": 5.14, + "learning_rate": 2.4329254970504697e-06, + "loss": 2.3409, + "step": 23549 + }, + { + "epoch": 5.14, + "learning_rate": 2.4328162551889887e-06, + "loss": 1.297, + "step": 23550 + }, + { + "epoch": 5.14, + "learning_rate": 2.4327070133275073e-06, + "loss": 1.5154, + "step": 23551 + }, + { + "epoch": 5.14, + "learning_rate": 2.432597771466026e-06, + "loss": 0.8283, + "step": 23552 + }, + { + "epoch": 5.14, + "learning_rate": 2.4324885296045446e-06, + "loss": 0.68, + "step": 23553 + }, + { + "epoch": 5.14, + "learning_rate": 2.432379287743063e-06, + "loss": 2.1906, + "step": 23554 + }, + { + "epoch": 5.14, + "learning_rate": 2.4322700458815822e-06, + "loss": 2.0996, + "step": 23555 + }, + { + "epoch": 5.14, + "learning_rate": 2.4321608040201004e-06, + "loss": 2.6814, + "step": 23556 + }, + { + "epoch": 5.14, + "learning_rate": 2.4320515621586194e-06, + "loss": 2.067, + "step": 23557 + }, + { + "epoch": 5.14, + "learning_rate": 2.431942320297138e-06, + "loss": 2.5126, + "step": 23558 + }, + { + "epoch": 5.14, + "learning_rate": 2.4318330784356567e-06, + "loss": 0.8401, + "step": 23559 + }, + { + "epoch": 5.14, + "learning_rate": 2.4317238365741757e-06, + "loss": 1.4408, + "step": 23560 + }, + { + "epoch": 5.14, + "learning_rate": 2.431614594712694e-06, + "loss": 1.9935, + "step": 23561 + }, + { + "epoch": 5.14, + "learning_rate": 2.431505352851213e-06, + "loss": 1.0401, + "step": 23562 + }, + { + "epoch": 5.14, + "learning_rate": 2.4313961109897315e-06, + "loss": 2.5424, + "step": 23563 + }, + { + "epoch": 5.14, + "learning_rate": 2.43128686912825e-06, + "loss": 2.111, + "step": 23564 + }, + { + "epoch": 5.14, + "learning_rate": 2.4311776272667687e-06, + "loss": 1.5897, + "step": 23565 + }, + { + "epoch": 5.14, + "learning_rate": 2.4310683854052874e-06, + "loss": 1.531, + "step": 23566 + }, + { + "epoch": 5.14, + "learning_rate": 2.4309591435438064e-06, + "loss": 0.6534, + "step": 23567 + }, + { + "epoch": 5.14, + "learning_rate": 2.4308499016823246e-06, + "loss": 0.754, + "step": 23568 + }, + { + "epoch": 5.14, + "learning_rate": 2.4307406598208436e-06, + "loss": 1.9067, + "step": 23569 + }, + { + "epoch": 5.14, + "learning_rate": 2.4306314179593622e-06, + "loss": 1.2916, + "step": 23570 + }, + { + "epoch": 5.14, + "learning_rate": 2.430522176097881e-06, + "loss": 1.6884, + "step": 23571 + }, + { + "epoch": 5.14, + "learning_rate": 2.4304129342364e-06, + "loss": 0.2442, + "step": 23572 + }, + { + "epoch": 5.14, + "learning_rate": 2.430303692374918e-06, + "loss": 1.5615, + "step": 23573 + }, + { + "epoch": 5.14, + "learning_rate": 2.430194450513437e-06, + "loss": 1.7166, + "step": 23574 + }, + { + "epoch": 5.15, + "learning_rate": 2.4300852086519557e-06, + "loss": 2.4644, + "step": 23575 + }, + { + "epoch": 5.15, + "learning_rate": 2.4299759667904743e-06, + "loss": 0.8982, + "step": 23576 + }, + { + "epoch": 5.15, + "learning_rate": 2.429866724928993e-06, + "loss": 1.7964, + "step": 23577 + }, + { + "epoch": 5.15, + "learning_rate": 2.4297574830675115e-06, + "loss": 1.4713, + "step": 23578 + }, + { + "epoch": 5.15, + "learning_rate": 2.4296482412060306e-06, + "loss": 2.3711, + "step": 23579 + }, + { + "epoch": 5.15, + "learning_rate": 2.4295389993445488e-06, + "loss": 1.8479, + "step": 23580 + }, + { + "epoch": 5.15, + "learning_rate": 2.429429757483068e-06, + "loss": 1.7241, + "step": 23581 + }, + { + "epoch": 5.15, + "learning_rate": 2.4293205156215864e-06, + "loss": 2.2256, + "step": 23582 + }, + { + "epoch": 5.15, + "learning_rate": 2.429211273760105e-06, + "loss": 1.5205, + "step": 23583 + }, + { + "epoch": 5.15, + "learning_rate": 2.429102031898624e-06, + "loss": 1.3928, + "step": 23584 + }, + { + "epoch": 5.15, + "learning_rate": 2.4289927900371422e-06, + "loss": 1.3301, + "step": 23585 + }, + { + "epoch": 5.15, + "learning_rate": 2.4288835481756613e-06, + "loss": 1.3899, + "step": 23586 + }, + { + "epoch": 5.15, + "learning_rate": 2.42877430631418e-06, + "loss": 1.9371, + "step": 23587 + }, + { + "epoch": 5.15, + "learning_rate": 2.4286650644526985e-06, + "loss": 2.4573, + "step": 23588 + }, + { + "epoch": 5.15, + "learning_rate": 2.428555822591217e-06, + "loss": 1.3523, + "step": 23589 + }, + { + "epoch": 5.15, + "learning_rate": 2.4284465807297357e-06, + "loss": 1.2464, + "step": 23590 + }, + { + "epoch": 5.15, + "learning_rate": 2.4283373388682548e-06, + "loss": 3.2012, + "step": 23591 + }, + { + "epoch": 5.15, + "learning_rate": 2.428228097006773e-06, + "loss": 1.6494, + "step": 23592 + }, + { + "epoch": 5.15, + "learning_rate": 2.428118855145292e-06, + "loss": 1.369, + "step": 23593 + }, + { + "epoch": 5.15, + "learning_rate": 2.4280096132838106e-06, + "loss": 1.2004, + "step": 23594 + }, + { + "epoch": 5.15, + "learning_rate": 2.427900371422329e-06, + "loss": 2.2446, + "step": 23595 + }, + { + "epoch": 5.15, + "learning_rate": 2.427791129560848e-06, + "loss": 1.3839, + "step": 23596 + }, + { + "epoch": 5.15, + "learning_rate": 2.4276818876993664e-06, + "loss": 0.9319, + "step": 23597 + }, + { + "epoch": 5.15, + "learning_rate": 2.4275726458378855e-06, + "loss": 1.3414, + "step": 23598 + }, + { + "epoch": 5.15, + "learning_rate": 2.427463403976404e-06, + "loss": 2.0319, + "step": 23599 + }, + { + "epoch": 5.15, + "learning_rate": 2.4273541621149227e-06, + "loss": 1.9229, + "step": 23600 + }, + { + "epoch": 5.15, + "learning_rate": 2.4272449202534413e-06, + "loss": 2.2889, + "step": 23601 + }, + { + "epoch": 5.15, + "learning_rate": 2.42713567839196e-06, + "loss": 0.9193, + "step": 23602 + }, + { + "epoch": 5.15, + "learning_rate": 2.427026436530479e-06, + "loss": 2.0088, + "step": 23603 + }, + { + "epoch": 5.15, + "learning_rate": 2.4269171946689975e-06, + "loss": 0.7865, + "step": 23604 + }, + { + "epoch": 5.15, + "learning_rate": 2.426807952807516e-06, + "loss": 0.9124, + "step": 23605 + }, + { + "epoch": 5.15, + "learning_rate": 2.4266987109460348e-06, + "loss": 3.337, + "step": 23606 + }, + { + "epoch": 5.15, + "learning_rate": 2.4265894690845534e-06, + "loss": 2.1368, + "step": 23607 + }, + { + "epoch": 5.15, + "learning_rate": 2.426480227223072e-06, + "loss": 1.6624, + "step": 23608 + }, + { + "epoch": 5.15, + "learning_rate": 2.4263709853615906e-06, + "loss": 2.7037, + "step": 23609 + }, + { + "epoch": 5.15, + "learning_rate": 2.4262617435001096e-06, + "loss": 2.329, + "step": 23610 + }, + { + "epoch": 5.15, + "learning_rate": 2.4261525016386282e-06, + "loss": 2.0092, + "step": 23611 + }, + { + "epoch": 5.15, + "learning_rate": 2.426043259777147e-06, + "loss": 0.7403, + "step": 23612 + }, + { + "epoch": 5.15, + "learning_rate": 2.4259340179156655e-06, + "loss": 0.2548, + "step": 23613 + }, + { + "epoch": 5.15, + "learning_rate": 2.425824776054184e-06, + "loss": 2.732, + "step": 23614 + }, + { + "epoch": 5.15, + "learning_rate": 2.4257155341927027e-06, + "loss": 1.4619, + "step": 23615 + }, + { + "epoch": 5.15, + "learning_rate": 2.4256062923312217e-06, + "loss": 0.5118, + "step": 23616 + }, + { + "epoch": 5.15, + "learning_rate": 2.4254970504697403e-06, + "loss": 1.1395, + "step": 23617 + }, + { + "epoch": 5.15, + "learning_rate": 2.425387808608259e-06, + "loss": 1.0462, + "step": 23618 + }, + { + "epoch": 5.15, + "learning_rate": 2.4252785667467776e-06, + "loss": 0.8349, + "step": 23619 + }, + { + "epoch": 5.15, + "learning_rate": 2.425169324885296e-06, + "loss": 0.9129, + "step": 23620 + }, + { + "epoch": 5.16, + "learning_rate": 2.4250600830238148e-06, + "loss": 1.2789, + "step": 23621 + }, + { + "epoch": 5.16, + "learning_rate": 2.4249508411623334e-06, + "loss": 2.3442, + "step": 23622 + }, + { + "epoch": 5.16, + "learning_rate": 2.4248415993008524e-06, + "loss": 1.0771, + "step": 23623 + }, + { + "epoch": 5.16, + "learning_rate": 2.424732357439371e-06, + "loss": 1.4629, + "step": 23624 + }, + { + "epoch": 5.16, + "learning_rate": 2.4246231155778896e-06, + "loss": 2.412, + "step": 23625 + }, + { + "epoch": 5.16, + "learning_rate": 2.4245138737164083e-06, + "loss": 0.7925, + "step": 23626 + }, + { + "epoch": 5.16, + "learning_rate": 2.424404631854927e-06, + "loss": 1.2349, + "step": 23627 + }, + { + "epoch": 5.16, + "learning_rate": 2.424295389993446e-06, + "loss": 1.9585, + "step": 23628 + }, + { + "epoch": 5.16, + "learning_rate": 2.424186148131964e-06, + "loss": 0.3174, + "step": 23629 + }, + { + "epoch": 5.16, + "learning_rate": 2.424076906270483e-06, + "loss": 1.9204, + "step": 23630 + }, + { + "epoch": 5.16, + "learning_rate": 2.4239676644090017e-06, + "loss": 2.2328, + "step": 23631 + }, + { + "epoch": 5.16, + "learning_rate": 2.4238584225475203e-06, + "loss": 1.5972, + "step": 23632 + }, + { + "epoch": 5.16, + "learning_rate": 2.423749180686039e-06, + "loss": 1.3492, + "step": 23633 + }, + { + "epoch": 5.16, + "learning_rate": 2.4236399388245576e-06, + "loss": 1.5547, + "step": 23634 + }, + { + "epoch": 5.16, + "learning_rate": 2.4235306969630766e-06, + "loss": 1.4001, + "step": 23635 + }, + { + "epoch": 5.16, + "learning_rate": 2.423421455101595e-06, + "loss": 1.9209, + "step": 23636 + }, + { + "epoch": 5.16, + "learning_rate": 2.423312213240114e-06, + "loss": 0.1609, + "step": 23637 + }, + { + "epoch": 5.16, + "learning_rate": 2.4232029713786324e-06, + "loss": 2.924, + "step": 23638 + }, + { + "epoch": 5.16, + "learning_rate": 2.423093729517151e-06, + "loss": 1.0286, + "step": 23639 + }, + { + "epoch": 5.16, + "learning_rate": 2.42298448765567e-06, + "loss": 1.1118, + "step": 23640 + }, + { + "epoch": 5.16, + "learning_rate": 2.4228752457941883e-06, + "loss": 2.0269, + "step": 23641 + }, + { + "epoch": 5.16, + "learning_rate": 2.4227660039327073e-06, + "loss": 0.2487, + "step": 23642 + }, + { + "epoch": 5.16, + "learning_rate": 2.422656762071226e-06, + "loss": 0.5686, + "step": 23643 + }, + { + "epoch": 5.16, + "learning_rate": 2.4225475202097445e-06, + "loss": 0.7327, + "step": 23644 + }, + { + "epoch": 5.16, + "learning_rate": 2.422438278348263e-06, + "loss": 2.102, + "step": 23645 + }, + { + "epoch": 5.16, + "learning_rate": 2.4223290364867817e-06, + "loss": 1.9361, + "step": 23646 + }, + { + "epoch": 5.16, + "learning_rate": 2.4222197946253008e-06, + "loss": 1.451, + "step": 23647 + }, + { + "epoch": 5.16, + "learning_rate": 2.422110552763819e-06, + "loss": 2.0144, + "step": 23648 + }, + { + "epoch": 5.16, + "learning_rate": 2.422001310902338e-06, + "loss": 0.9029, + "step": 23649 + }, + { + "epoch": 5.16, + "learning_rate": 2.4218920690408566e-06, + "loss": 2.9219, + "step": 23650 + }, + { + "epoch": 5.16, + "learning_rate": 2.4217828271793752e-06, + "loss": 0.8183, + "step": 23651 + }, + { + "epoch": 5.16, + "learning_rate": 2.4216735853178943e-06, + "loss": 1.0759, + "step": 23652 + }, + { + "epoch": 5.16, + "learning_rate": 2.4215643434564124e-06, + "loss": 0.9515, + "step": 23653 + }, + { + "epoch": 5.16, + "learning_rate": 2.4214551015949315e-06, + "loss": 1.077, + "step": 23654 + }, + { + "epoch": 5.16, + "learning_rate": 2.42134585973345e-06, + "loss": 0.5031, + "step": 23655 + }, + { + "epoch": 5.16, + "learning_rate": 2.4212366178719687e-06, + "loss": 0.7849, + "step": 23656 + }, + { + "epoch": 5.16, + "learning_rate": 2.4211273760104877e-06, + "loss": 1.3993, + "step": 23657 + }, + { + "epoch": 5.16, + "learning_rate": 2.421018134149006e-06, + "loss": 1.1073, + "step": 23658 + }, + { + "epoch": 5.16, + "learning_rate": 2.420908892287525e-06, + "loss": 0.9701, + "step": 23659 + }, + { + "epoch": 5.16, + "learning_rate": 2.4207996504260436e-06, + "loss": 1.1772, + "step": 23660 + }, + { + "epoch": 5.16, + "learning_rate": 2.420690408564562e-06, + "loss": 0.8515, + "step": 23661 + }, + { + "epoch": 5.16, + "learning_rate": 2.420581166703081e-06, + "loss": 1.3931, + "step": 23662 + }, + { + "epoch": 5.16, + "learning_rate": 2.4204719248415994e-06, + "loss": 2.0181, + "step": 23663 + }, + { + "epoch": 5.16, + "learning_rate": 2.4203626829801184e-06, + "loss": 1.463, + "step": 23664 + }, + { + "epoch": 5.16, + "learning_rate": 2.4202534411186366e-06, + "loss": 2.5468, + "step": 23665 + }, + { + "epoch": 5.16, + "learning_rate": 2.4201441992571557e-06, + "loss": 0.9918, + "step": 23666 + }, + { + "epoch": 5.17, + "learning_rate": 2.4200349573956743e-06, + "loss": 2.9683, + "step": 23667 + }, + { + "epoch": 5.17, + "learning_rate": 2.419925715534193e-06, + "loss": 0.6536, + "step": 23668 + }, + { + "epoch": 5.17, + "learning_rate": 2.419816473672712e-06, + "loss": 2.3192, + "step": 23669 + }, + { + "epoch": 5.17, + "learning_rate": 2.41970723181123e-06, + "loss": 0.4295, + "step": 23670 + }, + { + "epoch": 5.17, + "learning_rate": 2.419597989949749e-06, + "loss": 1.6267, + "step": 23671 + }, + { + "epoch": 5.17, + "learning_rate": 2.4194887480882677e-06, + "loss": 1.0054, + "step": 23672 + }, + { + "epoch": 5.17, + "learning_rate": 2.4193795062267864e-06, + "loss": 0.9892, + "step": 23673 + }, + { + "epoch": 5.17, + "learning_rate": 2.419270264365305e-06, + "loss": 1.3314, + "step": 23674 + }, + { + "epoch": 5.17, + "learning_rate": 2.4191610225038236e-06, + "loss": 2.8258, + "step": 23675 + }, + { + "epoch": 5.17, + "learning_rate": 2.4190517806423426e-06, + "loss": 2.7685, + "step": 23676 + }, + { + "epoch": 5.17, + "learning_rate": 2.418942538780861e-06, + "loss": 1.0353, + "step": 23677 + }, + { + "epoch": 5.17, + "learning_rate": 2.41883329691938e-06, + "loss": 0.8235, + "step": 23678 + }, + { + "epoch": 5.17, + "learning_rate": 2.4187240550578984e-06, + "loss": 1.3381, + "step": 23679 + }, + { + "epoch": 5.17, + "learning_rate": 2.418614813196417e-06, + "loss": 2.1414, + "step": 23680 + }, + { + "epoch": 5.17, + "learning_rate": 2.4185055713349357e-06, + "loss": 0.9589, + "step": 23681 + }, + { + "epoch": 5.17, + "learning_rate": 2.4183963294734543e-06, + "loss": 1.4842, + "step": 23682 + }, + { + "epoch": 5.17, + "learning_rate": 2.4182870876119733e-06, + "loss": 3.0553, + "step": 23683 + }, + { + "epoch": 5.17, + "learning_rate": 2.418177845750492e-06, + "loss": 1.8294, + "step": 23684 + }, + { + "epoch": 5.17, + "learning_rate": 2.4180686038890105e-06, + "loss": 1.5073, + "step": 23685 + }, + { + "epoch": 5.17, + "learning_rate": 2.417959362027529e-06, + "loss": 1.3798, + "step": 23686 + }, + { + "epoch": 5.17, + "learning_rate": 2.4178501201660478e-06, + "loss": 1.7428, + "step": 23687 + }, + { + "epoch": 5.17, + "learning_rate": 2.4177408783045664e-06, + "loss": 0.8672, + "step": 23688 + }, + { + "epoch": 5.17, + "learning_rate": 2.417631636443085e-06, + "loss": 2.2431, + "step": 23689 + }, + { + "epoch": 5.17, + "learning_rate": 2.417522394581604e-06, + "loss": 1.6167, + "step": 23690 + }, + { + "epoch": 5.17, + "learning_rate": 2.4174131527201226e-06, + "loss": 1.902, + "step": 23691 + }, + { + "epoch": 5.17, + "learning_rate": 2.4173039108586412e-06, + "loss": 3.0925, + "step": 23692 + }, + { + "epoch": 5.17, + "learning_rate": 2.41719466899716e-06, + "loss": 1.3803, + "step": 23693 + }, + { + "epoch": 5.17, + "learning_rate": 2.4170854271356785e-06, + "loss": 1.4639, + "step": 23694 + }, + { + "epoch": 5.17, + "learning_rate": 2.416976185274197e-06, + "loss": 1.3326, + "step": 23695 + }, + { + "epoch": 5.17, + "learning_rate": 2.416866943412716e-06, + "loss": 2.4317, + "step": 23696 + }, + { + "epoch": 5.17, + "learning_rate": 2.4167577015512347e-06, + "loss": 1.158, + "step": 23697 + }, + { + "epoch": 5.17, + "learning_rate": 2.4166484596897533e-06, + "loss": 1.8232, + "step": 23698 + }, + { + "epoch": 5.17, + "learning_rate": 2.416539217828272e-06, + "loss": 0.1185, + "step": 23699 + }, + { + "epoch": 5.17, + "learning_rate": 2.4164299759667905e-06, + "loss": 1.121, + "step": 23700 + }, + { + "epoch": 5.17, + "learning_rate": 2.4163207341053096e-06, + "loss": 1.8546, + "step": 23701 + }, + { + "epoch": 5.17, + "learning_rate": 2.4162114922438278e-06, + "loss": 3.3582, + "step": 23702 + }, + { + "epoch": 5.17, + "learning_rate": 2.416102250382347e-06, + "loss": 0.8715, + "step": 23703 + }, + { + "epoch": 5.17, + "learning_rate": 2.4159930085208654e-06, + "loss": 2.3736, + "step": 23704 + }, + { + "epoch": 5.17, + "learning_rate": 2.415883766659384e-06, + "loss": 0.2783, + "step": 23705 + }, + { + "epoch": 5.17, + "learning_rate": 2.4157745247979026e-06, + "loss": 0.5973, + "step": 23706 + }, + { + "epoch": 5.17, + "learning_rate": 2.4156652829364213e-06, + "loss": 1.7118, + "step": 23707 + }, + { + "epoch": 5.17, + "learning_rate": 2.4155560410749403e-06, + "loss": 0.7466, + "step": 23708 + }, + { + "epoch": 5.17, + "learning_rate": 2.415446799213459e-06, + "loss": 0.5218, + "step": 23709 + }, + { + "epoch": 5.17, + "learning_rate": 2.4153375573519775e-06, + "loss": 3.1695, + "step": 23710 + }, + { + "epoch": 5.17, + "learning_rate": 2.415228315490496e-06, + "loss": 2.7578, + "step": 23711 + }, + { + "epoch": 5.18, + "learning_rate": 2.4151190736290147e-06, + "loss": 2.2958, + "step": 23712 + }, + { + "epoch": 5.18, + "learning_rate": 2.4150098317675338e-06, + "loss": 1.8806, + "step": 23713 + }, + { + "epoch": 5.18, + "learning_rate": 2.414900589906052e-06, + "loss": 1.3622, + "step": 23714 + }, + { + "epoch": 5.18, + "learning_rate": 2.414791348044571e-06, + "loss": 0.868, + "step": 23715 + }, + { + "epoch": 5.18, + "learning_rate": 2.4146821061830896e-06, + "loss": 1.9009, + "step": 23716 + }, + { + "epoch": 5.18, + "learning_rate": 2.414572864321608e-06, + "loss": 1.8306, + "step": 23717 + }, + { + "epoch": 5.18, + "learning_rate": 2.414463622460127e-06, + "loss": 1.1735, + "step": 23718 + }, + { + "epoch": 5.18, + "learning_rate": 2.4143543805986454e-06, + "loss": 0.9166, + "step": 23719 + }, + { + "epoch": 5.18, + "learning_rate": 2.4142451387371645e-06, + "loss": 1.3341, + "step": 23720 + }, + { + "epoch": 5.18, + "learning_rate": 2.4141358968756827e-06, + "loss": 1.3866, + "step": 23721 + }, + { + "epoch": 5.18, + "learning_rate": 2.4140266550142017e-06, + "loss": 0.7002, + "step": 23722 + }, + { + "epoch": 5.18, + "learning_rate": 2.4139174131527203e-06, + "loss": 2.1485, + "step": 23723 + }, + { + "epoch": 5.18, + "learning_rate": 2.413808171291239e-06, + "loss": 0.8486, + "step": 23724 + }, + { + "epoch": 5.18, + "learning_rate": 2.413698929429758e-06, + "loss": 1.4361, + "step": 23725 + }, + { + "epoch": 5.18, + "learning_rate": 2.413589687568276e-06, + "loss": 0.8481, + "step": 23726 + }, + { + "epoch": 5.18, + "learning_rate": 2.413480445706795e-06, + "loss": 1.3143, + "step": 23727 + }, + { + "epoch": 5.18, + "learning_rate": 2.4133712038453138e-06, + "loss": 0.8616, + "step": 23728 + }, + { + "epoch": 5.18, + "learning_rate": 2.4132619619838324e-06, + "loss": 1.4272, + "step": 23729 + }, + { + "epoch": 5.18, + "learning_rate": 2.413152720122351e-06, + "loss": 1.2544, + "step": 23730 + }, + { + "epoch": 5.18, + "learning_rate": 2.4130434782608696e-06, + "loss": 1.6944, + "step": 23731 + }, + { + "epoch": 5.18, + "learning_rate": 2.4129342363993886e-06, + "loss": 1.0221, + "step": 23732 + }, + { + "epoch": 5.18, + "learning_rate": 2.412824994537907e-06, + "loss": 0.3484, + "step": 23733 + }, + { + "epoch": 5.18, + "learning_rate": 2.412715752676426e-06, + "loss": 1.452, + "step": 23734 + }, + { + "epoch": 5.18, + "learning_rate": 2.4126065108149445e-06, + "loss": 0.852, + "step": 23735 + }, + { + "epoch": 5.18, + "learning_rate": 2.412497268953463e-06, + "loss": 0.5226, + "step": 23736 + }, + { + "epoch": 5.18, + "learning_rate": 2.412388027091982e-06, + "loss": 1.6084, + "step": 23737 + }, + { + "epoch": 5.18, + "learning_rate": 2.4122787852305003e-06, + "loss": 1.5064, + "step": 23738 + }, + { + "epoch": 5.18, + "learning_rate": 2.4121695433690193e-06, + "loss": 1.4151, + "step": 23739 + }, + { + "epoch": 5.18, + "learning_rate": 2.412060301507538e-06, + "loss": 0.5951, + "step": 23740 + }, + { + "epoch": 5.18, + "learning_rate": 2.4119510596460566e-06, + "loss": 1.1671, + "step": 23741 + }, + { + "epoch": 5.18, + "learning_rate": 2.411841817784575e-06, + "loss": 0.2798, + "step": 23742 + }, + { + "epoch": 5.18, + "learning_rate": 2.4117325759230938e-06, + "loss": 0.9208, + "step": 23743 + }, + { + "epoch": 5.18, + "learning_rate": 2.411623334061613e-06, + "loss": 1.8599, + "step": 23744 + }, + { + "epoch": 5.18, + "learning_rate": 2.411514092200131e-06, + "loss": 1.201, + "step": 23745 + }, + { + "epoch": 5.18, + "learning_rate": 2.41140485033865e-06, + "loss": 2.1266, + "step": 23746 + }, + { + "epoch": 5.18, + "learning_rate": 2.4112956084771687e-06, + "loss": 0.3409, + "step": 23747 + }, + { + "epoch": 5.18, + "learning_rate": 2.4111863666156873e-06, + "loss": 2.037, + "step": 23748 + }, + { + "epoch": 5.18, + "learning_rate": 2.4110771247542063e-06, + "loss": 0.9189, + "step": 23749 + }, + { + "epoch": 5.18, + "learning_rate": 2.4109678828927245e-06, + "loss": 2.9112, + "step": 23750 + }, + { + "epoch": 5.18, + "learning_rate": 2.4108586410312435e-06, + "loss": 0.8161, + "step": 23751 + }, + { + "epoch": 5.18, + "learning_rate": 2.410749399169762e-06, + "loss": 1.749, + "step": 23752 + }, + { + "epoch": 5.18, + "learning_rate": 2.4106401573082807e-06, + "loss": 0.1165, + "step": 23753 + }, + { + "epoch": 5.18, + "learning_rate": 2.4105309154467994e-06, + "loss": 0.549, + "step": 23754 + }, + { + "epoch": 5.18, + "learning_rate": 2.410421673585318e-06, + "loss": 1.725, + "step": 23755 + }, + { + "epoch": 5.18, + "learning_rate": 2.410312431723837e-06, + "loss": 1.8755, + "step": 23756 + }, + { + "epoch": 5.18, + "learning_rate": 2.4102031898623556e-06, + "loss": 2.9915, + "step": 23757 + }, + { + "epoch": 5.19, + "learning_rate": 2.4100939480008742e-06, + "loss": 1.8479, + "step": 23758 + }, + { + "epoch": 5.19, + "learning_rate": 2.409984706139393e-06, + "loss": 0.9831, + "step": 23759 + }, + { + "epoch": 5.19, + "learning_rate": 2.4098754642779114e-06, + "loss": 2.1678, + "step": 23760 + }, + { + "epoch": 5.19, + "learning_rate": 2.40976622241643e-06, + "loss": 1.7573, + "step": 23761 + }, + { + "epoch": 5.19, + "learning_rate": 2.4096569805549487e-06, + "loss": 3.0992, + "step": 23762 + }, + { + "epoch": 5.19, + "learning_rate": 2.4095477386934677e-06, + "loss": 2.162, + "step": 23763 + }, + { + "epoch": 5.19, + "learning_rate": 2.4094384968319863e-06, + "loss": 2.4625, + "step": 23764 + }, + { + "epoch": 5.19, + "learning_rate": 2.409329254970505e-06, + "loss": 0.7041, + "step": 23765 + }, + { + "epoch": 5.19, + "learning_rate": 2.4092200131090235e-06, + "loss": 0.4748, + "step": 23766 + }, + { + "epoch": 5.19, + "learning_rate": 2.409110771247542e-06, + "loss": 1.4385, + "step": 23767 + }, + { + "epoch": 5.19, + "learning_rate": 2.4090015293860608e-06, + "loss": 2.4075, + "step": 23768 + }, + { + "epoch": 5.19, + "learning_rate": 2.40889228752458e-06, + "loss": 1.5657, + "step": 23769 + }, + { + "epoch": 5.19, + "learning_rate": 2.4087830456630984e-06, + "loss": 0.4113, + "step": 23770 + }, + { + "epoch": 5.19, + "learning_rate": 2.408673803801617e-06, + "loss": 2.366, + "step": 23771 + }, + { + "epoch": 5.19, + "learning_rate": 2.4085645619401356e-06, + "loss": 1.2027, + "step": 23772 + }, + { + "epoch": 5.19, + "learning_rate": 2.4084553200786542e-06, + "loss": 1.2552, + "step": 23773 + }, + { + "epoch": 5.19, + "learning_rate": 2.408346078217173e-06, + "loss": 1.5628, + "step": 23774 + }, + { + "epoch": 5.19, + "learning_rate": 2.408236836355692e-06, + "loss": 1.1994, + "step": 23775 + }, + { + "epoch": 5.19, + "learning_rate": 2.4081275944942105e-06, + "loss": 2.1468, + "step": 23776 + }, + { + "epoch": 5.19, + "learning_rate": 2.408018352632729e-06, + "loss": 1.4107, + "step": 23777 + }, + { + "epoch": 5.19, + "learning_rate": 2.4079091107712477e-06, + "loss": 1.7443, + "step": 23778 + }, + { + "epoch": 5.19, + "learning_rate": 2.4077998689097663e-06, + "loss": 0.5021, + "step": 23779 + }, + { + "epoch": 5.19, + "learning_rate": 2.407690627048285e-06, + "loss": 2.3854, + "step": 23780 + }, + { + "epoch": 5.19, + "learning_rate": 2.407581385186804e-06, + "loss": 0.9572, + "step": 23781 + }, + { + "epoch": 5.19, + "learning_rate": 2.4074721433253226e-06, + "loss": 1.9534, + "step": 23782 + }, + { + "epoch": 5.19, + "learning_rate": 2.407362901463841e-06, + "loss": 0.7949, + "step": 23783 + }, + { + "epoch": 5.19, + "learning_rate": 2.40725365960236e-06, + "loss": 1.2801, + "step": 23784 + }, + { + "epoch": 5.19, + "learning_rate": 2.4071444177408784e-06, + "loss": 0.9974, + "step": 23785 + }, + { + "epoch": 5.19, + "learning_rate": 2.407035175879397e-06, + "loss": 0.6364, + "step": 23786 + }, + { + "epoch": 5.19, + "learning_rate": 2.4069259340179156e-06, + "loss": 1.9725, + "step": 23787 + }, + { + "epoch": 5.19, + "learning_rate": 2.4068166921564347e-06, + "loss": 0.5524, + "step": 23788 + }, + { + "epoch": 5.19, + "learning_rate": 2.4067074502949533e-06, + "loss": 2.9255, + "step": 23789 + }, + { + "epoch": 5.19, + "learning_rate": 2.406598208433472e-06, + "loss": 2.248, + "step": 23790 + }, + { + "epoch": 5.19, + "learning_rate": 2.4064889665719905e-06, + "loss": 5.2791, + "step": 23791 + }, + { + "epoch": 5.19, + "learning_rate": 2.406379724710509e-06, + "loss": 2.4497, + "step": 23792 + }, + { + "epoch": 5.19, + "learning_rate": 2.406270482849028e-06, + "loss": 3.097, + "step": 23793 + }, + { + "epoch": 5.19, + "learning_rate": 2.4061612409875463e-06, + "loss": 2.3928, + "step": 23794 + }, + { + "epoch": 5.19, + "learning_rate": 2.4060519991260654e-06, + "loss": 1.5969, + "step": 23795 + }, + { + "epoch": 5.19, + "learning_rate": 2.405942757264584e-06, + "loss": 2.378, + "step": 23796 + }, + { + "epoch": 5.19, + "learning_rate": 2.4058335154031026e-06, + "loss": 2.6523, + "step": 23797 + }, + { + "epoch": 5.19, + "learning_rate": 2.405724273541621e-06, + "loss": 0.6931, + "step": 23798 + }, + { + "epoch": 5.19, + "learning_rate": 2.40561503168014e-06, + "loss": 1.4164, + "step": 23799 + }, + { + "epoch": 5.19, + "learning_rate": 2.405505789818659e-06, + "loss": 0.7704, + "step": 23800 + }, + { + "epoch": 5.19, + "learning_rate": 2.4053965479571775e-06, + "loss": 2.2214, + "step": 23801 + }, + { + "epoch": 5.19, + "learning_rate": 2.405287306095696e-06, + "loss": 1.534, + "step": 23802 + }, + { + "epoch": 5.19, + "learning_rate": 2.4051780642342147e-06, + "loss": 2.392, + "step": 23803 + }, + { + "epoch": 5.2, + "learning_rate": 2.4050688223727333e-06, + "loss": 1.3188, + "step": 23804 + }, + { + "epoch": 5.2, + "learning_rate": 2.4049595805112523e-06, + "loss": 2.3026, + "step": 23805 + }, + { + "epoch": 5.2, + "learning_rate": 2.4048503386497705e-06, + "loss": 1.3797, + "step": 23806 + }, + { + "epoch": 5.2, + "learning_rate": 2.4047410967882895e-06, + "loss": 2.4578, + "step": 23807 + }, + { + "epoch": 5.2, + "learning_rate": 2.404631854926808e-06, + "loss": 2.1907, + "step": 23808 + }, + { + "epoch": 5.2, + "learning_rate": 2.4045226130653268e-06, + "loss": 1.8244, + "step": 23809 + }, + { + "epoch": 5.2, + "learning_rate": 2.404413371203846e-06, + "loss": 2.0675, + "step": 23810 + }, + { + "epoch": 5.2, + "learning_rate": 2.404304129342364e-06, + "loss": 2.1123, + "step": 23811 + }, + { + "epoch": 5.2, + "learning_rate": 2.404194887480883e-06, + "loss": 2.0865, + "step": 23812 + }, + { + "epoch": 5.2, + "learning_rate": 2.4040856456194016e-06, + "loss": 1.0654, + "step": 23813 + }, + { + "epoch": 5.2, + "learning_rate": 2.4039764037579202e-06, + "loss": 2.0165, + "step": 23814 + }, + { + "epoch": 5.2, + "learning_rate": 2.403867161896439e-06, + "loss": 1.0714, + "step": 23815 + }, + { + "epoch": 5.2, + "learning_rate": 2.4037579200349575e-06, + "loss": 2.7378, + "step": 23816 + }, + { + "epoch": 5.2, + "learning_rate": 2.4036486781734765e-06, + "loss": 2.3964, + "step": 23817 + }, + { + "epoch": 5.2, + "learning_rate": 2.4035394363119947e-06, + "loss": 1.2946, + "step": 23818 + }, + { + "epoch": 5.2, + "learning_rate": 2.4034301944505137e-06, + "loss": 0.5721, + "step": 23819 + }, + { + "epoch": 5.2, + "learning_rate": 2.4033209525890323e-06, + "loss": 2.0777, + "step": 23820 + }, + { + "epoch": 5.2, + "learning_rate": 2.403211710727551e-06, + "loss": 1.6873, + "step": 23821 + }, + { + "epoch": 5.2, + "learning_rate": 2.40310246886607e-06, + "loss": 0.4962, + "step": 23822 + }, + { + "epoch": 5.2, + "learning_rate": 2.402993227004588e-06, + "loss": 1.483, + "step": 23823 + }, + { + "epoch": 5.2, + "learning_rate": 2.402883985143107e-06, + "loss": 0.6674, + "step": 23824 + }, + { + "epoch": 5.2, + "learning_rate": 2.402774743281626e-06, + "loss": 0.9768, + "step": 23825 + }, + { + "epoch": 5.2, + "learning_rate": 2.4026655014201444e-06, + "loss": 1.428, + "step": 23826 + }, + { + "epoch": 5.2, + "learning_rate": 2.402556259558663e-06, + "loss": 1.9154, + "step": 23827 + }, + { + "epoch": 5.2, + "learning_rate": 2.4024470176971817e-06, + "loss": 1.31, + "step": 23828 + }, + { + "epoch": 5.2, + "learning_rate": 2.4023377758357007e-06, + "loss": 2.5196, + "step": 23829 + }, + { + "epoch": 5.2, + "learning_rate": 2.402228533974219e-06, + "loss": 0.4299, + "step": 23830 + }, + { + "epoch": 5.2, + "learning_rate": 2.402119292112738e-06, + "loss": 1.2977, + "step": 23831 + }, + { + "epoch": 5.2, + "learning_rate": 2.4020100502512565e-06, + "loss": 0.4409, + "step": 23832 + }, + { + "epoch": 5.2, + "learning_rate": 2.401900808389775e-06, + "loss": 0.9629, + "step": 23833 + }, + { + "epoch": 5.2, + "learning_rate": 2.4017915665282937e-06, + "loss": 1.6626, + "step": 23834 + }, + { + "epoch": 5.2, + "learning_rate": 2.4016823246668124e-06, + "loss": 1.9861, + "step": 23835 + }, + { + "epoch": 5.2, + "learning_rate": 2.4015730828053314e-06, + "loss": 1.5431, + "step": 23836 + }, + { + "epoch": 5.2, + "learning_rate": 2.40146384094385e-06, + "loss": 1.4813, + "step": 23837 + }, + { + "epoch": 5.2, + "learning_rate": 2.4013545990823686e-06, + "loss": 4.2438, + "step": 23838 + }, + { + "epoch": 5.2, + "learning_rate": 2.4012453572208872e-06, + "loss": 0.8722, + "step": 23839 + }, + { + "epoch": 5.2, + "learning_rate": 2.401136115359406e-06, + "loss": 2.3003, + "step": 23840 + }, + { + "epoch": 5.2, + "learning_rate": 2.401026873497925e-06, + "loss": 0.6328, + "step": 23841 + }, + { + "epoch": 5.2, + "learning_rate": 2.400917631636443e-06, + "loss": 1.4902, + "step": 23842 + }, + { + "epoch": 5.2, + "learning_rate": 2.400808389774962e-06, + "loss": 1.7604, + "step": 23843 + }, + { + "epoch": 5.2, + "learning_rate": 2.4006991479134807e-06, + "loss": 1.7755, + "step": 23844 + }, + { + "epoch": 5.2, + "learning_rate": 2.4005899060519993e-06, + "loss": 0.5934, + "step": 23845 + }, + { + "epoch": 5.2, + "learning_rate": 2.400480664190518e-06, + "loss": 0.8498, + "step": 23846 + }, + { + "epoch": 5.2, + "learning_rate": 2.4003714223290365e-06, + "loss": 1.1871, + "step": 23847 + }, + { + "epoch": 5.2, + "learning_rate": 2.4002621804675556e-06, + "loss": 1.8407, + "step": 23848 + }, + { + "epoch": 5.2, + "learning_rate": 2.400152938606074e-06, + "loss": 1.3035, + "step": 23849 + }, + { + "epoch": 5.21, + "learning_rate": 2.4000436967445928e-06, + "loss": 1.8179, + "step": 23850 + }, + { + "epoch": 5.21, + "learning_rate": 2.3999344548831114e-06, + "loss": 1.3993, + "step": 23851 + }, + { + "epoch": 5.21, + "learning_rate": 2.39982521302163e-06, + "loss": 1.0573, + "step": 23852 + }, + { + "epoch": 5.21, + "learning_rate": 2.3997159711601486e-06, + "loss": 1.3698, + "step": 23853 + }, + { + "epoch": 5.21, + "learning_rate": 2.3996067292986677e-06, + "loss": 0.4927, + "step": 23854 + }, + { + "epoch": 5.21, + "learning_rate": 2.3994974874371863e-06, + "loss": 1.1531, + "step": 23855 + }, + { + "epoch": 5.21, + "learning_rate": 2.399388245575705e-06, + "loss": 2.7035, + "step": 23856 + }, + { + "epoch": 5.21, + "learning_rate": 2.3992790037142235e-06, + "loss": 0.2999, + "step": 23857 + }, + { + "epoch": 5.21, + "learning_rate": 2.399169761852742e-06, + "loss": 1.3783, + "step": 23858 + }, + { + "epoch": 5.21, + "learning_rate": 2.3990605199912607e-06, + "loss": 0.3216, + "step": 23859 + }, + { + "epoch": 5.21, + "learning_rate": 2.3989512781297793e-06, + "loss": 1.1074, + "step": 23860 + }, + { + "epoch": 5.21, + "learning_rate": 2.3988420362682984e-06, + "loss": 0.7143, + "step": 23861 + }, + { + "epoch": 5.21, + "learning_rate": 2.398732794406817e-06, + "loss": 2.5418, + "step": 23862 + }, + { + "epoch": 5.21, + "learning_rate": 2.3986235525453356e-06, + "loss": 2.157, + "step": 23863 + }, + { + "epoch": 5.21, + "learning_rate": 2.398514310683854e-06, + "loss": 3.541, + "step": 23864 + }, + { + "epoch": 5.21, + "learning_rate": 2.398405068822373e-06, + "loss": 0.94, + "step": 23865 + }, + { + "epoch": 5.21, + "learning_rate": 2.398295826960892e-06, + "loss": 2.5405, + "step": 23866 + }, + { + "epoch": 5.21, + "learning_rate": 2.39818658509941e-06, + "loss": 0.9754, + "step": 23867 + }, + { + "epoch": 5.21, + "learning_rate": 2.398077343237929e-06, + "loss": 3.1769, + "step": 23868 + }, + { + "epoch": 5.21, + "learning_rate": 2.3979681013764477e-06, + "loss": 0.6717, + "step": 23869 + }, + { + "epoch": 5.21, + "learning_rate": 2.3978588595149663e-06, + "loss": 0.7274, + "step": 23870 + }, + { + "epoch": 5.21, + "learning_rate": 2.397749617653485e-06, + "loss": 1.2979, + "step": 23871 + }, + { + "epoch": 5.21, + "learning_rate": 2.3976403757920035e-06, + "loss": 1.3936, + "step": 23872 + }, + { + "epoch": 5.21, + "learning_rate": 2.3975311339305225e-06, + "loss": 1.1108, + "step": 23873 + }, + { + "epoch": 5.21, + "learning_rate": 2.397421892069041e-06, + "loss": 1.2716, + "step": 23874 + }, + { + "epoch": 5.21, + "learning_rate": 2.3973126502075598e-06, + "loss": 1.9336, + "step": 23875 + }, + { + "epoch": 5.21, + "learning_rate": 2.3972034083460784e-06, + "loss": 0.1575, + "step": 23876 + }, + { + "epoch": 5.21, + "learning_rate": 2.397094166484597e-06, + "loss": 2.2549, + "step": 23877 + }, + { + "epoch": 5.21, + "learning_rate": 2.396984924623116e-06, + "loss": 1.2439, + "step": 23878 + }, + { + "epoch": 5.21, + "learning_rate": 2.396875682761634e-06, + "loss": 0.6482, + "step": 23879 + }, + { + "epoch": 5.21, + "learning_rate": 2.3967664409001532e-06, + "loss": 2.1925, + "step": 23880 + }, + { + "epoch": 5.21, + "learning_rate": 2.396657199038672e-06, + "loss": 2.1968, + "step": 23881 + }, + { + "epoch": 5.21, + "learning_rate": 2.3965479571771905e-06, + "loss": 2.0919, + "step": 23882 + }, + { + "epoch": 5.21, + "learning_rate": 2.396438715315709e-06, + "loss": 1.5422, + "step": 23883 + }, + { + "epoch": 5.21, + "learning_rate": 2.3963294734542277e-06, + "loss": 1.7936, + "step": 23884 + }, + { + "epoch": 5.21, + "learning_rate": 2.3962202315927467e-06, + "loss": 0.4229, + "step": 23885 + }, + { + "epoch": 5.21, + "learning_rate": 2.396110989731265e-06, + "loss": 2.2902, + "step": 23886 + }, + { + "epoch": 5.21, + "learning_rate": 2.396001747869784e-06, + "loss": 0.7887, + "step": 23887 + }, + { + "epoch": 5.21, + "learning_rate": 2.3958925060083025e-06, + "loss": 1.5392, + "step": 23888 + }, + { + "epoch": 5.21, + "learning_rate": 2.395783264146821e-06, + "loss": 1.6223, + "step": 23889 + }, + { + "epoch": 5.21, + "learning_rate": 2.39567402228534e-06, + "loss": 1.8858, + "step": 23890 + }, + { + "epoch": 5.21, + "learning_rate": 2.3955647804238584e-06, + "loss": 3.1792, + "step": 23891 + }, + { + "epoch": 5.21, + "learning_rate": 2.3954555385623774e-06, + "loss": 1.4296, + "step": 23892 + }, + { + "epoch": 5.21, + "learning_rate": 2.395346296700896e-06, + "loss": 0.9209, + "step": 23893 + }, + { + "epoch": 5.21, + "learning_rate": 2.3952370548394146e-06, + "loss": 1.0363, + "step": 23894 + }, + { + "epoch": 5.21, + "learning_rate": 2.3951278129779332e-06, + "loss": 1.2545, + "step": 23895 + }, + { + "epoch": 5.22, + "learning_rate": 2.395018571116452e-06, + "loss": 1.2985, + "step": 23896 + }, + { + "epoch": 5.22, + "learning_rate": 2.394909329254971e-06, + "loss": 0.9046, + "step": 23897 + }, + { + "epoch": 5.22, + "learning_rate": 2.3948000873934895e-06, + "loss": 2.2314, + "step": 23898 + }, + { + "epoch": 5.22, + "learning_rate": 2.394690845532008e-06, + "loss": 1.6803, + "step": 23899 + }, + { + "epoch": 5.22, + "learning_rate": 2.3945816036705267e-06, + "loss": 1.8432, + "step": 23900 + }, + { + "epoch": 5.22, + "learning_rate": 2.3944723618090453e-06, + "loss": 2.2983, + "step": 23901 + }, + { + "epoch": 5.22, + "learning_rate": 2.3943631199475644e-06, + "loss": 0.8947, + "step": 23902 + }, + { + "epoch": 5.22, + "learning_rate": 2.3942538780860826e-06, + "loss": 1.7931, + "step": 23903 + }, + { + "epoch": 5.22, + "learning_rate": 2.3941446362246016e-06, + "loss": 0.0898, + "step": 23904 + }, + { + "epoch": 5.22, + "learning_rate": 2.39403539436312e-06, + "loss": 0.7176, + "step": 23905 + }, + { + "epoch": 5.22, + "learning_rate": 2.393926152501639e-06, + "loss": 0.526, + "step": 23906 + }, + { + "epoch": 5.22, + "learning_rate": 2.3938169106401574e-06, + "loss": 1.4047, + "step": 23907 + }, + { + "epoch": 5.22, + "learning_rate": 2.393707668778676e-06, + "loss": 1.7583, + "step": 23908 + }, + { + "epoch": 5.22, + "learning_rate": 2.393598426917195e-06, + "loss": 0.7228, + "step": 23909 + }, + { + "epoch": 5.22, + "learning_rate": 2.3934891850557137e-06, + "loss": 0.768, + "step": 23910 + }, + { + "epoch": 5.22, + "learning_rate": 2.3933799431942323e-06, + "loss": 0.7081, + "step": 23911 + }, + { + "epoch": 5.22, + "learning_rate": 2.393270701332751e-06, + "loss": 2.4386, + "step": 23912 + }, + { + "epoch": 5.22, + "learning_rate": 2.3931614594712695e-06, + "loss": 0.4671, + "step": 23913 + }, + { + "epoch": 5.22, + "learning_rate": 2.3930522176097885e-06, + "loss": 3.1214, + "step": 23914 + }, + { + "epoch": 5.22, + "learning_rate": 2.3929429757483067e-06, + "loss": 1.2091, + "step": 23915 + }, + { + "epoch": 5.22, + "learning_rate": 2.3928337338868258e-06, + "loss": 0.4858, + "step": 23916 + }, + { + "epoch": 5.22, + "learning_rate": 2.3927244920253444e-06, + "loss": 2.4949, + "step": 23917 + }, + { + "epoch": 5.22, + "learning_rate": 2.392615250163863e-06, + "loss": 1.1358, + "step": 23918 + }, + { + "epoch": 5.22, + "learning_rate": 2.3925060083023816e-06, + "loss": 0.1542, + "step": 23919 + }, + { + "epoch": 5.22, + "learning_rate": 2.3923967664409002e-06, + "loss": 0.9612, + "step": 23920 + }, + { + "epoch": 5.22, + "learning_rate": 2.3922875245794192e-06, + "loss": 2.7511, + "step": 23921 + }, + { + "epoch": 5.22, + "learning_rate": 2.392178282717938e-06, + "loss": 0.4685, + "step": 23922 + }, + { + "epoch": 5.22, + "learning_rate": 2.3920690408564565e-06, + "loss": 3.5521, + "step": 23923 + }, + { + "epoch": 5.22, + "learning_rate": 2.391959798994975e-06, + "loss": 2.2415, + "step": 23924 + }, + { + "epoch": 5.22, + "learning_rate": 2.3918505571334937e-06, + "loss": 1.1062, + "step": 23925 + }, + { + "epoch": 5.22, + "learning_rate": 2.3917413152720123e-06, + "loss": 2.0295, + "step": 23926 + }, + { + "epoch": 5.22, + "learning_rate": 2.391632073410531e-06, + "loss": 2.1909, + "step": 23927 + }, + { + "epoch": 5.22, + "learning_rate": 2.39152283154905e-06, + "loss": 1.699, + "step": 23928 + }, + { + "epoch": 5.22, + "learning_rate": 2.3914135896875686e-06, + "loss": 1.1581, + "step": 23929 + }, + { + "epoch": 5.22, + "learning_rate": 2.391304347826087e-06, + "loss": 2.3672, + "step": 23930 + }, + { + "epoch": 5.22, + "learning_rate": 2.3911951059646058e-06, + "loss": 1.4332, + "step": 23931 + }, + { + "epoch": 5.22, + "learning_rate": 2.3910858641031244e-06, + "loss": 1.4066, + "step": 23932 + }, + { + "epoch": 5.22, + "learning_rate": 2.390976622241643e-06, + "loss": 1.976, + "step": 23933 + }, + { + "epoch": 5.22, + "learning_rate": 2.390867380380162e-06, + "loss": 2.0121, + "step": 23934 + }, + { + "epoch": 5.22, + "learning_rate": 2.3907581385186806e-06, + "loss": 1.1612, + "step": 23935 + }, + { + "epoch": 5.22, + "learning_rate": 2.3906488966571993e-06, + "loss": 0.7542, + "step": 23936 + }, + { + "epoch": 5.22, + "learning_rate": 2.390539654795718e-06, + "loss": 1.0715, + "step": 23937 + }, + { + "epoch": 5.22, + "learning_rate": 2.3904304129342365e-06, + "loss": 0.9816, + "step": 23938 + }, + { + "epoch": 5.22, + "learning_rate": 2.390321171072755e-06, + "loss": 2.9778, + "step": 23939 + }, + { + "epoch": 5.22, + "learning_rate": 2.3902119292112737e-06, + "loss": 0.6673, + "step": 23940 + }, + { + "epoch": 5.23, + "learning_rate": 2.3901026873497927e-06, + "loss": 0.7967, + "step": 23941 + }, + { + "epoch": 5.23, + "learning_rate": 2.3899934454883113e-06, + "loss": 0.7417, + "step": 23942 + }, + { + "epoch": 5.23, + "learning_rate": 2.38988420362683e-06, + "loss": 0.6169, + "step": 23943 + }, + { + "epoch": 5.23, + "learning_rate": 2.3897749617653486e-06, + "loss": 1.1435, + "step": 23944 + }, + { + "epoch": 5.23, + "learning_rate": 2.389665719903867e-06, + "loss": 1.4196, + "step": 23945 + }, + { + "epoch": 5.23, + "learning_rate": 2.3895564780423862e-06, + "loss": 1.425, + "step": 23946 + }, + { + "epoch": 5.23, + "learning_rate": 2.389447236180905e-06, + "loss": 1.277, + "step": 23947 + }, + { + "epoch": 5.23, + "learning_rate": 2.3893379943194234e-06, + "loss": 1.2442, + "step": 23948 + }, + { + "epoch": 5.23, + "learning_rate": 2.389228752457942e-06, + "loss": 1.8175, + "step": 23949 + }, + { + "epoch": 5.23, + "learning_rate": 2.3891195105964607e-06, + "loss": 1.8387, + "step": 23950 + }, + { + "epoch": 5.23, + "learning_rate": 2.3890102687349797e-06, + "loss": 1.2492, + "step": 23951 + }, + { + "epoch": 5.23, + "learning_rate": 2.388901026873498e-06, + "loss": 2.3594, + "step": 23952 + }, + { + "epoch": 5.23, + "learning_rate": 2.388791785012017e-06, + "loss": 1.3141, + "step": 23953 + }, + { + "epoch": 5.23, + "learning_rate": 2.3886825431505355e-06, + "loss": 1.548, + "step": 23954 + }, + { + "epoch": 5.23, + "learning_rate": 2.388573301289054e-06, + "loss": 1.0675, + "step": 23955 + }, + { + "epoch": 5.23, + "learning_rate": 2.3884640594275728e-06, + "loss": 2.1126, + "step": 23956 + }, + { + "epoch": 5.23, + "learning_rate": 2.3883548175660914e-06, + "loss": 0.1094, + "step": 23957 + }, + { + "epoch": 5.23, + "learning_rate": 2.3882455757046104e-06, + "loss": 1.2109, + "step": 23958 + }, + { + "epoch": 5.23, + "learning_rate": 2.3881363338431286e-06, + "loss": 0.229, + "step": 23959 + }, + { + "epoch": 5.23, + "learning_rate": 2.3880270919816476e-06, + "loss": 0.8546, + "step": 23960 + }, + { + "epoch": 5.23, + "learning_rate": 2.3879178501201662e-06, + "loss": 1.5611, + "step": 23961 + }, + { + "epoch": 5.23, + "learning_rate": 2.387808608258685e-06, + "loss": 3.0341, + "step": 23962 + }, + { + "epoch": 5.23, + "learning_rate": 2.387699366397204e-06, + "loss": 2.1026, + "step": 23963 + }, + { + "epoch": 5.23, + "learning_rate": 2.387590124535722e-06, + "loss": 0.8647, + "step": 23964 + }, + { + "epoch": 5.23, + "learning_rate": 2.387480882674241e-06, + "loss": 3.158, + "step": 23965 + }, + { + "epoch": 5.23, + "learning_rate": 2.3873716408127597e-06, + "loss": 1.2857, + "step": 23966 + }, + { + "epoch": 5.23, + "learning_rate": 2.3872623989512783e-06, + "loss": 0.9265, + "step": 23967 + }, + { + "epoch": 5.23, + "learning_rate": 2.387153157089797e-06, + "loss": 1.271, + "step": 23968 + }, + { + "epoch": 5.23, + "learning_rate": 2.3870439152283155e-06, + "loss": 0.9663, + "step": 23969 + }, + { + "epoch": 5.23, + "learning_rate": 2.3869346733668346e-06, + "loss": 1.4404, + "step": 23970 + }, + { + "epoch": 5.23, + "learning_rate": 2.3868254315053528e-06, + "loss": 2.1598, + "step": 23971 + }, + { + "epoch": 5.23, + "learning_rate": 2.386716189643872e-06, + "loss": 1.3517, + "step": 23972 + }, + { + "epoch": 5.23, + "learning_rate": 2.3866069477823904e-06, + "loss": 1.2614, + "step": 23973 + }, + { + "epoch": 5.23, + "learning_rate": 2.386497705920909e-06, + "loss": 0.3341, + "step": 23974 + }, + { + "epoch": 5.23, + "learning_rate": 2.386388464059428e-06, + "loss": 0.9919, + "step": 23975 + }, + { + "epoch": 5.23, + "learning_rate": 2.3862792221979462e-06, + "loss": 2.6547, + "step": 23976 + }, + { + "epoch": 5.23, + "learning_rate": 2.3861699803364653e-06, + "loss": 0.1675, + "step": 23977 + }, + { + "epoch": 5.23, + "learning_rate": 2.386060738474984e-06, + "loss": 1.0793, + "step": 23978 + }, + { + "epoch": 5.23, + "learning_rate": 2.3859514966135025e-06, + "loss": 1.2019, + "step": 23979 + }, + { + "epoch": 5.23, + "learning_rate": 2.385842254752021e-06, + "loss": 2.9729, + "step": 23980 + }, + { + "epoch": 5.23, + "learning_rate": 2.3857330128905397e-06, + "loss": 1.6866, + "step": 23981 + }, + { + "epoch": 5.23, + "learning_rate": 2.3856237710290588e-06, + "loss": 2.6635, + "step": 23982 + }, + { + "epoch": 5.23, + "learning_rate": 2.385514529167577e-06, + "loss": 1.7811, + "step": 23983 + }, + { + "epoch": 5.23, + "learning_rate": 2.385405287306096e-06, + "loss": 1.3467, + "step": 23984 + }, + { + "epoch": 5.23, + "learning_rate": 2.3852960454446146e-06, + "loss": 3.1112, + "step": 23985 + }, + { + "epoch": 5.23, + "learning_rate": 2.385186803583133e-06, + "loss": 2.0798, + "step": 23986 + }, + { + "epoch": 5.24, + "learning_rate": 2.3850775617216522e-06, + "loss": 1.7035, + "step": 23987 + }, + { + "epoch": 5.24, + "learning_rate": 2.3849683198601704e-06, + "loss": 2.6425, + "step": 23988 + }, + { + "epoch": 5.24, + "learning_rate": 2.3848590779986895e-06, + "loss": 1.9423, + "step": 23989 + }, + { + "epoch": 5.24, + "learning_rate": 2.384749836137208e-06, + "loss": 1.4019, + "step": 23990 + }, + { + "epoch": 5.24, + "learning_rate": 2.3846405942757267e-06, + "loss": 1.9993, + "step": 23991 + }, + { + "epoch": 5.24, + "learning_rate": 2.3845313524142453e-06, + "loss": 0.5931, + "step": 23992 + }, + { + "epoch": 5.24, + "learning_rate": 2.384422110552764e-06, + "loss": 2.3149, + "step": 23993 + }, + { + "epoch": 5.24, + "learning_rate": 2.384312868691283e-06, + "loss": 1.6284, + "step": 23994 + }, + { + "epoch": 5.24, + "learning_rate": 2.3842036268298015e-06, + "loss": 0.9192, + "step": 23995 + }, + { + "epoch": 5.24, + "learning_rate": 2.38409438496832e-06, + "loss": 0.7607, + "step": 23996 + }, + { + "epoch": 5.24, + "learning_rate": 2.3839851431068388e-06, + "loss": 0.7734, + "step": 23997 + }, + { + "epoch": 5.24, + "learning_rate": 2.3838759012453574e-06, + "loss": 1.3377, + "step": 23998 + }, + { + "epoch": 5.24, + "learning_rate": 2.383766659383876e-06, + "loss": 0.8814, + "step": 23999 + }, + { + "epoch": 5.24, + "learning_rate": 2.3836574175223946e-06, + "loss": 1.8165, + "step": 24000 + }, + { + "epoch": 5.24, + "learning_rate": 2.3835481756609136e-06, + "loss": 1.62, + "step": 24001 + }, + { + "epoch": 5.24, + "learning_rate": 2.3834389337994322e-06, + "loss": 1.2542, + "step": 24002 + }, + { + "epoch": 5.24, + "learning_rate": 2.383329691937951e-06, + "loss": 1.3371, + "step": 24003 + }, + { + "epoch": 5.24, + "learning_rate": 2.3832204500764695e-06, + "loss": 1.5246, + "step": 24004 + }, + { + "epoch": 5.24, + "learning_rate": 2.383111208214988e-06, + "loss": 2.0159, + "step": 24005 + }, + { + "epoch": 5.24, + "learning_rate": 2.3830019663535067e-06, + "loss": 1.324, + "step": 24006 + }, + { + "epoch": 5.24, + "learning_rate": 2.3828927244920257e-06, + "loss": 1.69, + "step": 24007 + }, + { + "epoch": 5.24, + "learning_rate": 2.3827834826305443e-06, + "loss": 1.5874, + "step": 24008 + }, + { + "epoch": 5.24, + "learning_rate": 2.382674240769063e-06, + "loss": 0.8224, + "step": 24009 + }, + { + "epoch": 5.24, + "learning_rate": 2.3825649989075816e-06, + "loss": 1.0268, + "step": 24010 + }, + { + "epoch": 5.24, + "learning_rate": 2.3824557570461e-06, + "loss": 1.6775, + "step": 24011 + }, + { + "epoch": 5.24, + "learning_rate": 2.3823465151846188e-06, + "loss": 2.6303, + "step": 24012 + }, + { + "epoch": 5.24, + "learning_rate": 2.382237273323138e-06, + "loss": 1.5748, + "step": 24013 + }, + { + "epoch": 5.24, + "learning_rate": 2.3821280314616564e-06, + "loss": 2.1498, + "step": 24014 + }, + { + "epoch": 5.24, + "learning_rate": 2.382018789600175e-06, + "loss": 1.9831, + "step": 24015 + }, + { + "epoch": 5.24, + "learning_rate": 2.3819095477386936e-06, + "loss": 1.4393, + "step": 24016 + }, + { + "epoch": 5.24, + "learning_rate": 2.3818003058772123e-06, + "loss": 1.2359, + "step": 24017 + }, + { + "epoch": 5.24, + "learning_rate": 2.381691064015731e-06, + "loss": 1.4297, + "step": 24018 + }, + { + "epoch": 5.24, + "learning_rate": 2.38158182215425e-06, + "loss": 0.7951, + "step": 24019 + }, + { + "epoch": 5.24, + "learning_rate": 2.3814725802927685e-06, + "loss": 1.7851, + "step": 24020 + }, + { + "epoch": 5.24, + "learning_rate": 2.381363338431287e-06, + "loss": 2.3622, + "step": 24021 + }, + { + "epoch": 5.24, + "learning_rate": 2.3812540965698057e-06, + "loss": 1.4358, + "step": 24022 + }, + { + "epoch": 5.24, + "learning_rate": 2.3811448547083243e-06, + "loss": 1.5247, + "step": 24023 + }, + { + "epoch": 5.24, + "learning_rate": 2.381035612846843e-06, + "loss": 1.619, + "step": 24024 + }, + { + "epoch": 5.24, + "learning_rate": 2.3809263709853616e-06, + "loss": 1.1326, + "step": 24025 + }, + { + "epoch": 5.24, + "learning_rate": 2.3808171291238806e-06, + "loss": 3.1474, + "step": 24026 + }, + { + "epoch": 5.24, + "learning_rate": 2.3807078872623992e-06, + "loss": 0.3658, + "step": 24027 + }, + { + "epoch": 5.24, + "learning_rate": 2.380598645400918e-06, + "loss": 0.5249, + "step": 24028 + }, + { + "epoch": 5.24, + "learning_rate": 2.3804894035394364e-06, + "loss": 0.8427, + "step": 24029 + }, + { + "epoch": 5.24, + "learning_rate": 2.380380161677955e-06, + "loss": 1.1084, + "step": 24030 + }, + { + "epoch": 5.24, + "learning_rate": 2.380270919816474e-06, + "loss": 2.485, + "step": 24031 + }, + { + "epoch": 5.24, + "learning_rate": 2.3801616779549923e-06, + "loss": 0.3109, + "step": 24032 + }, + { + "epoch": 5.25, + "learning_rate": 2.3800524360935113e-06, + "loss": 0.9241, + "step": 24033 + }, + { + "epoch": 5.25, + "learning_rate": 2.37994319423203e-06, + "loss": 1.0274, + "step": 24034 + }, + { + "epoch": 5.25, + "learning_rate": 2.3798339523705485e-06, + "loss": 1.7505, + "step": 24035 + }, + { + "epoch": 5.25, + "learning_rate": 2.379724710509067e-06, + "loss": 0.9055, + "step": 24036 + }, + { + "epoch": 5.25, + "learning_rate": 2.3796154686475857e-06, + "loss": 1.2745, + "step": 24037 + }, + { + "epoch": 5.25, + "learning_rate": 2.3795062267861048e-06, + "loss": 2.0331, + "step": 24038 + }, + { + "epoch": 5.25, + "learning_rate": 2.379396984924623e-06, + "loss": 0.7764, + "step": 24039 + }, + { + "epoch": 5.25, + "learning_rate": 2.379287743063142e-06, + "loss": 1.8743, + "step": 24040 + }, + { + "epoch": 5.25, + "learning_rate": 2.3791785012016606e-06, + "loss": 1.0469, + "step": 24041 + }, + { + "epoch": 5.25, + "learning_rate": 2.3790692593401792e-06, + "loss": 2.2339, + "step": 24042 + }, + { + "epoch": 5.25, + "learning_rate": 2.3789600174786983e-06, + "loss": 1.1977, + "step": 24043 + }, + { + "epoch": 5.25, + "learning_rate": 2.3788507756172164e-06, + "loss": 1.0345, + "step": 24044 + }, + { + "epoch": 5.25, + "learning_rate": 2.3787415337557355e-06, + "loss": 2.9511, + "step": 24045 + }, + { + "epoch": 5.25, + "learning_rate": 2.378632291894254e-06, + "loss": 1.5393, + "step": 24046 + }, + { + "epoch": 5.25, + "learning_rate": 2.3785230500327727e-06, + "loss": 1.4168, + "step": 24047 + }, + { + "epoch": 5.25, + "learning_rate": 2.3784138081712917e-06, + "loss": 0.9903, + "step": 24048 + }, + { + "epoch": 5.25, + "learning_rate": 2.37830456630981e-06, + "loss": 1.2633, + "step": 24049 + }, + { + "epoch": 5.25, + "learning_rate": 2.378195324448329e-06, + "loss": 2.6887, + "step": 24050 + }, + { + "epoch": 5.25, + "learning_rate": 2.3780860825868476e-06, + "loss": 1.3919, + "step": 24051 + }, + { + "epoch": 5.25, + "learning_rate": 2.377976840725366e-06, + "loss": 0.7245, + "step": 24052 + }, + { + "epoch": 5.25, + "learning_rate": 2.377867598863885e-06, + "loss": 1.3407, + "step": 24053 + }, + { + "epoch": 5.25, + "learning_rate": 2.3777583570024034e-06, + "loss": 1.4494, + "step": 24054 + }, + { + "epoch": 5.25, + "learning_rate": 2.3776491151409224e-06, + "loss": 0.1876, + "step": 24055 + }, + { + "epoch": 5.25, + "learning_rate": 2.3775398732794406e-06, + "loss": 1.2434, + "step": 24056 + }, + { + "epoch": 5.25, + "learning_rate": 2.3774306314179597e-06, + "loss": 2.1292, + "step": 24057 + }, + { + "epoch": 5.25, + "learning_rate": 2.3773213895564783e-06, + "loss": 1.8121, + "step": 24058 + }, + { + "epoch": 5.25, + "learning_rate": 2.377212147694997e-06, + "loss": 1.6107, + "step": 24059 + }, + { + "epoch": 5.25, + "learning_rate": 2.377102905833516e-06, + "loss": 0.4671, + "step": 24060 + }, + { + "epoch": 5.25, + "learning_rate": 2.376993663972034e-06, + "loss": 1.2052, + "step": 24061 + }, + { + "epoch": 5.25, + "learning_rate": 2.376884422110553e-06, + "loss": 2.6598, + "step": 24062 + }, + { + "epoch": 5.25, + "learning_rate": 2.3767751802490717e-06, + "loss": 2.412, + "step": 24063 + }, + { + "epoch": 5.25, + "learning_rate": 2.3766659383875904e-06, + "loss": 1.9554, + "step": 24064 + }, + { + "epoch": 5.25, + "learning_rate": 2.376556696526109e-06, + "loss": 1.5527, + "step": 24065 + }, + { + "epoch": 5.25, + "learning_rate": 2.3764474546646276e-06, + "loss": 2.468, + "step": 24066 + }, + { + "epoch": 5.25, + "learning_rate": 2.3763382128031466e-06, + "loss": 1.1955, + "step": 24067 + }, + { + "epoch": 5.25, + "learning_rate": 2.376228970941665e-06, + "loss": 0.5904, + "step": 24068 + }, + { + "epoch": 5.25, + "learning_rate": 2.376119729080184e-06, + "loss": 0.826, + "step": 24069 + }, + { + "epoch": 5.25, + "learning_rate": 2.3760104872187024e-06, + "loss": 2.5074, + "step": 24070 + }, + { + "epoch": 5.25, + "learning_rate": 2.375901245357221e-06, + "loss": 1.7005, + "step": 24071 + }, + { + "epoch": 5.25, + "learning_rate": 2.3757920034957397e-06, + "loss": 1.7229, + "step": 24072 + }, + { + "epoch": 5.25, + "learning_rate": 2.3756827616342583e-06, + "loss": 0.8303, + "step": 24073 + }, + { + "epoch": 5.25, + "learning_rate": 2.3755735197727773e-06, + "loss": 1.6799, + "step": 24074 + }, + { + "epoch": 5.25, + "learning_rate": 2.375464277911296e-06, + "loss": 0.9615, + "step": 24075 + }, + { + "epoch": 5.25, + "learning_rate": 2.3753550360498145e-06, + "loss": 1.1109, + "step": 24076 + }, + { + "epoch": 5.25, + "learning_rate": 2.375245794188333e-06, + "loss": 2.2894, + "step": 24077 + }, + { + "epoch": 5.25, + "learning_rate": 2.3751365523268518e-06, + "loss": 1.3892, + "step": 24078 + }, + { + "epoch": 5.26, + "learning_rate": 2.375027310465371e-06, + "loss": 1.8142, + "step": 24079 + }, + { + "epoch": 5.26, + "learning_rate": 2.374918068603889e-06, + "loss": 0.8282, + "step": 24080 + }, + { + "epoch": 5.26, + "learning_rate": 2.374808826742408e-06, + "loss": 1.0655, + "step": 24081 + }, + { + "epoch": 5.26, + "learning_rate": 2.3746995848809266e-06, + "loss": 1.9062, + "step": 24082 + }, + { + "epoch": 5.26, + "learning_rate": 2.3745903430194452e-06, + "loss": 2.7905, + "step": 24083 + }, + { + "epoch": 5.26, + "learning_rate": 2.374481101157964e-06, + "loss": 1.3424, + "step": 24084 + }, + { + "epoch": 5.26, + "learning_rate": 2.3743718592964825e-06, + "loss": 1.466, + "step": 24085 + }, + { + "epoch": 5.26, + "learning_rate": 2.3742626174350015e-06, + "loss": 1.9463, + "step": 24086 + }, + { + "epoch": 5.26, + "learning_rate": 2.37415337557352e-06, + "loss": 2.0728, + "step": 24087 + }, + { + "epoch": 5.26, + "learning_rate": 2.3740441337120387e-06, + "loss": 1.8221, + "step": 24088 + }, + { + "epoch": 5.26, + "learning_rate": 2.3739348918505573e-06, + "loss": 3.2687, + "step": 24089 + }, + { + "epoch": 5.26, + "learning_rate": 2.373825649989076e-06, + "loss": 0.7956, + "step": 24090 + }, + { + "epoch": 5.26, + "learning_rate": 2.3737164081275946e-06, + "loss": 1.6203, + "step": 24091 + }, + { + "epoch": 5.26, + "learning_rate": 2.373607166266113e-06, + "loss": 1.1514, + "step": 24092 + }, + { + "epoch": 5.26, + "learning_rate": 2.373497924404632e-06, + "loss": 1.2235, + "step": 24093 + }, + { + "epoch": 5.26, + "learning_rate": 2.373388682543151e-06, + "loss": 1.9672, + "step": 24094 + }, + { + "epoch": 5.26, + "learning_rate": 2.3732794406816694e-06, + "loss": 2.0551, + "step": 24095 + }, + { + "epoch": 5.26, + "learning_rate": 2.373170198820188e-06, + "loss": 1.72, + "step": 24096 + }, + { + "epoch": 5.26, + "learning_rate": 2.3730609569587066e-06, + "loss": 1.0761, + "step": 24097 + }, + { + "epoch": 5.26, + "learning_rate": 2.3729517150972253e-06, + "loss": 2.6643, + "step": 24098 + }, + { + "epoch": 5.26, + "learning_rate": 2.3728424732357443e-06, + "loss": 0.2373, + "step": 24099 + }, + { + "epoch": 5.26, + "learning_rate": 2.372733231374263e-06, + "loss": 0.7343, + "step": 24100 + }, + { + "epoch": 5.26, + "learning_rate": 2.3726239895127815e-06, + "loss": 1.1841, + "step": 24101 + }, + { + "epoch": 5.26, + "learning_rate": 2.3725147476513e-06, + "loss": 0.7975, + "step": 24102 + }, + { + "epoch": 5.26, + "learning_rate": 2.3724055057898187e-06, + "loss": 1.0818, + "step": 24103 + }, + { + "epoch": 5.26, + "learning_rate": 2.3722962639283378e-06, + "loss": 0.8218, + "step": 24104 + }, + { + "epoch": 5.26, + "learning_rate": 2.372187022066856e-06, + "loss": 1.2452, + "step": 24105 + }, + { + "epoch": 5.26, + "learning_rate": 2.372077780205375e-06, + "loss": 3.07, + "step": 24106 + }, + { + "epoch": 5.26, + "learning_rate": 2.3719685383438936e-06, + "loss": 1.9294, + "step": 24107 + }, + { + "epoch": 5.26, + "learning_rate": 2.371859296482412e-06, + "loss": 0.7028, + "step": 24108 + }, + { + "epoch": 5.26, + "learning_rate": 2.371750054620931e-06, + "loss": 1.0434, + "step": 24109 + }, + { + "epoch": 5.26, + "learning_rate": 2.3716408127594494e-06, + "loss": 1.3518, + "step": 24110 + }, + { + "epoch": 5.26, + "learning_rate": 2.3715315708979685e-06, + "loss": 2.273, + "step": 24111 + }, + { + "epoch": 5.26, + "learning_rate": 2.371422329036487e-06, + "loss": 2.474, + "step": 24112 + }, + { + "epoch": 5.26, + "learning_rate": 2.3713130871750057e-06, + "loss": 1.6953, + "step": 24113 + }, + { + "epoch": 5.26, + "learning_rate": 2.3712038453135243e-06, + "loss": 1.7796, + "step": 24114 + }, + { + "epoch": 5.26, + "learning_rate": 2.371094603452043e-06, + "loss": 0.8754, + "step": 24115 + }, + { + "epoch": 5.26, + "learning_rate": 2.370985361590562e-06, + "loss": 2.0274, + "step": 24116 + }, + { + "epoch": 5.26, + "learning_rate": 2.37087611972908e-06, + "loss": 1.8275, + "step": 24117 + }, + { + "epoch": 5.26, + "learning_rate": 2.370766877867599e-06, + "loss": 2.4978, + "step": 24118 + }, + { + "epoch": 5.26, + "learning_rate": 2.3706576360061178e-06, + "loss": 1.1307, + "step": 24119 + }, + { + "epoch": 5.26, + "learning_rate": 2.3705483941446364e-06, + "loss": 1.4586, + "step": 24120 + }, + { + "epoch": 5.26, + "learning_rate": 2.370439152283155e-06, + "loss": 1.3192, + "step": 24121 + }, + { + "epoch": 5.26, + "learning_rate": 2.3703299104216736e-06, + "loss": 1.8508, + "step": 24122 + }, + { + "epoch": 5.26, + "learning_rate": 2.3702206685601926e-06, + "loss": 1.8141, + "step": 24123 + }, + { + "epoch": 5.26, + "learning_rate": 2.370111426698711e-06, + "loss": 0.5718, + "step": 24124 + }, + { + "epoch": 5.27, + "learning_rate": 2.37000218483723e-06, + "loss": 2.9993, + "step": 24125 + }, + { + "epoch": 5.27, + "learning_rate": 2.3698929429757485e-06, + "loss": 1.9819, + "step": 24126 + }, + { + "epoch": 5.27, + "learning_rate": 2.369783701114267e-06, + "loss": 1.6582, + "step": 24127 + }, + { + "epoch": 5.27, + "learning_rate": 2.369674459252786e-06, + "loss": 0.5076, + "step": 24128 + }, + { + "epoch": 5.27, + "learning_rate": 2.3695652173913043e-06, + "loss": 1.5276, + "step": 24129 + }, + { + "epoch": 5.27, + "learning_rate": 2.3694559755298233e-06, + "loss": 1.3961, + "step": 24130 + }, + { + "epoch": 5.27, + "learning_rate": 2.369346733668342e-06, + "loss": 2.993, + "step": 24131 + }, + { + "epoch": 5.27, + "learning_rate": 2.3692374918068606e-06, + "loss": 0.5845, + "step": 24132 + }, + { + "epoch": 5.27, + "learning_rate": 2.369128249945379e-06, + "loss": 3.0525, + "step": 24133 + }, + { + "epoch": 5.27, + "learning_rate": 2.3690190080838978e-06, + "loss": 1.7662, + "step": 24134 + }, + { + "epoch": 5.27, + "learning_rate": 2.368909766222417e-06, + "loss": 0.4833, + "step": 24135 + }, + { + "epoch": 5.27, + "learning_rate": 2.368800524360935e-06, + "loss": 1.9244, + "step": 24136 + }, + { + "epoch": 5.27, + "learning_rate": 2.368691282499454e-06, + "loss": 1.5702, + "step": 24137 + }, + { + "epoch": 5.27, + "learning_rate": 2.3685820406379727e-06, + "loss": 0.507, + "step": 24138 + }, + { + "epoch": 5.27, + "learning_rate": 2.3684727987764913e-06, + "loss": 0.6681, + "step": 24139 + }, + { + "epoch": 5.27, + "learning_rate": 2.3683635569150103e-06, + "loss": 2.1131, + "step": 24140 + }, + { + "epoch": 5.27, + "learning_rate": 2.3682543150535285e-06, + "loss": 1.0934, + "step": 24141 + }, + { + "epoch": 5.27, + "learning_rate": 2.3681450731920475e-06, + "loss": 0.1184, + "step": 24142 + }, + { + "epoch": 5.27, + "learning_rate": 2.368035831330566e-06, + "loss": 1.5014, + "step": 24143 + }, + { + "epoch": 5.27, + "learning_rate": 2.3679265894690847e-06, + "loss": 1.8908, + "step": 24144 + }, + { + "epoch": 5.27, + "learning_rate": 2.3678173476076034e-06, + "loss": 2.3995, + "step": 24145 + }, + { + "epoch": 5.27, + "learning_rate": 2.367708105746122e-06, + "loss": 1.2521, + "step": 24146 + }, + { + "epoch": 5.27, + "learning_rate": 2.367598863884641e-06, + "loss": 0.6877, + "step": 24147 + }, + { + "epoch": 5.27, + "learning_rate": 2.3674896220231596e-06, + "loss": 0.6843, + "step": 24148 + }, + { + "epoch": 5.27, + "learning_rate": 2.3673803801616782e-06, + "loss": 1.5841, + "step": 24149 + }, + { + "epoch": 5.27, + "learning_rate": 2.367271138300197e-06, + "loss": 1.5069, + "step": 24150 + }, + { + "epoch": 5.27, + "learning_rate": 2.3671618964387154e-06, + "loss": 1.5066, + "step": 24151 + }, + { + "epoch": 5.27, + "learning_rate": 2.3670526545772345e-06, + "loss": 1.5381, + "step": 24152 + }, + { + "epoch": 5.27, + "learning_rate": 2.3669434127157527e-06, + "loss": 2.0559, + "step": 24153 + }, + { + "epoch": 5.27, + "learning_rate": 2.3668341708542717e-06, + "loss": 1.1258, + "step": 24154 + }, + { + "epoch": 5.27, + "learning_rate": 2.3667249289927903e-06, + "loss": 1.1179, + "step": 24155 + }, + { + "epoch": 5.27, + "learning_rate": 2.366615687131309e-06, + "loss": 0.8451, + "step": 24156 + }, + { + "epoch": 5.27, + "learning_rate": 2.3665064452698275e-06, + "loss": 1.1019, + "step": 24157 + }, + { + "epoch": 5.27, + "learning_rate": 2.366397203408346e-06, + "loss": 0.9537, + "step": 24158 + }, + { + "epoch": 5.27, + "learning_rate": 2.366287961546865e-06, + "loss": 1.8567, + "step": 24159 + }, + { + "epoch": 5.27, + "learning_rate": 2.366178719685384e-06, + "loss": 1.0766, + "step": 24160 + }, + { + "epoch": 5.27, + "learning_rate": 2.3660694778239024e-06, + "loss": 0.8532, + "step": 24161 + }, + { + "epoch": 5.27, + "learning_rate": 2.365960235962421e-06, + "loss": 0.9228, + "step": 24162 + }, + { + "epoch": 5.27, + "learning_rate": 2.3658509941009396e-06, + "loss": 2.2512, + "step": 24163 + }, + { + "epoch": 5.27, + "learning_rate": 2.3657417522394582e-06, + "loss": 3.4102, + "step": 24164 + }, + { + "epoch": 5.27, + "learning_rate": 2.365632510377977e-06, + "loss": 1.8132, + "step": 24165 + }, + { + "epoch": 5.27, + "learning_rate": 2.365523268516496e-06, + "loss": 1.1169, + "step": 24166 + }, + { + "epoch": 5.27, + "learning_rate": 2.3654140266550145e-06, + "loss": 1.2615, + "step": 24167 + }, + { + "epoch": 5.27, + "learning_rate": 2.365304784793533e-06, + "loss": 2.1591, + "step": 24168 + }, + { + "epoch": 5.27, + "learning_rate": 2.3651955429320517e-06, + "loss": 0.4407, + "step": 24169 + }, + { + "epoch": 5.27, + "learning_rate": 2.3650863010705703e-06, + "loss": 1.6001, + "step": 24170 + }, + { + "epoch": 5.28, + "learning_rate": 2.364977059209089e-06, + "loss": 1.736, + "step": 24171 + }, + { + "epoch": 5.28, + "learning_rate": 2.364867817347608e-06, + "loss": 1.3514, + "step": 24172 + }, + { + "epoch": 5.28, + "learning_rate": 2.3647585754861266e-06, + "loss": 1.1658, + "step": 24173 + }, + { + "epoch": 5.28, + "learning_rate": 2.364649333624645e-06, + "loss": 1.4221, + "step": 24174 + }, + { + "epoch": 5.28, + "learning_rate": 2.364540091763164e-06, + "loss": 2.3493, + "step": 24175 + }, + { + "epoch": 5.28, + "learning_rate": 2.3644308499016824e-06, + "loss": 2.65, + "step": 24176 + }, + { + "epoch": 5.28, + "learning_rate": 2.364321608040201e-06, + "loss": 1.8046, + "step": 24177 + }, + { + "epoch": 5.28, + "learning_rate": 2.3642123661787196e-06, + "loss": 0.7412, + "step": 24178 + }, + { + "epoch": 5.28, + "learning_rate": 2.3641031243172387e-06, + "loss": 1.6106, + "step": 24179 + }, + { + "epoch": 5.28, + "learning_rate": 2.3639938824557573e-06, + "loss": 0.9809, + "step": 24180 + }, + { + "epoch": 5.28, + "learning_rate": 2.363884640594276e-06, + "loss": 2.8371, + "step": 24181 + }, + { + "epoch": 5.28, + "learning_rate": 2.3637753987327945e-06, + "loss": 2.1439, + "step": 24182 + }, + { + "epoch": 5.28, + "learning_rate": 2.363666156871313e-06, + "loss": 1.8275, + "step": 24183 + }, + { + "epoch": 5.28, + "learning_rate": 2.363556915009832e-06, + "loss": 3.2555, + "step": 24184 + }, + { + "epoch": 5.28, + "learning_rate": 2.3634476731483508e-06, + "loss": 2.2767, + "step": 24185 + }, + { + "epoch": 5.28, + "learning_rate": 2.3633384312868694e-06, + "loss": 1.5721, + "step": 24186 + }, + { + "epoch": 5.28, + "learning_rate": 2.363229189425388e-06, + "loss": 1.2033, + "step": 24187 + }, + { + "epoch": 5.28, + "learning_rate": 2.3631199475639066e-06, + "loss": 1.2561, + "step": 24188 + }, + { + "epoch": 5.28, + "learning_rate": 2.363010705702425e-06, + "loss": 0.7039, + "step": 24189 + }, + { + "epoch": 5.28, + "learning_rate": 2.362901463840944e-06, + "loss": 1.2749, + "step": 24190 + }, + { + "epoch": 5.28, + "learning_rate": 2.362792221979463e-06, + "loss": 1.37, + "step": 24191 + }, + { + "epoch": 5.28, + "learning_rate": 2.3626829801179815e-06, + "loss": 1.712, + "step": 24192 + }, + { + "epoch": 5.28, + "learning_rate": 2.3625737382565e-06, + "loss": 0.9581, + "step": 24193 + }, + { + "epoch": 5.28, + "learning_rate": 2.3624644963950187e-06, + "loss": 0.2806, + "step": 24194 + }, + { + "epoch": 5.28, + "learning_rate": 2.3623552545335373e-06, + "loss": 2.0587, + "step": 24195 + }, + { + "epoch": 5.28, + "learning_rate": 2.3622460126720563e-06, + "loss": 1.8324, + "step": 24196 + }, + { + "epoch": 5.28, + "learning_rate": 2.3621367708105745e-06, + "loss": 0.9696, + "step": 24197 + }, + { + "epoch": 5.28, + "learning_rate": 2.3620275289490936e-06, + "loss": 1.5883, + "step": 24198 + }, + { + "epoch": 5.28, + "learning_rate": 2.361918287087612e-06, + "loss": 1.3153, + "step": 24199 + }, + { + "epoch": 5.28, + "learning_rate": 2.3618090452261308e-06, + "loss": 1.0532, + "step": 24200 + }, + { + "epoch": 5.28, + "learning_rate": 2.36169980336465e-06, + "loss": 1.2073, + "step": 24201 + }, + { + "epoch": 5.28, + "learning_rate": 2.361590561503168e-06, + "loss": 0.6903, + "step": 24202 + }, + { + "epoch": 5.28, + "learning_rate": 2.361481319641687e-06, + "loss": 1.3838, + "step": 24203 + }, + { + "epoch": 5.28, + "learning_rate": 2.3613720777802056e-06, + "loss": 0.6486, + "step": 24204 + }, + { + "epoch": 5.28, + "learning_rate": 2.3612628359187243e-06, + "loss": 1.0248, + "step": 24205 + }, + { + "epoch": 5.28, + "learning_rate": 2.361153594057243e-06, + "loss": 0.973, + "step": 24206 + }, + { + "epoch": 5.28, + "learning_rate": 2.3610443521957615e-06, + "loss": 1.1763, + "step": 24207 + }, + { + "epoch": 5.28, + "learning_rate": 2.3609351103342805e-06, + "loss": 1.4859, + "step": 24208 + }, + { + "epoch": 5.28, + "learning_rate": 2.3608258684727987e-06, + "loss": 0.3717, + "step": 24209 + }, + { + "epoch": 5.28, + "learning_rate": 2.3607166266113177e-06, + "loss": 0.4047, + "step": 24210 + }, + { + "epoch": 5.28, + "learning_rate": 2.3606073847498363e-06, + "loss": 2.5792, + "step": 24211 + }, + { + "epoch": 5.28, + "learning_rate": 2.360498142888355e-06, + "loss": 0.4886, + "step": 24212 + }, + { + "epoch": 5.28, + "learning_rate": 2.360388901026874e-06, + "loss": 0.8762, + "step": 24213 + }, + { + "epoch": 5.28, + "learning_rate": 2.360279659165392e-06, + "loss": 3.2786, + "step": 24214 + }, + { + "epoch": 5.28, + "learning_rate": 2.360170417303911e-06, + "loss": 1.5746, + "step": 24215 + }, + { + "epoch": 5.29, + "learning_rate": 2.36006117544243e-06, + "loss": 2.0178, + "step": 24216 + }, + { + "epoch": 5.29, + "learning_rate": 2.3599519335809484e-06, + "loss": 0.9878, + "step": 24217 + }, + { + "epoch": 5.29, + "learning_rate": 2.359842691719467e-06, + "loss": 1.8188, + "step": 24218 + }, + { + "epoch": 5.29, + "learning_rate": 2.3597334498579857e-06, + "loss": 0.6526, + "step": 24219 + }, + { + "epoch": 5.29, + "learning_rate": 2.3596242079965047e-06, + "loss": 1.2167, + "step": 24220 + }, + { + "epoch": 5.29, + "learning_rate": 2.359514966135023e-06, + "loss": 2.4786, + "step": 24221 + }, + { + "epoch": 5.29, + "learning_rate": 2.359405724273542e-06, + "loss": 0.7072, + "step": 24222 + }, + { + "epoch": 5.29, + "learning_rate": 2.3592964824120605e-06, + "loss": 1.4994, + "step": 24223 + }, + { + "epoch": 5.29, + "learning_rate": 2.359187240550579e-06, + "loss": 1.5717, + "step": 24224 + }, + { + "epoch": 5.29, + "learning_rate": 2.359077998689098e-06, + "loss": 2.2603, + "step": 24225 + }, + { + "epoch": 5.29, + "learning_rate": 2.3589687568276164e-06, + "loss": 2.6018, + "step": 24226 + }, + { + "epoch": 5.29, + "learning_rate": 2.3588595149661354e-06, + "loss": 1.1005, + "step": 24227 + }, + { + "epoch": 5.29, + "learning_rate": 2.358750273104654e-06, + "loss": 0.6641, + "step": 24228 + }, + { + "epoch": 5.29, + "learning_rate": 2.3586410312431726e-06, + "loss": 1.1635, + "step": 24229 + }, + { + "epoch": 5.29, + "learning_rate": 2.3585317893816912e-06, + "loss": 2.3099, + "step": 24230 + }, + { + "epoch": 5.29, + "learning_rate": 2.35842254752021e-06, + "loss": 3.4819, + "step": 24231 + }, + { + "epoch": 5.29, + "learning_rate": 2.358313305658729e-06, + "loss": 0.744, + "step": 24232 + }, + { + "epoch": 5.29, + "learning_rate": 2.358204063797247e-06, + "loss": 1.1821, + "step": 24233 + }, + { + "epoch": 5.29, + "learning_rate": 2.358094821935766e-06, + "loss": 2.0682, + "step": 24234 + }, + { + "epoch": 5.29, + "learning_rate": 2.3579855800742847e-06, + "loss": 2.2324, + "step": 24235 + }, + { + "epoch": 5.29, + "learning_rate": 2.3578763382128033e-06, + "loss": 1.7198, + "step": 24236 + }, + { + "epoch": 5.29, + "learning_rate": 2.357767096351322e-06, + "loss": 1.4856, + "step": 24237 + }, + { + "epoch": 5.29, + "learning_rate": 2.3576578544898405e-06, + "loss": 1.2382, + "step": 24238 + }, + { + "epoch": 5.29, + "learning_rate": 2.3575486126283596e-06, + "loss": 0.6245, + "step": 24239 + }, + { + "epoch": 5.29, + "learning_rate": 2.357439370766878e-06, + "loss": 1.6806, + "step": 24240 + }, + { + "epoch": 5.29, + "learning_rate": 2.3573301289053968e-06, + "loss": 2.9686, + "step": 24241 + }, + { + "epoch": 5.29, + "learning_rate": 2.3572208870439154e-06, + "loss": 2.9682, + "step": 24242 + }, + { + "epoch": 5.29, + "learning_rate": 2.357111645182434e-06, + "loss": 1.1451, + "step": 24243 + }, + { + "epoch": 5.29, + "learning_rate": 2.3570024033209526e-06, + "loss": 2.3568, + "step": 24244 + }, + { + "epoch": 5.29, + "learning_rate": 2.3568931614594717e-06, + "loss": 1.1929, + "step": 24245 + }, + { + "epoch": 5.29, + "learning_rate": 2.3567839195979903e-06, + "loss": 1.6614, + "step": 24246 + }, + { + "epoch": 5.29, + "learning_rate": 2.356674677736509e-06, + "loss": 1.5265, + "step": 24247 + }, + { + "epoch": 5.29, + "learning_rate": 2.3565654358750275e-06, + "loss": 1.7244, + "step": 24248 + }, + { + "epoch": 5.29, + "learning_rate": 2.356456194013546e-06, + "loss": 1.4361, + "step": 24249 + }, + { + "epoch": 5.29, + "learning_rate": 2.3563469521520647e-06, + "loss": 2.554, + "step": 24250 + }, + { + "epoch": 5.29, + "learning_rate": 2.3562377102905837e-06, + "loss": 3.401, + "step": 24251 + }, + { + "epoch": 5.29, + "learning_rate": 2.3561284684291024e-06, + "loss": 1.525, + "step": 24252 + }, + { + "epoch": 5.29, + "learning_rate": 2.356019226567621e-06, + "loss": 3.3212, + "step": 24253 + }, + { + "epoch": 5.29, + "learning_rate": 2.3559099847061396e-06, + "loss": 0.7883, + "step": 24254 + }, + { + "epoch": 5.29, + "learning_rate": 2.355800742844658e-06, + "loss": 1.4398, + "step": 24255 + }, + { + "epoch": 5.29, + "learning_rate": 2.355691500983177e-06, + "loss": 1.5034, + "step": 24256 + }, + { + "epoch": 5.29, + "learning_rate": 2.355582259121696e-06, + "loss": 1.2315, + "step": 24257 + }, + { + "epoch": 5.29, + "learning_rate": 2.3554730172602144e-06, + "loss": 1.3629, + "step": 24258 + }, + { + "epoch": 5.29, + "learning_rate": 2.355363775398733e-06, + "loss": 0.1283, + "step": 24259 + }, + { + "epoch": 5.29, + "learning_rate": 2.3552545335372517e-06, + "loss": 1.0624, + "step": 24260 + }, + { + "epoch": 5.29, + "learning_rate": 2.3551452916757703e-06, + "loss": 1.0599, + "step": 24261 + }, + { + "epoch": 5.3, + "learning_rate": 2.355036049814289e-06, + "loss": 1.5084, + "step": 24262 + }, + { + "epoch": 5.3, + "learning_rate": 2.3549268079528075e-06, + "loss": 0.9119, + "step": 24263 + }, + { + "epoch": 5.3, + "learning_rate": 2.3548175660913265e-06, + "loss": 2.5493, + "step": 24264 + }, + { + "epoch": 5.3, + "learning_rate": 2.354708324229845e-06, + "loss": 0.8568, + "step": 24265 + }, + { + "epoch": 5.3, + "learning_rate": 2.3545990823683638e-06, + "loss": 1.0725, + "step": 24266 + }, + { + "epoch": 5.3, + "learning_rate": 2.3544898405068824e-06, + "loss": 1.2139, + "step": 24267 + }, + { + "epoch": 5.3, + "learning_rate": 2.354380598645401e-06, + "loss": 4.1848, + "step": 24268 + }, + { + "epoch": 5.3, + "learning_rate": 2.35427135678392e-06, + "loss": 1.7039, + "step": 24269 + }, + { + "epoch": 5.3, + "learning_rate": 2.354162114922438e-06, + "loss": 1.3072, + "step": 24270 + }, + { + "epoch": 5.3, + "learning_rate": 2.3540528730609572e-06, + "loss": 0.5407, + "step": 24271 + }, + { + "epoch": 5.3, + "learning_rate": 2.353943631199476e-06, + "loss": 0.3733, + "step": 24272 + }, + { + "epoch": 5.3, + "learning_rate": 2.3538343893379945e-06, + "loss": 2.7564, + "step": 24273 + }, + { + "epoch": 5.3, + "learning_rate": 2.353725147476513e-06, + "loss": 1.9465, + "step": 24274 + }, + { + "epoch": 5.3, + "learning_rate": 2.3536159056150317e-06, + "loss": 1.1087, + "step": 24275 + }, + { + "epoch": 5.3, + "learning_rate": 2.3535066637535507e-06, + "loss": 1.5529, + "step": 24276 + }, + { + "epoch": 5.3, + "learning_rate": 2.353397421892069e-06, + "loss": 0.4141, + "step": 24277 + }, + { + "epoch": 5.3, + "learning_rate": 2.353288180030588e-06, + "loss": 2.3583, + "step": 24278 + }, + { + "epoch": 5.3, + "learning_rate": 2.3531789381691065e-06, + "loss": 3.4043, + "step": 24279 + }, + { + "epoch": 5.3, + "learning_rate": 2.353069696307625e-06, + "loss": 1.533, + "step": 24280 + }, + { + "epoch": 5.3, + "learning_rate": 2.352960454446144e-06, + "loss": 2.8006, + "step": 24281 + }, + { + "epoch": 5.3, + "learning_rate": 2.3528512125846624e-06, + "loss": 1.5326, + "step": 24282 + }, + { + "epoch": 5.3, + "learning_rate": 2.3527419707231814e-06, + "loss": 1.4547, + "step": 24283 + }, + { + "epoch": 5.3, + "learning_rate": 2.3526327288617e-06, + "loss": 1.409, + "step": 24284 + }, + { + "epoch": 5.3, + "learning_rate": 2.3525234870002186e-06, + "loss": 0.9523, + "step": 24285 + }, + { + "epoch": 5.3, + "learning_rate": 2.3524142451387372e-06, + "loss": 1.1147, + "step": 24286 + }, + { + "epoch": 5.3, + "learning_rate": 2.352305003277256e-06, + "loss": 0.8658, + "step": 24287 + }, + { + "epoch": 5.3, + "learning_rate": 2.352195761415775e-06, + "loss": 0.5922, + "step": 24288 + }, + { + "epoch": 5.3, + "learning_rate": 2.352086519554293e-06, + "loss": 1.3921, + "step": 24289 + }, + { + "epoch": 5.3, + "learning_rate": 2.351977277692812e-06, + "loss": 0.8817, + "step": 24290 + }, + { + "epoch": 5.3, + "learning_rate": 2.3518680358313307e-06, + "loss": 2.1863, + "step": 24291 + }, + { + "epoch": 5.3, + "learning_rate": 2.3517587939698493e-06, + "loss": 2.4135, + "step": 24292 + }, + { + "epoch": 5.3, + "learning_rate": 2.3516495521083684e-06, + "loss": 0.9887, + "step": 24293 + }, + { + "epoch": 5.3, + "learning_rate": 2.3515403102468866e-06, + "loss": 1.9618, + "step": 24294 + }, + { + "epoch": 5.3, + "learning_rate": 2.3514310683854056e-06, + "loss": 0.354, + "step": 24295 + }, + { + "epoch": 5.3, + "learning_rate": 2.351321826523924e-06, + "loss": 2.9052, + "step": 24296 + }, + { + "epoch": 5.3, + "learning_rate": 2.351212584662443e-06, + "loss": 1.5896, + "step": 24297 + }, + { + "epoch": 5.3, + "learning_rate": 2.351103342800962e-06, + "loss": 1.4493, + "step": 24298 + }, + { + "epoch": 5.3, + "learning_rate": 2.35099410093948e-06, + "loss": 0.7047, + "step": 24299 + }, + { + "epoch": 5.3, + "learning_rate": 2.350884859077999e-06, + "loss": 2.603, + "step": 24300 + }, + { + "epoch": 5.3, + "learning_rate": 2.3507756172165177e-06, + "loss": 2.6116, + "step": 24301 + }, + { + "epoch": 5.3, + "learning_rate": 2.3506663753550363e-06, + "loss": 2.2283, + "step": 24302 + }, + { + "epoch": 5.3, + "learning_rate": 2.350557133493555e-06, + "loss": 0.8169, + "step": 24303 + }, + { + "epoch": 5.3, + "learning_rate": 2.3504478916320735e-06, + "loss": 1.236, + "step": 24304 + }, + { + "epoch": 5.3, + "learning_rate": 2.3503386497705925e-06, + "loss": 1.711, + "step": 24305 + }, + { + "epoch": 5.3, + "learning_rate": 2.3502294079091107e-06, + "loss": 0.91, + "step": 24306 + }, + { + "epoch": 5.3, + "learning_rate": 2.3501201660476298e-06, + "loss": 2.0702, + "step": 24307 + }, + { + "epoch": 5.31, + "learning_rate": 2.3500109241861484e-06, + "loss": 1.4564, + "step": 24308 + }, + { + "epoch": 5.31, + "learning_rate": 2.349901682324667e-06, + "loss": 2.3618, + "step": 24309 + }, + { + "epoch": 5.31, + "learning_rate": 2.3497924404631856e-06, + "loss": 0.7975, + "step": 24310 + }, + { + "epoch": 5.31, + "learning_rate": 2.3496831986017042e-06, + "loss": 0.4538, + "step": 24311 + }, + { + "epoch": 5.31, + "learning_rate": 2.3495739567402232e-06, + "loss": 2.2361, + "step": 24312 + }, + { + "epoch": 5.31, + "learning_rate": 2.349464714878742e-06, + "loss": 0.5474, + "step": 24313 + }, + { + "epoch": 5.31, + "learning_rate": 2.3493554730172605e-06, + "loss": 1.0266, + "step": 24314 + }, + { + "epoch": 5.31, + "learning_rate": 2.349246231155779e-06, + "loss": 1.1075, + "step": 24315 + }, + { + "epoch": 5.31, + "learning_rate": 2.3491369892942977e-06, + "loss": 1.6397, + "step": 24316 + }, + { + "epoch": 5.31, + "learning_rate": 2.3490277474328167e-06, + "loss": 1.7964, + "step": 24317 + }, + { + "epoch": 5.31, + "learning_rate": 2.348918505571335e-06, + "loss": 0.8725, + "step": 24318 + }, + { + "epoch": 5.31, + "learning_rate": 2.348809263709854e-06, + "loss": 1.8008, + "step": 24319 + }, + { + "epoch": 5.31, + "learning_rate": 2.3487000218483726e-06, + "loss": 0.2791, + "step": 24320 + }, + { + "epoch": 5.31, + "learning_rate": 2.348590779986891e-06, + "loss": 1.8328, + "step": 24321 + }, + { + "epoch": 5.31, + "learning_rate": 2.3484815381254098e-06, + "loss": 1.2872, + "step": 24322 + }, + { + "epoch": 5.31, + "learning_rate": 2.3483722962639284e-06, + "loss": 0.2866, + "step": 24323 + }, + { + "epoch": 5.31, + "learning_rate": 2.3482630544024474e-06, + "loss": 0.9546, + "step": 24324 + }, + { + "epoch": 5.31, + "learning_rate": 2.348153812540966e-06, + "loss": 1.5512, + "step": 24325 + }, + { + "epoch": 5.31, + "learning_rate": 2.3480445706794847e-06, + "loss": 1.5141, + "step": 24326 + }, + { + "epoch": 5.31, + "learning_rate": 2.3479353288180033e-06, + "loss": 0.9191, + "step": 24327 + }, + { + "epoch": 5.31, + "learning_rate": 2.347826086956522e-06, + "loss": 1.3256, + "step": 24328 + }, + { + "epoch": 5.31, + "learning_rate": 2.3477168450950405e-06, + "loss": 2.4976, + "step": 24329 + }, + { + "epoch": 5.31, + "learning_rate": 2.347607603233559e-06, + "loss": 1.425, + "step": 24330 + }, + { + "epoch": 5.31, + "learning_rate": 2.347498361372078e-06, + "loss": 2.1757, + "step": 24331 + }, + { + "epoch": 5.31, + "learning_rate": 2.3473891195105967e-06, + "loss": 0.7816, + "step": 24332 + }, + { + "epoch": 5.31, + "learning_rate": 2.3472798776491154e-06, + "loss": 2.5198, + "step": 24333 + }, + { + "epoch": 5.31, + "learning_rate": 2.347170635787634e-06, + "loss": 1.1772, + "step": 24334 + }, + { + "epoch": 5.31, + "learning_rate": 2.3470613939261526e-06, + "loss": 2.299, + "step": 24335 + }, + { + "epoch": 5.31, + "learning_rate": 2.346952152064671e-06, + "loss": 0.6743, + "step": 24336 + }, + { + "epoch": 5.31, + "learning_rate": 2.3468429102031902e-06, + "loss": 2.2193, + "step": 24337 + }, + { + "epoch": 5.31, + "learning_rate": 2.346733668341709e-06, + "loss": 1.8327, + "step": 24338 + }, + { + "epoch": 5.31, + "learning_rate": 2.3466244264802274e-06, + "loss": 0.9566, + "step": 24339 + }, + { + "epoch": 5.31, + "learning_rate": 2.346515184618746e-06, + "loss": 0.2418, + "step": 24340 + }, + { + "epoch": 5.31, + "learning_rate": 2.3464059427572647e-06, + "loss": 2.481, + "step": 24341 + }, + { + "epoch": 5.31, + "learning_rate": 2.3462967008957837e-06, + "loss": 1.4955, + "step": 24342 + }, + { + "epoch": 5.31, + "learning_rate": 2.346187459034302e-06, + "loss": 1.3758, + "step": 24343 + }, + { + "epoch": 5.31, + "learning_rate": 2.346078217172821e-06, + "loss": 1.8151, + "step": 24344 + }, + { + "epoch": 5.31, + "learning_rate": 2.3459689753113395e-06, + "loss": 0.5546, + "step": 24345 + }, + { + "epoch": 5.31, + "learning_rate": 2.345859733449858e-06, + "loss": 3.2857, + "step": 24346 + }, + { + "epoch": 5.31, + "learning_rate": 2.3457504915883768e-06, + "loss": 0.7229, + "step": 24347 + }, + { + "epoch": 5.31, + "learning_rate": 2.3456412497268954e-06, + "loss": 0.5309, + "step": 24348 + }, + { + "epoch": 5.31, + "learning_rate": 2.3455320078654144e-06, + "loss": 1.1927, + "step": 24349 + }, + { + "epoch": 5.31, + "learning_rate": 2.3454227660039326e-06, + "loss": 0.2089, + "step": 24350 + }, + { + "epoch": 5.31, + "learning_rate": 2.3453135241424516e-06, + "loss": 1.5336, + "step": 24351 + }, + { + "epoch": 5.31, + "learning_rate": 2.3452042822809702e-06, + "loss": 1.9444, + "step": 24352 + }, + { + "epoch": 5.31, + "learning_rate": 2.345095040419489e-06, + "loss": 1.5296, + "step": 24353 + }, + { + "epoch": 5.32, + "learning_rate": 2.344985798558008e-06, + "loss": 0.9431, + "step": 24354 + }, + { + "epoch": 5.32, + "learning_rate": 2.344876556696526e-06, + "loss": 1.2116, + "step": 24355 + }, + { + "epoch": 5.32, + "learning_rate": 2.344767314835045e-06, + "loss": 2.5137, + "step": 24356 + }, + { + "epoch": 5.32, + "learning_rate": 2.3446580729735637e-06, + "loss": 0.8508, + "step": 24357 + }, + { + "epoch": 5.32, + "learning_rate": 2.3445488311120823e-06, + "loss": 2.5565, + "step": 24358 + }, + { + "epoch": 5.32, + "learning_rate": 2.344439589250601e-06, + "loss": 1.6614, + "step": 24359 + }, + { + "epoch": 5.32, + "learning_rate": 2.3443303473891195e-06, + "loss": 1.6919, + "step": 24360 + }, + { + "epoch": 5.32, + "learning_rate": 2.3442211055276386e-06, + "loss": 1.2654, + "step": 24361 + }, + { + "epoch": 5.32, + "learning_rate": 2.3441118636661568e-06, + "loss": 0.9389, + "step": 24362 + }, + { + "epoch": 5.32, + "learning_rate": 2.344002621804676e-06, + "loss": 1.9206, + "step": 24363 + }, + { + "epoch": 5.32, + "learning_rate": 2.3438933799431944e-06, + "loss": 0.8217, + "step": 24364 + }, + { + "epoch": 5.32, + "learning_rate": 2.343784138081713e-06, + "loss": 2.1167, + "step": 24365 + }, + { + "epoch": 5.32, + "learning_rate": 2.343674896220232e-06, + "loss": 0.4053, + "step": 24366 + }, + { + "epoch": 5.32, + "learning_rate": 2.3435656543587502e-06, + "loss": 0.816, + "step": 24367 + }, + { + "epoch": 5.32, + "learning_rate": 2.3434564124972693e-06, + "loss": 1.7314, + "step": 24368 + }, + { + "epoch": 5.32, + "learning_rate": 2.343347170635788e-06, + "loss": 1.6651, + "step": 24369 + }, + { + "epoch": 5.32, + "learning_rate": 2.3432379287743065e-06, + "loss": 3.0884, + "step": 24370 + }, + { + "epoch": 5.32, + "learning_rate": 2.343128686912825e-06, + "loss": 1.5337, + "step": 24371 + }, + { + "epoch": 5.32, + "learning_rate": 2.3430194450513437e-06, + "loss": 2.3994, + "step": 24372 + }, + { + "epoch": 5.32, + "learning_rate": 2.3429102031898628e-06, + "loss": 2.2569, + "step": 24373 + }, + { + "epoch": 5.32, + "learning_rate": 2.342800961328381e-06, + "loss": 1.4791, + "step": 24374 + }, + { + "epoch": 5.32, + "learning_rate": 2.3426917194669e-06, + "loss": 1.9218, + "step": 24375 + }, + { + "epoch": 5.32, + "learning_rate": 2.3425824776054186e-06, + "loss": 0.8439, + "step": 24376 + }, + { + "epoch": 5.32, + "learning_rate": 2.342473235743937e-06, + "loss": 1.1756, + "step": 24377 + }, + { + "epoch": 5.32, + "learning_rate": 2.3423639938824562e-06, + "loss": 2.141, + "step": 24378 + }, + { + "epoch": 5.32, + "learning_rate": 2.3422547520209744e-06, + "loss": 2.5263, + "step": 24379 + }, + { + "epoch": 5.32, + "learning_rate": 2.3421455101594935e-06, + "loss": 1.4138, + "step": 24380 + }, + { + "epoch": 5.32, + "learning_rate": 2.342036268298012e-06, + "loss": 2.5431, + "step": 24381 + }, + { + "epoch": 5.32, + "learning_rate": 2.3419270264365307e-06, + "loss": 2.906, + "step": 24382 + }, + { + "epoch": 5.32, + "learning_rate": 2.3418177845750493e-06, + "loss": 0.5339, + "step": 24383 + }, + { + "epoch": 5.32, + "learning_rate": 2.341708542713568e-06, + "loss": 0.628, + "step": 24384 + }, + { + "epoch": 5.32, + "learning_rate": 2.341599300852087e-06, + "loss": 1.0029, + "step": 24385 + }, + { + "epoch": 5.32, + "learning_rate": 2.341490058990605e-06, + "loss": 1.4965, + "step": 24386 + }, + { + "epoch": 5.32, + "learning_rate": 2.341380817129124e-06, + "loss": 0.8741, + "step": 24387 + }, + { + "epoch": 5.32, + "learning_rate": 2.3412715752676428e-06, + "loss": 1.8277, + "step": 24388 + }, + { + "epoch": 5.32, + "learning_rate": 2.3411623334061614e-06, + "loss": 1.6225, + "step": 24389 + }, + { + "epoch": 5.32, + "learning_rate": 2.3410530915446804e-06, + "loss": 1.2993, + "step": 24390 + }, + { + "epoch": 5.32, + "learning_rate": 2.3409438496831986e-06, + "loss": 1.4505, + "step": 24391 + }, + { + "epoch": 5.32, + "learning_rate": 2.3408346078217176e-06, + "loss": 1.372, + "step": 24392 + }, + { + "epoch": 5.32, + "learning_rate": 2.3407253659602362e-06, + "loss": 1.3651, + "step": 24393 + }, + { + "epoch": 5.32, + "learning_rate": 2.340616124098755e-06, + "loss": 1.5272, + "step": 24394 + }, + { + "epoch": 5.32, + "learning_rate": 2.3405068822372735e-06, + "loss": 2.1568, + "step": 24395 + }, + { + "epoch": 5.32, + "learning_rate": 2.340397640375792e-06, + "loss": 0.4566, + "step": 24396 + }, + { + "epoch": 5.32, + "learning_rate": 2.340288398514311e-06, + "loss": 1.6167, + "step": 24397 + }, + { + "epoch": 5.32, + "learning_rate": 2.3401791566528297e-06, + "loss": 2.1738, + "step": 24398 + }, + { + "epoch": 5.32, + "learning_rate": 2.3400699147913483e-06, + "loss": 1.3977, + "step": 24399 + }, + { + "epoch": 5.33, + "learning_rate": 2.339960672929867e-06, + "loss": 1.8121, + "step": 24400 + }, + { + "epoch": 5.33, + "learning_rate": 2.3398514310683856e-06, + "loss": 1.2075, + "step": 24401 + }, + { + "epoch": 5.33, + "learning_rate": 2.339742189206904e-06, + "loss": 2.2216, + "step": 24402 + }, + { + "epoch": 5.33, + "learning_rate": 2.3396329473454228e-06, + "loss": 0.481, + "step": 24403 + }, + { + "epoch": 5.33, + "learning_rate": 2.339523705483942e-06, + "loss": 2.1815, + "step": 24404 + }, + { + "epoch": 5.33, + "learning_rate": 2.3394144636224604e-06, + "loss": 0.6645, + "step": 24405 + }, + { + "epoch": 5.33, + "learning_rate": 2.339305221760979e-06, + "loss": 1.9886, + "step": 24406 + }, + { + "epoch": 5.33, + "learning_rate": 2.3391959798994976e-06, + "loss": 0.4485, + "step": 24407 + }, + { + "epoch": 5.33, + "learning_rate": 2.3390867380380163e-06, + "loss": 1.686, + "step": 24408 + }, + { + "epoch": 5.33, + "learning_rate": 2.338977496176535e-06, + "loss": 1.8111, + "step": 24409 + }, + { + "epoch": 5.33, + "learning_rate": 2.338868254315054e-06, + "loss": 1.1549, + "step": 24410 + }, + { + "epoch": 5.33, + "learning_rate": 2.3387590124535725e-06, + "loss": 1.6055, + "step": 24411 + }, + { + "epoch": 5.33, + "learning_rate": 2.338649770592091e-06, + "loss": 0.8986, + "step": 24412 + }, + { + "epoch": 5.33, + "learning_rate": 2.3385405287306097e-06, + "loss": 2.927, + "step": 24413 + }, + { + "epoch": 5.33, + "learning_rate": 2.3384312868691283e-06, + "loss": 1.479, + "step": 24414 + }, + { + "epoch": 5.33, + "learning_rate": 2.338322045007647e-06, + "loss": 1.2185, + "step": 24415 + }, + { + "epoch": 5.33, + "learning_rate": 2.3382128031461656e-06, + "loss": 0.4726, + "step": 24416 + }, + { + "epoch": 5.33, + "learning_rate": 2.3381035612846846e-06, + "loss": 1.6707, + "step": 24417 + }, + { + "epoch": 5.33, + "learning_rate": 2.3379943194232032e-06, + "loss": 2.4595, + "step": 24418 + }, + { + "epoch": 5.33, + "learning_rate": 2.337885077561722e-06, + "loss": 1.8815, + "step": 24419 + }, + { + "epoch": 5.33, + "learning_rate": 2.3377758357002404e-06, + "loss": 1.1669, + "step": 24420 + }, + { + "epoch": 5.33, + "learning_rate": 2.337666593838759e-06, + "loss": 2.0281, + "step": 24421 + }, + { + "epoch": 5.33, + "learning_rate": 2.337557351977278e-06, + "loss": 3.0643, + "step": 24422 + }, + { + "epoch": 5.33, + "learning_rate": 2.3374481101157967e-06, + "loss": 2.6664, + "step": 24423 + }, + { + "epoch": 5.33, + "learning_rate": 2.3373388682543153e-06, + "loss": 2.2409, + "step": 24424 + }, + { + "epoch": 5.33, + "learning_rate": 2.337229626392834e-06, + "loss": 2.873, + "step": 24425 + }, + { + "epoch": 5.33, + "learning_rate": 2.3371203845313525e-06, + "loss": 0.9638, + "step": 24426 + }, + { + "epoch": 5.33, + "learning_rate": 2.337011142669871e-06, + "loss": 0.4398, + "step": 24427 + }, + { + "epoch": 5.33, + "learning_rate": 2.3369019008083897e-06, + "loss": 2.3488, + "step": 24428 + }, + { + "epoch": 5.33, + "learning_rate": 2.3367926589469088e-06, + "loss": 1.6236, + "step": 24429 + }, + { + "epoch": 5.33, + "learning_rate": 2.3366834170854274e-06, + "loss": 1.24, + "step": 24430 + }, + { + "epoch": 5.33, + "learning_rate": 2.336574175223946e-06, + "loss": 1.2084, + "step": 24431 + }, + { + "epoch": 5.33, + "learning_rate": 2.3364649333624646e-06, + "loss": 0.532, + "step": 24432 + }, + { + "epoch": 5.33, + "learning_rate": 2.3363556915009832e-06, + "loss": 1.2391, + "step": 24433 + }, + { + "epoch": 5.33, + "learning_rate": 2.3362464496395023e-06, + "loss": 1.3715, + "step": 24434 + }, + { + "epoch": 5.33, + "learning_rate": 2.3361372077780204e-06, + "loss": 1.301, + "step": 24435 + }, + { + "epoch": 5.33, + "learning_rate": 2.3360279659165395e-06, + "loss": 1.8368, + "step": 24436 + }, + { + "epoch": 5.33, + "learning_rate": 2.335918724055058e-06, + "loss": 3.4399, + "step": 24437 + }, + { + "epoch": 5.33, + "learning_rate": 2.3358094821935767e-06, + "loss": 2.3564, + "step": 24438 + }, + { + "epoch": 5.33, + "learning_rate": 2.3357002403320953e-06, + "loss": 1.9835, + "step": 24439 + }, + { + "epoch": 5.33, + "learning_rate": 2.335590998470614e-06, + "loss": 1.4227, + "step": 24440 + }, + { + "epoch": 5.33, + "learning_rate": 2.335481756609133e-06, + "loss": 1.41, + "step": 24441 + }, + { + "epoch": 5.33, + "learning_rate": 2.3353725147476516e-06, + "loss": 2.5513, + "step": 24442 + }, + { + "epoch": 5.33, + "learning_rate": 2.33526327288617e-06, + "loss": 2.0342, + "step": 24443 + }, + { + "epoch": 5.33, + "learning_rate": 2.335154031024689e-06, + "loss": 0.4591, + "step": 24444 + }, + { + "epoch": 5.34, + "learning_rate": 2.3350447891632074e-06, + "loss": 1.9787, + "step": 24445 + }, + { + "epoch": 5.34, + "learning_rate": 2.3349355473017264e-06, + "loss": 1.1741, + "step": 24446 + }, + { + "epoch": 5.34, + "learning_rate": 2.3348263054402446e-06, + "loss": 1.0799, + "step": 24447 + }, + { + "epoch": 5.34, + "learning_rate": 2.3347170635787637e-06, + "loss": 1.2161, + "step": 24448 + }, + { + "epoch": 5.34, + "learning_rate": 2.3346078217172823e-06, + "loss": 2.5736, + "step": 24449 + }, + { + "epoch": 5.34, + "learning_rate": 2.334498579855801e-06, + "loss": 2.1072, + "step": 24450 + }, + { + "epoch": 5.34, + "learning_rate": 2.33438933799432e-06, + "loss": 0.3643, + "step": 24451 + }, + { + "epoch": 5.34, + "learning_rate": 2.334280096132838e-06, + "loss": 2.1356, + "step": 24452 + }, + { + "epoch": 5.34, + "learning_rate": 2.334170854271357e-06, + "loss": 1.9089, + "step": 24453 + }, + { + "epoch": 5.34, + "learning_rate": 2.3340616124098758e-06, + "loss": 1.1609, + "step": 24454 + }, + { + "epoch": 5.34, + "learning_rate": 2.3339523705483944e-06, + "loss": 1.5595, + "step": 24455 + }, + { + "epoch": 5.34, + "learning_rate": 2.333843128686913e-06, + "loss": 1.3622, + "step": 24456 + }, + { + "epoch": 5.34, + "learning_rate": 2.3337338868254316e-06, + "loss": 1.5014, + "step": 24457 + }, + { + "epoch": 5.34, + "learning_rate": 2.3336246449639506e-06, + "loss": 1.1905, + "step": 24458 + }, + { + "epoch": 5.34, + "learning_rate": 2.333515403102469e-06, + "loss": 1.837, + "step": 24459 + }, + { + "epoch": 5.34, + "learning_rate": 2.333406161240988e-06, + "loss": 1.4178, + "step": 24460 + }, + { + "epoch": 5.34, + "learning_rate": 2.3332969193795065e-06, + "loss": 1.5756, + "step": 24461 + }, + { + "epoch": 5.34, + "learning_rate": 2.333187677518025e-06, + "loss": 3.3335, + "step": 24462 + }, + { + "epoch": 5.34, + "learning_rate": 2.333078435656544e-06, + "loss": 1.0682, + "step": 24463 + }, + { + "epoch": 5.34, + "learning_rate": 2.3329691937950623e-06, + "loss": 1.9136, + "step": 24464 + }, + { + "epoch": 5.34, + "learning_rate": 2.3328599519335813e-06, + "loss": 1.2672, + "step": 24465 + }, + { + "epoch": 5.34, + "learning_rate": 2.3327507100721e-06, + "loss": 1.5418, + "step": 24466 + }, + { + "epoch": 5.34, + "learning_rate": 2.3326414682106185e-06, + "loss": 0.8149, + "step": 24467 + }, + { + "epoch": 5.34, + "learning_rate": 2.332532226349137e-06, + "loss": 0.7961, + "step": 24468 + }, + { + "epoch": 5.34, + "learning_rate": 2.3324229844876558e-06, + "loss": 1.0056, + "step": 24469 + }, + { + "epoch": 5.34, + "learning_rate": 2.332313742626175e-06, + "loss": 1.535, + "step": 24470 + }, + { + "epoch": 5.34, + "learning_rate": 2.332204500764693e-06, + "loss": 0.4502, + "step": 24471 + }, + { + "epoch": 5.34, + "learning_rate": 2.332095258903212e-06, + "loss": 0.8659, + "step": 24472 + }, + { + "epoch": 5.34, + "learning_rate": 2.3319860170417306e-06, + "loss": 1.578, + "step": 24473 + }, + { + "epoch": 5.34, + "learning_rate": 2.3318767751802492e-06, + "loss": 1.4491, + "step": 24474 + }, + { + "epoch": 5.34, + "learning_rate": 2.331767533318768e-06, + "loss": 2.1728, + "step": 24475 + }, + { + "epoch": 5.34, + "learning_rate": 2.3316582914572865e-06, + "loss": 1.1298, + "step": 24476 + }, + { + "epoch": 5.34, + "learning_rate": 2.3315490495958055e-06, + "loss": 1.7253, + "step": 24477 + }, + { + "epoch": 5.34, + "learning_rate": 2.331439807734324e-06, + "loss": 1.4211, + "step": 24478 + }, + { + "epoch": 5.34, + "learning_rate": 2.3313305658728427e-06, + "loss": 1.8212, + "step": 24479 + }, + { + "epoch": 5.34, + "learning_rate": 2.3312213240113613e-06, + "loss": 2.9288, + "step": 24480 + }, + { + "epoch": 5.34, + "learning_rate": 2.33111208214988e-06, + "loss": 1.352, + "step": 24481 + }, + { + "epoch": 5.34, + "learning_rate": 2.3310028402883986e-06, + "loss": 1.3366, + "step": 24482 + }, + { + "epoch": 5.34, + "learning_rate": 2.330893598426917e-06, + "loss": 1.3068, + "step": 24483 + }, + { + "epoch": 5.34, + "learning_rate": 2.330784356565436e-06, + "loss": 1.3809, + "step": 24484 + }, + { + "epoch": 5.34, + "learning_rate": 2.330675114703955e-06, + "loss": 2.4915, + "step": 24485 + }, + { + "epoch": 5.34, + "learning_rate": 2.3305658728424734e-06, + "loss": 1.1976, + "step": 24486 + }, + { + "epoch": 5.34, + "learning_rate": 2.330456630980992e-06, + "loss": 0.1359, + "step": 24487 + }, + { + "epoch": 5.34, + "learning_rate": 2.3303473891195106e-06, + "loss": 2.7914, + "step": 24488 + }, + { + "epoch": 5.34, + "learning_rate": 2.3302381472580297e-06, + "loss": 2.0339, + "step": 24489 + }, + { + "epoch": 5.34, + "learning_rate": 2.3301289053965483e-06, + "loss": 3.6422, + "step": 24490 + }, + { + "epoch": 5.35, + "learning_rate": 2.330019663535067e-06, + "loss": 0.7307, + "step": 24491 + }, + { + "epoch": 5.35, + "learning_rate": 2.3299104216735855e-06, + "loss": 1.8735, + "step": 24492 + }, + { + "epoch": 5.35, + "learning_rate": 2.329801179812104e-06, + "loss": 0.5784, + "step": 24493 + }, + { + "epoch": 5.35, + "learning_rate": 2.3296919379506227e-06, + "loss": 1.6385, + "step": 24494 + }, + { + "epoch": 5.35, + "learning_rate": 2.3295826960891418e-06, + "loss": 0.4909, + "step": 24495 + }, + { + "epoch": 5.35, + "learning_rate": 2.3294734542276604e-06, + "loss": 1.8756, + "step": 24496 + }, + { + "epoch": 5.35, + "learning_rate": 2.329364212366179e-06, + "loss": 0.8974, + "step": 24497 + }, + { + "epoch": 5.35, + "learning_rate": 2.3292549705046976e-06, + "loss": 3.0084, + "step": 24498 + }, + { + "epoch": 5.35, + "learning_rate": 2.329145728643216e-06, + "loss": 0.7447, + "step": 24499 + }, + { + "epoch": 5.35, + "learning_rate": 2.329036486781735e-06, + "loss": 1.9791, + "step": 24500 + }, + { + "epoch": 5.35, + "learning_rate": 2.3289272449202534e-06, + "loss": 0.9189, + "step": 24501 + }, + { + "epoch": 5.35, + "learning_rate": 2.3288180030587725e-06, + "loss": 2.9331, + "step": 24502 + }, + { + "epoch": 5.35, + "learning_rate": 2.328708761197291e-06, + "loss": 2.1368, + "step": 24503 + }, + { + "epoch": 5.35, + "learning_rate": 2.3285995193358097e-06, + "loss": 3.463, + "step": 24504 + }, + { + "epoch": 5.35, + "learning_rate": 2.3284902774743283e-06, + "loss": 2.04, + "step": 24505 + }, + { + "epoch": 5.35, + "learning_rate": 2.328381035612847e-06, + "loss": 1.1543, + "step": 24506 + }, + { + "epoch": 5.35, + "learning_rate": 2.328271793751366e-06, + "loss": 3.0924, + "step": 24507 + }, + { + "epoch": 5.35, + "learning_rate": 2.328162551889884e-06, + "loss": 2.268, + "step": 24508 + }, + { + "epoch": 5.35, + "learning_rate": 2.328053310028403e-06, + "loss": 1.6504, + "step": 24509 + }, + { + "epoch": 5.35, + "learning_rate": 2.3279440681669218e-06, + "loss": 0.8257, + "step": 24510 + }, + { + "epoch": 5.35, + "learning_rate": 2.3278348263054404e-06, + "loss": 0.8488, + "step": 24511 + }, + { + "epoch": 5.35, + "learning_rate": 2.327725584443959e-06, + "loss": 2.0032, + "step": 24512 + }, + { + "epoch": 5.35, + "learning_rate": 2.3276163425824776e-06, + "loss": 0.6192, + "step": 24513 + }, + { + "epoch": 5.35, + "learning_rate": 2.3275071007209966e-06, + "loss": 2.4786, + "step": 24514 + }, + { + "epoch": 5.35, + "learning_rate": 2.327397858859515e-06, + "loss": 1.8295, + "step": 24515 + }, + { + "epoch": 5.35, + "learning_rate": 2.327288616998034e-06, + "loss": 1.7901, + "step": 24516 + }, + { + "epoch": 5.35, + "learning_rate": 2.3271793751365525e-06, + "loss": 2.4862, + "step": 24517 + }, + { + "epoch": 5.35, + "learning_rate": 2.327070133275071e-06, + "loss": 0.9514, + "step": 24518 + }, + { + "epoch": 5.35, + "learning_rate": 2.32696089141359e-06, + "loss": 1.4592, + "step": 24519 + }, + { + "epoch": 5.35, + "learning_rate": 2.3268516495521083e-06, + "loss": 0.6674, + "step": 24520 + }, + { + "epoch": 5.35, + "learning_rate": 2.3267424076906273e-06, + "loss": 0.494, + "step": 24521 + }, + { + "epoch": 5.35, + "learning_rate": 2.326633165829146e-06, + "loss": 1.1015, + "step": 24522 + }, + { + "epoch": 5.35, + "learning_rate": 2.3265239239676646e-06, + "loss": 2.0644, + "step": 24523 + }, + { + "epoch": 5.35, + "learning_rate": 2.326414682106183e-06, + "loss": 0.8203, + "step": 24524 + }, + { + "epoch": 5.35, + "learning_rate": 2.326305440244702e-06, + "loss": 0.6928, + "step": 24525 + }, + { + "epoch": 5.35, + "learning_rate": 2.326196198383221e-06, + "loss": 2.6508, + "step": 24526 + }, + { + "epoch": 5.35, + "learning_rate": 2.326086956521739e-06, + "loss": 0.6761, + "step": 24527 + }, + { + "epoch": 5.35, + "learning_rate": 2.325977714660258e-06, + "loss": 1.3934, + "step": 24528 + }, + { + "epoch": 5.35, + "learning_rate": 2.3258684727987767e-06, + "loss": 2.3051, + "step": 24529 + }, + { + "epoch": 5.35, + "learning_rate": 2.3257592309372953e-06, + "loss": 1.2534, + "step": 24530 + }, + { + "epoch": 5.35, + "learning_rate": 2.3256499890758143e-06, + "loss": 1.795, + "step": 24531 + }, + { + "epoch": 5.35, + "learning_rate": 2.3255407472143325e-06, + "loss": 1.0536, + "step": 24532 + }, + { + "epoch": 5.35, + "learning_rate": 2.3254315053528515e-06, + "loss": 3.3851, + "step": 24533 + }, + { + "epoch": 5.35, + "learning_rate": 2.32532226349137e-06, + "loss": 1.5149, + "step": 24534 + }, + { + "epoch": 5.35, + "learning_rate": 2.3252130216298887e-06, + "loss": 0.6738, + "step": 24535 + }, + { + "epoch": 5.35, + "learning_rate": 2.3251037797684074e-06, + "loss": 1.7554, + "step": 24536 + }, + { + "epoch": 5.36, + "learning_rate": 2.324994537906926e-06, + "loss": 0.9272, + "step": 24537 + }, + { + "epoch": 5.36, + "learning_rate": 2.324885296045445e-06, + "loss": 2.3739, + "step": 24538 + }, + { + "epoch": 5.36, + "learning_rate": 2.3247760541839636e-06, + "loss": 3.3032, + "step": 24539 + }, + { + "epoch": 5.36, + "learning_rate": 2.3246668123224822e-06, + "loss": 1.7704, + "step": 24540 + }, + { + "epoch": 5.36, + "learning_rate": 2.324557570461001e-06, + "loss": 0.595, + "step": 24541 + }, + { + "epoch": 5.36, + "learning_rate": 2.3244483285995194e-06, + "loss": 2.0978, + "step": 24542 + }, + { + "epoch": 5.36, + "learning_rate": 2.3243390867380385e-06, + "loss": 1.5495, + "step": 24543 + }, + { + "epoch": 5.36, + "learning_rate": 2.3242298448765567e-06, + "loss": 2.2257, + "step": 24544 + }, + { + "epoch": 5.36, + "learning_rate": 2.3241206030150757e-06, + "loss": 1.7851, + "step": 24545 + }, + { + "epoch": 5.36, + "learning_rate": 2.3240113611535943e-06, + "loss": 0.7395, + "step": 24546 + }, + { + "epoch": 5.36, + "learning_rate": 2.323902119292113e-06, + "loss": 0.9575, + "step": 24547 + }, + { + "epoch": 5.36, + "learning_rate": 2.3237928774306315e-06, + "loss": 1.4712, + "step": 24548 + }, + { + "epoch": 5.36, + "learning_rate": 2.32368363556915e-06, + "loss": 1.759, + "step": 24549 + }, + { + "epoch": 5.36, + "learning_rate": 2.323574393707669e-06, + "loss": 1.196, + "step": 24550 + }, + { + "epoch": 5.36, + "learning_rate": 2.323465151846188e-06, + "loss": 0.8766, + "step": 24551 + }, + { + "epoch": 5.36, + "learning_rate": 2.3233559099847064e-06, + "loss": 2.5004, + "step": 24552 + }, + { + "epoch": 5.36, + "learning_rate": 2.323246668123225e-06, + "loss": 1.4832, + "step": 24553 + }, + { + "epoch": 5.36, + "learning_rate": 2.3231374262617436e-06, + "loss": 0.7317, + "step": 24554 + }, + { + "epoch": 5.36, + "learning_rate": 2.3230281844002622e-06, + "loss": 2.1191, + "step": 24555 + }, + { + "epoch": 5.36, + "learning_rate": 2.322918942538781e-06, + "loss": 1.0757, + "step": 24556 + }, + { + "epoch": 5.36, + "learning_rate": 2.3228097006773e-06, + "loss": 0.6468, + "step": 24557 + }, + { + "epoch": 5.36, + "learning_rate": 2.3227004588158185e-06, + "loss": 1.0947, + "step": 24558 + }, + { + "epoch": 5.36, + "learning_rate": 2.322591216954337e-06, + "loss": 2.1637, + "step": 24559 + }, + { + "epoch": 5.36, + "learning_rate": 2.3224819750928557e-06, + "loss": 0.8452, + "step": 24560 + }, + { + "epoch": 5.36, + "learning_rate": 2.3223727332313743e-06, + "loss": 1.5946, + "step": 24561 + }, + { + "epoch": 5.36, + "learning_rate": 2.3222634913698934e-06, + "loss": 1.3179, + "step": 24562 + }, + { + "epoch": 5.36, + "learning_rate": 2.322154249508412e-06, + "loss": 1.4554, + "step": 24563 + }, + { + "epoch": 5.36, + "learning_rate": 2.3220450076469306e-06, + "loss": 1.035, + "step": 24564 + }, + { + "epoch": 5.36, + "learning_rate": 2.321935765785449e-06, + "loss": 0.5753, + "step": 24565 + }, + { + "epoch": 5.36, + "learning_rate": 2.321826523923968e-06, + "loss": 1.5693, + "step": 24566 + }, + { + "epoch": 5.36, + "learning_rate": 2.3217172820624864e-06, + "loss": 1.1801, + "step": 24567 + }, + { + "epoch": 5.36, + "learning_rate": 2.321608040201005e-06, + "loss": 1.6, + "step": 24568 + }, + { + "epoch": 5.36, + "learning_rate": 2.321498798339524e-06, + "loss": 2.1624, + "step": 24569 + }, + { + "epoch": 5.36, + "learning_rate": 2.3213895564780427e-06, + "loss": 1.3416, + "step": 24570 + }, + { + "epoch": 5.36, + "learning_rate": 2.3212803146165613e-06, + "loss": 0.1498, + "step": 24571 + }, + { + "epoch": 5.36, + "learning_rate": 2.32117107275508e-06, + "loss": 0.9097, + "step": 24572 + }, + { + "epoch": 5.36, + "learning_rate": 2.3210618308935985e-06, + "loss": 2.123, + "step": 24573 + }, + { + "epoch": 5.36, + "learning_rate": 2.320952589032117e-06, + "loss": 0.6671, + "step": 24574 + }, + { + "epoch": 5.36, + "learning_rate": 2.320843347170636e-06, + "loss": 4.1545, + "step": 24575 + }, + { + "epoch": 5.36, + "learning_rate": 2.3207341053091548e-06, + "loss": 1.506, + "step": 24576 + }, + { + "epoch": 5.36, + "learning_rate": 2.3206248634476734e-06, + "loss": 1.2815, + "step": 24577 + }, + { + "epoch": 5.36, + "learning_rate": 2.320515621586192e-06, + "loss": 1.4872, + "step": 24578 + }, + { + "epoch": 5.36, + "learning_rate": 2.3204063797247106e-06, + "loss": 1.1503, + "step": 24579 + }, + { + "epoch": 5.36, + "learning_rate": 2.320297137863229e-06, + "loss": 1.3717, + "step": 24580 + }, + { + "epoch": 5.36, + "learning_rate": 2.320187896001748e-06, + "loss": 1.0665, + "step": 24581 + }, + { + "epoch": 5.36, + "learning_rate": 2.320078654140267e-06, + "loss": 1.3234, + "step": 24582 + }, + { + "epoch": 5.37, + "learning_rate": 2.3199694122787855e-06, + "loss": 0.4597, + "step": 24583 + }, + { + "epoch": 5.37, + "learning_rate": 2.319860170417304e-06, + "loss": 0.4801, + "step": 24584 + }, + { + "epoch": 5.37, + "learning_rate": 2.3197509285558227e-06, + "loss": 1.1912, + "step": 24585 + }, + { + "epoch": 5.37, + "learning_rate": 2.3196416866943413e-06, + "loss": 0.5755, + "step": 24586 + }, + { + "epoch": 5.37, + "learning_rate": 2.3195324448328603e-06, + "loss": 0.9348, + "step": 24587 + }, + { + "epoch": 5.37, + "learning_rate": 2.3194232029713785e-06, + "loss": 1.1749, + "step": 24588 + }, + { + "epoch": 5.37, + "learning_rate": 2.3193139611098976e-06, + "loss": 1.2635, + "step": 24589 + }, + { + "epoch": 5.37, + "learning_rate": 2.319204719248416e-06, + "loss": 1.2161, + "step": 24590 + }, + { + "epoch": 5.37, + "learning_rate": 2.3190954773869348e-06, + "loss": 2.8763, + "step": 24591 + }, + { + "epoch": 5.37, + "learning_rate": 2.318986235525454e-06, + "loss": 2.1935, + "step": 24592 + }, + { + "epoch": 5.37, + "learning_rate": 2.318876993663972e-06, + "loss": 1.5227, + "step": 24593 + }, + { + "epoch": 5.37, + "learning_rate": 2.318767751802491e-06, + "loss": 0.2405, + "step": 24594 + }, + { + "epoch": 5.37, + "learning_rate": 2.3186585099410096e-06, + "loss": 0.1219, + "step": 24595 + }, + { + "epoch": 5.37, + "learning_rate": 2.3185492680795283e-06, + "loss": 1.651, + "step": 24596 + }, + { + "epoch": 5.37, + "learning_rate": 2.318440026218047e-06, + "loss": 0.9494, + "step": 24597 + }, + { + "epoch": 5.37, + "learning_rate": 2.3183307843565655e-06, + "loss": 1.3242, + "step": 24598 + }, + { + "epoch": 5.37, + "learning_rate": 2.3182215424950845e-06, + "loss": 1.2185, + "step": 24599 + }, + { + "epoch": 5.37, + "learning_rate": 2.3181123006336027e-06, + "loss": 0.7631, + "step": 24600 + }, + { + "epoch": 5.37, + "learning_rate": 2.3180030587721217e-06, + "loss": 1.5377, + "step": 24601 + }, + { + "epoch": 5.37, + "learning_rate": 2.3178938169106403e-06, + "loss": 1.8427, + "step": 24602 + }, + { + "epoch": 5.37, + "learning_rate": 2.317784575049159e-06, + "loss": 1.484, + "step": 24603 + }, + { + "epoch": 5.37, + "learning_rate": 2.317675333187678e-06, + "loss": 2.2303, + "step": 24604 + }, + { + "epoch": 5.37, + "learning_rate": 2.317566091326196e-06, + "loss": 0.2302, + "step": 24605 + }, + { + "epoch": 5.37, + "learning_rate": 2.317456849464715e-06, + "loss": 1.7735, + "step": 24606 + }, + { + "epoch": 5.37, + "learning_rate": 2.317347607603234e-06, + "loss": 1.1243, + "step": 24607 + }, + { + "epoch": 5.37, + "learning_rate": 2.3172383657417524e-06, + "loss": 1.8827, + "step": 24608 + }, + { + "epoch": 5.37, + "learning_rate": 2.317129123880271e-06, + "loss": 3.3836, + "step": 24609 + }, + { + "epoch": 5.37, + "learning_rate": 2.3170198820187897e-06, + "loss": 1.3047, + "step": 24610 + }, + { + "epoch": 5.37, + "learning_rate": 2.3169106401573087e-06, + "loss": 1.622, + "step": 24611 + }, + { + "epoch": 5.37, + "learning_rate": 2.316801398295827e-06, + "loss": 1.1616, + "step": 24612 + }, + { + "epoch": 5.37, + "learning_rate": 2.316692156434346e-06, + "loss": 1.3127, + "step": 24613 + }, + { + "epoch": 5.37, + "learning_rate": 2.3165829145728645e-06, + "loss": 1.1175, + "step": 24614 + }, + { + "epoch": 5.37, + "learning_rate": 2.316473672711383e-06, + "loss": 1.9702, + "step": 24615 + }, + { + "epoch": 5.37, + "learning_rate": 2.316364430849902e-06, + "loss": 1.9894, + "step": 24616 + }, + { + "epoch": 5.37, + "learning_rate": 2.3162551889884204e-06, + "loss": 2.3355, + "step": 24617 + }, + { + "epoch": 5.37, + "learning_rate": 2.3161459471269394e-06, + "loss": 2.2092, + "step": 24618 + }, + { + "epoch": 5.37, + "learning_rate": 2.316036705265458e-06, + "loss": 1.9751, + "step": 24619 + }, + { + "epoch": 5.37, + "learning_rate": 2.3159274634039766e-06, + "loss": 1.2566, + "step": 24620 + }, + { + "epoch": 5.37, + "learning_rate": 2.3158182215424952e-06, + "loss": 1.2187, + "step": 24621 + }, + { + "epoch": 5.37, + "learning_rate": 2.315708979681014e-06, + "loss": 0.7247, + "step": 24622 + }, + { + "epoch": 5.37, + "learning_rate": 2.315599737819533e-06, + "loss": 1.3128, + "step": 24623 + }, + { + "epoch": 5.37, + "learning_rate": 2.315490495958051e-06, + "loss": 1.8797, + "step": 24624 + }, + { + "epoch": 5.37, + "learning_rate": 2.31538125409657e-06, + "loss": 2.3309, + "step": 24625 + }, + { + "epoch": 5.37, + "learning_rate": 2.3152720122350887e-06, + "loss": 0.527, + "step": 24626 + }, + { + "epoch": 5.37, + "learning_rate": 2.3151627703736073e-06, + "loss": 1.5523, + "step": 24627 + }, + { + "epoch": 5.37, + "learning_rate": 2.3150535285121263e-06, + "loss": 0.8182, + "step": 24628 + }, + { + "epoch": 5.38, + "learning_rate": 2.3149442866506445e-06, + "loss": 0.3606, + "step": 24629 + }, + { + "epoch": 5.38, + "learning_rate": 2.3148350447891636e-06, + "loss": 1.4485, + "step": 24630 + }, + { + "epoch": 5.38, + "learning_rate": 2.314725802927682e-06, + "loss": 1.0355, + "step": 24631 + }, + { + "epoch": 5.38, + "learning_rate": 2.314616561066201e-06, + "loss": 0.1532, + "step": 24632 + }, + { + "epoch": 5.38, + "learning_rate": 2.3145073192047194e-06, + "loss": 1.6351, + "step": 24633 + }, + { + "epoch": 5.38, + "learning_rate": 2.314398077343238e-06, + "loss": 2.2858, + "step": 24634 + }, + { + "epoch": 5.38, + "learning_rate": 2.314288835481757e-06, + "loss": 1.1204, + "step": 24635 + }, + { + "epoch": 5.38, + "learning_rate": 2.3141795936202757e-06, + "loss": 1.7552, + "step": 24636 + }, + { + "epoch": 5.38, + "learning_rate": 2.3140703517587943e-06, + "loss": 2.135, + "step": 24637 + }, + { + "epoch": 5.38, + "learning_rate": 2.313961109897313e-06, + "loss": 0.8654, + "step": 24638 + }, + { + "epoch": 5.38, + "learning_rate": 2.3138518680358315e-06, + "loss": 1.2986, + "step": 24639 + }, + { + "epoch": 5.38, + "learning_rate": 2.31374262617435e-06, + "loss": 0.7623, + "step": 24640 + }, + { + "epoch": 5.38, + "learning_rate": 2.3136333843128687e-06, + "loss": 2.8863, + "step": 24641 + }, + { + "epoch": 5.38, + "learning_rate": 2.3135241424513877e-06, + "loss": 1.6609, + "step": 24642 + }, + { + "epoch": 5.38, + "learning_rate": 2.3134149005899064e-06, + "loss": 0.9189, + "step": 24643 + }, + { + "epoch": 5.38, + "learning_rate": 2.313305658728425e-06, + "loss": 1.3097, + "step": 24644 + }, + { + "epoch": 5.38, + "learning_rate": 2.3131964168669436e-06, + "loss": 1.1571, + "step": 24645 + }, + { + "epoch": 5.38, + "learning_rate": 2.313087175005462e-06, + "loss": 1.8255, + "step": 24646 + }, + { + "epoch": 5.38, + "learning_rate": 2.312977933143981e-06, + "loss": 0.7827, + "step": 24647 + }, + { + "epoch": 5.38, + "learning_rate": 2.3128686912825e-06, + "loss": 1.0457, + "step": 24648 + }, + { + "epoch": 5.38, + "learning_rate": 2.3127594494210184e-06, + "loss": 0.7824, + "step": 24649 + }, + { + "epoch": 5.38, + "learning_rate": 2.312650207559537e-06, + "loss": 0.9137, + "step": 24650 + }, + { + "epoch": 5.38, + "learning_rate": 2.3125409656980557e-06, + "loss": 0.8116, + "step": 24651 + }, + { + "epoch": 5.38, + "learning_rate": 2.3124317238365743e-06, + "loss": 0.9718, + "step": 24652 + }, + { + "epoch": 5.38, + "learning_rate": 2.312322481975093e-06, + "loss": 1.6044, + "step": 24653 + }, + { + "epoch": 5.38, + "learning_rate": 2.3122132401136115e-06, + "loss": 1.5371, + "step": 24654 + }, + { + "epoch": 5.38, + "learning_rate": 2.3121039982521305e-06, + "loss": 1.6741, + "step": 24655 + }, + { + "epoch": 5.38, + "learning_rate": 2.311994756390649e-06, + "loss": 2.6759, + "step": 24656 + }, + { + "epoch": 5.38, + "learning_rate": 2.3118855145291678e-06, + "loss": 1.9751, + "step": 24657 + }, + { + "epoch": 5.38, + "learning_rate": 2.3117762726676864e-06, + "loss": 0.9593, + "step": 24658 + }, + { + "epoch": 5.38, + "learning_rate": 2.311667030806205e-06, + "loss": 2.0956, + "step": 24659 + }, + { + "epoch": 5.38, + "learning_rate": 2.311557788944724e-06, + "loss": 1.497, + "step": 24660 + }, + { + "epoch": 5.38, + "learning_rate": 2.3114485470832426e-06, + "loss": 0.7641, + "step": 24661 + }, + { + "epoch": 5.38, + "learning_rate": 2.3113393052217612e-06, + "loss": 1.2996, + "step": 24662 + }, + { + "epoch": 5.38, + "learning_rate": 2.31123006336028e-06, + "loss": 1.0429, + "step": 24663 + }, + { + "epoch": 5.38, + "learning_rate": 2.3111208214987985e-06, + "loss": 1.2661, + "step": 24664 + }, + { + "epoch": 5.38, + "learning_rate": 2.311011579637317e-06, + "loss": 3.1506, + "step": 24665 + }, + { + "epoch": 5.38, + "learning_rate": 2.3109023377758357e-06, + "loss": 0.2752, + "step": 24666 + }, + { + "epoch": 5.38, + "learning_rate": 2.3107930959143547e-06, + "loss": 2.879, + "step": 24667 + }, + { + "epoch": 5.38, + "learning_rate": 2.3106838540528733e-06, + "loss": 0.9439, + "step": 24668 + }, + { + "epoch": 5.38, + "learning_rate": 2.310574612191392e-06, + "loss": 1.1559, + "step": 24669 + }, + { + "epoch": 5.38, + "learning_rate": 2.3104653703299105e-06, + "loss": 2.3392, + "step": 24670 + }, + { + "epoch": 5.38, + "learning_rate": 2.310356128468429e-06, + "loss": 1.8336, + "step": 24671 + }, + { + "epoch": 5.38, + "learning_rate": 2.310246886606948e-06, + "loss": 1.3539, + "step": 24672 + }, + { + "epoch": 5.38, + "learning_rate": 2.3101376447454664e-06, + "loss": 1.4908, + "step": 24673 + }, + { + "epoch": 5.38, + "learning_rate": 2.3100284028839854e-06, + "loss": 0.8172, + "step": 24674 + }, + { + "epoch": 5.39, + "learning_rate": 2.309919161022504e-06, + "loss": 0.8954, + "step": 24675 + }, + { + "epoch": 5.39, + "learning_rate": 2.3098099191610226e-06, + "loss": 0.519, + "step": 24676 + }, + { + "epoch": 5.39, + "learning_rate": 2.3097006772995412e-06, + "loss": 0.8592, + "step": 24677 + }, + { + "epoch": 5.39, + "learning_rate": 2.30959143543806e-06, + "loss": 1.7604, + "step": 24678 + }, + { + "epoch": 5.39, + "learning_rate": 2.309482193576579e-06, + "loss": 1.701, + "step": 24679 + }, + { + "epoch": 5.39, + "learning_rate": 2.309372951715097e-06, + "loss": 1.4899, + "step": 24680 + }, + { + "epoch": 5.39, + "learning_rate": 2.309263709853616e-06, + "loss": 1.5386, + "step": 24681 + }, + { + "epoch": 5.39, + "learning_rate": 2.3091544679921347e-06, + "loss": 0.639, + "step": 24682 + }, + { + "epoch": 5.39, + "learning_rate": 2.3090452261306533e-06, + "loss": 0.4954, + "step": 24683 + }, + { + "epoch": 5.39, + "learning_rate": 2.3089359842691724e-06, + "loss": 2.4966, + "step": 24684 + }, + { + "epoch": 5.39, + "learning_rate": 2.3088267424076906e-06, + "loss": 2.253, + "step": 24685 + }, + { + "epoch": 5.39, + "learning_rate": 2.3087175005462096e-06, + "loss": 1.5635, + "step": 24686 + }, + { + "epoch": 5.39, + "learning_rate": 2.308608258684728e-06, + "loss": 1.5632, + "step": 24687 + }, + { + "epoch": 5.39, + "learning_rate": 2.308499016823247e-06, + "loss": 0.8743, + "step": 24688 + }, + { + "epoch": 5.39, + "learning_rate": 2.308389774961766e-06, + "loss": 0.481, + "step": 24689 + }, + { + "epoch": 5.39, + "learning_rate": 2.308280533100284e-06, + "loss": 1.359, + "step": 24690 + }, + { + "epoch": 5.39, + "learning_rate": 2.308171291238803e-06, + "loss": 2.1069, + "step": 24691 + }, + { + "epoch": 5.39, + "learning_rate": 2.3080620493773217e-06, + "loss": 0.9934, + "step": 24692 + }, + { + "epoch": 5.39, + "learning_rate": 2.3079528075158403e-06, + "loss": 0.2333, + "step": 24693 + }, + { + "epoch": 5.39, + "learning_rate": 2.307843565654359e-06, + "loss": 2.1386, + "step": 24694 + }, + { + "epoch": 5.39, + "learning_rate": 2.3077343237928775e-06, + "loss": 0.3564, + "step": 24695 + }, + { + "epoch": 5.39, + "learning_rate": 2.3076250819313966e-06, + "loss": 1.4296, + "step": 24696 + }, + { + "epoch": 5.39, + "learning_rate": 2.3075158400699147e-06, + "loss": 0.7847, + "step": 24697 + }, + { + "epoch": 5.39, + "learning_rate": 2.3074065982084338e-06, + "loss": 0.7825, + "step": 24698 + }, + { + "epoch": 5.39, + "learning_rate": 2.3072973563469524e-06, + "loss": 2.2832, + "step": 24699 + }, + { + "epoch": 5.39, + "learning_rate": 2.307188114485471e-06, + "loss": 1.4543, + "step": 24700 + }, + { + "epoch": 5.39, + "learning_rate": 2.30707887262399e-06, + "loss": 1.5539, + "step": 24701 + }, + { + "epoch": 5.39, + "learning_rate": 2.3069696307625082e-06, + "loss": 1.2316, + "step": 24702 + }, + { + "epoch": 5.39, + "learning_rate": 2.3068603889010273e-06, + "loss": 1.0417, + "step": 24703 + }, + { + "epoch": 5.39, + "learning_rate": 2.306751147039546e-06, + "loss": 0.5008, + "step": 24704 + }, + { + "epoch": 5.39, + "learning_rate": 2.3066419051780645e-06, + "loss": 0.919, + "step": 24705 + }, + { + "epoch": 5.39, + "learning_rate": 2.306532663316583e-06, + "loss": 2.3857, + "step": 24706 + }, + { + "epoch": 5.39, + "learning_rate": 2.3064234214551017e-06, + "loss": 2.3577, + "step": 24707 + }, + { + "epoch": 5.39, + "learning_rate": 2.3063141795936207e-06, + "loss": 2.3573, + "step": 24708 + }, + { + "epoch": 5.39, + "learning_rate": 2.306204937732139e-06, + "loss": 1.8214, + "step": 24709 + }, + { + "epoch": 5.39, + "learning_rate": 2.306095695870658e-06, + "loss": 0.7503, + "step": 24710 + }, + { + "epoch": 5.39, + "learning_rate": 2.3059864540091766e-06, + "loss": 2.064, + "step": 24711 + }, + { + "epoch": 5.39, + "learning_rate": 2.305877212147695e-06, + "loss": 0.4201, + "step": 24712 + }, + { + "epoch": 5.39, + "learning_rate": 2.3057679702862138e-06, + "loss": 0.3873, + "step": 24713 + }, + { + "epoch": 5.39, + "learning_rate": 2.3056587284247324e-06, + "loss": 1.7842, + "step": 24714 + }, + { + "epoch": 5.39, + "learning_rate": 2.3055494865632514e-06, + "loss": 1.5411, + "step": 24715 + }, + { + "epoch": 5.39, + "learning_rate": 2.30544024470177e-06, + "loss": 1.1574, + "step": 24716 + }, + { + "epoch": 5.39, + "learning_rate": 2.3053310028402887e-06, + "loss": 1.2576, + "step": 24717 + }, + { + "epoch": 5.39, + "learning_rate": 2.3052217609788073e-06, + "loss": 0.4423, + "step": 24718 + }, + { + "epoch": 5.39, + "learning_rate": 2.305112519117326e-06, + "loss": 0.4151, + "step": 24719 + }, + { + "epoch": 5.4, + "learning_rate": 2.3050032772558445e-06, + "loss": 0.8915, + "step": 24720 + }, + { + "epoch": 5.4, + "learning_rate": 2.304894035394363e-06, + "loss": 2.944, + "step": 24721 + }, + { + "epoch": 5.4, + "learning_rate": 2.304784793532882e-06, + "loss": 1.7331, + "step": 24722 + }, + { + "epoch": 5.4, + "learning_rate": 2.3046755516714007e-06, + "loss": 1.152, + "step": 24723 + }, + { + "epoch": 5.4, + "learning_rate": 2.3045663098099194e-06, + "loss": 0.8679, + "step": 24724 + }, + { + "epoch": 5.4, + "learning_rate": 2.304457067948438e-06, + "loss": 1.3731, + "step": 24725 + }, + { + "epoch": 5.4, + "learning_rate": 2.3043478260869566e-06, + "loss": 2.5318, + "step": 24726 + }, + { + "epoch": 5.4, + "learning_rate": 2.3042385842254756e-06, + "loss": 0.703, + "step": 24727 + }, + { + "epoch": 5.4, + "learning_rate": 2.3041293423639942e-06, + "loss": 0.9377, + "step": 24728 + }, + { + "epoch": 5.4, + "learning_rate": 2.304020100502513e-06, + "loss": 1.4458, + "step": 24729 + }, + { + "epoch": 5.4, + "learning_rate": 2.3039108586410314e-06, + "loss": 1.1379, + "step": 24730 + }, + { + "epoch": 5.4, + "learning_rate": 2.30380161677955e-06, + "loss": 0.8384, + "step": 24731 + }, + { + "epoch": 5.4, + "learning_rate": 2.3036923749180687e-06, + "loss": 0.6339, + "step": 24732 + }, + { + "epoch": 5.4, + "learning_rate": 2.3035831330565873e-06, + "loss": 0.9192, + "step": 24733 + }, + { + "epoch": 5.4, + "learning_rate": 2.3034738911951063e-06, + "loss": 0.2622, + "step": 24734 + }, + { + "epoch": 5.4, + "learning_rate": 2.303364649333625e-06, + "loss": 1.5121, + "step": 24735 + }, + { + "epoch": 5.4, + "learning_rate": 2.3032554074721435e-06, + "loss": 0.5949, + "step": 24736 + }, + { + "epoch": 5.4, + "learning_rate": 2.303146165610662e-06, + "loss": 1.4559, + "step": 24737 + }, + { + "epoch": 5.4, + "learning_rate": 2.3030369237491808e-06, + "loss": 2.0544, + "step": 24738 + }, + { + "epoch": 5.4, + "learning_rate": 2.3029276818876994e-06, + "loss": 1.8956, + "step": 24739 + }, + { + "epoch": 5.4, + "learning_rate": 2.3028184400262184e-06, + "loss": 1.1767, + "step": 24740 + }, + { + "epoch": 5.4, + "learning_rate": 2.302709198164737e-06, + "loss": 1.3093, + "step": 24741 + }, + { + "epoch": 5.4, + "learning_rate": 2.3025999563032556e-06, + "loss": 2.2526, + "step": 24742 + }, + { + "epoch": 5.4, + "learning_rate": 2.3024907144417742e-06, + "loss": 0.8472, + "step": 24743 + }, + { + "epoch": 5.4, + "learning_rate": 2.302381472580293e-06, + "loss": 2.5963, + "step": 24744 + }, + { + "epoch": 5.4, + "learning_rate": 2.302272230718812e-06, + "loss": 2.072, + "step": 24745 + }, + { + "epoch": 5.4, + "learning_rate": 2.30216298885733e-06, + "loss": 0.8391, + "step": 24746 + }, + { + "epoch": 5.4, + "learning_rate": 2.302053746995849e-06, + "loss": 1.1192, + "step": 24747 + }, + { + "epoch": 5.4, + "learning_rate": 2.3019445051343677e-06, + "loss": 1.009, + "step": 24748 + }, + { + "epoch": 5.4, + "learning_rate": 2.3018352632728863e-06, + "loss": 0.872, + "step": 24749 + }, + { + "epoch": 5.4, + "learning_rate": 2.301726021411405e-06, + "loss": 1.3277, + "step": 24750 + }, + { + "epoch": 5.4, + "learning_rate": 2.3016167795499235e-06, + "loss": 1.2279, + "step": 24751 + }, + { + "epoch": 5.4, + "learning_rate": 2.3015075376884426e-06, + "loss": 1.1798, + "step": 24752 + }, + { + "epoch": 5.4, + "learning_rate": 2.3013982958269608e-06, + "loss": 2.127, + "step": 24753 + }, + { + "epoch": 5.4, + "learning_rate": 2.30128905396548e-06, + "loss": 1.1271, + "step": 24754 + }, + { + "epoch": 5.4, + "learning_rate": 2.3011798121039984e-06, + "loss": 2.9431, + "step": 24755 + }, + { + "epoch": 5.4, + "learning_rate": 2.301070570242517e-06, + "loss": 2.5859, + "step": 24756 + }, + { + "epoch": 5.4, + "learning_rate": 2.300961328381036e-06, + "loss": 1.3477, + "step": 24757 + }, + { + "epoch": 5.4, + "learning_rate": 2.3008520865195542e-06, + "loss": 1.6534, + "step": 24758 + }, + { + "epoch": 5.4, + "learning_rate": 2.3007428446580733e-06, + "loss": 1.4582, + "step": 24759 + }, + { + "epoch": 5.4, + "learning_rate": 2.300633602796592e-06, + "loss": 0.5791, + "step": 24760 + }, + { + "epoch": 5.4, + "learning_rate": 2.3005243609351105e-06, + "loss": 0.9379, + "step": 24761 + }, + { + "epoch": 5.4, + "learning_rate": 2.300415119073629e-06, + "loss": 3.2608, + "step": 24762 + }, + { + "epoch": 5.4, + "learning_rate": 2.3003058772121477e-06, + "loss": 0.7993, + "step": 24763 + }, + { + "epoch": 5.4, + "learning_rate": 2.3001966353506668e-06, + "loss": 1.7433, + "step": 24764 + }, + { + "epoch": 5.4, + "learning_rate": 2.300087393489185e-06, + "loss": 0.739, + "step": 24765 + }, + { + "epoch": 5.41, + "learning_rate": 2.299978151627704e-06, + "loss": 1.8305, + "step": 24766 + }, + { + "epoch": 5.41, + "learning_rate": 2.2998689097662226e-06, + "loss": 1.437, + "step": 24767 + }, + { + "epoch": 5.41, + "learning_rate": 2.299759667904741e-06, + "loss": 1.6903, + "step": 24768 + }, + { + "epoch": 5.41, + "learning_rate": 2.2996504260432602e-06, + "loss": 0.6768, + "step": 24769 + }, + { + "epoch": 5.41, + "learning_rate": 2.2995411841817784e-06, + "loss": 1.2414, + "step": 24770 + }, + { + "epoch": 5.41, + "learning_rate": 2.2994319423202975e-06, + "loss": 1.6716, + "step": 24771 + }, + { + "epoch": 5.41, + "learning_rate": 2.299322700458816e-06, + "loss": 0.8069, + "step": 24772 + }, + { + "epoch": 5.41, + "learning_rate": 2.2992134585973347e-06, + "loss": 1.1548, + "step": 24773 + }, + { + "epoch": 5.41, + "learning_rate": 2.2991042167358533e-06, + "loss": 1.4348, + "step": 24774 + }, + { + "epoch": 5.41, + "learning_rate": 2.298994974874372e-06, + "loss": 2.7179, + "step": 24775 + }, + { + "epoch": 5.41, + "learning_rate": 2.298885733012891e-06, + "loss": 2.2841, + "step": 24776 + }, + { + "epoch": 5.41, + "learning_rate": 2.298776491151409e-06, + "loss": 1.2389, + "step": 24777 + }, + { + "epoch": 5.41, + "learning_rate": 2.298667249289928e-06, + "loss": 1.8327, + "step": 24778 + }, + { + "epoch": 5.41, + "learning_rate": 2.2985580074284468e-06, + "loss": 1.3546, + "step": 24779 + }, + { + "epoch": 5.41, + "learning_rate": 2.2984487655669654e-06, + "loss": 0.7315, + "step": 24780 + }, + { + "epoch": 5.41, + "learning_rate": 2.2983395237054844e-06, + "loss": 0.2944, + "step": 24781 + }, + { + "epoch": 5.41, + "learning_rate": 2.2982302818440026e-06, + "loss": 1.9793, + "step": 24782 + }, + { + "epoch": 5.41, + "learning_rate": 2.2981210399825216e-06, + "loss": 2.4149, + "step": 24783 + }, + { + "epoch": 5.41, + "learning_rate": 2.2980117981210402e-06, + "loss": 2.0213, + "step": 24784 + }, + { + "epoch": 5.41, + "learning_rate": 2.297902556259559e-06, + "loss": 1.5104, + "step": 24785 + }, + { + "epoch": 5.41, + "learning_rate": 2.2977933143980775e-06, + "loss": 1.1896, + "step": 24786 + }, + { + "epoch": 5.41, + "learning_rate": 2.297684072536596e-06, + "loss": 1.8494, + "step": 24787 + }, + { + "epoch": 5.41, + "learning_rate": 2.297574830675115e-06, + "loss": 1.4284, + "step": 24788 + }, + { + "epoch": 5.41, + "learning_rate": 2.2974655888136337e-06, + "loss": 4.2497, + "step": 24789 + }, + { + "epoch": 5.41, + "learning_rate": 2.2973563469521523e-06, + "loss": 1.1947, + "step": 24790 + }, + { + "epoch": 5.41, + "learning_rate": 2.297247105090671e-06, + "loss": 1.4523, + "step": 24791 + }, + { + "epoch": 5.41, + "learning_rate": 2.2971378632291896e-06, + "loss": 1.3458, + "step": 24792 + }, + { + "epoch": 5.41, + "learning_rate": 2.297028621367708e-06, + "loss": 1.4148, + "step": 24793 + }, + { + "epoch": 5.41, + "learning_rate": 2.2969193795062268e-06, + "loss": 1.4168, + "step": 24794 + }, + { + "epoch": 5.41, + "learning_rate": 2.296810137644746e-06, + "loss": 0.8973, + "step": 24795 + }, + { + "epoch": 5.41, + "learning_rate": 2.2967008957832644e-06, + "loss": 0.895, + "step": 24796 + }, + { + "epoch": 5.41, + "learning_rate": 2.296591653921783e-06, + "loss": 0.614, + "step": 24797 + }, + { + "epoch": 5.41, + "learning_rate": 2.2964824120603016e-06, + "loss": 0.7284, + "step": 24798 + }, + { + "epoch": 5.41, + "learning_rate": 2.2963731701988203e-06, + "loss": 0.8137, + "step": 24799 + }, + { + "epoch": 5.41, + "learning_rate": 2.2962639283373393e-06, + "loss": 0.306, + "step": 24800 + }, + { + "epoch": 5.41, + "learning_rate": 2.296154686475858e-06, + "loss": 1.6198, + "step": 24801 + }, + { + "epoch": 5.41, + "learning_rate": 2.2960454446143765e-06, + "loss": 1.0104, + "step": 24802 + }, + { + "epoch": 5.41, + "learning_rate": 2.295936202752895e-06, + "loss": 1.5204, + "step": 24803 + }, + { + "epoch": 5.41, + "learning_rate": 2.2958269608914137e-06, + "loss": 2.276, + "step": 24804 + }, + { + "epoch": 5.41, + "learning_rate": 2.2957177190299323e-06, + "loss": 2.1632, + "step": 24805 + }, + { + "epoch": 5.41, + "learning_rate": 2.295608477168451e-06, + "loss": 1.8172, + "step": 24806 + }, + { + "epoch": 5.41, + "learning_rate": 2.29549923530697e-06, + "loss": 0.9702, + "step": 24807 + }, + { + "epoch": 5.41, + "learning_rate": 2.2953899934454886e-06, + "loss": 1.7515, + "step": 24808 + }, + { + "epoch": 5.41, + "learning_rate": 2.2952807515840072e-06, + "loss": 3.0378, + "step": 24809 + }, + { + "epoch": 5.41, + "learning_rate": 2.295171509722526e-06, + "loss": 1.3084, + "step": 24810 + }, + { + "epoch": 5.41, + "learning_rate": 2.2950622678610444e-06, + "loss": 1.1233, + "step": 24811 + }, + { + "epoch": 5.42, + "learning_rate": 2.294953025999563e-06, + "loss": 2.0033, + "step": 24812 + }, + { + "epoch": 5.42, + "learning_rate": 2.294843784138082e-06, + "loss": 1.1155, + "step": 24813 + }, + { + "epoch": 5.42, + "learning_rate": 2.2947345422766007e-06, + "loss": 1.1954, + "step": 24814 + }, + { + "epoch": 5.42, + "learning_rate": 2.2946253004151193e-06, + "loss": 2.1541, + "step": 24815 + }, + { + "epoch": 5.42, + "learning_rate": 2.294516058553638e-06, + "loss": 2.3404, + "step": 24816 + }, + { + "epoch": 5.42, + "learning_rate": 2.2944068166921565e-06, + "loss": 3.0015, + "step": 24817 + }, + { + "epoch": 5.42, + "learning_rate": 2.294297574830675e-06, + "loss": 2.3697, + "step": 24818 + }, + { + "epoch": 5.42, + "learning_rate": 2.2941883329691937e-06, + "loss": 1.8422, + "step": 24819 + }, + { + "epoch": 5.42, + "learning_rate": 2.2940790911077128e-06, + "loss": 2.3892, + "step": 24820 + }, + { + "epoch": 5.42, + "learning_rate": 2.2939698492462314e-06, + "loss": 1.6423, + "step": 24821 + }, + { + "epoch": 5.42, + "learning_rate": 2.29386060738475e-06, + "loss": 1.5376, + "step": 24822 + }, + { + "epoch": 5.42, + "learning_rate": 2.2937513655232686e-06, + "loss": 2.6814, + "step": 24823 + }, + { + "epoch": 5.42, + "learning_rate": 2.2936421236617872e-06, + "loss": 1.6852, + "step": 24824 + }, + { + "epoch": 5.42, + "learning_rate": 2.2935328818003063e-06, + "loss": 1.1492, + "step": 24825 + }, + { + "epoch": 5.42, + "learning_rate": 2.2934236399388244e-06, + "loss": 0.7412, + "step": 24826 + }, + { + "epoch": 5.42, + "learning_rate": 2.2933143980773435e-06, + "loss": 1.041, + "step": 24827 + }, + { + "epoch": 5.42, + "learning_rate": 2.293205156215862e-06, + "loss": 0.7778, + "step": 24828 + }, + { + "epoch": 5.42, + "learning_rate": 2.2930959143543807e-06, + "loss": 1.6533, + "step": 24829 + }, + { + "epoch": 5.42, + "learning_rate": 2.2929866724928993e-06, + "loss": 1.0937, + "step": 24830 + }, + { + "epoch": 5.42, + "learning_rate": 2.292877430631418e-06, + "loss": 3.4194, + "step": 24831 + }, + { + "epoch": 5.42, + "learning_rate": 2.292768188769937e-06, + "loss": 1.5432, + "step": 24832 + }, + { + "epoch": 5.42, + "learning_rate": 2.2926589469084556e-06, + "loss": 0.6496, + "step": 24833 + }, + { + "epoch": 5.42, + "learning_rate": 2.292549705046974e-06, + "loss": 1.4479, + "step": 24834 + }, + { + "epoch": 5.42, + "learning_rate": 2.292440463185493e-06, + "loss": 2.011, + "step": 24835 + }, + { + "epoch": 5.42, + "learning_rate": 2.2923312213240114e-06, + "loss": 1.2648, + "step": 24836 + }, + { + "epoch": 5.42, + "learning_rate": 2.2922219794625304e-06, + "loss": 1.7353, + "step": 24837 + }, + { + "epoch": 5.42, + "learning_rate": 2.2921127376010486e-06, + "loss": 2.4188, + "step": 24838 + }, + { + "epoch": 5.42, + "learning_rate": 2.2920034957395677e-06, + "loss": 1.4047, + "step": 24839 + }, + { + "epoch": 5.42, + "learning_rate": 2.2918942538780863e-06, + "loss": 1.3111, + "step": 24840 + }, + { + "epoch": 5.42, + "learning_rate": 2.291785012016605e-06, + "loss": 0.9632, + "step": 24841 + }, + { + "epoch": 5.42, + "learning_rate": 2.291675770155124e-06, + "loss": 2.0982, + "step": 24842 + }, + { + "epoch": 5.42, + "learning_rate": 2.291566528293642e-06, + "loss": 0.8298, + "step": 24843 + }, + { + "epoch": 5.42, + "learning_rate": 2.291457286432161e-06, + "loss": 1.3178, + "step": 24844 + }, + { + "epoch": 5.42, + "learning_rate": 2.2913480445706798e-06, + "loss": 2.0909, + "step": 24845 + }, + { + "epoch": 5.42, + "learning_rate": 2.2912388027091984e-06, + "loss": 2.3926, + "step": 24846 + }, + { + "epoch": 5.42, + "learning_rate": 2.291129560847717e-06, + "loss": 1.6351, + "step": 24847 + }, + { + "epoch": 5.42, + "learning_rate": 2.2910203189862356e-06, + "loss": 1.6353, + "step": 24848 + }, + { + "epoch": 5.42, + "learning_rate": 2.2909110771247546e-06, + "loss": 1.42, + "step": 24849 + }, + { + "epoch": 5.42, + "learning_rate": 2.290801835263273e-06, + "loss": 1.0345, + "step": 24850 + }, + { + "epoch": 5.42, + "learning_rate": 2.290692593401792e-06, + "loss": 2.0446, + "step": 24851 + }, + { + "epoch": 5.42, + "learning_rate": 2.2905833515403105e-06, + "loss": 0.7762, + "step": 24852 + }, + { + "epoch": 5.42, + "learning_rate": 2.290474109678829e-06, + "loss": 1.0651, + "step": 24853 + }, + { + "epoch": 5.42, + "learning_rate": 2.290364867817348e-06, + "loss": 1.4642, + "step": 24854 + }, + { + "epoch": 5.42, + "learning_rate": 2.2902556259558663e-06, + "loss": 1.4692, + "step": 24855 + }, + { + "epoch": 5.42, + "learning_rate": 2.2901463840943853e-06, + "loss": 2.368, + "step": 24856 + }, + { + "epoch": 5.42, + "learning_rate": 2.290037142232904e-06, + "loss": 3.192, + "step": 24857 + }, + { + "epoch": 5.43, + "learning_rate": 2.2899279003714225e-06, + "loss": 2.5912, + "step": 24858 + }, + { + "epoch": 5.43, + "learning_rate": 2.289818658509941e-06, + "loss": 2.0482, + "step": 24859 + }, + { + "epoch": 5.43, + "learning_rate": 2.2897094166484598e-06, + "loss": 1.3976, + "step": 24860 + }, + { + "epoch": 5.43, + "learning_rate": 2.289600174786979e-06, + "loss": 1.6075, + "step": 24861 + }, + { + "epoch": 5.43, + "learning_rate": 2.289490932925497e-06, + "loss": 0.7012, + "step": 24862 + }, + { + "epoch": 5.43, + "learning_rate": 2.289381691064016e-06, + "loss": 1.4824, + "step": 24863 + }, + { + "epoch": 5.43, + "learning_rate": 2.2892724492025346e-06, + "loss": 3.2231, + "step": 24864 + }, + { + "epoch": 5.43, + "learning_rate": 2.2891632073410532e-06, + "loss": 2.8945, + "step": 24865 + }, + { + "epoch": 5.43, + "learning_rate": 2.2890539654795723e-06, + "loss": 1.8121, + "step": 24866 + }, + { + "epoch": 5.43, + "learning_rate": 2.2889447236180905e-06, + "loss": 3.915, + "step": 24867 + }, + { + "epoch": 5.43, + "learning_rate": 2.2888354817566095e-06, + "loss": 2.5516, + "step": 24868 + }, + { + "epoch": 5.43, + "learning_rate": 2.288726239895128e-06, + "loss": 1.8564, + "step": 24869 + }, + { + "epoch": 5.43, + "learning_rate": 2.2886169980336467e-06, + "loss": 1.7298, + "step": 24870 + }, + { + "epoch": 5.43, + "learning_rate": 2.2885077561721653e-06, + "loss": 1.0252, + "step": 24871 + }, + { + "epoch": 5.43, + "learning_rate": 2.288398514310684e-06, + "loss": 1.2209, + "step": 24872 + }, + { + "epoch": 5.43, + "learning_rate": 2.288289272449203e-06, + "loss": 0.4658, + "step": 24873 + }, + { + "epoch": 5.43, + "learning_rate": 2.288180030587721e-06, + "loss": 1.3613, + "step": 24874 + }, + { + "epoch": 5.43, + "learning_rate": 2.28807078872624e-06, + "loss": 1.8284, + "step": 24875 + }, + { + "epoch": 5.43, + "learning_rate": 2.287961546864759e-06, + "loss": 1.0836, + "step": 24876 + }, + { + "epoch": 5.43, + "learning_rate": 2.2878523050032774e-06, + "loss": 3.6882, + "step": 24877 + }, + { + "epoch": 5.43, + "learning_rate": 2.287743063141796e-06, + "loss": 1.8984, + "step": 24878 + }, + { + "epoch": 5.43, + "learning_rate": 2.2876338212803146e-06, + "loss": 1.8049, + "step": 24879 + }, + { + "epoch": 5.43, + "learning_rate": 2.2875245794188337e-06, + "loss": 2.9705, + "step": 24880 + }, + { + "epoch": 5.43, + "learning_rate": 2.2874153375573523e-06, + "loss": 0.8548, + "step": 24881 + }, + { + "epoch": 5.43, + "learning_rate": 2.287306095695871e-06, + "loss": 2.0651, + "step": 24882 + }, + { + "epoch": 5.43, + "learning_rate": 2.2871968538343895e-06, + "loss": 1.9181, + "step": 24883 + }, + { + "epoch": 5.43, + "learning_rate": 2.287087611972908e-06, + "loss": 1.1794, + "step": 24884 + }, + { + "epoch": 5.43, + "learning_rate": 2.2869783701114267e-06, + "loss": 0.966, + "step": 24885 + }, + { + "epoch": 5.43, + "learning_rate": 2.2868691282499458e-06, + "loss": 2.0096, + "step": 24886 + }, + { + "epoch": 5.43, + "learning_rate": 2.2867598863884644e-06, + "loss": 1.3157, + "step": 24887 + }, + { + "epoch": 5.43, + "learning_rate": 2.286650644526983e-06, + "loss": 1.003, + "step": 24888 + }, + { + "epoch": 5.43, + "learning_rate": 2.2865414026655016e-06, + "loss": 1.8863, + "step": 24889 + }, + { + "epoch": 5.43, + "learning_rate": 2.28643216080402e-06, + "loss": 0.4243, + "step": 24890 + }, + { + "epoch": 5.43, + "learning_rate": 2.286322918942539e-06, + "loss": 1.4754, + "step": 24891 + }, + { + "epoch": 5.43, + "learning_rate": 2.2862136770810574e-06, + "loss": 1.114, + "step": 24892 + }, + { + "epoch": 5.43, + "learning_rate": 2.2861044352195765e-06, + "loss": 1.6474, + "step": 24893 + }, + { + "epoch": 5.43, + "learning_rate": 2.285995193358095e-06, + "loss": 0.1942, + "step": 24894 + }, + { + "epoch": 5.43, + "learning_rate": 2.2858859514966137e-06, + "loss": 0.2432, + "step": 24895 + }, + { + "epoch": 5.43, + "learning_rate": 2.2857767096351323e-06, + "loss": 1.6506, + "step": 24896 + }, + { + "epoch": 5.43, + "learning_rate": 2.285667467773651e-06, + "loss": 2.5176, + "step": 24897 + }, + { + "epoch": 5.43, + "learning_rate": 2.28555822591217e-06, + "loss": 1.3583, + "step": 24898 + }, + { + "epoch": 5.43, + "learning_rate": 2.2854489840506886e-06, + "loss": 2.4135, + "step": 24899 + }, + { + "epoch": 5.43, + "learning_rate": 2.285339742189207e-06, + "loss": 1.3191, + "step": 24900 + }, + { + "epoch": 5.43, + "learning_rate": 2.2852305003277258e-06, + "loss": 2.4588, + "step": 24901 + }, + { + "epoch": 5.43, + "learning_rate": 2.2851212584662444e-06, + "loss": 2.3537, + "step": 24902 + }, + { + "epoch": 5.43, + "learning_rate": 2.285012016604763e-06, + "loss": 1.4786, + "step": 24903 + }, + { + "epoch": 5.44, + "learning_rate": 2.2849027747432816e-06, + "loss": 1.6662, + "step": 24904 + }, + { + "epoch": 5.44, + "learning_rate": 2.2847935328818006e-06, + "loss": 2.7711, + "step": 24905 + }, + { + "epoch": 5.44, + "learning_rate": 2.2846842910203193e-06, + "loss": 1.9925, + "step": 24906 + }, + { + "epoch": 5.44, + "learning_rate": 2.284575049158838e-06, + "loss": 1.0699, + "step": 24907 + }, + { + "epoch": 5.44, + "learning_rate": 2.2844658072973565e-06, + "loss": 0.9575, + "step": 24908 + }, + { + "epoch": 5.44, + "learning_rate": 2.284356565435875e-06, + "loss": 2.1052, + "step": 24909 + }, + { + "epoch": 5.44, + "learning_rate": 2.284247323574394e-06, + "loss": 0.1094, + "step": 24910 + }, + { + "epoch": 5.44, + "learning_rate": 2.2841380817129123e-06, + "loss": 1.0139, + "step": 24911 + }, + { + "epoch": 5.44, + "learning_rate": 2.2840288398514313e-06, + "loss": 1.43, + "step": 24912 + }, + { + "epoch": 5.44, + "learning_rate": 2.28391959798995e-06, + "loss": 0.4898, + "step": 24913 + }, + { + "epoch": 5.44, + "learning_rate": 2.2838103561284686e-06, + "loss": 2.2194, + "step": 24914 + }, + { + "epoch": 5.44, + "learning_rate": 2.283701114266987e-06, + "loss": 1.1713, + "step": 24915 + }, + { + "epoch": 5.44, + "learning_rate": 2.283591872405506e-06, + "loss": 2.7793, + "step": 24916 + }, + { + "epoch": 5.44, + "learning_rate": 2.283482630544025e-06, + "loss": 0.4268, + "step": 24917 + }, + { + "epoch": 5.44, + "learning_rate": 2.283373388682543e-06, + "loss": 0.4069, + "step": 24918 + }, + { + "epoch": 5.44, + "learning_rate": 2.283264146821062e-06, + "loss": 1.6606, + "step": 24919 + }, + { + "epoch": 5.44, + "learning_rate": 2.2831549049595807e-06, + "loss": 0.8849, + "step": 24920 + }, + { + "epoch": 5.44, + "learning_rate": 2.2830456630980993e-06, + "loss": 1.5331, + "step": 24921 + }, + { + "epoch": 5.44, + "learning_rate": 2.2829364212366183e-06, + "loss": 1.5324, + "step": 24922 + }, + { + "epoch": 5.44, + "learning_rate": 2.2828271793751365e-06, + "loss": 1.6612, + "step": 24923 + }, + { + "epoch": 5.44, + "learning_rate": 2.2827179375136555e-06, + "loss": 1.564, + "step": 24924 + }, + { + "epoch": 5.44, + "learning_rate": 2.282608695652174e-06, + "loss": 0.8796, + "step": 24925 + }, + { + "epoch": 5.44, + "learning_rate": 2.2824994537906927e-06, + "loss": 1.5434, + "step": 24926 + }, + { + "epoch": 5.44, + "learning_rate": 2.2823902119292114e-06, + "loss": 1.9307, + "step": 24927 + }, + { + "epoch": 5.44, + "learning_rate": 2.28228097006773e-06, + "loss": 3.3588, + "step": 24928 + }, + { + "epoch": 5.44, + "learning_rate": 2.282171728206249e-06, + "loss": 1.2492, + "step": 24929 + }, + { + "epoch": 5.44, + "learning_rate": 2.282062486344767e-06, + "loss": 1.5281, + "step": 24930 + }, + { + "epoch": 5.44, + "learning_rate": 2.2819532444832862e-06, + "loss": 0.5554, + "step": 24931 + }, + { + "epoch": 5.44, + "learning_rate": 2.281844002621805e-06, + "loss": 0.7121, + "step": 24932 + }, + { + "epoch": 5.44, + "learning_rate": 2.2817347607603234e-06, + "loss": 0.988, + "step": 24933 + }, + { + "epoch": 5.44, + "learning_rate": 2.2816255188988425e-06, + "loss": 0.8458, + "step": 24934 + }, + { + "epoch": 5.44, + "learning_rate": 2.2815162770373607e-06, + "loss": 2.3547, + "step": 24935 + }, + { + "epoch": 5.44, + "learning_rate": 2.2814070351758797e-06, + "loss": 0.4724, + "step": 24936 + }, + { + "epoch": 5.44, + "learning_rate": 2.2812977933143983e-06, + "loss": 1.2638, + "step": 24937 + }, + { + "epoch": 5.44, + "learning_rate": 2.281188551452917e-06, + "loss": 2.4711, + "step": 24938 + }, + { + "epoch": 5.44, + "learning_rate": 2.281079309591436e-06, + "loss": 1.4749, + "step": 24939 + }, + { + "epoch": 5.44, + "learning_rate": 2.280970067729954e-06, + "loss": 1.9491, + "step": 24940 + }, + { + "epoch": 5.44, + "learning_rate": 2.280860825868473e-06, + "loss": 2.1969, + "step": 24941 + }, + { + "epoch": 5.44, + "learning_rate": 2.280751584006992e-06, + "loss": 1.883, + "step": 24942 + }, + { + "epoch": 5.44, + "learning_rate": 2.2806423421455104e-06, + "loss": 2.5975, + "step": 24943 + }, + { + "epoch": 5.44, + "learning_rate": 2.280533100284029e-06, + "loss": 1.5287, + "step": 24944 + }, + { + "epoch": 5.44, + "learning_rate": 2.2804238584225476e-06, + "loss": 1.7925, + "step": 24945 + }, + { + "epoch": 5.44, + "learning_rate": 2.2803146165610667e-06, + "loss": 3.5773, + "step": 24946 + }, + { + "epoch": 5.44, + "learning_rate": 2.280205374699585e-06, + "loss": 1.9159, + "step": 24947 + }, + { + "epoch": 5.44, + "learning_rate": 2.280096132838104e-06, + "loss": 1.617, + "step": 24948 + }, + { + "epoch": 5.45, + "learning_rate": 2.2799868909766225e-06, + "loss": 1.3801, + "step": 24949 + }, + { + "epoch": 5.45, + "learning_rate": 2.279877649115141e-06, + "loss": 0.8562, + "step": 24950 + }, + { + "epoch": 5.45, + "learning_rate": 2.2797684072536597e-06, + "loss": 2.4732, + "step": 24951 + }, + { + "epoch": 5.45, + "learning_rate": 2.2796591653921783e-06, + "loss": 0.9493, + "step": 24952 + }, + { + "epoch": 5.45, + "learning_rate": 2.2795499235306974e-06, + "loss": 0.6292, + "step": 24953 + }, + { + "epoch": 5.45, + "learning_rate": 2.279440681669216e-06, + "loss": 1.7933, + "step": 24954 + }, + { + "epoch": 5.45, + "learning_rate": 2.2793314398077346e-06, + "loss": 1.6354, + "step": 24955 + }, + { + "epoch": 5.45, + "learning_rate": 2.279222197946253e-06, + "loss": 1.2266, + "step": 24956 + }, + { + "epoch": 5.45, + "learning_rate": 2.279112956084772e-06, + "loss": 1.8169, + "step": 24957 + }, + { + "epoch": 5.45, + "learning_rate": 2.2790037142232904e-06, + "loss": 0.9351, + "step": 24958 + }, + { + "epoch": 5.45, + "learning_rate": 2.278894472361809e-06, + "loss": 1.1386, + "step": 24959 + }, + { + "epoch": 5.45, + "learning_rate": 2.278785230500328e-06, + "loss": 0.7001, + "step": 24960 + }, + { + "epoch": 5.45, + "learning_rate": 2.2786759886388467e-06, + "loss": 0.8392, + "step": 24961 + }, + { + "epoch": 5.45, + "learning_rate": 2.2785667467773653e-06, + "loss": 1.9488, + "step": 24962 + }, + { + "epoch": 5.45, + "learning_rate": 2.278457504915884e-06, + "loss": 0.2791, + "step": 24963 + }, + { + "epoch": 5.45, + "learning_rate": 2.2783482630544025e-06, + "loss": 1.8124, + "step": 24964 + }, + { + "epoch": 5.45, + "learning_rate": 2.2782390211929215e-06, + "loss": 0.8434, + "step": 24965 + }, + { + "epoch": 5.45, + "learning_rate": 2.27812977933144e-06, + "loss": 3.468, + "step": 24966 + }, + { + "epoch": 5.45, + "learning_rate": 2.2780205374699588e-06, + "loss": 1.5531, + "step": 24967 + }, + { + "epoch": 5.45, + "learning_rate": 2.2779112956084774e-06, + "loss": 0.4284, + "step": 24968 + }, + { + "epoch": 5.45, + "learning_rate": 2.277802053746996e-06, + "loss": 1.8942, + "step": 24969 + }, + { + "epoch": 5.45, + "learning_rate": 2.2776928118855146e-06, + "loss": 1.8008, + "step": 24970 + }, + { + "epoch": 5.45, + "learning_rate": 2.277583570024033e-06, + "loss": 2.1698, + "step": 24971 + }, + { + "epoch": 5.45, + "learning_rate": 2.2774743281625522e-06, + "loss": 2.4949, + "step": 24972 + }, + { + "epoch": 5.45, + "learning_rate": 2.277365086301071e-06, + "loss": 1.6429, + "step": 24973 + }, + { + "epoch": 5.45, + "learning_rate": 2.2772558444395895e-06, + "loss": 0.6616, + "step": 24974 + }, + { + "epoch": 5.45, + "learning_rate": 2.277146602578108e-06, + "loss": 0.8354, + "step": 24975 + }, + { + "epoch": 5.45, + "learning_rate": 2.2770373607166267e-06, + "loss": 1.7372, + "step": 24976 + }, + { + "epoch": 5.45, + "learning_rate": 2.2769281188551453e-06, + "loss": 1.5883, + "step": 24977 + }, + { + "epoch": 5.45, + "learning_rate": 2.2768188769936643e-06, + "loss": 1.0152, + "step": 24978 + }, + { + "epoch": 5.45, + "learning_rate": 2.276709635132183e-06, + "loss": 1.4713, + "step": 24979 + }, + { + "epoch": 5.45, + "learning_rate": 2.2766003932707016e-06, + "loss": 1.086, + "step": 24980 + }, + { + "epoch": 5.45, + "learning_rate": 2.27649115140922e-06, + "loss": 5.2347, + "step": 24981 + }, + { + "epoch": 5.45, + "learning_rate": 2.2763819095477388e-06, + "loss": 0.7109, + "step": 24982 + }, + { + "epoch": 5.45, + "learning_rate": 2.276272667686258e-06, + "loss": 0.7821, + "step": 24983 + }, + { + "epoch": 5.45, + "learning_rate": 2.276163425824776e-06, + "loss": 1.6506, + "step": 24984 + }, + { + "epoch": 5.45, + "learning_rate": 2.276054183963295e-06, + "loss": 2.0175, + "step": 24985 + }, + { + "epoch": 5.45, + "learning_rate": 2.2759449421018136e-06, + "loss": 1.7909, + "step": 24986 + }, + { + "epoch": 5.45, + "learning_rate": 2.2758357002403323e-06, + "loss": 2.017, + "step": 24987 + }, + { + "epoch": 5.45, + "learning_rate": 2.275726458378851e-06, + "loss": 1.0572, + "step": 24988 + }, + { + "epoch": 5.45, + "learning_rate": 2.2756172165173695e-06, + "loss": 1.6084, + "step": 24989 + }, + { + "epoch": 5.45, + "learning_rate": 2.2755079746558885e-06, + "loss": 2.2734, + "step": 24990 + }, + { + "epoch": 5.45, + "learning_rate": 2.2753987327944067e-06, + "loss": 2.463, + "step": 24991 + }, + { + "epoch": 5.45, + "learning_rate": 2.2752894909329257e-06, + "loss": 2.9973, + "step": 24992 + }, + { + "epoch": 5.45, + "learning_rate": 2.2751802490714443e-06, + "loss": 1.1272, + "step": 24993 + }, + { + "epoch": 5.45, + "learning_rate": 2.275071007209963e-06, + "loss": 2.6329, + "step": 24994 + }, + { + "epoch": 5.46, + "learning_rate": 2.274961765348482e-06, + "loss": 0.191, + "step": 24995 + }, + { + "epoch": 5.46, + "learning_rate": 2.274852523487e-06, + "loss": 0.8204, + "step": 24996 + }, + { + "epoch": 5.46, + "learning_rate": 2.274743281625519e-06, + "loss": 1.5227, + "step": 24997 + }, + { + "epoch": 5.46, + "learning_rate": 2.274634039764038e-06, + "loss": 1.3625, + "step": 24998 + }, + { + "epoch": 5.46, + "learning_rate": 2.2745247979025564e-06, + "loss": 0.7279, + "step": 24999 + }, + { + "epoch": 5.46, + "learning_rate": 2.274415556041075e-06, + "loss": 1.962, + "step": 25000 + }, + { + "epoch": 5.46, + "learning_rate": 2.2743063141795937e-06, + "loss": 1.3434, + "step": 25001 + }, + { + "epoch": 5.46, + "learning_rate": 2.2741970723181127e-06, + "loss": 0.5869, + "step": 25002 + }, + { + "epoch": 5.46, + "learning_rate": 2.274087830456631e-06, + "loss": 1.4244, + "step": 25003 + }, + { + "epoch": 5.46, + "learning_rate": 2.27397858859515e-06, + "loss": 0.3521, + "step": 25004 + }, + { + "epoch": 5.46, + "learning_rate": 2.2738693467336685e-06, + "loss": 2.1831, + "step": 25005 + }, + { + "epoch": 5.46, + "learning_rate": 2.273760104872187e-06, + "loss": 0.2138, + "step": 25006 + }, + { + "epoch": 5.46, + "learning_rate": 2.273650863010706e-06, + "loss": 1.8917, + "step": 25007 + }, + { + "epoch": 5.46, + "learning_rate": 2.2735416211492244e-06, + "loss": 2.5581, + "step": 25008 + }, + { + "epoch": 5.46, + "learning_rate": 2.2734323792877434e-06, + "loss": 1.4782, + "step": 25009 + }, + { + "epoch": 5.46, + "learning_rate": 2.273323137426262e-06, + "loss": 1.3478, + "step": 25010 + }, + { + "epoch": 5.46, + "learning_rate": 2.2732138955647806e-06, + "loss": 2.8172, + "step": 25011 + }, + { + "epoch": 5.46, + "learning_rate": 2.2731046537032992e-06, + "loss": 1.0826, + "step": 25012 + }, + { + "epoch": 5.46, + "learning_rate": 2.272995411841818e-06, + "loss": 2.3753, + "step": 25013 + }, + { + "epoch": 5.46, + "learning_rate": 2.272886169980337e-06, + "loss": 1.459, + "step": 25014 + }, + { + "epoch": 5.46, + "learning_rate": 2.272776928118855e-06, + "loss": 1.4365, + "step": 25015 + }, + { + "epoch": 5.46, + "learning_rate": 2.272667686257374e-06, + "loss": 0.3798, + "step": 25016 + }, + { + "epoch": 5.46, + "learning_rate": 2.2725584443958927e-06, + "loss": 0.9219, + "step": 25017 + }, + { + "epoch": 5.46, + "learning_rate": 2.2724492025344113e-06, + "loss": 0.7056, + "step": 25018 + }, + { + "epoch": 5.46, + "learning_rate": 2.2723399606729303e-06, + "loss": 0.586, + "step": 25019 + }, + { + "epoch": 5.46, + "learning_rate": 2.2722307188114485e-06, + "loss": 1.8948, + "step": 25020 + }, + { + "epoch": 5.46, + "learning_rate": 2.2721214769499676e-06, + "loss": 1.7513, + "step": 25021 + }, + { + "epoch": 5.46, + "learning_rate": 2.272012235088486e-06, + "loss": 1.8968, + "step": 25022 + }, + { + "epoch": 5.46, + "learning_rate": 2.271902993227005e-06, + "loss": 0.7797, + "step": 25023 + }, + { + "epoch": 5.46, + "learning_rate": 2.2717937513655234e-06, + "loss": 1.4572, + "step": 25024 + }, + { + "epoch": 5.46, + "learning_rate": 2.271684509504042e-06, + "loss": 1.0346, + "step": 25025 + }, + { + "epoch": 5.46, + "learning_rate": 2.271575267642561e-06, + "loss": 2.4376, + "step": 25026 + }, + { + "epoch": 5.46, + "learning_rate": 2.2714660257810792e-06, + "loss": 1.075, + "step": 25027 + }, + { + "epoch": 5.46, + "learning_rate": 2.2713567839195983e-06, + "loss": 0.8294, + "step": 25028 + }, + { + "epoch": 5.46, + "learning_rate": 2.271247542058117e-06, + "loss": 1.5614, + "step": 25029 + }, + { + "epoch": 5.46, + "learning_rate": 2.2711383001966355e-06, + "loss": 0.613, + "step": 25030 + }, + { + "epoch": 5.46, + "learning_rate": 2.271029058335154e-06, + "loss": 1.3704, + "step": 25031 + }, + { + "epoch": 5.46, + "learning_rate": 2.2709198164736727e-06, + "loss": 1.1721, + "step": 25032 + }, + { + "epoch": 5.46, + "learning_rate": 2.2708105746121917e-06, + "loss": 1.0222, + "step": 25033 + }, + { + "epoch": 5.46, + "learning_rate": 2.2707013327507104e-06, + "loss": 2.0605, + "step": 25034 + }, + { + "epoch": 5.46, + "learning_rate": 2.270592090889229e-06, + "loss": 0.842, + "step": 25035 + }, + { + "epoch": 5.46, + "learning_rate": 2.2704828490277476e-06, + "loss": 1.6317, + "step": 25036 + }, + { + "epoch": 5.46, + "learning_rate": 2.270373607166266e-06, + "loss": 1.0472, + "step": 25037 + }, + { + "epoch": 5.46, + "learning_rate": 2.2702643653047852e-06, + "loss": 1.8446, + "step": 25038 + }, + { + "epoch": 5.46, + "learning_rate": 2.270155123443304e-06, + "loss": 1.8264, + "step": 25039 + }, + { + "epoch": 5.46, + "learning_rate": 2.2700458815818224e-06, + "loss": 0.9262, + "step": 25040 + }, + { + "epoch": 5.47, + "learning_rate": 2.269936639720341e-06, + "loss": 2.2189, + "step": 25041 + }, + { + "epoch": 5.47, + "learning_rate": 2.2698273978588597e-06, + "loss": 1.4284, + "step": 25042 + }, + { + "epoch": 5.47, + "learning_rate": 2.2697181559973783e-06, + "loss": 1.9942, + "step": 25043 + }, + { + "epoch": 5.47, + "learning_rate": 2.269608914135897e-06, + "loss": 1.7176, + "step": 25044 + }, + { + "epoch": 5.47, + "learning_rate": 2.269499672274416e-06, + "loss": 1.6214, + "step": 25045 + }, + { + "epoch": 5.47, + "learning_rate": 2.2693904304129345e-06, + "loss": 2.3777, + "step": 25046 + }, + { + "epoch": 5.47, + "learning_rate": 2.269281188551453e-06, + "loss": 1.4165, + "step": 25047 + }, + { + "epoch": 5.47, + "learning_rate": 2.2691719466899718e-06, + "loss": 1.5654, + "step": 25048 + }, + { + "epoch": 5.47, + "learning_rate": 2.2690627048284904e-06, + "loss": 0.9489, + "step": 25049 + }, + { + "epoch": 5.47, + "learning_rate": 2.268953462967009e-06, + "loss": 0.8117, + "step": 25050 + }, + { + "epoch": 5.47, + "learning_rate": 2.268844221105528e-06, + "loss": 0.8398, + "step": 25051 + }, + { + "epoch": 5.47, + "learning_rate": 2.2687349792440466e-06, + "loss": 0.3398, + "step": 25052 + }, + { + "epoch": 5.47, + "learning_rate": 2.2686257373825652e-06, + "loss": 0.7292, + "step": 25053 + }, + { + "epoch": 5.47, + "learning_rate": 2.268516495521084e-06, + "loss": 0.6649, + "step": 25054 + }, + { + "epoch": 5.47, + "learning_rate": 2.2684072536596025e-06, + "loss": 0.8, + "step": 25055 + }, + { + "epoch": 5.47, + "learning_rate": 2.268298011798121e-06, + "loss": 0.949, + "step": 25056 + }, + { + "epoch": 5.47, + "learning_rate": 2.2681887699366397e-06, + "loss": 0.7027, + "step": 25057 + }, + { + "epoch": 5.47, + "learning_rate": 2.2680795280751587e-06, + "loss": 1.7878, + "step": 25058 + }, + { + "epoch": 5.47, + "learning_rate": 2.2679702862136773e-06, + "loss": 1.5297, + "step": 25059 + }, + { + "epoch": 5.47, + "learning_rate": 2.267861044352196e-06, + "loss": 2.2018, + "step": 25060 + }, + { + "epoch": 5.47, + "learning_rate": 2.2677518024907145e-06, + "loss": 2.5744, + "step": 25061 + }, + { + "epoch": 5.47, + "learning_rate": 2.267642560629233e-06, + "loss": 1.6651, + "step": 25062 + }, + { + "epoch": 5.47, + "learning_rate": 2.267533318767752e-06, + "loss": 1.6027, + "step": 25063 + }, + { + "epoch": 5.47, + "learning_rate": 2.2674240769062704e-06, + "loss": 1.5769, + "step": 25064 + }, + { + "epoch": 5.47, + "learning_rate": 2.2673148350447894e-06, + "loss": 0.9782, + "step": 25065 + }, + { + "epoch": 5.47, + "learning_rate": 2.267205593183308e-06, + "loss": 2.1496, + "step": 25066 + }, + { + "epoch": 5.47, + "learning_rate": 2.2670963513218266e-06, + "loss": 1.0132, + "step": 25067 + }, + { + "epoch": 5.47, + "learning_rate": 2.2669871094603452e-06, + "loss": 1.1524, + "step": 25068 + }, + { + "epoch": 5.47, + "learning_rate": 2.266877867598864e-06, + "loss": 0.7168, + "step": 25069 + }, + { + "epoch": 5.47, + "learning_rate": 2.266768625737383e-06, + "loss": 1.3278, + "step": 25070 + }, + { + "epoch": 5.47, + "learning_rate": 2.2666593838759015e-06, + "loss": 0.6514, + "step": 25071 + }, + { + "epoch": 5.47, + "learning_rate": 2.26655014201442e-06, + "loss": 1.6181, + "step": 25072 + }, + { + "epoch": 5.47, + "learning_rate": 2.2664409001529387e-06, + "loss": 3.8469, + "step": 25073 + }, + { + "epoch": 5.47, + "learning_rate": 2.2663316582914573e-06, + "loss": 0.7242, + "step": 25074 + }, + { + "epoch": 5.47, + "learning_rate": 2.2662224164299764e-06, + "loss": 1.7479, + "step": 25075 + }, + { + "epoch": 5.47, + "learning_rate": 2.2661131745684946e-06, + "loss": 1.8911, + "step": 25076 + }, + { + "epoch": 5.47, + "learning_rate": 2.2660039327070136e-06, + "loss": 1.5408, + "step": 25077 + }, + { + "epoch": 5.47, + "learning_rate": 2.265894690845532e-06, + "loss": 0.7724, + "step": 25078 + }, + { + "epoch": 5.47, + "learning_rate": 2.265785448984051e-06, + "loss": 1.3749, + "step": 25079 + }, + { + "epoch": 5.47, + "learning_rate": 2.2656762071225694e-06, + "loss": 1.3825, + "step": 25080 + }, + { + "epoch": 5.47, + "learning_rate": 2.265566965261088e-06, + "loss": 0.8868, + "step": 25081 + }, + { + "epoch": 5.47, + "learning_rate": 2.265457723399607e-06, + "loss": 0.2522, + "step": 25082 + }, + { + "epoch": 5.47, + "learning_rate": 2.2653484815381257e-06, + "loss": 0.8666, + "step": 25083 + }, + { + "epoch": 5.47, + "learning_rate": 2.2652392396766443e-06, + "loss": 2.6387, + "step": 25084 + }, + { + "epoch": 5.47, + "learning_rate": 2.265129997815163e-06, + "loss": 1.3773, + "step": 25085 + }, + { + "epoch": 5.47, + "learning_rate": 2.2650207559536815e-06, + "loss": 0.6785, + "step": 25086 + }, + { + "epoch": 5.48, + "learning_rate": 2.2649115140922006e-06, + "loss": 2.4782, + "step": 25087 + }, + { + "epoch": 5.48, + "learning_rate": 2.2648022722307187e-06, + "loss": 0.7105, + "step": 25088 + }, + { + "epoch": 5.48, + "learning_rate": 2.2646930303692378e-06, + "loss": 0.3672, + "step": 25089 + }, + { + "epoch": 5.48, + "learning_rate": 2.2645837885077564e-06, + "loss": 3.0142, + "step": 25090 + }, + { + "epoch": 5.48, + "learning_rate": 2.264474546646275e-06, + "loss": 1.9685, + "step": 25091 + }, + { + "epoch": 5.48, + "learning_rate": 2.264365304784794e-06, + "loss": 1.4555, + "step": 25092 + }, + { + "epoch": 5.48, + "learning_rate": 2.2642560629233122e-06, + "loss": 0.7969, + "step": 25093 + }, + { + "epoch": 5.48, + "learning_rate": 2.2641468210618313e-06, + "loss": 1.5412, + "step": 25094 + }, + { + "epoch": 5.48, + "learning_rate": 2.26403757920035e-06, + "loss": 0.829, + "step": 25095 + }, + { + "epoch": 5.48, + "learning_rate": 2.2639283373388685e-06, + "loss": 1.9075, + "step": 25096 + }, + { + "epoch": 5.48, + "learning_rate": 2.263819095477387e-06, + "loss": 1.407, + "step": 25097 + }, + { + "epoch": 5.48, + "learning_rate": 2.2637098536159057e-06, + "loss": 0.738, + "step": 25098 + }, + { + "epoch": 5.48, + "learning_rate": 2.2636006117544247e-06, + "loss": 1.921, + "step": 25099 + }, + { + "epoch": 5.48, + "learning_rate": 2.263491369892943e-06, + "loss": 2.3892, + "step": 25100 + }, + { + "epoch": 5.48, + "learning_rate": 2.263382128031462e-06, + "loss": 0.435, + "step": 25101 + }, + { + "epoch": 5.48, + "learning_rate": 2.2632728861699806e-06, + "loss": 1.7655, + "step": 25102 + }, + { + "epoch": 5.48, + "learning_rate": 2.263163644308499e-06, + "loss": 1.7473, + "step": 25103 + }, + { + "epoch": 5.48, + "learning_rate": 2.263054402447018e-06, + "loss": 2.4989, + "step": 25104 + }, + { + "epoch": 5.48, + "learning_rate": 2.2629451605855364e-06, + "loss": 0.8343, + "step": 25105 + }, + { + "epoch": 5.48, + "learning_rate": 2.2628359187240554e-06, + "loss": 1.7324, + "step": 25106 + }, + { + "epoch": 5.48, + "learning_rate": 2.262726676862574e-06, + "loss": 2.629, + "step": 25107 + }, + { + "epoch": 5.48, + "learning_rate": 2.2626174350010927e-06, + "loss": 4.2078, + "step": 25108 + }, + { + "epoch": 5.48, + "learning_rate": 2.2625081931396113e-06, + "loss": 1.7899, + "step": 25109 + }, + { + "epoch": 5.48, + "learning_rate": 2.26239895127813e-06, + "loss": 2.0423, + "step": 25110 + }, + { + "epoch": 5.48, + "learning_rate": 2.262289709416649e-06, + "loss": 0.9649, + "step": 25111 + }, + { + "epoch": 5.48, + "learning_rate": 2.262180467555167e-06, + "loss": 0.7768, + "step": 25112 + }, + { + "epoch": 5.48, + "learning_rate": 2.262071225693686e-06, + "loss": 0.5665, + "step": 25113 + }, + { + "epoch": 5.48, + "learning_rate": 2.2619619838322047e-06, + "loss": 1.639, + "step": 25114 + }, + { + "epoch": 5.48, + "learning_rate": 2.2618527419707234e-06, + "loss": 1.0514, + "step": 25115 + }, + { + "epoch": 5.48, + "learning_rate": 2.261743500109242e-06, + "loss": 2.3054, + "step": 25116 + }, + { + "epoch": 5.48, + "learning_rate": 2.2616342582477606e-06, + "loss": 0.6364, + "step": 25117 + }, + { + "epoch": 5.48, + "learning_rate": 2.2615250163862796e-06, + "loss": 0.5724, + "step": 25118 + }, + { + "epoch": 5.48, + "learning_rate": 2.2614157745247982e-06, + "loss": 0.9746, + "step": 25119 + }, + { + "epoch": 5.48, + "learning_rate": 2.261306532663317e-06, + "loss": 1.1136, + "step": 25120 + }, + { + "epoch": 5.48, + "learning_rate": 2.2611972908018354e-06, + "loss": 1.6872, + "step": 25121 + }, + { + "epoch": 5.48, + "learning_rate": 2.261088048940354e-06, + "loss": 1.4484, + "step": 25122 + }, + { + "epoch": 5.48, + "learning_rate": 2.2609788070788727e-06, + "loss": 1.8559, + "step": 25123 + }, + { + "epoch": 5.48, + "learning_rate": 2.2608695652173913e-06, + "loss": 1.9733, + "step": 25124 + }, + { + "epoch": 5.48, + "learning_rate": 2.2607603233559103e-06, + "loss": 1.0776, + "step": 25125 + }, + { + "epoch": 5.48, + "learning_rate": 2.260651081494429e-06, + "loss": 2.5503, + "step": 25126 + }, + { + "epoch": 5.48, + "learning_rate": 2.2605418396329475e-06, + "loss": 2.4487, + "step": 25127 + }, + { + "epoch": 5.48, + "learning_rate": 2.260432597771466e-06, + "loss": 1.3741, + "step": 25128 + }, + { + "epoch": 5.48, + "learning_rate": 2.2603233559099848e-06, + "loss": 1.8622, + "step": 25129 + }, + { + "epoch": 5.48, + "learning_rate": 2.2602141140485034e-06, + "loss": 2.2011, + "step": 25130 + }, + { + "epoch": 5.48, + "learning_rate": 2.2601048721870224e-06, + "loss": 0.4427, + "step": 25131 + }, + { + "epoch": 5.48, + "learning_rate": 2.259995630325541e-06, + "loss": 2.2715, + "step": 25132 + }, + { + "epoch": 5.49, + "learning_rate": 2.2598863884640596e-06, + "loss": 1.9314, + "step": 25133 + }, + { + "epoch": 5.49, + "learning_rate": 2.2597771466025782e-06, + "loss": 1.1119, + "step": 25134 + }, + { + "epoch": 5.49, + "learning_rate": 2.259667904741097e-06, + "loss": 1.4899, + "step": 25135 + }, + { + "epoch": 5.49, + "learning_rate": 2.259558662879616e-06, + "loss": 0.9217, + "step": 25136 + }, + { + "epoch": 5.49, + "learning_rate": 2.2594494210181345e-06, + "loss": 2.1012, + "step": 25137 + }, + { + "epoch": 5.49, + "learning_rate": 2.259340179156653e-06, + "loss": 3.7309, + "step": 25138 + }, + { + "epoch": 5.49, + "learning_rate": 2.2592309372951717e-06, + "loss": 2.0998, + "step": 25139 + }, + { + "epoch": 5.49, + "learning_rate": 2.2591216954336903e-06, + "loss": 1.1385, + "step": 25140 + }, + { + "epoch": 5.49, + "learning_rate": 2.259012453572209e-06, + "loss": 0.2308, + "step": 25141 + }, + { + "epoch": 5.49, + "learning_rate": 2.2589032117107275e-06, + "loss": 3.0256, + "step": 25142 + }, + { + "epoch": 5.49, + "learning_rate": 2.2587939698492466e-06, + "loss": 4.28, + "step": 25143 + }, + { + "epoch": 5.49, + "learning_rate": 2.258684727987765e-06, + "loss": 0.1448, + "step": 25144 + }, + { + "epoch": 5.49, + "learning_rate": 2.258575486126284e-06, + "loss": 0.6993, + "step": 25145 + }, + { + "epoch": 5.49, + "learning_rate": 2.2584662442648024e-06, + "loss": 2.8021, + "step": 25146 + }, + { + "epoch": 5.49, + "learning_rate": 2.258357002403321e-06, + "loss": 1.5506, + "step": 25147 + }, + { + "epoch": 5.49, + "learning_rate": 2.25824776054184e-06, + "loss": 1.9521, + "step": 25148 + }, + { + "epoch": 5.49, + "learning_rate": 2.2581385186803582e-06, + "loss": 1.9779, + "step": 25149 + }, + { + "epoch": 5.49, + "learning_rate": 2.2580292768188773e-06, + "loss": 0.8337, + "step": 25150 + }, + { + "epoch": 5.49, + "learning_rate": 2.257920034957396e-06, + "loss": 1.0688, + "step": 25151 + }, + { + "epoch": 5.49, + "learning_rate": 2.2578107930959145e-06, + "loss": 0.2272, + "step": 25152 + }, + { + "epoch": 5.49, + "learning_rate": 2.257701551234433e-06, + "loss": 2.3592, + "step": 25153 + }, + { + "epoch": 5.49, + "learning_rate": 2.2575923093729517e-06, + "loss": 1.1581, + "step": 25154 + }, + { + "epoch": 5.49, + "learning_rate": 2.2574830675114708e-06, + "loss": 0.7543, + "step": 25155 + }, + { + "epoch": 5.49, + "learning_rate": 2.257373825649989e-06, + "loss": 1.5737, + "step": 25156 + }, + { + "epoch": 5.49, + "learning_rate": 2.257264583788508e-06, + "loss": 2.6863, + "step": 25157 + }, + { + "epoch": 5.49, + "learning_rate": 2.2571553419270266e-06, + "loss": 1.3639, + "step": 25158 + }, + { + "epoch": 5.49, + "learning_rate": 2.257046100065545e-06, + "loss": 1.4788, + "step": 25159 + }, + { + "epoch": 5.49, + "learning_rate": 2.2569368582040642e-06, + "loss": 3.2418, + "step": 25160 + }, + { + "epoch": 5.49, + "learning_rate": 2.2568276163425824e-06, + "loss": 0.5454, + "step": 25161 + }, + { + "epoch": 5.49, + "learning_rate": 2.2567183744811015e-06, + "loss": 0.6194, + "step": 25162 + }, + { + "epoch": 5.49, + "learning_rate": 2.25660913261962e-06, + "loss": 0.3616, + "step": 25163 + }, + { + "epoch": 5.49, + "learning_rate": 2.2564998907581387e-06, + "loss": 0.733, + "step": 25164 + }, + { + "epoch": 5.49, + "learning_rate": 2.2563906488966573e-06, + "loss": 1.8679, + "step": 25165 + }, + { + "epoch": 5.49, + "learning_rate": 2.256281407035176e-06, + "loss": 1.5652, + "step": 25166 + }, + { + "epoch": 5.49, + "learning_rate": 2.256172165173695e-06, + "loss": 1.9096, + "step": 25167 + }, + { + "epoch": 5.49, + "learning_rate": 2.256062923312213e-06, + "loss": 2.0451, + "step": 25168 + }, + { + "epoch": 5.49, + "learning_rate": 2.255953681450732e-06, + "loss": 1.6178, + "step": 25169 + }, + { + "epoch": 5.49, + "learning_rate": 2.2558444395892508e-06, + "loss": 0.3532, + "step": 25170 + }, + { + "epoch": 5.49, + "learning_rate": 2.2557351977277694e-06, + "loss": 1.4775, + "step": 25171 + }, + { + "epoch": 5.49, + "learning_rate": 2.2556259558662884e-06, + "loss": 0.4222, + "step": 25172 + }, + { + "epoch": 5.49, + "learning_rate": 2.2555167140048066e-06, + "loss": 1.4058, + "step": 25173 + }, + { + "epoch": 5.49, + "learning_rate": 2.2554074721433256e-06, + "loss": 0.4793, + "step": 25174 + }, + { + "epoch": 5.49, + "learning_rate": 2.2552982302818442e-06, + "loss": 1.4706, + "step": 25175 + }, + { + "epoch": 5.49, + "learning_rate": 2.255188988420363e-06, + "loss": 0.5163, + "step": 25176 + }, + { + "epoch": 5.49, + "learning_rate": 2.2550797465588815e-06, + "loss": 1.5697, + "step": 25177 + }, + { + "epoch": 5.49, + "learning_rate": 2.2549705046974e-06, + "loss": 1.9251, + "step": 25178 + }, + { + "epoch": 5.5, + "learning_rate": 2.254861262835919e-06, + "loss": 0.6274, + "step": 25179 + }, + { + "epoch": 5.5, + "learning_rate": 2.2547520209744377e-06, + "loss": 1.1466, + "step": 25180 + }, + { + "epoch": 5.5, + "learning_rate": 2.2546427791129563e-06, + "loss": 2.5115, + "step": 25181 + }, + { + "epoch": 5.5, + "learning_rate": 2.254533537251475e-06, + "loss": 2.2087, + "step": 25182 + }, + { + "epoch": 5.5, + "learning_rate": 2.2544242953899936e-06, + "loss": 1.3766, + "step": 25183 + }, + { + "epoch": 5.5, + "learning_rate": 2.2543150535285126e-06, + "loss": 1.0353, + "step": 25184 + }, + { + "epoch": 5.5, + "learning_rate": 2.2542058116670308e-06, + "loss": 1.0996, + "step": 25185 + }, + { + "epoch": 5.5, + "learning_rate": 2.25409656980555e-06, + "loss": 2.3153, + "step": 25186 + }, + { + "epoch": 5.5, + "learning_rate": 2.2539873279440684e-06, + "loss": 0.6415, + "step": 25187 + }, + { + "epoch": 5.5, + "learning_rate": 2.253878086082587e-06, + "loss": 0.3931, + "step": 25188 + }, + { + "epoch": 5.5, + "learning_rate": 2.2537688442211056e-06, + "loss": 2.4039, + "step": 25189 + }, + { + "epoch": 5.5, + "learning_rate": 2.2536596023596243e-06, + "loss": 0.89, + "step": 25190 + }, + { + "epoch": 5.5, + "learning_rate": 2.2535503604981433e-06, + "loss": 2.4429, + "step": 25191 + }, + { + "epoch": 5.5, + "learning_rate": 2.253441118636662e-06, + "loss": 1.7802, + "step": 25192 + }, + { + "epoch": 5.5, + "learning_rate": 2.2533318767751805e-06, + "loss": 1.2248, + "step": 25193 + }, + { + "epoch": 5.5, + "learning_rate": 2.253222634913699e-06, + "loss": 2.4415, + "step": 25194 + }, + { + "epoch": 5.5, + "learning_rate": 2.2531133930522177e-06, + "loss": 1.4097, + "step": 25195 + }, + { + "epoch": 5.5, + "learning_rate": 2.2530041511907363e-06, + "loss": 0.5731, + "step": 25196 + }, + { + "epoch": 5.5, + "learning_rate": 2.252894909329255e-06, + "loss": 0.1093, + "step": 25197 + }, + { + "epoch": 5.5, + "learning_rate": 2.252785667467774e-06, + "loss": 2.4657, + "step": 25198 + }, + { + "epoch": 5.5, + "learning_rate": 2.2526764256062926e-06, + "loss": 1.2743, + "step": 25199 + }, + { + "epoch": 5.5, + "learning_rate": 2.2525671837448112e-06, + "loss": 0.7731, + "step": 25200 + }, + { + "epoch": 5.5, + "learning_rate": 2.25245794188333e-06, + "loss": 1.3619, + "step": 25201 + }, + { + "epoch": 5.5, + "learning_rate": 2.2523487000218484e-06, + "loss": 2.9159, + "step": 25202 + }, + { + "epoch": 5.5, + "learning_rate": 2.252239458160367e-06, + "loss": 3.2292, + "step": 25203 + }, + { + "epoch": 5.5, + "learning_rate": 2.252130216298886e-06, + "loss": 1.4976, + "step": 25204 + }, + { + "epoch": 5.5, + "learning_rate": 2.2520209744374047e-06, + "loss": 1.5472, + "step": 25205 + }, + { + "epoch": 5.5, + "learning_rate": 2.2519117325759233e-06, + "loss": 1.0877, + "step": 25206 + }, + { + "epoch": 5.5, + "learning_rate": 2.251802490714442e-06, + "loss": 1.8961, + "step": 25207 + }, + { + "epoch": 5.5, + "learning_rate": 2.2516932488529605e-06, + "loss": 1.6651, + "step": 25208 + }, + { + "epoch": 5.5, + "learning_rate": 2.251584006991479e-06, + "loss": 1.4404, + "step": 25209 + }, + { + "epoch": 5.5, + "learning_rate": 2.251474765129998e-06, + "loss": 2.1268, + "step": 25210 + }, + { + "epoch": 5.5, + "learning_rate": 2.2513655232685168e-06, + "loss": 0.8568, + "step": 25211 + }, + { + "epoch": 5.5, + "learning_rate": 2.2512562814070354e-06, + "loss": 2.1364, + "step": 25212 + }, + { + "epoch": 5.5, + "learning_rate": 2.251147039545554e-06, + "loss": 1.3969, + "step": 25213 + }, + { + "epoch": 5.5, + "learning_rate": 2.2510377976840726e-06, + "loss": 3.2572, + "step": 25214 + }, + { + "epoch": 5.5, + "learning_rate": 2.2509285558225912e-06, + "loss": 1.7144, + "step": 25215 + }, + { + "epoch": 5.5, + "learning_rate": 2.2508193139611103e-06, + "loss": 2.6583, + "step": 25216 + }, + { + "epoch": 5.5, + "learning_rate": 2.250710072099629e-06, + "loss": 0.793, + "step": 25217 + }, + { + "epoch": 5.5, + "learning_rate": 2.2506008302381475e-06, + "loss": 1.925, + "step": 25218 + }, + { + "epoch": 5.5, + "learning_rate": 2.250491588376666e-06, + "loss": 1.1838, + "step": 25219 + }, + { + "epoch": 5.5, + "learning_rate": 2.2503823465151847e-06, + "loss": 1.1259, + "step": 25220 + }, + { + "epoch": 5.5, + "learning_rate": 2.2502731046537033e-06, + "loss": 0.9867, + "step": 25221 + }, + { + "epoch": 5.5, + "learning_rate": 2.250163862792222e-06, + "loss": 2.9232, + "step": 25222 + }, + { + "epoch": 5.5, + "learning_rate": 2.250054620930741e-06, + "loss": 1.0192, + "step": 25223 + }, + { + "epoch": 5.51, + "learning_rate": 2.2499453790692596e-06, + "loss": 2.2058, + "step": 25224 + }, + { + "epoch": 5.51, + "learning_rate": 2.249836137207778e-06, + "loss": 0.5061, + "step": 25225 + }, + { + "epoch": 5.51, + "learning_rate": 2.249726895346297e-06, + "loss": 2.2753, + "step": 25226 + }, + { + "epoch": 5.51, + "learning_rate": 2.2496176534848154e-06, + "loss": 1.2652, + "step": 25227 + }, + { + "epoch": 5.51, + "learning_rate": 2.2495084116233344e-06, + "loss": 0.8648, + "step": 25228 + }, + { + "epoch": 5.51, + "learning_rate": 2.2493991697618526e-06, + "loss": 1.4522, + "step": 25229 + }, + { + "epoch": 5.51, + "learning_rate": 2.2492899279003717e-06, + "loss": 1.4846, + "step": 25230 + }, + { + "epoch": 5.51, + "learning_rate": 2.2491806860388903e-06, + "loss": 1.9247, + "step": 25231 + }, + { + "epoch": 5.51, + "learning_rate": 2.249071444177409e-06, + "loss": 1.0655, + "step": 25232 + }, + { + "epoch": 5.51, + "learning_rate": 2.248962202315928e-06, + "loss": 0.6852, + "step": 25233 + }, + { + "epoch": 5.51, + "learning_rate": 2.248852960454446e-06, + "loss": 1.4594, + "step": 25234 + }, + { + "epoch": 5.51, + "learning_rate": 2.248743718592965e-06, + "loss": 0.9949, + "step": 25235 + }, + { + "epoch": 5.51, + "learning_rate": 2.2486344767314838e-06, + "loss": 0.6492, + "step": 25236 + }, + { + "epoch": 5.51, + "learning_rate": 2.2485252348700024e-06, + "loss": 1.3692, + "step": 25237 + }, + { + "epoch": 5.51, + "learning_rate": 2.248415993008521e-06, + "loss": 3.331, + "step": 25238 + }, + { + "epoch": 5.51, + "learning_rate": 2.2483067511470396e-06, + "loss": 0.5971, + "step": 25239 + }, + { + "epoch": 5.51, + "learning_rate": 2.2481975092855586e-06, + "loss": 1.6284, + "step": 25240 + }, + { + "epoch": 5.51, + "learning_rate": 2.248088267424077e-06, + "loss": 1.4126, + "step": 25241 + }, + { + "epoch": 5.51, + "learning_rate": 2.247979025562596e-06, + "loss": 0.9494, + "step": 25242 + }, + { + "epoch": 5.51, + "learning_rate": 2.2478697837011145e-06, + "loss": 2.1672, + "step": 25243 + }, + { + "epoch": 5.51, + "learning_rate": 2.247760541839633e-06, + "loss": 0.8827, + "step": 25244 + }, + { + "epoch": 5.51, + "learning_rate": 2.247651299978152e-06, + "loss": 1.2535, + "step": 25245 + }, + { + "epoch": 5.51, + "learning_rate": 2.2475420581166703e-06, + "loss": 1.3978, + "step": 25246 + }, + { + "epoch": 5.51, + "learning_rate": 2.2474328162551893e-06, + "loss": 0.7153, + "step": 25247 + }, + { + "epoch": 5.51, + "learning_rate": 2.247323574393708e-06, + "loss": 0.4962, + "step": 25248 + }, + { + "epoch": 5.51, + "learning_rate": 2.2472143325322265e-06, + "loss": 0.8264, + "step": 25249 + }, + { + "epoch": 5.51, + "learning_rate": 2.247105090670745e-06, + "loss": 2.3087, + "step": 25250 + }, + { + "epoch": 5.51, + "learning_rate": 2.2469958488092638e-06, + "loss": 2.3227, + "step": 25251 + }, + { + "epoch": 5.51, + "learning_rate": 2.246886606947783e-06, + "loss": 1.7411, + "step": 25252 + }, + { + "epoch": 5.51, + "learning_rate": 2.246777365086301e-06, + "loss": 1.5131, + "step": 25253 + }, + { + "epoch": 5.51, + "learning_rate": 2.24666812322482e-06, + "loss": 0.347, + "step": 25254 + }, + { + "epoch": 5.51, + "learning_rate": 2.2465588813633386e-06, + "loss": 1.9926, + "step": 25255 + }, + { + "epoch": 5.51, + "learning_rate": 2.2464496395018572e-06, + "loss": 1.2015, + "step": 25256 + }, + { + "epoch": 5.51, + "learning_rate": 2.2463403976403763e-06, + "loss": 0.4216, + "step": 25257 + }, + { + "epoch": 5.51, + "learning_rate": 2.2462311557788945e-06, + "loss": 1.1473, + "step": 25258 + }, + { + "epoch": 5.51, + "learning_rate": 2.2461219139174135e-06, + "loss": 1.5772, + "step": 25259 + }, + { + "epoch": 5.51, + "learning_rate": 2.246012672055932e-06, + "loss": 1.1517, + "step": 25260 + }, + { + "epoch": 5.51, + "learning_rate": 2.2459034301944507e-06, + "loss": 1.2538, + "step": 25261 + }, + { + "epoch": 5.51, + "learning_rate": 2.2457941883329693e-06, + "loss": 1.1256, + "step": 25262 + }, + { + "epoch": 5.51, + "learning_rate": 2.245684946471488e-06, + "loss": 0.3873, + "step": 25263 + }, + { + "epoch": 5.51, + "learning_rate": 2.245575704610007e-06, + "loss": 1.519, + "step": 25264 + }, + { + "epoch": 5.51, + "learning_rate": 2.245466462748525e-06, + "loss": 0.84, + "step": 25265 + }, + { + "epoch": 5.51, + "learning_rate": 2.245357220887044e-06, + "loss": 1.9226, + "step": 25266 + }, + { + "epoch": 5.51, + "learning_rate": 2.245247979025563e-06, + "loss": 1.4621, + "step": 25267 + }, + { + "epoch": 5.51, + "learning_rate": 2.2451387371640814e-06, + "loss": 0.7025, + "step": 25268 + }, + { + "epoch": 5.51, + "learning_rate": 2.2450294953026e-06, + "loss": 1.0403, + "step": 25269 + }, + { + "epoch": 5.52, + "learning_rate": 2.2449202534411186e-06, + "loss": 1.2832, + "step": 25270 + }, + { + "epoch": 5.52, + "learning_rate": 2.2448110115796377e-06, + "loss": 2.1281, + "step": 25271 + }, + { + "epoch": 5.52, + "learning_rate": 2.2447017697181563e-06, + "loss": 1.1851, + "step": 25272 + }, + { + "epoch": 5.52, + "learning_rate": 2.244592527856675e-06, + "loss": 1.8031, + "step": 25273 + }, + { + "epoch": 5.52, + "learning_rate": 2.2444832859951935e-06, + "loss": 1.1049, + "step": 25274 + }, + { + "epoch": 5.52, + "learning_rate": 2.244374044133712e-06, + "loss": 1.8755, + "step": 25275 + }, + { + "epoch": 5.52, + "learning_rate": 2.244264802272231e-06, + "loss": 0.4757, + "step": 25276 + }, + { + "epoch": 5.52, + "learning_rate": 2.2441555604107493e-06, + "loss": 2.1437, + "step": 25277 + }, + { + "epoch": 5.52, + "learning_rate": 2.2440463185492684e-06, + "loss": 1.9736, + "step": 25278 + }, + { + "epoch": 5.52, + "learning_rate": 2.243937076687787e-06, + "loss": 0.1469, + "step": 25279 + }, + { + "epoch": 5.52, + "learning_rate": 2.2438278348263056e-06, + "loss": 0.3629, + "step": 25280 + }, + { + "epoch": 5.52, + "learning_rate": 2.2437185929648242e-06, + "loss": 1.3777, + "step": 25281 + }, + { + "epoch": 5.52, + "learning_rate": 2.243609351103343e-06, + "loss": 1.2045, + "step": 25282 + }, + { + "epoch": 5.52, + "learning_rate": 2.243500109241862e-06, + "loss": 1.747, + "step": 25283 + }, + { + "epoch": 5.52, + "learning_rate": 2.2433908673803805e-06, + "loss": 1.036, + "step": 25284 + }, + { + "epoch": 5.52, + "learning_rate": 2.243281625518899e-06, + "loss": 1.818, + "step": 25285 + }, + { + "epoch": 5.52, + "learning_rate": 2.2431723836574177e-06, + "loss": 0.8168, + "step": 25286 + }, + { + "epoch": 5.52, + "learning_rate": 2.2430631417959363e-06, + "loss": 2.9547, + "step": 25287 + }, + { + "epoch": 5.52, + "learning_rate": 2.242953899934455e-06, + "loss": 1.9411, + "step": 25288 + }, + { + "epoch": 5.52, + "learning_rate": 2.242844658072974e-06, + "loss": 0.2223, + "step": 25289 + }, + { + "epoch": 5.52, + "learning_rate": 2.2427354162114926e-06, + "loss": 3.4073, + "step": 25290 + }, + { + "epoch": 5.52, + "learning_rate": 2.242626174350011e-06, + "loss": 1.3438, + "step": 25291 + }, + { + "epoch": 5.52, + "learning_rate": 2.2425169324885298e-06, + "loss": 2.7284, + "step": 25292 + }, + { + "epoch": 5.52, + "learning_rate": 2.2424076906270484e-06, + "loss": 0.9328, + "step": 25293 + }, + { + "epoch": 5.52, + "learning_rate": 2.242298448765567e-06, + "loss": 1.2585, + "step": 25294 + }, + { + "epoch": 5.52, + "learning_rate": 2.2421892069040856e-06, + "loss": 1.8858, + "step": 25295 + }, + { + "epoch": 5.52, + "learning_rate": 2.2420799650426046e-06, + "loss": 0.5688, + "step": 25296 + }, + { + "epoch": 5.52, + "learning_rate": 2.2419707231811233e-06, + "loss": 1.7903, + "step": 25297 + }, + { + "epoch": 5.52, + "learning_rate": 2.241861481319642e-06, + "loss": 1.4177, + "step": 25298 + }, + { + "epoch": 5.52, + "learning_rate": 2.2417522394581605e-06, + "loss": 2.3918, + "step": 25299 + }, + { + "epoch": 5.52, + "learning_rate": 2.241642997596679e-06, + "loss": 0.8104, + "step": 25300 + }, + { + "epoch": 5.52, + "learning_rate": 2.241533755735198e-06, + "loss": 2.1672, + "step": 25301 + }, + { + "epoch": 5.52, + "learning_rate": 2.2414245138737163e-06, + "loss": 0.8079, + "step": 25302 + }, + { + "epoch": 5.52, + "learning_rate": 2.2413152720122353e-06, + "loss": 0.9038, + "step": 25303 + }, + { + "epoch": 5.52, + "learning_rate": 2.241206030150754e-06, + "loss": 1.13, + "step": 25304 + }, + { + "epoch": 5.52, + "learning_rate": 2.2410967882892726e-06, + "loss": 0.7081, + "step": 25305 + }, + { + "epoch": 5.52, + "learning_rate": 2.240987546427791e-06, + "loss": 0.766, + "step": 25306 + }, + { + "epoch": 5.52, + "learning_rate": 2.24087830456631e-06, + "loss": 2.4615, + "step": 25307 + }, + { + "epoch": 5.52, + "learning_rate": 2.240769062704829e-06, + "loss": 1.8118, + "step": 25308 + }, + { + "epoch": 5.52, + "learning_rate": 2.2406598208433474e-06, + "loss": 1.5395, + "step": 25309 + }, + { + "epoch": 5.52, + "learning_rate": 2.240550578981866e-06, + "loss": 2.0005, + "step": 25310 + }, + { + "epoch": 5.52, + "learning_rate": 2.2404413371203847e-06, + "loss": 1.3206, + "step": 25311 + }, + { + "epoch": 5.52, + "learning_rate": 2.2403320952589033e-06, + "loss": 0.1453, + "step": 25312 + }, + { + "epoch": 5.52, + "learning_rate": 2.2402228533974223e-06, + "loss": 0.3721, + "step": 25313 + }, + { + "epoch": 5.52, + "learning_rate": 2.2401136115359405e-06, + "loss": 3.8433, + "step": 25314 + }, + { + "epoch": 5.52, + "learning_rate": 2.2400043696744595e-06, + "loss": 0.4836, + "step": 25315 + }, + { + "epoch": 5.53, + "learning_rate": 2.239895127812978e-06, + "loss": 1.5234, + "step": 25316 + }, + { + "epoch": 5.53, + "learning_rate": 2.2397858859514967e-06, + "loss": 0.591, + "step": 25317 + }, + { + "epoch": 5.53, + "learning_rate": 2.2396766440900154e-06, + "loss": 0.3361, + "step": 25318 + }, + { + "epoch": 5.53, + "learning_rate": 2.239567402228534e-06, + "loss": 0.907, + "step": 25319 + }, + { + "epoch": 5.53, + "learning_rate": 2.239458160367053e-06, + "loss": 1.4358, + "step": 25320 + }, + { + "epoch": 5.53, + "learning_rate": 2.239348918505571e-06, + "loss": 2.3037, + "step": 25321 + }, + { + "epoch": 5.53, + "learning_rate": 2.2392396766440902e-06, + "loss": 2.1624, + "step": 25322 + }, + { + "epoch": 5.53, + "learning_rate": 2.239130434782609e-06, + "loss": 0.8668, + "step": 25323 + }, + { + "epoch": 5.53, + "learning_rate": 2.2390211929211274e-06, + "loss": 0.6624, + "step": 25324 + }, + { + "epoch": 5.53, + "learning_rate": 2.2389119510596465e-06, + "loss": 1.4585, + "step": 25325 + }, + { + "epoch": 5.53, + "learning_rate": 2.2388027091981647e-06, + "loss": 1.9548, + "step": 25326 + }, + { + "epoch": 5.53, + "learning_rate": 2.2386934673366837e-06, + "loss": 1.4695, + "step": 25327 + }, + { + "epoch": 5.53, + "learning_rate": 2.2385842254752023e-06, + "loss": 1.1133, + "step": 25328 + }, + { + "epoch": 5.53, + "learning_rate": 2.238474983613721e-06, + "loss": 2.8733, + "step": 25329 + }, + { + "epoch": 5.53, + "learning_rate": 2.23836574175224e-06, + "loss": 1.7718, + "step": 25330 + }, + { + "epoch": 5.53, + "learning_rate": 2.238256499890758e-06, + "loss": 1.5267, + "step": 25331 + }, + { + "epoch": 5.53, + "learning_rate": 2.238147258029277e-06, + "loss": 1.9263, + "step": 25332 + }, + { + "epoch": 5.53, + "learning_rate": 2.238038016167796e-06, + "loss": 1.6625, + "step": 25333 + }, + { + "epoch": 5.53, + "learning_rate": 2.2379287743063144e-06, + "loss": 0.6583, + "step": 25334 + }, + { + "epoch": 5.53, + "learning_rate": 2.237819532444833e-06, + "loss": 0.6885, + "step": 25335 + }, + { + "epoch": 5.53, + "learning_rate": 2.2377102905833516e-06, + "loss": 0.4911, + "step": 25336 + }, + { + "epoch": 5.53, + "learning_rate": 2.2376010487218707e-06, + "loss": 0.9195, + "step": 25337 + }, + { + "epoch": 5.53, + "learning_rate": 2.237491806860389e-06, + "loss": 1.0235, + "step": 25338 + }, + { + "epoch": 5.53, + "learning_rate": 2.237382564998908e-06, + "loss": 2.6898, + "step": 25339 + }, + { + "epoch": 5.53, + "learning_rate": 2.2372733231374265e-06, + "loss": 0.8459, + "step": 25340 + }, + { + "epoch": 5.53, + "learning_rate": 2.237164081275945e-06, + "loss": 3.8273, + "step": 25341 + }, + { + "epoch": 5.53, + "learning_rate": 2.237054839414464e-06, + "loss": 1.7165, + "step": 25342 + }, + { + "epoch": 5.53, + "learning_rate": 2.2369455975529823e-06, + "loss": 1.3971, + "step": 25343 + }, + { + "epoch": 5.53, + "learning_rate": 2.2368363556915014e-06, + "loss": 1.8708, + "step": 25344 + }, + { + "epoch": 5.53, + "learning_rate": 2.23672711383002e-06, + "loss": 1.9372, + "step": 25345 + }, + { + "epoch": 5.53, + "learning_rate": 2.2366178719685386e-06, + "loss": 0.9045, + "step": 25346 + }, + { + "epoch": 5.53, + "learning_rate": 2.236508630107057e-06, + "loss": 1.0756, + "step": 25347 + }, + { + "epoch": 5.53, + "learning_rate": 2.236399388245576e-06, + "loss": 3.1305, + "step": 25348 + }, + { + "epoch": 5.53, + "learning_rate": 2.236290146384095e-06, + "loss": 1.109, + "step": 25349 + }, + { + "epoch": 5.53, + "learning_rate": 2.236180904522613e-06, + "loss": 0.9287, + "step": 25350 + }, + { + "epoch": 5.53, + "learning_rate": 2.236071662661132e-06, + "loss": 1.6635, + "step": 25351 + }, + { + "epoch": 5.53, + "learning_rate": 2.2359624207996507e-06, + "loss": 2.9703, + "step": 25352 + }, + { + "epoch": 5.53, + "learning_rate": 2.2358531789381693e-06, + "loss": 2.1379, + "step": 25353 + }, + { + "epoch": 5.53, + "learning_rate": 2.235743937076688e-06, + "loss": 2.0428, + "step": 25354 + }, + { + "epoch": 5.53, + "learning_rate": 2.2356346952152065e-06, + "loss": 2.5355, + "step": 25355 + }, + { + "epoch": 5.53, + "learning_rate": 2.2355254533537255e-06, + "loss": 1.9686, + "step": 25356 + }, + { + "epoch": 5.53, + "learning_rate": 2.235416211492244e-06, + "loss": 1.3299, + "step": 25357 + }, + { + "epoch": 5.53, + "learning_rate": 2.2353069696307628e-06, + "loss": 2.8919, + "step": 25358 + }, + { + "epoch": 5.53, + "learning_rate": 2.2351977277692814e-06, + "loss": 1.2617, + "step": 25359 + }, + { + "epoch": 5.53, + "learning_rate": 2.2350884859078e-06, + "loss": 0.293, + "step": 25360 + }, + { + "epoch": 5.53, + "learning_rate": 2.2349792440463186e-06, + "loss": 0.8588, + "step": 25361 + }, + { + "epoch": 5.54, + "learning_rate": 2.234870002184837e-06, + "loss": 1.433, + "step": 25362 + }, + { + "epoch": 5.54, + "learning_rate": 2.2347607603233562e-06, + "loss": 0.5337, + "step": 25363 + }, + { + "epoch": 5.54, + "learning_rate": 2.234651518461875e-06, + "loss": 1.8352, + "step": 25364 + }, + { + "epoch": 5.54, + "learning_rate": 2.2345422766003935e-06, + "loss": 1.6308, + "step": 25365 + }, + { + "epoch": 5.54, + "learning_rate": 2.234433034738912e-06, + "loss": 1.1943, + "step": 25366 + }, + { + "epoch": 5.54, + "learning_rate": 2.2343237928774307e-06, + "loss": 1.3641, + "step": 25367 + }, + { + "epoch": 5.54, + "learning_rate": 2.2342145510159493e-06, + "loss": 1.9036, + "step": 25368 + }, + { + "epoch": 5.54, + "learning_rate": 2.2341053091544683e-06, + "loss": 1.034, + "step": 25369 + }, + { + "epoch": 5.54, + "learning_rate": 2.233996067292987e-06, + "loss": 1.8348, + "step": 25370 + }, + { + "epoch": 5.54, + "learning_rate": 2.2338868254315056e-06, + "loss": 3.694, + "step": 25371 + }, + { + "epoch": 5.54, + "learning_rate": 2.233777583570024e-06, + "loss": 0.9965, + "step": 25372 + }, + { + "epoch": 5.54, + "learning_rate": 2.2336683417085428e-06, + "loss": 0.6105, + "step": 25373 + }, + { + "epoch": 5.54, + "learning_rate": 2.2335590998470614e-06, + "loss": 3.4445, + "step": 25374 + }, + { + "epoch": 5.54, + "learning_rate": 2.2334498579855804e-06, + "loss": 2.1906, + "step": 25375 + }, + { + "epoch": 5.54, + "learning_rate": 2.233340616124099e-06, + "loss": 0.743, + "step": 25376 + }, + { + "epoch": 5.54, + "learning_rate": 2.2332313742626176e-06, + "loss": 1.8148, + "step": 25377 + }, + { + "epoch": 5.54, + "learning_rate": 2.2331221324011363e-06, + "loss": 1.7976, + "step": 25378 + }, + { + "epoch": 5.54, + "learning_rate": 2.233012890539655e-06, + "loss": 2.6156, + "step": 25379 + }, + { + "epoch": 5.54, + "learning_rate": 2.2329036486781735e-06, + "loss": 1.3758, + "step": 25380 + }, + { + "epoch": 5.54, + "learning_rate": 2.2327944068166925e-06, + "loss": 2.6543, + "step": 25381 + }, + { + "epoch": 5.54, + "learning_rate": 2.232685164955211e-06, + "loss": 1.0441, + "step": 25382 + }, + { + "epoch": 5.54, + "learning_rate": 2.2325759230937297e-06, + "loss": 0.7354, + "step": 25383 + }, + { + "epoch": 5.54, + "learning_rate": 2.2324666812322483e-06, + "loss": 1.7189, + "step": 25384 + }, + { + "epoch": 5.54, + "learning_rate": 2.232357439370767e-06, + "loss": 1.1015, + "step": 25385 + }, + { + "epoch": 5.54, + "learning_rate": 2.232248197509286e-06, + "loss": 0.8251, + "step": 25386 + }, + { + "epoch": 5.54, + "learning_rate": 2.232138955647804e-06, + "loss": 3.5645, + "step": 25387 + }, + { + "epoch": 5.54, + "learning_rate": 2.232029713786323e-06, + "loss": 1.0904, + "step": 25388 + }, + { + "epoch": 5.54, + "learning_rate": 2.231920471924842e-06, + "loss": 0.5701, + "step": 25389 + }, + { + "epoch": 5.54, + "learning_rate": 2.2318112300633604e-06, + "loss": 2.0648, + "step": 25390 + }, + { + "epoch": 5.54, + "learning_rate": 2.231701988201879e-06, + "loss": 1.694, + "step": 25391 + }, + { + "epoch": 5.54, + "learning_rate": 2.2315927463403977e-06, + "loss": 1.7593, + "step": 25392 + }, + { + "epoch": 5.54, + "learning_rate": 2.2314835044789167e-06, + "loss": 1.4284, + "step": 25393 + }, + { + "epoch": 5.54, + "learning_rate": 2.231374262617435e-06, + "loss": 1.6552, + "step": 25394 + }, + { + "epoch": 5.54, + "learning_rate": 2.231265020755954e-06, + "loss": 0.9759, + "step": 25395 + }, + { + "epoch": 5.54, + "learning_rate": 2.2311557788944725e-06, + "loss": 2.4982, + "step": 25396 + }, + { + "epoch": 5.54, + "learning_rate": 2.231046537032991e-06, + "loss": 0.7901, + "step": 25397 + }, + { + "epoch": 5.54, + "learning_rate": 2.23093729517151e-06, + "loss": 1.5024, + "step": 25398 + }, + { + "epoch": 5.54, + "learning_rate": 2.2308280533100284e-06, + "loss": 0.4211, + "step": 25399 + }, + { + "epoch": 5.54, + "learning_rate": 2.2307188114485474e-06, + "loss": 1.1074, + "step": 25400 + }, + { + "epoch": 5.54, + "learning_rate": 2.230609569587066e-06, + "loss": 2.6644, + "step": 25401 + }, + { + "epoch": 5.54, + "learning_rate": 2.2305003277255846e-06, + "loss": 1.6901, + "step": 25402 + }, + { + "epoch": 5.54, + "learning_rate": 2.2303910858641032e-06, + "loss": 0.5228, + "step": 25403 + }, + { + "epoch": 5.54, + "learning_rate": 2.230281844002622e-06, + "loss": 1.9625, + "step": 25404 + }, + { + "epoch": 5.54, + "learning_rate": 2.230172602141141e-06, + "loss": 1.6128, + "step": 25405 + }, + { + "epoch": 5.54, + "learning_rate": 2.230063360279659e-06, + "loss": 0.6079, + "step": 25406 + }, + { + "epoch": 5.54, + "learning_rate": 2.229954118418178e-06, + "loss": 1.1148, + "step": 25407 + }, + { + "epoch": 5.55, + "learning_rate": 2.2298448765566967e-06, + "loss": 0.5514, + "step": 25408 + }, + { + "epoch": 5.55, + "learning_rate": 2.2297356346952153e-06, + "loss": 1.2648, + "step": 25409 + }, + { + "epoch": 5.55, + "learning_rate": 2.2296263928337343e-06, + "loss": 1.9776, + "step": 25410 + }, + { + "epoch": 5.55, + "learning_rate": 2.2295171509722525e-06, + "loss": 1.7463, + "step": 25411 + }, + { + "epoch": 5.55, + "learning_rate": 2.2294079091107716e-06, + "loss": 2.4373, + "step": 25412 + }, + { + "epoch": 5.55, + "learning_rate": 2.22929866724929e-06, + "loss": 1.7405, + "step": 25413 + }, + { + "epoch": 5.55, + "learning_rate": 2.229189425387809e-06, + "loss": 0.9081, + "step": 25414 + }, + { + "epoch": 5.55, + "learning_rate": 2.2290801835263274e-06, + "loss": 1.6541, + "step": 25415 + }, + { + "epoch": 5.55, + "learning_rate": 2.228970941664846e-06, + "loss": 1.177, + "step": 25416 + }, + { + "epoch": 5.55, + "learning_rate": 2.228861699803365e-06, + "loss": 1.8998, + "step": 25417 + }, + { + "epoch": 5.55, + "learning_rate": 2.2287524579418832e-06, + "loss": 1.3087, + "step": 25418 + }, + { + "epoch": 5.55, + "learning_rate": 2.2286432160804023e-06, + "loss": 0.8654, + "step": 25419 + }, + { + "epoch": 5.55, + "learning_rate": 2.228533974218921e-06, + "loss": 0.8853, + "step": 25420 + }, + { + "epoch": 5.55, + "learning_rate": 2.2284247323574395e-06, + "loss": 2.0844, + "step": 25421 + }, + { + "epoch": 5.55, + "learning_rate": 2.2283154904959585e-06, + "loss": 1.6888, + "step": 25422 + }, + { + "epoch": 5.55, + "learning_rate": 2.2282062486344767e-06, + "loss": 1.3184, + "step": 25423 + }, + { + "epoch": 5.55, + "learning_rate": 2.2280970067729957e-06, + "loss": 1.5364, + "step": 25424 + }, + { + "epoch": 5.55, + "learning_rate": 2.2279877649115144e-06, + "loss": 1.3035, + "step": 25425 + }, + { + "epoch": 5.55, + "learning_rate": 2.227878523050033e-06, + "loss": 2.3365, + "step": 25426 + }, + { + "epoch": 5.55, + "learning_rate": 2.2277692811885516e-06, + "loss": 2.4308, + "step": 25427 + }, + { + "epoch": 5.55, + "learning_rate": 2.22766003932707e-06, + "loss": 1.9449, + "step": 25428 + }, + { + "epoch": 5.55, + "learning_rate": 2.2275507974655892e-06, + "loss": 1.095, + "step": 25429 + }, + { + "epoch": 5.55, + "learning_rate": 2.227441555604108e-06, + "loss": 1.5886, + "step": 25430 + }, + { + "epoch": 5.55, + "learning_rate": 2.2273323137426264e-06, + "loss": 1.6035, + "step": 25431 + }, + { + "epoch": 5.55, + "learning_rate": 2.227223071881145e-06, + "loss": 3.0184, + "step": 25432 + }, + { + "epoch": 5.55, + "learning_rate": 2.2271138300196637e-06, + "loss": 2.566, + "step": 25433 + }, + { + "epoch": 5.55, + "learning_rate": 2.2270045881581823e-06, + "loss": 1.0575, + "step": 25434 + }, + { + "epoch": 5.55, + "learning_rate": 2.226895346296701e-06, + "loss": 2.7529, + "step": 25435 + }, + { + "epoch": 5.55, + "learning_rate": 2.22678610443522e-06, + "loss": 0.5084, + "step": 25436 + }, + { + "epoch": 5.55, + "learning_rate": 2.2266768625737385e-06, + "loss": 1.7719, + "step": 25437 + }, + { + "epoch": 5.55, + "learning_rate": 2.226567620712257e-06, + "loss": 1.5157, + "step": 25438 + }, + { + "epoch": 5.55, + "learning_rate": 2.2264583788507758e-06, + "loss": 1.5019, + "step": 25439 + }, + { + "epoch": 5.55, + "learning_rate": 2.2263491369892944e-06, + "loss": 0.1782, + "step": 25440 + }, + { + "epoch": 5.55, + "learning_rate": 2.226239895127813e-06, + "loss": 1.7902, + "step": 25441 + }, + { + "epoch": 5.55, + "learning_rate": 2.226130653266332e-06, + "loss": 1.1945, + "step": 25442 + }, + { + "epoch": 5.55, + "learning_rate": 2.2260214114048506e-06, + "loss": 2.0248, + "step": 25443 + }, + { + "epoch": 5.55, + "learning_rate": 2.2259121695433692e-06, + "loss": 0.7446, + "step": 25444 + }, + { + "epoch": 5.55, + "learning_rate": 2.225802927681888e-06, + "loss": 0.5025, + "step": 25445 + }, + { + "epoch": 5.55, + "learning_rate": 2.2256936858204065e-06, + "loss": 2.3951, + "step": 25446 + }, + { + "epoch": 5.55, + "learning_rate": 2.225584443958925e-06, + "loss": 0.3932, + "step": 25447 + }, + { + "epoch": 5.55, + "learning_rate": 2.225475202097444e-06, + "loss": 1.9478, + "step": 25448 + }, + { + "epoch": 5.55, + "learning_rate": 2.2253659602359627e-06, + "loss": 2.486, + "step": 25449 + }, + { + "epoch": 5.55, + "learning_rate": 2.2252567183744813e-06, + "loss": 1.0339, + "step": 25450 + }, + { + "epoch": 5.55, + "learning_rate": 2.225147476513e-06, + "loss": 1.6088, + "step": 25451 + }, + { + "epoch": 5.55, + "learning_rate": 2.2250382346515185e-06, + "loss": 2.0339, + "step": 25452 + }, + { + "epoch": 5.55, + "learning_rate": 2.224928992790037e-06, + "loss": 1.5242, + "step": 25453 + }, + { + "epoch": 5.56, + "learning_rate": 2.224819750928556e-06, + "loss": 1.6396, + "step": 25454 + }, + { + "epoch": 5.56, + "learning_rate": 2.224710509067075e-06, + "loss": 1.4612, + "step": 25455 + }, + { + "epoch": 5.56, + "learning_rate": 2.2246012672055934e-06, + "loss": 0.9339, + "step": 25456 + }, + { + "epoch": 5.56, + "learning_rate": 2.224492025344112e-06, + "loss": 2.6549, + "step": 25457 + }, + { + "epoch": 5.56, + "learning_rate": 2.2243827834826306e-06, + "loss": 1.7775, + "step": 25458 + }, + { + "epoch": 5.56, + "learning_rate": 2.2242735416211493e-06, + "loss": 4.3464, + "step": 25459 + }, + { + "epoch": 5.56, + "learning_rate": 2.224164299759668e-06, + "loss": 2.5996, + "step": 25460 + }, + { + "epoch": 5.56, + "learning_rate": 2.224055057898187e-06, + "loss": 1.5991, + "step": 25461 + }, + { + "epoch": 5.56, + "learning_rate": 2.2239458160367055e-06, + "loss": 0.3842, + "step": 25462 + }, + { + "epoch": 5.56, + "learning_rate": 2.223836574175224e-06, + "loss": 2.5636, + "step": 25463 + }, + { + "epoch": 5.56, + "learning_rate": 2.2237273323137427e-06, + "loss": 1.3042, + "step": 25464 + }, + { + "epoch": 5.56, + "learning_rate": 2.2236180904522613e-06, + "loss": 3.2234, + "step": 25465 + }, + { + "epoch": 5.56, + "learning_rate": 2.2235088485907804e-06, + "loss": 1.1709, + "step": 25466 + }, + { + "epoch": 5.56, + "learning_rate": 2.2233996067292986e-06, + "loss": 1.4379, + "step": 25467 + }, + { + "epoch": 5.56, + "learning_rate": 2.2232903648678176e-06, + "loss": 0.9796, + "step": 25468 + }, + { + "epoch": 5.56, + "learning_rate": 2.223181123006336e-06, + "loss": 1.658, + "step": 25469 + }, + { + "epoch": 5.56, + "learning_rate": 2.223071881144855e-06, + "loss": 0.7873, + "step": 25470 + }, + { + "epoch": 5.56, + "learning_rate": 2.2229626392833734e-06, + "loss": 0.9717, + "step": 25471 + }, + { + "epoch": 5.56, + "learning_rate": 2.222853397421892e-06, + "loss": 1.3388, + "step": 25472 + }, + { + "epoch": 5.56, + "learning_rate": 2.222744155560411e-06, + "loss": 1.7948, + "step": 25473 + }, + { + "epoch": 5.56, + "learning_rate": 2.2226349136989297e-06, + "loss": 1.639, + "step": 25474 + }, + { + "epoch": 5.56, + "learning_rate": 2.2225256718374483e-06, + "loss": 1.2025, + "step": 25475 + }, + { + "epoch": 5.56, + "learning_rate": 2.222416429975967e-06, + "loss": 0.8209, + "step": 25476 + }, + { + "epoch": 5.56, + "learning_rate": 2.2223071881144855e-06, + "loss": 2.0202, + "step": 25477 + }, + { + "epoch": 5.56, + "learning_rate": 2.2221979462530046e-06, + "loss": 2.2676, + "step": 25478 + }, + { + "epoch": 5.56, + "learning_rate": 2.2220887043915227e-06, + "loss": 1.7558, + "step": 25479 + }, + { + "epoch": 5.56, + "learning_rate": 2.2219794625300418e-06, + "loss": 1.3076, + "step": 25480 + }, + { + "epoch": 5.56, + "learning_rate": 2.2218702206685604e-06, + "loss": 1.8897, + "step": 25481 + }, + { + "epoch": 5.56, + "learning_rate": 2.221760978807079e-06, + "loss": 0.7151, + "step": 25482 + }, + { + "epoch": 5.56, + "learning_rate": 2.221651736945598e-06, + "loss": 1.4997, + "step": 25483 + }, + { + "epoch": 5.56, + "learning_rate": 2.2215424950841162e-06, + "loss": 1.2663, + "step": 25484 + }, + { + "epoch": 5.56, + "learning_rate": 2.2214332532226353e-06, + "loss": 2.8871, + "step": 25485 + }, + { + "epoch": 5.56, + "learning_rate": 2.221324011361154e-06, + "loss": 3.4716, + "step": 25486 + }, + { + "epoch": 5.56, + "learning_rate": 2.2212147694996725e-06, + "loss": 1.0518, + "step": 25487 + }, + { + "epoch": 5.56, + "learning_rate": 2.221105527638191e-06, + "loss": 1.9871, + "step": 25488 + }, + { + "epoch": 5.56, + "learning_rate": 2.2209962857767097e-06, + "loss": 1.7646, + "step": 25489 + }, + { + "epoch": 5.56, + "learning_rate": 2.2208870439152287e-06, + "loss": 0.6139, + "step": 25490 + }, + { + "epoch": 5.56, + "learning_rate": 2.220777802053747e-06, + "loss": 0.5364, + "step": 25491 + }, + { + "epoch": 5.56, + "learning_rate": 2.220668560192266e-06, + "loss": 1.4117, + "step": 25492 + }, + { + "epoch": 5.56, + "learning_rate": 2.2205593183307846e-06, + "loss": 0.8834, + "step": 25493 + }, + { + "epoch": 5.56, + "learning_rate": 2.220450076469303e-06, + "loss": 3.1873, + "step": 25494 + }, + { + "epoch": 5.56, + "learning_rate": 2.220340834607822e-06, + "loss": 2.444, + "step": 25495 + }, + { + "epoch": 5.56, + "learning_rate": 2.2202315927463404e-06, + "loss": 1.875, + "step": 25496 + }, + { + "epoch": 5.56, + "learning_rate": 2.2201223508848594e-06, + "loss": 1.3667, + "step": 25497 + }, + { + "epoch": 5.56, + "learning_rate": 2.220013109023378e-06, + "loss": 3.3091, + "step": 25498 + }, + { + "epoch": 5.57, + "learning_rate": 2.2199038671618967e-06, + "loss": 1.182, + "step": 25499 + }, + { + "epoch": 5.57, + "learning_rate": 2.2197946253004153e-06, + "loss": 2.9873, + "step": 25500 + }, + { + "epoch": 5.57, + "learning_rate": 2.219685383438934e-06, + "loss": 0.9504, + "step": 25501 + }, + { + "epoch": 5.57, + "learning_rate": 2.219576141577453e-06, + "loss": 0.1908, + "step": 25502 + }, + { + "epoch": 5.57, + "learning_rate": 2.219466899715971e-06, + "loss": 0.4704, + "step": 25503 + }, + { + "epoch": 5.57, + "learning_rate": 2.21935765785449e-06, + "loss": 2.1884, + "step": 25504 + }, + { + "epoch": 5.57, + "learning_rate": 2.2192484159930087e-06, + "loss": 1.2404, + "step": 25505 + }, + { + "epoch": 5.57, + "learning_rate": 2.2191391741315274e-06, + "loss": 1.1893, + "step": 25506 + }, + { + "epoch": 5.57, + "learning_rate": 2.219029932270046e-06, + "loss": 1.642, + "step": 25507 + }, + { + "epoch": 5.57, + "learning_rate": 2.2189206904085646e-06, + "loss": 1.3424, + "step": 25508 + }, + { + "epoch": 5.57, + "learning_rate": 2.2188114485470836e-06, + "loss": 1.5788, + "step": 25509 + }, + { + "epoch": 5.57, + "learning_rate": 2.2187022066856022e-06, + "loss": 0.9017, + "step": 25510 + }, + { + "epoch": 5.57, + "learning_rate": 2.218592964824121e-06, + "loss": 0.6963, + "step": 25511 + }, + { + "epoch": 5.57, + "learning_rate": 2.2184837229626394e-06, + "loss": 3.0369, + "step": 25512 + }, + { + "epoch": 5.57, + "learning_rate": 2.218374481101158e-06, + "loss": 0.6498, + "step": 25513 + }, + { + "epoch": 5.57, + "learning_rate": 2.218265239239677e-06, + "loss": 1.4221, + "step": 25514 + }, + { + "epoch": 5.57, + "learning_rate": 2.2181559973781953e-06, + "loss": 1.2287, + "step": 25515 + }, + { + "epoch": 5.57, + "learning_rate": 2.2180467555167143e-06, + "loss": 1.1952, + "step": 25516 + }, + { + "epoch": 5.57, + "learning_rate": 2.217937513655233e-06, + "loss": 1.4173, + "step": 25517 + }, + { + "epoch": 5.57, + "learning_rate": 2.2178282717937515e-06, + "loss": 0.8885, + "step": 25518 + }, + { + "epoch": 5.57, + "learning_rate": 2.21771902993227e-06, + "loss": 1.6737, + "step": 25519 + }, + { + "epoch": 5.57, + "learning_rate": 2.2176097880707888e-06, + "loss": 1.5397, + "step": 25520 + }, + { + "epoch": 5.57, + "learning_rate": 2.217500546209308e-06, + "loss": 1.7969, + "step": 25521 + }, + { + "epoch": 5.57, + "learning_rate": 2.2173913043478264e-06, + "loss": 2.0197, + "step": 25522 + }, + { + "epoch": 5.57, + "learning_rate": 2.217282062486345e-06, + "loss": 0.5928, + "step": 25523 + }, + { + "epoch": 5.57, + "learning_rate": 2.2171728206248636e-06, + "loss": 3.7262, + "step": 25524 + }, + { + "epoch": 5.57, + "learning_rate": 2.2170635787633822e-06, + "loss": 0.9637, + "step": 25525 + }, + { + "epoch": 5.57, + "learning_rate": 2.216954336901901e-06, + "loss": 1.042, + "step": 25526 + }, + { + "epoch": 5.57, + "learning_rate": 2.21684509504042e-06, + "loss": 1.3454, + "step": 25527 + }, + { + "epoch": 5.57, + "learning_rate": 2.2167358531789385e-06, + "loss": 1.8737, + "step": 25528 + }, + { + "epoch": 5.57, + "learning_rate": 2.216626611317457e-06, + "loss": 1.6821, + "step": 25529 + }, + { + "epoch": 5.57, + "learning_rate": 2.2165173694559757e-06, + "loss": 0.3849, + "step": 25530 + }, + { + "epoch": 5.57, + "learning_rate": 2.2164081275944943e-06, + "loss": 1.7604, + "step": 25531 + }, + { + "epoch": 5.57, + "learning_rate": 2.216298885733013e-06, + "loss": 0.7775, + "step": 25532 + }, + { + "epoch": 5.57, + "learning_rate": 2.2161896438715315e-06, + "loss": 1.9552, + "step": 25533 + }, + { + "epoch": 5.57, + "learning_rate": 2.2160804020100506e-06, + "loss": 4.0376, + "step": 25534 + }, + { + "epoch": 5.57, + "learning_rate": 2.215971160148569e-06, + "loss": 0.5315, + "step": 25535 + }, + { + "epoch": 5.57, + "learning_rate": 2.215861918287088e-06, + "loss": 1.61, + "step": 25536 + }, + { + "epoch": 5.57, + "learning_rate": 2.2157526764256064e-06, + "loss": 1.8887, + "step": 25537 + }, + { + "epoch": 5.57, + "learning_rate": 2.215643434564125e-06, + "loss": 0.7665, + "step": 25538 + }, + { + "epoch": 5.57, + "learning_rate": 2.215534192702644e-06, + "loss": 1.9272, + "step": 25539 + }, + { + "epoch": 5.57, + "learning_rate": 2.2154249508411622e-06, + "loss": 1.3899, + "step": 25540 + }, + { + "epoch": 5.57, + "learning_rate": 2.2153157089796813e-06, + "loss": 1.1421, + "step": 25541 + }, + { + "epoch": 5.57, + "learning_rate": 2.2152064671182e-06, + "loss": 0.9132, + "step": 25542 + }, + { + "epoch": 5.57, + "learning_rate": 2.2150972252567185e-06, + "loss": 0.2694, + "step": 25543 + }, + { + "epoch": 5.57, + "learning_rate": 2.214987983395237e-06, + "loss": 1.726, + "step": 25544 + }, + { + "epoch": 5.58, + "learning_rate": 2.2148787415337557e-06, + "loss": 2.074, + "step": 25545 + }, + { + "epoch": 5.58, + "learning_rate": 2.2147694996722748e-06, + "loss": 1.9214, + "step": 25546 + }, + { + "epoch": 5.58, + "learning_rate": 2.2146602578107934e-06, + "loss": 2.2979, + "step": 25547 + }, + { + "epoch": 5.58, + "learning_rate": 2.214551015949312e-06, + "loss": 1.5289, + "step": 25548 + }, + { + "epoch": 5.58, + "learning_rate": 2.2144417740878306e-06, + "loss": 2.6814, + "step": 25549 + }, + { + "epoch": 5.58, + "learning_rate": 2.214332532226349e-06, + "loss": 1.1248, + "step": 25550 + }, + { + "epoch": 5.58, + "learning_rate": 2.2142232903648682e-06, + "loss": 1.0419, + "step": 25551 + }, + { + "epoch": 5.58, + "learning_rate": 2.2141140485033864e-06, + "loss": 0.9007, + "step": 25552 + }, + { + "epoch": 5.58, + "learning_rate": 2.2140048066419055e-06, + "loss": 0.9741, + "step": 25553 + }, + { + "epoch": 5.58, + "learning_rate": 2.213895564780424e-06, + "loss": 0.6443, + "step": 25554 + }, + { + "epoch": 5.58, + "learning_rate": 2.2137863229189427e-06, + "loss": 2.4008, + "step": 25555 + }, + { + "epoch": 5.58, + "learning_rate": 2.2136770810574613e-06, + "loss": 1.0784, + "step": 25556 + }, + { + "epoch": 5.58, + "learning_rate": 2.21356783919598e-06, + "loss": 1.1598, + "step": 25557 + }, + { + "epoch": 5.58, + "learning_rate": 2.213458597334499e-06, + "loss": 1.4299, + "step": 25558 + }, + { + "epoch": 5.58, + "learning_rate": 2.213349355473017e-06, + "loss": 1.6426, + "step": 25559 + }, + { + "epoch": 5.58, + "learning_rate": 2.213240113611536e-06, + "loss": 1.1372, + "step": 25560 + }, + { + "epoch": 5.58, + "learning_rate": 2.2131308717500548e-06, + "loss": 1.2534, + "step": 25561 + }, + { + "epoch": 5.58, + "learning_rate": 2.2130216298885734e-06, + "loss": 1.0187, + "step": 25562 + }, + { + "epoch": 5.58, + "learning_rate": 2.2129123880270924e-06, + "loss": 0.5054, + "step": 25563 + }, + { + "epoch": 5.58, + "learning_rate": 2.2128031461656106e-06, + "loss": 1.3046, + "step": 25564 + }, + { + "epoch": 5.58, + "learning_rate": 2.2126939043041296e-06, + "loss": 2.6105, + "step": 25565 + }, + { + "epoch": 5.58, + "learning_rate": 2.2125846624426482e-06, + "loss": 1.82, + "step": 25566 + }, + { + "epoch": 5.58, + "learning_rate": 2.212475420581167e-06, + "loss": 1.9191, + "step": 25567 + }, + { + "epoch": 5.58, + "learning_rate": 2.2123661787196855e-06, + "loss": 1.24, + "step": 25568 + }, + { + "epoch": 5.58, + "learning_rate": 2.212256936858204e-06, + "loss": 1.1498, + "step": 25569 + }, + { + "epoch": 5.58, + "learning_rate": 2.212147694996723e-06, + "loss": 1.3244, + "step": 25570 + }, + { + "epoch": 5.58, + "learning_rate": 2.2120384531352413e-06, + "loss": 0.7428, + "step": 25571 + }, + { + "epoch": 5.58, + "learning_rate": 2.2119292112737603e-06, + "loss": 0.3542, + "step": 25572 + }, + { + "epoch": 5.58, + "learning_rate": 2.211819969412279e-06, + "loss": 2.2291, + "step": 25573 + }, + { + "epoch": 5.58, + "learning_rate": 2.2117107275507976e-06, + "loss": 2.1517, + "step": 25574 + }, + { + "epoch": 5.58, + "learning_rate": 2.2116014856893166e-06, + "loss": 2.1682, + "step": 25575 + }, + { + "epoch": 5.58, + "learning_rate": 2.2114922438278348e-06, + "loss": 1.7294, + "step": 25576 + }, + { + "epoch": 5.58, + "learning_rate": 2.211383001966354e-06, + "loss": 2.9499, + "step": 25577 + }, + { + "epoch": 5.58, + "learning_rate": 2.2112737601048724e-06, + "loss": 0.3555, + "step": 25578 + }, + { + "epoch": 5.58, + "learning_rate": 2.211164518243391e-06, + "loss": 0.258, + "step": 25579 + }, + { + "epoch": 5.58, + "learning_rate": 2.21105527638191e-06, + "loss": 3.4327, + "step": 25580 + }, + { + "epoch": 5.58, + "learning_rate": 2.2109460345204283e-06, + "loss": 1.0082, + "step": 25581 + }, + { + "epoch": 5.58, + "learning_rate": 2.2108367926589473e-06, + "loss": 1.8659, + "step": 25582 + }, + { + "epoch": 5.58, + "learning_rate": 2.210727550797466e-06, + "loss": 2.894, + "step": 25583 + }, + { + "epoch": 5.58, + "learning_rate": 2.2106183089359845e-06, + "loss": 1.1761, + "step": 25584 + }, + { + "epoch": 5.58, + "learning_rate": 2.210509067074503e-06, + "loss": 2.086, + "step": 25585 + }, + { + "epoch": 5.58, + "learning_rate": 2.2103998252130217e-06, + "loss": 1.6387, + "step": 25586 + }, + { + "epoch": 5.58, + "learning_rate": 2.2102905833515408e-06, + "loss": 1.5053, + "step": 25587 + }, + { + "epoch": 5.58, + "learning_rate": 2.210181341490059e-06, + "loss": 2.4702, + "step": 25588 + }, + { + "epoch": 5.58, + "learning_rate": 2.210072099628578e-06, + "loss": 1.0972, + "step": 25589 + }, + { + "epoch": 5.58, + "learning_rate": 2.2099628577670966e-06, + "loss": 0.6622, + "step": 25590 + }, + { + "epoch": 5.59, + "learning_rate": 2.2098536159056152e-06, + "loss": 1.4168, + "step": 25591 + }, + { + "epoch": 5.59, + "learning_rate": 2.209744374044134e-06, + "loss": 0.9162, + "step": 25592 + }, + { + "epoch": 5.59, + "learning_rate": 2.2096351321826524e-06, + "loss": 2.2161, + "step": 25593 + }, + { + "epoch": 5.59, + "learning_rate": 2.2095258903211715e-06, + "loss": 2.4009, + "step": 25594 + }, + { + "epoch": 5.59, + "learning_rate": 2.20941664845969e-06, + "loss": 0.925, + "step": 25595 + }, + { + "epoch": 5.59, + "learning_rate": 2.2093074065982087e-06, + "loss": 0.401, + "step": 25596 + }, + { + "epoch": 5.59, + "learning_rate": 2.2091981647367273e-06, + "loss": 1.646, + "step": 25597 + }, + { + "epoch": 5.59, + "learning_rate": 2.209088922875246e-06, + "loss": 1.2767, + "step": 25598 + }, + { + "epoch": 5.59, + "learning_rate": 2.2089796810137645e-06, + "loss": 1.0187, + "step": 25599 + }, + { + "epoch": 5.59, + "learning_rate": 2.208870439152283e-06, + "loss": 0.6617, + "step": 25600 + }, + { + "epoch": 5.59, + "learning_rate": 2.208761197290802e-06, + "loss": 0.2522, + "step": 25601 + }, + { + "epoch": 5.59, + "learning_rate": 2.2086519554293208e-06, + "loss": 0.5876, + "step": 25602 + }, + { + "epoch": 5.59, + "learning_rate": 2.2085427135678394e-06, + "loss": 1.5027, + "step": 25603 + }, + { + "epoch": 5.59, + "learning_rate": 2.208433471706358e-06, + "loss": 1.7794, + "step": 25604 + }, + { + "epoch": 5.59, + "learning_rate": 2.2083242298448766e-06, + "loss": 3.3188, + "step": 25605 + }, + { + "epoch": 5.59, + "learning_rate": 2.2082149879833952e-06, + "loss": 2.5562, + "step": 25606 + }, + { + "epoch": 5.59, + "learning_rate": 2.2081057461219143e-06, + "loss": 1.6011, + "step": 25607 + }, + { + "epoch": 5.59, + "learning_rate": 2.207996504260433e-06, + "loss": 0.6316, + "step": 25608 + }, + { + "epoch": 5.59, + "learning_rate": 2.2078872623989515e-06, + "loss": 0.9175, + "step": 25609 + }, + { + "epoch": 5.59, + "learning_rate": 2.20777802053747e-06, + "loss": 2.1514, + "step": 25610 + }, + { + "epoch": 5.59, + "learning_rate": 2.2076687786759887e-06, + "loss": 0.9183, + "step": 25611 + }, + { + "epoch": 5.59, + "learning_rate": 2.2075595368145073e-06, + "loss": 0.7464, + "step": 25612 + }, + { + "epoch": 5.59, + "learning_rate": 2.2074502949530264e-06, + "loss": 1.3269, + "step": 25613 + }, + { + "epoch": 5.59, + "learning_rate": 2.207341053091545e-06, + "loss": 0.8328, + "step": 25614 + }, + { + "epoch": 5.59, + "learning_rate": 2.2072318112300636e-06, + "loss": 1.6976, + "step": 25615 + }, + { + "epoch": 5.59, + "learning_rate": 2.207122569368582e-06, + "loss": 1.8343, + "step": 25616 + }, + { + "epoch": 5.59, + "learning_rate": 2.207013327507101e-06, + "loss": 0.4735, + "step": 25617 + }, + { + "epoch": 5.59, + "learning_rate": 2.2069040856456194e-06, + "loss": 2.9658, + "step": 25618 + }, + { + "epoch": 5.59, + "learning_rate": 2.2067948437841384e-06, + "loss": 1.7067, + "step": 25619 + }, + { + "epoch": 5.59, + "learning_rate": 2.206685601922657e-06, + "loss": 0.6878, + "step": 25620 + }, + { + "epoch": 5.59, + "learning_rate": 2.2065763600611757e-06, + "loss": 1.2928, + "step": 25621 + }, + { + "epoch": 5.59, + "learning_rate": 2.2064671181996943e-06, + "loss": 1.0026, + "step": 25622 + }, + { + "epoch": 5.59, + "learning_rate": 2.206357876338213e-06, + "loss": 0.9331, + "step": 25623 + }, + { + "epoch": 5.59, + "learning_rate": 2.206248634476732e-06, + "loss": 1.1345, + "step": 25624 + }, + { + "epoch": 5.59, + "learning_rate": 2.20613939261525e-06, + "loss": 2.2903, + "step": 25625 + }, + { + "epoch": 5.59, + "learning_rate": 2.206030150753769e-06, + "loss": 0.5965, + "step": 25626 + }, + { + "epoch": 5.59, + "learning_rate": 2.2059209088922878e-06, + "loss": 2.2438, + "step": 25627 + }, + { + "epoch": 5.59, + "learning_rate": 2.2058116670308064e-06, + "loss": 0.28, + "step": 25628 + }, + { + "epoch": 5.59, + "learning_rate": 2.205702425169325e-06, + "loss": 1.4529, + "step": 25629 + }, + { + "epoch": 5.59, + "learning_rate": 2.2055931833078436e-06, + "loss": 0.8062, + "step": 25630 + }, + { + "epoch": 5.59, + "learning_rate": 2.2054839414463626e-06, + "loss": 1.9526, + "step": 25631 + }, + { + "epoch": 5.59, + "learning_rate": 2.205374699584881e-06, + "loss": 2.355, + "step": 25632 + }, + { + "epoch": 5.59, + "learning_rate": 2.2052654577234e-06, + "loss": 1.9935, + "step": 25633 + }, + { + "epoch": 5.59, + "learning_rate": 2.2051562158619185e-06, + "loss": 1.867, + "step": 25634 + }, + { + "epoch": 5.59, + "learning_rate": 2.205046974000437e-06, + "loss": 1.9794, + "step": 25635 + }, + { + "epoch": 5.59, + "learning_rate": 2.204937732138956e-06, + "loss": 0.3964, + "step": 25636 + }, + { + "epoch": 5.6, + "learning_rate": 2.2048284902774743e-06, + "loss": 2.0068, + "step": 25637 + }, + { + "epoch": 5.6, + "learning_rate": 2.2047192484159933e-06, + "loss": 2.0486, + "step": 25638 + }, + { + "epoch": 5.6, + "learning_rate": 2.204610006554512e-06, + "loss": 1.4831, + "step": 25639 + }, + { + "epoch": 5.6, + "learning_rate": 2.2045007646930305e-06, + "loss": 1.9937, + "step": 25640 + }, + { + "epoch": 5.6, + "learning_rate": 2.204391522831549e-06, + "loss": 1.4206, + "step": 25641 + }, + { + "epoch": 5.6, + "learning_rate": 2.2042822809700678e-06, + "loss": 0.6884, + "step": 25642 + }, + { + "epoch": 5.6, + "learning_rate": 2.204173039108587e-06, + "loss": 2.3159, + "step": 25643 + }, + { + "epoch": 5.6, + "learning_rate": 2.204063797247105e-06, + "loss": 1.2436, + "step": 25644 + }, + { + "epoch": 5.6, + "learning_rate": 2.203954555385624e-06, + "loss": 3.5145, + "step": 25645 + }, + { + "epoch": 5.6, + "learning_rate": 2.2038453135241426e-06, + "loss": 1.1766, + "step": 25646 + }, + { + "epoch": 5.6, + "learning_rate": 2.2037360716626612e-06, + "loss": 3.04, + "step": 25647 + }, + { + "epoch": 5.6, + "learning_rate": 2.2036268298011803e-06, + "loss": 0.5942, + "step": 25648 + }, + { + "epoch": 5.6, + "learning_rate": 2.2035175879396985e-06, + "loss": 0.1769, + "step": 25649 + }, + { + "epoch": 5.6, + "learning_rate": 2.2034083460782175e-06, + "loss": 1.7077, + "step": 25650 + }, + { + "epoch": 5.6, + "learning_rate": 2.203299104216736e-06, + "loss": 1.591, + "step": 25651 + }, + { + "epoch": 5.6, + "learning_rate": 2.2031898623552547e-06, + "loss": 2.4245, + "step": 25652 + }, + { + "epoch": 5.6, + "learning_rate": 2.2030806204937733e-06, + "loss": 2.4296, + "step": 25653 + }, + { + "epoch": 5.6, + "learning_rate": 2.202971378632292e-06, + "loss": 0.116, + "step": 25654 + }, + { + "epoch": 5.6, + "learning_rate": 2.202862136770811e-06, + "loss": 3.1076, + "step": 25655 + }, + { + "epoch": 5.6, + "learning_rate": 2.202752894909329e-06, + "loss": 1.4589, + "step": 25656 + }, + { + "epoch": 5.6, + "learning_rate": 2.202643653047848e-06, + "loss": 2.0228, + "step": 25657 + }, + { + "epoch": 5.6, + "learning_rate": 2.202534411186367e-06, + "loss": 1.0074, + "step": 25658 + }, + { + "epoch": 5.6, + "learning_rate": 2.2024251693248854e-06, + "loss": 0.7307, + "step": 25659 + }, + { + "epoch": 5.6, + "learning_rate": 2.2023159274634045e-06, + "loss": 2.0156, + "step": 25660 + }, + { + "epoch": 5.6, + "learning_rate": 2.2022066856019226e-06, + "loss": 0.1195, + "step": 25661 + }, + { + "epoch": 5.6, + "learning_rate": 2.2020974437404417e-06, + "loss": 1.8855, + "step": 25662 + }, + { + "epoch": 5.6, + "learning_rate": 2.2019882018789603e-06, + "loss": 1.403, + "step": 25663 + }, + { + "epoch": 5.6, + "learning_rate": 2.201878960017479e-06, + "loss": 0.8445, + "step": 25664 + }, + { + "epoch": 5.6, + "learning_rate": 2.2017697181559975e-06, + "loss": 1.518, + "step": 25665 + }, + { + "epoch": 5.6, + "learning_rate": 2.201660476294516e-06, + "loss": 1.615, + "step": 25666 + }, + { + "epoch": 5.6, + "learning_rate": 2.201551234433035e-06, + "loss": 2.0096, + "step": 25667 + }, + { + "epoch": 5.6, + "learning_rate": 2.2014419925715533e-06, + "loss": 1.3816, + "step": 25668 + }, + { + "epoch": 5.6, + "learning_rate": 2.2013327507100724e-06, + "loss": 1.947, + "step": 25669 + }, + { + "epoch": 5.6, + "learning_rate": 2.201223508848591e-06, + "loss": 1.0304, + "step": 25670 + }, + { + "epoch": 5.6, + "learning_rate": 2.2011142669871096e-06, + "loss": 1.1102, + "step": 25671 + }, + { + "epoch": 5.6, + "learning_rate": 2.2010050251256282e-06, + "loss": 1.9031, + "step": 25672 + }, + { + "epoch": 5.6, + "learning_rate": 2.200895783264147e-06, + "loss": 2.8871, + "step": 25673 + }, + { + "epoch": 5.6, + "learning_rate": 2.200786541402666e-06, + "loss": 1.8317, + "step": 25674 + }, + { + "epoch": 5.6, + "learning_rate": 2.2006772995411845e-06, + "loss": 1.4237, + "step": 25675 + }, + { + "epoch": 5.6, + "learning_rate": 2.200568057679703e-06, + "loss": 0.2887, + "step": 25676 + }, + { + "epoch": 5.6, + "learning_rate": 2.2004588158182217e-06, + "loss": 1.3904, + "step": 25677 + }, + { + "epoch": 5.6, + "learning_rate": 2.2003495739567403e-06, + "loss": 1.7121, + "step": 25678 + }, + { + "epoch": 5.6, + "learning_rate": 2.200240332095259e-06, + "loss": 2.2675, + "step": 25679 + }, + { + "epoch": 5.6, + "learning_rate": 2.200131090233778e-06, + "loss": 0.9891, + "step": 25680 + }, + { + "epoch": 5.6, + "learning_rate": 2.2000218483722966e-06, + "loss": 1.7832, + "step": 25681 + }, + { + "epoch": 5.6, + "learning_rate": 2.199912606510815e-06, + "loss": 0.6937, + "step": 25682 + }, + { + "epoch": 5.61, + "learning_rate": 2.1998033646493338e-06, + "loss": 2.3611, + "step": 25683 + }, + { + "epoch": 5.61, + "learning_rate": 2.1996941227878524e-06, + "loss": 1.1027, + "step": 25684 + }, + { + "epoch": 5.61, + "learning_rate": 2.199584880926371e-06, + "loss": 0.4607, + "step": 25685 + }, + { + "epoch": 5.61, + "learning_rate": 2.19947563906489e-06, + "loss": 2.9897, + "step": 25686 + }, + { + "epoch": 5.61, + "learning_rate": 2.1993663972034086e-06, + "loss": 1.3467, + "step": 25687 + }, + { + "epoch": 5.61, + "learning_rate": 2.1992571553419273e-06, + "loss": 1.5296, + "step": 25688 + }, + { + "epoch": 5.61, + "learning_rate": 2.199147913480446e-06, + "loss": 1.1253, + "step": 25689 + }, + { + "epoch": 5.61, + "learning_rate": 2.1990386716189645e-06, + "loss": 2.7896, + "step": 25690 + }, + { + "epoch": 5.61, + "learning_rate": 2.198929429757483e-06, + "loss": 1.6166, + "step": 25691 + }, + { + "epoch": 5.61, + "learning_rate": 2.198820187896002e-06, + "loss": 0.5, + "step": 25692 + }, + { + "epoch": 5.61, + "learning_rate": 2.1987109460345207e-06, + "loss": 3.1291, + "step": 25693 + }, + { + "epoch": 5.61, + "learning_rate": 2.1986017041730393e-06, + "loss": 2.6137, + "step": 25694 + }, + { + "epoch": 5.61, + "learning_rate": 2.198492462311558e-06, + "loss": 0.8162, + "step": 25695 + }, + { + "epoch": 5.61, + "learning_rate": 2.1983832204500766e-06, + "loss": 1.7594, + "step": 25696 + }, + { + "epoch": 5.61, + "learning_rate": 2.198273978588595e-06, + "loss": 1.9711, + "step": 25697 + }, + { + "epoch": 5.61, + "learning_rate": 2.198164736727114e-06, + "loss": 1.4258, + "step": 25698 + }, + { + "epoch": 5.61, + "learning_rate": 2.198055494865633e-06, + "loss": 2.0343, + "step": 25699 + }, + { + "epoch": 5.61, + "learning_rate": 2.1979462530041514e-06, + "loss": 0.3264, + "step": 25700 + }, + { + "epoch": 5.61, + "learning_rate": 2.19783701114267e-06, + "loss": 1.8775, + "step": 25701 + }, + { + "epoch": 5.61, + "learning_rate": 2.1977277692811887e-06, + "loss": 2.4858, + "step": 25702 + }, + { + "epoch": 5.61, + "learning_rate": 2.1976185274197073e-06, + "loss": 0.205, + "step": 25703 + }, + { + "epoch": 5.61, + "learning_rate": 2.1975092855582263e-06, + "loss": 2.1227, + "step": 25704 + }, + { + "epoch": 5.61, + "learning_rate": 2.1974000436967445e-06, + "loss": 0.9438, + "step": 25705 + }, + { + "epoch": 5.61, + "learning_rate": 2.1972908018352635e-06, + "loss": 2.2669, + "step": 25706 + }, + { + "epoch": 5.61, + "learning_rate": 2.197181559973782e-06, + "loss": 1.1644, + "step": 25707 + }, + { + "epoch": 5.61, + "learning_rate": 2.1970723181123008e-06, + "loss": 1.3122, + "step": 25708 + }, + { + "epoch": 5.61, + "learning_rate": 2.1969630762508194e-06, + "loss": 0.9654, + "step": 25709 + }, + { + "epoch": 5.61, + "learning_rate": 2.196853834389338e-06, + "loss": 2.0589, + "step": 25710 + }, + { + "epoch": 5.61, + "learning_rate": 2.196744592527857e-06, + "loss": 0.8953, + "step": 25711 + }, + { + "epoch": 5.61, + "learning_rate": 2.196635350666375e-06, + "loss": 1.0743, + "step": 25712 + }, + { + "epoch": 5.61, + "learning_rate": 2.1965261088048942e-06, + "loss": 1.0811, + "step": 25713 + }, + { + "epoch": 5.61, + "learning_rate": 2.196416866943413e-06, + "loss": 2.3015, + "step": 25714 + }, + { + "epoch": 5.61, + "learning_rate": 2.1963076250819315e-06, + "loss": 3.0352, + "step": 25715 + }, + { + "epoch": 5.61, + "learning_rate": 2.1961983832204505e-06, + "loss": 0.6723, + "step": 25716 + }, + { + "epoch": 5.61, + "learning_rate": 2.1960891413589687e-06, + "loss": 2.4164, + "step": 25717 + }, + { + "epoch": 5.61, + "learning_rate": 2.1959798994974877e-06, + "loss": 1.2041, + "step": 25718 + }, + { + "epoch": 5.61, + "learning_rate": 2.1958706576360063e-06, + "loss": 1.7485, + "step": 25719 + }, + { + "epoch": 5.61, + "learning_rate": 2.195761415774525e-06, + "loss": 1.327, + "step": 25720 + }, + { + "epoch": 5.61, + "learning_rate": 2.1956521739130435e-06, + "loss": 2.5434, + "step": 25721 + }, + { + "epoch": 5.61, + "learning_rate": 2.195542932051562e-06, + "loss": 1.7602, + "step": 25722 + }, + { + "epoch": 5.61, + "learning_rate": 2.195433690190081e-06, + "loss": 1.5494, + "step": 25723 + }, + { + "epoch": 5.61, + "learning_rate": 2.1953244483286e-06, + "loss": 0.6084, + "step": 25724 + }, + { + "epoch": 5.61, + "learning_rate": 2.1952152064671184e-06, + "loss": 0.8076, + "step": 25725 + }, + { + "epoch": 5.61, + "learning_rate": 2.195105964605637e-06, + "loss": 0.8497, + "step": 25726 + }, + { + "epoch": 5.61, + "learning_rate": 2.1949967227441556e-06, + "loss": 3.1444, + "step": 25727 + }, + { + "epoch": 5.62, + "learning_rate": 2.1948874808826747e-06, + "loss": 1.1803, + "step": 25728 + }, + { + "epoch": 5.62, + "learning_rate": 2.194778239021193e-06, + "loss": 0.522, + "step": 25729 + }, + { + "epoch": 5.62, + "learning_rate": 2.194668997159712e-06, + "loss": 0.6295, + "step": 25730 + }, + { + "epoch": 5.62, + "learning_rate": 2.1945597552982305e-06, + "loss": 2.1757, + "step": 25731 + }, + { + "epoch": 5.62, + "learning_rate": 2.194450513436749e-06, + "loss": 0.9503, + "step": 25732 + }, + { + "epoch": 5.62, + "learning_rate": 2.194341271575268e-06, + "loss": 1.5004, + "step": 25733 + }, + { + "epoch": 5.62, + "learning_rate": 2.1942320297137863e-06, + "loss": 0.8718, + "step": 25734 + }, + { + "epoch": 5.62, + "learning_rate": 2.1941227878523054e-06, + "loss": 1.2306, + "step": 25735 + }, + { + "epoch": 5.62, + "learning_rate": 2.194013545990824e-06, + "loss": 0.6725, + "step": 25736 + }, + { + "epoch": 5.62, + "learning_rate": 2.1939043041293426e-06, + "loss": 1.3164, + "step": 25737 + }, + { + "epoch": 5.62, + "learning_rate": 2.193795062267861e-06, + "loss": 1.189, + "step": 25738 + }, + { + "epoch": 5.62, + "learning_rate": 2.19368582040638e-06, + "loss": 1.6856, + "step": 25739 + }, + { + "epoch": 5.62, + "learning_rate": 2.193576578544899e-06, + "loss": 1.4416, + "step": 25740 + }, + { + "epoch": 5.62, + "learning_rate": 2.193467336683417e-06, + "loss": 1.965, + "step": 25741 + }, + { + "epoch": 5.62, + "learning_rate": 2.193358094821936e-06, + "loss": 1.091, + "step": 25742 + }, + { + "epoch": 5.62, + "learning_rate": 2.1932488529604547e-06, + "loss": 1.5262, + "step": 25743 + }, + { + "epoch": 5.62, + "learning_rate": 2.1931396110989733e-06, + "loss": 0.4108, + "step": 25744 + }, + { + "epoch": 5.62, + "learning_rate": 2.193030369237492e-06, + "loss": 1.3433, + "step": 25745 + }, + { + "epoch": 5.62, + "learning_rate": 2.1929211273760105e-06, + "loss": 1.5099, + "step": 25746 + }, + { + "epoch": 5.62, + "learning_rate": 2.1928118855145295e-06, + "loss": 2.9323, + "step": 25747 + }, + { + "epoch": 5.62, + "learning_rate": 2.192702643653048e-06, + "loss": 1.4996, + "step": 25748 + }, + { + "epoch": 5.62, + "learning_rate": 2.1925934017915668e-06, + "loss": 1.2087, + "step": 25749 + }, + { + "epoch": 5.62, + "learning_rate": 2.1924841599300854e-06, + "loss": 0.8071, + "step": 25750 + }, + { + "epoch": 5.62, + "learning_rate": 2.192374918068604e-06, + "loss": 0.9002, + "step": 25751 + }, + { + "epoch": 5.62, + "learning_rate": 2.192265676207123e-06, + "loss": 0.7992, + "step": 25752 + }, + { + "epoch": 5.62, + "learning_rate": 2.192156434345641e-06, + "loss": 1.6133, + "step": 25753 + }, + { + "epoch": 5.62, + "learning_rate": 2.1920471924841602e-06, + "loss": 2.8992, + "step": 25754 + }, + { + "epoch": 5.62, + "learning_rate": 2.191937950622679e-06, + "loss": 1.4084, + "step": 25755 + }, + { + "epoch": 5.62, + "learning_rate": 2.1918287087611975e-06, + "loss": 0.6062, + "step": 25756 + }, + { + "epoch": 5.62, + "learning_rate": 2.191719466899716e-06, + "loss": 2.2235, + "step": 25757 + }, + { + "epoch": 5.62, + "learning_rate": 2.1916102250382347e-06, + "loss": 2.7372, + "step": 25758 + }, + { + "epoch": 5.62, + "learning_rate": 2.1915009831767537e-06, + "loss": 1.6794, + "step": 25759 + }, + { + "epoch": 5.62, + "learning_rate": 2.1913917413152723e-06, + "loss": 1.4287, + "step": 25760 + }, + { + "epoch": 5.62, + "learning_rate": 2.191282499453791e-06, + "loss": 1.6848, + "step": 25761 + }, + { + "epoch": 5.62, + "learning_rate": 2.1911732575923096e-06, + "loss": 0.6737, + "step": 25762 + }, + { + "epoch": 5.62, + "learning_rate": 2.191064015730828e-06, + "loss": 1.9796, + "step": 25763 + }, + { + "epoch": 5.62, + "learning_rate": 2.1909547738693468e-06, + "loss": 1.9192, + "step": 25764 + }, + { + "epoch": 5.62, + "learning_rate": 2.1908455320078654e-06, + "loss": 0.813, + "step": 25765 + }, + { + "epoch": 5.62, + "learning_rate": 2.1907362901463844e-06, + "loss": 1.1164, + "step": 25766 + }, + { + "epoch": 5.62, + "learning_rate": 2.190627048284903e-06, + "loss": 0.8023, + "step": 25767 + }, + { + "epoch": 5.62, + "learning_rate": 2.1905178064234216e-06, + "loss": 0.5902, + "step": 25768 + }, + { + "epoch": 5.62, + "learning_rate": 2.1904085645619403e-06, + "loss": 1.5314, + "step": 25769 + }, + { + "epoch": 5.62, + "learning_rate": 2.190299322700459e-06, + "loss": 0.8117, + "step": 25770 + }, + { + "epoch": 5.62, + "learning_rate": 2.1901900808389775e-06, + "loss": 0.3846, + "step": 25771 + }, + { + "epoch": 5.62, + "learning_rate": 2.1900808389774965e-06, + "loss": 1.2286, + "step": 25772 + }, + { + "epoch": 5.62, + "learning_rate": 2.189971597116015e-06, + "loss": 2.2048, + "step": 25773 + }, + { + "epoch": 5.63, + "learning_rate": 2.1898623552545337e-06, + "loss": 2.4834, + "step": 25774 + }, + { + "epoch": 5.63, + "learning_rate": 2.1897531133930523e-06, + "loss": 2.8095, + "step": 25775 + }, + { + "epoch": 5.63, + "learning_rate": 2.189643871531571e-06, + "loss": 0.4905, + "step": 25776 + }, + { + "epoch": 5.63, + "learning_rate": 2.18953462967009e-06, + "loss": 1.1412, + "step": 25777 + }, + { + "epoch": 5.63, + "learning_rate": 2.189425387808608e-06, + "loss": 2.5631, + "step": 25778 + }, + { + "epoch": 5.63, + "learning_rate": 2.1893161459471272e-06, + "loss": 0.4372, + "step": 25779 + }, + { + "epoch": 5.63, + "learning_rate": 2.189206904085646e-06, + "loss": 1.8409, + "step": 25780 + }, + { + "epoch": 5.63, + "learning_rate": 2.1890976622241644e-06, + "loss": 1.4809, + "step": 25781 + }, + { + "epoch": 5.63, + "learning_rate": 2.188988420362683e-06, + "loss": 2.2785, + "step": 25782 + }, + { + "epoch": 5.63, + "learning_rate": 2.1888791785012017e-06, + "loss": 0.5942, + "step": 25783 + }, + { + "epoch": 5.63, + "learning_rate": 2.1887699366397207e-06, + "loss": 1.0645, + "step": 25784 + }, + { + "epoch": 5.63, + "learning_rate": 2.1886606947782393e-06, + "loss": 0.5922, + "step": 25785 + }, + { + "epoch": 5.63, + "learning_rate": 2.188551452916758e-06, + "loss": 1.2867, + "step": 25786 + }, + { + "epoch": 5.63, + "learning_rate": 2.1884422110552765e-06, + "loss": 1.5843, + "step": 25787 + }, + { + "epoch": 5.63, + "learning_rate": 2.188332969193795e-06, + "loss": 2.7717, + "step": 25788 + }, + { + "epoch": 5.63, + "learning_rate": 2.188223727332314e-06, + "loss": 1.4704, + "step": 25789 + }, + { + "epoch": 5.63, + "learning_rate": 2.1881144854708324e-06, + "loss": 1.6035, + "step": 25790 + }, + { + "epoch": 5.63, + "learning_rate": 2.1880052436093514e-06, + "loss": 0.9709, + "step": 25791 + }, + { + "epoch": 5.63, + "learning_rate": 2.18789600174787e-06, + "loss": 2.4967, + "step": 25792 + }, + { + "epoch": 5.63, + "learning_rate": 2.1877867598863886e-06, + "loss": 1.1741, + "step": 25793 + }, + { + "epoch": 5.63, + "learning_rate": 2.1876775180249072e-06, + "loss": 1.3957, + "step": 25794 + }, + { + "epoch": 5.63, + "learning_rate": 2.187568276163426e-06, + "loss": 1.6717, + "step": 25795 + }, + { + "epoch": 5.63, + "learning_rate": 2.187459034301945e-06, + "loss": 2.0451, + "step": 25796 + }, + { + "epoch": 5.63, + "learning_rate": 2.187349792440463e-06, + "loss": 1.2703, + "step": 25797 + }, + { + "epoch": 5.63, + "learning_rate": 2.187240550578982e-06, + "loss": 1.6814, + "step": 25798 + }, + { + "epoch": 5.63, + "learning_rate": 2.1871313087175007e-06, + "loss": 1.6666, + "step": 25799 + }, + { + "epoch": 5.63, + "learning_rate": 2.1870220668560193e-06, + "loss": 1.7376, + "step": 25800 + }, + { + "epoch": 5.63, + "learning_rate": 2.1869128249945383e-06, + "loss": 0.9384, + "step": 25801 + }, + { + "epoch": 5.63, + "learning_rate": 2.1868035831330565e-06, + "loss": 1.8923, + "step": 25802 + }, + { + "epoch": 5.63, + "learning_rate": 2.1866943412715756e-06, + "loss": 2.8219, + "step": 25803 + }, + { + "epoch": 5.63, + "learning_rate": 2.186585099410094e-06, + "loss": 1.4678, + "step": 25804 + }, + { + "epoch": 5.63, + "learning_rate": 2.186475857548613e-06, + "loss": 1.3249, + "step": 25805 + }, + { + "epoch": 5.63, + "learning_rate": 2.1863666156871314e-06, + "loss": 0.8804, + "step": 25806 + }, + { + "epoch": 5.63, + "learning_rate": 2.18625737382565e-06, + "loss": 1.8487, + "step": 25807 + }, + { + "epoch": 5.63, + "learning_rate": 2.186148131964169e-06, + "loss": 0.8542, + "step": 25808 + }, + { + "epoch": 5.63, + "learning_rate": 2.1860388901026872e-06, + "loss": 1.4417, + "step": 25809 + }, + { + "epoch": 5.63, + "learning_rate": 2.1859296482412063e-06, + "loss": 1.5842, + "step": 25810 + }, + { + "epoch": 5.63, + "learning_rate": 2.185820406379725e-06, + "loss": 2.79, + "step": 25811 + }, + { + "epoch": 5.63, + "learning_rate": 2.1857111645182435e-06, + "loss": 1.2224, + "step": 25812 + }, + { + "epoch": 5.63, + "learning_rate": 2.1856019226567625e-06, + "loss": 0.2677, + "step": 25813 + }, + { + "epoch": 5.63, + "learning_rate": 2.1854926807952807e-06, + "loss": 1.2318, + "step": 25814 + }, + { + "epoch": 5.63, + "learning_rate": 2.1853834389337997e-06, + "loss": 1.607, + "step": 25815 + }, + { + "epoch": 5.63, + "learning_rate": 2.1852741970723184e-06, + "loss": 1.6405, + "step": 25816 + }, + { + "epoch": 5.63, + "learning_rate": 2.185164955210837e-06, + "loss": 0.8066, + "step": 25817 + }, + { + "epoch": 5.63, + "learning_rate": 2.1850557133493556e-06, + "loss": 1.0302, + "step": 25818 + }, + { + "epoch": 5.63, + "learning_rate": 2.184946471487874e-06, + "loss": 2.1225, + "step": 25819 + }, + { + "epoch": 5.64, + "learning_rate": 2.1848372296263932e-06, + "loss": 1.1544, + "step": 25820 + }, + { + "epoch": 5.64, + "learning_rate": 2.184727987764912e-06, + "loss": 1.4823, + "step": 25821 + }, + { + "epoch": 5.64, + "learning_rate": 2.1846187459034304e-06, + "loss": 1.2892, + "step": 25822 + }, + { + "epoch": 5.64, + "learning_rate": 2.184509504041949e-06, + "loss": 1.3313, + "step": 25823 + }, + { + "epoch": 5.64, + "learning_rate": 2.1844002621804677e-06, + "loss": 2.1844, + "step": 25824 + }, + { + "epoch": 5.64, + "learning_rate": 2.1842910203189867e-06, + "loss": 2.6209, + "step": 25825 + }, + { + "epoch": 5.64, + "learning_rate": 2.184181778457505e-06, + "loss": 1.2076, + "step": 25826 + }, + { + "epoch": 5.64, + "learning_rate": 2.184072536596024e-06, + "loss": 0.4219, + "step": 25827 + }, + { + "epoch": 5.64, + "learning_rate": 2.1839632947345425e-06, + "loss": 1.436, + "step": 25828 + }, + { + "epoch": 5.64, + "learning_rate": 2.183854052873061e-06, + "loss": 3.0317, + "step": 25829 + }, + { + "epoch": 5.64, + "learning_rate": 2.1837448110115798e-06, + "loss": 1.8903, + "step": 25830 + }, + { + "epoch": 5.64, + "learning_rate": 2.1836355691500984e-06, + "loss": 0.662, + "step": 25831 + }, + { + "epoch": 5.64, + "learning_rate": 2.1835263272886174e-06, + "loss": 1.0003, + "step": 25832 + }, + { + "epoch": 5.64, + "learning_rate": 2.183417085427136e-06, + "loss": 1.3695, + "step": 25833 + }, + { + "epoch": 5.64, + "learning_rate": 2.1833078435656546e-06, + "loss": 1.5446, + "step": 25834 + }, + { + "epoch": 5.64, + "learning_rate": 2.1831986017041732e-06, + "loss": 1.6513, + "step": 25835 + }, + { + "epoch": 5.64, + "learning_rate": 2.183089359842692e-06, + "loss": 0.2227, + "step": 25836 + }, + { + "epoch": 5.64, + "learning_rate": 2.1829801179812105e-06, + "loss": 0.5528, + "step": 25837 + }, + { + "epoch": 5.64, + "learning_rate": 2.182870876119729e-06, + "loss": 2.5529, + "step": 25838 + }, + { + "epoch": 5.64, + "learning_rate": 2.182761634258248e-06, + "loss": 0.759, + "step": 25839 + }, + { + "epoch": 5.64, + "learning_rate": 2.1826523923967667e-06, + "loss": 0.9277, + "step": 25840 + }, + { + "epoch": 5.64, + "learning_rate": 2.1825431505352853e-06, + "loss": 0.7566, + "step": 25841 + }, + { + "epoch": 5.64, + "learning_rate": 2.182433908673804e-06, + "loss": 2.3226, + "step": 25842 + }, + { + "epoch": 5.64, + "learning_rate": 2.1823246668123226e-06, + "loss": 3.1377, + "step": 25843 + }, + { + "epoch": 5.64, + "learning_rate": 2.182215424950841e-06, + "loss": 0.1014, + "step": 25844 + }, + { + "epoch": 5.64, + "learning_rate": 2.18210618308936e-06, + "loss": 1.7872, + "step": 25845 + }, + { + "epoch": 5.64, + "learning_rate": 2.181996941227879e-06, + "loss": 1.1753, + "step": 25846 + }, + { + "epoch": 5.64, + "learning_rate": 2.1818876993663974e-06, + "loss": 1.0481, + "step": 25847 + }, + { + "epoch": 5.64, + "learning_rate": 2.181778457504916e-06, + "loss": 1.0394, + "step": 25848 + }, + { + "epoch": 5.64, + "learning_rate": 2.1816692156434346e-06, + "loss": 2.2291, + "step": 25849 + }, + { + "epoch": 5.64, + "learning_rate": 2.1815599737819533e-06, + "loss": 0.5626, + "step": 25850 + }, + { + "epoch": 5.64, + "learning_rate": 2.181450731920472e-06, + "loss": 1.0699, + "step": 25851 + }, + { + "epoch": 5.64, + "learning_rate": 2.181341490058991e-06, + "loss": 2.4436, + "step": 25852 + }, + { + "epoch": 5.64, + "learning_rate": 2.1812322481975095e-06, + "loss": 3.0111, + "step": 25853 + }, + { + "epoch": 5.64, + "learning_rate": 2.181123006336028e-06, + "loss": 1.7583, + "step": 25854 + }, + { + "epoch": 5.64, + "learning_rate": 2.1810137644745467e-06, + "loss": 1.6322, + "step": 25855 + }, + { + "epoch": 5.64, + "learning_rate": 2.1809045226130653e-06, + "loss": 0.5408, + "step": 25856 + }, + { + "epoch": 5.64, + "learning_rate": 2.1807952807515844e-06, + "loss": 1.3398, + "step": 25857 + }, + { + "epoch": 5.64, + "learning_rate": 2.180686038890103e-06, + "loss": 1.4808, + "step": 25858 + }, + { + "epoch": 5.64, + "learning_rate": 2.1805767970286216e-06, + "loss": 1.075, + "step": 25859 + }, + { + "epoch": 5.64, + "learning_rate": 2.18046755516714e-06, + "loss": 1.7488, + "step": 25860 + }, + { + "epoch": 5.64, + "learning_rate": 2.180358313305659e-06, + "loss": 2.4096, + "step": 25861 + }, + { + "epoch": 5.64, + "learning_rate": 2.1802490714441774e-06, + "loss": 1.4705, + "step": 25862 + }, + { + "epoch": 5.64, + "learning_rate": 2.180139829582696e-06, + "loss": 2.212, + "step": 25863 + }, + { + "epoch": 5.64, + "learning_rate": 2.180030587721215e-06, + "loss": 2.6287, + "step": 25864 + }, + { + "epoch": 5.64, + "learning_rate": 2.1799213458597337e-06, + "loss": 2.2356, + "step": 25865 + }, + { + "epoch": 5.65, + "learning_rate": 2.1798121039982523e-06, + "loss": 0.4563, + "step": 25866 + }, + { + "epoch": 5.65, + "learning_rate": 2.179702862136771e-06, + "loss": 2.2814, + "step": 25867 + }, + { + "epoch": 5.65, + "learning_rate": 2.1795936202752895e-06, + "loss": 1.1172, + "step": 25868 + }, + { + "epoch": 5.65, + "learning_rate": 2.1794843784138086e-06, + "loss": 1.9011, + "step": 25869 + }, + { + "epoch": 5.65, + "learning_rate": 2.1793751365523267e-06, + "loss": 0.9823, + "step": 25870 + }, + { + "epoch": 5.65, + "learning_rate": 2.1792658946908458e-06, + "loss": 2.1835, + "step": 25871 + }, + { + "epoch": 5.65, + "learning_rate": 2.1791566528293644e-06, + "loss": 2.1614, + "step": 25872 + }, + { + "epoch": 5.65, + "learning_rate": 2.179047410967883e-06, + "loss": 2.4504, + "step": 25873 + }, + { + "epoch": 5.65, + "learning_rate": 2.178938169106402e-06, + "loss": 0.8503, + "step": 25874 + }, + { + "epoch": 5.65, + "learning_rate": 2.1788289272449202e-06, + "loss": 1.4254, + "step": 25875 + }, + { + "epoch": 5.65, + "learning_rate": 2.1787196853834393e-06, + "loss": 1.6327, + "step": 25876 + }, + { + "epoch": 5.65, + "learning_rate": 2.178610443521958e-06, + "loss": 2.4192, + "step": 25877 + }, + { + "epoch": 5.65, + "learning_rate": 2.1785012016604765e-06, + "loss": 2.0843, + "step": 25878 + }, + { + "epoch": 5.65, + "learning_rate": 2.178391959798995e-06, + "loss": 2.0644, + "step": 25879 + }, + { + "epoch": 5.65, + "learning_rate": 2.1782827179375137e-06, + "loss": 0.473, + "step": 25880 + }, + { + "epoch": 5.65, + "learning_rate": 2.1781734760760327e-06, + "loss": 1.9053, + "step": 25881 + }, + { + "epoch": 5.65, + "learning_rate": 2.178064234214551e-06, + "loss": 1.0292, + "step": 25882 + }, + { + "epoch": 5.65, + "learning_rate": 2.17795499235307e-06, + "loss": 1.5935, + "step": 25883 + }, + { + "epoch": 5.65, + "learning_rate": 2.1778457504915886e-06, + "loss": 0.9369, + "step": 25884 + }, + { + "epoch": 5.65, + "learning_rate": 2.177736508630107e-06, + "loss": 1.3784, + "step": 25885 + }, + { + "epoch": 5.65, + "learning_rate": 2.177627266768626e-06, + "loss": 1.6729, + "step": 25886 + }, + { + "epoch": 5.65, + "learning_rate": 2.1775180249071444e-06, + "loss": 2.3144, + "step": 25887 + }, + { + "epoch": 5.65, + "learning_rate": 2.1774087830456634e-06, + "loss": 1.2889, + "step": 25888 + }, + { + "epoch": 5.65, + "learning_rate": 2.177299541184182e-06, + "loss": 1.1874, + "step": 25889 + }, + { + "epoch": 5.65, + "learning_rate": 2.1771902993227007e-06, + "loss": 2.8939, + "step": 25890 + }, + { + "epoch": 5.65, + "learning_rate": 2.1770810574612193e-06, + "loss": 2.0773, + "step": 25891 + }, + { + "epoch": 5.65, + "learning_rate": 2.176971815599738e-06, + "loss": 0.2165, + "step": 25892 + }, + { + "epoch": 5.65, + "learning_rate": 2.176862573738257e-06, + "loss": 1.8784, + "step": 25893 + }, + { + "epoch": 5.65, + "learning_rate": 2.176753331876775e-06, + "loss": 0.7722, + "step": 25894 + }, + { + "epoch": 5.65, + "learning_rate": 2.176644090015294e-06, + "loss": 0.3714, + "step": 25895 + }, + { + "epoch": 5.65, + "learning_rate": 2.1765348481538127e-06, + "loss": 0.611, + "step": 25896 + }, + { + "epoch": 5.65, + "learning_rate": 2.1764256062923314e-06, + "loss": 1.9183, + "step": 25897 + }, + { + "epoch": 5.65, + "learning_rate": 2.1763163644308504e-06, + "loss": 2.663, + "step": 25898 + }, + { + "epoch": 5.65, + "learning_rate": 2.1762071225693686e-06, + "loss": 1.9163, + "step": 25899 + }, + { + "epoch": 5.65, + "learning_rate": 2.1760978807078876e-06, + "loss": 1.0206, + "step": 25900 + }, + { + "epoch": 5.65, + "learning_rate": 2.1759886388464062e-06, + "loss": 1.6557, + "step": 25901 + }, + { + "epoch": 5.65, + "learning_rate": 2.175879396984925e-06, + "loss": 0.8452, + "step": 25902 + }, + { + "epoch": 5.65, + "learning_rate": 2.1757701551234434e-06, + "loss": 1.0314, + "step": 25903 + }, + { + "epoch": 5.65, + "learning_rate": 2.175660913261962e-06, + "loss": 1.5333, + "step": 25904 + }, + { + "epoch": 5.65, + "learning_rate": 2.175551671400481e-06, + "loss": 0.7121, + "step": 25905 + }, + { + "epoch": 5.65, + "learning_rate": 2.1754424295389993e-06, + "loss": 0.3492, + "step": 25906 + }, + { + "epoch": 5.65, + "learning_rate": 2.1753331876775183e-06, + "loss": 2.1751, + "step": 25907 + }, + { + "epoch": 5.65, + "learning_rate": 2.175223945816037e-06, + "loss": 1.5964, + "step": 25908 + }, + { + "epoch": 5.65, + "learning_rate": 2.1751147039545555e-06, + "loss": 1.2924, + "step": 25909 + }, + { + "epoch": 5.65, + "learning_rate": 2.175005462093074e-06, + "loss": 2.3651, + "step": 25910 + }, + { + "epoch": 5.65, + "learning_rate": 2.1748962202315928e-06, + "loss": 0.8402, + "step": 25911 + }, + { + "epoch": 5.66, + "learning_rate": 2.174786978370112e-06, + "loss": 1.279, + "step": 25912 + }, + { + "epoch": 5.66, + "learning_rate": 2.1746777365086304e-06, + "loss": 3.3928, + "step": 25913 + }, + { + "epoch": 5.66, + "learning_rate": 2.174568494647149e-06, + "loss": 1.9487, + "step": 25914 + }, + { + "epoch": 5.66, + "learning_rate": 2.1744592527856676e-06, + "loss": 1.8635, + "step": 25915 + }, + { + "epoch": 5.66, + "learning_rate": 2.1743500109241862e-06, + "loss": 2.2895, + "step": 25916 + }, + { + "epoch": 5.66, + "learning_rate": 2.174240769062705e-06, + "loss": 2.0764, + "step": 25917 + }, + { + "epoch": 5.66, + "learning_rate": 2.1741315272012235e-06, + "loss": 1.2696, + "step": 25918 + }, + { + "epoch": 5.66, + "learning_rate": 2.1740222853397425e-06, + "loss": 1.264, + "step": 25919 + }, + { + "epoch": 5.66, + "learning_rate": 2.173913043478261e-06, + "loss": 0.2914, + "step": 25920 + }, + { + "epoch": 5.66, + "learning_rate": 2.1738038016167797e-06, + "loss": 2.3174, + "step": 25921 + }, + { + "epoch": 5.66, + "learning_rate": 2.1736945597552983e-06, + "loss": 0.9007, + "step": 25922 + }, + { + "epoch": 5.66, + "learning_rate": 2.173585317893817e-06, + "loss": 1.3697, + "step": 25923 + }, + { + "epoch": 5.66, + "learning_rate": 2.173476076032336e-06, + "loss": 1.0839, + "step": 25924 + }, + { + "epoch": 5.66, + "learning_rate": 2.1733668341708546e-06, + "loss": 0.8, + "step": 25925 + }, + { + "epoch": 5.66, + "learning_rate": 2.173257592309373e-06, + "loss": 1.0046, + "step": 25926 + }, + { + "epoch": 5.66, + "learning_rate": 2.173148350447892e-06, + "loss": 2.7801, + "step": 25927 + }, + { + "epoch": 5.66, + "learning_rate": 2.1730391085864104e-06, + "loss": 1.1572, + "step": 25928 + }, + { + "epoch": 5.66, + "learning_rate": 2.172929866724929e-06, + "loss": 0.8497, + "step": 25929 + }, + { + "epoch": 5.66, + "learning_rate": 2.172820624863448e-06, + "loss": 2.6124, + "step": 25930 + }, + { + "epoch": 5.66, + "learning_rate": 2.1727113830019667e-06, + "loss": 0.5472, + "step": 25931 + }, + { + "epoch": 5.66, + "learning_rate": 2.1726021411404853e-06, + "loss": 2.8002, + "step": 25932 + }, + { + "epoch": 5.66, + "learning_rate": 2.172492899279004e-06, + "loss": 1.645, + "step": 25933 + }, + { + "epoch": 5.66, + "learning_rate": 2.1723836574175225e-06, + "loss": 1.37, + "step": 25934 + }, + { + "epoch": 5.66, + "learning_rate": 2.172274415556041e-06, + "loss": 1.2953, + "step": 25935 + }, + { + "epoch": 5.66, + "learning_rate": 2.1721651736945597e-06, + "loss": 1.5768, + "step": 25936 + }, + { + "epoch": 5.66, + "learning_rate": 2.1720559318330788e-06, + "loss": 0.2636, + "step": 25937 + }, + { + "epoch": 5.66, + "learning_rate": 2.1719466899715974e-06, + "loss": 0.9596, + "step": 25938 + }, + { + "epoch": 5.66, + "learning_rate": 2.171837448110116e-06, + "loss": 0.7987, + "step": 25939 + }, + { + "epoch": 5.66, + "learning_rate": 2.1717282062486346e-06, + "loss": 0.8241, + "step": 25940 + }, + { + "epoch": 5.66, + "learning_rate": 2.171618964387153e-06, + "loss": 0.1786, + "step": 25941 + }, + { + "epoch": 5.66, + "learning_rate": 2.1715097225256722e-06, + "loss": 0.775, + "step": 25942 + }, + { + "epoch": 5.66, + "learning_rate": 2.1714004806641904e-06, + "loss": 1.0819, + "step": 25943 + }, + { + "epoch": 5.66, + "learning_rate": 2.1712912388027095e-06, + "loss": 1.4292, + "step": 25944 + }, + { + "epoch": 5.66, + "learning_rate": 2.171181996941228e-06, + "loss": 1.2249, + "step": 25945 + }, + { + "epoch": 5.66, + "learning_rate": 2.1710727550797467e-06, + "loss": 0.3897, + "step": 25946 + }, + { + "epoch": 5.66, + "learning_rate": 2.1709635132182653e-06, + "loss": 0.8316, + "step": 25947 + }, + { + "epoch": 5.66, + "learning_rate": 2.170854271356784e-06, + "loss": 1.7141, + "step": 25948 + }, + { + "epoch": 5.66, + "learning_rate": 2.170745029495303e-06, + "loss": 0.4137, + "step": 25949 + }, + { + "epoch": 5.66, + "learning_rate": 2.170635787633821e-06, + "loss": 2.9163, + "step": 25950 + }, + { + "epoch": 5.66, + "learning_rate": 2.17052654577234e-06, + "loss": 1.5336, + "step": 25951 + }, + { + "epoch": 5.66, + "learning_rate": 2.1704173039108588e-06, + "loss": 1.0332, + "step": 25952 + }, + { + "epoch": 5.66, + "learning_rate": 2.1703080620493774e-06, + "loss": 0.2918, + "step": 25953 + }, + { + "epoch": 5.66, + "learning_rate": 2.1701988201878964e-06, + "loss": 0.9107, + "step": 25954 + }, + { + "epoch": 5.66, + "learning_rate": 2.1700895783264146e-06, + "loss": 1.1416, + "step": 25955 + }, + { + "epoch": 5.66, + "learning_rate": 2.1699803364649336e-06, + "loss": 0.8782, + "step": 25956 + }, + { + "epoch": 5.66, + "learning_rate": 2.1698710946034523e-06, + "loss": 2.3498, + "step": 25957 + }, + { + "epoch": 5.67, + "learning_rate": 2.169761852741971e-06, + "loss": 1.364, + "step": 25958 + }, + { + "epoch": 5.67, + "learning_rate": 2.1696526108804895e-06, + "loss": 2.6231, + "step": 25959 + }, + { + "epoch": 5.67, + "learning_rate": 2.169543369019008e-06, + "loss": 2.7629, + "step": 25960 + }, + { + "epoch": 5.67, + "learning_rate": 2.169434127157527e-06, + "loss": 1.5385, + "step": 25961 + }, + { + "epoch": 5.67, + "learning_rate": 2.1693248852960453e-06, + "loss": 1.6625, + "step": 25962 + }, + { + "epoch": 5.67, + "learning_rate": 2.1692156434345643e-06, + "loss": 1.8297, + "step": 25963 + }, + { + "epoch": 5.67, + "learning_rate": 2.169106401573083e-06, + "loss": 0.4963, + "step": 25964 + }, + { + "epoch": 5.67, + "learning_rate": 2.1689971597116016e-06, + "loss": 1.103, + "step": 25965 + }, + { + "epoch": 5.67, + "learning_rate": 2.1688879178501206e-06, + "loss": 1.2297, + "step": 25966 + }, + { + "epoch": 5.67, + "learning_rate": 2.1687786759886388e-06, + "loss": 1.6095, + "step": 25967 + }, + { + "epoch": 5.67, + "learning_rate": 2.168669434127158e-06, + "loss": 1.7323, + "step": 25968 + }, + { + "epoch": 5.67, + "learning_rate": 2.1685601922656764e-06, + "loss": 1.586, + "step": 25969 + }, + { + "epoch": 5.67, + "learning_rate": 2.168450950404195e-06, + "loss": 0.8475, + "step": 25970 + }, + { + "epoch": 5.67, + "learning_rate": 2.168341708542714e-06, + "loss": 1.5866, + "step": 25971 + }, + { + "epoch": 5.67, + "learning_rate": 2.1682324666812323e-06, + "loss": 2.2325, + "step": 25972 + }, + { + "epoch": 5.67, + "learning_rate": 2.1681232248197513e-06, + "loss": 1.5103, + "step": 25973 + }, + { + "epoch": 5.67, + "learning_rate": 2.16801398295827e-06, + "loss": 1.1975, + "step": 25974 + }, + { + "epoch": 5.67, + "learning_rate": 2.1679047410967885e-06, + "loss": 0.8836, + "step": 25975 + }, + { + "epoch": 5.67, + "learning_rate": 2.167795499235307e-06, + "loss": 1.2731, + "step": 25976 + }, + { + "epoch": 5.67, + "learning_rate": 2.1676862573738257e-06, + "loss": 1.162, + "step": 25977 + }, + { + "epoch": 5.67, + "learning_rate": 2.1675770155123448e-06, + "loss": 1.0263, + "step": 25978 + }, + { + "epoch": 5.67, + "learning_rate": 2.167467773650863e-06, + "loss": 1.9891, + "step": 25979 + }, + { + "epoch": 5.67, + "learning_rate": 2.167358531789382e-06, + "loss": 1.0216, + "step": 25980 + }, + { + "epoch": 5.67, + "learning_rate": 2.1672492899279006e-06, + "loss": 0.7958, + "step": 25981 + }, + { + "epoch": 5.67, + "learning_rate": 2.1671400480664192e-06, + "loss": 2.1277, + "step": 25982 + }, + { + "epoch": 5.67, + "learning_rate": 2.167030806204938e-06, + "loss": 1.2361, + "step": 25983 + }, + { + "epoch": 5.67, + "learning_rate": 2.1669215643434564e-06, + "loss": 3.4705, + "step": 25984 + }, + { + "epoch": 5.67, + "learning_rate": 2.1668123224819755e-06, + "loss": 2.7298, + "step": 25985 + }, + { + "epoch": 5.67, + "learning_rate": 2.166703080620494e-06, + "loss": 1.8879, + "step": 25986 + }, + { + "epoch": 5.67, + "learning_rate": 2.1665938387590127e-06, + "loss": 1.8802, + "step": 25987 + }, + { + "epoch": 5.67, + "learning_rate": 2.1664845968975313e-06, + "loss": 1.4452, + "step": 25988 + }, + { + "epoch": 5.67, + "learning_rate": 2.16637535503605e-06, + "loss": 0.9145, + "step": 25989 + }, + { + "epoch": 5.67, + "learning_rate": 2.166266113174569e-06, + "loss": 1.8104, + "step": 25990 + }, + { + "epoch": 5.67, + "learning_rate": 2.166156871313087e-06, + "loss": 0.7414, + "step": 25991 + }, + { + "epoch": 5.67, + "learning_rate": 2.166047629451606e-06, + "loss": 0.8459, + "step": 25992 + }, + { + "epoch": 5.67, + "learning_rate": 2.1659383875901248e-06, + "loss": 2.2059, + "step": 25993 + }, + { + "epoch": 5.67, + "learning_rate": 2.1658291457286434e-06, + "loss": 0.8652, + "step": 25994 + }, + { + "epoch": 5.67, + "learning_rate": 2.165719903867162e-06, + "loss": 0.626, + "step": 25995 + }, + { + "epoch": 5.67, + "learning_rate": 2.1656106620056806e-06, + "loss": 0.7381, + "step": 25996 + }, + { + "epoch": 5.67, + "learning_rate": 2.1655014201441997e-06, + "loss": 1.3751, + "step": 25997 + }, + { + "epoch": 5.67, + "learning_rate": 2.1653921782827183e-06, + "loss": 1.6831, + "step": 25998 + }, + { + "epoch": 5.67, + "learning_rate": 2.165282936421237e-06, + "loss": 1.8175, + "step": 25999 + }, + { + "epoch": 5.67, + "learning_rate": 2.1651736945597555e-06, + "loss": 3.2582, + "step": 26000 + }, + { + "epoch": 5.67, + "learning_rate": 2.165064452698274e-06, + "loss": 1.7686, + "step": 26001 + }, + { + "epoch": 5.67, + "learning_rate": 2.1649552108367927e-06, + "loss": 2.171, + "step": 26002 + }, + { + "epoch": 5.68, + "learning_rate": 2.1648459689753113e-06, + "loss": 1.9929, + "step": 26003 + }, + { + "epoch": 5.68, + "learning_rate": 2.1647367271138304e-06, + "loss": 0.9636, + "step": 26004 + }, + { + "epoch": 5.68, + "learning_rate": 2.164627485252349e-06, + "loss": 1.8219, + "step": 26005 + }, + { + "epoch": 5.68, + "learning_rate": 2.1645182433908676e-06, + "loss": 1.9567, + "step": 26006 + }, + { + "epoch": 5.68, + "learning_rate": 2.164409001529386e-06, + "loss": 0.6964, + "step": 26007 + }, + { + "epoch": 5.68, + "learning_rate": 2.164299759667905e-06, + "loss": 0.6774, + "step": 26008 + }, + { + "epoch": 5.68, + "learning_rate": 2.1641905178064234e-06, + "loss": 0.8539, + "step": 26009 + }, + { + "epoch": 5.68, + "learning_rate": 2.1640812759449424e-06, + "loss": 1.2404, + "step": 26010 + }, + { + "epoch": 5.68, + "learning_rate": 2.163972034083461e-06, + "loss": 0.503, + "step": 26011 + }, + { + "epoch": 5.68, + "learning_rate": 2.1638627922219797e-06, + "loss": 1.0129, + "step": 26012 + }, + { + "epoch": 5.68, + "learning_rate": 2.1637535503604983e-06, + "loss": 2.3588, + "step": 26013 + }, + { + "epoch": 5.68, + "learning_rate": 2.163644308499017e-06, + "loss": 1.6452, + "step": 26014 + }, + { + "epoch": 5.68, + "learning_rate": 2.1635350666375355e-06, + "loss": 0.9835, + "step": 26015 + }, + { + "epoch": 5.68, + "learning_rate": 2.163425824776054e-06, + "loss": 2.1927, + "step": 26016 + }, + { + "epoch": 5.68, + "learning_rate": 2.163316582914573e-06, + "loss": 1.3591, + "step": 26017 + }, + { + "epoch": 5.68, + "learning_rate": 2.1632073410530918e-06, + "loss": 1.4454, + "step": 26018 + }, + { + "epoch": 5.68, + "learning_rate": 2.1630980991916104e-06, + "loss": 2.7958, + "step": 26019 + }, + { + "epoch": 5.68, + "learning_rate": 2.162988857330129e-06, + "loss": 1.8595, + "step": 26020 + }, + { + "epoch": 5.68, + "learning_rate": 2.1628796154686476e-06, + "loss": 1.1721, + "step": 26021 + }, + { + "epoch": 5.68, + "learning_rate": 2.1627703736071666e-06, + "loss": 1.3007, + "step": 26022 + }, + { + "epoch": 5.68, + "learning_rate": 2.1626611317456852e-06, + "loss": 1.9426, + "step": 26023 + }, + { + "epoch": 5.68, + "learning_rate": 2.162551889884204e-06, + "loss": 2.3522, + "step": 26024 + }, + { + "epoch": 5.68, + "learning_rate": 2.1624426480227225e-06, + "loss": 1.0235, + "step": 26025 + }, + { + "epoch": 5.68, + "learning_rate": 2.162333406161241e-06, + "loss": 1.6687, + "step": 26026 + }, + { + "epoch": 5.68, + "learning_rate": 2.16222416429976e-06, + "loss": 0.4126, + "step": 26027 + }, + { + "epoch": 5.68, + "learning_rate": 2.1621149224382783e-06, + "loss": 0.3191, + "step": 26028 + }, + { + "epoch": 5.68, + "learning_rate": 2.1620056805767973e-06, + "loss": 2.2076, + "step": 26029 + }, + { + "epoch": 5.68, + "learning_rate": 2.161896438715316e-06, + "loss": 1.1079, + "step": 26030 + }, + { + "epoch": 5.68, + "learning_rate": 2.1617871968538345e-06, + "loss": 1.0222, + "step": 26031 + }, + { + "epoch": 5.68, + "learning_rate": 2.161677954992353e-06, + "loss": 1.4158, + "step": 26032 + }, + { + "epoch": 5.68, + "learning_rate": 2.1615687131308718e-06, + "loss": 0.7417, + "step": 26033 + }, + { + "epoch": 5.68, + "learning_rate": 2.161459471269391e-06, + "loss": 1.1312, + "step": 26034 + }, + { + "epoch": 5.68, + "learning_rate": 2.161350229407909e-06, + "loss": 0.8984, + "step": 26035 + }, + { + "epoch": 5.68, + "learning_rate": 2.161240987546428e-06, + "loss": 0.4043, + "step": 26036 + }, + { + "epoch": 5.68, + "learning_rate": 2.1611317456849466e-06, + "loss": 1.0403, + "step": 26037 + }, + { + "epoch": 5.68, + "learning_rate": 2.1610225038234652e-06, + "loss": 1.3957, + "step": 26038 + }, + { + "epoch": 5.68, + "learning_rate": 2.1609132619619843e-06, + "loss": 2.3682, + "step": 26039 + }, + { + "epoch": 5.68, + "learning_rate": 2.1608040201005025e-06, + "loss": 0.9529, + "step": 26040 + }, + { + "epoch": 5.68, + "learning_rate": 2.1606947782390215e-06, + "loss": 1.5708, + "step": 26041 + }, + { + "epoch": 5.68, + "learning_rate": 2.16058553637754e-06, + "loss": 1.8763, + "step": 26042 + }, + { + "epoch": 5.68, + "learning_rate": 2.1604762945160587e-06, + "loss": 0.7323, + "step": 26043 + }, + { + "epoch": 5.68, + "learning_rate": 2.1603670526545773e-06, + "loss": 1.606, + "step": 26044 + }, + { + "epoch": 5.68, + "learning_rate": 2.160257810793096e-06, + "loss": 1.2634, + "step": 26045 + }, + { + "epoch": 5.68, + "learning_rate": 2.160148568931615e-06, + "loss": 1.7502, + "step": 26046 + }, + { + "epoch": 5.68, + "learning_rate": 2.160039327070133e-06, + "loss": 0.8328, + "step": 26047 + }, + { + "epoch": 5.68, + "learning_rate": 2.159930085208652e-06, + "loss": 1.3507, + "step": 26048 + }, + { + "epoch": 5.69, + "learning_rate": 2.159820843347171e-06, + "loss": 2.6949, + "step": 26049 + }, + { + "epoch": 5.69, + "learning_rate": 2.1597116014856894e-06, + "loss": 1.7243, + "step": 26050 + }, + { + "epoch": 5.69, + "learning_rate": 2.1596023596242085e-06, + "loss": 2.3944, + "step": 26051 + }, + { + "epoch": 5.69, + "learning_rate": 2.1594931177627266e-06, + "loss": 1.4, + "step": 26052 + }, + { + "epoch": 5.69, + "learning_rate": 2.1593838759012457e-06, + "loss": 0.3744, + "step": 26053 + }, + { + "epoch": 5.69, + "learning_rate": 2.1592746340397643e-06, + "loss": 2.2736, + "step": 26054 + }, + { + "epoch": 5.69, + "learning_rate": 2.159165392178283e-06, + "loss": 2.1644, + "step": 26055 + }, + { + "epoch": 5.69, + "learning_rate": 2.1590561503168015e-06, + "loss": 1.9763, + "step": 26056 + }, + { + "epoch": 5.69, + "learning_rate": 2.15894690845532e-06, + "loss": 1.6215, + "step": 26057 + }, + { + "epoch": 5.69, + "learning_rate": 2.158837666593839e-06, + "loss": 1.7713, + "step": 26058 + }, + { + "epoch": 5.69, + "learning_rate": 2.1587284247323573e-06, + "loss": 1.0328, + "step": 26059 + }, + { + "epoch": 5.69, + "learning_rate": 2.1586191828708764e-06, + "loss": 0.8926, + "step": 26060 + }, + { + "epoch": 5.69, + "learning_rate": 2.158509941009395e-06, + "loss": 0.3474, + "step": 26061 + }, + { + "epoch": 5.69, + "learning_rate": 2.1584006991479136e-06, + "loss": 1.2705, + "step": 26062 + }, + { + "epoch": 5.69, + "learning_rate": 2.1582914572864326e-06, + "loss": 0.5995, + "step": 26063 + }, + { + "epoch": 5.69, + "learning_rate": 2.158182215424951e-06, + "loss": 2.0028, + "step": 26064 + }, + { + "epoch": 5.69, + "learning_rate": 2.15807297356347e-06, + "loss": 1.5514, + "step": 26065 + }, + { + "epoch": 5.69, + "learning_rate": 2.1579637317019885e-06, + "loss": 1.1132, + "step": 26066 + }, + { + "epoch": 5.69, + "learning_rate": 2.157854489840507e-06, + "loss": 1.4353, + "step": 26067 + }, + { + "epoch": 5.69, + "learning_rate": 2.1577452479790257e-06, + "loss": 2.3202, + "step": 26068 + }, + { + "epoch": 5.69, + "learning_rate": 2.1576360061175443e-06, + "loss": 0.7776, + "step": 26069 + }, + { + "epoch": 5.69, + "learning_rate": 2.1575267642560633e-06, + "loss": 1.3204, + "step": 26070 + }, + { + "epoch": 5.69, + "learning_rate": 2.157417522394582e-06, + "loss": 1.0516, + "step": 26071 + }, + { + "epoch": 5.69, + "learning_rate": 2.1573082805331006e-06, + "loss": 0.6218, + "step": 26072 + }, + { + "epoch": 5.69, + "learning_rate": 2.157199038671619e-06, + "loss": 1.5001, + "step": 26073 + }, + { + "epoch": 5.69, + "learning_rate": 2.1570897968101378e-06, + "loss": 0.9974, + "step": 26074 + }, + { + "epoch": 5.69, + "learning_rate": 2.1569805549486564e-06, + "loss": 1.9019, + "step": 26075 + }, + { + "epoch": 5.69, + "learning_rate": 2.156871313087175e-06, + "loss": 0.4349, + "step": 26076 + }, + { + "epoch": 5.69, + "learning_rate": 2.156762071225694e-06, + "loss": 2.1117, + "step": 26077 + }, + { + "epoch": 5.69, + "learning_rate": 2.1566528293642127e-06, + "loss": 1.2961, + "step": 26078 + }, + { + "epoch": 5.69, + "learning_rate": 2.1565435875027313e-06, + "loss": 1.5399, + "step": 26079 + }, + { + "epoch": 5.69, + "learning_rate": 2.15643434564125e-06, + "loss": 0.7934, + "step": 26080 + }, + { + "epoch": 5.69, + "learning_rate": 2.1563251037797685e-06, + "loss": 2.6165, + "step": 26081 + }, + { + "epoch": 5.69, + "learning_rate": 2.156215861918287e-06, + "loss": 3.498, + "step": 26082 + }, + { + "epoch": 5.69, + "learning_rate": 2.156106620056806e-06, + "loss": 2.2821, + "step": 26083 + }, + { + "epoch": 5.69, + "learning_rate": 2.1559973781953247e-06, + "loss": 2.9621, + "step": 26084 + }, + { + "epoch": 5.69, + "learning_rate": 2.1558881363338434e-06, + "loss": 0.3031, + "step": 26085 + }, + { + "epoch": 5.69, + "learning_rate": 2.155778894472362e-06, + "loss": 1.143, + "step": 26086 + }, + { + "epoch": 5.69, + "learning_rate": 2.1556696526108806e-06, + "loss": 2.3824, + "step": 26087 + }, + { + "epoch": 5.69, + "learning_rate": 2.155560410749399e-06, + "loss": 1.1886, + "step": 26088 + }, + { + "epoch": 5.69, + "learning_rate": 2.155451168887918e-06, + "loss": 3.0414, + "step": 26089 + }, + { + "epoch": 5.69, + "learning_rate": 2.155341927026437e-06, + "loss": 1.9947, + "step": 26090 + }, + { + "epoch": 5.69, + "learning_rate": 2.1552326851649554e-06, + "loss": 0.7169, + "step": 26091 + }, + { + "epoch": 5.69, + "learning_rate": 2.155123443303474e-06, + "loss": 1.664, + "step": 26092 + }, + { + "epoch": 5.69, + "learning_rate": 2.1550142014419927e-06, + "loss": 0.1481, + "step": 26093 + }, + { + "epoch": 5.69, + "learning_rate": 2.1549049595805113e-06, + "loss": 1.6134, + "step": 26094 + }, + { + "epoch": 5.7, + "learning_rate": 2.1547957177190303e-06, + "loss": 0.8395, + "step": 26095 + }, + { + "epoch": 5.7, + "learning_rate": 2.154686475857549e-06, + "loss": 1.1834, + "step": 26096 + }, + { + "epoch": 5.7, + "learning_rate": 2.1545772339960675e-06, + "loss": 0.8682, + "step": 26097 + }, + { + "epoch": 5.7, + "learning_rate": 2.154467992134586e-06, + "loss": 1.3147, + "step": 26098 + }, + { + "epoch": 5.7, + "learning_rate": 2.1543587502731048e-06, + "loss": 0.1967, + "step": 26099 + }, + { + "epoch": 5.7, + "learning_rate": 2.1542495084116234e-06, + "loss": 2.2259, + "step": 26100 + }, + { + "epoch": 5.7, + "learning_rate": 2.154140266550142e-06, + "loss": 1.2768, + "step": 26101 + }, + { + "epoch": 5.7, + "learning_rate": 2.154031024688661e-06, + "loss": 1.349, + "step": 26102 + }, + { + "epoch": 5.7, + "learning_rate": 2.1539217828271796e-06, + "loss": 3.587, + "step": 26103 + }, + { + "epoch": 5.7, + "learning_rate": 2.1538125409656982e-06, + "loss": 1.548, + "step": 26104 + }, + { + "epoch": 5.7, + "learning_rate": 2.153703299104217e-06, + "loss": 2.326, + "step": 26105 + }, + { + "epoch": 5.7, + "learning_rate": 2.1535940572427355e-06, + "loss": 1.4737, + "step": 26106 + }, + { + "epoch": 5.7, + "learning_rate": 2.1534848153812545e-06, + "loss": 0.3403, + "step": 26107 + }, + { + "epoch": 5.7, + "learning_rate": 2.1533755735197727e-06, + "loss": 2.9393, + "step": 26108 + }, + { + "epoch": 5.7, + "learning_rate": 2.1532663316582917e-06, + "loss": 1.343, + "step": 26109 + }, + { + "epoch": 5.7, + "learning_rate": 2.1531570897968103e-06, + "loss": 0.5845, + "step": 26110 + }, + { + "epoch": 5.7, + "learning_rate": 2.153047847935329e-06, + "loss": 1.5383, + "step": 26111 + }, + { + "epoch": 5.7, + "learning_rate": 2.1529386060738475e-06, + "loss": 0.3912, + "step": 26112 + }, + { + "epoch": 5.7, + "learning_rate": 2.152829364212366e-06, + "loss": 2.5565, + "step": 26113 + }, + { + "epoch": 5.7, + "learning_rate": 2.152720122350885e-06, + "loss": 0.1785, + "step": 26114 + }, + { + "epoch": 5.7, + "learning_rate": 2.152610880489404e-06, + "loss": 1.806, + "step": 26115 + }, + { + "epoch": 5.7, + "learning_rate": 2.1525016386279224e-06, + "loss": 3.8742, + "step": 26116 + }, + { + "epoch": 5.7, + "learning_rate": 2.152392396766441e-06, + "loss": 2.1656, + "step": 26117 + }, + { + "epoch": 5.7, + "learning_rate": 2.1522831549049596e-06, + "loss": 1.7023, + "step": 26118 + }, + { + "epoch": 5.7, + "learning_rate": 2.1521739130434787e-06, + "loss": 0.9464, + "step": 26119 + }, + { + "epoch": 5.7, + "learning_rate": 2.152064671181997e-06, + "loss": 1.806, + "step": 26120 + }, + { + "epoch": 5.7, + "learning_rate": 2.151955429320516e-06, + "loss": 0.9398, + "step": 26121 + }, + { + "epoch": 5.7, + "learning_rate": 2.1518461874590345e-06, + "loss": 0.6591, + "step": 26122 + }, + { + "epoch": 5.7, + "learning_rate": 2.151736945597553e-06, + "loss": 1.1522, + "step": 26123 + }, + { + "epoch": 5.7, + "learning_rate": 2.151627703736072e-06, + "loss": 0.9256, + "step": 26124 + }, + { + "epoch": 5.7, + "learning_rate": 2.1515184618745903e-06, + "loss": 1.0935, + "step": 26125 + }, + { + "epoch": 5.7, + "learning_rate": 2.1514092200131094e-06, + "loss": 0.7835, + "step": 26126 + }, + { + "epoch": 5.7, + "learning_rate": 2.151299978151628e-06, + "loss": 0.6905, + "step": 26127 + }, + { + "epoch": 5.7, + "learning_rate": 2.1511907362901466e-06, + "loss": 2.1688, + "step": 26128 + }, + { + "epoch": 5.7, + "learning_rate": 2.151081494428665e-06, + "loss": 1.0377, + "step": 26129 + }, + { + "epoch": 5.7, + "learning_rate": 2.150972252567184e-06, + "loss": 0.8473, + "step": 26130 + }, + { + "epoch": 5.7, + "learning_rate": 2.150863010705703e-06, + "loss": 0.4435, + "step": 26131 + }, + { + "epoch": 5.7, + "learning_rate": 2.150753768844221e-06, + "loss": 1.9685, + "step": 26132 + }, + { + "epoch": 5.7, + "learning_rate": 2.15064452698274e-06, + "loss": 1.2699, + "step": 26133 + }, + { + "epoch": 5.7, + "learning_rate": 2.1505352851212587e-06, + "loss": 1.5612, + "step": 26134 + }, + { + "epoch": 5.7, + "learning_rate": 2.1504260432597773e-06, + "loss": 1.8914, + "step": 26135 + }, + { + "epoch": 5.7, + "learning_rate": 2.1503168013982963e-06, + "loss": 1.5604, + "step": 26136 + }, + { + "epoch": 5.7, + "learning_rate": 2.1502075595368145e-06, + "loss": 2.9837, + "step": 26137 + }, + { + "epoch": 5.7, + "learning_rate": 2.1500983176753335e-06, + "loss": 2.3563, + "step": 26138 + }, + { + "epoch": 5.7, + "learning_rate": 2.149989075813852e-06, + "loss": 0.8076, + "step": 26139 + }, + { + "epoch": 5.7, + "learning_rate": 2.1498798339523708e-06, + "loss": 0.4705, + "step": 26140 + }, + { + "epoch": 5.71, + "learning_rate": 2.1497705920908894e-06, + "loss": 0.3704, + "step": 26141 + }, + { + "epoch": 5.71, + "learning_rate": 2.149661350229408e-06, + "loss": 0.4704, + "step": 26142 + }, + { + "epoch": 5.71, + "learning_rate": 2.149552108367927e-06, + "loss": 0.4666, + "step": 26143 + }, + { + "epoch": 5.71, + "learning_rate": 2.149442866506445e-06, + "loss": 1.2908, + "step": 26144 + }, + { + "epoch": 5.71, + "learning_rate": 2.1493336246449642e-06, + "loss": 0.5914, + "step": 26145 + }, + { + "epoch": 5.71, + "learning_rate": 2.149224382783483e-06, + "loss": 1.2039, + "step": 26146 + }, + { + "epoch": 5.71, + "learning_rate": 2.1491151409220015e-06, + "loss": 1.0457, + "step": 26147 + }, + { + "epoch": 5.71, + "learning_rate": 2.14900589906052e-06, + "loss": 2.1409, + "step": 26148 + }, + { + "epoch": 5.71, + "learning_rate": 2.1488966571990387e-06, + "loss": 3.1208, + "step": 26149 + }, + { + "epoch": 5.71, + "learning_rate": 2.1487874153375577e-06, + "loss": 1.6978, + "step": 26150 + }, + { + "epoch": 5.71, + "learning_rate": 2.1486781734760763e-06, + "loss": 2.7597, + "step": 26151 + }, + { + "epoch": 5.71, + "learning_rate": 2.148568931614595e-06, + "loss": 0.4069, + "step": 26152 + }, + { + "epoch": 5.71, + "learning_rate": 2.1484596897531136e-06, + "loss": 2.6414, + "step": 26153 + }, + { + "epoch": 5.71, + "learning_rate": 2.148350447891632e-06, + "loss": 0.9435, + "step": 26154 + }, + { + "epoch": 5.71, + "learning_rate": 2.1482412060301508e-06, + "loss": 2.3367, + "step": 26155 + }, + { + "epoch": 5.71, + "learning_rate": 2.1481319641686694e-06, + "loss": 1.5328, + "step": 26156 + }, + { + "epoch": 5.71, + "learning_rate": 2.1480227223071884e-06, + "loss": 1.6402, + "step": 26157 + }, + { + "epoch": 5.71, + "learning_rate": 2.147913480445707e-06, + "loss": 1.7347, + "step": 26158 + }, + { + "epoch": 5.71, + "learning_rate": 2.1478042385842256e-06, + "loss": 0.8343, + "step": 26159 + }, + { + "epoch": 5.71, + "learning_rate": 2.1476949967227443e-06, + "loss": 1.2589, + "step": 26160 + }, + { + "epoch": 5.71, + "learning_rate": 2.147585754861263e-06, + "loss": 2.6347, + "step": 26161 + }, + { + "epoch": 5.71, + "learning_rate": 2.147476512999782e-06, + "loss": 2.4722, + "step": 26162 + }, + { + "epoch": 5.71, + "learning_rate": 2.1473672711383005e-06, + "loss": 2.177, + "step": 26163 + }, + { + "epoch": 5.71, + "learning_rate": 2.147258029276819e-06, + "loss": 1.9417, + "step": 26164 + }, + { + "epoch": 5.71, + "learning_rate": 2.1471487874153377e-06, + "loss": 0.2317, + "step": 26165 + }, + { + "epoch": 5.71, + "learning_rate": 2.1470395455538563e-06, + "loss": 1.0824, + "step": 26166 + }, + { + "epoch": 5.71, + "learning_rate": 2.146930303692375e-06, + "loss": 3.0108, + "step": 26167 + }, + { + "epoch": 5.71, + "learning_rate": 2.146821061830894e-06, + "loss": 3.0187, + "step": 26168 + }, + { + "epoch": 5.71, + "learning_rate": 2.1467118199694126e-06, + "loss": 2.8835, + "step": 26169 + }, + { + "epoch": 5.71, + "learning_rate": 2.1466025781079312e-06, + "loss": 1.4429, + "step": 26170 + }, + { + "epoch": 5.71, + "learning_rate": 2.14649333624645e-06, + "loss": 1.7321, + "step": 26171 + }, + { + "epoch": 5.71, + "learning_rate": 2.1463840943849684e-06, + "loss": 1.79, + "step": 26172 + }, + { + "epoch": 5.71, + "learning_rate": 2.146274852523487e-06, + "loss": 1.934, + "step": 26173 + }, + { + "epoch": 5.71, + "learning_rate": 2.1461656106620057e-06, + "loss": 1.7933, + "step": 26174 + }, + { + "epoch": 5.71, + "learning_rate": 2.1460563688005247e-06, + "loss": 2.2733, + "step": 26175 + }, + { + "epoch": 5.71, + "learning_rate": 2.1459471269390433e-06, + "loss": 0.4172, + "step": 26176 + }, + { + "epoch": 5.71, + "learning_rate": 2.145837885077562e-06, + "loss": 1.3339, + "step": 26177 + }, + { + "epoch": 5.71, + "learning_rate": 2.1457286432160805e-06, + "loss": 1.025, + "step": 26178 + }, + { + "epoch": 5.71, + "learning_rate": 2.145619401354599e-06, + "loss": 1.5314, + "step": 26179 + }, + { + "epoch": 5.71, + "learning_rate": 2.145510159493118e-06, + "loss": 1.3045, + "step": 26180 + }, + { + "epoch": 5.71, + "learning_rate": 2.1454009176316364e-06, + "loss": 0.8555, + "step": 26181 + }, + { + "epoch": 5.71, + "learning_rate": 2.1452916757701554e-06, + "loss": 1.0544, + "step": 26182 + }, + { + "epoch": 5.71, + "learning_rate": 2.145182433908674e-06, + "loss": 1.2218, + "step": 26183 + }, + { + "epoch": 5.71, + "learning_rate": 2.1450731920471926e-06, + "loss": 1.2683, + "step": 26184 + }, + { + "epoch": 5.71, + "learning_rate": 2.1449639501857112e-06, + "loss": 1.5116, + "step": 26185 + }, + { + "epoch": 5.71, + "learning_rate": 2.14485470832423e-06, + "loss": 2.2467, + "step": 26186 + }, + { + "epoch": 5.72, + "learning_rate": 2.144745466462749e-06, + "loss": 1.0825, + "step": 26187 + }, + { + "epoch": 5.72, + "learning_rate": 2.144636224601267e-06, + "loss": 0.8573, + "step": 26188 + }, + { + "epoch": 5.72, + "learning_rate": 2.144526982739786e-06, + "loss": 1.2784, + "step": 26189 + }, + { + "epoch": 5.72, + "learning_rate": 2.1444177408783047e-06, + "loss": 1.2391, + "step": 26190 + }, + { + "epoch": 5.72, + "learning_rate": 2.1443084990168233e-06, + "loss": 1.7257, + "step": 26191 + }, + { + "epoch": 5.72, + "learning_rate": 2.1441992571553423e-06, + "loss": 1.0392, + "step": 26192 + }, + { + "epoch": 5.72, + "learning_rate": 2.1440900152938605e-06, + "loss": 1.3206, + "step": 26193 + }, + { + "epoch": 5.72, + "learning_rate": 2.1439807734323796e-06, + "loss": 2.4979, + "step": 26194 + }, + { + "epoch": 5.72, + "learning_rate": 2.143871531570898e-06, + "loss": 1.1265, + "step": 26195 + }, + { + "epoch": 5.72, + "learning_rate": 2.143762289709417e-06, + "loss": 1.9252, + "step": 26196 + }, + { + "epoch": 5.72, + "learning_rate": 2.1436530478479354e-06, + "loss": 1.1372, + "step": 26197 + }, + { + "epoch": 5.72, + "learning_rate": 2.143543805986454e-06, + "loss": 1.2709, + "step": 26198 + }, + { + "epoch": 5.72, + "learning_rate": 2.143434564124973e-06, + "loss": 1.0139, + "step": 26199 + }, + { + "epoch": 5.72, + "learning_rate": 2.1433253222634912e-06, + "loss": 0.6992, + "step": 26200 + }, + { + "epoch": 5.72, + "learning_rate": 2.1432160804020103e-06, + "loss": 2.0407, + "step": 26201 + }, + { + "epoch": 5.72, + "learning_rate": 2.143106838540529e-06, + "loss": 2.4788, + "step": 26202 + }, + { + "epoch": 5.72, + "learning_rate": 2.1429975966790475e-06, + "loss": 1.6368, + "step": 26203 + }, + { + "epoch": 5.72, + "learning_rate": 2.1428883548175665e-06, + "loss": 1.2879, + "step": 26204 + }, + { + "epoch": 5.72, + "learning_rate": 2.1427791129560847e-06, + "loss": 0.8897, + "step": 26205 + }, + { + "epoch": 5.72, + "learning_rate": 2.1426698710946038e-06, + "loss": 2.9373, + "step": 26206 + }, + { + "epoch": 5.72, + "learning_rate": 2.1425606292331224e-06, + "loss": 1.2549, + "step": 26207 + }, + { + "epoch": 5.72, + "learning_rate": 2.142451387371641e-06, + "loss": 1.559, + "step": 26208 + }, + { + "epoch": 5.72, + "learning_rate": 2.1423421455101596e-06, + "loss": 1.4736, + "step": 26209 + }, + { + "epoch": 5.72, + "learning_rate": 2.142232903648678e-06, + "loss": 1.1523, + "step": 26210 + }, + { + "epoch": 5.72, + "learning_rate": 2.1421236617871972e-06, + "loss": 1.1176, + "step": 26211 + }, + { + "epoch": 5.72, + "learning_rate": 2.1420144199257154e-06, + "loss": 0.3632, + "step": 26212 + }, + { + "epoch": 5.72, + "learning_rate": 2.1419051780642345e-06, + "loss": 1.0571, + "step": 26213 + }, + { + "epoch": 5.72, + "learning_rate": 2.141795936202753e-06, + "loss": 2.2953, + "step": 26214 + }, + { + "epoch": 5.72, + "learning_rate": 2.1416866943412717e-06, + "loss": 1.4078, + "step": 26215 + }, + { + "epoch": 5.72, + "learning_rate": 2.1415774524797907e-06, + "loss": 1.2298, + "step": 26216 + }, + { + "epoch": 5.72, + "learning_rate": 2.141468210618309e-06, + "loss": 1.9097, + "step": 26217 + }, + { + "epoch": 5.72, + "learning_rate": 2.141358968756828e-06, + "loss": 0.7416, + "step": 26218 + }, + { + "epoch": 5.72, + "learning_rate": 2.1412497268953465e-06, + "loss": 0.6124, + "step": 26219 + }, + { + "epoch": 5.72, + "learning_rate": 2.141140485033865e-06, + "loss": 1.4099, + "step": 26220 + }, + { + "epoch": 5.72, + "learning_rate": 2.1410312431723838e-06, + "loss": 2.8278, + "step": 26221 + }, + { + "epoch": 5.72, + "learning_rate": 2.1409220013109024e-06, + "loss": 0.9721, + "step": 26222 + }, + { + "epoch": 5.72, + "learning_rate": 2.1408127594494214e-06, + "loss": 0.2274, + "step": 26223 + }, + { + "epoch": 5.72, + "learning_rate": 2.14070351758794e-06, + "loss": 2.0407, + "step": 26224 + }, + { + "epoch": 5.72, + "learning_rate": 2.1405942757264586e-06, + "loss": 1.8149, + "step": 26225 + }, + { + "epoch": 5.72, + "learning_rate": 2.1404850338649772e-06, + "loss": 1.4889, + "step": 26226 + }, + { + "epoch": 5.72, + "learning_rate": 2.140375792003496e-06, + "loss": 1.8486, + "step": 26227 + }, + { + "epoch": 5.72, + "learning_rate": 2.140266550142015e-06, + "loss": 0.6061, + "step": 26228 + }, + { + "epoch": 5.72, + "learning_rate": 2.140157308280533e-06, + "loss": 0.5815, + "step": 26229 + }, + { + "epoch": 5.72, + "learning_rate": 2.140048066419052e-06, + "loss": 0.7899, + "step": 26230 + }, + { + "epoch": 5.72, + "learning_rate": 2.1399388245575707e-06, + "loss": 0.9261, + "step": 26231 + }, + { + "epoch": 5.73, + "learning_rate": 2.1398295826960893e-06, + "loss": 2.0099, + "step": 26232 + }, + { + "epoch": 5.73, + "learning_rate": 2.139720340834608e-06, + "loss": 1.6481, + "step": 26233 + }, + { + "epoch": 5.73, + "learning_rate": 2.1396110989731266e-06, + "loss": 1.0216, + "step": 26234 + }, + { + "epoch": 5.73, + "learning_rate": 2.1395018571116456e-06, + "loss": 0.1642, + "step": 26235 + }, + { + "epoch": 5.73, + "learning_rate": 2.139392615250164e-06, + "loss": 2.025, + "step": 26236 + }, + { + "epoch": 5.73, + "learning_rate": 2.139283373388683e-06, + "loss": 0.2664, + "step": 26237 + }, + { + "epoch": 5.73, + "learning_rate": 2.1391741315272014e-06, + "loss": 3.0032, + "step": 26238 + }, + { + "epoch": 5.73, + "learning_rate": 2.13906488966572e-06, + "loss": 3.8105, + "step": 26239 + }, + { + "epoch": 5.73, + "learning_rate": 2.1389556478042386e-06, + "loss": 1.1011, + "step": 26240 + }, + { + "epoch": 5.73, + "learning_rate": 2.1388464059427573e-06, + "loss": 0.8291, + "step": 26241 + }, + { + "epoch": 5.73, + "learning_rate": 2.1387371640812763e-06, + "loss": 1.2375, + "step": 26242 + }, + { + "epoch": 5.73, + "learning_rate": 2.138627922219795e-06, + "loss": 1.9085, + "step": 26243 + }, + { + "epoch": 5.73, + "learning_rate": 2.1385186803583135e-06, + "loss": 0.5376, + "step": 26244 + }, + { + "epoch": 5.73, + "learning_rate": 2.138409438496832e-06, + "loss": 2.2134, + "step": 26245 + }, + { + "epoch": 5.73, + "learning_rate": 2.1383001966353507e-06, + "loss": 1.6276, + "step": 26246 + }, + { + "epoch": 5.73, + "learning_rate": 2.1381909547738693e-06, + "loss": 2.1453, + "step": 26247 + }, + { + "epoch": 5.73, + "learning_rate": 2.1380817129123884e-06, + "loss": 1.6552, + "step": 26248 + }, + { + "epoch": 5.73, + "learning_rate": 2.137972471050907e-06, + "loss": 2.3544, + "step": 26249 + }, + { + "epoch": 5.73, + "learning_rate": 2.1378632291894256e-06, + "loss": 0.5803, + "step": 26250 + }, + { + "epoch": 5.73, + "learning_rate": 2.137753987327944e-06, + "loss": 0.9109, + "step": 26251 + }, + { + "epoch": 5.73, + "learning_rate": 2.137644745466463e-06, + "loss": 0.8321, + "step": 26252 + }, + { + "epoch": 5.73, + "learning_rate": 2.1375355036049814e-06, + "loss": 0.1964, + "step": 26253 + }, + { + "epoch": 5.73, + "learning_rate": 2.1374262617435e-06, + "loss": 0.1548, + "step": 26254 + }, + { + "epoch": 5.73, + "learning_rate": 2.137317019882019e-06, + "loss": 0.2741, + "step": 26255 + }, + { + "epoch": 5.73, + "learning_rate": 2.1372077780205377e-06, + "loss": 1.8662, + "step": 26256 + }, + { + "epoch": 5.73, + "learning_rate": 2.1370985361590563e-06, + "loss": 2.3586, + "step": 26257 + }, + { + "epoch": 5.73, + "learning_rate": 2.136989294297575e-06, + "loss": 0.1203, + "step": 26258 + }, + { + "epoch": 5.73, + "learning_rate": 2.1368800524360935e-06, + "loss": 1.7669, + "step": 26259 + }, + { + "epoch": 5.73, + "learning_rate": 2.1367708105746126e-06, + "loss": 0.9525, + "step": 26260 + }, + { + "epoch": 5.73, + "learning_rate": 2.136661568713131e-06, + "loss": 1.9033, + "step": 26261 + }, + { + "epoch": 5.73, + "learning_rate": 2.1365523268516498e-06, + "loss": 0.1581, + "step": 26262 + }, + { + "epoch": 5.73, + "learning_rate": 2.1364430849901684e-06, + "loss": 1.3637, + "step": 26263 + }, + { + "epoch": 5.73, + "learning_rate": 2.136333843128687e-06, + "loss": 0.9231, + "step": 26264 + }, + { + "epoch": 5.73, + "learning_rate": 2.1362246012672056e-06, + "loss": 2.0933, + "step": 26265 + }, + { + "epoch": 5.73, + "learning_rate": 2.1361153594057242e-06, + "loss": 0.9434, + "step": 26266 + }, + { + "epoch": 5.73, + "learning_rate": 2.1360061175442433e-06, + "loss": 2.4756, + "step": 26267 + }, + { + "epoch": 5.73, + "learning_rate": 2.135896875682762e-06, + "loss": 2.116, + "step": 26268 + }, + { + "epoch": 5.73, + "learning_rate": 2.1357876338212805e-06, + "loss": 0.2638, + "step": 26269 + }, + { + "epoch": 5.73, + "learning_rate": 2.135678391959799e-06, + "loss": 1.9936, + "step": 26270 + }, + { + "epoch": 5.73, + "learning_rate": 2.1355691500983177e-06, + "loss": 3.3992, + "step": 26271 + }, + { + "epoch": 5.73, + "learning_rate": 2.1354599082368367e-06, + "loss": 2.2577, + "step": 26272 + }, + { + "epoch": 5.73, + "learning_rate": 2.135350666375355e-06, + "loss": 1.6947, + "step": 26273 + }, + { + "epoch": 5.73, + "learning_rate": 2.135241424513874e-06, + "loss": 1.1946, + "step": 26274 + }, + { + "epoch": 5.73, + "learning_rate": 2.1351321826523926e-06, + "loss": 1.85, + "step": 26275 + }, + { + "epoch": 5.73, + "learning_rate": 2.135022940790911e-06, + "loss": 1.4503, + "step": 26276 + }, + { + "epoch": 5.73, + "learning_rate": 2.1349136989294302e-06, + "loss": 0.5573, + "step": 26277 + }, + { + "epoch": 5.74, + "learning_rate": 2.1348044570679484e-06, + "loss": 0.6241, + "step": 26278 + }, + { + "epoch": 5.74, + "learning_rate": 2.1346952152064674e-06, + "loss": 0.6296, + "step": 26279 + }, + { + "epoch": 5.74, + "learning_rate": 2.134585973344986e-06, + "loss": 2.2154, + "step": 26280 + }, + { + "epoch": 5.74, + "learning_rate": 2.1344767314835047e-06, + "loss": 0.4993, + "step": 26281 + }, + { + "epoch": 5.74, + "learning_rate": 2.1343674896220233e-06, + "loss": 1.0286, + "step": 26282 + }, + { + "epoch": 5.74, + "learning_rate": 2.134258247760542e-06, + "loss": 4.3032, + "step": 26283 + }, + { + "epoch": 5.74, + "learning_rate": 2.134149005899061e-06, + "loss": 1.8606, + "step": 26284 + }, + { + "epoch": 5.74, + "learning_rate": 2.134039764037579e-06, + "loss": 0.2015, + "step": 26285 + }, + { + "epoch": 5.74, + "learning_rate": 2.133930522176098e-06, + "loss": 1.7109, + "step": 26286 + }, + { + "epoch": 5.74, + "learning_rate": 2.1338212803146167e-06, + "loss": 2.5794, + "step": 26287 + }, + { + "epoch": 5.74, + "learning_rate": 2.1337120384531354e-06, + "loss": 1.6247, + "step": 26288 + }, + { + "epoch": 5.74, + "learning_rate": 2.1336027965916544e-06, + "loss": 2.4191, + "step": 26289 + }, + { + "epoch": 5.74, + "learning_rate": 2.1334935547301726e-06, + "loss": 1.2397, + "step": 26290 + }, + { + "epoch": 5.74, + "learning_rate": 2.1333843128686916e-06, + "loss": 1.8996, + "step": 26291 + }, + { + "epoch": 5.74, + "learning_rate": 2.1332750710072102e-06, + "loss": 2.2027, + "step": 26292 + }, + { + "epoch": 5.74, + "learning_rate": 2.133165829145729e-06, + "loss": 0.823, + "step": 26293 + }, + { + "epoch": 5.74, + "learning_rate": 2.1330565872842474e-06, + "loss": 1.8, + "step": 26294 + }, + { + "epoch": 5.74, + "learning_rate": 2.132947345422766e-06, + "loss": 1.0911, + "step": 26295 + }, + { + "epoch": 5.74, + "learning_rate": 2.132838103561285e-06, + "loss": 1.7155, + "step": 26296 + }, + { + "epoch": 5.74, + "learning_rate": 2.1327288616998033e-06, + "loss": 1.8273, + "step": 26297 + }, + { + "epoch": 5.74, + "learning_rate": 2.1326196198383223e-06, + "loss": 0.6561, + "step": 26298 + }, + { + "epoch": 5.74, + "learning_rate": 2.132510377976841e-06, + "loss": 0.9251, + "step": 26299 + }, + { + "epoch": 5.74, + "learning_rate": 2.1324011361153595e-06, + "loss": 2.2263, + "step": 26300 + }, + { + "epoch": 5.74, + "learning_rate": 2.1322918942538786e-06, + "loss": 1.8192, + "step": 26301 + }, + { + "epoch": 5.74, + "learning_rate": 2.1321826523923968e-06, + "loss": 2.0213, + "step": 26302 + }, + { + "epoch": 5.74, + "learning_rate": 2.132073410530916e-06, + "loss": 3.1294, + "step": 26303 + }, + { + "epoch": 5.74, + "learning_rate": 2.1319641686694344e-06, + "loss": 0.3617, + "step": 26304 + }, + { + "epoch": 5.74, + "learning_rate": 2.131854926807953e-06, + "loss": 1.5157, + "step": 26305 + }, + { + "epoch": 5.74, + "learning_rate": 2.1317456849464716e-06, + "loss": 2.3033, + "step": 26306 + }, + { + "epoch": 5.74, + "learning_rate": 2.1316364430849902e-06, + "loss": 0.806, + "step": 26307 + }, + { + "epoch": 5.74, + "learning_rate": 2.1315272012235093e-06, + "loss": 2.9422, + "step": 26308 + }, + { + "epoch": 5.74, + "learning_rate": 2.1314179593620275e-06, + "loss": 1.3924, + "step": 26309 + }, + { + "epoch": 5.74, + "learning_rate": 2.1313087175005465e-06, + "loss": 1.1505, + "step": 26310 + }, + { + "epoch": 5.74, + "learning_rate": 2.131199475639065e-06, + "loss": 1.5213, + "step": 26311 + }, + { + "epoch": 5.74, + "learning_rate": 2.1310902337775837e-06, + "loss": 2.8015, + "step": 26312 + }, + { + "epoch": 5.74, + "learning_rate": 2.1309809919161023e-06, + "loss": 2.2453, + "step": 26313 + }, + { + "epoch": 5.74, + "learning_rate": 2.130871750054621e-06, + "loss": 1.1467, + "step": 26314 + }, + { + "epoch": 5.74, + "learning_rate": 2.13076250819314e-06, + "loss": 1.2992, + "step": 26315 + }, + { + "epoch": 5.74, + "learning_rate": 2.1306532663316586e-06, + "loss": 2.2567, + "step": 26316 + }, + { + "epoch": 5.74, + "learning_rate": 2.130544024470177e-06, + "loss": 0.377, + "step": 26317 + }, + { + "epoch": 5.74, + "learning_rate": 2.130434782608696e-06, + "loss": 2.3022, + "step": 26318 + }, + { + "epoch": 5.74, + "learning_rate": 2.1303255407472144e-06, + "loss": 0.4564, + "step": 26319 + }, + { + "epoch": 5.74, + "learning_rate": 2.130216298885733e-06, + "loss": 0.9578, + "step": 26320 + }, + { + "epoch": 5.74, + "learning_rate": 2.130107057024252e-06, + "loss": 1.933, + "step": 26321 + }, + { + "epoch": 5.74, + "learning_rate": 2.1299978151627707e-06, + "loss": 1.6992, + "step": 26322 + }, + { + "epoch": 5.74, + "learning_rate": 2.1298885733012893e-06, + "loss": 1.233, + "step": 26323 + }, + { + "epoch": 5.75, + "learning_rate": 2.129779331439808e-06, + "loss": 1.4789, + "step": 26324 + }, + { + "epoch": 5.75, + "learning_rate": 2.1296700895783265e-06, + "loss": 3.0073, + "step": 26325 + }, + { + "epoch": 5.75, + "learning_rate": 2.129560847716845e-06, + "loss": 2.2439, + "step": 26326 + }, + { + "epoch": 5.75, + "learning_rate": 2.1294516058553637e-06, + "loss": 0.1624, + "step": 26327 + }, + { + "epoch": 5.75, + "learning_rate": 2.1293423639938828e-06, + "loss": 0.561, + "step": 26328 + }, + { + "epoch": 5.75, + "learning_rate": 2.1292331221324014e-06, + "loss": 1.1918, + "step": 26329 + }, + { + "epoch": 5.75, + "learning_rate": 2.12912388027092e-06, + "loss": 2.3175, + "step": 26330 + }, + { + "epoch": 5.75, + "learning_rate": 2.1290146384094386e-06, + "loss": 1.6185, + "step": 26331 + }, + { + "epoch": 5.75, + "learning_rate": 2.128905396547957e-06, + "loss": 0.6983, + "step": 26332 + }, + { + "epoch": 5.75, + "learning_rate": 2.1287961546864762e-06, + "loss": 1.0126, + "step": 26333 + }, + { + "epoch": 5.75, + "learning_rate": 2.128686912824995e-06, + "loss": 1.5971, + "step": 26334 + }, + { + "epoch": 5.75, + "learning_rate": 2.1285776709635135e-06, + "loss": 1.042, + "step": 26335 + }, + { + "epoch": 5.75, + "learning_rate": 2.128468429102032e-06, + "loss": 2.3115, + "step": 26336 + }, + { + "epoch": 5.75, + "learning_rate": 2.1283591872405507e-06, + "loss": 1.1523, + "step": 26337 + }, + { + "epoch": 5.75, + "learning_rate": 2.1282499453790693e-06, + "loss": 0.7778, + "step": 26338 + }, + { + "epoch": 5.75, + "learning_rate": 2.128140703517588e-06, + "loss": 0.5187, + "step": 26339 + }, + { + "epoch": 5.75, + "learning_rate": 2.128031461656107e-06, + "loss": 2.2086, + "step": 26340 + }, + { + "epoch": 5.75, + "learning_rate": 2.1279222197946256e-06, + "loss": 1.5625, + "step": 26341 + }, + { + "epoch": 5.75, + "learning_rate": 2.127812977933144e-06, + "loss": 1.4705, + "step": 26342 + }, + { + "epoch": 5.75, + "learning_rate": 2.1277037360716628e-06, + "loss": 1.3726, + "step": 26343 + }, + { + "epoch": 5.75, + "learning_rate": 2.1275944942101814e-06, + "loss": 0.7654, + "step": 26344 + }, + { + "epoch": 5.75, + "learning_rate": 2.1274852523487004e-06, + "loss": 1.8195, + "step": 26345 + }, + { + "epoch": 5.75, + "learning_rate": 2.1273760104872186e-06, + "loss": 0.9317, + "step": 26346 + }, + { + "epoch": 5.75, + "learning_rate": 2.1272667686257376e-06, + "loss": 0.3532, + "step": 26347 + }, + { + "epoch": 5.75, + "learning_rate": 2.1271575267642563e-06, + "loss": 0.8605, + "step": 26348 + }, + { + "epoch": 5.75, + "learning_rate": 2.127048284902775e-06, + "loss": 2.629, + "step": 26349 + }, + { + "epoch": 5.75, + "learning_rate": 2.1269390430412935e-06, + "loss": 2.2822, + "step": 26350 + }, + { + "epoch": 5.75, + "learning_rate": 2.126829801179812e-06, + "loss": 0.6299, + "step": 26351 + }, + { + "epoch": 5.75, + "learning_rate": 2.126720559318331e-06, + "loss": 2.5155, + "step": 26352 + }, + { + "epoch": 5.75, + "learning_rate": 2.1266113174568493e-06, + "loss": 0.3575, + "step": 26353 + }, + { + "epoch": 5.75, + "learning_rate": 2.1265020755953683e-06, + "loss": 1.8302, + "step": 26354 + }, + { + "epoch": 5.75, + "learning_rate": 2.126392833733887e-06, + "loss": 1.5833, + "step": 26355 + }, + { + "epoch": 5.75, + "learning_rate": 2.1262835918724056e-06, + "loss": 0.5662, + "step": 26356 + }, + { + "epoch": 5.75, + "learning_rate": 2.1261743500109246e-06, + "loss": 1.8337, + "step": 26357 + }, + { + "epoch": 5.75, + "learning_rate": 2.1260651081494428e-06, + "loss": 1.4038, + "step": 26358 + }, + { + "epoch": 5.75, + "learning_rate": 2.125955866287962e-06, + "loss": 1.5519, + "step": 26359 + }, + { + "epoch": 5.75, + "learning_rate": 2.1258466244264804e-06, + "loss": 1.7988, + "step": 26360 + }, + { + "epoch": 5.75, + "learning_rate": 2.125737382564999e-06, + "loss": 2.7205, + "step": 26361 + }, + { + "epoch": 5.75, + "learning_rate": 2.1256281407035177e-06, + "loss": 0.7877, + "step": 26362 + }, + { + "epoch": 5.75, + "learning_rate": 2.1255188988420363e-06, + "loss": 0.6141, + "step": 26363 + }, + { + "epoch": 5.75, + "learning_rate": 2.1254096569805553e-06, + "loss": 2.849, + "step": 26364 + }, + { + "epoch": 5.75, + "learning_rate": 2.125300415119074e-06, + "loss": 0.8209, + "step": 26365 + }, + { + "epoch": 5.75, + "learning_rate": 2.1251911732575925e-06, + "loss": 1.3828, + "step": 26366 + }, + { + "epoch": 5.75, + "learning_rate": 2.125081931396111e-06, + "loss": 0.5439, + "step": 26367 + }, + { + "epoch": 5.75, + "learning_rate": 2.1249726895346297e-06, + "loss": 1.0252, + "step": 26368 + }, + { + "epoch": 5.75, + "learning_rate": 2.1248634476731488e-06, + "loss": 0.1954, + "step": 26369 + }, + { + "epoch": 5.76, + "learning_rate": 2.124754205811667e-06, + "loss": 1.6139, + "step": 26370 + }, + { + "epoch": 5.76, + "learning_rate": 2.124644963950186e-06, + "loss": 1.2422, + "step": 26371 + }, + { + "epoch": 5.76, + "learning_rate": 2.1245357220887046e-06, + "loss": 0.2422, + "step": 26372 + }, + { + "epoch": 5.76, + "learning_rate": 2.1244264802272232e-06, + "loss": 0.2558, + "step": 26373 + }, + { + "epoch": 5.76, + "learning_rate": 2.1243172383657423e-06, + "loss": 2.0939, + "step": 26374 + }, + { + "epoch": 5.76, + "learning_rate": 2.1242079965042604e-06, + "loss": 0.1706, + "step": 26375 + }, + { + "epoch": 5.76, + "learning_rate": 2.1240987546427795e-06, + "loss": 1.1442, + "step": 26376 + }, + { + "epoch": 5.76, + "learning_rate": 2.123989512781298e-06, + "loss": 0.7911, + "step": 26377 + }, + { + "epoch": 5.76, + "learning_rate": 2.1238802709198167e-06, + "loss": 1.5991, + "step": 26378 + }, + { + "epoch": 5.76, + "learning_rate": 2.1237710290583353e-06, + "loss": 2.2228, + "step": 26379 + }, + { + "epoch": 5.76, + "learning_rate": 2.123661787196854e-06, + "loss": 2.4673, + "step": 26380 + }, + { + "epoch": 5.76, + "learning_rate": 2.123552545335373e-06, + "loss": 0.3723, + "step": 26381 + }, + { + "epoch": 5.76, + "learning_rate": 2.123443303473891e-06, + "loss": 0.4727, + "step": 26382 + }, + { + "epoch": 5.76, + "learning_rate": 2.12333406161241e-06, + "loss": 2.5938, + "step": 26383 + }, + { + "epoch": 5.76, + "learning_rate": 2.1232248197509288e-06, + "loss": 2.8056, + "step": 26384 + }, + { + "epoch": 5.76, + "learning_rate": 2.1231155778894474e-06, + "loss": 1.0923, + "step": 26385 + }, + { + "epoch": 5.76, + "learning_rate": 2.123006336027966e-06, + "loss": 2.309, + "step": 26386 + }, + { + "epoch": 5.76, + "learning_rate": 2.1228970941664846e-06, + "loss": 1.4622, + "step": 26387 + }, + { + "epoch": 5.76, + "learning_rate": 2.1227878523050037e-06, + "loss": 1.6748, + "step": 26388 + }, + { + "epoch": 5.76, + "learning_rate": 2.1226786104435223e-06, + "loss": 1.6935, + "step": 26389 + }, + { + "epoch": 5.76, + "learning_rate": 2.122569368582041e-06, + "loss": 2.6587, + "step": 26390 + }, + { + "epoch": 5.76, + "learning_rate": 2.1224601267205595e-06, + "loss": 2.588, + "step": 26391 + }, + { + "epoch": 5.76, + "learning_rate": 2.122350884859078e-06, + "loss": 1.0026, + "step": 26392 + }, + { + "epoch": 5.76, + "learning_rate": 2.1222416429975967e-06, + "loss": 1.1152, + "step": 26393 + }, + { + "epoch": 5.76, + "learning_rate": 2.1221324011361153e-06, + "loss": 1.0124, + "step": 26394 + }, + { + "epoch": 5.76, + "learning_rate": 2.1220231592746344e-06, + "loss": 2.8632, + "step": 26395 + }, + { + "epoch": 5.76, + "learning_rate": 2.121913917413153e-06, + "loss": 1.4231, + "step": 26396 + }, + { + "epoch": 5.76, + "learning_rate": 2.1218046755516716e-06, + "loss": 1.9019, + "step": 26397 + }, + { + "epoch": 5.76, + "learning_rate": 2.12169543369019e-06, + "loss": 0.8818, + "step": 26398 + }, + { + "epoch": 5.76, + "learning_rate": 2.121586191828709e-06, + "loss": 0.968, + "step": 26399 + }, + { + "epoch": 5.76, + "learning_rate": 2.121476949967228e-06, + "loss": 0.673, + "step": 26400 + }, + { + "epoch": 5.76, + "learning_rate": 2.1213677081057464e-06, + "loss": 0.9196, + "step": 26401 + }, + { + "epoch": 5.76, + "learning_rate": 2.121258466244265e-06, + "loss": 1.1416, + "step": 26402 + }, + { + "epoch": 5.76, + "learning_rate": 2.1211492243827837e-06, + "loss": 1.1056, + "step": 26403 + }, + { + "epoch": 5.76, + "learning_rate": 2.1210399825213023e-06, + "loss": 1.3224, + "step": 26404 + }, + { + "epoch": 5.76, + "learning_rate": 2.120930740659821e-06, + "loss": 1.512, + "step": 26405 + }, + { + "epoch": 5.76, + "learning_rate": 2.1208214987983395e-06, + "loss": 1.522, + "step": 26406 + }, + { + "epoch": 5.76, + "learning_rate": 2.1207122569368585e-06, + "loss": 2.1548, + "step": 26407 + }, + { + "epoch": 5.76, + "learning_rate": 2.120603015075377e-06, + "loss": 0.9615, + "step": 26408 + }, + { + "epoch": 5.76, + "learning_rate": 2.1204937732138958e-06, + "loss": 2.0425, + "step": 26409 + }, + { + "epoch": 5.76, + "learning_rate": 2.1203845313524144e-06, + "loss": 1.7176, + "step": 26410 + }, + { + "epoch": 5.76, + "learning_rate": 2.120275289490933e-06, + "loss": 1.3524, + "step": 26411 + }, + { + "epoch": 5.76, + "learning_rate": 2.1201660476294516e-06, + "loss": 3.3437, + "step": 26412 + }, + { + "epoch": 5.76, + "learning_rate": 2.1200568057679706e-06, + "loss": 1.4076, + "step": 26413 + }, + { + "epoch": 5.76, + "learning_rate": 2.1199475639064892e-06, + "loss": 1.5969, + "step": 26414 + }, + { + "epoch": 5.76, + "learning_rate": 2.119838322045008e-06, + "loss": 2.118, + "step": 26415 + }, + { + "epoch": 5.77, + "learning_rate": 2.1197290801835265e-06, + "loss": 0.779, + "step": 26416 + }, + { + "epoch": 5.77, + "learning_rate": 2.119619838322045e-06, + "loss": 0.4581, + "step": 26417 + }, + { + "epoch": 5.77, + "learning_rate": 2.119510596460564e-06, + "loss": 1.8992, + "step": 26418 + }, + { + "epoch": 5.77, + "learning_rate": 2.1194013545990823e-06, + "loss": 2.7326, + "step": 26419 + }, + { + "epoch": 5.77, + "learning_rate": 2.1192921127376013e-06, + "loss": 0.9255, + "step": 26420 + }, + { + "epoch": 5.77, + "learning_rate": 2.11918287087612e-06, + "loss": 1.6852, + "step": 26421 + }, + { + "epoch": 5.77, + "learning_rate": 2.1190736290146385e-06, + "loss": 1.3345, + "step": 26422 + }, + { + "epoch": 5.77, + "learning_rate": 2.118964387153157e-06, + "loss": 2.9019, + "step": 26423 + }, + { + "epoch": 5.77, + "learning_rate": 2.1188551452916758e-06, + "loss": 0.1113, + "step": 26424 + }, + { + "epoch": 5.77, + "learning_rate": 2.118745903430195e-06, + "loss": 0.8728, + "step": 26425 + }, + { + "epoch": 5.77, + "learning_rate": 2.118636661568713e-06, + "loss": 0.9603, + "step": 26426 + }, + { + "epoch": 5.77, + "learning_rate": 2.118527419707232e-06, + "loss": 0.7023, + "step": 26427 + }, + { + "epoch": 5.77, + "learning_rate": 2.1184181778457506e-06, + "loss": 0.2964, + "step": 26428 + }, + { + "epoch": 5.77, + "learning_rate": 2.1183089359842692e-06, + "loss": 1.1405, + "step": 26429 + }, + { + "epoch": 5.77, + "learning_rate": 2.1181996941227883e-06, + "loss": 2.0932, + "step": 26430 + }, + { + "epoch": 5.77, + "learning_rate": 2.1180904522613065e-06, + "loss": 0.5622, + "step": 26431 + }, + { + "epoch": 5.77, + "learning_rate": 2.1179812103998255e-06, + "loss": 2.082, + "step": 26432 + }, + { + "epoch": 5.77, + "learning_rate": 2.117871968538344e-06, + "loss": 2.1149, + "step": 26433 + }, + { + "epoch": 5.77, + "learning_rate": 2.1177627266768627e-06, + "loss": 0.3079, + "step": 26434 + }, + { + "epoch": 5.77, + "learning_rate": 2.1176534848153813e-06, + "loss": 1.0659, + "step": 26435 + }, + { + "epoch": 5.77, + "learning_rate": 2.1175442429539e-06, + "loss": 1.8993, + "step": 26436 + }, + { + "epoch": 5.77, + "learning_rate": 2.117435001092419e-06, + "loss": 2.351, + "step": 26437 + }, + { + "epoch": 5.77, + "learning_rate": 2.117325759230937e-06, + "loss": 2.3804, + "step": 26438 + }, + { + "epoch": 5.77, + "learning_rate": 2.117216517369456e-06, + "loss": 1.4366, + "step": 26439 + }, + { + "epoch": 5.77, + "learning_rate": 2.117107275507975e-06, + "loss": 0.8554, + "step": 26440 + }, + { + "epoch": 5.77, + "learning_rate": 2.1169980336464934e-06, + "loss": 0.5013, + "step": 26441 + }, + { + "epoch": 5.77, + "learning_rate": 2.1168887917850125e-06, + "loss": 2.3818, + "step": 26442 + }, + { + "epoch": 5.77, + "learning_rate": 2.1167795499235306e-06, + "loss": 0.8729, + "step": 26443 + }, + { + "epoch": 5.77, + "learning_rate": 2.1166703080620497e-06, + "loss": 0.9967, + "step": 26444 + }, + { + "epoch": 5.77, + "learning_rate": 2.1165610662005683e-06, + "loss": 1.0234, + "step": 26445 + }, + { + "epoch": 5.77, + "learning_rate": 2.116451824339087e-06, + "loss": 1.5772, + "step": 26446 + }, + { + "epoch": 5.77, + "learning_rate": 2.1163425824776055e-06, + "loss": 0.8347, + "step": 26447 + }, + { + "epoch": 5.77, + "learning_rate": 2.116233340616124e-06, + "loss": 0.3115, + "step": 26448 + }, + { + "epoch": 5.77, + "learning_rate": 2.116124098754643e-06, + "loss": 1.6139, + "step": 26449 + }, + { + "epoch": 5.77, + "learning_rate": 2.1160148568931613e-06, + "loss": 1.1499, + "step": 26450 + }, + { + "epoch": 5.77, + "learning_rate": 2.1159056150316804e-06, + "loss": 1.6786, + "step": 26451 + }, + { + "epoch": 5.77, + "learning_rate": 2.115796373170199e-06, + "loss": 2.2288, + "step": 26452 + }, + { + "epoch": 5.77, + "learning_rate": 2.1156871313087176e-06, + "loss": 2.5083, + "step": 26453 + }, + { + "epoch": 5.77, + "learning_rate": 2.1155778894472366e-06, + "loss": 2.8299, + "step": 26454 + }, + { + "epoch": 5.77, + "learning_rate": 2.115468647585755e-06, + "loss": 1.7639, + "step": 26455 + }, + { + "epoch": 5.77, + "learning_rate": 2.115359405724274e-06, + "loss": 2.2889, + "step": 26456 + }, + { + "epoch": 5.77, + "learning_rate": 2.1152501638627925e-06, + "loss": 2.5437, + "step": 26457 + }, + { + "epoch": 5.77, + "learning_rate": 2.115140922001311e-06, + "loss": 1.7411, + "step": 26458 + }, + { + "epoch": 5.77, + "learning_rate": 2.1150316801398297e-06, + "loss": 3.1465, + "step": 26459 + }, + { + "epoch": 5.77, + "learning_rate": 2.1149224382783483e-06, + "loss": 2.7722, + "step": 26460 + }, + { + "epoch": 5.77, + "learning_rate": 2.1148131964168673e-06, + "loss": 2.0892, + "step": 26461 + }, + { + "epoch": 5.78, + "learning_rate": 2.114703954555386e-06, + "loss": 0.7211, + "step": 26462 + }, + { + "epoch": 5.78, + "learning_rate": 2.1145947126939046e-06, + "loss": 1.3303, + "step": 26463 + }, + { + "epoch": 5.78, + "learning_rate": 2.114485470832423e-06, + "loss": 1.1783, + "step": 26464 + }, + { + "epoch": 5.78, + "learning_rate": 2.1143762289709418e-06, + "loss": 1.2276, + "step": 26465 + }, + { + "epoch": 5.78, + "learning_rate": 2.114266987109461e-06, + "loss": 2.153, + "step": 26466 + }, + { + "epoch": 5.78, + "learning_rate": 2.114157745247979e-06, + "loss": 2.7366, + "step": 26467 + }, + { + "epoch": 5.78, + "learning_rate": 2.114048503386498e-06, + "loss": 1.7919, + "step": 26468 + }, + { + "epoch": 5.78, + "learning_rate": 2.1139392615250167e-06, + "loss": 0.8018, + "step": 26469 + }, + { + "epoch": 5.78, + "learning_rate": 2.1138300196635353e-06, + "loss": 1.6856, + "step": 26470 + }, + { + "epoch": 5.78, + "learning_rate": 2.113720777802054e-06, + "loss": 0.9069, + "step": 26471 + }, + { + "epoch": 5.78, + "learning_rate": 2.1136115359405725e-06, + "loss": 2.1529, + "step": 26472 + }, + { + "epoch": 5.78, + "learning_rate": 2.1135022940790915e-06, + "loss": 1.231, + "step": 26473 + }, + { + "epoch": 5.78, + "learning_rate": 2.11339305221761e-06, + "loss": 2.2253, + "step": 26474 + }, + { + "epoch": 5.78, + "learning_rate": 2.1132838103561287e-06, + "loss": 2.2803, + "step": 26475 + }, + { + "epoch": 5.78, + "learning_rate": 2.1131745684946474e-06, + "loss": 0.8475, + "step": 26476 + }, + { + "epoch": 5.78, + "learning_rate": 2.113065326633166e-06, + "loss": 1.6607, + "step": 26477 + }, + { + "epoch": 5.78, + "learning_rate": 2.1129560847716846e-06, + "loss": 0.9743, + "step": 26478 + }, + { + "epoch": 5.78, + "learning_rate": 2.112846842910203e-06, + "loss": 1.8733, + "step": 26479 + }, + { + "epoch": 5.78, + "learning_rate": 2.1127376010487222e-06, + "loss": 2.4848, + "step": 26480 + }, + { + "epoch": 5.78, + "learning_rate": 2.112628359187241e-06, + "loss": 0.786, + "step": 26481 + }, + { + "epoch": 5.78, + "learning_rate": 2.1125191173257594e-06, + "loss": 1.0728, + "step": 26482 + }, + { + "epoch": 5.78, + "learning_rate": 2.112409875464278e-06, + "loss": 1.6499, + "step": 26483 + }, + { + "epoch": 5.78, + "learning_rate": 2.1123006336027967e-06, + "loss": 1.0177, + "step": 26484 + }, + { + "epoch": 5.78, + "learning_rate": 2.1121913917413153e-06, + "loss": 2.073, + "step": 26485 + }, + { + "epoch": 5.78, + "learning_rate": 2.1120821498798343e-06, + "loss": 0.9523, + "step": 26486 + }, + { + "epoch": 5.78, + "learning_rate": 2.111972908018353e-06, + "loss": 1.7716, + "step": 26487 + }, + { + "epoch": 5.78, + "learning_rate": 2.1118636661568715e-06, + "loss": 1.1416, + "step": 26488 + }, + { + "epoch": 5.78, + "learning_rate": 2.11175442429539e-06, + "loss": 0.184, + "step": 26489 + }, + { + "epoch": 5.78, + "learning_rate": 2.1116451824339088e-06, + "loss": 2.2389, + "step": 26490 + }, + { + "epoch": 5.78, + "learning_rate": 2.1115359405724274e-06, + "loss": 1.5011, + "step": 26491 + }, + { + "epoch": 5.78, + "learning_rate": 2.111426698710946e-06, + "loss": 1.0494, + "step": 26492 + }, + { + "epoch": 5.78, + "learning_rate": 2.111317456849465e-06, + "loss": 2.8411, + "step": 26493 + }, + { + "epoch": 5.78, + "learning_rate": 2.1112082149879836e-06, + "loss": 1.4776, + "step": 26494 + }, + { + "epoch": 5.78, + "learning_rate": 2.1110989731265022e-06, + "loss": 2.2489, + "step": 26495 + }, + { + "epoch": 5.78, + "learning_rate": 2.110989731265021e-06, + "loss": 3.6548, + "step": 26496 + }, + { + "epoch": 5.78, + "learning_rate": 2.1108804894035395e-06, + "loss": 1.5531, + "step": 26497 + }, + { + "epoch": 5.78, + "learning_rate": 2.1107712475420585e-06, + "loss": 1.3269, + "step": 26498 + }, + { + "epoch": 5.78, + "learning_rate": 2.1106620056805767e-06, + "loss": 1.0568, + "step": 26499 + }, + { + "epoch": 5.78, + "learning_rate": 2.1105527638190957e-06, + "loss": 1.7942, + "step": 26500 + }, + { + "epoch": 5.78, + "learning_rate": 2.1104435219576143e-06, + "loss": 1.1358, + "step": 26501 + }, + { + "epoch": 5.78, + "learning_rate": 2.110334280096133e-06, + "loss": 0.9382, + "step": 26502 + }, + { + "epoch": 5.78, + "learning_rate": 2.1102250382346515e-06, + "loss": 1.075, + "step": 26503 + }, + { + "epoch": 5.78, + "learning_rate": 2.11011579637317e-06, + "loss": 2.4292, + "step": 26504 + }, + { + "epoch": 5.78, + "learning_rate": 2.110006554511689e-06, + "loss": 1.9123, + "step": 26505 + }, + { + "epoch": 5.78, + "learning_rate": 2.109897312650208e-06, + "loss": 0.8987, + "step": 26506 + }, + { + "epoch": 5.79, + "learning_rate": 2.1097880707887264e-06, + "loss": 0.6493, + "step": 26507 + }, + { + "epoch": 5.79, + "learning_rate": 2.109678828927245e-06, + "loss": 2.3572, + "step": 26508 + }, + { + "epoch": 5.79, + "learning_rate": 2.1095695870657636e-06, + "loss": 2.1011, + "step": 26509 + }, + { + "epoch": 5.79, + "learning_rate": 2.1094603452042827e-06, + "loss": 1.7159, + "step": 26510 + }, + { + "epoch": 5.79, + "learning_rate": 2.109351103342801e-06, + "loss": 1.7908, + "step": 26511 + }, + { + "epoch": 5.79, + "learning_rate": 2.10924186148132e-06, + "loss": 1.5946, + "step": 26512 + }, + { + "epoch": 5.79, + "learning_rate": 2.1091326196198385e-06, + "loss": 1.3589, + "step": 26513 + }, + { + "epoch": 5.79, + "learning_rate": 2.109023377758357e-06, + "loss": 0.3307, + "step": 26514 + }, + { + "epoch": 5.79, + "learning_rate": 2.108914135896876e-06, + "loss": 1.6758, + "step": 26515 + }, + { + "epoch": 5.79, + "learning_rate": 2.1088048940353943e-06, + "loss": 2.1327, + "step": 26516 + }, + { + "epoch": 5.79, + "learning_rate": 2.1086956521739134e-06, + "loss": 1.4713, + "step": 26517 + }, + { + "epoch": 5.79, + "learning_rate": 2.108586410312432e-06, + "loss": 0.4474, + "step": 26518 + }, + { + "epoch": 5.79, + "learning_rate": 2.1084771684509506e-06, + "loss": 0.1078, + "step": 26519 + }, + { + "epoch": 5.79, + "learning_rate": 2.108367926589469e-06, + "loss": 1.5024, + "step": 26520 + }, + { + "epoch": 5.79, + "learning_rate": 2.108258684727988e-06, + "loss": 0.2823, + "step": 26521 + }, + { + "epoch": 5.79, + "learning_rate": 2.108149442866507e-06, + "loss": 1.8039, + "step": 26522 + }, + { + "epoch": 5.79, + "learning_rate": 2.108040201005025e-06, + "loss": 1.4178, + "step": 26523 + }, + { + "epoch": 5.79, + "learning_rate": 2.107930959143544e-06, + "loss": 2.125, + "step": 26524 + }, + { + "epoch": 5.79, + "learning_rate": 2.1078217172820627e-06, + "loss": 1.1729, + "step": 26525 + }, + { + "epoch": 5.79, + "learning_rate": 2.1077124754205813e-06, + "loss": 1.2943, + "step": 26526 + }, + { + "epoch": 5.79, + "learning_rate": 2.1076032335591003e-06, + "loss": 0.4168, + "step": 26527 + }, + { + "epoch": 5.79, + "learning_rate": 2.1074939916976185e-06, + "loss": 1.5545, + "step": 26528 + }, + { + "epoch": 5.79, + "learning_rate": 2.1073847498361375e-06, + "loss": 2.2034, + "step": 26529 + }, + { + "epoch": 5.79, + "learning_rate": 2.107275507974656e-06, + "loss": 1.1719, + "step": 26530 + }, + { + "epoch": 5.79, + "learning_rate": 2.1071662661131748e-06, + "loss": 1.5726, + "step": 26531 + }, + { + "epoch": 5.79, + "learning_rate": 2.1070570242516934e-06, + "loss": 1.6615, + "step": 26532 + }, + { + "epoch": 5.79, + "learning_rate": 2.106947782390212e-06, + "loss": 2.0886, + "step": 26533 + }, + { + "epoch": 5.79, + "learning_rate": 2.106838540528731e-06, + "loss": 0.4956, + "step": 26534 + }, + { + "epoch": 5.79, + "learning_rate": 2.1067292986672492e-06, + "loss": 1.7805, + "step": 26535 + }, + { + "epoch": 5.79, + "learning_rate": 2.1066200568057682e-06, + "loss": 0.7052, + "step": 26536 + }, + { + "epoch": 5.79, + "learning_rate": 2.106510814944287e-06, + "loss": 1.1561, + "step": 26537 + }, + { + "epoch": 5.79, + "learning_rate": 2.1064015730828055e-06, + "loss": 2.2662, + "step": 26538 + }, + { + "epoch": 5.79, + "learning_rate": 2.1062923312213245e-06, + "loss": 0.5989, + "step": 26539 + }, + { + "epoch": 5.79, + "learning_rate": 2.1061830893598427e-06, + "loss": 1.5203, + "step": 26540 + }, + { + "epoch": 5.79, + "learning_rate": 2.1060738474983617e-06, + "loss": 0.3554, + "step": 26541 + }, + { + "epoch": 5.79, + "learning_rate": 2.1059646056368803e-06, + "loss": 0.4612, + "step": 26542 + }, + { + "epoch": 5.79, + "learning_rate": 2.105855363775399e-06, + "loss": 1.4175, + "step": 26543 + }, + { + "epoch": 5.79, + "learning_rate": 2.1057461219139176e-06, + "loss": 2.052, + "step": 26544 + }, + { + "epoch": 5.79, + "learning_rate": 2.105636880052436e-06, + "loss": 1.7003, + "step": 26545 + }, + { + "epoch": 5.79, + "learning_rate": 2.105527638190955e-06, + "loss": 1.2102, + "step": 26546 + }, + { + "epoch": 5.79, + "learning_rate": 2.1054183963294734e-06, + "loss": 1.5631, + "step": 26547 + }, + { + "epoch": 5.79, + "learning_rate": 2.1053091544679924e-06, + "loss": 1.1343, + "step": 26548 + }, + { + "epoch": 5.79, + "learning_rate": 2.105199912606511e-06, + "loss": 1.0839, + "step": 26549 + }, + { + "epoch": 5.79, + "learning_rate": 2.1050906707450296e-06, + "loss": 1.5379, + "step": 26550 + }, + { + "epoch": 5.79, + "learning_rate": 2.1049814288835483e-06, + "loss": 0.5523, + "step": 26551 + }, + { + "epoch": 5.79, + "learning_rate": 2.104872187022067e-06, + "loss": 2.0913, + "step": 26552 + }, + { + "epoch": 5.8, + "learning_rate": 2.104762945160586e-06, + "loss": 3.3805, + "step": 26553 + }, + { + "epoch": 5.8, + "learning_rate": 2.1046537032991045e-06, + "loss": 1.3276, + "step": 26554 + }, + { + "epoch": 5.8, + "learning_rate": 2.104544461437623e-06, + "loss": 2.5473, + "step": 26555 + }, + { + "epoch": 5.8, + "learning_rate": 2.1044352195761417e-06, + "loss": 1.2063, + "step": 26556 + }, + { + "epoch": 5.8, + "learning_rate": 2.1043259777146603e-06, + "loss": 2.4209, + "step": 26557 + }, + { + "epoch": 5.8, + "learning_rate": 2.104216735853179e-06, + "loss": 1.9009, + "step": 26558 + }, + { + "epoch": 5.8, + "learning_rate": 2.1041074939916976e-06, + "loss": 0.8547, + "step": 26559 + }, + { + "epoch": 5.8, + "learning_rate": 2.1039982521302166e-06, + "loss": 1.2013, + "step": 26560 + }, + { + "epoch": 5.8, + "learning_rate": 2.1038890102687352e-06, + "loss": 1.6625, + "step": 26561 + }, + { + "epoch": 5.8, + "learning_rate": 2.103779768407254e-06, + "loss": 1.4476, + "step": 26562 + }, + { + "epoch": 5.8, + "learning_rate": 2.1036705265457724e-06, + "loss": 1.1248, + "step": 26563 + }, + { + "epoch": 5.8, + "learning_rate": 2.103561284684291e-06, + "loss": 1.6072, + "step": 26564 + }, + { + "epoch": 5.8, + "learning_rate": 2.1034520428228097e-06, + "loss": 1.3626, + "step": 26565 + }, + { + "epoch": 5.8, + "learning_rate": 2.1033428009613287e-06, + "loss": 2.5728, + "step": 26566 + }, + { + "epoch": 5.8, + "learning_rate": 2.1032335590998473e-06, + "loss": 2.7595, + "step": 26567 + }, + { + "epoch": 5.8, + "learning_rate": 2.103124317238366e-06, + "loss": 0.2464, + "step": 26568 + }, + { + "epoch": 5.8, + "learning_rate": 2.1030150753768845e-06, + "loss": 2.0006, + "step": 26569 + }, + { + "epoch": 5.8, + "learning_rate": 2.102905833515403e-06, + "loss": 1.026, + "step": 26570 + }, + { + "epoch": 5.8, + "learning_rate": 2.102796591653922e-06, + "loss": 0.9375, + "step": 26571 + }, + { + "epoch": 5.8, + "learning_rate": 2.1026873497924408e-06, + "loss": 3.186, + "step": 26572 + }, + { + "epoch": 5.8, + "learning_rate": 2.1025781079309594e-06, + "loss": 1.3869, + "step": 26573 + }, + { + "epoch": 5.8, + "learning_rate": 2.102468866069478e-06, + "loss": 0.1632, + "step": 26574 + }, + { + "epoch": 5.8, + "learning_rate": 2.1023596242079966e-06, + "loss": 0.5122, + "step": 26575 + }, + { + "epoch": 5.8, + "learning_rate": 2.1022503823465152e-06, + "loss": 0.496, + "step": 26576 + }, + { + "epoch": 5.8, + "learning_rate": 2.102141140485034e-06, + "loss": 0.8617, + "step": 26577 + }, + { + "epoch": 5.8, + "learning_rate": 2.102031898623553e-06, + "loss": 0.3721, + "step": 26578 + }, + { + "epoch": 5.8, + "learning_rate": 2.1019226567620715e-06, + "loss": 1.1548, + "step": 26579 + }, + { + "epoch": 5.8, + "learning_rate": 2.10181341490059e-06, + "loss": 2.0686, + "step": 26580 + }, + { + "epoch": 5.8, + "learning_rate": 2.1017041730391087e-06, + "loss": 0.686, + "step": 26581 + }, + { + "epoch": 5.8, + "learning_rate": 2.1015949311776273e-06, + "loss": 1.549, + "step": 26582 + }, + { + "epoch": 5.8, + "learning_rate": 2.1014856893161464e-06, + "loss": 0.3412, + "step": 26583 + }, + { + "epoch": 5.8, + "learning_rate": 2.1013764474546645e-06, + "loss": 2.0031, + "step": 26584 + }, + { + "epoch": 5.8, + "learning_rate": 2.1012672055931836e-06, + "loss": 1.2709, + "step": 26585 + }, + { + "epoch": 5.8, + "learning_rate": 2.101157963731702e-06, + "loss": 1.8706, + "step": 26586 + }, + { + "epoch": 5.8, + "learning_rate": 2.101048721870221e-06, + "loss": 1.2758, + "step": 26587 + }, + { + "epoch": 5.8, + "learning_rate": 2.1009394800087394e-06, + "loss": 1.4237, + "step": 26588 + }, + { + "epoch": 5.8, + "learning_rate": 2.100830238147258e-06, + "loss": 3.1329, + "step": 26589 + }, + { + "epoch": 5.8, + "learning_rate": 2.100720996285777e-06, + "loss": 0.4212, + "step": 26590 + }, + { + "epoch": 5.8, + "learning_rate": 2.1006117544242952e-06, + "loss": 0.9876, + "step": 26591 + }, + { + "epoch": 5.8, + "learning_rate": 2.1005025125628143e-06, + "loss": 1.7496, + "step": 26592 + }, + { + "epoch": 5.8, + "learning_rate": 2.100393270701333e-06, + "loss": 0.4195, + "step": 26593 + }, + { + "epoch": 5.8, + "learning_rate": 2.1002840288398515e-06, + "loss": 1.2651, + "step": 26594 + }, + { + "epoch": 5.8, + "learning_rate": 2.1001747869783705e-06, + "loss": 1.3687, + "step": 26595 + }, + { + "epoch": 5.8, + "learning_rate": 2.1000655451168887e-06, + "loss": 2.373, + "step": 26596 + }, + { + "epoch": 5.8, + "learning_rate": 2.0999563032554078e-06, + "loss": 0.9001, + "step": 26597 + }, + { + "epoch": 5.8, + "learning_rate": 2.0998470613939264e-06, + "loss": 1.244, + "step": 26598 + }, + { + "epoch": 5.81, + "learning_rate": 2.099737819532445e-06, + "loss": 2.4409, + "step": 26599 + }, + { + "epoch": 5.81, + "learning_rate": 2.0996285776709636e-06, + "loss": 1.6735, + "step": 26600 + }, + { + "epoch": 5.81, + "learning_rate": 2.099519335809482e-06, + "loss": 1.8306, + "step": 26601 + }, + { + "epoch": 5.81, + "learning_rate": 2.0994100939480012e-06, + "loss": 1.1457, + "step": 26602 + }, + { + "epoch": 5.81, + "learning_rate": 2.0993008520865194e-06, + "loss": 0.3752, + "step": 26603 + }, + { + "epoch": 5.81, + "learning_rate": 2.0991916102250385e-06, + "loss": 0.569, + "step": 26604 + }, + { + "epoch": 5.81, + "learning_rate": 2.099082368363557e-06, + "loss": 0.7836, + "step": 26605 + }, + { + "epoch": 5.81, + "learning_rate": 2.0989731265020757e-06, + "loss": 0.265, + "step": 26606 + }, + { + "epoch": 5.81, + "learning_rate": 2.0988638846405947e-06, + "loss": 3.2481, + "step": 26607 + }, + { + "epoch": 5.81, + "learning_rate": 2.098754642779113e-06, + "loss": 1.3769, + "step": 26608 + }, + { + "epoch": 5.81, + "learning_rate": 2.098645400917632e-06, + "loss": 1.5291, + "step": 26609 + }, + { + "epoch": 5.81, + "learning_rate": 2.0985361590561505e-06, + "loss": 0.4823, + "step": 26610 + }, + { + "epoch": 5.81, + "learning_rate": 2.098426917194669e-06, + "loss": 0.9472, + "step": 26611 + }, + { + "epoch": 5.81, + "learning_rate": 2.098317675333188e-06, + "loss": 2.4299, + "step": 26612 + }, + { + "epoch": 5.81, + "learning_rate": 2.0982084334717064e-06, + "loss": 1.096, + "step": 26613 + }, + { + "epoch": 5.81, + "learning_rate": 2.0980991916102254e-06, + "loss": 1.4362, + "step": 26614 + }, + { + "epoch": 5.81, + "learning_rate": 2.097989949748744e-06, + "loss": 1.5969, + "step": 26615 + }, + { + "epoch": 5.81, + "learning_rate": 2.0978807078872626e-06, + "loss": 0.808, + "step": 26616 + }, + { + "epoch": 5.81, + "learning_rate": 2.0977714660257812e-06, + "loss": 0.7953, + "step": 26617 + }, + { + "epoch": 5.81, + "learning_rate": 2.0976622241643e-06, + "loss": 2.9102, + "step": 26618 + }, + { + "epoch": 5.81, + "learning_rate": 2.097552982302819e-06, + "loss": 0.5982, + "step": 26619 + }, + { + "epoch": 5.81, + "learning_rate": 2.097443740441337e-06, + "loss": 0.3341, + "step": 26620 + }, + { + "epoch": 5.81, + "learning_rate": 2.097334498579856e-06, + "loss": 1.0796, + "step": 26621 + }, + { + "epoch": 5.81, + "learning_rate": 2.0972252567183747e-06, + "loss": 1.0503, + "step": 26622 + }, + { + "epoch": 5.81, + "learning_rate": 2.0971160148568933e-06, + "loss": 0.5301, + "step": 26623 + }, + { + "epoch": 5.81, + "learning_rate": 2.097006772995412e-06, + "loss": 1.2774, + "step": 26624 + }, + { + "epoch": 5.81, + "learning_rate": 2.0968975311339306e-06, + "loss": 1.5655, + "step": 26625 + }, + { + "epoch": 5.81, + "learning_rate": 2.0967882892724496e-06, + "loss": 0.9413, + "step": 26626 + }, + { + "epoch": 5.81, + "learning_rate": 2.096679047410968e-06, + "loss": 1.6583, + "step": 26627 + }, + { + "epoch": 5.81, + "learning_rate": 2.096569805549487e-06, + "loss": 0.8656, + "step": 26628 + }, + { + "epoch": 5.81, + "learning_rate": 2.0964605636880054e-06, + "loss": 1.925, + "step": 26629 + }, + { + "epoch": 5.81, + "learning_rate": 2.096351321826524e-06, + "loss": 1.1497, + "step": 26630 + }, + { + "epoch": 5.81, + "learning_rate": 2.0962420799650426e-06, + "loss": 1.9595, + "step": 26631 + }, + { + "epoch": 5.81, + "learning_rate": 2.0961328381035613e-06, + "loss": 2.5336, + "step": 26632 + }, + { + "epoch": 5.81, + "learning_rate": 2.0960235962420803e-06, + "loss": 0.7879, + "step": 26633 + }, + { + "epoch": 5.81, + "learning_rate": 2.095914354380599e-06, + "loss": 1.1173, + "step": 26634 + }, + { + "epoch": 5.81, + "learning_rate": 2.0958051125191175e-06, + "loss": 1.2255, + "step": 26635 + }, + { + "epoch": 5.81, + "learning_rate": 2.095695870657636e-06, + "loss": 2.3229, + "step": 26636 + }, + { + "epoch": 5.81, + "learning_rate": 2.0955866287961547e-06, + "loss": 3.4544, + "step": 26637 + }, + { + "epoch": 5.81, + "learning_rate": 2.0954773869346738e-06, + "loss": 2.0193, + "step": 26638 + }, + { + "epoch": 5.81, + "learning_rate": 2.0953681450731924e-06, + "loss": 1.0809, + "step": 26639 + }, + { + "epoch": 5.81, + "learning_rate": 2.095258903211711e-06, + "loss": 0.596, + "step": 26640 + }, + { + "epoch": 5.81, + "learning_rate": 2.0951496613502296e-06, + "loss": 1.2749, + "step": 26641 + }, + { + "epoch": 5.81, + "learning_rate": 2.095040419488748e-06, + "loss": 2.2, + "step": 26642 + }, + { + "epoch": 5.81, + "learning_rate": 2.094931177627267e-06, + "loss": 1.7564, + "step": 26643 + }, + { + "epoch": 5.81, + "learning_rate": 2.0948219357657854e-06, + "loss": 0.9207, + "step": 26644 + }, + { + "epoch": 5.82, + "learning_rate": 2.0947126939043045e-06, + "loss": 0.6321, + "step": 26645 + }, + { + "epoch": 5.82, + "learning_rate": 2.094603452042823e-06, + "loss": 1.1693, + "step": 26646 + }, + { + "epoch": 5.82, + "learning_rate": 2.0944942101813417e-06, + "loss": 0.8514, + "step": 26647 + }, + { + "epoch": 5.82, + "learning_rate": 2.0943849683198603e-06, + "loss": 1.5926, + "step": 26648 + }, + { + "epoch": 5.82, + "learning_rate": 2.094275726458379e-06, + "loss": 1.7729, + "step": 26649 + }, + { + "epoch": 5.82, + "learning_rate": 2.0941664845968975e-06, + "loss": 0.7385, + "step": 26650 + }, + { + "epoch": 5.82, + "learning_rate": 2.0940572427354166e-06, + "loss": 3.2802, + "step": 26651 + }, + { + "epoch": 5.82, + "learning_rate": 2.093948000873935e-06, + "loss": 0.7785, + "step": 26652 + }, + { + "epoch": 5.82, + "learning_rate": 2.0938387590124538e-06, + "loss": 0.8686, + "step": 26653 + }, + { + "epoch": 5.82, + "learning_rate": 2.0937295171509724e-06, + "loss": 1.8864, + "step": 26654 + }, + { + "epoch": 5.82, + "learning_rate": 2.093620275289491e-06, + "loss": 2.2866, + "step": 26655 + }, + { + "epoch": 5.82, + "learning_rate": 2.0935110334280096e-06, + "loss": 0.541, + "step": 26656 + }, + { + "epoch": 5.82, + "learning_rate": 2.0934017915665282e-06, + "loss": 1.4386, + "step": 26657 + }, + { + "epoch": 5.82, + "learning_rate": 2.0932925497050473e-06, + "loss": 1.8821, + "step": 26658 + }, + { + "epoch": 5.82, + "learning_rate": 2.093183307843566e-06, + "loss": 1.8046, + "step": 26659 + }, + { + "epoch": 5.82, + "learning_rate": 2.0930740659820845e-06, + "loss": 1.6634, + "step": 26660 + }, + { + "epoch": 5.82, + "learning_rate": 2.092964824120603e-06, + "loss": 2.5023, + "step": 26661 + }, + { + "epoch": 5.82, + "learning_rate": 2.0928555822591217e-06, + "loss": 1.1838, + "step": 26662 + }, + { + "epoch": 5.82, + "learning_rate": 2.0927463403976407e-06, + "loss": 1.6052, + "step": 26663 + }, + { + "epoch": 5.82, + "learning_rate": 2.092637098536159e-06, + "loss": 1.5877, + "step": 26664 + }, + { + "epoch": 5.82, + "learning_rate": 2.092527856674678e-06, + "loss": 1.0925, + "step": 26665 + }, + { + "epoch": 5.82, + "learning_rate": 2.0924186148131966e-06, + "loss": 1.6404, + "step": 26666 + }, + { + "epoch": 5.82, + "learning_rate": 2.092309372951715e-06, + "loss": 1.5447, + "step": 26667 + }, + { + "epoch": 5.82, + "learning_rate": 2.0922001310902342e-06, + "loss": 1.375, + "step": 26668 + }, + { + "epoch": 5.82, + "learning_rate": 2.0920908892287524e-06, + "loss": 1.4101, + "step": 26669 + }, + { + "epoch": 5.82, + "learning_rate": 2.0919816473672714e-06, + "loss": 0.9735, + "step": 26670 + }, + { + "epoch": 5.82, + "learning_rate": 2.09187240550579e-06, + "loss": 1.9879, + "step": 26671 + }, + { + "epoch": 5.82, + "learning_rate": 2.0917631636443087e-06, + "loss": 2.341, + "step": 26672 + }, + { + "epoch": 5.82, + "learning_rate": 2.0916539217828273e-06, + "loss": 2.7776, + "step": 26673 + }, + { + "epoch": 5.82, + "learning_rate": 2.091544679921346e-06, + "loss": 1.1057, + "step": 26674 + }, + { + "epoch": 5.82, + "learning_rate": 2.091435438059865e-06, + "loss": 0.8305, + "step": 26675 + }, + { + "epoch": 5.82, + "learning_rate": 2.091326196198383e-06, + "loss": 1.1317, + "step": 26676 + }, + { + "epoch": 5.82, + "learning_rate": 2.091216954336902e-06, + "loss": 1.5274, + "step": 26677 + }, + { + "epoch": 5.82, + "learning_rate": 2.0911077124754207e-06, + "loss": 1.6613, + "step": 26678 + }, + { + "epoch": 5.82, + "learning_rate": 2.0909984706139394e-06, + "loss": 1.0073, + "step": 26679 + }, + { + "epoch": 5.82, + "learning_rate": 2.0908892287524584e-06, + "loss": 0.7802, + "step": 26680 + }, + { + "epoch": 5.82, + "learning_rate": 2.0907799868909766e-06, + "loss": 2.4535, + "step": 26681 + }, + { + "epoch": 5.82, + "learning_rate": 2.0906707450294956e-06, + "loss": 2.0117, + "step": 26682 + }, + { + "epoch": 5.82, + "learning_rate": 2.0905615031680142e-06, + "loss": 1.3311, + "step": 26683 + }, + { + "epoch": 5.82, + "learning_rate": 2.090452261306533e-06, + "loss": 1.4191, + "step": 26684 + }, + { + "epoch": 5.82, + "learning_rate": 2.0903430194450514e-06, + "loss": 0.7413, + "step": 26685 + }, + { + "epoch": 5.82, + "learning_rate": 2.09023377758357e-06, + "loss": 2.7565, + "step": 26686 + }, + { + "epoch": 5.82, + "learning_rate": 2.090124535722089e-06, + "loss": 1.5206, + "step": 26687 + }, + { + "epoch": 5.82, + "learning_rate": 2.0900152938606073e-06, + "loss": 1.3834, + "step": 26688 + }, + { + "epoch": 5.82, + "learning_rate": 2.0899060519991263e-06, + "loss": 0.9866, + "step": 26689 + }, + { + "epoch": 5.82, + "learning_rate": 2.089796810137645e-06, + "loss": 1.1186, + "step": 26690 + }, + { + "epoch": 5.83, + "learning_rate": 2.0896875682761635e-06, + "loss": 1.7314, + "step": 26691 + }, + { + "epoch": 5.83, + "learning_rate": 2.0895783264146826e-06, + "loss": 0.2819, + "step": 26692 + }, + { + "epoch": 5.83, + "learning_rate": 2.0894690845532008e-06, + "loss": 0.487, + "step": 26693 + }, + { + "epoch": 5.83, + "learning_rate": 2.08935984269172e-06, + "loss": 1.6423, + "step": 26694 + }, + { + "epoch": 5.83, + "learning_rate": 2.0892506008302384e-06, + "loss": 2.063, + "step": 26695 + }, + { + "epoch": 5.83, + "learning_rate": 2.089141358968757e-06, + "loss": 3.2876, + "step": 26696 + }, + { + "epoch": 5.83, + "learning_rate": 2.0890321171072756e-06, + "loss": 2.2207, + "step": 26697 + }, + { + "epoch": 5.83, + "learning_rate": 2.0889228752457942e-06, + "loss": 1.876, + "step": 26698 + }, + { + "epoch": 5.83, + "learning_rate": 2.0888136333843133e-06, + "loss": 1.1806, + "step": 26699 + }, + { + "epoch": 5.83, + "learning_rate": 2.0887043915228315e-06, + "loss": 2.0726, + "step": 26700 + }, + { + "epoch": 5.83, + "learning_rate": 2.0885951496613505e-06, + "loss": 3.9935, + "step": 26701 + }, + { + "epoch": 5.83, + "learning_rate": 2.088485907799869e-06, + "loss": 2.0478, + "step": 26702 + }, + { + "epoch": 5.83, + "learning_rate": 2.0883766659383877e-06, + "loss": 1.6516, + "step": 26703 + }, + { + "epoch": 5.83, + "learning_rate": 2.0882674240769068e-06, + "loss": 0.6265, + "step": 26704 + }, + { + "epoch": 5.83, + "learning_rate": 2.088158182215425e-06, + "loss": 1.6903, + "step": 26705 + }, + { + "epoch": 5.83, + "learning_rate": 2.088048940353944e-06, + "loss": 1.2477, + "step": 26706 + }, + { + "epoch": 5.83, + "learning_rate": 2.0879396984924626e-06, + "loss": 1.9164, + "step": 26707 + }, + { + "epoch": 5.83, + "learning_rate": 2.087830456630981e-06, + "loss": 2.309, + "step": 26708 + }, + { + "epoch": 5.83, + "learning_rate": 2.0877212147695e-06, + "loss": 2.5185, + "step": 26709 + }, + { + "epoch": 5.83, + "learning_rate": 2.0876119729080184e-06, + "loss": 0.7738, + "step": 26710 + }, + { + "epoch": 5.83, + "learning_rate": 2.0875027310465375e-06, + "loss": 0.1113, + "step": 26711 + }, + { + "epoch": 5.83, + "learning_rate": 2.087393489185056e-06, + "loss": 2.4377, + "step": 26712 + }, + { + "epoch": 5.83, + "learning_rate": 2.0872842473235747e-06, + "loss": 2.653, + "step": 26713 + }, + { + "epoch": 5.83, + "learning_rate": 2.0871750054620933e-06, + "loss": 1.0841, + "step": 26714 + }, + { + "epoch": 5.83, + "learning_rate": 2.087065763600612e-06, + "loss": 1.5476, + "step": 26715 + }, + { + "epoch": 5.83, + "learning_rate": 2.0869565217391305e-06, + "loss": 2.5861, + "step": 26716 + }, + { + "epoch": 5.83, + "learning_rate": 2.086847279877649e-06, + "loss": 0.4717, + "step": 26717 + }, + { + "epoch": 5.83, + "learning_rate": 2.086738038016168e-06, + "loss": 2.5809, + "step": 26718 + }, + { + "epoch": 5.83, + "learning_rate": 2.0866287961546868e-06, + "loss": 2.6553, + "step": 26719 + }, + { + "epoch": 5.83, + "learning_rate": 2.0865195542932054e-06, + "loss": 3.5458, + "step": 26720 + }, + { + "epoch": 5.83, + "learning_rate": 2.086410312431724e-06, + "loss": 0.9344, + "step": 26721 + }, + { + "epoch": 5.83, + "learning_rate": 2.0863010705702426e-06, + "loss": 1.2499, + "step": 26722 + }, + { + "epoch": 5.83, + "learning_rate": 2.086191828708761e-06, + "loss": 1.811, + "step": 26723 + }, + { + "epoch": 5.83, + "learning_rate": 2.0860825868472802e-06, + "loss": 2.7134, + "step": 26724 + }, + { + "epoch": 5.83, + "learning_rate": 2.085973344985799e-06, + "loss": 0.8112, + "step": 26725 + }, + { + "epoch": 5.83, + "learning_rate": 2.0858641031243175e-06, + "loss": 1.6707, + "step": 26726 + }, + { + "epoch": 5.83, + "learning_rate": 2.085754861262836e-06, + "loss": 0.7296, + "step": 26727 + }, + { + "epoch": 5.83, + "learning_rate": 2.0856456194013547e-06, + "loss": 0.7219, + "step": 26728 + }, + { + "epoch": 5.83, + "learning_rate": 2.0855363775398733e-06, + "loss": 1.9872, + "step": 26729 + }, + { + "epoch": 5.83, + "learning_rate": 2.085427135678392e-06, + "loss": 1.407, + "step": 26730 + }, + { + "epoch": 5.83, + "learning_rate": 2.085317893816911e-06, + "loss": 0.1569, + "step": 26731 + }, + { + "epoch": 5.83, + "learning_rate": 2.0852086519554296e-06, + "loss": 0.8506, + "step": 26732 + }, + { + "epoch": 5.83, + "learning_rate": 2.085099410093948e-06, + "loss": 1.4535, + "step": 26733 + }, + { + "epoch": 5.83, + "learning_rate": 2.0849901682324668e-06, + "loss": 1.4414, + "step": 26734 + }, + { + "epoch": 5.83, + "learning_rate": 2.0848809263709854e-06, + "loss": 1.3468, + "step": 26735 + }, + { + "epoch": 5.84, + "learning_rate": 2.0847716845095044e-06, + "loss": 2.79, + "step": 26736 + }, + { + "epoch": 5.84, + "learning_rate": 2.0846624426480226e-06, + "loss": 2.2827, + "step": 26737 + }, + { + "epoch": 5.84, + "learning_rate": 2.0845532007865416e-06, + "loss": 1.0232, + "step": 26738 + }, + { + "epoch": 5.84, + "learning_rate": 2.0844439589250603e-06, + "loss": 2.1716, + "step": 26739 + }, + { + "epoch": 5.84, + "learning_rate": 2.084334717063579e-06, + "loss": 1.2393, + "step": 26740 + }, + { + "epoch": 5.84, + "learning_rate": 2.0842254752020975e-06, + "loss": 2.2176, + "step": 26741 + }, + { + "epoch": 5.84, + "learning_rate": 2.084116233340616e-06, + "loss": 2.0192, + "step": 26742 + }, + { + "epoch": 5.84, + "learning_rate": 2.084006991479135e-06, + "loss": 1.3672, + "step": 26743 + }, + { + "epoch": 5.84, + "learning_rate": 2.0838977496176537e-06, + "loss": 1.2239, + "step": 26744 + }, + { + "epoch": 5.84, + "learning_rate": 2.0837885077561723e-06, + "loss": 2.4496, + "step": 26745 + }, + { + "epoch": 5.84, + "learning_rate": 2.083679265894691e-06, + "loss": 3.1606, + "step": 26746 + }, + { + "epoch": 5.84, + "learning_rate": 2.0835700240332096e-06, + "loss": 0.9111, + "step": 26747 + }, + { + "epoch": 5.84, + "learning_rate": 2.0834607821717286e-06, + "loss": 0.8775, + "step": 26748 + }, + { + "epoch": 5.84, + "learning_rate": 2.0833515403102468e-06, + "loss": 0.2348, + "step": 26749 + }, + { + "epoch": 5.84, + "learning_rate": 2.083242298448766e-06, + "loss": 1.477, + "step": 26750 + }, + { + "epoch": 5.84, + "learning_rate": 2.0831330565872844e-06, + "loss": 1.2263, + "step": 26751 + }, + { + "epoch": 5.84, + "learning_rate": 2.083023814725803e-06, + "loss": 1.7415, + "step": 26752 + }, + { + "epoch": 5.84, + "learning_rate": 2.0829145728643217e-06, + "loss": 0.1022, + "step": 26753 + }, + { + "epoch": 5.84, + "learning_rate": 2.0828053310028403e-06, + "loss": 1.2411, + "step": 26754 + }, + { + "epoch": 5.84, + "learning_rate": 2.0826960891413593e-06, + "loss": 2.1438, + "step": 26755 + }, + { + "epoch": 5.84, + "learning_rate": 2.082586847279878e-06, + "loss": 2.4547, + "step": 26756 + }, + { + "epoch": 5.84, + "learning_rate": 2.0824776054183965e-06, + "loss": 2.96, + "step": 26757 + }, + { + "epoch": 5.84, + "learning_rate": 2.082368363556915e-06, + "loss": 0.639, + "step": 26758 + }, + { + "epoch": 5.84, + "learning_rate": 2.0822591216954337e-06, + "loss": 2.0704, + "step": 26759 + }, + { + "epoch": 5.84, + "learning_rate": 2.0821498798339528e-06, + "loss": 1.0994, + "step": 26760 + }, + { + "epoch": 5.84, + "learning_rate": 2.082040637972471e-06, + "loss": 1.9664, + "step": 26761 + }, + { + "epoch": 5.84, + "learning_rate": 2.08193139611099e-06, + "loss": 1.7283, + "step": 26762 + }, + { + "epoch": 5.84, + "learning_rate": 2.0818221542495086e-06, + "loss": 2.742, + "step": 26763 + }, + { + "epoch": 5.84, + "learning_rate": 2.0817129123880272e-06, + "loss": 0.5502, + "step": 26764 + }, + { + "epoch": 5.84, + "learning_rate": 2.0816036705265463e-06, + "loss": 1.5495, + "step": 26765 + }, + { + "epoch": 5.84, + "learning_rate": 2.0814944286650644e-06, + "loss": 2.0648, + "step": 26766 + }, + { + "epoch": 5.84, + "learning_rate": 2.0813851868035835e-06, + "loss": 1.7401, + "step": 26767 + }, + { + "epoch": 5.84, + "learning_rate": 2.081275944942102e-06, + "loss": 2.7209, + "step": 26768 + }, + { + "epoch": 5.84, + "learning_rate": 2.0811667030806207e-06, + "loss": 2.2651, + "step": 26769 + }, + { + "epoch": 5.84, + "learning_rate": 2.0810574612191393e-06, + "loss": 1.6363, + "step": 26770 + }, + { + "epoch": 5.84, + "learning_rate": 2.080948219357658e-06, + "loss": 1.838, + "step": 26771 + }, + { + "epoch": 5.84, + "learning_rate": 2.080838977496177e-06, + "loss": 2.3174, + "step": 26772 + }, + { + "epoch": 5.84, + "learning_rate": 2.080729735634695e-06, + "loss": 2.186, + "step": 26773 + }, + { + "epoch": 5.84, + "learning_rate": 2.080620493773214e-06, + "loss": 0.7267, + "step": 26774 + }, + { + "epoch": 5.84, + "learning_rate": 2.080511251911733e-06, + "loss": 1.8911, + "step": 26775 + }, + { + "epoch": 5.84, + "learning_rate": 2.0804020100502514e-06, + "loss": 0.1938, + "step": 26776 + }, + { + "epoch": 5.84, + "learning_rate": 2.0802927681887704e-06, + "loss": 1.4444, + "step": 26777 + }, + { + "epoch": 5.84, + "learning_rate": 2.0801835263272886e-06, + "loss": 1.1947, + "step": 26778 + }, + { + "epoch": 5.84, + "learning_rate": 2.0800742844658077e-06, + "loss": 1.3936, + "step": 26779 + }, + { + "epoch": 5.84, + "learning_rate": 2.0799650426043263e-06, + "loss": 2.1033, + "step": 26780 + }, + { + "epoch": 5.84, + "learning_rate": 2.079855800742845e-06, + "loss": 1.1928, + "step": 26781 + }, + { + "epoch": 5.85, + "learning_rate": 2.0797465588813635e-06, + "loss": 1.8299, + "step": 26782 + }, + { + "epoch": 5.85, + "learning_rate": 2.079637317019882e-06, + "loss": 1.8416, + "step": 26783 + }, + { + "epoch": 5.85, + "learning_rate": 2.079528075158401e-06, + "loss": 1.2393, + "step": 26784 + }, + { + "epoch": 5.85, + "learning_rate": 2.0794188332969193e-06, + "loss": 1.4625, + "step": 26785 + }, + { + "epoch": 5.85, + "learning_rate": 2.0793095914354384e-06, + "loss": 1.3344, + "step": 26786 + }, + { + "epoch": 5.85, + "learning_rate": 2.079200349573957e-06, + "loss": 2.5209, + "step": 26787 + }, + { + "epoch": 5.85, + "learning_rate": 2.0790911077124756e-06, + "loss": 1.8305, + "step": 26788 + }, + { + "epoch": 5.85, + "learning_rate": 2.078981865850994e-06, + "loss": 0.3909, + "step": 26789 + }, + { + "epoch": 5.85, + "learning_rate": 2.078872623989513e-06, + "loss": 1.5011, + "step": 26790 + }, + { + "epoch": 5.85, + "learning_rate": 2.078763382128032e-06, + "loss": 1.5631, + "step": 26791 + }, + { + "epoch": 5.85, + "learning_rate": 2.0786541402665504e-06, + "loss": 1.6159, + "step": 26792 + }, + { + "epoch": 5.85, + "learning_rate": 2.078544898405069e-06, + "loss": 0.7758, + "step": 26793 + }, + { + "epoch": 5.85, + "learning_rate": 2.0784356565435877e-06, + "loss": 0.3354, + "step": 26794 + }, + { + "epoch": 5.85, + "learning_rate": 2.0783264146821063e-06, + "loss": 1.6245, + "step": 26795 + }, + { + "epoch": 5.85, + "learning_rate": 2.078217172820625e-06, + "loss": 1.8886, + "step": 26796 + }, + { + "epoch": 5.85, + "learning_rate": 2.0781079309591435e-06, + "loss": 0.6747, + "step": 26797 + }, + { + "epoch": 5.85, + "learning_rate": 2.0779986890976625e-06, + "loss": 0.3502, + "step": 26798 + }, + { + "epoch": 5.85, + "learning_rate": 2.077889447236181e-06, + "loss": 0.4946, + "step": 26799 + }, + { + "epoch": 5.85, + "learning_rate": 2.0777802053746998e-06, + "loss": 1.2846, + "step": 26800 + }, + { + "epoch": 5.85, + "learning_rate": 2.0776709635132184e-06, + "loss": 0.124, + "step": 26801 + }, + { + "epoch": 5.85, + "learning_rate": 2.077561721651737e-06, + "loss": 1.2217, + "step": 26802 + }, + { + "epoch": 5.85, + "learning_rate": 2.0774524797902556e-06, + "loss": 2.8234, + "step": 26803 + }, + { + "epoch": 5.85, + "learning_rate": 2.0773432379287746e-06, + "loss": 2.8153, + "step": 26804 + }, + { + "epoch": 5.85, + "learning_rate": 2.0772339960672932e-06, + "loss": 0.7174, + "step": 26805 + }, + { + "epoch": 5.85, + "learning_rate": 2.077124754205812e-06, + "loss": 0.9972, + "step": 26806 + }, + { + "epoch": 5.85, + "learning_rate": 2.0770155123443305e-06, + "loss": 2.2018, + "step": 26807 + }, + { + "epoch": 5.85, + "learning_rate": 2.076906270482849e-06, + "loss": 2.2775, + "step": 26808 + }, + { + "epoch": 5.85, + "learning_rate": 2.076797028621368e-06, + "loss": 2.6938, + "step": 26809 + }, + { + "epoch": 5.85, + "learning_rate": 2.0766877867598867e-06, + "loss": 1.0745, + "step": 26810 + }, + { + "epoch": 5.85, + "learning_rate": 2.0765785448984053e-06, + "loss": 1.4719, + "step": 26811 + }, + { + "epoch": 5.85, + "learning_rate": 2.076469303036924e-06, + "loss": 1.1578, + "step": 26812 + }, + { + "epoch": 5.85, + "learning_rate": 2.0763600611754425e-06, + "loss": 0.6624, + "step": 26813 + }, + { + "epoch": 5.85, + "learning_rate": 2.076250819313961e-06, + "loss": 0.3406, + "step": 26814 + }, + { + "epoch": 5.85, + "learning_rate": 2.0761415774524798e-06, + "loss": 1.112, + "step": 26815 + }, + { + "epoch": 5.85, + "learning_rate": 2.076032335590999e-06, + "loss": 0.853, + "step": 26816 + }, + { + "epoch": 5.85, + "learning_rate": 2.0759230937295174e-06, + "loss": 0.8946, + "step": 26817 + }, + { + "epoch": 5.85, + "learning_rate": 2.075813851868036e-06, + "loss": 0.4605, + "step": 26818 + }, + { + "epoch": 5.85, + "learning_rate": 2.0757046100065546e-06, + "loss": 1.0008, + "step": 26819 + }, + { + "epoch": 5.85, + "learning_rate": 2.0755953681450732e-06, + "loss": 1.0098, + "step": 26820 + }, + { + "epoch": 5.85, + "learning_rate": 2.0754861262835923e-06, + "loss": 1.1936, + "step": 26821 + }, + { + "epoch": 5.85, + "learning_rate": 2.0753768844221105e-06, + "loss": 1.795, + "step": 26822 + }, + { + "epoch": 5.85, + "learning_rate": 2.0752676425606295e-06, + "loss": 1.2567, + "step": 26823 + }, + { + "epoch": 5.85, + "learning_rate": 2.075158400699148e-06, + "loss": 1.6874, + "step": 26824 + }, + { + "epoch": 5.85, + "learning_rate": 2.0750491588376667e-06, + "loss": 1.3176, + "step": 26825 + }, + { + "epoch": 5.85, + "learning_rate": 2.0749399169761853e-06, + "loss": 1.2445, + "step": 26826 + }, + { + "epoch": 5.85, + "learning_rate": 2.074830675114704e-06, + "loss": 2.3046, + "step": 26827 + }, + { + "epoch": 5.86, + "learning_rate": 2.074721433253223e-06, + "loss": 1.6823, + "step": 26828 + }, + { + "epoch": 5.86, + "learning_rate": 2.074612191391741e-06, + "loss": 2.1902, + "step": 26829 + }, + { + "epoch": 5.86, + "learning_rate": 2.07450294953026e-06, + "loss": 1.6373, + "step": 26830 + }, + { + "epoch": 5.86, + "learning_rate": 2.074393707668779e-06, + "loss": 0.9618, + "step": 26831 + }, + { + "epoch": 5.86, + "learning_rate": 2.0742844658072974e-06, + "loss": 0.9704, + "step": 26832 + }, + { + "epoch": 5.86, + "learning_rate": 2.0741752239458165e-06, + "loss": 2.5846, + "step": 26833 + }, + { + "epoch": 5.86, + "learning_rate": 2.0740659820843346e-06, + "loss": 0.6257, + "step": 26834 + }, + { + "epoch": 5.86, + "learning_rate": 2.0739567402228537e-06, + "loss": 1.8278, + "step": 26835 + }, + { + "epoch": 5.86, + "learning_rate": 2.0738474983613723e-06, + "loss": 2.5356, + "step": 26836 + }, + { + "epoch": 5.86, + "learning_rate": 2.073738256499891e-06, + "loss": 1.5115, + "step": 26837 + }, + { + "epoch": 5.86, + "learning_rate": 2.0736290146384095e-06, + "loss": 1.0546, + "step": 26838 + }, + { + "epoch": 5.86, + "learning_rate": 2.073519772776928e-06, + "loss": 1.3056, + "step": 26839 + }, + { + "epoch": 5.86, + "learning_rate": 2.073410530915447e-06, + "loss": 0.3673, + "step": 26840 + }, + { + "epoch": 5.86, + "learning_rate": 2.0733012890539653e-06, + "loss": 1.0817, + "step": 26841 + }, + { + "epoch": 5.86, + "learning_rate": 2.0731920471924844e-06, + "loss": 1.5473, + "step": 26842 + }, + { + "epoch": 5.86, + "learning_rate": 2.073082805331003e-06, + "loss": 0.3927, + "step": 26843 + }, + { + "epoch": 5.86, + "learning_rate": 2.0729735634695216e-06, + "loss": 1.1427, + "step": 26844 + }, + { + "epoch": 5.86, + "learning_rate": 2.0728643216080406e-06, + "loss": 1.1635, + "step": 26845 + }, + { + "epoch": 5.86, + "learning_rate": 2.072755079746559e-06, + "loss": 1.5995, + "step": 26846 + }, + { + "epoch": 5.86, + "learning_rate": 2.072645837885078e-06, + "loss": 1.3201, + "step": 26847 + }, + { + "epoch": 5.86, + "learning_rate": 2.0725365960235965e-06, + "loss": 0.8234, + "step": 26848 + }, + { + "epoch": 5.86, + "learning_rate": 2.072427354162115e-06, + "loss": 1.3992, + "step": 26849 + }, + { + "epoch": 5.86, + "learning_rate": 2.0723181123006337e-06, + "loss": 1.0111, + "step": 26850 + }, + { + "epoch": 5.86, + "learning_rate": 2.0722088704391523e-06, + "loss": 0.7477, + "step": 26851 + }, + { + "epoch": 5.86, + "learning_rate": 2.0720996285776713e-06, + "loss": 1.2878, + "step": 26852 + }, + { + "epoch": 5.86, + "learning_rate": 2.0719903867161895e-06, + "loss": 1.1827, + "step": 26853 + }, + { + "epoch": 5.86, + "learning_rate": 2.0718811448547086e-06, + "loss": 2.5651, + "step": 26854 + }, + { + "epoch": 5.86, + "learning_rate": 2.071771902993227e-06, + "loss": 1.2156, + "step": 26855 + }, + { + "epoch": 5.86, + "learning_rate": 2.0716626611317458e-06, + "loss": 0.4655, + "step": 26856 + }, + { + "epoch": 5.86, + "learning_rate": 2.071553419270265e-06, + "loss": 0.6862, + "step": 26857 + }, + { + "epoch": 5.86, + "learning_rate": 2.071444177408783e-06, + "loss": 0.5922, + "step": 26858 + }, + { + "epoch": 5.86, + "learning_rate": 2.071334935547302e-06, + "loss": 1.2787, + "step": 26859 + }, + { + "epoch": 5.86, + "learning_rate": 2.0712256936858207e-06, + "loss": 0.4173, + "step": 26860 + }, + { + "epoch": 5.86, + "learning_rate": 2.0711164518243393e-06, + "loss": 1.2953, + "step": 26861 + }, + { + "epoch": 5.86, + "learning_rate": 2.071007209962858e-06, + "loss": 0.3935, + "step": 26862 + }, + { + "epoch": 5.86, + "learning_rate": 2.0708979681013765e-06, + "loss": 1.1658, + "step": 26863 + }, + { + "epoch": 5.86, + "learning_rate": 2.0707887262398955e-06, + "loss": 1.2859, + "step": 26864 + }, + { + "epoch": 5.86, + "learning_rate": 2.070679484378414e-06, + "loss": 0.5509, + "step": 26865 + }, + { + "epoch": 5.86, + "learning_rate": 2.0705702425169327e-06, + "loss": 1.7448, + "step": 26866 + }, + { + "epoch": 5.86, + "learning_rate": 2.0704610006554514e-06, + "loss": 1.1485, + "step": 26867 + }, + { + "epoch": 5.86, + "learning_rate": 2.07035175879397e-06, + "loss": 0.7314, + "step": 26868 + }, + { + "epoch": 5.86, + "learning_rate": 2.0702425169324886e-06, + "loss": 1.6772, + "step": 26869 + }, + { + "epoch": 5.86, + "learning_rate": 2.070133275071007e-06, + "loss": 0.4951, + "step": 26870 + }, + { + "epoch": 5.86, + "learning_rate": 2.0700240332095262e-06, + "loss": 0.8683, + "step": 26871 + }, + { + "epoch": 5.86, + "learning_rate": 2.069914791348045e-06, + "loss": 1.6722, + "step": 26872 + }, + { + "epoch": 5.86, + "learning_rate": 2.0698055494865634e-06, + "loss": 3.1372, + "step": 26873 + }, + { + "epoch": 5.87, + "learning_rate": 2.069696307625082e-06, + "loss": 1.1394, + "step": 26874 + }, + { + "epoch": 5.87, + "learning_rate": 2.0695870657636007e-06, + "loss": 0.7617, + "step": 26875 + }, + { + "epoch": 5.87, + "learning_rate": 2.0694778239021197e-06, + "loss": 1.686, + "step": 26876 + }, + { + "epoch": 5.87, + "learning_rate": 2.0693685820406383e-06, + "loss": 1.0496, + "step": 26877 + }, + { + "epoch": 5.87, + "learning_rate": 2.069259340179157e-06, + "loss": 0.9014, + "step": 26878 + }, + { + "epoch": 5.87, + "learning_rate": 2.0691500983176755e-06, + "loss": 2.8233, + "step": 26879 + }, + { + "epoch": 5.87, + "learning_rate": 2.069040856456194e-06, + "loss": 0.8203, + "step": 26880 + }, + { + "epoch": 5.87, + "learning_rate": 2.0689316145947128e-06, + "loss": 1.6327, + "step": 26881 + }, + { + "epoch": 5.87, + "learning_rate": 2.0688223727332314e-06, + "loss": 1.5265, + "step": 26882 + }, + { + "epoch": 5.87, + "learning_rate": 2.0687131308717504e-06, + "loss": 0.7856, + "step": 26883 + }, + { + "epoch": 5.87, + "learning_rate": 2.068603889010269e-06, + "loss": 1.3408, + "step": 26884 + }, + { + "epoch": 5.87, + "learning_rate": 2.0684946471487876e-06, + "loss": 1.41, + "step": 26885 + }, + { + "epoch": 5.87, + "learning_rate": 2.0683854052873062e-06, + "loss": 1.0789, + "step": 26886 + }, + { + "epoch": 5.87, + "learning_rate": 2.068276163425825e-06, + "loss": 1.4164, + "step": 26887 + }, + { + "epoch": 5.87, + "learning_rate": 2.0681669215643435e-06, + "loss": 0.7795, + "step": 26888 + }, + { + "epoch": 5.87, + "learning_rate": 2.0680576797028625e-06, + "loss": 1.1137, + "step": 26889 + }, + { + "epoch": 5.87, + "learning_rate": 2.067948437841381e-06, + "loss": 3.0053, + "step": 26890 + }, + { + "epoch": 5.87, + "learning_rate": 2.0678391959798997e-06, + "loss": 1.4674, + "step": 26891 + }, + { + "epoch": 5.87, + "learning_rate": 2.0677299541184183e-06, + "loss": 0.7733, + "step": 26892 + }, + { + "epoch": 5.87, + "learning_rate": 2.067620712256937e-06, + "loss": 1.7747, + "step": 26893 + }, + { + "epoch": 5.87, + "learning_rate": 2.0675114703954555e-06, + "loss": 1.5981, + "step": 26894 + }, + { + "epoch": 5.87, + "learning_rate": 2.067402228533974e-06, + "loss": 1.8142, + "step": 26895 + }, + { + "epoch": 5.87, + "learning_rate": 2.067292986672493e-06, + "loss": 2.2752, + "step": 26896 + }, + { + "epoch": 5.87, + "learning_rate": 2.067183744811012e-06, + "loss": 2.4033, + "step": 26897 + }, + { + "epoch": 5.87, + "learning_rate": 2.0670745029495304e-06, + "loss": 0.6503, + "step": 26898 + }, + { + "epoch": 5.87, + "learning_rate": 2.066965261088049e-06, + "loss": 2.6229, + "step": 26899 + }, + { + "epoch": 5.87, + "learning_rate": 2.0668560192265676e-06, + "loss": 1.308, + "step": 26900 + }, + { + "epoch": 5.87, + "learning_rate": 2.0667467773650867e-06, + "loss": 0.7178, + "step": 26901 + }, + { + "epoch": 5.87, + "learning_rate": 2.066637535503605e-06, + "loss": 0.3329, + "step": 26902 + }, + { + "epoch": 5.87, + "learning_rate": 2.066528293642124e-06, + "loss": 1.2975, + "step": 26903 + }, + { + "epoch": 5.87, + "learning_rate": 2.0664190517806425e-06, + "loss": 2.4321, + "step": 26904 + }, + { + "epoch": 5.87, + "learning_rate": 2.066309809919161e-06, + "loss": 1.3677, + "step": 26905 + }, + { + "epoch": 5.87, + "learning_rate": 2.0662005680576797e-06, + "loss": 1.0154, + "step": 26906 + }, + { + "epoch": 5.87, + "learning_rate": 2.0660913261961983e-06, + "loss": 1.0117, + "step": 26907 + }, + { + "epoch": 5.87, + "learning_rate": 2.0659820843347174e-06, + "loss": 0.4556, + "step": 26908 + }, + { + "epoch": 5.87, + "learning_rate": 2.065872842473236e-06, + "loss": 2.6469, + "step": 26909 + }, + { + "epoch": 5.87, + "learning_rate": 2.0657636006117546e-06, + "loss": 1.036, + "step": 26910 + }, + { + "epoch": 5.87, + "learning_rate": 2.065654358750273e-06, + "loss": 2.3191, + "step": 26911 + }, + { + "epoch": 5.87, + "learning_rate": 2.065545116888792e-06, + "loss": 0.7303, + "step": 26912 + }, + { + "epoch": 5.87, + "learning_rate": 2.065435875027311e-06, + "loss": 1.0852, + "step": 26913 + }, + { + "epoch": 5.87, + "learning_rate": 2.065326633165829e-06, + "loss": 2.2491, + "step": 26914 + }, + { + "epoch": 5.87, + "learning_rate": 2.065217391304348e-06, + "loss": 0.8671, + "step": 26915 + }, + { + "epoch": 5.87, + "learning_rate": 2.0651081494428667e-06, + "loss": 1.9398, + "step": 26916 + }, + { + "epoch": 5.87, + "learning_rate": 2.0649989075813853e-06, + "loss": 0.5925, + "step": 26917 + }, + { + "epoch": 5.87, + "learning_rate": 2.0648896657199043e-06, + "loss": 0.7627, + "step": 26918 + }, + { + "epoch": 5.87, + "learning_rate": 2.0647804238584225e-06, + "loss": 1.6017, + "step": 26919 + }, + { + "epoch": 5.88, + "learning_rate": 2.0646711819969415e-06, + "loss": 1.4005, + "step": 26920 + }, + { + "epoch": 5.88, + "learning_rate": 2.06456194013546e-06, + "loss": 0.5805, + "step": 26921 + }, + { + "epoch": 5.88, + "learning_rate": 2.0644526982739788e-06, + "loss": 0.699, + "step": 26922 + }, + { + "epoch": 5.88, + "learning_rate": 2.0643434564124974e-06, + "loss": 2.4818, + "step": 26923 + }, + { + "epoch": 5.88, + "learning_rate": 2.064234214551016e-06, + "loss": 1.0861, + "step": 26924 + }, + { + "epoch": 5.88, + "learning_rate": 2.064124972689535e-06, + "loss": 1.03, + "step": 26925 + }, + { + "epoch": 5.88, + "learning_rate": 2.0640157308280532e-06, + "loss": 1.4481, + "step": 26926 + }, + { + "epoch": 5.88, + "learning_rate": 2.0639064889665722e-06, + "loss": 2.472, + "step": 26927 + }, + { + "epoch": 5.88, + "learning_rate": 2.063797247105091e-06, + "loss": 1.2173, + "step": 26928 + }, + { + "epoch": 5.88, + "learning_rate": 2.0636880052436095e-06, + "loss": 2.1308, + "step": 26929 + }, + { + "epoch": 5.88, + "learning_rate": 2.0635787633821285e-06, + "loss": 1.8578, + "step": 26930 + }, + { + "epoch": 5.88, + "learning_rate": 2.0634695215206467e-06, + "loss": 1.5274, + "step": 26931 + }, + { + "epoch": 5.88, + "learning_rate": 2.0633602796591657e-06, + "loss": 0.3716, + "step": 26932 + }, + { + "epoch": 5.88, + "learning_rate": 2.0632510377976843e-06, + "loss": 3.0074, + "step": 26933 + }, + { + "epoch": 5.88, + "learning_rate": 2.063141795936203e-06, + "loss": 0.1349, + "step": 26934 + }, + { + "epoch": 5.88, + "learning_rate": 2.0630325540747216e-06, + "loss": 0.8331, + "step": 26935 + }, + { + "epoch": 5.88, + "learning_rate": 2.06292331221324e-06, + "loss": 0.5541, + "step": 26936 + }, + { + "epoch": 5.88, + "learning_rate": 2.062814070351759e-06, + "loss": 1.278, + "step": 26937 + }, + { + "epoch": 5.88, + "learning_rate": 2.0627048284902774e-06, + "loss": 2.2309, + "step": 26938 + }, + { + "epoch": 5.88, + "learning_rate": 2.0625955866287964e-06, + "loss": 0.7106, + "step": 26939 + }, + { + "epoch": 5.88, + "learning_rate": 2.062486344767315e-06, + "loss": 2.0797, + "step": 26940 + }, + { + "epoch": 5.88, + "learning_rate": 2.0623771029058336e-06, + "loss": 1.0276, + "step": 26941 + }, + { + "epoch": 5.88, + "learning_rate": 2.0622678610443523e-06, + "loss": 2.0663, + "step": 26942 + }, + { + "epoch": 5.88, + "learning_rate": 2.062158619182871e-06, + "loss": 0.696, + "step": 26943 + }, + { + "epoch": 5.88, + "learning_rate": 2.06204937732139e-06, + "loss": 0.2196, + "step": 26944 + }, + { + "epoch": 5.88, + "learning_rate": 2.0619401354599085e-06, + "loss": 1.4193, + "step": 26945 + }, + { + "epoch": 5.88, + "learning_rate": 2.061830893598427e-06, + "loss": 0.8751, + "step": 26946 + }, + { + "epoch": 5.88, + "learning_rate": 2.0617216517369457e-06, + "loss": 0.7212, + "step": 26947 + }, + { + "epoch": 5.88, + "learning_rate": 2.0616124098754643e-06, + "loss": 2.5571, + "step": 26948 + }, + { + "epoch": 5.88, + "learning_rate": 2.0615031680139834e-06, + "loss": 2.4869, + "step": 26949 + }, + { + "epoch": 5.88, + "learning_rate": 2.0613939261525016e-06, + "loss": 1.941, + "step": 26950 + }, + { + "epoch": 5.88, + "learning_rate": 2.0612846842910206e-06, + "loss": 0.9299, + "step": 26951 + }, + { + "epoch": 5.88, + "learning_rate": 2.0611754424295392e-06, + "loss": 1.9981, + "step": 26952 + }, + { + "epoch": 5.88, + "learning_rate": 2.061066200568058e-06, + "loss": 1.1809, + "step": 26953 + }, + { + "epoch": 5.88, + "learning_rate": 2.0609569587065764e-06, + "loss": 1.9259, + "step": 26954 + }, + { + "epoch": 5.88, + "learning_rate": 2.060847716845095e-06, + "loss": 2.7922, + "step": 26955 + }, + { + "epoch": 5.88, + "learning_rate": 2.060738474983614e-06, + "loss": 1.1866, + "step": 26956 + }, + { + "epoch": 5.88, + "learning_rate": 2.0606292331221327e-06, + "loss": 2.1624, + "step": 26957 + }, + { + "epoch": 5.88, + "learning_rate": 2.0605199912606513e-06, + "loss": 0.882, + "step": 26958 + }, + { + "epoch": 5.88, + "learning_rate": 2.06041074939917e-06, + "loss": 2.2515, + "step": 26959 + }, + { + "epoch": 5.88, + "learning_rate": 2.0603015075376885e-06, + "loss": 2.1428, + "step": 26960 + }, + { + "epoch": 5.88, + "learning_rate": 2.060192265676207e-06, + "loss": 0.89, + "step": 26961 + }, + { + "epoch": 5.88, + "learning_rate": 2.060083023814726e-06, + "loss": 1.3606, + "step": 26962 + }, + { + "epoch": 5.88, + "learning_rate": 2.0599737819532448e-06, + "loss": 1.6072, + "step": 26963 + }, + { + "epoch": 5.88, + "learning_rate": 2.0598645400917634e-06, + "loss": 1.8343, + "step": 26964 + }, + { + "epoch": 5.88, + "learning_rate": 2.059755298230282e-06, + "loss": 0.341, + "step": 26965 + }, + { + "epoch": 5.89, + "learning_rate": 2.0596460563688006e-06, + "loss": 0.5673, + "step": 26966 + }, + { + "epoch": 5.89, + "learning_rate": 2.0595368145073192e-06, + "loss": 1.8025, + "step": 26967 + }, + { + "epoch": 5.89, + "learning_rate": 2.059427572645838e-06, + "loss": 0.9047, + "step": 26968 + }, + { + "epoch": 5.89, + "learning_rate": 2.059318330784357e-06, + "loss": 1.3587, + "step": 26969 + }, + { + "epoch": 5.89, + "learning_rate": 2.0592090889228755e-06, + "loss": 1.3416, + "step": 26970 + }, + { + "epoch": 5.89, + "learning_rate": 2.059099847061394e-06, + "loss": 0.9382, + "step": 26971 + }, + { + "epoch": 5.89, + "learning_rate": 2.0589906051999127e-06, + "loss": 1.6591, + "step": 26972 + }, + { + "epoch": 5.89, + "learning_rate": 2.0588813633384313e-06, + "loss": 1.4662, + "step": 26973 + }, + { + "epoch": 5.89, + "learning_rate": 2.0587721214769504e-06, + "loss": 0.8233, + "step": 26974 + }, + { + "epoch": 5.89, + "learning_rate": 2.0586628796154685e-06, + "loss": 2.6099, + "step": 26975 + }, + { + "epoch": 5.89, + "learning_rate": 2.0585536377539876e-06, + "loss": 3.0134, + "step": 26976 + }, + { + "epoch": 5.89, + "learning_rate": 2.058444395892506e-06, + "loss": 1.4773, + "step": 26977 + }, + { + "epoch": 5.89, + "learning_rate": 2.058335154031025e-06, + "loss": 1.2838, + "step": 26978 + }, + { + "epoch": 5.89, + "learning_rate": 2.0582259121695434e-06, + "loss": 0.8177, + "step": 26979 + }, + { + "epoch": 5.89, + "learning_rate": 2.058116670308062e-06, + "loss": 1.6487, + "step": 26980 + }, + { + "epoch": 5.89, + "learning_rate": 2.058007428446581e-06, + "loss": 1.5401, + "step": 26981 + }, + { + "epoch": 5.89, + "learning_rate": 2.0578981865850997e-06, + "loss": 0.4363, + "step": 26982 + }, + { + "epoch": 5.89, + "learning_rate": 2.0577889447236183e-06, + "loss": 2.3594, + "step": 26983 + }, + { + "epoch": 5.89, + "learning_rate": 2.057679702862137e-06, + "loss": 1.4113, + "step": 26984 + }, + { + "epoch": 5.89, + "learning_rate": 2.0575704610006555e-06, + "loss": 1.1781, + "step": 26985 + }, + { + "epoch": 5.89, + "learning_rate": 2.0574612191391745e-06, + "loss": 2.5781, + "step": 26986 + }, + { + "epoch": 5.89, + "learning_rate": 2.0573519772776927e-06, + "loss": 1.8195, + "step": 26987 + }, + { + "epoch": 5.89, + "learning_rate": 2.0572427354162118e-06, + "loss": 0.6539, + "step": 26988 + }, + { + "epoch": 5.89, + "learning_rate": 2.0571334935547304e-06, + "loss": 1.8194, + "step": 26989 + }, + { + "epoch": 5.89, + "learning_rate": 2.057024251693249e-06, + "loss": 0.194, + "step": 26990 + }, + { + "epoch": 5.89, + "learning_rate": 2.0569150098317676e-06, + "loss": 1.8941, + "step": 26991 + }, + { + "epoch": 5.89, + "learning_rate": 2.056805767970286e-06, + "loss": 1.0197, + "step": 26992 + }, + { + "epoch": 5.89, + "learning_rate": 2.0566965261088052e-06, + "loss": 1.0518, + "step": 26993 + }, + { + "epoch": 5.89, + "learning_rate": 2.0565872842473234e-06, + "loss": 0.6815, + "step": 26994 + }, + { + "epoch": 5.89, + "learning_rate": 2.0564780423858425e-06, + "loss": 2.8, + "step": 26995 + }, + { + "epoch": 5.89, + "learning_rate": 2.056368800524361e-06, + "loss": 2.8669, + "step": 26996 + }, + { + "epoch": 5.89, + "learning_rate": 2.0562595586628797e-06, + "loss": 2.5188, + "step": 26997 + }, + { + "epoch": 5.89, + "learning_rate": 2.0561503168013987e-06, + "loss": 1.8277, + "step": 26998 + }, + { + "epoch": 5.89, + "learning_rate": 2.056041074939917e-06, + "loss": 0.7922, + "step": 26999 + }, + { + "epoch": 5.89, + "learning_rate": 2.055931833078436e-06, + "loss": 2.5506, + "step": 27000 + }, + { + "epoch": 5.89, + "learning_rate": 2.0558225912169545e-06, + "loss": 0.5501, + "step": 27001 + }, + { + "epoch": 5.89, + "learning_rate": 2.055713349355473e-06, + "loss": 2.5043, + "step": 27002 + }, + { + "epoch": 5.89, + "learning_rate": 2.0556041074939918e-06, + "loss": 1.2241, + "step": 27003 + }, + { + "epoch": 5.89, + "learning_rate": 2.0554948656325104e-06, + "loss": 0.843, + "step": 27004 + }, + { + "epoch": 5.89, + "learning_rate": 2.0553856237710294e-06, + "loss": 0.565, + "step": 27005 + }, + { + "epoch": 5.89, + "learning_rate": 2.055276381909548e-06, + "loss": 1.4203, + "step": 27006 + }, + { + "epoch": 5.89, + "learning_rate": 2.0551671400480666e-06, + "loss": 0.366, + "step": 27007 + }, + { + "epoch": 5.89, + "learning_rate": 2.0550578981865852e-06, + "loss": 2.1761, + "step": 27008 + }, + { + "epoch": 5.89, + "learning_rate": 2.054948656325104e-06, + "loss": 2.0477, + "step": 27009 + }, + { + "epoch": 5.89, + "learning_rate": 2.054839414463623e-06, + "loss": 2.0799, + "step": 27010 + }, + { + "epoch": 5.9, + "learning_rate": 2.054730172602141e-06, + "loss": 0.8355, + "step": 27011 + }, + { + "epoch": 5.9, + "learning_rate": 2.05462093074066e-06, + "loss": 0.7686, + "step": 27012 + }, + { + "epoch": 5.9, + "learning_rate": 2.0545116888791787e-06, + "loss": 0.1021, + "step": 27013 + }, + { + "epoch": 5.9, + "learning_rate": 2.0544024470176973e-06, + "loss": 1.2157, + "step": 27014 + }, + { + "epoch": 5.9, + "learning_rate": 2.0542932051562164e-06, + "loss": 1.3296, + "step": 27015 + }, + { + "epoch": 5.9, + "learning_rate": 2.0541839632947346e-06, + "loss": 0.7824, + "step": 27016 + }, + { + "epoch": 5.9, + "learning_rate": 2.0540747214332536e-06, + "loss": 0.8149, + "step": 27017 + }, + { + "epoch": 5.9, + "learning_rate": 2.053965479571772e-06, + "loss": 0.5641, + "step": 27018 + }, + { + "epoch": 5.9, + "learning_rate": 2.053856237710291e-06, + "loss": 1.5471, + "step": 27019 + }, + { + "epoch": 5.9, + "learning_rate": 2.0537469958488094e-06, + "loss": 0.4981, + "step": 27020 + }, + { + "epoch": 5.9, + "learning_rate": 2.053637753987328e-06, + "loss": 2.026, + "step": 27021 + }, + { + "epoch": 5.9, + "learning_rate": 2.053528512125847e-06, + "loss": 1.9101, + "step": 27022 + }, + { + "epoch": 5.9, + "learning_rate": 2.0534192702643653e-06, + "loss": 1.6702, + "step": 27023 + }, + { + "epoch": 5.9, + "learning_rate": 2.0533100284028843e-06, + "loss": 0.8772, + "step": 27024 + }, + { + "epoch": 5.9, + "learning_rate": 2.053200786541403e-06, + "loss": 0.5659, + "step": 27025 + }, + { + "epoch": 5.9, + "learning_rate": 2.0530915446799215e-06, + "loss": 0.9572, + "step": 27026 + }, + { + "epoch": 5.9, + "learning_rate": 2.05298230281844e-06, + "loss": 1.2521, + "step": 27027 + }, + { + "epoch": 5.9, + "learning_rate": 2.0528730609569587e-06, + "loss": 1.2034, + "step": 27028 + }, + { + "epoch": 5.9, + "learning_rate": 2.0527638190954778e-06, + "loss": 2.2959, + "step": 27029 + }, + { + "epoch": 5.9, + "learning_rate": 2.0526545772339964e-06, + "loss": 3.2567, + "step": 27030 + }, + { + "epoch": 5.9, + "learning_rate": 2.052545335372515e-06, + "loss": 0.5893, + "step": 27031 + }, + { + "epoch": 5.9, + "learning_rate": 2.0524360935110336e-06, + "loss": 1.8263, + "step": 27032 + }, + { + "epoch": 5.9, + "learning_rate": 2.0523268516495522e-06, + "loss": 1.0202, + "step": 27033 + }, + { + "epoch": 5.9, + "learning_rate": 2.052217609788071e-06, + "loss": 3.058, + "step": 27034 + }, + { + "epoch": 5.9, + "learning_rate": 2.0521083679265894e-06, + "loss": 2.6869, + "step": 27035 + }, + { + "epoch": 5.9, + "learning_rate": 2.0519991260651085e-06, + "loss": 3.577, + "step": 27036 + }, + { + "epoch": 5.9, + "learning_rate": 2.051889884203627e-06, + "loss": 1.2577, + "step": 27037 + }, + { + "epoch": 5.9, + "learning_rate": 2.0517806423421457e-06, + "loss": 2.2874, + "step": 27038 + }, + { + "epoch": 5.9, + "learning_rate": 2.0516714004806643e-06, + "loss": 1.8263, + "step": 27039 + }, + { + "epoch": 5.9, + "learning_rate": 2.051562158619183e-06, + "loss": 1.5489, + "step": 27040 + }, + { + "epoch": 5.9, + "learning_rate": 2.0514529167577015e-06, + "loss": 1.4767, + "step": 27041 + }, + { + "epoch": 5.9, + "learning_rate": 2.0513436748962206e-06, + "loss": 1.4029, + "step": 27042 + }, + { + "epoch": 5.9, + "learning_rate": 2.051234433034739e-06, + "loss": 2.11, + "step": 27043 + }, + { + "epoch": 5.9, + "learning_rate": 2.0511251911732578e-06, + "loss": 0.8677, + "step": 27044 + }, + { + "epoch": 5.9, + "learning_rate": 2.0510159493117764e-06, + "loss": 2.0321, + "step": 27045 + }, + { + "epoch": 5.9, + "learning_rate": 2.050906707450295e-06, + "loss": 0.4071, + "step": 27046 + }, + { + "epoch": 5.9, + "learning_rate": 2.0507974655888136e-06, + "loss": 0.1851, + "step": 27047 + }, + { + "epoch": 5.9, + "learning_rate": 2.0506882237273326e-06, + "loss": 1.2171, + "step": 27048 + }, + { + "epoch": 5.9, + "learning_rate": 2.0505789818658513e-06, + "loss": 1.2547, + "step": 27049 + }, + { + "epoch": 5.9, + "learning_rate": 2.05046974000437e-06, + "loss": 2.55, + "step": 27050 + }, + { + "epoch": 5.9, + "learning_rate": 2.0503604981428885e-06, + "loss": 1.8183, + "step": 27051 + }, + { + "epoch": 5.9, + "learning_rate": 2.050251256281407e-06, + "loss": 2.226, + "step": 27052 + }, + { + "epoch": 5.9, + "learning_rate": 2.0501420144199257e-06, + "loss": 1.9445, + "step": 27053 + }, + { + "epoch": 5.9, + "learning_rate": 2.0500327725584447e-06, + "loss": 3.1154, + "step": 27054 + }, + { + "epoch": 5.9, + "learning_rate": 2.0499235306969633e-06, + "loss": 2.4247, + "step": 27055 + }, + { + "epoch": 5.9, + "learning_rate": 2.049814288835482e-06, + "loss": 0.973, + "step": 27056 + }, + { + "epoch": 5.91, + "learning_rate": 2.0497050469740006e-06, + "loss": 2.0545, + "step": 27057 + }, + { + "epoch": 5.91, + "learning_rate": 2.049595805112519e-06, + "loss": 0.752, + "step": 27058 + }, + { + "epoch": 5.91, + "learning_rate": 2.0494865632510382e-06, + "loss": 0.8665, + "step": 27059 + }, + { + "epoch": 5.91, + "learning_rate": 2.0493773213895564e-06, + "loss": 1.2822, + "step": 27060 + }, + { + "epoch": 5.91, + "learning_rate": 2.0492680795280754e-06, + "loss": 1.3894, + "step": 27061 + }, + { + "epoch": 5.91, + "learning_rate": 2.049158837666594e-06, + "loss": 1.4192, + "step": 27062 + }, + { + "epoch": 5.91, + "learning_rate": 2.0490495958051127e-06, + "loss": 2.407, + "step": 27063 + }, + { + "epoch": 5.91, + "learning_rate": 2.0489403539436313e-06, + "loss": 2.5941, + "step": 27064 + }, + { + "epoch": 5.91, + "learning_rate": 2.04883111208215e-06, + "loss": 3.7841, + "step": 27065 + }, + { + "epoch": 5.91, + "learning_rate": 2.048721870220669e-06, + "loss": 1.2522, + "step": 27066 + }, + { + "epoch": 5.91, + "learning_rate": 2.048612628359187e-06, + "loss": 0.2382, + "step": 27067 + }, + { + "epoch": 5.91, + "learning_rate": 2.048503386497706e-06, + "loss": 2.0448, + "step": 27068 + }, + { + "epoch": 5.91, + "learning_rate": 2.0483941446362247e-06, + "loss": 2.0525, + "step": 27069 + }, + { + "epoch": 5.91, + "learning_rate": 2.0482849027747434e-06, + "loss": 2.5227, + "step": 27070 + }, + { + "epoch": 5.91, + "learning_rate": 2.0481756609132624e-06, + "loss": 1.634, + "step": 27071 + }, + { + "epoch": 5.91, + "learning_rate": 2.0480664190517806e-06, + "loss": 0.4613, + "step": 27072 + }, + { + "epoch": 5.91, + "learning_rate": 2.0479571771902996e-06, + "loss": 2.0292, + "step": 27073 + }, + { + "epoch": 5.91, + "learning_rate": 2.0478479353288182e-06, + "loss": 1.2979, + "step": 27074 + }, + { + "epoch": 5.91, + "learning_rate": 2.047738693467337e-06, + "loss": 1.2857, + "step": 27075 + }, + { + "epoch": 5.91, + "learning_rate": 2.0476294516058554e-06, + "loss": 0.7769, + "step": 27076 + }, + { + "epoch": 5.91, + "learning_rate": 2.047520209744374e-06, + "loss": 1.5317, + "step": 27077 + }, + { + "epoch": 5.91, + "learning_rate": 2.047410967882893e-06, + "loss": 0.442, + "step": 27078 + }, + { + "epoch": 5.91, + "learning_rate": 2.0473017260214113e-06, + "loss": 0.5146, + "step": 27079 + }, + { + "epoch": 5.91, + "learning_rate": 2.0471924841599303e-06, + "loss": 1.6988, + "step": 27080 + }, + { + "epoch": 5.91, + "learning_rate": 2.047083242298449e-06, + "loss": 0.7948, + "step": 27081 + }, + { + "epoch": 5.91, + "learning_rate": 2.0469740004369675e-06, + "loss": 1.4571, + "step": 27082 + }, + { + "epoch": 5.91, + "learning_rate": 2.0468647585754866e-06, + "loss": 0.0847, + "step": 27083 + }, + { + "epoch": 5.91, + "learning_rate": 2.0467555167140048e-06, + "loss": 0.9429, + "step": 27084 + }, + { + "epoch": 5.91, + "learning_rate": 2.046646274852524e-06, + "loss": 0.5941, + "step": 27085 + }, + { + "epoch": 5.91, + "learning_rate": 2.0465370329910424e-06, + "loss": 0.4504, + "step": 27086 + }, + { + "epoch": 5.91, + "learning_rate": 2.046427791129561e-06, + "loss": 1.3404, + "step": 27087 + }, + { + "epoch": 5.91, + "learning_rate": 2.0463185492680796e-06, + "loss": 0.7218, + "step": 27088 + }, + { + "epoch": 5.91, + "learning_rate": 2.0462093074065982e-06, + "loss": 0.3483, + "step": 27089 + }, + { + "epoch": 5.91, + "learning_rate": 2.0461000655451173e-06, + "loss": 1.487, + "step": 27090 + }, + { + "epoch": 5.91, + "learning_rate": 2.0459908236836355e-06, + "loss": 0.3225, + "step": 27091 + }, + { + "epoch": 5.91, + "learning_rate": 2.0458815818221545e-06, + "loss": 0.9006, + "step": 27092 + }, + { + "epoch": 5.91, + "learning_rate": 2.045772339960673e-06, + "loss": 0.0612, + "step": 27093 + }, + { + "epoch": 5.91, + "learning_rate": 2.0456630980991917e-06, + "loss": 1.2624, + "step": 27094 + }, + { + "epoch": 5.91, + "learning_rate": 2.0455538562377108e-06, + "loss": 1.6805, + "step": 27095 + }, + { + "epoch": 5.91, + "learning_rate": 2.045444614376229e-06, + "loss": 1.1171, + "step": 27096 + }, + { + "epoch": 5.91, + "learning_rate": 2.045335372514748e-06, + "loss": 0.2621, + "step": 27097 + }, + { + "epoch": 5.91, + "learning_rate": 2.0452261306532666e-06, + "loss": 1.5921, + "step": 27098 + }, + { + "epoch": 5.91, + "learning_rate": 2.045116888791785e-06, + "loss": 2.468, + "step": 27099 + }, + { + "epoch": 5.91, + "learning_rate": 2.045007646930304e-06, + "loss": 0.6879, + "step": 27100 + }, + { + "epoch": 5.91, + "learning_rate": 2.0448984050688224e-06, + "loss": 0.3047, + "step": 27101 + }, + { + "epoch": 5.91, + "learning_rate": 2.0447891632073415e-06, + "loss": 1.819, + "step": 27102 + }, + { + "epoch": 5.92, + "learning_rate": 2.04467992134586e-06, + "loss": 1.5064, + "step": 27103 + }, + { + "epoch": 5.92, + "learning_rate": 2.0445706794843787e-06, + "loss": 1.2823, + "step": 27104 + }, + { + "epoch": 5.92, + "learning_rate": 2.0444614376228973e-06, + "loss": 1.0146, + "step": 27105 + }, + { + "epoch": 5.92, + "learning_rate": 2.044352195761416e-06, + "loss": 0.5748, + "step": 27106 + }, + { + "epoch": 5.92, + "learning_rate": 2.0442429538999345e-06, + "loss": 1.5417, + "step": 27107 + }, + { + "epoch": 5.92, + "learning_rate": 2.044133712038453e-06, + "loss": 2.3411, + "step": 27108 + }, + { + "epoch": 5.92, + "learning_rate": 2.044024470176972e-06, + "loss": 2.6607, + "step": 27109 + }, + { + "epoch": 5.92, + "learning_rate": 2.0439152283154908e-06, + "loss": 0.7014, + "step": 27110 + }, + { + "epoch": 5.92, + "learning_rate": 2.0438059864540094e-06, + "loss": 0.8088, + "step": 27111 + }, + { + "epoch": 5.92, + "learning_rate": 2.043696744592528e-06, + "loss": 1.498, + "step": 27112 + }, + { + "epoch": 5.92, + "learning_rate": 2.0435875027310466e-06, + "loss": 1.1889, + "step": 27113 + }, + { + "epoch": 5.92, + "learning_rate": 2.0434782608695656e-06, + "loss": 1.3065, + "step": 27114 + }, + { + "epoch": 5.92, + "learning_rate": 2.0433690190080842e-06, + "loss": 0.9544, + "step": 27115 + }, + { + "epoch": 5.92, + "learning_rate": 2.043259777146603e-06, + "loss": 1.7797, + "step": 27116 + }, + { + "epoch": 5.92, + "learning_rate": 2.0431505352851215e-06, + "loss": 0.9618, + "step": 27117 + }, + { + "epoch": 5.92, + "learning_rate": 2.04304129342364e-06, + "loss": 1.7403, + "step": 27118 + }, + { + "epoch": 5.92, + "learning_rate": 2.0429320515621587e-06, + "loss": 1.516, + "step": 27119 + }, + { + "epoch": 5.92, + "learning_rate": 2.0428228097006773e-06, + "loss": 0.2756, + "step": 27120 + }, + { + "epoch": 5.92, + "learning_rate": 2.0427135678391963e-06, + "loss": 0.6784, + "step": 27121 + }, + { + "epoch": 5.92, + "learning_rate": 2.042604325977715e-06, + "loss": 1.5449, + "step": 27122 + }, + { + "epoch": 5.92, + "learning_rate": 2.0424950841162336e-06, + "loss": 1.6946, + "step": 27123 + }, + { + "epoch": 5.92, + "learning_rate": 2.042385842254752e-06, + "loss": 0.8451, + "step": 27124 + }, + { + "epoch": 5.92, + "learning_rate": 2.0422766003932708e-06, + "loss": 1.2777, + "step": 27125 + }, + { + "epoch": 5.92, + "learning_rate": 2.0421673585317894e-06, + "loss": 0.9593, + "step": 27126 + }, + { + "epoch": 5.92, + "learning_rate": 2.0420581166703084e-06, + "loss": 1.5997, + "step": 27127 + }, + { + "epoch": 5.92, + "learning_rate": 2.041948874808827e-06, + "loss": 2.7625, + "step": 27128 + }, + { + "epoch": 5.92, + "learning_rate": 2.0418396329473456e-06, + "loss": 0.3785, + "step": 27129 + }, + { + "epoch": 5.92, + "learning_rate": 2.0417303910858643e-06, + "loss": 1.5165, + "step": 27130 + }, + { + "epoch": 5.92, + "learning_rate": 2.041621149224383e-06, + "loss": 0.76, + "step": 27131 + }, + { + "epoch": 5.92, + "learning_rate": 2.0415119073629015e-06, + "loss": 0.7849, + "step": 27132 + }, + { + "epoch": 5.92, + "learning_rate": 2.04140266550142e-06, + "loss": 0.5294, + "step": 27133 + }, + { + "epoch": 5.92, + "learning_rate": 2.041293423639939e-06, + "loss": 1.2894, + "step": 27134 + }, + { + "epoch": 5.92, + "learning_rate": 2.0411841817784577e-06, + "loss": 1.8236, + "step": 27135 + }, + { + "epoch": 5.92, + "learning_rate": 2.0410749399169763e-06, + "loss": 1.0373, + "step": 27136 + }, + { + "epoch": 5.92, + "learning_rate": 2.040965698055495e-06, + "loss": 0.4121, + "step": 27137 + }, + { + "epoch": 5.92, + "learning_rate": 2.0408564561940136e-06, + "loss": 1.3022, + "step": 27138 + }, + { + "epoch": 5.92, + "learning_rate": 2.0407472143325326e-06, + "loss": 0.6816, + "step": 27139 + }, + { + "epoch": 5.92, + "learning_rate": 2.0406379724710508e-06, + "loss": 1.5002, + "step": 27140 + }, + { + "epoch": 5.92, + "learning_rate": 2.04052873060957e-06, + "loss": 1.3306, + "step": 27141 + }, + { + "epoch": 5.92, + "learning_rate": 2.0404194887480884e-06, + "loss": 1.8676, + "step": 27142 + }, + { + "epoch": 5.92, + "learning_rate": 2.040310246886607e-06, + "loss": 0.4129, + "step": 27143 + }, + { + "epoch": 5.92, + "learning_rate": 2.0402010050251257e-06, + "loss": 0.1548, + "step": 27144 + }, + { + "epoch": 5.92, + "learning_rate": 2.0400917631636443e-06, + "loss": 1.245, + "step": 27145 + }, + { + "epoch": 5.92, + "learning_rate": 2.0399825213021633e-06, + "loss": 0.4967, + "step": 27146 + }, + { + "epoch": 5.92, + "learning_rate": 2.0398732794406815e-06, + "loss": 1.0519, + "step": 27147 + }, + { + "epoch": 5.92, + "learning_rate": 2.0397640375792005e-06, + "loss": 0.8866, + "step": 27148 + }, + { + "epoch": 5.93, + "learning_rate": 2.039654795717719e-06, + "loss": 0.2508, + "step": 27149 + }, + { + "epoch": 5.93, + "learning_rate": 2.0395455538562377e-06, + "loss": 1.3857, + "step": 27150 + }, + { + "epoch": 5.93, + "learning_rate": 2.0394363119947568e-06, + "loss": 1.3957, + "step": 27151 + }, + { + "epoch": 5.93, + "learning_rate": 2.039327070133275e-06, + "loss": 1.007, + "step": 27152 + }, + { + "epoch": 5.93, + "learning_rate": 2.039217828271794e-06, + "loss": 0.4655, + "step": 27153 + }, + { + "epoch": 5.93, + "learning_rate": 2.0391085864103126e-06, + "loss": 0.6947, + "step": 27154 + }, + { + "epoch": 5.93, + "learning_rate": 2.0389993445488312e-06, + "loss": 1.2079, + "step": 27155 + }, + { + "epoch": 5.93, + "learning_rate": 2.0388901026873503e-06, + "loss": 0.8357, + "step": 27156 + }, + { + "epoch": 5.93, + "learning_rate": 2.0387808608258684e-06, + "loss": 1.5037, + "step": 27157 + }, + { + "epoch": 5.93, + "learning_rate": 2.0386716189643875e-06, + "loss": 0.7763, + "step": 27158 + }, + { + "epoch": 5.93, + "learning_rate": 2.038562377102906e-06, + "loss": 0.6749, + "step": 27159 + }, + { + "epoch": 5.93, + "learning_rate": 2.0384531352414247e-06, + "loss": 2.1702, + "step": 27160 + }, + { + "epoch": 5.93, + "learning_rate": 2.0383438933799433e-06, + "loss": 4.0165, + "step": 27161 + }, + { + "epoch": 5.93, + "learning_rate": 2.038234651518462e-06, + "loss": 1.0283, + "step": 27162 + }, + { + "epoch": 5.93, + "learning_rate": 2.038125409656981e-06, + "loss": 0.8228, + "step": 27163 + }, + { + "epoch": 5.93, + "learning_rate": 2.038016167795499e-06, + "loss": 1.0739, + "step": 27164 + }, + { + "epoch": 5.93, + "learning_rate": 2.037906925934018e-06, + "loss": 3.0841, + "step": 27165 + }, + { + "epoch": 5.93, + "learning_rate": 2.037797684072537e-06, + "loss": 0.9406, + "step": 27166 + }, + { + "epoch": 5.93, + "learning_rate": 2.0376884422110554e-06, + "loss": 2.2339, + "step": 27167 + }, + { + "epoch": 5.93, + "learning_rate": 2.0375792003495744e-06, + "loss": 1.5065, + "step": 27168 + }, + { + "epoch": 5.93, + "learning_rate": 2.0374699584880926e-06, + "loss": 0.7328, + "step": 27169 + }, + { + "epoch": 5.93, + "learning_rate": 2.0373607166266117e-06, + "loss": 0.6098, + "step": 27170 + }, + { + "epoch": 5.93, + "learning_rate": 2.0372514747651303e-06, + "loss": 0.1409, + "step": 27171 + }, + { + "epoch": 5.93, + "learning_rate": 2.037142232903649e-06, + "loss": 2.9375, + "step": 27172 + }, + { + "epoch": 5.93, + "learning_rate": 2.0370329910421675e-06, + "loss": 0.7689, + "step": 27173 + }, + { + "epoch": 5.93, + "learning_rate": 2.036923749180686e-06, + "loss": 2.3128, + "step": 27174 + }, + { + "epoch": 5.93, + "learning_rate": 2.036814507319205e-06, + "loss": 1.2992, + "step": 27175 + }, + { + "epoch": 5.93, + "learning_rate": 2.0367052654577233e-06, + "loss": 0.7056, + "step": 27176 + }, + { + "epoch": 5.93, + "learning_rate": 2.0365960235962424e-06, + "loss": 3.5062, + "step": 27177 + }, + { + "epoch": 5.93, + "learning_rate": 2.036486781734761e-06, + "loss": 0.6471, + "step": 27178 + }, + { + "epoch": 5.93, + "learning_rate": 2.0363775398732796e-06, + "loss": 0.6005, + "step": 27179 + }, + { + "epoch": 5.93, + "learning_rate": 2.036268298011798e-06, + "loss": 0.479, + "step": 27180 + }, + { + "epoch": 5.93, + "learning_rate": 2.036159056150317e-06, + "loss": 1.7434, + "step": 27181 + }, + { + "epoch": 5.93, + "learning_rate": 2.036049814288836e-06, + "loss": 1.3514, + "step": 27182 + }, + { + "epoch": 5.93, + "learning_rate": 2.0359405724273544e-06, + "loss": 1.6387, + "step": 27183 + }, + { + "epoch": 5.93, + "learning_rate": 2.035831330565873e-06, + "loss": 2.4014, + "step": 27184 + }, + { + "epoch": 5.93, + "learning_rate": 2.0357220887043917e-06, + "loss": 1.6385, + "step": 27185 + }, + { + "epoch": 5.93, + "learning_rate": 2.0356128468429103e-06, + "loss": 1.6432, + "step": 27186 + }, + { + "epoch": 5.93, + "learning_rate": 2.0355036049814293e-06, + "loss": 2.4491, + "step": 27187 + }, + { + "epoch": 5.93, + "learning_rate": 2.0353943631199475e-06, + "loss": 1.6222, + "step": 27188 + }, + { + "epoch": 5.93, + "learning_rate": 2.0352851212584665e-06, + "loss": 1.7222, + "step": 27189 + }, + { + "epoch": 5.93, + "learning_rate": 2.035175879396985e-06, + "loss": 0.7833, + "step": 27190 + }, + { + "epoch": 5.93, + "learning_rate": 2.0350666375355038e-06, + "loss": 1.0156, + "step": 27191 + }, + { + "epoch": 5.93, + "learning_rate": 2.0349573956740224e-06, + "loss": 2.432, + "step": 27192 + }, + { + "epoch": 5.93, + "learning_rate": 2.034848153812541e-06, + "loss": 0.2824, + "step": 27193 + }, + { + "epoch": 5.93, + "learning_rate": 2.03473891195106e-06, + "loss": 0.4572, + "step": 27194 + }, + { + "epoch": 5.94, + "learning_rate": 2.0346296700895786e-06, + "loss": 2.1348, + "step": 27195 + }, + { + "epoch": 5.94, + "learning_rate": 2.0345204282280972e-06, + "loss": 0.1008, + "step": 27196 + }, + { + "epoch": 5.94, + "learning_rate": 2.034411186366616e-06, + "loss": 1.7984, + "step": 27197 + }, + { + "epoch": 5.94, + "learning_rate": 2.0343019445051345e-06, + "loss": 1.3945, + "step": 27198 + }, + { + "epoch": 5.94, + "learning_rate": 2.034192702643653e-06, + "loss": 1.5167, + "step": 27199 + }, + { + "epoch": 5.94, + "learning_rate": 2.0340834607821717e-06, + "loss": 0.7138, + "step": 27200 + }, + { + "epoch": 5.94, + "learning_rate": 2.0339742189206907e-06, + "loss": 0.8526, + "step": 27201 + }, + { + "epoch": 5.94, + "learning_rate": 2.0338649770592093e-06, + "loss": 0.8831, + "step": 27202 + }, + { + "epoch": 5.94, + "learning_rate": 2.033755735197728e-06, + "loss": 1.7759, + "step": 27203 + }, + { + "epoch": 5.94, + "learning_rate": 2.0336464933362465e-06, + "loss": 0.8827, + "step": 27204 + }, + { + "epoch": 5.94, + "learning_rate": 2.033537251474765e-06, + "loss": 2.3455, + "step": 27205 + }, + { + "epoch": 5.94, + "learning_rate": 2.0334280096132838e-06, + "loss": 1.5656, + "step": 27206 + }, + { + "epoch": 5.94, + "learning_rate": 2.033318767751803e-06, + "loss": 1.2275, + "step": 27207 + }, + { + "epoch": 5.94, + "learning_rate": 2.0332095258903214e-06, + "loss": 0.319, + "step": 27208 + }, + { + "epoch": 5.94, + "learning_rate": 2.03310028402884e-06, + "loss": 2.3492, + "step": 27209 + }, + { + "epoch": 5.94, + "learning_rate": 2.0329910421673586e-06, + "loss": 2.4497, + "step": 27210 + }, + { + "epoch": 5.94, + "learning_rate": 2.0328818003058772e-06, + "loss": 2.1559, + "step": 27211 + }, + { + "epoch": 5.94, + "learning_rate": 2.0327725584443963e-06, + "loss": 1.736, + "step": 27212 + }, + { + "epoch": 5.94, + "learning_rate": 2.0326633165829145e-06, + "loss": 1.3544, + "step": 27213 + }, + { + "epoch": 5.94, + "learning_rate": 2.0325540747214335e-06, + "loss": 1.5402, + "step": 27214 + }, + { + "epoch": 5.94, + "learning_rate": 2.032444832859952e-06, + "loss": 1.9147, + "step": 27215 + }, + { + "epoch": 5.94, + "learning_rate": 2.0323355909984707e-06, + "loss": 1.6492, + "step": 27216 + }, + { + "epoch": 5.94, + "learning_rate": 2.0322263491369893e-06, + "loss": 0.9832, + "step": 27217 + }, + { + "epoch": 5.94, + "learning_rate": 2.032117107275508e-06, + "loss": 1.4777, + "step": 27218 + }, + { + "epoch": 5.94, + "learning_rate": 2.032007865414027e-06, + "loss": 0.8071, + "step": 27219 + }, + { + "epoch": 5.94, + "learning_rate": 2.0318986235525456e-06, + "loss": 2.5182, + "step": 27220 + }, + { + "epoch": 5.94, + "learning_rate": 2.031789381691064e-06, + "loss": 1.178, + "step": 27221 + }, + { + "epoch": 5.94, + "learning_rate": 2.031680139829583e-06, + "loss": 0.6226, + "step": 27222 + }, + { + "epoch": 5.94, + "learning_rate": 2.0315708979681014e-06, + "loss": 0.6667, + "step": 27223 + }, + { + "epoch": 5.94, + "learning_rate": 2.0314616561066205e-06, + "loss": 0.7288, + "step": 27224 + }, + { + "epoch": 5.94, + "learning_rate": 2.0313524142451387e-06, + "loss": 1.3501, + "step": 27225 + }, + { + "epoch": 5.94, + "learning_rate": 2.0312431723836577e-06, + "loss": 0.1495, + "step": 27226 + }, + { + "epoch": 5.94, + "learning_rate": 2.0311339305221763e-06, + "loss": 0.8524, + "step": 27227 + }, + { + "epoch": 5.94, + "learning_rate": 2.031024688660695e-06, + "loss": 1.3481, + "step": 27228 + }, + { + "epoch": 5.94, + "learning_rate": 2.0309154467992135e-06, + "loss": 1.6501, + "step": 27229 + }, + { + "epoch": 5.94, + "learning_rate": 2.030806204937732e-06, + "loss": 1.3582, + "step": 27230 + }, + { + "epoch": 5.94, + "learning_rate": 2.030696963076251e-06, + "loss": 1.7993, + "step": 27231 + }, + { + "epoch": 5.94, + "learning_rate": 2.0305877212147694e-06, + "loss": 1.8938, + "step": 27232 + }, + { + "epoch": 5.94, + "learning_rate": 2.0304784793532884e-06, + "loss": 1.1646, + "step": 27233 + }, + { + "epoch": 5.94, + "learning_rate": 2.030369237491807e-06, + "loss": 2.5732, + "step": 27234 + }, + { + "epoch": 5.94, + "learning_rate": 2.0302599956303256e-06, + "loss": 2.4306, + "step": 27235 + }, + { + "epoch": 5.94, + "learning_rate": 2.0301507537688446e-06, + "loss": 1.8013, + "step": 27236 + }, + { + "epoch": 5.94, + "learning_rate": 2.030041511907363e-06, + "loss": 1.512, + "step": 27237 + }, + { + "epoch": 5.94, + "learning_rate": 2.029932270045882e-06, + "loss": 1.2821, + "step": 27238 + }, + { + "epoch": 5.94, + "learning_rate": 2.0298230281844005e-06, + "loss": 0.7545, + "step": 27239 + }, + { + "epoch": 5.95, + "learning_rate": 2.029713786322919e-06, + "loss": 1.9612, + "step": 27240 + }, + { + "epoch": 5.95, + "learning_rate": 2.0296045444614377e-06, + "loss": 1.7699, + "step": 27241 + }, + { + "epoch": 5.95, + "learning_rate": 2.0294953025999563e-06, + "loss": 3.3814, + "step": 27242 + }, + { + "epoch": 5.95, + "learning_rate": 2.0293860607384753e-06, + "loss": 1.2854, + "step": 27243 + }, + { + "epoch": 5.95, + "learning_rate": 2.0292768188769935e-06, + "loss": 1.9686, + "step": 27244 + }, + { + "epoch": 5.95, + "learning_rate": 2.0291675770155126e-06, + "loss": 1.2438, + "step": 27245 + }, + { + "epoch": 5.95, + "learning_rate": 2.029058335154031e-06, + "loss": 2.6267, + "step": 27246 + }, + { + "epoch": 5.95, + "learning_rate": 2.0289490932925498e-06, + "loss": 1.6078, + "step": 27247 + }, + { + "epoch": 5.95, + "learning_rate": 2.028839851431069e-06, + "loss": 2.8869, + "step": 27248 + }, + { + "epoch": 5.95, + "learning_rate": 2.028730609569587e-06, + "loss": 1.3415, + "step": 27249 + }, + { + "epoch": 5.95, + "learning_rate": 2.028621367708106e-06, + "loss": 2.6398, + "step": 27250 + }, + { + "epoch": 5.95, + "learning_rate": 2.0285121258466247e-06, + "loss": 1.2289, + "step": 27251 + }, + { + "epoch": 5.95, + "learning_rate": 2.0284028839851433e-06, + "loss": 0.6954, + "step": 27252 + }, + { + "epoch": 5.95, + "learning_rate": 2.028293642123662e-06, + "loss": 0.9158, + "step": 27253 + }, + { + "epoch": 5.95, + "learning_rate": 2.0281844002621805e-06, + "loss": 0.7949, + "step": 27254 + }, + { + "epoch": 5.95, + "learning_rate": 2.0280751584006995e-06, + "loss": 1.5598, + "step": 27255 + }, + { + "epoch": 5.95, + "learning_rate": 2.027965916539218e-06, + "loss": 1.6132, + "step": 27256 + }, + { + "epoch": 5.95, + "learning_rate": 2.0278566746777367e-06, + "loss": 3.5069, + "step": 27257 + }, + { + "epoch": 5.95, + "learning_rate": 2.0277474328162554e-06, + "loss": 0.8929, + "step": 27258 + }, + { + "epoch": 5.95, + "learning_rate": 2.027638190954774e-06, + "loss": 0.1546, + "step": 27259 + }, + { + "epoch": 5.95, + "learning_rate": 2.027528949093293e-06, + "loss": 1.3183, + "step": 27260 + }, + { + "epoch": 5.95, + "learning_rate": 2.027419707231811e-06, + "loss": 0.888, + "step": 27261 + }, + { + "epoch": 5.95, + "learning_rate": 2.0273104653703302e-06, + "loss": 0.8389, + "step": 27262 + }, + { + "epoch": 5.95, + "learning_rate": 2.027201223508849e-06, + "loss": 1.6647, + "step": 27263 + }, + { + "epoch": 5.95, + "learning_rate": 2.0270919816473674e-06, + "loss": 1.5114, + "step": 27264 + }, + { + "epoch": 5.95, + "learning_rate": 2.026982739785886e-06, + "loss": 1.011, + "step": 27265 + }, + { + "epoch": 5.95, + "learning_rate": 2.0268734979244047e-06, + "loss": 0.9039, + "step": 27266 + }, + { + "epoch": 5.95, + "learning_rate": 2.0267642560629237e-06, + "loss": 0.2198, + "step": 27267 + }, + { + "epoch": 5.95, + "learning_rate": 2.0266550142014423e-06, + "loss": 1.7206, + "step": 27268 + }, + { + "epoch": 5.95, + "learning_rate": 2.026545772339961e-06, + "loss": 0.5087, + "step": 27269 + }, + { + "epoch": 5.95, + "learning_rate": 2.0264365304784795e-06, + "loss": 1.1105, + "step": 27270 + }, + { + "epoch": 5.95, + "learning_rate": 2.026327288616998e-06, + "loss": 3.1661, + "step": 27271 + }, + { + "epoch": 5.95, + "learning_rate": 2.0262180467555168e-06, + "loss": 2.39, + "step": 27272 + }, + { + "epoch": 5.95, + "learning_rate": 2.0261088048940354e-06, + "loss": 1.8415, + "step": 27273 + }, + { + "epoch": 5.95, + "learning_rate": 2.0259995630325544e-06, + "loss": 0.8518, + "step": 27274 + }, + { + "epoch": 5.95, + "learning_rate": 2.025890321171073e-06, + "loss": 3.0158, + "step": 27275 + }, + { + "epoch": 5.95, + "learning_rate": 2.0257810793095916e-06, + "loss": 2.1652, + "step": 27276 + }, + { + "epoch": 5.95, + "learning_rate": 2.0256718374481102e-06, + "loss": 2.8664, + "step": 27277 + }, + { + "epoch": 5.95, + "learning_rate": 2.025562595586629e-06, + "loss": 0.5854, + "step": 27278 + }, + { + "epoch": 5.95, + "learning_rate": 2.0254533537251475e-06, + "loss": 2.133, + "step": 27279 + }, + { + "epoch": 5.95, + "learning_rate": 2.0253441118636665e-06, + "loss": 2.4146, + "step": 27280 + }, + { + "epoch": 5.95, + "learning_rate": 2.025234870002185e-06, + "loss": 1.3645, + "step": 27281 + }, + { + "epoch": 5.95, + "learning_rate": 2.0251256281407037e-06, + "loss": 0.2336, + "step": 27282 + }, + { + "epoch": 5.95, + "learning_rate": 2.0250163862792223e-06, + "loss": 0.0848, + "step": 27283 + }, + { + "epoch": 5.95, + "learning_rate": 2.024907144417741e-06, + "loss": 2.4629, + "step": 27284 + }, + { + "epoch": 5.95, + "learning_rate": 2.0247979025562595e-06, + "loss": 0.9204, + "step": 27285 + }, + { + "epoch": 5.96, + "learning_rate": 2.0246886606947786e-06, + "loss": 1.9083, + "step": 27286 + }, + { + "epoch": 5.96, + "learning_rate": 2.024579418833297e-06, + "loss": 1.4351, + "step": 27287 + }, + { + "epoch": 5.96, + "learning_rate": 2.024470176971816e-06, + "loss": 1.454, + "step": 27288 + }, + { + "epoch": 5.96, + "learning_rate": 2.0243609351103344e-06, + "loss": 0.4676, + "step": 27289 + }, + { + "epoch": 5.96, + "learning_rate": 2.024251693248853e-06, + "loss": 0.893, + "step": 27290 + }, + { + "epoch": 5.96, + "learning_rate": 2.0241424513873716e-06, + "loss": 1.941, + "step": 27291 + }, + { + "epoch": 5.96, + "learning_rate": 2.0240332095258907e-06, + "loss": 2.3708, + "step": 27292 + }, + { + "epoch": 5.96, + "learning_rate": 2.0239239676644093e-06, + "loss": 2.3484, + "step": 27293 + }, + { + "epoch": 5.96, + "learning_rate": 2.023814725802928e-06, + "loss": 1.5586, + "step": 27294 + }, + { + "epoch": 5.96, + "learning_rate": 2.0237054839414465e-06, + "loss": 0.9756, + "step": 27295 + }, + { + "epoch": 5.96, + "learning_rate": 2.023596242079965e-06, + "loss": 1.5741, + "step": 27296 + }, + { + "epoch": 5.96, + "learning_rate": 2.0234870002184837e-06, + "loss": 1.47, + "step": 27297 + }, + { + "epoch": 5.96, + "learning_rate": 2.0233777583570023e-06, + "loss": 2.3712, + "step": 27298 + }, + { + "epoch": 5.96, + "learning_rate": 2.0232685164955214e-06, + "loss": 0.9668, + "step": 27299 + }, + { + "epoch": 5.96, + "learning_rate": 2.02315927463404e-06, + "loss": 2.1332, + "step": 27300 + }, + { + "epoch": 5.96, + "learning_rate": 2.0230500327725586e-06, + "loss": 1.67, + "step": 27301 + }, + { + "epoch": 5.96, + "learning_rate": 2.022940790911077e-06, + "loss": 0.9396, + "step": 27302 + }, + { + "epoch": 5.96, + "learning_rate": 2.022831549049596e-06, + "loss": 1.6782, + "step": 27303 + }, + { + "epoch": 5.96, + "learning_rate": 2.022722307188115e-06, + "loss": 1.2735, + "step": 27304 + }, + { + "epoch": 5.96, + "learning_rate": 2.022613065326633e-06, + "loss": 1.5545, + "step": 27305 + }, + { + "epoch": 5.96, + "learning_rate": 2.022503823465152e-06, + "loss": 0.3368, + "step": 27306 + }, + { + "epoch": 5.96, + "learning_rate": 2.0223945816036707e-06, + "loss": 2.6639, + "step": 27307 + }, + { + "epoch": 5.96, + "learning_rate": 2.0222853397421893e-06, + "loss": 1.6774, + "step": 27308 + }, + { + "epoch": 5.96, + "learning_rate": 2.0221760978807083e-06, + "loss": 1.6769, + "step": 27309 + }, + { + "epoch": 5.96, + "learning_rate": 2.0220668560192265e-06, + "loss": 2.0354, + "step": 27310 + }, + { + "epoch": 5.96, + "learning_rate": 2.0219576141577455e-06, + "loss": 1.0098, + "step": 27311 + }, + { + "epoch": 5.96, + "learning_rate": 2.021848372296264e-06, + "loss": 1.2875, + "step": 27312 + }, + { + "epoch": 5.96, + "learning_rate": 2.0217391304347828e-06, + "loss": 0.5526, + "step": 27313 + }, + { + "epoch": 5.96, + "learning_rate": 2.0216298885733014e-06, + "loss": 1.1841, + "step": 27314 + }, + { + "epoch": 5.96, + "learning_rate": 2.02152064671182e-06, + "loss": 1.4428, + "step": 27315 + }, + { + "epoch": 5.96, + "learning_rate": 2.021411404850339e-06, + "loss": 1.192, + "step": 27316 + }, + { + "epoch": 5.96, + "learning_rate": 2.0213021629888572e-06, + "loss": 1.895, + "step": 27317 + }, + { + "epoch": 5.96, + "learning_rate": 2.0211929211273762e-06, + "loss": 1.9429, + "step": 27318 + }, + { + "epoch": 5.96, + "learning_rate": 2.021083679265895e-06, + "loss": 2.3128, + "step": 27319 + }, + { + "epoch": 5.96, + "learning_rate": 2.0209744374044135e-06, + "loss": 1.6724, + "step": 27320 + }, + { + "epoch": 5.96, + "learning_rate": 2.0208651955429325e-06, + "loss": 1.2019, + "step": 27321 + }, + { + "epoch": 5.96, + "learning_rate": 2.0207559536814507e-06, + "loss": 1.0286, + "step": 27322 + }, + { + "epoch": 5.96, + "learning_rate": 2.0206467118199697e-06, + "loss": 0.7807, + "step": 27323 + }, + { + "epoch": 5.96, + "learning_rate": 2.0205374699584883e-06, + "loss": 1.4623, + "step": 27324 + }, + { + "epoch": 5.96, + "learning_rate": 2.020428228097007e-06, + "loss": 1.5363, + "step": 27325 + }, + { + "epoch": 5.96, + "learning_rate": 2.0203189862355256e-06, + "loss": 1.8053, + "step": 27326 + }, + { + "epoch": 5.96, + "learning_rate": 2.020209744374044e-06, + "loss": 2.3533, + "step": 27327 + }, + { + "epoch": 5.96, + "learning_rate": 2.020100502512563e-06, + "loss": 0.9982, + "step": 27328 + }, + { + "epoch": 5.96, + "learning_rate": 2.0199912606510814e-06, + "loss": 0.7932, + "step": 27329 + }, + { + "epoch": 5.96, + "learning_rate": 2.0198820187896004e-06, + "loss": 1.9755, + "step": 27330 + }, + { + "epoch": 5.96, + "learning_rate": 2.019772776928119e-06, + "loss": 0.1799, + "step": 27331 + }, + { + "epoch": 5.97, + "learning_rate": 2.0196635350666376e-06, + "loss": 3.0416, + "step": 27332 + }, + { + "epoch": 5.97, + "learning_rate": 2.0195542932051567e-06, + "loss": 2.9732, + "step": 27333 + }, + { + "epoch": 5.97, + "learning_rate": 2.019445051343675e-06, + "loss": 0.7905, + "step": 27334 + }, + { + "epoch": 5.97, + "learning_rate": 2.019335809482194e-06, + "loss": 1.0778, + "step": 27335 + }, + { + "epoch": 5.97, + "learning_rate": 2.0192265676207125e-06, + "loss": 1.6148, + "step": 27336 + }, + { + "epoch": 5.97, + "learning_rate": 2.019117325759231e-06, + "loss": 1.9641, + "step": 27337 + }, + { + "epoch": 5.97, + "learning_rate": 2.0190080838977497e-06, + "loss": 2.0916, + "step": 27338 + }, + { + "epoch": 5.97, + "learning_rate": 2.0188988420362684e-06, + "loss": 1.0919, + "step": 27339 + }, + { + "epoch": 5.97, + "learning_rate": 2.0187896001747874e-06, + "loss": 0.818, + "step": 27340 + }, + { + "epoch": 5.97, + "learning_rate": 2.0186803583133056e-06, + "loss": 1.131, + "step": 27341 + }, + { + "epoch": 5.97, + "learning_rate": 2.0185711164518246e-06, + "loss": 1.6812, + "step": 27342 + }, + { + "epoch": 5.97, + "learning_rate": 2.0184618745903432e-06, + "loss": 1.9088, + "step": 27343 + }, + { + "epoch": 5.97, + "learning_rate": 2.018352632728862e-06, + "loss": 0.6167, + "step": 27344 + }, + { + "epoch": 5.97, + "learning_rate": 2.0182433908673804e-06, + "loss": 2.2183, + "step": 27345 + }, + { + "epoch": 5.97, + "learning_rate": 2.018134149005899e-06, + "loss": 1.284, + "step": 27346 + }, + { + "epoch": 5.97, + "learning_rate": 2.018024907144418e-06, + "loss": 1.3217, + "step": 27347 + }, + { + "epoch": 5.97, + "learning_rate": 2.0179156652829367e-06, + "loss": 2.7251, + "step": 27348 + }, + { + "epoch": 5.97, + "learning_rate": 2.0178064234214553e-06, + "loss": 1.1121, + "step": 27349 + }, + { + "epoch": 5.97, + "learning_rate": 2.017697181559974e-06, + "loss": 1.9062, + "step": 27350 + }, + { + "epoch": 5.97, + "learning_rate": 2.0175879396984925e-06, + "loss": 1.1382, + "step": 27351 + }, + { + "epoch": 5.97, + "learning_rate": 2.0174786978370116e-06, + "loss": 1.5536, + "step": 27352 + }, + { + "epoch": 5.97, + "learning_rate": 2.01736945597553e-06, + "loss": 2.4125, + "step": 27353 + }, + { + "epoch": 5.97, + "learning_rate": 2.0172602141140488e-06, + "loss": 1.1556, + "step": 27354 + }, + { + "epoch": 5.97, + "learning_rate": 2.0171509722525674e-06, + "loss": 0.6623, + "step": 27355 + }, + { + "epoch": 5.97, + "learning_rate": 2.017041730391086e-06, + "loss": 1.8927, + "step": 27356 + }, + { + "epoch": 5.97, + "learning_rate": 2.0169324885296046e-06, + "loss": 1.0053, + "step": 27357 + }, + { + "epoch": 5.97, + "learning_rate": 2.0168232466681232e-06, + "loss": 1.9928, + "step": 27358 + }, + { + "epoch": 5.97, + "learning_rate": 2.0167140048066423e-06, + "loss": 0.7391, + "step": 27359 + }, + { + "epoch": 5.97, + "learning_rate": 2.016604762945161e-06, + "loss": 0.3096, + "step": 27360 + }, + { + "epoch": 5.97, + "learning_rate": 2.0164955210836795e-06, + "loss": 1.342, + "step": 27361 + }, + { + "epoch": 5.97, + "learning_rate": 2.016386279222198e-06, + "loss": 3.4344, + "step": 27362 + }, + { + "epoch": 5.97, + "learning_rate": 2.0162770373607167e-06, + "loss": 1.2257, + "step": 27363 + }, + { + "epoch": 5.97, + "learning_rate": 2.0161677954992353e-06, + "loss": 2.4518, + "step": 27364 + }, + { + "epoch": 5.97, + "learning_rate": 2.0160585536377544e-06, + "loss": 0.9868, + "step": 27365 + }, + { + "epoch": 5.97, + "learning_rate": 2.015949311776273e-06, + "loss": 1.3706, + "step": 27366 + }, + { + "epoch": 5.97, + "learning_rate": 2.0158400699147916e-06, + "loss": 2.8902, + "step": 27367 + }, + { + "epoch": 5.97, + "learning_rate": 2.01573082805331e-06, + "loss": 0.626, + "step": 27368 + }, + { + "epoch": 5.97, + "learning_rate": 2.015621586191829e-06, + "loss": 1.835, + "step": 27369 + }, + { + "epoch": 5.97, + "learning_rate": 2.0155123443303474e-06, + "loss": 2.0297, + "step": 27370 + }, + { + "epoch": 5.97, + "learning_rate": 2.015403102468866e-06, + "loss": 0.5369, + "step": 27371 + }, + { + "epoch": 5.97, + "learning_rate": 2.015293860607385e-06, + "loss": 0.3992, + "step": 27372 + }, + { + "epoch": 5.97, + "learning_rate": 2.0151846187459037e-06, + "loss": 2.083, + "step": 27373 + }, + { + "epoch": 5.97, + "learning_rate": 2.0150753768844223e-06, + "loss": 1.6388, + "step": 27374 + }, + { + "epoch": 5.97, + "learning_rate": 2.014966135022941e-06, + "loss": 1.3049, + "step": 27375 + }, + { + "epoch": 5.97, + "learning_rate": 2.0148568931614595e-06, + "loss": 1.1819, + "step": 27376 + }, + { + "epoch": 5.97, + "learning_rate": 2.0147476512999785e-06, + "loss": 0.4737, + "step": 27377 + }, + { + "epoch": 5.98, + "learning_rate": 2.0146384094384967e-06, + "loss": 2.1287, + "step": 27378 + }, + { + "epoch": 5.98, + "learning_rate": 2.0145291675770158e-06, + "loss": 2.1833, + "step": 27379 + }, + { + "epoch": 5.98, + "learning_rate": 2.0144199257155344e-06, + "loss": 1.7481, + "step": 27380 + }, + { + "epoch": 5.98, + "learning_rate": 2.014310683854053e-06, + "loss": 2.3835, + "step": 27381 + }, + { + "epoch": 5.98, + "learning_rate": 2.0142014419925716e-06, + "loss": 0.3584, + "step": 27382 + }, + { + "epoch": 5.98, + "learning_rate": 2.01409220013109e-06, + "loss": 1.9311, + "step": 27383 + }, + { + "epoch": 5.98, + "learning_rate": 2.0139829582696092e-06, + "loss": 2.304, + "step": 27384 + }, + { + "epoch": 5.98, + "learning_rate": 2.0138737164081274e-06, + "loss": 1.2831, + "step": 27385 + }, + { + "epoch": 5.98, + "learning_rate": 2.0137644745466465e-06, + "loss": 3.1837, + "step": 27386 + }, + { + "epoch": 5.98, + "learning_rate": 2.013655232685165e-06, + "loss": 1.9647, + "step": 27387 + }, + { + "epoch": 5.98, + "learning_rate": 2.0135459908236837e-06, + "loss": 0.8289, + "step": 27388 + }, + { + "epoch": 5.98, + "learning_rate": 2.0134367489622027e-06, + "loss": 0.7895, + "step": 27389 + }, + { + "epoch": 5.98, + "learning_rate": 2.013327507100721e-06, + "loss": 1.2861, + "step": 27390 + }, + { + "epoch": 5.98, + "learning_rate": 2.01321826523924e-06, + "loss": 2.4994, + "step": 27391 + }, + { + "epoch": 5.98, + "learning_rate": 2.0131090233777585e-06, + "loss": 1.135, + "step": 27392 + }, + { + "epoch": 5.98, + "learning_rate": 2.012999781516277e-06, + "loss": 1.8571, + "step": 27393 + }, + { + "epoch": 5.98, + "learning_rate": 2.0128905396547958e-06, + "loss": 1.8692, + "step": 27394 + }, + { + "epoch": 5.98, + "learning_rate": 2.0127812977933144e-06, + "loss": 1.8584, + "step": 27395 + }, + { + "epoch": 5.98, + "learning_rate": 2.0126720559318334e-06, + "loss": 1.0431, + "step": 27396 + }, + { + "epoch": 5.98, + "learning_rate": 2.012562814070352e-06, + "loss": 0.9837, + "step": 27397 + }, + { + "epoch": 5.98, + "learning_rate": 2.0124535722088706e-06, + "loss": 1.7164, + "step": 27398 + }, + { + "epoch": 5.98, + "learning_rate": 2.0123443303473892e-06, + "loss": 1.2416, + "step": 27399 + }, + { + "epoch": 5.98, + "learning_rate": 2.012235088485908e-06, + "loss": 0.8163, + "step": 27400 + }, + { + "epoch": 5.98, + "learning_rate": 2.012125846624427e-06, + "loss": 1.4398, + "step": 27401 + }, + { + "epoch": 5.98, + "learning_rate": 2.012016604762945e-06, + "loss": 1.4551, + "step": 27402 + }, + { + "epoch": 5.98, + "learning_rate": 2.011907362901464e-06, + "loss": 1.5614, + "step": 27403 + }, + { + "epoch": 5.98, + "learning_rate": 2.0117981210399827e-06, + "loss": 2.7175, + "step": 27404 + }, + { + "epoch": 5.98, + "learning_rate": 2.0116888791785013e-06, + "loss": 1.6579, + "step": 27405 + }, + { + "epoch": 5.98, + "learning_rate": 2.0115796373170204e-06, + "loss": 1.809, + "step": 27406 + }, + { + "epoch": 5.98, + "learning_rate": 2.0114703954555386e-06, + "loss": 0.799, + "step": 27407 + }, + { + "epoch": 5.98, + "learning_rate": 2.0113611535940576e-06, + "loss": 0.6327, + "step": 27408 + }, + { + "epoch": 5.98, + "learning_rate": 2.011251911732576e-06, + "loss": 2.0375, + "step": 27409 + }, + { + "epoch": 5.98, + "learning_rate": 2.011142669871095e-06, + "loss": 1.3515, + "step": 27410 + }, + { + "epoch": 5.98, + "learning_rate": 2.0110334280096134e-06, + "loss": 3.2625, + "step": 27411 + }, + { + "epoch": 5.98, + "learning_rate": 2.010924186148132e-06, + "loss": 0.9459, + "step": 27412 + }, + { + "epoch": 5.98, + "learning_rate": 2.010814944286651e-06, + "loss": 2.2668, + "step": 27413 + }, + { + "epoch": 5.98, + "learning_rate": 2.0107057024251693e-06, + "loss": 1.3102, + "step": 27414 + }, + { + "epoch": 5.98, + "learning_rate": 2.0105964605636883e-06, + "loss": 3.4135, + "step": 27415 + }, + { + "epoch": 5.98, + "learning_rate": 2.010487218702207e-06, + "loss": 1.4832, + "step": 27416 + }, + { + "epoch": 5.98, + "learning_rate": 2.0103779768407255e-06, + "loss": 1.1653, + "step": 27417 + }, + { + "epoch": 5.98, + "learning_rate": 2.010268734979244e-06, + "loss": 0.4053, + "step": 27418 + }, + { + "epoch": 5.98, + "learning_rate": 2.0101594931177627e-06, + "loss": 1.2456, + "step": 27419 + }, + { + "epoch": 5.98, + "learning_rate": 2.0100502512562818e-06, + "loss": 1.1389, + "step": 27420 + }, + { + "epoch": 5.98, + "learning_rate": 2.0099410093948004e-06, + "loss": 1.826, + "step": 27421 + }, + { + "epoch": 5.98, + "learning_rate": 2.009831767533319e-06, + "loss": 2.0014, + "step": 27422 + }, + { + "epoch": 5.98, + "learning_rate": 2.0097225256718376e-06, + "loss": 1.5005, + "step": 27423 + }, + { + "epoch": 5.99, + "learning_rate": 2.0096132838103562e-06, + "loss": 1.5826, + "step": 27424 + }, + { + "epoch": 5.99, + "learning_rate": 2.0095040419488752e-06, + "loss": 0.6912, + "step": 27425 + }, + { + "epoch": 5.99, + "learning_rate": 2.0093948000873934e-06, + "loss": 1.0403, + "step": 27426 + }, + { + "epoch": 5.99, + "learning_rate": 2.0092855582259125e-06, + "loss": 0.9654, + "step": 27427 + }, + { + "epoch": 5.99, + "learning_rate": 2.009176316364431e-06, + "loss": 1.922, + "step": 27428 + }, + { + "epoch": 5.99, + "learning_rate": 2.0090670745029497e-06, + "loss": 1.0259, + "step": 27429 + }, + { + "epoch": 5.99, + "learning_rate": 2.0089578326414683e-06, + "loss": 1.185, + "step": 27430 + }, + { + "epoch": 5.99, + "learning_rate": 2.008848590779987e-06, + "loss": 1.1523, + "step": 27431 + }, + { + "epoch": 5.99, + "learning_rate": 2.008739348918506e-06, + "loss": 1.3655, + "step": 27432 + }, + { + "epoch": 5.99, + "learning_rate": 2.0086301070570246e-06, + "loss": 2.3605, + "step": 27433 + }, + { + "epoch": 5.99, + "learning_rate": 2.008520865195543e-06, + "loss": 0.3172, + "step": 27434 + }, + { + "epoch": 5.99, + "learning_rate": 2.0084116233340618e-06, + "loss": 0.6912, + "step": 27435 + }, + { + "epoch": 5.99, + "learning_rate": 2.0083023814725804e-06, + "loss": 0.6166, + "step": 27436 + }, + { + "epoch": 5.99, + "learning_rate": 2.008193139611099e-06, + "loss": 1.3486, + "step": 27437 + }, + { + "epoch": 5.99, + "learning_rate": 2.0080838977496176e-06, + "loss": 0.7756, + "step": 27438 + }, + { + "epoch": 5.99, + "learning_rate": 2.0079746558881366e-06, + "loss": 1.1215, + "step": 27439 + }, + { + "epoch": 5.99, + "learning_rate": 2.0078654140266553e-06, + "loss": 2.1133, + "step": 27440 + }, + { + "epoch": 5.99, + "learning_rate": 2.007756172165174e-06, + "loss": 1.1948, + "step": 27441 + }, + { + "epoch": 5.99, + "learning_rate": 2.0076469303036925e-06, + "loss": 1.9586, + "step": 27442 + }, + { + "epoch": 5.99, + "learning_rate": 2.007537688442211e-06, + "loss": 0.7026, + "step": 27443 + }, + { + "epoch": 5.99, + "learning_rate": 2.0074284465807297e-06, + "loss": 0.5403, + "step": 27444 + }, + { + "epoch": 5.99, + "learning_rate": 2.0073192047192487e-06, + "loss": 0.5494, + "step": 27445 + }, + { + "epoch": 5.99, + "learning_rate": 2.0072099628577673e-06, + "loss": 0.4859, + "step": 27446 + }, + { + "epoch": 5.99, + "learning_rate": 2.007100720996286e-06, + "loss": 0.2117, + "step": 27447 + }, + { + "epoch": 5.99, + "learning_rate": 2.0069914791348046e-06, + "loss": 1.8622, + "step": 27448 + }, + { + "epoch": 5.99, + "learning_rate": 2.006882237273323e-06, + "loss": 0.9987, + "step": 27449 + }, + { + "epoch": 5.99, + "learning_rate": 2.0067729954118422e-06, + "loss": 1.5453, + "step": 27450 + }, + { + "epoch": 5.99, + "learning_rate": 2.0066637535503604e-06, + "loss": 1.0009, + "step": 27451 + }, + { + "epoch": 5.99, + "learning_rate": 2.0065545116888794e-06, + "loss": 1.8451, + "step": 27452 + }, + { + "epoch": 5.99, + "learning_rate": 2.006445269827398e-06, + "loss": 1.7474, + "step": 27453 + }, + { + "epoch": 5.99, + "learning_rate": 2.0063360279659167e-06, + "loss": 1.8399, + "step": 27454 + }, + { + "epoch": 5.99, + "learning_rate": 2.0062267861044353e-06, + "loss": 0.722, + "step": 27455 + }, + { + "epoch": 5.99, + "learning_rate": 2.006117544242954e-06, + "loss": 1.8634, + "step": 27456 + }, + { + "epoch": 5.99, + "learning_rate": 2.006008302381473e-06, + "loss": 0.3631, + "step": 27457 + }, + { + "epoch": 5.99, + "learning_rate": 2.0058990605199915e-06, + "loss": 0.9192, + "step": 27458 + }, + { + "epoch": 5.99, + "learning_rate": 2.00578981865851e-06, + "loss": 1.4758, + "step": 27459 + }, + { + "epoch": 5.99, + "learning_rate": 2.0056805767970287e-06, + "loss": 1.6127, + "step": 27460 + }, + { + "epoch": 5.99, + "learning_rate": 2.0055713349355474e-06, + "loss": 1.5543, + "step": 27461 + }, + { + "epoch": 5.99, + "learning_rate": 2.0054620930740664e-06, + "loss": 1.4333, + "step": 27462 + }, + { + "epoch": 5.99, + "learning_rate": 2.0053528512125846e-06, + "loss": 0.8393, + "step": 27463 + }, + { + "epoch": 5.99, + "learning_rate": 2.0052436093511036e-06, + "loss": 0.6177, + "step": 27464 + }, + { + "epoch": 5.99, + "learning_rate": 2.0051343674896222e-06, + "loss": 0.1567, + "step": 27465 + }, + { + "epoch": 5.99, + "learning_rate": 2.005025125628141e-06, + "loss": 0.6181, + "step": 27466 + }, + { + "epoch": 5.99, + "learning_rate": 2.0049158837666595e-06, + "loss": 1.2196, + "step": 27467 + }, + { + "epoch": 5.99, + "learning_rate": 2.004806641905178e-06, + "loss": 1.2683, + "step": 27468 + }, + { + "epoch": 5.99, + "learning_rate": 2.004697400043697e-06, + "loss": 0.9725, + "step": 27469 + }, + { + "epoch": 6.0, + "learning_rate": 2.0045881581822153e-06, + "loss": 2.5632, + "step": 27470 + }, + { + "epoch": 6.0, + "learning_rate": 2.0044789163207343e-06, + "loss": 1.2486, + "step": 27471 + }, + { + "epoch": 6.0, + "learning_rate": 2.004369674459253e-06, + "loss": 1.3682, + "step": 27472 + }, + { + "epoch": 6.0, + "learning_rate": 2.0042604325977715e-06, + "loss": 1.8834, + "step": 27473 + }, + { + "epoch": 6.0, + "learning_rate": 2.0041511907362906e-06, + "loss": 1.7478, + "step": 27474 + }, + { + "epoch": 6.0, + "learning_rate": 2.0040419488748088e-06, + "loss": 1.8163, + "step": 27475 + }, + { + "epoch": 6.0, + "learning_rate": 2.003932707013328e-06, + "loss": 4.2143, + "step": 27476 + }, + { + "epoch": 6.0, + "learning_rate": 2.0038234651518464e-06, + "loss": 1.3179, + "step": 27477 + }, + { + "epoch": 6.0, + "learning_rate": 2.003714223290365e-06, + "loss": 1.3086, + "step": 27478 + }, + { + "epoch": 6.0, + "learning_rate": 2.0036049814288836e-06, + "loss": 0.8983, + "step": 27479 + }, + { + "epoch": 6.0, + "learning_rate": 2.0034957395674022e-06, + "loss": 2.1526, + "step": 27480 + }, + { + "epoch": 6.0, + "learning_rate": 2.0033864977059213e-06, + "loss": 2.0628, + "step": 27481 + }, + { + "epoch": 6.0, + "learning_rate": 2.0032772558444395e-06, + "loss": 1.371, + "step": 27482 + }, + { + "epoch": 6.0, + "learning_rate": 2.0031680139829585e-06, + "loss": 0.4847, + "step": 27483 + }, + { + "epoch": 6.0, + "learning_rate": 2.003058772121477e-06, + "loss": 2.2806, + "step": 27484 + }, + { + "epoch": 6.0, + "learning_rate": 2.0029495302599957e-06, + "loss": 2.3487, + "step": 27485 + }, + { + "epoch": 6.0, + "learning_rate": 2.0028402883985148e-06, + "loss": 1.6104, + "step": 27486 + }, + { + "epoch": 6.0, + "learning_rate": 2.002731046537033e-06, + "loss": 0.6764, + "step": 27487 + }, + { + "epoch": 6.0, + "learning_rate": 2.002621804675552e-06, + "loss": 3.0269, + "step": 27488 + }, + { + "epoch": 6.0, + "learning_rate": 2.0025125628140706e-06, + "loss": 1.0724, + "step": 27489 + }, + { + "epoch": 6.0, + "learning_rate": 2.002403320952589e-06, + "loss": 0.5679, + "step": 27490 + }, + { + "epoch": 6.0, + "learning_rate": 2.002294079091108e-06, + "loss": 1.1289, + "step": 27491 + }, + { + "epoch": 6.0, + "learning_rate": 2.0021848372296264e-06, + "loss": 0.8704, + "step": 27492 + }, + { + "epoch": 6.0, + "eval_accuracy": 0.6639934533551555, + "eval_loss": 1.386324167251587, + "eval_runtime": 331.1247, + "eval_samples_per_second": 18.452, + "eval_steps_per_second": 4.615, + "step": 27492 + }, + { + "epoch": 6.0, + "learning_rate": 2.0020755953681455e-06, + "loss": 1.1124, + "step": 27493 + }, + { + "epoch": 6.0, + "learning_rate": 2.0019663535066636e-06, + "loss": 1.495, + "step": 27494 + }, + { + "epoch": 6.0, + "learning_rate": 2.0018571116451827e-06, + "loss": 1.3964, + "step": 27495 + }, + { + "epoch": 6.0, + "learning_rate": 2.0017478697837013e-06, + "loss": 0.9483, + "step": 27496 + }, + { + "epoch": 6.0, + "learning_rate": 2.00163862792222e-06, + "loss": 2.3877, + "step": 27497 + }, + { + "epoch": 6.0, + "learning_rate": 2.001529386060739e-06, + "loss": 0.7902, + "step": 27498 + }, + { + "epoch": 6.0, + "learning_rate": 2.001420144199257e-06, + "loss": 0.271, + "step": 27499 + }, + { + "epoch": 6.0, + "learning_rate": 2.001310902337776e-06, + "loss": 2.0558, + "step": 27500 + }, + { + "epoch": 6.0, + "learning_rate": 2.0012016604762948e-06, + "loss": 1.0575, + "step": 27501 + }, + { + "epoch": 6.0, + "learning_rate": 2.0010924186148134e-06, + "loss": 2.1248, + "step": 27502 + }, + { + "epoch": 6.0, + "learning_rate": 2.000983176753332e-06, + "loss": 0.2608, + "step": 27503 + }, + { + "epoch": 6.0, + "learning_rate": 2.0008739348918506e-06, + "loss": 0.348, + "step": 27504 + }, + { + "epoch": 6.0, + "learning_rate": 2.0007646930303696e-06, + "loss": 0.4637, + "step": 27505 + }, + { + "epoch": 6.0, + "learning_rate": 2.0006554511688882e-06, + "loss": 1.1053, + "step": 27506 + }, + { + "epoch": 6.0, + "learning_rate": 2.000546209307407e-06, + "loss": 1.1635, + "step": 27507 + }, + { + "epoch": 6.0, + "learning_rate": 2.0004369674459255e-06, + "loss": 1.3504, + "step": 27508 + }, + { + "epoch": 6.0, + "learning_rate": 2.000327725584444e-06, + "loss": 0.6325, + "step": 27509 + }, + { + "epoch": 6.0, + "learning_rate": 2.0002184837229627e-06, + "loss": 0.5804, + "step": 27510 + }, + { + "epoch": 6.0, + "learning_rate": 2.0001092418614813e-06, + "loss": 0.4255, + "step": 27511 + }, + { + "epoch": 6.0, + "learning_rate": 2.0000000000000003e-06, + "loss": 1.6221, + "step": 27512 + }, + { + "epoch": 6.0, + "learning_rate": 1.999890758138519e-06, + "loss": 1.7256, + "step": 27513 + }, + { + "epoch": 6.0, + "learning_rate": 1.9997815162770376e-06, + "loss": 1.6094, + "step": 27514 + }, + { + "epoch": 6.01, + "learning_rate": 1.999672274415556e-06, + "loss": 0.4871, + "step": 27515 + }, + { + "epoch": 6.01, + "learning_rate": 1.9995630325540748e-06, + "loss": 0.2273, + "step": 27516 + }, + { + "epoch": 6.01, + "learning_rate": 1.9994537906925934e-06, + "loss": 1.1009, + "step": 27517 + }, + { + "epoch": 6.01, + "learning_rate": 1.9993445488311124e-06, + "loss": 1.1669, + "step": 27518 + }, + { + "epoch": 6.01, + "learning_rate": 1.999235306969631e-06, + "loss": 1.1375, + "step": 27519 + }, + { + "epoch": 6.01, + "learning_rate": 1.9991260651081496e-06, + "loss": 1.8791, + "step": 27520 + }, + { + "epoch": 6.01, + "learning_rate": 1.9990168232466683e-06, + "loss": 1.1506, + "step": 27521 + }, + { + "epoch": 6.01, + "learning_rate": 1.998907581385187e-06, + "loss": 0.4447, + "step": 27522 + }, + { + "epoch": 6.01, + "learning_rate": 1.9987983395237055e-06, + "loss": 1.9873, + "step": 27523 + }, + { + "epoch": 6.01, + "learning_rate": 1.9986890976622245e-06, + "loss": 1.6827, + "step": 27524 + }, + { + "epoch": 6.01, + "learning_rate": 1.998579855800743e-06, + "loss": 0.8757, + "step": 27525 + }, + { + "epoch": 6.01, + "learning_rate": 1.9984706139392617e-06, + "loss": 1.1496, + "step": 27526 + }, + { + "epoch": 6.01, + "learning_rate": 1.9983613720777803e-06, + "loss": 0.8204, + "step": 27527 + }, + { + "epoch": 6.01, + "learning_rate": 1.998252130216299e-06, + "loss": 1.1121, + "step": 27528 + }, + { + "epoch": 6.01, + "learning_rate": 1.9981428883548176e-06, + "loss": 0.9484, + "step": 27529 + }, + { + "epoch": 6.01, + "learning_rate": 1.9980336464933366e-06, + "loss": 1.2191, + "step": 27530 + }, + { + "epoch": 6.01, + "learning_rate": 1.9979244046318552e-06, + "loss": 1.4679, + "step": 27531 + }, + { + "epoch": 6.01, + "learning_rate": 1.997815162770374e-06, + "loss": 1.7731, + "step": 27532 + }, + { + "epoch": 6.01, + "learning_rate": 1.9977059209088924e-06, + "loss": 1.8815, + "step": 27533 + }, + { + "epoch": 6.01, + "learning_rate": 1.997596679047411e-06, + "loss": 1.0981, + "step": 27534 + }, + { + "epoch": 6.01, + "learning_rate": 1.9974874371859297e-06, + "loss": 0.9969, + "step": 27535 + }, + { + "epoch": 6.01, + "learning_rate": 1.9973781953244483e-06, + "loss": 1.5916, + "step": 27536 + }, + { + "epoch": 6.01, + "learning_rate": 1.9972689534629673e-06, + "loss": 0.3016, + "step": 27537 + }, + { + "epoch": 6.01, + "learning_rate": 1.997159711601486e-06, + "loss": 1.7808, + "step": 27538 + }, + { + "epoch": 6.01, + "learning_rate": 1.9970504697400045e-06, + "loss": 1.0267, + "step": 27539 + }, + { + "epoch": 6.01, + "learning_rate": 1.996941227878523e-06, + "loss": 3.0827, + "step": 27540 + }, + { + "epoch": 6.01, + "learning_rate": 1.9968319860170417e-06, + "loss": 0.6976, + "step": 27541 + }, + { + "epoch": 6.01, + "learning_rate": 1.9967227441555608e-06, + "loss": 1.2231, + "step": 27542 + }, + { + "epoch": 6.01, + "learning_rate": 1.996613502294079e-06, + "loss": 1.0298, + "step": 27543 + }, + { + "epoch": 6.01, + "learning_rate": 1.996504260432598e-06, + "loss": 0.8786, + "step": 27544 + }, + { + "epoch": 6.01, + "learning_rate": 1.9963950185711166e-06, + "loss": 2.6051, + "step": 27545 + }, + { + "epoch": 6.01, + "learning_rate": 1.9962857767096352e-06, + "loss": 1.6145, + "step": 27546 + }, + { + "epoch": 6.01, + "learning_rate": 1.996176534848154e-06, + "loss": 0.0979, + "step": 27547 + }, + { + "epoch": 6.01, + "learning_rate": 1.9960672929866724e-06, + "loss": 1.9837, + "step": 27548 + }, + { + "epoch": 6.01, + "learning_rate": 1.9959580511251915e-06, + "loss": 1.1482, + "step": 27549 + }, + { + "epoch": 6.01, + "learning_rate": 1.99584880926371e-06, + "loss": 0.1424, + "step": 27550 + }, + { + "epoch": 6.01, + "learning_rate": 1.9957395674022287e-06, + "loss": 1.763, + "step": 27551 + }, + { + "epoch": 6.01, + "learning_rate": 1.9956303255407473e-06, + "loss": 0.9332, + "step": 27552 + }, + { + "epoch": 6.01, + "learning_rate": 1.995521083679266e-06, + "loss": 1.6508, + "step": 27553 + }, + { + "epoch": 6.01, + "learning_rate": 1.995411841817785e-06, + "loss": 0.8124, + "step": 27554 + }, + { + "epoch": 6.01, + "learning_rate": 1.995302599956303e-06, + "loss": 1.5287, + "step": 27555 + }, + { + "epoch": 6.01, + "learning_rate": 1.995193358094822e-06, + "loss": 1.6442, + "step": 27556 + }, + { + "epoch": 6.01, + "learning_rate": 1.995084116233341e-06, + "loss": 0.7916, + "step": 27557 + }, + { + "epoch": 6.01, + "learning_rate": 1.9949748743718594e-06, + "loss": 0.612, + "step": 27558 + }, + { + "epoch": 6.01, + "learning_rate": 1.9948656325103784e-06, + "loss": 2.0112, + "step": 27559 + }, + { + "epoch": 6.01, + "learning_rate": 1.9947563906488966e-06, + "loss": 1.4192, + "step": 27560 + }, + { + "epoch": 6.02, + "learning_rate": 1.9946471487874157e-06, + "loss": 1.1925, + "step": 27561 + }, + { + "epoch": 6.02, + "learning_rate": 1.9945379069259343e-06, + "loss": 1.3557, + "step": 27562 + }, + { + "epoch": 6.02, + "learning_rate": 1.994428665064453e-06, + "loss": 1.1317, + "step": 27563 + }, + { + "epoch": 6.02, + "learning_rate": 1.9943194232029715e-06, + "loss": 0.6556, + "step": 27564 + }, + { + "epoch": 6.02, + "learning_rate": 1.99421018134149e-06, + "loss": 2.538, + "step": 27565 + }, + { + "epoch": 6.02, + "learning_rate": 1.994100939480009e-06, + "loss": 1.8251, + "step": 27566 + }, + { + "epoch": 6.02, + "learning_rate": 1.9939916976185273e-06, + "loss": 1.0156, + "step": 27567 + }, + { + "epoch": 6.02, + "learning_rate": 1.9938824557570464e-06, + "loss": 2.6205, + "step": 27568 + }, + { + "epoch": 6.02, + "learning_rate": 1.993773213895565e-06, + "loss": 0.5995, + "step": 27569 + }, + { + "epoch": 6.02, + "learning_rate": 1.9936639720340836e-06, + "loss": 2.3712, + "step": 27570 + }, + { + "epoch": 6.02, + "learning_rate": 1.9935547301726026e-06, + "loss": 1.6811, + "step": 27571 + }, + { + "epoch": 6.02, + "learning_rate": 1.993445488311121e-06, + "loss": 1.5147, + "step": 27572 + }, + { + "epoch": 6.02, + "learning_rate": 1.99333624644964e-06, + "loss": 1.0479, + "step": 27573 + }, + { + "epoch": 6.02, + "learning_rate": 1.9932270045881584e-06, + "loss": 0.5235, + "step": 27574 + }, + { + "epoch": 6.02, + "learning_rate": 1.993117762726677e-06, + "loss": 2.2984, + "step": 27575 + }, + { + "epoch": 6.02, + "learning_rate": 1.9930085208651957e-06, + "loss": 1.5961, + "step": 27576 + }, + { + "epoch": 6.02, + "learning_rate": 1.9928992790037143e-06, + "loss": 1.6381, + "step": 27577 + }, + { + "epoch": 6.02, + "learning_rate": 1.9927900371422333e-06, + "loss": 1.277, + "step": 27578 + }, + { + "epoch": 6.02, + "learning_rate": 1.9926807952807515e-06, + "loss": 1.0361, + "step": 27579 + }, + { + "epoch": 6.02, + "learning_rate": 1.9925715534192705e-06, + "loss": 0.9578, + "step": 27580 + }, + { + "epoch": 6.02, + "learning_rate": 1.992462311557789e-06, + "loss": 0.5318, + "step": 27581 + }, + { + "epoch": 6.02, + "learning_rate": 1.9923530696963078e-06, + "loss": 0.8418, + "step": 27582 + }, + { + "epoch": 6.02, + "learning_rate": 1.9922438278348264e-06, + "loss": 1.0092, + "step": 27583 + }, + { + "epoch": 6.02, + "learning_rate": 1.992134585973345e-06, + "loss": 0.5771, + "step": 27584 + }, + { + "epoch": 6.02, + "learning_rate": 1.992025344111864e-06, + "loss": 1.95, + "step": 27585 + }, + { + "epoch": 6.02, + "learning_rate": 1.9919161022503826e-06, + "loss": 0.4365, + "step": 27586 + }, + { + "epoch": 6.02, + "learning_rate": 1.9918068603889012e-06, + "loss": 1.6331, + "step": 27587 + }, + { + "epoch": 6.02, + "learning_rate": 1.99169761852742e-06, + "loss": 0.4547, + "step": 27588 + }, + { + "epoch": 6.02, + "learning_rate": 1.9915883766659385e-06, + "loss": 0.2959, + "step": 27589 + }, + { + "epoch": 6.02, + "learning_rate": 1.991479134804457e-06, + "loss": 3.1761, + "step": 27590 + }, + { + "epoch": 6.02, + "learning_rate": 1.9913698929429757e-06, + "loss": 1.3905, + "step": 27591 + }, + { + "epoch": 6.02, + "learning_rate": 1.9912606510814947e-06, + "loss": 0.8626, + "step": 27592 + }, + { + "epoch": 6.02, + "learning_rate": 1.9911514092200133e-06, + "loss": 0.9281, + "step": 27593 + }, + { + "epoch": 6.02, + "learning_rate": 1.991042167358532e-06, + "loss": 1.1374, + "step": 27594 + }, + { + "epoch": 6.02, + "learning_rate": 1.9909329254970506e-06, + "loss": 0.8651, + "step": 27595 + }, + { + "epoch": 6.02, + "learning_rate": 1.990823683635569e-06, + "loss": 1.0344, + "step": 27596 + }, + { + "epoch": 6.02, + "learning_rate": 1.990714441774088e-06, + "loss": 0.3083, + "step": 27597 + }, + { + "epoch": 6.02, + "learning_rate": 1.990605199912607e-06, + "loss": 2.7691, + "step": 27598 + }, + { + "epoch": 6.02, + "learning_rate": 1.9904959580511254e-06, + "loss": 1.0378, + "step": 27599 + }, + { + "epoch": 6.02, + "learning_rate": 1.990386716189644e-06, + "loss": 1.6699, + "step": 27600 + }, + { + "epoch": 6.02, + "learning_rate": 1.9902774743281626e-06, + "loss": 2.9621, + "step": 27601 + }, + { + "epoch": 6.02, + "learning_rate": 1.9901682324666813e-06, + "loss": 0.4284, + "step": 27602 + }, + { + "epoch": 6.02, + "learning_rate": 1.9900589906052003e-06, + "loss": 1.0439, + "step": 27603 + }, + { + "epoch": 6.02, + "learning_rate": 1.989949748743719e-06, + "loss": 2.7597, + "step": 27604 + }, + { + "epoch": 6.02, + "learning_rate": 1.9898405068822375e-06, + "loss": 0.8367, + "step": 27605 + }, + { + "epoch": 6.02, + "learning_rate": 1.989731265020756e-06, + "loss": 0.7368, + "step": 27606 + }, + { + "epoch": 6.03, + "learning_rate": 1.9896220231592747e-06, + "loss": 1.3315, + "step": 27607 + }, + { + "epoch": 6.03, + "learning_rate": 1.9895127812977933e-06, + "loss": 1.3419, + "step": 27608 + }, + { + "epoch": 6.03, + "learning_rate": 1.989403539436312e-06, + "loss": 0.8391, + "step": 27609 + }, + { + "epoch": 6.03, + "learning_rate": 1.989294297574831e-06, + "loss": 1.6772, + "step": 27610 + }, + { + "epoch": 6.03, + "learning_rate": 1.9891850557133496e-06, + "loss": 2.03, + "step": 27611 + }, + { + "epoch": 6.03, + "learning_rate": 1.989075813851868e-06, + "loss": 1.6917, + "step": 27612 + }, + { + "epoch": 6.03, + "learning_rate": 1.988966571990387e-06, + "loss": 1.1984, + "step": 27613 + }, + { + "epoch": 6.03, + "learning_rate": 1.9888573301289054e-06, + "loss": 1.6265, + "step": 27614 + }, + { + "epoch": 6.03, + "learning_rate": 1.9887480882674245e-06, + "loss": 0.998, + "step": 27615 + }, + { + "epoch": 6.03, + "learning_rate": 1.9886388464059427e-06, + "loss": 2.0368, + "step": 27616 + }, + { + "epoch": 6.03, + "learning_rate": 1.9885296045444617e-06, + "loss": 1.7026, + "step": 27617 + }, + { + "epoch": 6.03, + "learning_rate": 1.9884203626829803e-06, + "loss": 1.3351, + "step": 27618 + }, + { + "epoch": 6.03, + "learning_rate": 1.988311120821499e-06, + "loss": 1.8, + "step": 27619 + }, + { + "epoch": 6.03, + "learning_rate": 1.9882018789600175e-06, + "loss": 0.2271, + "step": 27620 + }, + { + "epoch": 6.03, + "learning_rate": 1.988092637098536e-06, + "loss": 0.9478, + "step": 27621 + }, + { + "epoch": 6.03, + "learning_rate": 1.987983395237055e-06, + "loss": 2.2726, + "step": 27622 + }, + { + "epoch": 6.03, + "learning_rate": 1.9878741533755734e-06, + "loss": 2.5476, + "step": 27623 + }, + { + "epoch": 6.03, + "learning_rate": 1.9877649115140924e-06, + "loss": 1.9579, + "step": 27624 + }, + { + "epoch": 6.03, + "learning_rate": 1.987655669652611e-06, + "loss": 1.328, + "step": 27625 + }, + { + "epoch": 6.03, + "learning_rate": 1.9875464277911296e-06, + "loss": 1.8757, + "step": 27626 + }, + { + "epoch": 6.03, + "learning_rate": 1.9874371859296486e-06, + "loss": 0.6051, + "step": 27627 + }, + { + "epoch": 6.03, + "learning_rate": 1.987327944068167e-06, + "loss": 1.8602, + "step": 27628 + }, + { + "epoch": 6.03, + "learning_rate": 1.987218702206686e-06, + "loss": 1.7139, + "step": 27629 + }, + { + "epoch": 6.03, + "learning_rate": 1.9871094603452045e-06, + "loss": 1.006, + "step": 27630 + }, + { + "epoch": 6.03, + "learning_rate": 1.987000218483723e-06, + "loss": 1.6102, + "step": 27631 + }, + { + "epoch": 6.03, + "learning_rate": 1.9868909766222417e-06, + "loss": 2.0735, + "step": 27632 + }, + { + "epoch": 6.03, + "learning_rate": 1.9867817347607603e-06, + "loss": 1.8834, + "step": 27633 + }, + { + "epoch": 6.03, + "learning_rate": 1.9866724928992793e-06, + "loss": 1.6381, + "step": 27634 + }, + { + "epoch": 6.03, + "learning_rate": 1.9865632510377975e-06, + "loss": 1.1778, + "step": 27635 + }, + { + "epoch": 6.03, + "learning_rate": 1.9864540091763166e-06, + "loss": 1.9161, + "step": 27636 + }, + { + "epoch": 6.03, + "learning_rate": 1.986344767314835e-06, + "loss": 2.3091, + "step": 27637 + }, + { + "epoch": 6.03, + "learning_rate": 1.9862355254533538e-06, + "loss": 0.8344, + "step": 27638 + }, + { + "epoch": 6.03, + "learning_rate": 1.986126283591873e-06, + "loss": 1.1937, + "step": 27639 + }, + { + "epoch": 6.03, + "learning_rate": 1.986017041730391e-06, + "loss": 0.9608, + "step": 27640 + }, + { + "epoch": 6.03, + "learning_rate": 1.98590779986891e-06, + "loss": 1.3295, + "step": 27641 + }, + { + "epoch": 6.03, + "learning_rate": 1.9857985580074287e-06, + "loss": 0.2167, + "step": 27642 + }, + { + "epoch": 6.03, + "learning_rate": 1.9856893161459473e-06, + "loss": 1.7271, + "step": 27643 + }, + { + "epoch": 6.03, + "learning_rate": 1.985580074284466e-06, + "loss": 1.0851, + "step": 27644 + }, + { + "epoch": 6.03, + "learning_rate": 1.9854708324229845e-06, + "loss": 1.984, + "step": 27645 + }, + { + "epoch": 6.03, + "learning_rate": 1.9853615905615035e-06, + "loss": 1.1414, + "step": 27646 + }, + { + "epoch": 6.03, + "learning_rate": 1.985252348700022e-06, + "loss": 1.1135, + "step": 27647 + }, + { + "epoch": 6.03, + "learning_rate": 1.9851431068385407e-06, + "loss": 2.0112, + "step": 27648 + }, + { + "epoch": 6.03, + "learning_rate": 1.9850338649770594e-06, + "loss": 1.8682, + "step": 27649 + }, + { + "epoch": 6.03, + "learning_rate": 1.984924623115578e-06, + "loss": 1.9505, + "step": 27650 + }, + { + "epoch": 6.03, + "learning_rate": 1.984815381254097e-06, + "loss": 1.2657, + "step": 27651 + }, + { + "epoch": 6.03, + "learning_rate": 1.984706139392615e-06, + "loss": 0.7395, + "step": 27652 + }, + { + "epoch": 6.04, + "learning_rate": 1.9845968975311342e-06, + "loss": 0.9004, + "step": 27653 + }, + { + "epoch": 6.04, + "learning_rate": 1.984487655669653e-06, + "loss": 1.2672, + "step": 27654 + }, + { + "epoch": 6.04, + "learning_rate": 1.9843784138081714e-06, + "loss": 1.3674, + "step": 27655 + }, + { + "epoch": 6.04, + "learning_rate": 1.98426917194669e-06, + "loss": 1.8228, + "step": 27656 + }, + { + "epoch": 6.04, + "learning_rate": 1.9841599300852087e-06, + "loss": 2.3775, + "step": 27657 + }, + { + "epoch": 6.04, + "learning_rate": 1.9840506882237277e-06, + "loss": 2.2027, + "step": 27658 + }, + { + "epoch": 6.04, + "learning_rate": 1.9839414463622463e-06, + "loss": 2.441, + "step": 27659 + }, + { + "epoch": 6.04, + "learning_rate": 1.983832204500765e-06, + "loss": 3.0012, + "step": 27660 + }, + { + "epoch": 6.04, + "learning_rate": 1.9837229626392835e-06, + "loss": 1.2071, + "step": 27661 + }, + { + "epoch": 6.04, + "learning_rate": 1.983613720777802e-06, + "loss": 1.5016, + "step": 27662 + }, + { + "epoch": 6.04, + "learning_rate": 1.983504478916321e-06, + "loss": 2.0806, + "step": 27663 + }, + { + "epoch": 6.04, + "learning_rate": 1.9833952370548394e-06, + "loss": 1.7513, + "step": 27664 + }, + { + "epoch": 6.04, + "learning_rate": 1.9832859951933584e-06, + "loss": 3.0424, + "step": 27665 + }, + { + "epoch": 6.04, + "learning_rate": 1.983176753331877e-06, + "loss": 1.3888, + "step": 27666 + }, + { + "epoch": 6.04, + "learning_rate": 1.9830675114703956e-06, + "loss": 2.451, + "step": 27667 + }, + { + "epoch": 6.04, + "learning_rate": 1.9829582696089142e-06, + "loss": 2.3457, + "step": 27668 + }, + { + "epoch": 6.04, + "learning_rate": 1.982849027747433e-06, + "loss": 1.0649, + "step": 27669 + }, + { + "epoch": 6.04, + "learning_rate": 1.982739785885952e-06, + "loss": 1.4831, + "step": 27670 + }, + { + "epoch": 6.04, + "learning_rate": 1.9826305440244705e-06, + "loss": 2.5459, + "step": 27671 + }, + { + "epoch": 6.04, + "learning_rate": 1.982521302162989e-06, + "loss": 0.6637, + "step": 27672 + }, + { + "epoch": 6.04, + "learning_rate": 1.9824120603015077e-06, + "loss": 0.6706, + "step": 27673 + }, + { + "epoch": 6.04, + "learning_rate": 1.9823028184400263e-06, + "loss": 1.0729, + "step": 27674 + }, + { + "epoch": 6.04, + "learning_rate": 1.982193576578545e-06, + "loss": 1.3786, + "step": 27675 + }, + { + "epoch": 6.04, + "learning_rate": 1.9820843347170635e-06, + "loss": 2.5228, + "step": 27676 + }, + { + "epoch": 6.04, + "learning_rate": 1.9819750928555826e-06, + "loss": 1.3634, + "step": 27677 + }, + { + "epoch": 6.04, + "learning_rate": 1.981865850994101e-06, + "loss": 1.5085, + "step": 27678 + }, + { + "epoch": 6.04, + "learning_rate": 1.98175660913262e-06, + "loss": 1.4604, + "step": 27679 + }, + { + "epoch": 6.04, + "learning_rate": 1.9816473672711384e-06, + "loss": 0.966, + "step": 27680 + }, + { + "epoch": 6.04, + "learning_rate": 1.981538125409657e-06, + "loss": 1.5955, + "step": 27681 + }, + { + "epoch": 6.04, + "learning_rate": 1.9814288835481756e-06, + "loss": 1.7323, + "step": 27682 + }, + { + "epoch": 6.04, + "learning_rate": 1.9813196416866947e-06, + "loss": 0.6151, + "step": 27683 + }, + { + "epoch": 6.04, + "learning_rate": 1.9812103998252133e-06, + "loss": 0.7557, + "step": 27684 + }, + { + "epoch": 6.04, + "learning_rate": 1.981101157963732e-06, + "loss": 1.5355, + "step": 27685 + }, + { + "epoch": 6.04, + "learning_rate": 1.9809919161022505e-06, + "loss": 1.6511, + "step": 27686 + }, + { + "epoch": 6.04, + "learning_rate": 1.980882674240769e-06, + "loss": 2.5824, + "step": 27687 + }, + { + "epoch": 6.04, + "learning_rate": 1.9807734323792877e-06, + "loss": 1.442, + "step": 27688 + }, + { + "epoch": 6.04, + "learning_rate": 1.9806641905178063e-06, + "loss": 1.4842, + "step": 27689 + }, + { + "epoch": 6.04, + "learning_rate": 1.9805549486563254e-06, + "loss": 0.7712, + "step": 27690 + }, + { + "epoch": 6.04, + "learning_rate": 1.980445706794844e-06, + "loss": 1.6955, + "step": 27691 + }, + { + "epoch": 6.04, + "learning_rate": 1.9803364649333626e-06, + "loss": 2.0124, + "step": 27692 + }, + { + "epoch": 6.04, + "learning_rate": 1.980227223071881e-06, + "loss": 1.5578, + "step": 27693 + }, + { + "epoch": 6.04, + "learning_rate": 1.9801179812104e-06, + "loss": 1.2773, + "step": 27694 + }, + { + "epoch": 6.04, + "learning_rate": 1.980008739348919e-06, + "loss": 2.0333, + "step": 27695 + }, + { + "epoch": 6.04, + "learning_rate": 1.9798994974874375e-06, + "loss": 1.6708, + "step": 27696 + }, + { + "epoch": 6.04, + "learning_rate": 1.979790255625956e-06, + "loss": 1.192, + "step": 27697 + }, + { + "epoch": 6.04, + "learning_rate": 1.9796810137644747e-06, + "loss": 1.1248, + "step": 27698 + }, + { + "epoch": 6.05, + "learning_rate": 1.9795717719029933e-06, + "loss": 2.2991, + "step": 27699 + }, + { + "epoch": 6.05, + "learning_rate": 1.9794625300415123e-06, + "loss": 1.6438, + "step": 27700 + }, + { + "epoch": 6.05, + "learning_rate": 1.9793532881800305e-06, + "loss": 1.9396, + "step": 27701 + }, + { + "epoch": 6.05, + "learning_rate": 1.9792440463185495e-06, + "loss": 1.7202, + "step": 27702 + }, + { + "epoch": 6.05, + "learning_rate": 1.979134804457068e-06, + "loss": 0.697, + "step": 27703 + }, + { + "epoch": 6.05, + "learning_rate": 1.9790255625955868e-06, + "loss": 1.6396, + "step": 27704 + }, + { + "epoch": 6.05, + "learning_rate": 1.9789163207341054e-06, + "loss": 1.3608, + "step": 27705 + }, + { + "epoch": 6.05, + "learning_rate": 1.978807078872624e-06, + "loss": 0.8399, + "step": 27706 + }, + { + "epoch": 6.05, + "learning_rate": 1.978697837011143e-06, + "loss": 2.6582, + "step": 27707 + }, + { + "epoch": 6.05, + "learning_rate": 1.9785885951496612e-06, + "loss": 0.3004, + "step": 27708 + }, + { + "epoch": 6.05, + "learning_rate": 1.9784793532881803e-06, + "loss": 1.0385, + "step": 27709 + }, + { + "epoch": 6.05, + "learning_rate": 1.978370111426699e-06, + "loss": 2.492, + "step": 27710 + }, + { + "epoch": 6.05, + "learning_rate": 1.9782608695652175e-06, + "loss": 0.1399, + "step": 27711 + }, + { + "epoch": 6.05, + "learning_rate": 1.9781516277037365e-06, + "loss": 1.6186, + "step": 27712 + }, + { + "epoch": 6.05, + "learning_rate": 1.9780423858422547e-06, + "loss": 0.6633, + "step": 27713 + }, + { + "epoch": 6.05, + "learning_rate": 1.9779331439807737e-06, + "loss": 0.7347, + "step": 27714 + }, + { + "epoch": 6.05, + "learning_rate": 1.9778239021192923e-06, + "loss": 0.8823, + "step": 27715 + }, + { + "epoch": 6.05, + "learning_rate": 1.977714660257811e-06, + "loss": 1.0244, + "step": 27716 + }, + { + "epoch": 6.05, + "learning_rate": 1.9776054183963296e-06, + "loss": 3.1525, + "step": 27717 + }, + { + "epoch": 6.05, + "learning_rate": 1.977496176534848e-06, + "loss": 1.188, + "step": 27718 + }, + { + "epoch": 6.05, + "learning_rate": 1.977386934673367e-06, + "loss": 0.7128, + "step": 27719 + }, + { + "epoch": 6.05, + "learning_rate": 1.9772776928118854e-06, + "loss": 1.6905, + "step": 27720 + }, + { + "epoch": 6.05, + "learning_rate": 1.9771684509504044e-06, + "loss": 0.8509, + "step": 27721 + }, + { + "epoch": 6.05, + "learning_rate": 1.977059209088923e-06, + "loss": 1.291, + "step": 27722 + }, + { + "epoch": 6.05, + "learning_rate": 1.9769499672274417e-06, + "loss": 1.0871, + "step": 27723 + }, + { + "epoch": 6.05, + "learning_rate": 1.9768407253659607e-06, + "loss": 0.7043, + "step": 27724 + }, + { + "epoch": 6.05, + "learning_rate": 1.976731483504479e-06, + "loss": 1.1922, + "step": 27725 + }, + { + "epoch": 6.05, + "learning_rate": 1.976622241642998e-06, + "loss": 1.9701, + "step": 27726 + }, + { + "epoch": 6.05, + "learning_rate": 1.9765129997815165e-06, + "loss": 1.5698, + "step": 27727 + }, + { + "epoch": 6.05, + "learning_rate": 1.976403757920035e-06, + "loss": 0.8578, + "step": 27728 + }, + { + "epoch": 6.05, + "learning_rate": 1.9762945160585537e-06, + "loss": 1.2191, + "step": 27729 + }, + { + "epoch": 6.05, + "learning_rate": 1.9761852741970724e-06, + "loss": 0.5327, + "step": 27730 + }, + { + "epoch": 6.05, + "learning_rate": 1.9760760323355914e-06, + "loss": 0.6358, + "step": 27731 + }, + { + "epoch": 6.05, + "learning_rate": 1.9759667904741096e-06, + "loss": 1.464, + "step": 27732 + }, + { + "epoch": 6.05, + "learning_rate": 1.9758575486126286e-06, + "loss": 1.4392, + "step": 27733 + }, + { + "epoch": 6.05, + "learning_rate": 1.9757483067511472e-06, + "loss": 0.4451, + "step": 27734 + }, + { + "epoch": 6.05, + "learning_rate": 1.975639064889666e-06, + "loss": 1.0545, + "step": 27735 + }, + { + "epoch": 6.05, + "learning_rate": 1.975529823028185e-06, + "loss": 0.8473, + "step": 27736 + }, + { + "epoch": 6.05, + "learning_rate": 1.975420581166703e-06, + "loss": 0.8787, + "step": 27737 + }, + { + "epoch": 6.05, + "learning_rate": 1.975311339305222e-06, + "loss": 0.7661, + "step": 27738 + }, + { + "epoch": 6.05, + "learning_rate": 1.9752020974437407e-06, + "loss": 0.8681, + "step": 27739 + }, + { + "epoch": 6.05, + "learning_rate": 1.9750928555822593e-06, + "loss": 0.2436, + "step": 27740 + }, + { + "epoch": 6.05, + "learning_rate": 1.974983613720778e-06, + "loss": 0.1248, + "step": 27741 + }, + { + "epoch": 6.05, + "learning_rate": 1.9748743718592965e-06, + "loss": 0.6222, + "step": 27742 + }, + { + "epoch": 6.05, + "learning_rate": 1.9747651299978156e-06, + "loss": 2.3757, + "step": 27743 + }, + { + "epoch": 6.05, + "learning_rate": 1.974655888136334e-06, + "loss": 1.5589, + "step": 27744 + }, + { + "epoch": 6.06, + "learning_rate": 1.9745466462748528e-06, + "loss": 1.4337, + "step": 27745 + }, + { + "epoch": 6.06, + "learning_rate": 1.9744374044133714e-06, + "loss": 1.0062, + "step": 27746 + }, + { + "epoch": 6.06, + "learning_rate": 1.97432816255189e-06, + "loss": 0.2041, + "step": 27747 + }, + { + "epoch": 6.06, + "learning_rate": 1.9742189206904086e-06, + "loss": 2.1682, + "step": 27748 + }, + { + "epoch": 6.06, + "learning_rate": 1.9741096788289272e-06, + "loss": 1.4322, + "step": 27749 + }, + { + "epoch": 6.06, + "learning_rate": 1.9740004369674463e-06, + "loss": 1.4344, + "step": 27750 + }, + { + "epoch": 6.06, + "learning_rate": 1.973891195105965e-06, + "loss": 1.6888, + "step": 27751 + }, + { + "epoch": 6.06, + "learning_rate": 1.9737819532444835e-06, + "loss": 1.3648, + "step": 27752 + }, + { + "epoch": 6.06, + "learning_rate": 1.973672711383002e-06, + "loss": 1.2781, + "step": 27753 + }, + { + "epoch": 6.06, + "learning_rate": 1.9735634695215207e-06, + "loss": 1.9659, + "step": 27754 + }, + { + "epoch": 6.06, + "learning_rate": 1.9734542276600393e-06, + "loss": 2.7406, + "step": 27755 + }, + { + "epoch": 6.06, + "learning_rate": 1.9733449857985584e-06, + "loss": 1.6124, + "step": 27756 + }, + { + "epoch": 6.06, + "learning_rate": 1.973235743937077e-06, + "loss": 1.4622, + "step": 27757 + }, + { + "epoch": 6.06, + "learning_rate": 1.9731265020755956e-06, + "loss": 0.9679, + "step": 27758 + }, + { + "epoch": 6.06, + "learning_rate": 1.973017260214114e-06, + "loss": 2.5428, + "step": 27759 + }, + { + "epoch": 6.06, + "learning_rate": 1.972908018352633e-06, + "loss": 0.7626, + "step": 27760 + }, + { + "epoch": 6.06, + "learning_rate": 1.9727987764911514e-06, + "loss": 0.9718, + "step": 27761 + }, + { + "epoch": 6.06, + "learning_rate": 1.9726895346296704e-06, + "loss": 1.0637, + "step": 27762 + }, + { + "epoch": 6.06, + "learning_rate": 1.972580292768189e-06, + "loss": 1.3822, + "step": 27763 + }, + { + "epoch": 6.06, + "learning_rate": 1.9724710509067077e-06, + "loss": 2.3398, + "step": 27764 + }, + { + "epoch": 6.06, + "learning_rate": 1.9723618090452263e-06, + "loss": 1.1078, + "step": 27765 + }, + { + "epoch": 6.06, + "learning_rate": 1.972252567183745e-06, + "loss": 0.5895, + "step": 27766 + }, + { + "epoch": 6.06, + "learning_rate": 1.9721433253222635e-06, + "loss": 0.0619, + "step": 27767 + }, + { + "epoch": 6.06, + "learning_rate": 1.9720340834607825e-06, + "loss": 1.7559, + "step": 27768 + }, + { + "epoch": 6.06, + "learning_rate": 1.971924841599301e-06, + "loss": 1.7242, + "step": 27769 + }, + { + "epoch": 6.06, + "learning_rate": 1.9718155997378198e-06, + "loss": 1.1524, + "step": 27770 + }, + { + "epoch": 6.06, + "learning_rate": 1.9717063578763384e-06, + "loss": 1.6579, + "step": 27771 + }, + { + "epoch": 6.06, + "learning_rate": 1.971597116014857e-06, + "loss": 1.1558, + "step": 27772 + }, + { + "epoch": 6.06, + "learning_rate": 1.9714878741533756e-06, + "loss": 1.1159, + "step": 27773 + }, + { + "epoch": 6.06, + "learning_rate": 1.971378632291894e-06, + "loss": 1.2729, + "step": 27774 + }, + { + "epoch": 6.06, + "learning_rate": 1.9712693904304132e-06, + "loss": 0.9837, + "step": 27775 + }, + { + "epoch": 6.06, + "learning_rate": 1.971160148568932e-06, + "loss": 2.0215, + "step": 27776 + }, + { + "epoch": 6.06, + "learning_rate": 1.9710509067074505e-06, + "loss": 0.5844, + "step": 27777 + }, + { + "epoch": 6.06, + "learning_rate": 1.970941664845969e-06, + "loss": 1.7818, + "step": 27778 + }, + { + "epoch": 6.06, + "learning_rate": 1.9708324229844877e-06, + "loss": 1.4596, + "step": 27779 + }, + { + "epoch": 6.06, + "learning_rate": 1.9707231811230067e-06, + "loss": 1.7213, + "step": 27780 + }, + { + "epoch": 6.06, + "learning_rate": 1.970613939261525e-06, + "loss": 1.3962, + "step": 27781 + }, + { + "epoch": 6.06, + "learning_rate": 1.970504697400044e-06, + "loss": 0.746, + "step": 27782 + }, + { + "epoch": 6.06, + "learning_rate": 1.9703954555385625e-06, + "loss": 0.535, + "step": 27783 + }, + { + "epoch": 6.06, + "learning_rate": 1.970286213677081e-06, + "loss": 0.8286, + "step": 27784 + }, + { + "epoch": 6.06, + "learning_rate": 1.9701769718155998e-06, + "loss": 1.1483, + "step": 27785 + }, + { + "epoch": 6.06, + "learning_rate": 1.9700677299541184e-06, + "loss": 1.2559, + "step": 27786 + }, + { + "epoch": 6.06, + "learning_rate": 1.9699584880926374e-06, + "loss": 1.1497, + "step": 27787 + }, + { + "epoch": 6.06, + "learning_rate": 1.9698492462311556e-06, + "loss": 0.7933, + "step": 27788 + }, + { + "epoch": 6.06, + "learning_rate": 1.9697400043696746e-06, + "loss": 1.0411, + "step": 27789 + }, + { + "epoch": 6.07, + "learning_rate": 1.9696307625081932e-06, + "loss": 2.1891, + "step": 27790 + }, + { + "epoch": 6.07, + "learning_rate": 1.969521520646712e-06, + "loss": 1.2161, + "step": 27791 + }, + { + "epoch": 6.07, + "learning_rate": 1.969412278785231e-06, + "loss": 0.9938, + "step": 27792 + }, + { + "epoch": 6.07, + "learning_rate": 1.969303036923749e-06, + "loss": 1.4564, + "step": 27793 + }, + { + "epoch": 6.07, + "learning_rate": 1.969193795062268e-06, + "loss": 1.8714, + "step": 27794 + }, + { + "epoch": 6.07, + "learning_rate": 1.9690845532007867e-06, + "loss": 2.2103, + "step": 27795 + }, + { + "epoch": 6.07, + "learning_rate": 1.9689753113393053e-06, + "loss": 0.8195, + "step": 27796 + }, + { + "epoch": 6.07, + "learning_rate": 1.9688660694778244e-06, + "loss": 0.9783, + "step": 27797 + }, + { + "epoch": 6.07, + "learning_rate": 1.9687568276163426e-06, + "loss": 0.5896, + "step": 27798 + }, + { + "epoch": 6.07, + "learning_rate": 1.9686475857548616e-06, + "loss": 1.488, + "step": 27799 + }, + { + "epoch": 6.07, + "learning_rate": 1.96853834389338e-06, + "loss": 0.2589, + "step": 27800 + }, + { + "epoch": 6.07, + "learning_rate": 1.968429102031899e-06, + "loss": 1.1717, + "step": 27801 + }, + { + "epoch": 6.07, + "learning_rate": 1.9683198601704174e-06, + "loss": 1.6506, + "step": 27802 + }, + { + "epoch": 6.07, + "learning_rate": 1.968210618308936e-06, + "loss": 0.923, + "step": 27803 + }, + { + "epoch": 6.07, + "learning_rate": 1.968101376447455e-06, + "loss": 0.9055, + "step": 27804 + }, + { + "epoch": 6.07, + "learning_rate": 1.9679921345859733e-06, + "loss": 0.7387, + "step": 27805 + }, + { + "epoch": 6.07, + "learning_rate": 1.9678828927244923e-06, + "loss": 0.9069, + "step": 27806 + }, + { + "epoch": 6.07, + "learning_rate": 1.967773650863011e-06, + "loss": 0.5886, + "step": 27807 + }, + { + "epoch": 6.07, + "learning_rate": 1.9676644090015295e-06, + "loss": 4.1412, + "step": 27808 + }, + { + "epoch": 6.07, + "learning_rate": 1.9675551671400485e-06, + "loss": 0.8543, + "step": 27809 + }, + { + "epoch": 6.07, + "learning_rate": 1.9674459252785667e-06, + "loss": 1.5318, + "step": 27810 + }, + { + "epoch": 6.07, + "learning_rate": 1.9673366834170858e-06, + "loss": 2.0718, + "step": 27811 + }, + { + "epoch": 6.07, + "learning_rate": 1.9672274415556044e-06, + "loss": 0.5718, + "step": 27812 + }, + { + "epoch": 6.07, + "learning_rate": 1.967118199694123e-06, + "loss": 0.8558, + "step": 27813 + }, + { + "epoch": 6.07, + "learning_rate": 1.9670089578326416e-06, + "loss": 1.9113, + "step": 27814 + }, + { + "epoch": 6.07, + "learning_rate": 1.9668997159711602e-06, + "loss": 1.4274, + "step": 27815 + }, + { + "epoch": 6.07, + "learning_rate": 1.9667904741096792e-06, + "loss": 1.7145, + "step": 27816 + }, + { + "epoch": 6.07, + "learning_rate": 1.9666812322481974e-06, + "loss": 0.7741, + "step": 27817 + }, + { + "epoch": 6.07, + "learning_rate": 1.9665719903867165e-06, + "loss": 1.6311, + "step": 27818 + }, + { + "epoch": 6.07, + "learning_rate": 1.966462748525235e-06, + "loss": 1.5349, + "step": 27819 + }, + { + "epoch": 6.07, + "learning_rate": 1.9663535066637537e-06, + "loss": 0.6416, + "step": 27820 + }, + { + "epoch": 6.07, + "learning_rate": 1.9662442648022723e-06, + "loss": 2.2268, + "step": 27821 + }, + { + "epoch": 6.07, + "learning_rate": 1.966135022940791e-06, + "loss": 1.3646, + "step": 27822 + }, + { + "epoch": 6.07, + "learning_rate": 1.96602578107931e-06, + "loss": 1.3529, + "step": 27823 + }, + { + "epoch": 6.07, + "learning_rate": 1.9659165392178286e-06, + "loss": 0.6372, + "step": 27824 + }, + { + "epoch": 6.07, + "learning_rate": 1.965807297356347e-06, + "loss": 0.6317, + "step": 27825 + }, + { + "epoch": 6.07, + "learning_rate": 1.9656980554948658e-06, + "loss": 1.5056, + "step": 27826 + }, + { + "epoch": 6.07, + "learning_rate": 1.9655888136333844e-06, + "loss": 2.2223, + "step": 27827 + }, + { + "epoch": 6.07, + "learning_rate": 1.965479571771903e-06, + "loss": 2.2489, + "step": 27828 + }, + { + "epoch": 6.07, + "learning_rate": 1.9653703299104216e-06, + "loss": 1.2024, + "step": 27829 + }, + { + "epoch": 6.07, + "learning_rate": 1.9652610880489406e-06, + "loss": 1.9474, + "step": 27830 + }, + { + "epoch": 6.07, + "learning_rate": 1.9651518461874593e-06, + "loss": 0.5577, + "step": 27831 + }, + { + "epoch": 6.07, + "learning_rate": 1.965042604325978e-06, + "loss": 1.7444, + "step": 27832 + }, + { + "epoch": 6.07, + "learning_rate": 1.9649333624644965e-06, + "loss": 0.2492, + "step": 27833 + }, + { + "epoch": 6.07, + "learning_rate": 1.964824120603015e-06, + "loss": 1.7047, + "step": 27834 + }, + { + "epoch": 6.07, + "learning_rate": 1.964714878741534e-06, + "loss": 1.2102, + "step": 27835 + }, + { + "epoch": 6.08, + "learning_rate": 1.9646056368800527e-06, + "loss": 1.205, + "step": 27836 + }, + { + "epoch": 6.08, + "learning_rate": 1.9644963950185714e-06, + "loss": 1.9303, + "step": 27837 + }, + { + "epoch": 6.08, + "learning_rate": 1.96438715315709e-06, + "loss": 1.0808, + "step": 27838 + }, + { + "epoch": 6.08, + "learning_rate": 1.9642779112956086e-06, + "loss": 1.6614, + "step": 27839 + }, + { + "epoch": 6.08, + "learning_rate": 1.964168669434127e-06, + "loss": 2.1156, + "step": 27840 + }, + { + "epoch": 6.08, + "learning_rate": 1.964059427572646e-06, + "loss": 1.5832, + "step": 27841 + }, + { + "epoch": 6.08, + "learning_rate": 1.963950185711165e-06, + "loss": 1.8059, + "step": 27842 + }, + { + "epoch": 6.08, + "learning_rate": 1.9638409438496834e-06, + "loss": 2.8158, + "step": 27843 + }, + { + "epoch": 6.08, + "learning_rate": 1.963731701988202e-06, + "loss": 3.0769, + "step": 27844 + }, + { + "epoch": 6.08, + "learning_rate": 1.9636224601267207e-06, + "loss": 1.5012, + "step": 27845 + }, + { + "epoch": 6.08, + "learning_rate": 1.9635132182652393e-06, + "loss": 0.8602, + "step": 27846 + }, + { + "epoch": 6.08, + "learning_rate": 1.963403976403758e-06, + "loss": 1.7233, + "step": 27847 + }, + { + "epoch": 6.08, + "learning_rate": 1.963294734542277e-06, + "loss": 1.32, + "step": 27848 + }, + { + "epoch": 6.08, + "learning_rate": 1.9631854926807955e-06, + "loss": 0.3295, + "step": 27849 + }, + { + "epoch": 6.08, + "learning_rate": 1.963076250819314e-06, + "loss": 0.6971, + "step": 27850 + }, + { + "epoch": 6.08, + "learning_rate": 1.9629670089578328e-06, + "loss": 1.9295, + "step": 27851 + }, + { + "epoch": 6.08, + "learning_rate": 1.9628577670963514e-06, + "loss": 1.7959, + "step": 27852 + }, + { + "epoch": 6.08, + "learning_rate": 1.9627485252348704e-06, + "loss": 1.6086, + "step": 27853 + }, + { + "epoch": 6.08, + "learning_rate": 1.9626392833733886e-06, + "loss": 1.9413, + "step": 27854 + }, + { + "epoch": 6.08, + "learning_rate": 1.9625300415119076e-06, + "loss": 2.1699, + "step": 27855 + }, + { + "epoch": 6.08, + "learning_rate": 1.9624207996504262e-06, + "loss": 0.6875, + "step": 27856 + }, + { + "epoch": 6.08, + "learning_rate": 1.962311557788945e-06, + "loss": 1.3811, + "step": 27857 + }, + { + "epoch": 6.08, + "learning_rate": 1.9622023159274635e-06, + "loss": 0.4373, + "step": 27858 + }, + { + "epoch": 6.08, + "learning_rate": 1.962093074065982e-06, + "loss": 1.0002, + "step": 27859 + }, + { + "epoch": 6.08, + "learning_rate": 1.961983832204501e-06, + "loss": 3.8304, + "step": 27860 + }, + { + "epoch": 6.08, + "learning_rate": 1.9618745903430193e-06, + "loss": 1.0413, + "step": 27861 + }, + { + "epoch": 6.08, + "learning_rate": 1.9617653484815383e-06, + "loss": 0.5733, + "step": 27862 + }, + { + "epoch": 6.08, + "learning_rate": 1.961656106620057e-06, + "loss": 0.9489, + "step": 27863 + }, + { + "epoch": 6.08, + "learning_rate": 1.9615468647585755e-06, + "loss": 0.3023, + "step": 27864 + }, + { + "epoch": 6.08, + "learning_rate": 1.9614376228970946e-06, + "loss": 2.4322, + "step": 27865 + }, + { + "epoch": 6.08, + "learning_rate": 1.9613283810356128e-06, + "loss": 1.6074, + "step": 27866 + }, + { + "epoch": 6.08, + "learning_rate": 1.961219139174132e-06, + "loss": 0.0906, + "step": 27867 + }, + { + "epoch": 6.08, + "learning_rate": 1.9611098973126504e-06, + "loss": 2.038, + "step": 27868 + }, + { + "epoch": 6.08, + "learning_rate": 1.961000655451169e-06, + "loss": 1.5147, + "step": 27869 + }, + { + "epoch": 6.08, + "learning_rate": 1.9608914135896876e-06, + "loss": 1.083, + "step": 27870 + }, + { + "epoch": 6.08, + "learning_rate": 1.9607821717282062e-06, + "loss": 0.8101, + "step": 27871 + }, + { + "epoch": 6.08, + "learning_rate": 1.9606729298667253e-06, + "loss": 0.1524, + "step": 27872 + }, + { + "epoch": 6.08, + "learning_rate": 1.9605636880052435e-06, + "loss": 0.4168, + "step": 27873 + }, + { + "epoch": 6.08, + "learning_rate": 1.9604544461437625e-06, + "loss": 0.5219, + "step": 27874 + }, + { + "epoch": 6.08, + "learning_rate": 1.960345204282281e-06, + "loss": 1.3447, + "step": 27875 + }, + { + "epoch": 6.08, + "learning_rate": 1.9602359624207997e-06, + "loss": 2.7069, + "step": 27876 + }, + { + "epoch": 6.08, + "learning_rate": 1.9601267205593188e-06, + "loss": 1.4066, + "step": 27877 + }, + { + "epoch": 6.08, + "learning_rate": 1.960017478697837e-06, + "loss": 2.951, + "step": 27878 + }, + { + "epoch": 6.08, + "learning_rate": 1.959908236836356e-06, + "loss": 1.7697, + "step": 27879 + }, + { + "epoch": 6.08, + "learning_rate": 1.9597989949748746e-06, + "loss": 2.2392, + "step": 27880 + }, + { + "epoch": 6.08, + "learning_rate": 1.959689753113393e-06, + "loss": 2.6123, + "step": 27881 + }, + { + "epoch": 6.09, + "learning_rate": 1.959580511251912e-06, + "loss": 1.2096, + "step": 27882 + }, + { + "epoch": 6.09, + "learning_rate": 1.9594712693904304e-06, + "loss": 0.783, + "step": 27883 + }, + { + "epoch": 6.09, + "learning_rate": 1.9593620275289495e-06, + "loss": 0.0943, + "step": 27884 + }, + { + "epoch": 6.09, + "learning_rate": 1.9592527856674676e-06, + "loss": 0.5265, + "step": 27885 + }, + { + "epoch": 6.09, + "learning_rate": 1.9591435438059867e-06, + "loss": 1.0528, + "step": 27886 + }, + { + "epoch": 6.09, + "learning_rate": 1.9590343019445053e-06, + "loss": 1.0238, + "step": 27887 + }, + { + "epoch": 6.09, + "learning_rate": 1.958925060083024e-06, + "loss": 1.3948, + "step": 27888 + }, + { + "epoch": 6.09, + "learning_rate": 1.958815818221543e-06, + "loss": 0.4722, + "step": 27889 + }, + { + "epoch": 6.09, + "learning_rate": 1.958706576360061e-06, + "loss": 0.8731, + "step": 27890 + }, + { + "epoch": 6.09, + "learning_rate": 1.95859733449858e-06, + "loss": 1.7117, + "step": 27891 + }, + { + "epoch": 6.09, + "learning_rate": 1.9584880926370988e-06, + "loss": 1.0195, + "step": 27892 + }, + { + "epoch": 6.09, + "learning_rate": 1.9583788507756174e-06, + "loss": 3.4775, + "step": 27893 + }, + { + "epoch": 6.09, + "learning_rate": 1.958269608914136e-06, + "loss": 1.5804, + "step": 27894 + }, + { + "epoch": 6.09, + "learning_rate": 1.9581603670526546e-06, + "loss": 2.2876, + "step": 27895 + }, + { + "epoch": 6.09, + "learning_rate": 1.9580511251911736e-06, + "loss": 1.504, + "step": 27896 + }, + { + "epoch": 6.09, + "learning_rate": 1.9579418833296922e-06, + "loss": 0.3191, + "step": 27897 + }, + { + "epoch": 6.09, + "learning_rate": 1.957832641468211e-06, + "loss": 1.0768, + "step": 27898 + }, + { + "epoch": 6.09, + "learning_rate": 1.9577233996067295e-06, + "loss": 2.9247, + "step": 27899 + }, + { + "epoch": 6.09, + "learning_rate": 1.957614157745248e-06, + "loss": 1.5828, + "step": 27900 + }, + { + "epoch": 6.09, + "learning_rate": 1.957504915883767e-06, + "loss": 1.66, + "step": 27901 + }, + { + "epoch": 6.09, + "learning_rate": 1.9573956740222853e-06, + "loss": 1.0865, + "step": 27902 + }, + { + "epoch": 6.09, + "learning_rate": 1.9572864321608043e-06, + "loss": 0.7911, + "step": 27903 + }, + { + "epoch": 6.09, + "learning_rate": 1.957177190299323e-06, + "loss": 1.0277, + "step": 27904 + }, + { + "epoch": 6.09, + "learning_rate": 1.9570679484378416e-06, + "loss": 2.7892, + "step": 27905 + }, + { + "epoch": 6.09, + "learning_rate": 1.95695870657636e-06, + "loss": 2.3658, + "step": 27906 + }, + { + "epoch": 6.09, + "learning_rate": 1.9568494647148788e-06, + "loss": 2.4672, + "step": 27907 + }, + { + "epoch": 6.09, + "learning_rate": 1.956740222853398e-06, + "loss": 2.675, + "step": 27908 + }, + { + "epoch": 6.09, + "learning_rate": 1.9566309809919164e-06, + "loss": 1.7546, + "step": 27909 + }, + { + "epoch": 6.09, + "learning_rate": 1.956521739130435e-06, + "loss": 2.2868, + "step": 27910 + }, + { + "epoch": 6.09, + "learning_rate": 1.9564124972689536e-06, + "loss": 1.9532, + "step": 27911 + }, + { + "epoch": 6.09, + "learning_rate": 1.9563032554074723e-06, + "loss": 2.8132, + "step": 27912 + }, + { + "epoch": 6.09, + "learning_rate": 1.956194013545991e-06, + "loss": 1.025, + "step": 27913 + }, + { + "epoch": 6.09, + "learning_rate": 1.9560847716845095e-06, + "loss": 0.7375, + "step": 27914 + }, + { + "epoch": 6.09, + "learning_rate": 1.9559755298230285e-06, + "loss": 1.555, + "step": 27915 + }, + { + "epoch": 6.09, + "learning_rate": 1.955866287961547e-06, + "loss": 1.6232, + "step": 27916 + }, + { + "epoch": 6.09, + "learning_rate": 1.9557570461000657e-06, + "loss": 1.0264, + "step": 27917 + }, + { + "epoch": 6.09, + "learning_rate": 1.9556478042385843e-06, + "loss": 0.6981, + "step": 27918 + }, + { + "epoch": 6.09, + "learning_rate": 1.955538562377103e-06, + "loss": 2.3357, + "step": 27919 + }, + { + "epoch": 6.09, + "learning_rate": 1.9554293205156216e-06, + "loss": 0.9613, + "step": 27920 + }, + { + "epoch": 6.09, + "learning_rate": 1.9553200786541406e-06, + "loss": 1.1925, + "step": 27921 + }, + { + "epoch": 6.09, + "learning_rate": 1.9552108367926592e-06, + "loss": 0.5187, + "step": 27922 + }, + { + "epoch": 6.09, + "learning_rate": 1.955101594931178e-06, + "loss": 1.5934, + "step": 27923 + }, + { + "epoch": 6.09, + "learning_rate": 1.9549923530696964e-06, + "loss": 1.4741, + "step": 27924 + }, + { + "epoch": 6.09, + "learning_rate": 1.954883111208215e-06, + "loss": 1.3594, + "step": 27925 + }, + { + "epoch": 6.09, + "learning_rate": 1.9547738693467337e-06, + "loss": 1.2037, + "step": 27926 + }, + { + "epoch": 6.09, + "learning_rate": 1.9546646274852523e-06, + "loss": 2.2497, + "step": 27927 + }, + { + "epoch": 6.1, + "learning_rate": 1.9545553856237713e-06, + "loss": 1.2109, + "step": 27928 + }, + { + "epoch": 6.1, + "learning_rate": 1.95444614376229e-06, + "loss": 2.2266, + "step": 27929 + }, + { + "epoch": 6.1, + "learning_rate": 1.9543369019008085e-06, + "loss": 1.3823, + "step": 27930 + }, + { + "epoch": 6.1, + "learning_rate": 1.954227660039327e-06, + "loss": 0.7886, + "step": 27931 + }, + { + "epoch": 6.1, + "learning_rate": 1.9541184181778457e-06, + "loss": 2.0217, + "step": 27932 + }, + { + "epoch": 6.1, + "learning_rate": 1.9540091763163648e-06, + "loss": 0.4398, + "step": 27933 + }, + { + "epoch": 6.1, + "learning_rate": 1.9538999344548834e-06, + "loss": 1.8927, + "step": 27934 + }, + { + "epoch": 6.1, + "learning_rate": 1.953790692593402e-06, + "loss": 0.9317, + "step": 27935 + }, + { + "epoch": 6.1, + "learning_rate": 1.9536814507319206e-06, + "loss": 1.7022, + "step": 27936 + }, + { + "epoch": 6.1, + "learning_rate": 1.9535722088704392e-06, + "loss": 1.6577, + "step": 27937 + }, + { + "epoch": 6.1, + "learning_rate": 1.953462967008958e-06, + "loss": 2.1611, + "step": 27938 + }, + { + "epoch": 6.1, + "learning_rate": 1.9533537251474764e-06, + "loss": 1.6227, + "step": 27939 + }, + { + "epoch": 6.1, + "learning_rate": 1.9532444832859955e-06, + "loss": 0.8618, + "step": 27940 + }, + { + "epoch": 6.1, + "learning_rate": 1.953135241424514e-06, + "loss": 1.4916, + "step": 27941 + }, + { + "epoch": 6.1, + "learning_rate": 1.9530259995630327e-06, + "loss": 1.1387, + "step": 27942 + }, + { + "epoch": 6.1, + "learning_rate": 1.9529167577015513e-06, + "loss": 0.444, + "step": 27943 + }, + { + "epoch": 6.1, + "learning_rate": 1.95280751584007e-06, + "loss": 1.1914, + "step": 27944 + }, + { + "epoch": 6.1, + "learning_rate": 1.952698273978589e-06, + "loss": 2.2317, + "step": 27945 + }, + { + "epoch": 6.1, + "learning_rate": 1.952589032117107e-06, + "loss": 0.6578, + "step": 27946 + }, + { + "epoch": 6.1, + "learning_rate": 1.952479790255626e-06, + "loss": 0.5375, + "step": 27947 + }, + { + "epoch": 6.1, + "learning_rate": 1.952370548394145e-06, + "loss": 2.1715, + "step": 27948 + }, + { + "epoch": 6.1, + "learning_rate": 1.9522613065326634e-06, + "loss": 1.6063, + "step": 27949 + }, + { + "epoch": 6.1, + "learning_rate": 1.9521520646711824e-06, + "loss": 1.6623, + "step": 27950 + }, + { + "epoch": 6.1, + "learning_rate": 1.9520428228097006e-06, + "loss": 2.0014, + "step": 27951 + }, + { + "epoch": 6.1, + "learning_rate": 1.9519335809482197e-06, + "loss": 1.3516, + "step": 27952 + }, + { + "epoch": 6.1, + "learning_rate": 1.9518243390867383e-06, + "loss": 1.4706, + "step": 27953 + }, + { + "epoch": 6.1, + "learning_rate": 1.951715097225257e-06, + "loss": 2.2227, + "step": 27954 + }, + { + "epoch": 6.1, + "learning_rate": 1.9516058553637755e-06, + "loss": 0.9507, + "step": 27955 + }, + { + "epoch": 6.1, + "learning_rate": 1.951496613502294e-06, + "loss": 1.0225, + "step": 27956 + }, + { + "epoch": 6.1, + "learning_rate": 1.951387371640813e-06, + "loss": 1.7366, + "step": 27957 + }, + { + "epoch": 6.1, + "learning_rate": 1.9512781297793313e-06, + "loss": 0.7926, + "step": 27958 + }, + { + "epoch": 6.1, + "learning_rate": 1.9511688879178504e-06, + "loss": 1.2373, + "step": 27959 + }, + { + "epoch": 6.1, + "learning_rate": 1.951059646056369e-06, + "loss": 1.3212, + "step": 27960 + }, + { + "epoch": 6.1, + "learning_rate": 1.9509504041948876e-06, + "loss": 0.6086, + "step": 27961 + }, + { + "epoch": 6.1, + "learning_rate": 1.9508411623334066e-06, + "loss": 0.6291, + "step": 27962 + }, + { + "epoch": 6.1, + "learning_rate": 1.950731920471925e-06, + "loss": 1.322, + "step": 27963 + }, + { + "epoch": 6.1, + "learning_rate": 1.950622678610444e-06, + "loss": 1.0591, + "step": 27964 + }, + { + "epoch": 6.1, + "learning_rate": 1.9505134367489625e-06, + "loss": 0.8773, + "step": 27965 + }, + { + "epoch": 6.1, + "learning_rate": 1.950404194887481e-06, + "loss": 1.694, + "step": 27966 + }, + { + "epoch": 6.1, + "learning_rate": 1.9502949530259997e-06, + "loss": 2.1003, + "step": 27967 + }, + { + "epoch": 6.1, + "learning_rate": 1.9501857111645183e-06, + "loss": 1.8049, + "step": 27968 + }, + { + "epoch": 6.1, + "learning_rate": 1.9500764693030373e-06, + "loss": 2.0877, + "step": 27969 + }, + { + "epoch": 6.1, + "learning_rate": 1.9499672274415555e-06, + "loss": 2.3614, + "step": 27970 + }, + { + "epoch": 6.1, + "learning_rate": 1.9498579855800745e-06, + "loss": 1.2344, + "step": 27971 + }, + { + "epoch": 6.1, + "learning_rate": 1.949748743718593e-06, + "loss": 0.4025, + "step": 27972 + }, + { + "epoch": 6.1, + "learning_rate": 1.9496395018571118e-06, + "loss": 1.0297, + "step": 27973 + }, + { + "epoch": 6.11, + "learning_rate": 1.949530259995631e-06, + "loss": 1.3778, + "step": 27974 + }, + { + "epoch": 6.11, + "learning_rate": 1.949421018134149e-06, + "loss": 2.6806, + "step": 27975 + }, + { + "epoch": 6.11, + "learning_rate": 1.949311776272668e-06, + "loss": 0.9768, + "step": 27976 + }, + { + "epoch": 6.11, + "learning_rate": 1.9492025344111866e-06, + "loss": 1.0721, + "step": 27977 + }, + { + "epoch": 6.11, + "learning_rate": 1.9490932925497052e-06, + "loss": 1.6316, + "step": 27978 + }, + { + "epoch": 6.11, + "learning_rate": 1.948984050688224e-06, + "loss": 1.5683, + "step": 27979 + }, + { + "epoch": 6.11, + "learning_rate": 1.9488748088267425e-06, + "loss": 1.2013, + "step": 27980 + }, + { + "epoch": 6.11, + "learning_rate": 1.9487655669652615e-06, + "loss": 0.2167, + "step": 27981 + }, + { + "epoch": 6.11, + "learning_rate": 1.9486563251037797e-06, + "loss": 0.5443, + "step": 27982 + }, + { + "epoch": 6.11, + "learning_rate": 1.9485470832422987e-06, + "loss": 0.8594, + "step": 27983 + }, + { + "epoch": 6.11, + "learning_rate": 1.9484378413808173e-06, + "loss": 2.0862, + "step": 27984 + }, + { + "epoch": 6.11, + "learning_rate": 1.948328599519336e-06, + "loss": 1.8828, + "step": 27985 + }, + { + "epoch": 6.11, + "learning_rate": 1.9482193576578546e-06, + "loss": 0.1607, + "step": 27986 + }, + { + "epoch": 6.11, + "learning_rate": 1.948110115796373e-06, + "loss": 0.8936, + "step": 27987 + }, + { + "epoch": 6.11, + "learning_rate": 1.948000873934892e-06, + "loss": 0.1578, + "step": 27988 + }, + { + "epoch": 6.11, + "learning_rate": 1.947891632073411e-06, + "loss": 1.8526, + "step": 27989 + }, + { + "epoch": 6.11, + "learning_rate": 1.9477823902119294e-06, + "loss": 1.56, + "step": 27990 + }, + { + "epoch": 6.11, + "learning_rate": 1.947673148350448e-06, + "loss": 0.7464, + "step": 27991 + }, + { + "epoch": 6.11, + "learning_rate": 1.9475639064889666e-06, + "loss": 0.2822, + "step": 27992 + }, + { + "epoch": 6.11, + "learning_rate": 1.9474546646274853e-06, + "loss": 0.4197, + "step": 27993 + }, + { + "epoch": 6.11, + "learning_rate": 1.9473454227660043e-06, + "loss": 1.169, + "step": 27994 + }, + { + "epoch": 6.11, + "learning_rate": 1.947236180904523e-06, + "loss": 0.8337, + "step": 27995 + }, + { + "epoch": 6.11, + "learning_rate": 1.9471269390430415e-06, + "loss": 2.9809, + "step": 27996 + }, + { + "epoch": 6.11, + "learning_rate": 1.94701769718156e-06, + "loss": 0.9151, + "step": 27997 + }, + { + "epoch": 6.11, + "learning_rate": 1.9469084553200787e-06, + "loss": 2.0573, + "step": 27998 + }, + { + "epoch": 6.11, + "learning_rate": 1.9467992134585973e-06, + "loss": 0.2655, + "step": 27999 + }, + { + "epoch": 6.11, + "learning_rate": 1.9466899715971164e-06, + "loss": 1.5032, + "step": 28000 + }, + { + "epoch": 6.11, + "learning_rate": 1.946580729735635e-06, + "loss": 2.108, + "step": 28001 + }, + { + "epoch": 6.11, + "learning_rate": 1.9464714878741536e-06, + "loss": 0.7873, + "step": 28002 + }, + { + "epoch": 6.11, + "learning_rate": 1.946362246012672e-06, + "loss": 0.2072, + "step": 28003 + }, + { + "epoch": 6.11, + "learning_rate": 1.946253004151191e-06, + "loss": 1.3374, + "step": 28004 + }, + { + "epoch": 6.11, + "learning_rate": 1.9461437622897094e-06, + "loss": 0.4723, + "step": 28005 + }, + { + "epoch": 6.11, + "learning_rate": 1.9460345204282285e-06, + "loss": 0.3892, + "step": 28006 + }, + { + "epoch": 6.11, + "learning_rate": 1.945925278566747e-06, + "loss": 1.7044, + "step": 28007 + }, + { + "epoch": 6.11, + "learning_rate": 1.9458160367052657e-06, + "loss": 1.3211, + "step": 28008 + }, + { + "epoch": 6.11, + "learning_rate": 1.9457067948437843e-06, + "loss": 0.979, + "step": 28009 + }, + { + "epoch": 6.11, + "learning_rate": 1.945597552982303e-06, + "loss": 1.2257, + "step": 28010 + }, + { + "epoch": 6.11, + "learning_rate": 1.9454883111208215e-06, + "loss": 1.1453, + "step": 28011 + }, + { + "epoch": 6.11, + "learning_rate": 1.94537906925934e-06, + "loss": 0.9467, + "step": 28012 + }, + { + "epoch": 6.11, + "learning_rate": 1.945269827397859e-06, + "loss": 3.703, + "step": 28013 + }, + { + "epoch": 6.11, + "learning_rate": 1.9451605855363778e-06, + "loss": 0.8309, + "step": 28014 + }, + { + "epoch": 6.11, + "learning_rate": 1.9450513436748964e-06, + "loss": 1.737, + "step": 28015 + }, + { + "epoch": 6.11, + "learning_rate": 1.944942101813415e-06, + "loss": 1.6944, + "step": 28016 + }, + { + "epoch": 6.11, + "learning_rate": 1.9448328599519336e-06, + "loss": 3.3178, + "step": 28017 + }, + { + "epoch": 6.11, + "learning_rate": 1.9447236180904526e-06, + "loss": 2.2204, + "step": 28018 + }, + { + "epoch": 6.12, + "learning_rate": 1.944614376228971e-06, + "loss": 0.7456, + "step": 28019 + }, + { + "epoch": 6.12, + "learning_rate": 1.94450513436749e-06, + "loss": 0.8148, + "step": 28020 + }, + { + "epoch": 6.12, + "learning_rate": 1.9443958925060085e-06, + "loss": 1.5484, + "step": 28021 + }, + { + "epoch": 6.12, + "learning_rate": 1.944286650644527e-06, + "loss": 1.5687, + "step": 28022 + }, + { + "epoch": 6.12, + "learning_rate": 1.9441774087830457e-06, + "loss": 0.2832, + "step": 28023 + }, + { + "epoch": 6.12, + "learning_rate": 1.9440681669215643e-06, + "loss": 1.2393, + "step": 28024 + }, + { + "epoch": 6.12, + "learning_rate": 1.9439589250600833e-06, + "loss": 1.8841, + "step": 28025 + }, + { + "epoch": 6.12, + "learning_rate": 1.9438496831986015e-06, + "loss": 0.2649, + "step": 28026 + }, + { + "epoch": 6.12, + "learning_rate": 1.9437404413371206e-06, + "loss": 1.3487, + "step": 28027 + }, + { + "epoch": 6.12, + "learning_rate": 1.943631199475639e-06, + "loss": 2.3069, + "step": 28028 + }, + { + "epoch": 6.12, + "learning_rate": 1.943521957614158e-06, + "loss": 0.4209, + "step": 28029 + }, + { + "epoch": 6.12, + "learning_rate": 1.943412715752677e-06, + "loss": 0.6048, + "step": 28030 + }, + { + "epoch": 6.12, + "learning_rate": 1.943303473891195e-06, + "loss": 1.0475, + "step": 28031 + }, + { + "epoch": 6.12, + "learning_rate": 1.943194232029714e-06, + "loss": 0.8035, + "step": 28032 + }, + { + "epoch": 6.12, + "learning_rate": 1.9430849901682327e-06, + "loss": 0.701, + "step": 28033 + }, + { + "epoch": 6.12, + "learning_rate": 1.9429757483067513e-06, + "loss": 0.9166, + "step": 28034 + }, + { + "epoch": 6.12, + "learning_rate": 1.94286650644527e-06, + "loss": 1.8133, + "step": 28035 + }, + { + "epoch": 6.12, + "learning_rate": 1.9427572645837885e-06, + "loss": 0.3865, + "step": 28036 + }, + { + "epoch": 6.12, + "learning_rate": 1.9426480227223075e-06, + "loss": 0.6764, + "step": 28037 + }, + { + "epoch": 6.12, + "learning_rate": 1.942538780860826e-06, + "loss": 0.9943, + "step": 28038 + }, + { + "epoch": 6.12, + "learning_rate": 1.9424295389993447e-06, + "loss": 1.8424, + "step": 28039 + }, + { + "epoch": 6.12, + "learning_rate": 1.9423202971378634e-06, + "loss": 1.2255, + "step": 28040 + }, + { + "epoch": 6.12, + "learning_rate": 1.942211055276382e-06, + "loss": 1.9565, + "step": 28041 + }, + { + "epoch": 6.12, + "learning_rate": 1.942101813414901e-06, + "loss": 0.3039, + "step": 28042 + }, + { + "epoch": 6.12, + "learning_rate": 1.941992571553419e-06, + "loss": 1.38, + "step": 28043 + }, + { + "epoch": 6.12, + "learning_rate": 1.9418833296919382e-06, + "loss": 0.2717, + "step": 28044 + }, + { + "epoch": 6.12, + "learning_rate": 1.941774087830457e-06, + "loss": 1.7955, + "step": 28045 + }, + { + "epoch": 6.12, + "learning_rate": 1.9416648459689754e-06, + "loss": 0.6011, + "step": 28046 + }, + { + "epoch": 6.12, + "learning_rate": 1.9415556041074945e-06, + "loss": 1.5066, + "step": 28047 + }, + { + "epoch": 6.12, + "learning_rate": 1.9414463622460127e-06, + "loss": 1.3055, + "step": 28048 + }, + { + "epoch": 6.12, + "learning_rate": 1.9413371203845317e-06, + "loss": 2.2648, + "step": 28049 + }, + { + "epoch": 6.12, + "learning_rate": 1.9412278785230503e-06, + "loss": 1.4092, + "step": 28050 + }, + { + "epoch": 6.12, + "learning_rate": 1.941118636661569e-06, + "loss": 0.7188, + "step": 28051 + }, + { + "epoch": 6.12, + "learning_rate": 1.9410093948000875e-06, + "loss": 0.4428, + "step": 28052 + }, + { + "epoch": 6.12, + "learning_rate": 1.940900152938606e-06, + "loss": 1.9023, + "step": 28053 + }, + { + "epoch": 6.12, + "learning_rate": 1.940790911077125e-06, + "loss": 0.1262, + "step": 28054 + }, + { + "epoch": 6.12, + "learning_rate": 1.9406816692156434e-06, + "loss": 1.2112, + "step": 28055 + }, + { + "epoch": 6.12, + "learning_rate": 1.9405724273541624e-06, + "loss": 1.5995, + "step": 28056 + }, + { + "epoch": 6.12, + "learning_rate": 1.940463185492681e-06, + "loss": 0.9995, + "step": 28057 + }, + { + "epoch": 6.12, + "learning_rate": 1.9403539436311996e-06, + "loss": 0.1781, + "step": 28058 + }, + { + "epoch": 6.12, + "learning_rate": 1.9402447017697182e-06, + "loss": 0.615, + "step": 28059 + }, + { + "epoch": 6.12, + "learning_rate": 1.940135459908237e-06, + "loss": 2.5219, + "step": 28060 + }, + { + "epoch": 6.12, + "learning_rate": 1.940026218046756e-06, + "loss": 1.0016, + "step": 28061 + }, + { + "epoch": 6.12, + "learning_rate": 1.9399169761852745e-06, + "loss": 3.0585, + "step": 28062 + }, + { + "epoch": 6.12, + "learning_rate": 1.939807734323793e-06, + "loss": 0.5566, + "step": 28063 + }, + { + "epoch": 6.12, + "learning_rate": 1.9396984924623117e-06, + "loss": 0.6804, + "step": 28064 + }, + { + "epoch": 6.13, + "learning_rate": 1.9395892506008303e-06, + "loss": 0.89, + "step": 28065 + }, + { + "epoch": 6.13, + "learning_rate": 1.939480008739349e-06, + "loss": 1.4151, + "step": 28066 + }, + { + "epoch": 6.13, + "learning_rate": 1.9393707668778675e-06, + "loss": 1.1448, + "step": 28067 + }, + { + "epoch": 6.13, + "learning_rate": 1.9392615250163866e-06, + "loss": 1.6465, + "step": 28068 + }, + { + "epoch": 6.13, + "learning_rate": 1.939152283154905e-06, + "loss": 0.9004, + "step": 28069 + }, + { + "epoch": 6.13, + "learning_rate": 1.939043041293424e-06, + "loss": 0.2496, + "step": 28070 + }, + { + "epoch": 6.13, + "learning_rate": 1.9389337994319424e-06, + "loss": 2.133, + "step": 28071 + }, + { + "epoch": 6.13, + "learning_rate": 1.938824557570461e-06, + "loss": 1.0956, + "step": 28072 + }, + { + "epoch": 6.13, + "learning_rate": 1.93871531570898e-06, + "loss": 1.4733, + "step": 28073 + }, + { + "epoch": 6.13, + "learning_rate": 1.9386060738474987e-06, + "loss": 0.8119, + "step": 28074 + }, + { + "epoch": 6.13, + "learning_rate": 1.9384968319860173e-06, + "loss": 2.955, + "step": 28075 + }, + { + "epoch": 6.13, + "learning_rate": 1.938387590124536e-06, + "loss": 1.3415, + "step": 28076 + }, + { + "epoch": 6.13, + "learning_rate": 1.9382783482630545e-06, + "loss": 1.0982, + "step": 28077 + }, + { + "epoch": 6.13, + "learning_rate": 1.938169106401573e-06, + "loss": 1.1048, + "step": 28078 + }, + { + "epoch": 6.13, + "learning_rate": 1.9380598645400917e-06, + "loss": 1.0432, + "step": 28079 + }, + { + "epoch": 6.13, + "learning_rate": 1.9379506226786108e-06, + "loss": 2.3225, + "step": 28080 + }, + { + "epoch": 6.13, + "learning_rate": 1.9378413808171294e-06, + "loss": 1.6587, + "step": 28081 + }, + { + "epoch": 6.13, + "learning_rate": 1.937732138955648e-06, + "loss": 0.6695, + "step": 28082 + }, + { + "epoch": 6.13, + "learning_rate": 1.9376228970941666e-06, + "loss": 0.9926, + "step": 28083 + }, + { + "epoch": 6.13, + "learning_rate": 1.937513655232685e-06, + "loss": 1.2834, + "step": 28084 + }, + { + "epoch": 6.13, + "learning_rate": 1.937404413371204e-06, + "loss": 1.071, + "step": 28085 + }, + { + "epoch": 6.13, + "learning_rate": 1.937295171509723e-06, + "loss": 0.5756, + "step": 28086 + }, + { + "epoch": 6.13, + "learning_rate": 1.9371859296482415e-06, + "loss": 0.5791, + "step": 28087 + }, + { + "epoch": 6.13, + "learning_rate": 1.93707668778676e-06, + "loss": 1.9929, + "step": 28088 + }, + { + "epoch": 6.13, + "learning_rate": 1.9369674459252787e-06, + "loss": 2.9223, + "step": 28089 + }, + { + "epoch": 6.13, + "learning_rate": 1.9368582040637973e-06, + "loss": 1.6354, + "step": 28090 + }, + { + "epoch": 6.13, + "learning_rate": 1.9367489622023163e-06, + "loss": 1.0788, + "step": 28091 + }, + { + "epoch": 6.13, + "learning_rate": 1.9366397203408345e-06, + "loss": 3.0412, + "step": 28092 + }, + { + "epoch": 6.13, + "learning_rate": 1.9365304784793536e-06, + "loss": 1.5226, + "step": 28093 + }, + { + "epoch": 6.13, + "learning_rate": 1.936421236617872e-06, + "loss": 0.3569, + "step": 28094 + }, + { + "epoch": 6.13, + "learning_rate": 1.9363119947563908e-06, + "loss": 0.5131, + "step": 28095 + }, + { + "epoch": 6.13, + "learning_rate": 1.9362027528949094e-06, + "loss": 1.2383, + "step": 28096 + }, + { + "epoch": 6.13, + "learning_rate": 1.936093511033428e-06, + "loss": 2.9908, + "step": 28097 + }, + { + "epoch": 6.13, + "learning_rate": 1.935984269171947e-06, + "loss": 0.9883, + "step": 28098 + }, + { + "epoch": 6.13, + "learning_rate": 1.9358750273104652e-06, + "loss": 1.6688, + "step": 28099 + }, + { + "epoch": 6.13, + "learning_rate": 1.9357657854489843e-06, + "loss": 1.0122, + "step": 28100 + }, + { + "epoch": 6.13, + "learning_rate": 1.935656543587503e-06, + "loss": 1.642, + "step": 28101 + }, + { + "epoch": 6.13, + "learning_rate": 1.9355473017260215e-06, + "loss": 1.5167, + "step": 28102 + }, + { + "epoch": 6.13, + "learning_rate": 1.9354380598645405e-06, + "loss": 1.1658, + "step": 28103 + }, + { + "epoch": 6.13, + "learning_rate": 1.9353288180030587e-06, + "loss": 1.6929, + "step": 28104 + }, + { + "epoch": 6.13, + "learning_rate": 1.9352195761415777e-06, + "loss": 0.6055, + "step": 28105 + }, + { + "epoch": 6.13, + "learning_rate": 1.9351103342800963e-06, + "loss": 0.1643, + "step": 28106 + }, + { + "epoch": 6.13, + "learning_rate": 1.935001092418615e-06, + "loss": 0.1873, + "step": 28107 + }, + { + "epoch": 6.13, + "learning_rate": 1.9348918505571336e-06, + "loss": 2.9532, + "step": 28108 + }, + { + "epoch": 6.13, + "learning_rate": 1.934782608695652e-06, + "loss": 1.8924, + "step": 28109 + }, + { + "epoch": 6.13, + "learning_rate": 1.934673366834171e-06, + "loss": 0.343, + "step": 28110 + }, + { + "epoch": 6.14, + "learning_rate": 1.9345641249726894e-06, + "loss": 2.2633, + "step": 28111 + }, + { + "epoch": 6.14, + "learning_rate": 1.9344548831112084e-06, + "loss": 1.452, + "step": 28112 + }, + { + "epoch": 6.14, + "learning_rate": 1.934345641249727e-06, + "loss": 1.0404, + "step": 28113 + }, + { + "epoch": 6.14, + "learning_rate": 1.9342363993882457e-06, + "loss": 1.0575, + "step": 28114 + }, + { + "epoch": 6.14, + "learning_rate": 1.9341271575267647e-06, + "loss": 1.0153, + "step": 28115 + }, + { + "epoch": 6.14, + "learning_rate": 1.934017915665283e-06, + "loss": 0.2635, + "step": 28116 + }, + { + "epoch": 6.14, + "learning_rate": 1.933908673803802e-06, + "loss": 1.3969, + "step": 28117 + }, + { + "epoch": 6.14, + "learning_rate": 1.9337994319423205e-06, + "loss": 0.8664, + "step": 28118 + }, + { + "epoch": 6.14, + "learning_rate": 1.933690190080839e-06, + "loss": 1.7604, + "step": 28119 + }, + { + "epoch": 6.14, + "learning_rate": 1.9335809482193577e-06, + "loss": 2.585, + "step": 28120 + }, + { + "epoch": 6.14, + "learning_rate": 1.9334717063578764e-06, + "loss": 0.4164, + "step": 28121 + }, + { + "epoch": 6.14, + "learning_rate": 1.9333624644963954e-06, + "loss": 1.3398, + "step": 28122 + }, + { + "epoch": 6.14, + "learning_rate": 1.9332532226349136e-06, + "loss": 1.1999, + "step": 28123 + }, + { + "epoch": 6.14, + "learning_rate": 1.9331439807734326e-06, + "loss": 0.798, + "step": 28124 + }, + { + "epoch": 6.14, + "learning_rate": 1.9330347389119512e-06, + "loss": 0.8822, + "step": 28125 + }, + { + "epoch": 6.14, + "learning_rate": 1.93292549705047e-06, + "loss": 2.3868, + "step": 28126 + }, + { + "epoch": 6.14, + "learning_rate": 1.932816255188989e-06, + "loss": 1.9217, + "step": 28127 + }, + { + "epoch": 6.14, + "learning_rate": 1.932707013327507e-06, + "loss": 0.7943, + "step": 28128 + }, + { + "epoch": 6.14, + "learning_rate": 1.932597771466026e-06, + "loss": 2.1522, + "step": 28129 + }, + { + "epoch": 6.14, + "learning_rate": 1.9324885296045447e-06, + "loss": 0.162, + "step": 28130 + }, + { + "epoch": 6.14, + "learning_rate": 1.9323792877430633e-06, + "loss": 1.3007, + "step": 28131 + }, + { + "epoch": 6.14, + "learning_rate": 1.932270045881582e-06, + "loss": 4.3981, + "step": 28132 + }, + { + "epoch": 6.14, + "learning_rate": 1.9321608040201005e-06, + "loss": 1.589, + "step": 28133 + }, + { + "epoch": 6.14, + "learning_rate": 1.9320515621586196e-06, + "loss": 2.2295, + "step": 28134 + }, + { + "epoch": 6.14, + "learning_rate": 1.9319423202971378e-06, + "loss": 1.8422, + "step": 28135 + }, + { + "epoch": 6.14, + "learning_rate": 1.9318330784356568e-06, + "loss": 0.5323, + "step": 28136 + }, + { + "epoch": 6.14, + "learning_rate": 1.9317238365741754e-06, + "loss": 0.31, + "step": 28137 + }, + { + "epoch": 6.14, + "learning_rate": 1.931614594712694e-06, + "loss": 1.1162, + "step": 28138 + }, + { + "epoch": 6.14, + "learning_rate": 1.931505352851213e-06, + "loss": 1.2714, + "step": 28139 + }, + { + "epoch": 6.14, + "learning_rate": 1.9313961109897312e-06, + "loss": 1.3506, + "step": 28140 + }, + { + "epoch": 6.14, + "learning_rate": 1.9312868691282503e-06, + "loss": 1.7174, + "step": 28141 + }, + { + "epoch": 6.14, + "learning_rate": 1.931177627266769e-06, + "loss": 2.1939, + "step": 28142 + }, + { + "epoch": 6.14, + "learning_rate": 1.9310683854052875e-06, + "loss": 0.7648, + "step": 28143 + }, + { + "epoch": 6.14, + "learning_rate": 1.930959143543806e-06, + "loss": 1.2431, + "step": 28144 + }, + { + "epoch": 6.14, + "learning_rate": 1.9308499016823247e-06, + "loss": 1.467, + "step": 28145 + }, + { + "epoch": 6.14, + "learning_rate": 1.9307406598208437e-06, + "loss": 1.0421, + "step": 28146 + }, + { + "epoch": 6.14, + "learning_rate": 1.9306314179593624e-06, + "loss": 1.7462, + "step": 28147 + }, + { + "epoch": 6.14, + "learning_rate": 1.930522176097881e-06, + "loss": 1.8431, + "step": 28148 + }, + { + "epoch": 6.14, + "learning_rate": 1.9304129342363996e-06, + "loss": 1.5619, + "step": 28149 + }, + { + "epoch": 6.14, + "learning_rate": 1.930303692374918e-06, + "loss": 0.4995, + "step": 28150 + }, + { + "epoch": 6.14, + "learning_rate": 1.930194450513437e-06, + "loss": 0.0899, + "step": 28151 + }, + { + "epoch": 6.14, + "learning_rate": 1.9300852086519554e-06, + "loss": 0.5692, + "step": 28152 + }, + { + "epoch": 6.14, + "learning_rate": 1.9299759667904744e-06, + "loss": 1.4412, + "step": 28153 + }, + { + "epoch": 6.14, + "learning_rate": 1.929866724928993e-06, + "loss": 2.6758, + "step": 28154 + }, + { + "epoch": 6.14, + "learning_rate": 1.9297574830675117e-06, + "loss": 1.1125, + "step": 28155 + }, + { + "epoch": 6.14, + "learning_rate": 1.9296482412060303e-06, + "loss": 0.7234, + "step": 28156 + }, + { + "epoch": 6.15, + "learning_rate": 1.929538999344549e-06, + "loss": 3.3647, + "step": 28157 + }, + { + "epoch": 6.15, + "learning_rate": 1.9294297574830675e-06, + "loss": 1.1668, + "step": 28158 + }, + { + "epoch": 6.15, + "learning_rate": 1.9293205156215865e-06, + "loss": 1.3209, + "step": 28159 + }, + { + "epoch": 6.15, + "learning_rate": 1.929211273760105e-06, + "loss": 1.4054, + "step": 28160 + }, + { + "epoch": 6.15, + "learning_rate": 1.9291020318986238e-06, + "loss": 1.1932, + "step": 28161 + }, + { + "epoch": 6.15, + "learning_rate": 1.9289927900371424e-06, + "loss": 0.324, + "step": 28162 + }, + { + "epoch": 6.15, + "learning_rate": 1.928883548175661e-06, + "loss": 0.8094, + "step": 28163 + }, + { + "epoch": 6.15, + "learning_rate": 1.9287743063141796e-06, + "loss": 1.7704, + "step": 28164 + }, + { + "epoch": 6.15, + "learning_rate": 1.928665064452698e-06, + "loss": 0.6176, + "step": 28165 + }, + { + "epoch": 6.15, + "learning_rate": 1.9285558225912172e-06, + "loss": 1.2233, + "step": 28166 + }, + { + "epoch": 6.15, + "learning_rate": 1.928446580729736e-06, + "loss": 1.6543, + "step": 28167 + }, + { + "epoch": 6.15, + "learning_rate": 1.9283373388682545e-06, + "loss": 2.3391, + "step": 28168 + }, + { + "epoch": 6.15, + "learning_rate": 1.928228097006773e-06, + "loss": 1.6139, + "step": 28169 + }, + { + "epoch": 6.15, + "learning_rate": 1.9281188551452917e-06, + "loss": 2.6373, + "step": 28170 + }, + { + "epoch": 6.15, + "learning_rate": 1.9280096132838107e-06, + "loss": 0.316, + "step": 28171 + }, + { + "epoch": 6.15, + "learning_rate": 1.9279003714223293e-06, + "loss": 0.6799, + "step": 28172 + }, + { + "epoch": 6.15, + "learning_rate": 1.927791129560848e-06, + "loss": 1.0221, + "step": 28173 + }, + { + "epoch": 6.15, + "learning_rate": 1.9276818876993665e-06, + "loss": 3.0752, + "step": 28174 + }, + { + "epoch": 6.15, + "learning_rate": 1.927572645837885e-06, + "loss": 0.8991, + "step": 28175 + }, + { + "epoch": 6.15, + "learning_rate": 1.9274634039764038e-06, + "loss": 1.8382, + "step": 28176 + }, + { + "epoch": 6.15, + "learning_rate": 1.9273541621149224e-06, + "loss": 1.4244, + "step": 28177 + }, + { + "epoch": 6.15, + "learning_rate": 1.9272449202534414e-06, + "loss": 2.0908, + "step": 28178 + }, + { + "epoch": 6.15, + "learning_rate": 1.92713567839196e-06, + "loss": 2.0448, + "step": 28179 + }, + { + "epoch": 6.15, + "learning_rate": 1.9270264365304786e-06, + "loss": 1.3776, + "step": 28180 + }, + { + "epoch": 6.15, + "learning_rate": 1.9269171946689972e-06, + "loss": 1.0406, + "step": 28181 + }, + { + "epoch": 6.15, + "learning_rate": 1.926807952807516e-06, + "loss": 1.1912, + "step": 28182 + }, + { + "epoch": 6.15, + "learning_rate": 1.926698710946035e-06, + "loss": 0.9808, + "step": 28183 + }, + { + "epoch": 6.15, + "learning_rate": 1.926589469084553e-06, + "loss": 0.8624, + "step": 28184 + }, + { + "epoch": 6.15, + "learning_rate": 1.926480227223072e-06, + "loss": 2.098, + "step": 28185 + }, + { + "epoch": 6.15, + "learning_rate": 1.9263709853615907e-06, + "loss": 2.5902, + "step": 28186 + }, + { + "epoch": 6.15, + "learning_rate": 1.9262617435001093e-06, + "loss": 2.4301, + "step": 28187 + }, + { + "epoch": 6.15, + "learning_rate": 1.926152501638628e-06, + "loss": 0.6152, + "step": 28188 + }, + { + "epoch": 6.15, + "learning_rate": 1.9260432597771466e-06, + "loss": 0.7819, + "step": 28189 + }, + { + "epoch": 6.15, + "learning_rate": 1.9259340179156656e-06, + "loss": 1.0456, + "step": 28190 + }, + { + "epoch": 6.15, + "learning_rate": 1.925824776054184e-06, + "loss": 0.9323, + "step": 28191 + }, + { + "epoch": 6.15, + "learning_rate": 1.925715534192703e-06, + "loss": 3.1997, + "step": 28192 + }, + { + "epoch": 6.15, + "learning_rate": 1.9256062923312214e-06, + "loss": 0.7856, + "step": 28193 + }, + { + "epoch": 6.15, + "learning_rate": 1.92549705046974e-06, + "loss": 1.2451, + "step": 28194 + }, + { + "epoch": 6.15, + "learning_rate": 1.925387808608259e-06, + "loss": 0.9632, + "step": 28195 + }, + { + "epoch": 6.15, + "learning_rate": 1.9252785667467773e-06, + "loss": 1.465, + "step": 28196 + }, + { + "epoch": 6.15, + "learning_rate": 1.9251693248852963e-06, + "loss": 0.8801, + "step": 28197 + }, + { + "epoch": 6.15, + "learning_rate": 1.925060083023815e-06, + "loss": 0.6964, + "step": 28198 + }, + { + "epoch": 6.15, + "learning_rate": 1.9249508411623335e-06, + "loss": 0.276, + "step": 28199 + }, + { + "epoch": 6.15, + "learning_rate": 1.9248415993008525e-06, + "loss": 0.8772, + "step": 28200 + }, + { + "epoch": 6.15, + "learning_rate": 1.9247323574393707e-06, + "loss": 0.5454, + "step": 28201 + }, + { + "epoch": 6.15, + "learning_rate": 1.9246231155778898e-06, + "loss": 2.4522, + "step": 28202 + }, + { + "epoch": 6.16, + "learning_rate": 1.9245138737164084e-06, + "loss": 0.6365, + "step": 28203 + }, + { + "epoch": 6.16, + "learning_rate": 1.924404631854927e-06, + "loss": 0.5729, + "step": 28204 + }, + { + "epoch": 6.16, + "learning_rate": 1.9242953899934456e-06, + "loss": 3.2796, + "step": 28205 + }, + { + "epoch": 6.16, + "learning_rate": 1.9241861481319642e-06, + "loss": 0.8862, + "step": 28206 + }, + { + "epoch": 6.16, + "learning_rate": 1.9240769062704833e-06, + "loss": 1.1167, + "step": 28207 + }, + { + "epoch": 6.16, + "learning_rate": 1.9239676644090014e-06, + "loss": 2.16, + "step": 28208 + }, + { + "epoch": 6.16, + "learning_rate": 1.9238584225475205e-06, + "loss": 2.1634, + "step": 28209 + }, + { + "epoch": 6.16, + "learning_rate": 1.923749180686039e-06, + "loss": 0.8039, + "step": 28210 + }, + { + "epoch": 6.16, + "learning_rate": 1.9236399388245577e-06, + "loss": 1.1767, + "step": 28211 + }, + { + "epoch": 6.16, + "learning_rate": 1.9235306969630767e-06, + "loss": 1.1469, + "step": 28212 + }, + { + "epoch": 6.16, + "learning_rate": 1.923421455101595e-06, + "loss": 0.1512, + "step": 28213 + }, + { + "epoch": 6.16, + "learning_rate": 1.923312213240114e-06, + "loss": 1.8669, + "step": 28214 + }, + { + "epoch": 6.16, + "learning_rate": 1.9232029713786326e-06, + "loss": 0.7554, + "step": 28215 + }, + { + "epoch": 6.16, + "learning_rate": 1.923093729517151e-06, + "loss": 2.3638, + "step": 28216 + }, + { + "epoch": 6.16, + "learning_rate": 1.9229844876556698e-06, + "loss": 2.3813, + "step": 28217 + }, + { + "epoch": 6.16, + "learning_rate": 1.9228752457941884e-06, + "loss": 1.4167, + "step": 28218 + }, + { + "epoch": 6.16, + "learning_rate": 1.9227660039327074e-06, + "loss": 2.5764, + "step": 28219 + }, + { + "epoch": 6.16, + "learning_rate": 1.9226567620712256e-06, + "loss": 1.981, + "step": 28220 + }, + { + "epoch": 6.16, + "learning_rate": 1.9225475202097447e-06, + "loss": 1.2646, + "step": 28221 + }, + { + "epoch": 6.16, + "learning_rate": 1.9224382783482633e-06, + "loss": 1.0698, + "step": 28222 + }, + { + "epoch": 6.16, + "learning_rate": 1.922329036486782e-06, + "loss": 0.7931, + "step": 28223 + }, + { + "epoch": 6.16, + "learning_rate": 1.9222197946253005e-06, + "loss": 1.2915, + "step": 28224 + }, + { + "epoch": 6.16, + "learning_rate": 1.922110552763819e-06, + "loss": 0.6668, + "step": 28225 + }, + { + "epoch": 6.16, + "learning_rate": 1.922001310902338e-06, + "loss": 1.6075, + "step": 28226 + }, + { + "epoch": 6.16, + "learning_rate": 1.9218920690408567e-06, + "loss": 1.7302, + "step": 28227 + }, + { + "epoch": 6.16, + "learning_rate": 1.9217828271793754e-06, + "loss": 0.7206, + "step": 28228 + }, + { + "epoch": 6.16, + "learning_rate": 1.921673585317894e-06, + "loss": 2.0411, + "step": 28229 + }, + { + "epoch": 6.16, + "learning_rate": 1.9215643434564126e-06, + "loss": 1.9801, + "step": 28230 + }, + { + "epoch": 6.16, + "learning_rate": 1.921455101594931e-06, + "loss": 0.2844, + "step": 28231 + }, + { + "epoch": 6.16, + "learning_rate": 1.92134585973345e-06, + "loss": 1.7118, + "step": 28232 + }, + { + "epoch": 6.16, + "learning_rate": 1.921236617871969e-06, + "loss": 2.071, + "step": 28233 + }, + { + "epoch": 6.16, + "learning_rate": 1.9211273760104874e-06, + "loss": 2.5106, + "step": 28234 + }, + { + "epoch": 6.16, + "learning_rate": 1.921018134149006e-06, + "loss": 2.4694, + "step": 28235 + }, + { + "epoch": 6.16, + "learning_rate": 1.9209088922875247e-06, + "loss": 1.3529, + "step": 28236 + }, + { + "epoch": 6.16, + "learning_rate": 1.9207996504260433e-06, + "loss": 2.5955, + "step": 28237 + }, + { + "epoch": 6.16, + "learning_rate": 1.920690408564562e-06, + "loss": 2.4698, + "step": 28238 + }, + { + "epoch": 6.16, + "learning_rate": 1.920581166703081e-06, + "loss": 1.7033, + "step": 28239 + }, + { + "epoch": 6.16, + "learning_rate": 1.9204719248415995e-06, + "loss": 0.5849, + "step": 28240 + }, + { + "epoch": 6.16, + "learning_rate": 1.920362682980118e-06, + "loss": 0.8928, + "step": 28241 + }, + { + "epoch": 6.16, + "learning_rate": 1.9202534411186368e-06, + "loss": 0.6665, + "step": 28242 + }, + { + "epoch": 6.16, + "learning_rate": 1.9201441992571554e-06, + "loss": 0.3808, + "step": 28243 + }, + { + "epoch": 6.16, + "learning_rate": 1.9200349573956744e-06, + "loss": 0.8328, + "step": 28244 + }, + { + "epoch": 6.16, + "learning_rate": 1.919925715534193e-06, + "loss": 1.451, + "step": 28245 + }, + { + "epoch": 6.16, + "learning_rate": 1.9198164736727116e-06, + "loss": 1.2073, + "step": 28246 + }, + { + "epoch": 6.16, + "learning_rate": 1.9197072318112302e-06, + "loss": 0.8678, + "step": 28247 + }, + { + "epoch": 6.16, + "learning_rate": 1.919597989949749e-06, + "loss": 0.6532, + "step": 28248 + }, + { + "epoch": 6.17, + "learning_rate": 1.9194887480882675e-06, + "loss": 1.3159, + "step": 28249 + }, + { + "epoch": 6.17, + "learning_rate": 1.919379506226786e-06, + "loss": 1.4056, + "step": 28250 + }, + { + "epoch": 6.17, + "learning_rate": 1.919270264365305e-06, + "loss": 0.1886, + "step": 28251 + }, + { + "epoch": 6.17, + "learning_rate": 1.9191610225038237e-06, + "loss": 2.0281, + "step": 28252 + }, + { + "epoch": 6.17, + "learning_rate": 1.9190517806423423e-06, + "loss": 0.8857, + "step": 28253 + }, + { + "epoch": 6.17, + "learning_rate": 1.918942538780861e-06, + "loss": 0.1666, + "step": 28254 + }, + { + "epoch": 6.17, + "learning_rate": 1.9188332969193795e-06, + "loss": 0.8013, + "step": 28255 + }, + { + "epoch": 6.17, + "learning_rate": 1.9187240550578986e-06, + "loss": 0.9942, + "step": 28256 + }, + { + "epoch": 6.17, + "learning_rate": 1.9186148131964168e-06, + "loss": 0.6138, + "step": 28257 + }, + { + "epoch": 6.17, + "learning_rate": 1.918505571334936e-06, + "loss": 1.757, + "step": 28258 + }, + { + "epoch": 6.17, + "learning_rate": 1.9183963294734544e-06, + "loss": 2.8301, + "step": 28259 + }, + { + "epoch": 6.17, + "learning_rate": 1.918287087611973e-06, + "loss": 1.0082, + "step": 28260 + }, + { + "epoch": 6.17, + "learning_rate": 1.9181778457504916e-06, + "loss": 1.1297, + "step": 28261 + }, + { + "epoch": 6.17, + "learning_rate": 1.9180686038890102e-06, + "loss": 1.6229, + "step": 28262 + }, + { + "epoch": 6.17, + "learning_rate": 1.9179593620275293e-06, + "loss": 0.7017, + "step": 28263 + }, + { + "epoch": 6.17, + "learning_rate": 1.9178501201660475e-06, + "loss": 2.4103, + "step": 28264 + }, + { + "epoch": 6.17, + "learning_rate": 1.9177408783045665e-06, + "loss": 1.4051, + "step": 28265 + }, + { + "epoch": 6.17, + "learning_rate": 1.917631636443085e-06, + "loss": 2.0052, + "step": 28266 + }, + { + "epoch": 6.17, + "learning_rate": 1.9175223945816037e-06, + "loss": 1.0478, + "step": 28267 + }, + { + "epoch": 6.17, + "learning_rate": 1.9174131527201228e-06, + "loss": 0.315, + "step": 28268 + }, + { + "epoch": 6.17, + "learning_rate": 1.917303910858641e-06, + "loss": 3.2084, + "step": 28269 + }, + { + "epoch": 6.17, + "learning_rate": 1.91719466899716e-06, + "loss": 0.9109, + "step": 28270 + }, + { + "epoch": 6.17, + "learning_rate": 1.9170854271356786e-06, + "loss": 1.5833, + "step": 28271 + }, + { + "epoch": 6.17, + "learning_rate": 1.916976185274197e-06, + "loss": 1.3658, + "step": 28272 + }, + { + "epoch": 6.17, + "learning_rate": 1.916866943412716e-06, + "loss": 1.6088, + "step": 28273 + }, + { + "epoch": 6.17, + "learning_rate": 1.9167577015512344e-06, + "loss": 0.5691, + "step": 28274 + }, + { + "epoch": 6.17, + "learning_rate": 1.9166484596897535e-06, + "loss": 1.4954, + "step": 28275 + }, + { + "epoch": 6.17, + "learning_rate": 1.9165392178282716e-06, + "loss": 0.6944, + "step": 28276 + }, + { + "epoch": 6.17, + "learning_rate": 1.9164299759667907e-06, + "loss": 2.1308, + "step": 28277 + }, + { + "epoch": 6.17, + "learning_rate": 1.9163207341053093e-06, + "loss": 0.5256, + "step": 28278 + }, + { + "epoch": 6.17, + "learning_rate": 1.916211492243828e-06, + "loss": 1.6266, + "step": 28279 + }, + { + "epoch": 6.17, + "learning_rate": 1.916102250382347e-06, + "loss": 1.1476, + "step": 28280 + }, + { + "epoch": 6.17, + "learning_rate": 1.915993008520865e-06, + "loss": 0.4507, + "step": 28281 + }, + { + "epoch": 6.17, + "learning_rate": 1.915883766659384e-06, + "loss": 0.3563, + "step": 28282 + }, + { + "epoch": 6.17, + "learning_rate": 1.9157745247979028e-06, + "loss": 1.4033, + "step": 28283 + }, + { + "epoch": 6.17, + "learning_rate": 1.9156652829364214e-06, + "loss": 1.734, + "step": 28284 + }, + { + "epoch": 6.17, + "learning_rate": 1.91555604107494e-06, + "loss": 2.705, + "step": 28285 + }, + { + "epoch": 6.17, + "learning_rate": 1.9154467992134586e-06, + "loss": 1.4976, + "step": 28286 + }, + { + "epoch": 6.17, + "learning_rate": 1.9153375573519776e-06, + "loss": 1.958, + "step": 28287 + }, + { + "epoch": 6.17, + "learning_rate": 1.9152283154904962e-06, + "loss": 0.246, + "step": 28288 + }, + { + "epoch": 6.17, + "learning_rate": 1.915119073629015e-06, + "loss": 1.2789, + "step": 28289 + }, + { + "epoch": 6.17, + "learning_rate": 1.9150098317675335e-06, + "loss": 1.981, + "step": 28290 + }, + { + "epoch": 6.17, + "learning_rate": 1.914900589906052e-06, + "loss": 1.438, + "step": 28291 + }, + { + "epoch": 6.17, + "learning_rate": 1.914791348044571e-06, + "loss": 0.7408, + "step": 28292 + }, + { + "epoch": 6.17, + "learning_rate": 1.9146821061830893e-06, + "loss": 0.1977, + "step": 28293 + }, + { + "epoch": 6.18, + "learning_rate": 1.9145728643216083e-06, + "loss": 1.6921, + "step": 28294 + }, + { + "epoch": 6.18, + "learning_rate": 1.914463622460127e-06, + "loss": 1.3852, + "step": 28295 + }, + { + "epoch": 6.18, + "learning_rate": 1.9143543805986456e-06, + "loss": 0.9611, + "step": 28296 + }, + { + "epoch": 6.18, + "learning_rate": 1.914245138737164e-06, + "loss": 1.4884, + "step": 28297 + }, + { + "epoch": 6.18, + "learning_rate": 1.9141358968756828e-06, + "loss": 1.5447, + "step": 28298 + }, + { + "epoch": 6.18, + "learning_rate": 1.914026655014202e-06, + "loss": 0.155, + "step": 28299 + }, + { + "epoch": 6.18, + "learning_rate": 1.9139174131527204e-06, + "loss": 0.8806, + "step": 28300 + }, + { + "epoch": 6.18, + "learning_rate": 1.913808171291239e-06, + "loss": 1.9166, + "step": 28301 + }, + { + "epoch": 6.18, + "learning_rate": 1.9136989294297576e-06, + "loss": 0.9264, + "step": 28302 + }, + { + "epoch": 6.18, + "learning_rate": 1.9135896875682763e-06, + "loss": 0.2828, + "step": 28303 + }, + { + "epoch": 6.18, + "learning_rate": 1.913480445706795e-06, + "loss": 0.8685, + "step": 28304 + }, + { + "epoch": 6.18, + "learning_rate": 1.9133712038453135e-06, + "loss": 0.7142, + "step": 28305 + }, + { + "epoch": 6.18, + "learning_rate": 1.9132619619838325e-06, + "loss": 1.585, + "step": 28306 + }, + { + "epoch": 6.18, + "learning_rate": 1.913152720122351e-06, + "loss": 1.684, + "step": 28307 + }, + { + "epoch": 6.18, + "learning_rate": 1.9130434782608697e-06, + "loss": 1.6958, + "step": 28308 + }, + { + "epoch": 6.18, + "learning_rate": 1.9129342363993883e-06, + "loss": 1.8808, + "step": 28309 + }, + { + "epoch": 6.18, + "learning_rate": 1.912824994537907e-06, + "loss": 1.2717, + "step": 28310 + }, + { + "epoch": 6.18, + "learning_rate": 1.912715752676426e-06, + "loss": 0.7861, + "step": 28311 + }, + { + "epoch": 6.18, + "learning_rate": 1.9126065108149446e-06, + "loss": 1.9651, + "step": 28312 + }, + { + "epoch": 6.18, + "learning_rate": 1.9124972689534632e-06, + "loss": 1.6339, + "step": 28313 + }, + { + "epoch": 6.18, + "learning_rate": 1.912388027091982e-06, + "loss": 1.6864, + "step": 28314 + }, + { + "epoch": 6.18, + "learning_rate": 1.9122787852305004e-06, + "loss": 1.7177, + "step": 28315 + }, + { + "epoch": 6.18, + "learning_rate": 1.912169543369019e-06, + "loss": 1.3049, + "step": 28316 + }, + { + "epoch": 6.18, + "learning_rate": 1.9120603015075377e-06, + "loss": 1.5121, + "step": 28317 + }, + { + "epoch": 6.18, + "learning_rate": 1.9119510596460567e-06, + "loss": 2.5312, + "step": 28318 + }, + { + "epoch": 6.18, + "learning_rate": 1.9118418177845753e-06, + "loss": 0.365, + "step": 28319 + }, + { + "epoch": 6.18, + "learning_rate": 1.911732575923094e-06, + "loss": 0.8738, + "step": 28320 + }, + { + "epoch": 6.18, + "learning_rate": 1.9116233340616125e-06, + "loss": 0.5438, + "step": 28321 + }, + { + "epoch": 6.18, + "learning_rate": 1.911514092200131e-06, + "loss": 0.956, + "step": 28322 + }, + { + "epoch": 6.18, + "learning_rate": 1.9114048503386497e-06, + "loss": 0.5305, + "step": 28323 + }, + { + "epoch": 6.18, + "learning_rate": 1.9112956084771688e-06, + "loss": 1.7378, + "step": 28324 + }, + { + "epoch": 6.18, + "learning_rate": 1.9111863666156874e-06, + "loss": 2.0045, + "step": 28325 + }, + { + "epoch": 6.18, + "learning_rate": 1.911077124754206e-06, + "loss": 0.9356, + "step": 28326 + }, + { + "epoch": 6.18, + "learning_rate": 1.9109678828927246e-06, + "loss": 1.9222, + "step": 28327 + }, + { + "epoch": 6.18, + "learning_rate": 1.9108586410312432e-06, + "loss": 3.0882, + "step": 28328 + }, + { + "epoch": 6.18, + "learning_rate": 1.910749399169762e-06, + "loss": 1.6576, + "step": 28329 + }, + { + "epoch": 6.18, + "learning_rate": 1.9106401573082804e-06, + "loss": 1.0425, + "step": 28330 + }, + { + "epoch": 6.18, + "learning_rate": 1.9105309154467995e-06, + "loss": 0.1907, + "step": 28331 + }, + { + "epoch": 6.18, + "learning_rate": 1.910421673585318e-06, + "loss": 1.1006, + "step": 28332 + }, + { + "epoch": 6.18, + "learning_rate": 1.9103124317238367e-06, + "loss": 1.7654, + "step": 28333 + }, + { + "epoch": 6.18, + "learning_rate": 1.9102031898623553e-06, + "loss": 0.6791, + "step": 28334 + }, + { + "epoch": 6.18, + "learning_rate": 1.910093948000874e-06, + "loss": 1.03, + "step": 28335 + }, + { + "epoch": 6.18, + "learning_rate": 1.909984706139393e-06, + "loss": 1.1965, + "step": 28336 + }, + { + "epoch": 6.18, + "learning_rate": 1.909875464277911e-06, + "loss": 2.067, + "step": 28337 + }, + { + "epoch": 6.18, + "learning_rate": 1.90976622241643e-06, + "loss": 1.5237, + "step": 28338 + }, + { + "epoch": 6.18, + "learning_rate": 1.909656980554949e-06, + "loss": 1.2626, + "step": 28339 + }, + { + "epoch": 6.19, + "learning_rate": 1.9095477386934674e-06, + "loss": 2.1445, + "step": 28340 + }, + { + "epoch": 6.19, + "learning_rate": 1.9094384968319864e-06, + "loss": 1.3109, + "step": 28341 + }, + { + "epoch": 6.19, + "learning_rate": 1.9093292549705046e-06, + "loss": 1.352, + "step": 28342 + }, + { + "epoch": 6.19, + "learning_rate": 1.9092200131090237e-06, + "loss": 2.8537, + "step": 28343 + }, + { + "epoch": 6.19, + "learning_rate": 1.9091107712475423e-06, + "loss": 1.6008, + "step": 28344 + }, + { + "epoch": 6.19, + "learning_rate": 1.909001529386061e-06, + "loss": 2.5566, + "step": 28345 + }, + { + "epoch": 6.19, + "learning_rate": 1.9088922875245795e-06, + "loss": 1.8301, + "step": 28346 + }, + { + "epoch": 6.19, + "learning_rate": 1.908783045663098e-06, + "loss": 0.4671, + "step": 28347 + }, + { + "epoch": 6.19, + "learning_rate": 1.908673803801617e-06, + "loss": 2.341, + "step": 28348 + }, + { + "epoch": 6.19, + "learning_rate": 1.9085645619401353e-06, + "loss": 1.5629, + "step": 28349 + }, + { + "epoch": 6.19, + "learning_rate": 1.9084553200786544e-06, + "loss": 1.1377, + "step": 28350 + }, + { + "epoch": 6.19, + "learning_rate": 1.908346078217173e-06, + "loss": 0.5731, + "step": 28351 + }, + { + "epoch": 6.19, + "learning_rate": 1.9082368363556916e-06, + "loss": 0.1558, + "step": 28352 + }, + { + "epoch": 6.19, + "learning_rate": 1.9081275944942106e-06, + "loss": 0.7041, + "step": 28353 + }, + { + "epoch": 6.19, + "learning_rate": 1.908018352632729e-06, + "loss": 1.176, + "step": 28354 + }, + { + "epoch": 6.19, + "learning_rate": 1.907909110771248e-06, + "loss": 2.7085, + "step": 28355 + }, + { + "epoch": 6.19, + "learning_rate": 1.9077998689097665e-06, + "loss": 0.6241, + "step": 28356 + }, + { + "epoch": 6.19, + "learning_rate": 1.907690627048285e-06, + "loss": 0.2783, + "step": 28357 + }, + { + "epoch": 6.19, + "learning_rate": 1.9075813851868037e-06, + "loss": 1.1643, + "step": 28358 + }, + { + "epoch": 6.19, + "learning_rate": 1.9074721433253223e-06, + "loss": 2.2375, + "step": 28359 + }, + { + "epoch": 6.19, + "learning_rate": 1.9073629014638413e-06, + "loss": 1.4339, + "step": 28360 + }, + { + "epoch": 6.19, + "learning_rate": 1.9072536596023597e-06, + "loss": 1.9247, + "step": 28361 + }, + { + "epoch": 6.19, + "learning_rate": 1.9071444177408785e-06, + "loss": 1.4481, + "step": 28362 + }, + { + "epoch": 6.19, + "learning_rate": 1.907035175879397e-06, + "loss": 1.9964, + "step": 28363 + }, + { + "epoch": 6.19, + "learning_rate": 1.9069259340179158e-06, + "loss": 1.5719, + "step": 28364 + }, + { + "epoch": 6.19, + "learning_rate": 1.9068166921564346e-06, + "loss": 1.5104, + "step": 28365 + }, + { + "epoch": 6.19, + "learning_rate": 1.9067074502949532e-06, + "loss": 2.5358, + "step": 28366 + }, + { + "epoch": 6.19, + "learning_rate": 1.906598208433472e-06, + "loss": 2.3379, + "step": 28367 + }, + { + "epoch": 6.19, + "learning_rate": 1.9064889665719904e-06, + "loss": 1.4957, + "step": 28368 + }, + { + "epoch": 6.19, + "learning_rate": 1.9063797247105092e-06, + "loss": 0.192, + "step": 28369 + }, + { + "epoch": 6.19, + "learning_rate": 1.9062704828490279e-06, + "loss": 1.1582, + "step": 28370 + }, + { + "epoch": 6.19, + "learning_rate": 1.9061612409875465e-06, + "loss": 1.3595, + "step": 28371 + }, + { + "epoch": 6.19, + "learning_rate": 1.9060519991260653e-06, + "loss": 1.995, + "step": 28372 + }, + { + "epoch": 6.19, + "learning_rate": 1.905942757264584e-06, + "loss": 1.2613, + "step": 28373 + }, + { + "epoch": 6.19, + "learning_rate": 1.9058335154031027e-06, + "loss": 2.0203, + "step": 28374 + }, + { + "epoch": 6.19, + "learning_rate": 1.9057242735416211e-06, + "loss": 1.9399, + "step": 28375 + }, + { + "epoch": 6.19, + "learning_rate": 1.90561503168014e-06, + "loss": 0.292, + "step": 28376 + }, + { + "epoch": 6.19, + "learning_rate": 1.9055057898186588e-06, + "loss": 2.015, + "step": 28377 + }, + { + "epoch": 6.19, + "learning_rate": 1.9053965479571774e-06, + "loss": 1.7097, + "step": 28378 + }, + { + "epoch": 6.19, + "learning_rate": 1.9052873060956962e-06, + "loss": 1.8681, + "step": 28379 + }, + { + "epoch": 6.19, + "learning_rate": 1.9051780642342146e-06, + "loss": 0.8916, + "step": 28380 + }, + { + "epoch": 6.19, + "learning_rate": 1.9050688223727334e-06, + "loss": 1.0293, + "step": 28381 + }, + { + "epoch": 6.19, + "learning_rate": 1.904959580511252e-06, + "loss": 1.4527, + "step": 28382 + }, + { + "epoch": 6.19, + "learning_rate": 1.9048503386497709e-06, + "loss": 1.4109, + "step": 28383 + }, + { + "epoch": 6.19, + "learning_rate": 1.9047410967882895e-06, + "loss": 1.9547, + "step": 28384 + }, + { + "epoch": 6.19, + "learning_rate": 1.904631854926808e-06, + "loss": 1.8301, + "step": 28385 + }, + { + "epoch": 6.2, + "learning_rate": 1.904522613065327e-06, + "loss": 0.2287, + "step": 28386 + }, + { + "epoch": 6.2, + "learning_rate": 1.9044133712038453e-06, + "loss": 1.3591, + "step": 28387 + }, + { + "epoch": 6.2, + "learning_rate": 1.9043041293423641e-06, + "loss": 2.6501, + "step": 28388 + }, + { + "epoch": 6.2, + "learning_rate": 1.9041948874808827e-06, + "loss": 0.9483, + "step": 28389 + }, + { + "epoch": 6.2, + "learning_rate": 1.9040856456194016e-06, + "loss": 1.1917, + "step": 28390 + }, + { + "epoch": 6.2, + "learning_rate": 1.9039764037579204e-06, + "loss": 1.188, + "step": 28391 + }, + { + "epoch": 6.2, + "learning_rate": 1.9038671618964388e-06, + "loss": 0.8092, + "step": 28392 + }, + { + "epoch": 6.2, + "learning_rate": 1.9037579200349576e-06, + "loss": 0.8537, + "step": 28393 + }, + { + "epoch": 6.2, + "learning_rate": 1.9036486781734762e-06, + "loss": 5.1564, + "step": 28394 + }, + { + "epoch": 6.2, + "learning_rate": 1.903539436311995e-06, + "loss": 0.7402, + "step": 28395 + }, + { + "epoch": 6.2, + "learning_rate": 1.9034301944505134e-06, + "loss": 0.4182, + "step": 28396 + }, + { + "epoch": 6.2, + "learning_rate": 1.9033209525890323e-06, + "loss": 2.3391, + "step": 28397 + }, + { + "epoch": 6.2, + "learning_rate": 1.903211710727551e-06, + "loss": 0.821, + "step": 28398 + }, + { + "epoch": 6.2, + "learning_rate": 1.9031024688660697e-06, + "loss": 1.1332, + "step": 28399 + }, + { + "epoch": 6.2, + "learning_rate": 1.9029932270045883e-06, + "loss": 0.6513, + "step": 28400 + }, + { + "epoch": 6.2, + "learning_rate": 1.902883985143107e-06, + "loss": 1.5559, + "step": 28401 + }, + { + "epoch": 6.2, + "learning_rate": 1.9027747432816257e-06, + "loss": 1.7344, + "step": 28402 + }, + { + "epoch": 6.2, + "learning_rate": 1.9026655014201441e-06, + "loss": 0.8902, + "step": 28403 + }, + { + "epoch": 6.2, + "learning_rate": 1.902556259558663e-06, + "loss": 1.2911, + "step": 28404 + }, + { + "epoch": 6.2, + "learning_rate": 1.9024470176971818e-06, + "loss": 0.7996, + "step": 28405 + }, + { + "epoch": 6.2, + "learning_rate": 1.9023377758357004e-06, + "loss": 2.6599, + "step": 28406 + }, + { + "epoch": 6.2, + "learning_rate": 1.9022285339742192e-06, + "loss": 1.7542, + "step": 28407 + }, + { + "epoch": 6.2, + "learning_rate": 1.9021192921127376e-06, + "loss": 1.0354, + "step": 28408 + }, + { + "epoch": 6.2, + "learning_rate": 1.9020100502512564e-06, + "loss": 1.6377, + "step": 28409 + }, + { + "epoch": 6.2, + "learning_rate": 1.9019008083897753e-06, + "loss": 2.0277, + "step": 28410 + }, + { + "epoch": 6.2, + "learning_rate": 1.9017915665282939e-06, + "loss": 1.8375, + "step": 28411 + }, + { + "epoch": 6.2, + "learning_rate": 1.9016823246668125e-06, + "loss": 2.2913, + "step": 28412 + }, + { + "epoch": 6.2, + "learning_rate": 1.901573082805331e-06, + "loss": 2.9889, + "step": 28413 + }, + { + "epoch": 6.2, + "learning_rate": 1.90146384094385e-06, + "loss": 0.9819, + "step": 28414 + }, + { + "epoch": 6.2, + "learning_rate": 1.9013545990823683e-06, + "loss": 0.9604, + "step": 28415 + }, + { + "epoch": 6.2, + "learning_rate": 1.9012453572208871e-06, + "loss": 2.1613, + "step": 28416 + }, + { + "epoch": 6.2, + "learning_rate": 1.901136115359406e-06, + "loss": 0.3078, + "step": 28417 + }, + { + "epoch": 6.2, + "learning_rate": 1.9010268734979246e-06, + "loss": 1.8453, + "step": 28418 + }, + { + "epoch": 6.2, + "learning_rate": 1.9009176316364434e-06, + "loss": 0.9852, + "step": 28419 + }, + { + "epoch": 6.2, + "learning_rate": 1.9008083897749618e-06, + "loss": 0.859, + "step": 28420 + }, + { + "epoch": 6.2, + "learning_rate": 1.9006991479134806e-06, + "loss": 1.7469, + "step": 28421 + }, + { + "epoch": 6.2, + "learning_rate": 1.9005899060519992e-06, + "loss": 0.2663, + "step": 28422 + }, + { + "epoch": 6.2, + "learning_rate": 1.900480664190518e-06, + "loss": 2.3153, + "step": 28423 + }, + { + "epoch": 6.2, + "learning_rate": 1.9003714223290367e-06, + "loss": 2.4086, + "step": 28424 + }, + { + "epoch": 6.2, + "learning_rate": 1.9002621804675553e-06, + "loss": 2.2883, + "step": 28425 + }, + { + "epoch": 6.2, + "learning_rate": 1.900152938606074e-06, + "loss": 2.1069, + "step": 28426 + }, + { + "epoch": 6.2, + "learning_rate": 1.9000436967445927e-06, + "loss": 0.3828, + "step": 28427 + }, + { + "epoch": 6.2, + "learning_rate": 1.8999344548831113e-06, + "loss": 0.0959, + "step": 28428 + }, + { + "epoch": 6.2, + "learning_rate": 1.89982521302163e-06, + "loss": 0.7397, + "step": 28429 + }, + { + "epoch": 6.2, + "learning_rate": 1.8997159711601487e-06, + "loss": 0.2733, + "step": 28430 + }, + { + "epoch": 6.2, + "learning_rate": 1.8996067292986676e-06, + "loss": 2.0204, + "step": 28431 + }, + { + "epoch": 6.21, + "learning_rate": 1.899497487437186e-06, + "loss": 1.5398, + "step": 28432 + }, + { + "epoch": 6.21, + "learning_rate": 1.8993882455757048e-06, + "loss": 1.5742, + "step": 28433 + }, + { + "epoch": 6.21, + "learning_rate": 1.8992790037142234e-06, + "loss": 0.9396, + "step": 28434 + }, + { + "epoch": 6.21, + "learning_rate": 1.8991697618527422e-06, + "loss": 1.4186, + "step": 28435 + }, + { + "epoch": 6.21, + "learning_rate": 1.8990605199912606e-06, + "loss": 0.4779, + "step": 28436 + }, + { + "epoch": 6.21, + "learning_rate": 1.8989512781297794e-06, + "loss": 0.7535, + "step": 28437 + }, + { + "epoch": 6.21, + "learning_rate": 1.8988420362682983e-06, + "loss": 1.6829, + "step": 28438 + }, + { + "epoch": 6.21, + "learning_rate": 1.8987327944068169e-06, + "loss": 1.2507, + "step": 28439 + }, + { + "epoch": 6.21, + "learning_rate": 1.8986235525453355e-06, + "loss": 1.4823, + "step": 28440 + }, + { + "epoch": 6.21, + "learning_rate": 1.898514310683854e-06, + "loss": 1.911, + "step": 28441 + }, + { + "epoch": 6.21, + "learning_rate": 1.898405068822373e-06, + "loss": 0.3723, + "step": 28442 + }, + { + "epoch": 6.21, + "learning_rate": 1.8982958269608913e-06, + "loss": 0.5641, + "step": 28443 + }, + { + "epoch": 6.21, + "learning_rate": 1.8981865850994101e-06, + "loss": 2.0527, + "step": 28444 + }, + { + "epoch": 6.21, + "learning_rate": 1.898077343237929e-06, + "loss": 1.0815, + "step": 28445 + }, + { + "epoch": 6.21, + "learning_rate": 1.8979681013764476e-06, + "loss": 1.2317, + "step": 28446 + }, + { + "epoch": 6.21, + "learning_rate": 1.8978588595149664e-06, + "loss": 1.2368, + "step": 28447 + }, + { + "epoch": 6.21, + "learning_rate": 1.8977496176534848e-06, + "loss": 1.1428, + "step": 28448 + }, + { + "epoch": 6.21, + "learning_rate": 1.8976403757920036e-06, + "loss": 0.446, + "step": 28449 + }, + { + "epoch": 6.21, + "learning_rate": 1.8975311339305224e-06, + "loss": 1.2951, + "step": 28450 + }, + { + "epoch": 6.21, + "learning_rate": 1.897421892069041e-06, + "loss": 0.7085, + "step": 28451 + }, + { + "epoch": 6.21, + "learning_rate": 1.8973126502075599e-06, + "loss": 0.2884, + "step": 28452 + }, + { + "epoch": 6.21, + "learning_rate": 1.8972034083460783e-06, + "loss": 0.9172, + "step": 28453 + }, + { + "epoch": 6.21, + "learning_rate": 1.897094166484597e-06, + "loss": 0.8013, + "step": 28454 + }, + { + "epoch": 6.21, + "learning_rate": 1.8969849246231157e-06, + "loss": 0.9501, + "step": 28455 + }, + { + "epoch": 6.21, + "learning_rate": 1.8968756827616343e-06, + "loss": 1.1969, + "step": 28456 + }, + { + "epoch": 6.21, + "learning_rate": 1.8967664409001531e-06, + "loss": 0.811, + "step": 28457 + }, + { + "epoch": 6.21, + "learning_rate": 1.8966571990386718e-06, + "loss": 0.5554, + "step": 28458 + }, + { + "epoch": 6.21, + "learning_rate": 1.8965479571771906e-06, + "loss": 0.5225, + "step": 28459 + }, + { + "epoch": 6.21, + "learning_rate": 1.896438715315709e-06, + "loss": 0.7024, + "step": 28460 + }, + { + "epoch": 6.21, + "learning_rate": 1.8963294734542278e-06, + "loss": 0.9076, + "step": 28461 + }, + { + "epoch": 6.21, + "learning_rate": 1.8962202315927464e-06, + "loss": 1.278, + "step": 28462 + }, + { + "epoch": 6.21, + "learning_rate": 1.8961109897312652e-06, + "loss": 1.5916, + "step": 28463 + }, + { + "epoch": 6.21, + "learning_rate": 1.896001747869784e-06, + "loss": 0.9427, + "step": 28464 + }, + { + "epoch": 6.21, + "learning_rate": 1.8958925060083025e-06, + "loss": 1.9292, + "step": 28465 + }, + { + "epoch": 6.21, + "learning_rate": 1.8957832641468213e-06, + "loss": 1.5005, + "step": 28466 + }, + { + "epoch": 6.21, + "learning_rate": 1.89567402228534e-06, + "loss": 1.3446, + "step": 28467 + }, + { + "epoch": 6.21, + "learning_rate": 1.8955647804238585e-06, + "loss": 0.5224, + "step": 28468 + }, + { + "epoch": 6.21, + "learning_rate": 1.8954555385623771e-06, + "loss": 0.901, + "step": 28469 + }, + { + "epoch": 6.21, + "learning_rate": 1.895346296700896e-06, + "loss": 0.9958, + "step": 28470 + }, + { + "epoch": 6.21, + "learning_rate": 1.8952370548394148e-06, + "loss": 1.7763, + "step": 28471 + }, + { + "epoch": 6.21, + "learning_rate": 1.8951278129779332e-06, + "loss": 1.2414, + "step": 28472 + }, + { + "epoch": 6.21, + "learning_rate": 1.895018571116452e-06, + "loss": 0.8889, + "step": 28473 + }, + { + "epoch": 6.21, + "learning_rate": 1.8949093292549706e-06, + "loss": 2.0869, + "step": 28474 + }, + { + "epoch": 6.21, + "learning_rate": 1.8948000873934894e-06, + "loss": 0.9028, + "step": 28475 + }, + { + "epoch": 6.21, + "learning_rate": 1.8946908455320078e-06, + "loss": 3.3001, + "step": 28476 + }, + { + "epoch": 6.21, + "learning_rate": 1.8945816036705266e-06, + "loss": 3.387, + "step": 28477 + }, + { + "epoch": 6.22, + "learning_rate": 1.8944723618090455e-06, + "loss": 1.5291, + "step": 28478 + }, + { + "epoch": 6.22, + "learning_rate": 1.894363119947564e-06, + "loss": 1.1778, + "step": 28479 + }, + { + "epoch": 6.22, + "learning_rate": 1.894253878086083e-06, + "loss": 2.8908, + "step": 28480 + }, + { + "epoch": 6.22, + "learning_rate": 1.8941446362246013e-06, + "loss": 1.9614, + "step": 28481 + }, + { + "epoch": 6.22, + "learning_rate": 1.8940353943631201e-06, + "loss": 2.7087, + "step": 28482 + }, + { + "epoch": 6.22, + "learning_rate": 1.893926152501639e-06, + "loss": 1.952, + "step": 28483 + }, + { + "epoch": 6.22, + "learning_rate": 1.8938169106401573e-06, + "loss": 1.5274, + "step": 28484 + }, + { + "epoch": 6.22, + "learning_rate": 1.8937076687786762e-06, + "loss": 0.4312, + "step": 28485 + }, + { + "epoch": 6.22, + "learning_rate": 1.8935984269171948e-06, + "loss": 3.3054, + "step": 28486 + }, + { + "epoch": 6.22, + "learning_rate": 1.8934891850557136e-06, + "loss": 1.5802, + "step": 28487 + }, + { + "epoch": 6.22, + "learning_rate": 1.893379943194232e-06, + "loss": 1.5993, + "step": 28488 + }, + { + "epoch": 6.22, + "learning_rate": 1.8932707013327508e-06, + "loss": 0.9671, + "step": 28489 + }, + { + "epoch": 6.22, + "learning_rate": 1.8931614594712696e-06, + "loss": 1.2704, + "step": 28490 + }, + { + "epoch": 6.22, + "learning_rate": 1.8930522176097883e-06, + "loss": 1.4542, + "step": 28491 + }, + { + "epoch": 6.22, + "learning_rate": 1.892942975748307e-06, + "loss": 1.0613, + "step": 28492 + }, + { + "epoch": 6.22, + "learning_rate": 1.8928337338868255e-06, + "loss": 0.2209, + "step": 28493 + }, + { + "epoch": 6.22, + "learning_rate": 1.8927244920253443e-06, + "loss": 0.962, + "step": 28494 + }, + { + "epoch": 6.22, + "learning_rate": 1.892615250163863e-06, + "loss": 1.257, + "step": 28495 + }, + { + "epoch": 6.22, + "learning_rate": 1.8925060083023817e-06, + "loss": 2.4649, + "step": 28496 + }, + { + "epoch": 6.22, + "learning_rate": 1.8923967664409003e-06, + "loss": 1.2342, + "step": 28497 + }, + { + "epoch": 6.22, + "learning_rate": 1.892287524579419e-06, + "loss": 0.9641, + "step": 28498 + }, + { + "epoch": 6.22, + "learning_rate": 1.8921782827179378e-06, + "loss": 1.4704, + "step": 28499 + }, + { + "epoch": 6.22, + "learning_rate": 1.8920690408564562e-06, + "loss": 1.3723, + "step": 28500 + }, + { + "epoch": 6.22, + "learning_rate": 1.891959798994975e-06, + "loss": 2.723, + "step": 28501 + }, + { + "epoch": 6.22, + "learning_rate": 1.8918505571334936e-06, + "loss": 0.416, + "step": 28502 + }, + { + "epoch": 6.22, + "learning_rate": 1.8917413152720124e-06, + "loss": 2.5375, + "step": 28503 + }, + { + "epoch": 6.22, + "learning_rate": 1.8916320734105313e-06, + "loss": 1.805, + "step": 28504 + }, + { + "epoch": 6.22, + "learning_rate": 1.8915228315490497e-06, + "loss": 1.8682, + "step": 28505 + }, + { + "epoch": 6.22, + "learning_rate": 1.8914135896875685e-06, + "loss": 1.7854, + "step": 28506 + }, + { + "epoch": 6.22, + "learning_rate": 1.891304347826087e-06, + "loss": 1.0021, + "step": 28507 + }, + { + "epoch": 6.22, + "learning_rate": 1.891195105964606e-06, + "loss": 1.3295, + "step": 28508 + }, + { + "epoch": 6.22, + "learning_rate": 1.8910858641031243e-06, + "loss": 0.8563, + "step": 28509 + }, + { + "epoch": 6.22, + "learning_rate": 1.8909766222416431e-06, + "loss": 2.1838, + "step": 28510 + }, + { + "epoch": 6.22, + "learning_rate": 1.890867380380162e-06, + "loss": 0.7221, + "step": 28511 + }, + { + "epoch": 6.22, + "learning_rate": 1.8907581385186804e-06, + "loss": 1.8605, + "step": 28512 + }, + { + "epoch": 6.22, + "learning_rate": 1.8906488966571992e-06, + "loss": 1.2593, + "step": 28513 + }, + { + "epoch": 6.22, + "learning_rate": 1.8905396547957178e-06, + "loss": 1.6001, + "step": 28514 + }, + { + "epoch": 6.22, + "learning_rate": 1.8904304129342366e-06, + "loss": 1.788, + "step": 28515 + }, + { + "epoch": 6.22, + "learning_rate": 1.8903211710727554e-06, + "loss": 1.1115, + "step": 28516 + }, + { + "epoch": 6.22, + "learning_rate": 1.8902119292112738e-06, + "loss": 2.3414, + "step": 28517 + }, + { + "epoch": 6.22, + "learning_rate": 1.8901026873497927e-06, + "loss": 2.2453, + "step": 28518 + }, + { + "epoch": 6.22, + "learning_rate": 1.8899934454883113e-06, + "loss": 2.0639, + "step": 28519 + }, + { + "epoch": 6.22, + "learning_rate": 1.88988420362683e-06, + "loss": 1.1103, + "step": 28520 + }, + { + "epoch": 6.22, + "learning_rate": 1.8897749617653485e-06, + "loss": 1.3498, + "step": 28521 + }, + { + "epoch": 6.22, + "learning_rate": 1.8896657199038673e-06, + "loss": 0.8138, + "step": 28522 + }, + { + "epoch": 6.23, + "learning_rate": 1.8895564780423861e-06, + "loss": 1.1689, + "step": 28523 + }, + { + "epoch": 6.23, + "learning_rate": 1.8894472361809047e-06, + "loss": 1.1517, + "step": 28524 + }, + { + "epoch": 6.23, + "learning_rate": 1.8893379943194234e-06, + "loss": 1.3415, + "step": 28525 + }, + { + "epoch": 6.23, + "learning_rate": 1.889228752457942e-06, + "loss": 0.1892, + "step": 28526 + }, + { + "epoch": 6.23, + "learning_rate": 1.8891195105964608e-06, + "loss": 1.0542, + "step": 28527 + }, + { + "epoch": 6.23, + "learning_rate": 1.8890102687349792e-06, + "loss": 0.7898, + "step": 28528 + }, + { + "epoch": 6.23, + "learning_rate": 1.888901026873498e-06, + "loss": 1.5502, + "step": 28529 + }, + { + "epoch": 6.23, + "learning_rate": 1.8887917850120168e-06, + "loss": 3.4802, + "step": 28530 + }, + { + "epoch": 6.23, + "learning_rate": 1.8886825431505354e-06, + "loss": 1.4061, + "step": 28531 + }, + { + "epoch": 6.23, + "learning_rate": 1.8885733012890543e-06, + "loss": 0.9541, + "step": 28532 + }, + { + "epoch": 6.23, + "learning_rate": 1.8884640594275727e-06, + "loss": 1.8615, + "step": 28533 + }, + { + "epoch": 6.23, + "learning_rate": 1.8883548175660915e-06, + "loss": 0.2001, + "step": 28534 + }, + { + "epoch": 6.23, + "learning_rate": 1.88824557570461e-06, + "loss": 2.2489, + "step": 28535 + }, + { + "epoch": 6.23, + "learning_rate": 1.888136333843129e-06, + "loss": 1.1155, + "step": 28536 + }, + { + "epoch": 6.23, + "learning_rate": 1.8880270919816475e-06, + "loss": 0.3389, + "step": 28537 + }, + { + "epoch": 6.23, + "learning_rate": 1.8879178501201661e-06, + "loss": 0.7294, + "step": 28538 + }, + { + "epoch": 6.23, + "learning_rate": 1.887808608258685e-06, + "loss": 1.0814, + "step": 28539 + }, + { + "epoch": 6.23, + "learning_rate": 1.8876993663972034e-06, + "loss": 1.7404, + "step": 28540 + }, + { + "epoch": 6.23, + "learning_rate": 1.8875901245357222e-06, + "loss": 1.1329, + "step": 28541 + }, + { + "epoch": 6.23, + "learning_rate": 1.8874808826742408e-06, + "loss": 0.9711, + "step": 28542 + }, + { + "epoch": 6.23, + "learning_rate": 1.8873716408127596e-06, + "loss": 1.5254, + "step": 28543 + }, + { + "epoch": 6.23, + "learning_rate": 1.8872623989512784e-06, + "loss": 1.0776, + "step": 28544 + }, + { + "epoch": 6.23, + "learning_rate": 1.8871531570897968e-06, + "loss": 2.3325, + "step": 28545 + }, + { + "epoch": 6.23, + "learning_rate": 1.8870439152283157e-06, + "loss": 1.2874, + "step": 28546 + }, + { + "epoch": 6.23, + "learning_rate": 1.8869346733668343e-06, + "loss": 1.6651, + "step": 28547 + }, + { + "epoch": 6.23, + "learning_rate": 1.886825431505353e-06, + "loss": 1.1851, + "step": 28548 + }, + { + "epoch": 6.23, + "learning_rate": 1.886716189643872e-06, + "loss": 0.6876, + "step": 28549 + }, + { + "epoch": 6.23, + "learning_rate": 1.8866069477823903e-06, + "loss": 0.4935, + "step": 28550 + }, + { + "epoch": 6.23, + "learning_rate": 1.8864977059209091e-06, + "loss": 1.6878, + "step": 28551 + }, + { + "epoch": 6.23, + "learning_rate": 1.8863884640594278e-06, + "loss": 1.9617, + "step": 28552 + }, + { + "epoch": 6.23, + "learning_rate": 1.8862792221979464e-06, + "loss": 1.6065, + "step": 28553 + }, + { + "epoch": 6.23, + "learning_rate": 1.886169980336465e-06, + "loss": 0.6469, + "step": 28554 + }, + { + "epoch": 6.23, + "learning_rate": 1.8860607384749838e-06, + "loss": 0.8278, + "step": 28555 + }, + { + "epoch": 6.23, + "learning_rate": 1.8859514966135026e-06, + "loss": 2.535, + "step": 28556 + }, + { + "epoch": 6.23, + "learning_rate": 1.885842254752021e-06, + "loss": 1.4906, + "step": 28557 + }, + { + "epoch": 6.23, + "learning_rate": 1.8857330128905398e-06, + "loss": 1.1773, + "step": 28558 + }, + { + "epoch": 6.23, + "learning_rate": 1.8856237710290585e-06, + "loss": 0.7628, + "step": 28559 + }, + { + "epoch": 6.23, + "learning_rate": 1.8855145291675773e-06, + "loss": 3.6441, + "step": 28560 + }, + { + "epoch": 6.23, + "learning_rate": 1.8854052873060957e-06, + "loss": 1.1811, + "step": 28561 + }, + { + "epoch": 6.23, + "learning_rate": 1.8852960454446145e-06, + "loss": 0.131, + "step": 28562 + }, + { + "epoch": 6.23, + "learning_rate": 1.8851868035831333e-06, + "loss": 1.1322, + "step": 28563 + }, + { + "epoch": 6.23, + "learning_rate": 1.885077561721652e-06, + "loss": 0.5187, + "step": 28564 + }, + { + "epoch": 6.23, + "learning_rate": 1.8849683198601705e-06, + "loss": 1.2612, + "step": 28565 + }, + { + "epoch": 6.23, + "learning_rate": 1.8848590779986892e-06, + "loss": 0.7791, + "step": 28566 + }, + { + "epoch": 6.23, + "learning_rate": 1.884749836137208e-06, + "loss": 3.1893, + "step": 28567 + }, + { + "epoch": 6.23, + "learning_rate": 1.8846405942757266e-06, + "loss": 1.2361, + "step": 28568 + }, + { + "epoch": 6.24, + "learning_rate": 1.8845313524142452e-06, + "loss": 1.8434, + "step": 28569 + }, + { + "epoch": 6.24, + "learning_rate": 1.884422110552764e-06, + "loss": 0.9788, + "step": 28570 + }, + { + "epoch": 6.24, + "learning_rate": 1.8843128686912826e-06, + "loss": 0.106, + "step": 28571 + }, + { + "epoch": 6.24, + "learning_rate": 1.8842036268298015e-06, + "loss": 0.8869, + "step": 28572 + }, + { + "epoch": 6.24, + "learning_rate": 1.8840943849683199e-06, + "loss": 0.0973, + "step": 28573 + }, + { + "epoch": 6.24, + "learning_rate": 1.8839851431068387e-06, + "loss": 1.2018, + "step": 28574 + }, + { + "epoch": 6.24, + "learning_rate": 1.8838759012453573e-06, + "loss": 2.287, + "step": 28575 + }, + { + "epoch": 6.24, + "learning_rate": 1.8837666593838761e-06, + "loss": 1.5557, + "step": 28576 + }, + { + "epoch": 6.24, + "learning_rate": 1.883657417522395e-06, + "loss": 0.3798, + "step": 28577 + }, + { + "epoch": 6.24, + "learning_rate": 1.8835481756609133e-06, + "loss": 0.9843, + "step": 28578 + }, + { + "epoch": 6.24, + "learning_rate": 1.8834389337994322e-06, + "loss": 0.7481, + "step": 28579 + }, + { + "epoch": 6.24, + "learning_rate": 1.8833296919379508e-06, + "loss": 0.6716, + "step": 28580 + }, + { + "epoch": 6.24, + "learning_rate": 1.8832204500764694e-06, + "loss": 0.4218, + "step": 28581 + }, + { + "epoch": 6.24, + "learning_rate": 1.8831112082149882e-06, + "loss": 1.1052, + "step": 28582 + }, + { + "epoch": 6.24, + "learning_rate": 1.8830019663535068e-06, + "loss": 1.0323, + "step": 28583 + }, + { + "epoch": 6.24, + "learning_rate": 1.8828927244920256e-06, + "loss": 1.1557, + "step": 28584 + }, + { + "epoch": 6.24, + "learning_rate": 1.882783482630544e-06, + "loss": 1.1965, + "step": 28585 + }, + { + "epoch": 6.24, + "learning_rate": 1.8826742407690629e-06, + "loss": 0.5122, + "step": 28586 + }, + { + "epoch": 6.24, + "learning_rate": 1.8825649989075815e-06, + "loss": 1.2955, + "step": 28587 + }, + { + "epoch": 6.24, + "learning_rate": 1.8824557570461003e-06, + "loss": 2.8671, + "step": 28588 + }, + { + "epoch": 6.24, + "learning_rate": 1.8823465151846191e-06, + "loss": 1.1468, + "step": 28589 + }, + { + "epoch": 6.24, + "learning_rate": 1.8822372733231375e-06, + "loss": 0.8224, + "step": 28590 + }, + { + "epoch": 6.24, + "learning_rate": 1.8821280314616563e-06, + "loss": 0.1961, + "step": 28591 + }, + { + "epoch": 6.24, + "learning_rate": 1.882018789600175e-06, + "loss": 1.9887, + "step": 28592 + }, + { + "epoch": 6.24, + "learning_rate": 1.8819095477386936e-06, + "loss": 1.7463, + "step": 28593 + }, + { + "epoch": 6.24, + "learning_rate": 1.8818003058772122e-06, + "loss": 1.3565, + "step": 28594 + }, + { + "epoch": 6.24, + "learning_rate": 1.881691064015731e-06, + "loss": 1.7145, + "step": 28595 + }, + { + "epoch": 6.24, + "learning_rate": 1.8815818221542498e-06, + "loss": 1.6363, + "step": 28596 + }, + { + "epoch": 6.24, + "learning_rate": 1.8814725802927682e-06, + "loss": 0.789, + "step": 28597 + }, + { + "epoch": 6.24, + "learning_rate": 1.881363338431287e-06, + "loss": 0.8183, + "step": 28598 + }, + { + "epoch": 6.24, + "learning_rate": 1.8812540965698057e-06, + "loss": 1.3697, + "step": 28599 + }, + { + "epoch": 6.24, + "learning_rate": 1.8811448547083245e-06, + "loss": 0.8488, + "step": 28600 + }, + { + "epoch": 6.24, + "learning_rate": 1.8810356128468429e-06, + "loss": 1.202, + "step": 28601 + }, + { + "epoch": 6.24, + "learning_rate": 1.8809263709853617e-06, + "loss": 0.5656, + "step": 28602 + }, + { + "epoch": 6.24, + "learning_rate": 1.8808171291238805e-06, + "loss": 1.5327, + "step": 28603 + }, + { + "epoch": 6.24, + "learning_rate": 1.8807078872623991e-06, + "loss": 0.6678, + "step": 28604 + }, + { + "epoch": 6.24, + "learning_rate": 1.880598645400918e-06, + "loss": 1.5219, + "step": 28605 + }, + { + "epoch": 6.24, + "learning_rate": 1.8804894035394364e-06, + "loss": 1.7183, + "step": 28606 + }, + { + "epoch": 6.24, + "learning_rate": 1.8803801616779552e-06, + "loss": 0.9303, + "step": 28607 + }, + { + "epoch": 6.24, + "learning_rate": 1.8802709198164738e-06, + "loss": 1.5433, + "step": 28608 + }, + { + "epoch": 6.24, + "learning_rate": 1.8801616779549924e-06, + "loss": 0.1975, + "step": 28609 + }, + { + "epoch": 6.24, + "learning_rate": 1.8800524360935112e-06, + "loss": 1.3435, + "step": 28610 + }, + { + "epoch": 6.24, + "learning_rate": 1.8799431942320298e-06, + "loss": 1.413, + "step": 28611 + }, + { + "epoch": 6.24, + "learning_rate": 1.8798339523705487e-06, + "loss": 2.5531, + "step": 28612 + }, + { + "epoch": 6.24, + "learning_rate": 1.879724710509067e-06, + "loss": 1.2473, + "step": 28613 + }, + { + "epoch": 6.24, + "learning_rate": 1.8796154686475859e-06, + "loss": 2.7761, + "step": 28614 + }, + { + "epoch": 6.25, + "learning_rate": 1.8795062267861047e-06, + "loss": 0.4892, + "step": 28615 + }, + { + "epoch": 6.25, + "learning_rate": 1.8793969849246233e-06, + "loss": 1.7914, + "step": 28616 + }, + { + "epoch": 6.25, + "learning_rate": 1.8792877430631421e-06, + "loss": 1.1255, + "step": 28617 + }, + { + "epoch": 6.25, + "learning_rate": 1.8791785012016605e-06, + "loss": 1.7483, + "step": 28618 + }, + { + "epoch": 6.25, + "learning_rate": 1.8790692593401794e-06, + "loss": 1.9331, + "step": 28619 + }, + { + "epoch": 6.25, + "learning_rate": 1.878960017478698e-06, + "loss": 2.3129, + "step": 28620 + }, + { + "epoch": 6.25, + "learning_rate": 1.8788507756172168e-06, + "loss": 1.2871, + "step": 28621 + }, + { + "epoch": 6.25, + "learning_rate": 1.8787415337557354e-06, + "loss": 1.6463, + "step": 28622 + }, + { + "epoch": 6.25, + "learning_rate": 1.878632291894254e-06, + "loss": 2.1383, + "step": 28623 + }, + { + "epoch": 6.25, + "learning_rate": 1.8785230500327728e-06, + "loss": 2.6774, + "step": 28624 + }, + { + "epoch": 6.25, + "learning_rate": 1.8784138081712912e-06, + "loss": 1.6575, + "step": 28625 + }, + { + "epoch": 6.25, + "learning_rate": 1.87830456630981e-06, + "loss": 2.5315, + "step": 28626 + }, + { + "epoch": 6.25, + "learning_rate": 1.8781953244483287e-06, + "loss": 1.0348, + "step": 28627 + }, + { + "epoch": 6.25, + "learning_rate": 1.8780860825868475e-06, + "loss": 1.97, + "step": 28628 + }, + { + "epoch": 6.25, + "learning_rate": 1.8779768407253663e-06, + "loss": 2.4151, + "step": 28629 + }, + { + "epoch": 6.25, + "learning_rate": 1.8778675988638847e-06, + "loss": 1.1898, + "step": 28630 + }, + { + "epoch": 6.25, + "learning_rate": 1.8777583570024035e-06, + "loss": 2.7152, + "step": 28631 + }, + { + "epoch": 6.25, + "learning_rate": 1.8776491151409221e-06, + "loss": 2.3913, + "step": 28632 + }, + { + "epoch": 6.25, + "learning_rate": 1.877539873279441e-06, + "loss": 0.8539, + "step": 28633 + }, + { + "epoch": 6.25, + "learning_rate": 1.8774306314179594e-06, + "loss": 1.1851, + "step": 28634 + }, + { + "epoch": 6.25, + "learning_rate": 1.8773213895564782e-06, + "loss": 0.4426, + "step": 28635 + }, + { + "epoch": 6.25, + "learning_rate": 1.877212147694997e-06, + "loss": 2.5641, + "step": 28636 + }, + { + "epoch": 6.25, + "learning_rate": 1.8771029058335154e-06, + "loss": 1.0809, + "step": 28637 + }, + { + "epoch": 6.25, + "learning_rate": 1.8769936639720342e-06, + "loss": 2.7459, + "step": 28638 + }, + { + "epoch": 6.25, + "learning_rate": 1.8768844221105528e-06, + "loss": 0.9432, + "step": 28639 + }, + { + "epoch": 6.25, + "learning_rate": 1.8767751802490717e-06, + "loss": 1.4037, + "step": 28640 + }, + { + "epoch": 6.25, + "learning_rate": 1.87666593838759e-06, + "loss": 1.1001, + "step": 28641 + }, + { + "epoch": 6.25, + "learning_rate": 1.8765566965261089e-06, + "loss": 2.7863, + "step": 28642 + }, + { + "epoch": 6.25, + "learning_rate": 1.8764474546646277e-06, + "loss": 1.9015, + "step": 28643 + }, + { + "epoch": 6.25, + "learning_rate": 1.8763382128031463e-06, + "loss": 2.1695, + "step": 28644 + }, + { + "epoch": 6.25, + "learning_rate": 1.8762289709416651e-06, + "loss": 0.7725, + "step": 28645 + }, + { + "epoch": 6.25, + "learning_rate": 1.8761197290801835e-06, + "loss": 3.2625, + "step": 28646 + }, + { + "epoch": 6.25, + "learning_rate": 1.8760104872187024e-06, + "loss": 1.4154, + "step": 28647 + }, + { + "epoch": 6.25, + "learning_rate": 1.8759012453572212e-06, + "loss": 1.7607, + "step": 28648 + }, + { + "epoch": 6.25, + "learning_rate": 1.8757920034957398e-06, + "loss": 0.9565, + "step": 28649 + }, + { + "epoch": 6.25, + "learning_rate": 1.8756827616342584e-06, + "loss": 1.6225, + "step": 28650 + }, + { + "epoch": 6.25, + "learning_rate": 1.875573519772777e-06, + "loss": 1.8299, + "step": 28651 + }, + { + "epoch": 6.25, + "learning_rate": 1.8754642779112958e-06, + "loss": 1.7124, + "step": 28652 + }, + { + "epoch": 6.25, + "learning_rate": 1.8753550360498142e-06, + "loss": 0.3919, + "step": 28653 + }, + { + "epoch": 6.25, + "learning_rate": 1.875245794188333e-06, + "loss": 1.2449, + "step": 28654 + }, + { + "epoch": 6.25, + "learning_rate": 1.8751365523268519e-06, + "loss": 0.3089, + "step": 28655 + }, + { + "epoch": 6.25, + "learning_rate": 1.8750273104653705e-06, + "loss": 1.3819, + "step": 28656 + }, + { + "epoch": 6.25, + "learning_rate": 1.8749180686038893e-06, + "loss": 1.0977, + "step": 28657 + }, + { + "epoch": 6.25, + "learning_rate": 1.8748088267424077e-06, + "loss": 2.0636, + "step": 28658 + }, + { + "epoch": 6.25, + "learning_rate": 1.8746995848809265e-06, + "loss": 3.3499, + "step": 28659 + }, + { + "epoch": 6.25, + "learning_rate": 1.8745903430194452e-06, + "loss": 1.2373, + "step": 28660 + }, + { + "epoch": 6.26, + "learning_rate": 1.874481101157964e-06, + "loss": 1.0869, + "step": 28661 + }, + { + "epoch": 6.26, + "learning_rate": 1.8743718592964826e-06, + "loss": 1.1026, + "step": 28662 + }, + { + "epoch": 6.26, + "learning_rate": 1.8742626174350012e-06, + "loss": 1.0691, + "step": 28663 + }, + { + "epoch": 6.26, + "learning_rate": 1.87415337557352e-06, + "loss": 0.796, + "step": 28664 + }, + { + "epoch": 6.26, + "learning_rate": 1.8740441337120384e-06, + "loss": 0.907, + "step": 28665 + }, + { + "epoch": 6.26, + "learning_rate": 1.8739348918505572e-06, + "loss": 1.2661, + "step": 28666 + }, + { + "epoch": 6.26, + "learning_rate": 1.8738256499890759e-06, + "loss": 2.0727, + "step": 28667 + }, + { + "epoch": 6.26, + "learning_rate": 1.8737164081275947e-06, + "loss": 1.6156, + "step": 28668 + }, + { + "epoch": 6.26, + "learning_rate": 1.8736071662661135e-06, + "loss": 0.6295, + "step": 28669 + }, + { + "epoch": 6.26, + "learning_rate": 1.873497924404632e-06, + "loss": 1.6071, + "step": 28670 + }, + { + "epoch": 6.26, + "learning_rate": 1.8733886825431507e-06, + "loss": 2.2629, + "step": 28671 + }, + { + "epoch": 6.26, + "learning_rate": 1.8732794406816693e-06, + "loss": 1.4344, + "step": 28672 + }, + { + "epoch": 6.26, + "learning_rate": 1.8731701988201882e-06, + "loss": 2.901, + "step": 28673 + }, + { + "epoch": 6.26, + "learning_rate": 1.8730609569587066e-06, + "loss": 0.3152, + "step": 28674 + }, + { + "epoch": 6.26, + "learning_rate": 1.8729517150972254e-06, + "loss": 1.283, + "step": 28675 + }, + { + "epoch": 6.26, + "learning_rate": 1.8728424732357442e-06, + "loss": 2.0087, + "step": 28676 + }, + { + "epoch": 6.26, + "learning_rate": 1.8727332313742628e-06, + "loss": 1.8499, + "step": 28677 + }, + { + "epoch": 6.26, + "learning_rate": 1.8726239895127814e-06, + "loss": 1.7221, + "step": 28678 + }, + { + "epoch": 6.26, + "learning_rate": 1.8725147476513e-06, + "loss": 2.2333, + "step": 28679 + }, + { + "epoch": 6.26, + "learning_rate": 1.8724055057898189e-06, + "loss": 0.3606, + "step": 28680 + }, + { + "epoch": 6.26, + "learning_rate": 1.8722962639283373e-06, + "loss": 0.9982, + "step": 28681 + }, + { + "epoch": 6.26, + "learning_rate": 1.872187022066856e-06, + "loss": 1.0708, + "step": 28682 + }, + { + "epoch": 6.26, + "learning_rate": 1.872077780205375e-06, + "loss": 2.4544, + "step": 28683 + }, + { + "epoch": 6.26, + "learning_rate": 1.8719685383438935e-06, + "loss": 1.8757, + "step": 28684 + }, + { + "epoch": 6.26, + "learning_rate": 1.8718592964824123e-06, + "loss": 2.0063, + "step": 28685 + }, + { + "epoch": 6.26, + "learning_rate": 1.8717500546209307e-06, + "loss": 1.9247, + "step": 28686 + }, + { + "epoch": 6.26, + "learning_rate": 1.8716408127594496e-06, + "loss": 1.9417, + "step": 28687 + }, + { + "epoch": 6.26, + "learning_rate": 1.8715315708979684e-06, + "loss": 1.1042, + "step": 28688 + }, + { + "epoch": 6.26, + "learning_rate": 1.871422329036487e-06, + "loss": 1.1486, + "step": 28689 + }, + { + "epoch": 6.26, + "learning_rate": 1.8713130871750056e-06, + "loss": 1.7824, + "step": 28690 + }, + { + "epoch": 6.26, + "learning_rate": 1.8712038453135242e-06, + "loss": 2.1569, + "step": 28691 + }, + { + "epoch": 6.26, + "learning_rate": 1.871094603452043e-06, + "loss": 0.9385, + "step": 28692 + }, + { + "epoch": 6.26, + "learning_rate": 1.8709853615905616e-06, + "loss": 1.1427, + "step": 28693 + }, + { + "epoch": 6.26, + "learning_rate": 1.8708761197290803e-06, + "loss": 1.1601, + "step": 28694 + }, + { + "epoch": 6.26, + "learning_rate": 1.870766877867599e-06, + "loss": 1.0151, + "step": 28695 + }, + { + "epoch": 6.26, + "learning_rate": 1.8706576360061177e-06, + "loss": 0.0899, + "step": 28696 + }, + { + "epoch": 6.26, + "learning_rate": 1.8705483941446365e-06, + "loss": 4.3747, + "step": 28697 + }, + { + "epoch": 6.26, + "learning_rate": 1.870439152283155e-06, + "loss": 0.6809, + "step": 28698 + }, + { + "epoch": 6.26, + "learning_rate": 1.8703299104216737e-06, + "loss": 1.4778, + "step": 28699 + }, + { + "epoch": 6.26, + "learning_rate": 1.8702206685601923e-06, + "loss": 1.7101, + "step": 28700 + }, + { + "epoch": 6.26, + "learning_rate": 1.8701114266987112e-06, + "loss": 0.6561, + "step": 28701 + }, + { + "epoch": 6.26, + "learning_rate": 1.87000218483723e-06, + "loss": 0.0754, + "step": 28702 + }, + { + "epoch": 6.26, + "learning_rate": 1.8698929429757484e-06, + "loss": 2.1881, + "step": 28703 + }, + { + "epoch": 6.26, + "learning_rate": 1.8697837011142672e-06, + "loss": 0.666, + "step": 28704 + }, + { + "epoch": 6.26, + "learning_rate": 1.8696744592527858e-06, + "loss": 1.5239, + "step": 28705 + }, + { + "epoch": 6.26, + "learning_rate": 1.8695652173913044e-06, + "loss": 1.4099, + "step": 28706 + }, + { + "epoch": 6.27, + "learning_rate": 1.869455975529823e-06, + "loss": 1.3556, + "step": 28707 + }, + { + "epoch": 6.27, + "learning_rate": 1.8693467336683419e-06, + "loss": 2.2017, + "step": 28708 + }, + { + "epoch": 6.27, + "learning_rate": 1.8692374918068607e-06, + "loss": 1.3269, + "step": 28709 + }, + { + "epoch": 6.27, + "learning_rate": 1.869128249945379e-06, + "loss": 1.0428, + "step": 28710 + }, + { + "epoch": 6.27, + "learning_rate": 1.869019008083898e-06, + "loss": 1.946, + "step": 28711 + }, + { + "epoch": 6.27, + "learning_rate": 1.8689097662224165e-06, + "loss": 0.6871, + "step": 28712 + }, + { + "epoch": 6.27, + "learning_rate": 1.8688005243609354e-06, + "loss": 3.7428, + "step": 28713 + }, + { + "epoch": 6.27, + "learning_rate": 1.8686912824994537e-06, + "loss": 1.7279, + "step": 28714 + }, + { + "epoch": 6.27, + "learning_rate": 1.8685820406379726e-06, + "loss": 1.3675, + "step": 28715 + }, + { + "epoch": 6.27, + "learning_rate": 1.8684727987764914e-06, + "loss": 0.7324, + "step": 28716 + }, + { + "epoch": 6.27, + "learning_rate": 1.86836355691501e-06, + "loss": 0.3848, + "step": 28717 + }, + { + "epoch": 6.27, + "learning_rate": 1.8682543150535286e-06, + "loss": 0.8466, + "step": 28718 + }, + { + "epoch": 6.27, + "learning_rate": 1.8681450731920472e-06, + "loss": 2.3936, + "step": 28719 + }, + { + "epoch": 6.27, + "learning_rate": 1.868035831330566e-06, + "loss": 0.85, + "step": 28720 + }, + { + "epoch": 6.27, + "learning_rate": 1.8679265894690849e-06, + "loss": 1.8839, + "step": 28721 + }, + { + "epoch": 6.27, + "learning_rate": 1.8678173476076033e-06, + "loss": 0.864, + "step": 28722 + }, + { + "epoch": 6.27, + "learning_rate": 1.867708105746122e-06, + "loss": 2.6167, + "step": 28723 + }, + { + "epoch": 6.27, + "learning_rate": 1.8675988638846407e-06, + "loss": 0.9108, + "step": 28724 + }, + { + "epoch": 6.27, + "learning_rate": 1.8674896220231595e-06, + "loss": 0.2547, + "step": 28725 + }, + { + "epoch": 6.27, + "learning_rate": 1.867380380161678e-06, + "loss": 0.7635, + "step": 28726 + }, + { + "epoch": 6.27, + "learning_rate": 1.8672711383001968e-06, + "loss": 1.8246, + "step": 28727 + }, + { + "epoch": 6.27, + "learning_rate": 1.8671618964387156e-06, + "loss": 0.5404, + "step": 28728 + }, + { + "epoch": 6.27, + "learning_rate": 1.8670526545772342e-06, + "loss": 2.1143, + "step": 28729 + }, + { + "epoch": 6.27, + "learning_rate": 1.866943412715753e-06, + "loss": 1.0441, + "step": 28730 + }, + { + "epoch": 6.27, + "learning_rate": 1.8668341708542714e-06, + "loss": 0.3965, + "step": 28731 + }, + { + "epoch": 6.27, + "learning_rate": 1.8667249289927902e-06, + "loss": 0.5494, + "step": 28732 + }, + { + "epoch": 6.27, + "learning_rate": 1.8666156871313088e-06, + "loss": 2.6282, + "step": 28733 + }, + { + "epoch": 6.27, + "learning_rate": 1.8665064452698275e-06, + "loss": 1.7844, + "step": 28734 + }, + { + "epoch": 6.27, + "learning_rate": 1.8663972034083463e-06, + "loss": 0.7424, + "step": 28735 + }, + { + "epoch": 6.27, + "learning_rate": 1.8662879615468649e-06, + "loss": 0.9335, + "step": 28736 + }, + { + "epoch": 6.27, + "learning_rate": 1.8661787196853837e-06, + "loss": 2.4159, + "step": 28737 + }, + { + "epoch": 6.27, + "learning_rate": 1.8660694778239021e-06, + "loss": 3.2627, + "step": 28738 + }, + { + "epoch": 6.27, + "learning_rate": 1.865960235962421e-06, + "loss": 1.12, + "step": 28739 + }, + { + "epoch": 6.27, + "learning_rate": 1.8658509941009395e-06, + "loss": 1.507, + "step": 28740 + }, + { + "epoch": 6.27, + "learning_rate": 1.8657417522394584e-06, + "loss": 1.4329, + "step": 28741 + }, + { + "epoch": 6.27, + "learning_rate": 1.8656325103779772e-06, + "loss": 4.536, + "step": 28742 + }, + { + "epoch": 6.27, + "learning_rate": 1.8655232685164956e-06, + "loss": 1.6451, + "step": 28743 + }, + { + "epoch": 6.27, + "learning_rate": 1.8654140266550144e-06, + "loss": 2.2997, + "step": 28744 + }, + { + "epoch": 6.27, + "learning_rate": 1.865304784793533e-06, + "loss": 0.6012, + "step": 28745 + }, + { + "epoch": 6.27, + "learning_rate": 1.8651955429320518e-06, + "loss": 1.4553, + "step": 28746 + }, + { + "epoch": 6.27, + "learning_rate": 1.8650863010705702e-06, + "loss": 1.0876, + "step": 28747 + }, + { + "epoch": 6.27, + "learning_rate": 1.864977059209089e-06, + "loss": 0.4603, + "step": 28748 + }, + { + "epoch": 6.27, + "learning_rate": 1.8648678173476079e-06, + "loss": 1.8933, + "step": 28749 + }, + { + "epoch": 6.27, + "learning_rate": 1.8647585754861263e-06, + "loss": 1.0714, + "step": 28750 + }, + { + "epoch": 6.27, + "learning_rate": 1.8646493336246451e-06, + "loss": 2.6828, + "step": 28751 + }, + { + "epoch": 6.27, + "learning_rate": 1.8645400917631637e-06, + "loss": 1.0535, + "step": 28752 + }, + { + "epoch": 6.28, + "learning_rate": 1.8644308499016825e-06, + "loss": 1.615, + "step": 28753 + }, + { + "epoch": 6.28, + "learning_rate": 1.8643216080402014e-06, + "loss": 0.9331, + "step": 28754 + }, + { + "epoch": 6.28, + "learning_rate": 1.8642123661787198e-06, + "loss": 2.2009, + "step": 28755 + }, + { + "epoch": 6.28, + "learning_rate": 1.8641031243172386e-06, + "loss": 1.7831, + "step": 28756 + }, + { + "epoch": 6.28, + "learning_rate": 1.8639938824557572e-06, + "loss": 0.896, + "step": 28757 + }, + { + "epoch": 6.28, + "learning_rate": 1.863884640594276e-06, + "loss": 1.3755, + "step": 28758 + }, + { + "epoch": 6.28, + "learning_rate": 1.8637753987327944e-06, + "loss": 0.9694, + "step": 28759 + }, + { + "epoch": 6.28, + "learning_rate": 1.8636661568713132e-06, + "loss": 0.6036, + "step": 28760 + }, + { + "epoch": 6.28, + "learning_rate": 1.863556915009832e-06, + "loss": 1.9098, + "step": 28761 + }, + { + "epoch": 6.28, + "learning_rate": 1.8634476731483505e-06, + "loss": 1.1581, + "step": 28762 + }, + { + "epoch": 6.28, + "learning_rate": 1.8633384312868693e-06, + "loss": 1.6568, + "step": 28763 + }, + { + "epoch": 6.28, + "learning_rate": 1.863229189425388e-06, + "loss": 0.7066, + "step": 28764 + }, + { + "epoch": 6.28, + "learning_rate": 1.8631199475639067e-06, + "loss": 1.1124, + "step": 28765 + }, + { + "epoch": 6.28, + "learning_rate": 1.8630107057024251e-06, + "loss": 1.1973, + "step": 28766 + }, + { + "epoch": 6.28, + "learning_rate": 1.862901463840944e-06, + "loss": 1.1812, + "step": 28767 + }, + { + "epoch": 6.28, + "learning_rate": 1.8627922219794628e-06, + "loss": 1.3636, + "step": 28768 + }, + { + "epoch": 6.28, + "learning_rate": 1.8626829801179814e-06, + "loss": 0.6693, + "step": 28769 + }, + { + "epoch": 6.28, + "learning_rate": 1.8625737382565002e-06, + "loss": 0.7933, + "step": 28770 + }, + { + "epoch": 6.28, + "learning_rate": 1.8624644963950186e-06, + "loss": 2.9622, + "step": 28771 + }, + { + "epoch": 6.28, + "learning_rate": 1.8623552545335374e-06, + "loss": 0.9777, + "step": 28772 + }, + { + "epoch": 6.28, + "learning_rate": 1.862246012672056e-06, + "loss": 0.577, + "step": 28773 + }, + { + "epoch": 6.28, + "learning_rate": 1.8621367708105749e-06, + "loss": 0.8391, + "step": 28774 + }, + { + "epoch": 6.28, + "learning_rate": 1.8620275289490935e-06, + "loss": 1.5222, + "step": 28775 + }, + { + "epoch": 6.28, + "learning_rate": 1.861918287087612e-06, + "loss": 1.5299, + "step": 28776 + }, + { + "epoch": 6.28, + "learning_rate": 1.861809045226131e-06, + "loss": 4.129, + "step": 28777 + }, + { + "epoch": 6.28, + "learning_rate": 1.8616998033646493e-06, + "loss": 1.0394, + "step": 28778 + }, + { + "epoch": 6.28, + "learning_rate": 1.8615905615031681e-06, + "loss": 1.2045, + "step": 28779 + }, + { + "epoch": 6.28, + "learning_rate": 1.8614813196416867e-06, + "loss": 0.4639, + "step": 28780 + }, + { + "epoch": 6.28, + "learning_rate": 1.8613720777802056e-06, + "loss": 1.4872, + "step": 28781 + }, + { + "epoch": 6.28, + "learning_rate": 1.8612628359187244e-06, + "loss": 0.3838, + "step": 28782 + }, + { + "epoch": 6.28, + "learning_rate": 1.8611535940572428e-06, + "loss": 2.3639, + "step": 28783 + }, + { + "epoch": 6.28, + "learning_rate": 1.8610443521957616e-06, + "loss": 1.2139, + "step": 28784 + }, + { + "epoch": 6.28, + "learning_rate": 1.8609351103342802e-06, + "loss": 1.1231, + "step": 28785 + }, + { + "epoch": 6.28, + "learning_rate": 1.860825868472799e-06, + "loss": 0.6378, + "step": 28786 + }, + { + "epoch": 6.28, + "learning_rate": 1.8607166266113176e-06, + "loss": 1.1684, + "step": 28787 + }, + { + "epoch": 6.28, + "learning_rate": 1.8606073847498363e-06, + "loss": 1.3164, + "step": 28788 + }, + { + "epoch": 6.28, + "learning_rate": 1.860498142888355e-06, + "loss": 1.7164, + "step": 28789 + }, + { + "epoch": 6.28, + "learning_rate": 1.8603889010268735e-06, + "loss": 0.512, + "step": 28790 + }, + { + "epoch": 6.28, + "learning_rate": 1.8602796591653923e-06, + "loss": 0.4346, + "step": 28791 + }, + { + "epoch": 6.28, + "learning_rate": 1.860170417303911e-06, + "loss": 2.6496, + "step": 28792 + }, + { + "epoch": 6.28, + "learning_rate": 1.8600611754424297e-06, + "loss": 2.6244, + "step": 28793 + }, + { + "epoch": 6.28, + "learning_rate": 1.8599519335809486e-06, + "loss": 1.4158, + "step": 28794 + }, + { + "epoch": 6.28, + "learning_rate": 1.859842691719467e-06, + "loss": 1.4913, + "step": 28795 + }, + { + "epoch": 6.28, + "learning_rate": 1.8597334498579858e-06, + "loss": 1.3539, + "step": 28796 + }, + { + "epoch": 6.28, + "learning_rate": 1.8596242079965044e-06, + "loss": 1.4753, + "step": 28797 + }, + { + "epoch": 6.29, + "learning_rate": 1.8595149661350232e-06, + "loss": 1.5074, + "step": 28798 + }, + { + "epoch": 6.29, + "learning_rate": 1.8594057242735416e-06, + "loss": 0.8472, + "step": 28799 + }, + { + "epoch": 6.29, + "learning_rate": 1.8592964824120604e-06, + "loss": 2.884, + "step": 28800 + }, + { + "epoch": 6.29, + "learning_rate": 1.8591872405505793e-06, + "loss": 1.6536, + "step": 28801 + }, + { + "epoch": 6.29, + "learning_rate": 1.8590779986890979e-06, + "loss": 1.7976, + "step": 28802 + }, + { + "epoch": 6.29, + "learning_rate": 1.8589687568276165e-06, + "loss": 0.1273, + "step": 28803 + }, + { + "epoch": 6.29, + "learning_rate": 1.858859514966135e-06, + "loss": 1.4765, + "step": 28804 + }, + { + "epoch": 6.29, + "learning_rate": 1.858750273104654e-06, + "loss": 0.8978, + "step": 28805 + }, + { + "epoch": 6.29, + "learning_rate": 1.8586410312431723e-06, + "loss": 1.2957, + "step": 28806 + }, + { + "epoch": 6.29, + "learning_rate": 1.8585317893816911e-06, + "loss": 1.117, + "step": 28807 + }, + { + "epoch": 6.29, + "learning_rate": 1.85842254752021e-06, + "loss": 2.0048, + "step": 28808 + }, + { + "epoch": 6.29, + "learning_rate": 1.8583133056587286e-06, + "loss": 1.7951, + "step": 28809 + }, + { + "epoch": 6.29, + "learning_rate": 1.8582040637972474e-06, + "loss": 0.1662, + "step": 28810 + }, + { + "epoch": 6.29, + "learning_rate": 1.8580948219357658e-06, + "loss": 0.8262, + "step": 28811 + }, + { + "epoch": 6.29, + "learning_rate": 1.8579855800742846e-06, + "loss": 0.6335, + "step": 28812 + }, + { + "epoch": 6.29, + "learning_rate": 1.8578763382128032e-06, + "loss": 1.1864, + "step": 28813 + }, + { + "epoch": 6.29, + "learning_rate": 1.857767096351322e-06, + "loss": 1.3908, + "step": 28814 + }, + { + "epoch": 6.29, + "learning_rate": 1.8576578544898407e-06, + "loss": 1.0185, + "step": 28815 + }, + { + "epoch": 6.29, + "learning_rate": 1.8575486126283593e-06, + "loss": 0.6781, + "step": 28816 + }, + { + "epoch": 6.29, + "learning_rate": 1.857439370766878e-06, + "loss": 2.2298, + "step": 28817 + }, + { + "epoch": 6.29, + "learning_rate": 1.8573301289053967e-06, + "loss": 1.1243, + "step": 28818 + }, + { + "epoch": 6.29, + "learning_rate": 1.8572208870439153e-06, + "loss": 0.6918, + "step": 28819 + }, + { + "epoch": 6.29, + "learning_rate": 1.8571116451824341e-06, + "loss": 1.9947, + "step": 28820 + }, + { + "epoch": 6.29, + "learning_rate": 1.8570024033209527e-06, + "loss": 3.9219, + "step": 28821 + }, + { + "epoch": 6.29, + "learning_rate": 1.8568931614594716e-06, + "loss": 1.1605, + "step": 28822 + }, + { + "epoch": 6.29, + "learning_rate": 1.85678391959799e-06, + "loss": 1.0441, + "step": 28823 + }, + { + "epoch": 6.29, + "learning_rate": 1.8566746777365088e-06, + "loss": 1.4546, + "step": 28824 + }, + { + "epoch": 6.29, + "learning_rate": 1.8565654358750274e-06, + "loss": 1.9035, + "step": 28825 + }, + { + "epoch": 6.29, + "learning_rate": 1.8564561940135462e-06, + "loss": 0.9627, + "step": 28826 + }, + { + "epoch": 6.29, + "learning_rate": 1.856346952152065e-06, + "loss": 0.7592, + "step": 28827 + }, + { + "epoch": 6.29, + "learning_rate": 1.8562377102905834e-06, + "loss": 0.8267, + "step": 28828 + }, + { + "epoch": 6.29, + "learning_rate": 1.8561284684291023e-06, + "loss": 0.7226, + "step": 28829 + }, + { + "epoch": 6.29, + "learning_rate": 1.8560192265676209e-06, + "loss": 2.489, + "step": 28830 + }, + { + "epoch": 6.29, + "learning_rate": 1.8559099847061395e-06, + "loss": 1.7312, + "step": 28831 + }, + { + "epoch": 6.29, + "learning_rate": 1.855800742844658e-06, + "loss": 0.193, + "step": 28832 + }, + { + "epoch": 6.29, + "learning_rate": 1.855691500983177e-06, + "loss": 0.3652, + "step": 28833 + }, + { + "epoch": 6.29, + "learning_rate": 1.8555822591216958e-06, + "loss": 1.1326, + "step": 28834 + }, + { + "epoch": 6.29, + "learning_rate": 1.8554730172602141e-06, + "loss": 1.3245, + "step": 28835 + }, + { + "epoch": 6.29, + "learning_rate": 1.855363775398733e-06, + "loss": 1.5019, + "step": 28836 + }, + { + "epoch": 6.29, + "learning_rate": 1.8552545335372516e-06, + "loss": 2.5053, + "step": 28837 + }, + { + "epoch": 6.29, + "learning_rate": 1.8551452916757704e-06, + "loss": 2.0923, + "step": 28838 + }, + { + "epoch": 6.29, + "learning_rate": 1.8550360498142888e-06, + "loss": 1.5005, + "step": 28839 + }, + { + "epoch": 6.29, + "learning_rate": 1.8549268079528076e-06, + "loss": 0.796, + "step": 28840 + }, + { + "epoch": 6.29, + "learning_rate": 1.8548175660913265e-06, + "loss": 2.3321, + "step": 28841 + }, + { + "epoch": 6.29, + "learning_rate": 1.854708324229845e-06, + "loss": 1.0185, + "step": 28842 + }, + { + "epoch": 6.29, + "learning_rate": 1.8545990823683639e-06, + "loss": 0.969, + "step": 28843 + }, + { + "epoch": 6.3, + "learning_rate": 1.8544898405068823e-06, + "loss": 0.7503, + "step": 28844 + }, + { + "epoch": 6.3, + "learning_rate": 1.8543805986454011e-06, + "loss": 1.2137, + "step": 28845 + }, + { + "epoch": 6.3, + "learning_rate": 1.8542713567839197e-06, + "loss": 1.427, + "step": 28846 + }, + { + "epoch": 6.3, + "learning_rate": 1.8541621149224383e-06, + "loss": 2.5078, + "step": 28847 + }, + { + "epoch": 6.3, + "learning_rate": 1.8540528730609572e-06, + "loss": 2.1458, + "step": 28848 + }, + { + "epoch": 6.3, + "learning_rate": 1.8539436311994758e-06, + "loss": 1.0983, + "step": 28849 + }, + { + "epoch": 6.3, + "learning_rate": 1.8538343893379946e-06, + "loss": 1.0389, + "step": 28850 + }, + { + "epoch": 6.3, + "learning_rate": 1.853725147476513e-06, + "loss": 1.5038, + "step": 28851 + }, + { + "epoch": 6.3, + "learning_rate": 1.8536159056150318e-06, + "loss": 0.1231, + "step": 28852 + }, + { + "epoch": 6.3, + "learning_rate": 1.8535066637535506e-06, + "loss": 0.3527, + "step": 28853 + }, + { + "epoch": 6.3, + "learning_rate": 1.8533974218920692e-06, + "loss": 1.4471, + "step": 28854 + }, + { + "epoch": 6.3, + "learning_rate": 1.853288180030588e-06, + "loss": 1.459, + "step": 28855 + }, + { + "epoch": 6.3, + "learning_rate": 1.8531789381691065e-06, + "loss": 0.8898, + "step": 28856 + }, + { + "epoch": 6.3, + "learning_rate": 1.8530696963076253e-06, + "loss": 1.8587, + "step": 28857 + }, + { + "epoch": 6.3, + "learning_rate": 1.852960454446144e-06, + "loss": 2.4326, + "step": 28858 + }, + { + "epoch": 6.3, + "learning_rate": 1.8528512125846625e-06, + "loss": 0.9214, + "step": 28859 + }, + { + "epoch": 6.3, + "learning_rate": 1.8527419707231813e-06, + "loss": 1.633, + "step": 28860 + }, + { + "epoch": 6.3, + "learning_rate": 1.8526327288617e-06, + "loss": 1.1167, + "step": 28861 + }, + { + "epoch": 6.3, + "learning_rate": 1.8525234870002188e-06, + "loss": 1.0457, + "step": 28862 + }, + { + "epoch": 6.3, + "learning_rate": 1.8524142451387372e-06, + "loss": 1.2756, + "step": 28863 + }, + { + "epoch": 6.3, + "learning_rate": 1.852305003277256e-06, + "loss": 0.7861, + "step": 28864 + }, + { + "epoch": 6.3, + "learning_rate": 1.8521957614157746e-06, + "loss": 3.2943, + "step": 28865 + }, + { + "epoch": 6.3, + "learning_rate": 1.8520865195542934e-06, + "loss": 0.4247, + "step": 28866 + }, + { + "epoch": 6.3, + "learning_rate": 1.8519772776928122e-06, + "loss": 1.1998, + "step": 28867 + }, + { + "epoch": 6.3, + "learning_rate": 1.8518680358313306e-06, + "loss": 0.9087, + "step": 28868 + }, + { + "epoch": 6.3, + "learning_rate": 1.8517587939698495e-06, + "loss": 0.9351, + "step": 28869 + }, + { + "epoch": 6.3, + "learning_rate": 1.851649552108368e-06, + "loss": 1.7442, + "step": 28870 + }, + { + "epoch": 6.3, + "learning_rate": 1.851540310246887e-06, + "loss": 0.5051, + "step": 28871 + }, + { + "epoch": 6.3, + "learning_rate": 1.8514310683854053e-06, + "loss": 1.5027, + "step": 28872 + }, + { + "epoch": 6.3, + "learning_rate": 1.8513218265239241e-06, + "loss": 1.7484, + "step": 28873 + }, + { + "epoch": 6.3, + "learning_rate": 1.851212584662443e-06, + "loss": 2.0953, + "step": 28874 + }, + { + "epoch": 6.3, + "learning_rate": 1.8511033428009613e-06, + "loss": 0.6247, + "step": 28875 + }, + { + "epoch": 6.3, + "learning_rate": 1.8509941009394802e-06, + "loss": 2.2278, + "step": 28876 + }, + { + "epoch": 6.3, + "learning_rate": 1.8508848590779988e-06, + "loss": 1.8599, + "step": 28877 + }, + { + "epoch": 6.3, + "learning_rate": 1.8507756172165176e-06, + "loss": 1.4905, + "step": 28878 + }, + { + "epoch": 6.3, + "learning_rate": 1.850666375355036e-06, + "loss": 0.9636, + "step": 28879 + }, + { + "epoch": 6.3, + "learning_rate": 1.8505571334935548e-06, + "loss": 2.2038, + "step": 28880 + }, + { + "epoch": 6.3, + "learning_rate": 1.8504478916320736e-06, + "loss": 2.0201, + "step": 28881 + }, + { + "epoch": 6.3, + "learning_rate": 1.8503386497705923e-06, + "loss": 1.3336, + "step": 28882 + }, + { + "epoch": 6.3, + "learning_rate": 1.850229407909111e-06, + "loss": 0.8993, + "step": 28883 + }, + { + "epoch": 6.3, + "learning_rate": 1.8501201660476295e-06, + "loss": 0.1236, + "step": 28884 + }, + { + "epoch": 6.3, + "learning_rate": 1.8500109241861483e-06, + "loss": 2.0017, + "step": 28885 + }, + { + "epoch": 6.3, + "learning_rate": 1.849901682324667e-06, + "loss": 2.223, + "step": 28886 + }, + { + "epoch": 6.3, + "learning_rate": 1.8497924404631855e-06, + "loss": 2.1289, + "step": 28887 + }, + { + "epoch": 6.3, + "learning_rate": 1.8496831986017043e-06, + "loss": 1.5504, + "step": 28888 + }, + { + "epoch": 6.3, + "learning_rate": 1.849573956740223e-06, + "loss": 1.0322, + "step": 28889 + }, + { + "epoch": 6.31, + "learning_rate": 1.8494647148787418e-06, + "loss": 1.808, + "step": 28890 + }, + { + "epoch": 6.31, + "learning_rate": 1.8493554730172602e-06, + "loss": 0.4787, + "step": 28891 + }, + { + "epoch": 6.31, + "learning_rate": 1.849246231155779e-06, + "loss": 1.2608, + "step": 28892 + }, + { + "epoch": 6.31, + "learning_rate": 1.8491369892942978e-06, + "loss": 0.6074, + "step": 28893 + }, + { + "epoch": 6.31, + "learning_rate": 1.8490277474328164e-06, + "loss": 1.5498, + "step": 28894 + }, + { + "epoch": 6.31, + "learning_rate": 1.8489185055713353e-06, + "loss": 1.0186, + "step": 28895 + }, + { + "epoch": 6.31, + "learning_rate": 1.8488092637098537e-06, + "loss": 4.1233, + "step": 28896 + }, + { + "epoch": 6.31, + "learning_rate": 1.8487000218483725e-06, + "loss": 1.8885, + "step": 28897 + }, + { + "epoch": 6.31, + "learning_rate": 1.848590779986891e-06, + "loss": 0.6239, + "step": 28898 + }, + { + "epoch": 6.31, + "learning_rate": 1.84848153812541e-06, + "loss": 0.1105, + "step": 28899 + }, + { + "epoch": 6.31, + "learning_rate": 1.8483722962639285e-06, + "loss": 0.901, + "step": 28900 + }, + { + "epoch": 6.31, + "learning_rate": 1.8482630544024471e-06, + "loss": 1.563, + "step": 28901 + }, + { + "epoch": 6.31, + "learning_rate": 1.848153812540966e-06, + "loss": 1.1512, + "step": 28902 + }, + { + "epoch": 6.31, + "learning_rate": 1.8480445706794844e-06, + "loss": 1.6636, + "step": 28903 + }, + { + "epoch": 6.31, + "learning_rate": 1.8479353288180032e-06, + "loss": 3.013, + "step": 28904 + }, + { + "epoch": 6.31, + "learning_rate": 1.8478260869565218e-06, + "loss": 1.7701, + "step": 28905 + }, + { + "epoch": 6.31, + "learning_rate": 1.8477168450950406e-06, + "loss": 1.2479, + "step": 28906 + }, + { + "epoch": 6.31, + "learning_rate": 1.8476076032335594e-06, + "loss": 0.5508, + "step": 28907 + }, + { + "epoch": 6.31, + "learning_rate": 1.8474983613720778e-06, + "loss": 0.8077, + "step": 28908 + }, + { + "epoch": 6.31, + "learning_rate": 1.8473891195105967e-06, + "loss": 1.1898, + "step": 28909 + }, + { + "epoch": 6.31, + "learning_rate": 1.8472798776491153e-06, + "loss": 0.2429, + "step": 28910 + }, + { + "epoch": 6.31, + "learning_rate": 1.847170635787634e-06, + "loss": 1.6144, + "step": 28911 + }, + { + "epoch": 6.31, + "learning_rate": 1.8470613939261525e-06, + "loss": 0.6145, + "step": 28912 + }, + { + "epoch": 6.31, + "learning_rate": 1.8469521520646713e-06, + "loss": 3.338, + "step": 28913 + }, + { + "epoch": 6.31, + "learning_rate": 1.8468429102031901e-06, + "loss": 2.3118, + "step": 28914 + }, + { + "epoch": 6.31, + "learning_rate": 1.8467336683417087e-06, + "loss": 1.0917, + "step": 28915 + }, + { + "epoch": 6.31, + "learning_rate": 1.8466244264802274e-06, + "loss": 1.3221, + "step": 28916 + }, + { + "epoch": 6.31, + "learning_rate": 1.846515184618746e-06, + "loss": 0.8042, + "step": 28917 + }, + { + "epoch": 6.31, + "learning_rate": 1.8464059427572648e-06, + "loss": 0.5216, + "step": 28918 + }, + { + "epoch": 6.31, + "learning_rate": 1.8462967008957832e-06, + "loss": 0.5626, + "step": 28919 + }, + { + "epoch": 6.31, + "learning_rate": 1.846187459034302e-06, + "loss": 1.1835, + "step": 28920 + }, + { + "epoch": 6.31, + "learning_rate": 1.8460782171728208e-06, + "loss": 1.8759, + "step": 28921 + }, + { + "epoch": 6.31, + "learning_rate": 1.8459689753113394e-06, + "loss": 1.8573, + "step": 28922 + }, + { + "epoch": 6.31, + "learning_rate": 1.8458597334498583e-06, + "loss": 0.6561, + "step": 28923 + }, + { + "epoch": 6.31, + "learning_rate": 1.8457504915883767e-06, + "loss": 1.2407, + "step": 28924 + }, + { + "epoch": 6.31, + "learning_rate": 1.8456412497268955e-06, + "loss": 0.398, + "step": 28925 + }, + { + "epoch": 6.31, + "learning_rate": 1.8455320078654143e-06, + "loss": 1.6432, + "step": 28926 + }, + { + "epoch": 6.31, + "learning_rate": 1.845422766003933e-06, + "loss": 1.9623, + "step": 28927 + }, + { + "epoch": 6.31, + "learning_rate": 1.8453135241424515e-06, + "loss": 1.2218, + "step": 28928 + }, + { + "epoch": 6.31, + "learning_rate": 1.8452042822809701e-06, + "loss": 1.2587, + "step": 28929 + }, + { + "epoch": 6.31, + "learning_rate": 1.845095040419489e-06, + "loss": 0.3321, + "step": 28930 + }, + { + "epoch": 6.31, + "learning_rate": 1.8449857985580074e-06, + "loss": 0.9278, + "step": 28931 + }, + { + "epoch": 6.31, + "learning_rate": 1.8448765566965262e-06, + "loss": 3.3695, + "step": 28932 + }, + { + "epoch": 6.31, + "learning_rate": 1.844767314835045e-06, + "loss": 2.4015, + "step": 28933 + }, + { + "epoch": 6.31, + "learning_rate": 1.8446580729735636e-06, + "loss": 0.7692, + "step": 28934 + }, + { + "epoch": 6.31, + "learning_rate": 1.8445488311120824e-06, + "loss": 0.3395, + "step": 28935 + }, + { + "epoch": 6.32, + "learning_rate": 1.8444395892506008e-06, + "loss": 1.594, + "step": 28936 + }, + { + "epoch": 6.32, + "learning_rate": 1.8443303473891197e-06, + "loss": 0.8441, + "step": 28937 + }, + { + "epoch": 6.32, + "learning_rate": 1.8442211055276383e-06, + "loss": 1.3678, + "step": 28938 + }, + { + "epoch": 6.32, + "learning_rate": 1.844111863666157e-06, + "loss": 1.1903, + "step": 28939 + }, + { + "epoch": 6.32, + "learning_rate": 1.8440026218046757e-06, + "loss": 2.4177, + "step": 28940 + }, + { + "epoch": 6.32, + "learning_rate": 1.8438933799431943e-06, + "loss": 1.4786, + "step": 28941 + }, + { + "epoch": 6.32, + "learning_rate": 1.8437841380817131e-06, + "loss": 0.6197, + "step": 28942 + }, + { + "epoch": 6.32, + "learning_rate": 1.8436748962202318e-06, + "loss": 2.6892, + "step": 28943 + }, + { + "epoch": 6.32, + "learning_rate": 1.8435656543587504e-06, + "loss": 2.7306, + "step": 28944 + }, + { + "epoch": 6.32, + "learning_rate": 1.843456412497269e-06, + "loss": 1.5229, + "step": 28945 + }, + { + "epoch": 6.32, + "learning_rate": 1.8433471706357878e-06, + "loss": 2.5927, + "step": 28946 + }, + { + "epoch": 6.32, + "learning_rate": 1.8432379287743066e-06, + "loss": 0.6291, + "step": 28947 + }, + { + "epoch": 6.32, + "learning_rate": 1.843128686912825e-06, + "loss": 1.7876, + "step": 28948 + }, + { + "epoch": 6.32, + "learning_rate": 1.8430194450513438e-06, + "loss": 0.8897, + "step": 28949 + }, + { + "epoch": 6.32, + "learning_rate": 1.8429102031898625e-06, + "loss": 0.7602, + "step": 28950 + }, + { + "epoch": 6.32, + "learning_rate": 1.8428009613283813e-06, + "loss": 1.6789, + "step": 28951 + }, + { + "epoch": 6.32, + "learning_rate": 1.8426917194668997e-06, + "loss": 1.4922, + "step": 28952 + }, + { + "epoch": 6.32, + "learning_rate": 1.8425824776054185e-06, + "loss": 0.2368, + "step": 28953 + }, + { + "epoch": 6.32, + "learning_rate": 1.8424732357439373e-06, + "loss": 1.5654, + "step": 28954 + }, + { + "epoch": 6.32, + "learning_rate": 1.842363993882456e-06, + "loss": 0.7427, + "step": 28955 + }, + { + "epoch": 6.32, + "learning_rate": 1.8422547520209745e-06, + "loss": 2.9704, + "step": 28956 + }, + { + "epoch": 6.32, + "learning_rate": 1.8421455101594932e-06, + "loss": 2.0326, + "step": 28957 + }, + { + "epoch": 6.32, + "learning_rate": 1.842036268298012e-06, + "loss": 1.7177, + "step": 28958 + }, + { + "epoch": 6.32, + "learning_rate": 1.8419270264365308e-06, + "loss": 2.485, + "step": 28959 + }, + { + "epoch": 6.32, + "learning_rate": 1.8418177845750492e-06, + "loss": 1.0949, + "step": 28960 + }, + { + "epoch": 6.32, + "learning_rate": 1.841708542713568e-06, + "loss": 0.9065, + "step": 28961 + }, + { + "epoch": 6.32, + "learning_rate": 1.8415993008520866e-06, + "loss": 1.5163, + "step": 28962 + }, + { + "epoch": 6.32, + "learning_rate": 1.8414900589906055e-06, + "loss": 1.7093, + "step": 28963 + }, + { + "epoch": 6.32, + "learning_rate": 1.8413808171291239e-06, + "loss": 0.8213, + "step": 28964 + }, + { + "epoch": 6.32, + "learning_rate": 1.8412715752676427e-06, + "loss": 1.1334, + "step": 28965 + }, + { + "epoch": 6.32, + "learning_rate": 1.8411623334061615e-06, + "loss": 0.533, + "step": 28966 + }, + { + "epoch": 6.32, + "learning_rate": 1.8410530915446801e-06, + "loss": 1.7876, + "step": 28967 + }, + { + "epoch": 6.32, + "learning_rate": 1.840943849683199e-06, + "loss": 0.4974, + "step": 28968 + }, + { + "epoch": 6.32, + "learning_rate": 1.8408346078217173e-06, + "loss": 1.1226, + "step": 28969 + }, + { + "epoch": 6.32, + "learning_rate": 1.8407253659602362e-06, + "loss": 1.7294, + "step": 28970 + }, + { + "epoch": 6.32, + "learning_rate": 1.8406161240987548e-06, + "loss": 0.8609, + "step": 28971 + }, + { + "epoch": 6.32, + "learning_rate": 1.8405068822372734e-06, + "loss": 1.388, + "step": 28972 + }, + { + "epoch": 6.32, + "learning_rate": 1.8403976403757922e-06, + "loss": 1.8884, + "step": 28973 + }, + { + "epoch": 6.32, + "learning_rate": 1.8402883985143108e-06, + "loss": 0.7704, + "step": 28974 + }, + { + "epoch": 6.32, + "learning_rate": 1.8401791566528296e-06, + "loss": 0.9465, + "step": 28975 + }, + { + "epoch": 6.32, + "learning_rate": 1.840069914791348e-06, + "loss": 1.1267, + "step": 28976 + }, + { + "epoch": 6.32, + "learning_rate": 1.8399606729298669e-06, + "loss": 0.9902, + "step": 28977 + }, + { + "epoch": 6.32, + "learning_rate": 1.8398514310683855e-06, + "loss": 1.4335, + "step": 28978 + }, + { + "epoch": 6.32, + "learning_rate": 1.8397421892069043e-06, + "loss": 2.5094, + "step": 28979 + }, + { + "epoch": 6.32, + "learning_rate": 1.8396329473454231e-06, + "loss": 1.9809, + "step": 28980 + }, + { + "epoch": 6.32, + "learning_rate": 1.8395237054839415e-06, + "loss": 0.2642, + "step": 28981 + }, + { + "epoch": 6.33, + "learning_rate": 1.8394144636224603e-06, + "loss": 1.0345, + "step": 28982 + }, + { + "epoch": 6.33, + "learning_rate": 1.839305221760979e-06, + "loss": 0.444, + "step": 28983 + }, + { + "epoch": 6.33, + "learning_rate": 1.8391959798994976e-06, + "loss": 1.7175, + "step": 28984 + }, + { + "epoch": 6.33, + "learning_rate": 1.8390867380380162e-06, + "loss": 2.0639, + "step": 28985 + }, + { + "epoch": 6.33, + "learning_rate": 1.838977496176535e-06, + "loss": 0.5145, + "step": 28986 + }, + { + "epoch": 6.33, + "learning_rate": 1.8388682543150538e-06, + "loss": 0.9968, + "step": 28987 + }, + { + "epoch": 6.33, + "learning_rate": 1.8387590124535722e-06, + "loss": 1.2431, + "step": 28988 + }, + { + "epoch": 6.33, + "learning_rate": 1.838649770592091e-06, + "loss": 1.5036, + "step": 28989 + }, + { + "epoch": 6.33, + "learning_rate": 1.8385405287306097e-06, + "loss": 1.3254, + "step": 28990 + }, + { + "epoch": 6.33, + "learning_rate": 1.8384312868691285e-06, + "loss": 0.8659, + "step": 28991 + }, + { + "epoch": 6.33, + "learning_rate": 1.8383220450076473e-06, + "loss": 0.5923, + "step": 28992 + }, + { + "epoch": 6.33, + "learning_rate": 1.8382128031461657e-06, + "loss": 0.4333, + "step": 28993 + }, + { + "epoch": 6.33, + "learning_rate": 1.8381035612846845e-06, + "loss": 1.1948, + "step": 28994 + }, + { + "epoch": 6.33, + "learning_rate": 1.8379943194232031e-06, + "loss": 1.7935, + "step": 28995 + }, + { + "epoch": 6.33, + "learning_rate": 1.837885077561722e-06, + "loss": 1.424, + "step": 28996 + }, + { + "epoch": 6.33, + "learning_rate": 1.8377758357002404e-06, + "loss": 1.177, + "step": 28997 + }, + { + "epoch": 6.33, + "learning_rate": 1.8376665938387592e-06, + "loss": 1.8155, + "step": 28998 + }, + { + "epoch": 6.33, + "learning_rate": 1.837557351977278e-06, + "loss": 1.3965, + "step": 28999 + }, + { + "epoch": 6.33, + "learning_rate": 1.8374481101157964e-06, + "loss": 1.1008, + "step": 29000 + }, + { + "epoch": 6.33, + "learning_rate": 1.8373388682543152e-06, + "loss": 1.0786, + "step": 29001 + }, + { + "epoch": 6.33, + "learning_rate": 1.8372296263928338e-06, + "loss": 0.2294, + "step": 29002 + }, + { + "epoch": 6.33, + "learning_rate": 1.8371203845313527e-06, + "loss": 0.7624, + "step": 29003 + }, + { + "epoch": 6.33, + "learning_rate": 1.837011142669871e-06, + "loss": 0.9666, + "step": 29004 + }, + { + "epoch": 6.33, + "learning_rate": 1.8369019008083899e-06, + "loss": 0.82, + "step": 29005 + }, + { + "epoch": 6.33, + "learning_rate": 1.8367926589469087e-06, + "loss": 0.5622, + "step": 29006 + }, + { + "epoch": 6.33, + "learning_rate": 1.8366834170854273e-06, + "loss": 1.7179, + "step": 29007 + }, + { + "epoch": 6.33, + "learning_rate": 1.8365741752239461e-06, + "loss": 0.982, + "step": 29008 + }, + { + "epoch": 6.33, + "learning_rate": 1.8364649333624645e-06, + "loss": 1.0035, + "step": 29009 + }, + { + "epoch": 6.33, + "learning_rate": 1.8363556915009834e-06, + "loss": 1.2668, + "step": 29010 + }, + { + "epoch": 6.33, + "learning_rate": 1.836246449639502e-06, + "loss": 1.289, + "step": 29011 + }, + { + "epoch": 6.33, + "learning_rate": 1.8361372077780206e-06, + "loss": 1.004, + "step": 29012 + }, + { + "epoch": 6.33, + "learning_rate": 1.8360279659165394e-06, + "loss": 2.0747, + "step": 29013 + }, + { + "epoch": 6.33, + "learning_rate": 1.835918724055058e-06, + "loss": 0.7918, + "step": 29014 + }, + { + "epoch": 6.33, + "learning_rate": 1.8358094821935768e-06, + "loss": 1.2119, + "step": 29015 + }, + { + "epoch": 6.33, + "learning_rate": 1.8357002403320952e-06, + "loss": 2.2527, + "step": 29016 + }, + { + "epoch": 6.33, + "learning_rate": 1.835590998470614e-06, + "loss": 1.8836, + "step": 29017 + }, + { + "epoch": 6.33, + "learning_rate": 1.8354817566091327e-06, + "loss": 0.7447, + "step": 29018 + }, + { + "epoch": 6.33, + "learning_rate": 1.8353725147476515e-06, + "loss": 1.0907, + "step": 29019 + }, + { + "epoch": 6.33, + "learning_rate": 1.8352632728861703e-06, + "loss": 0.6041, + "step": 29020 + }, + { + "epoch": 6.33, + "learning_rate": 1.8351540310246887e-06, + "loss": 1.5359, + "step": 29021 + }, + { + "epoch": 6.33, + "learning_rate": 1.8350447891632075e-06, + "loss": 3.7165, + "step": 29022 + }, + { + "epoch": 6.33, + "learning_rate": 1.8349355473017261e-06, + "loss": 0.535, + "step": 29023 + }, + { + "epoch": 6.33, + "learning_rate": 1.834826305440245e-06, + "loss": 2.2189, + "step": 29024 + }, + { + "epoch": 6.33, + "learning_rate": 1.8347170635787636e-06, + "loss": 1.4165, + "step": 29025 + }, + { + "epoch": 6.33, + "learning_rate": 1.8346078217172822e-06, + "loss": 1.3401, + "step": 29026 + }, + { + "epoch": 6.34, + "learning_rate": 1.834498579855801e-06, + "loss": 2.577, + "step": 29027 + }, + { + "epoch": 6.34, + "learning_rate": 1.8343893379943194e-06, + "loss": 0.7386, + "step": 29028 + }, + { + "epoch": 6.34, + "learning_rate": 1.8342800961328382e-06, + "loss": 0.5146, + "step": 29029 + }, + { + "epoch": 6.34, + "learning_rate": 1.8341708542713568e-06, + "loss": 1.5383, + "step": 29030 + }, + { + "epoch": 6.34, + "learning_rate": 1.8340616124098757e-06, + "loss": 1.5166, + "step": 29031 + }, + { + "epoch": 6.34, + "learning_rate": 1.8339523705483945e-06, + "loss": 0.3767, + "step": 29032 + }, + { + "epoch": 6.34, + "learning_rate": 1.8338431286869129e-06, + "loss": 1.1601, + "step": 29033 + }, + { + "epoch": 6.34, + "learning_rate": 1.8337338868254317e-06, + "loss": 1.4022, + "step": 29034 + }, + { + "epoch": 6.34, + "learning_rate": 1.8336246449639503e-06, + "loss": 0.8671, + "step": 29035 + }, + { + "epoch": 6.34, + "learning_rate": 1.8335154031024691e-06, + "loss": 1.9611, + "step": 29036 + }, + { + "epoch": 6.34, + "learning_rate": 1.8334061612409875e-06, + "loss": 1.353, + "step": 29037 + }, + { + "epoch": 6.34, + "learning_rate": 1.8332969193795064e-06, + "loss": 1.1127, + "step": 29038 + }, + { + "epoch": 6.34, + "learning_rate": 1.8331876775180252e-06, + "loss": 1.7615, + "step": 29039 + }, + { + "epoch": 6.34, + "learning_rate": 1.8330784356565438e-06, + "loss": 0.4989, + "step": 29040 + }, + { + "epoch": 6.34, + "learning_rate": 1.8329691937950624e-06, + "loss": 1.5608, + "step": 29041 + }, + { + "epoch": 6.34, + "learning_rate": 1.832859951933581e-06, + "loss": 0.9342, + "step": 29042 + }, + { + "epoch": 6.34, + "learning_rate": 1.8327507100720998e-06, + "loss": 1.0698, + "step": 29043 + }, + { + "epoch": 6.34, + "learning_rate": 1.8326414682106182e-06, + "loss": 0.1791, + "step": 29044 + }, + { + "epoch": 6.34, + "learning_rate": 1.832532226349137e-06, + "loss": 0.338, + "step": 29045 + }, + { + "epoch": 6.34, + "learning_rate": 1.8324229844876559e-06, + "loss": 2.67, + "step": 29046 + }, + { + "epoch": 6.34, + "learning_rate": 1.8323137426261745e-06, + "loss": 1.7073, + "step": 29047 + }, + { + "epoch": 6.34, + "learning_rate": 1.8322045007646933e-06, + "loss": 1.6984, + "step": 29048 + }, + { + "epoch": 6.34, + "learning_rate": 1.8320952589032117e-06, + "loss": 0.8773, + "step": 29049 + }, + { + "epoch": 6.34, + "learning_rate": 1.8319860170417305e-06, + "loss": 0.5795, + "step": 29050 + }, + { + "epoch": 6.34, + "learning_rate": 1.8318767751802492e-06, + "loss": 1.0809, + "step": 29051 + }, + { + "epoch": 6.34, + "learning_rate": 1.831767533318768e-06, + "loss": 1.0666, + "step": 29052 + }, + { + "epoch": 6.34, + "learning_rate": 1.8316582914572866e-06, + "loss": 1.8497, + "step": 29053 + }, + { + "epoch": 6.34, + "learning_rate": 1.8315490495958052e-06, + "loss": 2.4419, + "step": 29054 + }, + { + "epoch": 6.34, + "learning_rate": 1.831439807734324e-06, + "loss": 2.6846, + "step": 29055 + }, + { + "epoch": 6.34, + "learning_rate": 1.8313305658728424e-06, + "loss": 0.2095, + "step": 29056 + }, + { + "epoch": 6.34, + "learning_rate": 1.8312213240113612e-06, + "loss": 0.6457, + "step": 29057 + }, + { + "epoch": 6.34, + "learning_rate": 1.83111208214988e-06, + "loss": 0.7758, + "step": 29058 + }, + { + "epoch": 6.34, + "learning_rate": 1.8310028402883987e-06, + "loss": 1.1502, + "step": 29059 + }, + { + "epoch": 6.34, + "learning_rate": 1.8308935984269175e-06, + "loss": 2.1986, + "step": 29060 + }, + { + "epoch": 6.34, + "learning_rate": 1.830784356565436e-06, + "loss": 1.0777, + "step": 29061 + }, + { + "epoch": 6.34, + "learning_rate": 1.8306751147039547e-06, + "loss": 0.8301, + "step": 29062 + }, + { + "epoch": 6.34, + "learning_rate": 1.8305658728424733e-06, + "loss": 0.3502, + "step": 29063 + }, + { + "epoch": 6.34, + "learning_rate": 1.8304566309809922e-06, + "loss": 1.7105, + "step": 29064 + }, + { + "epoch": 6.34, + "learning_rate": 1.8303473891195108e-06, + "loss": 1.3911, + "step": 29065 + }, + { + "epoch": 6.34, + "learning_rate": 1.8302381472580294e-06, + "loss": 0.6067, + "step": 29066 + }, + { + "epoch": 6.34, + "learning_rate": 1.8301289053965482e-06, + "loss": 1.5197, + "step": 29067 + }, + { + "epoch": 6.34, + "learning_rate": 1.8300196635350668e-06, + "loss": 3.3572, + "step": 29068 + }, + { + "epoch": 6.34, + "learning_rate": 1.8299104216735854e-06, + "loss": 0.6718, + "step": 29069 + }, + { + "epoch": 6.34, + "learning_rate": 1.829801179812104e-06, + "loss": 1.1423, + "step": 29070 + }, + { + "epoch": 6.34, + "learning_rate": 1.8296919379506229e-06, + "loss": 1.4564, + "step": 29071 + }, + { + "epoch": 6.34, + "learning_rate": 1.8295826960891417e-06, + "loss": 1.9817, + "step": 29072 + }, + { + "epoch": 6.35, + "learning_rate": 1.82947345422766e-06, + "loss": 0.8844, + "step": 29073 + }, + { + "epoch": 6.35, + "learning_rate": 1.829364212366179e-06, + "loss": 0.6014, + "step": 29074 + }, + { + "epoch": 6.35, + "learning_rate": 1.8292549705046975e-06, + "loss": 0.6174, + "step": 29075 + }, + { + "epoch": 6.35, + "learning_rate": 1.8291457286432163e-06, + "loss": 1.0686, + "step": 29076 + }, + { + "epoch": 6.35, + "learning_rate": 1.8290364867817347e-06, + "loss": 1.453, + "step": 29077 + }, + { + "epoch": 6.35, + "learning_rate": 1.8289272449202536e-06, + "loss": 2.191, + "step": 29078 + }, + { + "epoch": 6.35, + "learning_rate": 1.8288180030587724e-06, + "loss": 1.8824, + "step": 29079 + }, + { + "epoch": 6.35, + "learning_rate": 1.828708761197291e-06, + "loss": 1.321, + "step": 29080 + }, + { + "epoch": 6.35, + "learning_rate": 1.8285995193358096e-06, + "loss": 0.9449, + "step": 29081 + }, + { + "epoch": 6.35, + "learning_rate": 1.8284902774743282e-06, + "loss": 0.9647, + "step": 29082 + }, + { + "epoch": 6.35, + "learning_rate": 1.828381035612847e-06, + "loss": 0.334, + "step": 29083 + }, + { + "epoch": 6.35, + "learning_rate": 1.8282717937513654e-06, + "loss": 0.4498, + "step": 29084 + }, + { + "epoch": 6.35, + "learning_rate": 1.8281625518898843e-06, + "loss": 1.1449, + "step": 29085 + }, + { + "epoch": 6.35, + "learning_rate": 1.828053310028403e-06, + "loss": 0.8803, + "step": 29086 + }, + { + "epoch": 6.35, + "learning_rate": 1.8279440681669217e-06, + "loss": 2.4651, + "step": 29087 + }, + { + "epoch": 6.35, + "learning_rate": 1.8278348263054405e-06, + "loss": 0.355, + "step": 29088 + }, + { + "epoch": 6.35, + "learning_rate": 1.827725584443959e-06, + "loss": 0.6273, + "step": 29089 + }, + { + "epoch": 6.35, + "learning_rate": 1.8276163425824777e-06, + "loss": 0.5834, + "step": 29090 + }, + { + "epoch": 6.35, + "learning_rate": 1.8275071007209963e-06, + "loss": 1.2273, + "step": 29091 + }, + { + "epoch": 6.35, + "learning_rate": 1.8273978588595152e-06, + "loss": 0.5933, + "step": 29092 + }, + { + "epoch": 6.35, + "learning_rate": 1.827288616998034e-06, + "loss": 0.4162, + "step": 29093 + }, + { + "epoch": 6.35, + "learning_rate": 1.8271793751365524e-06, + "loss": 1.327, + "step": 29094 + }, + { + "epoch": 6.35, + "learning_rate": 1.8270701332750712e-06, + "loss": 0.3704, + "step": 29095 + }, + { + "epoch": 6.35, + "learning_rate": 1.8269608914135898e-06, + "loss": 2.598, + "step": 29096 + }, + { + "epoch": 6.35, + "learning_rate": 1.8268516495521084e-06, + "loss": 1.5101, + "step": 29097 + }, + { + "epoch": 6.35, + "learning_rate": 1.8267424076906273e-06, + "loss": 1.1356, + "step": 29098 + }, + { + "epoch": 6.35, + "learning_rate": 1.8266331658291459e-06, + "loss": 0.8703, + "step": 29099 + }, + { + "epoch": 6.35, + "learning_rate": 1.8265239239676647e-06, + "loss": 1.4204, + "step": 29100 + }, + { + "epoch": 6.35, + "learning_rate": 1.826414682106183e-06, + "loss": 1.2982, + "step": 29101 + }, + { + "epoch": 6.35, + "learning_rate": 1.826305440244702e-06, + "loss": 0.2889, + "step": 29102 + }, + { + "epoch": 6.35, + "learning_rate": 1.8261961983832205e-06, + "loss": 0.1531, + "step": 29103 + }, + { + "epoch": 6.35, + "learning_rate": 1.8260869565217394e-06, + "loss": 0.1351, + "step": 29104 + }, + { + "epoch": 6.35, + "learning_rate": 1.8259777146602582e-06, + "loss": 1.4588, + "step": 29105 + }, + { + "epoch": 6.35, + "learning_rate": 1.8258684727987766e-06, + "loss": 0.7504, + "step": 29106 + }, + { + "epoch": 6.35, + "learning_rate": 1.8257592309372954e-06, + "loss": 0.6128, + "step": 29107 + }, + { + "epoch": 6.35, + "learning_rate": 1.825649989075814e-06, + "loss": 2.2053, + "step": 29108 + }, + { + "epoch": 6.35, + "learning_rate": 1.8255407472143326e-06, + "loss": 2.3949, + "step": 29109 + }, + { + "epoch": 6.35, + "learning_rate": 1.8254315053528512e-06, + "loss": 1.1446, + "step": 29110 + }, + { + "epoch": 6.35, + "learning_rate": 1.82532226349137e-06, + "loss": 0.3267, + "step": 29111 + }, + { + "epoch": 6.35, + "learning_rate": 1.8252130216298889e-06, + "loss": 2.0856, + "step": 29112 + }, + { + "epoch": 6.35, + "learning_rate": 1.8251037797684073e-06, + "loss": 1.251, + "step": 29113 + }, + { + "epoch": 6.35, + "learning_rate": 1.824994537906926e-06, + "loss": 1.4582, + "step": 29114 + }, + { + "epoch": 6.35, + "learning_rate": 1.8248852960454447e-06, + "loss": 2.1048, + "step": 29115 + }, + { + "epoch": 6.35, + "learning_rate": 1.8247760541839635e-06, + "loss": 0.4039, + "step": 29116 + }, + { + "epoch": 6.35, + "learning_rate": 1.824666812322482e-06, + "loss": 1.5158, + "step": 29117 + }, + { + "epoch": 6.35, + "learning_rate": 1.8245575704610008e-06, + "loss": 1.6035, + "step": 29118 + }, + { + "epoch": 6.36, + "learning_rate": 1.8244483285995196e-06, + "loss": 2.0798, + "step": 29119 + }, + { + "epoch": 6.36, + "learning_rate": 1.8243390867380382e-06, + "loss": 1.2675, + "step": 29120 + }, + { + "epoch": 6.36, + "learning_rate": 1.824229844876557e-06, + "loss": 0.8666, + "step": 29121 + }, + { + "epoch": 6.36, + "learning_rate": 1.8241206030150754e-06, + "loss": 2.4593, + "step": 29122 + }, + { + "epoch": 6.36, + "learning_rate": 1.8240113611535942e-06, + "loss": 0.1913, + "step": 29123 + }, + { + "epoch": 6.36, + "learning_rate": 1.8239021192921128e-06, + "loss": 1.7028, + "step": 29124 + }, + { + "epoch": 6.36, + "learning_rate": 1.8237928774306315e-06, + "loss": 0.8082, + "step": 29125 + }, + { + "epoch": 6.36, + "learning_rate": 1.8236836355691503e-06, + "loss": 1.1079, + "step": 29126 + }, + { + "epoch": 6.36, + "learning_rate": 1.8235743937076689e-06, + "loss": 1.7946, + "step": 29127 + }, + { + "epoch": 6.36, + "learning_rate": 1.8234651518461877e-06, + "loss": 0.1475, + "step": 29128 + }, + { + "epoch": 6.36, + "learning_rate": 1.8233559099847061e-06, + "loss": 1.6916, + "step": 29129 + }, + { + "epoch": 6.36, + "learning_rate": 1.823246668123225e-06, + "loss": 0.7987, + "step": 29130 + }, + { + "epoch": 6.36, + "learning_rate": 1.8231374262617438e-06, + "loss": 0.8456, + "step": 29131 + }, + { + "epoch": 6.36, + "learning_rate": 1.8230281844002624e-06, + "loss": 1.1849, + "step": 29132 + }, + { + "epoch": 6.36, + "learning_rate": 1.8229189425387812e-06, + "loss": 2.6584, + "step": 29133 + }, + { + "epoch": 6.36, + "learning_rate": 1.8228097006772996e-06, + "loss": 0.8566, + "step": 29134 + }, + { + "epoch": 6.36, + "learning_rate": 1.8227004588158184e-06, + "loss": 1.0085, + "step": 29135 + }, + { + "epoch": 6.36, + "learning_rate": 1.822591216954337e-06, + "loss": 1.1119, + "step": 29136 + }, + { + "epoch": 6.36, + "learning_rate": 1.8224819750928556e-06, + "loss": 1.4004, + "step": 29137 + }, + { + "epoch": 6.36, + "learning_rate": 1.8223727332313745e-06, + "loss": 0.7979, + "step": 29138 + }, + { + "epoch": 6.36, + "learning_rate": 1.822263491369893e-06, + "loss": 0.754, + "step": 29139 + }, + { + "epoch": 6.36, + "learning_rate": 1.8221542495084119e-06, + "loss": 0.9063, + "step": 29140 + }, + { + "epoch": 6.36, + "learning_rate": 1.8220450076469303e-06, + "loss": 0.8223, + "step": 29141 + }, + { + "epoch": 6.36, + "learning_rate": 1.8219357657854491e-06, + "loss": 0.6223, + "step": 29142 + }, + { + "epoch": 6.36, + "learning_rate": 1.8218265239239677e-06, + "loss": 1.0326, + "step": 29143 + }, + { + "epoch": 6.36, + "learning_rate": 1.8217172820624865e-06, + "loss": 0.1953, + "step": 29144 + }, + { + "epoch": 6.36, + "learning_rate": 1.8216080402010054e-06, + "loss": 1.3952, + "step": 29145 + }, + { + "epoch": 6.36, + "learning_rate": 1.8214987983395238e-06, + "loss": 0.8414, + "step": 29146 + }, + { + "epoch": 6.36, + "learning_rate": 1.8213895564780426e-06, + "loss": 2.0969, + "step": 29147 + }, + { + "epoch": 6.36, + "learning_rate": 1.8212803146165612e-06, + "loss": 1.5717, + "step": 29148 + }, + { + "epoch": 6.36, + "learning_rate": 1.82117107275508e-06, + "loss": 1.3659, + "step": 29149 + }, + { + "epoch": 6.36, + "learning_rate": 1.8210618308935984e-06, + "loss": 0.9751, + "step": 29150 + }, + { + "epoch": 6.36, + "learning_rate": 1.8209525890321172e-06, + "loss": 1.1456, + "step": 29151 + }, + { + "epoch": 6.36, + "learning_rate": 1.820843347170636e-06, + "loss": 0.9109, + "step": 29152 + }, + { + "epoch": 6.36, + "learning_rate": 1.8207341053091545e-06, + "loss": 0.9597, + "step": 29153 + }, + { + "epoch": 6.36, + "learning_rate": 1.8206248634476733e-06, + "loss": 1.1263, + "step": 29154 + }, + { + "epoch": 6.36, + "learning_rate": 1.820515621586192e-06, + "loss": 1.8872, + "step": 29155 + }, + { + "epoch": 6.36, + "learning_rate": 1.8204063797247107e-06, + "loss": 1.5994, + "step": 29156 + }, + { + "epoch": 6.36, + "learning_rate": 1.8202971378632291e-06, + "loss": 0.307, + "step": 29157 + }, + { + "epoch": 6.36, + "learning_rate": 1.820187896001748e-06, + "loss": 1.0537, + "step": 29158 + }, + { + "epoch": 6.36, + "learning_rate": 1.8200786541402668e-06, + "loss": 0.6502, + "step": 29159 + }, + { + "epoch": 6.36, + "learning_rate": 1.8199694122787854e-06, + "loss": 1.536, + "step": 29160 + }, + { + "epoch": 6.36, + "learning_rate": 1.8198601704173042e-06, + "loss": 0.4982, + "step": 29161 + }, + { + "epoch": 6.36, + "learning_rate": 1.8197509285558226e-06, + "loss": 0.4916, + "step": 29162 + }, + { + "epoch": 6.36, + "learning_rate": 1.8196416866943414e-06, + "loss": 1.4728, + "step": 29163 + }, + { + "epoch": 6.36, + "learning_rate": 1.8195324448328602e-06, + "loss": 0.6342, + "step": 29164 + }, + { + "epoch": 6.37, + "learning_rate": 1.8194232029713789e-06, + "loss": 0.6351, + "step": 29165 + }, + { + "epoch": 6.37, + "learning_rate": 1.8193139611098975e-06, + "loss": 2.1445, + "step": 29166 + }, + { + "epoch": 6.37, + "learning_rate": 1.819204719248416e-06, + "loss": 1.2671, + "step": 29167 + }, + { + "epoch": 6.37, + "learning_rate": 1.819095477386935e-06, + "loss": 0.4644, + "step": 29168 + }, + { + "epoch": 6.37, + "learning_rate": 1.8189862355254533e-06, + "loss": 1.0647, + "step": 29169 + }, + { + "epoch": 6.37, + "learning_rate": 1.8188769936639721e-06, + "loss": 0.9693, + "step": 29170 + }, + { + "epoch": 6.37, + "learning_rate": 1.818767751802491e-06, + "loss": 1.2654, + "step": 29171 + }, + { + "epoch": 6.37, + "learning_rate": 1.8186585099410096e-06, + "loss": 1.433, + "step": 29172 + }, + { + "epoch": 6.37, + "learning_rate": 1.8185492680795284e-06, + "loss": 0.1308, + "step": 29173 + }, + { + "epoch": 6.37, + "learning_rate": 1.8184400262180468e-06, + "loss": 1.16, + "step": 29174 + }, + { + "epoch": 6.37, + "learning_rate": 1.8183307843565656e-06, + "loss": 2.7585, + "step": 29175 + }, + { + "epoch": 6.37, + "learning_rate": 1.8182215424950842e-06, + "loss": 1.2119, + "step": 29176 + }, + { + "epoch": 6.37, + "learning_rate": 1.818112300633603e-06, + "loss": 1.356, + "step": 29177 + }, + { + "epoch": 6.37, + "learning_rate": 1.8180030587721216e-06, + "loss": 1.0515, + "step": 29178 + }, + { + "epoch": 6.37, + "learning_rate": 1.8178938169106403e-06, + "loss": 2.9703, + "step": 29179 + }, + { + "epoch": 6.37, + "learning_rate": 1.817784575049159e-06, + "loss": 0.4735, + "step": 29180 + }, + { + "epoch": 6.37, + "learning_rate": 1.8176753331876775e-06, + "loss": 1.7137, + "step": 29181 + }, + { + "epoch": 6.37, + "learning_rate": 1.8175660913261963e-06, + "loss": 1.7863, + "step": 29182 + }, + { + "epoch": 6.37, + "learning_rate": 1.817456849464715e-06, + "loss": 1.5543, + "step": 29183 + }, + { + "epoch": 6.37, + "learning_rate": 1.8173476076032337e-06, + "loss": 1.2218, + "step": 29184 + }, + { + "epoch": 6.37, + "learning_rate": 1.8172383657417526e-06, + "loss": 2.1461, + "step": 29185 + }, + { + "epoch": 6.37, + "learning_rate": 1.817129123880271e-06, + "loss": 0.1957, + "step": 29186 + }, + { + "epoch": 6.37, + "learning_rate": 1.8170198820187898e-06, + "loss": 0.9742, + "step": 29187 + }, + { + "epoch": 6.37, + "learning_rate": 1.8169106401573084e-06, + "loss": 1.0271, + "step": 29188 + }, + { + "epoch": 6.37, + "learning_rate": 1.8168013982958272e-06, + "loss": 1.0322, + "step": 29189 + }, + { + "epoch": 6.37, + "learning_rate": 1.8166921564343456e-06, + "loss": 1.7495, + "step": 29190 + }, + { + "epoch": 6.37, + "learning_rate": 1.8165829145728644e-06, + "loss": 1.24, + "step": 29191 + }, + { + "epoch": 6.37, + "learning_rate": 1.8164736727113833e-06, + "loss": 1.309, + "step": 29192 + }, + { + "epoch": 6.37, + "learning_rate": 1.8163644308499019e-06, + "loss": 3.0804, + "step": 29193 + }, + { + "epoch": 6.37, + "learning_rate": 1.8162551889884205e-06, + "loss": 1.1598, + "step": 29194 + }, + { + "epoch": 6.37, + "learning_rate": 1.816145947126939e-06, + "loss": 1.0708, + "step": 29195 + }, + { + "epoch": 6.37, + "learning_rate": 1.816036705265458e-06, + "loss": 1.3534, + "step": 29196 + }, + { + "epoch": 6.37, + "learning_rate": 1.8159274634039767e-06, + "loss": 0.8007, + "step": 29197 + }, + { + "epoch": 6.37, + "learning_rate": 1.8158182215424951e-06, + "loss": 2.0168, + "step": 29198 + }, + { + "epoch": 6.37, + "learning_rate": 1.815708979681014e-06, + "loss": 0.8762, + "step": 29199 + }, + { + "epoch": 6.37, + "learning_rate": 1.8155997378195326e-06, + "loss": 1.4412, + "step": 29200 + }, + { + "epoch": 6.37, + "learning_rate": 1.8154904959580514e-06, + "loss": 1.2095, + "step": 29201 + }, + { + "epoch": 6.37, + "learning_rate": 1.8153812540965698e-06, + "loss": 1.4888, + "step": 29202 + }, + { + "epoch": 6.37, + "learning_rate": 1.8152720122350886e-06, + "loss": 1.4426, + "step": 29203 + }, + { + "epoch": 6.37, + "learning_rate": 1.8151627703736074e-06, + "loss": 1.407, + "step": 29204 + }, + { + "epoch": 6.37, + "learning_rate": 1.815053528512126e-06, + "loss": 1.8974, + "step": 29205 + }, + { + "epoch": 6.37, + "learning_rate": 1.8149442866506447e-06, + "loss": 0.5707, + "step": 29206 + }, + { + "epoch": 6.37, + "learning_rate": 1.8148350447891633e-06, + "loss": 2.3646, + "step": 29207 + }, + { + "epoch": 6.37, + "learning_rate": 1.814725802927682e-06, + "loss": 0.3429, + "step": 29208 + }, + { + "epoch": 6.37, + "learning_rate": 1.8146165610662007e-06, + "loss": 2.9604, + "step": 29209 + }, + { + "epoch": 6.37, + "learning_rate": 1.8145073192047193e-06, + "loss": 0.5369, + "step": 29210 + }, + { + "epoch": 6.38, + "learning_rate": 1.8143980773432381e-06, + "loss": 0.1435, + "step": 29211 + }, + { + "epoch": 6.38, + "learning_rate": 1.8142888354817567e-06, + "loss": 2.0725, + "step": 29212 + }, + { + "epoch": 6.38, + "learning_rate": 1.8141795936202756e-06, + "loss": 1.0355, + "step": 29213 + }, + { + "epoch": 6.38, + "learning_rate": 1.814070351758794e-06, + "loss": 1.3306, + "step": 29214 + }, + { + "epoch": 6.38, + "learning_rate": 1.8139611098973128e-06, + "loss": 1.3037, + "step": 29215 + }, + { + "epoch": 6.38, + "learning_rate": 1.8138518680358314e-06, + "loss": 0.6012, + "step": 29216 + }, + { + "epoch": 6.38, + "learning_rate": 1.8137426261743502e-06, + "loss": 2.2677, + "step": 29217 + }, + { + "epoch": 6.38, + "learning_rate": 1.813633384312869e-06, + "loss": 0.7008, + "step": 29218 + }, + { + "epoch": 6.38, + "learning_rate": 1.8135241424513875e-06, + "loss": 2.7616, + "step": 29219 + }, + { + "epoch": 6.38, + "learning_rate": 1.8134149005899063e-06, + "loss": 1.6324, + "step": 29220 + }, + { + "epoch": 6.38, + "learning_rate": 1.8133056587284249e-06, + "loss": 2.7299, + "step": 29221 + }, + { + "epoch": 6.38, + "learning_rate": 1.8131964168669435e-06, + "loss": 0.8435, + "step": 29222 + }, + { + "epoch": 6.38, + "learning_rate": 1.8130871750054621e-06, + "loss": 1.3389, + "step": 29223 + }, + { + "epoch": 6.38, + "learning_rate": 1.812977933143981e-06, + "loss": 1.1797, + "step": 29224 + }, + { + "epoch": 6.38, + "learning_rate": 1.8128686912824998e-06, + "loss": 1.6632, + "step": 29225 + }, + { + "epoch": 6.38, + "learning_rate": 1.8127594494210182e-06, + "loss": 0.5189, + "step": 29226 + }, + { + "epoch": 6.38, + "learning_rate": 1.812650207559537e-06, + "loss": 2.5355, + "step": 29227 + }, + { + "epoch": 6.38, + "learning_rate": 1.8125409656980556e-06, + "loss": 1.006, + "step": 29228 + }, + { + "epoch": 6.38, + "learning_rate": 1.8124317238365744e-06, + "loss": 1.0112, + "step": 29229 + }, + { + "epoch": 6.38, + "learning_rate": 1.8123224819750932e-06, + "loss": 0.1367, + "step": 29230 + }, + { + "epoch": 6.38, + "learning_rate": 1.8122132401136116e-06, + "loss": 1.1832, + "step": 29231 + }, + { + "epoch": 6.38, + "learning_rate": 1.8121039982521305e-06, + "loss": 1.1605, + "step": 29232 + }, + { + "epoch": 6.38, + "learning_rate": 1.811994756390649e-06, + "loss": 1.6463, + "step": 29233 + }, + { + "epoch": 6.38, + "learning_rate": 1.8118855145291677e-06, + "loss": 2.3004, + "step": 29234 + }, + { + "epoch": 6.38, + "learning_rate": 1.8117762726676863e-06, + "loss": 1.3367, + "step": 29235 + }, + { + "epoch": 6.38, + "learning_rate": 1.8116670308062051e-06, + "loss": 1.3239, + "step": 29236 + }, + { + "epoch": 6.38, + "learning_rate": 1.811557788944724e-06, + "loss": 1.8505, + "step": 29237 + }, + { + "epoch": 6.38, + "learning_rate": 1.8114485470832423e-06, + "loss": 0.749, + "step": 29238 + }, + { + "epoch": 6.38, + "learning_rate": 1.8113393052217612e-06, + "loss": 1.5766, + "step": 29239 + }, + { + "epoch": 6.38, + "learning_rate": 1.8112300633602798e-06, + "loss": 1.6587, + "step": 29240 + }, + { + "epoch": 6.38, + "learning_rate": 1.8111208214987986e-06, + "loss": 0.739, + "step": 29241 + }, + { + "epoch": 6.38, + "learning_rate": 1.811011579637317e-06, + "loss": 0.2518, + "step": 29242 + }, + { + "epoch": 6.38, + "learning_rate": 1.8109023377758358e-06, + "loss": 2.0694, + "step": 29243 + }, + { + "epoch": 6.38, + "learning_rate": 1.8107930959143546e-06, + "loss": 1.0988, + "step": 29244 + }, + { + "epoch": 6.38, + "learning_rate": 1.8106838540528732e-06, + "loss": 0.2639, + "step": 29245 + }, + { + "epoch": 6.38, + "learning_rate": 1.810574612191392e-06, + "loss": 0.5444, + "step": 29246 + }, + { + "epoch": 6.38, + "learning_rate": 1.8104653703299105e-06, + "loss": 1.3191, + "step": 29247 + }, + { + "epoch": 6.38, + "learning_rate": 1.8103561284684293e-06, + "loss": 1.5554, + "step": 29248 + }, + { + "epoch": 6.38, + "learning_rate": 1.810246886606948e-06, + "loss": 1.4648, + "step": 29249 + }, + { + "epoch": 6.38, + "learning_rate": 1.8101376447454665e-06, + "loss": 2.4358, + "step": 29250 + }, + { + "epoch": 6.38, + "learning_rate": 1.8100284028839853e-06, + "loss": 1.8609, + "step": 29251 + }, + { + "epoch": 6.38, + "learning_rate": 1.809919161022504e-06, + "loss": 0.7411, + "step": 29252 + }, + { + "epoch": 6.38, + "learning_rate": 1.8098099191610228e-06, + "loss": 1.2007, + "step": 29253 + }, + { + "epoch": 6.38, + "learning_rate": 1.8097006772995412e-06, + "loss": 2.1515, + "step": 29254 + }, + { + "epoch": 6.38, + "learning_rate": 1.80959143543806e-06, + "loss": 1.292, + "step": 29255 + }, + { + "epoch": 6.38, + "learning_rate": 1.8094821935765786e-06, + "loss": 2.3978, + "step": 29256 + }, + { + "epoch": 6.39, + "learning_rate": 1.8093729517150974e-06, + "loss": 1.97, + "step": 29257 + }, + { + "epoch": 6.39, + "learning_rate": 1.8092637098536162e-06, + "loss": 1.2429, + "step": 29258 + }, + { + "epoch": 6.39, + "learning_rate": 1.8091544679921346e-06, + "loss": 2.374, + "step": 29259 + }, + { + "epoch": 6.39, + "learning_rate": 1.8090452261306535e-06, + "loss": 0.8783, + "step": 29260 + }, + { + "epoch": 6.39, + "learning_rate": 1.808935984269172e-06, + "loss": 0.9909, + "step": 29261 + }, + { + "epoch": 6.39, + "learning_rate": 1.808826742407691e-06, + "loss": 1.1233, + "step": 29262 + }, + { + "epoch": 6.39, + "learning_rate": 1.8087175005462095e-06, + "loss": 2.4655, + "step": 29263 + }, + { + "epoch": 6.39, + "learning_rate": 1.8086082586847281e-06, + "loss": 0.7687, + "step": 29264 + }, + { + "epoch": 6.39, + "learning_rate": 1.808499016823247e-06, + "loss": 1.1149, + "step": 29265 + }, + { + "epoch": 6.39, + "learning_rate": 1.8083897749617653e-06, + "loss": 0.6565, + "step": 29266 + }, + { + "epoch": 6.39, + "learning_rate": 1.8082805331002842e-06, + "loss": 0.5829, + "step": 29267 + }, + { + "epoch": 6.39, + "learning_rate": 1.8081712912388028e-06, + "loss": 1.9766, + "step": 29268 + }, + { + "epoch": 6.39, + "learning_rate": 1.8080620493773216e-06, + "loss": 0.6504, + "step": 29269 + }, + { + "epoch": 6.39, + "learning_rate": 1.8079528075158404e-06, + "loss": 1.163, + "step": 29270 + }, + { + "epoch": 6.39, + "learning_rate": 1.8078435656543588e-06, + "loss": 0.5838, + "step": 29271 + }, + { + "epoch": 6.39, + "learning_rate": 1.8077343237928776e-06, + "loss": 1.1326, + "step": 29272 + }, + { + "epoch": 6.39, + "learning_rate": 1.8076250819313963e-06, + "loss": 1.1406, + "step": 29273 + }, + { + "epoch": 6.39, + "learning_rate": 1.807515840069915e-06, + "loss": 0.2927, + "step": 29274 + }, + { + "epoch": 6.39, + "learning_rate": 1.8074065982084335e-06, + "loss": 0.3087, + "step": 29275 + }, + { + "epoch": 6.39, + "learning_rate": 1.8072973563469523e-06, + "loss": 1.2729, + "step": 29276 + }, + { + "epoch": 6.39, + "learning_rate": 1.8071881144854711e-06, + "loss": 1.4472, + "step": 29277 + }, + { + "epoch": 6.39, + "learning_rate": 1.8070788726239895e-06, + "loss": 0.4115, + "step": 29278 + }, + { + "epoch": 6.39, + "learning_rate": 1.8069696307625083e-06, + "loss": 1.0094, + "step": 29279 + }, + { + "epoch": 6.39, + "learning_rate": 1.806860388901027e-06, + "loss": 1.5119, + "step": 29280 + }, + { + "epoch": 6.39, + "learning_rate": 1.8067511470395458e-06, + "loss": 1.9239, + "step": 29281 + }, + { + "epoch": 6.39, + "learning_rate": 1.8066419051780642e-06, + "loss": 1.2354, + "step": 29282 + }, + { + "epoch": 6.39, + "learning_rate": 1.806532663316583e-06, + "loss": 1.7321, + "step": 29283 + }, + { + "epoch": 6.39, + "learning_rate": 1.8064234214551018e-06, + "loss": 1.0449, + "step": 29284 + }, + { + "epoch": 6.39, + "learning_rate": 1.8063141795936204e-06, + "loss": 1.0747, + "step": 29285 + }, + { + "epoch": 6.39, + "learning_rate": 1.8062049377321393e-06, + "loss": 1.0354, + "step": 29286 + }, + { + "epoch": 6.39, + "learning_rate": 1.8060956958706577e-06, + "loss": 0.6835, + "step": 29287 + }, + { + "epoch": 6.39, + "learning_rate": 1.8059864540091765e-06, + "loss": 2.0946, + "step": 29288 + }, + { + "epoch": 6.39, + "learning_rate": 1.805877212147695e-06, + "loss": 1.3928, + "step": 29289 + }, + { + "epoch": 6.39, + "learning_rate": 1.805767970286214e-06, + "loss": 2.3027, + "step": 29290 + }, + { + "epoch": 6.39, + "learning_rate": 1.8056587284247325e-06, + "loss": 0.2621, + "step": 29291 + }, + { + "epoch": 6.39, + "learning_rate": 1.8055494865632511e-06, + "loss": 1.608, + "step": 29292 + }, + { + "epoch": 6.39, + "learning_rate": 1.80544024470177e-06, + "loss": 1.3813, + "step": 29293 + }, + { + "epoch": 6.39, + "learning_rate": 1.8053310028402884e-06, + "loss": 0.6758, + "step": 29294 + }, + { + "epoch": 6.39, + "learning_rate": 1.8052217609788072e-06, + "loss": 1.5657, + "step": 29295 + }, + { + "epoch": 6.39, + "learning_rate": 1.805112519117326e-06, + "loss": 1.3836, + "step": 29296 + }, + { + "epoch": 6.39, + "learning_rate": 1.8050032772558446e-06, + "loss": 0.2471, + "step": 29297 + }, + { + "epoch": 6.39, + "learning_rate": 1.8048940353943634e-06, + "loss": 1.0175, + "step": 29298 + }, + { + "epoch": 6.39, + "learning_rate": 1.8047847935328818e-06, + "loss": 0.3769, + "step": 29299 + }, + { + "epoch": 6.39, + "learning_rate": 1.8046755516714007e-06, + "loss": 1.2903, + "step": 29300 + }, + { + "epoch": 6.39, + "learning_rate": 1.8045663098099193e-06, + "loss": 1.5885, + "step": 29301 + }, + { + "epoch": 6.4, + "learning_rate": 1.804457067948438e-06, + "loss": 0.4848, + "step": 29302 + }, + { + "epoch": 6.4, + "learning_rate": 1.8043478260869567e-06, + "loss": 1.0172, + "step": 29303 + }, + { + "epoch": 6.4, + "learning_rate": 1.8042385842254753e-06, + "loss": 1.3633, + "step": 29304 + }, + { + "epoch": 6.4, + "learning_rate": 1.8041293423639941e-06, + "loss": 0.8977, + "step": 29305 + }, + { + "epoch": 6.4, + "learning_rate": 1.8040201005025125e-06, + "loss": 2.9092, + "step": 29306 + }, + { + "epoch": 6.4, + "learning_rate": 1.8039108586410314e-06, + "loss": 2.6226, + "step": 29307 + }, + { + "epoch": 6.4, + "learning_rate": 1.80380161677955e-06, + "loss": 1.9681, + "step": 29308 + }, + { + "epoch": 6.4, + "learning_rate": 1.8036923749180688e-06, + "loss": 1.1905, + "step": 29309 + }, + { + "epoch": 6.4, + "learning_rate": 1.8035831330565876e-06, + "loss": 1.7181, + "step": 29310 + }, + { + "epoch": 6.4, + "learning_rate": 1.803473891195106e-06, + "loss": 0.7429, + "step": 29311 + }, + { + "epoch": 6.4, + "learning_rate": 1.8033646493336248e-06, + "loss": 0.602, + "step": 29312 + }, + { + "epoch": 6.4, + "learning_rate": 1.8032554074721434e-06, + "loss": 1.5594, + "step": 29313 + }, + { + "epoch": 6.4, + "learning_rate": 1.8031461656106623e-06, + "loss": 1.2, + "step": 29314 + }, + { + "epoch": 6.4, + "learning_rate": 1.8030369237491807e-06, + "loss": 2.6668, + "step": 29315 + }, + { + "epoch": 6.4, + "learning_rate": 1.8029276818876995e-06, + "loss": 0.5946, + "step": 29316 + }, + { + "epoch": 6.4, + "learning_rate": 1.8028184400262183e-06, + "loss": 1.0977, + "step": 29317 + }, + { + "epoch": 6.4, + "learning_rate": 1.802709198164737e-06, + "loss": 1.1643, + "step": 29318 + }, + { + "epoch": 6.4, + "learning_rate": 1.8025999563032555e-06, + "loss": 0.8386, + "step": 29319 + }, + { + "epoch": 6.4, + "learning_rate": 1.8024907144417741e-06, + "loss": 1.8397, + "step": 29320 + }, + { + "epoch": 6.4, + "learning_rate": 1.802381472580293e-06, + "loss": 1.2908, + "step": 29321 + }, + { + "epoch": 6.4, + "learning_rate": 1.8022722307188114e-06, + "loss": 0.8947, + "step": 29322 + }, + { + "epoch": 6.4, + "learning_rate": 1.8021629888573302e-06, + "loss": 1.1986, + "step": 29323 + }, + { + "epoch": 6.4, + "learning_rate": 1.802053746995849e-06, + "loss": 3.9325, + "step": 29324 + }, + { + "epoch": 6.4, + "learning_rate": 1.8019445051343676e-06, + "loss": 0.7381, + "step": 29325 + }, + { + "epoch": 6.4, + "learning_rate": 1.8018352632728864e-06, + "loss": 0.9728, + "step": 29326 + }, + { + "epoch": 6.4, + "learning_rate": 1.8017260214114048e-06, + "loss": 1.189, + "step": 29327 + }, + { + "epoch": 6.4, + "learning_rate": 1.8016167795499237e-06, + "loss": 2.0715, + "step": 29328 + }, + { + "epoch": 6.4, + "learning_rate": 1.8015075376884423e-06, + "loss": 1.1408, + "step": 29329 + }, + { + "epoch": 6.4, + "learning_rate": 1.801398295826961e-06, + "loss": 0.2957, + "step": 29330 + }, + { + "epoch": 6.4, + "learning_rate": 1.8012890539654797e-06, + "loss": 1.4951, + "step": 29331 + }, + { + "epoch": 6.4, + "learning_rate": 1.8011798121039983e-06, + "loss": 3.5169, + "step": 29332 + }, + { + "epoch": 6.4, + "learning_rate": 1.8010705702425171e-06, + "loss": 0.8164, + "step": 29333 + }, + { + "epoch": 6.4, + "learning_rate": 1.8009613283810358e-06, + "loss": 1.4501, + "step": 29334 + }, + { + "epoch": 6.4, + "learning_rate": 1.8008520865195544e-06, + "loss": 1.5787, + "step": 29335 + }, + { + "epoch": 6.4, + "learning_rate": 1.8007428446580732e-06, + "loss": 2.2175, + "step": 29336 + }, + { + "epoch": 6.4, + "learning_rate": 1.8006336027965918e-06, + "loss": 0.7106, + "step": 29337 + }, + { + "epoch": 6.4, + "learning_rate": 1.8005243609351106e-06, + "loss": 1.4276, + "step": 29338 + }, + { + "epoch": 6.4, + "learning_rate": 1.800415119073629e-06, + "loss": 1.1997, + "step": 29339 + }, + { + "epoch": 6.4, + "learning_rate": 1.8003058772121479e-06, + "loss": 1.1406, + "step": 29340 + }, + { + "epoch": 6.4, + "learning_rate": 1.8001966353506665e-06, + "loss": 0.4449, + "step": 29341 + }, + { + "epoch": 6.4, + "learning_rate": 1.8000873934891853e-06, + "loss": 0.195, + "step": 29342 + }, + { + "epoch": 6.4, + "learning_rate": 1.7999781516277041e-06, + "loss": 0.096, + "step": 29343 + }, + { + "epoch": 6.4, + "learning_rate": 1.7998689097662225e-06, + "loss": 1.3423, + "step": 29344 + }, + { + "epoch": 6.4, + "learning_rate": 1.7997596679047413e-06, + "loss": 2.3703, + "step": 29345 + }, + { + "epoch": 6.4, + "learning_rate": 1.79965042604326e-06, + "loss": 0.0979, + "step": 29346 + }, + { + "epoch": 6.4, + "learning_rate": 1.7995411841817786e-06, + "loss": 0.7734, + "step": 29347 + }, + { + "epoch": 6.41, + "learning_rate": 1.7994319423202972e-06, + "loss": 1.9568, + "step": 29348 + }, + { + "epoch": 6.41, + "learning_rate": 1.799322700458816e-06, + "loss": 1.1987, + "step": 29349 + }, + { + "epoch": 6.41, + "learning_rate": 1.7992134585973348e-06, + "loss": 1.3324, + "step": 29350 + }, + { + "epoch": 6.41, + "learning_rate": 1.7991042167358532e-06, + "loss": 0.4292, + "step": 29351 + }, + { + "epoch": 6.41, + "learning_rate": 1.798994974874372e-06, + "loss": 1.4977, + "step": 29352 + }, + { + "epoch": 6.41, + "learning_rate": 1.7988857330128906e-06, + "loss": 0.5106, + "step": 29353 + }, + { + "epoch": 6.41, + "learning_rate": 1.7987764911514095e-06, + "loss": 0.1825, + "step": 29354 + }, + { + "epoch": 6.41, + "learning_rate": 1.7986672492899279e-06, + "loss": 0.8347, + "step": 29355 + }, + { + "epoch": 6.41, + "learning_rate": 1.7985580074284467e-06, + "loss": 0.5947, + "step": 29356 + }, + { + "epoch": 6.41, + "learning_rate": 1.7984487655669655e-06, + "loss": 1.7236, + "step": 29357 + }, + { + "epoch": 6.41, + "learning_rate": 1.7983395237054841e-06, + "loss": 3.3799, + "step": 29358 + }, + { + "epoch": 6.41, + "learning_rate": 1.7982302818440027e-06, + "loss": 1.0855, + "step": 29359 + }, + { + "epoch": 6.41, + "learning_rate": 1.7981210399825213e-06, + "loss": 1.3533, + "step": 29360 + }, + { + "epoch": 6.41, + "learning_rate": 1.7980117981210402e-06, + "loss": 1.3571, + "step": 29361 + }, + { + "epoch": 6.41, + "learning_rate": 1.7979025562595588e-06, + "loss": 0.8103, + "step": 29362 + }, + { + "epoch": 6.41, + "learning_rate": 1.7977933143980774e-06, + "loss": 1.5122, + "step": 29363 + }, + { + "epoch": 6.41, + "learning_rate": 1.7976840725365962e-06, + "loss": 0.3325, + "step": 29364 + }, + { + "epoch": 6.41, + "learning_rate": 1.7975748306751148e-06, + "loss": 0.4864, + "step": 29365 + }, + { + "epoch": 6.41, + "learning_rate": 1.7974655888136336e-06, + "loss": 1.9531, + "step": 29366 + }, + { + "epoch": 6.41, + "learning_rate": 1.797356346952152e-06, + "loss": 1.0742, + "step": 29367 + }, + { + "epoch": 6.41, + "learning_rate": 1.7972471050906709e-06, + "loss": 1.8265, + "step": 29368 + }, + { + "epoch": 6.41, + "learning_rate": 1.7971378632291897e-06, + "loss": 1.4335, + "step": 29369 + }, + { + "epoch": 6.41, + "learning_rate": 1.7970286213677083e-06, + "loss": 2.0338, + "step": 29370 + }, + { + "epoch": 6.41, + "learning_rate": 1.7969193795062271e-06, + "loss": 0.7312, + "step": 29371 + }, + { + "epoch": 6.41, + "learning_rate": 1.7968101376447455e-06, + "loss": 2.3202, + "step": 29372 + }, + { + "epoch": 6.41, + "learning_rate": 1.7967008957832643e-06, + "loss": 1.8, + "step": 29373 + }, + { + "epoch": 6.41, + "learning_rate": 1.796591653921783e-06, + "loss": 1.1258, + "step": 29374 + }, + { + "epoch": 6.41, + "learning_rate": 1.7964824120603016e-06, + "loss": 1.1152, + "step": 29375 + }, + { + "epoch": 6.41, + "learning_rate": 1.7963731701988204e-06, + "loss": 2.219, + "step": 29376 + }, + { + "epoch": 6.41, + "learning_rate": 1.796263928337339e-06, + "loss": 1.0071, + "step": 29377 + }, + { + "epoch": 6.41, + "learning_rate": 1.7961546864758578e-06, + "loss": 0.7098, + "step": 29378 + }, + { + "epoch": 6.41, + "learning_rate": 1.7960454446143762e-06, + "loss": 1.5988, + "step": 29379 + }, + { + "epoch": 6.41, + "learning_rate": 1.795936202752895e-06, + "loss": 0.5863, + "step": 29380 + }, + { + "epoch": 6.41, + "learning_rate": 1.7958269608914137e-06, + "loss": 2.0403, + "step": 29381 + }, + { + "epoch": 6.41, + "learning_rate": 1.7957177190299325e-06, + "loss": 1.5265, + "step": 29382 + }, + { + "epoch": 6.41, + "learning_rate": 1.7956084771684513e-06, + "loss": 1.0133, + "step": 29383 + }, + { + "epoch": 6.41, + "learning_rate": 1.7954992353069697e-06, + "loss": 0.896, + "step": 29384 + }, + { + "epoch": 6.41, + "learning_rate": 1.7953899934454885e-06, + "loss": 1.3447, + "step": 29385 + }, + { + "epoch": 6.41, + "learning_rate": 1.7952807515840071e-06, + "loss": 2.2794, + "step": 29386 + }, + { + "epoch": 6.41, + "learning_rate": 1.795171509722526e-06, + "loss": 1.1184, + "step": 29387 + }, + { + "epoch": 6.41, + "learning_rate": 1.7950622678610444e-06, + "loss": 1.0368, + "step": 29388 + }, + { + "epoch": 6.41, + "learning_rate": 1.7949530259995632e-06, + "loss": 0.8219, + "step": 29389 + }, + { + "epoch": 6.41, + "learning_rate": 1.794843784138082e-06, + "loss": 1.683, + "step": 29390 + }, + { + "epoch": 6.41, + "learning_rate": 1.7947345422766004e-06, + "loss": 1.8451, + "step": 29391 + }, + { + "epoch": 6.41, + "learning_rate": 1.7946253004151192e-06, + "loss": 1.8756, + "step": 29392 + }, + { + "epoch": 6.41, + "learning_rate": 1.7945160585536378e-06, + "loss": 1.8333, + "step": 29393 + }, + { + "epoch": 6.42, + "learning_rate": 1.7944068166921567e-06, + "loss": 0.5167, + "step": 29394 + }, + { + "epoch": 6.42, + "learning_rate": 1.794297574830675e-06, + "loss": 1.1855, + "step": 29395 + }, + { + "epoch": 6.42, + "learning_rate": 1.7941883329691939e-06, + "loss": 1.5673, + "step": 29396 + }, + { + "epoch": 6.42, + "learning_rate": 1.7940790911077127e-06, + "loss": 1.5314, + "step": 29397 + }, + { + "epoch": 6.42, + "learning_rate": 1.7939698492462313e-06, + "loss": 1.3416, + "step": 29398 + }, + { + "epoch": 6.42, + "learning_rate": 1.7938606073847501e-06, + "loss": 1.2401, + "step": 29399 + }, + { + "epoch": 6.42, + "learning_rate": 1.7937513655232685e-06, + "loss": 0.255, + "step": 29400 + }, + { + "epoch": 6.42, + "learning_rate": 1.7936421236617874e-06, + "loss": 0.3153, + "step": 29401 + }, + { + "epoch": 6.42, + "learning_rate": 1.7935328818003062e-06, + "loss": 1.3011, + "step": 29402 + }, + { + "epoch": 6.42, + "learning_rate": 1.7934236399388246e-06, + "loss": 2.2324, + "step": 29403 + }, + { + "epoch": 6.42, + "learning_rate": 1.7933143980773434e-06, + "loss": 0.8647, + "step": 29404 + }, + { + "epoch": 6.42, + "learning_rate": 1.793205156215862e-06, + "loss": 0.6775, + "step": 29405 + }, + { + "epoch": 6.42, + "learning_rate": 1.7930959143543808e-06, + "loss": 1.0472, + "step": 29406 + }, + { + "epoch": 6.42, + "learning_rate": 1.7929866724928992e-06, + "loss": 1.2918, + "step": 29407 + }, + { + "epoch": 6.42, + "learning_rate": 1.792877430631418e-06, + "loss": 0.5268, + "step": 29408 + }, + { + "epoch": 6.42, + "learning_rate": 1.7927681887699369e-06, + "loss": 1.3381, + "step": 29409 + }, + { + "epoch": 6.42, + "learning_rate": 1.7926589469084555e-06, + "loss": 2.259, + "step": 29410 + }, + { + "epoch": 6.42, + "learning_rate": 1.7925497050469743e-06, + "loss": 3.1582, + "step": 29411 + }, + { + "epoch": 6.42, + "learning_rate": 1.7924404631854927e-06, + "loss": 0.8347, + "step": 29412 + }, + { + "epoch": 6.42, + "learning_rate": 1.7923312213240115e-06, + "loss": 0.482, + "step": 29413 + }, + { + "epoch": 6.42, + "learning_rate": 1.7922219794625301e-06, + "loss": 1.1847, + "step": 29414 + }, + { + "epoch": 6.42, + "learning_rate": 1.792112737601049e-06, + "loss": 1.1453, + "step": 29415 + }, + { + "epoch": 6.42, + "learning_rate": 1.7920034957395676e-06, + "loss": 3.4268, + "step": 29416 + }, + { + "epoch": 6.42, + "learning_rate": 1.7918942538780862e-06, + "loss": 1.0412, + "step": 29417 + }, + { + "epoch": 6.42, + "learning_rate": 1.791785012016605e-06, + "loss": 1.0378, + "step": 29418 + }, + { + "epoch": 6.42, + "learning_rate": 1.7916757701551234e-06, + "loss": 0.7612, + "step": 29419 + }, + { + "epoch": 6.42, + "learning_rate": 1.7915665282936422e-06, + "loss": 0.3392, + "step": 29420 + }, + { + "epoch": 6.42, + "learning_rate": 1.7914572864321608e-06, + "loss": 0.9896, + "step": 29421 + }, + { + "epoch": 6.42, + "learning_rate": 1.7913480445706797e-06, + "loss": 0.8171, + "step": 29422 + }, + { + "epoch": 6.42, + "learning_rate": 1.7912388027091985e-06, + "loss": 3.3962, + "step": 29423 + }, + { + "epoch": 6.42, + "learning_rate": 1.7911295608477169e-06, + "loss": 0.1688, + "step": 29424 + }, + { + "epoch": 6.42, + "learning_rate": 1.7910203189862357e-06, + "loss": 1.7761, + "step": 29425 + }, + { + "epoch": 6.42, + "learning_rate": 1.7909110771247543e-06, + "loss": 1.8045, + "step": 29426 + }, + { + "epoch": 6.42, + "learning_rate": 1.7908018352632731e-06, + "loss": 1.3323, + "step": 29427 + }, + { + "epoch": 6.42, + "learning_rate": 1.7906925934017915e-06, + "loss": 0.6643, + "step": 29428 + }, + { + "epoch": 6.42, + "learning_rate": 1.7905833515403104e-06, + "loss": 1.153, + "step": 29429 + }, + { + "epoch": 6.42, + "learning_rate": 1.7904741096788292e-06, + "loss": 1.2835, + "step": 29430 + }, + { + "epoch": 6.42, + "learning_rate": 1.7903648678173476e-06, + "loss": 1.6953, + "step": 29431 + }, + { + "epoch": 6.42, + "learning_rate": 1.7902556259558664e-06, + "loss": 1.6734, + "step": 29432 + }, + { + "epoch": 6.42, + "learning_rate": 1.790146384094385e-06, + "loss": 1.0108, + "step": 29433 + }, + { + "epoch": 6.42, + "learning_rate": 1.7900371422329038e-06, + "loss": 0.1717, + "step": 29434 + }, + { + "epoch": 6.42, + "learning_rate": 1.7899279003714227e-06, + "loss": 1.3135, + "step": 29435 + }, + { + "epoch": 6.42, + "learning_rate": 1.789818658509941e-06, + "loss": 0.5783, + "step": 29436 + }, + { + "epoch": 6.42, + "learning_rate": 1.7897094166484599e-06, + "loss": 1.5288, + "step": 29437 + }, + { + "epoch": 6.42, + "learning_rate": 1.7896001747869785e-06, + "loss": 0.9827, + "step": 29438 + }, + { + "epoch": 6.42, + "learning_rate": 1.7894909329254973e-06, + "loss": 1.3705, + "step": 29439 + }, + { + "epoch": 6.43, + "learning_rate": 1.7893816910640157e-06, + "loss": 0.3519, + "step": 29440 + }, + { + "epoch": 6.43, + "learning_rate": 1.7892724492025345e-06, + "loss": 0.4709, + "step": 29441 + }, + { + "epoch": 6.43, + "learning_rate": 1.7891632073410534e-06, + "loss": 2.0891, + "step": 29442 + }, + { + "epoch": 6.43, + "learning_rate": 1.789053965479572e-06, + "loss": 1.8086, + "step": 29443 + }, + { + "epoch": 6.43, + "learning_rate": 1.7889447236180906e-06, + "loss": 0.8266, + "step": 29444 + }, + { + "epoch": 6.43, + "learning_rate": 1.7888354817566092e-06, + "loss": 1.3165, + "step": 29445 + }, + { + "epoch": 6.43, + "learning_rate": 1.788726239895128e-06, + "loss": 1.4579, + "step": 29446 + }, + { + "epoch": 6.43, + "learning_rate": 1.7886169980336464e-06, + "loss": 1.9561, + "step": 29447 + }, + { + "epoch": 6.43, + "learning_rate": 1.7885077561721652e-06, + "loss": 0.9307, + "step": 29448 + }, + { + "epoch": 6.43, + "learning_rate": 1.788398514310684e-06, + "loss": 0.5968, + "step": 29449 + }, + { + "epoch": 6.43, + "learning_rate": 1.7882892724492027e-06, + "loss": 0.1836, + "step": 29450 + }, + { + "epoch": 6.43, + "learning_rate": 1.7881800305877215e-06, + "loss": 3.041, + "step": 29451 + }, + { + "epoch": 6.43, + "learning_rate": 1.78807078872624e-06, + "loss": 3.9524, + "step": 29452 + }, + { + "epoch": 6.43, + "learning_rate": 1.7879615468647587e-06, + "loss": 1.6988, + "step": 29453 + }, + { + "epoch": 6.43, + "learning_rate": 1.7878523050032773e-06, + "loss": 0.7692, + "step": 29454 + }, + { + "epoch": 6.43, + "learning_rate": 1.7877430631417962e-06, + "loss": 0.2784, + "step": 29455 + }, + { + "epoch": 6.43, + "learning_rate": 1.7876338212803148e-06, + "loss": 1.0678, + "step": 29456 + }, + { + "epoch": 6.43, + "learning_rate": 1.7875245794188334e-06, + "loss": 2.6653, + "step": 29457 + }, + { + "epoch": 6.43, + "learning_rate": 1.7874153375573522e-06, + "loss": 0.9168, + "step": 29458 + }, + { + "epoch": 6.43, + "learning_rate": 1.7873060956958708e-06, + "loss": 1.8803, + "step": 29459 + }, + { + "epoch": 6.43, + "learning_rate": 1.7871968538343894e-06, + "loss": 1.1773, + "step": 29460 + }, + { + "epoch": 6.43, + "learning_rate": 1.787087611972908e-06, + "loss": 2.0135, + "step": 29461 + }, + { + "epoch": 6.43, + "learning_rate": 1.7869783701114269e-06, + "loss": 0.9402, + "step": 29462 + }, + { + "epoch": 6.43, + "learning_rate": 1.7868691282499457e-06, + "loss": 0.9887, + "step": 29463 + }, + { + "epoch": 6.43, + "learning_rate": 1.786759886388464e-06, + "loss": 0.4924, + "step": 29464 + }, + { + "epoch": 6.43, + "learning_rate": 1.786650644526983e-06, + "loss": 1.5797, + "step": 29465 + }, + { + "epoch": 6.43, + "learning_rate": 1.7865414026655015e-06, + "loss": 1.4771, + "step": 29466 + }, + { + "epoch": 6.43, + "learning_rate": 1.7864321608040203e-06, + "loss": 1.0655, + "step": 29467 + }, + { + "epoch": 6.43, + "learning_rate": 1.7863229189425392e-06, + "loss": 1.1069, + "step": 29468 + }, + { + "epoch": 6.43, + "learning_rate": 1.7862136770810576e-06, + "loss": 4.5632, + "step": 29469 + }, + { + "epoch": 6.43, + "learning_rate": 1.7861044352195764e-06, + "loss": 2.3454, + "step": 29470 + }, + { + "epoch": 6.43, + "learning_rate": 1.785995193358095e-06, + "loss": 1.7894, + "step": 29471 + }, + { + "epoch": 6.43, + "learning_rate": 1.7858859514966136e-06, + "loss": 1.0317, + "step": 29472 + }, + { + "epoch": 6.43, + "learning_rate": 1.7857767096351322e-06, + "loss": 0.5762, + "step": 29473 + }, + { + "epoch": 6.43, + "learning_rate": 1.785667467773651e-06, + "loss": 1.4192, + "step": 29474 + }, + { + "epoch": 6.43, + "learning_rate": 1.7855582259121699e-06, + "loss": 0.9121, + "step": 29475 + }, + { + "epoch": 6.43, + "learning_rate": 1.7854489840506883e-06, + "loss": 1.4739, + "step": 29476 + }, + { + "epoch": 6.43, + "learning_rate": 1.785339742189207e-06, + "loss": 3.1126, + "step": 29477 + }, + { + "epoch": 6.43, + "learning_rate": 1.7852305003277257e-06, + "loss": 0.2706, + "step": 29478 + }, + { + "epoch": 6.43, + "learning_rate": 1.7851212584662445e-06, + "loss": 2.1442, + "step": 29479 + }, + { + "epoch": 6.43, + "learning_rate": 1.785012016604763e-06, + "loss": 1.0641, + "step": 29480 + }, + { + "epoch": 6.43, + "learning_rate": 1.7849027747432817e-06, + "loss": 0.1423, + "step": 29481 + }, + { + "epoch": 6.43, + "learning_rate": 1.7847935328818006e-06, + "loss": 0.2084, + "step": 29482 + }, + { + "epoch": 6.43, + "learning_rate": 1.7846842910203192e-06, + "loss": 1.5457, + "step": 29483 + }, + { + "epoch": 6.43, + "learning_rate": 1.784575049158838e-06, + "loss": 3.0251, + "step": 29484 + }, + { + "epoch": 6.43, + "learning_rate": 1.7844658072973564e-06, + "loss": 1.5195, + "step": 29485 + }, + { + "epoch": 6.44, + "learning_rate": 1.7843565654358752e-06, + "loss": 1.0678, + "step": 29486 + }, + { + "epoch": 6.44, + "learning_rate": 1.7842473235743938e-06, + "loss": 2.1062, + "step": 29487 + }, + { + "epoch": 6.44, + "learning_rate": 1.7841380817129124e-06, + "loss": 1.2162, + "step": 29488 + }, + { + "epoch": 6.44, + "learning_rate": 1.7840288398514313e-06, + "loss": 0.1866, + "step": 29489 + }, + { + "epoch": 6.44, + "learning_rate": 1.7839195979899499e-06, + "loss": 2.7454, + "step": 29490 + }, + { + "epoch": 6.44, + "learning_rate": 1.7838103561284687e-06, + "loss": 0.4392, + "step": 29491 + }, + { + "epoch": 6.44, + "learning_rate": 1.783701114266987e-06, + "loss": 0.8036, + "step": 29492 + }, + { + "epoch": 6.44, + "learning_rate": 1.783591872405506e-06, + "loss": 0.8494, + "step": 29493 + }, + { + "epoch": 6.44, + "learning_rate": 1.7834826305440245e-06, + "loss": 1.6026, + "step": 29494 + }, + { + "epoch": 6.44, + "learning_rate": 1.7833733886825434e-06, + "loss": 0.6132, + "step": 29495 + }, + { + "epoch": 6.44, + "learning_rate": 1.7832641468210622e-06, + "loss": 1.1525, + "step": 29496 + }, + { + "epoch": 6.44, + "learning_rate": 1.7831549049595806e-06, + "loss": 0.5693, + "step": 29497 + }, + { + "epoch": 6.44, + "learning_rate": 1.7830456630980994e-06, + "loss": 1.4018, + "step": 29498 + }, + { + "epoch": 6.44, + "learning_rate": 1.782936421236618e-06, + "loss": 1.8238, + "step": 29499 + }, + { + "epoch": 6.44, + "learning_rate": 1.7828271793751366e-06, + "loss": 1.6561, + "step": 29500 + }, + { + "epoch": 6.44, + "learning_rate": 1.7827179375136554e-06, + "loss": 2.9725, + "step": 29501 + }, + { + "epoch": 6.44, + "learning_rate": 1.782608695652174e-06, + "loss": 0.1374, + "step": 29502 + }, + { + "epoch": 6.44, + "learning_rate": 1.7824994537906929e-06, + "loss": 2.0068, + "step": 29503 + }, + { + "epoch": 6.44, + "learning_rate": 1.7823902119292113e-06, + "loss": 1.1899, + "step": 29504 + }, + { + "epoch": 6.44, + "learning_rate": 1.78228097006773e-06, + "loss": 0.9879, + "step": 29505 + }, + { + "epoch": 6.44, + "learning_rate": 1.7821717282062487e-06, + "loss": 2.6068, + "step": 29506 + }, + { + "epoch": 6.44, + "learning_rate": 1.7820624863447675e-06, + "loss": 0.3561, + "step": 29507 + }, + { + "epoch": 6.44, + "learning_rate": 1.7819532444832864e-06, + "loss": 1.5194, + "step": 29508 + }, + { + "epoch": 6.44, + "learning_rate": 1.7818440026218048e-06, + "loss": 1.8204, + "step": 29509 + }, + { + "epoch": 6.44, + "learning_rate": 1.7817347607603236e-06, + "loss": 0.3486, + "step": 29510 + }, + { + "epoch": 6.44, + "learning_rate": 1.7816255188988422e-06, + "loss": 2.0574, + "step": 29511 + }, + { + "epoch": 6.44, + "learning_rate": 1.781516277037361e-06, + "loss": 2.5915, + "step": 29512 + }, + { + "epoch": 6.44, + "learning_rate": 1.7814070351758794e-06, + "loss": 1.1398, + "step": 29513 + }, + { + "epoch": 6.44, + "learning_rate": 1.7812977933143982e-06, + "loss": 1.727, + "step": 29514 + }, + { + "epoch": 6.44, + "learning_rate": 1.781188551452917e-06, + "loss": 0.6365, + "step": 29515 + }, + { + "epoch": 6.44, + "learning_rate": 1.7810793095914355e-06, + "loss": 1.1479, + "step": 29516 + }, + { + "epoch": 6.44, + "learning_rate": 1.7809700677299543e-06, + "loss": 1.1018, + "step": 29517 + }, + { + "epoch": 6.44, + "learning_rate": 1.7808608258684729e-06, + "loss": 1.1727, + "step": 29518 + }, + { + "epoch": 6.44, + "learning_rate": 1.7807515840069917e-06, + "loss": 2.1457, + "step": 29519 + }, + { + "epoch": 6.44, + "learning_rate": 1.7806423421455101e-06, + "loss": 1.4198, + "step": 29520 + }, + { + "epoch": 6.44, + "learning_rate": 1.780533100284029e-06, + "loss": 0.8431, + "step": 29521 + }, + { + "epoch": 6.44, + "learning_rate": 1.7804238584225478e-06, + "loss": 1.3452, + "step": 29522 + }, + { + "epoch": 6.44, + "learning_rate": 1.7803146165610664e-06, + "loss": 1.3201, + "step": 29523 + }, + { + "epoch": 6.44, + "learning_rate": 1.7802053746995852e-06, + "loss": 1.0208, + "step": 29524 + }, + { + "epoch": 6.44, + "learning_rate": 1.7800961328381036e-06, + "loss": 1.8376, + "step": 29525 + }, + { + "epoch": 6.44, + "learning_rate": 1.7799868909766224e-06, + "loss": 1.966, + "step": 29526 + }, + { + "epoch": 6.44, + "learning_rate": 1.779877649115141e-06, + "loss": 2.2602, + "step": 29527 + }, + { + "epoch": 6.44, + "learning_rate": 1.7797684072536596e-06, + "loss": 1.6333, + "step": 29528 + }, + { + "epoch": 6.44, + "learning_rate": 1.7796591653921785e-06, + "loss": 1.6103, + "step": 29529 + }, + { + "epoch": 6.44, + "learning_rate": 1.779549923530697e-06, + "loss": 1.7754, + "step": 29530 + }, + { + "epoch": 6.45, + "learning_rate": 1.7794406816692159e-06, + "loss": 1.6546, + "step": 29531 + }, + { + "epoch": 6.45, + "learning_rate": 1.7793314398077343e-06, + "loss": 0.7374, + "step": 29532 + }, + { + "epoch": 6.45, + "learning_rate": 1.7792221979462531e-06, + "loss": 0.2345, + "step": 29533 + }, + { + "epoch": 6.45, + "learning_rate": 1.7791129560847717e-06, + "loss": 1.7035, + "step": 29534 + }, + { + "epoch": 6.45, + "learning_rate": 1.7790037142232905e-06, + "loss": 1.7033, + "step": 29535 + }, + { + "epoch": 6.45, + "learning_rate": 1.7788944723618094e-06, + "loss": 0.7441, + "step": 29536 + }, + { + "epoch": 6.45, + "learning_rate": 1.7787852305003278e-06, + "loss": 0.8556, + "step": 29537 + }, + { + "epoch": 6.45, + "learning_rate": 1.7786759886388466e-06, + "loss": 1.1846, + "step": 29538 + }, + { + "epoch": 6.45, + "learning_rate": 1.7785667467773652e-06, + "loss": 2.1967, + "step": 29539 + }, + { + "epoch": 6.45, + "learning_rate": 1.778457504915884e-06, + "loss": 1.3052, + "step": 29540 + }, + { + "epoch": 6.45, + "learning_rate": 1.7783482630544026e-06, + "loss": 0.881, + "step": 29541 + }, + { + "epoch": 6.45, + "learning_rate": 1.7782390211929212e-06, + "loss": 1.5211, + "step": 29542 + }, + { + "epoch": 6.45, + "learning_rate": 1.77812977933144e-06, + "loss": 0.1695, + "step": 29543 + }, + { + "epoch": 6.45, + "learning_rate": 1.7780205374699585e-06, + "loss": 1.1082, + "step": 29544 + }, + { + "epoch": 6.45, + "learning_rate": 1.7779112956084773e-06, + "loss": 1.3617, + "step": 29545 + }, + { + "epoch": 6.45, + "learning_rate": 1.777802053746996e-06, + "loss": 1.3104, + "step": 29546 + }, + { + "epoch": 6.45, + "learning_rate": 1.7776928118855147e-06, + "loss": 1.6998, + "step": 29547 + }, + { + "epoch": 6.45, + "learning_rate": 1.7775835700240335e-06, + "loss": 0.3736, + "step": 29548 + }, + { + "epoch": 6.45, + "learning_rate": 1.777474328162552e-06, + "loss": 1.3954, + "step": 29549 + }, + { + "epoch": 6.45, + "learning_rate": 1.7773650863010708e-06, + "loss": 1.6353, + "step": 29550 + }, + { + "epoch": 6.45, + "learning_rate": 1.7772558444395894e-06, + "loss": 0.4871, + "step": 29551 + }, + { + "epoch": 6.45, + "learning_rate": 1.7771466025781082e-06, + "loss": 1.0874, + "step": 29552 + }, + { + "epoch": 6.45, + "learning_rate": 1.7770373607166266e-06, + "loss": 1.3987, + "step": 29553 + }, + { + "epoch": 6.45, + "learning_rate": 1.7769281188551454e-06, + "loss": 0.8087, + "step": 29554 + }, + { + "epoch": 6.45, + "learning_rate": 1.7768188769936642e-06, + "loss": 0.8909, + "step": 29555 + }, + { + "epoch": 6.45, + "learning_rate": 1.7767096351321829e-06, + "loss": 2.2114, + "step": 29556 + }, + { + "epoch": 6.45, + "learning_rate": 1.7766003932707015e-06, + "loss": 0.7702, + "step": 29557 + }, + { + "epoch": 6.45, + "learning_rate": 1.77649115140922e-06, + "loss": 3.2528, + "step": 29558 + }, + { + "epoch": 6.45, + "learning_rate": 1.776381909547739e-06, + "loss": 0.6432, + "step": 29559 + }, + { + "epoch": 6.45, + "learning_rate": 1.7762726676862573e-06, + "loss": 0.7757, + "step": 29560 + }, + { + "epoch": 6.45, + "learning_rate": 1.7761634258247761e-06, + "loss": 1.4338, + "step": 29561 + }, + { + "epoch": 6.45, + "learning_rate": 1.776054183963295e-06, + "loss": 1.3345, + "step": 29562 + }, + { + "epoch": 6.45, + "learning_rate": 1.7759449421018136e-06, + "loss": 1.3008, + "step": 29563 + }, + { + "epoch": 6.45, + "learning_rate": 1.7758357002403324e-06, + "loss": 0.5737, + "step": 29564 + }, + { + "epoch": 6.45, + "learning_rate": 1.7757264583788508e-06, + "loss": 0.5881, + "step": 29565 + }, + { + "epoch": 6.45, + "learning_rate": 1.7756172165173696e-06, + "loss": 1.8868, + "step": 29566 + }, + { + "epoch": 6.45, + "learning_rate": 1.7755079746558882e-06, + "loss": 2.9551, + "step": 29567 + }, + { + "epoch": 6.45, + "learning_rate": 1.775398732794407e-06, + "loss": 1.8702, + "step": 29568 + }, + { + "epoch": 6.45, + "learning_rate": 1.7752894909329256e-06, + "loss": 0.3287, + "step": 29569 + }, + { + "epoch": 6.45, + "learning_rate": 1.7751802490714443e-06, + "loss": 3.0801, + "step": 29570 + }, + { + "epoch": 6.45, + "learning_rate": 1.775071007209963e-06, + "loss": 1.6696, + "step": 29571 + }, + { + "epoch": 6.45, + "learning_rate": 1.7749617653484815e-06, + "loss": 1.9386, + "step": 29572 + }, + { + "epoch": 6.45, + "learning_rate": 1.7748525234870003e-06, + "loss": 1.7996, + "step": 29573 + }, + { + "epoch": 6.45, + "learning_rate": 1.7747432816255191e-06, + "loss": 0.6462, + "step": 29574 + }, + { + "epoch": 6.45, + "learning_rate": 1.7746340397640377e-06, + "loss": 1.7612, + "step": 29575 + }, + { + "epoch": 6.45, + "learning_rate": 1.7745247979025566e-06, + "loss": 0.3697, + "step": 29576 + }, + { + "epoch": 6.46, + "learning_rate": 1.774415556041075e-06, + "loss": 1.9063, + "step": 29577 + }, + { + "epoch": 6.46, + "learning_rate": 1.7743063141795938e-06, + "loss": 0.6342, + "step": 29578 + }, + { + "epoch": 6.46, + "learning_rate": 1.7741970723181124e-06, + "loss": 2.1375, + "step": 29579 + }, + { + "epoch": 6.46, + "learning_rate": 1.7740878304566312e-06, + "loss": 0.8696, + "step": 29580 + }, + { + "epoch": 6.46, + "learning_rate": 1.7739785885951498e-06, + "loss": 0.6412, + "step": 29581 + }, + { + "epoch": 6.46, + "learning_rate": 1.7738693467336684e-06, + "loss": 0.7561, + "step": 29582 + }, + { + "epoch": 6.46, + "learning_rate": 1.7737601048721873e-06, + "loss": 2.9013, + "step": 29583 + }, + { + "epoch": 6.46, + "learning_rate": 1.7736508630107059e-06, + "loss": 0.3625, + "step": 29584 + }, + { + "epoch": 6.46, + "learning_rate": 1.7735416211492245e-06, + "loss": 0.8956, + "step": 29585 + }, + { + "epoch": 6.46, + "learning_rate": 1.773432379287743e-06, + "loss": 2.7016, + "step": 29586 + }, + { + "epoch": 6.46, + "learning_rate": 1.773323137426262e-06, + "loss": 1.5017, + "step": 29587 + }, + { + "epoch": 6.46, + "learning_rate": 1.7732138955647807e-06, + "loss": 1.8716, + "step": 29588 + }, + { + "epoch": 6.46, + "learning_rate": 1.7731046537032991e-06, + "loss": 1.0445, + "step": 29589 + }, + { + "epoch": 6.46, + "learning_rate": 1.772995411841818e-06, + "loss": 1.1612, + "step": 29590 + }, + { + "epoch": 6.46, + "learning_rate": 1.7728861699803366e-06, + "loss": 2.0207, + "step": 29591 + }, + { + "epoch": 6.46, + "learning_rate": 1.7727769281188554e-06, + "loss": 1.9626, + "step": 29592 + }, + { + "epoch": 6.46, + "learning_rate": 1.7726676862573738e-06, + "loss": 1.2499, + "step": 29593 + }, + { + "epoch": 6.46, + "learning_rate": 1.7725584443958926e-06, + "loss": 1.7742, + "step": 29594 + }, + { + "epoch": 6.46, + "learning_rate": 1.7724492025344114e-06, + "loss": 0.9311, + "step": 29595 + }, + { + "epoch": 6.46, + "learning_rate": 1.77233996067293e-06, + "loss": 1.7176, + "step": 29596 + }, + { + "epoch": 6.46, + "learning_rate": 1.7722307188114487e-06, + "loss": 0.0664, + "step": 29597 + }, + { + "epoch": 6.46, + "learning_rate": 1.7721214769499673e-06, + "loss": 1.3701, + "step": 29598 + }, + { + "epoch": 6.46, + "learning_rate": 1.772012235088486e-06, + "loss": 2.4511, + "step": 29599 + }, + { + "epoch": 6.46, + "learning_rate": 1.7719029932270045e-06, + "loss": 1.2075, + "step": 29600 + }, + { + "epoch": 6.46, + "learning_rate": 1.7717937513655233e-06, + "loss": 1.2946, + "step": 29601 + }, + { + "epoch": 6.46, + "learning_rate": 1.7716845095040421e-06, + "loss": 1.9321, + "step": 29602 + }, + { + "epoch": 6.46, + "learning_rate": 1.7715752676425608e-06, + "loss": 1.4712, + "step": 29603 + }, + { + "epoch": 6.46, + "learning_rate": 1.7714660257810796e-06, + "loss": 1.1813, + "step": 29604 + }, + { + "epoch": 6.46, + "learning_rate": 1.771356783919598e-06, + "loss": 0.449, + "step": 29605 + }, + { + "epoch": 6.46, + "learning_rate": 1.7712475420581168e-06, + "loss": 0.0788, + "step": 29606 + }, + { + "epoch": 6.46, + "learning_rate": 1.7711383001966356e-06, + "loss": 2.1344, + "step": 29607 + }, + { + "epoch": 6.46, + "learning_rate": 1.7710290583351542e-06, + "loss": 1.717, + "step": 29608 + }, + { + "epoch": 6.46, + "learning_rate": 1.770919816473673e-06, + "loss": 2.0439, + "step": 29609 + }, + { + "epoch": 6.46, + "learning_rate": 1.7708105746121915e-06, + "loss": 1.754, + "step": 29610 + }, + { + "epoch": 6.46, + "learning_rate": 1.7707013327507103e-06, + "loss": 1.1938, + "step": 29611 + }, + { + "epoch": 6.46, + "learning_rate": 1.7705920908892289e-06, + "loss": 1.9134, + "step": 29612 + }, + { + "epoch": 6.46, + "learning_rate": 1.7704828490277475e-06, + "loss": 2.3824, + "step": 29613 + }, + { + "epoch": 6.46, + "learning_rate": 1.7703736071662663e-06, + "loss": 1.2612, + "step": 29614 + }, + { + "epoch": 6.46, + "learning_rate": 1.770264365304785e-06, + "loss": 2.3453, + "step": 29615 + }, + { + "epoch": 6.46, + "learning_rate": 1.7701551234433038e-06, + "loss": 2.0725, + "step": 29616 + }, + { + "epoch": 6.46, + "learning_rate": 1.7700458815818222e-06, + "loss": 1.3873, + "step": 29617 + }, + { + "epoch": 6.46, + "learning_rate": 1.769936639720341e-06, + "loss": 1.4718, + "step": 29618 + }, + { + "epoch": 6.46, + "learning_rate": 1.7698273978588596e-06, + "loss": 2.3296, + "step": 29619 + }, + { + "epoch": 6.46, + "learning_rate": 1.7697181559973784e-06, + "loss": 1.4436, + "step": 29620 + }, + { + "epoch": 6.46, + "learning_rate": 1.7696089141358972e-06, + "loss": 1.6156, + "step": 29621 + }, + { + "epoch": 6.46, + "learning_rate": 1.7694996722744156e-06, + "loss": 1.4219, + "step": 29622 + }, + { + "epoch": 6.47, + "learning_rate": 1.7693904304129345e-06, + "loss": 0.1809, + "step": 29623 + }, + { + "epoch": 6.47, + "learning_rate": 1.769281188551453e-06, + "loss": 0.5846, + "step": 29624 + }, + { + "epoch": 6.47, + "learning_rate": 1.7691719466899717e-06, + "loss": 1.042, + "step": 29625 + }, + { + "epoch": 6.47, + "learning_rate": 1.7690627048284903e-06, + "loss": 1.5675, + "step": 29626 + }, + { + "epoch": 6.47, + "learning_rate": 1.7689534629670091e-06, + "loss": 0.9628, + "step": 29627 + }, + { + "epoch": 6.47, + "learning_rate": 1.768844221105528e-06, + "loss": 0.8233, + "step": 29628 + }, + { + "epoch": 6.47, + "learning_rate": 1.7687349792440463e-06, + "loss": 1.5164, + "step": 29629 + }, + { + "epoch": 6.47, + "learning_rate": 1.7686257373825652e-06, + "loss": 0.3193, + "step": 29630 + }, + { + "epoch": 6.47, + "learning_rate": 1.7685164955210838e-06, + "loss": 2.0367, + "step": 29631 + }, + { + "epoch": 6.47, + "learning_rate": 1.7684072536596026e-06, + "loss": 2.0367, + "step": 29632 + }, + { + "epoch": 6.47, + "learning_rate": 1.768298011798121e-06, + "loss": 0.2221, + "step": 29633 + }, + { + "epoch": 6.47, + "learning_rate": 1.7681887699366398e-06, + "loss": 0.8674, + "step": 29634 + }, + { + "epoch": 6.47, + "learning_rate": 1.7680795280751586e-06, + "loss": 0.4519, + "step": 29635 + }, + { + "epoch": 6.47, + "learning_rate": 1.7679702862136772e-06, + "loss": 0.4532, + "step": 29636 + }, + { + "epoch": 6.47, + "learning_rate": 1.767861044352196e-06, + "loss": 1.5979, + "step": 29637 + }, + { + "epoch": 6.47, + "learning_rate": 1.7677518024907145e-06, + "loss": 1.4089, + "step": 29638 + }, + { + "epoch": 6.47, + "learning_rate": 1.7676425606292333e-06, + "loss": 0.6372, + "step": 29639 + }, + { + "epoch": 6.47, + "learning_rate": 1.7675333187677521e-06, + "loss": 0.7425, + "step": 29640 + }, + { + "epoch": 6.47, + "learning_rate": 1.7674240769062705e-06, + "loss": 1.4851, + "step": 29641 + }, + { + "epoch": 6.47, + "learning_rate": 1.7673148350447893e-06, + "loss": 1.4392, + "step": 29642 + }, + { + "epoch": 6.47, + "learning_rate": 1.767205593183308e-06, + "loss": 1.4437, + "step": 29643 + }, + { + "epoch": 6.47, + "learning_rate": 1.7670963513218268e-06, + "loss": 1.9257, + "step": 29644 + }, + { + "epoch": 6.47, + "learning_rate": 1.7669871094603452e-06, + "loss": 0.785, + "step": 29645 + }, + { + "epoch": 6.47, + "learning_rate": 1.766877867598864e-06, + "loss": 1.3708, + "step": 29646 + }, + { + "epoch": 6.47, + "learning_rate": 1.7667686257373828e-06, + "loss": 0.8113, + "step": 29647 + }, + { + "epoch": 6.47, + "learning_rate": 1.7666593838759014e-06, + "loss": 0.2774, + "step": 29648 + }, + { + "epoch": 6.47, + "learning_rate": 1.7665501420144202e-06, + "loss": 0.7174, + "step": 29649 + }, + { + "epoch": 6.47, + "learning_rate": 1.7664409001529386e-06, + "loss": 2.8077, + "step": 29650 + }, + { + "epoch": 6.47, + "learning_rate": 1.7663316582914575e-06, + "loss": 2.2288, + "step": 29651 + }, + { + "epoch": 6.47, + "learning_rate": 1.766222416429976e-06, + "loss": 1.2246, + "step": 29652 + }, + { + "epoch": 6.47, + "learning_rate": 1.7661131745684947e-06, + "loss": 2.2968, + "step": 29653 + }, + { + "epoch": 6.47, + "learning_rate": 1.7660039327070135e-06, + "loss": 1.1511, + "step": 29654 + }, + { + "epoch": 6.47, + "learning_rate": 1.7658946908455321e-06, + "loss": 0.4568, + "step": 29655 + }, + { + "epoch": 6.47, + "learning_rate": 1.765785448984051e-06, + "loss": 2.3117, + "step": 29656 + }, + { + "epoch": 6.47, + "learning_rate": 1.7656762071225693e-06, + "loss": 2.3976, + "step": 29657 + }, + { + "epoch": 6.47, + "learning_rate": 1.7655669652610882e-06, + "loss": 0.614, + "step": 29658 + }, + { + "epoch": 6.47, + "learning_rate": 1.7654577233996068e-06, + "loss": 1.4492, + "step": 29659 + }, + { + "epoch": 6.47, + "learning_rate": 1.7653484815381256e-06, + "loss": 0.8061, + "step": 29660 + }, + { + "epoch": 6.47, + "learning_rate": 1.7652392396766444e-06, + "loss": 1.4683, + "step": 29661 + }, + { + "epoch": 6.47, + "learning_rate": 1.7651299978151628e-06, + "loss": 1.1448, + "step": 29662 + }, + { + "epoch": 6.47, + "learning_rate": 1.7650207559536816e-06, + "loss": 1.2976, + "step": 29663 + }, + { + "epoch": 6.47, + "learning_rate": 1.7649115140922003e-06, + "loss": 0.8199, + "step": 29664 + }, + { + "epoch": 6.47, + "learning_rate": 1.764802272230719e-06, + "loss": 1.1347, + "step": 29665 + }, + { + "epoch": 6.47, + "learning_rate": 1.7646930303692375e-06, + "loss": 0.7135, + "step": 29666 + }, + { + "epoch": 6.47, + "learning_rate": 1.7645837885077563e-06, + "loss": 1.4587, + "step": 29667 + }, + { + "epoch": 6.47, + "learning_rate": 1.7644745466462751e-06, + "loss": 1.0149, + "step": 29668 + }, + { + "epoch": 6.48, + "learning_rate": 1.7643653047847935e-06, + "loss": 1.9961, + "step": 29669 + }, + { + "epoch": 6.48, + "learning_rate": 1.7642560629233123e-06, + "loss": 0.9058, + "step": 29670 + }, + { + "epoch": 6.48, + "learning_rate": 1.764146821061831e-06, + "loss": 1.4518, + "step": 29671 + }, + { + "epoch": 6.48, + "learning_rate": 1.7640375792003498e-06, + "loss": 0.6422, + "step": 29672 + }, + { + "epoch": 6.48, + "learning_rate": 1.7639283373388686e-06, + "loss": 1.2667, + "step": 29673 + }, + { + "epoch": 6.48, + "learning_rate": 1.763819095477387e-06, + "loss": 1.6288, + "step": 29674 + }, + { + "epoch": 6.48, + "learning_rate": 1.7637098536159058e-06, + "loss": 1.4176, + "step": 29675 + }, + { + "epoch": 6.48, + "learning_rate": 1.7636006117544244e-06, + "loss": 1.6249, + "step": 29676 + }, + { + "epoch": 6.48, + "learning_rate": 1.7634913698929433e-06, + "loss": 2.3611, + "step": 29677 + }, + { + "epoch": 6.48, + "learning_rate": 1.7633821280314617e-06, + "loss": 1.4584, + "step": 29678 + }, + { + "epoch": 6.48, + "learning_rate": 1.7632728861699805e-06, + "loss": 1.1104, + "step": 29679 + }, + { + "epoch": 6.48, + "learning_rate": 1.7631636443084993e-06, + "loss": 1.264, + "step": 29680 + }, + { + "epoch": 6.48, + "learning_rate": 1.763054402447018e-06, + "loss": 2.3376, + "step": 29681 + }, + { + "epoch": 6.48, + "learning_rate": 1.7629451605855365e-06, + "loss": 1.4393, + "step": 29682 + }, + { + "epoch": 6.48, + "learning_rate": 1.7628359187240551e-06, + "loss": 1.0823, + "step": 29683 + }, + { + "epoch": 6.48, + "learning_rate": 1.762726676862574e-06, + "loss": 0.2181, + "step": 29684 + }, + { + "epoch": 6.48, + "learning_rate": 1.7626174350010924e-06, + "loss": 1.9658, + "step": 29685 + }, + { + "epoch": 6.48, + "learning_rate": 1.7625081931396112e-06, + "loss": 2.1811, + "step": 29686 + }, + { + "epoch": 6.48, + "learning_rate": 1.76239895127813e-06, + "loss": 1.1368, + "step": 29687 + }, + { + "epoch": 6.48, + "learning_rate": 1.7622897094166486e-06, + "loss": 1.064, + "step": 29688 + }, + { + "epoch": 6.48, + "learning_rate": 1.7621804675551674e-06, + "loss": 0.9455, + "step": 29689 + }, + { + "epoch": 6.48, + "learning_rate": 1.7620712256936858e-06, + "loss": 0.4172, + "step": 29690 + }, + { + "epoch": 6.48, + "learning_rate": 1.7619619838322047e-06, + "loss": 1.3104, + "step": 29691 + }, + { + "epoch": 6.48, + "learning_rate": 1.7618527419707233e-06, + "loss": 1.2873, + "step": 29692 + }, + { + "epoch": 6.48, + "learning_rate": 1.761743500109242e-06, + "loss": 0.2378, + "step": 29693 + }, + { + "epoch": 6.48, + "learning_rate": 1.7616342582477607e-06, + "loss": 2.0691, + "step": 29694 + }, + { + "epoch": 6.48, + "learning_rate": 1.7615250163862793e-06, + "loss": 0.7368, + "step": 29695 + }, + { + "epoch": 6.48, + "learning_rate": 1.7614157745247981e-06, + "loss": 1.2186, + "step": 29696 + }, + { + "epoch": 6.48, + "learning_rate": 1.7613065326633165e-06, + "loss": 1.1144, + "step": 29697 + }, + { + "epoch": 6.48, + "learning_rate": 1.7611972908018354e-06, + "loss": 0.1307, + "step": 29698 + }, + { + "epoch": 6.48, + "learning_rate": 1.761088048940354e-06, + "loss": 1.7453, + "step": 29699 + }, + { + "epoch": 6.48, + "learning_rate": 1.7609788070788728e-06, + "loss": 2.4723, + "step": 29700 + }, + { + "epoch": 6.48, + "learning_rate": 1.7608695652173916e-06, + "loss": 1.0182, + "step": 29701 + }, + { + "epoch": 6.48, + "learning_rate": 1.76076032335591e-06, + "loss": 2.9154, + "step": 29702 + }, + { + "epoch": 6.48, + "learning_rate": 1.7606510814944288e-06, + "loss": 0.8439, + "step": 29703 + }, + { + "epoch": 6.48, + "learning_rate": 1.7605418396329474e-06, + "loss": 1.5344, + "step": 29704 + }, + { + "epoch": 6.48, + "learning_rate": 1.7604325977714663e-06, + "loss": 1.9036, + "step": 29705 + }, + { + "epoch": 6.48, + "learning_rate": 1.7603233559099849e-06, + "loss": 0.1486, + "step": 29706 + }, + { + "epoch": 6.48, + "learning_rate": 1.7602141140485035e-06, + "loss": 1.6426, + "step": 29707 + }, + { + "epoch": 6.48, + "learning_rate": 1.7601048721870223e-06, + "loss": 0.8906, + "step": 29708 + }, + { + "epoch": 6.48, + "learning_rate": 1.759995630325541e-06, + "loss": 1.4055, + "step": 29709 + }, + { + "epoch": 6.48, + "learning_rate": 1.7598863884640595e-06, + "loss": 1.0814, + "step": 29710 + }, + { + "epoch": 6.48, + "learning_rate": 1.7597771466025781e-06, + "loss": 1.9307, + "step": 29711 + }, + { + "epoch": 6.48, + "learning_rate": 1.759667904741097e-06, + "loss": 1.7009, + "step": 29712 + }, + { + "epoch": 6.48, + "learning_rate": 1.7595586628796158e-06, + "loss": 0.5216, + "step": 29713 + }, + { + "epoch": 6.48, + "learning_rate": 1.7594494210181342e-06, + "loss": 1.3182, + "step": 29714 + }, + { + "epoch": 6.49, + "learning_rate": 1.759340179156653e-06, + "loss": 0.4815, + "step": 29715 + }, + { + "epoch": 6.49, + "learning_rate": 1.7592309372951716e-06, + "loss": 1.1421, + "step": 29716 + }, + { + "epoch": 6.49, + "learning_rate": 1.7591216954336905e-06, + "loss": 1.126, + "step": 29717 + }, + { + "epoch": 6.49, + "learning_rate": 1.7590124535722088e-06, + "loss": 1.9155, + "step": 29718 + }, + { + "epoch": 6.49, + "learning_rate": 1.7589032117107277e-06, + "loss": 0.758, + "step": 29719 + }, + { + "epoch": 6.49, + "learning_rate": 1.7587939698492465e-06, + "loss": 2.5634, + "step": 29720 + }, + { + "epoch": 6.49, + "learning_rate": 1.7586847279877651e-06, + "loss": 0.624, + "step": 29721 + }, + { + "epoch": 6.49, + "learning_rate": 1.7585754861262837e-06, + "loss": 2.1229, + "step": 29722 + }, + { + "epoch": 6.49, + "learning_rate": 1.7584662442648023e-06, + "loss": 2.0255, + "step": 29723 + }, + { + "epoch": 6.49, + "learning_rate": 1.7583570024033212e-06, + "loss": 0.1298, + "step": 29724 + }, + { + "epoch": 6.49, + "learning_rate": 1.7582477605418396e-06, + "loss": 1.9579, + "step": 29725 + }, + { + "epoch": 6.49, + "learning_rate": 1.7581385186803584e-06, + "loss": 1.0208, + "step": 29726 + }, + { + "epoch": 6.49, + "learning_rate": 1.7580292768188772e-06, + "loss": 1.9536, + "step": 29727 + }, + { + "epoch": 6.49, + "learning_rate": 1.7579200349573958e-06, + "loss": 1.284, + "step": 29728 + }, + { + "epoch": 6.49, + "learning_rate": 1.7578107930959146e-06, + "loss": 2.2973, + "step": 29729 + }, + { + "epoch": 6.49, + "learning_rate": 1.757701551234433e-06, + "loss": 2.2878, + "step": 29730 + }, + { + "epoch": 6.49, + "learning_rate": 1.7575923093729519e-06, + "loss": 0.3904, + "step": 29731 + }, + { + "epoch": 6.49, + "learning_rate": 1.7574830675114705e-06, + "loss": 0.7018, + "step": 29732 + }, + { + "epoch": 6.49, + "learning_rate": 1.7573738256499893e-06, + "loss": 1.2978, + "step": 29733 + }, + { + "epoch": 6.49, + "learning_rate": 1.7572645837885081e-06, + "loss": 1.6872, + "step": 29734 + }, + { + "epoch": 6.49, + "learning_rate": 1.7571553419270265e-06, + "loss": 1.8848, + "step": 29735 + }, + { + "epoch": 6.49, + "learning_rate": 1.7570461000655453e-06, + "loss": 2.1671, + "step": 29736 + }, + { + "epoch": 6.49, + "learning_rate": 1.756936858204064e-06, + "loss": 0.5273, + "step": 29737 + }, + { + "epoch": 6.49, + "learning_rate": 1.7568276163425826e-06, + "loss": 1.2854, + "step": 29738 + }, + { + "epoch": 6.49, + "learning_rate": 1.7567183744811012e-06, + "loss": 1.6469, + "step": 29739 + }, + { + "epoch": 6.49, + "learning_rate": 1.75660913261962e-06, + "loss": 0.739, + "step": 29740 + }, + { + "epoch": 6.49, + "learning_rate": 1.7564998907581388e-06, + "loss": 2.2903, + "step": 29741 + }, + { + "epoch": 6.49, + "learning_rate": 1.7563906488966572e-06, + "loss": 1.1375, + "step": 29742 + }, + { + "epoch": 6.49, + "learning_rate": 1.756281407035176e-06, + "loss": 0.4372, + "step": 29743 + }, + { + "epoch": 6.49, + "learning_rate": 1.7561721651736946e-06, + "loss": 0.7311, + "step": 29744 + }, + { + "epoch": 6.49, + "learning_rate": 1.7560629233122135e-06, + "loss": 1.532, + "step": 29745 + }, + { + "epoch": 6.49, + "learning_rate": 1.7559536814507323e-06, + "loss": 0.736, + "step": 29746 + }, + { + "epoch": 6.49, + "learning_rate": 1.7558444395892507e-06, + "loss": 0.3411, + "step": 29747 + }, + { + "epoch": 6.49, + "learning_rate": 1.7557351977277695e-06, + "loss": 1.9154, + "step": 29748 + }, + { + "epoch": 6.49, + "learning_rate": 1.7556259558662881e-06, + "loss": 0.4621, + "step": 29749 + }, + { + "epoch": 6.49, + "learning_rate": 1.7555167140048067e-06, + "loss": 0.6508, + "step": 29750 + }, + { + "epoch": 6.49, + "learning_rate": 1.7554074721433253e-06, + "loss": 2.1108, + "step": 29751 + }, + { + "epoch": 6.49, + "learning_rate": 1.7552982302818442e-06, + "loss": 0.9754, + "step": 29752 + }, + { + "epoch": 6.49, + "learning_rate": 1.755188988420363e-06, + "loss": 0.3417, + "step": 29753 + }, + { + "epoch": 6.49, + "learning_rate": 1.7550797465588814e-06, + "loss": 0.3513, + "step": 29754 + }, + { + "epoch": 6.49, + "learning_rate": 1.7549705046974002e-06, + "loss": 1.0461, + "step": 29755 + }, + { + "epoch": 6.49, + "learning_rate": 1.7548612628359188e-06, + "loss": 2.6609, + "step": 29756 + }, + { + "epoch": 6.49, + "learning_rate": 1.7547520209744376e-06, + "loss": 2.1151, + "step": 29757 + }, + { + "epoch": 6.49, + "learning_rate": 1.754642779112956e-06, + "loss": 0.7024, + "step": 29758 + }, + { + "epoch": 6.49, + "learning_rate": 1.7545335372514749e-06, + "loss": 0.7793, + "step": 29759 + }, + { + "epoch": 6.49, + "learning_rate": 1.7544242953899937e-06, + "loss": 1.7482, + "step": 29760 + }, + { + "epoch": 6.5, + "learning_rate": 1.7543150535285123e-06, + "loss": 0.8443, + "step": 29761 + }, + { + "epoch": 6.5, + "learning_rate": 1.7542058116670311e-06, + "loss": 0.8343, + "step": 29762 + }, + { + "epoch": 6.5, + "learning_rate": 1.7540965698055495e-06, + "loss": 0.8093, + "step": 29763 + }, + { + "epoch": 6.5, + "learning_rate": 1.7539873279440683e-06, + "loss": 0.9146, + "step": 29764 + }, + { + "epoch": 6.5, + "learning_rate": 1.753878086082587e-06, + "loss": 2.1589, + "step": 29765 + }, + { + "epoch": 6.5, + "learning_rate": 1.7537688442211056e-06, + "loss": 0.6273, + "step": 29766 + }, + { + "epoch": 6.5, + "learning_rate": 1.7536596023596244e-06, + "loss": 0.7099, + "step": 29767 + }, + { + "epoch": 6.5, + "learning_rate": 1.753550360498143e-06, + "loss": 1.0549, + "step": 29768 + }, + { + "epoch": 6.5, + "learning_rate": 1.7534411186366618e-06, + "loss": 1.1971, + "step": 29769 + }, + { + "epoch": 6.5, + "learning_rate": 1.7533318767751802e-06, + "loss": 1.6372, + "step": 29770 + }, + { + "epoch": 6.5, + "learning_rate": 1.753222634913699e-06, + "loss": 1.409, + "step": 29771 + }, + { + "epoch": 6.5, + "learning_rate": 1.7531133930522177e-06, + "loss": 2.0452, + "step": 29772 + }, + { + "epoch": 6.5, + "learning_rate": 1.7530041511907365e-06, + "loss": 0.937, + "step": 29773 + }, + { + "epoch": 6.5, + "learning_rate": 1.7528949093292553e-06, + "loss": 0.1981, + "step": 29774 + }, + { + "epoch": 6.5, + "learning_rate": 1.7527856674677737e-06, + "loss": 1.013, + "step": 29775 + }, + { + "epoch": 6.5, + "learning_rate": 1.7526764256062925e-06, + "loss": 1.4493, + "step": 29776 + }, + { + "epoch": 6.5, + "learning_rate": 1.7525671837448111e-06, + "loss": 2.1338, + "step": 29777 + }, + { + "epoch": 6.5, + "learning_rate": 1.7524579418833297e-06, + "loss": 0.8972, + "step": 29778 + }, + { + "epoch": 6.5, + "learning_rate": 1.7523487000218486e-06, + "loss": 0.992, + "step": 29779 + }, + { + "epoch": 6.5, + "learning_rate": 1.7522394581603672e-06, + "loss": 2.358, + "step": 29780 + }, + { + "epoch": 6.5, + "learning_rate": 1.752130216298886e-06, + "loss": 1.5371, + "step": 29781 + }, + { + "epoch": 6.5, + "learning_rate": 1.7520209744374044e-06, + "loss": 2.6586, + "step": 29782 + }, + { + "epoch": 6.5, + "learning_rate": 1.7519117325759232e-06, + "loss": 1.3785, + "step": 29783 + }, + { + "epoch": 6.5, + "learning_rate": 1.7518024907144418e-06, + "loss": 1.2133, + "step": 29784 + }, + { + "epoch": 6.5, + "learning_rate": 1.7516932488529607e-06, + "loss": 1.2894, + "step": 29785 + }, + { + "epoch": 6.5, + "learning_rate": 1.7515840069914795e-06, + "loss": 1.3005, + "step": 29786 + }, + { + "epoch": 6.5, + "learning_rate": 1.7514747651299979e-06, + "loss": 1.8179, + "step": 29787 + }, + { + "epoch": 6.5, + "learning_rate": 1.7513655232685167e-06, + "loss": 0.6847, + "step": 29788 + }, + { + "epoch": 6.5, + "learning_rate": 1.7512562814070353e-06, + "loss": 2.6792, + "step": 29789 + }, + { + "epoch": 6.5, + "learning_rate": 1.7511470395455541e-06, + "loss": 1.6236, + "step": 29790 + }, + { + "epoch": 6.5, + "learning_rate": 1.7510377976840725e-06, + "loss": 2.3334, + "step": 29791 + }, + { + "epoch": 6.5, + "learning_rate": 1.7509285558225914e-06, + "loss": 2.1054, + "step": 29792 + }, + { + "epoch": 6.5, + "learning_rate": 1.7508193139611102e-06, + "loss": 2.1718, + "step": 29793 + }, + { + "epoch": 6.5, + "learning_rate": 1.7507100720996286e-06, + "loss": 2.8776, + "step": 29794 + }, + { + "epoch": 6.5, + "learning_rate": 1.7506008302381474e-06, + "loss": 0.4923, + "step": 29795 + }, + { + "epoch": 6.5, + "learning_rate": 1.750491588376666e-06, + "loss": 1.9516, + "step": 29796 + }, + { + "epoch": 6.5, + "learning_rate": 1.7503823465151848e-06, + "loss": 1.2379, + "step": 29797 + }, + { + "epoch": 6.5, + "learning_rate": 1.7502731046537032e-06, + "loss": 1.1045, + "step": 29798 + }, + { + "epoch": 6.5, + "learning_rate": 1.750163862792222e-06, + "loss": 1.2639, + "step": 29799 + }, + { + "epoch": 6.5, + "learning_rate": 1.7500546209307409e-06, + "loss": 0.4095, + "step": 29800 + }, + { + "epoch": 6.5, + "learning_rate": 1.7499453790692595e-06, + "loss": 0.1908, + "step": 29801 + }, + { + "epoch": 6.5, + "learning_rate": 1.7498361372077783e-06, + "loss": 0.0454, + "step": 29802 + }, + { + "epoch": 6.5, + "learning_rate": 1.7497268953462967e-06, + "loss": 0.9952, + "step": 29803 + }, + { + "epoch": 6.5, + "learning_rate": 1.7496176534848155e-06, + "loss": 1.2236, + "step": 29804 + }, + { + "epoch": 6.5, + "learning_rate": 1.7495084116233341e-06, + "loss": 1.7121, + "step": 29805 + }, + { + "epoch": 6.51, + "learning_rate": 1.749399169761853e-06, + "loss": 2.1408, + "step": 29806 + }, + { + "epoch": 6.51, + "learning_rate": 1.7492899279003716e-06, + "loss": 0.5901, + "step": 29807 + }, + { + "epoch": 6.51, + "learning_rate": 1.7491806860388902e-06, + "loss": 2.1177, + "step": 29808 + }, + { + "epoch": 6.51, + "learning_rate": 1.749071444177409e-06, + "loss": 2.2304, + "step": 29809 + }, + { + "epoch": 6.51, + "learning_rate": 1.7489622023159274e-06, + "loss": 0.7811, + "step": 29810 + }, + { + "epoch": 6.51, + "learning_rate": 1.7488529604544462e-06, + "loss": 2.0058, + "step": 29811 + }, + { + "epoch": 6.51, + "learning_rate": 1.748743718592965e-06, + "loss": 1.2117, + "step": 29812 + }, + { + "epoch": 6.51, + "learning_rate": 1.7486344767314837e-06, + "loss": 0.8373, + "step": 29813 + }, + { + "epoch": 6.51, + "learning_rate": 1.7485252348700025e-06, + "loss": 1.34, + "step": 29814 + }, + { + "epoch": 6.51, + "learning_rate": 1.7484159930085209e-06, + "loss": 0.238, + "step": 29815 + }, + { + "epoch": 6.51, + "learning_rate": 1.7483067511470397e-06, + "loss": 1.4046, + "step": 29816 + }, + { + "epoch": 6.51, + "learning_rate": 1.7481975092855583e-06, + "loss": 1.0977, + "step": 29817 + }, + { + "epoch": 6.51, + "learning_rate": 1.7480882674240771e-06, + "loss": 0.7426, + "step": 29818 + }, + { + "epoch": 6.51, + "learning_rate": 1.7479790255625958e-06, + "loss": 1.1131, + "step": 29819 + }, + { + "epoch": 6.51, + "learning_rate": 1.7478697837011144e-06, + "loss": 0.2829, + "step": 29820 + }, + { + "epoch": 6.51, + "learning_rate": 1.7477605418396332e-06, + "loss": 1.6579, + "step": 29821 + }, + { + "epoch": 6.51, + "learning_rate": 1.7476512999781516e-06, + "loss": 1.9122, + "step": 29822 + }, + { + "epoch": 6.51, + "learning_rate": 1.7475420581166704e-06, + "loss": 1.4676, + "step": 29823 + }, + { + "epoch": 6.51, + "learning_rate": 1.747432816255189e-06, + "loss": 1.2127, + "step": 29824 + }, + { + "epoch": 6.51, + "learning_rate": 1.7473235743937078e-06, + "loss": 0.9261, + "step": 29825 + }, + { + "epoch": 6.51, + "learning_rate": 1.7472143325322267e-06, + "loss": 0.4366, + "step": 29826 + }, + { + "epoch": 6.51, + "learning_rate": 1.747105090670745e-06, + "loss": 1.5979, + "step": 29827 + }, + { + "epoch": 6.51, + "learning_rate": 1.7469958488092639e-06, + "loss": 0.3468, + "step": 29828 + }, + { + "epoch": 6.51, + "learning_rate": 1.7468866069477825e-06, + "loss": 1.3415, + "step": 29829 + }, + { + "epoch": 6.51, + "learning_rate": 1.7467773650863013e-06, + "loss": 1.7248, + "step": 29830 + }, + { + "epoch": 6.51, + "learning_rate": 1.7466681232248197e-06, + "loss": 1.5366, + "step": 29831 + }, + { + "epoch": 6.51, + "learning_rate": 1.7465588813633385e-06, + "loss": 3.0141, + "step": 29832 + }, + { + "epoch": 6.51, + "learning_rate": 1.7464496395018574e-06, + "loss": 0.2216, + "step": 29833 + }, + { + "epoch": 6.51, + "learning_rate": 1.746340397640376e-06, + "loss": 2.2337, + "step": 29834 + }, + { + "epoch": 6.51, + "learning_rate": 1.7462311557788946e-06, + "loss": 0.5298, + "step": 29835 + }, + { + "epoch": 6.51, + "learning_rate": 1.7461219139174132e-06, + "loss": 2.4136, + "step": 29836 + }, + { + "epoch": 6.51, + "learning_rate": 1.746012672055932e-06, + "loss": 2.0511, + "step": 29837 + }, + { + "epoch": 6.51, + "learning_rate": 1.7459034301944504e-06, + "loss": 1.9122, + "step": 29838 + }, + { + "epoch": 6.51, + "learning_rate": 1.7457941883329692e-06, + "loss": 0.343, + "step": 29839 + }, + { + "epoch": 6.51, + "learning_rate": 1.745684946471488e-06, + "loss": 2.3074, + "step": 29840 + }, + { + "epoch": 6.51, + "learning_rate": 1.7455757046100067e-06, + "loss": 1.7816, + "step": 29841 + }, + { + "epoch": 6.51, + "learning_rate": 1.7454664627485255e-06, + "loss": 1.5105, + "step": 29842 + }, + { + "epoch": 6.51, + "learning_rate": 1.745357220887044e-06, + "loss": 2.5533, + "step": 29843 + }, + { + "epoch": 6.51, + "learning_rate": 1.7452479790255627e-06, + "loss": 1.7935, + "step": 29844 + }, + { + "epoch": 6.51, + "learning_rate": 1.7451387371640816e-06, + "loss": 1.9564, + "step": 29845 + }, + { + "epoch": 6.51, + "learning_rate": 1.7450294953026002e-06, + "loss": 0.6027, + "step": 29846 + }, + { + "epoch": 6.51, + "learning_rate": 1.7449202534411188e-06, + "loss": 1.4657, + "step": 29847 + }, + { + "epoch": 6.51, + "learning_rate": 1.7448110115796374e-06, + "loss": 0.7733, + "step": 29848 + }, + { + "epoch": 6.51, + "learning_rate": 1.7447017697181562e-06, + "loss": 2.4953, + "step": 29849 + }, + { + "epoch": 6.51, + "learning_rate": 1.7445925278566746e-06, + "loss": 1.7114, + "step": 29850 + }, + { + "epoch": 6.51, + "learning_rate": 1.7444832859951934e-06, + "loss": 0.7327, + "step": 29851 + }, + { + "epoch": 6.52, + "learning_rate": 1.7443740441337123e-06, + "loss": 0.8497, + "step": 29852 + }, + { + "epoch": 6.52, + "learning_rate": 1.7442648022722309e-06, + "loss": 1.3954, + "step": 29853 + }, + { + "epoch": 6.52, + "learning_rate": 1.7441555604107497e-06, + "loss": 1.465, + "step": 29854 + }, + { + "epoch": 6.52, + "learning_rate": 1.744046318549268e-06, + "loss": 1.0956, + "step": 29855 + }, + { + "epoch": 6.52, + "learning_rate": 1.743937076687787e-06, + "loss": 0.6636, + "step": 29856 + }, + { + "epoch": 6.52, + "learning_rate": 1.7438278348263055e-06, + "loss": 1.0204, + "step": 29857 + }, + { + "epoch": 6.52, + "learning_rate": 1.7437185929648243e-06, + "loss": 1.5588, + "step": 29858 + }, + { + "epoch": 6.52, + "learning_rate": 1.7436093511033432e-06, + "loss": 1.42, + "step": 29859 + }, + { + "epoch": 6.52, + "learning_rate": 1.7435001092418616e-06, + "loss": 1.0087, + "step": 29860 + }, + { + "epoch": 6.52, + "learning_rate": 1.7433908673803804e-06, + "loss": 1.3983, + "step": 29861 + }, + { + "epoch": 6.52, + "learning_rate": 1.743281625518899e-06, + "loss": 1.9298, + "step": 29862 + }, + { + "epoch": 6.52, + "learning_rate": 1.7431723836574176e-06, + "loss": 2.353, + "step": 29863 + }, + { + "epoch": 6.52, + "learning_rate": 1.7430631417959362e-06, + "loss": 1.7367, + "step": 29864 + }, + { + "epoch": 6.52, + "learning_rate": 1.742953899934455e-06, + "loss": 0.3587, + "step": 29865 + }, + { + "epoch": 6.52, + "learning_rate": 1.7428446580729739e-06, + "loss": 0.4899, + "step": 29866 + }, + { + "epoch": 6.52, + "learning_rate": 1.7427354162114923e-06, + "loss": 2.4555, + "step": 29867 + }, + { + "epoch": 6.52, + "learning_rate": 1.742626174350011e-06, + "loss": 0.1298, + "step": 29868 + }, + { + "epoch": 6.52, + "learning_rate": 1.7425169324885297e-06, + "loss": 0.7106, + "step": 29869 + }, + { + "epoch": 6.52, + "learning_rate": 1.7424076906270485e-06, + "loss": 2.7801, + "step": 29870 + }, + { + "epoch": 6.52, + "learning_rate": 1.742298448765567e-06, + "loss": 1.7756, + "step": 29871 + }, + { + "epoch": 6.52, + "learning_rate": 1.7421892069040857e-06, + "loss": 0.8444, + "step": 29872 + }, + { + "epoch": 6.52, + "learning_rate": 1.7420799650426046e-06, + "loss": 0.666, + "step": 29873 + }, + { + "epoch": 6.52, + "learning_rate": 1.7419707231811232e-06, + "loss": 1.6805, + "step": 29874 + }, + { + "epoch": 6.52, + "learning_rate": 1.7418614813196418e-06, + "loss": 1.1543, + "step": 29875 + }, + { + "epoch": 6.52, + "learning_rate": 1.7417522394581604e-06, + "loss": 0.8559, + "step": 29876 + }, + { + "epoch": 6.52, + "learning_rate": 1.7416429975966792e-06, + "loss": 2.1732, + "step": 29877 + }, + { + "epoch": 6.52, + "learning_rate": 1.741533755735198e-06, + "loss": 0.5973, + "step": 29878 + }, + { + "epoch": 6.52, + "learning_rate": 1.7414245138737164e-06, + "loss": 1.6812, + "step": 29879 + }, + { + "epoch": 6.52, + "learning_rate": 1.7413152720122353e-06, + "loss": 2.5175, + "step": 29880 + }, + { + "epoch": 6.52, + "learning_rate": 1.7412060301507539e-06, + "loss": 0.7417, + "step": 29881 + }, + { + "epoch": 6.52, + "learning_rate": 1.7410967882892727e-06, + "loss": 0.2971, + "step": 29882 + }, + { + "epoch": 6.52, + "learning_rate": 1.740987546427791e-06, + "loss": 0.3159, + "step": 29883 + }, + { + "epoch": 6.52, + "learning_rate": 1.74087830456631e-06, + "loss": 1.3912, + "step": 29884 + }, + { + "epoch": 6.52, + "learning_rate": 1.7407690627048287e-06, + "loss": 1.8179, + "step": 29885 + }, + { + "epoch": 6.52, + "learning_rate": 1.7406598208433474e-06, + "loss": 2.1673, + "step": 29886 + }, + { + "epoch": 6.52, + "learning_rate": 1.7405505789818662e-06, + "loss": 1.4272, + "step": 29887 + }, + { + "epoch": 6.52, + "learning_rate": 1.7404413371203846e-06, + "loss": 2.6352, + "step": 29888 + }, + { + "epoch": 6.52, + "learning_rate": 1.7403320952589034e-06, + "loss": 1.3297, + "step": 29889 + }, + { + "epoch": 6.52, + "learning_rate": 1.740222853397422e-06, + "loss": 1.5492, + "step": 29890 + }, + { + "epoch": 6.52, + "learning_rate": 1.7401136115359406e-06, + "loss": 1.8215, + "step": 29891 + }, + { + "epoch": 6.52, + "learning_rate": 1.7400043696744594e-06, + "loss": 2.2544, + "step": 29892 + }, + { + "epoch": 6.52, + "learning_rate": 1.739895127812978e-06, + "loss": 1.4455, + "step": 29893 + }, + { + "epoch": 6.52, + "learning_rate": 1.7397858859514969e-06, + "loss": 1.0842, + "step": 29894 + }, + { + "epoch": 6.52, + "learning_rate": 1.7396766440900153e-06, + "loss": 1.5969, + "step": 29895 + }, + { + "epoch": 6.52, + "learning_rate": 1.739567402228534e-06, + "loss": 0.9079, + "step": 29896 + }, + { + "epoch": 6.52, + "learning_rate": 1.7394581603670527e-06, + "loss": 1.533, + "step": 29897 + }, + { + "epoch": 6.53, + "learning_rate": 1.7393489185055715e-06, + "loss": 1.4162, + "step": 29898 + }, + { + "epoch": 6.53, + "learning_rate": 1.7392396766440904e-06, + "loss": 0.8148, + "step": 29899 + }, + { + "epoch": 6.53, + "learning_rate": 1.7391304347826088e-06, + "loss": 2.0662, + "step": 29900 + }, + { + "epoch": 6.53, + "learning_rate": 1.7390211929211276e-06, + "loss": 1.6195, + "step": 29901 + }, + { + "epoch": 6.53, + "learning_rate": 1.7389119510596462e-06, + "loss": 0.9518, + "step": 29902 + }, + { + "epoch": 6.53, + "learning_rate": 1.738802709198165e-06, + "loss": 3.2777, + "step": 29903 + }, + { + "epoch": 6.53, + "learning_rate": 1.7386934673366834e-06, + "loss": 1.3709, + "step": 29904 + }, + { + "epoch": 6.53, + "learning_rate": 1.7385842254752022e-06, + "loss": 1.4259, + "step": 29905 + }, + { + "epoch": 6.53, + "learning_rate": 1.738474983613721e-06, + "loss": 1.3949, + "step": 29906 + }, + { + "epoch": 6.53, + "learning_rate": 1.7383657417522395e-06, + "loss": 1.9632, + "step": 29907 + }, + { + "epoch": 6.53, + "learning_rate": 1.7382564998907583e-06, + "loss": 0.7227, + "step": 29908 + }, + { + "epoch": 6.53, + "learning_rate": 1.7381472580292769e-06, + "loss": 1.1724, + "step": 29909 + }, + { + "epoch": 6.53, + "learning_rate": 1.7380380161677957e-06, + "loss": 1.4128, + "step": 29910 + }, + { + "epoch": 6.53, + "learning_rate": 1.7379287743063145e-06, + "loss": 1.3147, + "step": 29911 + }, + { + "epoch": 6.53, + "learning_rate": 1.737819532444833e-06, + "loss": 2.4898, + "step": 29912 + }, + { + "epoch": 6.53, + "learning_rate": 1.7377102905833518e-06, + "loss": 1.4416, + "step": 29913 + }, + { + "epoch": 6.53, + "learning_rate": 1.7376010487218704e-06, + "loss": 0.9391, + "step": 29914 + }, + { + "epoch": 6.53, + "learning_rate": 1.7374918068603892e-06, + "loss": 3.41, + "step": 29915 + }, + { + "epoch": 6.53, + "learning_rate": 1.7373825649989076e-06, + "loss": 1.8762, + "step": 29916 + }, + { + "epoch": 6.53, + "learning_rate": 1.7372733231374264e-06, + "loss": 1.2323, + "step": 29917 + }, + { + "epoch": 6.53, + "learning_rate": 1.7371640812759452e-06, + "loss": 0.5285, + "step": 29918 + }, + { + "epoch": 6.53, + "learning_rate": 1.7370548394144636e-06, + "loss": 1.6513, + "step": 29919 + }, + { + "epoch": 6.53, + "learning_rate": 1.7369455975529825e-06, + "loss": 1.3629, + "step": 29920 + }, + { + "epoch": 6.53, + "learning_rate": 1.736836355691501e-06, + "loss": 1.2143, + "step": 29921 + }, + { + "epoch": 6.53, + "learning_rate": 1.7367271138300199e-06, + "loss": 0.7741, + "step": 29922 + }, + { + "epoch": 6.53, + "learning_rate": 1.7366178719685383e-06, + "loss": 0.4761, + "step": 29923 + }, + { + "epoch": 6.53, + "learning_rate": 1.7365086301070571e-06, + "loss": 0.2164, + "step": 29924 + }, + { + "epoch": 6.53, + "learning_rate": 1.736399388245576e-06, + "loss": 3.3383, + "step": 29925 + }, + { + "epoch": 6.53, + "learning_rate": 1.7362901463840945e-06, + "loss": 1.2705, + "step": 29926 + }, + { + "epoch": 6.53, + "learning_rate": 1.7361809045226134e-06, + "loss": 0.423, + "step": 29927 + }, + { + "epoch": 6.53, + "learning_rate": 1.7360716626611318e-06, + "loss": 0.5894, + "step": 29928 + }, + { + "epoch": 6.53, + "learning_rate": 1.7359624207996506e-06, + "loss": 0.3885, + "step": 29929 + }, + { + "epoch": 6.53, + "learning_rate": 1.7358531789381692e-06, + "loss": 0.3892, + "step": 29930 + }, + { + "epoch": 6.53, + "learning_rate": 1.735743937076688e-06, + "loss": 0.4694, + "step": 29931 + }, + { + "epoch": 6.53, + "learning_rate": 1.7356346952152066e-06, + "loss": 0.9819, + "step": 29932 + }, + { + "epoch": 6.53, + "learning_rate": 1.7355254533537252e-06, + "loss": 1.56, + "step": 29933 + }, + { + "epoch": 6.53, + "learning_rate": 1.735416211492244e-06, + "loss": 1.1975, + "step": 29934 + }, + { + "epoch": 6.53, + "learning_rate": 1.7353069696307625e-06, + "loss": 1.9927, + "step": 29935 + }, + { + "epoch": 6.53, + "learning_rate": 1.7351977277692813e-06, + "loss": 0.2175, + "step": 29936 + }, + { + "epoch": 6.53, + "learning_rate": 1.7350884859078e-06, + "loss": 1.9072, + "step": 29937 + }, + { + "epoch": 6.53, + "learning_rate": 1.7349792440463187e-06, + "loss": 0.5661, + "step": 29938 + }, + { + "epoch": 6.53, + "learning_rate": 1.7348700021848375e-06, + "loss": 1.4453, + "step": 29939 + }, + { + "epoch": 6.53, + "learning_rate": 1.734760760323356e-06, + "loss": 1.7911, + "step": 29940 + }, + { + "epoch": 6.53, + "learning_rate": 1.7346515184618748e-06, + "loss": 0.9637, + "step": 29941 + }, + { + "epoch": 6.53, + "learning_rate": 1.7345422766003934e-06, + "loss": 1.1279, + "step": 29942 + }, + { + "epoch": 6.53, + "learning_rate": 1.7344330347389122e-06, + "loss": 0.5803, + "step": 29943 + }, + { + "epoch": 6.54, + "learning_rate": 1.7343237928774308e-06, + "loss": 0.3456, + "step": 29944 + }, + { + "epoch": 6.54, + "learning_rate": 1.7342145510159494e-06, + "loss": 0.7337, + "step": 29945 + }, + { + "epoch": 6.54, + "learning_rate": 1.7341053091544682e-06, + "loss": 0.386, + "step": 29946 + }, + { + "epoch": 6.54, + "learning_rate": 1.7339960672929866e-06, + "loss": 1.8307, + "step": 29947 + }, + { + "epoch": 6.54, + "learning_rate": 1.7338868254315055e-06, + "loss": 1.0302, + "step": 29948 + }, + { + "epoch": 6.54, + "learning_rate": 1.733777583570024e-06, + "loss": 3.1563, + "step": 29949 + }, + { + "epoch": 6.54, + "learning_rate": 1.733668341708543e-06, + "loss": 1.3244, + "step": 29950 + }, + { + "epoch": 6.54, + "learning_rate": 1.7335590998470617e-06, + "loss": 1.1758, + "step": 29951 + }, + { + "epoch": 6.54, + "learning_rate": 1.7334498579855801e-06, + "loss": 1.2717, + "step": 29952 + }, + { + "epoch": 6.54, + "learning_rate": 1.733340616124099e-06, + "loss": 1.3441, + "step": 29953 + }, + { + "epoch": 6.54, + "learning_rate": 1.7332313742626176e-06, + "loss": 1.1242, + "step": 29954 + }, + { + "epoch": 6.54, + "learning_rate": 1.7331221324011364e-06, + "loss": 2.3724, + "step": 29955 + }, + { + "epoch": 6.54, + "learning_rate": 1.7330128905396548e-06, + "loss": 1.2195, + "step": 29956 + }, + { + "epoch": 6.54, + "learning_rate": 1.7329036486781736e-06, + "loss": 1.3351, + "step": 29957 + }, + { + "epoch": 6.54, + "learning_rate": 1.7327944068166924e-06, + "loss": 1.2161, + "step": 29958 + }, + { + "epoch": 6.54, + "learning_rate": 1.732685164955211e-06, + "loss": 0.6504, + "step": 29959 + }, + { + "epoch": 6.54, + "learning_rate": 1.7325759230937296e-06, + "loss": 1.3431, + "step": 29960 + }, + { + "epoch": 6.54, + "learning_rate": 1.7324666812322483e-06, + "loss": 1.4847, + "step": 29961 + }, + { + "epoch": 6.54, + "learning_rate": 1.732357439370767e-06, + "loss": 1.3357, + "step": 29962 + }, + { + "epoch": 6.54, + "learning_rate": 1.7322481975092855e-06, + "loss": 0.8708, + "step": 29963 + }, + { + "epoch": 6.54, + "learning_rate": 1.7321389556478043e-06, + "loss": 0.8349, + "step": 29964 + }, + { + "epoch": 6.54, + "learning_rate": 1.7320297137863231e-06, + "loss": 1.4304, + "step": 29965 + }, + { + "epoch": 6.54, + "learning_rate": 1.7319204719248417e-06, + "loss": 1.147, + "step": 29966 + }, + { + "epoch": 6.54, + "learning_rate": 1.7318112300633606e-06, + "loss": 1.4207, + "step": 29967 + }, + { + "epoch": 6.54, + "learning_rate": 1.731701988201879e-06, + "loss": 2.1158, + "step": 29968 + }, + { + "epoch": 6.54, + "learning_rate": 1.7315927463403978e-06, + "loss": 0.8278, + "step": 29969 + }, + { + "epoch": 6.54, + "learning_rate": 1.7314835044789164e-06, + "loss": 0.3164, + "step": 29970 + }, + { + "epoch": 6.54, + "learning_rate": 1.7313742626174352e-06, + "loss": 0.5387, + "step": 29971 + }, + { + "epoch": 6.54, + "learning_rate": 1.7312650207559538e-06, + "loss": 1.1863, + "step": 29972 + }, + { + "epoch": 6.54, + "learning_rate": 1.7311557788944724e-06, + "loss": 0.2902, + "step": 29973 + }, + { + "epoch": 6.54, + "learning_rate": 1.7310465370329913e-06, + "loss": 1.403, + "step": 29974 + }, + { + "epoch": 6.54, + "learning_rate": 1.7309372951715099e-06, + "loss": 0.8492, + "step": 29975 + }, + { + "epoch": 6.54, + "learning_rate": 1.7308280533100285e-06, + "loss": 1.2503, + "step": 29976 + }, + { + "epoch": 6.54, + "learning_rate": 1.730718811448547e-06, + "loss": 1.3102, + "step": 29977 + }, + { + "epoch": 6.54, + "learning_rate": 1.730609569587066e-06, + "loss": 1.6327, + "step": 29978 + }, + { + "epoch": 6.54, + "learning_rate": 1.7305003277255847e-06, + "loss": 0.7775, + "step": 29979 + }, + { + "epoch": 6.54, + "learning_rate": 1.7303910858641031e-06, + "loss": 2.2502, + "step": 29980 + }, + { + "epoch": 6.54, + "learning_rate": 1.730281844002622e-06, + "loss": 1.6438, + "step": 29981 + }, + { + "epoch": 6.54, + "learning_rate": 1.7301726021411406e-06, + "loss": 1.0195, + "step": 29982 + }, + { + "epoch": 6.54, + "learning_rate": 1.7300633602796594e-06, + "loss": 0.9147, + "step": 29983 + }, + { + "epoch": 6.54, + "learning_rate": 1.7299541184181782e-06, + "loss": 0.7762, + "step": 29984 + }, + { + "epoch": 6.54, + "learning_rate": 1.7298448765566966e-06, + "loss": 1.1502, + "step": 29985 + }, + { + "epoch": 6.54, + "learning_rate": 1.7297356346952154e-06, + "loss": 1.3301, + "step": 29986 + }, + { + "epoch": 6.54, + "learning_rate": 1.729626392833734e-06, + "loss": 1.1047, + "step": 29987 + }, + { + "epoch": 6.54, + "learning_rate": 1.7295171509722527e-06, + "loss": 2.008, + "step": 29988 + }, + { + "epoch": 6.54, + "learning_rate": 1.7294079091107713e-06, + "loss": 0.7417, + "step": 29989 + }, + { + "epoch": 6.55, + "learning_rate": 1.72929866724929e-06, + "loss": 1.3444, + "step": 29990 + }, + { + "epoch": 6.55, + "learning_rate": 1.729189425387809e-06, + "loss": 1.4018, + "step": 29991 + }, + { + "epoch": 6.55, + "learning_rate": 1.7290801835263273e-06, + "loss": 1.2996, + "step": 29992 + }, + { + "epoch": 6.55, + "learning_rate": 1.7289709416648461e-06, + "loss": 0.6148, + "step": 29993 + }, + { + "epoch": 6.55, + "learning_rate": 1.7288616998033648e-06, + "loss": 1.8381, + "step": 29994 + }, + { + "epoch": 6.55, + "learning_rate": 1.7287524579418836e-06, + "loss": 2.5877, + "step": 29995 + }, + { + "epoch": 6.55, + "learning_rate": 1.728643216080402e-06, + "loss": 0.4272, + "step": 29996 + }, + { + "epoch": 6.55, + "learning_rate": 1.7285339742189208e-06, + "loss": 0.956, + "step": 29997 + }, + { + "epoch": 6.55, + "learning_rate": 1.7284247323574396e-06, + "loss": 0.3559, + "step": 29998 + }, + { + "epoch": 6.55, + "learning_rate": 1.7283154904959582e-06, + "loss": 1.9005, + "step": 29999 + }, + { + "epoch": 6.55, + "learning_rate": 1.7282062486344768e-06, + "loss": 0.9479, + "step": 30000 + }, + { + "epoch": 6.55, + "learning_rate": 1.7280970067729955e-06, + "loss": 0.7882, + "step": 30001 + }, + { + "epoch": 6.55, + "learning_rate": 1.7279877649115143e-06, + "loss": 0.596, + "step": 30002 + }, + { + "epoch": 6.55, + "learning_rate": 1.7278785230500329e-06, + "loss": 2.8908, + "step": 30003 + }, + { + "epoch": 6.55, + "learning_rate": 1.7277692811885515e-06, + "loss": 2.5516, + "step": 30004 + }, + { + "epoch": 6.55, + "learning_rate": 1.7276600393270703e-06, + "loss": 1.0857, + "step": 30005 + }, + { + "epoch": 6.55, + "learning_rate": 1.727550797465589e-06, + "loss": 0.7796, + "step": 30006 + }, + { + "epoch": 6.55, + "learning_rate": 1.7274415556041078e-06, + "loss": 0.8057, + "step": 30007 + }, + { + "epoch": 6.55, + "learning_rate": 1.7273323137426262e-06, + "loss": 1.4674, + "step": 30008 + }, + { + "epoch": 6.55, + "learning_rate": 1.727223071881145e-06, + "loss": 1.4489, + "step": 30009 + }, + { + "epoch": 6.55, + "learning_rate": 1.7271138300196636e-06, + "loss": 0.551, + "step": 30010 + }, + { + "epoch": 6.55, + "learning_rate": 1.7270045881581824e-06, + "loss": 1.9275, + "step": 30011 + }, + { + "epoch": 6.55, + "learning_rate": 1.7268953462967012e-06, + "loss": 2.7102, + "step": 30012 + }, + { + "epoch": 6.55, + "learning_rate": 1.7267861044352196e-06, + "loss": 1.7303, + "step": 30013 + }, + { + "epoch": 6.55, + "learning_rate": 1.7266768625737385e-06, + "loss": 0.8151, + "step": 30014 + }, + { + "epoch": 6.55, + "learning_rate": 1.726567620712257e-06, + "loss": 1.8108, + "step": 30015 + }, + { + "epoch": 6.55, + "learning_rate": 1.7264583788507757e-06, + "loss": 0.8623, + "step": 30016 + }, + { + "epoch": 6.55, + "learning_rate": 1.7263491369892945e-06, + "loss": 1.5862, + "step": 30017 + }, + { + "epoch": 6.55, + "learning_rate": 1.7262398951278131e-06, + "loss": 1.9382, + "step": 30018 + }, + { + "epoch": 6.55, + "learning_rate": 1.726130653266332e-06, + "loss": 1.7484, + "step": 30019 + }, + { + "epoch": 6.55, + "learning_rate": 1.7260214114048503e-06, + "loss": 0.9127, + "step": 30020 + }, + { + "epoch": 6.55, + "learning_rate": 1.7259121695433692e-06, + "loss": 1.5992, + "step": 30021 + }, + { + "epoch": 6.55, + "learning_rate": 1.7258029276818878e-06, + "loss": 0.7631, + "step": 30022 + }, + { + "epoch": 6.55, + "learning_rate": 1.7256936858204066e-06, + "loss": 1.5089, + "step": 30023 + }, + { + "epoch": 6.55, + "learning_rate": 1.7255844439589254e-06, + "loss": 1.7251, + "step": 30024 + }, + { + "epoch": 6.55, + "learning_rate": 1.7254752020974438e-06, + "loss": 1.2667, + "step": 30025 + }, + { + "epoch": 6.55, + "learning_rate": 1.7253659602359626e-06, + "loss": 1.2329, + "step": 30026 + }, + { + "epoch": 6.55, + "learning_rate": 1.7252567183744812e-06, + "loss": 0.4807, + "step": 30027 + }, + { + "epoch": 6.55, + "learning_rate": 1.725147476513e-06, + "loss": 1.6345, + "step": 30028 + }, + { + "epoch": 6.55, + "learning_rate": 1.7250382346515185e-06, + "loss": 2.4803, + "step": 30029 + }, + { + "epoch": 6.55, + "learning_rate": 1.7249289927900373e-06, + "loss": 1.1105, + "step": 30030 + }, + { + "epoch": 6.55, + "learning_rate": 1.7248197509285561e-06, + "loss": 1.3393, + "step": 30031 + }, + { + "epoch": 6.55, + "learning_rate": 1.7247105090670745e-06, + "loss": 1.6341, + "step": 30032 + }, + { + "epoch": 6.55, + "learning_rate": 1.7246012672055933e-06, + "loss": 1.436, + "step": 30033 + }, + { + "epoch": 6.55, + "learning_rate": 1.724492025344112e-06, + "loss": 2.0444, + "step": 30034 + }, + { + "epoch": 6.55, + "learning_rate": 1.7243827834826308e-06, + "loss": 1.3671, + "step": 30035 + }, + { + "epoch": 6.56, + "learning_rate": 1.7242735416211492e-06, + "loss": 0.8718, + "step": 30036 + }, + { + "epoch": 6.56, + "learning_rate": 1.724164299759668e-06, + "loss": 0.5068, + "step": 30037 + }, + { + "epoch": 6.56, + "learning_rate": 1.7240550578981868e-06, + "loss": 1.9653, + "step": 30038 + }, + { + "epoch": 6.56, + "learning_rate": 1.7239458160367054e-06, + "loss": 0.7861, + "step": 30039 + }, + { + "epoch": 6.56, + "learning_rate": 1.7238365741752242e-06, + "loss": 2.6634, + "step": 30040 + }, + { + "epoch": 6.56, + "learning_rate": 1.7237273323137426e-06, + "loss": 1.6857, + "step": 30041 + }, + { + "epoch": 6.56, + "learning_rate": 1.7236180904522615e-06, + "loss": 0.5333, + "step": 30042 + }, + { + "epoch": 6.56, + "learning_rate": 1.72350884859078e-06, + "loss": 1.1594, + "step": 30043 + }, + { + "epoch": 6.56, + "learning_rate": 1.7233996067292987e-06, + "loss": 0.8769, + "step": 30044 + }, + { + "epoch": 6.56, + "learning_rate": 1.7232903648678175e-06, + "loss": 1.001, + "step": 30045 + }, + { + "epoch": 6.56, + "learning_rate": 1.7231811230063361e-06, + "loss": 0.5773, + "step": 30046 + }, + { + "epoch": 6.56, + "learning_rate": 1.723071881144855e-06, + "loss": 0.1669, + "step": 30047 + }, + { + "epoch": 6.56, + "learning_rate": 1.7229626392833733e-06, + "loss": 1.4076, + "step": 30048 + }, + { + "epoch": 6.56, + "learning_rate": 1.7228533974218922e-06, + "loss": 1.1342, + "step": 30049 + }, + { + "epoch": 6.56, + "learning_rate": 1.722744155560411e-06, + "loss": 0.8853, + "step": 30050 + }, + { + "epoch": 6.56, + "learning_rate": 1.7226349136989296e-06, + "loss": 1.6519, + "step": 30051 + }, + { + "epoch": 6.56, + "learning_rate": 1.7225256718374484e-06, + "loss": 1.1135, + "step": 30052 + }, + { + "epoch": 6.56, + "learning_rate": 1.7224164299759668e-06, + "loss": 0.5904, + "step": 30053 + }, + { + "epoch": 6.56, + "learning_rate": 1.7223071881144856e-06, + "loss": 0.8445, + "step": 30054 + }, + { + "epoch": 6.56, + "learning_rate": 1.7221979462530043e-06, + "loss": 1.3225, + "step": 30055 + }, + { + "epoch": 6.56, + "learning_rate": 1.722088704391523e-06, + "loss": 1.9798, + "step": 30056 + }, + { + "epoch": 6.56, + "learning_rate": 1.7219794625300417e-06, + "loss": 1.8086, + "step": 30057 + }, + { + "epoch": 6.56, + "learning_rate": 1.7218702206685603e-06, + "loss": 2.7711, + "step": 30058 + }, + { + "epoch": 6.56, + "learning_rate": 1.7217609788070791e-06, + "loss": 0.7474, + "step": 30059 + }, + { + "epoch": 6.56, + "learning_rate": 1.7216517369455975e-06, + "loss": 1.1696, + "step": 30060 + }, + { + "epoch": 6.56, + "learning_rate": 1.7215424950841163e-06, + "loss": 1.4981, + "step": 30061 + }, + { + "epoch": 6.56, + "learning_rate": 1.721433253222635e-06, + "loss": 1.1027, + "step": 30062 + }, + { + "epoch": 6.56, + "learning_rate": 1.7213240113611538e-06, + "loss": 1.1432, + "step": 30063 + }, + { + "epoch": 6.56, + "learning_rate": 1.7212147694996726e-06, + "loss": 1.0645, + "step": 30064 + }, + { + "epoch": 6.56, + "learning_rate": 1.721105527638191e-06, + "loss": 0.2609, + "step": 30065 + }, + { + "epoch": 6.56, + "learning_rate": 1.7209962857767098e-06, + "loss": 2.4655, + "step": 30066 + }, + { + "epoch": 6.56, + "learning_rate": 1.7208870439152284e-06, + "loss": 1.4971, + "step": 30067 + }, + { + "epoch": 6.56, + "learning_rate": 1.7207778020537473e-06, + "loss": 2.7147, + "step": 30068 + }, + { + "epoch": 6.56, + "learning_rate": 1.7206685601922657e-06, + "loss": 2.0327, + "step": 30069 + }, + { + "epoch": 6.56, + "learning_rate": 1.7205593183307845e-06, + "loss": 0.5354, + "step": 30070 + }, + { + "epoch": 6.56, + "learning_rate": 1.7204500764693033e-06, + "loss": 1.3453, + "step": 30071 + }, + { + "epoch": 6.56, + "learning_rate": 1.7203408346078217e-06, + "loss": 0.7993, + "step": 30072 + }, + { + "epoch": 6.56, + "learning_rate": 1.7202315927463405e-06, + "loss": 2.2347, + "step": 30073 + }, + { + "epoch": 6.56, + "learning_rate": 1.7201223508848591e-06, + "loss": 1.384, + "step": 30074 + }, + { + "epoch": 6.56, + "learning_rate": 1.720013109023378e-06, + "loss": 0.0689, + "step": 30075 + }, + { + "epoch": 6.56, + "learning_rate": 1.7199038671618964e-06, + "loss": 1.4664, + "step": 30076 + }, + { + "epoch": 6.56, + "learning_rate": 1.7197946253004152e-06, + "loss": 1.4072, + "step": 30077 + }, + { + "epoch": 6.56, + "learning_rate": 1.719685383438934e-06, + "loss": 1.1339, + "step": 30078 + }, + { + "epoch": 6.56, + "learning_rate": 1.7195761415774526e-06, + "loss": 0.7578, + "step": 30079 + }, + { + "epoch": 6.56, + "learning_rate": 1.7194668997159714e-06, + "loss": 1.0844, + "step": 30080 + }, + { + "epoch": 6.57, + "learning_rate": 1.7193576578544898e-06, + "loss": 1.4862, + "step": 30081 + }, + { + "epoch": 6.57, + "learning_rate": 1.7192484159930087e-06, + "loss": 1.8454, + "step": 30082 + }, + { + "epoch": 6.57, + "learning_rate": 1.7191391741315275e-06, + "loss": 1.674, + "step": 30083 + }, + { + "epoch": 6.57, + "learning_rate": 1.719029932270046e-06, + "loss": 1.881, + "step": 30084 + }, + { + "epoch": 6.57, + "learning_rate": 1.7189206904085647e-06, + "loss": 1.3572, + "step": 30085 + }, + { + "epoch": 6.57, + "learning_rate": 1.7188114485470833e-06, + "loss": 0.9575, + "step": 30086 + }, + { + "epoch": 6.57, + "learning_rate": 1.7187022066856021e-06, + "loss": 2.1852, + "step": 30087 + }, + { + "epoch": 6.57, + "learning_rate": 1.7185929648241205e-06, + "loss": 3.8947, + "step": 30088 + }, + { + "epoch": 6.57, + "learning_rate": 1.7184837229626394e-06, + "loss": 1.2163, + "step": 30089 + }, + { + "epoch": 6.57, + "learning_rate": 1.7183744811011582e-06, + "loss": 0.5698, + "step": 30090 + }, + { + "epoch": 6.57, + "learning_rate": 1.7182652392396768e-06, + "loss": 1.2821, + "step": 30091 + }, + { + "epoch": 6.57, + "learning_rate": 1.7181559973781956e-06, + "loss": 2.5259, + "step": 30092 + }, + { + "epoch": 6.57, + "learning_rate": 1.718046755516714e-06, + "loss": 0.9996, + "step": 30093 + }, + { + "epoch": 6.57, + "learning_rate": 1.7179375136552328e-06, + "loss": 0.8925, + "step": 30094 + }, + { + "epoch": 6.57, + "learning_rate": 1.7178282717937515e-06, + "loss": 0.8075, + "step": 30095 + }, + { + "epoch": 6.57, + "learning_rate": 1.7177190299322703e-06, + "loss": 1.9486, + "step": 30096 + }, + { + "epoch": 6.57, + "learning_rate": 1.7176097880707889e-06, + "loss": 0.5472, + "step": 30097 + }, + { + "epoch": 6.57, + "learning_rate": 1.7175005462093075e-06, + "loss": 1.2241, + "step": 30098 + }, + { + "epoch": 6.57, + "learning_rate": 1.7173913043478263e-06, + "loss": 0.6468, + "step": 30099 + }, + { + "epoch": 6.57, + "learning_rate": 1.717282062486345e-06, + "loss": 1.1336, + "step": 30100 + }, + { + "epoch": 6.57, + "learning_rate": 1.7171728206248635e-06, + "loss": 0.945, + "step": 30101 + }, + { + "epoch": 6.57, + "learning_rate": 1.7170635787633822e-06, + "loss": 1.5413, + "step": 30102 + }, + { + "epoch": 6.57, + "learning_rate": 1.716954336901901e-06, + "loss": 2.1387, + "step": 30103 + }, + { + "epoch": 6.57, + "learning_rate": 1.7168450950404198e-06, + "loss": 4.3618, + "step": 30104 + }, + { + "epoch": 6.57, + "learning_rate": 1.7167358531789382e-06, + "loss": 1.0257, + "step": 30105 + }, + { + "epoch": 6.57, + "learning_rate": 1.716626611317457e-06, + "loss": 1.1719, + "step": 30106 + }, + { + "epoch": 6.57, + "learning_rate": 1.7165173694559756e-06, + "loss": 0.816, + "step": 30107 + }, + { + "epoch": 6.57, + "learning_rate": 1.7164081275944945e-06, + "loss": 1.2014, + "step": 30108 + }, + { + "epoch": 6.57, + "learning_rate": 1.7162988857330129e-06, + "loss": 0.9382, + "step": 30109 + }, + { + "epoch": 6.57, + "learning_rate": 1.7161896438715317e-06, + "loss": 0.2237, + "step": 30110 + }, + { + "epoch": 6.57, + "learning_rate": 1.7160804020100505e-06, + "loss": 1.7319, + "step": 30111 + }, + { + "epoch": 6.57, + "learning_rate": 1.7159711601485691e-06, + "loss": 0.5592, + "step": 30112 + }, + { + "epoch": 6.57, + "learning_rate": 1.7158619182870877e-06, + "loss": 1.6358, + "step": 30113 + }, + { + "epoch": 6.57, + "learning_rate": 1.7157526764256063e-06, + "loss": 0.8274, + "step": 30114 + }, + { + "epoch": 6.57, + "learning_rate": 1.7156434345641252e-06, + "loss": 1.0238, + "step": 30115 + }, + { + "epoch": 6.57, + "learning_rate": 1.715534192702644e-06, + "loss": 0.9148, + "step": 30116 + }, + { + "epoch": 6.57, + "learning_rate": 1.7154249508411624e-06, + "loss": 2.8501, + "step": 30117 + }, + { + "epoch": 6.57, + "learning_rate": 1.7153157089796812e-06, + "loss": 0.2359, + "step": 30118 + }, + { + "epoch": 6.57, + "learning_rate": 1.7152064671181998e-06, + "loss": 1.0442, + "step": 30119 + }, + { + "epoch": 6.57, + "learning_rate": 1.7150972252567186e-06, + "loss": 2.4694, + "step": 30120 + }, + { + "epoch": 6.57, + "learning_rate": 1.714987983395237e-06, + "loss": 2.6479, + "step": 30121 + }, + { + "epoch": 6.57, + "learning_rate": 1.7148787415337559e-06, + "loss": 2.3317, + "step": 30122 + }, + { + "epoch": 6.57, + "learning_rate": 1.7147694996722747e-06, + "loss": 1.6511, + "step": 30123 + }, + { + "epoch": 6.57, + "learning_rate": 1.7146602578107933e-06, + "loss": 0.4058, + "step": 30124 + }, + { + "epoch": 6.57, + "learning_rate": 1.714551015949312e-06, + "loss": 2.6269, + "step": 30125 + }, + { + "epoch": 6.57, + "learning_rate": 1.7144417740878305e-06, + "loss": 2.1828, + "step": 30126 + }, + { + "epoch": 6.58, + "learning_rate": 1.7143325322263493e-06, + "loss": 1.4124, + "step": 30127 + }, + { + "epoch": 6.58, + "learning_rate": 1.714223290364868e-06, + "loss": 0.9081, + "step": 30128 + }, + { + "epoch": 6.58, + "learning_rate": 1.7141140485033866e-06, + "loss": 1.9125, + "step": 30129 + }, + { + "epoch": 6.58, + "learning_rate": 1.7140048066419054e-06, + "loss": 1.2449, + "step": 30130 + }, + { + "epoch": 6.58, + "learning_rate": 1.713895564780424e-06, + "loss": 1.4328, + "step": 30131 + }, + { + "epoch": 6.58, + "learning_rate": 1.7137863229189428e-06, + "loss": 2.0965, + "step": 30132 + }, + { + "epoch": 6.58, + "learning_rate": 1.7136770810574612e-06, + "loss": 1.3448, + "step": 30133 + }, + { + "epoch": 6.58, + "learning_rate": 1.71356783919598e-06, + "loss": 1.081, + "step": 30134 + }, + { + "epoch": 6.58, + "learning_rate": 1.7134585973344986e-06, + "loss": 1.2284, + "step": 30135 + }, + { + "epoch": 6.58, + "learning_rate": 1.7133493554730175e-06, + "loss": 0.8329, + "step": 30136 + }, + { + "epoch": 6.58, + "learning_rate": 1.7132401136115363e-06, + "loss": 1.3589, + "step": 30137 + }, + { + "epoch": 6.58, + "learning_rate": 1.7131308717500547e-06, + "loss": 1.4582, + "step": 30138 + }, + { + "epoch": 6.58, + "learning_rate": 1.7130216298885735e-06, + "loss": 1.0364, + "step": 30139 + }, + { + "epoch": 6.58, + "learning_rate": 1.7129123880270921e-06, + "loss": 2.4703, + "step": 30140 + }, + { + "epoch": 6.58, + "learning_rate": 1.7128031461656107e-06, + "loss": 1.7855, + "step": 30141 + }, + { + "epoch": 6.58, + "learning_rate": 1.7126939043041293e-06, + "loss": 0.3165, + "step": 30142 + }, + { + "epoch": 6.58, + "learning_rate": 1.7125846624426482e-06, + "loss": 1.5236, + "step": 30143 + }, + { + "epoch": 6.58, + "learning_rate": 1.712475420581167e-06, + "loss": 1.4878, + "step": 30144 + }, + { + "epoch": 6.58, + "learning_rate": 1.7123661787196854e-06, + "loss": 1.676, + "step": 30145 + }, + { + "epoch": 6.58, + "learning_rate": 1.7122569368582042e-06, + "loss": 1.8918, + "step": 30146 + }, + { + "epoch": 6.58, + "learning_rate": 1.7121476949967228e-06, + "loss": 1.4468, + "step": 30147 + }, + { + "epoch": 6.58, + "learning_rate": 1.7120384531352416e-06, + "loss": 1.5622, + "step": 30148 + }, + { + "epoch": 6.58, + "learning_rate": 1.7119292112737605e-06, + "loss": 0.2146, + "step": 30149 + }, + { + "epoch": 6.58, + "learning_rate": 1.7118199694122789e-06, + "loss": 1.3801, + "step": 30150 + }, + { + "epoch": 6.58, + "learning_rate": 1.7117107275507977e-06, + "loss": 1.6527, + "step": 30151 + }, + { + "epoch": 6.58, + "learning_rate": 1.7116014856893163e-06, + "loss": 1.7236, + "step": 30152 + }, + { + "epoch": 6.58, + "learning_rate": 1.7114922438278351e-06, + "loss": 1.3878, + "step": 30153 + }, + { + "epoch": 6.58, + "learning_rate": 1.7113830019663535e-06, + "loss": 0.6382, + "step": 30154 + }, + { + "epoch": 6.58, + "learning_rate": 1.7112737601048723e-06, + "loss": 0.7018, + "step": 30155 + }, + { + "epoch": 6.58, + "learning_rate": 1.7111645182433912e-06, + "loss": 0.9449, + "step": 30156 + }, + { + "epoch": 6.58, + "learning_rate": 1.7110552763819096e-06, + "loss": 0.6963, + "step": 30157 + }, + { + "epoch": 6.58, + "learning_rate": 1.7109460345204284e-06, + "loss": 1.2187, + "step": 30158 + }, + { + "epoch": 6.58, + "learning_rate": 1.710836792658947e-06, + "loss": 2.2937, + "step": 30159 + }, + { + "epoch": 6.58, + "learning_rate": 1.7107275507974658e-06, + "loss": 1.0582, + "step": 30160 + }, + { + "epoch": 6.58, + "learning_rate": 1.7106183089359842e-06, + "loss": 1.4716, + "step": 30161 + }, + { + "epoch": 6.58, + "learning_rate": 1.710509067074503e-06, + "loss": 1.3531, + "step": 30162 + }, + { + "epoch": 6.58, + "learning_rate": 1.7103998252130219e-06, + "loss": 1.5689, + "step": 30163 + }, + { + "epoch": 6.58, + "learning_rate": 1.7102905833515405e-06, + "loss": 2.0024, + "step": 30164 + }, + { + "epoch": 6.58, + "learning_rate": 1.7101813414900593e-06, + "loss": 0.8302, + "step": 30165 + }, + { + "epoch": 6.58, + "learning_rate": 1.7100720996285777e-06, + "loss": 1.5206, + "step": 30166 + }, + { + "epoch": 6.58, + "learning_rate": 1.7099628577670965e-06, + "loss": 1.294, + "step": 30167 + }, + { + "epoch": 6.58, + "learning_rate": 1.7098536159056151e-06, + "loss": 0.3327, + "step": 30168 + }, + { + "epoch": 6.58, + "learning_rate": 1.7097443740441337e-06, + "loss": 0.6122, + "step": 30169 + }, + { + "epoch": 6.58, + "learning_rate": 1.7096351321826526e-06, + "loss": 0.7657, + "step": 30170 + }, + { + "epoch": 6.58, + "learning_rate": 1.7095258903211712e-06, + "loss": 1.09, + "step": 30171 + }, + { + "epoch": 6.58, + "learning_rate": 1.70941664845969e-06, + "loss": 2.3821, + "step": 30172 + }, + { + "epoch": 6.59, + "learning_rate": 1.7093074065982084e-06, + "loss": 0.4432, + "step": 30173 + }, + { + "epoch": 6.59, + "learning_rate": 1.7091981647367272e-06, + "loss": 1.9325, + "step": 30174 + }, + { + "epoch": 6.59, + "learning_rate": 1.7090889228752458e-06, + "loss": 0.5614, + "step": 30175 + }, + { + "epoch": 6.59, + "learning_rate": 1.7089796810137647e-06, + "loss": 0.6417, + "step": 30176 + }, + { + "epoch": 6.59, + "learning_rate": 1.7088704391522835e-06, + "loss": 0.4116, + "step": 30177 + }, + { + "epoch": 6.59, + "learning_rate": 1.7087611972908019e-06, + "loss": 0.8213, + "step": 30178 + }, + { + "epoch": 6.59, + "learning_rate": 1.7086519554293207e-06, + "loss": 0.8888, + "step": 30179 + }, + { + "epoch": 6.59, + "learning_rate": 1.7085427135678393e-06, + "loss": 1.9109, + "step": 30180 + }, + { + "epoch": 6.59, + "learning_rate": 1.7084334717063581e-06, + "loss": 1.5046, + "step": 30181 + }, + { + "epoch": 6.59, + "learning_rate": 1.7083242298448765e-06, + "loss": 1.1779, + "step": 30182 + }, + { + "epoch": 6.59, + "learning_rate": 1.7082149879833954e-06, + "loss": 1.3427, + "step": 30183 + }, + { + "epoch": 6.59, + "learning_rate": 1.7081057461219142e-06, + "loss": 0.4688, + "step": 30184 + }, + { + "epoch": 6.59, + "learning_rate": 1.7079965042604326e-06, + "loss": 1.7108, + "step": 30185 + }, + { + "epoch": 6.59, + "learning_rate": 1.7078872623989514e-06, + "loss": 1.4455, + "step": 30186 + }, + { + "epoch": 6.59, + "learning_rate": 1.70777802053747e-06, + "loss": 1.8206, + "step": 30187 + }, + { + "epoch": 6.59, + "learning_rate": 1.7076687786759888e-06, + "loss": 2.1155, + "step": 30188 + }, + { + "epoch": 6.59, + "learning_rate": 1.7075595368145077e-06, + "loss": 0.2479, + "step": 30189 + }, + { + "epoch": 6.59, + "learning_rate": 1.707450294953026e-06, + "loss": 1.4359, + "step": 30190 + }, + { + "epoch": 6.59, + "learning_rate": 1.7073410530915449e-06, + "loss": 0.7602, + "step": 30191 + }, + { + "epoch": 6.59, + "learning_rate": 1.7072318112300635e-06, + "loss": 2.1767, + "step": 30192 + }, + { + "epoch": 6.59, + "learning_rate": 1.7071225693685823e-06, + "loss": 0.0739, + "step": 30193 + }, + { + "epoch": 6.59, + "learning_rate": 1.7070133275071007e-06, + "loss": 1.3416, + "step": 30194 + }, + { + "epoch": 6.59, + "learning_rate": 1.7069040856456195e-06, + "loss": 1.0634, + "step": 30195 + }, + { + "epoch": 6.59, + "learning_rate": 1.7067948437841384e-06, + "loss": 2.0419, + "step": 30196 + }, + { + "epoch": 6.59, + "learning_rate": 1.706685601922657e-06, + "loss": 1.1803, + "step": 30197 + }, + { + "epoch": 6.59, + "learning_rate": 1.7065763600611756e-06, + "loss": 0.9779, + "step": 30198 + }, + { + "epoch": 6.59, + "learning_rate": 1.7064671181996942e-06, + "loss": 0.7052, + "step": 30199 + }, + { + "epoch": 6.59, + "learning_rate": 1.706357876338213e-06, + "loss": 0.8987, + "step": 30200 + }, + { + "epoch": 6.59, + "learning_rate": 1.7062486344767314e-06, + "loss": 0.8997, + "step": 30201 + }, + { + "epoch": 6.59, + "learning_rate": 1.7061393926152502e-06, + "loss": 0.3632, + "step": 30202 + }, + { + "epoch": 6.59, + "learning_rate": 1.706030150753769e-06, + "loss": 1.1212, + "step": 30203 + }, + { + "epoch": 6.59, + "learning_rate": 1.7059209088922877e-06, + "loss": 1.0959, + "step": 30204 + }, + { + "epoch": 6.59, + "learning_rate": 1.7058116670308065e-06, + "loss": 1.135, + "step": 30205 + }, + { + "epoch": 6.59, + "learning_rate": 1.7057024251693249e-06, + "loss": 0.2588, + "step": 30206 + }, + { + "epoch": 6.59, + "learning_rate": 1.7055931833078437e-06, + "loss": 1.3289, + "step": 30207 + }, + { + "epoch": 6.59, + "learning_rate": 1.7054839414463623e-06, + "loss": 1.1383, + "step": 30208 + }, + { + "epoch": 6.59, + "learning_rate": 1.7053746995848811e-06, + "loss": 1.6695, + "step": 30209 + }, + { + "epoch": 6.59, + "learning_rate": 1.7052654577233998e-06, + "loss": 0.7667, + "step": 30210 + }, + { + "epoch": 6.59, + "learning_rate": 1.7051562158619184e-06, + "loss": 0.1367, + "step": 30211 + }, + { + "epoch": 6.59, + "learning_rate": 1.7050469740004372e-06, + "loss": 1.6304, + "step": 30212 + }, + { + "epoch": 6.59, + "learning_rate": 1.7049377321389556e-06, + "loss": 0.3359, + "step": 30213 + }, + { + "epoch": 6.59, + "learning_rate": 1.7048284902774744e-06, + "loss": 0.8179, + "step": 30214 + }, + { + "epoch": 6.59, + "learning_rate": 1.704719248415993e-06, + "loss": 0.8531, + "step": 30215 + }, + { + "epoch": 6.59, + "learning_rate": 1.7046100065545119e-06, + "loss": 0.4943, + "step": 30216 + }, + { + "epoch": 6.59, + "learning_rate": 1.7045007646930307e-06, + "loss": 0.5693, + "step": 30217 + }, + { + "epoch": 6.59, + "learning_rate": 1.704391522831549e-06, + "loss": 0.7926, + "step": 30218 + }, + { + "epoch": 6.6, + "learning_rate": 1.704282280970068e-06, + "loss": 1.9382, + "step": 30219 + }, + { + "epoch": 6.6, + "learning_rate": 1.7041730391085865e-06, + "loss": 0.2627, + "step": 30220 + }, + { + "epoch": 6.6, + "learning_rate": 1.7040637972471053e-06, + "loss": 1.1344, + "step": 30221 + }, + { + "epoch": 6.6, + "learning_rate": 1.703954555385624e-06, + "loss": 2.5987, + "step": 30222 + }, + { + "epoch": 6.6, + "learning_rate": 1.7038453135241426e-06, + "loss": 1.407, + "step": 30223 + }, + { + "epoch": 6.6, + "learning_rate": 1.7037360716626614e-06, + "loss": 1.1597, + "step": 30224 + }, + { + "epoch": 6.6, + "learning_rate": 1.70362682980118e-06, + "loss": 1.725, + "step": 30225 + }, + { + "epoch": 6.6, + "learning_rate": 1.7035175879396986e-06, + "loss": 2.1566, + "step": 30226 + }, + { + "epoch": 6.6, + "learning_rate": 1.7034083460782172e-06, + "loss": 0.3849, + "step": 30227 + }, + { + "epoch": 6.6, + "learning_rate": 1.703299104216736e-06, + "loss": 0.646, + "step": 30228 + }, + { + "epoch": 6.6, + "learning_rate": 1.7031898623552549e-06, + "loss": 0.8929, + "step": 30229 + }, + { + "epoch": 6.6, + "learning_rate": 1.7030806204937733e-06, + "loss": 1.2727, + "step": 30230 + }, + { + "epoch": 6.6, + "learning_rate": 1.702971378632292e-06, + "loss": 0.8161, + "step": 30231 + }, + { + "epoch": 6.6, + "learning_rate": 1.7028621367708107e-06, + "loss": 0.8342, + "step": 30232 + }, + { + "epoch": 6.6, + "learning_rate": 1.7027528949093295e-06, + "loss": 1.5196, + "step": 30233 + }, + { + "epoch": 6.6, + "learning_rate": 1.702643653047848e-06, + "loss": 0.6011, + "step": 30234 + }, + { + "epoch": 6.6, + "learning_rate": 1.7025344111863667e-06, + "loss": 0.8235, + "step": 30235 + }, + { + "epoch": 6.6, + "learning_rate": 1.7024251693248856e-06, + "loss": 0.9225, + "step": 30236 + }, + { + "epoch": 6.6, + "learning_rate": 1.7023159274634042e-06, + "loss": 1.4939, + "step": 30237 + }, + { + "epoch": 6.6, + "learning_rate": 1.7022066856019228e-06, + "loss": 0.8935, + "step": 30238 + }, + { + "epoch": 6.6, + "learning_rate": 1.7020974437404414e-06, + "loss": 1.2526, + "step": 30239 + }, + { + "epoch": 6.6, + "learning_rate": 1.7019882018789602e-06, + "loss": 1.042, + "step": 30240 + }, + { + "epoch": 6.6, + "learning_rate": 1.7018789600174786e-06, + "loss": 1.2848, + "step": 30241 + }, + { + "epoch": 6.6, + "learning_rate": 1.7017697181559974e-06, + "loss": 1.2614, + "step": 30242 + }, + { + "epoch": 6.6, + "learning_rate": 1.7016604762945163e-06, + "loss": 1.9319, + "step": 30243 + }, + { + "epoch": 6.6, + "learning_rate": 1.7015512344330349e-06, + "loss": 2.7607, + "step": 30244 + }, + { + "epoch": 6.6, + "learning_rate": 1.7014419925715537e-06, + "loss": 1.4589, + "step": 30245 + }, + { + "epoch": 6.6, + "learning_rate": 1.701332750710072e-06, + "loss": 1.0115, + "step": 30246 + }, + { + "epoch": 6.6, + "learning_rate": 1.701223508848591e-06, + "loss": 0.3625, + "step": 30247 + }, + { + "epoch": 6.6, + "learning_rate": 1.7011142669871095e-06, + "loss": 0.683, + "step": 30248 + }, + { + "epoch": 6.6, + "learning_rate": 1.7010050251256283e-06, + "loss": 1.9954, + "step": 30249 + }, + { + "epoch": 6.6, + "learning_rate": 1.7008957832641472e-06, + "loss": 1.1226, + "step": 30250 + }, + { + "epoch": 6.6, + "learning_rate": 1.7007865414026656e-06, + "loss": 2.3452, + "step": 30251 + }, + { + "epoch": 6.6, + "learning_rate": 1.7006772995411844e-06, + "loss": 0.3493, + "step": 30252 + }, + { + "epoch": 6.6, + "learning_rate": 1.700568057679703e-06, + "loss": 0.7466, + "step": 30253 + }, + { + "epoch": 6.6, + "learning_rate": 1.7004588158182216e-06, + "loss": 2.5791, + "step": 30254 + }, + { + "epoch": 6.6, + "learning_rate": 1.7003495739567404e-06, + "loss": 0.351, + "step": 30255 + }, + { + "epoch": 6.6, + "learning_rate": 1.700240332095259e-06, + "loss": 1.255, + "step": 30256 + }, + { + "epoch": 6.6, + "learning_rate": 1.7001310902337779e-06, + "loss": 0.9974, + "step": 30257 + }, + { + "epoch": 6.6, + "learning_rate": 1.7000218483722963e-06, + "loss": 2.8208, + "step": 30258 + }, + { + "epoch": 6.6, + "learning_rate": 1.699912606510815e-06, + "loss": 1.8331, + "step": 30259 + }, + { + "epoch": 6.6, + "learning_rate": 1.6998033646493337e-06, + "loss": 1.1339, + "step": 30260 + }, + { + "epoch": 6.6, + "learning_rate": 1.6996941227878525e-06, + "loss": 1.0644, + "step": 30261 + }, + { + "epoch": 6.6, + "learning_rate": 1.6995848809263713e-06, + "loss": 2.8128, + "step": 30262 + }, + { + "epoch": 6.6, + "learning_rate": 1.6994756390648897e-06, + "loss": 0.2481, + "step": 30263 + }, + { + "epoch": 6.6, + "learning_rate": 1.6993663972034086e-06, + "loss": 1.5221, + "step": 30264 + }, + { + "epoch": 6.61, + "learning_rate": 1.6992571553419272e-06, + "loss": 1.6214, + "step": 30265 + }, + { + "epoch": 6.61, + "learning_rate": 1.6991479134804458e-06, + "loss": 1.2237, + "step": 30266 + }, + { + "epoch": 6.61, + "learning_rate": 1.6990386716189644e-06, + "loss": 0.6526, + "step": 30267 + }, + { + "epoch": 6.61, + "learning_rate": 1.6989294297574832e-06, + "loss": 1.36, + "step": 30268 + }, + { + "epoch": 6.61, + "learning_rate": 1.698820187896002e-06, + "loss": 1.4185, + "step": 30269 + }, + { + "epoch": 6.61, + "learning_rate": 1.6987109460345204e-06, + "loss": 1.2984, + "step": 30270 + }, + { + "epoch": 6.61, + "learning_rate": 1.6986017041730393e-06, + "loss": 0.2277, + "step": 30271 + }, + { + "epoch": 6.61, + "learning_rate": 1.6984924623115579e-06, + "loss": 2.001, + "step": 30272 + }, + { + "epoch": 6.61, + "learning_rate": 1.6983832204500767e-06, + "loss": 0.1778, + "step": 30273 + }, + { + "epoch": 6.61, + "learning_rate": 1.698273978588595e-06, + "loss": 0.4903, + "step": 30274 + }, + { + "epoch": 6.61, + "learning_rate": 1.698164736727114e-06, + "loss": 1.0557, + "step": 30275 + }, + { + "epoch": 6.61, + "learning_rate": 1.6980554948656327e-06, + "loss": 0.2108, + "step": 30276 + }, + { + "epoch": 6.61, + "learning_rate": 1.6979462530041514e-06, + "loss": 1.6832, + "step": 30277 + }, + { + "epoch": 6.61, + "learning_rate": 1.6978370111426702e-06, + "loss": 3.177, + "step": 30278 + }, + { + "epoch": 6.61, + "learning_rate": 1.6977277692811886e-06, + "loss": 2.6542, + "step": 30279 + }, + { + "epoch": 6.61, + "learning_rate": 1.6976185274197074e-06, + "loss": 2.8591, + "step": 30280 + }, + { + "epoch": 6.61, + "learning_rate": 1.697509285558226e-06, + "loss": 0.5842, + "step": 30281 + }, + { + "epoch": 6.61, + "learning_rate": 1.6974000436967446e-06, + "loss": 2.4898, + "step": 30282 + }, + { + "epoch": 6.61, + "learning_rate": 1.6972908018352634e-06, + "loss": 1.1344, + "step": 30283 + }, + { + "epoch": 6.61, + "learning_rate": 1.697181559973782e-06, + "loss": 0.8569, + "step": 30284 + }, + { + "epoch": 6.61, + "learning_rate": 1.6970723181123009e-06, + "loss": 1.9139, + "step": 30285 + }, + { + "epoch": 6.61, + "learning_rate": 1.6969630762508193e-06, + "loss": 2.0251, + "step": 30286 + }, + { + "epoch": 6.61, + "learning_rate": 1.696853834389338e-06, + "loss": 2.3917, + "step": 30287 + }, + { + "epoch": 6.61, + "learning_rate": 1.696744592527857e-06, + "loss": 0.1685, + "step": 30288 + }, + { + "epoch": 6.61, + "learning_rate": 1.6966353506663755e-06, + "loss": 1.7495, + "step": 30289 + }, + { + "epoch": 6.61, + "learning_rate": 1.6965261088048944e-06, + "loss": 1.4118, + "step": 30290 + }, + { + "epoch": 6.61, + "learning_rate": 1.6964168669434128e-06, + "loss": 2.0201, + "step": 30291 + }, + { + "epoch": 6.61, + "learning_rate": 1.6963076250819316e-06, + "loss": 0.6758, + "step": 30292 + }, + { + "epoch": 6.61, + "learning_rate": 1.6961983832204502e-06, + "loss": 2.6447, + "step": 30293 + }, + { + "epoch": 6.61, + "learning_rate": 1.6960891413589688e-06, + "loss": 1.1048, + "step": 30294 + }, + { + "epoch": 6.61, + "learning_rate": 1.6959798994974876e-06, + "loss": 1.4641, + "step": 30295 + }, + { + "epoch": 6.61, + "learning_rate": 1.6958706576360062e-06, + "loss": 1.4279, + "step": 30296 + }, + { + "epoch": 6.61, + "learning_rate": 1.695761415774525e-06, + "loss": 2.2678, + "step": 30297 + }, + { + "epoch": 6.61, + "learning_rate": 1.6956521739130435e-06, + "loss": 1.066, + "step": 30298 + }, + { + "epoch": 6.61, + "learning_rate": 1.6955429320515623e-06, + "loss": 0.9921, + "step": 30299 + }, + { + "epoch": 6.61, + "learning_rate": 1.6954336901900809e-06, + "loss": 1.8891, + "step": 30300 + }, + { + "epoch": 6.61, + "learning_rate": 1.6953244483285997e-06, + "loss": 2.7223, + "step": 30301 + }, + { + "epoch": 6.61, + "learning_rate": 1.6952152064671185e-06, + "loss": 1.4137, + "step": 30302 + }, + { + "epoch": 6.61, + "learning_rate": 1.695105964605637e-06, + "loss": 2.1123, + "step": 30303 + }, + { + "epoch": 6.61, + "learning_rate": 1.6949967227441558e-06, + "loss": 0.7461, + "step": 30304 + }, + { + "epoch": 6.61, + "learning_rate": 1.6948874808826744e-06, + "loss": 2.0207, + "step": 30305 + }, + { + "epoch": 6.61, + "learning_rate": 1.6947782390211932e-06, + "loss": 2.591, + "step": 30306 + }, + { + "epoch": 6.61, + "learning_rate": 1.6946689971597116e-06, + "loss": 0.1516, + "step": 30307 + }, + { + "epoch": 6.61, + "learning_rate": 1.6945597552982304e-06, + "loss": 0.3964, + "step": 30308 + }, + { + "epoch": 6.61, + "learning_rate": 1.6944505134367492e-06, + "loss": 0.7591, + "step": 30309 + }, + { + "epoch": 6.62, + "learning_rate": 1.6943412715752676e-06, + "loss": 1.8553, + "step": 30310 + }, + { + "epoch": 6.62, + "learning_rate": 1.6942320297137865e-06, + "loss": 0.6764, + "step": 30311 + }, + { + "epoch": 6.62, + "learning_rate": 1.694122787852305e-06, + "loss": 0.732, + "step": 30312 + }, + { + "epoch": 6.62, + "learning_rate": 1.6940135459908239e-06, + "loss": 0.1787, + "step": 30313 + }, + { + "epoch": 6.62, + "learning_rate": 1.6939043041293423e-06, + "loss": 0.866, + "step": 30314 + }, + { + "epoch": 6.62, + "learning_rate": 1.6937950622678611e-06, + "loss": 0.5882, + "step": 30315 + }, + { + "epoch": 6.62, + "learning_rate": 1.69368582040638e-06, + "loss": 0.9957, + "step": 30316 + }, + { + "epoch": 6.62, + "learning_rate": 1.6935765785448985e-06, + "loss": 1.5495, + "step": 30317 + }, + { + "epoch": 6.62, + "learning_rate": 1.6934673366834174e-06, + "loss": 2.2114, + "step": 30318 + }, + { + "epoch": 6.62, + "learning_rate": 1.6933580948219358e-06, + "loss": 1.7749, + "step": 30319 + }, + { + "epoch": 6.62, + "learning_rate": 1.6932488529604546e-06, + "loss": 0.2918, + "step": 30320 + }, + { + "epoch": 6.62, + "learning_rate": 1.6931396110989734e-06, + "loss": 2.5238, + "step": 30321 + }, + { + "epoch": 6.62, + "learning_rate": 1.693030369237492e-06, + "loss": 1.594, + "step": 30322 + }, + { + "epoch": 6.62, + "learning_rate": 1.6929211273760106e-06, + "loss": 2.0888, + "step": 30323 + }, + { + "epoch": 6.62, + "learning_rate": 1.6928118855145292e-06, + "loss": 1.3515, + "step": 30324 + }, + { + "epoch": 6.62, + "learning_rate": 1.692702643653048e-06, + "loss": 1.7933, + "step": 30325 + }, + { + "epoch": 6.62, + "learning_rate": 1.6925934017915665e-06, + "loss": 0.7629, + "step": 30326 + }, + { + "epoch": 6.62, + "learning_rate": 1.6924841599300853e-06, + "loss": 0.8595, + "step": 30327 + }, + { + "epoch": 6.62, + "learning_rate": 1.6923749180686041e-06, + "loss": 1.0246, + "step": 30328 + }, + { + "epoch": 6.62, + "learning_rate": 1.6922656762071227e-06, + "loss": 2.5041, + "step": 30329 + }, + { + "epoch": 6.62, + "learning_rate": 1.6921564343456415e-06, + "loss": 1.4036, + "step": 30330 + }, + { + "epoch": 6.62, + "learning_rate": 1.69204719248416e-06, + "loss": 1.7778, + "step": 30331 + }, + { + "epoch": 6.62, + "learning_rate": 1.6919379506226788e-06, + "loss": 2.7576, + "step": 30332 + }, + { + "epoch": 6.62, + "learning_rate": 1.6918287087611974e-06, + "loss": 0.6517, + "step": 30333 + }, + { + "epoch": 6.62, + "learning_rate": 1.6917194668997162e-06, + "loss": 1.881, + "step": 30334 + }, + { + "epoch": 6.62, + "learning_rate": 1.6916102250382348e-06, + "loss": 2.1158, + "step": 30335 + }, + { + "epoch": 6.62, + "learning_rate": 1.6915009831767534e-06, + "loss": 0.6441, + "step": 30336 + }, + { + "epoch": 6.62, + "learning_rate": 1.6913917413152722e-06, + "loss": 0.3026, + "step": 30337 + }, + { + "epoch": 6.62, + "learning_rate": 1.6912824994537906e-06, + "loss": 0.425, + "step": 30338 + }, + { + "epoch": 6.62, + "learning_rate": 1.6911732575923095e-06, + "loss": 1.0368, + "step": 30339 + }, + { + "epoch": 6.62, + "learning_rate": 1.691064015730828e-06, + "loss": 1.3885, + "step": 30340 + }, + { + "epoch": 6.62, + "learning_rate": 1.690954773869347e-06, + "loss": 1.283, + "step": 30341 + }, + { + "epoch": 6.62, + "learning_rate": 1.6908455320078657e-06, + "loss": 2.647, + "step": 30342 + }, + { + "epoch": 6.62, + "learning_rate": 1.6907362901463841e-06, + "loss": 0.9217, + "step": 30343 + }, + { + "epoch": 6.62, + "learning_rate": 1.690627048284903e-06, + "loss": 1.7467, + "step": 30344 + }, + { + "epoch": 6.62, + "learning_rate": 1.6905178064234216e-06, + "loss": 1.4701, + "step": 30345 + }, + { + "epoch": 6.62, + "learning_rate": 1.6904085645619404e-06, + "loss": 0.3072, + "step": 30346 + }, + { + "epoch": 6.62, + "learning_rate": 1.6902993227004588e-06, + "loss": 1.1798, + "step": 30347 + }, + { + "epoch": 6.62, + "learning_rate": 1.6901900808389776e-06, + "loss": 1.7029, + "step": 30348 + }, + { + "epoch": 6.62, + "learning_rate": 1.6900808389774964e-06, + "loss": 2.3302, + "step": 30349 + }, + { + "epoch": 6.62, + "learning_rate": 1.689971597116015e-06, + "loss": 1.0726, + "step": 30350 + }, + { + "epoch": 6.62, + "learning_rate": 1.6898623552545337e-06, + "loss": 0.4719, + "step": 30351 + }, + { + "epoch": 6.62, + "learning_rate": 1.6897531133930523e-06, + "loss": 3.0964, + "step": 30352 + }, + { + "epoch": 6.62, + "learning_rate": 1.689643871531571e-06, + "loss": 1.2726, + "step": 30353 + }, + { + "epoch": 6.62, + "learning_rate": 1.68953462967009e-06, + "loss": 2.0165, + "step": 30354 + }, + { + "epoch": 6.62, + "learning_rate": 1.6894253878086083e-06, + "loss": 1.5681, + "step": 30355 + }, + { + "epoch": 6.63, + "learning_rate": 1.6893161459471271e-06, + "loss": 3.0214, + "step": 30356 + }, + { + "epoch": 6.63, + "learning_rate": 1.6892069040856457e-06, + "loss": 1.8235, + "step": 30357 + }, + { + "epoch": 6.63, + "learning_rate": 1.6890976622241646e-06, + "loss": 1.47, + "step": 30358 + }, + { + "epoch": 6.63, + "learning_rate": 1.688988420362683e-06, + "loss": 1.9261, + "step": 30359 + }, + { + "epoch": 6.63, + "learning_rate": 1.6888791785012018e-06, + "loss": 0.7902, + "step": 30360 + }, + { + "epoch": 6.63, + "learning_rate": 1.6887699366397206e-06, + "loss": 1.1254, + "step": 30361 + }, + { + "epoch": 6.63, + "learning_rate": 1.6886606947782392e-06, + "loss": 1.2027, + "step": 30362 + }, + { + "epoch": 6.63, + "learning_rate": 1.6885514529167578e-06, + "loss": 0.4666, + "step": 30363 + }, + { + "epoch": 6.63, + "learning_rate": 1.6884422110552764e-06, + "loss": 0.9386, + "step": 30364 + }, + { + "epoch": 6.63, + "learning_rate": 1.6883329691937953e-06, + "loss": 1.5288, + "step": 30365 + }, + { + "epoch": 6.63, + "learning_rate": 1.6882237273323137e-06, + "loss": 1.7989, + "step": 30366 + }, + { + "epoch": 6.63, + "learning_rate": 1.6881144854708325e-06, + "loss": 1.5537, + "step": 30367 + }, + { + "epoch": 6.63, + "learning_rate": 1.6880052436093513e-06, + "loss": 1.3094, + "step": 30368 + }, + { + "epoch": 6.63, + "learning_rate": 1.68789600174787e-06, + "loss": 1.18, + "step": 30369 + }, + { + "epoch": 6.63, + "learning_rate": 1.6877867598863887e-06, + "loss": 0.7959, + "step": 30370 + }, + { + "epoch": 6.63, + "learning_rate": 1.6876775180249071e-06, + "loss": 1.2333, + "step": 30371 + }, + { + "epoch": 6.63, + "learning_rate": 1.687568276163426e-06, + "loss": 0.7187, + "step": 30372 + }, + { + "epoch": 6.63, + "learning_rate": 1.6874590343019446e-06, + "loss": 1.3314, + "step": 30373 + }, + { + "epoch": 6.63, + "learning_rate": 1.6873497924404634e-06, + "loss": 3.2734, + "step": 30374 + }, + { + "epoch": 6.63, + "learning_rate": 1.6872405505789822e-06, + "loss": 2.5479, + "step": 30375 + }, + { + "epoch": 6.63, + "learning_rate": 1.6871313087175006e-06, + "loss": 0.5922, + "step": 30376 + }, + { + "epoch": 6.63, + "learning_rate": 1.6870220668560194e-06, + "loss": 2.0796, + "step": 30377 + }, + { + "epoch": 6.63, + "learning_rate": 1.686912824994538e-06, + "loss": 2.8841, + "step": 30378 + }, + { + "epoch": 6.63, + "learning_rate": 1.6868035831330567e-06, + "loss": 2.8803, + "step": 30379 + }, + { + "epoch": 6.63, + "learning_rate": 1.6866943412715753e-06, + "loss": 1.3715, + "step": 30380 + }, + { + "epoch": 6.63, + "learning_rate": 1.686585099410094e-06, + "loss": 0.4389, + "step": 30381 + }, + { + "epoch": 6.63, + "learning_rate": 1.686475857548613e-06, + "loss": 1.6329, + "step": 30382 + }, + { + "epoch": 6.63, + "learning_rate": 1.6863666156871313e-06, + "loss": 0.7936, + "step": 30383 + }, + { + "epoch": 6.63, + "learning_rate": 1.6862573738256501e-06, + "loss": 2.0431, + "step": 30384 + }, + { + "epoch": 6.63, + "learning_rate": 1.6861481319641688e-06, + "loss": 0.526, + "step": 30385 + }, + { + "epoch": 6.63, + "learning_rate": 1.6860388901026876e-06, + "loss": 0.6882, + "step": 30386 + }, + { + "epoch": 6.63, + "learning_rate": 1.685929648241206e-06, + "loss": 1.0902, + "step": 30387 + }, + { + "epoch": 6.63, + "learning_rate": 1.6858204063797248e-06, + "loss": 3.519, + "step": 30388 + }, + { + "epoch": 6.63, + "learning_rate": 1.6857111645182436e-06, + "loss": 0.4042, + "step": 30389 + }, + { + "epoch": 6.63, + "learning_rate": 1.6856019226567622e-06, + "loss": 0.8508, + "step": 30390 + }, + { + "epoch": 6.63, + "learning_rate": 1.6854926807952808e-06, + "loss": 0.6442, + "step": 30391 + }, + { + "epoch": 6.63, + "learning_rate": 1.6853834389337995e-06, + "loss": 2.327, + "step": 30392 + }, + { + "epoch": 6.63, + "learning_rate": 1.6852741970723183e-06, + "loss": 1.993, + "step": 30393 + }, + { + "epoch": 6.63, + "learning_rate": 1.685164955210837e-06, + "loss": 1.8591, + "step": 30394 + }, + { + "epoch": 6.63, + "learning_rate": 1.6850557133493555e-06, + "loss": 1.4956, + "step": 30395 + }, + { + "epoch": 6.63, + "learning_rate": 1.6849464714878743e-06, + "loss": 0.7837, + "step": 30396 + }, + { + "epoch": 6.63, + "learning_rate": 1.684837229626393e-06, + "loss": 1.5887, + "step": 30397 + }, + { + "epoch": 6.63, + "learning_rate": 1.6847279877649118e-06, + "loss": 3.1676, + "step": 30398 + }, + { + "epoch": 6.63, + "learning_rate": 1.6846187459034302e-06, + "loss": 2.3876, + "step": 30399 + }, + { + "epoch": 6.63, + "learning_rate": 1.684509504041949e-06, + "loss": 1.2245, + "step": 30400 + }, + { + "epoch": 6.63, + "learning_rate": 1.6844002621804678e-06, + "loss": 1.4313, + "step": 30401 + }, + { + "epoch": 6.64, + "learning_rate": 1.6842910203189864e-06, + "loss": 1.5751, + "step": 30402 + }, + { + "epoch": 6.64, + "learning_rate": 1.6841817784575052e-06, + "loss": 0.9766, + "step": 30403 + }, + { + "epoch": 6.64, + "learning_rate": 1.6840725365960236e-06, + "loss": 1.0825, + "step": 30404 + }, + { + "epoch": 6.64, + "learning_rate": 1.6839632947345425e-06, + "loss": 1.1936, + "step": 30405 + }, + { + "epoch": 6.64, + "learning_rate": 1.683854052873061e-06, + "loss": 1.5529, + "step": 30406 + }, + { + "epoch": 6.64, + "learning_rate": 1.6837448110115797e-06, + "loss": 1.3977, + "step": 30407 + }, + { + "epoch": 6.64, + "learning_rate": 1.6836355691500985e-06, + "loss": 1.1693, + "step": 30408 + }, + { + "epoch": 6.64, + "learning_rate": 1.6835263272886171e-06, + "loss": 0.4665, + "step": 30409 + }, + { + "epoch": 6.64, + "learning_rate": 1.683417085427136e-06, + "loss": 0.549, + "step": 30410 + }, + { + "epoch": 6.64, + "learning_rate": 1.6833078435656543e-06, + "loss": 1.4904, + "step": 30411 + }, + { + "epoch": 6.64, + "learning_rate": 1.6831986017041732e-06, + "loss": 2.9547, + "step": 30412 + }, + { + "epoch": 6.64, + "learning_rate": 1.6830893598426918e-06, + "loss": 1.1916, + "step": 30413 + }, + { + "epoch": 6.64, + "learning_rate": 1.6829801179812106e-06, + "loss": 0.1412, + "step": 30414 + }, + { + "epoch": 6.64, + "learning_rate": 1.6828708761197294e-06, + "loss": 0.5538, + "step": 30415 + }, + { + "epoch": 6.64, + "learning_rate": 1.6827616342582478e-06, + "loss": 0.8279, + "step": 30416 + }, + { + "epoch": 6.64, + "learning_rate": 1.6826523923967666e-06, + "loss": 0.8201, + "step": 30417 + }, + { + "epoch": 6.64, + "learning_rate": 1.6825431505352852e-06, + "loss": 0.9342, + "step": 30418 + }, + { + "epoch": 6.64, + "learning_rate": 1.6824339086738039e-06, + "loss": 0.6401, + "step": 30419 + }, + { + "epoch": 6.64, + "learning_rate": 1.6823246668123225e-06, + "loss": 1.0866, + "step": 30420 + }, + { + "epoch": 6.64, + "learning_rate": 1.6822154249508413e-06, + "loss": 1.2852, + "step": 30421 + }, + { + "epoch": 6.64, + "learning_rate": 1.6821061830893601e-06, + "loss": 1.1839, + "step": 30422 + }, + { + "epoch": 6.64, + "learning_rate": 1.6819969412278785e-06, + "loss": 0.6154, + "step": 30423 + }, + { + "epoch": 6.64, + "learning_rate": 1.6818876993663973e-06, + "loss": 0.7343, + "step": 30424 + }, + { + "epoch": 6.64, + "learning_rate": 1.681778457504916e-06, + "loss": 0.4748, + "step": 30425 + }, + { + "epoch": 6.64, + "learning_rate": 1.6816692156434348e-06, + "loss": 1.6689, + "step": 30426 + }, + { + "epoch": 6.64, + "learning_rate": 1.6815599737819536e-06, + "loss": 1.6195, + "step": 30427 + }, + { + "epoch": 6.64, + "learning_rate": 1.681450731920472e-06, + "loss": 0.3297, + "step": 30428 + }, + { + "epoch": 6.64, + "learning_rate": 1.6813414900589908e-06, + "loss": 1.1555, + "step": 30429 + }, + { + "epoch": 6.64, + "learning_rate": 1.6812322481975094e-06, + "loss": 1.7297, + "step": 30430 + }, + { + "epoch": 6.64, + "learning_rate": 1.6811230063360282e-06, + "loss": 0.7661, + "step": 30431 + }, + { + "epoch": 6.64, + "learning_rate": 1.6810137644745466e-06, + "loss": 1.4666, + "step": 30432 + }, + { + "epoch": 6.64, + "learning_rate": 1.6809045226130655e-06, + "loss": 0.5526, + "step": 30433 + }, + { + "epoch": 6.64, + "learning_rate": 1.6807952807515843e-06, + "loss": 1.2757, + "step": 30434 + }, + { + "epoch": 6.64, + "learning_rate": 1.6806860388901027e-06, + "loss": 1.0783, + "step": 30435 + }, + { + "epoch": 6.64, + "learning_rate": 1.6805767970286215e-06, + "loss": 1.4798, + "step": 30436 + }, + { + "epoch": 6.64, + "learning_rate": 1.6804675551671401e-06, + "loss": 0.5635, + "step": 30437 + }, + { + "epoch": 6.64, + "learning_rate": 1.680358313305659e-06, + "loss": 1.1867, + "step": 30438 + }, + { + "epoch": 6.64, + "learning_rate": 1.6802490714441773e-06, + "loss": 1.5071, + "step": 30439 + }, + { + "epoch": 6.64, + "learning_rate": 1.6801398295826962e-06, + "loss": 2.1395, + "step": 30440 + }, + { + "epoch": 6.64, + "learning_rate": 1.680030587721215e-06, + "loss": 2.0585, + "step": 30441 + }, + { + "epoch": 6.64, + "learning_rate": 1.6799213458597336e-06, + "loss": 0.1528, + "step": 30442 + }, + { + "epoch": 6.64, + "learning_rate": 1.6798121039982524e-06, + "loss": 1.0059, + "step": 30443 + }, + { + "epoch": 6.64, + "learning_rate": 1.6797028621367708e-06, + "loss": 0.7386, + "step": 30444 + }, + { + "epoch": 6.64, + "learning_rate": 1.6795936202752896e-06, + "loss": 0.6736, + "step": 30445 + }, + { + "epoch": 6.64, + "learning_rate": 1.6794843784138083e-06, + "loss": 3.1506, + "step": 30446 + }, + { + "epoch": 6.64, + "learning_rate": 1.679375136552327e-06, + "loss": 1.7921, + "step": 30447 + }, + { + "epoch": 6.65, + "learning_rate": 1.6792658946908457e-06, + "loss": 1.4719, + "step": 30448 + }, + { + "epoch": 6.65, + "learning_rate": 1.6791566528293643e-06, + "loss": 1.2569, + "step": 30449 + }, + { + "epoch": 6.65, + "learning_rate": 1.6790474109678831e-06, + "loss": 0.3408, + "step": 30450 + }, + { + "epoch": 6.65, + "learning_rate": 1.6789381691064015e-06, + "loss": 3.6844, + "step": 30451 + }, + { + "epoch": 6.65, + "learning_rate": 1.6788289272449203e-06, + "loss": 0.1722, + "step": 30452 + }, + { + "epoch": 6.65, + "learning_rate": 1.678719685383439e-06, + "loss": 0.8546, + "step": 30453 + }, + { + "epoch": 6.65, + "learning_rate": 1.6786104435219578e-06, + "loss": 1.8914, + "step": 30454 + }, + { + "epoch": 6.65, + "learning_rate": 1.6785012016604766e-06, + "loss": 2.1518, + "step": 30455 + }, + { + "epoch": 6.65, + "learning_rate": 1.678391959798995e-06, + "loss": 0.5373, + "step": 30456 + }, + { + "epoch": 6.65, + "learning_rate": 1.6782827179375138e-06, + "loss": 0.2909, + "step": 30457 + }, + { + "epoch": 6.65, + "learning_rate": 1.6781734760760324e-06, + "loss": 1.7826, + "step": 30458 + }, + { + "epoch": 6.65, + "learning_rate": 1.6780642342145513e-06, + "loss": 2.8184, + "step": 30459 + }, + { + "epoch": 6.65, + "learning_rate": 1.6779549923530699e-06, + "loss": 3.9902, + "step": 30460 + }, + { + "epoch": 6.65, + "learning_rate": 1.6778457504915885e-06, + "loss": 0.488, + "step": 30461 + }, + { + "epoch": 6.65, + "learning_rate": 1.6777365086301073e-06, + "loss": 1.1986, + "step": 30462 + }, + { + "epoch": 6.65, + "learning_rate": 1.6776272667686257e-06, + "loss": 0.7425, + "step": 30463 + }, + { + "epoch": 6.65, + "learning_rate": 1.6775180249071445e-06, + "loss": 0.8665, + "step": 30464 + }, + { + "epoch": 6.65, + "learning_rate": 1.6774087830456631e-06, + "loss": 2.4303, + "step": 30465 + }, + { + "epoch": 6.65, + "learning_rate": 1.677299541184182e-06, + "loss": 1.1167, + "step": 30466 + }, + { + "epoch": 6.65, + "learning_rate": 1.6771902993227008e-06, + "loss": 0.3176, + "step": 30467 + }, + { + "epoch": 6.65, + "learning_rate": 1.6770810574612192e-06, + "loss": 2.0047, + "step": 30468 + }, + { + "epoch": 6.65, + "learning_rate": 1.676971815599738e-06, + "loss": 1.133, + "step": 30469 + }, + { + "epoch": 6.65, + "learning_rate": 1.6768625737382566e-06, + "loss": 3.2692, + "step": 30470 + }, + { + "epoch": 6.65, + "learning_rate": 1.6767533318767754e-06, + "loss": 2.5726, + "step": 30471 + }, + { + "epoch": 6.65, + "learning_rate": 1.6766440900152938e-06, + "loss": 1.8045, + "step": 30472 + }, + { + "epoch": 6.65, + "learning_rate": 1.6765348481538127e-06, + "loss": 0.6241, + "step": 30473 + }, + { + "epoch": 6.65, + "learning_rate": 1.6764256062923315e-06, + "loss": 1.2711, + "step": 30474 + }, + { + "epoch": 6.65, + "learning_rate": 1.67631636443085e-06, + "loss": 1.8341, + "step": 30475 + }, + { + "epoch": 6.65, + "learning_rate": 1.6762071225693687e-06, + "loss": 2.7778, + "step": 30476 + }, + { + "epoch": 6.65, + "learning_rate": 1.6760978807078873e-06, + "loss": 1.0566, + "step": 30477 + }, + { + "epoch": 6.65, + "learning_rate": 1.6759886388464061e-06, + "loss": 3.0196, + "step": 30478 + }, + { + "epoch": 6.65, + "learning_rate": 1.6758793969849245e-06, + "loss": 0.3428, + "step": 30479 + }, + { + "epoch": 6.65, + "learning_rate": 1.6757701551234434e-06, + "loss": 1.057, + "step": 30480 + }, + { + "epoch": 6.65, + "learning_rate": 1.6756609132619622e-06, + "loss": 0.6308, + "step": 30481 + }, + { + "epoch": 6.65, + "learning_rate": 1.6755516714004808e-06, + "loss": 1.5186, + "step": 30482 + }, + { + "epoch": 6.65, + "learning_rate": 1.6754424295389996e-06, + "loss": 1.0094, + "step": 30483 + }, + { + "epoch": 6.65, + "learning_rate": 1.675333187677518e-06, + "loss": 0.7419, + "step": 30484 + }, + { + "epoch": 6.65, + "learning_rate": 1.6752239458160368e-06, + "loss": 0.7884, + "step": 30485 + }, + { + "epoch": 6.65, + "learning_rate": 1.6751147039545555e-06, + "loss": 1.7169, + "step": 30486 + }, + { + "epoch": 6.65, + "learning_rate": 1.6750054620930743e-06, + "loss": 0.3544, + "step": 30487 + }, + { + "epoch": 6.65, + "learning_rate": 1.6748962202315929e-06, + "loss": 0.3622, + "step": 30488 + }, + { + "epoch": 6.65, + "learning_rate": 1.6747869783701115e-06, + "loss": 1.5272, + "step": 30489 + }, + { + "epoch": 6.65, + "learning_rate": 1.6746777365086303e-06, + "loss": 0.6784, + "step": 30490 + }, + { + "epoch": 6.65, + "learning_rate": 1.6745684946471487e-06, + "loss": 0.1246, + "step": 30491 + }, + { + "epoch": 6.65, + "learning_rate": 1.6744592527856675e-06, + "loss": 1.1366, + "step": 30492 + }, + { + "epoch": 6.65, + "learning_rate": 1.6743500109241864e-06, + "loss": 1.3709, + "step": 30493 + }, + { + "epoch": 6.66, + "learning_rate": 1.674240769062705e-06, + "loss": 1.9067, + "step": 30494 + }, + { + "epoch": 6.66, + "learning_rate": 1.6741315272012238e-06, + "loss": 0.5205, + "step": 30495 + }, + { + "epoch": 6.66, + "learning_rate": 1.6740222853397422e-06, + "loss": 1.5932, + "step": 30496 + }, + { + "epoch": 6.66, + "learning_rate": 1.673913043478261e-06, + "loss": 0.8553, + "step": 30497 + }, + { + "epoch": 6.66, + "learning_rate": 1.6738038016167796e-06, + "loss": 0.4655, + "step": 30498 + }, + { + "epoch": 6.66, + "learning_rate": 1.6736945597552985e-06, + "loss": 0.9439, + "step": 30499 + }, + { + "epoch": 6.66, + "learning_rate": 1.6735853178938173e-06, + "loss": 1.4498, + "step": 30500 + }, + { + "epoch": 6.66, + "learning_rate": 1.6734760760323357e-06, + "loss": 0.618, + "step": 30501 + }, + { + "epoch": 6.66, + "learning_rate": 1.6733668341708545e-06, + "loss": 1.2177, + "step": 30502 + }, + { + "epoch": 6.66, + "learning_rate": 1.6732575923093731e-06, + "loss": 1.0802, + "step": 30503 + }, + { + "epoch": 6.66, + "learning_rate": 1.6731483504478917e-06, + "loss": 1.5746, + "step": 30504 + }, + { + "epoch": 6.66, + "learning_rate": 1.6730391085864103e-06, + "loss": 0.344, + "step": 30505 + }, + { + "epoch": 6.66, + "learning_rate": 1.6729298667249292e-06, + "loss": 3.1733, + "step": 30506 + }, + { + "epoch": 6.66, + "learning_rate": 1.672820624863448e-06, + "loss": 2.2741, + "step": 30507 + }, + { + "epoch": 6.66, + "learning_rate": 1.6727113830019664e-06, + "loss": 1.1356, + "step": 30508 + }, + { + "epoch": 6.66, + "learning_rate": 1.6726021411404852e-06, + "loss": 1.1787, + "step": 30509 + }, + { + "epoch": 6.66, + "learning_rate": 1.6724928992790038e-06, + "loss": 0.276, + "step": 30510 + }, + { + "epoch": 6.66, + "learning_rate": 1.6723836574175226e-06, + "loss": 1.6495, + "step": 30511 + }, + { + "epoch": 6.66, + "learning_rate": 1.672274415556041e-06, + "loss": 0.6664, + "step": 30512 + }, + { + "epoch": 6.66, + "learning_rate": 1.6721651736945599e-06, + "loss": 0.1984, + "step": 30513 + }, + { + "epoch": 6.66, + "learning_rate": 1.6720559318330787e-06, + "loss": 0.8184, + "step": 30514 + }, + { + "epoch": 6.66, + "learning_rate": 1.6719466899715973e-06, + "loss": 0.3807, + "step": 30515 + }, + { + "epoch": 6.66, + "learning_rate": 1.671837448110116e-06, + "loss": 1.8222, + "step": 30516 + }, + { + "epoch": 6.66, + "learning_rate": 1.6717282062486345e-06, + "loss": 1.5406, + "step": 30517 + }, + { + "epoch": 6.66, + "learning_rate": 1.6716189643871533e-06, + "loss": 2.2345, + "step": 30518 + }, + { + "epoch": 6.66, + "learning_rate": 1.671509722525672e-06, + "loss": 0.559, + "step": 30519 + }, + { + "epoch": 6.66, + "learning_rate": 1.6714004806641906e-06, + "loss": 1.3459, + "step": 30520 + }, + { + "epoch": 6.66, + "learning_rate": 1.6712912388027094e-06, + "loss": 1.7548, + "step": 30521 + }, + { + "epoch": 6.66, + "learning_rate": 1.671181996941228e-06, + "loss": 0.7268, + "step": 30522 + }, + { + "epoch": 6.66, + "learning_rate": 1.6710727550797468e-06, + "loss": 1.5433, + "step": 30523 + }, + { + "epoch": 6.66, + "learning_rate": 1.6709635132182652e-06, + "loss": 0.7821, + "step": 30524 + }, + { + "epoch": 6.66, + "learning_rate": 1.670854271356784e-06, + "loss": 0.6845, + "step": 30525 + }, + { + "epoch": 6.66, + "learning_rate": 1.6707450294953029e-06, + "loss": 1.8434, + "step": 30526 + }, + { + "epoch": 6.66, + "learning_rate": 1.6706357876338215e-06, + "loss": 0.4457, + "step": 30527 + }, + { + "epoch": 6.66, + "learning_rate": 1.6705265457723403e-06, + "loss": 0.7013, + "step": 30528 + }, + { + "epoch": 6.66, + "learning_rate": 1.6704173039108587e-06, + "loss": 1.4698, + "step": 30529 + }, + { + "epoch": 6.66, + "learning_rate": 1.6703080620493775e-06, + "loss": 2.2453, + "step": 30530 + }, + { + "epoch": 6.66, + "learning_rate": 1.6701988201878961e-06, + "loss": 0.5373, + "step": 30531 + }, + { + "epoch": 6.66, + "learning_rate": 1.6700895783264147e-06, + "loss": 1.4919, + "step": 30532 + }, + { + "epoch": 6.66, + "learning_rate": 1.6699803364649336e-06, + "loss": 1.2701, + "step": 30533 + }, + { + "epoch": 6.66, + "learning_rate": 1.6698710946034522e-06, + "loss": 1.5415, + "step": 30534 + }, + { + "epoch": 6.66, + "learning_rate": 1.669761852741971e-06, + "loss": 0.4012, + "step": 30535 + }, + { + "epoch": 6.66, + "learning_rate": 1.6696526108804894e-06, + "loss": 2.6843, + "step": 30536 + }, + { + "epoch": 6.66, + "learning_rate": 1.6695433690190082e-06, + "loss": 2.2312, + "step": 30537 + }, + { + "epoch": 6.66, + "learning_rate": 1.6694341271575268e-06, + "loss": 1.1731, + "step": 30538 + }, + { + "epoch": 6.66, + "learning_rate": 1.6693248852960456e-06, + "loss": 4.1824, + "step": 30539 + }, + { + "epoch": 6.67, + "learning_rate": 1.6692156434345645e-06, + "loss": 1.533, + "step": 30540 + }, + { + "epoch": 6.67, + "learning_rate": 1.6691064015730829e-06, + "loss": 2.4682, + "step": 30541 + }, + { + "epoch": 6.67, + "learning_rate": 1.6689971597116017e-06, + "loss": 2.4977, + "step": 30542 + }, + { + "epoch": 6.67, + "learning_rate": 1.6688879178501203e-06, + "loss": 1.5858, + "step": 30543 + }, + { + "epoch": 6.67, + "learning_rate": 1.6687786759886391e-06, + "loss": 1.3287, + "step": 30544 + }, + { + "epoch": 6.67, + "learning_rate": 1.6686694341271575e-06, + "loss": 2.1105, + "step": 30545 + }, + { + "epoch": 6.67, + "learning_rate": 1.6685601922656763e-06, + "loss": 0.6926, + "step": 30546 + }, + { + "epoch": 6.67, + "learning_rate": 1.6684509504041952e-06, + "loss": 2.2154, + "step": 30547 + }, + { + "epoch": 6.67, + "learning_rate": 1.6683417085427136e-06, + "loss": 0.2695, + "step": 30548 + }, + { + "epoch": 6.67, + "learning_rate": 1.6682324666812324e-06, + "loss": 1.1646, + "step": 30549 + }, + { + "epoch": 6.67, + "learning_rate": 1.668123224819751e-06, + "loss": 1.7615, + "step": 30550 + }, + { + "epoch": 6.67, + "learning_rate": 1.6680139829582698e-06, + "loss": 2.9997, + "step": 30551 + }, + { + "epoch": 6.67, + "learning_rate": 1.6679047410967882e-06, + "loss": 1.8373, + "step": 30552 + }, + { + "epoch": 6.67, + "learning_rate": 1.667795499235307e-06, + "loss": 1.0471, + "step": 30553 + }, + { + "epoch": 6.67, + "learning_rate": 1.6676862573738259e-06, + "loss": 1.3549, + "step": 30554 + }, + { + "epoch": 6.67, + "learning_rate": 1.6675770155123445e-06, + "loss": 0.1299, + "step": 30555 + }, + { + "epoch": 6.67, + "learning_rate": 1.6674677736508633e-06, + "loss": 2.8038, + "step": 30556 + }, + { + "epoch": 6.67, + "learning_rate": 1.6673585317893817e-06, + "loss": 2.7069, + "step": 30557 + }, + { + "epoch": 6.67, + "learning_rate": 1.6672492899279005e-06, + "loss": 0.572, + "step": 30558 + }, + { + "epoch": 6.67, + "learning_rate": 1.6671400480664193e-06, + "loss": 1.9624, + "step": 30559 + }, + { + "epoch": 6.67, + "learning_rate": 1.6670308062049377e-06, + "loss": 1.6101, + "step": 30560 + }, + { + "epoch": 6.67, + "learning_rate": 1.6669215643434566e-06, + "loss": 0.8362, + "step": 30561 + }, + { + "epoch": 6.67, + "learning_rate": 1.6668123224819752e-06, + "loss": 0.4231, + "step": 30562 + }, + { + "epoch": 6.67, + "learning_rate": 1.666703080620494e-06, + "loss": 0.6695, + "step": 30563 + }, + { + "epoch": 6.67, + "learning_rate": 1.6665938387590124e-06, + "loss": 0.6432, + "step": 30564 + }, + { + "epoch": 6.67, + "learning_rate": 1.6664845968975312e-06, + "loss": 2.891, + "step": 30565 + }, + { + "epoch": 6.67, + "learning_rate": 1.66637535503605e-06, + "loss": 1.0507, + "step": 30566 + }, + { + "epoch": 6.67, + "learning_rate": 1.6662661131745687e-06, + "loss": 3.0881, + "step": 30567 + }, + { + "epoch": 6.67, + "learning_rate": 1.6661568713130875e-06, + "loss": 3.0984, + "step": 30568 + }, + { + "epoch": 6.67, + "learning_rate": 1.6660476294516059e-06, + "loss": 1.7131, + "step": 30569 + }, + { + "epoch": 6.67, + "learning_rate": 1.6659383875901247e-06, + "loss": 2.5472, + "step": 30570 + }, + { + "epoch": 6.67, + "learning_rate": 1.6658291457286433e-06, + "loss": 1.4841, + "step": 30571 + }, + { + "epoch": 6.67, + "learning_rate": 1.6657199038671621e-06, + "loss": 1.0742, + "step": 30572 + }, + { + "epoch": 6.67, + "learning_rate": 1.6656106620056807e-06, + "loss": 0.5379, + "step": 30573 + }, + { + "epoch": 6.67, + "learning_rate": 1.6655014201441994e-06, + "loss": 1.3329, + "step": 30574 + }, + { + "epoch": 6.67, + "learning_rate": 1.6653921782827182e-06, + "loss": 2.2404, + "step": 30575 + }, + { + "epoch": 6.67, + "learning_rate": 1.6652829364212366e-06, + "loss": 0.7156, + "step": 30576 + }, + { + "epoch": 6.67, + "learning_rate": 1.6651736945597554e-06, + "loss": 1.4502, + "step": 30577 + }, + { + "epoch": 6.67, + "learning_rate": 1.665064452698274e-06, + "loss": 0.1595, + "step": 30578 + }, + { + "epoch": 6.67, + "learning_rate": 1.6649552108367928e-06, + "loss": 1.3987, + "step": 30579 + }, + { + "epoch": 6.67, + "learning_rate": 1.6648459689753117e-06, + "loss": 0.2698, + "step": 30580 + }, + { + "epoch": 6.67, + "learning_rate": 1.66473672711383e-06, + "loss": 2.2105, + "step": 30581 + }, + { + "epoch": 6.67, + "learning_rate": 1.6646274852523489e-06, + "loss": 0.936, + "step": 30582 + }, + { + "epoch": 6.67, + "learning_rate": 1.6645182433908675e-06, + "loss": 2.0714, + "step": 30583 + }, + { + "epoch": 6.67, + "learning_rate": 1.6644090015293863e-06, + "loss": 1.5466, + "step": 30584 + }, + { + "epoch": 6.68, + "learning_rate": 1.6642997596679047e-06, + "loss": 0.9266, + "step": 30585 + }, + { + "epoch": 6.68, + "learning_rate": 1.6641905178064235e-06, + "loss": 1.9576, + "step": 30586 + }, + { + "epoch": 6.68, + "learning_rate": 1.6640812759449424e-06, + "loss": 1.1742, + "step": 30587 + }, + { + "epoch": 6.68, + "learning_rate": 1.6639720340834608e-06, + "loss": 1.2542, + "step": 30588 + }, + { + "epoch": 6.68, + "learning_rate": 1.6638627922219796e-06, + "loss": 1.3153, + "step": 30589 + }, + { + "epoch": 6.68, + "learning_rate": 1.6637535503604982e-06, + "loss": 2.1876, + "step": 30590 + }, + { + "epoch": 6.68, + "learning_rate": 1.663644308499017e-06, + "loss": 1.6391, + "step": 30591 + }, + { + "epoch": 6.68, + "learning_rate": 1.6635350666375358e-06, + "loss": 2.4747, + "step": 30592 + }, + { + "epoch": 6.68, + "learning_rate": 1.6634258247760542e-06, + "loss": 0.4168, + "step": 30593 + }, + { + "epoch": 6.68, + "learning_rate": 1.663316582914573e-06, + "loss": 1.4145, + "step": 30594 + }, + { + "epoch": 6.68, + "learning_rate": 1.6632073410530917e-06, + "loss": 2.6083, + "step": 30595 + }, + { + "epoch": 6.68, + "learning_rate": 1.6630980991916105e-06, + "loss": 1.0123, + "step": 30596 + }, + { + "epoch": 6.68, + "learning_rate": 1.662988857330129e-06, + "loss": 0.9754, + "step": 30597 + }, + { + "epoch": 6.68, + "learning_rate": 1.6628796154686477e-06, + "loss": 0.9428, + "step": 30598 + }, + { + "epoch": 6.68, + "learning_rate": 1.6627703736071665e-06, + "loss": 1.9893, + "step": 30599 + }, + { + "epoch": 6.68, + "learning_rate": 1.6626611317456852e-06, + "loss": 0.9136, + "step": 30600 + }, + { + "epoch": 6.68, + "learning_rate": 1.6625518898842038e-06, + "loss": 1.1332, + "step": 30601 + }, + { + "epoch": 6.68, + "learning_rate": 1.6624426480227224e-06, + "loss": 2.1317, + "step": 30602 + }, + { + "epoch": 6.68, + "learning_rate": 1.6623334061612412e-06, + "loss": 0.553, + "step": 30603 + }, + { + "epoch": 6.68, + "learning_rate": 1.6622241642997596e-06, + "loss": 0.3833, + "step": 30604 + }, + { + "epoch": 6.68, + "learning_rate": 1.6621149224382784e-06, + "loss": 0.158, + "step": 30605 + }, + { + "epoch": 6.68, + "learning_rate": 1.6620056805767972e-06, + "loss": 0.8849, + "step": 30606 + }, + { + "epoch": 6.68, + "learning_rate": 1.6618964387153159e-06, + "loss": 2.3536, + "step": 30607 + }, + { + "epoch": 6.68, + "learning_rate": 1.6617871968538347e-06, + "loss": 0.8935, + "step": 30608 + }, + { + "epoch": 6.68, + "learning_rate": 1.661677954992353e-06, + "loss": 0.8463, + "step": 30609 + }, + { + "epoch": 6.68, + "learning_rate": 1.661568713130872e-06, + "loss": 1.6975, + "step": 30610 + }, + { + "epoch": 6.68, + "learning_rate": 1.6614594712693905e-06, + "loss": 0.4527, + "step": 30611 + }, + { + "epoch": 6.68, + "learning_rate": 1.6613502294079093e-06, + "loss": 1.0677, + "step": 30612 + }, + { + "epoch": 6.68, + "learning_rate": 1.661240987546428e-06, + "loss": 0.9388, + "step": 30613 + }, + { + "epoch": 6.68, + "learning_rate": 1.6611317456849466e-06, + "loss": 1.2525, + "step": 30614 + }, + { + "epoch": 6.68, + "learning_rate": 1.6610225038234654e-06, + "loss": 0.86, + "step": 30615 + }, + { + "epoch": 6.68, + "learning_rate": 1.660913261961984e-06, + "loss": 0.8099, + "step": 30616 + }, + { + "epoch": 6.68, + "learning_rate": 1.6608040201005026e-06, + "loss": 0.8494, + "step": 30617 + }, + { + "epoch": 6.68, + "learning_rate": 1.6606947782390212e-06, + "loss": 1.7791, + "step": 30618 + }, + { + "epoch": 6.68, + "learning_rate": 1.66058553637754e-06, + "loss": 2.1684, + "step": 30619 + }, + { + "epoch": 6.68, + "learning_rate": 1.6604762945160589e-06, + "loss": 1.2696, + "step": 30620 + }, + { + "epoch": 6.68, + "learning_rate": 1.6603670526545773e-06, + "loss": 0.4522, + "step": 30621 + }, + { + "epoch": 6.68, + "learning_rate": 1.660257810793096e-06, + "loss": 2.2468, + "step": 30622 + }, + { + "epoch": 6.68, + "learning_rate": 1.6601485689316147e-06, + "loss": 1.9776, + "step": 30623 + }, + { + "epoch": 6.68, + "learning_rate": 1.6600393270701335e-06, + "loss": 1.1214, + "step": 30624 + }, + { + "epoch": 6.68, + "learning_rate": 1.659930085208652e-06, + "loss": 1.1331, + "step": 30625 + }, + { + "epoch": 6.68, + "learning_rate": 1.6598208433471707e-06, + "loss": 1.3454, + "step": 30626 + }, + { + "epoch": 6.68, + "learning_rate": 1.6597116014856896e-06, + "loss": 0.8303, + "step": 30627 + }, + { + "epoch": 6.68, + "learning_rate": 1.6596023596242082e-06, + "loss": 2.1882, + "step": 30628 + }, + { + "epoch": 6.68, + "learning_rate": 1.6594931177627268e-06, + "loss": 0.7087, + "step": 30629 + }, + { + "epoch": 6.68, + "learning_rate": 1.6593838759012454e-06, + "loss": 1.6352, + "step": 30630 + }, + { + "epoch": 6.69, + "learning_rate": 1.6592746340397642e-06, + "loss": 1.215, + "step": 30631 + }, + { + "epoch": 6.69, + "learning_rate": 1.659165392178283e-06, + "loss": 2.0759, + "step": 30632 + }, + { + "epoch": 6.69, + "learning_rate": 1.6590561503168014e-06, + "loss": 0.6581, + "step": 30633 + }, + { + "epoch": 6.69, + "learning_rate": 1.6589469084553203e-06, + "loss": 1.734, + "step": 30634 + }, + { + "epoch": 6.69, + "learning_rate": 1.6588376665938389e-06, + "loss": 0.0829, + "step": 30635 + }, + { + "epoch": 6.69, + "learning_rate": 1.6587284247323577e-06, + "loss": 1.2406, + "step": 30636 + }, + { + "epoch": 6.69, + "learning_rate": 1.658619182870876e-06, + "loss": 2.1023, + "step": 30637 + }, + { + "epoch": 6.69, + "learning_rate": 1.658509941009395e-06, + "loss": 0.5287, + "step": 30638 + }, + { + "epoch": 6.69, + "learning_rate": 1.6584006991479137e-06, + "loss": 1.7557, + "step": 30639 + }, + { + "epoch": 6.69, + "learning_rate": 1.6582914572864323e-06, + "loss": 0.8384, + "step": 30640 + }, + { + "epoch": 6.69, + "learning_rate": 1.658182215424951e-06, + "loss": 1.0311, + "step": 30641 + }, + { + "epoch": 6.69, + "learning_rate": 1.6580729735634696e-06, + "loss": 0.8193, + "step": 30642 + }, + { + "epoch": 6.69, + "learning_rate": 1.6579637317019884e-06, + "loss": 0.508, + "step": 30643 + }, + { + "epoch": 6.69, + "learning_rate": 1.657854489840507e-06, + "loss": 1.0882, + "step": 30644 + }, + { + "epoch": 6.69, + "learning_rate": 1.6577452479790256e-06, + "loss": 1.3203, + "step": 30645 + }, + { + "epoch": 6.69, + "learning_rate": 1.6576360061175444e-06, + "loss": 2.2558, + "step": 30646 + }, + { + "epoch": 6.69, + "learning_rate": 1.657526764256063e-06, + "loss": 1.4967, + "step": 30647 + }, + { + "epoch": 6.69, + "learning_rate": 1.6574175223945819e-06, + "loss": 1.4132, + "step": 30648 + }, + { + "epoch": 6.69, + "learning_rate": 1.6573082805331003e-06, + "loss": 1.2424, + "step": 30649 + }, + { + "epoch": 6.69, + "learning_rate": 1.657199038671619e-06, + "loss": 2.1177, + "step": 30650 + }, + { + "epoch": 6.69, + "learning_rate": 1.6570897968101377e-06, + "loss": 0.6079, + "step": 30651 + }, + { + "epoch": 6.69, + "learning_rate": 1.6569805549486565e-06, + "loss": 0.1715, + "step": 30652 + }, + { + "epoch": 6.69, + "learning_rate": 1.6568713130871753e-06, + "loss": 1.3883, + "step": 30653 + }, + { + "epoch": 6.69, + "learning_rate": 1.6567620712256937e-06, + "loss": 0.9804, + "step": 30654 + }, + { + "epoch": 6.69, + "learning_rate": 1.6566528293642126e-06, + "loss": 1.3978, + "step": 30655 + }, + { + "epoch": 6.69, + "learning_rate": 1.6565435875027312e-06, + "loss": 0.5957, + "step": 30656 + }, + { + "epoch": 6.69, + "learning_rate": 1.6564343456412498e-06, + "loss": 1.5762, + "step": 30657 + }, + { + "epoch": 6.69, + "learning_rate": 1.6563251037797684e-06, + "loss": 0.3467, + "step": 30658 + }, + { + "epoch": 6.69, + "learning_rate": 1.6562158619182872e-06, + "loss": 0.9597, + "step": 30659 + }, + { + "epoch": 6.69, + "learning_rate": 1.656106620056806e-06, + "loss": 1.2707, + "step": 30660 + }, + { + "epoch": 6.69, + "learning_rate": 1.6559973781953244e-06, + "loss": 2.8491, + "step": 30661 + }, + { + "epoch": 6.69, + "learning_rate": 1.6558881363338433e-06, + "loss": 2.2507, + "step": 30662 + }, + { + "epoch": 6.69, + "learning_rate": 1.6557788944723619e-06, + "loss": 0.5199, + "step": 30663 + }, + { + "epoch": 6.69, + "learning_rate": 1.6556696526108807e-06, + "loss": 0.3834, + "step": 30664 + }, + { + "epoch": 6.69, + "learning_rate": 1.6555604107493995e-06, + "loss": 1.28, + "step": 30665 + }, + { + "epoch": 6.69, + "learning_rate": 1.655451168887918e-06, + "loss": 1.0755, + "step": 30666 + }, + { + "epoch": 6.69, + "learning_rate": 1.6553419270264367e-06, + "loss": 0.7493, + "step": 30667 + }, + { + "epoch": 6.69, + "learning_rate": 1.6552326851649554e-06, + "loss": 1.5468, + "step": 30668 + }, + { + "epoch": 6.69, + "learning_rate": 1.6551234433034742e-06, + "loss": 2.0826, + "step": 30669 + }, + { + "epoch": 6.69, + "learning_rate": 1.6550142014419926e-06, + "loss": 1.6697, + "step": 30670 + }, + { + "epoch": 6.69, + "learning_rate": 1.6549049595805114e-06, + "loss": 0.1454, + "step": 30671 + }, + { + "epoch": 6.69, + "learning_rate": 1.6547957177190302e-06, + "loss": 0.8884, + "step": 30672 + }, + { + "epoch": 6.69, + "learning_rate": 1.6546864758575486e-06, + "loss": 1.8006, + "step": 30673 + }, + { + "epoch": 6.69, + "learning_rate": 1.6545772339960674e-06, + "loss": 3.0765, + "step": 30674 + }, + { + "epoch": 6.69, + "learning_rate": 1.654467992134586e-06, + "loss": 0.2799, + "step": 30675 + }, + { + "epoch": 6.69, + "learning_rate": 1.6543587502731049e-06, + "loss": 2.0858, + "step": 30676 + }, + { + "epoch": 6.7, + "learning_rate": 1.6542495084116233e-06, + "loss": 2.1273, + "step": 30677 + }, + { + "epoch": 6.7, + "learning_rate": 1.654140266550142e-06, + "loss": 3.0358, + "step": 30678 + }, + { + "epoch": 6.7, + "learning_rate": 1.654031024688661e-06, + "loss": 1.8831, + "step": 30679 + }, + { + "epoch": 6.7, + "learning_rate": 1.6539217828271795e-06, + "loss": 1.1656, + "step": 30680 + }, + { + "epoch": 6.7, + "learning_rate": 1.6538125409656984e-06, + "loss": 0.8397, + "step": 30681 + }, + { + "epoch": 6.7, + "learning_rate": 1.6537032991042168e-06, + "loss": 1.3871, + "step": 30682 + }, + { + "epoch": 6.7, + "learning_rate": 1.6535940572427356e-06, + "loss": 0.2063, + "step": 30683 + }, + { + "epoch": 6.7, + "learning_rate": 1.6534848153812542e-06, + "loss": 2.9829, + "step": 30684 + }, + { + "epoch": 6.7, + "learning_rate": 1.6533755735197728e-06, + "loss": 1.1819, + "step": 30685 + }, + { + "epoch": 6.7, + "learning_rate": 1.6532663316582916e-06, + "loss": 0.9933, + "step": 30686 + }, + { + "epoch": 6.7, + "learning_rate": 1.6531570897968102e-06, + "loss": 3.3827, + "step": 30687 + }, + { + "epoch": 6.7, + "learning_rate": 1.653047847935329e-06, + "loss": 1.3317, + "step": 30688 + }, + { + "epoch": 6.7, + "learning_rate": 1.6529386060738475e-06, + "loss": 0.9951, + "step": 30689 + }, + { + "epoch": 6.7, + "learning_rate": 1.6528293642123663e-06, + "loss": 1.1058, + "step": 30690 + }, + { + "epoch": 6.7, + "learning_rate": 1.6527201223508849e-06, + "loss": 1.49, + "step": 30691 + }, + { + "epoch": 6.7, + "learning_rate": 1.6526108804894037e-06, + "loss": 2.2204, + "step": 30692 + }, + { + "epoch": 6.7, + "learning_rate": 1.6525016386279225e-06, + "loss": 1.8528, + "step": 30693 + }, + { + "epoch": 6.7, + "learning_rate": 1.652392396766441e-06, + "loss": 1.9593, + "step": 30694 + }, + { + "epoch": 6.7, + "learning_rate": 1.6522831549049598e-06, + "loss": 1.3163, + "step": 30695 + }, + { + "epoch": 6.7, + "learning_rate": 1.6521739130434784e-06, + "loss": 1.6953, + "step": 30696 + }, + { + "epoch": 6.7, + "learning_rate": 1.6520646711819972e-06, + "loss": 0.7362, + "step": 30697 + }, + { + "epoch": 6.7, + "learning_rate": 1.6519554293205158e-06, + "loss": 1.715, + "step": 30698 + }, + { + "epoch": 6.7, + "learning_rate": 1.6518461874590344e-06, + "loss": 1.0844, + "step": 30699 + }, + { + "epoch": 6.7, + "learning_rate": 1.6517369455975532e-06, + "loss": 0.7139, + "step": 30700 + }, + { + "epoch": 6.7, + "learning_rate": 1.6516277037360716e-06, + "loss": 3.7079, + "step": 30701 + }, + { + "epoch": 6.7, + "learning_rate": 1.6515184618745905e-06, + "loss": 0.5058, + "step": 30702 + }, + { + "epoch": 6.7, + "learning_rate": 1.651409220013109e-06, + "loss": 3.1226, + "step": 30703 + }, + { + "epoch": 6.7, + "learning_rate": 1.6512999781516279e-06, + "loss": 1.5152, + "step": 30704 + }, + { + "epoch": 6.7, + "learning_rate": 1.6511907362901467e-06, + "loss": 0.9883, + "step": 30705 + }, + { + "epoch": 6.7, + "learning_rate": 1.6510814944286651e-06, + "loss": 1.0041, + "step": 30706 + }, + { + "epoch": 6.7, + "learning_rate": 1.650972252567184e-06, + "loss": 0.8089, + "step": 30707 + }, + { + "epoch": 6.7, + "learning_rate": 1.6508630107057025e-06, + "loss": 2.483, + "step": 30708 + }, + { + "epoch": 6.7, + "learning_rate": 1.6507537688442214e-06, + "loss": 1.1068, + "step": 30709 + }, + { + "epoch": 6.7, + "learning_rate": 1.6506445269827398e-06, + "loss": 1.5975, + "step": 30710 + }, + { + "epoch": 6.7, + "learning_rate": 1.6505352851212586e-06, + "loss": 1.1939, + "step": 30711 + }, + { + "epoch": 6.7, + "learning_rate": 1.6504260432597774e-06, + "loss": 1.6036, + "step": 30712 + }, + { + "epoch": 6.7, + "learning_rate": 1.6503168013982958e-06, + "loss": 0.4237, + "step": 30713 + }, + { + "epoch": 6.7, + "learning_rate": 1.6502075595368146e-06, + "loss": 1.1571, + "step": 30714 + }, + { + "epoch": 6.7, + "learning_rate": 1.6500983176753332e-06, + "loss": 2.5202, + "step": 30715 + }, + { + "epoch": 6.7, + "learning_rate": 1.649989075813852e-06, + "loss": 1.0903, + "step": 30716 + }, + { + "epoch": 6.7, + "learning_rate": 1.6498798339523705e-06, + "loss": 0.9916, + "step": 30717 + }, + { + "epoch": 6.7, + "learning_rate": 1.6497705920908893e-06, + "loss": 0.7391, + "step": 30718 + }, + { + "epoch": 6.7, + "learning_rate": 1.6496613502294081e-06, + "loss": 1.0877, + "step": 30719 + }, + { + "epoch": 6.7, + "learning_rate": 1.6495521083679267e-06, + "loss": 1.545, + "step": 30720 + }, + { + "epoch": 6.7, + "learning_rate": 1.6494428665064456e-06, + "loss": 1.6315, + "step": 30721 + }, + { + "epoch": 6.7, + "learning_rate": 1.649333624644964e-06, + "loss": 2.4287, + "step": 30722 + }, + { + "epoch": 6.71, + "learning_rate": 1.6492243827834828e-06, + "loss": 1.983, + "step": 30723 + }, + { + "epoch": 6.71, + "learning_rate": 1.6491151409220014e-06, + "loss": 1.575, + "step": 30724 + }, + { + "epoch": 6.71, + "learning_rate": 1.6490058990605202e-06, + "loss": 1.6809, + "step": 30725 + }, + { + "epoch": 6.71, + "learning_rate": 1.6488966571990388e-06, + "loss": 1.747, + "step": 30726 + }, + { + "epoch": 6.71, + "learning_rate": 1.6487874153375574e-06, + "loss": 1.3387, + "step": 30727 + }, + { + "epoch": 6.71, + "learning_rate": 1.6486781734760763e-06, + "loss": 3.2112, + "step": 30728 + }, + { + "epoch": 6.71, + "learning_rate": 1.6485689316145947e-06, + "loss": 1.5482, + "step": 30729 + }, + { + "epoch": 6.71, + "learning_rate": 1.6484596897531135e-06, + "loss": 2.1733, + "step": 30730 + }, + { + "epoch": 6.71, + "learning_rate": 1.6483504478916323e-06, + "loss": 2.2199, + "step": 30731 + }, + { + "epoch": 6.71, + "learning_rate": 1.648241206030151e-06, + "loss": 0.6504, + "step": 30732 + }, + { + "epoch": 6.71, + "learning_rate": 1.6481319641686697e-06, + "loss": 3.3662, + "step": 30733 + }, + { + "epoch": 6.71, + "learning_rate": 1.6480227223071881e-06, + "loss": 2.0935, + "step": 30734 + }, + { + "epoch": 6.71, + "learning_rate": 1.647913480445707e-06, + "loss": 2.5072, + "step": 30735 + }, + { + "epoch": 6.71, + "learning_rate": 1.6478042385842256e-06, + "loss": 0.3237, + "step": 30736 + }, + { + "epoch": 6.71, + "learning_rate": 1.6476949967227444e-06, + "loss": 0.855, + "step": 30737 + }, + { + "epoch": 6.71, + "learning_rate": 1.647585754861263e-06, + "loss": 2.5045, + "step": 30738 + }, + { + "epoch": 6.71, + "learning_rate": 1.6474765129997816e-06, + "loss": 0.2183, + "step": 30739 + }, + { + "epoch": 6.71, + "learning_rate": 1.6473672711383004e-06, + "loss": 1.0502, + "step": 30740 + }, + { + "epoch": 6.71, + "learning_rate": 1.647258029276819e-06, + "loss": 1.0442, + "step": 30741 + }, + { + "epoch": 6.71, + "learning_rate": 1.6471487874153377e-06, + "loss": 1.351, + "step": 30742 + }, + { + "epoch": 6.71, + "learning_rate": 1.6470395455538563e-06, + "loss": 0.995, + "step": 30743 + }, + { + "epoch": 6.71, + "learning_rate": 1.646930303692375e-06, + "loss": 1.5337, + "step": 30744 + }, + { + "epoch": 6.71, + "learning_rate": 1.646821061830894e-06, + "loss": 1.2735, + "step": 30745 + }, + { + "epoch": 6.71, + "learning_rate": 1.6467118199694123e-06, + "loss": 2.4249, + "step": 30746 + }, + { + "epoch": 6.71, + "learning_rate": 1.6466025781079311e-06, + "loss": 0.3077, + "step": 30747 + }, + { + "epoch": 6.71, + "learning_rate": 1.6464933362464497e-06, + "loss": 3.2919, + "step": 30748 + }, + { + "epoch": 6.71, + "learning_rate": 1.6463840943849686e-06, + "loss": 0.9311, + "step": 30749 + }, + { + "epoch": 6.71, + "learning_rate": 1.646274852523487e-06, + "loss": 0.1345, + "step": 30750 + }, + { + "epoch": 6.71, + "learning_rate": 1.6461656106620058e-06, + "loss": 1.1724, + "step": 30751 + }, + { + "epoch": 6.71, + "learning_rate": 1.6460563688005246e-06, + "loss": 1.7697, + "step": 30752 + }, + { + "epoch": 6.71, + "learning_rate": 1.6459471269390432e-06, + "loss": 1.354, + "step": 30753 + }, + { + "epoch": 6.71, + "learning_rate": 1.6458378850775618e-06, + "loss": 2.7637, + "step": 30754 + }, + { + "epoch": 6.71, + "learning_rate": 1.6457286432160804e-06, + "loss": 1.8041, + "step": 30755 + }, + { + "epoch": 6.71, + "learning_rate": 1.6456194013545993e-06, + "loss": 0.9094, + "step": 30756 + }, + { + "epoch": 6.71, + "learning_rate": 1.6455101594931177e-06, + "loss": 1.1484, + "step": 30757 + }, + { + "epoch": 6.71, + "learning_rate": 1.6454009176316365e-06, + "loss": 1.7116, + "step": 30758 + }, + { + "epoch": 6.71, + "learning_rate": 1.6452916757701553e-06, + "loss": 1.8583, + "step": 30759 + }, + { + "epoch": 6.71, + "learning_rate": 1.645182433908674e-06, + "loss": 1.0736, + "step": 30760 + }, + { + "epoch": 6.71, + "learning_rate": 1.6450731920471927e-06, + "loss": 0.1082, + "step": 30761 + }, + { + "epoch": 6.71, + "learning_rate": 1.6449639501857111e-06, + "loss": 1.2303, + "step": 30762 + }, + { + "epoch": 6.71, + "learning_rate": 1.64485470832423e-06, + "loss": 1.5433, + "step": 30763 + }, + { + "epoch": 6.71, + "learning_rate": 1.6447454664627488e-06, + "loss": 0.695, + "step": 30764 + }, + { + "epoch": 6.71, + "learning_rate": 1.6446362246012674e-06, + "loss": 0.7632, + "step": 30765 + }, + { + "epoch": 6.71, + "learning_rate": 1.644526982739786e-06, + "loss": 0.4522, + "step": 30766 + }, + { + "epoch": 6.71, + "learning_rate": 1.6444177408783046e-06, + "loss": 1.2327, + "step": 30767 + }, + { + "epoch": 6.71, + "learning_rate": 1.6443084990168234e-06, + "loss": 1.0288, + "step": 30768 + }, + { + "epoch": 6.72, + "learning_rate": 1.644199257155342e-06, + "loss": 3.4782, + "step": 30769 + }, + { + "epoch": 6.72, + "learning_rate": 1.6440900152938607e-06, + "loss": 1.8109, + "step": 30770 + }, + { + "epoch": 6.72, + "learning_rate": 1.6439807734323795e-06, + "loss": 1.3597, + "step": 30771 + }, + { + "epoch": 6.72, + "learning_rate": 1.643871531570898e-06, + "loss": 1.3987, + "step": 30772 + }, + { + "epoch": 6.72, + "learning_rate": 1.643762289709417e-06, + "loss": 0.9603, + "step": 30773 + }, + { + "epoch": 6.72, + "learning_rate": 1.6436530478479353e-06, + "loss": 0.714, + "step": 30774 + }, + { + "epoch": 6.72, + "learning_rate": 1.6435438059864541e-06, + "loss": 0.1005, + "step": 30775 + }, + { + "epoch": 6.72, + "learning_rate": 1.6434345641249728e-06, + "loss": 1.0215, + "step": 30776 + }, + { + "epoch": 6.72, + "learning_rate": 1.6433253222634916e-06, + "loss": 0.6944, + "step": 30777 + }, + { + "epoch": 6.72, + "learning_rate": 1.6432160804020104e-06, + "loss": 1.486, + "step": 30778 + }, + { + "epoch": 6.72, + "learning_rate": 1.6431068385405288e-06, + "loss": 1.2813, + "step": 30779 + }, + { + "epoch": 6.72, + "learning_rate": 1.6429975966790476e-06, + "loss": 1.4648, + "step": 30780 + }, + { + "epoch": 6.72, + "learning_rate": 1.6428883548175662e-06, + "loss": 0.8779, + "step": 30781 + }, + { + "epoch": 6.72, + "learning_rate": 1.6427791129560848e-06, + "loss": 2.8236, + "step": 30782 + }, + { + "epoch": 6.72, + "learning_rate": 1.6426698710946035e-06, + "loss": 1.1451, + "step": 30783 + }, + { + "epoch": 6.72, + "learning_rate": 1.6425606292331223e-06, + "loss": 0.8091, + "step": 30784 + }, + { + "epoch": 6.72, + "learning_rate": 1.642451387371641e-06, + "loss": 1.8692, + "step": 30785 + }, + { + "epoch": 6.72, + "learning_rate": 1.6423421455101595e-06, + "loss": 0.6839, + "step": 30786 + }, + { + "epoch": 6.72, + "learning_rate": 1.6422329036486783e-06, + "loss": 1.5868, + "step": 30787 + }, + { + "epoch": 6.72, + "learning_rate": 1.642123661787197e-06, + "loss": 1.0042, + "step": 30788 + }, + { + "epoch": 6.72, + "learning_rate": 1.6420144199257158e-06, + "loss": 2.9808, + "step": 30789 + }, + { + "epoch": 6.72, + "learning_rate": 1.6419051780642342e-06, + "loss": 1.6105, + "step": 30790 + }, + { + "epoch": 6.72, + "learning_rate": 1.641795936202753e-06, + "loss": 1.6951, + "step": 30791 + }, + { + "epoch": 6.72, + "learning_rate": 1.6416866943412718e-06, + "loss": 1.4236, + "step": 30792 + }, + { + "epoch": 6.72, + "learning_rate": 1.6415774524797904e-06, + "loss": 2.5149, + "step": 30793 + }, + { + "epoch": 6.72, + "learning_rate": 1.6414682106183092e-06, + "loss": 0.6583, + "step": 30794 + }, + { + "epoch": 6.72, + "learning_rate": 1.6413589687568276e-06, + "loss": 0.8806, + "step": 30795 + }, + { + "epoch": 6.72, + "learning_rate": 1.6412497268953465e-06, + "loss": 0.7861, + "step": 30796 + }, + { + "epoch": 6.72, + "learning_rate": 1.6411404850338653e-06, + "loss": 1.7512, + "step": 30797 + }, + { + "epoch": 6.72, + "learning_rate": 1.6410312431723837e-06, + "loss": 0.9858, + "step": 30798 + }, + { + "epoch": 6.72, + "learning_rate": 1.6409220013109025e-06, + "loss": 1.028, + "step": 30799 + }, + { + "epoch": 6.72, + "learning_rate": 1.6408127594494211e-06, + "loss": 2.0405, + "step": 30800 + }, + { + "epoch": 6.72, + "learning_rate": 1.64070351758794e-06, + "loss": 1.7064, + "step": 30801 + }, + { + "epoch": 6.72, + "learning_rate": 1.6405942757264583e-06, + "loss": 1.1072, + "step": 30802 + }, + { + "epoch": 6.72, + "learning_rate": 1.6404850338649772e-06, + "loss": 3.572, + "step": 30803 + }, + { + "epoch": 6.72, + "learning_rate": 1.640375792003496e-06, + "loss": 1.7551, + "step": 30804 + }, + { + "epoch": 6.72, + "learning_rate": 1.6402665501420146e-06, + "loss": 0.4908, + "step": 30805 + }, + { + "epoch": 6.72, + "learning_rate": 1.6401573082805334e-06, + "loss": 1.1544, + "step": 30806 + }, + { + "epoch": 6.72, + "learning_rate": 1.6400480664190518e-06, + "loss": 2.3406, + "step": 30807 + }, + { + "epoch": 6.72, + "learning_rate": 1.6399388245575706e-06, + "loss": 1.8616, + "step": 30808 + }, + { + "epoch": 6.72, + "learning_rate": 1.6398295826960892e-06, + "loss": 0.7306, + "step": 30809 + }, + { + "epoch": 6.72, + "learning_rate": 1.6397203408346079e-06, + "loss": 1.282, + "step": 30810 + }, + { + "epoch": 6.72, + "learning_rate": 1.6396110989731267e-06, + "loss": 0.8113, + "step": 30811 + }, + { + "epoch": 6.72, + "learning_rate": 1.6395018571116453e-06, + "loss": 1.4015, + "step": 30812 + }, + { + "epoch": 6.72, + "learning_rate": 1.6393926152501641e-06, + "loss": 0.7831, + "step": 30813 + }, + { + "epoch": 6.73, + "learning_rate": 1.6392833733886825e-06, + "loss": 1.5142, + "step": 30814 + }, + { + "epoch": 6.73, + "learning_rate": 1.6391741315272013e-06, + "loss": 1.046, + "step": 30815 + }, + { + "epoch": 6.73, + "learning_rate": 1.63906488966572e-06, + "loss": 2.6829, + "step": 30816 + }, + { + "epoch": 6.73, + "learning_rate": 1.6389556478042388e-06, + "loss": 3.1308, + "step": 30817 + }, + { + "epoch": 6.73, + "learning_rate": 1.6388464059427576e-06, + "loss": 0.2169, + "step": 30818 + }, + { + "epoch": 6.73, + "learning_rate": 1.638737164081276e-06, + "loss": 1.2806, + "step": 30819 + }, + { + "epoch": 6.73, + "learning_rate": 1.6386279222197948e-06, + "loss": 1.2121, + "step": 30820 + }, + { + "epoch": 6.73, + "learning_rate": 1.6385186803583134e-06, + "loss": 0.6159, + "step": 30821 + }, + { + "epoch": 6.73, + "learning_rate": 1.6384094384968322e-06, + "loss": 2.2032, + "step": 30822 + }, + { + "epoch": 6.73, + "learning_rate": 1.6383001966353506e-06, + "loss": 1.2961, + "step": 30823 + }, + { + "epoch": 6.73, + "learning_rate": 1.6381909547738695e-06, + "loss": 1.7721, + "step": 30824 + }, + { + "epoch": 6.73, + "learning_rate": 1.6380817129123883e-06, + "loss": 0.9472, + "step": 30825 + }, + { + "epoch": 6.73, + "learning_rate": 1.6379724710509067e-06, + "loss": 0.3982, + "step": 30826 + }, + { + "epoch": 6.73, + "learning_rate": 1.6378632291894255e-06, + "loss": 1.6781, + "step": 30827 + }, + { + "epoch": 6.73, + "learning_rate": 1.6377539873279441e-06, + "loss": 1.8595, + "step": 30828 + }, + { + "epoch": 6.73, + "learning_rate": 1.637644745466463e-06, + "loss": 0.51, + "step": 30829 + }, + { + "epoch": 6.73, + "learning_rate": 1.6375355036049813e-06, + "loss": 1.037, + "step": 30830 + }, + { + "epoch": 6.73, + "learning_rate": 1.6374262617435002e-06, + "loss": 0.9516, + "step": 30831 + }, + { + "epoch": 6.73, + "learning_rate": 1.637317019882019e-06, + "loss": 0.7441, + "step": 30832 + }, + { + "epoch": 6.73, + "learning_rate": 1.6372077780205376e-06, + "loss": 0.4736, + "step": 30833 + }, + { + "epoch": 6.73, + "learning_rate": 1.6370985361590564e-06, + "loss": 1.1661, + "step": 30834 + }, + { + "epoch": 6.73, + "learning_rate": 1.6369892942975748e-06, + "loss": 1.1736, + "step": 30835 + }, + { + "epoch": 6.73, + "learning_rate": 1.6368800524360936e-06, + "loss": 1.2771, + "step": 30836 + }, + { + "epoch": 6.73, + "learning_rate": 1.6367708105746125e-06, + "loss": 1.4776, + "step": 30837 + }, + { + "epoch": 6.73, + "learning_rate": 1.6366615687131309e-06, + "loss": 1.7984, + "step": 30838 + }, + { + "epoch": 6.73, + "learning_rate": 1.6365523268516497e-06, + "loss": 0.4283, + "step": 30839 + }, + { + "epoch": 6.73, + "learning_rate": 1.6364430849901683e-06, + "loss": 1.7206, + "step": 30840 + }, + { + "epoch": 6.73, + "learning_rate": 1.6363338431286871e-06, + "loss": 0.8582, + "step": 30841 + }, + { + "epoch": 6.73, + "learning_rate": 1.6362246012672055e-06, + "loss": 0.0701, + "step": 30842 + }, + { + "epoch": 6.73, + "learning_rate": 1.6361153594057243e-06, + "loss": 2.2561, + "step": 30843 + }, + { + "epoch": 6.73, + "learning_rate": 1.6360061175442432e-06, + "loss": 1.4846, + "step": 30844 + }, + { + "epoch": 6.73, + "learning_rate": 1.6358968756827618e-06, + "loss": 0.5277, + "step": 30845 + }, + { + "epoch": 6.73, + "learning_rate": 1.6357876338212806e-06, + "loss": 1.9232, + "step": 30846 + }, + { + "epoch": 6.73, + "learning_rate": 1.635678391959799e-06, + "loss": 1.9052, + "step": 30847 + }, + { + "epoch": 6.73, + "learning_rate": 1.6355691500983178e-06, + "loss": 2.0117, + "step": 30848 + }, + { + "epoch": 6.73, + "learning_rate": 1.6354599082368364e-06, + "loss": 1.3097, + "step": 30849 + }, + { + "epoch": 6.73, + "learning_rate": 1.6353506663753553e-06, + "loss": 0.2744, + "step": 30850 + }, + { + "epoch": 6.73, + "learning_rate": 1.6352414245138739e-06, + "loss": 0.1115, + "step": 30851 + }, + { + "epoch": 6.73, + "learning_rate": 1.6351321826523925e-06, + "loss": 2.4957, + "step": 30852 + }, + { + "epoch": 6.73, + "learning_rate": 1.6350229407909113e-06, + "loss": 1.7094, + "step": 30853 + }, + { + "epoch": 6.73, + "learning_rate": 1.6349136989294297e-06, + "loss": 1.9135, + "step": 30854 + }, + { + "epoch": 6.73, + "learning_rate": 1.6348044570679485e-06, + "loss": 1.0733, + "step": 30855 + }, + { + "epoch": 6.73, + "learning_rate": 1.6346952152064671e-06, + "loss": 0.7289, + "step": 30856 + }, + { + "epoch": 6.73, + "learning_rate": 1.634585973344986e-06, + "loss": 0.6853, + "step": 30857 + }, + { + "epoch": 6.73, + "learning_rate": 1.6344767314835048e-06, + "loss": 0.7331, + "step": 30858 + }, + { + "epoch": 6.73, + "learning_rate": 1.6343674896220232e-06, + "loss": 1.3192, + "step": 30859 + }, + { + "epoch": 6.74, + "learning_rate": 1.634258247760542e-06, + "loss": 1.8505, + "step": 30860 + }, + { + "epoch": 6.74, + "learning_rate": 1.6341490058990606e-06, + "loss": 1.1565, + "step": 30861 + }, + { + "epoch": 6.74, + "learning_rate": 1.6340397640375794e-06, + "loss": 0.8512, + "step": 30862 + }, + { + "epoch": 6.74, + "learning_rate": 1.6339305221760978e-06, + "loss": 1.6488, + "step": 30863 + }, + { + "epoch": 6.74, + "learning_rate": 1.6338212803146167e-06, + "loss": 0.9128, + "step": 30864 + }, + { + "epoch": 6.74, + "learning_rate": 1.6337120384531355e-06, + "loss": 2.8737, + "step": 30865 + }, + { + "epoch": 6.74, + "learning_rate": 1.633602796591654e-06, + "loss": 1.3227, + "step": 30866 + }, + { + "epoch": 6.74, + "learning_rate": 1.6334935547301727e-06, + "loss": 0.1169, + "step": 30867 + }, + { + "epoch": 6.74, + "learning_rate": 1.6333843128686913e-06, + "loss": 0.8652, + "step": 30868 + }, + { + "epoch": 6.74, + "learning_rate": 1.6332750710072101e-06, + "loss": 0.7119, + "step": 30869 + }, + { + "epoch": 6.74, + "learning_rate": 1.633165829145729e-06, + "loss": 1.8728, + "step": 30870 + }, + { + "epoch": 6.74, + "learning_rate": 1.6330565872842474e-06, + "loss": 1.0919, + "step": 30871 + }, + { + "epoch": 6.74, + "learning_rate": 1.6329473454227662e-06, + "loss": 2.3853, + "step": 30872 + }, + { + "epoch": 6.74, + "learning_rate": 1.6328381035612848e-06, + "loss": 1.5426, + "step": 30873 + }, + { + "epoch": 6.74, + "learning_rate": 1.6327288616998036e-06, + "loss": 1.8736, + "step": 30874 + }, + { + "epoch": 6.74, + "learning_rate": 1.632619619838322e-06, + "loss": 1.4102, + "step": 30875 + }, + { + "epoch": 6.74, + "learning_rate": 1.6325103779768408e-06, + "loss": 0.6886, + "step": 30876 + }, + { + "epoch": 6.74, + "learning_rate": 1.6324011361153597e-06, + "loss": 2.0621, + "step": 30877 + }, + { + "epoch": 6.74, + "learning_rate": 1.6322918942538783e-06, + "loss": 0.3224, + "step": 30878 + }, + { + "epoch": 6.74, + "learning_rate": 1.6321826523923969e-06, + "loss": 0.0886, + "step": 30879 + }, + { + "epoch": 6.74, + "learning_rate": 1.6320734105309155e-06, + "loss": 0.7129, + "step": 30880 + }, + { + "epoch": 6.74, + "learning_rate": 1.6319641686694343e-06, + "loss": 1.9041, + "step": 30881 + }, + { + "epoch": 6.74, + "learning_rate": 1.6318549268079527e-06, + "loss": 0.7577, + "step": 30882 + }, + { + "epoch": 6.74, + "learning_rate": 1.6317456849464715e-06, + "loss": 0.4001, + "step": 30883 + }, + { + "epoch": 6.74, + "learning_rate": 1.6316364430849904e-06, + "loss": 1.7432, + "step": 30884 + }, + { + "epoch": 6.74, + "learning_rate": 1.631527201223509e-06, + "loss": 0.2125, + "step": 30885 + }, + { + "epoch": 6.74, + "learning_rate": 1.6314179593620278e-06, + "loss": 1.1251, + "step": 30886 + }, + { + "epoch": 6.74, + "learning_rate": 1.6313087175005462e-06, + "loss": 1.3198, + "step": 30887 + }, + { + "epoch": 6.74, + "learning_rate": 1.631199475639065e-06, + "loss": 0.5299, + "step": 30888 + }, + { + "epoch": 6.74, + "learning_rate": 1.6310902337775836e-06, + "loss": 1.4959, + "step": 30889 + }, + { + "epoch": 6.74, + "learning_rate": 1.6309809919161025e-06, + "loss": 1.0698, + "step": 30890 + }, + { + "epoch": 6.74, + "learning_rate": 1.6308717500546213e-06, + "loss": 1.1704, + "step": 30891 + }, + { + "epoch": 6.74, + "learning_rate": 1.6307625081931397e-06, + "loss": 0.6157, + "step": 30892 + }, + { + "epoch": 6.74, + "learning_rate": 1.6306532663316585e-06, + "loss": 1.7569, + "step": 30893 + }, + { + "epoch": 6.74, + "learning_rate": 1.6305440244701771e-06, + "loss": 1.399, + "step": 30894 + }, + { + "epoch": 6.74, + "learning_rate": 1.6304347826086957e-06, + "loss": 0.3176, + "step": 30895 + }, + { + "epoch": 6.74, + "learning_rate": 1.6303255407472143e-06, + "loss": 0.9664, + "step": 30896 + }, + { + "epoch": 6.74, + "learning_rate": 1.6302162988857332e-06, + "loss": 1.9261, + "step": 30897 + }, + { + "epoch": 6.74, + "learning_rate": 1.630107057024252e-06, + "loss": 1.742, + "step": 30898 + }, + { + "epoch": 6.74, + "learning_rate": 1.6299978151627704e-06, + "loss": 1.4983, + "step": 30899 + }, + { + "epoch": 6.74, + "learning_rate": 1.6298885733012892e-06, + "loss": 1.9668, + "step": 30900 + }, + { + "epoch": 6.74, + "learning_rate": 1.6297793314398078e-06, + "loss": 1.6003, + "step": 30901 + }, + { + "epoch": 6.74, + "learning_rate": 1.6296700895783266e-06, + "loss": 1.1708, + "step": 30902 + }, + { + "epoch": 6.74, + "learning_rate": 1.6295608477168455e-06, + "loss": 2.9585, + "step": 30903 + }, + { + "epoch": 6.74, + "learning_rate": 1.6294516058553639e-06, + "loss": 0.8896, + "step": 30904 + }, + { + "epoch": 6.74, + "learning_rate": 1.6293423639938827e-06, + "loss": 1.0306, + "step": 30905 + }, + { + "epoch": 6.75, + "learning_rate": 1.6292331221324013e-06, + "loss": 2.3221, + "step": 30906 + }, + { + "epoch": 6.75, + "learning_rate": 1.62912388027092e-06, + "loss": 1.6626, + "step": 30907 + }, + { + "epoch": 6.75, + "learning_rate": 1.6290146384094385e-06, + "loss": 2.3595, + "step": 30908 + }, + { + "epoch": 6.75, + "learning_rate": 1.6289053965479573e-06, + "loss": 1.3247, + "step": 30909 + }, + { + "epoch": 6.75, + "learning_rate": 1.6287961546864762e-06, + "loss": 0.2863, + "step": 30910 + }, + { + "epoch": 6.75, + "learning_rate": 1.6286869128249946e-06, + "loss": 1.9591, + "step": 30911 + }, + { + "epoch": 6.75, + "learning_rate": 1.6285776709635134e-06, + "loss": 1.6001, + "step": 30912 + }, + { + "epoch": 6.75, + "learning_rate": 1.628468429102032e-06, + "loss": 0.8559, + "step": 30913 + }, + { + "epoch": 6.75, + "learning_rate": 1.6283591872405508e-06, + "loss": 1.1481, + "step": 30914 + }, + { + "epoch": 6.75, + "learning_rate": 1.6282499453790692e-06, + "loss": 1.4556, + "step": 30915 + }, + { + "epoch": 6.75, + "learning_rate": 1.628140703517588e-06, + "loss": 1.4211, + "step": 30916 + }, + { + "epoch": 6.75, + "learning_rate": 1.6280314616561069e-06, + "loss": 0.1926, + "step": 30917 + }, + { + "epoch": 6.75, + "learning_rate": 1.6279222197946255e-06, + "loss": 0.7154, + "step": 30918 + }, + { + "epoch": 6.75, + "learning_rate": 1.6278129779331443e-06, + "loss": 2.0882, + "step": 30919 + }, + { + "epoch": 6.75, + "learning_rate": 1.6277037360716627e-06, + "loss": 1.3223, + "step": 30920 + }, + { + "epoch": 6.75, + "learning_rate": 1.6275944942101815e-06, + "loss": 0.4405, + "step": 30921 + }, + { + "epoch": 6.75, + "learning_rate": 1.6274852523487001e-06, + "loss": 3.9315, + "step": 30922 + }, + { + "epoch": 6.75, + "learning_rate": 1.6273760104872187e-06, + "loss": 1.2139, + "step": 30923 + }, + { + "epoch": 6.75, + "learning_rate": 1.6272667686257376e-06, + "loss": 0.3668, + "step": 30924 + }, + { + "epoch": 6.75, + "learning_rate": 1.6271575267642562e-06, + "loss": 1.2422, + "step": 30925 + }, + { + "epoch": 6.75, + "learning_rate": 1.627048284902775e-06, + "loss": 1.6177, + "step": 30926 + }, + { + "epoch": 6.75, + "learning_rate": 1.6269390430412934e-06, + "loss": 0.8861, + "step": 30927 + }, + { + "epoch": 6.75, + "learning_rate": 1.6268298011798122e-06, + "loss": 0.7542, + "step": 30928 + }, + { + "epoch": 6.75, + "learning_rate": 1.6267205593183308e-06, + "loss": 1.7238, + "step": 30929 + }, + { + "epoch": 6.75, + "learning_rate": 1.6266113174568496e-06, + "loss": 1.5666, + "step": 30930 + }, + { + "epoch": 6.75, + "learning_rate": 1.6265020755953685e-06, + "loss": 1.3243, + "step": 30931 + }, + { + "epoch": 6.75, + "learning_rate": 1.6263928337338869e-06, + "loss": 1.602, + "step": 30932 + }, + { + "epoch": 6.75, + "learning_rate": 1.6262835918724057e-06, + "loss": 2.3442, + "step": 30933 + }, + { + "epoch": 6.75, + "learning_rate": 1.6261743500109243e-06, + "loss": 0.6771, + "step": 30934 + }, + { + "epoch": 6.75, + "learning_rate": 1.626065108149443e-06, + "loss": 1.547, + "step": 30935 + }, + { + "epoch": 6.75, + "learning_rate": 1.6259558662879617e-06, + "loss": 2.314, + "step": 30936 + }, + { + "epoch": 6.75, + "learning_rate": 1.6258466244264803e-06, + "loss": 3.626, + "step": 30937 + }, + { + "epoch": 6.75, + "learning_rate": 1.6257373825649992e-06, + "loss": 1.1064, + "step": 30938 + }, + { + "epoch": 6.75, + "learning_rate": 1.6256281407035176e-06, + "loss": 0.7576, + "step": 30939 + }, + { + "epoch": 6.75, + "learning_rate": 1.6255188988420364e-06, + "loss": 0.6254, + "step": 30940 + }, + { + "epoch": 6.75, + "learning_rate": 1.625409656980555e-06, + "loss": 0.8358, + "step": 30941 + }, + { + "epoch": 6.75, + "learning_rate": 1.6253004151190738e-06, + "loss": 1.11, + "step": 30942 + }, + { + "epoch": 6.75, + "learning_rate": 1.6251911732575926e-06, + "loss": 0.9422, + "step": 30943 + }, + { + "epoch": 6.75, + "learning_rate": 1.625081931396111e-06, + "loss": 1.4116, + "step": 30944 + }, + { + "epoch": 6.75, + "learning_rate": 1.6249726895346299e-06, + "loss": 0.4505, + "step": 30945 + }, + { + "epoch": 6.75, + "learning_rate": 1.6248634476731485e-06, + "loss": 0.7677, + "step": 30946 + }, + { + "epoch": 6.75, + "learning_rate": 1.6247542058116673e-06, + "loss": 1.4882, + "step": 30947 + }, + { + "epoch": 6.75, + "learning_rate": 1.6246449639501857e-06, + "loss": 2.4949, + "step": 30948 + }, + { + "epoch": 6.75, + "learning_rate": 1.6245357220887045e-06, + "loss": 0.7889, + "step": 30949 + }, + { + "epoch": 6.75, + "learning_rate": 1.6244264802272233e-06, + "loss": 1.6142, + "step": 30950 + }, + { + "epoch": 6.75, + "learning_rate": 1.6243172383657417e-06, + "loss": 0.8864, + "step": 30951 + }, + { + "epoch": 6.76, + "learning_rate": 1.6242079965042606e-06, + "loss": 0.6112, + "step": 30952 + }, + { + "epoch": 6.76, + "learning_rate": 1.6240987546427792e-06, + "loss": 0.4086, + "step": 30953 + }, + { + "epoch": 6.76, + "learning_rate": 1.623989512781298e-06, + "loss": 2.1172, + "step": 30954 + }, + { + "epoch": 6.76, + "learning_rate": 1.6238802709198164e-06, + "loss": 2.8511, + "step": 30955 + }, + { + "epoch": 6.76, + "learning_rate": 1.6237710290583352e-06, + "loss": 1.0177, + "step": 30956 + }, + { + "epoch": 6.76, + "learning_rate": 1.623661787196854e-06, + "loss": 1.1452, + "step": 30957 + }, + { + "epoch": 6.76, + "learning_rate": 1.6235525453353727e-06, + "loss": 0.5718, + "step": 30958 + }, + { + "epoch": 6.76, + "learning_rate": 1.6234433034738915e-06, + "loss": 0.5727, + "step": 30959 + }, + { + "epoch": 6.76, + "learning_rate": 1.6233340616124099e-06, + "loss": 1.4171, + "step": 30960 + }, + { + "epoch": 6.76, + "learning_rate": 1.6232248197509287e-06, + "loss": 0.8453, + "step": 30961 + }, + { + "epoch": 6.76, + "learning_rate": 1.6231155778894473e-06, + "loss": 0.6035, + "step": 30962 + }, + { + "epoch": 6.76, + "learning_rate": 1.6230063360279661e-06, + "loss": 1.4627, + "step": 30963 + }, + { + "epoch": 6.76, + "learning_rate": 1.6228970941664847e-06, + "loss": 0.1658, + "step": 30964 + }, + { + "epoch": 6.76, + "learning_rate": 1.6227878523050034e-06, + "loss": 0.7818, + "step": 30965 + }, + { + "epoch": 6.76, + "learning_rate": 1.6226786104435222e-06, + "loss": 0.2638, + "step": 30966 + }, + { + "epoch": 6.76, + "learning_rate": 1.6225693685820406e-06, + "loss": 2.3052, + "step": 30967 + }, + { + "epoch": 6.76, + "learning_rate": 1.6224601267205594e-06, + "loss": 1.119, + "step": 30968 + }, + { + "epoch": 6.76, + "learning_rate": 1.6223508848590782e-06, + "loss": 1.8088, + "step": 30969 + }, + { + "epoch": 6.76, + "learning_rate": 1.6222416429975968e-06, + "loss": 2.8824, + "step": 30970 + }, + { + "epoch": 6.76, + "learning_rate": 1.6221324011361157e-06, + "loss": 1.5229, + "step": 30971 + }, + { + "epoch": 6.76, + "learning_rate": 1.622023159274634e-06, + "loss": 4.2436, + "step": 30972 + }, + { + "epoch": 6.76, + "learning_rate": 1.6219139174131529e-06, + "loss": 0.6287, + "step": 30973 + }, + { + "epoch": 6.76, + "learning_rate": 1.6218046755516715e-06, + "loss": 0.5606, + "step": 30974 + }, + { + "epoch": 6.76, + "learning_rate": 1.6216954336901903e-06, + "loss": 1.2708, + "step": 30975 + }, + { + "epoch": 6.76, + "learning_rate": 1.621586191828709e-06, + "loss": 1.0847, + "step": 30976 + }, + { + "epoch": 6.76, + "learning_rate": 1.6214769499672275e-06, + "loss": 1.5523, + "step": 30977 + }, + { + "epoch": 6.76, + "learning_rate": 1.6213677081057464e-06, + "loss": 0.1792, + "step": 30978 + }, + { + "epoch": 6.76, + "learning_rate": 1.6212584662442648e-06, + "loss": 0.1764, + "step": 30979 + }, + { + "epoch": 6.76, + "learning_rate": 1.6211492243827836e-06, + "loss": 0.3695, + "step": 30980 + }, + { + "epoch": 6.76, + "learning_rate": 1.6210399825213022e-06, + "loss": 2.2186, + "step": 30981 + }, + { + "epoch": 6.76, + "learning_rate": 1.620930740659821e-06, + "loss": 1.9893, + "step": 30982 + }, + { + "epoch": 6.76, + "learning_rate": 1.6208214987983398e-06, + "loss": 0.4786, + "step": 30983 + }, + { + "epoch": 6.76, + "learning_rate": 1.6207122569368582e-06, + "loss": 2.2025, + "step": 30984 + }, + { + "epoch": 6.76, + "learning_rate": 1.620603015075377e-06, + "loss": 1.3189, + "step": 30985 + }, + { + "epoch": 6.76, + "learning_rate": 1.6204937732138957e-06, + "loss": 1.9629, + "step": 30986 + }, + { + "epoch": 6.76, + "learning_rate": 1.6203845313524145e-06, + "loss": 1.3771, + "step": 30987 + }, + { + "epoch": 6.76, + "learning_rate": 1.620275289490933e-06, + "loss": 1.2535, + "step": 30988 + }, + { + "epoch": 6.76, + "learning_rate": 1.6201660476294517e-06, + "loss": 1.6978, + "step": 30989 + }, + { + "epoch": 6.76, + "learning_rate": 1.6200568057679705e-06, + "loss": 0.0648, + "step": 30990 + }, + { + "epoch": 6.76, + "learning_rate": 1.6199475639064892e-06, + "loss": 1.2381, + "step": 30991 + }, + { + "epoch": 6.76, + "learning_rate": 1.6198383220450078e-06, + "loss": 1.6515, + "step": 30992 + }, + { + "epoch": 6.76, + "learning_rate": 1.6197290801835264e-06, + "loss": 0.5448, + "step": 30993 + }, + { + "epoch": 6.76, + "learning_rate": 1.6196198383220452e-06, + "loss": 1.0225, + "step": 30994 + }, + { + "epoch": 6.76, + "learning_rate": 1.6195105964605636e-06, + "loss": 0.8971, + "step": 30995 + }, + { + "epoch": 6.76, + "learning_rate": 1.6194013545990824e-06, + "loss": 1.5661, + "step": 30996 + }, + { + "epoch": 6.76, + "learning_rate": 1.6192921127376012e-06, + "loss": 2.2963, + "step": 30997 + }, + { + "epoch": 6.77, + "learning_rate": 1.6191828708761199e-06, + "loss": 1.1679, + "step": 30998 + }, + { + "epoch": 6.77, + "learning_rate": 1.6190736290146387e-06, + "loss": 1.2506, + "step": 30999 + }, + { + "epoch": 6.77, + "learning_rate": 1.618964387153157e-06, + "loss": 1.9783, + "step": 31000 + }, + { + "epoch": 6.77, + "learning_rate": 1.618855145291676e-06, + "loss": 2.1075, + "step": 31001 + }, + { + "epoch": 6.77, + "learning_rate": 1.6187459034301947e-06, + "loss": 0.9005, + "step": 31002 + }, + { + "epoch": 6.77, + "learning_rate": 1.6186366615687133e-06, + "loss": 1.6099, + "step": 31003 + }, + { + "epoch": 6.77, + "learning_rate": 1.618527419707232e-06, + "loss": 2.0752, + "step": 31004 + }, + { + "epoch": 6.77, + "learning_rate": 1.6184181778457506e-06, + "loss": 2.0081, + "step": 31005 + }, + { + "epoch": 6.77, + "learning_rate": 1.6183089359842694e-06, + "loss": 1.1459, + "step": 31006 + }, + { + "epoch": 6.77, + "learning_rate": 1.6181996941227878e-06, + "loss": 1.8148, + "step": 31007 + }, + { + "epoch": 6.77, + "learning_rate": 1.6180904522613066e-06, + "loss": 0.6239, + "step": 31008 + }, + { + "epoch": 6.77, + "learning_rate": 1.6179812103998254e-06, + "loss": 1.305, + "step": 31009 + }, + { + "epoch": 6.77, + "learning_rate": 1.617871968538344e-06, + "loss": 0.1603, + "step": 31010 + }, + { + "epoch": 6.77, + "learning_rate": 1.6177627266768629e-06, + "loss": 1.8388, + "step": 31011 + }, + { + "epoch": 6.77, + "learning_rate": 1.6176534848153813e-06, + "loss": 2.7913, + "step": 31012 + }, + { + "epoch": 6.77, + "learning_rate": 1.6175442429539e-06, + "loss": 2.6269, + "step": 31013 + }, + { + "epoch": 6.77, + "learning_rate": 1.6174350010924187e-06, + "loss": 1.5621, + "step": 31014 + }, + { + "epoch": 6.77, + "learning_rate": 1.6173257592309375e-06, + "loss": 1.0127, + "step": 31015 + }, + { + "epoch": 6.77, + "learning_rate": 1.6172165173694563e-06, + "loss": 1.8958, + "step": 31016 + }, + { + "epoch": 6.77, + "learning_rate": 1.6171072755079747e-06, + "loss": 1.9486, + "step": 31017 + }, + { + "epoch": 6.77, + "learning_rate": 1.6169980336464936e-06, + "loss": 1.1599, + "step": 31018 + }, + { + "epoch": 6.77, + "learning_rate": 1.6168887917850122e-06, + "loss": 2.3444, + "step": 31019 + }, + { + "epoch": 6.77, + "learning_rate": 1.6167795499235308e-06, + "loss": 2.5546, + "step": 31020 + }, + { + "epoch": 6.77, + "learning_rate": 1.6166703080620494e-06, + "loss": 1.2252, + "step": 31021 + }, + { + "epoch": 6.77, + "learning_rate": 1.6165610662005682e-06, + "loss": 1.1432, + "step": 31022 + }, + { + "epoch": 6.77, + "learning_rate": 1.616451824339087e-06, + "loss": 1.6663, + "step": 31023 + }, + { + "epoch": 6.77, + "learning_rate": 1.6163425824776054e-06, + "loss": 0.6292, + "step": 31024 + }, + { + "epoch": 6.77, + "learning_rate": 1.6162333406161243e-06, + "loss": 1.8436, + "step": 31025 + }, + { + "epoch": 6.77, + "learning_rate": 1.6161240987546429e-06, + "loss": 1.732, + "step": 31026 + }, + { + "epoch": 6.77, + "learning_rate": 1.6160148568931617e-06, + "loss": 2.5079, + "step": 31027 + }, + { + "epoch": 6.77, + "learning_rate": 1.61590561503168e-06, + "loss": 0.2267, + "step": 31028 + }, + { + "epoch": 6.77, + "learning_rate": 1.615796373170199e-06, + "loss": 2.1542, + "step": 31029 + }, + { + "epoch": 6.77, + "learning_rate": 1.6156871313087177e-06, + "loss": 2.2173, + "step": 31030 + }, + { + "epoch": 6.77, + "learning_rate": 1.6155778894472363e-06, + "loss": 0.2302, + "step": 31031 + }, + { + "epoch": 6.77, + "learning_rate": 1.615468647585755e-06, + "loss": 1.9653, + "step": 31032 + }, + { + "epoch": 6.77, + "learning_rate": 1.6153594057242736e-06, + "loss": 2.5018, + "step": 31033 + }, + { + "epoch": 6.77, + "learning_rate": 1.6152501638627924e-06, + "loss": 2.8941, + "step": 31034 + }, + { + "epoch": 6.77, + "learning_rate": 1.615140922001311e-06, + "loss": 2.4569, + "step": 31035 + }, + { + "epoch": 6.77, + "learning_rate": 1.6150316801398296e-06, + "loss": 1.8069, + "step": 31036 + }, + { + "epoch": 6.77, + "learning_rate": 1.6149224382783484e-06, + "loss": 2.257, + "step": 31037 + }, + { + "epoch": 6.77, + "learning_rate": 1.614813196416867e-06, + "loss": 0.0822, + "step": 31038 + }, + { + "epoch": 6.77, + "learning_rate": 1.6147039545553859e-06, + "loss": 1.1184, + "step": 31039 + }, + { + "epoch": 6.77, + "learning_rate": 1.6145947126939043e-06, + "loss": 1.484, + "step": 31040 + }, + { + "epoch": 6.77, + "learning_rate": 1.614485470832423e-06, + "loss": 1.0037, + "step": 31041 + }, + { + "epoch": 6.77, + "learning_rate": 1.614376228970942e-06, + "loss": 0.6203, + "step": 31042 + }, + { + "epoch": 6.77, + "learning_rate": 1.6142669871094605e-06, + "loss": 1.847, + "step": 31043 + }, + { + "epoch": 6.78, + "learning_rate": 1.6141577452479793e-06, + "loss": 1.8259, + "step": 31044 + }, + { + "epoch": 6.78, + "learning_rate": 1.6140485033864977e-06, + "loss": 0.4565, + "step": 31045 + }, + { + "epoch": 6.78, + "learning_rate": 1.6139392615250166e-06, + "loss": 1.9001, + "step": 31046 + }, + { + "epoch": 6.78, + "learning_rate": 1.6138300196635352e-06, + "loss": 0.689, + "step": 31047 + }, + { + "epoch": 6.78, + "learning_rate": 1.6137207778020538e-06, + "loss": 1.5223, + "step": 31048 + }, + { + "epoch": 6.78, + "learning_rate": 1.6136115359405726e-06, + "loss": 1.9418, + "step": 31049 + }, + { + "epoch": 6.78, + "learning_rate": 1.6135022940790912e-06, + "loss": 3.2989, + "step": 31050 + }, + { + "epoch": 6.78, + "learning_rate": 1.61339305221761e-06, + "loss": 2.5677, + "step": 31051 + }, + { + "epoch": 6.78, + "learning_rate": 1.6132838103561284e-06, + "loss": 0.7371, + "step": 31052 + }, + { + "epoch": 6.78, + "learning_rate": 1.6131745684946473e-06, + "loss": 1.0873, + "step": 31053 + }, + { + "epoch": 6.78, + "learning_rate": 1.6130653266331659e-06, + "loss": 1.6246, + "step": 31054 + }, + { + "epoch": 6.78, + "learning_rate": 1.6129560847716847e-06, + "loss": 1.336, + "step": 31055 + }, + { + "epoch": 6.78, + "learning_rate": 1.6128468429102035e-06, + "loss": 1.492, + "step": 31056 + }, + { + "epoch": 6.78, + "learning_rate": 1.612737601048722e-06, + "loss": 0.4093, + "step": 31057 + }, + { + "epoch": 6.78, + "learning_rate": 1.6126283591872407e-06, + "loss": 1.4748, + "step": 31058 + }, + { + "epoch": 6.78, + "learning_rate": 1.6125191173257594e-06, + "loss": 1.4727, + "step": 31059 + }, + { + "epoch": 6.78, + "learning_rate": 1.612409875464278e-06, + "loss": 1.5665, + "step": 31060 + }, + { + "epoch": 6.78, + "learning_rate": 1.6123006336027966e-06, + "loss": 0.781, + "step": 31061 + }, + { + "epoch": 6.78, + "learning_rate": 1.6121913917413154e-06, + "loss": 0.8846, + "step": 31062 + }, + { + "epoch": 6.78, + "learning_rate": 1.6120821498798342e-06, + "loss": 2.0033, + "step": 31063 + }, + { + "epoch": 6.78, + "learning_rate": 1.6119729080183526e-06, + "loss": 0.2697, + "step": 31064 + }, + { + "epoch": 6.78, + "learning_rate": 1.6118636661568714e-06, + "loss": 3.0133, + "step": 31065 + }, + { + "epoch": 6.78, + "learning_rate": 1.61175442429539e-06, + "loss": 0.8833, + "step": 31066 + }, + { + "epoch": 6.78, + "learning_rate": 1.6116451824339089e-06, + "loss": 0.7134, + "step": 31067 + }, + { + "epoch": 6.78, + "learning_rate": 1.6115359405724273e-06, + "loss": 2.1255, + "step": 31068 + }, + { + "epoch": 6.78, + "learning_rate": 1.611426698710946e-06, + "loss": 0.8408, + "step": 31069 + }, + { + "epoch": 6.78, + "learning_rate": 1.611317456849465e-06, + "loss": 1.1624, + "step": 31070 + }, + { + "epoch": 6.78, + "learning_rate": 1.6112082149879835e-06, + "loss": 1.4522, + "step": 31071 + }, + { + "epoch": 6.78, + "learning_rate": 1.6110989731265024e-06, + "loss": 3.8791, + "step": 31072 + }, + { + "epoch": 6.78, + "learning_rate": 1.6109897312650208e-06, + "loss": 2.6054, + "step": 31073 + }, + { + "epoch": 6.78, + "learning_rate": 1.6108804894035396e-06, + "loss": 0.242, + "step": 31074 + }, + { + "epoch": 6.78, + "learning_rate": 1.6107712475420584e-06, + "loss": 1.1295, + "step": 31075 + }, + { + "epoch": 6.78, + "learning_rate": 1.6106620056805768e-06, + "loss": 2.0906, + "step": 31076 + }, + { + "epoch": 6.78, + "learning_rate": 1.6105527638190956e-06, + "loss": 0.6794, + "step": 31077 + }, + { + "epoch": 6.78, + "learning_rate": 1.6104435219576142e-06, + "loss": 1.5072, + "step": 31078 + }, + { + "epoch": 6.78, + "learning_rate": 1.610334280096133e-06, + "loss": 1.5737, + "step": 31079 + }, + { + "epoch": 6.78, + "learning_rate": 1.6102250382346515e-06, + "loss": 1.147, + "step": 31080 + }, + { + "epoch": 6.78, + "learning_rate": 1.6101157963731703e-06, + "loss": 0.6305, + "step": 31081 + }, + { + "epoch": 6.78, + "learning_rate": 1.610006554511689e-06, + "loss": 2.4686, + "step": 31082 + }, + { + "epoch": 6.78, + "learning_rate": 1.6098973126502077e-06, + "loss": 1.7635, + "step": 31083 + }, + { + "epoch": 6.78, + "learning_rate": 1.6097880707887265e-06, + "loss": 1.6578, + "step": 31084 + }, + { + "epoch": 6.78, + "learning_rate": 1.609678828927245e-06, + "loss": 0.1675, + "step": 31085 + }, + { + "epoch": 6.78, + "learning_rate": 1.6095695870657638e-06, + "loss": 0.6908, + "step": 31086 + }, + { + "epoch": 6.78, + "learning_rate": 1.6094603452042824e-06, + "loss": 1.3284, + "step": 31087 + }, + { + "epoch": 6.78, + "learning_rate": 1.6093511033428012e-06, + "loss": 0.1621, + "step": 31088 + }, + { + "epoch": 6.79, + "learning_rate": 1.6092418614813198e-06, + "loss": 0.6004, + "step": 31089 + }, + { + "epoch": 6.79, + "learning_rate": 1.6091326196198384e-06, + "loss": 1.1075, + "step": 31090 + }, + { + "epoch": 6.79, + "learning_rate": 1.6090233777583572e-06, + "loss": 0.7635, + "step": 31091 + }, + { + "epoch": 6.79, + "learning_rate": 1.6089141358968756e-06, + "loss": 2.1785, + "step": 31092 + }, + { + "epoch": 6.79, + "learning_rate": 1.6088048940353945e-06, + "loss": 0.9449, + "step": 31093 + }, + { + "epoch": 6.79, + "learning_rate": 1.608695652173913e-06, + "loss": 2.0583, + "step": 31094 + }, + { + "epoch": 6.79, + "learning_rate": 1.608586410312432e-06, + "loss": 0.8759, + "step": 31095 + }, + { + "epoch": 6.79, + "learning_rate": 1.6084771684509507e-06, + "loss": 1.2931, + "step": 31096 + }, + { + "epoch": 6.79, + "learning_rate": 1.6083679265894691e-06, + "loss": 1.2854, + "step": 31097 + }, + { + "epoch": 6.79, + "learning_rate": 1.608258684727988e-06, + "loss": 1.9056, + "step": 31098 + }, + { + "epoch": 6.79, + "learning_rate": 1.6081494428665066e-06, + "loss": 3.0033, + "step": 31099 + }, + { + "epoch": 6.79, + "learning_rate": 1.6080402010050254e-06, + "loss": 0.632, + "step": 31100 + }, + { + "epoch": 6.79, + "learning_rate": 1.6079309591435438e-06, + "loss": 0.9555, + "step": 31101 + }, + { + "epoch": 6.79, + "learning_rate": 1.6078217172820626e-06, + "loss": 1.1059, + "step": 31102 + }, + { + "epoch": 6.79, + "learning_rate": 1.6077124754205814e-06, + "loss": 0.5967, + "step": 31103 + }, + { + "epoch": 6.79, + "learning_rate": 1.6076032335590998e-06, + "loss": 1.2634, + "step": 31104 + }, + { + "epoch": 6.79, + "learning_rate": 1.6074939916976186e-06, + "loss": 2.2724, + "step": 31105 + }, + { + "epoch": 6.79, + "learning_rate": 1.6073847498361373e-06, + "loss": 0.6326, + "step": 31106 + }, + { + "epoch": 6.79, + "learning_rate": 1.607275507974656e-06, + "loss": 1.2147, + "step": 31107 + }, + { + "epoch": 6.79, + "learning_rate": 1.607166266113175e-06, + "loss": 0.6458, + "step": 31108 + }, + { + "epoch": 6.79, + "learning_rate": 1.6070570242516933e-06, + "loss": 1.0021, + "step": 31109 + }, + { + "epoch": 6.79, + "learning_rate": 1.6069477823902121e-06, + "loss": 2.8096, + "step": 31110 + }, + { + "epoch": 6.79, + "learning_rate": 1.6068385405287307e-06, + "loss": 1.5434, + "step": 31111 + }, + { + "epoch": 6.79, + "learning_rate": 1.6067292986672496e-06, + "loss": 1.0648, + "step": 31112 + }, + { + "epoch": 6.79, + "learning_rate": 1.606620056805768e-06, + "loss": 1.2919, + "step": 31113 + }, + { + "epoch": 6.79, + "learning_rate": 1.6065108149442868e-06, + "loss": 1.4182, + "step": 31114 + }, + { + "epoch": 6.79, + "learning_rate": 1.6064015730828056e-06, + "loss": 1.3262, + "step": 31115 + }, + { + "epoch": 6.79, + "learning_rate": 1.6062923312213242e-06, + "loss": 0.8098, + "step": 31116 + }, + { + "epoch": 6.79, + "learning_rate": 1.6061830893598428e-06, + "loss": 1.6187, + "step": 31117 + }, + { + "epoch": 6.79, + "learning_rate": 1.6060738474983614e-06, + "loss": 0.8223, + "step": 31118 + }, + { + "epoch": 6.79, + "learning_rate": 1.6059646056368803e-06, + "loss": 0.5036, + "step": 31119 + }, + { + "epoch": 6.79, + "learning_rate": 1.6058553637753987e-06, + "loss": 1.152, + "step": 31120 + }, + { + "epoch": 6.79, + "learning_rate": 1.6057461219139175e-06, + "loss": 1.2837, + "step": 31121 + }, + { + "epoch": 6.79, + "learning_rate": 1.6056368800524363e-06, + "loss": 0.4736, + "step": 31122 + }, + { + "epoch": 6.79, + "learning_rate": 1.605527638190955e-06, + "loss": 0.3502, + "step": 31123 + }, + { + "epoch": 6.79, + "learning_rate": 1.6054183963294737e-06, + "loss": 1.6986, + "step": 31124 + }, + { + "epoch": 6.79, + "learning_rate": 1.6053091544679921e-06, + "loss": 0.9584, + "step": 31125 + }, + { + "epoch": 6.79, + "learning_rate": 1.605199912606511e-06, + "loss": 2.1094, + "step": 31126 + }, + { + "epoch": 6.79, + "learning_rate": 1.6050906707450296e-06, + "loss": 1.9309, + "step": 31127 + }, + { + "epoch": 6.79, + "learning_rate": 1.6049814288835484e-06, + "loss": 0.9444, + "step": 31128 + }, + { + "epoch": 6.79, + "learning_rate": 1.604872187022067e-06, + "loss": 1.4391, + "step": 31129 + }, + { + "epoch": 6.79, + "learning_rate": 1.6047629451605856e-06, + "loss": 1.2182, + "step": 31130 + }, + { + "epoch": 6.79, + "learning_rate": 1.6046537032991044e-06, + "loss": 0.8586, + "step": 31131 + }, + { + "epoch": 6.79, + "learning_rate": 1.6045444614376228e-06, + "loss": 2.0568, + "step": 31132 + }, + { + "epoch": 6.79, + "learning_rate": 1.6044352195761417e-06, + "loss": 1.8488, + "step": 31133 + }, + { + "epoch": 6.79, + "learning_rate": 1.6043259777146603e-06, + "loss": 1.0318, + "step": 31134 + }, + { + "epoch": 6.8, + "learning_rate": 1.604216735853179e-06, + "loss": 0.167, + "step": 31135 + }, + { + "epoch": 6.8, + "learning_rate": 1.604107493991698e-06, + "loss": 0.8039, + "step": 31136 + }, + { + "epoch": 6.8, + "learning_rate": 1.6039982521302163e-06, + "loss": 0.9985, + "step": 31137 + }, + { + "epoch": 6.8, + "learning_rate": 1.6038890102687351e-06, + "loss": 2.3615, + "step": 31138 + }, + { + "epoch": 6.8, + "learning_rate": 1.6037797684072537e-06, + "loss": 2.3684, + "step": 31139 + }, + { + "epoch": 6.8, + "learning_rate": 1.6036705265457726e-06, + "loss": 0.4149, + "step": 31140 + }, + { + "epoch": 6.8, + "learning_rate": 1.6035612846842914e-06, + "loss": 2.3125, + "step": 31141 + }, + { + "epoch": 6.8, + "learning_rate": 1.6034520428228098e-06, + "loss": 2.8057, + "step": 31142 + }, + { + "epoch": 6.8, + "learning_rate": 1.6033428009613286e-06, + "loss": 4.4376, + "step": 31143 + }, + { + "epoch": 6.8, + "learning_rate": 1.6032335590998472e-06, + "loss": 1.703, + "step": 31144 + }, + { + "epoch": 6.8, + "learning_rate": 1.6031243172383658e-06, + "loss": 1.6452, + "step": 31145 + }, + { + "epoch": 6.8, + "learning_rate": 1.6030150753768844e-06, + "loss": 1.1172, + "step": 31146 + }, + { + "epoch": 6.8, + "learning_rate": 1.6029058335154033e-06, + "loss": 2.4138, + "step": 31147 + }, + { + "epoch": 6.8, + "learning_rate": 1.602796591653922e-06, + "loss": 0.2756, + "step": 31148 + }, + { + "epoch": 6.8, + "learning_rate": 1.6026873497924405e-06, + "loss": 0.5682, + "step": 31149 + }, + { + "epoch": 6.8, + "learning_rate": 1.6025781079309593e-06, + "loss": 0.7517, + "step": 31150 + }, + { + "epoch": 6.8, + "learning_rate": 1.602468866069478e-06, + "loss": 1.3858, + "step": 31151 + }, + { + "epoch": 6.8, + "learning_rate": 1.6023596242079967e-06, + "loss": 1.8659, + "step": 31152 + }, + { + "epoch": 6.8, + "learning_rate": 1.6022503823465151e-06, + "loss": 0.1432, + "step": 31153 + }, + { + "epoch": 6.8, + "learning_rate": 1.602141140485034e-06, + "loss": 1.4645, + "step": 31154 + }, + { + "epoch": 6.8, + "learning_rate": 1.6020318986235528e-06, + "loss": 0.6441, + "step": 31155 + }, + { + "epoch": 6.8, + "learning_rate": 1.6019226567620714e-06, + "loss": 0.8866, + "step": 31156 + }, + { + "epoch": 6.8, + "learning_rate": 1.60181341490059e-06, + "loss": 1.9036, + "step": 31157 + }, + { + "epoch": 6.8, + "learning_rate": 1.6017041730391086e-06, + "loss": 0.7781, + "step": 31158 + }, + { + "epoch": 6.8, + "learning_rate": 1.6015949311776274e-06, + "loss": 0.7993, + "step": 31159 + }, + { + "epoch": 6.8, + "learning_rate": 1.601485689316146e-06, + "loss": 0.8966, + "step": 31160 + }, + { + "epoch": 6.8, + "learning_rate": 1.6013764474546647e-06, + "loss": 1.0743, + "step": 31161 + }, + { + "epoch": 6.8, + "learning_rate": 1.6012672055931835e-06, + "loss": 0.0987, + "step": 31162 + }, + { + "epoch": 6.8, + "learning_rate": 1.601157963731702e-06, + "loss": 1.8817, + "step": 31163 + }, + { + "epoch": 6.8, + "learning_rate": 1.601048721870221e-06, + "loss": 1.2303, + "step": 31164 + }, + { + "epoch": 6.8, + "learning_rate": 1.6009394800087393e-06, + "loss": 0.7046, + "step": 31165 + }, + { + "epoch": 6.8, + "learning_rate": 1.6008302381472581e-06, + "loss": 0.4336, + "step": 31166 + }, + { + "epoch": 6.8, + "learning_rate": 1.6007209962857768e-06, + "loss": 1.0386, + "step": 31167 + }, + { + "epoch": 6.8, + "learning_rate": 1.6006117544242956e-06, + "loss": 0.2107, + "step": 31168 + }, + { + "epoch": 6.8, + "learning_rate": 1.6005025125628144e-06, + "loss": 2.0763, + "step": 31169 + }, + { + "epoch": 6.8, + "learning_rate": 1.6003932707013328e-06, + "loss": 1.5481, + "step": 31170 + }, + { + "epoch": 6.8, + "learning_rate": 1.6002840288398516e-06, + "loss": 0.819, + "step": 31171 + }, + { + "epoch": 6.8, + "learning_rate": 1.6001747869783702e-06, + "loss": 0.2368, + "step": 31172 + }, + { + "epoch": 6.8, + "learning_rate": 1.6000655451168888e-06, + "loss": 1.0564, + "step": 31173 + }, + { + "epoch": 6.8, + "learning_rate": 1.5999563032554077e-06, + "loss": 0.6368, + "step": 31174 + }, + { + "epoch": 6.8, + "learning_rate": 1.5998470613939263e-06, + "loss": 0.5255, + "step": 31175 + }, + { + "epoch": 6.8, + "learning_rate": 1.599737819532445e-06, + "loss": 1.8803, + "step": 31176 + }, + { + "epoch": 6.8, + "learning_rate": 1.5996285776709635e-06, + "loss": 0.6508, + "step": 31177 + }, + { + "epoch": 6.8, + "learning_rate": 1.5995193358094823e-06, + "loss": 1.4721, + "step": 31178 + }, + { + "epoch": 6.8, + "learning_rate": 1.599410093948001e-06, + "loss": 1.5337, + "step": 31179 + }, + { + "epoch": 6.8, + "learning_rate": 1.5993008520865198e-06, + "loss": 1.0087, + "step": 31180 + }, + { + "epoch": 6.81, + "learning_rate": 1.5991916102250386e-06, + "loss": 1.3505, + "step": 31181 + }, + { + "epoch": 6.81, + "learning_rate": 1.599082368363557e-06, + "loss": 2.0296, + "step": 31182 + }, + { + "epoch": 6.81, + "learning_rate": 1.5989731265020758e-06, + "loss": 0.7826, + "step": 31183 + }, + { + "epoch": 6.81, + "learning_rate": 1.5988638846405944e-06, + "loss": 2.8408, + "step": 31184 + }, + { + "epoch": 6.81, + "learning_rate": 1.5987546427791132e-06, + "loss": 0.6327, + "step": 31185 + }, + { + "epoch": 6.81, + "learning_rate": 1.5986454009176316e-06, + "loss": 0.8526, + "step": 31186 + }, + { + "epoch": 6.81, + "learning_rate": 1.5985361590561505e-06, + "loss": 1.1467, + "step": 31187 + }, + { + "epoch": 6.81, + "learning_rate": 1.5984269171946693e-06, + "loss": 0.8274, + "step": 31188 + }, + { + "epoch": 6.81, + "learning_rate": 1.5983176753331877e-06, + "loss": 1.3226, + "step": 31189 + }, + { + "epoch": 6.81, + "learning_rate": 1.5982084334717065e-06, + "loss": 1.3681, + "step": 31190 + }, + { + "epoch": 6.81, + "learning_rate": 1.5980991916102251e-06, + "loss": 0.6145, + "step": 31191 + }, + { + "epoch": 6.81, + "learning_rate": 1.597989949748744e-06, + "loss": 0.771, + "step": 31192 + }, + { + "epoch": 6.81, + "learning_rate": 1.5978807078872623e-06, + "loss": 1.9043, + "step": 31193 + }, + { + "epoch": 6.81, + "learning_rate": 1.5977714660257812e-06, + "loss": 1.3915, + "step": 31194 + }, + { + "epoch": 6.81, + "learning_rate": 1.5976622241643e-06, + "loss": 0.8093, + "step": 31195 + }, + { + "epoch": 6.81, + "learning_rate": 1.5975529823028186e-06, + "loss": 0.951, + "step": 31196 + }, + { + "epoch": 6.81, + "learning_rate": 1.5974437404413374e-06, + "loss": 1.3083, + "step": 31197 + }, + { + "epoch": 6.81, + "learning_rate": 1.5973344985798558e-06, + "loss": 0.5491, + "step": 31198 + }, + { + "epoch": 6.81, + "learning_rate": 1.5972252567183746e-06, + "loss": 1.7993, + "step": 31199 + }, + { + "epoch": 6.81, + "learning_rate": 1.5971160148568932e-06, + "loss": 0.7784, + "step": 31200 + }, + { + "epoch": 6.81, + "learning_rate": 1.5970067729954119e-06, + "loss": 2.4147, + "step": 31201 + }, + { + "epoch": 6.81, + "learning_rate": 1.5968975311339307e-06, + "loss": 1.7364, + "step": 31202 + }, + { + "epoch": 6.81, + "learning_rate": 1.5967882892724493e-06, + "loss": 0.3713, + "step": 31203 + }, + { + "epoch": 6.81, + "learning_rate": 1.5966790474109681e-06, + "loss": 1.2515, + "step": 31204 + }, + { + "epoch": 6.81, + "learning_rate": 1.5965698055494865e-06, + "loss": 1.3635, + "step": 31205 + }, + { + "epoch": 6.81, + "learning_rate": 1.5964605636880053e-06, + "loss": 1.3672, + "step": 31206 + }, + { + "epoch": 6.81, + "learning_rate": 1.5963513218265242e-06, + "loss": 3.5141, + "step": 31207 + }, + { + "epoch": 6.81, + "learning_rate": 1.5962420799650428e-06, + "loss": 1.9553, + "step": 31208 + }, + { + "epoch": 6.81, + "learning_rate": 1.5961328381035616e-06, + "loss": 0.3404, + "step": 31209 + }, + { + "epoch": 6.81, + "learning_rate": 1.59602359624208e-06, + "loss": 0.9933, + "step": 31210 + }, + { + "epoch": 6.81, + "learning_rate": 1.5959143543805988e-06, + "loss": 0.2739, + "step": 31211 + }, + { + "epoch": 6.81, + "learning_rate": 1.5958051125191174e-06, + "loss": 0.5432, + "step": 31212 + }, + { + "epoch": 6.81, + "learning_rate": 1.5956958706576362e-06, + "loss": 1.3851, + "step": 31213 + }, + { + "epoch": 6.81, + "learning_rate": 1.5955866287961549e-06, + "loss": 2.1057, + "step": 31214 + }, + { + "epoch": 6.81, + "learning_rate": 1.5954773869346735e-06, + "loss": 1.5858, + "step": 31215 + }, + { + "epoch": 6.81, + "learning_rate": 1.5953681450731923e-06, + "loss": 1.5254, + "step": 31216 + }, + { + "epoch": 6.81, + "learning_rate": 1.5952589032117107e-06, + "loss": 1.938, + "step": 31217 + }, + { + "epoch": 6.81, + "learning_rate": 1.5951496613502295e-06, + "loss": 2.8236, + "step": 31218 + }, + { + "epoch": 6.81, + "learning_rate": 1.5950404194887481e-06, + "loss": 0.514, + "step": 31219 + }, + { + "epoch": 6.81, + "learning_rate": 1.594931177627267e-06, + "loss": 2.8311, + "step": 31220 + }, + { + "epoch": 6.81, + "learning_rate": 1.5948219357657858e-06, + "loss": 0.5789, + "step": 31221 + }, + { + "epoch": 6.81, + "learning_rate": 1.5947126939043042e-06, + "loss": 0.4733, + "step": 31222 + }, + { + "epoch": 6.81, + "learning_rate": 1.594603452042823e-06, + "loss": 0.7124, + "step": 31223 + }, + { + "epoch": 6.81, + "learning_rate": 1.5944942101813416e-06, + "loss": 1.0005, + "step": 31224 + }, + { + "epoch": 6.81, + "learning_rate": 1.5943849683198604e-06, + "loss": 1.25, + "step": 31225 + }, + { + "epoch": 6.81, + "learning_rate": 1.5942757264583788e-06, + "loss": 0.369, + "step": 31226 + }, + { + "epoch": 6.82, + "learning_rate": 1.5941664845968977e-06, + "loss": 0.3875, + "step": 31227 + }, + { + "epoch": 6.82, + "learning_rate": 1.5940572427354165e-06, + "loss": 1.8972, + "step": 31228 + }, + { + "epoch": 6.82, + "learning_rate": 1.5939480008739349e-06, + "loss": 2.9179, + "step": 31229 + }, + { + "epoch": 6.82, + "learning_rate": 1.5938387590124537e-06, + "loss": 1.1834, + "step": 31230 + }, + { + "epoch": 6.82, + "learning_rate": 1.5937295171509723e-06, + "loss": 1.2776, + "step": 31231 + }, + { + "epoch": 6.82, + "learning_rate": 1.5936202752894911e-06, + "loss": 1.6635, + "step": 31232 + }, + { + "epoch": 6.82, + "learning_rate": 1.5935110334280095e-06, + "loss": 1.6464, + "step": 31233 + }, + { + "epoch": 6.82, + "learning_rate": 1.5934017915665284e-06, + "loss": 2.4791, + "step": 31234 + }, + { + "epoch": 6.82, + "learning_rate": 1.5932925497050472e-06, + "loss": 1.6146, + "step": 31235 + }, + { + "epoch": 6.82, + "learning_rate": 1.5931833078435658e-06, + "loss": 0.6674, + "step": 31236 + }, + { + "epoch": 6.82, + "learning_rate": 1.5930740659820846e-06, + "loss": 0.3502, + "step": 31237 + }, + { + "epoch": 6.82, + "learning_rate": 1.592964824120603e-06, + "loss": 0.4531, + "step": 31238 + }, + { + "epoch": 6.82, + "learning_rate": 1.5928555822591218e-06, + "loss": 1.7322, + "step": 31239 + }, + { + "epoch": 6.82, + "learning_rate": 1.5927463403976407e-06, + "loss": 1.0979, + "step": 31240 + }, + { + "epoch": 6.82, + "learning_rate": 1.5926370985361593e-06, + "loss": 0.9036, + "step": 31241 + }, + { + "epoch": 6.82, + "learning_rate": 1.5925278566746779e-06, + "loss": 2.7706, + "step": 31242 + }, + { + "epoch": 6.82, + "learning_rate": 1.5924186148131965e-06, + "loss": 1.2263, + "step": 31243 + }, + { + "epoch": 6.82, + "learning_rate": 1.5923093729517153e-06, + "loss": 1.2666, + "step": 31244 + }, + { + "epoch": 6.82, + "learning_rate": 1.5922001310902337e-06, + "loss": 1.2153, + "step": 31245 + }, + { + "epoch": 6.82, + "learning_rate": 1.5920908892287525e-06, + "loss": 1.5554, + "step": 31246 + }, + { + "epoch": 6.82, + "learning_rate": 1.5919816473672714e-06, + "loss": 1.6149, + "step": 31247 + }, + { + "epoch": 6.82, + "learning_rate": 1.59187240550579e-06, + "loss": 0.822, + "step": 31248 + }, + { + "epoch": 6.82, + "learning_rate": 1.5917631636443088e-06, + "loss": 2.2781, + "step": 31249 + }, + { + "epoch": 6.82, + "learning_rate": 1.5916539217828272e-06, + "loss": 1.3178, + "step": 31250 + }, + { + "epoch": 6.82, + "learning_rate": 1.591544679921346e-06, + "loss": 0.5009, + "step": 31251 + }, + { + "epoch": 6.82, + "learning_rate": 1.5914354380598646e-06, + "loss": 1.6096, + "step": 31252 + }, + { + "epoch": 6.82, + "learning_rate": 1.5913261961983834e-06, + "loss": 1.5627, + "step": 31253 + }, + { + "epoch": 6.82, + "learning_rate": 1.591216954336902e-06, + "loss": 1.7111, + "step": 31254 + }, + { + "epoch": 6.82, + "learning_rate": 1.5911077124754207e-06, + "loss": 0.8512, + "step": 31255 + }, + { + "epoch": 6.82, + "learning_rate": 1.5909984706139395e-06, + "loss": 0.2967, + "step": 31256 + }, + { + "epoch": 6.82, + "learning_rate": 1.590889228752458e-06, + "loss": 2.2175, + "step": 31257 + }, + { + "epoch": 6.82, + "learning_rate": 1.5907799868909767e-06, + "loss": 0.9722, + "step": 31258 + }, + { + "epoch": 6.82, + "learning_rate": 1.5906707450294953e-06, + "loss": 1.2065, + "step": 31259 + }, + { + "epoch": 6.82, + "learning_rate": 1.5905615031680141e-06, + "loss": 2.435, + "step": 31260 + }, + { + "epoch": 6.82, + "learning_rate": 1.590452261306533e-06, + "loss": 1.3132, + "step": 31261 + }, + { + "epoch": 6.82, + "learning_rate": 1.5903430194450514e-06, + "loss": 2.0045, + "step": 31262 + }, + { + "epoch": 6.82, + "learning_rate": 1.5902337775835702e-06, + "loss": 2.299, + "step": 31263 + }, + { + "epoch": 6.82, + "learning_rate": 1.5901245357220888e-06, + "loss": 0.5794, + "step": 31264 + }, + { + "epoch": 6.82, + "learning_rate": 1.5900152938606076e-06, + "loss": 1.265, + "step": 31265 + }, + { + "epoch": 6.82, + "learning_rate": 1.589906051999126e-06, + "loss": 0.6735, + "step": 31266 + }, + { + "epoch": 6.82, + "learning_rate": 1.5897968101376448e-06, + "loss": 0.937, + "step": 31267 + }, + { + "epoch": 6.82, + "learning_rate": 1.5896875682761637e-06, + "loss": 2.3703, + "step": 31268 + }, + { + "epoch": 6.82, + "learning_rate": 1.5895783264146823e-06, + "loss": 0.5775, + "step": 31269 + }, + { + "epoch": 6.82, + "learning_rate": 1.5894690845532009e-06, + "loss": 1.5118, + "step": 31270 + }, + { + "epoch": 6.82, + "learning_rate": 1.5893598426917195e-06, + "loss": 0.5241, + "step": 31271 + }, + { + "epoch": 6.82, + "learning_rate": 1.5892506008302383e-06, + "loss": 2.8666, + "step": 31272 + }, + { + "epoch": 6.83, + "learning_rate": 1.5891413589687567e-06, + "loss": 0.7439, + "step": 31273 + }, + { + "epoch": 6.83, + "learning_rate": 1.5890321171072755e-06, + "loss": 1.6691, + "step": 31274 + }, + { + "epoch": 6.83, + "learning_rate": 1.5889228752457944e-06, + "loss": 1.3534, + "step": 31275 + }, + { + "epoch": 6.83, + "learning_rate": 1.588813633384313e-06, + "loss": 1.4684, + "step": 31276 + }, + { + "epoch": 6.83, + "learning_rate": 1.5887043915228318e-06, + "loss": 0.7053, + "step": 31277 + }, + { + "epoch": 6.83, + "learning_rate": 1.5885951496613502e-06, + "loss": 0.4806, + "step": 31278 + }, + { + "epoch": 6.83, + "learning_rate": 1.588485907799869e-06, + "loss": 2.4197, + "step": 31279 + }, + { + "epoch": 6.83, + "learning_rate": 1.5883766659383878e-06, + "loss": 0.178, + "step": 31280 + }, + { + "epoch": 6.83, + "learning_rate": 1.5882674240769065e-06, + "loss": 1.687, + "step": 31281 + }, + { + "epoch": 6.83, + "learning_rate": 1.588158182215425e-06, + "loss": 1.122, + "step": 31282 + }, + { + "epoch": 6.83, + "learning_rate": 1.5880489403539437e-06, + "loss": 1.6724, + "step": 31283 + }, + { + "epoch": 6.83, + "learning_rate": 1.5879396984924625e-06, + "loss": 2.9093, + "step": 31284 + }, + { + "epoch": 6.83, + "learning_rate": 1.5878304566309811e-06, + "loss": 1.1251, + "step": 31285 + }, + { + "epoch": 6.83, + "learning_rate": 1.5877212147694997e-06, + "loss": 0.6539, + "step": 31286 + }, + { + "epoch": 6.83, + "learning_rate": 1.5876119729080185e-06, + "loss": 1.1944, + "step": 31287 + }, + { + "epoch": 6.83, + "learning_rate": 1.5875027310465372e-06, + "loss": 1.3083, + "step": 31288 + }, + { + "epoch": 6.83, + "learning_rate": 1.587393489185056e-06, + "loss": 0.8814, + "step": 31289 + }, + { + "epoch": 6.83, + "learning_rate": 1.5872842473235744e-06, + "loss": 1.4014, + "step": 31290 + }, + { + "epoch": 6.83, + "learning_rate": 1.5871750054620932e-06, + "loss": 1.3618, + "step": 31291 + }, + { + "epoch": 6.83, + "learning_rate": 1.5870657636006118e-06, + "loss": 1.338, + "step": 31292 + }, + { + "epoch": 6.83, + "learning_rate": 1.5869565217391306e-06, + "loss": 2.5817, + "step": 31293 + }, + { + "epoch": 6.83, + "learning_rate": 1.5868472798776495e-06, + "loss": 2.674, + "step": 31294 + }, + { + "epoch": 6.83, + "learning_rate": 1.5867380380161679e-06, + "loss": 0.7216, + "step": 31295 + }, + { + "epoch": 6.83, + "learning_rate": 1.5866287961546867e-06, + "loss": 0.7865, + "step": 31296 + }, + { + "epoch": 6.83, + "learning_rate": 1.5865195542932053e-06, + "loss": 1.4115, + "step": 31297 + }, + { + "epoch": 6.83, + "learning_rate": 1.586410312431724e-06, + "loss": 1.5299, + "step": 31298 + }, + { + "epoch": 6.83, + "learning_rate": 1.5863010705702425e-06, + "loss": 1.1895, + "step": 31299 + }, + { + "epoch": 6.83, + "learning_rate": 1.5861918287087613e-06, + "loss": 1.5795, + "step": 31300 + }, + { + "epoch": 6.83, + "learning_rate": 1.5860825868472802e-06, + "loss": 0.3258, + "step": 31301 + }, + { + "epoch": 6.83, + "learning_rate": 1.5859733449857986e-06, + "loss": 0.7326, + "step": 31302 + }, + { + "epoch": 6.83, + "learning_rate": 1.5858641031243174e-06, + "loss": 0.8891, + "step": 31303 + }, + { + "epoch": 6.83, + "learning_rate": 1.585754861262836e-06, + "loss": 0.5134, + "step": 31304 + }, + { + "epoch": 6.83, + "learning_rate": 1.5856456194013548e-06, + "loss": 2.2911, + "step": 31305 + }, + { + "epoch": 6.83, + "learning_rate": 1.5855363775398732e-06, + "loss": 1.4442, + "step": 31306 + }, + { + "epoch": 6.83, + "learning_rate": 1.585427135678392e-06, + "loss": 1.4837, + "step": 31307 + }, + { + "epoch": 6.83, + "learning_rate": 1.5853178938169109e-06, + "loss": 0.3155, + "step": 31308 + }, + { + "epoch": 6.83, + "learning_rate": 1.5852086519554295e-06, + "loss": 1.1377, + "step": 31309 + }, + { + "epoch": 6.83, + "learning_rate": 1.5850994100939483e-06, + "loss": 2.5238, + "step": 31310 + }, + { + "epoch": 6.83, + "learning_rate": 1.5849901682324667e-06, + "loss": 0.0961, + "step": 31311 + }, + { + "epoch": 6.83, + "learning_rate": 1.5848809263709855e-06, + "loss": 1.4064, + "step": 31312 + }, + { + "epoch": 6.83, + "learning_rate": 1.5847716845095043e-06, + "loss": 4.0927, + "step": 31313 + }, + { + "epoch": 6.83, + "learning_rate": 1.5846624426480227e-06, + "loss": 1.3873, + "step": 31314 + }, + { + "epoch": 6.83, + "learning_rate": 1.5845532007865416e-06, + "loss": 0.6718, + "step": 31315 + }, + { + "epoch": 6.83, + "learning_rate": 1.5844439589250602e-06, + "loss": 0.6638, + "step": 31316 + }, + { + "epoch": 6.83, + "learning_rate": 1.584334717063579e-06, + "loss": 1.6799, + "step": 31317 + }, + { + "epoch": 6.84, + "learning_rate": 1.5842254752020974e-06, + "loss": 2.0504, + "step": 31318 + }, + { + "epoch": 6.84, + "learning_rate": 1.5841162333406162e-06, + "loss": 1.7605, + "step": 31319 + }, + { + "epoch": 6.84, + "learning_rate": 1.584006991479135e-06, + "loss": 2.5866, + "step": 31320 + }, + { + "epoch": 6.84, + "learning_rate": 1.5838977496176536e-06, + "loss": 0.0557, + "step": 31321 + }, + { + "epoch": 6.84, + "learning_rate": 1.5837885077561725e-06, + "loss": 2.7373, + "step": 31322 + }, + { + "epoch": 6.84, + "learning_rate": 1.5836792658946909e-06, + "loss": 0.3145, + "step": 31323 + }, + { + "epoch": 6.84, + "learning_rate": 1.5835700240332097e-06, + "loss": 1.3811, + "step": 31324 + }, + { + "epoch": 6.84, + "learning_rate": 1.5834607821717283e-06, + "loss": 1.518, + "step": 31325 + }, + { + "epoch": 6.84, + "learning_rate": 1.583351540310247e-06, + "loss": 1.2232, + "step": 31326 + }, + { + "epoch": 6.84, + "learning_rate": 1.5832422984487657e-06, + "loss": 1.4102, + "step": 31327 + }, + { + "epoch": 6.84, + "learning_rate": 1.5831330565872843e-06, + "loss": 0.8116, + "step": 31328 + }, + { + "epoch": 6.84, + "learning_rate": 1.5830238147258032e-06, + "loss": 0.1773, + "step": 31329 + }, + { + "epoch": 6.84, + "learning_rate": 1.5829145728643216e-06, + "loss": 1.5392, + "step": 31330 + }, + { + "epoch": 6.84, + "learning_rate": 1.5828053310028404e-06, + "loss": 0.4456, + "step": 31331 + }, + { + "epoch": 6.84, + "learning_rate": 1.582696089141359e-06, + "loss": 0.3514, + "step": 31332 + }, + { + "epoch": 6.84, + "learning_rate": 1.5825868472798778e-06, + "loss": 1.0273, + "step": 31333 + }, + { + "epoch": 6.84, + "learning_rate": 1.5824776054183966e-06, + "loss": 0.3208, + "step": 31334 + }, + { + "epoch": 6.84, + "learning_rate": 1.582368363556915e-06, + "loss": 0.4579, + "step": 31335 + }, + { + "epoch": 6.84, + "learning_rate": 1.5822591216954339e-06, + "loss": 0.0705, + "step": 31336 + }, + { + "epoch": 6.84, + "learning_rate": 1.5821498798339525e-06, + "loss": 0.7733, + "step": 31337 + }, + { + "epoch": 6.84, + "learning_rate": 1.5820406379724713e-06, + "loss": 0.4925, + "step": 31338 + }, + { + "epoch": 6.84, + "learning_rate": 1.5819313961109897e-06, + "loss": 2.0034, + "step": 31339 + }, + { + "epoch": 6.84, + "learning_rate": 1.5818221542495085e-06, + "loss": 1.2427, + "step": 31340 + }, + { + "epoch": 6.84, + "learning_rate": 1.5817129123880274e-06, + "loss": 1.3437, + "step": 31341 + }, + { + "epoch": 6.84, + "learning_rate": 1.5816036705265457e-06, + "loss": 1.6272, + "step": 31342 + }, + { + "epoch": 6.84, + "learning_rate": 1.5814944286650646e-06, + "loss": 0.5201, + "step": 31343 + }, + { + "epoch": 6.84, + "learning_rate": 1.5813851868035832e-06, + "loss": 2.3762, + "step": 31344 + }, + { + "epoch": 6.84, + "learning_rate": 1.581275944942102e-06, + "loss": 1.6145, + "step": 31345 + }, + { + "epoch": 6.84, + "learning_rate": 1.5811667030806208e-06, + "loss": 0.4908, + "step": 31346 + }, + { + "epoch": 6.84, + "learning_rate": 1.5810574612191392e-06, + "loss": 1.5093, + "step": 31347 + }, + { + "epoch": 6.84, + "learning_rate": 1.580948219357658e-06, + "loss": 1.9014, + "step": 31348 + }, + { + "epoch": 6.84, + "learning_rate": 1.5808389774961767e-06, + "loss": 0.7959, + "step": 31349 + }, + { + "epoch": 6.84, + "learning_rate": 1.5807297356346955e-06, + "loss": 0.8369, + "step": 31350 + }, + { + "epoch": 6.84, + "learning_rate": 1.5806204937732139e-06, + "loss": 0.7155, + "step": 31351 + }, + { + "epoch": 6.84, + "learning_rate": 1.5805112519117327e-06, + "loss": 1.0575, + "step": 31352 + }, + { + "epoch": 6.84, + "learning_rate": 1.5804020100502515e-06, + "loss": 0.2916, + "step": 31353 + }, + { + "epoch": 6.84, + "learning_rate": 1.58029276818877e-06, + "loss": 2.1267, + "step": 31354 + }, + { + "epoch": 6.84, + "learning_rate": 1.5801835263272888e-06, + "loss": 0.4421, + "step": 31355 + }, + { + "epoch": 6.84, + "learning_rate": 1.5800742844658074e-06, + "loss": 1.5511, + "step": 31356 + }, + { + "epoch": 6.84, + "learning_rate": 1.5799650426043262e-06, + "loss": 1.8184, + "step": 31357 + }, + { + "epoch": 6.84, + "learning_rate": 1.5798558007428446e-06, + "loss": 2.4704, + "step": 31358 + }, + { + "epoch": 6.84, + "learning_rate": 1.5797465588813634e-06, + "loss": 1.161, + "step": 31359 + }, + { + "epoch": 6.84, + "learning_rate": 1.5796373170198822e-06, + "loss": 1.1562, + "step": 31360 + }, + { + "epoch": 6.84, + "learning_rate": 1.5795280751584008e-06, + "loss": 2.1586, + "step": 31361 + }, + { + "epoch": 6.84, + "learning_rate": 1.5794188332969197e-06, + "loss": 1.4967, + "step": 31362 + }, + { + "epoch": 6.84, + "learning_rate": 1.579309591435438e-06, + "loss": 0.8462, + "step": 31363 + }, + { + "epoch": 6.85, + "learning_rate": 1.5792003495739569e-06, + "loss": 1.001, + "step": 31364 + }, + { + "epoch": 6.85, + "learning_rate": 1.5790911077124755e-06, + "loss": 1.4323, + "step": 31365 + }, + { + "epoch": 6.85, + "learning_rate": 1.5789818658509943e-06, + "loss": 1.3696, + "step": 31366 + }, + { + "epoch": 6.85, + "learning_rate": 1.578872623989513e-06, + "loss": 1.5131, + "step": 31367 + }, + { + "epoch": 6.85, + "learning_rate": 1.5787633821280315e-06, + "loss": 1.0414, + "step": 31368 + }, + { + "epoch": 6.85, + "learning_rate": 1.5786541402665504e-06, + "loss": 3.3207, + "step": 31369 + }, + { + "epoch": 6.85, + "learning_rate": 1.5785448984050688e-06, + "loss": 3.1661, + "step": 31370 + }, + { + "epoch": 6.85, + "learning_rate": 1.5784356565435876e-06, + "loss": 0.6953, + "step": 31371 + }, + { + "epoch": 6.85, + "learning_rate": 1.5783264146821062e-06, + "loss": 1.3487, + "step": 31372 + }, + { + "epoch": 6.85, + "learning_rate": 1.578217172820625e-06, + "loss": 1.2578, + "step": 31373 + }, + { + "epoch": 6.85, + "learning_rate": 1.5781079309591438e-06, + "loss": 2.5847, + "step": 31374 + }, + { + "epoch": 6.85, + "learning_rate": 1.5779986890976622e-06, + "loss": 2.0918, + "step": 31375 + }, + { + "epoch": 6.85, + "learning_rate": 1.577889447236181e-06, + "loss": 0.9851, + "step": 31376 + }, + { + "epoch": 6.85, + "learning_rate": 1.5777802053746997e-06, + "loss": 0.5834, + "step": 31377 + }, + { + "epoch": 6.85, + "learning_rate": 1.5776709635132185e-06, + "loss": 1.4203, + "step": 31378 + }, + { + "epoch": 6.85, + "learning_rate": 1.5775617216517371e-06, + "loss": 3.0363, + "step": 31379 + }, + { + "epoch": 6.85, + "learning_rate": 1.5774524797902557e-06, + "loss": 1.1111, + "step": 31380 + }, + { + "epoch": 6.85, + "learning_rate": 1.5773432379287745e-06, + "loss": 1.4995, + "step": 31381 + }, + { + "epoch": 6.85, + "learning_rate": 1.5772339960672932e-06, + "loss": 2.5377, + "step": 31382 + }, + { + "epoch": 6.85, + "learning_rate": 1.5771247542058118e-06, + "loss": 0.5893, + "step": 31383 + }, + { + "epoch": 6.85, + "learning_rate": 1.5770155123443304e-06, + "loss": 0.6701, + "step": 31384 + }, + { + "epoch": 6.85, + "learning_rate": 1.5769062704828492e-06, + "loss": 0.6255, + "step": 31385 + }, + { + "epoch": 6.85, + "learning_rate": 1.576797028621368e-06, + "loss": 0.1021, + "step": 31386 + }, + { + "epoch": 6.85, + "learning_rate": 1.5766877867598864e-06, + "loss": 0.2545, + "step": 31387 + }, + { + "epoch": 6.85, + "learning_rate": 1.5765785448984052e-06, + "loss": 2.454, + "step": 31388 + }, + { + "epoch": 6.85, + "learning_rate": 1.5764693030369239e-06, + "loss": 0.4111, + "step": 31389 + }, + { + "epoch": 6.85, + "learning_rate": 1.5763600611754427e-06, + "loss": 0.5869, + "step": 31390 + }, + { + "epoch": 6.85, + "learning_rate": 1.576250819313961e-06, + "loss": 1.5545, + "step": 31391 + }, + { + "epoch": 6.85, + "learning_rate": 1.57614157745248e-06, + "loss": 0.1545, + "step": 31392 + }, + { + "epoch": 6.85, + "learning_rate": 1.5760323355909987e-06, + "loss": 2.5726, + "step": 31393 + }, + { + "epoch": 6.85, + "learning_rate": 1.5759230937295173e-06, + "loss": 0.9516, + "step": 31394 + }, + { + "epoch": 6.85, + "learning_rate": 1.575813851868036e-06, + "loss": 0.8947, + "step": 31395 + }, + { + "epoch": 6.85, + "learning_rate": 1.5757046100065546e-06, + "loss": 2.3179, + "step": 31396 + }, + { + "epoch": 6.85, + "learning_rate": 1.5755953681450734e-06, + "loss": 2.0792, + "step": 31397 + }, + { + "epoch": 6.85, + "learning_rate": 1.5754861262835918e-06, + "loss": 1.9877, + "step": 31398 + }, + { + "epoch": 6.85, + "learning_rate": 1.5753768844221106e-06, + "loss": 1.1212, + "step": 31399 + }, + { + "epoch": 6.85, + "learning_rate": 1.5752676425606294e-06, + "loss": 1.7548, + "step": 31400 + }, + { + "epoch": 6.85, + "learning_rate": 1.575158400699148e-06, + "loss": 1.5084, + "step": 31401 + }, + { + "epoch": 6.85, + "learning_rate": 1.5750491588376669e-06, + "loss": 0.8376, + "step": 31402 + }, + { + "epoch": 6.85, + "learning_rate": 1.5749399169761853e-06, + "loss": 1.6431, + "step": 31403 + }, + { + "epoch": 6.85, + "learning_rate": 1.574830675114704e-06, + "loss": 0.4703, + "step": 31404 + }, + { + "epoch": 6.85, + "learning_rate": 1.5747214332532227e-06, + "loss": 1.8345, + "step": 31405 + }, + { + "epoch": 6.85, + "learning_rate": 1.5746121913917415e-06, + "loss": 0.8835, + "step": 31406 + }, + { + "epoch": 6.85, + "learning_rate": 1.5745029495302601e-06, + "loss": 0.9383, + "step": 31407 + }, + { + "epoch": 6.85, + "learning_rate": 1.5743937076687787e-06, + "loss": 2.2089, + "step": 31408 + }, + { + "epoch": 6.85, + "learning_rate": 1.5742844658072976e-06, + "loss": 1.3098, + "step": 31409 + }, + { + "epoch": 6.86, + "learning_rate": 1.5741752239458162e-06, + "loss": 1.4957, + "step": 31410 + }, + { + "epoch": 6.86, + "learning_rate": 1.5740659820843348e-06, + "loss": 0.5754, + "step": 31411 + }, + { + "epoch": 6.86, + "learning_rate": 1.5739567402228536e-06, + "loss": 1.8355, + "step": 31412 + }, + { + "epoch": 6.86, + "learning_rate": 1.5738474983613722e-06, + "loss": 2.6538, + "step": 31413 + }, + { + "epoch": 6.86, + "learning_rate": 1.573738256499891e-06, + "loss": 0.556, + "step": 31414 + }, + { + "epoch": 6.86, + "learning_rate": 1.5736290146384094e-06, + "loss": 2.5614, + "step": 31415 + }, + { + "epoch": 6.86, + "learning_rate": 1.5735197727769283e-06, + "loss": 3.0252, + "step": 31416 + }, + { + "epoch": 6.86, + "learning_rate": 1.5734105309154469e-06, + "loss": 1.2613, + "step": 31417 + }, + { + "epoch": 6.86, + "learning_rate": 1.5733012890539657e-06, + "loss": 1.9058, + "step": 31418 + }, + { + "epoch": 6.86, + "learning_rate": 1.5731920471924845e-06, + "loss": 1.891, + "step": 31419 + }, + { + "epoch": 6.86, + "learning_rate": 1.573082805331003e-06, + "loss": 0.3659, + "step": 31420 + }, + { + "epoch": 6.86, + "learning_rate": 1.5729735634695217e-06, + "loss": 0.064, + "step": 31421 + }, + { + "epoch": 6.86, + "learning_rate": 1.5728643216080403e-06, + "loss": 1.3602, + "step": 31422 + }, + { + "epoch": 6.86, + "learning_rate": 1.572755079746559e-06, + "loss": 2.2497, + "step": 31423 + }, + { + "epoch": 6.86, + "learning_rate": 1.5726458378850776e-06, + "loss": 1.278, + "step": 31424 + }, + { + "epoch": 6.86, + "learning_rate": 1.5725365960235964e-06, + "loss": 0.2207, + "step": 31425 + }, + { + "epoch": 6.86, + "learning_rate": 1.5724273541621152e-06, + "loss": 1.2409, + "step": 31426 + }, + { + "epoch": 6.86, + "learning_rate": 1.5723181123006336e-06, + "loss": 0.9466, + "step": 31427 + }, + { + "epoch": 6.86, + "learning_rate": 1.5722088704391524e-06, + "loss": 0.9293, + "step": 31428 + }, + { + "epoch": 6.86, + "learning_rate": 1.572099628577671e-06, + "loss": 0.3992, + "step": 31429 + }, + { + "epoch": 6.86, + "learning_rate": 1.5719903867161899e-06, + "loss": 0.9536, + "step": 31430 + }, + { + "epoch": 6.86, + "learning_rate": 1.5718811448547083e-06, + "loss": 2.3232, + "step": 31431 + }, + { + "epoch": 6.86, + "learning_rate": 1.571771902993227e-06, + "loss": 0.8125, + "step": 31432 + }, + { + "epoch": 6.86, + "learning_rate": 1.571662661131746e-06, + "loss": 0.7028, + "step": 31433 + }, + { + "epoch": 6.86, + "learning_rate": 1.5715534192702645e-06, + "loss": 0.519, + "step": 31434 + }, + { + "epoch": 6.86, + "learning_rate": 1.5714441774087833e-06, + "loss": 2.5039, + "step": 31435 + }, + { + "epoch": 6.86, + "learning_rate": 1.5713349355473017e-06, + "loss": 0.3954, + "step": 31436 + }, + { + "epoch": 6.86, + "learning_rate": 1.5712256936858206e-06, + "loss": 2.6953, + "step": 31437 + }, + { + "epoch": 6.86, + "learning_rate": 1.5711164518243392e-06, + "loss": 1.1573, + "step": 31438 + }, + { + "epoch": 6.86, + "learning_rate": 1.5710072099628578e-06, + "loss": 1.305, + "step": 31439 + }, + { + "epoch": 6.86, + "learning_rate": 1.5708979681013766e-06, + "loss": 2.3177, + "step": 31440 + }, + { + "epoch": 6.86, + "learning_rate": 1.5707887262398952e-06, + "loss": 0.6924, + "step": 31441 + }, + { + "epoch": 6.86, + "learning_rate": 1.570679484378414e-06, + "loss": 2.1942, + "step": 31442 + }, + { + "epoch": 6.86, + "learning_rate": 1.5705702425169324e-06, + "loss": 1.3167, + "step": 31443 + }, + { + "epoch": 6.86, + "learning_rate": 1.5704610006554513e-06, + "loss": 2.6034, + "step": 31444 + }, + { + "epoch": 6.86, + "learning_rate": 1.57035175879397e-06, + "loss": 0.8464, + "step": 31445 + }, + { + "epoch": 6.86, + "learning_rate": 1.5702425169324887e-06, + "loss": 0.3574, + "step": 31446 + }, + { + "epoch": 6.86, + "learning_rate": 1.5701332750710075e-06, + "loss": 1.175, + "step": 31447 + }, + { + "epoch": 6.86, + "learning_rate": 1.570024033209526e-06, + "loss": 2.2499, + "step": 31448 + }, + { + "epoch": 6.86, + "learning_rate": 1.5699147913480447e-06, + "loss": 0.2338, + "step": 31449 + }, + { + "epoch": 6.86, + "learning_rate": 1.5698055494865634e-06, + "loss": 0.6017, + "step": 31450 + }, + { + "epoch": 6.86, + "learning_rate": 1.569696307625082e-06, + "loss": 0.2538, + "step": 31451 + }, + { + "epoch": 6.86, + "learning_rate": 1.5695870657636008e-06, + "loss": 1.4709, + "step": 31452 + }, + { + "epoch": 6.86, + "learning_rate": 1.5694778239021194e-06, + "loss": 1.051, + "step": 31453 + }, + { + "epoch": 6.86, + "learning_rate": 1.5693685820406382e-06, + "loss": 1.2834, + "step": 31454 + }, + { + "epoch": 6.86, + "learning_rate": 1.5692593401791566e-06, + "loss": 0.5083, + "step": 31455 + }, + { + "epoch": 6.87, + "learning_rate": 1.5691500983176754e-06, + "loss": 1.1545, + "step": 31456 + }, + { + "epoch": 6.87, + "learning_rate": 1.569040856456194e-06, + "loss": 1.7368, + "step": 31457 + }, + { + "epoch": 6.87, + "learning_rate": 1.5689316145947129e-06, + "loss": 0.3915, + "step": 31458 + }, + { + "epoch": 6.87, + "learning_rate": 1.5688223727332317e-06, + "loss": 2.8743, + "step": 31459 + }, + { + "epoch": 6.87, + "learning_rate": 1.56871313087175e-06, + "loss": 0.8615, + "step": 31460 + }, + { + "epoch": 6.87, + "learning_rate": 1.568603889010269e-06, + "loss": 0.6421, + "step": 31461 + }, + { + "epoch": 6.87, + "learning_rate": 1.5684946471487875e-06, + "loss": 0.5868, + "step": 31462 + }, + { + "epoch": 6.87, + "learning_rate": 1.5683854052873064e-06, + "loss": 3.1846, + "step": 31463 + }, + { + "epoch": 6.87, + "learning_rate": 1.5682761634258248e-06, + "loss": 0.7674, + "step": 31464 + }, + { + "epoch": 6.87, + "learning_rate": 1.5681669215643436e-06, + "loss": 0.9523, + "step": 31465 + }, + { + "epoch": 6.87, + "learning_rate": 1.5680576797028624e-06, + "loss": 0.5888, + "step": 31466 + }, + { + "epoch": 6.87, + "learning_rate": 1.5679484378413808e-06, + "loss": 2.0015, + "step": 31467 + }, + { + "epoch": 6.87, + "learning_rate": 1.5678391959798996e-06, + "loss": 2.4128, + "step": 31468 + }, + { + "epoch": 6.87, + "learning_rate": 1.5677299541184182e-06, + "loss": 1.5592, + "step": 31469 + }, + { + "epoch": 6.87, + "learning_rate": 1.567620712256937e-06, + "loss": 1.0519, + "step": 31470 + }, + { + "epoch": 6.87, + "learning_rate": 1.5675114703954555e-06, + "loss": 0.7858, + "step": 31471 + }, + { + "epoch": 6.87, + "learning_rate": 1.5674022285339743e-06, + "loss": 1.7559, + "step": 31472 + }, + { + "epoch": 6.87, + "learning_rate": 1.5672929866724931e-06, + "loss": 0.5334, + "step": 31473 + }, + { + "epoch": 6.87, + "learning_rate": 1.5671837448110117e-06, + "loss": 2.9223, + "step": 31474 + }, + { + "epoch": 6.87, + "learning_rate": 1.5670745029495305e-06, + "loss": 1.2624, + "step": 31475 + }, + { + "epoch": 6.87, + "learning_rate": 1.566965261088049e-06, + "loss": 1.3813, + "step": 31476 + }, + { + "epoch": 6.87, + "learning_rate": 1.5668560192265678e-06, + "loss": 0.0958, + "step": 31477 + }, + { + "epoch": 6.87, + "learning_rate": 1.5667467773650864e-06, + "loss": 0.5558, + "step": 31478 + }, + { + "epoch": 6.87, + "learning_rate": 1.566637535503605e-06, + "loss": 2.1437, + "step": 31479 + }, + { + "epoch": 6.87, + "learning_rate": 1.5665282936421238e-06, + "loss": 1.1144, + "step": 31480 + }, + { + "epoch": 6.87, + "learning_rate": 1.5664190517806424e-06, + "loss": 0.3557, + "step": 31481 + }, + { + "epoch": 6.87, + "learning_rate": 1.5663098099191612e-06, + "loss": 2.3455, + "step": 31482 + }, + { + "epoch": 6.87, + "learning_rate": 1.5662005680576796e-06, + "loss": 1.7685, + "step": 31483 + }, + { + "epoch": 6.87, + "learning_rate": 1.5660913261961985e-06, + "loss": 1.7372, + "step": 31484 + }, + { + "epoch": 6.87, + "learning_rate": 1.5659820843347173e-06, + "loss": 1.6207, + "step": 31485 + }, + { + "epoch": 6.87, + "learning_rate": 1.565872842473236e-06, + "loss": 0.5835, + "step": 31486 + }, + { + "epoch": 6.87, + "learning_rate": 1.5657636006117547e-06, + "loss": 1.6177, + "step": 31487 + }, + { + "epoch": 6.87, + "learning_rate": 1.5656543587502731e-06, + "loss": 4.535, + "step": 31488 + }, + { + "epoch": 6.87, + "learning_rate": 1.565545116888792e-06, + "loss": 1.8865, + "step": 31489 + }, + { + "epoch": 6.87, + "learning_rate": 1.5654358750273106e-06, + "loss": 1.5062, + "step": 31490 + }, + { + "epoch": 6.87, + "learning_rate": 1.5653266331658294e-06, + "loss": 0.3708, + "step": 31491 + }, + { + "epoch": 6.87, + "learning_rate": 1.565217391304348e-06, + "loss": 1.1268, + "step": 31492 + }, + { + "epoch": 6.87, + "learning_rate": 1.5651081494428666e-06, + "loss": 0.9442, + "step": 31493 + }, + { + "epoch": 6.87, + "learning_rate": 1.5649989075813854e-06, + "loss": 1.9962, + "step": 31494 + }, + { + "epoch": 6.87, + "learning_rate": 1.5648896657199038e-06, + "loss": 0.3603, + "step": 31495 + }, + { + "epoch": 6.87, + "learning_rate": 1.5647804238584226e-06, + "loss": 1.7543, + "step": 31496 + }, + { + "epoch": 6.87, + "learning_rate": 1.5646711819969413e-06, + "loss": 1.0525, + "step": 31497 + }, + { + "epoch": 6.87, + "learning_rate": 1.56456194013546e-06, + "loss": 1.1237, + "step": 31498 + }, + { + "epoch": 6.87, + "learning_rate": 1.564452698273979e-06, + "loss": 1.901, + "step": 31499 + }, + { + "epoch": 6.87, + "learning_rate": 1.5643434564124973e-06, + "loss": 0.4965, + "step": 31500 + }, + { + "epoch": 6.87, + "learning_rate": 1.5642342145510161e-06, + "loss": 0.9261, + "step": 31501 + }, + { + "epoch": 6.88, + "learning_rate": 1.5641249726895347e-06, + "loss": 0.3095, + "step": 31502 + }, + { + "epoch": 6.88, + "learning_rate": 1.5640157308280536e-06, + "loss": 1.1061, + "step": 31503 + }, + { + "epoch": 6.88, + "learning_rate": 1.563906488966572e-06, + "loss": 1.4264, + "step": 31504 + }, + { + "epoch": 6.88, + "learning_rate": 1.5637972471050908e-06, + "loss": 1.324, + "step": 31505 + }, + { + "epoch": 6.88, + "learning_rate": 1.5636880052436096e-06, + "loss": 3.4431, + "step": 31506 + }, + { + "epoch": 6.88, + "learning_rate": 1.5635787633821282e-06, + "loss": 0.2523, + "step": 31507 + }, + { + "epoch": 6.88, + "learning_rate": 1.5634695215206468e-06, + "loss": 1.4395, + "step": 31508 + }, + { + "epoch": 6.88, + "learning_rate": 1.5633602796591654e-06, + "loss": 2.9036, + "step": 31509 + }, + { + "epoch": 6.88, + "learning_rate": 1.5632510377976843e-06, + "loss": 3.3769, + "step": 31510 + }, + { + "epoch": 6.88, + "learning_rate": 1.5631417959362027e-06, + "loss": 0.5331, + "step": 31511 + }, + { + "epoch": 6.88, + "learning_rate": 1.5630325540747215e-06, + "loss": 1.6112, + "step": 31512 + }, + { + "epoch": 6.88, + "learning_rate": 1.5629233122132403e-06, + "loss": 0.8539, + "step": 31513 + }, + { + "epoch": 6.88, + "learning_rate": 1.562814070351759e-06, + "loss": 1.2924, + "step": 31514 + }, + { + "epoch": 6.88, + "learning_rate": 1.5627048284902777e-06, + "loss": 0.7396, + "step": 31515 + }, + { + "epoch": 6.88, + "learning_rate": 1.5625955866287961e-06, + "loss": 0.613, + "step": 31516 + }, + { + "epoch": 6.88, + "learning_rate": 1.562486344767315e-06, + "loss": 0.6334, + "step": 31517 + }, + { + "epoch": 6.88, + "learning_rate": 1.5623771029058338e-06, + "loss": 1.2603, + "step": 31518 + }, + { + "epoch": 6.88, + "learning_rate": 1.5622678610443524e-06, + "loss": 0.2153, + "step": 31519 + }, + { + "epoch": 6.88, + "learning_rate": 1.562158619182871e-06, + "loss": 2.3076, + "step": 31520 + }, + { + "epoch": 6.88, + "learning_rate": 1.5620493773213896e-06, + "loss": 0.6674, + "step": 31521 + }, + { + "epoch": 6.88, + "learning_rate": 1.5619401354599084e-06, + "loss": 0.7566, + "step": 31522 + }, + { + "epoch": 6.88, + "learning_rate": 1.5618308935984268e-06, + "loss": 1.0685, + "step": 31523 + }, + { + "epoch": 6.88, + "learning_rate": 1.5617216517369457e-06, + "loss": 0.5068, + "step": 31524 + }, + { + "epoch": 6.88, + "learning_rate": 1.5616124098754645e-06, + "loss": 1.549, + "step": 31525 + }, + { + "epoch": 6.88, + "learning_rate": 1.561503168013983e-06, + "loss": 1.2867, + "step": 31526 + }, + { + "epoch": 6.88, + "learning_rate": 1.561393926152502e-06, + "loss": 2.0554, + "step": 31527 + }, + { + "epoch": 6.88, + "learning_rate": 1.5612846842910203e-06, + "loss": 1.3999, + "step": 31528 + }, + { + "epoch": 6.88, + "learning_rate": 1.5611754424295391e-06, + "loss": 2.1167, + "step": 31529 + }, + { + "epoch": 6.88, + "learning_rate": 1.5610662005680577e-06, + "loss": 0.7092, + "step": 31530 + }, + { + "epoch": 6.88, + "learning_rate": 1.5609569587065766e-06, + "loss": 2.7948, + "step": 31531 + }, + { + "epoch": 6.88, + "learning_rate": 1.5608477168450954e-06, + "loss": 0.1355, + "step": 31532 + }, + { + "epoch": 6.88, + "learning_rate": 1.5607384749836138e-06, + "loss": 0.702, + "step": 31533 + }, + { + "epoch": 6.88, + "learning_rate": 1.5606292331221326e-06, + "loss": 0.266, + "step": 31534 + }, + { + "epoch": 6.88, + "learning_rate": 1.5605199912606512e-06, + "loss": 2.8718, + "step": 31535 + }, + { + "epoch": 6.88, + "learning_rate": 1.5604107493991698e-06, + "loss": 2.0955, + "step": 31536 + }, + { + "epoch": 6.88, + "learning_rate": 1.5603015075376884e-06, + "loss": 0.2087, + "step": 31537 + }, + { + "epoch": 6.88, + "learning_rate": 1.5601922656762073e-06, + "loss": 2.3744, + "step": 31538 + }, + { + "epoch": 6.88, + "learning_rate": 1.560083023814726e-06, + "loss": 1.6113, + "step": 31539 + }, + { + "epoch": 6.88, + "learning_rate": 1.5599737819532445e-06, + "loss": 1.1338, + "step": 31540 + }, + { + "epoch": 6.88, + "learning_rate": 1.5598645400917633e-06, + "loss": 2.3451, + "step": 31541 + }, + { + "epoch": 6.88, + "learning_rate": 1.559755298230282e-06, + "loss": 1.247, + "step": 31542 + }, + { + "epoch": 6.88, + "learning_rate": 1.5596460563688007e-06, + "loss": 0.5347, + "step": 31543 + }, + { + "epoch": 6.88, + "learning_rate": 1.5595368145073191e-06, + "loss": 2.5086, + "step": 31544 + }, + { + "epoch": 6.88, + "learning_rate": 1.559427572645838e-06, + "loss": 0.2819, + "step": 31545 + }, + { + "epoch": 6.88, + "learning_rate": 1.5593183307843568e-06, + "loss": 2.0041, + "step": 31546 + }, + { + "epoch": 6.88, + "learning_rate": 1.5592090889228754e-06, + "loss": 2.6242, + "step": 31547 + }, + { + "epoch": 6.89, + "learning_rate": 1.559099847061394e-06, + "loss": 1.7026, + "step": 31548 + }, + { + "epoch": 6.89, + "learning_rate": 1.5589906051999126e-06, + "loss": 1.4264, + "step": 31549 + }, + { + "epoch": 6.89, + "learning_rate": 1.5588813633384314e-06, + "loss": 1.79, + "step": 31550 + }, + { + "epoch": 6.89, + "learning_rate": 1.5587721214769503e-06, + "loss": 2.2772, + "step": 31551 + }, + { + "epoch": 6.89, + "learning_rate": 1.5586628796154687e-06, + "loss": 0.5621, + "step": 31552 + }, + { + "epoch": 6.89, + "learning_rate": 1.5585536377539875e-06, + "loss": 0.726, + "step": 31553 + }, + { + "epoch": 6.89, + "learning_rate": 1.558444395892506e-06, + "loss": 0.9789, + "step": 31554 + }, + { + "epoch": 6.89, + "learning_rate": 1.558335154031025e-06, + "loss": 0.7993, + "step": 31555 + }, + { + "epoch": 6.89, + "learning_rate": 1.5582259121695433e-06, + "loss": 1.689, + "step": 31556 + }, + { + "epoch": 6.89, + "learning_rate": 1.5581166703080621e-06, + "loss": 0.6432, + "step": 31557 + }, + { + "epoch": 6.89, + "learning_rate": 1.558007428446581e-06, + "loss": 0.9655, + "step": 31558 + }, + { + "epoch": 6.89, + "learning_rate": 1.5578981865850996e-06, + "loss": 1.039, + "step": 31559 + }, + { + "epoch": 6.89, + "learning_rate": 1.5577889447236184e-06, + "loss": 2.6711, + "step": 31560 + }, + { + "epoch": 6.89, + "learning_rate": 1.5576797028621368e-06, + "loss": 1.5113, + "step": 31561 + }, + { + "epoch": 6.89, + "learning_rate": 1.5575704610006556e-06, + "loss": 0.8598, + "step": 31562 + }, + { + "epoch": 6.89, + "learning_rate": 1.5574612191391742e-06, + "loss": 1.2319, + "step": 31563 + }, + { + "epoch": 6.89, + "learning_rate": 1.5573519772776928e-06, + "loss": 1.061, + "step": 31564 + }, + { + "epoch": 6.89, + "learning_rate": 1.5572427354162117e-06, + "loss": 1.1219, + "step": 31565 + }, + { + "epoch": 6.89, + "learning_rate": 1.5571334935547303e-06, + "loss": 0.1476, + "step": 31566 + }, + { + "epoch": 6.89, + "learning_rate": 1.557024251693249e-06, + "loss": 1.2412, + "step": 31567 + }, + { + "epoch": 6.89, + "learning_rate": 1.5569150098317675e-06, + "loss": 3.2296, + "step": 31568 + }, + { + "epoch": 6.89, + "learning_rate": 1.5568057679702863e-06, + "loss": 1.5056, + "step": 31569 + }, + { + "epoch": 6.89, + "learning_rate": 1.556696526108805e-06, + "loss": 0.235, + "step": 31570 + }, + { + "epoch": 6.89, + "learning_rate": 1.5565872842473238e-06, + "loss": 2.1217, + "step": 31571 + }, + { + "epoch": 6.89, + "learning_rate": 1.5564780423858426e-06, + "loss": 1.6124, + "step": 31572 + }, + { + "epoch": 6.89, + "learning_rate": 1.556368800524361e-06, + "loss": 0.2317, + "step": 31573 + }, + { + "epoch": 6.89, + "learning_rate": 1.5562595586628798e-06, + "loss": 1.9452, + "step": 31574 + }, + { + "epoch": 6.89, + "learning_rate": 1.5561503168013984e-06, + "loss": 2.0299, + "step": 31575 + }, + { + "epoch": 6.89, + "learning_rate": 1.556041074939917e-06, + "loss": 1.9734, + "step": 31576 + }, + { + "epoch": 6.89, + "learning_rate": 1.5559318330784356e-06, + "loss": 0.8635, + "step": 31577 + }, + { + "epoch": 6.89, + "learning_rate": 1.5558225912169545e-06, + "loss": 1.2744, + "step": 31578 + }, + { + "epoch": 6.89, + "learning_rate": 1.5557133493554733e-06, + "loss": 1.6536, + "step": 31579 + }, + { + "epoch": 6.89, + "learning_rate": 1.5556041074939917e-06, + "loss": 1.6694, + "step": 31580 + }, + { + "epoch": 6.89, + "learning_rate": 1.5554948656325105e-06, + "loss": 0.8832, + "step": 31581 + }, + { + "epoch": 6.89, + "learning_rate": 1.5553856237710291e-06, + "loss": 0.7912, + "step": 31582 + }, + { + "epoch": 6.89, + "learning_rate": 1.555276381909548e-06, + "loss": 3.0102, + "step": 31583 + }, + { + "epoch": 6.89, + "learning_rate": 1.5551671400480668e-06, + "loss": 0.6425, + "step": 31584 + }, + { + "epoch": 6.89, + "learning_rate": 1.5550578981865852e-06, + "loss": 0.8764, + "step": 31585 + }, + { + "epoch": 6.89, + "learning_rate": 1.554948656325104e-06, + "loss": 0.6416, + "step": 31586 + }, + { + "epoch": 6.89, + "learning_rate": 1.5548394144636226e-06, + "loss": 0.852, + "step": 31587 + }, + { + "epoch": 6.89, + "learning_rate": 1.5547301726021414e-06, + "loss": 1.3776, + "step": 31588 + }, + { + "epoch": 6.89, + "learning_rate": 1.5546209307406598e-06, + "loss": 1.674, + "step": 31589 + }, + { + "epoch": 6.89, + "learning_rate": 1.5545116888791786e-06, + "loss": 2.0579, + "step": 31590 + }, + { + "epoch": 6.89, + "learning_rate": 1.5544024470176975e-06, + "loss": 1.5793, + "step": 31591 + }, + { + "epoch": 6.89, + "learning_rate": 1.5542932051562159e-06, + "loss": 1.4357, + "step": 31592 + }, + { + "epoch": 6.9, + "learning_rate": 1.5541839632947347e-06, + "loss": 0.6104, + "step": 31593 + }, + { + "epoch": 6.9, + "learning_rate": 1.5540747214332533e-06, + "loss": 1.5315, + "step": 31594 + }, + { + "epoch": 6.9, + "learning_rate": 1.5539654795717721e-06, + "loss": 0.4997, + "step": 31595 + }, + { + "epoch": 6.9, + "learning_rate": 1.5538562377102905e-06, + "loss": 1.7071, + "step": 31596 + }, + { + "epoch": 6.9, + "learning_rate": 1.5537469958488093e-06, + "loss": 1.482, + "step": 31597 + }, + { + "epoch": 6.9, + "learning_rate": 1.5536377539873282e-06, + "loss": 1.0815, + "step": 31598 + }, + { + "epoch": 6.9, + "learning_rate": 1.5535285121258468e-06, + "loss": 0.9571, + "step": 31599 + }, + { + "epoch": 6.9, + "learning_rate": 1.5534192702643656e-06, + "loss": 0.6787, + "step": 31600 + }, + { + "epoch": 6.9, + "learning_rate": 1.553310028402884e-06, + "loss": 0.1756, + "step": 31601 + }, + { + "epoch": 6.9, + "learning_rate": 1.5532007865414028e-06, + "loss": 0.0387, + "step": 31602 + }, + { + "epoch": 6.9, + "learning_rate": 1.5530915446799214e-06, + "loss": 0.4903, + "step": 31603 + }, + { + "epoch": 6.9, + "learning_rate": 1.5529823028184403e-06, + "loss": 2.579, + "step": 31604 + }, + { + "epoch": 6.9, + "learning_rate": 1.5528730609569589e-06, + "loss": 1.3643, + "step": 31605 + }, + { + "epoch": 6.9, + "learning_rate": 1.5527638190954775e-06, + "loss": 1.7548, + "step": 31606 + }, + { + "epoch": 6.9, + "learning_rate": 1.5526545772339963e-06, + "loss": 0.5535, + "step": 31607 + }, + { + "epoch": 6.9, + "learning_rate": 1.5525453353725147e-06, + "loss": 1.4921, + "step": 31608 + }, + { + "epoch": 6.9, + "learning_rate": 1.5524360935110335e-06, + "loss": 1.5775, + "step": 31609 + }, + { + "epoch": 6.9, + "learning_rate": 1.5523268516495521e-06, + "loss": 1.2459, + "step": 31610 + }, + { + "epoch": 6.9, + "learning_rate": 1.552217609788071e-06, + "loss": 1.7813, + "step": 31611 + }, + { + "epoch": 6.9, + "learning_rate": 1.5521083679265898e-06, + "loss": 1.2244, + "step": 31612 + }, + { + "epoch": 6.9, + "learning_rate": 1.5519991260651082e-06, + "loss": 2.3573, + "step": 31613 + }, + { + "epoch": 6.9, + "learning_rate": 1.551889884203627e-06, + "loss": 0.7304, + "step": 31614 + }, + { + "epoch": 6.9, + "learning_rate": 1.5517806423421456e-06, + "loss": 1.3925, + "step": 31615 + }, + { + "epoch": 6.9, + "learning_rate": 1.5516714004806644e-06, + "loss": 2.1479, + "step": 31616 + }, + { + "epoch": 6.9, + "learning_rate": 1.551562158619183e-06, + "loss": 2.1319, + "step": 31617 + }, + { + "epoch": 6.9, + "learning_rate": 1.5514529167577017e-06, + "loss": 1.1356, + "step": 31618 + }, + { + "epoch": 6.9, + "learning_rate": 1.5513436748962205e-06, + "loss": 1.6261, + "step": 31619 + }, + { + "epoch": 6.9, + "learning_rate": 1.5512344330347389e-06, + "loss": 1.1942, + "step": 31620 + }, + { + "epoch": 6.9, + "learning_rate": 1.5511251911732577e-06, + "loss": 0.0522, + "step": 31621 + }, + { + "epoch": 6.9, + "learning_rate": 1.5510159493117763e-06, + "loss": 0.2651, + "step": 31622 + }, + { + "epoch": 6.9, + "learning_rate": 1.5509067074502951e-06, + "loss": 0.2703, + "step": 31623 + }, + { + "epoch": 6.9, + "learning_rate": 1.550797465588814e-06, + "loss": 1.3255, + "step": 31624 + }, + { + "epoch": 6.9, + "learning_rate": 1.5506882237273324e-06, + "loss": 2.4031, + "step": 31625 + }, + { + "epoch": 6.9, + "learning_rate": 1.5505789818658512e-06, + "loss": 0.1594, + "step": 31626 + }, + { + "epoch": 6.9, + "learning_rate": 1.5504697400043698e-06, + "loss": 1.1673, + "step": 31627 + }, + { + "epoch": 6.9, + "learning_rate": 1.5503604981428886e-06, + "loss": 2.2783, + "step": 31628 + }, + { + "epoch": 6.9, + "learning_rate": 1.550251256281407e-06, + "loss": 1.8124, + "step": 31629 + }, + { + "epoch": 6.9, + "learning_rate": 1.5501420144199258e-06, + "loss": 1.6486, + "step": 31630 + }, + { + "epoch": 6.9, + "learning_rate": 1.5500327725584447e-06, + "loss": 1.0656, + "step": 31631 + }, + { + "epoch": 6.9, + "learning_rate": 1.5499235306969633e-06, + "loss": 1.5035, + "step": 31632 + }, + { + "epoch": 6.9, + "learning_rate": 1.5498142888354819e-06, + "loss": 0.8321, + "step": 31633 + }, + { + "epoch": 6.9, + "learning_rate": 1.5497050469740005e-06, + "loss": 0.9214, + "step": 31634 + }, + { + "epoch": 6.9, + "learning_rate": 1.5495958051125193e-06, + "loss": 1.3602, + "step": 31635 + }, + { + "epoch": 6.9, + "learning_rate": 1.5494865632510377e-06, + "loss": 2.0741, + "step": 31636 + }, + { + "epoch": 6.9, + "learning_rate": 1.5493773213895565e-06, + "loss": 1.3377, + "step": 31637 + }, + { + "epoch": 6.9, + "learning_rate": 1.5492680795280754e-06, + "loss": 0.6128, + "step": 31638 + }, + { + "epoch": 6.91, + "learning_rate": 1.549158837666594e-06, + "loss": 0.7618, + "step": 31639 + }, + { + "epoch": 6.91, + "learning_rate": 1.5490495958051128e-06, + "loss": 0.6825, + "step": 31640 + }, + { + "epoch": 6.91, + "learning_rate": 1.5489403539436312e-06, + "loss": 0.5239, + "step": 31641 + }, + { + "epoch": 6.91, + "learning_rate": 1.54883111208215e-06, + "loss": 0.6241, + "step": 31642 + }, + { + "epoch": 6.91, + "learning_rate": 1.5487218702206686e-06, + "loss": 1.2746, + "step": 31643 + }, + { + "epoch": 6.91, + "learning_rate": 1.5486126283591874e-06, + "loss": 1.8969, + "step": 31644 + }, + { + "epoch": 6.91, + "learning_rate": 1.548503386497706e-06, + "loss": 2.3304, + "step": 31645 + }, + { + "epoch": 6.91, + "learning_rate": 1.5483941446362247e-06, + "loss": 2.0107, + "step": 31646 + }, + { + "epoch": 6.91, + "learning_rate": 1.5482849027747435e-06, + "loss": 1.6798, + "step": 31647 + }, + { + "epoch": 6.91, + "learning_rate": 1.5481756609132619e-06, + "loss": 1.5844, + "step": 31648 + }, + { + "epoch": 6.91, + "learning_rate": 1.5480664190517807e-06, + "loss": 0.3696, + "step": 31649 + }, + { + "epoch": 6.91, + "learning_rate": 1.5479571771902995e-06, + "loss": 2.6766, + "step": 31650 + }, + { + "epoch": 6.91, + "learning_rate": 1.5478479353288181e-06, + "loss": 1.8873, + "step": 31651 + }, + { + "epoch": 6.91, + "learning_rate": 1.547738693467337e-06, + "loss": 0.4647, + "step": 31652 + }, + { + "epoch": 6.91, + "learning_rate": 1.5476294516058554e-06, + "loss": 0.3928, + "step": 31653 + }, + { + "epoch": 6.91, + "learning_rate": 1.5475202097443742e-06, + "loss": 1.325, + "step": 31654 + }, + { + "epoch": 6.91, + "learning_rate": 1.5474109678828928e-06, + "loss": 1.3962, + "step": 31655 + }, + { + "epoch": 6.91, + "learning_rate": 1.5473017260214116e-06, + "loss": 2.7848, + "step": 31656 + }, + { + "epoch": 6.91, + "learning_rate": 1.5471924841599304e-06, + "loss": 2.1428, + "step": 31657 + }, + { + "epoch": 6.91, + "learning_rate": 1.5470832422984488e-06, + "loss": 0.0491, + "step": 31658 + }, + { + "epoch": 6.91, + "learning_rate": 1.5469740004369677e-06, + "loss": 2.1222, + "step": 31659 + }, + { + "epoch": 6.91, + "learning_rate": 1.5468647585754863e-06, + "loss": 1.3886, + "step": 31660 + }, + { + "epoch": 6.91, + "learning_rate": 1.5467555167140049e-06, + "loss": 1.0284, + "step": 31661 + }, + { + "epoch": 6.91, + "learning_rate": 1.5466462748525235e-06, + "loss": 2.2144, + "step": 31662 + }, + { + "epoch": 6.91, + "learning_rate": 1.5465370329910423e-06, + "loss": 1.2152, + "step": 31663 + }, + { + "epoch": 6.91, + "learning_rate": 1.5464277911295611e-06, + "loss": 0.6156, + "step": 31664 + }, + { + "epoch": 6.91, + "learning_rate": 1.5463185492680795e-06, + "loss": 1.5459, + "step": 31665 + }, + { + "epoch": 6.91, + "learning_rate": 1.5462093074065984e-06, + "loss": 0.904, + "step": 31666 + }, + { + "epoch": 6.91, + "learning_rate": 1.546100065545117e-06, + "loss": 0.9132, + "step": 31667 + }, + { + "epoch": 6.91, + "learning_rate": 1.5459908236836358e-06, + "loss": 0.9077, + "step": 31668 + }, + { + "epoch": 6.91, + "learning_rate": 1.5458815818221542e-06, + "loss": 0.688, + "step": 31669 + }, + { + "epoch": 6.91, + "learning_rate": 1.545772339960673e-06, + "loss": 2.2576, + "step": 31670 + }, + { + "epoch": 6.91, + "learning_rate": 1.5456630980991918e-06, + "loss": 1.296, + "step": 31671 + }, + { + "epoch": 6.91, + "learning_rate": 1.5455538562377105e-06, + "loss": 2.5849, + "step": 31672 + }, + { + "epoch": 6.91, + "learning_rate": 1.545444614376229e-06, + "loss": 0.4542, + "step": 31673 + }, + { + "epoch": 6.91, + "learning_rate": 1.5453353725147477e-06, + "loss": 0.419, + "step": 31674 + }, + { + "epoch": 6.91, + "learning_rate": 1.5452261306532665e-06, + "loss": 1.4756, + "step": 31675 + }, + { + "epoch": 6.91, + "learning_rate": 1.5451168887917851e-06, + "loss": 0.7485, + "step": 31676 + }, + { + "epoch": 6.91, + "learning_rate": 1.5450076469303037e-06, + "loss": 2.1457, + "step": 31677 + }, + { + "epoch": 6.91, + "learning_rate": 1.5448984050688225e-06, + "loss": 1.6121, + "step": 31678 + }, + { + "epoch": 6.91, + "learning_rate": 1.5447891632073412e-06, + "loss": 2.4275, + "step": 31679 + }, + { + "epoch": 6.91, + "learning_rate": 1.54467992134586e-06, + "loss": 1.0834, + "step": 31680 + }, + { + "epoch": 6.91, + "learning_rate": 1.5445706794843784e-06, + "loss": 0.6577, + "step": 31681 + }, + { + "epoch": 6.91, + "learning_rate": 1.5444614376228972e-06, + "loss": 0.4262, + "step": 31682 + }, + { + "epoch": 6.91, + "learning_rate": 1.5443521957614158e-06, + "loss": 1.0963, + "step": 31683 + }, + { + "epoch": 6.91, + "learning_rate": 1.5442429538999346e-06, + "loss": 3.5606, + "step": 31684 + }, + { + "epoch": 6.92, + "learning_rate": 1.5441337120384535e-06, + "loss": 1.2126, + "step": 31685 + }, + { + "epoch": 6.92, + "learning_rate": 1.5440244701769719e-06, + "loss": 0.885, + "step": 31686 + }, + { + "epoch": 6.92, + "learning_rate": 1.5439152283154907e-06, + "loss": 1.8175, + "step": 31687 + }, + { + "epoch": 6.92, + "learning_rate": 1.5438059864540093e-06, + "loss": 0.0626, + "step": 31688 + }, + { + "epoch": 6.92, + "learning_rate": 1.543696744592528e-06, + "loss": 0.2483, + "step": 31689 + }, + { + "epoch": 6.92, + "learning_rate": 1.5435875027310467e-06, + "loss": 2.4476, + "step": 31690 + }, + { + "epoch": 6.92, + "learning_rate": 1.5434782608695653e-06, + "loss": 3.1093, + "step": 31691 + }, + { + "epoch": 6.92, + "learning_rate": 1.5433690190080842e-06, + "loss": 0.6829, + "step": 31692 + }, + { + "epoch": 6.92, + "learning_rate": 1.5432597771466026e-06, + "loss": 0.7401, + "step": 31693 + }, + { + "epoch": 6.92, + "learning_rate": 1.5431505352851214e-06, + "loss": 1.5175, + "step": 31694 + }, + { + "epoch": 6.92, + "learning_rate": 1.54304129342364e-06, + "loss": 1.4132, + "step": 31695 + }, + { + "epoch": 6.92, + "learning_rate": 1.5429320515621588e-06, + "loss": 0.4596, + "step": 31696 + }, + { + "epoch": 6.92, + "learning_rate": 1.5428228097006776e-06, + "loss": 0.328, + "step": 31697 + }, + { + "epoch": 6.92, + "learning_rate": 1.542713567839196e-06, + "loss": 2.0988, + "step": 31698 + }, + { + "epoch": 6.92, + "learning_rate": 1.5426043259777149e-06, + "loss": 1.9394, + "step": 31699 + }, + { + "epoch": 6.92, + "learning_rate": 1.5424950841162335e-06, + "loss": 0.4631, + "step": 31700 + }, + { + "epoch": 6.92, + "learning_rate": 1.542385842254752e-06, + "loss": 1.713, + "step": 31701 + }, + { + "epoch": 6.92, + "learning_rate": 1.5422766003932707e-06, + "loss": 0.5202, + "step": 31702 + }, + { + "epoch": 6.92, + "learning_rate": 1.5421673585317895e-06, + "loss": 1.1979, + "step": 31703 + }, + { + "epoch": 6.92, + "learning_rate": 1.5420581166703083e-06, + "loss": 2.2069, + "step": 31704 + }, + { + "epoch": 6.92, + "learning_rate": 1.5419488748088267e-06, + "loss": 1.1295, + "step": 31705 + }, + { + "epoch": 6.92, + "learning_rate": 1.5418396329473456e-06, + "loss": 1.2797, + "step": 31706 + }, + { + "epoch": 6.92, + "learning_rate": 1.5417303910858642e-06, + "loss": 0.6569, + "step": 31707 + }, + { + "epoch": 6.92, + "learning_rate": 1.541621149224383e-06, + "loss": 1.1085, + "step": 31708 + }, + { + "epoch": 6.92, + "learning_rate": 1.5415119073629014e-06, + "loss": 0.4932, + "step": 31709 + }, + { + "epoch": 6.92, + "learning_rate": 1.5414026655014202e-06, + "loss": 2.8642, + "step": 31710 + }, + { + "epoch": 6.92, + "learning_rate": 1.541293423639939e-06, + "loss": 1.7632, + "step": 31711 + }, + { + "epoch": 6.92, + "learning_rate": 1.5411841817784576e-06, + "loss": 1.8171, + "step": 31712 + }, + { + "epoch": 6.92, + "learning_rate": 1.5410749399169765e-06, + "loss": 1.1449, + "step": 31713 + }, + { + "epoch": 6.92, + "learning_rate": 1.5409656980554949e-06, + "loss": 1.8279, + "step": 31714 + }, + { + "epoch": 6.92, + "learning_rate": 1.5408564561940137e-06, + "loss": 1.2236, + "step": 31715 + }, + { + "epoch": 6.92, + "learning_rate": 1.5407472143325323e-06, + "loss": 0.7497, + "step": 31716 + }, + { + "epoch": 6.92, + "learning_rate": 1.540637972471051e-06, + "loss": 0.5617, + "step": 31717 + }, + { + "epoch": 6.92, + "learning_rate": 1.5405287306095697e-06, + "loss": 2.1691, + "step": 31718 + }, + { + "epoch": 6.92, + "learning_rate": 1.5404194887480883e-06, + "loss": 1.3583, + "step": 31719 + }, + { + "epoch": 6.92, + "learning_rate": 1.5403102468866072e-06, + "loss": 1.7937, + "step": 31720 + }, + { + "epoch": 6.92, + "learning_rate": 1.5402010050251256e-06, + "loss": 1.071, + "step": 31721 + }, + { + "epoch": 6.92, + "learning_rate": 1.5400917631636444e-06, + "loss": 1.7301, + "step": 31722 + }, + { + "epoch": 6.92, + "learning_rate": 1.5399825213021632e-06, + "loss": 0.9915, + "step": 31723 + }, + { + "epoch": 6.92, + "learning_rate": 1.5398732794406818e-06, + "loss": 2.151, + "step": 31724 + }, + { + "epoch": 6.92, + "learning_rate": 1.5397640375792007e-06, + "loss": 2.5007, + "step": 31725 + }, + { + "epoch": 6.92, + "learning_rate": 1.539654795717719e-06, + "loss": 0.1046, + "step": 31726 + }, + { + "epoch": 6.92, + "learning_rate": 1.5395455538562379e-06, + "loss": 1.5637, + "step": 31727 + }, + { + "epoch": 6.92, + "learning_rate": 1.5394363119947565e-06, + "loss": 0.3093, + "step": 31728 + }, + { + "epoch": 6.92, + "learning_rate": 1.5393270701332753e-06, + "loss": 0.5516, + "step": 31729 + }, + { + "epoch": 6.92, + "learning_rate": 1.539217828271794e-06, + "loss": 0.5009, + "step": 31730 + }, + { + "epoch": 6.93, + "learning_rate": 1.5391085864103125e-06, + "loss": 0.9551, + "step": 31731 + }, + { + "epoch": 6.93, + "learning_rate": 1.5389993445488314e-06, + "loss": 1.0688, + "step": 31732 + }, + { + "epoch": 6.93, + "learning_rate": 1.5388901026873498e-06, + "loss": 1.8278, + "step": 31733 + }, + { + "epoch": 6.93, + "learning_rate": 1.5387808608258686e-06, + "loss": 1.888, + "step": 31734 + }, + { + "epoch": 6.93, + "learning_rate": 1.5386716189643872e-06, + "loss": 2.0129, + "step": 31735 + }, + { + "epoch": 6.93, + "learning_rate": 1.538562377102906e-06, + "loss": 3.198, + "step": 31736 + }, + { + "epoch": 6.93, + "learning_rate": 1.5384531352414248e-06, + "loss": 0.6492, + "step": 31737 + }, + { + "epoch": 6.93, + "learning_rate": 1.5383438933799432e-06, + "loss": 1.378, + "step": 31738 + }, + { + "epoch": 6.93, + "learning_rate": 1.538234651518462e-06, + "loss": 1.2995, + "step": 31739 + }, + { + "epoch": 6.93, + "learning_rate": 1.5381254096569807e-06, + "loss": 0.3123, + "step": 31740 + }, + { + "epoch": 6.93, + "learning_rate": 1.5380161677954995e-06, + "loss": 1.1691, + "step": 31741 + }, + { + "epoch": 6.93, + "learning_rate": 1.5379069259340179e-06, + "loss": 1.0701, + "step": 31742 + }, + { + "epoch": 6.93, + "learning_rate": 1.5377976840725367e-06, + "loss": 0.5875, + "step": 31743 + }, + { + "epoch": 6.93, + "learning_rate": 1.5376884422110555e-06, + "loss": 1.0492, + "step": 31744 + }, + { + "epoch": 6.93, + "learning_rate": 1.537579200349574e-06, + "loss": 1.0646, + "step": 31745 + }, + { + "epoch": 6.93, + "learning_rate": 1.5374699584880928e-06, + "loss": 1.6549, + "step": 31746 + }, + { + "epoch": 6.93, + "learning_rate": 1.5373607166266114e-06, + "loss": 0.4687, + "step": 31747 + }, + { + "epoch": 6.93, + "learning_rate": 1.5372514747651302e-06, + "loss": 2.7236, + "step": 31748 + }, + { + "epoch": 6.93, + "learning_rate": 1.5371422329036486e-06, + "loss": 0.4614, + "step": 31749 + }, + { + "epoch": 6.93, + "learning_rate": 1.5370329910421674e-06, + "loss": 1.3309, + "step": 31750 + }, + { + "epoch": 6.93, + "learning_rate": 1.5369237491806862e-06, + "loss": 1.4741, + "step": 31751 + }, + { + "epoch": 6.93, + "learning_rate": 1.5368145073192048e-06, + "loss": 0.8817, + "step": 31752 + }, + { + "epoch": 6.93, + "learning_rate": 1.5367052654577237e-06, + "loss": 2.7001, + "step": 31753 + }, + { + "epoch": 6.93, + "learning_rate": 1.536596023596242e-06, + "loss": 2.3765, + "step": 31754 + }, + { + "epoch": 6.93, + "learning_rate": 1.5364867817347609e-06, + "loss": 2.8349, + "step": 31755 + }, + { + "epoch": 6.93, + "learning_rate": 1.5363775398732797e-06, + "loss": 1.3514, + "step": 31756 + }, + { + "epoch": 6.93, + "learning_rate": 1.5362682980117983e-06, + "loss": 2.4408, + "step": 31757 + }, + { + "epoch": 6.93, + "learning_rate": 1.536159056150317e-06, + "loss": 1.3834, + "step": 31758 + }, + { + "epoch": 6.93, + "learning_rate": 1.5360498142888355e-06, + "loss": 1.3125, + "step": 31759 + }, + { + "epoch": 6.93, + "learning_rate": 1.5359405724273544e-06, + "loss": 3.0983, + "step": 31760 + }, + { + "epoch": 6.93, + "learning_rate": 1.5358313305658728e-06, + "loss": 1.9924, + "step": 31761 + }, + { + "epoch": 6.93, + "learning_rate": 1.5357220887043916e-06, + "loss": 0.7593, + "step": 31762 + }, + { + "epoch": 6.93, + "learning_rate": 1.5356128468429104e-06, + "loss": 0.4537, + "step": 31763 + }, + { + "epoch": 6.93, + "learning_rate": 1.535503604981429e-06, + "loss": 1.0582, + "step": 31764 + }, + { + "epoch": 6.93, + "learning_rate": 1.5353943631199478e-06, + "loss": 1.6916, + "step": 31765 + }, + { + "epoch": 6.93, + "learning_rate": 1.5352851212584662e-06, + "loss": 1.9477, + "step": 31766 + }, + { + "epoch": 6.93, + "learning_rate": 1.535175879396985e-06, + "loss": 0.7645, + "step": 31767 + }, + { + "epoch": 6.93, + "learning_rate": 1.5350666375355037e-06, + "loss": 1.7914, + "step": 31768 + }, + { + "epoch": 6.93, + "learning_rate": 1.5349573956740225e-06, + "loss": 0.5727, + "step": 31769 + }, + { + "epoch": 6.93, + "learning_rate": 1.5348481538125411e-06, + "loss": 0.2186, + "step": 31770 + }, + { + "epoch": 6.93, + "learning_rate": 1.5347389119510597e-06, + "loss": 1.9518, + "step": 31771 + }, + { + "epoch": 6.93, + "learning_rate": 1.5346296700895785e-06, + "loss": 1.0334, + "step": 31772 + }, + { + "epoch": 6.93, + "learning_rate": 1.534520428228097e-06, + "loss": 0.0833, + "step": 31773 + }, + { + "epoch": 6.93, + "learning_rate": 1.5344111863666158e-06, + "loss": 0.7739, + "step": 31774 + }, + { + "epoch": 6.93, + "learning_rate": 1.5343019445051344e-06, + "loss": 0.806, + "step": 31775 + }, + { + "epoch": 6.93, + "learning_rate": 1.5341927026436532e-06, + "loss": 2.4303, + "step": 31776 + }, + { + "epoch": 6.94, + "learning_rate": 1.534083460782172e-06, + "loss": 0.8369, + "step": 31777 + }, + { + "epoch": 6.94, + "learning_rate": 1.5339742189206904e-06, + "loss": 1.7743, + "step": 31778 + }, + { + "epoch": 6.94, + "learning_rate": 1.5338649770592092e-06, + "loss": 2.088, + "step": 31779 + }, + { + "epoch": 6.94, + "learning_rate": 1.5337557351977279e-06, + "loss": 1.1878, + "step": 31780 + }, + { + "epoch": 6.94, + "learning_rate": 1.5336464933362467e-06, + "loss": 1.0349, + "step": 31781 + }, + { + "epoch": 6.94, + "learning_rate": 1.533537251474765e-06, + "loss": 1.9382, + "step": 31782 + }, + { + "epoch": 6.94, + "learning_rate": 1.533428009613284e-06, + "loss": 0.7005, + "step": 31783 + }, + { + "epoch": 6.94, + "learning_rate": 1.5333187677518027e-06, + "loss": 1.6677, + "step": 31784 + }, + { + "epoch": 6.94, + "learning_rate": 1.5332095258903213e-06, + "loss": 1.6352, + "step": 31785 + }, + { + "epoch": 6.94, + "learning_rate": 1.53310028402884e-06, + "loss": 2.0065, + "step": 31786 + }, + { + "epoch": 6.94, + "learning_rate": 1.5329910421673586e-06, + "loss": 0.8398, + "step": 31787 + }, + { + "epoch": 6.94, + "learning_rate": 1.5328818003058774e-06, + "loss": 0.9878, + "step": 31788 + }, + { + "epoch": 6.94, + "learning_rate": 1.5327725584443962e-06, + "loss": 1.4476, + "step": 31789 + }, + { + "epoch": 6.94, + "learning_rate": 1.5326633165829146e-06, + "loss": 1.4877, + "step": 31790 + }, + { + "epoch": 6.94, + "learning_rate": 1.5325540747214334e-06, + "loss": 0.8808, + "step": 31791 + }, + { + "epoch": 6.94, + "learning_rate": 1.532444832859952e-06, + "loss": 1.2237, + "step": 31792 + }, + { + "epoch": 6.94, + "learning_rate": 1.5323355909984709e-06, + "loss": 2.2816, + "step": 31793 + }, + { + "epoch": 6.94, + "learning_rate": 1.5322263491369893e-06, + "loss": 1.7229, + "step": 31794 + }, + { + "epoch": 6.94, + "learning_rate": 1.532117107275508e-06, + "loss": 2.7337, + "step": 31795 + }, + { + "epoch": 6.94, + "learning_rate": 1.532007865414027e-06, + "loss": 2.5655, + "step": 31796 + }, + { + "epoch": 6.94, + "learning_rate": 1.5318986235525455e-06, + "loss": 2.1851, + "step": 31797 + }, + { + "epoch": 6.94, + "learning_rate": 1.5317893816910641e-06, + "loss": 1.8365, + "step": 31798 + }, + { + "epoch": 6.94, + "learning_rate": 1.5316801398295827e-06, + "loss": 0.7569, + "step": 31799 + }, + { + "epoch": 6.94, + "learning_rate": 1.5315708979681016e-06, + "loss": 0.7131, + "step": 31800 + }, + { + "epoch": 6.94, + "learning_rate": 1.5314616561066202e-06, + "loss": 2.4332, + "step": 31801 + }, + { + "epoch": 6.94, + "learning_rate": 1.5313524142451388e-06, + "loss": 0.3794, + "step": 31802 + }, + { + "epoch": 6.94, + "learning_rate": 1.5312431723836576e-06, + "loss": 0.2093, + "step": 31803 + }, + { + "epoch": 6.94, + "learning_rate": 1.5311339305221762e-06, + "loss": 1.3797, + "step": 31804 + }, + { + "epoch": 6.94, + "learning_rate": 1.531024688660695e-06, + "loss": 2.143, + "step": 31805 + }, + { + "epoch": 6.94, + "learning_rate": 1.5309154467992134e-06, + "loss": 1.5764, + "step": 31806 + }, + { + "epoch": 6.94, + "learning_rate": 1.5308062049377323e-06, + "loss": 1.2943, + "step": 31807 + }, + { + "epoch": 6.94, + "learning_rate": 1.5306969630762509e-06, + "loss": 0.5856, + "step": 31808 + }, + { + "epoch": 6.94, + "learning_rate": 1.5305877212147697e-06, + "loss": 1.8105, + "step": 31809 + }, + { + "epoch": 6.94, + "learning_rate": 1.5304784793532885e-06, + "loss": 1.5851, + "step": 31810 + }, + { + "epoch": 6.94, + "learning_rate": 1.530369237491807e-06, + "loss": 0.5203, + "step": 31811 + }, + { + "epoch": 6.94, + "learning_rate": 1.5302599956303257e-06, + "loss": 1.1661, + "step": 31812 + }, + { + "epoch": 6.94, + "learning_rate": 1.5301507537688443e-06, + "loss": 1.6239, + "step": 31813 + }, + { + "epoch": 6.94, + "learning_rate": 1.530041511907363e-06, + "loss": 2.4663, + "step": 31814 + }, + { + "epoch": 6.94, + "learning_rate": 1.5299322700458816e-06, + "loss": 1.2165, + "step": 31815 + }, + { + "epoch": 6.94, + "learning_rate": 1.5298230281844004e-06, + "loss": 0.1467, + "step": 31816 + }, + { + "epoch": 6.94, + "learning_rate": 1.5297137863229192e-06, + "loss": 1.1344, + "step": 31817 + }, + { + "epoch": 6.94, + "learning_rate": 1.5296045444614376e-06, + "loss": 0.3826, + "step": 31818 + }, + { + "epoch": 6.94, + "learning_rate": 1.5294953025999564e-06, + "loss": 1.963, + "step": 31819 + }, + { + "epoch": 6.94, + "learning_rate": 1.529386060738475e-06, + "loss": 0.4951, + "step": 31820 + }, + { + "epoch": 6.94, + "learning_rate": 1.5292768188769939e-06, + "loss": 1.1907, + "step": 31821 + }, + { + "epoch": 6.95, + "learning_rate": 1.5291675770155127e-06, + "loss": 0.8884, + "step": 31822 + }, + { + "epoch": 6.95, + "learning_rate": 1.529058335154031e-06, + "loss": 0.6447, + "step": 31823 + }, + { + "epoch": 6.95, + "learning_rate": 1.52894909329255e-06, + "loss": 0.9863, + "step": 31824 + }, + { + "epoch": 6.95, + "learning_rate": 1.5288398514310685e-06, + "loss": 1.0096, + "step": 31825 + }, + { + "epoch": 6.95, + "learning_rate": 1.5287306095695871e-06, + "loss": 0.2036, + "step": 31826 + }, + { + "epoch": 6.95, + "learning_rate": 1.5286213677081057e-06, + "loss": 0.1519, + "step": 31827 + }, + { + "epoch": 6.95, + "learning_rate": 1.5285121258466246e-06, + "loss": 1.3124, + "step": 31828 + }, + { + "epoch": 6.95, + "learning_rate": 1.5284028839851434e-06, + "loss": 2.3588, + "step": 31829 + }, + { + "epoch": 6.95, + "learning_rate": 1.5282936421236618e-06, + "loss": 1.2198, + "step": 31830 + }, + { + "epoch": 6.95, + "learning_rate": 1.5281844002621806e-06, + "loss": 3.0505, + "step": 31831 + }, + { + "epoch": 6.95, + "learning_rate": 1.5280751584006992e-06, + "loss": 0.5288, + "step": 31832 + }, + { + "epoch": 6.95, + "learning_rate": 1.527965916539218e-06, + "loss": 0.4634, + "step": 31833 + }, + { + "epoch": 6.95, + "learning_rate": 1.5278566746777364e-06, + "loss": 1.7429, + "step": 31834 + }, + { + "epoch": 6.95, + "learning_rate": 1.5277474328162553e-06, + "loss": 0.7311, + "step": 31835 + }, + { + "epoch": 6.95, + "learning_rate": 1.527638190954774e-06, + "loss": 2.0664, + "step": 31836 + }, + { + "epoch": 6.95, + "learning_rate": 1.5275289490932927e-06, + "loss": 0.9859, + "step": 31837 + }, + { + "epoch": 6.95, + "learning_rate": 1.5274197072318115e-06, + "loss": 1.7188, + "step": 31838 + }, + { + "epoch": 6.95, + "learning_rate": 1.52731046537033e-06, + "loss": 0.2272, + "step": 31839 + }, + { + "epoch": 6.95, + "learning_rate": 1.5272012235088487e-06, + "loss": 0.6568, + "step": 31840 + }, + { + "epoch": 6.95, + "learning_rate": 1.5270919816473674e-06, + "loss": 1.1304, + "step": 31841 + }, + { + "epoch": 6.95, + "learning_rate": 1.526982739785886e-06, + "loss": 1.0706, + "step": 31842 + }, + { + "epoch": 6.95, + "learning_rate": 1.5268734979244048e-06, + "loss": 1.1849, + "step": 31843 + }, + { + "epoch": 6.95, + "learning_rate": 1.5267642560629234e-06, + "loss": 2.0025, + "step": 31844 + }, + { + "epoch": 6.95, + "learning_rate": 1.5266550142014422e-06, + "loss": 1.5016, + "step": 31845 + }, + { + "epoch": 6.95, + "learning_rate": 1.5265457723399606e-06, + "loss": 0.2616, + "step": 31846 + }, + { + "epoch": 6.95, + "learning_rate": 1.5264365304784795e-06, + "loss": 1.7447, + "step": 31847 + }, + { + "epoch": 6.95, + "learning_rate": 1.526327288616998e-06, + "loss": 0.4033, + "step": 31848 + }, + { + "epoch": 6.95, + "learning_rate": 1.5262180467555169e-06, + "loss": 0.4553, + "step": 31849 + }, + { + "epoch": 6.95, + "learning_rate": 1.5261088048940357e-06, + "loss": 1.3825, + "step": 31850 + }, + { + "epoch": 6.95, + "learning_rate": 1.525999563032554e-06, + "loss": 0.1672, + "step": 31851 + }, + { + "epoch": 6.95, + "learning_rate": 1.525890321171073e-06, + "loss": 1.5582, + "step": 31852 + }, + { + "epoch": 6.95, + "learning_rate": 1.5257810793095915e-06, + "loss": 0.0866, + "step": 31853 + }, + { + "epoch": 6.95, + "learning_rate": 1.5256718374481104e-06, + "loss": 0.1048, + "step": 31854 + }, + { + "epoch": 6.95, + "learning_rate": 1.525562595586629e-06, + "loss": 0.0891, + "step": 31855 + }, + { + "epoch": 6.95, + "learning_rate": 1.5254533537251476e-06, + "loss": 0.5978, + "step": 31856 + }, + { + "epoch": 6.95, + "learning_rate": 1.5253441118636664e-06, + "loss": 1.7045, + "step": 31857 + }, + { + "epoch": 6.95, + "learning_rate": 1.5252348700021848e-06, + "loss": 2.8329, + "step": 31858 + }, + { + "epoch": 6.95, + "learning_rate": 1.5251256281407036e-06, + "loss": 4.1922, + "step": 31859 + }, + { + "epoch": 6.95, + "learning_rate": 1.5250163862792222e-06, + "loss": 0.9353, + "step": 31860 + }, + { + "epoch": 6.95, + "learning_rate": 1.524907144417741e-06, + "loss": 0.0592, + "step": 31861 + }, + { + "epoch": 6.95, + "learning_rate": 1.5247979025562599e-06, + "loss": 1.5209, + "step": 31862 + }, + { + "epoch": 6.95, + "learning_rate": 1.5246886606947783e-06, + "loss": 0.5477, + "step": 31863 + }, + { + "epoch": 6.95, + "learning_rate": 1.5245794188332971e-06, + "loss": 1.8297, + "step": 31864 + }, + { + "epoch": 6.95, + "learning_rate": 1.5244701769718157e-06, + "loss": 1.5386, + "step": 31865 + }, + { + "epoch": 6.95, + "learning_rate": 1.5243609351103345e-06, + "loss": 1.5245, + "step": 31866 + }, + { + "epoch": 6.95, + "learning_rate": 1.524251693248853e-06, + "loss": 0.9307, + "step": 31867 + }, + { + "epoch": 6.96, + "learning_rate": 1.5241424513873718e-06, + "loss": 1.5287, + "step": 31868 + }, + { + "epoch": 6.96, + "learning_rate": 1.5240332095258906e-06, + "loss": 1.8815, + "step": 31869 + }, + { + "epoch": 6.96, + "learning_rate": 1.523923967664409e-06, + "loss": 1.2983, + "step": 31870 + }, + { + "epoch": 6.96, + "learning_rate": 1.5238147258029278e-06, + "loss": 2.408, + "step": 31871 + }, + { + "epoch": 6.96, + "learning_rate": 1.5237054839414464e-06, + "loss": 1.5396, + "step": 31872 + }, + { + "epoch": 6.96, + "learning_rate": 1.5235962420799652e-06, + "loss": 1.5278, + "step": 31873 + }, + { + "epoch": 6.96, + "learning_rate": 1.5234870002184836e-06, + "loss": 2.0284, + "step": 31874 + }, + { + "epoch": 6.96, + "learning_rate": 1.5233777583570025e-06, + "loss": 1.798, + "step": 31875 + }, + { + "epoch": 6.96, + "learning_rate": 1.5232685164955213e-06, + "loss": 1.9926, + "step": 31876 + }, + { + "epoch": 6.96, + "learning_rate": 1.52315927463404e-06, + "loss": 2.7017, + "step": 31877 + }, + { + "epoch": 6.96, + "learning_rate": 1.5230500327725587e-06, + "loss": 2.0468, + "step": 31878 + }, + { + "epoch": 6.96, + "learning_rate": 1.5229407909110771e-06, + "loss": 1.5143, + "step": 31879 + }, + { + "epoch": 6.96, + "learning_rate": 1.522831549049596e-06, + "loss": 1.2502, + "step": 31880 + }, + { + "epoch": 6.96, + "learning_rate": 1.5227223071881146e-06, + "loss": 0.2835, + "step": 31881 + }, + { + "epoch": 6.96, + "learning_rate": 1.5226130653266334e-06, + "loss": 1.4477, + "step": 31882 + }, + { + "epoch": 6.96, + "learning_rate": 1.522503823465152e-06, + "loss": 1.6054, + "step": 31883 + }, + { + "epoch": 6.96, + "learning_rate": 1.5223945816036706e-06, + "loss": 0.8448, + "step": 31884 + }, + { + "epoch": 6.96, + "learning_rate": 1.5222853397421894e-06, + "loss": 1.7627, + "step": 31885 + }, + { + "epoch": 6.96, + "learning_rate": 1.5221760978807078e-06, + "loss": 2.4375, + "step": 31886 + }, + { + "epoch": 6.96, + "learning_rate": 1.5220668560192266e-06, + "loss": 1.4646, + "step": 31887 + }, + { + "epoch": 6.96, + "learning_rate": 1.5219576141577455e-06, + "loss": 0.3899, + "step": 31888 + }, + { + "epoch": 6.96, + "learning_rate": 1.521848372296264e-06, + "loss": 0.9116, + "step": 31889 + }, + { + "epoch": 6.96, + "learning_rate": 1.521739130434783e-06, + "loss": 0.8844, + "step": 31890 + }, + { + "epoch": 6.96, + "learning_rate": 1.5216298885733013e-06, + "loss": 0.8681, + "step": 31891 + }, + { + "epoch": 6.96, + "learning_rate": 1.5215206467118201e-06, + "loss": 0.7833, + "step": 31892 + }, + { + "epoch": 6.96, + "learning_rate": 1.5214114048503387e-06, + "loss": 1.7074, + "step": 31893 + }, + { + "epoch": 6.96, + "learning_rate": 1.5213021629888576e-06, + "loss": 0.931, + "step": 31894 + }, + { + "epoch": 6.96, + "learning_rate": 1.5211929211273762e-06, + "loss": 3.918, + "step": 31895 + }, + { + "epoch": 6.96, + "learning_rate": 1.5210836792658948e-06, + "loss": 1.2405, + "step": 31896 + }, + { + "epoch": 6.96, + "learning_rate": 1.5209744374044136e-06, + "loss": 0.8019, + "step": 31897 + }, + { + "epoch": 6.96, + "learning_rate": 1.5208651955429322e-06, + "loss": 2.2774, + "step": 31898 + }, + { + "epoch": 6.96, + "learning_rate": 1.5207559536814508e-06, + "loss": 0.5945, + "step": 31899 + }, + { + "epoch": 6.96, + "learning_rate": 1.5206467118199694e-06, + "loss": 2.1715, + "step": 31900 + }, + { + "epoch": 6.96, + "learning_rate": 1.5205374699584883e-06, + "loss": 0.5973, + "step": 31901 + }, + { + "epoch": 6.96, + "learning_rate": 1.520428228097007e-06, + "loss": 0.7222, + "step": 31902 + }, + { + "epoch": 6.96, + "learning_rate": 1.5203189862355255e-06, + "loss": 1.2293, + "step": 31903 + }, + { + "epoch": 6.96, + "learning_rate": 1.5202097443740443e-06, + "loss": 0.9707, + "step": 31904 + }, + { + "epoch": 6.96, + "learning_rate": 1.520100502512563e-06, + "loss": 0.9359, + "step": 31905 + }, + { + "epoch": 6.96, + "learning_rate": 1.5199912606510817e-06, + "loss": 2.0789, + "step": 31906 + }, + { + "epoch": 6.96, + "learning_rate": 1.5198820187896001e-06, + "loss": 1.5571, + "step": 31907 + }, + { + "epoch": 6.96, + "learning_rate": 1.519772776928119e-06, + "loss": 1.6564, + "step": 31908 + }, + { + "epoch": 6.96, + "learning_rate": 1.5196635350666378e-06, + "loss": 1.4672, + "step": 31909 + }, + { + "epoch": 6.96, + "learning_rate": 1.5195542932051564e-06, + "loss": 1.8085, + "step": 31910 + }, + { + "epoch": 6.96, + "learning_rate": 1.519445051343675e-06, + "loss": 0.6041, + "step": 31911 + }, + { + "epoch": 6.96, + "learning_rate": 1.5193358094821936e-06, + "loss": 1.4736, + "step": 31912 + }, + { + "epoch": 6.96, + "learning_rate": 1.5192265676207124e-06, + "loss": 1.3328, + "step": 31913 + }, + { + "epoch": 6.97, + "learning_rate": 1.5191173257592308e-06, + "loss": 1.2329, + "step": 31914 + }, + { + "epoch": 6.97, + "learning_rate": 1.5190080838977497e-06, + "loss": 0.7333, + "step": 31915 + }, + { + "epoch": 6.97, + "learning_rate": 1.5188988420362685e-06, + "loss": 1.6484, + "step": 31916 + }, + { + "epoch": 6.97, + "learning_rate": 1.518789600174787e-06, + "loss": 0.205, + "step": 31917 + }, + { + "epoch": 6.97, + "learning_rate": 1.518680358313306e-06, + "loss": 0.1509, + "step": 31918 + }, + { + "epoch": 6.97, + "learning_rate": 1.5185711164518243e-06, + "loss": 0.6055, + "step": 31919 + }, + { + "epoch": 6.97, + "learning_rate": 1.5184618745903431e-06, + "loss": 0.7103, + "step": 31920 + }, + { + "epoch": 6.97, + "learning_rate": 1.5183526327288617e-06, + "loss": 1.317, + "step": 31921 + }, + { + "epoch": 6.97, + "learning_rate": 1.5182433908673806e-06, + "loss": 2.1237, + "step": 31922 + }, + { + "epoch": 6.97, + "learning_rate": 1.5181341490058992e-06, + "loss": 0.7716, + "step": 31923 + }, + { + "epoch": 6.97, + "learning_rate": 1.5180249071444178e-06, + "loss": 0.1542, + "step": 31924 + }, + { + "epoch": 6.97, + "learning_rate": 1.5179156652829366e-06, + "loss": 2.0748, + "step": 31925 + }, + { + "epoch": 6.97, + "learning_rate": 1.5178064234214552e-06, + "loss": 0.2738, + "step": 31926 + }, + { + "epoch": 6.97, + "learning_rate": 1.5176971815599738e-06, + "loss": 0.7184, + "step": 31927 + }, + { + "epoch": 6.97, + "learning_rate": 1.5175879396984927e-06, + "loss": 0.9184, + "step": 31928 + }, + { + "epoch": 6.97, + "learning_rate": 1.5174786978370113e-06, + "loss": 1.5542, + "step": 31929 + }, + { + "epoch": 6.97, + "learning_rate": 1.51736945597553e-06, + "loss": 2.0865, + "step": 31930 + }, + { + "epoch": 6.97, + "learning_rate": 1.5172602141140485e-06, + "loss": 1.6832, + "step": 31931 + }, + { + "epoch": 6.97, + "learning_rate": 1.5171509722525673e-06, + "loss": 0.1557, + "step": 31932 + }, + { + "epoch": 6.97, + "learning_rate": 1.517041730391086e-06, + "loss": 0.0936, + "step": 31933 + }, + { + "epoch": 6.97, + "learning_rate": 1.5169324885296047e-06, + "loss": 2.3714, + "step": 31934 + }, + { + "epoch": 6.97, + "learning_rate": 1.5168232466681236e-06, + "loss": 1.4256, + "step": 31935 + }, + { + "epoch": 6.97, + "learning_rate": 1.516714004806642e-06, + "loss": 1.5894, + "step": 31936 + }, + { + "epoch": 6.97, + "learning_rate": 1.5166047629451608e-06, + "loss": 0.9513, + "step": 31937 + }, + { + "epoch": 6.97, + "learning_rate": 1.5164955210836794e-06, + "loss": 1.603, + "step": 31938 + }, + { + "epoch": 6.97, + "learning_rate": 1.516386279222198e-06, + "loss": 0.511, + "step": 31939 + }, + { + "epoch": 6.97, + "learning_rate": 1.5162770373607166e-06, + "loss": 0.4009, + "step": 31940 + }, + { + "epoch": 6.97, + "learning_rate": 1.5161677954992354e-06, + "loss": 0.7861, + "step": 31941 + }, + { + "epoch": 6.97, + "learning_rate": 1.5160585536377543e-06, + "loss": 0.8947, + "step": 31942 + }, + { + "epoch": 6.97, + "learning_rate": 1.5159493117762727e-06, + "loss": 1.4001, + "step": 31943 + }, + { + "epoch": 6.97, + "learning_rate": 1.5158400699147915e-06, + "loss": 1.1604, + "step": 31944 + }, + { + "epoch": 6.97, + "learning_rate": 1.51573082805331e-06, + "loss": 1.0106, + "step": 31945 + }, + { + "epoch": 6.97, + "learning_rate": 1.515621586191829e-06, + "loss": 1.8628, + "step": 31946 + }, + { + "epoch": 6.97, + "learning_rate": 1.5155123443303473e-06, + "loss": 0.8697, + "step": 31947 + }, + { + "epoch": 6.97, + "learning_rate": 1.5154031024688661e-06, + "loss": 1.145, + "step": 31948 + }, + { + "epoch": 6.97, + "learning_rate": 1.515293860607385e-06, + "loss": 3.03, + "step": 31949 + }, + { + "epoch": 6.97, + "learning_rate": 1.5151846187459036e-06, + "loss": 0.9274, + "step": 31950 + }, + { + "epoch": 6.97, + "learning_rate": 1.5150753768844224e-06, + "loss": 0.5045, + "step": 31951 + }, + { + "epoch": 6.97, + "learning_rate": 1.5149661350229408e-06, + "loss": 1.654, + "step": 31952 + }, + { + "epoch": 6.97, + "learning_rate": 1.5148568931614596e-06, + "loss": 0.6786, + "step": 31953 + }, + { + "epoch": 6.97, + "learning_rate": 1.5147476512999782e-06, + "loss": 1.0103, + "step": 31954 + }, + { + "epoch": 6.97, + "learning_rate": 1.5146384094384968e-06, + "loss": 1.8566, + "step": 31955 + }, + { + "epoch": 6.97, + "learning_rate": 1.5145291675770157e-06, + "loss": 1.4076, + "step": 31956 + }, + { + "epoch": 6.97, + "learning_rate": 1.5144199257155343e-06, + "loss": 0.9458, + "step": 31957 + }, + { + "epoch": 6.97, + "learning_rate": 1.514310683854053e-06, + "loss": 1.0101, + "step": 31958 + }, + { + "epoch": 6.97, + "learning_rate": 1.5142014419925715e-06, + "loss": 0.8549, + "step": 31959 + }, + { + "epoch": 6.98, + "learning_rate": 1.5140922001310903e-06, + "loss": 0.9003, + "step": 31960 + }, + { + "epoch": 6.98, + "learning_rate": 1.5139829582696091e-06, + "loss": 0.9055, + "step": 31961 + }, + { + "epoch": 6.98, + "learning_rate": 1.5138737164081278e-06, + "loss": 1.1138, + "step": 31962 + }, + { + "epoch": 6.98, + "learning_rate": 1.5137644745466466e-06, + "loss": 0.5995, + "step": 31963 + }, + { + "epoch": 6.98, + "learning_rate": 1.513655232685165e-06, + "loss": 1.1435, + "step": 31964 + }, + { + "epoch": 6.98, + "learning_rate": 1.5135459908236838e-06, + "loss": 2.9094, + "step": 31965 + }, + { + "epoch": 6.98, + "learning_rate": 1.5134367489622024e-06, + "loss": 1.815, + "step": 31966 + }, + { + "epoch": 6.98, + "learning_rate": 1.513327507100721e-06, + "loss": 0.5481, + "step": 31967 + }, + { + "epoch": 6.98, + "learning_rate": 1.5132182652392398e-06, + "loss": 0.1491, + "step": 31968 + }, + { + "epoch": 6.98, + "learning_rate": 1.5131090233777585e-06, + "loss": 0.3712, + "step": 31969 + }, + { + "epoch": 6.98, + "learning_rate": 1.5129997815162773e-06, + "loss": 0.1091, + "step": 31970 + }, + { + "epoch": 6.98, + "learning_rate": 1.5128905396547957e-06, + "loss": 0.5828, + "step": 31971 + }, + { + "epoch": 6.98, + "learning_rate": 1.5127812977933145e-06, + "loss": 0.8597, + "step": 31972 + }, + { + "epoch": 6.98, + "learning_rate": 1.5126720559318331e-06, + "loss": 1.1523, + "step": 31973 + }, + { + "epoch": 6.98, + "learning_rate": 1.512562814070352e-06, + "loss": 0.8018, + "step": 31974 + }, + { + "epoch": 6.98, + "learning_rate": 1.5124535722088708e-06, + "loss": 1.4352, + "step": 31975 + }, + { + "epoch": 6.98, + "learning_rate": 1.5123443303473892e-06, + "loss": 1.3886, + "step": 31976 + }, + { + "epoch": 6.98, + "learning_rate": 1.512235088485908e-06, + "loss": 2.5384, + "step": 31977 + }, + { + "epoch": 6.98, + "learning_rate": 1.5121258466244266e-06, + "loss": 1.5893, + "step": 31978 + }, + { + "epoch": 6.98, + "learning_rate": 1.5120166047629454e-06, + "loss": 0.7534, + "step": 31979 + }, + { + "epoch": 6.98, + "learning_rate": 1.5119073629014638e-06, + "loss": 1.2642, + "step": 31980 + }, + { + "epoch": 6.98, + "learning_rate": 1.5117981210399826e-06, + "loss": 1.7678, + "step": 31981 + }, + { + "epoch": 6.98, + "learning_rate": 1.5116888791785015e-06, + "loss": 0.6476, + "step": 31982 + }, + { + "epoch": 6.98, + "learning_rate": 1.5115796373170199e-06, + "loss": 0.1931, + "step": 31983 + }, + { + "epoch": 6.98, + "learning_rate": 1.5114703954555387e-06, + "loss": 1.4171, + "step": 31984 + }, + { + "epoch": 6.98, + "learning_rate": 1.5113611535940573e-06, + "loss": 1.6111, + "step": 31985 + }, + { + "epoch": 6.98, + "learning_rate": 1.5112519117325761e-06, + "loss": 1.7384, + "step": 31986 + }, + { + "epoch": 6.98, + "learning_rate": 1.5111426698710945e-06, + "loss": 1.777, + "step": 31987 + }, + { + "epoch": 6.98, + "learning_rate": 1.5110334280096133e-06, + "loss": 1.0429, + "step": 31988 + }, + { + "epoch": 6.98, + "learning_rate": 1.5109241861481322e-06, + "loss": 0.8454, + "step": 31989 + }, + { + "epoch": 6.98, + "learning_rate": 1.5108149442866508e-06, + "loss": 0.9518, + "step": 31990 + }, + { + "epoch": 6.98, + "learning_rate": 1.5107057024251696e-06, + "loss": 2.3961, + "step": 31991 + }, + { + "epoch": 6.98, + "learning_rate": 1.510596460563688e-06, + "loss": 0.3999, + "step": 31992 + }, + { + "epoch": 6.98, + "learning_rate": 1.5104872187022068e-06, + "loss": 0.8501, + "step": 31993 + }, + { + "epoch": 6.98, + "learning_rate": 1.5103779768407256e-06, + "loss": 2.3219, + "step": 31994 + }, + { + "epoch": 6.98, + "learning_rate": 1.510268734979244e-06, + "loss": 0.8595, + "step": 31995 + }, + { + "epoch": 6.98, + "learning_rate": 1.5101594931177629e-06, + "loss": 1.7653, + "step": 31996 + }, + { + "epoch": 6.98, + "learning_rate": 1.5100502512562815e-06, + "loss": 2.4248, + "step": 31997 + }, + { + "epoch": 6.98, + "learning_rate": 1.5099410093948003e-06, + "loss": 1.8003, + "step": 31998 + }, + { + "epoch": 6.98, + "learning_rate": 1.5098317675333187e-06, + "loss": 1.6322, + "step": 31999 + }, + { + "epoch": 6.98, + "learning_rate": 1.5097225256718375e-06, + "loss": 2.3482, + "step": 32000 + }, + { + "epoch": 6.98, + "learning_rate": 1.5096132838103563e-06, + "loss": 0.4375, + "step": 32001 + }, + { + "epoch": 6.98, + "learning_rate": 1.509504041948875e-06, + "loss": 0.1573, + "step": 32002 + }, + { + "epoch": 6.98, + "learning_rate": 1.5093948000873938e-06, + "loss": 1.4058, + "step": 32003 + }, + { + "epoch": 6.98, + "learning_rate": 1.5092855582259122e-06, + "loss": 1.3269, + "step": 32004 + }, + { + "epoch": 6.98, + "learning_rate": 1.509176316364431e-06, + "loss": 1.5982, + "step": 32005 + }, + { + "epoch": 6.99, + "learning_rate": 1.5090670745029496e-06, + "loss": 2.887, + "step": 32006 + }, + { + "epoch": 6.99, + "learning_rate": 1.5089578326414684e-06, + "loss": 1.4388, + "step": 32007 + }, + { + "epoch": 6.99, + "learning_rate": 1.508848590779987e-06, + "loss": 0.6204, + "step": 32008 + }, + { + "epoch": 6.99, + "learning_rate": 1.5087393489185057e-06, + "loss": 0.1829, + "step": 32009 + }, + { + "epoch": 6.99, + "learning_rate": 1.5086301070570245e-06, + "loss": 1.6605, + "step": 32010 + }, + { + "epoch": 6.99, + "learning_rate": 1.5085208651955429e-06, + "loss": 2.0749, + "step": 32011 + }, + { + "epoch": 6.99, + "learning_rate": 1.5084116233340617e-06, + "loss": 2.0206, + "step": 32012 + }, + { + "epoch": 6.99, + "learning_rate": 1.5083023814725803e-06, + "loss": 0.9751, + "step": 32013 + }, + { + "epoch": 6.99, + "learning_rate": 1.5081931396110991e-06, + "loss": 0.6287, + "step": 32014 + }, + { + "epoch": 6.99, + "learning_rate": 1.508083897749618e-06, + "loss": 0.224, + "step": 32015 + }, + { + "epoch": 6.99, + "learning_rate": 1.5079746558881364e-06, + "loss": 1.3198, + "step": 32016 + }, + { + "epoch": 6.99, + "learning_rate": 1.5078654140266552e-06, + "loss": 1.4653, + "step": 32017 + }, + { + "epoch": 6.99, + "learning_rate": 1.5077561721651738e-06, + "loss": 0.9118, + "step": 32018 + }, + { + "epoch": 6.99, + "learning_rate": 1.5076469303036926e-06, + "loss": 0.3237, + "step": 32019 + }, + { + "epoch": 6.99, + "learning_rate": 1.507537688442211e-06, + "loss": 2.9962, + "step": 32020 + }, + { + "epoch": 6.99, + "learning_rate": 1.5074284465807298e-06, + "loss": 0.1494, + "step": 32021 + }, + { + "epoch": 6.99, + "learning_rate": 1.5073192047192487e-06, + "loss": 1.3096, + "step": 32022 + }, + { + "epoch": 6.99, + "learning_rate": 1.5072099628577673e-06, + "loss": 1.3295, + "step": 32023 + }, + { + "epoch": 6.99, + "learning_rate": 1.5071007209962859e-06, + "loss": 1.3184, + "step": 32024 + }, + { + "epoch": 6.99, + "learning_rate": 1.5069914791348045e-06, + "loss": 0.9292, + "step": 32025 + }, + { + "epoch": 6.99, + "learning_rate": 1.5068822372733233e-06, + "loss": 1.2194, + "step": 32026 + }, + { + "epoch": 6.99, + "learning_rate": 1.5067729954118421e-06, + "loss": 1.1339, + "step": 32027 + }, + { + "epoch": 6.99, + "learning_rate": 1.5066637535503605e-06, + "loss": 0.8136, + "step": 32028 + }, + { + "epoch": 6.99, + "learning_rate": 1.5065545116888794e-06, + "loss": 0.3466, + "step": 32029 + }, + { + "epoch": 6.99, + "learning_rate": 1.506445269827398e-06, + "loss": 0.8653, + "step": 32030 + }, + { + "epoch": 6.99, + "learning_rate": 1.5063360279659168e-06, + "loss": 1.1436, + "step": 32031 + }, + { + "epoch": 6.99, + "learning_rate": 1.5062267861044352e-06, + "loss": 1.9722, + "step": 32032 + }, + { + "epoch": 6.99, + "learning_rate": 1.506117544242954e-06, + "loss": 2.9945, + "step": 32033 + }, + { + "epoch": 6.99, + "learning_rate": 1.5060083023814728e-06, + "loss": 1.5846, + "step": 32034 + }, + { + "epoch": 6.99, + "learning_rate": 1.5058990605199914e-06, + "loss": 2.1644, + "step": 32035 + }, + { + "epoch": 6.99, + "learning_rate": 1.50578981865851e-06, + "loss": 1.776, + "step": 32036 + }, + { + "epoch": 6.99, + "learning_rate": 1.5056805767970287e-06, + "loss": 1.6506, + "step": 32037 + }, + { + "epoch": 6.99, + "learning_rate": 1.5055713349355475e-06, + "loss": 1.1966, + "step": 32038 + }, + { + "epoch": 6.99, + "learning_rate": 1.5054620930740659e-06, + "loss": 1.998, + "step": 32039 + }, + { + "epoch": 6.99, + "learning_rate": 1.5053528512125847e-06, + "loss": 0.0486, + "step": 32040 + }, + { + "epoch": 6.99, + "learning_rate": 1.5052436093511035e-06, + "loss": 0.3405, + "step": 32041 + }, + { + "epoch": 6.99, + "learning_rate": 1.5051343674896221e-06, + "loss": 2.5911, + "step": 32042 + }, + { + "epoch": 6.99, + "learning_rate": 1.505025125628141e-06, + "loss": 1.4005, + "step": 32043 + }, + { + "epoch": 6.99, + "learning_rate": 1.5049158837666594e-06, + "loss": 1.9165, + "step": 32044 + }, + { + "epoch": 6.99, + "learning_rate": 1.5048066419051782e-06, + "loss": 0.8442, + "step": 32045 + }, + { + "epoch": 6.99, + "learning_rate": 1.5046974000436968e-06, + "loss": 1.1426, + "step": 32046 + }, + { + "epoch": 6.99, + "learning_rate": 1.5045881581822156e-06, + "loss": 1.5804, + "step": 32047 + }, + { + "epoch": 6.99, + "learning_rate": 1.5044789163207342e-06, + "loss": 1.3872, + "step": 32048 + }, + { + "epoch": 6.99, + "learning_rate": 1.5043696744592528e-06, + "loss": 0.6769, + "step": 32049 + }, + { + "epoch": 6.99, + "learning_rate": 1.5042604325977717e-06, + "loss": 0.5816, + "step": 32050 + }, + { + "epoch": 6.99, + "learning_rate": 1.5041511907362903e-06, + "loss": 1.6596, + "step": 32051 + }, + { + "epoch": 7.0, + "learning_rate": 1.5040419488748089e-06, + "loss": 1.0465, + "step": 32052 + }, + { + "epoch": 7.0, + "learning_rate": 1.5039327070133275e-06, + "loss": 0.284, + "step": 32053 + }, + { + "epoch": 7.0, + "learning_rate": 1.5038234651518463e-06, + "loss": 2.0463, + "step": 32054 + }, + { + "epoch": 7.0, + "learning_rate": 1.5037142232903651e-06, + "loss": 0.4761, + "step": 32055 + }, + { + "epoch": 7.0, + "learning_rate": 1.5036049814288835e-06, + "loss": 0.3564, + "step": 32056 + }, + { + "epoch": 7.0, + "learning_rate": 1.5034957395674024e-06, + "loss": 2.4768, + "step": 32057 + }, + { + "epoch": 7.0, + "learning_rate": 1.503386497705921e-06, + "loss": 0.5417, + "step": 32058 + }, + { + "epoch": 7.0, + "learning_rate": 1.5032772558444398e-06, + "loss": 2.001, + "step": 32059 + }, + { + "epoch": 7.0, + "learning_rate": 1.5031680139829586e-06, + "loss": 0.1071, + "step": 32060 + }, + { + "epoch": 7.0, + "learning_rate": 1.503058772121477e-06, + "loss": 0.6432, + "step": 32061 + }, + { + "epoch": 7.0, + "learning_rate": 1.5029495302599958e-06, + "loss": 1.4462, + "step": 32062 + }, + { + "epoch": 7.0, + "learning_rate": 1.5028402883985145e-06, + "loss": 1.3865, + "step": 32063 + }, + { + "epoch": 7.0, + "learning_rate": 1.502731046537033e-06, + "loss": 1.0353, + "step": 32064 + }, + { + "epoch": 7.0, + "learning_rate": 1.5026218046755517e-06, + "loss": 0.7049, + "step": 32065 + }, + { + "epoch": 7.0, + "learning_rate": 1.5025125628140705e-06, + "loss": 0.7723, + "step": 32066 + }, + { + "epoch": 7.0, + "learning_rate": 1.5024033209525893e-06, + "loss": 1.4966, + "step": 32067 + }, + { + "epoch": 7.0, + "learning_rate": 1.5022940790911077e-06, + "loss": 1.467, + "step": 32068 + }, + { + "epoch": 7.0, + "learning_rate": 1.5021848372296265e-06, + "loss": 2.1717, + "step": 32069 + }, + { + "epoch": 7.0, + "learning_rate": 1.5020755953681452e-06, + "loss": 1.9127, + "step": 32070 + }, + { + "epoch": 7.0, + "learning_rate": 1.501966353506664e-06, + "loss": 1.3548, + "step": 32071 + }, + { + "epoch": 7.0, + "learning_rate": 1.5018571116451824e-06, + "loss": 0.2169, + "step": 32072 + }, + { + "epoch": 7.0, + "learning_rate": 1.5017478697837012e-06, + "loss": 0.9019, + "step": 32073 + }, + { + "epoch": 7.0, + "learning_rate": 1.50163862792222e-06, + "loss": 1.2869, + "step": 32074 + }, + { + "epoch": 7.0, + "eval_accuracy": 0.6898527004909983, + "eval_loss": 1.2499380111694336, + "eval_runtime": 337.0829, + "eval_samples_per_second": 18.126, + "eval_steps_per_second": 4.533, + "step": 32074 + }, + { + "epoch": 7.0, + "learning_rate": 1.5015293860607386e-06, + "loss": 2.8022, + "step": 32075 + }, + { + "epoch": 7.0, + "learning_rate": 1.5014201441992575e-06, + "loss": 1.1073, + "step": 32076 + }, + { + "epoch": 7.0, + "learning_rate": 1.5013109023377759e-06, + "loss": 0.5094, + "step": 32077 + }, + { + "epoch": 7.0, + "learning_rate": 1.5012016604762947e-06, + "loss": 0.8281, + "step": 32078 + }, + { + "epoch": 7.0, + "learning_rate": 1.5010924186148133e-06, + "loss": 1.2156, + "step": 32079 + }, + { + "epoch": 7.0, + "learning_rate": 1.500983176753332e-06, + "loss": 0.6895, + "step": 32080 + }, + { + "epoch": 7.0, + "learning_rate": 1.5008739348918507e-06, + "loss": 0.3171, + "step": 32081 + }, + { + "epoch": 7.0, + "learning_rate": 1.5007646930303693e-06, + "loss": 0.803, + "step": 32082 + }, + { + "epoch": 7.0, + "learning_rate": 1.5006554511688882e-06, + "loss": 1.1984, + "step": 32083 + }, + { + "epoch": 7.0, + "learning_rate": 1.5005462093074066e-06, + "loss": 1.3666, + "step": 32084 + }, + { + "epoch": 7.0, + "learning_rate": 1.5004369674459254e-06, + "loss": 0.2043, + "step": 32085 + }, + { + "epoch": 7.0, + "learning_rate": 1.500327725584444e-06, + "loss": 0.1871, + "step": 32086 + }, + { + "epoch": 7.0, + "learning_rate": 1.5002184837229628e-06, + "loss": 0.9851, + "step": 32087 + }, + { + "epoch": 7.0, + "learning_rate": 1.5001092418614816e-06, + "loss": 1.8329, + "step": 32088 + }, + { + "epoch": 7.0, + "learning_rate": 1.5e-06, + "loss": 1.139, + "step": 32089 + }, + { + "epoch": 7.0, + "learning_rate": 1.4998907581385189e-06, + "loss": 1.244, + "step": 32090 + }, + { + "epoch": 7.0, + "learning_rate": 1.4997815162770375e-06, + "loss": 0.9962, + "step": 32091 + }, + { + "epoch": 7.0, + "learning_rate": 1.499672274415556e-06, + "loss": 0.2753, + "step": 32092 + }, + { + "epoch": 7.0, + "learning_rate": 1.499563032554075e-06, + "loss": 0.4087, + "step": 32093 + }, + { + "epoch": 7.0, + "learning_rate": 1.4994537906925935e-06, + "loss": 2.0207, + "step": 32094 + }, + { + "epoch": 7.0, + "learning_rate": 1.4993445488311123e-06, + "loss": 0.8412, + "step": 32095 + }, + { + "epoch": 7.0, + "learning_rate": 1.4992353069696307e-06, + "loss": 1.2069, + "step": 32096 + }, + { + "epoch": 7.01, + "learning_rate": 1.4991260651081496e-06, + "loss": 1.0317, + "step": 32097 + }, + { + "epoch": 7.01, + "learning_rate": 1.4990168232466682e-06, + "loss": 1.9884, + "step": 32098 + }, + { + "epoch": 7.01, + "learning_rate": 1.498907581385187e-06, + "loss": 1.2259, + "step": 32099 + }, + { + "epoch": 7.01, + "learning_rate": 1.4987983395237058e-06, + "loss": 0.5461, + "step": 32100 + }, + { + "epoch": 7.01, + "learning_rate": 1.4986890976622242e-06, + "loss": 1.4353, + "step": 32101 + }, + { + "epoch": 7.01, + "learning_rate": 1.498579855800743e-06, + "loss": 1.3734, + "step": 32102 + }, + { + "epoch": 7.01, + "learning_rate": 1.4984706139392617e-06, + "loss": 0.8884, + "step": 32103 + }, + { + "epoch": 7.01, + "learning_rate": 1.4983613720777805e-06, + "loss": 1.1312, + "step": 32104 + }, + { + "epoch": 7.01, + "learning_rate": 1.4982521302162989e-06, + "loss": 0.6515, + "step": 32105 + }, + { + "epoch": 7.01, + "learning_rate": 1.4981428883548177e-06, + "loss": 0.8743, + "step": 32106 + }, + { + "epoch": 7.01, + "learning_rate": 1.4980336464933365e-06, + "loss": 0.5296, + "step": 32107 + }, + { + "epoch": 7.01, + "learning_rate": 1.497924404631855e-06, + "loss": 0.4579, + "step": 32108 + }, + { + "epoch": 7.01, + "learning_rate": 1.4978151627703737e-06, + "loss": 0.9109, + "step": 32109 + }, + { + "epoch": 7.01, + "learning_rate": 1.4977059209088924e-06, + "loss": 1.0898, + "step": 32110 + }, + { + "epoch": 7.01, + "learning_rate": 1.4975966790474112e-06, + "loss": 1.725, + "step": 32111 + }, + { + "epoch": 7.01, + "learning_rate": 1.4974874371859296e-06, + "loss": 0.0701, + "step": 32112 + }, + { + "epoch": 7.01, + "learning_rate": 1.4973781953244484e-06, + "loss": 1.4151, + "step": 32113 + }, + { + "epoch": 7.01, + "learning_rate": 1.4972689534629672e-06, + "loss": 0.2899, + "step": 32114 + }, + { + "epoch": 7.01, + "learning_rate": 1.4971597116014858e-06, + "loss": 0.3199, + "step": 32115 + }, + { + "epoch": 7.01, + "learning_rate": 1.4970504697400047e-06, + "loss": 0.1805, + "step": 32116 + }, + { + "epoch": 7.01, + "learning_rate": 1.496941227878523e-06, + "loss": 1.0312, + "step": 32117 + }, + { + "epoch": 7.01, + "learning_rate": 1.4968319860170419e-06, + "loss": 0.3005, + "step": 32118 + }, + { + "epoch": 7.01, + "learning_rate": 1.4967227441555605e-06, + "loss": 2.0878, + "step": 32119 + }, + { + "epoch": 7.01, + "learning_rate": 1.496613502294079e-06, + "loss": 1.2094, + "step": 32120 + }, + { + "epoch": 7.01, + "learning_rate": 1.496504260432598e-06, + "loss": 1.2358, + "step": 32121 + }, + { + "epoch": 7.01, + "learning_rate": 1.4963950185711165e-06, + "loss": 1.5567, + "step": 32122 + }, + { + "epoch": 7.01, + "learning_rate": 1.4962857767096354e-06, + "loss": 2.3821, + "step": 32123 + }, + { + "epoch": 7.01, + "learning_rate": 1.4961765348481538e-06, + "loss": 0.778, + "step": 32124 + }, + { + "epoch": 7.01, + "learning_rate": 1.4960672929866726e-06, + "loss": 0.6017, + "step": 32125 + }, + { + "epoch": 7.01, + "learning_rate": 1.4959580511251912e-06, + "loss": 1.1353, + "step": 32126 + }, + { + "epoch": 7.01, + "learning_rate": 1.49584880926371e-06, + "loss": 1.3271, + "step": 32127 + }, + { + "epoch": 7.01, + "learning_rate": 1.4957395674022288e-06, + "loss": 0.657, + "step": 32128 + }, + { + "epoch": 7.01, + "learning_rate": 1.4956303255407472e-06, + "loss": 0.954, + "step": 32129 + }, + { + "epoch": 7.01, + "learning_rate": 1.495521083679266e-06, + "loss": 0.0974, + "step": 32130 + }, + { + "epoch": 7.01, + "learning_rate": 1.4954118418177847e-06, + "loss": 1.1968, + "step": 32131 + }, + { + "epoch": 7.01, + "learning_rate": 1.4953025999563035e-06, + "loss": 0.6094, + "step": 32132 + }, + { + "epoch": 7.01, + "learning_rate": 1.495193358094822e-06, + "loss": 0.4398, + "step": 32133 + }, + { + "epoch": 7.01, + "learning_rate": 1.4950841162333407e-06, + "loss": 0.9253, + "step": 32134 + }, + { + "epoch": 7.01, + "learning_rate": 1.4949748743718595e-06, + "loss": 0.8375, + "step": 32135 + }, + { + "epoch": 7.01, + "learning_rate": 1.494865632510378e-06, + "loss": 0.8705, + "step": 32136 + }, + { + "epoch": 7.01, + "learning_rate": 1.4947563906488968e-06, + "loss": 0.2423, + "step": 32137 + }, + { + "epoch": 7.01, + "learning_rate": 1.4946471487874154e-06, + "loss": 3.8184, + "step": 32138 + }, + { + "epoch": 7.01, + "learning_rate": 1.4945379069259342e-06, + "loss": 2.424, + "step": 32139 + }, + { + "epoch": 7.01, + "learning_rate": 1.494428665064453e-06, + "loss": 1.339, + "step": 32140 + }, + { + "epoch": 7.01, + "learning_rate": 1.4943194232029714e-06, + "loss": 1.4406, + "step": 32141 + }, + { + "epoch": 7.01, + "learning_rate": 1.4942101813414902e-06, + "loss": 1.8406, + "step": 32142 + }, + { + "epoch": 7.02, + "learning_rate": 1.4941009394800088e-06, + "loss": 1.1873, + "step": 32143 + }, + { + "epoch": 7.02, + "learning_rate": 1.4939916976185277e-06, + "loss": 1.13, + "step": 32144 + }, + { + "epoch": 7.02, + "learning_rate": 1.493882455757046e-06, + "loss": 0.5787, + "step": 32145 + }, + { + "epoch": 7.02, + "learning_rate": 1.4937732138955649e-06, + "loss": 0.1407, + "step": 32146 + }, + { + "epoch": 7.02, + "learning_rate": 1.4936639720340837e-06, + "loss": 1.7422, + "step": 32147 + }, + { + "epoch": 7.02, + "learning_rate": 1.4935547301726023e-06, + "loss": 2.1457, + "step": 32148 + }, + { + "epoch": 7.02, + "learning_rate": 1.493445488311121e-06, + "loss": 1.1216, + "step": 32149 + }, + { + "epoch": 7.02, + "learning_rate": 1.4933362464496395e-06, + "loss": 2.2299, + "step": 32150 + }, + { + "epoch": 7.02, + "learning_rate": 1.4932270045881584e-06, + "loss": 1.4502, + "step": 32151 + }, + { + "epoch": 7.02, + "learning_rate": 1.4931177627266768e-06, + "loss": 0.2435, + "step": 32152 + }, + { + "epoch": 7.02, + "learning_rate": 1.4930085208651956e-06, + "loss": 2.0656, + "step": 32153 + }, + { + "epoch": 7.02, + "learning_rate": 1.4928992790037144e-06, + "loss": 1.3211, + "step": 32154 + }, + { + "epoch": 7.02, + "learning_rate": 1.492790037142233e-06, + "loss": 0.4826, + "step": 32155 + }, + { + "epoch": 7.02, + "learning_rate": 1.4926807952807518e-06, + "loss": 2.0529, + "step": 32156 + }, + { + "epoch": 7.02, + "learning_rate": 1.4925715534192702e-06, + "loss": 3.2735, + "step": 32157 + }, + { + "epoch": 7.02, + "learning_rate": 1.492462311557789e-06, + "loss": 0.1009, + "step": 32158 + }, + { + "epoch": 7.02, + "learning_rate": 1.4923530696963077e-06, + "loss": 0.9382, + "step": 32159 + }, + { + "epoch": 7.02, + "learning_rate": 1.4922438278348265e-06, + "loss": 0.9955, + "step": 32160 + }, + { + "epoch": 7.02, + "learning_rate": 1.4921345859733451e-06, + "loss": 3.0226, + "step": 32161 + }, + { + "epoch": 7.02, + "learning_rate": 1.4920253441118637e-06, + "loss": 2.7498, + "step": 32162 + }, + { + "epoch": 7.02, + "learning_rate": 1.4919161022503825e-06, + "loss": 1.1622, + "step": 32163 + }, + { + "epoch": 7.02, + "learning_rate": 1.491806860388901e-06, + "loss": 0.2743, + "step": 32164 + }, + { + "epoch": 7.02, + "learning_rate": 1.4916976185274198e-06, + "loss": 1.1262, + "step": 32165 + }, + { + "epoch": 7.02, + "learning_rate": 1.4915883766659386e-06, + "loss": 0.6237, + "step": 32166 + }, + { + "epoch": 7.02, + "learning_rate": 1.4914791348044572e-06, + "loss": 1.7242, + "step": 32167 + }, + { + "epoch": 7.02, + "learning_rate": 1.491369892942976e-06, + "loss": 1.6464, + "step": 32168 + }, + { + "epoch": 7.02, + "learning_rate": 1.4912606510814944e-06, + "loss": 1.9559, + "step": 32169 + }, + { + "epoch": 7.02, + "learning_rate": 1.4911514092200132e-06, + "loss": 2.1852, + "step": 32170 + }, + { + "epoch": 7.02, + "learning_rate": 1.4910421673585319e-06, + "loss": 1.7446, + "step": 32171 + }, + { + "epoch": 7.02, + "learning_rate": 1.4909329254970507e-06, + "loss": 1.0016, + "step": 32172 + }, + { + "epoch": 7.02, + "learning_rate": 1.4908236836355695e-06, + "loss": 1.8089, + "step": 32173 + }, + { + "epoch": 7.02, + "learning_rate": 1.490714441774088e-06, + "loss": 1.0328, + "step": 32174 + }, + { + "epoch": 7.02, + "learning_rate": 1.4906051999126067e-06, + "loss": 1.6912, + "step": 32175 + }, + { + "epoch": 7.02, + "learning_rate": 1.4904959580511253e-06, + "loss": 0.2123, + "step": 32176 + }, + { + "epoch": 7.02, + "learning_rate": 1.490386716189644e-06, + "loss": 0.5985, + "step": 32177 + }, + { + "epoch": 7.02, + "learning_rate": 1.4902774743281626e-06, + "loss": 0.3189, + "step": 32178 + }, + { + "epoch": 7.02, + "learning_rate": 1.4901682324666814e-06, + "loss": 0.9223, + "step": 32179 + }, + { + "epoch": 7.02, + "learning_rate": 1.4900589906052002e-06, + "loss": 1.6164, + "step": 32180 + }, + { + "epoch": 7.02, + "learning_rate": 1.4899497487437186e-06, + "loss": 0.668, + "step": 32181 + }, + { + "epoch": 7.02, + "learning_rate": 1.4898405068822374e-06, + "loss": 1.2631, + "step": 32182 + }, + { + "epoch": 7.02, + "learning_rate": 1.489731265020756e-06, + "loss": 0.1641, + "step": 32183 + }, + { + "epoch": 7.02, + "learning_rate": 1.4896220231592749e-06, + "loss": 2.0793, + "step": 32184 + }, + { + "epoch": 7.02, + "learning_rate": 1.4895127812977933e-06, + "loss": 1.1081, + "step": 32185 + }, + { + "epoch": 7.02, + "learning_rate": 1.489403539436312e-06, + "loss": 2.54, + "step": 32186 + }, + { + "epoch": 7.02, + "learning_rate": 1.489294297574831e-06, + "loss": 1.2748, + "step": 32187 + }, + { + "epoch": 7.02, + "learning_rate": 1.4891850557133495e-06, + "loss": 1.6564, + "step": 32188 + }, + { + "epoch": 7.03, + "learning_rate": 1.4890758138518681e-06, + "loss": 0.134, + "step": 32189 + }, + { + "epoch": 7.03, + "learning_rate": 1.4889665719903867e-06, + "loss": 1.5397, + "step": 32190 + }, + { + "epoch": 7.03, + "learning_rate": 1.4888573301289056e-06, + "loss": 2.6358, + "step": 32191 + }, + { + "epoch": 7.03, + "learning_rate": 1.488748088267424e-06, + "loss": 1.1103, + "step": 32192 + }, + { + "epoch": 7.03, + "learning_rate": 1.4886388464059428e-06, + "loss": 2.4788, + "step": 32193 + }, + { + "epoch": 7.03, + "learning_rate": 1.4885296045444616e-06, + "loss": 0.4051, + "step": 32194 + }, + { + "epoch": 7.03, + "learning_rate": 1.4884203626829802e-06, + "loss": 2.6525, + "step": 32195 + }, + { + "epoch": 7.03, + "learning_rate": 1.488311120821499e-06, + "loss": 0.9831, + "step": 32196 + }, + { + "epoch": 7.03, + "learning_rate": 1.4882018789600174e-06, + "loss": 0.2234, + "step": 32197 + }, + { + "epoch": 7.03, + "learning_rate": 1.4880926370985363e-06, + "loss": 1.7125, + "step": 32198 + }, + { + "epoch": 7.03, + "learning_rate": 1.487983395237055e-06, + "loss": 2.9795, + "step": 32199 + }, + { + "epoch": 7.03, + "learning_rate": 1.4878741533755737e-06, + "loss": 0.3035, + "step": 32200 + }, + { + "epoch": 7.03, + "learning_rate": 1.4877649115140925e-06, + "loss": 1.1894, + "step": 32201 + }, + { + "epoch": 7.03, + "learning_rate": 1.487655669652611e-06, + "loss": 0.9764, + "step": 32202 + }, + { + "epoch": 7.03, + "learning_rate": 1.4875464277911297e-06, + "loss": 1.1623, + "step": 32203 + }, + { + "epoch": 7.03, + "learning_rate": 1.4874371859296483e-06, + "loss": 1.3279, + "step": 32204 + }, + { + "epoch": 7.03, + "learning_rate": 1.487327944068167e-06, + "loss": 0.6506, + "step": 32205 + }, + { + "epoch": 7.03, + "learning_rate": 1.4872187022066858e-06, + "loss": 2.6096, + "step": 32206 + }, + { + "epoch": 7.03, + "learning_rate": 1.4871094603452044e-06, + "loss": 0.374, + "step": 32207 + }, + { + "epoch": 7.03, + "learning_rate": 1.4870002184837232e-06, + "loss": 1.3179, + "step": 32208 + }, + { + "epoch": 7.03, + "learning_rate": 1.4868909766222416e-06, + "loss": 1.4407, + "step": 32209 + }, + { + "epoch": 7.03, + "learning_rate": 1.4867817347607604e-06, + "loss": 1.2234, + "step": 32210 + }, + { + "epoch": 7.03, + "learning_rate": 1.486672492899279e-06, + "loss": 1.9646, + "step": 32211 + }, + { + "epoch": 7.03, + "learning_rate": 1.4865632510377979e-06, + "loss": 2.5537, + "step": 32212 + }, + { + "epoch": 7.03, + "learning_rate": 1.4864540091763167e-06, + "loss": 1.5306, + "step": 32213 + }, + { + "epoch": 7.03, + "learning_rate": 1.486344767314835e-06, + "loss": 0.623, + "step": 32214 + }, + { + "epoch": 7.03, + "learning_rate": 1.486235525453354e-06, + "loss": 1.9189, + "step": 32215 + }, + { + "epoch": 7.03, + "learning_rate": 1.4861262835918725e-06, + "loss": 2.5831, + "step": 32216 + }, + { + "epoch": 7.03, + "learning_rate": 1.4860170417303911e-06, + "loss": 0.4991, + "step": 32217 + }, + { + "epoch": 7.03, + "learning_rate": 1.4859077998689097e-06, + "loss": 0.7124, + "step": 32218 + }, + { + "epoch": 7.03, + "learning_rate": 1.4857985580074286e-06, + "loss": 1.9332, + "step": 32219 + }, + { + "epoch": 7.03, + "learning_rate": 1.4856893161459474e-06, + "loss": 0.9382, + "step": 32220 + }, + { + "epoch": 7.03, + "learning_rate": 1.4855800742844658e-06, + "loss": 0.8464, + "step": 32221 + }, + { + "epoch": 7.03, + "learning_rate": 1.4854708324229846e-06, + "loss": 0.4467, + "step": 32222 + }, + { + "epoch": 7.03, + "learning_rate": 1.4853615905615032e-06, + "loss": 1.4188, + "step": 32223 + }, + { + "epoch": 7.03, + "learning_rate": 1.485252348700022e-06, + "loss": 0.8061, + "step": 32224 + }, + { + "epoch": 7.03, + "learning_rate": 1.4851431068385404e-06, + "loss": 1.0717, + "step": 32225 + }, + { + "epoch": 7.03, + "learning_rate": 1.4850338649770593e-06, + "loss": 0.7271, + "step": 32226 + }, + { + "epoch": 7.03, + "learning_rate": 1.484924623115578e-06, + "loss": 1.8155, + "step": 32227 + }, + { + "epoch": 7.03, + "learning_rate": 1.4848153812540967e-06, + "loss": 2.5114, + "step": 32228 + }, + { + "epoch": 7.03, + "learning_rate": 1.4847061393926155e-06, + "loss": 1.0976, + "step": 32229 + }, + { + "epoch": 7.03, + "learning_rate": 1.484596897531134e-06, + "loss": 1.7569, + "step": 32230 + }, + { + "epoch": 7.03, + "learning_rate": 1.4844876556696528e-06, + "loss": 0.7982, + "step": 32231 + }, + { + "epoch": 7.03, + "learning_rate": 1.4843784138081716e-06, + "loss": 2.7463, + "step": 32232 + }, + { + "epoch": 7.03, + "learning_rate": 1.48426917194669e-06, + "loss": 1.3478, + "step": 32233 + }, + { + "epoch": 7.03, + "learning_rate": 1.4841599300852088e-06, + "loss": 1.8375, + "step": 32234 + }, + { + "epoch": 7.04, + "learning_rate": 1.4840506882237274e-06, + "loss": 0.3706, + "step": 32235 + }, + { + "epoch": 7.04, + "learning_rate": 1.4839414463622462e-06, + "loss": 1.2027, + "step": 32236 + }, + { + "epoch": 7.04, + "learning_rate": 1.4838322045007646e-06, + "loss": 0.7987, + "step": 32237 + }, + { + "epoch": 7.04, + "learning_rate": 1.4837229626392835e-06, + "loss": 0.9195, + "step": 32238 + }, + { + "epoch": 7.04, + "learning_rate": 1.4836137207778023e-06, + "loss": 0.7264, + "step": 32239 + }, + { + "epoch": 7.04, + "learning_rate": 1.4835044789163209e-06, + "loss": 0.496, + "step": 32240 + }, + { + "epoch": 7.04, + "learning_rate": 1.4833952370548397e-06, + "loss": 3.2096, + "step": 32241 + }, + { + "epoch": 7.04, + "learning_rate": 1.4832859951933581e-06, + "loss": 1.5616, + "step": 32242 + }, + { + "epoch": 7.04, + "learning_rate": 1.483176753331877e-06, + "loss": 1.5405, + "step": 32243 + }, + { + "epoch": 7.04, + "learning_rate": 1.4830675114703955e-06, + "loss": 1.1154, + "step": 32244 + }, + { + "epoch": 7.04, + "learning_rate": 1.4829582696089144e-06, + "loss": 2.2138, + "step": 32245 + }, + { + "epoch": 7.04, + "learning_rate": 1.482849027747433e-06, + "loss": 0.4129, + "step": 32246 + }, + { + "epoch": 7.04, + "learning_rate": 1.4827397858859516e-06, + "loss": 1.3571, + "step": 32247 + }, + { + "epoch": 7.04, + "learning_rate": 1.4826305440244704e-06, + "loss": 2.453, + "step": 32248 + }, + { + "epoch": 7.04, + "learning_rate": 1.4825213021629888e-06, + "loss": 0.1774, + "step": 32249 + }, + { + "epoch": 7.04, + "learning_rate": 1.4824120603015076e-06, + "loss": 2.2817, + "step": 32250 + }, + { + "epoch": 7.04, + "learning_rate": 1.4823028184400262e-06, + "loss": 1.7632, + "step": 32251 + }, + { + "epoch": 7.04, + "learning_rate": 1.482193576578545e-06, + "loss": 1.1579, + "step": 32252 + }, + { + "epoch": 7.04, + "learning_rate": 1.4820843347170639e-06, + "loss": 1.6973, + "step": 32253 + }, + { + "epoch": 7.04, + "learning_rate": 1.4819750928555823e-06, + "loss": 0.9045, + "step": 32254 + }, + { + "epoch": 7.04, + "learning_rate": 1.4818658509941011e-06, + "loss": 0.9046, + "step": 32255 + }, + { + "epoch": 7.04, + "learning_rate": 1.4817566091326197e-06, + "loss": 0.9321, + "step": 32256 + }, + { + "epoch": 7.04, + "learning_rate": 1.4816473672711385e-06, + "loss": 2.9209, + "step": 32257 + }, + { + "epoch": 7.04, + "learning_rate": 1.481538125409657e-06, + "loss": 2.1248, + "step": 32258 + }, + { + "epoch": 7.04, + "learning_rate": 1.4814288835481758e-06, + "loss": 2.1382, + "step": 32259 + }, + { + "epoch": 7.04, + "learning_rate": 1.4813196416866946e-06, + "loss": 0.3542, + "step": 32260 + }, + { + "epoch": 7.04, + "learning_rate": 1.481210399825213e-06, + "loss": 0.5167, + "step": 32261 + }, + { + "epoch": 7.04, + "learning_rate": 1.4811011579637318e-06, + "loss": 1.0843, + "step": 32262 + }, + { + "epoch": 7.04, + "learning_rate": 1.4809919161022504e-06, + "loss": 0.8911, + "step": 32263 + }, + { + "epoch": 7.04, + "learning_rate": 1.4808826742407692e-06, + "loss": 1.454, + "step": 32264 + }, + { + "epoch": 7.04, + "learning_rate": 1.480773432379288e-06, + "loss": 1.8758, + "step": 32265 + }, + { + "epoch": 7.04, + "learning_rate": 1.4806641905178065e-06, + "loss": 1.4332, + "step": 32266 + }, + { + "epoch": 7.04, + "learning_rate": 1.4805549486563253e-06, + "loss": 0.8472, + "step": 32267 + }, + { + "epoch": 7.04, + "learning_rate": 1.480445706794844e-06, + "loss": 0.9688, + "step": 32268 + }, + { + "epoch": 7.04, + "learning_rate": 1.4803364649333627e-06, + "loss": 2.7334, + "step": 32269 + }, + { + "epoch": 7.04, + "learning_rate": 1.4802272230718811e-06, + "loss": 0.1557, + "step": 32270 + }, + { + "epoch": 7.04, + "learning_rate": 1.4801179812104e-06, + "loss": 0.1205, + "step": 32271 + }, + { + "epoch": 7.04, + "learning_rate": 1.4800087393489188e-06, + "loss": 2.2719, + "step": 32272 + }, + { + "epoch": 7.04, + "learning_rate": 1.4798994974874374e-06, + "loss": 1.3309, + "step": 32273 + }, + { + "epoch": 7.04, + "learning_rate": 1.479790255625956e-06, + "loss": 2.0883, + "step": 32274 + }, + { + "epoch": 7.04, + "learning_rate": 1.4796810137644746e-06, + "loss": 0.4136, + "step": 32275 + }, + { + "epoch": 7.04, + "learning_rate": 1.4795717719029934e-06, + "loss": 0.5729, + "step": 32276 + }, + { + "epoch": 7.04, + "learning_rate": 1.4794625300415118e-06, + "loss": 0.0894, + "step": 32277 + }, + { + "epoch": 7.04, + "learning_rate": 1.4793532881800306e-06, + "loss": 0.3773, + "step": 32278 + }, + { + "epoch": 7.04, + "learning_rate": 1.4792440463185495e-06, + "loss": 2.4703, + "step": 32279 + }, + { + "epoch": 7.04, + "learning_rate": 1.479134804457068e-06, + "loss": 0.3883, + "step": 32280 + }, + { + "epoch": 7.05, + "learning_rate": 1.479025562595587e-06, + "loss": 1.7168, + "step": 32281 + }, + { + "epoch": 7.05, + "learning_rate": 1.4789163207341053e-06, + "loss": 0.295, + "step": 32282 + }, + { + "epoch": 7.05, + "learning_rate": 1.4788070788726241e-06, + "loss": 1.6535, + "step": 32283 + }, + { + "epoch": 7.05, + "learning_rate": 1.4786978370111427e-06, + "loss": 1.1472, + "step": 32284 + }, + { + "epoch": 7.05, + "learning_rate": 1.4785885951496616e-06, + "loss": 1.8773, + "step": 32285 + }, + { + "epoch": 7.05, + "learning_rate": 1.4784793532881802e-06, + "loss": 2.1192, + "step": 32286 + }, + { + "epoch": 7.05, + "learning_rate": 1.4783701114266988e-06, + "loss": 1.9107, + "step": 32287 + }, + { + "epoch": 7.05, + "learning_rate": 1.4782608695652176e-06, + "loss": 1.0023, + "step": 32288 + }, + { + "epoch": 7.05, + "learning_rate": 1.478151627703736e-06, + "loss": 1.5786, + "step": 32289 + }, + { + "epoch": 7.05, + "learning_rate": 1.4780423858422548e-06, + "loss": 1.0251, + "step": 32290 + }, + { + "epoch": 7.05, + "learning_rate": 1.4779331439807734e-06, + "loss": 1.6046, + "step": 32291 + }, + { + "epoch": 7.05, + "learning_rate": 1.4778239021192923e-06, + "loss": 2.0301, + "step": 32292 + }, + { + "epoch": 7.05, + "learning_rate": 1.477714660257811e-06, + "loss": 0.7264, + "step": 32293 + }, + { + "epoch": 7.05, + "learning_rate": 1.4776054183963295e-06, + "loss": 1.4047, + "step": 32294 + }, + { + "epoch": 7.05, + "learning_rate": 1.4774961765348483e-06, + "loss": 1.3227, + "step": 32295 + }, + { + "epoch": 7.05, + "learning_rate": 1.477386934673367e-06, + "loss": 1.2155, + "step": 32296 + }, + { + "epoch": 7.05, + "learning_rate": 1.4772776928118857e-06, + "loss": 2.532, + "step": 32297 + }, + { + "epoch": 7.05, + "learning_rate": 1.4771684509504046e-06, + "loss": 0.5699, + "step": 32298 + }, + { + "epoch": 7.05, + "learning_rate": 1.477059209088923e-06, + "loss": 1.1273, + "step": 32299 + }, + { + "epoch": 7.05, + "learning_rate": 1.4769499672274418e-06, + "loss": 0.6453, + "step": 32300 + }, + { + "epoch": 7.05, + "learning_rate": 1.4768407253659604e-06, + "loss": 1.9862, + "step": 32301 + }, + { + "epoch": 7.05, + "learning_rate": 1.476731483504479e-06, + "loss": 1.1758, + "step": 32302 + }, + { + "epoch": 7.05, + "learning_rate": 1.4766222416429976e-06, + "loss": 1.1804, + "step": 32303 + }, + { + "epoch": 7.05, + "learning_rate": 1.4765129997815164e-06, + "loss": 1.1304, + "step": 32304 + }, + { + "epoch": 7.05, + "learning_rate": 1.4764037579200353e-06, + "loss": 0.6971, + "step": 32305 + }, + { + "epoch": 7.05, + "learning_rate": 1.4762945160585537e-06, + "loss": 0.7197, + "step": 32306 + }, + { + "epoch": 7.05, + "learning_rate": 1.4761852741970725e-06, + "loss": 1.03, + "step": 32307 + }, + { + "epoch": 7.05, + "learning_rate": 1.476076032335591e-06, + "loss": 0.5937, + "step": 32308 + }, + { + "epoch": 7.05, + "learning_rate": 1.47596679047411e-06, + "loss": 1.1799, + "step": 32309 + }, + { + "epoch": 7.05, + "learning_rate": 1.4758575486126283e-06, + "loss": 1.3055, + "step": 32310 + }, + { + "epoch": 7.05, + "learning_rate": 1.4757483067511471e-06, + "loss": 0.6753, + "step": 32311 + }, + { + "epoch": 7.05, + "learning_rate": 1.475639064889666e-06, + "loss": 1.1447, + "step": 32312 + }, + { + "epoch": 7.05, + "learning_rate": 1.4755298230281846e-06, + "loss": 0.5147, + "step": 32313 + }, + { + "epoch": 7.05, + "learning_rate": 1.4754205811667032e-06, + "loss": 3.1662, + "step": 32314 + }, + { + "epoch": 7.05, + "learning_rate": 1.4753113393052218e-06, + "loss": 0.8356, + "step": 32315 + }, + { + "epoch": 7.05, + "learning_rate": 1.4752020974437406e-06, + "loss": 1.6647, + "step": 32316 + }, + { + "epoch": 7.05, + "learning_rate": 1.4750928555822592e-06, + "loss": 1.5487, + "step": 32317 + }, + { + "epoch": 7.05, + "learning_rate": 1.4749836137207778e-06, + "loss": 0.3485, + "step": 32318 + }, + { + "epoch": 7.05, + "learning_rate": 1.4748743718592967e-06, + "loss": 0.956, + "step": 32319 + }, + { + "epoch": 7.05, + "learning_rate": 1.4747651299978153e-06, + "loss": 0.575, + "step": 32320 + }, + { + "epoch": 7.05, + "learning_rate": 1.474655888136334e-06, + "loss": 1.2057, + "step": 32321 + }, + { + "epoch": 7.05, + "learning_rate": 1.4745466462748525e-06, + "loss": 0.8069, + "step": 32322 + }, + { + "epoch": 7.05, + "learning_rate": 1.4744374044133713e-06, + "loss": 0.9884, + "step": 32323 + }, + { + "epoch": 7.05, + "learning_rate": 1.47432816255189e-06, + "loss": 1.8157, + "step": 32324 + }, + { + "epoch": 7.05, + "learning_rate": 1.4742189206904087e-06, + "loss": 0.6974, + "step": 32325 + }, + { + "epoch": 7.05, + "learning_rate": 1.4741096788289276e-06, + "loss": 1.675, + "step": 32326 + }, + { + "epoch": 7.06, + "learning_rate": 1.474000436967446e-06, + "loss": 0.6202, + "step": 32327 + }, + { + "epoch": 7.06, + "learning_rate": 1.4738911951059648e-06, + "loss": 1.7682, + "step": 32328 + }, + { + "epoch": 7.06, + "learning_rate": 1.4737819532444834e-06, + "loss": 1.1625, + "step": 32329 + }, + { + "epoch": 7.06, + "learning_rate": 1.473672711383002e-06, + "loss": 0.1439, + "step": 32330 + }, + { + "epoch": 7.06, + "learning_rate": 1.4735634695215206e-06, + "loss": 1.0203, + "step": 32331 + }, + { + "epoch": 7.06, + "learning_rate": 1.4734542276600394e-06, + "loss": 0.4463, + "step": 32332 + }, + { + "epoch": 7.06, + "learning_rate": 1.4733449857985583e-06, + "loss": 1.0246, + "step": 32333 + }, + { + "epoch": 7.06, + "learning_rate": 1.4732357439370767e-06, + "loss": 3.1814, + "step": 32334 + }, + { + "epoch": 7.06, + "learning_rate": 1.4731265020755955e-06, + "loss": 1.926, + "step": 32335 + }, + { + "epoch": 7.06, + "learning_rate": 1.473017260214114e-06, + "loss": 0.9935, + "step": 32336 + }, + { + "epoch": 7.06, + "learning_rate": 1.472908018352633e-06, + "loss": 0.7886, + "step": 32337 + }, + { + "epoch": 7.06, + "learning_rate": 1.4727987764911517e-06, + "loss": 2.4623, + "step": 32338 + }, + { + "epoch": 7.06, + "learning_rate": 1.4726895346296701e-06, + "loss": 0.6169, + "step": 32339 + }, + { + "epoch": 7.06, + "learning_rate": 1.472580292768189e-06, + "loss": 1.4874, + "step": 32340 + }, + { + "epoch": 7.06, + "learning_rate": 1.4724710509067076e-06, + "loss": 1.491, + "step": 32341 + }, + { + "epoch": 7.06, + "learning_rate": 1.4723618090452262e-06, + "loss": 1.2821, + "step": 32342 + }, + { + "epoch": 7.06, + "learning_rate": 1.4722525671837448e-06, + "loss": 0.1477, + "step": 32343 + }, + { + "epoch": 7.06, + "learning_rate": 1.4721433253222636e-06, + "loss": 0.3086, + "step": 32344 + }, + { + "epoch": 7.06, + "learning_rate": 1.4720340834607825e-06, + "loss": 1.0699, + "step": 32345 + }, + { + "epoch": 7.06, + "learning_rate": 1.4719248415993008e-06, + "loss": 1.2661, + "step": 32346 + }, + { + "epoch": 7.06, + "learning_rate": 1.4718155997378197e-06, + "loss": 0.8239, + "step": 32347 + }, + { + "epoch": 7.06, + "learning_rate": 1.4717063578763383e-06, + "loss": 1.9042, + "step": 32348 + }, + { + "epoch": 7.06, + "learning_rate": 1.4715971160148571e-06, + "loss": 1.1952, + "step": 32349 + }, + { + "epoch": 7.06, + "learning_rate": 1.4714878741533755e-06, + "loss": 1.1576, + "step": 32350 + }, + { + "epoch": 7.06, + "learning_rate": 1.4713786322918943e-06, + "loss": 1.2668, + "step": 32351 + }, + { + "epoch": 7.06, + "learning_rate": 1.4712693904304132e-06, + "loss": 1.0784, + "step": 32352 + }, + { + "epoch": 7.06, + "learning_rate": 1.4711601485689318e-06, + "loss": 0.3546, + "step": 32353 + }, + { + "epoch": 7.06, + "learning_rate": 1.4710509067074506e-06, + "loss": 0.0887, + "step": 32354 + }, + { + "epoch": 7.06, + "learning_rate": 1.470941664845969e-06, + "loss": 1.5399, + "step": 32355 + }, + { + "epoch": 7.06, + "learning_rate": 1.4708324229844878e-06, + "loss": 1.4282, + "step": 32356 + }, + { + "epoch": 7.06, + "learning_rate": 1.4707231811230064e-06, + "loss": 1.0149, + "step": 32357 + }, + { + "epoch": 7.06, + "learning_rate": 1.470613939261525e-06, + "loss": 1.5317, + "step": 32358 + }, + { + "epoch": 7.06, + "learning_rate": 1.4705046974000439e-06, + "loss": 1.0068, + "step": 32359 + }, + { + "epoch": 7.06, + "learning_rate": 1.4703954555385625e-06, + "loss": 1.9725, + "step": 32360 + }, + { + "epoch": 7.06, + "learning_rate": 1.4702862136770813e-06, + "loss": 1.2392, + "step": 32361 + }, + { + "epoch": 7.06, + "learning_rate": 1.4701769718155997e-06, + "loss": 1.8822, + "step": 32362 + }, + { + "epoch": 7.06, + "learning_rate": 1.4700677299541185e-06, + "loss": 0.6266, + "step": 32363 + }, + { + "epoch": 7.06, + "learning_rate": 1.4699584880926371e-06, + "loss": 1.9611, + "step": 32364 + }, + { + "epoch": 7.06, + "learning_rate": 1.469849246231156e-06, + "loss": 0.4764, + "step": 32365 + }, + { + "epoch": 7.06, + "learning_rate": 1.4697400043696748e-06, + "loss": 2.7848, + "step": 32366 + }, + { + "epoch": 7.06, + "learning_rate": 1.4696307625081932e-06, + "loss": 1.0554, + "step": 32367 + }, + { + "epoch": 7.06, + "learning_rate": 1.469521520646712e-06, + "loss": 0.6928, + "step": 32368 + }, + { + "epoch": 7.06, + "learning_rate": 1.4694122787852306e-06, + "loss": 1.0795, + "step": 32369 + }, + { + "epoch": 7.06, + "learning_rate": 1.4693030369237494e-06, + "loss": 2.187, + "step": 32370 + }, + { + "epoch": 7.06, + "learning_rate": 1.469193795062268e-06, + "loss": 1.7975, + "step": 32371 + }, + { + "epoch": 7.07, + "learning_rate": 1.4690845532007866e-06, + "loss": 1.5752, + "step": 32372 + }, + { + "epoch": 7.07, + "learning_rate": 1.4689753113393055e-06, + "loss": 0.3509, + "step": 32373 + }, + { + "epoch": 7.07, + "learning_rate": 1.4688660694778239e-06, + "loss": 0.2115, + "step": 32374 + }, + { + "epoch": 7.07, + "learning_rate": 1.4687568276163427e-06, + "loss": 0.8833, + "step": 32375 + }, + { + "epoch": 7.07, + "learning_rate": 1.4686475857548613e-06, + "loss": 0.8108, + "step": 32376 + }, + { + "epoch": 7.07, + "learning_rate": 1.4685383438933801e-06, + "loss": 1.3353, + "step": 32377 + }, + { + "epoch": 7.07, + "learning_rate": 1.468429102031899e-06, + "loss": 0.2427, + "step": 32378 + }, + { + "epoch": 7.07, + "learning_rate": 1.4683198601704173e-06, + "loss": 1.4583, + "step": 32379 + }, + { + "epoch": 7.07, + "learning_rate": 1.4682106183089362e-06, + "loss": 0.0931, + "step": 32380 + }, + { + "epoch": 7.07, + "learning_rate": 1.4681013764474548e-06, + "loss": 2.4179, + "step": 32381 + }, + { + "epoch": 7.07, + "learning_rate": 1.4679921345859736e-06, + "loss": 3.1125, + "step": 32382 + }, + { + "epoch": 7.07, + "learning_rate": 1.467882892724492e-06, + "loss": 0.5066, + "step": 32383 + }, + { + "epoch": 7.07, + "learning_rate": 1.4677736508630108e-06, + "loss": 1.3629, + "step": 32384 + }, + { + "epoch": 7.07, + "learning_rate": 1.4676644090015296e-06, + "loss": 0.6852, + "step": 32385 + }, + { + "epoch": 7.07, + "learning_rate": 1.467555167140048e-06, + "loss": 1.4131, + "step": 32386 + }, + { + "epoch": 7.07, + "learning_rate": 1.4674459252785669e-06, + "loss": 0.9721, + "step": 32387 + }, + { + "epoch": 7.07, + "learning_rate": 1.4673366834170855e-06, + "loss": 0.6965, + "step": 32388 + }, + { + "epoch": 7.07, + "learning_rate": 1.4672274415556043e-06, + "loss": 3.0549, + "step": 32389 + }, + { + "epoch": 7.07, + "learning_rate": 1.4671181996941227e-06, + "loss": 2.0472, + "step": 32390 + }, + { + "epoch": 7.07, + "learning_rate": 1.4670089578326415e-06, + "loss": 0.5306, + "step": 32391 + }, + { + "epoch": 7.07, + "learning_rate": 1.4668997159711603e-06, + "loss": 0.3557, + "step": 32392 + }, + { + "epoch": 7.07, + "learning_rate": 1.466790474109679e-06, + "loss": 1.1776, + "step": 32393 + }, + { + "epoch": 7.07, + "learning_rate": 1.4666812322481978e-06, + "loss": 0.9233, + "step": 32394 + }, + { + "epoch": 7.07, + "learning_rate": 1.4665719903867162e-06, + "loss": 0.7514, + "step": 32395 + }, + { + "epoch": 7.07, + "learning_rate": 1.466462748525235e-06, + "loss": 0.1432, + "step": 32396 + }, + { + "epoch": 7.07, + "learning_rate": 1.4663535066637536e-06, + "loss": 1.4626, + "step": 32397 + }, + { + "epoch": 7.07, + "learning_rate": 1.4662442648022724e-06, + "loss": 0.8296, + "step": 32398 + }, + { + "epoch": 7.07, + "learning_rate": 1.466135022940791e-06, + "loss": 0.6166, + "step": 32399 + }, + { + "epoch": 7.07, + "learning_rate": 1.4660257810793097e-06, + "loss": 1.9492, + "step": 32400 + }, + { + "epoch": 7.07, + "learning_rate": 1.4659165392178285e-06, + "loss": 1.0762, + "step": 32401 + }, + { + "epoch": 7.07, + "learning_rate": 1.4658072973563469e-06, + "loss": 0.4236, + "step": 32402 + }, + { + "epoch": 7.07, + "learning_rate": 1.4656980554948657e-06, + "loss": 0.84, + "step": 32403 + }, + { + "epoch": 7.07, + "learning_rate": 1.4655888136333845e-06, + "loss": 1.6427, + "step": 32404 + }, + { + "epoch": 7.07, + "learning_rate": 1.4654795717719031e-06, + "loss": 2.5346, + "step": 32405 + }, + { + "epoch": 7.07, + "learning_rate": 1.465370329910422e-06, + "loss": 0.909, + "step": 32406 + }, + { + "epoch": 7.07, + "learning_rate": 1.4652610880489404e-06, + "loss": 2.9451, + "step": 32407 + }, + { + "epoch": 7.07, + "learning_rate": 1.4651518461874592e-06, + "loss": 2.7226, + "step": 32408 + }, + { + "epoch": 7.07, + "learning_rate": 1.4650426043259778e-06, + "loss": 0.8276, + "step": 32409 + }, + { + "epoch": 7.07, + "learning_rate": 1.4649333624644966e-06, + "loss": 2.3486, + "step": 32410 + }, + { + "epoch": 7.07, + "learning_rate": 1.4648241206030152e-06, + "loss": 1.8333, + "step": 32411 + }, + { + "epoch": 7.07, + "learning_rate": 1.4647148787415338e-06, + "loss": 0.8018, + "step": 32412 + }, + { + "epoch": 7.07, + "learning_rate": 1.4646056368800527e-06, + "loss": 0.5047, + "step": 32413 + }, + { + "epoch": 7.07, + "learning_rate": 1.464496395018571e-06, + "loss": 0.6204, + "step": 32414 + }, + { + "epoch": 7.07, + "learning_rate": 1.4643871531570899e-06, + "loss": 0.5556, + "step": 32415 + }, + { + "epoch": 7.07, + "learning_rate": 1.4642779112956085e-06, + "loss": 1.6183, + "step": 32416 + }, + { + "epoch": 7.07, + "learning_rate": 1.4641686694341273e-06, + "loss": 0.7149, + "step": 32417 + }, + { + "epoch": 7.08, + "learning_rate": 1.4640594275726461e-06, + "loss": 1.3847, + "step": 32418 + }, + { + "epoch": 7.08, + "learning_rate": 1.4639501857111645e-06, + "loss": 0.9503, + "step": 32419 + }, + { + "epoch": 7.08, + "learning_rate": 1.4638409438496834e-06, + "loss": 1.1792, + "step": 32420 + }, + { + "epoch": 7.08, + "learning_rate": 1.463731701988202e-06, + "loss": 1.4542, + "step": 32421 + }, + { + "epoch": 7.08, + "learning_rate": 1.4636224601267208e-06, + "loss": 1.4195, + "step": 32422 + }, + { + "epoch": 7.08, + "learning_rate": 1.4635132182652392e-06, + "loss": 0.7975, + "step": 32423 + }, + { + "epoch": 7.08, + "learning_rate": 1.463403976403758e-06, + "loss": 1.337, + "step": 32424 + }, + { + "epoch": 7.08, + "learning_rate": 1.4632947345422768e-06, + "loss": 2.4649, + "step": 32425 + }, + { + "epoch": 7.08, + "learning_rate": 1.4631854926807954e-06, + "loss": 1.9926, + "step": 32426 + }, + { + "epoch": 7.08, + "learning_rate": 1.463076250819314e-06, + "loss": 0.4184, + "step": 32427 + }, + { + "epoch": 7.08, + "learning_rate": 1.4629670089578327e-06, + "loss": 0.1684, + "step": 32428 + }, + { + "epoch": 7.08, + "learning_rate": 1.4628577670963515e-06, + "loss": 1.2623, + "step": 32429 + }, + { + "epoch": 7.08, + "learning_rate": 1.4627485252348699e-06, + "loss": 0.5928, + "step": 32430 + }, + { + "epoch": 7.08, + "learning_rate": 1.4626392833733887e-06, + "loss": 0.6437, + "step": 32431 + }, + { + "epoch": 7.08, + "learning_rate": 1.4625300415119075e-06, + "loss": 1.3381, + "step": 32432 + }, + { + "epoch": 7.08, + "learning_rate": 1.4624207996504261e-06, + "loss": 0.726, + "step": 32433 + }, + { + "epoch": 7.08, + "learning_rate": 1.462311557788945e-06, + "loss": 0.4225, + "step": 32434 + }, + { + "epoch": 7.08, + "learning_rate": 1.4622023159274634e-06, + "loss": 1.8991, + "step": 32435 + }, + { + "epoch": 7.08, + "learning_rate": 1.4620930740659822e-06, + "loss": 1.0911, + "step": 32436 + }, + { + "epoch": 7.08, + "learning_rate": 1.461983832204501e-06, + "loss": 2.6763, + "step": 32437 + }, + { + "epoch": 7.08, + "learning_rate": 1.4618745903430196e-06, + "loss": 0.5251, + "step": 32438 + }, + { + "epoch": 7.08, + "learning_rate": 1.4617653484815382e-06, + "loss": 0.6487, + "step": 32439 + }, + { + "epoch": 7.08, + "learning_rate": 1.4616561066200568e-06, + "loss": 1.0867, + "step": 32440 + }, + { + "epoch": 7.08, + "learning_rate": 1.4615468647585757e-06, + "loss": 2.2778, + "step": 32441 + }, + { + "epoch": 7.08, + "learning_rate": 1.4614376228970943e-06, + "loss": 1.4626, + "step": 32442 + }, + { + "epoch": 7.08, + "learning_rate": 1.4613283810356129e-06, + "loss": 1.7626, + "step": 32443 + }, + { + "epoch": 7.08, + "learning_rate": 1.4612191391741317e-06, + "loss": 1.9289, + "step": 32444 + }, + { + "epoch": 7.08, + "learning_rate": 1.4611098973126503e-06, + "loss": 3.7034, + "step": 32445 + }, + { + "epoch": 7.08, + "learning_rate": 1.4610006554511691e-06, + "loss": 1.5504, + "step": 32446 + }, + { + "epoch": 7.08, + "learning_rate": 1.4608914135896875e-06, + "loss": 0.8297, + "step": 32447 + }, + { + "epoch": 7.08, + "learning_rate": 1.4607821717282064e-06, + "loss": 0.3718, + "step": 32448 + }, + { + "epoch": 7.08, + "learning_rate": 1.460672929866725e-06, + "loss": 1.0151, + "step": 32449 + }, + { + "epoch": 7.08, + "learning_rate": 1.4605636880052438e-06, + "loss": 0.8627, + "step": 32450 + }, + { + "epoch": 7.08, + "learning_rate": 1.4604544461437626e-06, + "loss": 1.8261, + "step": 32451 + }, + { + "epoch": 7.08, + "learning_rate": 1.460345204282281e-06, + "loss": 1.5638, + "step": 32452 + }, + { + "epoch": 7.08, + "learning_rate": 1.4602359624207998e-06, + "loss": 1.1424, + "step": 32453 + }, + { + "epoch": 7.08, + "learning_rate": 1.4601267205593185e-06, + "loss": 0.5637, + "step": 32454 + }, + { + "epoch": 7.08, + "learning_rate": 1.460017478697837e-06, + "loss": 1.6682, + "step": 32455 + }, + { + "epoch": 7.08, + "learning_rate": 1.4599082368363557e-06, + "loss": 0.7471, + "step": 32456 + }, + { + "epoch": 7.08, + "learning_rate": 1.4597989949748745e-06, + "loss": 0.9334, + "step": 32457 + }, + { + "epoch": 7.08, + "learning_rate": 1.4596897531133933e-06, + "loss": 1.3161, + "step": 32458 + }, + { + "epoch": 7.08, + "learning_rate": 1.4595805112519117e-06, + "loss": 0.0653, + "step": 32459 + }, + { + "epoch": 7.08, + "learning_rate": 1.4594712693904305e-06, + "loss": 2.3314, + "step": 32460 + }, + { + "epoch": 7.08, + "learning_rate": 1.4593620275289492e-06, + "loss": 1.3722, + "step": 32461 + }, + { + "epoch": 7.08, + "learning_rate": 1.459252785667468e-06, + "loss": 0.6892, + "step": 32462 + }, + { + "epoch": 7.08, + "learning_rate": 1.4591435438059864e-06, + "loss": 0.3838, + "step": 32463 + }, + { + "epoch": 7.09, + "learning_rate": 1.4590343019445052e-06, + "loss": 0.5988, + "step": 32464 + }, + { + "epoch": 7.09, + "learning_rate": 1.458925060083024e-06, + "loss": 1.4847, + "step": 32465 + }, + { + "epoch": 7.09, + "learning_rate": 1.4588158182215426e-06, + "loss": 1.4182, + "step": 32466 + }, + { + "epoch": 7.09, + "learning_rate": 1.4587065763600612e-06, + "loss": 1.8683, + "step": 32467 + }, + { + "epoch": 7.09, + "learning_rate": 1.4585973344985799e-06, + "loss": 2.2242, + "step": 32468 + }, + { + "epoch": 7.09, + "learning_rate": 1.4584880926370987e-06, + "loss": 1.8342, + "step": 32469 + }, + { + "epoch": 7.09, + "learning_rate": 1.4583788507756175e-06, + "loss": 2.3421, + "step": 32470 + }, + { + "epoch": 7.09, + "learning_rate": 1.458269608914136e-06, + "loss": 3.495, + "step": 32471 + }, + { + "epoch": 7.09, + "learning_rate": 1.4581603670526547e-06, + "loss": 0.2772, + "step": 32472 + }, + { + "epoch": 7.09, + "learning_rate": 1.4580511251911733e-06, + "loss": 0.2972, + "step": 32473 + }, + { + "epoch": 7.09, + "learning_rate": 1.4579418833296922e-06, + "loss": 2.5701, + "step": 32474 + }, + { + "epoch": 7.09, + "learning_rate": 1.4578326414682106e-06, + "loss": 1.1945, + "step": 32475 + }, + { + "epoch": 7.09, + "learning_rate": 1.4577233996067294e-06, + "loss": 1.4892, + "step": 32476 + }, + { + "epoch": 7.09, + "learning_rate": 1.4576141577452482e-06, + "loss": 0.1777, + "step": 32477 + }, + { + "epoch": 7.09, + "learning_rate": 1.4575049158837668e-06, + "loss": 1.2957, + "step": 32478 + }, + { + "epoch": 7.09, + "learning_rate": 1.4573956740222856e-06, + "loss": 0.4262, + "step": 32479 + }, + { + "epoch": 7.09, + "learning_rate": 1.457286432160804e-06, + "loss": 2.155, + "step": 32480 + }, + { + "epoch": 7.09, + "learning_rate": 1.4571771902993229e-06, + "loss": 2.6439, + "step": 32481 + }, + { + "epoch": 7.09, + "learning_rate": 1.4570679484378415e-06, + "loss": 0.9737, + "step": 32482 + }, + { + "epoch": 7.09, + "learning_rate": 1.45695870657636e-06, + "loss": 1.6235, + "step": 32483 + }, + { + "epoch": 7.09, + "learning_rate": 1.456849464714879e-06, + "loss": 1.1679, + "step": 32484 + }, + { + "epoch": 7.09, + "learning_rate": 1.4567402228533975e-06, + "loss": 1.0566, + "step": 32485 + }, + { + "epoch": 7.09, + "learning_rate": 1.4566309809919163e-06, + "loss": 0.9, + "step": 32486 + }, + { + "epoch": 7.09, + "learning_rate": 1.4565217391304347e-06, + "loss": 1.6323, + "step": 32487 + }, + { + "epoch": 7.09, + "learning_rate": 1.4564124972689536e-06, + "loss": 1.1968, + "step": 32488 + }, + { + "epoch": 7.09, + "learning_rate": 1.4563032554074722e-06, + "loss": 0.3103, + "step": 32489 + }, + { + "epoch": 7.09, + "learning_rate": 1.456194013545991e-06, + "loss": 1.2445, + "step": 32490 + }, + { + "epoch": 7.09, + "learning_rate": 1.4560847716845098e-06, + "loss": 2.5879, + "step": 32491 + }, + { + "epoch": 7.09, + "learning_rate": 1.4559755298230282e-06, + "loss": 1.0613, + "step": 32492 + }, + { + "epoch": 7.09, + "learning_rate": 1.455866287961547e-06, + "loss": 0.8135, + "step": 32493 + }, + { + "epoch": 7.09, + "learning_rate": 1.4557570461000657e-06, + "loss": 0.7962, + "step": 32494 + }, + { + "epoch": 7.09, + "learning_rate": 1.4556478042385845e-06, + "loss": 0.589, + "step": 32495 + }, + { + "epoch": 7.09, + "learning_rate": 1.4555385623771029e-06, + "loss": 1.2843, + "step": 32496 + }, + { + "epoch": 7.09, + "learning_rate": 1.4554293205156217e-06, + "loss": 1.1746, + "step": 32497 + }, + { + "epoch": 7.09, + "learning_rate": 1.4553200786541405e-06, + "loss": 1.2167, + "step": 32498 + }, + { + "epoch": 7.09, + "learning_rate": 1.455210836792659e-06, + "loss": 1.7341, + "step": 32499 + }, + { + "epoch": 7.09, + "learning_rate": 1.4551015949311777e-06, + "loss": 1.5432, + "step": 32500 + }, + { + "epoch": 7.09, + "learning_rate": 1.4549923530696964e-06, + "loss": 0.3092, + "step": 32501 + }, + { + "epoch": 7.09, + "learning_rate": 1.4548831112082152e-06, + "loss": 0.4195, + "step": 32502 + }, + { + "epoch": 7.09, + "learning_rate": 1.454773869346734e-06, + "loss": 1.0463, + "step": 32503 + }, + { + "epoch": 7.09, + "learning_rate": 1.4546646274852524e-06, + "loss": 0.5791, + "step": 32504 + }, + { + "epoch": 7.09, + "learning_rate": 1.4545553856237712e-06, + "loss": 0.8994, + "step": 32505 + }, + { + "epoch": 7.09, + "learning_rate": 1.4544461437622898e-06, + "loss": 0.8261, + "step": 32506 + }, + { + "epoch": 7.09, + "learning_rate": 1.4543369019008087e-06, + "loss": 0.2998, + "step": 32507 + }, + { + "epoch": 7.09, + "learning_rate": 1.454227660039327e-06, + "loss": 0.9465, + "step": 32508 + }, + { + "epoch": 7.09, + "learning_rate": 1.4541184181778459e-06, + "loss": 0.6796, + "step": 32509 + }, + { + "epoch": 7.1, + "learning_rate": 1.4540091763163647e-06, + "loss": 0.2335, + "step": 32510 + }, + { + "epoch": 7.1, + "learning_rate": 1.453899934454883e-06, + "loss": 1.8344, + "step": 32511 + }, + { + "epoch": 7.1, + "learning_rate": 1.453790692593402e-06, + "loss": 1.8523, + "step": 32512 + }, + { + "epoch": 7.1, + "learning_rate": 1.4536814507319205e-06, + "loss": 1.6843, + "step": 32513 + }, + { + "epoch": 7.1, + "learning_rate": 1.4535722088704394e-06, + "loss": 0.9309, + "step": 32514 + }, + { + "epoch": 7.1, + "learning_rate": 1.4534629670089578e-06, + "loss": 1.6433, + "step": 32515 + }, + { + "epoch": 7.1, + "learning_rate": 1.4533537251474766e-06, + "loss": 1.1716, + "step": 32516 + }, + { + "epoch": 7.1, + "learning_rate": 1.4532444832859954e-06, + "loss": 1.4174, + "step": 32517 + }, + { + "epoch": 7.1, + "learning_rate": 1.453135241424514e-06, + "loss": 1.3377, + "step": 32518 + }, + { + "epoch": 7.1, + "learning_rate": 1.4530259995630328e-06, + "loss": 2.5095, + "step": 32519 + }, + { + "epoch": 7.1, + "learning_rate": 1.4529167577015512e-06, + "loss": 0.819, + "step": 32520 + }, + { + "epoch": 7.1, + "learning_rate": 1.45280751584007e-06, + "loss": 0.8751, + "step": 32521 + }, + { + "epoch": 7.1, + "learning_rate": 1.4526982739785887e-06, + "loss": 1.1591, + "step": 32522 + }, + { + "epoch": 7.1, + "learning_rate": 1.4525890321171075e-06, + "loss": 0.8148, + "step": 32523 + }, + { + "epoch": 7.1, + "learning_rate": 1.452479790255626e-06, + "loss": 1.1069, + "step": 32524 + }, + { + "epoch": 7.1, + "learning_rate": 1.4523705483941447e-06, + "loss": 1.4783, + "step": 32525 + }, + { + "epoch": 7.1, + "learning_rate": 1.4522613065326635e-06, + "loss": 1.4541, + "step": 32526 + }, + { + "epoch": 7.1, + "learning_rate": 1.452152064671182e-06, + "loss": 0.9413, + "step": 32527 + }, + { + "epoch": 7.1, + "learning_rate": 1.4520428228097008e-06, + "loss": 0.6889, + "step": 32528 + }, + { + "epoch": 7.1, + "learning_rate": 1.4519335809482194e-06, + "loss": 2.1238, + "step": 32529 + }, + { + "epoch": 7.1, + "learning_rate": 1.4518243390867382e-06, + "loss": 1.4926, + "step": 32530 + }, + { + "epoch": 7.1, + "learning_rate": 1.451715097225257e-06, + "loss": 1.4964, + "step": 32531 + }, + { + "epoch": 7.1, + "learning_rate": 1.4516058553637754e-06, + "loss": 0.7721, + "step": 32532 + }, + { + "epoch": 7.1, + "learning_rate": 1.4514966135022942e-06, + "loss": 1.7156, + "step": 32533 + }, + { + "epoch": 7.1, + "learning_rate": 1.4513873716408128e-06, + "loss": 0.8834, + "step": 32534 + }, + { + "epoch": 7.1, + "learning_rate": 1.4512781297793317e-06, + "loss": 1.1313, + "step": 32535 + }, + { + "epoch": 7.1, + "learning_rate": 1.4511688879178503e-06, + "loss": 0.7074, + "step": 32536 + }, + { + "epoch": 7.1, + "learning_rate": 1.4510596460563689e-06, + "loss": 2.4547, + "step": 32537 + }, + { + "epoch": 7.1, + "learning_rate": 1.4509504041948877e-06, + "loss": 1.4031, + "step": 32538 + }, + { + "epoch": 7.1, + "learning_rate": 1.4508411623334061e-06, + "loss": 1.2266, + "step": 32539 + }, + { + "epoch": 7.1, + "learning_rate": 1.450731920471925e-06, + "loss": 1.6095, + "step": 32540 + }, + { + "epoch": 7.1, + "learning_rate": 1.4506226786104435e-06, + "loss": 1.6594, + "step": 32541 + }, + { + "epoch": 7.1, + "learning_rate": 1.4505134367489624e-06, + "loss": 1.7953, + "step": 32542 + }, + { + "epoch": 7.1, + "learning_rate": 1.4504041948874812e-06, + "loss": 1.203, + "step": 32543 + }, + { + "epoch": 7.1, + "learning_rate": 1.4502949530259996e-06, + "loss": 1.3748, + "step": 32544 + }, + { + "epoch": 7.1, + "learning_rate": 1.4501857111645184e-06, + "loss": 0.2768, + "step": 32545 + }, + { + "epoch": 7.1, + "learning_rate": 1.450076469303037e-06, + "loss": 0.9021, + "step": 32546 + }, + { + "epoch": 7.1, + "learning_rate": 1.4499672274415558e-06, + "loss": 2.301, + "step": 32547 + }, + { + "epoch": 7.1, + "learning_rate": 1.4498579855800742e-06, + "loss": 1.7226, + "step": 32548 + }, + { + "epoch": 7.1, + "learning_rate": 1.449748743718593e-06, + "loss": 1.4425, + "step": 32549 + }, + { + "epoch": 7.1, + "learning_rate": 1.4496395018571119e-06, + "loss": 0.5019, + "step": 32550 + }, + { + "epoch": 7.1, + "learning_rate": 1.4495302599956305e-06, + "loss": 0.5843, + "step": 32551 + }, + { + "epoch": 7.1, + "learning_rate": 1.4494210181341491e-06, + "loss": 0.1449, + "step": 32552 + }, + { + "epoch": 7.1, + "learning_rate": 1.4493117762726677e-06, + "loss": 1.2641, + "step": 32553 + }, + { + "epoch": 7.1, + "learning_rate": 1.4492025344111865e-06, + "loss": 2.2185, + "step": 32554 + }, + { + "epoch": 7.1, + "learning_rate": 1.449093292549705e-06, + "loss": 1.7962, + "step": 32555 + }, + { + "epoch": 7.11, + "learning_rate": 1.4489840506882238e-06, + "loss": 1.751, + "step": 32556 + }, + { + "epoch": 7.11, + "learning_rate": 1.4488748088267426e-06, + "loss": 1.7492, + "step": 32557 + }, + { + "epoch": 7.11, + "learning_rate": 1.4487655669652612e-06, + "loss": 0.1364, + "step": 32558 + }, + { + "epoch": 7.11, + "learning_rate": 1.44865632510378e-06, + "loss": 0.9049, + "step": 32559 + }, + { + "epoch": 7.11, + "learning_rate": 1.4485470832422984e-06, + "loss": 0.7755, + "step": 32560 + }, + { + "epoch": 7.11, + "learning_rate": 1.4484378413808172e-06, + "loss": 1.0761, + "step": 32561 + }, + { + "epoch": 7.11, + "learning_rate": 1.4483285995193359e-06, + "loss": 0.3053, + "step": 32562 + }, + { + "epoch": 7.11, + "learning_rate": 1.4482193576578547e-06, + "loss": 1.8238, + "step": 32563 + }, + { + "epoch": 7.11, + "learning_rate": 1.4481101157963733e-06, + "loss": 1.7012, + "step": 32564 + }, + { + "epoch": 7.11, + "learning_rate": 1.448000873934892e-06, + "loss": 1.5871, + "step": 32565 + }, + { + "epoch": 7.11, + "learning_rate": 1.4478916320734107e-06, + "loss": 0.8153, + "step": 32566 + }, + { + "epoch": 7.11, + "learning_rate": 1.4477823902119293e-06, + "loss": 0.5961, + "step": 32567 + }, + { + "epoch": 7.11, + "learning_rate": 1.447673148350448e-06, + "loss": 0.4722, + "step": 32568 + }, + { + "epoch": 7.11, + "learning_rate": 1.4475639064889666e-06, + "loss": 0.3436, + "step": 32569 + }, + { + "epoch": 7.11, + "learning_rate": 1.4474546646274854e-06, + "loss": 0.8684, + "step": 32570 + }, + { + "epoch": 7.11, + "learning_rate": 1.4473454227660042e-06, + "loss": 1.8272, + "step": 32571 + }, + { + "epoch": 7.11, + "learning_rate": 1.4472361809045226e-06, + "loss": 0.8626, + "step": 32572 + }, + { + "epoch": 7.11, + "learning_rate": 1.4471269390430414e-06, + "loss": 0.9319, + "step": 32573 + }, + { + "epoch": 7.11, + "learning_rate": 1.44701769718156e-06, + "loss": 0.7207, + "step": 32574 + }, + { + "epoch": 7.11, + "learning_rate": 1.4469084553200789e-06, + "loss": 0.1703, + "step": 32575 + }, + { + "epoch": 7.11, + "learning_rate": 1.4467992134585977e-06, + "loss": 1.6554, + "step": 32576 + }, + { + "epoch": 7.11, + "learning_rate": 1.446689971597116e-06, + "loss": 0.9617, + "step": 32577 + }, + { + "epoch": 7.11, + "learning_rate": 1.446580729735635e-06, + "loss": 1.551, + "step": 32578 + }, + { + "epoch": 7.11, + "learning_rate": 1.4464714878741535e-06, + "loss": 1.2309, + "step": 32579 + }, + { + "epoch": 7.11, + "learning_rate": 1.4463622460126721e-06, + "loss": 0.6916, + "step": 32580 + }, + { + "epoch": 7.11, + "learning_rate": 1.4462530041511907e-06, + "loss": 1.2147, + "step": 32581 + }, + { + "epoch": 7.11, + "learning_rate": 1.4461437622897096e-06, + "loss": 2.7118, + "step": 32582 + }, + { + "epoch": 7.11, + "learning_rate": 1.4460345204282284e-06, + "loss": 1.0486, + "step": 32583 + }, + { + "epoch": 7.11, + "learning_rate": 1.4459252785667468e-06, + "loss": 0.73, + "step": 32584 + }, + { + "epoch": 7.11, + "learning_rate": 1.4458160367052656e-06, + "loss": 2.2579, + "step": 32585 + }, + { + "epoch": 7.11, + "learning_rate": 1.4457067948437842e-06, + "loss": 2.1943, + "step": 32586 + }, + { + "epoch": 7.11, + "learning_rate": 1.445597552982303e-06, + "loss": 1.0887, + "step": 32587 + }, + { + "epoch": 7.11, + "learning_rate": 1.4454883111208214e-06, + "loss": 0.4819, + "step": 32588 + }, + { + "epoch": 7.11, + "learning_rate": 1.4453790692593403e-06, + "loss": 0.7854, + "step": 32589 + }, + { + "epoch": 7.11, + "learning_rate": 1.445269827397859e-06, + "loss": 1.3898, + "step": 32590 + }, + { + "epoch": 7.11, + "learning_rate": 1.4451605855363777e-06, + "loss": 2.7807, + "step": 32591 + }, + { + "epoch": 7.11, + "learning_rate": 1.4450513436748965e-06, + "loss": 1.5182, + "step": 32592 + }, + { + "epoch": 7.11, + "learning_rate": 1.444942101813415e-06, + "loss": 1.457, + "step": 32593 + }, + { + "epoch": 7.11, + "learning_rate": 1.4448328599519337e-06, + "loss": 2.084, + "step": 32594 + }, + { + "epoch": 7.11, + "learning_rate": 1.4447236180904523e-06, + "loss": 2.8623, + "step": 32595 + }, + { + "epoch": 7.11, + "learning_rate": 1.444614376228971e-06, + "loss": 2.4392, + "step": 32596 + }, + { + "epoch": 7.11, + "learning_rate": 1.4445051343674898e-06, + "loss": 4.1114, + "step": 32597 + }, + { + "epoch": 7.11, + "learning_rate": 1.4443958925060084e-06, + "loss": 0.8582, + "step": 32598 + }, + { + "epoch": 7.11, + "learning_rate": 1.4442866506445272e-06, + "loss": 1.5386, + "step": 32599 + }, + { + "epoch": 7.11, + "learning_rate": 1.4441774087830456e-06, + "loss": 1.8339, + "step": 32600 + }, + { + "epoch": 7.12, + "learning_rate": 1.4440681669215644e-06, + "loss": 1.0063, + "step": 32601 + }, + { + "epoch": 7.12, + "learning_rate": 1.443958925060083e-06, + "loss": 0.5748, + "step": 32602 + }, + { + "epoch": 7.12, + "learning_rate": 1.4438496831986019e-06, + "loss": 3.1183, + "step": 32603 + }, + { + "epoch": 7.12, + "learning_rate": 1.4437404413371207e-06, + "loss": 0.9215, + "step": 32604 + }, + { + "epoch": 7.12, + "learning_rate": 1.443631199475639e-06, + "loss": 0.3705, + "step": 32605 + }, + { + "epoch": 7.12, + "learning_rate": 1.443521957614158e-06, + "loss": 1.0698, + "step": 32606 + }, + { + "epoch": 7.12, + "learning_rate": 1.4434127157526765e-06, + "loss": 3.1107, + "step": 32607 + }, + { + "epoch": 7.12, + "learning_rate": 1.4433034738911951e-06, + "loss": 0.8693, + "step": 32608 + }, + { + "epoch": 7.12, + "learning_rate": 1.443194232029714e-06, + "loss": 1.196, + "step": 32609 + }, + { + "epoch": 7.12, + "learning_rate": 1.4430849901682326e-06, + "loss": 0.5343, + "step": 32610 + }, + { + "epoch": 7.12, + "learning_rate": 1.4429757483067514e-06, + "loss": 0.3165, + "step": 32611 + }, + { + "epoch": 7.12, + "learning_rate": 1.4428665064452698e-06, + "loss": 2.9236, + "step": 32612 + }, + { + "epoch": 7.12, + "learning_rate": 1.4427572645837886e-06, + "loss": 0.4859, + "step": 32613 + }, + { + "epoch": 7.12, + "learning_rate": 1.4426480227223072e-06, + "loss": 0.3914, + "step": 32614 + }, + { + "epoch": 7.12, + "learning_rate": 1.442538780860826e-06, + "loss": 1.4769, + "step": 32615 + }, + { + "epoch": 7.12, + "learning_rate": 1.4424295389993449e-06, + "loss": 1.7602, + "step": 32616 + }, + { + "epoch": 7.12, + "learning_rate": 1.4423202971378633e-06, + "loss": 0.7509, + "step": 32617 + }, + { + "epoch": 7.12, + "learning_rate": 1.442211055276382e-06, + "loss": 1.1218, + "step": 32618 + }, + { + "epoch": 7.12, + "learning_rate": 1.4421018134149007e-06, + "loss": 1.4792, + "step": 32619 + }, + { + "epoch": 7.12, + "learning_rate": 1.4419925715534195e-06, + "loss": 1.3564, + "step": 32620 + }, + { + "epoch": 7.12, + "learning_rate": 1.441883329691938e-06, + "loss": 0.6551, + "step": 32621 + }, + { + "epoch": 7.12, + "learning_rate": 1.4417740878304568e-06, + "loss": 1.1324, + "step": 32622 + }, + { + "epoch": 7.12, + "learning_rate": 1.4416648459689756e-06, + "loss": 0.7819, + "step": 32623 + }, + { + "epoch": 7.12, + "learning_rate": 1.441555604107494e-06, + "loss": 0.3514, + "step": 32624 + }, + { + "epoch": 7.12, + "learning_rate": 1.4414463622460128e-06, + "loss": 1.1352, + "step": 32625 + }, + { + "epoch": 7.12, + "learning_rate": 1.4413371203845314e-06, + "loss": 2.0198, + "step": 32626 + }, + { + "epoch": 7.12, + "learning_rate": 1.4412278785230502e-06, + "loss": 1.14, + "step": 32627 + }, + { + "epoch": 7.12, + "learning_rate": 1.4411186366615686e-06, + "loss": 1.7409, + "step": 32628 + }, + { + "epoch": 7.12, + "learning_rate": 1.4410093948000875e-06, + "loss": 0.727, + "step": 32629 + }, + { + "epoch": 7.12, + "learning_rate": 1.4409001529386063e-06, + "loss": 1.4035, + "step": 32630 + }, + { + "epoch": 7.12, + "learning_rate": 1.4407909110771249e-06, + "loss": 1.5388, + "step": 32631 + }, + { + "epoch": 7.12, + "learning_rate": 1.4406816692156437e-06, + "loss": 1.4202, + "step": 32632 + }, + { + "epoch": 7.12, + "learning_rate": 1.4405724273541621e-06, + "loss": 2.5354, + "step": 32633 + }, + { + "epoch": 7.12, + "learning_rate": 1.440463185492681e-06, + "loss": 0.9069, + "step": 32634 + }, + { + "epoch": 7.12, + "learning_rate": 1.4403539436311995e-06, + "loss": 0.791, + "step": 32635 + }, + { + "epoch": 7.12, + "learning_rate": 1.4402447017697182e-06, + "loss": 0.221, + "step": 32636 + }, + { + "epoch": 7.12, + "learning_rate": 1.440135459908237e-06, + "loss": 1.6503, + "step": 32637 + }, + { + "epoch": 7.12, + "learning_rate": 1.4400262180467556e-06, + "loss": 0.8175, + "step": 32638 + }, + { + "epoch": 7.12, + "learning_rate": 1.4399169761852744e-06, + "loss": 1.6151, + "step": 32639 + }, + { + "epoch": 7.12, + "learning_rate": 1.4398077343237928e-06, + "loss": 2.5913, + "step": 32640 + }, + { + "epoch": 7.12, + "learning_rate": 1.4396984924623116e-06, + "loss": 1.4356, + "step": 32641 + }, + { + "epoch": 7.12, + "learning_rate": 1.4395892506008305e-06, + "loss": 0.9739, + "step": 32642 + }, + { + "epoch": 7.12, + "learning_rate": 1.439480008739349e-06, + "loss": 1.5622, + "step": 32643 + }, + { + "epoch": 7.12, + "learning_rate": 1.4393707668778679e-06, + "loss": 0.4723, + "step": 32644 + }, + { + "epoch": 7.12, + "learning_rate": 1.4392615250163863e-06, + "loss": 1.2751, + "step": 32645 + }, + { + "epoch": 7.12, + "learning_rate": 1.4391522831549051e-06, + "loss": 3.464, + "step": 32646 + }, + { + "epoch": 7.13, + "learning_rate": 1.4390430412934237e-06, + "loss": 0.3699, + "step": 32647 + }, + { + "epoch": 7.13, + "learning_rate": 1.4389337994319425e-06, + "loss": 2.344, + "step": 32648 + }, + { + "epoch": 7.13, + "learning_rate": 1.4388245575704612e-06, + "loss": 1.1958, + "step": 32649 + }, + { + "epoch": 7.13, + "learning_rate": 1.4387153157089798e-06, + "loss": 1.2459, + "step": 32650 + }, + { + "epoch": 7.13, + "learning_rate": 1.4386060738474986e-06, + "loss": 0.817, + "step": 32651 + }, + { + "epoch": 7.13, + "learning_rate": 1.438496831986017e-06, + "loss": 1.0234, + "step": 32652 + }, + { + "epoch": 7.13, + "learning_rate": 1.4383875901245358e-06, + "loss": 1.0951, + "step": 32653 + }, + { + "epoch": 7.13, + "learning_rate": 1.4382783482630544e-06, + "loss": 1.0271, + "step": 32654 + }, + { + "epoch": 7.13, + "learning_rate": 1.4381691064015732e-06, + "loss": 0.5803, + "step": 32655 + }, + { + "epoch": 7.13, + "learning_rate": 1.438059864540092e-06, + "loss": 0.627, + "step": 32656 + }, + { + "epoch": 7.13, + "learning_rate": 1.4379506226786105e-06, + "loss": 2.1949, + "step": 32657 + }, + { + "epoch": 7.13, + "learning_rate": 1.4378413808171293e-06, + "loss": 1.5186, + "step": 32658 + }, + { + "epoch": 7.13, + "learning_rate": 1.437732138955648e-06, + "loss": 1.0386, + "step": 32659 + }, + { + "epoch": 7.13, + "learning_rate": 1.4376228970941667e-06, + "loss": 0.084, + "step": 32660 + }, + { + "epoch": 7.13, + "learning_rate": 1.4375136552326851e-06, + "loss": 2.6072, + "step": 32661 + }, + { + "epoch": 7.13, + "learning_rate": 1.437404413371204e-06, + "loss": 2.7882, + "step": 32662 + }, + { + "epoch": 7.13, + "learning_rate": 1.4372951715097228e-06, + "loss": 3.094, + "step": 32663 + }, + { + "epoch": 7.13, + "learning_rate": 1.4371859296482414e-06, + "loss": 1.3927, + "step": 32664 + }, + { + "epoch": 7.13, + "learning_rate": 1.43707668778676e-06, + "loss": 1.7033, + "step": 32665 + }, + { + "epoch": 7.13, + "learning_rate": 1.4369674459252786e-06, + "loss": 2.3231, + "step": 32666 + }, + { + "epoch": 7.13, + "learning_rate": 1.4368582040637974e-06, + "loss": 1.7728, + "step": 32667 + }, + { + "epoch": 7.13, + "learning_rate": 1.4367489622023158e-06, + "loss": 1.0401, + "step": 32668 + }, + { + "epoch": 7.13, + "learning_rate": 1.4366397203408346e-06, + "loss": 1.0898, + "step": 32669 + }, + { + "epoch": 7.13, + "learning_rate": 1.4365304784793535e-06, + "loss": 1.4764, + "step": 32670 + }, + { + "epoch": 7.13, + "learning_rate": 1.436421236617872e-06, + "loss": 0.9982, + "step": 32671 + }, + { + "epoch": 7.13, + "learning_rate": 1.436311994756391e-06, + "loss": 1.6048, + "step": 32672 + }, + { + "epoch": 7.13, + "learning_rate": 1.4362027528949093e-06, + "loss": 1.813, + "step": 32673 + }, + { + "epoch": 7.13, + "learning_rate": 1.4360935110334281e-06, + "loss": 0.5996, + "step": 32674 + }, + { + "epoch": 7.13, + "learning_rate": 1.435984269171947e-06, + "loss": 0.2067, + "step": 32675 + }, + { + "epoch": 7.13, + "learning_rate": 1.4358750273104656e-06, + "loss": 1.5871, + "step": 32676 + }, + { + "epoch": 7.13, + "learning_rate": 1.4357657854489842e-06, + "loss": 2.3293, + "step": 32677 + }, + { + "epoch": 7.13, + "learning_rate": 1.4356565435875028e-06, + "loss": 0.7126, + "step": 32678 + }, + { + "epoch": 7.13, + "learning_rate": 1.4355473017260216e-06, + "loss": 1.2672, + "step": 32679 + }, + { + "epoch": 7.13, + "learning_rate": 1.43543805986454e-06, + "loss": 2.1007, + "step": 32680 + }, + { + "epoch": 7.13, + "learning_rate": 1.4353288180030588e-06, + "loss": 1.3539, + "step": 32681 + }, + { + "epoch": 7.13, + "learning_rate": 1.4352195761415776e-06, + "loss": 0.9413, + "step": 32682 + }, + { + "epoch": 7.13, + "learning_rate": 1.4351103342800963e-06, + "loss": 1.2858, + "step": 32683 + }, + { + "epoch": 7.13, + "learning_rate": 1.435001092418615e-06, + "loss": 3.0338, + "step": 32684 + }, + { + "epoch": 7.13, + "learning_rate": 1.4348918505571335e-06, + "loss": 1.2563, + "step": 32685 + }, + { + "epoch": 7.13, + "learning_rate": 1.4347826086956523e-06, + "loss": 0.8368, + "step": 32686 + }, + { + "epoch": 7.13, + "learning_rate": 1.434673366834171e-06, + "loss": 1.1388, + "step": 32687 + }, + { + "epoch": 7.13, + "learning_rate": 1.4345641249726897e-06, + "loss": 2.3572, + "step": 32688 + }, + { + "epoch": 7.13, + "learning_rate": 1.4344548831112083e-06, + "loss": 0.2329, + "step": 32689 + }, + { + "epoch": 7.13, + "learning_rate": 1.434345641249727e-06, + "loss": 1.7957, + "step": 32690 + }, + { + "epoch": 7.13, + "learning_rate": 1.4342363993882458e-06, + "loss": 0.6184, + "step": 32691 + }, + { + "epoch": 7.13, + "learning_rate": 1.4341271575267644e-06, + "loss": 1.7211, + "step": 32692 + }, + { + "epoch": 7.14, + "learning_rate": 1.434017915665283e-06, + "loss": 0.038, + "step": 32693 + }, + { + "epoch": 7.14, + "learning_rate": 1.4339086738038016e-06, + "loss": 1.4357, + "step": 32694 + }, + { + "epoch": 7.14, + "learning_rate": 1.4337994319423204e-06, + "loss": 2.5654, + "step": 32695 + }, + { + "epoch": 7.14, + "learning_rate": 1.4336901900808393e-06, + "loss": 0.2813, + "step": 32696 + }, + { + "epoch": 7.14, + "learning_rate": 1.4335809482193577e-06, + "loss": 0.7263, + "step": 32697 + }, + { + "epoch": 7.14, + "learning_rate": 1.4334717063578765e-06, + "loss": 1.1763, + "step": 32698 + }, + { + "epoch": 7.14, + "learning_rate": 1.433362464496395e-06, + "loss": 2.0136, + "step": 32699 + }, + { + "epoch": 7.14, + "learning_rate": 1.433253222634914e-06, + "loss": 0.0698, + "step": 32700 + }, + { + "epoch": 7.14, + "learning_rate": 1.4331439807734323e-06, + "loss": 2.1701, + "step": 32701 + }, + { + "epoch": 7.14, + "learning_rate": 1.4330347389119511e-06, + "loss": 1.8868, + "step": 32702 + }, + { + "epoch": 7.14, + "learning_rate": 1.43292549705047e-06, + "loss": 2.3031, + "step": 32703 + }, + { + "epoch": 7.14, + "learning_rate": 1.4328162551889886e-06, + "loss": 0.6237, + "step": 32704 + }, + { + "epoch": 7.14, + "learning_rate": 1.4327070133275072e-06, + "loss": 0.2519, + "step": 32705 + }, + { + "epoch": 7.14, + "learning_rate": 1.4325977714660258e-06, + "loss": 1.273, + "step": 32706 + }, + { + "epoch": 7.14, + "learning_rate": 1.4324885296045446e-06, + "loss": 0.0573, + "step": 32707 + }, + { + "epoch": 7.14, + "learning_rate": 1.4323792877430634e-06, + "loss": 2.3015, + "step": 32708 + }, + { + "epoch": 7.14, + "learning_rate": 1.4322700458815818e-06, + "loss": 0.3607, + "step": 32709 + }, + { + "epoch": 7.14, + "learning_rate": 1.4321608040201007e-06, + "loss": 0.2112, + "step": 32710 + }, + { + "epoch": 7.14, + "learning_rate": 1.4320515621586193e-06, + "loss": 2.0513, + "step": 32711 + }, + { + "epoch": 7.14, + "learning_rate": 1.431942320297138e-06, + "loss": 4.8936, + "step": 32712 + }, + { + "epoch": 7.14, + "learning_rate": 1.4318330784356565e-06, + "loss": 1.1055, + "step": 32713 + }, + { + "epoch": 7.14, + "learning_rate": 1.4317238365741753e-06, + "loss": 1.6036, + "step": 32714 + }, + { + "epoch": 7.14, + "learning_rate": 1.4316145947126941e-06, + "loss": 2.5825, + "step": 32715 + }, + { + "epoch": 7.14, + "learning_rate": 1.4315053528512127e-06, + "loss": 1.4571, + "step": 32716 + }, + { + "epoch": 7.14, + "learning_rate": 1.4313961109897316e-06, + "loss": 1.5965, + "step": 32717 + }, + { + "epoch": 7.14, + "learning_rate": 1.43128686912825e-06, + "loss": 2.942, + "step": 32718 + }, + { + "epoch": 7.14, + "learning_rate": 1.4311776272667688e-06, + "loss": 1.7995, + "step": 32719 + }, + { + "epoch": 7.14, + "learning_rate": 1.4310683854052874e-06, + "loss": 0.5897, + "step": 32720 + }, + { + "epoch": 7.14, + "learning_rate": 1.430959143543806e-06, + "loss": 0.6761, + "step": 32721 + }, + { + "epoch": 7.14, + "learning_rate": 1.4308499016823248e-06, + "loss": 0.8673, + "step": 32722 + }, + { + "epoch": 7.14, + "learning_rate": 1.4307406598208434e-06, + "loss": 1.6139, + "step": 32723 + }, + { + "epoch": 7.14, + "learning_rate": 1.4306314179593623e-06, + "loss": 1.5571, + "step": 32724 + }, + { + "epoch": 7.14, + "learning_rate": 1.4305221760978807e-06, + "loss": 0.415, + "step": 32725 + }, + { + "epoch": 7.14, + "learning_rate": 1.4304129342363995e-06, + "loss": 0.3206, + "step": 32726 + }, + { + "epoch": 7.14, + "learning_rate": 1.430303692374918e-06, + "loss": 2.1008, + "step": 32727 + }, + { + "epoch": 7.14, + "learning_rate": 1.430194450513437e-06, + "loss": 0.4248, + "step": 32728 + }, + { + "epoch": 7.14, + "learning_rate": 1.4300852086519558e-06, + "loss": 1.463, + "step": 32729 + }, + { + "epoch": 7.14, + "learning_rate": 1.4299759667904742e-06, + "loss": 2.1061, + "step": 32730 + }, + { + "epoch": 7.14, + "learning_rate": 1.429866724928993e-06, + "loss": 1.0478, + "step": 32731 + }, + { + "epoch": 7.14, + "learning_rate": 1.4297574830675116e-06, + "loss": 1.3235, + "step": 32732 + }, + { + "epoch": 7.14, + "learning_rate": 1.4296482412060302e-06, + "loss": 1.1175, + "step": 32733 + }, + { + "epoch": 7.14, + "learning_rate": 1.4295389993445488e-06, + "loss": 2.1208, + "step": 32734 + }, + { + "epoch": 7.14, + "learning_rate": 1.4294297574830676e-06, + "loss": 1.1918, + "step": 32735 + }, + { + "epoch": 7.14, + "learning_rate": 1.4293205156215865e-06, + "loss": 0.6098, + "step": 32736 + }, + { + "epoch": 7.14, + "learning_rate": 1.4292112737601049e-06, + "loss": 1.2798, + "step": 32737 + }, + { + "epoch": 7.14, + "learning_rate": 1.4291020318986237e-06, + "loss": 0.3723, + "step": 32738 + }, + { + "epoch": 7.15, + "learning_rate": 1.4289927900371423e-06, + "loss": 2.6629, + "step": 32739 + }, + { + "epoch": 7.15, + "learning_rate": 1.4288835481756611e-06, + "loss": 3.5522, + "step": 32740 + }, + { + "epoch": 7.15, + "learning_rate": 1.42877430631418e-06, + "loss": 1.9893, + "step": 32741 + }, + { + "epoch": 7.15, + "learning_rate": 1.4286650644526983e-06, + "loss": 1.4762, + "step": 32742 + }, + { + "epoch": 7.15, + "learning_rate": 1.4285558225912172e-06, + "loss": 1.7485, + "step": 32743 + }, + { + "epoch": 7.15, + "learning_rate": 1.4284465807297358e-06, + "loss": 1.0509, + "step": 32744 + }, + { + "epoch": 7.15, + "learning_rate": 1.4283373388682546e-06, + "loss": 1.5179, + "step": 32745 + }, + { + "epoch": 7.15, + "learning_rate": 1.428228097006773e-06, + "loss": 1.831, + "step": 32746 + }, + { + "epoch": 7.15, + "learning_rate": 1.4281188551452918e-06, + "loss": 0.7718, + "step": 32747 + }, + { + "epoch": 7.15, + "learning_rate": 1.4280096132838106e-06, + "loss": 1.4086, + "step": 32748 + }, + { + "epoch": 7.15, + "learning_rate": 1.427900371422329e-06, + "loss": 0.4278, + "step": 32749 + }, + { + "epoch": 7.15, + "learning_rate": 1.4277911295608479e-06, + "loss": 1.9001, + "step": 32750 + }, + { + "epoch": 7.15, + "learning_rate": 1.4276818876993665e-06, + "loss": 0.2393, + "step": 32751 + }, + { + "epoch": 7.15, + "learning_rate": 1.4275726458378853e-06, + "loss": 1.2255, + "step": 32752 + }, + { + "epoch": 7.15, + "learning_rate": 1.4274634039764037e-06, + "loss": 1.0197, + "step": 32753 + }, + { + "epoch": 7.15, + "learning_rate": 1.4273541621149225e-06, + "loss": 1.2661, + "step": 32754 + }, + { + "epoch": 7.15, + "learning_rate": 1.4272449202534413e-06, + "loss": 2.1875, + "step": 32755 + }, + { + "epoch": 7.15, + "learning_rate": 1.42713567839196e-06, + "loss": 0.7927, + "step": 32756 + }, + { + "epoch": 7.15, + "learning_rate": 1.4270264365304788e-06, + "loss": 1.3843, + "step": 32757 + }, + { + "epoch": 7.15, + "learning_rate": 1.4269171946689972e-06, + "loss": 1.9846, + "step": 32758 + }, + { + "epoch": 7.15, + "learning_rate": 1.426807952807516e-06, + "loss": 0.4258, + "step": 32759 + }, + { + "epoch": 7.15, + "learning_rate": 1.4266987109460346e-06, + "loss": 1.1182, + "step": 32760 + }, + { + "epoch": 7.15, + "learning_rate": 1.4265894690845532e-06, + "loss": 0.8938, + "step": 32761 + }, + { + "epoch": 7.15, + "learning_rate": 1.426480227223072e-06, + "loss": 2.3135, + "step": 32762 + }, + { + "epoch": 7.15, + "learning_rate": 1.4263709853615906e-06, + "loss": 1.8334, + "step": 32763 + }, + { + "epoch": 7.15, + "learning_rate": 1.4262617435001095e-06, + "loss": 0.8154, + "step": 32764 + }, + { + "epoch": 7.15, + "learning_rate": 1.4261525016386279e-06, + "loss": 2.0944, + "step": 32765 + }, + { + "epoch": 7.15, + "learning_rate": 1.4260432597771467e-06, + "loss": 1.6671, + "step": 32766 + }, + { + "epoch": 7.15, + "learning_rate": 1.4259340179156653e-06, + "loss": 1.1572, + "step": 32767 + }, + { + "epoch": 7.15, + "learning_rate": 1.4258247760541841e-06, + "loss": 0.3949, + "step": 32768 + }, + { + "epoch": 7.15, + "learning_rate": 1.425715534192703e-06, + "loss": 0.2027, + "step": 32769 + }, + { + "epoch": 7.15, + "learning_rate": 1.4256062923312213e-06, + "loss": 2.2821, + "step": 32770 + }, + { + "epoch": 7.15, + "learning_rate": 1.4254970504697402e-06, + "loss": 1.2735, + "step": 32771 + }, + { + "epoch": 7.15, + "learning_rate": 1.4253878086082588e-06, + "loss": 1.7384, + "step": 32772 + }, + { + "epoch": 7.15, + "learning_rate": 1.4252785667467776e-06, + "loss": 0.5977, + "step": 32773 + }, + { + "epoch": 7.15, + "learning_rate": 1.425169324885296e-06, + "loss": 0.2221, + "step": 32774 + }, + { + "epoch": 7.15, + "learning_rate": 1.4250600830238148e-06, + "loss": 0.1712, + "step": 32775 + }, + { + "epoch": 7.15, + "learning_rate": 1.4249508411623336e-06, + "loss": 0.6583, + "step": 32776 + }, + { + "epoch": 7.15, + "learning_rate": 1.424841599300852e-06, + "loss": 2.2788, + "step": 32777 + }, + { + "epoch": 7.15, + "learning_rate": 1.4247323574393709e-06, + "loss": 1.3697, + "step": 32778 + }, + { + "epoch": 7.15, + "learning_rate": 1.4246231155778895e-06, + "loss": 0.9693, + "step": 32779 + }, + { + "epoch": 7.15, + "learning_rate": 1.4245138737164083e-06, + "loss": 0.7467, + "step": 32780 + }, + { + "epoch": 7.15, + "learning_rate": 1.4244046318549271e-06, + "loss": 1.5869, + "step": 32781 + }, + { + "epoch": 7.15, + "learning_rate": 1.4242953899934455e-06, + "loss": 0.6572, + "step": 32782 + }, + { + "epoch": 7.15, + "learning_rate": 1.4241861481319643e-06, + "loss": 0.5551, + "step": 32783 + }, + { + "epoch": 7.15, + "learning_rate": 1.424076906270483e-06, + "loss": 0.6677, + "step": 32784 + }, + { + "epoch": 7.16, + "learning_rate": 1.4239676644090018e-06, + "loss": 2.4529, + "step": 32785 + }, + { + "epoch": 7.16, + "learning_rate": 1.4238584225475202e-06, + "loss": 1.5941, + "step": 32786 + }, + { + "epoch": 7.16, + "learning_rate": 1.423749180686039e-06, + "loss": 1.5266, + "step": 32787 + }, + { + "epoch": 7.16, + "learning_rate": 1.4236399388245578e-06, + "loss": 1.3838, + "step": 32788 + }, + { + "epoch": 7.16, + "learning_rate": 1.4235306969630764e-06, + "loss": 0.2706, + "step": 32789 + }, + { + "epoch": 7.16, + "learning_rate": 1.423421455101595e-06, + "loss": 0.974, + "step": 32790 + }, + { + "epoch": 7.16, + "learning_rate": 1.4233122132401137e-06, + "loss": 0.781, + "step": 32791 + }, + { + "epoch": 7.16, + "learning_rate": 1.4232029713786325e-06, + "loss": 1.5369, + "step": 32792 + }, + { + "epoch": 7.16, + "learning_rate": 1.4230937295171509e-06, + "loss": 1.7073, + "step": 32793 + }, + { + "epoch": 7.16, + "learning_rate": 1.4229844876556697e-06, + "loss": 1.8615, + "step": 32794 + }, + { + "epoch": 7.16, + "learning_rate": 1.4228752457941885e-06, + "loss": 2.1116, + "step": 32795 + }, + { + "epoch": 7.16, + "learning_rate": 1.4227660039327071e-06, + "loss": 0.2087, + "step": 32796 + }, + { + "epoch": 7.16, + "learning_rate": 1.422656762071226e-06, + "loss": 1.7189, + "step": 32797 + }, + { + "epoch": 7.16, + "learning_rate": 1.4225475202097444e-06, + "loss": 0.6862, + "step": 32798 + }, + { + "epoch": 7.16, + "learning_rate": 1.4224382783482632e-06, + "loss": 1.8291, + "step": 32799 + }, + { + "epoch": 7.16, + "learning_rate": 1.4223290364867818e-06, + "loss": 0.5606, + "step": 32800 + }, + { + "epoch": 7.16, + "learning_rate": 1.4222197946253006e-06, + "loss": 1.3229, + "step": 32801 + }, + { + "epoch": 7.16, + "learning_rate": 1.4221105527638192e-06, + "loss": 1.2197, + "step": 32802 + }, + { + "epoch": 7.16, + "learning_rate": 1.4220013109023378e-06, + "loss": 1.5839, + "step": 32803 + }, + { + "epoch": 7.16, + "learning_rate": 1.4218920690408567e-06, + "loss": 1.8294, + "step": 32804 + }, + { + "epoch": 7.16, + "learning_rate": 1.421782827179375e-06, + "loss": 3.1424, + "step": 32805 + }, + { + "epoch": 7.16, + "learning_rate": 1.4216735853178939e-06, + "loss": 1.1029, + "step": 32806 + }, + { + "epoch": 7.16, + "learning_rate": 1.4215643434564125e-06, + "loss": 2.1787, + "step": 32807 + }, + { + "epoch": 7.16, + "learning_rate": 1.4214551015949313e-06, + "loss": 1.1942, + "step": 32808 + }, + { + "epoch": 7.16, + "learning_rate": 1.4213458597334501e-06, + "loss": 1.7356, + "step": 32809 + }, + { + "epoch": 7.16, + "learning_rate": 1.4212366178719685e-06, + "loss": 1.6391, + "step": 32810 + }, + { + "epoch": 7.16, + "learning_rate": 1.4211273760104874e-06, + "loss": 0.8228, + "step": 32811 + }, + { + "epoch": 7.16, + "learning_rate": 1.421018134149006e-06, + "loss": 1.2154, + "step": 32812 + }, + { + "epoch": 7.16, + "learning_rate": 1.4209088922875248e-06, + "loss": 2.3287, + "step": 32813 + }, + { + "epoch": 7.16, + "learning_rate": 1.4207996504260436e-06, + "loss": 0.1654, + "step": 32814 + }, + { + "epoch": 7.16, + "learning_rate": 1.420690408564562e-06, + "loss": 0.5075, + "step": 32815 + }, + { + "epoch": 7.16, + "learning_rate": 1.4205811667030808e-06, + "loss": 1.4459, + "step": 32816 + }, + { + "epoch": 7.16, + "learning_rate": 1.4204719248415994e-06, + "loss": 1.4117, + "step": 32817 + }, + { + "epoch": 7.16, + "learning_rate": 1.420362682980118e-06, + "loss": 0.5841, + "step": 32818 + }, + { + "epoch": 7.16, + "learning_rate": 1.4202534411186367e-06, + "loss": 2.7669, + "step": 32819 + }, + { + "epoch": 7.16, + "learning_rate": 1.4201441992571555e-06, + "loss": 1.5077, + "step": 32820 + }, + { + "epoch": 7.16, + "learning_rate": 1.4200349573956743e-06, + "loss": 1.751, + "step": 32821 + }, + { + "epoch": 7.16, + "learning_rate": 1.4199257155341927e-06, + "loss": 0.6633, + "step": 32822 + }, + { + "epoch": 7.16, + "learning_rate": 1.4198164736727115e-06, + "loss": 0.773, + "step": 32823 + }, + { + "epoch": 7.16, + "learning_rate": 1.4197072318112301e-06, + "loss": 1.6982, + "step": 32824 + }, + { + "epoch": 7.16, + "learning_rate": 1.419597989949749e-06, + "loss": 0.9843, + "step": 32825 + }, + { + "epoch": 7.16, + "learning_rate": 1.4194887480882674e-06, + "loss": 1.1081, + "step": 32826 + }, + { + "epoch": 7.16, + "learning_rate": 1.4193795062267862e-06, + "loss": 0.3625, + "step": 32827 + }, + { + "epoch": 7.16, + "learning_rate": 1.419270264365305e-06, + "loss": 0.8396, + "step": 32828 + }, + { + "epoch": 7.16, + "learning_rate": 1.4191610225038236e-06, + "loss": 0.6116, + "step": 32829 + }, + { + "epoch": 7.16, + "learning_rate": 1.4190517806423422e-06, + "loss": 0.7066, + "step": 32830 + }, + { + "epoch": 7.17, + "learning_rate": 1.4189425387808608e-06, + "loss": 1.1076, + "step": 32831 + }, + { + "epoch": 7.17, + "learning_rate": 1.4188332969193797e-06, + "loss": 1.3352, + "step": 32832 + }, + { + "epoch": 7.17, + "learning_rate": 1.418724055057898e-06, + "loss": 1.2174, + "step": 32833 + }, + { + "epoch": 7.17, + "learning_rate": 1.4186148131964169e-06, + "loss": 0.7415, + "step": 32834 + }, + { + "epoch": 7.17, + "learning_rate": 1.4185055713349357e-06, + "loss": 0.7732, + "step": 32835 + }, + { + "epoch": 7.17, + "learning_rate": 1.4183963294734543e-06, + "loss": 1.686, + "step": 32836 + }, + { + "epoch": 7.17, + "learning_rate": 1.4182870876119731e-06, + "loss": 1.0732, + "step": 32837 + }, + { + "epoch": 7.17, + "learning_rate": 1.4181778457504915e-06, + "loss": 0.4401, + "step": 32838 + }, + { + "epoch": 7.17, + "learning_rate": 1.4180686038890104e-06, + "loss": 1.0653, + "step": 32839 + }, + { + "epoch": 7.17, + "learning_rate": 1.417959362027529e-06, + "loss": 0.4157, + "step": 32840 + }, + { + "epoch": 7.17, + "learning_rate": 1.4178501201660478e-06, + "loss": 1.8639, + "step": 32841 + }, + { + "epoch": 7.17, + "learning_rate": 1.4177408783045666e-06, + "loss": 1.819, + "step": 32842 + }, + { + "epoch": 7.17, + "learning_rate": 1.417631636443085e-06, + "loss": 0.8669, + "step": 32843 + }, + { + "epoch": 7.17, + "learning_rate": 1.4175223945816038e-06, + "loss": 0.7296, + "step": 32844 + }, + { + "epoch": 7.17, + "learning_rate": 1.4174131527201225e-06, + "loss": 3.8365, + "step": 32845 + }, + { + "epoch": 7.17, + "learning_rate": 1.417303910858641e-06, + "loss": 1.636, + "step": 32846 + }, + { + "epoch": 7.17, + "learning_rate": 1.41719466899716e-06, + "loss": 0.6948, + "step": 32847 + }, + { + "epoch": 7.17, + "learning_rate": 1.4170854271356785e-06, + "loss": 0.4794, + "step": 32848 + }, + { + "epoch": 7.17, + "learning_rate": 1.4169761852741973e-06, + "loss": 0.2105, + "step": 32849 + }, + { + "epoch": 7.17, + "learning_rate": 1.4168669434127157e-06, + "loss": 0.8478, + "step": 32850 + }, + { + "epoch": 7.17, + "learning_rate": 1.4167577015512346e-06, + "loss": 1.2799, + "step": 32851 + }, + { + "epoch": 7.17, + "learning_rate": 1.4166484596897532e-06, + "loss": 1.0466, + "step": 32852 + }, + { + "epoch": 7.17, + "learning_rate": 1.416539217828272e-06, + "loss": 2.0663, + "step": 32853 + }, + { + "epoch": 7.17, + "learning_rate": 1.4164299759667908e-06, + "loss": 0.951, + "step": 32854 + }, + { + "epoch": 7.17, + "learning_rate": 1.4163207341053092e-06, + "loss": 1.9877, + "step": 32855 + }, + { + "epoch": 7.17, + "learning_rate": 1.416211492243828e-06, + "loss": 1.6647, + "step": 32856 + }, + { + "epoch": 7.17, + "learning_rate": 1.4161022503823466e-06, + "loss": 1.4928, + "step": 32857 + }, + { + "epoch": 7.17, + "learning_rate": 1.4159930085208653e-06, + "loss": 0.5359, + "step": 32858 + }, + { + "epoch": 7.17, + "learning_rate": 1.4158837666593839e-06, + "loss": 2.9044, + "step": 32859 + }, + { + "epoch": 7.17, + "learning_rate": 1.4157745247979027e-06, + "loss": 2.3054, + "step": 32860 + }, + { + "epoch": 7.17, + "learning_rate": 1.4156652829364215e-06, + "loss": 0.0937, + "step": 32861 + }, + { + "epoch": 7.17, + "learning_rate": 1.41555604107494e-06, + "loss": 1.7607, + "step": 32862 + }, + { + "epoch": 7.17, + "learning_rate": 1.4154467992134587e-06, + "loss": 1.288, + "step": 32863 + }, + { + "epoch": 7.17, + "learning_rate": 1.4153375573519773e-06, + "loss": 0.6986, + "step": 32864 + }, + { + "epoch": 7.17, + "learning_rate": 1.4152283154904962e-06, + "loss": 1.9654, + "step": 32865 + }, + { + "epoch": 7.17, + "learning_rate": 1.4151190736290146e-06, + "loss": 0.9366, + "step": 32866 + }, + { + "epoch": 7.17, + "learning_rate": 1.4150098317675334e-06, + "loss": 0.8327, + "step": 32867 + }, + { + "epoch": 7.17, + "learning_rate": 1.4149005899060522e-06, + "loss": 0.6669, + "step": 32868 + }, + { + "epoch": 7.17, + "learning_rate": 1.4147913480445708e-06, + "loss": 1.9043, + "step": 32869 + }, + { + "epoch": 7.17, + "learning_rate": 1.4146821061830896e-06, + "loss": 1.2859, + "step": 32870 + }, + { + "epoch": 7.17, + "learning_rate": 1.414572864321608e-06, + "loss": 1.2634, + "step": 32871 + }, + { + "epoch": 7.17, + "learning_rate": 1.4144636224601269e-06, + "loss": 1.4385, + "step": 32872 + }, + { + "epoch": 7.17, + "learning_rate": 1.4143543805986455e-06, + "loss": 0.1813, + "step": 32873 + }, + { + "epoch": 7.17, + "learning_rate": 1.414245138737164e-06, + "loss": 0.3587, + "step": 32874 + }, + { + "epoch": 7.17, + "learning_rate": 1.414135896875683e-06, + "loss": 2.9249, + "step": 32875 + }, + { + "epoch": 7.18, + "learning_rate": 1.4140266550142015e-06, + "loss": 1.9325, + "step": 32876 + }, + { + "epoch": 7.18, + "learning_rate": 1.4139174131527203e-06, + "loss": 1.1122, + "step": 32877 + }, + { + "epoch": 7.18, + "learning_rate": 1.4138081712912387e-06, + "loss": 1.4233, + "step": 32878 + }, + { + "epoch": 7.18, + "learning_rate": 1.4136989294297576e-06, + "loss": 0.5353, + "step": 32879 + }, + { + "epoch": 7.18, + "learning_rate": 1.4135896875682764e-06, + "loss": 0.7612, + "step": 32880 + }, + { + "epoch": 7.18, + "learning_rate": 1.413480445706795e-06, + "loss": 1.0527, + "step": 32881 + }, + { + "epoch": 7.18, + "learning_rate": 1.4133712038453138e-06, + "loss": 2.048, + "step": 32882 + }, + { + "epoch": 7.18, + "learning_rate": 1.4132619619838322e-06, + "loss": 1.9428, + "step": 32883 + }, + { + "epoch": 7.18, + "learning_rate": 1.413152720122351e-06, + "loss": 1.654, + "step": 32884 + }, + { + "epoch": 7.18, + "learning_rate": 1.4130434782608697e-06, + "loss": 0.2902, + "step": 32885 + }, + { + "epoch": 7.18, + "learning_rate": 1.4129342363993885e-06, + "loss": 0.6753, + "step": 32886 + }, + { + "epoch": 7.18, + "learning_rate": 1.412824994537907e-06, + "loss": 1.4681, + "step": 32887 + }, + { + "epoch": 7.18, + "learning_rate": 1.4127157526764257e-06, + "loss": 1.7996, + "step": 32888 + }, + { + "epoch": 7.18, + "learning_rate": 1.4126065108149445e-06, + "loss": 1.242, + "step": 32889 + }, + { + "epoch": 7.18, + "learning_rate": 1.412497268953463e-06, + "loss": 1.4343, + "step": 32890 + }, + { + "epoch": 7.18, + "learning_rate": 1.4123880270919817e-06, + "loss": 0.1227, + "step": 32891 + }, + { + "epoch": 7.18, + "learning_rate": 1.4122787852305004e-06, + "loss": 1.8468, + "step": 32892 + }, + { + "epoch": 7.18, + "learning_rate": 1.4121695433690192e-06, + "loss": 2.1202, + "step": 32893 + }, + { + "epoch": 7.18, + "learning_rate": 1.412060301507538e-06, + "loss": 0.9182, + "step": 32894 + }, + { + "epoch": 7.18, + "learning_rate": 1.4119510596460564e-06, + "loss": 0.8107, + "step": 32895 + }, + { + "epoch": 7.18, + "learning_rate": 1.4118418177845752e-06, + "loss": 1.3654, + "step": 32896 + }, + { + "epoch": 7.18, + "learning_rate": 1.4117325759230938e-06, + "loss": 2.3899, + "step": 32897 + }, + { + "epoch": 7.18, + "learning_rate": 1.4116233340616127e-06, + "loss": 0.7517, + "step": 32898 + }, + { + "epoch": 7.18, + "learning_rate": 1.411514092200131e-06, + "loss": 0.6869, + "step": 32899 + }, + { + "epoch": 7.18, + "learning_rate": 1.4114048503386499e-06, + "loss": 1.2773, + "step": 32900 + }, + { + "epoch": 7.18, + "learning_rate": 1.4112956084771687e-06, + "loss": 0.7756, + "step": 32901 + }, + { + "epoch": 7.18, + "learning_rate": 1.411186366615687e-06, + "loss": 0.9229, + "step": 32902 + }, + { + "epoch": 7.18, + "learning_rate": 1.411077124754206e-06, + "loss": 0.6926, + "step": 32903 + }, + { + "epoch": 7.18, + "learning_rate": 1.4109678828927245e-06, + "loss": 1.4617, + "step": 32904 + }, + { + "epoch": 7.18, + "learning_rate": 1.4108586410312434e-06, + "loss": 1.6154, + "step": 32905 + }, + { + "epoch": 7.18, + "learning_rate": 1.4107493991697618e-06, + "loss": 2.1702, + "step": 32906 + }, + { + "epoch": 7.18, + "learning_rate": 1.4106401573082806e-06, + "loss": 0.8324, + "step": 32907 + }, + { + "epoch": 7.18, + "learning_rate": 1.4105309154467994e-06, + "loss": 1.4013, + "step": 32908 + }, + { + "epoch": 7.18, + "learning_rate": 1.410421673585318e-06, + "loss": 0.4294, + "step": 32909 + }, + { + "epoch": 7.18, + "learning_rate": 1.4103124317238368e-06, + "loss": 0.3912, + "step": 32910 + }, + { + "epoch": 7.18, + "learning_rate": 1.4102031898623552e-06, + "loss": 1.5361, + "step": 32911 + }, + { + "epoch": 7.18, + "learning_rate": 1.410093948000874e-06, + "loss": 1.2575, + "step": 32912 + }, + { + "epoch": 7.18, + "learning_rate": 1.4099847061393929e-06, + "loss": 0.7042, + "step": 32913 + }, + { + "epoch": 7.18, + "learning_rate": 1.4098754642779115e-06, + "loss": 0.8857, + "step": 32914 + }, + { + "epoch": 7.18, + "learning_rate": 1.40976622241643e-06, + "loss": 4.2146, + "step": 32915 + }, + { + "epoch": 7.18, + "learning_rate": 1.4096569805549487e-06, + "loss": 1.1904, + "step": 32916 + }, + { + "epoch": 7.18, + "learning_rate": 1.4095477386934675e-06, + "loss": 0.1212, + "step": 32917 + }, + { + "epoch": 7.18, + "learning_rate": 1.409438496831986e-06, + "loss": 1.8823, + "step": 32918 + }, + { + "epoch": 7.18, + "learning_rate": 1.4093292549705048e-06, + "loss": 1.0435, + "step": 32919 + }, + { + "epoch": 7.18, + "learning_rate": 1.4092200131090236e-06, + "loss": 0.4238, + "step": 32920 + }, + { + "epoch": 7.18, + "learning_rate": 1.4091107712475422e-06, + "loss": 2.7661, + "step": 32921 + }, + { + "epoch": 7.19, + "learning_rate": 1.409001529386061e-06, + "loss": 2.4955, + "step": 32922 + }, + { + "epoch": 7.19, + "learning_rate": 1.4088922875245794e-06, + "loss": 2.0002, + "step": 32923 + }, + { + "epoch": 7.19, + "learning_rate": 1.4087830456630982e-06, + "loss": 2.0032, + "step": 32924 + }, + { + "epoch": 7.19, + "learning_rate": 1.4086738038016168e-06, + "loss": 1.1679, + "step": 32925 + }, + { + "epoch": 7.19, + "learning_rate": 1.4085645619401357e-06, + "loss": 0.9389, + "step": 32926 + }, + { + "epoch": 7.19, + "learning_rate": 1.4084553200786543e-06, + "loss": 0.9385, + "step": 32927 + }, + { + "epoch": 7.19, + "learning_rate": 1.4083460782171729e-06, + "loss": 2.2488, + "step": 32928 + }, + { + "epoch": 7.19, + "learning_rate": 1.4082368363556917e-06, + "loss": 0.2251, + "step": 32929 + }, + { + "epoch": 7.19, + "learning_rate": 1.4081275944942101e-06, + "loss": 1.0737, + "step": 32930 + }, + { + "epoch": 7.19, + "learning_rate": 1.408018352632729e-06, + "loss": 0.209, + "step": 32931 + }, + { + "epoch": 7.19, + "learning_rate": 1.4079091107712475e-06, + "loss": 1.0598, + "step": 32932 + }, + { + "epoch": 7.19, + "learning_rate": 1.4077998689097664e-06, + "loss": 0.3435, + "step": 32933 + }, + { + "epoch": 7.19, + "learning_rate": 1.4076906270482852e-06, + "loss": 0.6424, + "step": 32934 + }, + { + "epoch": 7.19, + "learning_rate": 1.4075813851868036e-06, + "loss": 1.6379, + "step": 32935 + }, + { + "epoch": 7.19, + "learning_rate": 1.4074721433253224e-06, + "loss": 0.2255, + "step": 32936 + }, + { + "epoch": 7.19, + "learning_rate": 1.407362901463841e-06, + "loss": 2.1323, + "step": 32937 + }, + { + "epoch": 7.19, + "learning_rate": 1.4072536596023598e-06, + "loss": 1.2489, + "step": 32938 + }, + { + "epoch": 7.19, + "learning_rate": 1.4071444177408782e-06, + "loss": 0.5758, + "step": 32939 + }, + { + "epoch": 7.19, + "learning_rate": 1.407035175879397e-06, + "loss": 1.7394, + "step": 32940 + }, + { + "epoch": 7.19, + "learning_rate": 1.4069259340179159e-06, + "loss": 3.509, + "step": 32941 + }, + { + "epoch": 7.19, + "learning_rate": 1.4068166921564345e-06, + "loss": 1.0367, + "step": 32942 + }, + { + "epoch": 7.19, + "learning_rate": 1.4067074502949531e-06, + "loss": 1.2686, + "step": 32943 + }, + { + "epoch": 7.19, + "learning_rate": 1.4065982084334717e-06, + "loss": 1.7134, + "step": 32944 + }, + { + "epoch": 7.19, + "learning_rate": 1.4064889665719905e-06, + "loss": 1.8464, + "step": 32945 + }, + { + "epoch": 7.19, + "learning_rate": 1.4063797247105094e-06, + "loss": 0.922, + "step": 32946 + }, + { + "epoch": 7.19, + "learning_rate": 1.4062704828490278e-06, + "loss": 0.5759, + "step": 32947 + }, + { + "epoch": 7.19, + "learning_rate": 1.4061612409875466e-06, + "loss": 1.6759, + "step": 32948 + }, + { + "epoch": 7.19, + "learning_rate": 1.4060519991260652e-06, + "loss": 1.4034, + "step": 32949 + }, + { + "epoch": 7.19, + "learning_rate": 1.405942757264584e-06, + "loss": 0.5225, + "step": 32950 + }, + { + "epoch": 7.19, + "learning_rate": 1.4058335154031024e-06, + "loss": 1.7991, + "step": 32951 + }, + { + "epoch": 7.19, + "learning_rate": 1.4057242735416212e-06, + "loss": 1.2333, + "step": 32952 + }, + { + "epoch": 7.19, + "learning_rate": 1.40561503168014e-06, + "loss": 1.0878, + "step": 32953 + }, + { + "epoch": 7.19, + "learning_rate": 1.4055057898186587e-06, + "loss": 1.8951, + "step": 32954 + }, + { + "epoch": 7.19, + "learning_rate": 1.4053965479571773e-06, + "loss": 0.7998, + "step": 32955 + }, + { + "epoch": 7.19, + "learning_rate": 1.405287306095696e-06, + "loss": 0.9146, + "step": 32956 + }, + { + "epoch": 7.19, + "learning_rate": 1.4051780642342147e-06, + "loss": 1.4984, + "step": 32957 + }, + { + "epoch": 7.19, + "learning_rate": 1.4050688223727333e-06, + "loss": 2.0559, + "step": 32958 + }, + { + "epoch": 7.19, + "learning_rate": 1.404959580511252e-06, + "loss": 0.5573, + "step": 32959 + }, + { + "epoch": 7.19, + "learning_rate": 1.4048503386497708e-06, + "loss": 1.3126, + "step": 32960 + }, + { + "epoch": 7.19, + "learning_rate": 1.4047410967882894e-06, + "loss": 2.1828, + "step": 32961 + }, + { + "epoch": 7.19, + "learning_rate": 1.4046318549268082e-06, + "loss": 0.9124, + "step": 32962 + }, + { + "epoch": 7.19, + "learning_rate": 1.4045226130653266e-06, + "loss": 0.2755, + "step": 32963 + }, + { + "epoch": 7.19, + "learning_rate": 1.4044133712038454e-06, + "loss": 1.8652, + "step": 32964 + }, + { + "epoch": 7.19, + "learning_rate": 1.404304129342364e-06, + "loss": 2.303, + "step": 32965 + }, + { + "epoch": 7.19, + "learning_rate": 1.4041948874808829e-06, + "loss": 1.4361, + "step": 32966 + }, + { + "epoch": 7.19, + "learning_rate": 1.4040856456194017e-06, + "loss": 0.1168, + "step": 32967 + }, + { + "epoch": 7.2, + "learning_rate": 1.40397640375792e-06, + "loss": 0.3364, + "step": 32968 + }, + { + "epoch": 7.2, + "learning_rate": 1.403867161896439e-06, + "loss": 1.8457, + "step": 32969 + }, + { + "epoch": 7.2, + "learning_rate": 1.4037579200349575e-06, + "loss": 0.7345, + "step": 32970 + }, + { + "epoch": 7.2, + "learning_rate": 1.4036486781734761e-06, + "loss": 2.4335, + "step": 32971 + }, + { + "epoch": 7.2, + "learning_rate": 1.4035394363119947e-06, + "loss": 2.6572, + "step": 32972 + }, + { + "epoch": 7.2, + "learning_rate": 1.4034301944505136e-06, + "loss": 1.0896, + "step": 32973 + }, + { + "epoch": 7.2, + "learning_rate": 1.4033209525890324e-06, + "loss": 1.6282, + "step": 32974 + }, + { + "epoch": 7.2, + "learning_rate": 1.4032117107275508e-06, + "loss": 0.4355, + "step": 32975 + }, + { + "epoch": 7.2, + "learning_rate": 1.4031024688660696e-06, + "loss": 1.4059, + "step": 32976 + }, + { + "epoch": 7.2, + "learning_rate": 1.4029932270045882e-06, + "loss": 1.4378, + "step": 32977 + }, + { + "epoch": 7.2, + "learning_rate": 1.402883985143107e-06, + "loss": 1.0528, + "step": 32978 + }, + { + "epoch": 7.2, + "learning_rate": 1.4027747432816254e-06, + "loss": 2.8148, + "step": 32979 + }, + { + "epoch": 7.2, + "learning_rate": 1.4026655014201443e-06, + "loss": 0.1426, + "step": 32980 + }, + { + "epoch": 7.2, + "learning_rate": 1.402556259558663e-06, + "loss": 1.5474, + "step": 32981 + }, + { + "epoch": 7.2, + "learning_rate": 1.4024470176971817e-06, + "loss": 1.5066, + "step": 32982 + }, + { + "epoch": 7.2, + "learning_rate": 1.4023377758357003e-06, + "loss": 1.3409, + "step": 32983 + }, + { + "epoch": 7.2, + "learning_rate": 1.402228533974219e-06, + "loss": 1.2617, + "step": 32984 + }, + { + "epoch": 7.2, + "learning_rate": 1.4021192921127377e-06, + "loss": 2.7123, + "step": 32985 + }, + { + "epoch": 7.2, + "learning_rate": 1.4020100502512566e-06, + "loss": 0.1211, + "step": 32986 + }, + { + "epoch": 7.2, + "learning_rate": 1.401900808389775e-06, + "loss": 1.37, + "step": 32987 + }, + { + "epoch": 7.2, + "learning_rate": 1.4017915665282938e-06, + "loss": 2.2967, + "step": 32988 + }, + { + "epoch": 7.2, + "learning_rate": 1.4016823246668124e-06, + "loss": 0.6713, + "step": 32989 + }, + { + "epoch": 7.2, + "learning_rate": 1.4015730828053312e-06, + "loss": 0.2357, + "step": 32990 + }, + { + "epoch": 7.2, + "learning_rate": 1.4014638409438496e-06, + "loss": 0.8562, + "step": 32991 + }, + { + "epoch": 7.2, + "learning_rate": 1.4013545990823684e-06, + "loss": 3.2525, + "step": 32992 + }, + { + "epoch": 7.2, + "learning_rate": 1.4012453572208873e-06, + "loss": 1.3318, + "step": 32993 + }, + { + "epoch": 7.2, + "learning_rate": 1.4011361153594059e-06, + "loss": 1.6971, + "step": 32994 + }, + { + "epoch": 7.2, + "learning_rate": 1.4010268734979247e-06, + "loss": 0.9078, + "step": 32995 + }, + { + "epoch": 7.2, + "learning_rate": 1.400917631636443e-06, + "loss": 0.7942, + "step": 32996 + }, + { + "epoch": 7.2, + "learning_rate": 1.400808389774962e-06, + "loss": 0.2589, + "step": 32997 + }, + { + "epoch": 7.2, + "learning_rate": 1.4006991479134805e-06, + "loss": 1.2626, + "step": 32998 + }, + { + "epoch": 7.2, + "learning_rate": 1.4005899060519991e-06, + "loss": 2.3567, + "step": 32999 + }, + { + "epoch": 7.2, + "learning_rate": 1.400480664190518e-06, + "loss": 1.0551, + "step": 33000 + }, + { + "epoch": 7.2, + "learning_rate": 1.4003714223290366e-06, + "loss": 3.0015, + "step": 33001 + }, + { + "epoch": 7.2, + "learning_rate": 1.4002621804675554e-06, + "loss": 1.2775, + "step": 33002 + }, + { + "epoch": 7.2, + "learning_rate": 1.4001529386060738e-06, + "loss": 0.8596, + "step": 33003 + }, + { + "epoch": 7.2, + "learning_rate": 1.4000436967445926e-06, + "loss": 0.8838, + "step": 33004 + }, + { + "epoch": 7.2, + "learning_rate": 1.3999344548831112e-06, + "loss": 0.7863, + "step": 33005 + }, + { + "epoch": 7.2, + "learning_rate": 1.39982521302163e-06, + "loss": 0.114, + "step": 33006 + }, + { + "epoch": 7.2, + "learning_rate": 1.3997159711601489e-06, + "loss": 0.4388, + "step": 33007 + }, + { + "epoch": 7.2, + "learning_rate": 1.3996067292986673e-06, + "loss": 0.4995, + "step": 33008 + }, + { + "epoch": 7.2, + "learning_rate": 1.399497487437186e-06, + "loss": 1.5675, + "step": 33009 + }, + { + "epoch": 7.2, + "learning_rate": 1.3993882455757047e-06, + "loss": 1.4305, + "step": 33010 + }, + { + "epoch": 7.2, + "learning_rate": 1.3992790037142235e-06, + "loss": 1.6553, + "step": 33011 + }, + { + "epoch": 7.2, + "learning_rate": 1.399169761852742e-06, + "loss": 1.9774, + "step": 33012 + }, + { + "epoch": 7.2, + "learning_rate": 1.3990605199912608e-06, + "loss": 0.4848, + "step": 33013 + }, + { + "epoch": 7.21, + "learning_rate": 1.3989512781297796e-06, + "loss": 1.8496, + "step": 33014 + }, + { + "epoch": 7.21, + "learning_rate": 1.398842036268298e-06, + "loss": 1.8661, + "step": 33015 + }, + { + "epoch": 7.21, + "learning_rate": 1.3987327944068168e-06, + "loss": 1.9537, + "step": 33016 + }, + { + "epoch": 7.21, + "learning_rate": 1.3986235525453354e-06, + "loss": 0.4762, + "step": 33017 + }, + { + "epoch": 7.21, + "learning_rate": 1.3985143106838542e-06, + "loss": 1.861, + "step": 33018 + }, + { + "epoch": 7.21, + "learning_rate": 1.398405068822373e-06, + "loss": 0.1581, + "step": 33019 + }, + { + "epoch": 7.21, + "learning_rate": 1.3982958269608915e-06, + "loss": 2.4644, + "step": 33020 + }, + { + "epoch": 7.21, + "learning_rate": 1.3981865850994103e-06, + "loss": 0.9324, + "step": 33021 + }, + { + "epoch": 7.21, + "learning_rate": 1.3980773432379289e-06, + "loss": 0.7796, + "step": 33022 + }, + { + "epoch": 7.21, + "learning_rate": 1.3979681013764477e-06, + "loss": 1.3622, + "step": 33023 + }, + { + "epoch": 7.21, + "learning_rate": 1.3978588595149661e-06, + "loss": 1.5346, + "step": 33024 + }, + { + "epoch": 7.21, + "learning_rate": 1.397749617653485e-06, + "loss": 0.7868, + "step": 33025 + }, + { + "epoch": 7.21, + "learning_rate": 1.3976403757920038e-06, + "loss": 1.0613, + "step": 33026 + }, + { + "epoch": 7.21, + "learning_rate": 1.3975311339305222e-06, + "loss": 1.4333, + "step": 33027 + }, + { + "epoch": 7.21, + "learning_rate": 1.397421892069041e-06, + "loss": 1.6632, + "step": 33028 + }, + { + "epoch": 7.21, + "learning_rate": 1.3973126502075596e-06, + "loss": 1.2329, + "step": 33029 + }, + { + "epoch": 7.21, + "learning_rate": 1.3972034083460784e-06, + "loss": 0.5122, + "step": 33030 + }, + { + "epoch": 7.21, + "learning_rate": 1.3970941664845968e-06, + "loss": 0.6631, + "step": 33031 + }, + { + "epoch": 7.21, + "learning_rate": 1.3969849246231156e-06, + "loss": 0.2668, + "step": 33032 + }, + { + "epoch": 7.21, + "learning_rate": 1.3968756827616345e-06, + "loss": 0.629, + "step": 33033 + }, + { + "epoch": 7.21, + "learning_rate": 1.396766440900153e-06, + "loss": 0.6763, + "step": 33034 + }, + { + "epoch": 7.21, + "learning_rate": 1.3966571990386719e-06, + "loss": 0.9718, + "step": 33035 + }, + { + "epoch": 7.21, + "learning_rate": 1.3965479571771903e-06, + "loss": 1.281, + "step": 33036 + }, + { + "epoch": 7.21, + "learning_rate": 1.3964387153157091e-06, + "loss": 1.1649, + "step": 33037 + }, + { + "epoch": 7.21, + "learning_rate": 1.3963294734542277e-06, + "loss": 1.4646, + "step": 33038 + }, + { + "epoch": 7.21, + "learning_rate": 1.3962202315927465e-06, + "loss": 0.9361, + "step": 33039 + }, + { + "epoch": 7.21, + "learning_rate": 1.3961109897312652e-06, + "loss": 0.586, + "step": 33040 + }, + { + "epoch": 7.21, + "learning_rate": 1.3960017478697838e-06, + "loss": 1.0098, + "step": 33041 + }, + { + "epoch": 7.21, + "learning_rate": 1.3958925060083026e-06, + "loss": 0.6193, + "step": 33042 + }, + { + "epoch": 7.21, + "learning_rate": 1.395783264146821e-06, + "loss": 1.6846, + "step": 33043 + }, + { + "epoch": 7.21, + "learning_rate": 1.3956740222853398e-06, + "loss": 3.7065, + "step": 33044 + }, + { + "epoch": 7.21, + "learning_rate": 1.3955647804238584e-06, + "loss": 1.0944, + "step": 33045 + }, + { + "epoch": 7.21, + "learning_rate": 1.3954555385623772e-06, + "loss": 2.1776, + "step": 33046 + }, + { + "epoch": 7.21, + "learning_rate": 1.395346296700896e-06, + "loss": 1.8213, + "step": 33047 + }, + { + "epoch": 7.21, + "learning_rate": 1.3952370548394145e-06, + "loss": 1.4508, + "step": 33048 + }, + { + "epoch": 7.21, + "learning_rate": 1.3951278129779333e-06, + "loss": 1.5074, + "step": 33049 + }, + { + "epoch": 7.21, + "learning_rate": 1.395018571116452e-06, + "loss": 1.9787, + "step": 33050 + }, + { + "epoch": 7.21, + "learning_rate": 1.3949093292549707e-06, + "loss": 1.316, + "step": 33051 + }, + { + "epoch": 7.21, + "learning_rate": 1.3948000873934893e-06, + "loss": 1.8669, + "step": 33052 + }, + { + "epoch": 7.21, + "learning_rate": 1.394690845532008e-06, + "loss": 1.2694, + "step": 33053 + }, + { + "epoch": 7.21, + "learning_rate": 1.3945816036705268e-06, + "loss": 2.164, + "step": 33054 + }, + { + "epoch": 7.21, + "learning_rate": 1.3944723618090452e-06, + "loss": 1.0653, + "step": 33055 + }, + { + "epoch": 7.21, + "learning_rate": 1.394363119947564e-06, + "loss": 0.7662, + "step": 33056 + }, + { + "epoch": 7.21, + "learning_rate": 1.3942538780860826e-06, + "loss": 1.2425, + "step": 33057 + }, + { + "epoch": 7.21, + "learning_rate": 1.3941446362246014e-06, + "loss": 0.8015, + "step": 33058 + }, + { + "epoch": 7.21, + "learning_rate": 1.3940353943631202e-06, + "loss": 1.856, + "step": 33059 + }, + { + "epoch": 7.22, + "learning_rate": 1.3939261525016386e-06, + "loss": 1.0068, + "step": 33060 + }, + { + "epoch": 7.22, + "learning_rate": 1.3938169106401575e-06, + "loss": 0.9445, + "step": 33061 + }, + { + "epoch": 7.22, + "learning_rate": 1.393707668778676e-06, + "loss": 1.426, + "step": 33062 + }, + { + "epoch": 7.22, + "learning_rate": 1.393598426917195e-06, + "loss": 2.2227, + "step": 33063 + }, + { + "epoch": 7.22, + "learning_rate": 1.3934891850557133e-06, + "loss": 0.5756, + "step": 33064 + }, + { + "epoch": 7.22, + "learning_rate": 1.3933799431942321e-06, + "loss": 1.6862, + "step": 33065 + }, + { + "epoch": 7.22, + "learning_rate": 1.393270701332751e-06, + "loss": 1.2607, + "step": 33066 + }, + { + "epoch": 7.22, + "learning_rate": 1.3931614594712696e-06, + "loss": 2.3171, + "step": 33067 + }, + { + "epoch": 7.22, + "learning_rate": 1.3930522176097882e-06, + "loss": 2.8239, + "step": 33068 + }, + { + "epoch": 7.22, + "learning_rate": 1.3929429757483068e-06, + "loss": 1.8883, + "step": 33069 + }, + { + "epoch": 7.22, + "learning_rate": 1.3928337338868256e-06, + "loss": 1.356, + "step": 33070 + }, + { + "epoch": 7.22, + "learning_rate": 1.392724492025344e-06, + "loss": 1.1577, + "step": 33071 + }, + { + "epoch": 7.22, + "learning_rate": 1.3926152501638628e-06, + "loss": 1.0917, + "step": 33072 + }, + { + "epoch": 7.22, + "learning_rate": 1.3925060083023816e-06, + "loss": 0.7803, + "step": 33073 + }, + { + "epoch": 7.22, + "learning_rate": 1.3923967664409003e-06, + "loss": 0.959, + "step": 33074 + }, + { + "epoch": 7.22, + "learning_rate": 1.392287524579419e-06, + "loss": 0.1672, + "step": 33075 + }, + { + "epoch": 7.22, + "learning_rate": 1.3921782827179375e-06, + "loss": 1.2842, + "step": 33076 + }, + { + "epoch": 7.22, + "learning_rate": 1.3920690408564563e-06, + "loss": 1.2915, + "step": 33077 + }, + { + "epoch": 7.22, + "learning_rate": 1.391959798994975e-06, + "loss": 0.7416, + "step": 33078 + }, + { + "epoch": 7.22, + "learning_rate": 1.3918505571334937e-06, + "loss": 0.6969, + "step": 33079 + }, + { + "epoch": 7.22, + "learning_rate": 1.3917413152720123e-06, + "loss": 3.1416, + "step": 33080 + }, + { + "epoch": 7.22, + "learning_rate": 1.391632073410531e-06, + "loss": 2.89, + "step": 33081 + }, + { + "epoch": 7.22, + "learning_rate": 1.3915228315490498e-06, + "loss": 1.5433, + "step": 33082 + }, + { + "epoch": 7.22, + "learning_rate": 1.3914135896875684e-06, + "loss": 1.1003, + "step": 33083 + }, + { + "epoch": 7.22, + "learning_rate": 1.391304347826087e-06, + "loss": 0.4556, + "step": 33084 + }, + { + "epoch": 7.22, + "learning_rate": 1.3911951059646058e-06, + "loss": 2.7897, + "step": 33085 + }, + { + "epoch": 7.22, + "learning_rate": 1.3910858641031244e-06, + "loss": 0.6194, + "step": 33086 + }, + { + "epoch": 7.22, + "learning_rate": 1.3909766222416433e-06, + "loss": 0.8917, + "step": 33087 + }, + { + "epoch": 7.22, + "learning_rate": 1.3908673803801617e-06, + "loss": 1.1108, + "step": 33088 + }, + { + "epoch": 7.22, + "learning_rate": 1.3907581385186805e-06, + "loss": 1.1566, + "step": 33089 + }, + { + "epoch": 7.22, + "learning_rate": 1.390648896657199e-06, + "loss": 3.1624, + "step": 33090 + }, + { + "epoch": 7.22, + "learning_rate": 1.390539654795718e-06, + "loss": 0.3491, + "step": 33091 + }, + { + "epoch": 7.22, + "learning_rate": 1.3904304129342367e-06, + "loss": 1.365, + "step": 33092 + }, + { + "epoch": 7.22, + "learning_rate": 1.3903211710727551e-06, + "loss": 1.7004, + "step": 33093 + }, + { + "epoch": 7.22, + "learning_rate": 1.390211929211274e-06, + "loss": 1.0868, + "step": 33094 + }, + { + "epoch": 7.22, + "learning_rate": 1.3901026873497926e-06, + "loss": 1.0359, + "step": 33095 + }, + { + "epoch": 7.22, + "learning_rate": 1.3899934454883112e-06, + "loss": 1.9442, + "step": 33096 + }, + { + "epoch": 7.22, + "learning_rate": 1.3898842036268298e-06, + "loss": 1.0859, + "step": 33097 + }, + { + "epoch": 7.22, + "learning_rate": 1.3897749617653486e-06, + "loss": 2.6046, + "step": 33098 + }, + { + "epoch": 7.22, + "learning_rate": 1.3896657199038674e-06, + "loss": 1.2921, + "step": 33099 + }, + { + "epoch": 7.22, + "learning_rate": 1.3895564780423858e-06, + "loss": 1.9549, + "step": 33100 + }, + { + "epoch": 7.22, + "learning_rate": 1.3894472361809047e-06, + "loss": 0.5717, + "step": 33101 + }, + { + "epoch": 7.22, + "learning_rate": 1.3893379943194233e-06, + "loss": 1.5224, + "step": 33102 + }, + { + "epoch": 7.22, + "learning_rate": 1.389228752457942e-06, + "loss": 0.6854, + "step": 33103 + }, + { + "epoch": 7.22, + "learning_rate": 1.3891195105964605e-06, + "loss": 1.5142, + "step": 33104 + }, + { + "epoch": 7.23, + "learning_rate": 1.3890102687349793e-06, + "loss": 2.0617, + "step": 33105 + }, + { + "epoch": 7.23, + "learning_rate": 1.3889010268734981e-06, + "loss": 2.965, + "step": 33106 + }, + { + "epoch": 7.23, + "learning_rate": 1.3887917850120168e-06, + "loss": 1.36, + "step": 33107 + }, + { + "epoch": 7.23, + "learning_rate": 1.3886825431505354e-06, + "loss": 1.1399, + "step": 33108 + }, + { + "epoch": 7.23, + "learning_rate": 1.388573301289054e-06, + "loss": 1.0707, + "step": 33109 + }, + { + "epoch": 7.23, + "learning_rate": 1.3884640594275728e-06, + "loss": 0.7078, + "step": 33110 + }, + { + "epoch": 7.23, + "learning_rate": 1.3883548175660914e-06, + "loss": 0.2835, + "step": 33111 + }, + { + "epoch": 7.23, + "learning_rate": 1.38824557570461e-06, + "loss": 1.8078, + "step": 33112 + }, + { + "epoch": 7.23, + "learning_rate": 1.3881363338431288e-06, + "loss": 2.251, + "step": 33113 + }, + { + "epoch": 7.23, + "learning_rate": 1.3880270919816475e-06, + "loss": 0.9787, + "step": 33114 + }, + { + "epoch": 7.23, + "learning_rate": 1.3879178501201663e-06, + "loss": 0.1033, + "step": 33115 + }, + { + "epoch": 7.23, + "learning_rate": 1.3878086082586847e-06, + "loss": 0.7765, + "step": 33116 + }, + { + "epoch": 7.23, + "learning_rate": 1.3876993663972035e-06, + "loss": 1.1944, + "step": 33117 + }, + { + "epoch": 7.23, + "learning_rate": 1.3875901245357223e-06, + "loss": 0.6092, + "step": 33118 + }, + { + "epoch": 7.23, + "learning_rate": 1.387480882674241e-06, + "loss": 0.8569, + "step": 33119 + }, + { + "epoch": 7.23, + "learning_rate": 1.3873716408127598e-06, + "loss": 2.1181, + "step": 33120 + }, + { + "epoch": 7.23, + "learning_rate": 1.3872623989512782e-06, + "loss": 1.4471, + "step": 33121 + }, + { + "epoch": 7.23, + "learning_rate": 1.387153157089797e-06, + "loss": 0.8363, + "step": 33122 + }, + { + "epoch": 7.23, + "learning_rate": 1.3870439152283156e-06, + "loss": 1.9751, + "step": 33123 + }, + { + "epoch": 7.23, + "learning_rate": 1.3869346733668342e-06, + "loss": 1.6951, + "step": 33124 + }, + { + "epoch": 7.23, + "learning_rate": 1.386825431505353e-06, + "loss": 1.2327, + "step": 33125 + }, + { + "epoch": 7.23, + "learning_rate": 1.3867161896438716e-06, + "loss": 1.9014, + "step": 33126 + }, + { + "epoch": 7.23, + "learning_rate": 1.3866069477823905e-06, + "loss": 1.7886, + "step": 33127 + }, + { + "epoch": 7.23, + "learning_rate": 1.3864977059209089e-06, + "loss": 2.7743, + "step": 33128 + }, + { + "epoch": 7.23, + "learning_rate": 1.3863884640594277e-06, + "loss": 1.3796, + "step": 33129 + }, + { + "epoch": 7.23, + "learning_rate": 1.3862792221979463e-06, + "loss": 1.0889, + "step": 33130 + }, + { + "epoch": 7.23, + "learning_rate": 1.3861699803364651e-06, + "loss": 1.3166, + "step": 33131 + }, + { + "epoch": 7.23, + "learning_rate": 1.386060738474984e-06, + "loss": 1.8247, + "step": 33132 + }, + { + "epoch": 7.23, + "learning_rate": 1.3859514966135023e-06, + "loss": 0.2262, + "step": 33133 + }, + { + "epoch": 7.23, + "learning_rate": 1.3858422547520212e-06, + "loss": 1.5075, + "step": 33134 + }, + { + "epoch": 7.23, + "learning_rate": 1.3857330128905398e-06, + "loss": 0.4052, + "step": 33135 + }, + { + "epoch": 7.23, + "learning_rate": 1.3856237710290586e-06, + "loss": 1.9148, + "step": 33136 + }, + { + "epoch": 7.23, + "learning_rate": 1.385514529167577e-06, + "loss": 1.7791, + "step": 33137 + }, + { + "epoch": 7.23, + "learning_rate": 1.3854052873060958e-06, + "loss": 2.0356, + "step": 33138 + }, + { + "epoch": 7.23, + "learning_rate": 1.3852960454446146e-06, + "loss": 0.2784, + "step": 33139 + }, + { + "epoch": 7.23, + "learning_rate": 1.385186803583133e-06, + "loss": 2.121, + "step": 33140 + }, + { + "epoch": 7.23, + "learning_rate": 1.3850775617216519e-06, + "loss": 1.2165, + "step": 33141 + }, + { + "epoch": 7.23, + "learning_rate": 1.3849683198601705e-06, + "loss": 1.1941, + "step": 33142 + }, + { + "epoch": 7.23, + "learning_rate": 1.3848590779986893e-06, + "loss": 2.8789, + "step": 33143 + }, + { + "epoch": 7.23, + "learning_rate": 1.3847498361372077e-06, + "loss": 0.9983, + "step": 33144 + }, + { + "epoch": 7.23, + "learning_rate": 1.3846405942757265e-06, + "loss": 1.8683, + "step": 33145 + }, + { + "epoch": 7.23, + "learning_rate": 1.3845313524142453e-06, + "loss": 0.5942, + "step": 33146 + }, + { + "epoch": 7.23, + "learning_rate": 1.384422110552764e-06, + "loss": 1.0027, + "step": 33147 + }, + { + "epoch": 7.23, + "learning_rate": 1.3843128686912828e-06, + "loss": 1.5221, + "step": 33148 + }, + { + "epoch": 7.23, + "learning_rate": 1.3842036268298012e-06, + "loss": 1.9176, + "step": 33149 + }, + { + "epoch": 7.23, + "learning_rate": 1.38409438496832e-06, + "loss": 2.2172, + "step": 33150 + }, + { + "epoch": 7.24, + "learning_rate": 1.3839851431068388e-06, + "loss": 0.2049, + "step": 33151 + }, + { + "epoch": 7.24, + "learning_rate": 1.3838759012453572e-06, + "loss": 0.3962, + "step": 33152 + }, + { + "epoch": 7.24, + "learning_rate": 1.383766659383876e-06, + "loss": 0.4953, + "step": 33153 + }, + { + "epoch": 7.24, + "learning_rate": 1.3836574175223946e-06, + "loss": 0.4264, + "step": 33154 + }, + { + "epoch": 7.24, + "learning_rate": 1.3835481756609135e-06, + "loss": 1.9275, + "step": 33155 + }, + { + "epoch": 7.24, + "learning_rate": 1.3834389337994319e-06, + "loss": 0.1913, + "step": 33156 + }, + { + "epoch": 7.24, + "learning_rate": 1.3833296919379507e-06, + "loss": 1.2243, + "step": 33157 + }, + { + "epoch": 7.24, + "learning_rate": 1.3832204500764695e-06, + "loss": 0.9113, + "step": 33158 + }, + { + "epoch": 7.24, + "learning_rate": 1.3831112082149881e-06, + "loss": 0.1204, + "step": 33159 + }, + { + "epoch": 7.24, + "learning_rate": 1.383001966353507e-06, + "loss": 0.1487, + "step": 33160 + }, + { + "epoch": 7.24, + "learning_rate": 1.3828927244920253e-06, + "loss": 0.8077, + "step": 33161 + }, + { + "epoch": 7.24, + "learning_rate": 1.3827834826305442e-06, + "loss": 1.587, + "step": 33162 + }, + { + "epoch": 7.24, + "learning_rate": 1.3826742407690628e-06, + "loss": 1.3918, + "step": 33163 + }, + { + "epoch": 7.24, + "learning_rate": 1.3825649989075816e-06, + "loss": 1.4196, + "step": 33164 + }, + { + "epoch": 7.24, + "learning_rate": 1.3824557570461002e-06, + "loss": 0.7635, + "step": 33165 + }, + { + "epoch": 7.24, + "learning_rate": 1.3823465151846188e-06, + "loss": 1.4197, + "step": 33166 + }, + { + "epoch": 7.24, + "learning_rate": 1.3822372733231376e-06, + "loss": 1.2055, + "step": 33167 + }, + { + "epoch": 7.24, + "learning_rate": 1.382128031461656e-06, + "loss": 1.7185, + "step": 33168 + }, + { + "epoch": 7.24, + "learning_rate": 1.3820187896001749e-06, + "loss": 1.3955, + "step": 33169 + }, + { + "epoch": 7.24, + "learning_rate": 1.3819095477386935e-06, + "loss": 0.742, + "step": 33170 + }, + { + "epoch": 7.24, + "learning_rate": 1.3818003058772123e-06, + "loss": 2.1704, + "step": 33171 + }, + { + "epoch": 7.24, + "learning_rate": 1.3816910640157311e-06, + "loss": 1.5398, + "step": 33172 + }, + { + "epoch": 7.24, + "learning_rate": 1.3815818221542495e-06, + "loss": 1.0513, + "step": 33173 + }, + { + "epoch": 7.24, + "learning_rate": 1.3814725802927683e-06, + "loss": 1.6687, + "step": 33174 + }, + { + "epoch": 7.24, + "learning_rate": 1.381363338431287e-06, + "loss": 1.9633, + "step": 33175 + }, + { + "epoch": 7.24, + "learning_rate": 1.3812540965698058e-06, + "loss": 1.308, + "step": 33176 + }, + { + "epoch": 7.24, + "learning_rate": 1.3811448547083242e-06, + "loss": 0.8648, + "step": 33177 + }, + { + "epoch": 7.24, + "learning_rate": 1.381035612846843e-06, + "loss": 2.1502, + "step": 33178 + }, + { + "epoch": 7.24, + "learning_rate": 1.3809263709853618e-06, + "loss": 2.7985, + "step": 33179 + }, + { + "epoch": 7.24, + "learning_rate": 1.3808171291238802e-06, + "loss": 0.8236, + "step": 33180 + }, + { + "epoch": 7.24, + "learning_rate": 1.380707887262399e-06, + "loss": 1.4813, + "step": 33181 + }, + { + "epoch": 7.24, + "learning_rate": 1.3805986454009177e-06, + "loss": 1.4554, + "step": 33182 + }, + { + "epoch": 7.24, + "learning_rate": 1.3804894035394365e-06, + "loss": 1.413, + "step": 33183 + }, + { + "epoch": 7.24, + "learning_rate": 1.3803801616779553e-06, + "loss": 2.8906, + "step": 33184 + }, + { + "epoch": 7.24, + "learning_rate": 1.3802709198164737e-06, + "loss": 1.392, + "step": 33185 + }, + { + "epoch": 7.24, + "learning_rate": 1.3801616779549925e-06, + "loss": 3.3403, + "step": 33186 + }, + { + "epoch": 7.24, + "learning_rate": 1.3800524360935111e-06, + "loss": 1.3319, + "step": 33187 + }, + { + "epoch": 7.24, + "learning_rate": 1.37994319423203e-06, + "loss": 1.8159, + "step": 33188 + }, + { + "epoch": 7.24, + "learning_rate": 1.3798339523705484e-06, + "loss": 0.9762, + "step": 33189 + }, + { + "epoch": 7.24, + "learning_rate": 1.3797247105090672e-06, + "loss": 1.8358, + "step": 33190 + }, + { + "epoch": 7.24, + "learning_rate": 1.379615468647586e-06, + "loss": 0.1326, + "step": 33191 + }, + { + "epoch": 7.24, + "learning_rate": 1.3795062267861046e-06, + "loss": 1.0564, + "step": 33192 + }, + { + "epoch": 7.24, + "learning_rate": 1.3793969849246232e-06, + "loss": 1.5411, + "step": 33193 + }, + { + "epoch": 7.24, + "learning_rate": 1.3792877430631418e-06, + "loss": 2.3421, + "step": 33194 + }, + { + "epoch": 7.24, + "learning_rate": 1.3791785012016607e-06, + "loss": 1.5769, + "step": 33195 + }, + { + "epoch": 7.24, + "learning_rate": 1.379069259340179e-06, + "loss": 0.4866, + "step": 33196 + }, + { + "epoch": 7.25, + "learning_rate": 1.3789600174786979e-06, + "loss": 1.0338, + "step": 33197 + }, + { + "epoch": 7.25, + "learning_rate": 1.3788507756172167e-06, + "loss": 0.8417, + "step": 33198 + }, + { + "epoch": 7.25, + "learning_rate": 1.3787415337557353e-06, + "loss": 1.4652, + "step": 33199 + }, + { + "epoch": 7.25, + "learning_rate": 1.3786322918942541e-06, + "loss": 1.4099, + "step": 33200 + }, + { + "epoch": 7.25, + "learning_rate": 1.3785230500327725e-06, + "loss": 1.9694, + "step": 33201 + }, + { + "epoch": 7.25, + "learning_rate": 1.3784138081712914e-06, + "loss": 1.0009, + "step": 33202 + }, + { + "epoch": 7.25, + "learning_rate": 1.37830456630981e-06, + "loss": 3.1037, + "step": 33203 + }, + { + "epoch": 7.25, + "learning_rate": 1.3781953244483288e-06, + "loss": 1.8898, + "step": 33204 + }, + { + "epoch": 7.25, + "learning_rate": 1.3780860825868474e-06, + "loss": 1.1786, + "step": 33205 + }, + { + "epoch": 7.25, + "learning_rate": 1.377976840725366e-06, + "loss": 1.1133, + "step": 33206 + }, + { + "epoch": 7.25, + "learning_rate": 1.3778675988638848e-06, + "loss": 0.858, + "step": 33207 + }, + { + "epoch": 7.25, + "learning_rate": 1.3777583570024034e-06, + "loss": 0.5075, + "step": 33208 + }, + { + "epoch": 7.25, + "learning_rate": 1.377649115140922e-06, + "loss": 2.6416, + "step": 33209 + }, + { + "epoch": 7.25, + "learning_rate": 1.3775398732794407e-06, + "loss": 1.5219, + "step": 33210 + }, + { + "epoch": 7.25, + "learning_rate": 1.3774306314179595e-06, + "loss": 1.1321, + "step": 33211 + }, + { + "epoch": 7.25, + "learning_rate": 1.3773213895564783e-06, + "loss": 0.3085, + "step": 33212 + }, + { + "epoch": 7.25, + "learning_rate": 1.3772121476949967e-06, + "loss": 0.7174, + "step": 33213 + }, + { + "epoch": 7.25, + "learning_rate": 1.3771029058335155e-06, + "loss": 0.9741, + "step": 33214 + }, + { + "epoch": 7.25, + "learning_rate": 1.3769936639720341e-06, + "loss": 0.7621, + "step": 33215 + }, + { + "epoch": 7.25, + "learning_rate": 1.376884422110553e-06, + "loss": 1.0517, + "step": 33216 + }, + { + "epoch": 7.25, + "learning_rate": 1.3767751802490714e-06, + "loss": 0.8705, + "step": 33217 + }, + { + "epoch": 7.25, + "learning_rate": 1.3766659383875902e-06, + "loss": 0.7199, + "step": 33218 + }, + { + "epoch": 7.25, + "learning_rate": 1.376556696526109e-06, + "loss": 0.2767, + "step": 33219 + }, + { + "epoch": 7.25, + "learning_rate": 1.3764474546646276e-06, + "loss": 1.2412, + "step": 33220 + }, + { + "epoch": 7.25, + "learning_rate": 1.3763382128031462e-06, + "loss": 1.3404, + "step": 33221 + }, + { + "epoch": 7.25, + "learning_rate": 1.3762289709416648e-06, + "loss": 0.8296, + "step": 33222 + }, + { + "epoch": 7.25, + "learning_rate": 1.3761197290801837e-06, + "loss": 0.382, + "step": 33223 + }, + { + "epoch": 7.25, + "learning_rate": 1.3760104872187025e-06, + "loss": 0.413, + "step": 33224 + }, + { + "epoch": 7.25, + "learning_rate": 1.375901245357221e-06, + "loss": 0.2502, + "step": 33225 + }, + { + "epoch": 7.25, + "learning_rate": 1.3757920034957397e-06, + "loss": 1.8095, + "step": 33226 + }, + { + "epoch": 7.25, + "learning_rate": 1.3756827616342583e-06, + "loss": 2.0211, + "step": 33227 + }, + { + "epoch": 7.25, + "learning_rate": 1.3755735197727772e-06, + "loss": 0.6624, + "step": 33228 + }, + { + "epoch": 7.25, + "learning_rate": 1.3754642779112955e-06, + "loss": 1.6159, + "step": 33229 + }, + { + "epoch": 7.25, + "learning_rate": 1.3753550360498144e-06, + "loss": 1.8204, + "step": 33230 + }, + { + "epoch": 7.25, + "learning_rate": 1.3752457941883332e-06, + "loss": 1.8746, + "step": 33231 + }, + { + "epoch": 7.25, + "learning_rate": 1.3751365523268518e-06, + "loss": 1.6211, + "step": 33232 + }, + { + "epoch": 7.25, + "learning_rate": 1.3750273104653706e-06, + "loss": 0.55, + "step": 33233 + }, + { + "epoch": 7.25, + "learning_rate": 1.374918068603889e-06, + "loss": 1.1179, + "step": 33234 + }, + { + "epoch": 7.25, + "learning_rate": 1.3748088267424079e-06, + "loss": 1.165, + "step": 33235 + }, + { + "epoch": 7.25, + "learning_rate": 1.3746995848809265e-06, + "loss": 0.3645, + "step": 33236 + }, + { + "epoch": 7.25, + "learning_rate": 1.374590343019445e-06, + "loss": 1.2764, + "step": 33237 + }, + { + "epoch": 7.25, + "learning_rate": 1.374481101157964e-06, + "loss": 1.6863, + "step": 33238 + }, + { + "epoch": 7.25, + "learning_rate": 1.3743718592964825e-06, + "loss": 0.6905, + "step": 33239 + }, + { + "epoch": 7.25, + "learning_rate": 1.3742626174350013e-06, + "loss": 3.3591, + "step": 33240 + }, + { + "epoch": 7.25, + "learning_rate": 1.3741533755735197e-06, + "loss": 1.2143, + "step": 33241 + }, + { + "epoch": 7.25, + "learning_rate": 1.3740441337120386e-06, + "loss": 0.9042, + "step": 33242 + }, + { + "epoch": 7.26, + "learning_rate": 1.3739348918505572e-06, + "loss": 1.8072, + "step": 33243 + }, + { + "epoch": 7.26, + "learning_rate": 1.373825649989076e-06, + "loss": 2.077, + "step": 33244 + }, + { + "epoch": 7.26, + "learning_rate": 1.3737164081275948e-06, + "loss": 1.7303, + "step": 33245 + }, + { + "epoch": 7.26, + "learning_rate": 1.3736071662661132e-06, + "loss": 1.6634, + "step": 33246 + }, + { + "epoch": 7.26, + "learning_rate": 1.373497924404632e-06, + "loss": 1.2315, + "step": 33247 + }, + { + "epoch": 7.26, + "learning_rate": 1.3733886825431506e-06, + "loss": 0.8719, + "step": 33248 + }, + { + "epoch": 7.26, + "learning_rate": 1.3732794406816693e-06, + "loss": 1.0989, + "step": 33249 + }, + { + "epoch": 7.26, + "learning_rate": 1.3731701988201879e-06, + "loss": 2.9692, + "step": 33250 + }, + { + "epoch": 7.26, + "learning_rate": 1.3730609569587067e-06, + "loss": 1.6822, + "step": 33251 + }, + { + "epoch": 7.26, + "learning_rate": 1.3729517150972255e-06, + "loss": 4.48, + "step": 33252 + }, + { + "epoch": 7.26, + "learning_rate": 1.372842473235744e-06, + "loss": 1.0755, + "step": 33253 + }, + { + "epoch": 7.26, + "learning_rate": 1.3727332313742627e-06, + "loss": 0.5066, + "step": 33254 + }, + { + "epoch": 7.26, + "learning_rate": 1.3726239895127813e-06, + "loss": 0.2878, + "step": 33255 + }, + { + "epoch": 7.26, + "learning_rate": 1.3725147476513002e-06, + "loss": 2.5158, + "step": 33256 + }, + { + "epoch": 7.26, + "learning_rate": 1.372405505789819e-06, + "loss": 2.2485, + "step": 33257 + }, + { + "epoch": 7.26, + "learning_rate": 1.3722962639283374e-06, + "loss": 1.8982, + "step": 33258 + }, + { + "epoch": 7.26, + "learning_rate": 1.3721870220668562e-06, + "loss": 0.9645, + "step": 33259 + }, + { + "epoch": 7.26, + "learning_rate": 1.3720777802053748e-06, + "loss": 0.7092, + "step": 33260 + }, + { + "epoch": 7.26, + "learning_rate": 1.3719685383438936e-06, + "loss": 0.6133, + "step": 33261 + }, + { + "epoch": 7.26, + "learning_rate": 1.371859296482412e-06, + "loss": 1.137, + "step": 33262 + }, + { + "epoch": 7.26, + "learning_rate": 1.3717500546209309e-06, + "loss": 1.9553, + "step": 33263 + }, + { + "epoch": 7.26, + "learning_rate": 1.3716408127594497e-06, + "loss": 0.5208, + "step": 33264 + }, + { + "epoch": 7.26, + "learning_rate": 1.371531570897968e-06, + "loss": 0.3738, + "step": 33265 + }, + { + "epoch": 7.26, + "learning_rate": 1.371422329036487e-06, + "loss": 1.1415, + "step": 33266 + }, + { + "epoch": 7.26, + "learning_rate": 1.3713130871750055e-06, + "loss": 1.0969, + "step": 33267 + }, + { + "epoch": 7.26, + "learning_rate": 1.3712038453135243e-06, + "loss": 2.5132, + "step": 33268 + }, + { + "epoch": 7.26, + "learning_rate": 1.3710946034520427e-06, + "loss": 0.6457, + "step": 33269 + }, + { + "epoch": 7.26, + "learning_rate": 1.3709853615905616e-06, + "loss": 1.8974, + "step": 33270 + }, + { + "epoch": 7.26, + "learning_rate": 1.3708761197290804e-06, + "loss": 0.0801, + "step": 33271 + }, + { + "epoch": 7.26, + "learning_rate": 1.370766877867599e-06, + "loss": 1.5995, + "step": 33272 + }, + { + "epoch": 7.26, + "learning_rate": 1.3706576360061178e-06, + "loss": 0.2907, + "step": 33273 + }, + { + "epoch": 7.26, + "learning_rate": 1.3705483941446362e-06, + "loss": 1.1314, + "step": 33274 + }, + { + "epoch": 7.26, + "learning_rate": 1.370439152283155e-06, + "loss": 0.9388, + "step": 33275 + }, + { + "epoch": 7.26, + "learning_rate": 1.3703299104216737e-06, + "loss": 0.2493, + "step": 33276 + }, + { + "epoch": 7.26, + "learning_rate": 1.3702206685601923e-06, + "loss": 0.8608, + "step": 33277 + }, + { + "epoch": 7.26, + "learning_rate": 1.370111426698711e-06, + "loss": 0.7036, + "step": 33278 + }, + { + "epoch": 7.26, + "learning_rate": 1.3700021848372297e-06, + "loss": 1.5508, + "step": 33279 + }, + { + "epoch": 7.26, + "learning_rate": 1.3698929429757485e-06, + "loss": 1.6145, + "step": 33280 + }, + { + "epoch": 7.26, + "learning_rate": 1.369783701114267e-06, + "loss": 1.0447, + "step": 33281 + }, + { + "epoch": 7.26, + "learning_rate": 1.3696744592527857e-06, + "loss": 2.4374, + "step": 33282 + }, + { + "epoch": 7.26, + "learning_rate": 1.3695652173913044e-06, + "loss": 0.4415, + "step": 33283 + }, + { + "epoch": 7.26, + "learning_rate": 1.3694559755298232e-06, + "loss": 1.6155, + "step": 33284 + }, + { + "epoch": 7.26, + "learning_rate": 1.369346733668342e-06, + "loss": 1.1795, + "step": 33285 + }, + { + "epoch": 7.26, + "learning_rate": 1.3692374918068604e-06, + "loss": 0.2619, + "step": 33286 + }, + { + "epoch": 7.26, + "learning_rate": 1.3691282499453792e-06, + "loss": 0.9513, + "step": 33287 + }, + { + "epoch": 7.26, + "learning_rate": 1.3690190080838978e-06, + "loss": 0.9792, + "step": 33288 + }, + { + "epoch": 7.27, + "learning_rate": 1.3689097662224167e-06, + "loss": 1.3012, + "step": 33289 + }, + { + "epoch": 7.27, + "learning_rate": 1.3688005243609353e-06, + "loss": 1.414, + "step": 33290 + }, + { + "epoch": 7.27, + "learning_rate": 1.3686912824994539e-06, + "loss": 1.3914, + "step": 33291 + }, + { + "epoch": 7.27, + "learning_rate": 1.3685820406379727e-06, + "loss": 1.3429, + "step": 33292 + }, + { + "epoch": 7.27, + "learning_rate": 1.368472798776491e-06, + "loss": 0.0794, + "step": 33293 + }, + { + "epoch": 7.27, + "learning_rate": 1.36836355691501e-06, + "loss": 0.9035, + "step": 33294 + }, + { + "epoch": 7.27, + "learning_rate": 1.3682543150535285e-06, + "loss": 0.8135, + "step": 33295 + }, + { + "epoch": 7.27, + "learning_rate": 1.3681450731920474e-06, + "loss": 0.8329, + "step": 33296 + }, + { + "epoch": 7.27, + "learning_rate": 1.3680358313305662e-06, + "loss": 0.6678, + "step": 33297 + }, + { + "epoch": 7.27, + "learning_rate": 1.3679265894690846e-06, + "loss": 1.8337, + "step": 33298 + }, + { + "epoch": 7.27, + "learning_rate": 1.3678173476076034e-06, + "loss": 1.168, + "step": 33299 + }, + { + "epoch": 7.27, + "learning_rate": 1.367708105746122e-06, + "loss": 1.6205, + "step": 33300 + }, + { + "epoch": 7.27, + "learning_rate": 1.3675988638846408e-06, + "loss": 0.6024, + "step": 33301 + }, + { + "epoch": 7.27, + "learning_rate": 1.3674896220231592e-06, + "loss": 0.9758, + "step": 33302 + }, + { + "epoch": 7.27, + "learning_rate": 1.367380380161678e-06, + "loss": 2.3108, + "step": 33303 + }, + { + "epoch": 7.27, + "learning_rate": 1.3672711383001969e-06, + "loss": 1.07, + "step": 33304 + }, + { + "epoch": 7.27, + "learning_rate": 1.3671618964387155e-06, + "loss": 2.6738, + "step": 33305 + }, + { + "epoch": 7.27, + "learning_rate": 1.367052654577234e-06, + "loss": 4.1256, + "step": 33306 + }, + { + "epoch": 7.27, + "learning_rate": 1.3669434127157527e-06, + "loss": 0.1049, + "step": 33307 + }, + { + "epoch": 7.27, + "learning_rate": 1.3668341708542715e-06, + "loss": 0.6668, + "step": 33308 + }, + { + "epoch": 7.27, + "learning_rate": 1.36672492899279e-06, + "loss": 1.8411, + "step": 33309 + }, + { + "epoch": 7.27, + "learning_rate": 1.3666156871313088e-06, + "loss": 0.4708, + "step": 33310 + }, + { + "epoch": 7.27, + "learning_rate": 1.3665064452698276e-06, + "loss": 1.5035, + "step": 33311 + }, + { + "epoch": 7.27, + "learning_rate": 1.3663972034083462e-06, + "loss": 1.0206, + "step": 33312 + }, + { + "epoch": 7.27, + "learning_rate": 1.366287961546865e-06, + "loss": 0.6557, + "step": 33313 + }, + { + "epoch": 7.27, + "learning_rate": 1.3661787196853834e-06, + "loss": 1.8996, + "step": 33314 + }, + { + "epoch": 7.27, + "learning_rate": 1.3660694778239022e-06, + "loss": 1.0043, + "step": 33315 + }, + { + "epoch": 7.27, + "learning_rate": 1.3659602359624208e-06, + "loss": 1.6372, + "step": 33316 + }, + { + "epoch": 7.27, + "learning_rate": 1.3658509941009397e-06, + "loss": 0.4657, + "step": 33317 + }, + { + "epoch": 7.27, + "learning_rate": 1.3657417522394583e-06, + "loss": 0.4012, + "step": 33318 + }, + { + "epoch": 7.27, + "learning_rate": 1.3656325103779769e-06, + "loss": 0.5947, + "step": 33319 + }, + { + "epoch": 7.27, + "learning_rate": 1.3655232685164957e-06, + "loss": 1.6467, + "step": 33320 + }, + { + "epoch": 7.27, + "learning_rate": 1.3654140266550141e-06, + "loss": 1.7254, + "step": 33321 + }, + { + "epoch": 7.27, + "learning_rate": 1.365304784793533e-06, + "loss": 1.9111, + "step": 33322 + }, + { + "epoch": 7.27, + "learning_rate": 1.3651955429320518e-06, + "loss": 0.6976, + "step": 33323 + }, + { + "epoch": 7.27, + "learning_rate": 1.3650863010705704e-06, + "loss": 1.7497, + "step": 33324 + }, + { + "epoch": 7.27, + "learning_rate": 1.3649770592090892e-06, + "loss": 1.8702, + "step": 33325 + }, + { + "epoch": 7.27, + "learning_rate": 1.3648678173476076e-06, + "loss": 1.1213, + "step": 33326 + }, + { + "epoch": 7.27, + "learning_rate": 1.3647585754861264e-06, + "loss": 1.5568, + "step": 33327 + }, + { + "epoch": 7.27, + "learning_rate": 1.364649333624645e-06, + "loss": 1.451, + "step": 33328 + }, + { + "epoch": 7.27, + "learning_rate": 1.3645400917631638e-06, + "loss": 0.4471, + "step": 33329 + }, + { + "epoch": 7.27, + "learning_rate": 1.3644308499016825e-06, + "loss": 1.5146, + "step": 33330 + }, + { + "epoch": 7.27, + "learning_rate": 1.364321608040201e-06, + "loss": 0.8925, + "step": 33331 + }, + { + "epoch": 7.27, + "learning_rate": 1.3642123661787199e-06, + "loss": 1.89, + "step": 33332 + }, + { + "epoch": 7.27, + "learning_rate": 1.3641031243172385e-06, + "loss": 0.9784, + "step": 33333 + }, + { + "epoch": 7.27, + "learning_rate": 1.3639938824557571e-06, + "loss": 0.6234, + "step": 33334 + }, + { + "epoch": 7.28, + "learning_rate": 1.3638846405942757e-06, + "loss": 0.4575, + "step": 33335 + }, + { + "epoch": 7.28, + "learning_rate": 1.3637753987327945e-06, + "loss": 0.6387, + "step": 33336 + }, + { + "epoch": 7.28, + "learning_rate": 1.3636661568713134e-06, + "loss": 1.4554, + "step": 33337 + }, + { + "epoch": 7.28, + "learning_rate": 1.3635569150098318e-06, + "loss": 0.499, + "step": 33338 + }, + { + "epoch": 7.28, + "learning_rate": 1.3634476731483506e-06, + "loss": 0.1754, + "step": 33339 + }, + { + "epoch": 7.28, + "learning_rate": 1.3633384312868692e-06, + "loss": 0.8031, + "step": 33340 + }, + { + "epoch": 7.28, + "learning_rate": 1.363229189425388e-06, + "loss": 0.6239, + "step": 33341 + }, + { + "epoch": 7.28, + "learning_rate": 1.3631199475639064e-06, + "loss": 2.5482, + "step": 33342 + }, + { + "epoch": 7.28, + "learning_rate": 1.3630107057024252e-06, + "loss": 1.6477, + "step": 33343 + }, + { + "epoch": 7.28, + "learning_rate": 1.362901463840944e-06, + "loss": 1.9729, + "step": 33344 + }, + { + "epoch": 7.28, + "learning_rate": 1.3627922219794627e-06, + "loss": 2.3529, + "step": 33345 + }, + { + "epoch": 7.28, + "learning_rate": 1.3626829801179813e-06, + "loss": 2.4865, + "step": 33346 + }, + { + "epoch": 7.28, + "learning_rate": 1.3625737382565e-06, + "loss": 1.989, + "step": 33347 + }, + { + "epoch": 7.28, + "learning_rate": 1.3624644963950187e-06, + "loss": 1.1911, + "step": 33348 + }, + { + "epoch": 7.28, + "learning_rate": 1.3623552545335371e-06, + "loss": 0.1115, + "step": 33349 + }, + { + "epoch": 7.28, + "learning_rate": 1.362246012672056e-06, + "loss": 1.1943, + "step": 33350 + }, + { + "epoch": 7.28, + "learning_rate": 1.3621367708105748e-06, + "loss": 1.369, + "step": 33351 + }, + { + "epoch": 7.28, + "learning_rate": 1.3620275289490934e-06, + "loss": 0.4372, + "step": 33352 + }, + { + "epoch": 7.28, + "learning_rate": 1.3619182870876122e-06, + "loss": 1.3675, + "step": 33353 + }, + { + "epoch": 7.28, + "learning_rate": 1.3618090452261306e-06, + "loss": 1.3064, + "step": 33354 + }, + { + "epoch": 7.28, + "learning_rate": 1.3616998033646494e-06, + "loss": 1.8781, + "step": 33355 + }, + { + "epoch": 7.28, + "learning_rate": 1.3615905615031683e-06, + "loss": 0.9459, + "step": 33356 + }, + { + "epoch": 7.28, + "learning_rate": 1.3614813196416869e-06, + "loss": 2.2816, + "step": 33357 + }, + { + "epoch": 7.28, + "learning_rate": 1.3613720777802057e-06, + "loss": 2.1422, + "step": 33358 + }, + { + "epoch": 7.28, + "learning_rate": 1.361262835918724e-06, + "loss": 1.0577, + "step": 33359 + }, + { + "epoch": 7.28, + "learning_rate": 1.361153594057243e-06, + "loss": 0.9178, + "step": 33360 + }, + { + "epoch": 7.28, + "learning_rate": 1.3610443521957615e-06, + "loss": 1.5471, + "step": 33361 + }, + { + "epoch": 7.28, + "learning_rate": 1.3609351103342801e-06, + "loss": 1.077, + "step": 33362 + }, + { + "epoch": 7.28, + "learning_rate": 1.360825868472799e-06, + "loss": 0.7825, + "step": 33363 + }, + { + "epoch": 7.28, + "learning_rate": 1.3607166266113176e-06, + "loss": 1.0649, + "step": 33364 + }, + { + "epoch": 7.28, + "learning_rate": 1.3606073847498364e-06, + "loss": 0.6591, + "step": 33365 + }, + { + "epoch": 7.28, + "learning_rate": 1.3604981428883548e-06, + "loss": 1.2849, + "step": 33366 + }, + { + "epoch": 7.28, + "learning_rate": 1.3603889010268736e-06, + "loss": 1.5291, + "step": 33367 + }, + { + "epoch": 7.28, + "learning_rate": 1.3602796591653922e-06, + "loss": 1.7115, + "step": 33368 + }, + { + "epoch": 7.28, + "learning_rate": 1.360170417303911e-06, + "loss": 0.4602, + "step": 33369 + }, + { + "epoch": 7.28, + "learning_rate": 1.3600611754424299e-06, + "loss": 1.6266, + "step": 33370 + }, + { + "epoch": 7.28, + "learning_rate": 1.3599519335809483e-06, + "loss": 1.9, + "step": 33371 + }, + { + "epoch": 7.28, + "learning_rate": 1.359842691719467e-06, + "loss": 0.2499, + "step": 33372 + }, + { + "epoch": 7.28, + "learning_rate": 1.3597334498579857e-06, + "loss": 0.4984, + "step": 33373 + }, + { + "epoch": 7.28, + "learning_rate": 1.3596242079965043e-06, + "loss": 0.3925, + "step": 33374 + }, + { + "epoch": 7.28, + "learning_rate": 1.359514966135023e-06, + "loss": 1.3281, + "step": 33375 + }, + { + "epoch": 7.28, + "learning_rate": 1.3594057242735417e-06, + "loss": 1.5524, + "step": 33376 + }, + { + "epoch": 7.28, + "learning_rate": 1.3592964824120606e-06, + "loss": 2.849, + "step": 33377 + }, + { + "epoch": 7.28, + "learning_rate": 1.359187240550579e-06, + "loss": 2.215, + "step": 33378 + }, + { + "epoch": 7.28, + "learning_rate": 1.3590779986890978e-06, + "loss": 1.145, + "step": 33379 + }, + { + "epoch": 7.29, + "learning_rate": 1.3589687568276164e-06, + "loss": 1.825, + "step": 33380 + }, + { + "epoch": 7.29, + "learning_rate": 1.3588595149661352e-06, + "loss": 2.3713, + "step": 33381 + }, + { + "epoch": 7.29, + "learning_rate": 1.3587502731046536e-06, + "loss": 0.8935, + "step": 33382 + }, + { + "epoch": 7.29, + "learning_rate": 1.3586410312431724e-06, + "loss": 1.6953, + "step": 33383 + }, + { + "epoch": 7.29, + "learning_rate": 1.3585317893816913e-06, + "loss": 1.6625, + "step": 33384 + }, + { + "epoch": 7.29, + "learning_rate": 1.3584225475202099e-06, + "loss": 0.5604, + "step": 33385 + }, + { + "epoch": 7.29, + "learning_rate": 1.3583133056587287e-06, + "loss": 0.9076, + "step": 33386 + }, + { + "epoch": 7.29, + "learning_rate": 1.358204063797247e-06, + "loss": 0.4737, + "step": 33387 + }, + { + "epoch": 7.29, + "learning_rate": 1.358094821935766e-06, + "loss": 1.5912, + "step": 33388 + }, + { + "epoch": 7.29, + "learning_rate": 1.3579855800742847e-06, + "loss": 0.6375, + "step": 33389 + }, + { + "epoch": 7.29, + "learning_rate": 1.3578763382128031e-06, + "loss": 1.2822, + "step": 33390 + }, + { + "epoch": 7.29, + "learning_rate": 1.357767096351322e-06, + "loss": 0.9804, + "step": 33391 + }, + { + "epoch": 7.29, + "learning_rate": 1.3576578544898406e-06, + "loss": 1.7126, + "step": 33392 + }, + { + "epoch": 7.29, + "learning_rate": 1.3575486126283594e-06, + "loss": 0.6253, + "step": 33393 + }, + { + "epoch": 7.29, + "learning_rate": 1.3574393707668778e-06, + "loss": 1.4394, + "step": 33394 + }, + { + "epoch": 7.29, + "learning_rate": 1.3573301289053966e-06, + "loss": 0.7763, + "step": 33395 + }, + { + "epoch": 7.29, + "learning_rate": 1.3572208870439154e-06, + "loss": 1.9142, + "step": 33396 + }, + { + "epoch": 7.29, + "learning_rate": 1.357111645182434e-06, + "loss": 1.6292, + "step": 33397 + }, + { + "epoch": 7.29, + "learning_rate": 1.3570024033209529e-06, + "loss": 1.0835, + "step": 33398 + }, + { + "epoch": 7.29, + "learning_rate": 1.3568931614594713e-06, + "loss": 1.4652, + "step": 33399 + }, + { + "epoch": 7.29, + "learning_rate": 1.35678391959799e-06, + "loss": 1.0151, + "step": 33400 + }, + { + "epoch": 7.29, + "learning_rate": 1.3566746777365087e-06, + "loss": 1.488, + "step": 33401 + }, + { + "epoch": 7.29, + "learning_rate": 1.3565654358750273e-06, + "loss": 0.9357, + "step": 33402 + }, + { + "epoch": 7.29, + "learning_rate": 1.3564561940135461e-06, + "loss": 1.3117, + "step": 33403 + }, + { + "epoch": 7.29, + "learning_rate": 1.3563469521520648e-06, + "loss": 0.6891, + "step": 33404 + }, + { + "epoch": 7.29, + "learning_rate": 1.3562377102905836e-06, + "loss": 2.2449, + "step": 33405 + }, + { + "epoch": 7.29, + "learning_rate": 1.356128468429102e-06, + "loss": 0.991, + "step": 33406 + }, + { + "epoch": 7.29, + "learning_rate": 1.3560192265676208e-06, + "loss": 1.3852, + "step": 33407 + }, + { + "epoch": 7.29, + "learning_rate": 1.3559099847061394e-06, + "loss": 0.2805, + "step": 33408 + }, + { + "epoch": 7.29, + "learning_rate": 1.3558007428446582e-06, + "loss": 3.398, + "step": 33409 + }, + { + "epoch": 7.29, + "learning_rate": 1.355691500983177e-06, + "loss": 0.662, + "step": 33410 + }, + { + "epoch": 7.29, + "learning_rate": 1.3555822591216955e-06, + "loss": 0.4287, + "step": 33411 + }, + { + "epoch": 7.29, + "learning_rate": 1.3554730172602143e-06, + "loss": 0.5651, + "step": 33412 + }, + { + "epoch": 7.29, + "learning_rate": 1.3553637753987329e-06, + "loss": 0.5247, + "step": 33413 + }, + { + "epoch": 7.29, + "learning_rate": 1.3552545335372517e-06, + "loss": 0.5995, + "step": 33414 + }, + { + "epoch": 7.29, + "learning_rate": 1.3551452916757701e-06, + "loss": 2.1394, + "step": 33415 + }, + { + "epoch": 7.29, + "learning_rate": 1.355036049814289e-06, + "loss": 0.5231, + "step": 33416 + }, + { + "epoch": 7.29, + "learning_rate": 1.3549268079528078e-06, + "loss": 1.0344, + "step": 33417 + }, + { + "epoch": 7.29, + "learning_rate": 1.3548175660913262e-06, + "loss": 1.138, + "step": 33418 + }, + { + "epoch": 7.29, + "learning_rate": 1.354708324229845e-06, + "loss": 0.8172, + "step": 33419 + }, + { + "epoch": 7.29, + "learning_rate": 1.3545990823683636e-06, + "loss": 0.6524, + "step": 33420 + }, + { + "epoch": 7.29, + "learning_rate": 1.3544898405068824e-06, + "loss": 0.6456, + "step": 33421 + }, + { + "epoch": 7.29, + "learning_rate": 1.3543805986454008e-06, + "loss": 0.5896, + "step": 33422 + }, + { + "epoch": 7.29, + "learning_rate": 1.3542713567839196e-06, + "loss": 0.2047, + "step": 33423 + }, + { + "epoch": 7.29, + "learning_rate": 1.3541621149224385e-06, + "loss": 0.9875, + "step": 33424 + }, + { + "epoch": 7.29, + "learning_rate": 1.354052873060957e-06, + "loss": 1.2671, + "step": 33425 + }, + { + "epoch": 7.3, + "learning_rate": 1.3539436311994759e-06, + "loss": 1.2152, + "step": 33426 + }, + { + "epoch": 7.3, + "learning_rate": 1.3538343893379943e-06, + "loss": 0.5986, + "step": 33427 + }, + { + "epoch": 7.3, + "learning_rate": 1.3537251474765131e-06, + "loss": 0.5266, + "step": 33428 + }, + { + "epoch": 7.3, + "learning_rate": 1.353615905615032e-06, + "loss": 1.4109, + "step": 33429 + }, + { + "epoch": 7.3, + "learning_rate": 1.3535066637535505e-06, + "loss": 0.7591, + "step": 33430 + }, + { + "epoch": 7.3, + "learning_rate": 1.3533974218920692e-06, + "loss": 1.2269, + "step": 33431 + }, + { + "epoch": 7.3, + "learning_rate": 1.3532881800305878e-06, + "loss": 0.5042, + "step": 33432 + }, + { + "epoch": 7.3, + "learning_rate": 1.3531789381691066e-06, + "loss": 0.7943, + "step": 33433 + }, + { + "epoch": 7.3, + "learning_rate": 1.353069696307625e-06, + "loss": 1.2571, + "step": 33434 + }, + { + "epoch": 7.3, + "learning_rate": 1.3529604544461438e-06, + "loss": 0.8879, + "step": 33435 + }, + { + "epoch": 7.3, + "learning_rate": 1.3528512125846626e-06, + "loss": 3.1975, + "step": 33436 + }, + { + "epoch": 7.3, + "learning_rate": 1.3527419707231812e-06, + "loss": 0.6457, + "step": 33437 + }, + { + "epoch": 7.3, + "learning_rate": 1.3526327288617e-06, + "loss": 1.0377, + "step": 33438 + }, + { + "epoch": 7.3, + "learning_rate": 1.3525234870002185e-06, + "loss": 1.4618, + "step": 33439 + }, + { + "epoch": 7.3, + "learning_rate": 1.3524142451387373e-06, + "loss": 0.7383, + "step": 33440 + }, + { + "epoch": 7.3, + "learning_rate": 1.352305003277256e-06, + "loss": 1.8744, + "step": 33441 + }, + { + "epoch": 7.3, + "learning_rate": 1.3521957614157747e-06, + "loss": 1.0437, + "step": 33442 + }, + { + "epoch": 7.3, + "learning_rate": 1.3520865195542933e-06, + "loss": 1.011, + "step": 33443 + }, + { + "epoch": 7.3, + "learning_rate": 1.351977277692812e-06, + "loss": 2.2248, + "step": 33444 + }, + { + "epoch": 7.3, + "learning_rate": 1.3518680358313308e-06, + "loss": 3.0564, + "step": 33445 + }, + { + "epoch": 7.3, + "learning_rate": 1.3517587939698492e-06, + "loss": 0.528, + "step": 33446 + }, + { + "epoch": 7.3, + "learning_rate": 1.351649552108368e-06, + "loss": 1.9687, + "step": 33447 + }, + { + "epoch": 7.3, + "learning_rate": 1.3515403102468866e-06, + "loss": 2.4069, + "step": 33448 + }, + { + "epoch": 7.3, + "learning_rate": 1.3514310683854054e-06, + "loss": 0.2628, + "step": 33449 + }, + { + "epoch": 7.3, + "learning_rate": 1.3513218265239242e-06, + "loss": 1.3943, + "step": 33450 + }, + { + "epoch": 7.3, + "learning_rate": 1.3512125846624426e-06, + "loss": 1.399, + "step": 33451 + }, + { + "epoch": 7.3, + "learning_rate": 1.3511033428009615e-06, + "loss": 0.5548, + "step": 33452 + }, + { + "epoch": 7.3, + "learning_rate": 1.35099410093948e-06, + "loss": 0.1897, + "step": 33453 + }, + { + "epoch": 7.3, + "learning_rate": 1.350884859077999e-06, + "loss": 2.5877, + "step": 33454 + }, + { + "epoch": 7.3, + "learning_rate": 1.3507756172165173e-06, + "loss": 1.5753, + "step": 33455 + }, + { + "epoch": 7.3, + "learning_rate": 1.3506663753550361e-06, + "loss": 0.7001, + "step": 33456 + }, + { + "epoch": 7.3, + "learning_rate": 1.350557133493555e-06, + "loss": 1.893, + "step": 33457 + }, + { + "epoch": 7.3, + "learning_rate": 1.3504478916320736e-06, + "loss": 0.2801, + "step": 33458 + }, + { + "epoch": 7.3, + "learning_rate": 1.3503386497705922e-06, + "loss": 1.6555, + "step": 33459 + }, + { + "epoch": 7.3, + "learning_rate": 1.3502294079091108e-06, + "loss": 1.2939, + "step": 33460 + }, + { + "epoch": 7.3, + "learning_rate": 1.3501201660476296e-06, + "loss": 2.3651, + "step": 33461 + }, + { + "epoch": 7.3, + "learning_rate": 1.3500109241861484e-06, + "loss": 1.5786, + "step": 33462 + }, + { + "epoch": 7.3, + "learning_rate": 1.3499016823246668e-06, + "loss": 2.0476, + "step": 33463 + }, + { + "epoch": 7.3, + "learning_rate": 1.3497924404631856e-06, + "loss": 2.4268, + "step": 33464 + }, + { + "epoch": 7.3, + "learning_rate": 1.3496831986017043e-06, + "loss": 1.7642, + "step": 33465 + }, + { + "epoch": 7.3, + "learning_rate": 1.349573956740223e-06, + "loss": 0.1491, + "step": 33466 + }, + { + "epoch": 7.3, + "learning_rate": 1.3494647148787415e-06, + "loss": 0.7725, + "step": 33467 + }, + { + "epoch": 7.3, + "learning_rate": 1.3493554730172603e-06, + "loss": 1.3044, + "step": 33468 + }, + { + "epoch": 7.3, + "learning_rate": 1.3492462311557791e-06, + "loss": 2.2489, + "step": 33469 + }, + { + "epoch": 7.3, + "learning_rate": 1.3491369892942977e-06, + "loss": 0.632, + "step": 33470 + }, + { + "epoch": 7.3, + "learning_rate": 1.3490277474328163e-06, + "loss": 0.187, + "step": 33471 + }, + { + "epoch": 7.31, + "learning_rate": 1.348918505571335e-06, + "loss": 0.5413, + "step": 33472 + }, + { + "epoch": 7.31, + "learning_rate": 1.3488092637098538e-06, + "loss": 0.9762, + "step": 33473 + }, + { + "epoch": 7.31, + "learning_rate": 1.3487000218483722e-06, + "loss": 1.2119, + "step": 33474 + }, + { + "epoch": 7.31, + "learning_rate": 1.348590779986891e-06, + "loss": 2.6992, + "step": 33475 + }, + { + "epoch": 7.31, + "learning_rate": 1.3484815381254098e-06, + "loss": 0.8248, + "step": 33476 + }, + { + "epoch": 7.31, + "learning_rate": 1.3483722962639284e-06, + "loss": 2.497, + "step": 33477 + }, + { + "epoch": 7.31, + "learning_rate": 1.3482630544024473e-06, + "loss": 0.3227, + "step": 33478 + }, + { + "epoch": 7.31, + "learning_rate": 1.3481538125409657e-06, + "loss": 1.9944, + "step": 33479 + }, + { + "epoch": 7.31, + "learning_rate": 1.3480445706794845e-06, + "loss": 1.3384, + "step": 33480 + }, + { + "epoch": 7.31, + "learning_rate": 1.347935328818003e-06, + "loss": 1.8564, + "step": 33481 + }, + { + "epoch": 7.31, + "learning_rate": 1.347826086956522e-06, + "loss": 0.621, + "step": 33482 + }, + { + "epoch": 7.31, + "learning_rate": 1.3477168450950407e-06, + "loss": 0.4476, + "step": 33483 + }, + { + "epoch": 7.31, + "learning_rate": 1.3476076032335591e-06, + "loss": 1.1314, + "step": 33484 + }, + { + "epoch": 7.31, + "learning_rate": 1.347498361372078e-06, + "loss": 0.9008, + "step": 33485 + }, + { + "epoch": 7.31, + "learning_rate": 1.3473891195105966e-06, + "loss": 0.4796, + "step": 33486 + }, + { + "epoch": 7.31, + "learning_rate": 1.3472798776491152e-06, + "loss": 0.8644, + "step": 33487 + }, + { + "epoch": 7.31, + "learning_rate": 1.3471706357876338e-06, + "loss": 0.8924, + "step": 33488 + }, + { + "epoch": 7.31, + "learning_rate": 1.3470613939261526e-06, + "loss": 0.171, + "step": 33489 + }, + { + "epoch": 7.31, + "learning_rate": 1.3469521520646714e-06, + "loss": 1.2149, + "step": 33490 + }, + { + "epoch": 7.31, + "learning_rate": 1.3468429102031898e-06, + "loss": 0.6589, + "step": 33491 + }, + { + "epoch": 7.31, + "learning_rate": 1.3467336683417087e-06, + "loss": 0.7628, + "step": 33492 + }, + { + "epoch": 7.31, + "learning_rate": 1.3466244264802273e-06, + "loss": 2.0521, + "step": 33493 + }, + { + "epoch": 7.31, + "learning_rate": 1.346515184618746e-06, + "loss": 0.7258, + "step": 33494 + }, + { + "epoch": 7.31, + "learning_rate": 1.346405942757265e-06, + "loss": 1.4641, + "step": 33495 + }, + { + "epoch": 7.31, + "learning_rate": 1.3462967008957833e-06, + "loss": 1.3055, + "step": 33496 + }, + { + "epoch": 7.31, + "learning_rate": 1.3461874590343021e-06, + "loss": 0.6482, + "step": 33497 + }, + { + "epoch": 7.31, + "learning_rate": 1.3460782171728208e-06, + "loss": 0.8714, + "step": 33498 + }, + { + "epoch": 7.31, + "learning_rate": 1.3459689753113394e-06, + "loss": 0.4099, + "step": 33499 + }, + { + "epoch": 7.31, + "learning_rate": 1.345859733449858e-06, + "loss": 0.2702, + "step": 33500 + }, + { + "epoch": 7.31, + "learning_rate": 1.3457504915883768e-06, + "loss": 0.9992, + "step": 33501 + }, + { + "epoch": 7.31, + "learning_rate": 1.3456412497268956e-06, + "loss": 2.1919, + "step": 33502 + }, + { + "epoch": 7.31, + "learning_rate": 1.345532007865414e-06, + "loss": 1.2028, + "step": 33503 + }, + { + "epoch": 7.31, + "learning_rate": 1.3454227660039328e-06, + "loss": 1.978, + "step": 33504 + }, + { + "epoch": 7.31, + "learning_rate": 1.3453135241424515e-06, + "loss": 1.4279, + "step": 33505 + }, + { + "epoch": 7.31, + "learning_rate": 1.3452042822809703e-06, + "loss": 1.0145, + "step": 33506 + }, + { + "epoch": 7.31, + "learning_rate": 1.3450950404194887e-06, + "loss": 0.26, + "step": 33507 + }, + { + "epoch": 7.31, + "learning_rate": 1.3449857985580075e-06, + "loss": 2.3154, + "step": 33508 + }, + { + "epoch": 7.31, + "learning_rate": 1.3448765566965263e-06, + "loss": 0.7625, + "step": 33509 + }, + { + "epoch": 7.31, + "learning_rate": 1.344767314835045e-06, + "loss": 2.2812, + "step": 33510 + }, + { + "epoch": 7.31, + "learning_rate": 1.3446580729735638e-06, + "loss": 1.0102, + "step": 33511 + }, + { + "epoch": 7.31, + "learning_rate": 1.3445488311120822e-06, + "loss": 0.5578, + "step": 33512 + }, + { + "epoch": 7.31, + "learning_rate": 1.344439589250601e-06, + "loss": 1.26, + "step": 33513 + }, + { + "epoch": 7.31, + "learning_rate": 1.3443303473891196e-06, + "loss": 1.4963, + "step": 33514 + }, + { + "epoch": 7.31, + "learning_rate": 1.3442211055276382e-06, + "loss": 0.7286, + "step": 33515 + }, + { + "epoch": 7.31, + "learning_rate": 1.344111863666157e-06, + "loss": 1.8742, + "step": 33516 + }, + { + "epoch": 7.31, + "learning_rate": 1.3440026218046756e-06, + "loss": 1.9849, + "step": 33517 + }, + { + "epoch": 7.32, + "learning_rate": 1.3438933799431945e-06, + "loss": 1.357, + "step": 33518 + }, + { + "epoch": 7.32, + "learning_rate": 1.3437841380817129e-06, + "loss": 1.9871, + "step": 33519 + }, + { + "epoch": 7.32, + "learning_rate": 1.3436748962202317e-06, + "loss": 1.6461, + "step": 33520 + }, + { + "epoch": 7.32, + "learning_rate": 1.3435656543587503e-06, + "loss": 1.4735, + "step": 33521 + }, + { + "epoch": 7.32, + "learning_rate": 1.3434564124972691e-06, + "loss": 0.2457, + "step": 33522 + }, + { + "epoch": 7.32, + "learning_rate": 1.343347170635788e-06, + "loss": 0.9377, + "step": 33523 + }, + { + "epoch": 7.32, + "learning_rate": 1.3432379287743063e-06, + "loss": 1.8671, + "step": 33524 + }, + { + "epoch": 7.32, + "learning_rate": 1.3431286869128252e-06, + "loss": 1.5276, + "step": 33525 + }, + { + "epoch": 7.32, + "learning_rate": 1.3430194450513438e-06, + "loss": 1.1462, + "step": 33526 + }, + { + "epoch": 7.32, + "learning_rate": 1.3429102031898624e-06, + "loss": 1.2624, + "step": 33527 + }, + { + "epoch": 7.32, + "learning_rate": 1.3428009613283812e-06, + "loss": 0.3845, + "step": 33528 + }, + { + "epoch": 7.32, + "learning_rate": 1.3426917194668998e-06, + "loss": 2.4312, + "step": 33529 + }, + { + "epoch": 7.32, + "learning_rate": 1.3425824776054186e-06, + "loss": 0.4056, + "step": 33530 + }, + { + "epoch": 7.32, + "learning_rate": 1.342473235743937e-06, + "loss": 1.5306, + "step": 33531 + }, + { + "epoch": 7.32, + "learning_rate": 1.3423639938824559e-06, + "loss": 0.5829, + "step": 33532 + }, + { + "epoch": 7.32, + "learning_rate": 1.3422547520209745e-06, + "loss": 3.3851, + "step": 33533 + }, + { + "epoch": 7.32, + "learning_rate": 1.3421455101594933e-06, + "loss": 0.4831, + "step": 33534 + }, + { + "epoch": 7.32, + "learning_rate": 1.3420362682980121e-06, + "loss": 0.6399, + "step": 33535 + }, + { + "epoch": 7.32, + "learning_rate": 1.3419270264365305e-06, + "loss": 2.1473, + "step": 33536 + }, + { + "epoch": 7.32, + "learning_rate": 1.3418177845750493e-06, + "loss": 1.9189, + "step": 33537 + }, + { + "epoch": 7.32, + "learning_rate": 1.341708542713568e-06, + "loss": 2.2326, + "step": 33538 + }, + { + "epoch": 7.32, + "learning_rate": 1.3415993008520868e-06, + "loss": 1.2387, + "step": 33539 + }, + { + "epoch": 7.32, + "learning_rate": 1.3414900589906052e-06, + "loss": 0.1048, + "step": 33540 + }, + { + "epoch": 7.32, + "learning_rate": 1.341380817129124e-06, + "loss": 1.1047, + "step": 33541 + }, + { + "epoch": 7.32, + "learning_rate": 1.3412715752676428e-06, + "loss": 1.2849, + "step": 33542 + }, + { + "epoch": 7.32, + "learning_rate": 1.3411623334061612e-06, + "loss": 1.0685, + "step": 33543 + }, + { + "epoch": 7.32, + "learning_rate": 1.34105309154468e-06, + "loss": 0.5443, + "step": 33544 + }, + { + "epoch": 7.32, + "learning_rate": 1.3409438496831986e-06, + "loss": 1.6734, + "step": 33545 + }, + { + "epoch": 7.32, + "learning_rate": 1.3408346078217175e-06, + "loss": 1.5661, + "step": 33546 + }, + { + "epoch": 7.32, + "learning_rate": 1.3407253659602359e-06, + "loss": 1.8969, + "step": 33547 + }, + { + "epoch": 7.32, + "learning_rate": 1.3406161240987547e-06, + "loss": 1.6967, + "step": 33548 + }, + { + "epoch": 7.32, + "learning_rate": 1.3405068822372735e-06, + "loss": 3.5459, + "step": 33549 + }, + { + "epoch": 7.32, + "learning_rate": 1.3403976403757921e-06, + "loss": 0.7871, + "step": 33550 + }, + { + "epoch": 7.32, + "learning_rate": 1.340288398514311e-06, + "loss": 0.2223, + "step": 33551 + }, + { + "epoch": 7.32, + "learning_rate": 1.3401791566528293e-06, + "loss": 0.4461, + "step": 33552 + }, + { + "epoch": 7.32, + "learning_rate": 1.3400699147913482e-06, + "loss": 0.7302, + "step": 33553 + }, + { + "epoch": 7.32, + "learning_rate": 1.3399606729298668e-06, + "loss": 1.6631, + "step": 33554 + }, + { + "epoch": 7.32, + "learning_rate": 1.3398514310683856e-06, + "loss": 2.1684, + "step": 33555 + }, + { + "epoch": 7.32, + "learning_rate": 1.3397421892069042e-06, + "loss": 0.9398, + "step": 33556 + }, + { + "epoch": 7.32, + "learning_rate": 1.3396329473454228e-06, + "loss": 0.9028, + "step": 33557 + }, + { + "epoch": 7.32, + "learning_rate": 1.3395237054839416e-06, + "loss": 0.959, + "step": 33558 + }, + { + "epoch": 7.32, + "learning_rate": 1.33941446362246e-06, + "loss": 2.1612, + "step": 33559 + }, + { + "epoch": 7.32, + "learning_rate": 1.3393052217609789e-06, + "loss": 1.3725, + "step": 33560 + }, + { + "epoch": 7.32, + "learning_rate": 1.3391959798994977e-06, + "loss": 1.1401, + "step": 33561 + }, + { + "epoch": 7.32, + "learning_rate": 1.3390867380380163e-06, + "loss": 0.9891, + "step": 33562 + }, + { + "epoch": 7.32, + "learning_rate": 1.3389774961765351e-06, + "loss": 1.7061, + "step": 33563 + }, + { + "epoch": 7.33, + "learning_rate": 1.3388682543150535e-06, + "loss": 0.9302, + "step": 33564 + }, + { + "epoch": 7.33, + "learning_rate": 1.3387590124535723e-06, + "loss": 0.2143, + "step": 33565 + }, + { + "epoch": 7.33, + "learning_rate": 1.338649770592091e-06, + "loss": 0.7826, + "step": 33566 + }, + { + "epoch": 7.33, + "learning_rate": 1.3385405287306098e-06, + "loss": 1.4477, + "step": 33567 + }, + { + "epoch": 7.33, + "learning_rate": 1.3384312868691284e-06, + "loss": 0.1739, + "step": 33568 + }, + { + "epoch": 7.33, + "learning_rate": 1.338322045007647e-06, + "loss": 2.5844, + "step": 33569 + }, + { + "epoch": 7.33, + "learning_rate": 1.3382128031461658e-06, + "loss": 3.2268, + "step": 33570 + }, + { + "epoch": 7.33, + "learning_rate": 1.3381035612846842e-06, + "loss": 1.146, + "step": 33571 + }, + { + "epoch": 7.33, + "learning_rate": 1.337994319423203e-06, + "loss": 2.4654, + "step": 33572 + }, + { + "epoch": 7.33, + "learning_rate": 1.3378850775617217e-06, + "loss": 1.2293, + "step": 33573 + }, + { + "epoch": 7.33, + "learning_rate": 1.3377758357002405e-06, + "loss": 1.1445, + "step": 33574 + }, + { + "epoch": 7.33, + "learning_rate": 1.3376665938387593e-06, + "loss": 0.1546, + "step": 33575 + }, + { + "epoch": 7.33, + "learning_rate": 1.3375573519772777e-06, + "loss": 1.8903, + "step": 33576 + }, + { + "epoch": 7.33, + "learning_rate": 1.3374481101157965e-06, + "loss": 0.3295, + "step": 33577 + }, + { + "epoch": 7.33, + "learning_rate": 1.3373388682543151e-06, + "loss": 2.7156, + "step": 33578 + }, + { + "epoch": 7.33, + "learning_rate": 1.337229626392834e-06, + "loss": 0.8282, + "step": 33579 + }, + { + "epoch": 7.33, + "learning_rate": 1.3371203845313524e-06, + "loss": 0.3838, + "step": 33580 + }, + { + "epoch": 7.33, + "learning_rate": 1.3370111426698712e-06, + "loss": 2.0525, + "step": 33581 + }, + { + "epoch": 7.33, + "learning_rate": 1.33690190080839e-06, + "loss": 1.8412, + "step": 33582 + }, + { + "epoch": 7.33, + "learning_rate": 1.3367926589469086e-06, + "loss": 0.147, + "step": 33583 + }, + { + "epoch": 7.33, + "learning_rate": 1.3366834170854272e-06, + "loss": 1.1527, + "step": 33584 + }, + { + "epoch": 7.33, + "learning_rate": 1.3365741752239458e-06, + "loss": 0.4713, + "step": 33585 + }, + { + "epoch": 7.33, + "learning_rate": 1.3364649333624647e-06, + "loss": 1.3327, + "step": 33586 + }, + { + "epoch": 7.33, + "learning_rate": 1.336355691500983e-06, + "loss": 1.0328, + "step": 33587 + }, + { + "epoch": 7.33, + "learning_rate": 1.3362464496395019e-06, + "loss": 0.7794, + "step": 33588 + }, + { + "epoch": 7.33, + "learning_rate": 1.3361372077780207e-06, + "loss": 0.2039, + "step": 33589 + }, + { + "epoch": 7.33, + "learning_rate": 1.3360279659165393e-06, + "loss": 1.5072, + "step": 33590 + }, + { + "epoch": 7.33, + "learning_rate": 1.3359187240550581e-06, + "loss": 0.1769, + "step": 33591 + }, + { + "epoch": 7.33, + "learning_rate": 1.3358094821935765e-06, + "loss": 1.624, + "step": 33592 + }, + { + "epoch": 7.33, + "learning_rate": 1.3357002403320954e-06, + "loss": 2.1097, + "step": 33593 + }, + { + "epoch": 7.33, + "learning_rate": 1.3355909984706142e-06, + "loss": 0.1677, + "step": 33594 + }, + { + "epoch": 7.33, + "learning_rate": 1.3354817566091328e-06, + "loss": 1.8063, + "step": 33595 + }, + { + "epoch": 7.33, + "learning_rate": 1.3353725147476514e-06, + "loss": 1.3619, + "step": 33596 + }, + { + "epoch": 7.33, + "learning_rate": 1.33526327288617e-06, + "loss": 1.2624, + "step": 33597 + }, + { + "epoch": 7.33, + "learning_rate": 1.3351540310246888e-06, + "loss": 0.1094, + "step": 33598 + }, + { + "epoch": 7.33, + "learning_rate": 1.3350447891632074e-06, + "loss": 1.8993, + "step": 33599 + }, + { + "epoch": 7.33, + "learning_rate": 1.334935547301726e-06, + "loss": 0.6096, + "step": 33600 + }, + { + "epoch": 7.33, + "learning_rate": 1.3348263054402449e-06, + "loss": 1.2727, + "step": 33601 + }, + { + "epoch": 7.33, + "learning_rate": 1.3347170635787635e-06, + "loss": 0.2567, + "step": 33602 + }, + { + "epoch": 7.33, + "learning_rate": 1.3346078217172823e-06, + "loss": 1.5734, + "step": 33603 + }, + { + "epoch": 7.33, + "learning_rate": 1.3344985798558007e-06, + "loss": 0.904, + "step": 33604 + }, + { + "epoch": 7.33, + "learning_rate": 1.3343893379943195e-06, + "loss": 0.5986, + "step": 33605 + }, + { + "epoch": 7.33, + "learning_rate": 1.3342800961328382e-06, + "loss": 1.6562, + "step": 33606 + }, + { + "epoch": 7.33, + "learning_rate": 1.334170854271357e-06, + "loss": 0.1732, + "step": 33607 + }, + { + "epoch": 7.33, + "learning_rate": 1.3340616124098758e-06, + "loss": 2.2547, + "step": 33608 + }, + { + "epoch": 7.34, + "learning_rate": 1.3339523705483942e-06, + "loss": 1.6785, + "step": 33609 + }, + { + "epoch": 7.34, + "learning_rate": 1.333843128686913e-06, + "loss": 1.4138, + "step": 33610 + }, + { + "epoch": 7.34, + "learning_rate": 1.3337338868254316e-06, + "loss": 1.9906, + "step": 33611 + }, + { + "epoch": 7.34, + "learning_rate": 1.3336246449639502e-06, + "loss": 2.7124, + "step": 33612 + }, + { + "epoch": 7.34, + "learning_rate": 1.3335154031024689e-06, + "loss": 0.3656, + "step": 33613 + }, + { + "epoch": 7.34, + "learning_rate": 1.3334061612409877e-06, + "loss": 0.3043, + "step": 33614 + }, + { + "epoch": 7.34, + "learning_rate": 1.3332969193795065e-06, + "loss": 0.9707, + "step": 33615 + }, + { + "epoch": 7.34, + "learning_rate": 1.333187677518025e-06, + "loss": 0.5625, + "step": 33616 + }, + { + "epoch": 7.34, + "learning_rate": 1.3330784356565437e-06, + "loss": 1.1651, + "step": 33617 + }, + { + "epoch": 7.34, + "learning_rate": 1.3329691937950623e-06, + "loss": 0.6339, + "step": 33618 + }, + { + "epoch": 7.34, + "learning_rate": 1.3328599519335812e-06, + "loss": 1.1497, + "step": 33619 + }, + { + "epoch": 7.34, + "learning_rate": 1.3327507100720996e-06, + "loss": 2.5452, + "step": 33620 + }, + { + "epoch": 7.34, + "learning_rate": 1.3326414682106184e-06, + "loss": 1.9022, + "step": 33621 + }, + { + "epoch": 7.34, + "learning_rate": 1.3325322263491372e-06, + "loss": 1.7478, + "step": 33622 + }, + { + "epoch": 7.34, + "learning_rate": 1.3324229844876558e-06, + "loss": 0.8834, + "step": 33623 + }, + { + "epoch": 7.34, + "learning_rate": 1.3323137426261744e-06, + "loss": 0.4473, + "step": 33624 + }, + { + "epoch": 7.34, + "learning_rate": 1.332204500764693e-06, + "loss": 0.7727, + "step": 33625 + }, + { + "epoch": 7.34, + "learning_rate": 1.3320952589032119e-06, + "loss": 2.9833, + "step": 33626 + }, + { + "epoch": 7.34, + "learning_rate": 1.3319860170417305e-06, + "loss": 0.9995, + "step": 33627 + }, + { + "epoch": 7.34, + "learning_rate": 1.331876775180249e-06, + "loss": 1.0617, + "step": 33628 + }, + { + "epoch": 7.34, + "learning_rate": 1.331767533318768e-06, + "loss": 1.9093, + "step": 33629 + }, + { + "epoch": 7.34, + "learning_rate": 1.3316582914572865e-06, + "loss": 1.1631, + "step": 33630 + }, + { + "epoch": 7.34, + "learning_rate": 1.3315490495958053e-06, + "loss": 0.4808, + "step": 33631 + }, + { + "epoch": 7.34, + "learning_rate": 1.3314398077343237e-06, + "loss": 0.1795, + "step": 33632 + }, + { + "epoch": 7.34, + "learning_rate": 1.3313305658728426e-06, + "loss": 0.9124, + "step": 33633 + }, + { + "epoch": 7.34, + "learning_rate": 1.3312213240113614e-06, + "loss": 0.3654, + "step": 33634 + }, + { + "epoch": 7.34, + "learning_rate": 1.33111208214988e-06, + "loss": 0.7939, + "step": 33635 + }, + { + "epoch": 7.34, + "learning_rate": 1.3310028402883988e-06, + "loss": 0.9558, + "step": 33636 + }, + { + "epoch": 7.34, + "learning_rate": 1.3308935984269172e-06, + "loss": 1.0636, + "step": 33637 + }, + { + "epoch": 7.34, + "learning_rate": 1.330784356565436e-06, + "loss": 0.5197, + "step": 33638 + }, + { + "epoch": 7.34, + "learning_rate": 1.3306751147039546e-06, + "loss": 0.7352, + "step": 33639 + }, + { + "epoch": 7.34, + "learning_rate": 1.3305658728424733e-06, + "loss": 0.5593, + "step": 33640 + }, + { + "epoch": 7.34, + "learning_rate": 1.330456630980992e-06, + "loss": 1.4999, + "step": 33641 + }, + { + "epoch": 7.34, + "learning_rate": 1.3303473891195107e-06, + "loss": 2.545, + "step": 33642 + }, + { + "epoch": 7.34, + "learning_rate": 1.3302381472580295e-06, + "loss": 0.8049, + "step": 33643 + }, + { + "epoch": 7.34, + "learning_rate": 1.330128905396548e-06, + "loss": 1.5525, + "step": 33644 + }, + { + "epoch": 7.34, + "learning_rate": 1.3300196635350667e-06, + "loss": 0.8383, + "step": 33645 + }, + { + "epoch": 7.34, + "learning_rate": 1.3299104216735853e-06, + "loss": 1.4566, + "step": 33646 + }, + { + "epoch": 7.34, + "learning_rate": 1.3298011798121042e-06, + "loss": 1.208, + "step": 33647 + }, + { + "epoch": 7.34, + "learning_rate": 1.329691937950623e-06, + "loss": 0.2725, + "step": 33648 + }, + { + "epoch": 7.34, + "learning_rate": 1.3295826960891414e-06, + "loss": 1.2348, + "step": 33649 + }, + { + "epoch": 7.34, + "learning_rate": 1.3294734542276602e-06, + "loss": 2.0147, + "step": 33650 + }, + { + "epoch": 7.34, + "learning_rate": 1.3293642123661788e-06, + "loss": 1.2257, + "step": 33651 + }, + { + "epoch": 7.34, + "learning_rate": 1.3292549705046976e-06, + "loss": 2.8299, + "step": 33652 + }, + { + "epoch": 7.34, + "learning_rate": 1.329145728643216e-06, + "loss": 0.4455, + "step": 33653 + }, + { + "epoch": 7.34, + "learning_rate": 1.3290364867817349e-06, + "loss": 1.3868, + "step": 33654 + }, + { + "epoch": 7.35, + "learning_rate": 1.3289272449202537e-06, + "loss": 0.8403, + "step": 33655 + }, + { + "epoch": 7.35, + "learning_rate": 1.328818003058772e-06, + "loss": 1.5373, + "step": 33656 + }, + { + "epoch": 7.35, + "learning_rate": 1.328708761197291e-06, + "loss": 0.0935, + "step": 33657 + }, + { + "epoch": 7.35, + "learning_rate": 1.3285995193358095e-06, + "loss": 1.2261, + "step": 33658 + }, + { + "epoch": 7.35, + "learning_rate": 1.3284902774743283e-06, + "loss": 0.9207, + "step": 33659 + }, + { + "epoch": 7.35, + "learning_rate": 1.3283810356128467e-06, + "loss": 1.7916, + "step": 33660 + }, + { + "epoch": 7.35, + "learning_rate": 1.3282717937513656e-06, + "loss": 1.7648, + "step": 33661 + }, + { + "epoch": 7.35, + "learning_rate": 1.3281625518898844e-06, + "loss": 0.6776, + "step": 33662 + }, + { + "epoch": 7.35, + "learning_rate": 1.328053310028403e-06, + "loss": 0.6047, + "step": 33663 + }, + { + "epoch": 7.35, + "learning_rate": 1.3279440681669218e-06, + "loss": 0.1261, + "step": 33664 + }, + { + "epoch": 7.35, + "learning_rate": 1.3278348263054402e-06, + "loss": 0.8199, + "step": 33665 + }, + { + "epoch": 7.35, + "learning_rate": 1.327725584443959e-06, + "loss": 1.0244, + "step": 33666 + }, + { + "epoch": 7.35, + "learning_rate": 1.3276163425824779e-06, + "loss": 1.4078, + "step": 33667 + }, + { + "epoch": 7.35, + "learning_rate": 1.3275071007209963e-06, + "loss": 1.8315, + "step": 33668 + }, + { + "epoch": 7.35, + "learning_rate": 1.327397858859515e-06, + "loss": 1.3386, + "step": 33669 + }, + { + "epoch": 7.35, + "learning_rate": 1.3272886169980337e-06, + "loss": 1.7883, + "step": 33670 + }, + { + "epoch": 7.35, + "learning_rate": 1.3271793751365525e-06, + "loss": 0.1614, + "step": 33671 + }, + { + "epoch": 7.35, + "learning_rate": 1.327070133275071e-06, + "loss": 0.7399, + "step": 33672 + }, + { + "epoch": 7.35, + "learning_rate": 1.3269608914135897e-06, + "loss": 0.5768, + "step": 33673 + }, + { + "epoch": 7.35, + "learning_rate": 1.3268516495521086e-06, + "loss": 1.0138, + "step": 33674 + }, + { + "epoch": 7.35, + "learning_rate": 1.3267424076906272e-06, + "loss": 0.2599, + "step": 33675 + }, + { + "epoch": 7.35, + "learning_rate": 1.326633165829146e-06, + "loss": 0.3952, + "step": 33676 + }, + { + "epoch": 7.35, + "learning_rate": 1.3265239239676644e-06, + "loss": 1.6374, + "step": 33677 + }, + { + "epoch": 7.35, + "learning_rate": 1.3264146821061832e-06, + "loss": 0.7754, + "step": 33678 + }, + { + "epoch": 7.35, + "learning_rate": 1.3263054402447018e-06, + "loss": 0.7196, + "step": 33679 + }, + { + "epoch": 7.35, + "learning_rate": 1.3261961983832207e-06, + "loss": 0.8608, + "step": 33680 + }, + { + "epoch": 7.35, + "learning_rate": 1.3260869565217393e-06, + "loss": 0.6148, + "step": 33681 + }, + { + "epoch": 7.35, + "learning_rate": 1.3259777146602579e-06, + "loss": 1.6625, + "step": 33682 + }, + { + "epoch": 7.35, + "learning_rate": 1.3258684727987767e-06, + "loss": 1.3639, + "step": 33683 + }, + { + "epoch": 7.35, + "learning_rate": 1.325759230937295e-06, + "loss": 1.4624, + "step": 33684 + }, + { + "epoch": 7.35, + "learning_rate": 1.325649989075814e-06, + "loss": 1.1633, + "step": 33685 + }, + { + "epoch": 7.35, + "learning_rate": 1.3255407472143325e-06, + "loss": 0.9736, + "step": 33686 + }, + { + "epoch": 7.35, + "learning_rate": 1.3254315053528514e-06, + "loss": 0.3556, + "step": 33687 + }, + { + "epoch": 7.35, + "learning_rate": 1.3253222634913702e-06, + "loss": 0.6494, + "step": 33688 + }, + { + "epoch": 7.35, + "learning_rate": 1.3252130216298886e-06, + "loss": 1.1737, + "step": 33689 + }, + { + "epoch": 7.35, + "learning_rate": 1.3251037797684074e-06, + "loss": 2.3172, + "step": 33690 + }, + { + "epoch": 7.35, + "learning_rate": 1.324994537906926e-06, + "loss": 0.5522, + "step": 33691 + }, + { + "epoch": 7.35, + "learning_rate": 1.3248852960454448e-06, + "loss": 0.9353, + "step": 33692 + }, + { + "epoch": 7.35, + "learning_rate": 1.3247760541839632e-06, + "loss": 2.0535, + "step": 33693 + }, + { + "epoch": 7.35, + "learning_rate": 1.324666812322482e-06, + "loss": 1.7037, + "step": 33694 + }, + { + "epoch": 7.35, + "learning_rate": 1.3245575704610009e-06, + "loss": 2.134, + "step": 33695 + }, + { + "epoch": 7.35, + "learning_rate": 1.3244483285995193e-06, + "loss": 2.0132, + "step": 33696 + }, + { + "epoch": 7.35, + "learning_rate": 1.324339086738038e-06, + "loss": 0.2115, + "step": 33697 + }, + { + "epoch": 7.35, + "learning_rate": 1.3242298448765567e-06, + "loss": 2.7547, + "step": 33698 + }, + { + "epoch": 7.35, + "learning_rate": 1.3241206030150755e-06, + "loss": 0.7021, + "step": 33699 + }, + { + "epoch": 7.35, + "learning_rate": 1.3240113611535944e-06, + "loss": 1.0016, + "step": 33700 + }, + { + "epoch": 7.36, + "learning_rate": 1.3239021192921128e-06, + "loss": 0.8531, + "step": 33701 + }, + { + "epoch": 7.36, + "learning_rate": 1.3237928774306316e-06, + "loss": 0.8246, + "step": 33702 + }, + { + "epoch": 7.36, + "learning_rate": 1.3236836355691502e-06, + "loss": 1.0723, + "step": 33703 + }, + { + "epoch": 7.36, + "learning_rate": 1.323574393707669e-06, + "loss": 1.0969, + "step": 33704 + }, + { + "epoch": 7.36, + "learning_rate": 1.3234651518461874e-06, + "loss": 0.979, + "step": 33705 + }, + { + "epoch": 7.36, + "learning_rate": 1.3233559099847062e-06, + "loss": 0.5873, + "step": 33706 + }, + { + "epoch": 7.36, + "learning_rate": 1.323246668123225e-06, + "loss": 1.5876, + "step": 33707 + }, + { + "epoch": 7.36, + "learning_rate": 1.3231374262617437e-06, + "loss": 1.1803, + "step": 33708 + }, + { + "epoch": 7.36, + "learning_rate": 1.3230281844002623e-06, + "loss": 0.7031, + "step": 33709 + }, + { + "epoch": 7.36, + "learning_rate": 1.3229189425387809e-06, + "loss": 2.2167, + "step": 33710 + }, + { + "epoch": 7.36, + "learning_rate": 1.3228097006772997e-06, + "loss": 0.9801, + "step": 33711 + }, + { + "epoch": 7.36, + "learning_rate": 1.3227004588158181e-06, + "loss": 0.7072, + "step": 33712 + }, + { + "epoch": 7.36, + "learning_rate": 1.322591216954337e-06, + "loss": 1.3543, + "step": 33713 + }, + { + "epoch": 7.36, + "learning_rate": 1.3224819750928558e-06, + "loss": 0.1559, + "step": 33714 + }, + { + "epoch": 7.36, + "learning_rate": 1.3223727332313744e-06, + "loss": 0.9709, + "step": 33715 + }, + { + "epoch": 7.36, + "learning_rate": 1.3222634913698932e-06, + "loss": 0.7549, + "step": 33716 + }, + { + "epoch": 7.36, + "learning_rate": 1.3221542495084116e-06, + "loss": 1.7125, + "step": 33717 + }, + { + "epoch": 7.36, + "learning_rate": 1.3220450076469304e-06, + "loss": 2.1018, + "step": 33718 + }, + { + "epoch": 7.36, + "learning_rate": 1.321935765785449e-06, + "loss": 0.933, + "step": 33719 + }, + { + "epoch": 7.36, + "learning_rate": 1.3218265239239678e-06, + "loss": 1.1837, + "step": 33720 + }, + { + "epoch": 7.36, + "learning_rate": 1.3217172820624865e-06, + "loss": 2.021, + "step": 33721 + }, + { + "epoch": 7.36, + "learning_rate": 1.321608040201005e-06, + "loss": 0.2267, + "step": 33722 + }, + { + "epoch": 7.36, + "learning_rate": 1.321498798339524e-06, + "loss": 1.8091, + "step": 33723 + }, + { + "epoch": 7.36, + "learning_rate": 1.3213895564780425e-06, + "loss": 0.4943, + "step": 33724 + }, + { + "epoch": 7.36, + "learning_rate": 1.3212803146165611e-06, + "loss": 0.1545, + "step": 33725 + }, + { + "epoch": 7.36, + "learning_rate": 1.3211710727550797e-06, + "loss": 1.9999, + "step": 33726 + }, + { + "epoch": 7.36, + "learning_rate": 1.3210618308935986e-06, + "loss": 1.2987, + "step": 33727 + }, + { + "epoch": 7.36, + "learning_rate": 1.3209525890321174e-06, + "loss": 0.7247, + "step": 33728 + }, + { + "epoch": 7.36, + "learning_rate": 1.3208433471706358e-06, + "loss": 0.5315, + "step": 33729 + }, + { + "epoch": 7.36, + "learning_rate": 1.3207341053091546e-06, + "loss": 1.2517, + "step": 33730 + }, + { + "epoch": 7.36, + "learning_rate": 1.3206248634476732e-06, + "loss": 1.2082, + "step": 33731 + }, + { + "epoch": 7.36, + "learning_rate": 1.320515621586192e-06, + "loss": 0.5307, + "step": 33732 + }, + { + "epoch": 7.36, + "learning_rate": 1.3204063797247109e-06, + "loss": 1.549, + "step": 33733 + }, + { + "epoch": 7.36, + "learning_rate": 1.3202971378632293e-06, + "loss": 1.8768, + "step": 33734 + }, + { + "epoch": 7.36, + "learning_rate": 1.320187896001748e-06, + "loss": 1.2157, + "step": 33735 + }, + { + "epoch": 7.36, + "learning_rate": 1.3200786541402667e-06, + "loss": 0.751, + "step": 33736 + }, + { + "epoch": 7.36, + "learning_rate": 1.3199694122787853e-06, + "loss": 1.7722, + "step": 33737 + }, + { + "epoch": 7.36, + "learning_rate": 1.319860170417304e-06, + "loss": 1.4746, + "step": 33738 + }, + { + "epoch": 7.36, + "learning_rate": 1.3197509285558227e-06, + "loss": 2.2839, + "step": 33739 + }, + { + "epoch": 7.36, + "learning_rate": 1.3196416866943416e-06, + "loss": 2.0662, + "step": 33740 + }, + { + "epoch": 7.36, + "learning_rate": 1.31953244483286e-06, + "loss": 0.2322, + "step": 33741 + }, + { + "epoch": 7.36, + "learning_rate": 1.3194232029713788e-06, + "loss": 1.0303, + "step": 33742 + }, + { + "epoch": 7.36, + "learning_rate": 1.3193139611098974e-06, + "loss": 1.3538, + "step": 33743 + }, + { + "epoch": 7.36, + "learning_rate": 1.3192047192484162e-06, + "loss": 1.0791, + "step": 33744 + }, + { + "epoch": 7.36, + "learning_rate": 1.3190954773869346e-06, + "loss": 0.9127, + "step": 33745 + }, + { + "epoch": 7.36, + "learning_rate": 1.3189862355254534e-06, + "loss": 0.262, + "step": 33746 + }, + { + "epoch": 7.37, + "learning_rate": 1.3188769936639723e-06, + "loss": 0.9335, + "step": 33747 + }, + { + "epoch": 7.37, + "learning_rate": 1.3187677518024909e-06, + "loss": 2.2047, + "step": 33748 + }, + { + "epoch": 7.37, + "learning_rate": 1.3186585099410095e-06, + "loss": 1.2558, + "step": 33749 + }, + { + "epoch": 7.37, + "learning_rate": 1.318549268079528e-06, + "loss": 0.6063, + "step": 33750 + }, + { + "epoch": 7.37, + "learning_rate": 1.318440026218047e-06, + "loss": 1.3393, + "step": 33751 + }, + { + "epoch": 7.37, + "learning_rate": 1.3183307843565655e-06, + "loss": 0.5288, + "step": 33752 + }, + { + "epoch": 7.37, + "learning_rate": 1.3182215424950841e-06, + "loss": 2.6004, + "step": 33753 + }, + { + "epoch": 7.37, + "learning_rate": 1.318112300633603e-06, + "loss": 1.6729, + "step": 33754 + }, + { + "epoch": 7.37, + "learning_rate": 1.3180030587721216e-06, + "loss": 0.9793, + "step": 33755 + }, + { + "epoch": 7.37, + "learning_rate": 1.3178938169106404e-06, + "loss": 0.1334, + "step": 33756 + }, + { + "epoch": 7.37, + "learning_rate": 1.3177845750491588e-06, + "loss": 1.1833, + "step": 33757 + }, + { + "epoch": 7.37, + "learning_rate": 1.3176753331876776e-06, + "loss": 1.9539, + "step": 33758 + }, + { + "epoch": 7.37, + "learning_rate": 1.3175660913261962e-06, + "loss": 1.7332, + "step": 33759 + }, + { + "epoch": 7.37, + "learning_rate": 1.317456849464715e-06, + "loss": 2.7168, + "step": 33760 + }, + { + "epoch": 7.37, + "learning_rate": 1.3173476076032339e-06, + "loss": 1.0908, + "step": 33761 + }, + { + "epoch": 7.37, + "learning_rate": 1.3172383657417523e-06, + "loss": 0.4363, + "step": 33762 + }, + { + "epoch": 7.37, + "learning_rate": 1.317129123880271e-06, + "loss": 2.0423, + "step": 33763 + }, + { + "epoch": 7.37, + "learning_rate": 1.3170198820187897e-06, + "loss": 1.1445, + "step": 33764 + }, + { + "epoch": 7.37, + "learning_rate": 1.3169106401573083e-06, + "loss": 0.5842, + "step": 33765 + }, + { + "epoch": 7.37, + "learning_rate": 1.3168013982958271e-06, + "loss": 0.3999, + "step": 33766 + }, + { + "epoch": 7.37, + "learning_rate": 1.3166921564343457e-06, + "loss": 0.3063, + "step": 33767 + }, + { + "epoch": 7.37, + "learning_rate": 1.3165829145728646e-06, + "loss": 2.037, + "step": 33768 + }, + { + "epoch": 7.37, + "learning_rate": 1.316473672711383e-06, + "loss": 0.2689, + "step": 33769 + }, + { + "epoch": 7.37, + "learning_rate": 1.3163644308499018e-06, + "loss": 0.8959, + "step": 33770 + }, + { + "epoch": 7.37, + "learning_rate": 1.3162551889884204e-06, + "loss": 1.0398, + "step": 33771 + }, + { + "epoch": 7.37, + "learning_rate": 1.3161459471269392e-06, + "loss": 3.1713, + "step": 33772 + }, + { + "epoch": 7.37, + "learning_rate": 1.316036705265458e-06, + "loss": 2.2525, + "step": 33773 + }, + { + "epoch": 7.37, + "learning_rate": 1.3159274634039764e-06, + "loss": 2.7423, + "step": 33774 + }, + { + "epoch": 7.37, + "learning_rate": 1.3158182215424953e-06, + "loss": 0.7743, + "step": 33775 + }, + { + "epoch": 7.37, + "learning_rate": 1.3157089796810139e-06, + "loss": 1.1252, + "step": 33776 + }, + { + "epoch": 7.37, + "learning_rate": 1.3155997378195327e-06, + "loss": 2.2196, + "step": 33777 + }, + { + "epoch": 7.37, + "learning_rate": 1.315490495958051e-06, + "loss": 1.3658, + "step": 33778 + }, + { + "epoch": 7.37, + "learning_rate": 1.31538125409657e-06, + "loss": 2.2705, + "step": 33779 + }, + { + "epoch": 7.37, + "learning_rate": 1.3152720122350887e-06, + "loss": 2.1384, + "step": 33780 + }, + { + "epoch": 7.37, + "learning_rate": 1.3151627703736071e-06, + "loss": 0.5793, + "step": 33781 + }, + { + "epoch": 7.37, + "learning_rate": 1.315053528512126e-06, + "loss": 0.1789, + "step": 33782 + }, + { + "epoch": 7.37, + "learning_rate": 1.3149442866506446e-06, + "loss": 1.0648, + "step": 33783 + }, + { + "epoch": 7.37, + "learning_rate": 1.3148350447891634e-06, + "loss": 2.2843, + "step": 33784 + }, + { + "epoch": 7.37, + "learning_rate": 1.3147258029276818e-06, + "loss": 1.6702, + "step": 33785 + }, + { + "epoch": 7.37, + "learning_rate": 1.3146165610662006e-06, + "loss": 2.1016, + "step": 33786 + }, + { + "epoch": 7.37, + "learning_rate": 1.3145073192047194e-06, + "loss": 0.3177, + "step": 33787 + }, + { + "epoch": 7.37, + "learning_rate": 1.314398077343238e-06, + "loss": 0.7294, + "step": 33788 + }, + { + "epoch": 7.37, + "learning_rate": 1.3142888354817569e-06, + "loss": 0.9044, + "step": 33789 + }, + { + "epoch": 7.37, + "learning_rate": 1.3141795936202753e-06, + "loss": 1.658, + "step": 33790 + }, + { + "epoch": 7.37, + "learning_rate": 1.314070351758794e-06, + "loss": 1.419, + "step": 33791 + }, + { + "epoch": 7.37, + "learning_rate": 1.3139611098973127e-06, + "loss": 1.4249, + "step": 33792 + }, + { + "epoch": 7.38, + "learning_rate": 1.3138518680358313e-06, + "loss": 0.5562, + "step": 33793 + }, + { + "epoch": 7.38, + "learning_rate": 1.3137426261743501e-06, + "loss": 1.9268, + "step": 33794 + }, + { + "epoch": 7.38, + "learning_rate": 1.3136333843128688e-06, + "loss": 0.5776, + "step": 33795 + }, + { + "epoch": 7.38, + "learning_rate": 1.3135241424513876e-06, + "loss": 1.8276, + "step": 33796 + }, + { + "epoch": 7.38, + "learning_rate": 1.313414900589906e-06, + "loss": 1.178, + "step": 33797 + }, + { + "epoch": 7.38, + "learning_rate": 1.3133056587284248e-06, + "loss": 0.5477, + "step": 33798 + }, + { + "epoch": 7.38, + "learning_rate": 1.3131964168669436e-06, + "loss": 1.6117, + "step": 33799 + }, + { + "epoch": 7.38, + "learning_rate": 1.3130871750054622e-06, + "loss": 3.4335, + "step": 33800 + }, + { + "epoch": 7.38, + "learning_rate": 1.312977933143981e-06, + "loss": 0.585, + "step": 33801 + }, + { + "epoch": 7.38, + "learning_rate": 1.3128686912824995e-06, + "loss": 0.9376, + "step": 33802 + }, + { + "epoch": 7.38, + "learning_rate": 1.3127594494210183e-06, + "loss": 0.2816, + "step": 33803 + }, + { + "epoch": 7.38, + "learning_rate": 1.3126502075595369e-06, + "loss": 1.5917, + "step": 33804 + }, + { + "epoch": 7.38, + "learning_rate": 1.3125409656980557e-06, + "loss": 0.9094, + "step": 33805 + }, + { + "epoch": 7.38, + "learning_rate": 1.3124317238365743e-06, + "loss": 1.2974, + "step": 33806 + }, + { + "epoch": 7.38, + "learning_rate": 1.312322481975093e-06, + "loss": 1.5558, + "step": 33807 + }, + { + "epoch": 7.38, + "learning_rate": 1.3122132401136118e-06, + "loss": 1.0852, + "step": 33808 + }, + { + "epoch": 7.38, + "learning_rate": 1.3121039982521302e-06, + "loss": 0.7079, + "step": 33809 + }, + { + "epoch": 7.38, + "learning_rate": 1.311994756390649e-06, + "loss": 0.6434, + "step": 33810 + }, + { + "epoch": 7.38, + "learning_rate": 1.3118855145291676e-06, + "loss": 2.9982, + "step": 33811 + }, + { + "epoch": 7.38, + "learning_rate": 1.3117762726676864e-06, + "loss": 1.0074, + "step": 33812 + }, + { + "epoch": 7.38, + "learning_rate": 1.3116670308062052e-06, + "loss": 0.5192, + "step": 33813 + }, + { + "epoch": 7.38, + "learning_rate": 1.3115577889447236e-06, + "loss": 1.3235, + "step": 33814 + }, + { + "epoch": 7.38, + "learning_rate": 1.3114485470832425e-06, + "loss": 0.3555, + "step": 33815 + }, + { + "epoch": 7.38, + "learning_rate": 1.311339305221761e-06, + "loss": 2.0711, + "step": 33816 + }, + { + "epoch": 7.38, + "learning_rate": 1.3112300633602799e-06, + "loss": 0.4962, + "step": 33817 + }, + { + "epoch": 7.38, + "learning_rate": 1.3111208214987983e-06, + "loss": 1.2509, + "step": 33818 + }, + { + "epoch": 7.38, + "learning_rate": 1.3110115796373171e-06, + "loss": 1.177, + "step": 33819 + }, + { + "epoch": 7.38, + "learning_rate": 1.310902337775836e-06, + "loss": 1.6249, + "step": 33820 + }, + { + "epoch": 7.38, + "learning_rate": 1.3107930959143543e-06, + "loss": 1.7498, + "step": 33821 + }, + { + "epoch": 7.38, + "learning_rate": 1.3106838540528732e-06, + "loss": 0.831, + "step": 33822 + }, + { + "epoch": 7.38, + "learning_rate": 1.3105746121913918e-06, + "loss": 0.9101, + "step": 33823 + }, + { + "epoch": 7.38, + "learning_rate": 1.3104653703299106e-06, + "loss": 2.0839, + "step": 33824 + }, + { + "epoch": 7.38, + "learning_rate": 1.310356128468429e-06, + "loss": 1.052, + "step": 33825 + }, + { + "epoch": 7.38, + "learning_rate": 1.3102468866069478e-06, + "loss": 1.1501, + "step": 33826 + }, + { + "epoch": 7.38, + "learning_rate": 1.3101376447454666e-06, + "loss": 1.3076, + "step": 33827 + }, + { + "epoch": 7.38, + "learning_rate": 1.3100284028839852e-06, + "loss": 0.5292, + "step": 33828 + }, + { + "epoch": 7.38, + "learning_rate": 1.309919161022504e-06, + "loss": 1.1447, + "step": 33829 + }, + { + "epoch": 7.38, + "learning_rate": 1.3098099191610225e-06, + "loss": 0.4022, + "step": 33830 + }, + { + "epoch": 7.38, + "learning_rate": 1.3097006772995413e-06, + "loss": 0.4428, + "step": 33831 + }, + { + "epoch": 7.38, + "learning_rate": 1.3095914354380601e-06, + "loss": 0.1788, + "step": 33832 + }, + { + "epoch": 7.38, + "learning_rate": 1.3094821935765787e-06, + "loss": 1.659, + "step": 33833 + }, + { + "epoch": 7.38, + "learning_rate": 1.3093729517150973e-06, + "loss": 1.8709, + "step": 33834 + }, + { + "epoch": 7.38, + "learning_rate": 1.309263709853616e-06, + "loss": 0.4662, + "step": 33835 + }, + { + "epoch": 7.38, + "learning_rate": 1.3091544679921348e-06, + "loss": 0.3135, + "step": 33836 + }, + { + "epoch": 7.38, + "learning_rate": 1.3090452261306532e-06, + "loss": 1.7219, + "step": 33837 + }, + { + "epoch": 7.38, + "learning_rate": 1.308935984269172e-06, + "loss": 1.2352, + "step": 33838 + }, + { + "epoch": 7.39, + "learning_rate": 1.3088267424076908e-06, + "loss": 1.2488, + "step": 33839 + }, + { + "epoch": 7.39, + "learning_rate": 1.3087175005462094e-06, + "loss": 1.1276, + "step": 33840 + }, + { + "epoch": 7.39, + "learning_rate": 1.3086082586847282e-06, + "loss": 0.8522, + "step": 33841 + }, + { + "epoch": 7.39, + "learning_rate": 1.3084990168232466e-06, + "loss": 0.917, + "step": 33842 + }, + { + "epoch": 7.39, + "learning_rate": 1.3083897749617655e-06, + "loss": 1.284, + "step": 33843 + }, + { + "epoch": 7.39, + "learning_rate": 1.308280533100284e-06, + "loss": 0.7354, + "step": 33844 + }, + { + "epoch": 7.39, + "learning_rate": 1.308171291238803e-06, + "loss": 2.5132, + "step": 33845 + }, + { + "epoch": 7.39, + "learning_rate": 1.3080620493773215e-06, + "loss": 0.4664, + "step": 33846 + }, + { + "epoch": 7.39, + "learning_rate": 1.3079528075158401e-06, + "loss": 0.5016, + "step": 33847 + }, + { + "epoch": 7.39, + "learning_rate": 1.307843565654359e-06, + "loss": 0.616, + "step": 33848 + }, + { + "epoch": 7.39, + "learning_rate": 1.3077343237928776e-06, + "loss": 1.4492, + "step": 33849 + }, + { + "epoch": 7.39, + "learning_rate": 1.3076250819313962e-06, + "loss": 2.1591, + "step": 33850 + }, + { + "epoch": 7.39, + "learning_rate": 1.3075158400699148e-06, + "loss": 0.8428, + "step": 33851 + }, + { + "epoch": 7.39, + "learning_rate": 1.3074065982084336e-06, + "loss": 2.4334, + "step": 33852 + }, + { + "epoch": 7.39, + "learning_rate": 1.3072973563469524e-06, + "loss": 1.76, + "step": 33853 + }, + { + "epoch": 7.39, + "learning_rate": 1.3071881144854708e-06, + "loss": 1.3953, + "step": 33854 + }, + { + "epoch": 7.39, + "learning_rate": 1.3070788726239897e-06, + "loss": 0.3921, + "step": 33855 + }, + { + "epoch": 7.39, + "learning_rate": 1.3069696307625083e-06, + "loss": 0.3392, + "step": 33856 + }, + { + "epoch": 7.39, + "learning_rate": 1.306860388901027e-06, + "loss": 0.2037, + "step": 33857 + }, + { + "epoch": 7.39, + "learning_rate": 1.3067511470395455e-06, + "loss": 2.1652, + "step": 33858 + }, + { + "epoch": 7.39, + "learning_rate": 1.3066419051780643e-06, + "loss": 0.3724, + "step": 33859 + }, + { + "epoch": 7.39, + "learning_rate": 1.3065326633165831e-06, + "loss": 1.1338, + "step": 33860 + }, + { + "epoch": 7.39, + "learning_rate": 1.3064234214551017e-06, + "loss": 0.9344, + "step": 33861 + }, + { + "epoch": 7.39, + "learning_rate": 1.3063141795936204e-06, + "loss": 2.0468, + "step": 33862 + }, + { + "epoch": 7.39, + "learning_rate": 1.306204937732139e-06, + "loss": 1.0886, + "step": 33863 + }, + { + "epoch": 7.39, + "learning_rate": 1.3060956958706578e-06, + "loss": 1.0701, + "step": 33864 + }, + { + "epoch": 7.39, + "learning_rate": 1.3059864540091762e-06, + "loss": 1.6836, + "step": 33865 + }, + { + "epoch": 7.39, + "learning_rate": 1.305877212147695e-06, + "loss": 0.8284, + "step": 33866 + }, + { + "epoch": 7.39, + "learning_rate": 1.3057679702862138e-06, + "loss": 0.5709, + "step": 33867 + }, + { + "epoch": 7.39, + "learning_rate": 1.3056587284247324e-06, + "loss": 0.9445, + "step": 33868 + }, + { + "epoch": 7.39, + "learning_rate": 1.3055494865632513e-06, + "loss": 1.5696, + "step": 33869 + }, + { + "epoch": 7.39, + "learning_rate": 1.3054402447017697e-06, + "loss": 0.0953, + "step": 33870 + }, + { + "epoch": 7.39, + "learning_rate": 1.3053310028402885e-06, + "loss": 1.0858, + "step": 33871 + }, + { + "epoch": 7.39, + "learning_rate": 1.3052217609788073e-06, + "loss": 0.7833, + "step": 33872 + }, + { + "epoch": 7.39, + "learning_rate": 1.305112519117326e-06, + "loss": 1.021, + "step": 33873 + }, + { + "epoch": 7.39, + "learning_rate": 1.3050032772558447e-06, + "loss": 1.8993, + "step": 33874 + }, + { + "epoch": 7.39, + "learning_rate": 1.3048940353943631e-06, + "loss": 1.303, + "step": 33875 + }, + { + "epoch": 7.39, + "learning_rate": 1.304784793532882e-06, + "loss": 2.0861, + "step": 33876 + }, + { + "epoch": 7.39, + "learning_rate": 1.3046755516714006e-06, + "loss": 0.811, + "step": 33877 + }, + { + "epoch": 7.39, + "learning_rate": 1.3045663098099192e-06, + "loss": 1.2181, + "step": 33878 + }, + { + "epoch": 7.39, + "learning_rate": 1.304457067948438e-06, + "loss": 0.1631, + "step": 33879 + }, + { + "epoch": 7.39, + "learning_rate": 1.3043478260869566e-06, + "loss": 1.2191, + "step": 33880 + }, + { + "epoch": 7.39, + "learning_rate": 1.3042385842254754e-06, + "loss": 0.744, + "step": 33881 + }, + { + "epoch": 7.39, + "learning_rate": 1.3041293423639938e-06, + "loss": 4.0641, + "step": 33882 + }, + { + "epoch": 7.39, + "learning_rate": 1.3040201005025127e-06, + "loss": 2.0158, + "step": 33883 + }, + { + "epoch": 7.4, + "learning_rate": 1.3039108586410313e-06, + "loss": 0.4676, + "step": 33884 + }, + { + "epoch": 7.4, + "learning_rate": 1.30380161677955e-06, + "loss": 1.6838, + "step": 33885 + }, + { + "epoch": 7.4, + "learning_rate": 1.303692374918069e-06, + "loss": 1.3455, + "step": 33886 + }, + { + "epoch": 7.4, + "learning_rate": 1.3035831330565873e-06, + "loss": 1.333, + "step": 33887 + }, + { + "epoch": 7.4, + "learning_rate": 1.3034738911951061e-06, + "loss": 0.6543, + "step": 33888 + }, + { + "epoch": 7.4, + "learning_rate": 1.3033646493336248e-06, + "loss": 0.5093, + "step": 33889 + }, + { + "epoch": 7.4, + "learning_rate": 1.3032554074721434e-06, + "loss": 1.2176, + "step": 33890 + }, + { + "epoch": 7.4, + "learning_rate": 1.303146165610662e-06, + "loss": 1.7628, + "step": 33891 + }, + { + "epoch": 7.4, + "learning_rate": 1.3030369237491808e-06, + "loss": 1.5264, + "step": 33892 + }, + { + "epoch": 7.4, + "learning_rate": 1.3029276818876996e-06, + "loss": 0.1954, + "step": 33893 + }, + { + "epoch": 7.4, + "learning_rate": 1.302818440026218e-06, + "loss": 2.7871, + "step": 33894 + }, + { + "epoch": 7.4, + "learning_rate": 1.3027091981647368e-06, + "loss": 1.4965, + "step": 33895 + }, + { + "epoch": 7.4, + "learning_rate": 1.3025999563032555e-06, + "loss": 0.869, + "step": 33896 + }, + { + "epoch": 7.4, + "learning_rate": 1.3024907144417743e-06, + "loss": 2.3967, + "step": 33897 + }, + { + "epoch": 7.4, + "learning_rate": 1.3023814725802927e-06, + "loss": 0.1202, + "step": 33898 + }, + { + "epoch": 7.4, + "learning_rate": 1.3022722307188115e-06, + "loss": 1.883, + "step": 33899 + }, + { + "epoch": 7.4, + "learning_rate": 1.3021629888573303e-06, + "loss": 1.7798, + "step": 33900 + }, + { + "epoch": 7.4, + "learning_rate": 1.302053746995849e-06, + "loss": 1.3176, + "step": 33901 + }, + { + "epoch": 7.4, + "learning_rate": 1.3019445051343678e-06, + "loss": 1.1627, + "step": 33902 + }, + { + "epoch": 7.4, + "learning_rate": 1.3018352632728862e-06, + "loss": 2.1698, + "step": 33903 + }, + { + "epoch": 7.4, + "learning_rate": 1.301726021411405e-06, + "loss": 2.3538, + "step": 33904 + }, + { + "epoch": 7.4, + "learning_rate": 1.3016167795499238e-06, + "loss": 1.0406, + "step": 33905 + }, + { + "epoch": 7.4, + "learning_rate": 1.3015075376884422e-06, + "loss": 0.4905, + "step": 33906 + }, + { + "epoch": 7.4, + "learning_rate": 1.301398295826961e-06, + "loss": 2.2412, + "step": 33907 + }, + { + "epoch": 7.4, + "learning_rate": 1.3012890539654796e-06, + "loss": 1.7986, + "step": 33908 + }, + { + "epoch": 7.4, + "learning_rate": 1.3011798121039985e-06, + "loss": 0.9768, + "step": 33909 + }, + { + "epoch": 7.4, + "learning_rate": 1.3010705702425169e-06, + "loss": 0.9852, + "step": 33910 + }, + { + "epoch": 7.4, + "learning_rate": 1.3009613283810357e-06, + "loss": 1.2397, + "step": 33911 + }, + { + "epoch": 7.4, + "learning_rate": 1.3008520865195545e-06, + "loss": 0.6758, + "step": 33912 + }, + { + "epoch": 7.4, + "learning_rate": 1.3007428446580731e-06, + "loss": 0.8179, + "step": 33913 + }, + { + "epoch": 7.4, + "learning_rate": 1.300633602796592e-06, + "loss": 0.4317, + "step": 33914 + }, + { + "epoch": 7.4, + "learning_rate": 1.3005243609351103e-06, + "loss": 0.7135, + "step": 33915 + }, + { + "epoch": 7.4, + "learning_rate": 1.3004151190736292e-06, + "loss": 0.294, + "step": 33916 + }, + { + "epoch": 7.4, + "learning_rate": 1.3003058772121478e-06, + "loss": 1.1836, + "step": 33917 + }, + { + "epoch": 7.4, + "learning_rate": 1.3001966353506664e-06, + "loss": 0.8053, + "step": 33918 + }, + { + "epoch": 7.4, + "learning_rate": 1.3000873934891852e-06, + "loss": 1.073, + "step": 33919 + }, + { + "epoch": 7.4, + "learning_rate": 1.2999781516277038e-06, + "loss": 0.4657, + "step": 33920 + }, + { + "epoch": 7.4, + "learning_rate": 1.2998689097662226e-06, + "loss": 0.9875, + "step": 33921 + }, + { + "epoch": 7.4, + "learning_rate": 1.299759667904741e-06, + "loss": 0.7619, + "step": 33922 + }, + { + "epoch": 7.4, + "learning_rate": 1.2996504260432599e-06, + "loss": 2.2005, + "step": 33923 + }, + { + "epoch": 7.4, + "learning_rate": 1.2995411841817785e-06, + "loss": 2.1891, + "step": 33924 + }, + { + "epoch": 7.4, + "learning_rate": 1.2994319423202973e-06, + "loss": 1.505, + "step": 33925 + }, + { + "epoch": 7.4, + "learning_rate": 1.2993227004588161e-06, + "loss": 0.9656, + "step": 33926 + }, + { + "epoch": 7.4, + "learning_rate": 1.2992134585973345e-06, + "loss": 1.0654, + "step": 33927 + }, + { + "epoch": 7.4, + "learning_rate": 1.2991042167358533e-06, + "loss": 0.9444, + "step": 33928 + }, + { + "epoch": 7.4, + "learning_rate": 1.298994974874372e-06, + "loss": 0.6464, + "step": 33929 + }, + { + "epoch": 7.41, + "learning_rate": 1.2988857330128908e-06, + "loss": 1.9894, + "step": 33930 + }, + { + "epoch": 7.41, + "learning_rate": 1.2987764911514092e-06, + "loss": 2.2342, + "step": 33931 + }, + { + "epoch": 7.41, + "learning_rate": 1.298667249289928e-06, + "loss": 1.285, + "step": 33932 + }, + { + "epoch": 7.41, + "learning_rate": 1.2985580074284468e-06, + "loss": 0.1928, + "step": 33933 + }, + { + "epoch": 7.41, + "learning_rate": 1.2984487655669652e-06, + "loss": 0.958, + "step": 33934 + }, + { + "epoch": 7.41, + "learning_rate": 1.298339523705484e-06, + "loss": 0.6457, + "step": 33935 + }, + { + "epoch": 7.41, + "learning_rate": 1.2982302818440026e-06, + "loss": 0.976, + "step": 33936 + }, + { + "epoch": 7.41, + "learning_rate": 1.2981210399825215e-06, + "loss": 0.5759, + "step": 33937 + }, + { + "epoch": 7.41, + "learning_rate": 1.2980117981210403e-06, + "loss": 2.6931, + "step": 33938 + }, + { + "epoch": 7.41, + "learning_rate": 1.2979025562595587e-06, + "loss": 3.5366, + "step": 33939 + }, + { + "epoch": 7.41, + "learning_rate": 1.2977933143980775e-06, + "loss": 1.3232, + "step": 33940 + }, + { + "epoch": 7.41, + "learning_rate": 1.2976840725365961e-06, + "loss": 0.2629, + "step": 33941 + }, + { + "epoch": 7.41, + "learning_rate": 1.297574830675115e-06, + "loss": 0.6852, + "step": 33942 + }, + { + "epoch": 7.41, + "learning_rate": 1.2974655888136333e-06, + "loss": 1.5928, + "step": 33943 + }, + { + "epoch": 7.41, + "learning_rate": 1.2973563469521522e-06, + "loss": 1.3274, + "step": 33944 + }, + { + "epoch": 7.41, + "learning_rate": 1.297247105090671e-06, + "loss": 1.6376, + "step": 33945 + }, + { + "epoch": 7.41, + "learning_rate": 1.2971378632291896e-06, + "loss": 1.4058, + "step": 33946 + }, + { + "epoch": 7.41, + "learning_rate": 1.2970286213677082e-06, + "loss": 0.9759, + "step": 33947 + }, + { + "epoch": 7.41, + "learning_rate": 1.2969193795062268e-06, + "loss": 1.5381, + "step": 33948 + }, + { + "epoch": 7.41, + "learning_rate": 1.2968101376447456e-06, + "loss": 0.7976, + "step": 33949 + }, + { + "epoch": 7.41, + "learning_rate": 1.296700895783264e-06, + "loss": 1.0977, + "step": 33950 + }, + { + "epoch": 7.41, + "learning_rate": 1.2965916539217829e-06, + "loss": 1.6714, + "step": 33951 + }, + { + "epoch": 7.41, + "learning_rate": 1.2964824120603017e-06, + "loss": 2.5222, + "step": 33952 + }, + { + "epoch": 7.41, + "learning_rate": 1.2963731701988203e-06, + "loss": 0.9293, + "step": 33953 + }, + { + "epoch": 7.41, + "learning_rate": 1.2962639283373391e-06, + "loss": 1.6989, + "step": 33954 + }, + { + "epoch": 7.41, + "learning_rate": 1.2961546864758575e-06, + "loss": 2.1475, + "step": 33955 + }, + { + "epoch": 7.41, + "learning_rate": 1.2960454446143763e-06, + "loss": 0.3528, + "step": 33956 + }, + { + "epoch": 7.41, + "learning_rate": 1.295936202752895e-06, + "loss": 0.9763, + "step": 33957 + }, + { + "epoch": 7.41, + "learning_rate": 1.2958269608914138e-06, + "loss": 0.5994, + "step": 33958 + }, + { + "epoch": 7.41, + "learning_rate": 1.2957177190299324e-06, + "loss": 0.3011, + "step": 33959 + }, + { + "epoch": 7.41, + "learning_rate": 1.295608477168451e-06, + "loss": 1.9259, + "step": 33960 + }, + { + "epoch": 7.41, + "learning_rate": 1.2954992353069698e-06, + "loss": 2.5474, + "step": 33961 + }, + { + "epoch": 7.41, + "learning_rate": 1.2953899934454882e-06, + "loss": 1.1939, + "step": 33962 + }, + { + "epoch": 7.41, + "learning_rate": 1.295280751584007e-06, + "loss": 0.1266, + "step": 33963 + }, + { + "epoch": 7.41, + "learning_rate": 1.2951715097225257e-06, + "loss": 0.1245, + "step": 33964 + }, + { + "epoch": 7.41, + "learning_rate": 1.2950622678610445e-06, + "loss": 0.1573, + "step": 33965 + }, + { + "epoch": 7.41, + "learning_rate": 1.2949530259995633e-06, + "loss": 2.0278, + "step": 33966 + }, + { + "epoch": 7.41, + "learning_rate": 1.2948437841380817e-06, + "loss": 0.9349, + "step": 33967 + }, + { + "epoch": 7.41, + "learning_rate": 1.2947345422766005e-06, + "loss": 2.6978, + "step": 33968 + }, + { + "epoch": 7.41, + "learning_rate": 1.2946253004151191e-06, + "loss": 1.8657, + "step": 33969 + }, + { + "epoch": 7.41, + "learning_rate": 1.294516058553638e-06, + "loss": 2.6213, + "step": 33970 + }, + { + "epoch": 7.41, + "learning_rate": 1.2944068166921566e-06, + "loss": 0.9881, + "step": 33971 + }, + { + "epoch": 7.41, + "learning_rate": 1.2942975748306752e-06, + "loss": 0.1858, + "step": 33972 + }, + { + "epoch": 7.41, + "learning_rate": 1.294188332969194e-06, + "loss": 0.4947, + "step": 33973 + }, + { + "epoch": 7.41, + "learning_rate": 1.2940790911077126e-06, + "loss": 0.7557, + "step": 33974 + }, + { + "epoch": 7.41, + "learning_rate": 1.2939698492462312e-06, + "loss": 1.5724, + "step": 33975 + }, + { + "epoch": 7.42, + "learning_rate": 1.2938606073847498e-06, + "loss": 0.6559, + "step": 33976 + }, + { + "epoch": 7.42, + "learning_rate": 1.2937513655232687e-06, + "loss": 0.781, + "step": 33977 + }, + { + "epoch": 7.42, + "learning_rate": 1.2936421236617875e-06, + "loss": 1.2281, + "step": 33978 + }, + { + "epoch": 7.42, + "learning_rate": 1.2935328818003059e-06, + "loss": 3.3653, + "step": 33979 + }, + { + "epoch": 7.42, + "learning_rate": 1.2934236399388247e-06, + "loss": 0.6658, + "step": 33980 + }, + { + "epoch": 7.42, + "learning_rate": 1.2933143980773433e-06, + "loss": 0.7409, + "step": 33981 + }, + { + "epoch": 7.42, + "learning_rate": 1.2932051562158621e-06, + "loss": 2.0781, + "step": 33982 + }, + { + "epoch": 7.42, + "learning_rate": 1.2930959143543805e-06, + "loss": 0.8895, + "step": 33983 + }, + { + "epoch": 7.42, + "learning_rate": 1.2929866724928994e-06, + "loss": 1.7399, + "step": 33984 + }, + { + "epoch": 7.42, + "learning_rate": 1.2928774306314182e-06, + "loss": 0.6355, + "step": 33985 + }, + { + "epoch": 7.42, + "learning_rate": 1.2927681887699368e-06, + "loss": 0.9997, + "step": 33986 + }, + { + "epoch": 7.42, + "learning_rate": 1.2926589469084554e-06, + "loss": 0.8355, + "step": 33987 + }, + { + "epoch": 7.42, + "learning_rate": 1.292549705046974e-06, + "loss": 1.1569, + "step": 33988 + }, + { + "epoch": 7.42, + "learning_rate": 1.2924404631854928e-06, + "loss": 1.5255, + "step": 33989 + }, + { + "epoch": 7.42, + "learning_rate": 1.2923312213240112e-06, + "loss": 1.3689, + "step": 33990 + }, + { + "epoch": 7.42, + "learning_rate": 1.29222197946253e-06, + "loss": 1.9128, + "step": 33991 + }, + { + "epoch": 7.42, + "learning_rate": 1.2921127376010489e-06, + "loss": 1.0793, + "step": 33992 + }, + { + "epoch": 7.42, + "learning_rate": 1.2920034957395675e-06, + "loss": 0.8143, + "step": 33993 + }, + { + "epoch": 7.42, + "learning_rate": 1.2918942538780863e-06, + "loss": 1.6511, + "step": 33994 + }, + { + "epoch": 7.42, + "learning_rate": 1.2917850120166047e-06, + "loss": 2.7378, + "step": 33995 + }, + { + "epoch": 7.42, + "learning_rate": 1.2916757701551235e-06, + "loss": 2.0087, + "step": 33996 + }, + { + "epoch": 7.42, + "learning_rate": 1.2915665282936422e-06, + "loss": 0.205, + "step": 33997 + }, + { + "epoch": 7.42, + "learning_rate": 1.291457286432161e-06, + "loss": 1.6162, + "step": 33998 + }, + { + "epoch": 7.42, + "learning_rate": 1.2913480445706798e-06, + "loss": 1.5243, + "step": 33999 + }, + { + "epoch": 7.42, + "learning_rate": 1.2912388027091982e-06, + "loss": 0.5448, + "step": 34000 + }, + { + "epoch": 7.42, + "learning_rate": 1.291129560847717e-06, + "loss": 2.5305, + "step": 34001 + }, + { + "epoch": 7.42, + "learning_rate": 1.2910203189862356e-06, + "loss": 2.122, + "step": 34002 + }, + { + "epoch": 7.42, + "learning_rate": 1.2909110771247542e-06, + "loss": 2.5175, + "step": 34003 + }, + { + "epoch": 7.42, + "learning_rate": 1.290801835263273e-06, + "loss": 0.4435, + "step": 34004 + }, + { + "epoch": 7.42, + "learning_rate": 1.2906925934017917e-06, + "loss": 1.0231, + "step": 34005 + }, + { + "epoch": 7.42, + "learning_rate": 1.2905833515403105e-06, + "loss": 1.4606, + "step": 34006 + }, + { + "epoch": 7.42, + "learning_rate": 1.290474109678829e-06, + "loss": 0.1153, + "step": 34007 + }, + { + "epoch": 7.42, + "learning_rate": 1.2903648678173477e-06, + "loss": 0.6849, + "step": 34008 + }, + { + "epoch": 7.42, + "learning_rate": 1.2902556259558663e-06, + "loss": 0.6393, + "step": 34009 + }, + { + "epoch": 7.42, + "learning_rate": 1.2901463840943852e-06, + "loss": 0.358, + "step": 34010 + }, + { + "epoch": 7.42, + "learning_rate": 1.290037142232904e-06, + "loss": 1.3013, + "step": 34011 + }, + { + "epoch": 7.42, + "learning_rate": 1.2899279003714224e-06, + "loss": 0.5787, + "step": 34012 + }, + { + "epoch": 7.42, + "learning_rate": 1.2898186585099412e-06, + "loss": 2.0879, + "step": 34013 + }, + { + "epoch": 7.42, + "learning_rate": 1.2897094166484598e-06, + "loss": 1.3769, + "step": 34014 + }, + { + "epoch": 7.42, + "learning_rate": 1.2896001747869784e-06, + "loss": 1.9214, + "step": 34015 + }, + { + "epoch": 7.42, + "learning_rate": 1.289490932925497e-06, + "loss": 1.1557, + "step": 34016 + }, + { + "epoch": 7.42, + "learning_rate": 1.2893816910640159e-06, + "loss": 1.835, + "step": 34017 + }, + { + "epoch": 7.42, + "learning_rate": 1.2892724492025347e-06, + "loss": 2.5492, + "step": 34018 + }, + { + "epoch": 7.42, + "learning_rate": 1.289163207341053e-06, + "loss": 0.1818, + "step": 34019 + }, + { + "epoch": 7.42, + "learning_rate": 1.289053965479572e-06, + "loss": 1.1085, + "step": 34020 + }, + { + "epoch": 7.42, + "learning_rate": 1.2889447236180905e-06, + "loss": 1.3556, + "step": 34021 + }, + { + "epoch": 7.43, + "learning_rate": 1.2888354817566093e-06, + "loss": 1.5734, + "step": 34022 + }, + { + "epoch": 7.43, + "learning_rate": 1.2887262398951277e-06, + "loss": 1.4324, + "step": 34023 + }, + { + "epoch": 7.43, + "learning_rate": 1.2886169980336466e-06, + "loss": 1.2179, + "step": 34024 + }, + { + "epoch": 7.43, + "learning_rate": 1.2885077561721654e-06, + "loss": 0.1524, + "step": 34025 + }, + { + "epoch": 7.43, + "learning_rate": 1.288398514310684e-06, + "loss": 1.347, + "step": 34026 + }, + { + "epoch": 7.43, + "learning_rate": 1.2882892724492028e-06, + "loss": 0.2902, + "step": 34027 + }, + { + "epoch": 7.43, + "learning_rate": 1.2881800305877212e-06, + "loss": 0.6398, + "step": 34028 + }, + { + "epoch": 7.43, + "learning_rate": 1.28807078872624e-06, + "loss": 0.9394, + "step": 34029 + }, + { + "epoch": 7.43, + "learning_rate": 1.2879615468647586e-06, + "loss": 0.828, + "step": 34030 + }, + { + "epoch": 7.43, + "learning_rate": 1.2878523050032773e-06, + "loss": 1.8275, + "step": 34031 + }, + { + "epoch": 7.43, + "learning_rate": 1.287743063141796e-06, + "loss": 0.7548, + "step": 34032 + }, + { + "epoch": 7.43, + "learning_rate": 1.2876338212803147e-06, + "loss": 2.796, + "step": 34033 + }, + { + "epoch": 7.43, + "learning_rate": 1.2875245794188335e-06, + "loss": 1.4321, + "step": 34034 + }, + { + "epoch": 7.43, + "learning_rate": 1.287415337557352e-06, + "loss": 0.6586, + "step": 34035 + }, + { + "epoch": 7.43, + "learning_rate": 1.2873060956958707e-06, + "loss": 1.3799, + "step": 34036 + }, + { + "epoch": 7.43, + "learning_rate": 1.2871968538343896e-06, + "loss": 0.131, + "step": 34037 + }, + { + "epoch": 7.43, + "learning_rate": 1.2870876119729082e-06, + "loss": 1.9081, + "step": 34038 + }, + { + "epoch": 7.43, + "learning_rate": 1.286978370111427e-06, + "loss": 0.7989, + "step": 34039 + }, + { + "epoch": 7.43, + "learning_rate": 1.2868691282499454e-06, + "loss": 0.271, + "step": 34040 + }, + { + "epoch": 7.43, + "learning_rate": 1.2867598863884642e-06, + "loss": 1.9594, + "step": 34041 + }, + { + "epoch": 7.43, + "learning_rate": 1.2866506445269828e-06, + "loss": 1.2734, + "step": 34042 + }, + { + "epoch": 7.43, + "learning_rate": 1.2865414026655014e-06, + "loss": 0.7347, + "step": 34043 + }, + { + "epoch": 7.43, + "learning_rate": 1.2864321608040203e-06, + "loss": 0.8484, + "step": 34044 + }, + { + "epoch": 7.43, + "learning_rate": 1.2863229189425389e-06, + "loss": 2.0809, + "step": 34045 + }, + { + "epoch": 7.43, + "learning_rate": 1.2862136770810577e-06, + "loss": 2.6891, + "step": 34046 + }, + { + "epoch": 7.43, + "learning_rate": 1.286104435219576e-06, + "loss": 0.4305, + "step": 34047 + }, + { + "epoch": 7.43, + "learning_rate": 1.285995193358095e-06, + "loss": 0.6783, + "step": 34048 + }, + { + "epoch": 7.43, + "learning_rate": 1.2858859514966135e-06, + "loss": 0.3186, + "step": 34049 + }, + { + "epoch": 7.43, + "learning_rate": 1.2857767096351323e-06, + "loss": 3.2359, + "step": 34050 + }, + { + "epoch": 7.43, + "learning_rate": 1.2856674677736512e-06, + "loss": 0.2966, + "step": 34051 + }, + { + "epoch": 7.43, + "learning_rate": 1.2855582259121696e-06, + "loss": 0.9753, + "step": 34052 + }, + { + "epoch": 7.43, + "learning_rate": 1.2854489840506884e-06, + "loss": 0.7294, + "step": 34053 + }, + { + "epoch": 7.43, + "learning_rate": 1.285339742189207e-06, + "loss": 2.1016, + "step": 34054 + }, + { + "epoch": 7.43, + "learning_rate": 1.2852305003277258e-06, + "loss": 1.5869, + "step": 34055 + }, + { + "epoch": 7.43, + "learning_rate": 1.2851212584662442e-06, + "loss": 0.5289, + "step": 34056 + }, + { + "epoch": 7.43, + "learning_rate": 1.285012016604763e-06, + "loss": 0.2088, + "step": 34057 + }, + { + "epoch": 7.43, + "learning_rate": 1.2849027747432819e-06, + "loss": 1.1979, + "step": 34058 + }, + { + "epoch": 7.43, + "learning_rate": 1.2847935328818003e-06, + "loss": 0.6522, + "step": 34059 + }, + { + "epoch": 7.43, + "learning_rate": 1.284684291020319e-06, + "loss": 2.299, + "step": 34060 + }, + { + "epoch": 7.43, + "learning_rate": 1.2845750491588377e-06, + "loss": 0.4748, + "step": 34061 + }, + { + "epoch": 7.43, + "learning_rate": 1.2844658072973565e-06, + "loss": 1.1284, + "step": 34062 + }, + { + "epoch": 7.43, + "learning_rate": 1.284356565435875e-06, + "loss": 0.2581, + "step": 34063 + }, + { + "epoch": 7.43, + "learning_rate": 1.2842473235743937e-06, + "loss": 2.011, + "step": 34064 + }, + { + "epoch": 7.43, + "learning_rate": 1.2841380817129126e-06, + "loss": 0.5076, + "step": 34065 + }, + { + "epoch": 7.43, + "learning_rate": 1.2840288398514312e-06, + "loss": 0.7389, + "step": 34066 + }, + { + "epoch": 7.43, + "learning_rate": 1.28391959798995e-06, + "loss": 1.1225, + "step": 34067 + }, + { + "epoch": 7.44, + "learning_rate": 1.2838103561284684e-06, + "loss": 2.7717, + "step": 34068 + }, + { + "epoch": 7.44, + "learning_rate": 1.2837011142669872e-06, + "loss": 0.8775, + "step": 34069 + }, + { + "epoch": 7.44, + "learning_rate": 1.2835918724055058e-06, + "loss": 0.8918, + "step": 34070 + }, + { + "epoch": 7.44, + "learning_rate": 1.2834826305440247e-06, + "loss": 0.2227, + "step": 34071 + }, + { + "epoch": 7.44, + "learning_rate": 1.2833733886825433e-06, + "loss": 1.538, + "step": 34072 + }, + { + "epoch": 7.44, + "learning_rate": 1.2832641468210619e-06, + "loss": 2.0251, + "step": 34073 + }, + { + "epoch": 7.44, + "learning_rate": 1.2831549049595807e-06, + "loss": 2.0304, + "step": 34074 + }, + { + "epoch": 7.44, + "learning_rate": 1.283045663098099e-06, + "loss": 1.8422, + "step": 34075 + }, + { + "epoch": 7.44, + "learning_rate": 1.282936421236618e-06, + "loss": 0.924, + "step": 34076 + }, + { + "epoch": 7.44, + "learning_rate": 1.2828271793751367e-06, + "loss": 0.6327, + "step": 34077 + }, + { + "epoch": 7.44, + "learning_rate": 1.2827179375136554e-06, + "loss": 0.2369, + "step": 34078 + }, + { + "epoch": 7.44, + "learning_rate": 1.2826086956521742e-06, + "loss": 1.6383, + "step": 34079 + }, + { + "epoch": 7.44, + "learning_rate": 1.2824994537906926e-06, + "loss": 1.9053, + "step": 34080 + }, + { + "epoch": 7.44, + "learning_rate": 1.2823902119292114e-06, + "loss": 2.6352, + "step": 34081 + }, + { + "epoch": 7.44, + "learning_rate": 1.28228097006773e-06, + "loss": 1.2831, + "step": 34082 + }, + { + "epoch": 7.44, + "learning_rate": 1.2821717282062488e-06, + "loss": 1.3449, + "step": 34083 + }, + { + "epoch": 7.44, + "learning_rate": 1.2820624863447674e-06, + "loss": 1.6199, + "step": 34084 + }, + { + "epoch": 7.44, + "learning_rate": 1.281953244483286e-06, + "loss": 1.4785, + "step": 34085 + }, + { + "epoch": 7.44, + "learning_rate": 1.2818440026218049e-06, + "loss": 1.2154, + "step": 34086 + }, + { + "epoch": 7.44, + "learning_rate": 1.2817347607603233e-06, + "loss": 1.0038, + "step": 34087 + }, + { + "epoch": 7.44, + "learning_rate": 1.281625518898842e-06, + "loss": 1.0152, + "step": 34088 + }, + { + "epoch": 7.44, + "learning_rate": 1.2815162770373607e-06, + "loss": 1.0789, + "step": 34089 + }, + { + "epoch": 7.44, + "learning_rate": 1.2814070351758795e-06, + "loss": 0.2451, + "step": 34090 + }, + { + "epoch": 7.44, + "learning_rate": 1.2812977933143984e-06, + "loss": 1.8019, + "step": 34091 + }, + { + "epoch": 7.44, + "learning_rate": 1.2811885514529168e-06, + "loss": 1.4219, + "step": 34092 + }, + { + "epoch": 7.44, + "learning_rate": 1.2810793095914356e-06, + "loss": 1.0459, + "step": 34093 + }, + { + "epoch": 7.44, + "learning_rate": 1.2809700677299542e-06, + "loss": 1.9949, + "step": 34094 + }, + { + "epoch": 7.44, + "learning_rate": 1.280860825868473e-06, + "loss": 0.9343, + "step": 34095 + }, + { + "epoch": 7.44, + "learning_rate": 1.2807515840069914e-06, + "loss": 0.7284, + "step": 34096 + }, + { + "epoch": 7.44, + "learning_rate": 1.2806423421455102e-06, + "loss": 0.8285, + "step": 34097 + }, + { + "epoch": 7.44, + "learning_rate": 1.280533100284029e-06, + "loss": 2.5361, + "step": 34098 + }, + { + "epoch": 7.44, + "learning_rate": 1.2804238584225477e-06, + "loss": 1.7673, + "step": 34099 + }, + { + "epoch": 7.44, + "learning_rate": 1.2803146165610663e-06, + "loss": 0.6756, + "step": 34100 + }, + { + "epoch": 7.44, + "learning_rate": 1.280205374699585e-06, + "loss": 1.3252, + "step": 34101 + }, + { + "epoch": 7.44, + "learning_rate": 1.2800961328381037e-06, + "loss": 0.3086, + "step": 34102 + }, + { + "epoch": 7.44, + "learning_rate": 1.2799868909766221e-06, + "loss": 1.0607, + "step": 34103 + }, + { + "epoch": 7.44, + "learning_rate": 1.279877649115141e-06, + "loss": 0.9455, + "step": 34104 + }, + { + "epoch": 7.44, + "learning_rate": 1.2797684072536598e-06, + "loss": 0.2072, + "step": 34105 + }, + { + "epoch": 7.44, + "learning_rate": 1.2796591653921784e-06, + "loss": 1.2194, + "step": 34106 + }, + { + "epoch": 7.44, + "learning_rate": 1.2795499235306972e-06, + "loss": 2.6355, + "step": 34107 + }, + { + "epoch": 7.44, + "learning_rate": 1.2794406816692156e-06, + "loss": 0.0914, + "step": 34108 + }, + { + "epoch": 7.44, + "learning_rate": 1.2793314398077344e-06, + "loss": 0.6834, + "step": 34109 + }, + { + "epoch": 7.44, + "learning_rate": 1.2792221979462532e-06, + "loss": 0.1331, + "step": 34110 + }, + { + "epoch": 7.44, + "learning_rate": 1.2791129560847719e-06, + "loss": 0.6296, + "step": 34111 + }, + { + "epoch": 7.44, + "learning_rate": 1.2790037142232905e-06, + "loss": 1.6123, + "step": 34112 + }, + { + "epoch": 7.45, + "learning_rate": 1.278894472361809e-06, + "loss": 0.6829, + "step": 34113 + }, + { + "epoch": 7.45, + "learning_rate": 1.278785230500328e-06, + "loss": 1.4462, + "step": 34114 + }, + { + "epoch": 7.45, + "learning_rate": 1.2786759886388463e-06, + "loss": 1.9849, + "step": 34115 + }, + { + "epoch": 7.45, + "learning_rate": 1.2785667467773651e-06, + "loss": 1.5645, + "step": 34116 + }, + { + "epoch": 7.45, + "learning_rate": 1.278457504915884e-06, + "loss": 1.1647, + "step": 34117 + }, + { + "epoch": 7.45, + "learning_rate": 1.2783482630544026e-06, + "loss": 0.9663, + "step": 34118 + }, + { + "epoch": 7.45, + "learning_rate": 1.2782390211929214e-06, + "loss": 0.5203, + "step": 34119 + }, + { + "epoch": 7.45, + "learning_rate": 1.2781297793314398e-06, + "loss": 1.0191, + "step": 34120 + }, + { + "epoch": 7.45, + "learning_rate": 1.2780205374699586e-06, + "loss": 0.9213, + "step": 34121 + }, + { + "epoch": 7.45, + "learning_rate": 1.2779112956084772e-06, + "loss": 2.6393, + "step": 34122 + }, + { + "epoch": 7.45, + "learning_rate": 1.277802053746996e-06, + "loss": 2.2317, + "step": 34123 + }, + { + "epoch": 7.45, + "learning_rate": 1.2776928118855149e-06, + "loss": 0.7664, + "step": 34124 + }, + { + "epoch": 7.45, + "learning_rate": 1.2775835700240333e-06, + "loss": 0.8415, + "step": 34125 + }, + { + "epoch": 7.45, + "learning_rate": 1.277474328162552e-06, + "loss": 1.0498, + "step": 34126 + }, + { + "epoch": 7.45, + "learning_rate": 1.2773650863010707e-06, + "loss": 2.7511, + "step": 34127 + }, + { + "epoch": 7.45, + "learning_rate": 1.2772558444395893e-06, + "loss": 1.0985, + "step": 34128 + }, + { + "epoch": 7.45, + "learning_rate": 1.277146602578108e-06, + "loss": 0.8212, + "step": 34129 + }, + { + "epoch": 7.45, + "learning_rate": 1.2770373607166267e-06, + "loss": 1.2655, + "step": 34130 + }, + { + "epoch": 7.45, + "learning_rate": 1.2769281188551456e-06, + "loss": 0.8928, + "step": 34131 + }, + { + "epoch": 7.45, + "learning_rate": 1.276818876993664e-06, + "loss": 1.8865, + "step": 34132 + }, + { + "epoch": 7.45, + "learning_rate": 1.2767096351321828e-06, + "loss": 3.5443, + "step": 34133 + }, + { + "epoch": 7.45, + "learning_rate": 1.2766003932707014e-06, + "loss": 2.8461, + "step": 34134 + }, + { + "epoch": 7.45, + "learning_rate": 1.2764911514092202e-06, + "loss": 0.1085, + "step": 34135 + }, + { + "epoch": 7.45, + "learning_rate": 1.2763819095477386e-06, + "loss": 0.4674, + "step": 34136 + }, + { + "epoch": 7.45, + "learning_rate": 1.2762726676862574e-06, + "loss": 2.3907, + "step": 34137 + }, + { + "epoch": 7.45, + "learning_rate": 1.2761634258247763e-06, + "loss": 0.1922, + "step": 34138 + }, + { + "epoch": 7.45, + "learning_rate": 1.2760541839632949e-06, + "loss": 0.6709, + "step": 34139 + }, + { + "epoch": 7.45, + "learning_rate": 1.2759449421018135e-06, + "loss": 0.4966, + "step": 34140 + }, + { + "epoch": 7.45, + "learning_rate": 1.275835700240332e-06, + "loss": 2.0144, + "step": 34141 + }, + { + "epoch": 7.45, + "learning_rate": 1.275726458378851e-06, + "loss": 0.12, + "step": 34142 + }, + { + "epoch": 7.45, + "learning_rate": 1.2756172165173697e-06, + "loss": 0.8427, + "step": 34143 + }, + { + "epoch": 7.45, + "learning_rate": 1.2755079746558881e-06, + "loss": 1.7389, + "step": 34144 + }, + { + "epoch": 7.45, + "learning_rate": 1.275398732794407e-06, + "loss": 3.4668, + "step": 34145 + }, + { + "epoch": 7.45, + "learning_rate": 1.2752894909329256e-06, + "loss": 1.147, + "step": 34146 + }, + { + "epoch": 7.45, + "learning_rate": 1.2751802490714444e-06, + "loss": 0.6266, + "step": 34147 + }, + { + "epoch": 7.45, + "learning_rate": 1.2750710072099628e-06, + "loss": 0.516, + "step": 34148 + }, + { + "epoch": 7.45, + "learning_rate": 1.2749617653484816e-06, + "loss": 1.9446, + "step": 34149 + }, + { + "epoch": 7.45, + "learning_rate": 1.2748525234870004e-06, + "loss": 1.1561, + "step": 34150 + }, + { + "epoch": 7.45, + "learning_rate": 1.274743281625519e-06, + "loss": 0.7893, + "step": 34151 + }, + { + "epoch": 7.45, + "learning_rate": 1.2746340397640379e-06, + "loss": 1.3222, + "step": 34152 + }, + { + "epoch": 7.45, + "learning_rate": 1.2745247979025563e-06, + "loss": 0.5613, + "step": 34153 + }, + { + "epoch": 7.45, + "learning_rate": 1.274415556041075e-06, + "loss": 0.059, + "step": 34154 + }, + { + "epoch": 7.45, + "learning_rate": 1.2743063141795937e-06, + "loss": 0.514, + "step": 34155 + }, + { + "epoch": 7.45, + "learning_rate": 1.2741970723181123e-06, + "loss": 1.9738, + "step": 34156 + }, + { + "epoch": 7.45, + "learning_rate": 1.2740878304566311e-06, + "loss": 1.0061, + "step": 34157 + }, + { + "epoch": 7.45, + "learning_rate": 1.2739785885951497e-06, + "loss": 0.0725, + "step": 34158 + }, + { + "epoch": 7.46, + "learning_rate": 1.2738693467336686e-06, + "loss": 1.4129, + "step": 34159 + }, + { + "epoch": 7.46, + "learning_rate": 1.273760104872187e-06, + "loss": 0.6213, + "step": 34160 + }, + { + "epoch": 7.46, + "learning_rate": 1.2736508630107058e-06, + "loss": 1.5319, + "step": 34161 + }, + { + "epoch": 7.46, + "learning_rate": 1.2735416211492244e-06, + "loss": 2.5201, + "step": 34162 + }, + { + "epoch": 7.46, + "learning_rate": 1.2734323792877432e-06, + "loss": 0.6934, + "step": 34163 + }, + { + "epoch": 7.46, + "learning_rate": 1.273323137426262e-06, + "loss": 0.9056, + "step": 34164 + }, + { + "epoch": 7.46, + "learning_rate": 1.2732138955647804e-06, + "loss": 0.947, + "step": 34165 + }, + { + "epoch": 7.46, + "learning_rate": 1.2731046537032993e-06, + "loss": 0.5312, + "step": 34166 + }, + { + "epoch": 7.46, + "learning_rate": 1.2729954118418179e-06, + "loss": 0.1101, + "step": 34167 + }, + { + "epoch": 7.46, + "learning_rate": 1.2728861699803365e-06, + "loss": 0.7439, + "step": 34168 + }, + { + "epoch": 7.46, + "learning_rate": 1.272776928118855e-06, + "loss": 0.7063, + "step": 34169 + }, + { + "epoch": 7.46, + "learning_rate": 1.272667686257374e-06, + "loss": 0.7527, + "step": 34170 + }, + { + "epoch": 7.46, + "learning_rate": 1.2725584443958927e-06, + "loss": 1.2371, + "step": 34171 + }, + { + "epoch": 7.46, + "learning_rate": 1.2724492025344111e-06, + "loss": 0.4794, + "step": 34172 + }, + { + "epoch": 7.46, + "learning_rate": 1.27233996067293e-06, + "loss": 1.2603, + "step": 34173 + }, + { + "epoch": 7.46, + "learning_rate": 1.2722307188114486e-06, + "loss": 0.1975, + "step": 34174 + }, + { + "epoch": 7.46, + "learning_rate": 1.2721214769499674e-06, + "loss": 1.239, + "step": 34175 + }, + { + "epoch": 7.46, + "learning_rate": 1.2720122350884862e-06, + "loss": 1.0761, + "step": 34176 + }, + { + "epoch": 7.46, + "learning_rate": 1.2719029932270046e-06, + "loss": 1.1364, + "step": 34177 + }, + { + "epoch": 7.46, + "learning_rate": 1.2717937513655234e-06, + "loss": 0.6662, + "step": 34178 + }, + { + "epoch": 7.46, + "learning_rate": 1.271684509504042e-06, + "loss": 0.8712, + "step": 34179 + }, + { + "epoch": 7.46, + "learning_rate": 1.2715752676425609e-06, + "loss": 1.2142, + "step": 34180 + }, + { + "epoch": 7.46, + "learning_rate": 1.2714660257810793e-06, + "loss": 1.2642, + "step": 34181 + }, + { + "epoch": 7.46, + "learning_rate": 1.271356783919598e-06, + "loss": 0.4003, + "step": 34182 + }, + { + "epoch": 7.46, + "learning_rate": 1.271247542058117e-06, + "loss": 0.9053, + "step": 34183 + }, + { + "epoch": 7.46, + "learning_rate": 1.2711383001966353e-06, + "loss": 0.5828, + "step": 34184 + }, + { + "epoch": 7.46, + "learning_rate": 1.2710290583351541e-06, + "loss": 1.5662, + "step": 34185 + }, + { + "epoch": 7.46, + "learning_rate": 1.2709198164736728e-06, + "loss": 0.6033, + "step": 34186 + }, + { + "epoch": 7.46, + "learning_rate": 1.2708105746121916e-06, + "loss": 1.3161, + "step": 34187 + }, + { + "epoch": 7.46, + "learning_rate": 1.27070133275071e-06, + "loss": 1.6709, + "step": 34188 + }, + { + "epoch": 7.46, + "learning_rate": 1.2705920908892288e-06, + "loss": 2.5837, + "step": 34189 + }, + { + "epoch": 7.46, + "learning_rate": 1.2704828490277476e-06, + "loss": 2.3299, + "step": 34190 + }, + { + "epoch": 7.46, + "learning_rate": 1.2703736071662662e-06, + "loss": 1.5322, + "step": 34191 + }, + { + "epoch": 7.46, + "learning_rate": 1.270264365304785e-06, + "loss": 0.3935, + "step": 34192 + }, + { + "epoch": 7.46, + "learning_rate": 1.2701551234433035e-06, + "loss": 2.0126, + "step": 34193 + }, + { + "epoch": 7.46, + "learning_rate": 1.2700458815818223e-06, + "loss": 0.7485, + "step": 34194 + }, + { + "epoch": 7.46, + "learning_rate": 1.2699366397203409e-06, + "loss": 1.8556, + "step": 34195 + }, + { + "epoch": 7.46, + "learning_rate": 1.2698273978588597e-06, + "loss": 0.6004, + "step": 34196 + }, + { + "epoch": 7.46, + "learning_rate": 1.2697181559973783e-06, + "loss": 1.1805, + "step": 34197 + }, + { + "epoch": 7.46, + "learning_rate": 1.269608914135897e-06, + "loss": 1.6683, + "step": 34198 + }, + { + "epoch": 7.46, + "learning_rate": 1.2694996722744158e-06, + "loss": 1.6628, + "step": 34199 + }, + { + "epoch": 7.46, + "learning_rate": 1.2693904304129342e-06, + "loss": 1.025, + "step": 34200 + }, + { + "epoch": 7.46, + "learning_rate": 1.269281188551453e-06, + "loss": 0.559, + "step": 34201 + }, + { + "epoch": 7.46, + "learning_rate": 1.2691719466899716e-06, + "loss": 0.4251, + "step": 34202 + }, + { + "epoch": 7.46, + "learning_rate": 1.2690627048284904e-06, + "loss": 0.989, + "step": 34203 + }, + { + "epoch": 7.46, + "learning_rate": 1.2689534629670092e-06, + "loss": 0.5515, + "step": 34204 + }, + { + "epoch": 7.47, + "learning_rate": 1.2688442211055276e-06, + "loss": 1.528, + "step": 34205 + }, + { + "epoch": 7.47, + "learning_rate": 1.2687349792440465e-06, + "loss": 1.0908, + "step": 34206 + }, + { + "epoch": 7.47, + "learning_rate": 1.268625737382565e-06, + "loss": 0.6875, + "step": 34207 + }, + { + "epoch": 7.47, + "learning_rate": 1.2685164955210839e-06, + "loss": 0.9886, + "step": 34208 + }, + { + "epoch": 7.47, + "learning_rate": 1.2684072536596025e-06, + "loss": 0.5704, + "step": 34209 + }, + { + "epoch": 7.47, + "learning_rate": 1.2682980117981211e-06, + "loss": 1.2269, + "step": 34210 + }, + { + "epoch": 7.47, + "learning_rate": 1.26818876993664e-06, + "loss": 0.7286, + "step": 34211 + }, + { + "epoch": 7.47, + "learning_rate": 1.2680795280751583e-06, + "loss": 1.4623, + "step": 34212 + }, + { + "epoch": 7.47, + "learning_rate": 1.2679702862136772e-06, + "loss": 0.4716, + "step": 34213 + }, + { + "epoch": 7.47, + "learning_rate": 1.2678610443521958e-06, + "loss": 0.3146, + "step": 34214 + }, + { + "epoch": 7.47, + "learning_rate": 1.2677518024907146e-06, + "loss": 2.8187, + "step": 34215 + }, + { + "epoch": 7.47, + "learning_rate": 1.2676425606292334e-06, + "loss": 0.9254, + "step": 34216 + }, + { + "epoch": 7.47, + "learning_rate": 1.2675333187677518e-06, + "loss": 2.0976, + "step": 34217 + }, + { + "epoch": 7.47, + "learning_rate": 1.2674240769062706e-06, + "loss": 2.26, + "step": 34218 + }, + { + "epoch": 7.47, + "learning_rate": 1.2673148350447892e-06, + "loss": 1.0625, + "step": 34219 + }, + { + "epoch": 7.47, + "learning_rate": 1.267205593183308e-06, + "loss": 1.9699, + "step": 34220 + }, + { + "epoch": 7.47, + "learning_rate": 1.2670963513218265e-06, + "loss": 0.8166, + "step": 34221 + }, + { + "epoch": 7.47, + "learning_rate": 1.2669871094603453e-06, + "loss": 0.7596, + "step": 34222 + }, + { + "epoch": 7.47, + "learning_rate": 1.2668778675988641e-06, + "loss": 0.4253, + "step": 34223 + }, + { + "epoch": 7.47, + "learning_rate": 1.2667686257373827e-06, + "loss": 1.3132, + "step": 34224 + }, + { + "epoch": 7.47, + "learning_rate": 1.2666593838759013e-06, + "loss": 1.9075, + "step": 34225 + }, + { + "epoch": 7.47, + "learning_rate": 1.26655014201442e-06, + "loss": 1.1687, + "step": 34226 + }, + { + "epoch": 7.47, + "learning_rate": 1.2664409001529388e-06, + "loss": 0.8256, + "step": 34227 + }, + { + "epoch": 7.47, + "learning_rate": 1.2663316582914572e-06, + "loss": 1.913, + "step": 34228 + }, + { + "epoch": 7.47, + "learning_rate": 1.266222416429976e-06, + "loss": 1.2136, + "step": 34229 + }, + { + "epoch": 7.47, + "learning_rate": 1.2661131745684948e-06, + "loss": 0.909, + "step": 34230 + }, + { + "epoch": 7.47, + "learning_rate": 1.2660039327070134e-06, + "loss": 0.8463, + "step": 34231 + }, + { + "epoch": 7.47, + "learning_rate": 1.2658946908455323e-06, + "loss": 2.8406, + "step": 34232 + }, + { + "epoch": 7.47, + "learning_rate": 1.2657854489840507e-06, + "loss": 2.1899, + "step": 34233 + }, + { + "epoch": 7.47, + "learning_rate": 1.2656762071225695e-06, + "loss": 2.1246, + "step": 34234 + }, + { + "epoch": 7.47, + "learning_rate": 1.265566965261088e-06, + "loss": 1.1535, + "step": 34235 + }, + { + "epoch": 7.47, + "learning_rate": 1.265457723399607e-06, + "loss": 1.3891, + "step": 34236 + }, + { + "epoch": 7.47, + "learning_rate": 1.2653484815381255e-06, + "loss": 1.1584, + "step": 34237 + }, + { + "epoch": 7.47, + "learning_rate": 1.2652392396766441e-06, + "loss": 0.5119, + "step": 34238 + }, + { + "epoch": 7.47, + "learning_rate": 1.265129997815163e-06, + "loss": 1.1777, + "step": 34239 + }, + { + "epoch": 7.47, + "learning_rate": 1.2650207559536814e-06, + "loss": 2.0427, + "step": 34240 + }, + { + "epoch": 7.47, + "learning_rate": 1.2649115140922002e-06, + "loss": 1.332, + "step": 34241 + }, + { + "epoch": 7.47, + "learning_rate": 1.264802272230719e-06, + "loss": 1.1035, + "step": 34242 + }, + { + "epoch": 7.47, + "learning_rate": 1.2646930303692376e-06, + "loss": 2.3885, + "step": 34243 + }, + { + "epoch": 7.47, + "learning_rate": 1.2645837885077564e-06, + "loss": 0.4872, + "step": 34244 + }, + { + "epoch": 7.47, + "learning_rate": 1.2644745466462748e-06, + "loss": 0.6996, + "step": 34245 + }, + { + "epoch": 7.47, + "learning_rate": 1.2643653047847937e-06, + "loss": 2.2694, + "step": 34246 + }, + { + "epoch": 7.47, + "learning_rate": 1.2642560629233123e-06, + "loss": 1.4498, + "step": 34247 + }, + { + "epoch": 7.47, + "learning_rate": 1.264146821061831e-06, + "loss": 1.1388, + "step": 34248 + }, + { + "epoch": 7.47, + "learning_rate": 1.26403757920035e-06, + "loss": 1.1147, + "step": 34249 + }, + { + "epoch": 7.47, + "learning_rate": 1.2639283373388683e-06, + "loss": 1.9816, + "step": 34250 + }, + { + "epoch": 7.48, + "learning_rate": 1.2638190954773871e-06, + "loss": 1.2253, + "step": 34251 + }, + { + "epoch": 7.48, + "learning_rate": 1.2637098536159057e-06, + "loss": 1.9024, + "step": 34252 + }, + { + "epoch": 7.48, + "learning_rate": 1.2636006117544244e-06, + "loss": 1.2413, + "step": 34253 + }, + { + "epoch": 7.48, + "learning_rate": 1.263491369892943e-06, + "loss": 1.7953, + "step": 34254 + }, + { + "epoch": 7.48, + "learning_rate": 1.2633821280314618e-06, + "loss": 0.2543, + "step": 34255 + }, + { + "epoch": 7.48, + "learning_rate": 1.2632728861699806e-06, + "loss": 1.4676, + "step": 34256 + }, + { + "epoch": 7.48, + "learning_rate": 1.263163644308499e-06, + "loss": 1.3934, + "step": 34257 + }, + { + "epoch": 7.48, + "learning_rate": 1.2630544024470178e-06, + "loss": 1.3071, + "step": 34258 + }, + { + "epoch": 7.48, + "learning_rate": 1.2629451605855364e-06, + "loss": 1.3041, + "step": 34259 + }, + { + "epoch": 7.48, + "learning_rate": 1.2628359187240553e-06, + "loss": 2.4555, + "step": 34260 + }, + { + "epoch": 7.48, + "learning_rate": 1.2627266768625737e-06, + "loss": 3.3444, + "step": 34261 + }, + { + "epoch": 7.48, + "learning_rate": 1.2626174350010925e-06, + "loss": 1.8057, + "step": 34262 + }, + { + "epoch": 7.48, + "learning_rate": 1.2625081931396113e-06, + "loss": 1.1246, + "step": 34263 + }, + { + "epoch": 7.48, + "learning_rate": 1.26239895127813e-06, + "loss": 1.1815, + "step": 34264 + }, + { + "epoch": 7.48, + "learning_rate": 1.2622897094166485e-06, + "loss": 0.7214, + "step": 34265 + }, + { + "epoch": 7.48, + "learning_rate": 1.2621804675551671e-06, + "loss": 1.7781, + "step": 34266 + }, + { + "epoch": 7.48, + "learning_rate": 1.262071225693686e-06, + "loss": 0.7595, + "step": 34267 + }, + { + "epoch": 7.48, + "learning_rate": 1.2619619838322046e-06, + "loss": 0.7385, + "step": 34268 + }, + { + "epoch": 7.48, + "learning_rate": 1.2618527419707232e-06, + "loss": 2.5903, + "step": 34269 + }, + { + "epoch": 7.48, + "learning_rate": 1.261743500109242e-06, + "loss": 1.6397, + "step": 34270 + }, + { + "epoch": 7.48, + "learning_rate": 1.2616342582477606e-06, + "loss": 0.6804, + "step": 34271 + }, + { + "epoch": 7.48, + "learning_rate": 1.2615250163862794e-06, + "loss": 0.3309, + "step": 34272 + }, + { + "epoch": 7.48, + "learning_rate": 1.2614157745247978e-06, + "loss": 1.4835, + "step": 34273 + }, + { + "epoch": 7.48, + "learning_rate": 1.2613065326633167e-06, + "loss": 0.4363, + "step": 34274 + }, + { + "epoch": 7.48, + "learning_rate": 1.2611972908018353e-06, + "loss": 0.6849, + "step": 34275 + }, + { + "epoch": 7.48, + "learning_rate": 1.261088048940354e-06, + "loss": 1.2871, + "step": 34276 + }, + { + "epoch": 7.48, + "learning_rate": 1.260978807078873e-06, + "loss": 0.4466, + "step": 34277 + }, + { + "epoch": 7.48, + "learning_rate": 1.2608695652173913e-06, + "loss": 1.0726, + "step": 34278 + }, + { + "epoch": 7.48, + "learning_rate": 1.2607603233559101e-06, + "loss": 1.7928, + "step": 34279 + }, + { + "epoch": 7.48, + "learning_rate": 1.2606510814944288e-06, + "loss": 0.6745, + "step": 34280 + }, + { + "epoch": 7.48, + "learning_rate": 1.2605418396329474e-06, + "loss": 1.4953, + "step": 34281 + }, + { + "epoch": 7.48, + "learning_rate": 1.2604325977714662e-06, + "loss": 0.7714, + "step": 34282 + }, + { + "epoch": 7.48, + "learning_rate": 1.2603233559099848e-06, + "loss": 0.911, + "step": 34283 + }, + { + "epoch": 7.48, + "learning_rate": 1.2602141140485036e-06, + "loss": 2.1539, + "step": 34284 + }, + { + "epoch": 7.48, + "learning_rate": 1.260104872187022e-06, + "loss": 2.3007, + "step": 34285 + }, + { + "epoch": 7.48, + "learning_rate": 1.2599956303255408e-06, + "loss": 0.8565, + "step": 34286 + }, + { + "epoch": 7.48, + "learning_rate": 1.2598863884640595e-06, + "loss": 1.9671, + "step": 34287 + }, + { + "epoch": 7.48, + "learning_rate": 1.2597771466025783e-06, + "loss": 0.3684, + "step": 34288 + }, + { + "epoch": 7.48, + "learning_rate": 1.259667904741097e-06, + "loss": 1.9269, + "step": 34289 + }, + { + "epoch": 7.48, + "learning_rate": 1.2595586628796155e-06, + "loss": 3.6993, + "step": 34290 + }, + { + "epoch": 7.48, + "learning_rate": 1.2594494210181343e-06, + "loss": 0.8394, + "step": 34291 + }, + { + "epoch": 7.48, + "learning_rate": 1.259340179156653e-06, + "loss": 1.2429, + "step": 34292 + }, + { + "epoch": 7.48, + "learning_rate": 1.2592309372951718e-06, + "loss": 1.2605, + "step": 34293 + }, + { + "epoch": 7.48, + "learning_rate": 1.2591216954336902e-06, + "loss": 0.4407, + "step": 34294 + }, + { + "epoch": 7.48, + "learning_rate": 1.259012453572209e-06, + "loss": 1.3367, + "step": 34295 + }, + { + "epoch": 7.48, + "learning_rate": 1.2589032117107278e-06, + "loss": 1.292, + "step": 34296 + }, + { + "epoch": 7.49, + "learning_rate": 1.2587939698492462e-06, + "loss": 2.301, + "step": 34297 + }, + { + "epoch": 7.49, + "learning_rate": 1.258684727987765e-06, + "loss": 1.9802, + "step": 34298 + }, + { + "epoch": 7.49, + "learning_rate": 1.2585754861262836e-06, + "loss": 1.8431, + "step": 34299 + }, + { + "epoch": 7.49, + "learning_rate": 1.2584662442648025e-06, + "loss": 1.27, + "step": 34300 + }, + { + "epoch": 7.49, + "learning_rate": 1.2583570024033209e-06, + "loss": 0.597, + "step": 34301 + }, + { + "epoch": 7.49, + "learning_rate": 1.2582477605418397e-06, + "loss": 2.9588, + "step": 34302 + }, + { + "epoch": 7.49, + "learning_rate": 1.2581385186803585e-06, + "loss": 1.4335, + "step": 34303 + }, + { + "epoch": 7.49, + "learning_rate": 1.2580292768188771e-06, + "loss": 1.1911, + "step": 34304 + }, + { + "epoch": 7.49, + "learning_rate": 1.257920034957396e-06, + "loss": 0.9302, + "step": 34305 + }, + { + "epoch": 7.49, + "learning_rate": 1.2578107930959143e-06, + "loss": 1.4137, + "step": 34306 + }, + { + "epoch": 7.49, + "learning_rate": 1.2577015512344332e-06, + "loss": 0.2589, + "step": 34307 + }, + { + "epoch": 7.49, + "learning_rate": 1.2575923093729518e-06, + "loss": 0.4016, + "step": 34308 + }, + { + "epoch": 7.49, + "learning_rate": 1.2574830675114704e-06, + "loss": 1.5537, + "step": 34309 + }, + { + "epoch": 7.49, + "learning_rate": 1.2573738256499892e-06, + "loss": 1.3299, + "step": 34310 + }, + { + "epoch": 7.49, + "learning_rate": 1.2572645837885078e-06, + "loss": 0.7054, + "step": 34311 + }, + { + "epoch": 7.49, + "learning_rate": 1.2571553419270266e-06, + "loss": 2.7836, + "step": 34312 + }, + { + "epoch": 7.49, + "learning_rate": 1.257046100065545e-06, + "loss": 2.1117, + "step": 34313 + }, + { + "epoch": 7.49, + "learning_rate": 1.2569368582040639e-06, + "loss": 1.0984, + "step": 34314 + }, + { + "epoch": 7.49, + "learning_rate": 1.2568276163425827e-06, + "loss": 0.7138, + "step": 34315 + }, + { + "epoch": 7.49, + "learning_rate": 1.2567183744811013e-06, + "loss": 0.2017, + "step": 34316 + }, + { + "epoch": 7.49, + "learning_rate": 1.2566091326196201e-06, + "loss": 0.7857, + "step": 34317 + }, + { + "epoch": 7.49, + "learning_rate": 1.2564998907581385e-06, + "loss": 0.1214, + "step": 34318 + }, + { + "epoch": 7.49, + "learning_rate": 1.2563906488966573e-06, + "loss": 1.2499, + "step": 34319 + }, + { + "epoch": 7.49, + "learning_rate": 1.256281407035176e-06, + "loss": 3.8203, + "step": 34320 + }, + { + "epoch": 7.49, + "learning_rate": 1.2561721651736948e-06, + "loss": 0.2996, + "step": 34321 + }, + { + "epoch": 7.49, + "learning_rate": 1.2560629233122134e-06, + "loss": 1.9838, + "step": 34322 + }, + { + "epoch": 7.49, + "learning_rate": 1.255953681450732e-06, + "loss": 1.6556, + "step": 34323 + }, + { + "epoch": 7.49, + "learning_rate": 1.2558444395892508e-06, + "loss": 1.6127, + "step": 34324 + }, + { + "epoch": 7.49, + "learning_rate": 1.2557351977277692e-06, + "loss": 0.6201, + "step": 34325 + }, + { + "epoch": 7.49, + "learning_rate": 1.255625955866288e-06, + "loss": 1.2218, + "step": 34326 + }, + { + "epoch": 7.49, + "learning_rate": 1.2555167140048066e-06, + "loss": 0.424, + "step": 34327 + }, + { + "epoch": 7.49, + "learning_rate": 1.2554074721433255e-06, + "loss": 2.0701, + "step": 34328 + }, + { + "epoch": 7.49, + "learning_rate": 1.2552982302818443e-06, + "loss": 0.5906, + "step": 34329 + }, + { + "epoch": 7.49, + "learning_rate": 1.2551889884203627e-06, + "loss": 0.9442, + "step": 34330 + }, + { + "epoch": 7.49, + "learning_rate": 1.2550797465588815e-06, + "loss": 1.1828, + "step": 34331 + }, + { + "epoch": 7.49, + "learning_rate": 1.2549705046974001e-06, + "loss": 0.5727, + "step": 34332 + }, + { + "epoch": 7.49, + "learning_rate": 1.254861262835919e-06, + "loss": 0.876, + "step": 34333 + }, + { + "epoch": 7.49, + "learning_rate": 1.2547520209744373e-06, + "loss": 1.2004, + "step": 34334 + }, + { + "epoch": 7.49, + "learning_rate": 1.2546427791129562e-06, + "loss": 0.485, + "step": 34335 + }, + { + "epoch": 7.49, + "learning_rate": 1.254533537251475e-06, + "loss": 0.0831, + "step": 34336 + }, + { + "epoch": 7.49, + "learning_rate": 1.2544242953899934e-06, + "loss": 0.59, + "step": 34337 + }, + { + "epoch": 7.49, + "learning_rate": 1.2543150535285122e-06, + "loss": 1.8524, + "step": 34338 + }, + { + "epoch": 7.49, + "learning_rate": 1.2542058116670308e-06, + "loss": 0.9147, + "step": 34339 + }, + { + "epoch": 7.49, + "learning_rate": 1.2540965698055496e-06, + "loss": 1.1747, + "step": 34340 + }, + { + "epoch": 7.49, + "learning_rate": 1.253987327944068e-06, + "loss": 1.9553, + "step": 34341 + }, + { + "epoch": 7.49, + "learning_rate": 1.2538780860825869e-06, + "loss": 2.1047, + "step": 34342 + }, + { + "epoch": 7.5, + "learning_rate": 1.2537688442211057e-06, + "loss": 2.8507, + "step": 34343 + }, + { + "epoch": 7.5, + "learning_rate": 1.2536596023596243e-06, + "loss": 1.287, + "step": 34344 + }, + { + "epoch": 7.5, + "learning_rate": 1.2535503604981431e-06, + "loss": 1.3895, + "step": 34345 + }, + { + "epoch": 7.5, + "learning_rate": 1.2534411186366615e-06, + "loss": 1.1766, + "step": 34346 + }, + { + "epoch": 7.5, + "learning_rate": 1.2533318767751803e-06, + "loss": 1.262, + "step": 34347 + }, + { + "epoch": 7.5, + "learning_rate": 1.2532226349136992e-06, + "loss": 1.5727, + "step": 34348 + }, + { + "epoch": 7.5, + "learning_rate": 1.2531133930522178e-06, + "loss": 0.6702, + "step": 34349 + }, + { + "epoch": 7.5, + "learning_rate": 1.2530041511907364e-06, + "loss": 1.4718, + "step": 34350 + }, + { + "epoch": 7.5, + "learning_rate": 1.252894909329255e-06, + "loss": 2.5391, + "step": 34351 + }, + { + "epoch": 7.5, + "learning_rate": 1.2527856674677738e-06, + "loss": 2.7136, + "step": 34352 + }, + { + "epoch": 7.5, + "learning_rate": 1.2526764256062922e-06, + "loss": 0.772, + "step": 34353 + }, + { + "epoch": 7.5, + "learning_rate": 1.252567183744811e-06, + "loss": 1.9779, + "step": 34354 + }, + { + "epoch": 7.5, + "learning_rate": 1.2524579418833299e-06, + "loss": 0.7207, + "step": 34355 + }, + { + "epoch": 7.5, + "learning_rate": 1.2523487000218485e-06, + "loss": 0.665, + "step": 34356 + }, + { + "epoch": 7.5, + "learning_rate": 1.2522394581603673e-06, + "loss": 1.4495, + "step": 34357 + }, + { + "epoch": 7.5, + "learning_rate": 1.2521302162988857e-06, + "loss": 1.0097, + "step": 34358 + }, + { + "epoch": 7.5, + "learning_rate": 1.2520209744374045e-06, + "loss": 1.1588, + "step": 34359 + }, + { + "epoch": 7.5, + "learning_rate": 1.2519117325759231e-06, + "loss": 1.6261, + "step": 34360 + }, + { + "epoch": 7.5, + "learning_rate": 1.251802490714442e-06, + "loss": 1.3081, + "step": 34361 + }, + { + "epoch": 7.5, + "learning_rate": 1.2516932488529606e-06, + "loss": 1.0467, + "step": 34362 + }, + { + "epoch": 7.5, + "learning_rate": 1.2515840069914792e-06, + "loss": 0.9774, + "step": 34363 + }, + { + "epoch": 7.5, + "learning_rate": 1.251474765129998e-06, + "loss": 0.9117, + "step": 34364 + }, + { + "epoch": 7.5, + "learning_rate": 1.2513655232685166e-06, + "loss": 2.4296, + "step": 34365 + }, + { + "epoch": 7.5, + "learning_rate": 1.2512562814070352e-06, + "loss": 2.19, + "step": 34366 + }, + { + "epoch": 7.5, + "learning_rate": 1.2511470395455538e-06, + "loss": 0.5373, + "step": 34367 + }, + { + "epoch": 7.5, + "learning_rate": 1.2510377976840727e-06, + "loss": 1.7304, + "step": 34368 + }, + { + "epoch": 7.5, + "learning_rate": 1.2509285558225915e-06, + "loss": 0.7807, + "step": 34369 + }, + { + "epoch": 7.5, + "learning_rate": 1.2508193139611099e-06, + "loss": 1.9903, + "step": 34370 + }, + { + "epoch": 7.5, + "learning_rate": 1.2507100720996287e-06, + "loss": 1.5283, + "step": 34371 + }, + { + "epoch": 7.5, + "learning_rate": 1.2506008302381473e-06, + "loss": 1.3053, + "step": 34372 + }, + { + "epoch": 7.5, + "learning_rate": 1.2504915883766661e-06, + "loss": 1.2851, + "step": 34373 + }, + { + "epoch": 7.5, + "learning_rate": 1.2503823465151845e-06, + "loss": 0.6031, + "step": 34374 + }, + { + "epoch": 7.5, + "learning_rate": 1.2502731046537034e-06, + "loss": 0.9408, + "step": 34375 + }, + { + "epoch": 7.5, + "learning_rate": 1.2501638627922222e-06, + "loss": 0.6441, + "step": 34376 + }, + { + "epoch": 7.5, + "learning_rate": 1.2500546209307408e-06, + "loss": 0.1897, + "step": 34377 + }, + { + "epoch": 7.5, + "learning_rate": 1.2499453790692594e-06, + "loss": 1.2938, + "step": 34378 + }, + { + "epoch": 7.5, + "learning_rate": 1.2498361372077782e-06, + "loss": 1.093, + "step": 34379 + }, + { + "epoch": 7.5, + "learning_rate": 1.2497268953462968e-06, + "loss": 0.1482, + "step": 34380 + }, + { + "epoch": 7.5, + "learning_rate": 1.2496176534848155e-06, + "loss": 2.7225, + "step": 34381 + }, + { + "epoch": 7.5, + "learning_rate": 1.249508411623334e-06, + "loss": 0.935, + "step": 34382 + }, + { + "epoch": 7.5, + "learning_rate": 1.2493991697618529e-06, + "loss": 2.201, + "step": 34383 + }, + { + "epoch": 7.5, + "learning_rate": 1.2492899279003715e-06, + "loss": 1.1532, + "step": 34384 + }, + { + "epoch": 7.5, + "learning_rate": 1.2491806860388903e-06, + "loss": 0.2101, + "step": 34385 + }, + { + "epoch": 7.5, + "learning_rate": 1.249071444177409e-06, + "loss": 2.4054, + "step": 34386 + }, + { + "epoch": 7.5, + "learning_rate": 1.2489622023159275e-06, + "loss": 2.0198, + "step": 34387 + }, + { + "epoch": 7.51, + "learning_rate": 1.2488529604544462e-06, + "loss": 1.4586, + "step": 34388 + }, + { + "epoch": 7.51, + "learning_rate": 1.248743718592965e-06, + "loss": 1.3184, + "step": 34389 + }, + { + "epoch": 7.51, + "learning_rate": 1.2486344767314836e-06, + "loss": 1.606, + "step": 34390 + }, + { + "epoch": 7.51, + "learning_rate": 1.2485252348700024e-06, + "loss": 1.2729, + "step": 34391 + }, + { + "epoch": 7.51, + "learning_rate": 1.248415993008521e-06, + "loss": 0.6387, + "step": 34392 + }, + { + "epoch": 7.51, + "learning_rate": 1.2483067511470396e-06, + "loss": 0.6245, + "step": 34393 + }, + { + "epoch": 7.51, + "learning_rate": 1.2481975092855582e-06, + "loss": 0.5253, + "step": 34394 + }, + { + "epoch": 7.51, + "learning_rate": 1.248088267424077e-06, + "loss": 0.998, + "step": 34395 + }, + { + "epoch": 7.51, + "learning_rate": 1.2479790255625957e-06, + "loss": 1.1818, + "step": 34396 + }, + { + "epoch": 7.51, + "learning_rate": 1.2478697837011143e-06, + "loss": 1.7802, + "step": 34397 + }, + { + "epoch": 7.51, + "learning_rate": 1.2477605418396331e-06, + "loss": 1.0973, + "step": 34398 + }, + { + "epoch": 7.51, + "learning_rate": 1.2476512999781517e-06, + "loss": 0.8394, + "step": 34399 + }, + { + "epoch": 7.51, + "learning_rate": 1.2475420581166703e-06, + "loss": 0.1652, + "step": 34400 + }, + { + "epoch": 7.51, + "learning_rate": 1.2474328162551892e-06, + "loss": 0.2319, + "step": 34401 + }, + { + "epoch": 7.51, + "learning_rate": 1.2473235743937078e-06, + "loss": 0.8534, + "step": 34402 + }, + { + "epoch": 7.51, + "learning_rate": 1.2472143325322264e-06, + "loss": 2.1978, + "step": 34403 + }, + { + "epoch": 7.51, + "learning_rate": 1.2471050906707452e-06, + "loss": 0.5565, + "step": 34404 + }, + { + "epoch": 7.51, + "learning_rate": 1.2469958488092638e-06, + "loss": 0.5429, + "step": 34405 + }, + { + "epoch": 7.51, + "learning_rate": 1.2468866069477824e-06, + "loss": 4.8954, + "step": 34406 + }, + { + "epoch": 7.51, + "learning_rate": 1.2467773650863012e-06, + "loss": 2.0732, + "step": 34407 + }, + { + "epoch": 7.51, + "learning_rate": 1.2466681232248199e-06, + "loss": 0.5671, + "step": 34408 + }, + { + "epoch": 7.51, + "learning_rate": 1.2465588813633385e-06, + "loss": 0.1193, + "step": 34409 + }, + { + "epoch": 7.51, + "learning_rate": 1.246449639501857e-06, + "loss": 0.9592, + "step": 34410 + }, + { + "epoch": 7.51, + "learning_rate": 1.246340397640376e-06, + "loss": 0.3409, + "step": 34411 + }, + { + "epoch": 7.51, + "learning_rate": 1.2462311557788945e-06, + "loss": 2.4819, + "step": 34412 + }, + { + "epoch": 7.51, + "learning_rate": 1.2461219139174133e-06, + "loss": 0.5636, + "step": 34413 + }, + { + "epoch": 7.51, + "learning_rate": 1.246012672055932e-06, + "loss": 0.9395, + "step": 34414 + }, + { + "epoch": 7.51, + "learning_rate": 1.2459034301944506e-06, + "loss": 1.3111, + "step": 34415 + }, + { + "epoch": 7.51, + "learning_rate": 1.2457941883329692e-06, + "loss": 1.519, + "step": 34416 + }, + { + "epoch": 7.51, + "learning_rate": 1.245684946471488e-06, + "loss": 0.8124, + "step": 34417 + }, + { + "epoch": 7.51, + "learning_rate": 1.2455757046100068e-06, + "loss": 1.3265, + "step": 34418 + }, + { + "epoch": 7.51, + "learning_rate": 1.2454664627485254e-06, + "loss": 1.8171, + "step": 34419 + }, + { + "epoch": 7.51, + "learning_rate": 1.245357220887044e-06, + "loss": 0.6834, + "step": 34420 + }, + { + "epoch": 7.51, + "learning_rate": 1.2452479790255626e-06, + "loss": 0.551, + "step": 34421 + }, + { + "epoch": 7.51, + "learning_rate": 1.2451387371640813e-06, + "loss": 0.31, + "step": 34422 + }, + { + "epoch": 7.51, + "learning_rate": 1.2450294953026e-06, + "loss": 0.5148, + "step": 34423 + }, + { + "epoch": 7.51, + "learning_rate": 1.244920253441119e-06, + "loss": 2.0126, + "step": 34424 + }, + { + "epoch": 7.51, + "learning_rate": 1.2448110115796375e-06, + "loss": 0.0824, + "step": 34425 + }, + { + "epoch": 7.51, + "learning_rate": 1.2447017697181561e-06, + "loss": 2.2892, + "step": 34426 + }, + { + "epoch": 7.51, + "learning_rate": 1.2445925278566747e-06, + "loss": 3.5682, + "step": 34427 + }, + { + "epoch": 7.51, + "learning_rate": 1.2444832859951933e-06, + "loss": 0.7202, + "step": 34428 + }, + { + "epoch": 7.51, + "learning_rate": 1.2443740441337122e-06, + "loss": 0.5094, + "step": 34429 + }, + { + "epoch": 7.51, + "learning_rate": 1.2442648022722308e-06, + "loss": 0.4696, + "step": 34430 + }, + { + "epoch": 7.51, + "learning_rate": 1.2441555604107496e-06, + "loss": 1.6587, + "step": 34431 + }, + { + "epoch": 7.51, + "learning_rate": 1.2440463185492682e-06, + "loss": 1.5508, + "step": 34432 + }, + { + "epoch": 7.51, + "learning_rate": 1.2439370766877868e-06, + "loss": 0.6591, + "step": 34433 + }, + { + "epoch": 7.52, + "learning_rate": 1.2438278348263054e-06, + "loss": 0.6199, + "step": 34434 + }, + { + "epoch": 7.52, + "learning_rate": 1.2437185929648243e-06, + "loss": 0.3541, + "step": 34435 + }, + { + "epoch": 7.52, + "learning_rate": 1.2436093511033429e-06, + "loss": 1.2161, + "step": 34436 + }, + { + "epoch": 7.52, + "learning_rate": 1.2435001092418615e-06, + "loss": 1.465, + "step": 34437 + }, + { + "epoch": 7.52, + "learning_rate": 1.2433908673803803e-06, + "loss": 1.5554, + "step": 34438 + }, + { + "epoch": 7.52, + "learning_rate": 1.243281625518899e-06, + "loss": 0.9194, + "step": 34439 + }, + { + "epoch": 7.52, + "learning_rate": 1.2431723836574177e-06, + "loss": 1.2599, + "step": 34440 + }, + { + "epoch": 7.52, + "learning_rate": 1.2430631417959363e-06, + "loss": 2.6199, + "step": 34441 + }, + { + "epoch": 7.52, + "learning_rate": 1.242953899934455e-06, + "loss": 1.615, + "step": 34442 + }, + { + "epoch": 7.52, + "learning_rate": 1.2428446580729736e-06, + "loss": 1.0429, + "step": 34443 + }, + { + "epoch": 7.52, + "learning_rate": 1.2427354162114924e-06, + "loss": 2.6697, + "step": 34444 + }, + { + "epoch": 7.52, + "learning_rate": 1.242626174350011e-06, + "loss": 0.1421, + "step": 34445 + }, + { + "epoch": 7.52, + "learning_rate": 1.2425169324885298e-06, + "loss": 0.934, + "step": 34446 + }, + { + "epoch": 7.52, + "learning_rate": 1.2424076906270484e-06, + "loss": 0.9337, + "step": 34447 + }, + { + "epoch": 7.52, + "learning_rate": 1.242298448765567e-06, + "loss": 1.2498, + "step": 34448 + }, + { + "epoch": 7.52, + "learning_rate": 1.2421892069040857e-06, + "loss": 0.6466, + "step": 34449 + }, + { + "epoch": 7.52, + "learning_rate": 1.2420799650426043e-06, + "loss": 0.9714, + "step": 34450 + }, + { + "epoch": 7.52, + "learning_rate": 1.241970723181123e-06, + "loss": 1.4867, + "step": 34451 + }, + { + "epoch": 7.52, + "learning_rate": 1.241861481319642e-06, + "loss": 2.2668, + "step": 34452 + }, + { + "epoch": 7.52, + "learning_rate": 1.2417522394581605e-06, + "loss": 1.9846, + "step": 34453 + }, + { + "epoch": 7.52, + "learning_rate": 1.2416429975966791e-06, + "loss": 1.4542, + "step": 34454 + }, + { + "epoch": 7.52, + "learning_rate": 1.2415337557351977e-06, + "loss": 0.544, + "step": 34455 + }, + { + "epoch": 7.52, + "learning_rate": 1.2414245138737164e-06, + "loss": 0.6495, + "step": 34456 + }, + { + "epoch": 7.52, + "learning_rate": 1.2413152720122352e-06, + "loss": 1.1198, + "step": 34457 + }, + { + "epoch": 7.52, + "learning_rate": 1.241206030150754e-06, + "loss": 0.8192, + "step": 34458 + }, + { + "epoch": 7.52, + "learning_rate": 1.2410967882892726e-06, + "loss": 1.2289, + "step": 34459 + }, + { + "epoch": 7.52, + "learning_rate": 1.2409875464277912e-06, + "loss": 0.1487, + "step": 34460 + }, + { + "epoch": 7.52, + "learning_rate": 1.2408783045663098e-06, + "loss": 1.0675, + "step": 34461 + }, + { + "epoch": 7.52, + "learning_rate": 1.2407690627048284e-06, + "loss": 3.0235, + "step": 34462 + }, + { + "epoch": 7.52, + "learning_rate": 1.2406598208433473e-06, + "loss": 0.3572, + "step": 34463 + }, + { + "epoch": 7.52, + "learning_rate": 1.240550578981866e-06, + "loss": 0.2296, + "step": 34464 + }, + { + "epoch": 7.52, + "learning_rate": 1.2404413371203847e-06, + "loss": 0.4155, + "step": 34465 + }, + { + "epoch": 7.52, + "learning_rate": 1.2403320952589033e-06, + "loss": 2.4427, + "step": 34466 + }, + { + "epoch": 7.52, + "learning_rate": 1.240222853397422e-06, + "loss": 0.4563, + "step": 34467 + }, + { + "epoch": 7.52, + "learning_rate": 1.2401136115359407e-06, + "loss": 1.368, + "step": 34468 + }, + { + "epoch": 7.52, + "learning_rate": 1.2400043696744594e-06, + "loss": 1.8075, + "step": 34469 + }, + { + "epoch": 7.52, + "learning_rate": 1.239895127812978e-06, + "loss": 0.836, + "step": 34470 + }, + { + "epoch": 7.52, + "learning_rate": 1.2397858859514968e-06, + "loss": 2.2024, + "step": 34471 + }, + { + "epoch": 7.52, + "learning_rate": 1.2396766440900154e-06, + "loss": 0.6907, + "step": 34472 + }, + { + "epoch": 7.52, + "learning_rate": 1.239567402228534e-06, + "loss": 0.0823, + "step": 34473 + }, + { + "epoch": 7.52, + "learning_rate": 1.2394581603670528e-06, + "loss": 2.5015, + "step": 34474 + }, + { + "epoch": 7.52, + "learning_rate": 1.2393489185055714e-06, + "loss": 1.0425, + "step": 34475 + }, + { + "epoch": 7.52, + "learning_rate": 1.23923967664409e-06, + "loss": 1.7608, + "step": 34476 + }, + { + "epoch": 7.52, + "learning_rate": 1.2391304347826089e-06, + "loss": 0.5793, + "step": 34477 + }, + { + "epoch": 7.52, + "learning_rate": 1.2390211929211275e-06, + "loss": 0.7388, + "step": 34478 + }, + { + "epoch": 7.52, + "learning_rate": 1.238911951059646e-06, + "loss": 1.7863, + "step": 34479 + }, + { + "epoch": 7.53, + "learning_rate": 1.238802709198165e-06, + "loss": 2.7562, + "step": 34480 + }, + { + "epoch": 7.53, + "learning_rate": 1.2386934673366835e-06, + "loss": 0.3577, + "step": 34481 + }, + { + "epoch": 7.53, + "learning_rate": 1.2385842254752022e-06, + "loss": 2.2873, + "step": 34482 + }, + { + "epoch": 7.53, + "learning_rate": 1.2384749836137208e-06, + "loss": 1.0746, + "step": 34483 + }, + { + "epoch": 7.53, + "learning_rate": 1.2383657417522396e-06, + "loss": 1.8582, + "step": 34484 + }, + { + "epoch": 7.53, + "learning_rate": 1.2382564998907582e-06, + "loss": 0.7932, + "step": 34485 + }, + { + "epoch": 7.53, + "learning_rate": 1.238147258029277e-06, + "loss": 1.024, + "step": 34486 + }, + { + "epoch": 7.53, + "learning_rate": 1.2380380161677956e-06, + "loss": 0.3269, + "step": 34487 + }, + { + "epoch": 7.53, + "learning_rate": 1.2379287743063142e-06, + "loss": 1.0872, + "step": 34488 + }, + { + "epoch": 7.53, + "learning_rate": 1.2378195324448329e-06, + "loss": 0.6267, + "step": 34489 + }, + { + "epoch": 7.53, + "learning_rate": 1.2377102905833517e-06, + "loss": 1.0925, + "step": 34490 + }, + { + "epoch": 7.53, + "learning_rate": 1.2376010487218703e-06, + "loss": 0.9038, + "step": 34491 + }, + { + "epoch": 7.53, + "learning_rate": 1.2374918068603891e-06, + "loss": 1.3922, + "step": 34492 + }, + { + "epoch": 7.53, + "learning_rate": 1.2373825649989077e-06, + "loss": 1.0673, + "step": 34493 + }, + { + "epoch": 7.53, + "learning_rate": 1.2372733231374263e-06, + "loss": 0.8824, + "step": 34494 + }, + { + "epoch": 7.53, + "learning_rate": 1.237164081275945e-06, + "loss": 0.5353, + "step": 34495 + }, + { + "epoch": 7.53, + "learning_rate": 1.2370548394144638e-06, + "loss": 1.9056, + "step": 34496 + }, + { + "epoch": 7.53, + "learning_rate": 1.2369455975529824e-06, + "loss": 2.7615, + "step": 34497 + }, + { + "epoch": 7.53, + "learning_rate": 1.2368363556915012e-06, + "loss": 0.1259, + "step": 34498 + }, + { + "epoch": 7.53, + "learning_rate": 1.2367271138300198e-06, + "loss": 0.1474, + "step": 34499 + }, + { + "epoch": 7.53, + "learning_rate": 1.2366178719685384e-06, + "loss": 0.7487, + "step": 34500 + }, + { + "epoch": 7.53, + "learning_rate": 1.236508630107057e-06, + "loss": 0.3913, + "step": 34501 + }, + { + "epoch": 7.53, + "learning_rate": 1.2363993882455759e-06, + "loss": 1.7454, + "step": 34502 + }, + { + "epoch": 7.53, + "learning_rate": 1.2362901463840945e-06, + "loss": 2.5899, + "step": 34503 + }, + { + "epoch": 7.53, + "learning_rate": 1.2361809045226133e-06, + "loss": 0.4811, + "step": 34504 + }, + { + "epoch": 7.53, + "learning_rate": 1.236071662661132e-06, + "loss": 1.146, + "step": 34505 + }, + { + "epoch": 7.53, + "learning_rate": 1.2359624207996505e-06, + "loss": 1.1072, + "step": 34506 + }, + { + "epoch": 7.53, + "learning_rate": 1.2358531789381691e-06, + "loss": 1.584, + "step": 34507 + }, + { + "epoch": 7.53, + "learning_rate": 1.235743937076688e-06, + "loss": 1.6804, + "step": 34508 + }, + { + "epoch": 7.53, + "learning_rate": 1.2356346952152066e-06, + "loss": 0.8119, + "step": 34509 + }, + { + "epoch": 7.53, + "learning_rate": 1.2355254533537254e-06, + "loss": 1.6111, + "step": 34510 + }, + { + "epoch": 7.53, + "learning_rate": 1.235416211492244e-06, + "loss": 1.5807, + "step": 34511 + }, + { + "epoch": 7.53, + "learning_rate": 1.2353069696307626e-06, + "loss": 2.1065, + "step": 34512 + }, + { + "epoch": 7.53, + "learning_rate": 1.2351977277692812e-06, + "loss": 2.2873, + "step": 34513 + }, + { + "epoch": 7.53, + "learning_rate": 1.2350884859078e-06, + "loss": 1.6264, + "step": 34514 + }, + { + "epoch": 7.53, + "learning_rate": 1.2349792440463186e-06, + "loss": 0.803, + "step": 34515 + }, + { + "epoch": 7.53, + "learning_rate": 1.2348700021848373e-06, + "loss": 0.7925, + "step": 34516 + }, + { + "epoch": 7.53, + "learning_rate": 1.234760760323356e-06, + "loss": 1.4645, + "step": 34517 + }, + { + "epoch": 7.53, + "learning_rate": 1.2346515184618747e-06, + "loss": 0.2182, + "step": 34518 + }, + { + "epoch": 7.53, + "learning_rate": 1.2345422766003933e-06, + "loss": 1.0148, + "step": 34519 + }, + { + "epoch": 7.53, + "learning_rate": 1.2344330347389121e-06, + "loss": 0.6904, + "step": 34520 + }, + { + "epoch": 7.53, + "learning_rate": 1.2343237928774307e-06, + "loss": 0.6782, + "step": 34521 + }, + { + "epoch": 7.53, + "learning_rate": 1.2342145510159493e-06, + "loss": 2.0964, + "step": 34522 + }, + { + "epoch": 7.53, + "learning_rate": 1.2341053091544682e-06, + "loss": 1.6878, + "step": 34523 + }, + { + "epoch": 7.53, + "learning_rate": 1.2339960672929868e-06, + "loss": 0.9304, + "step": 34524 + }, + { + "epoch": 7.53, + "learning_rate": 1.2338868254315054e-06, + "loss": 1.2513, + "step": 34525 + }, + { + "epoch": 7.54, + "learning_rate": 1.2337775835700242e-06, + "loss": 0.9702, + "step": 34526 + }, + { + "epoch": 7.54, + "learning_rate": 1.2336683417085428e-06, + "loss": 2.0501, + "step": 34527 + }, + { + "epoch": 7.54, + "learning_rate": 1.2335590998470614e-06, + "loss": 0.871, + "step": 34528 + }, + { + "epoch": 7.54, + "learning_rate": 1.23344985798558e-06, + "loss": 0.9021, + "step": 34529 + }, + { + "epoch": 7.54, + "learning_rate": 1.2333406161240989e-06, + "loss": 0.7507, + "step": 34530 + }, + { + "epoch": 7.54, + "learning_rate": 1.2332313742626175e-06, + "loss": 0.2176, + "step": 34531 + }, + { + "epoch": 7.54, + "learning_rate": 1.2331221324011363e-06, + "loss": 0.8851, + "step": 34532 + }, + { + "epoch": 7.54, + "learning_rate": 1.233012890539655e-06, + "loss": 1.3394, + "step": 34533 + }, + { + "epoch": 7.54, + "learning_rate": 1.2329036486781735e-06, + "loss": 1.0012, + "step": 34534 + }, + { + "epoch": 7.54, + "learning_rate": 1.2327944068166921e-06, + "loss": 0.55, + "step": 34535 + }, + { + "epoch": 7.54, + "learning_rate": 1.232685164955211e-06, + "loss": 1.0235, + "step": 34536 + }, + { + "epoch": 7.54, + "learning_rate": 1.2325759230937296e-06, + "loss": 1.145, + "step": 34537 + }, + { + "epoch": 7.54, + "learning_rate": 1.2324666812322484e-06, + "loss": 1.4394, + "step": 34538 + }, + { + "epoch": 7.54, + "learning_rate": 1.232357439370767e-06, + "loss": 0.8038, + "step": 34539 + }, + { + "epoch": 7.54, + "learning_rate": 1.2322481975092856e-06, + "loss": 1.2202, + "step": 34540 + }, + { + "epoch": 7.54, + "learning_rate": 1.2321389556478042e-06, + "loss": 0.3484, + "step": 34541 + }, + { + "epoch": 7.54, + "learning_rate": 1.232029713786323e-06, + "loss": 0.2692, + "step": 34542 + }, + { + "epoch": 7.54, + "learning_rate": 1.2319204719248419e-06, + "loss": 3.5893, + "step": 34543 + }, + { + "epoch": 7.54, + "learning_rate": 1.2318112300633605e-06, + "loss": 1.7365, + "step": 34544 + }, + { + "epoch": 7.54, + "learning_rate": 1.231701988201879e-06, + "loss": 0.5067, + "step": 34545 + }, + { + "epoch": 7.54, + "learning_rate": 1.2315927463403977e-06, + "loss": 0.8167, + "step": 34546 + }, + { + "epoch": 7.54, + "learning_rate": 1.2314835044789163e-06, + "loss": 1.0952, + "step": 34547 + }, + { + "epoch": 7.54, + "learning_rate": 1.2313742626174351e-06, + "loss": 1.7325, + "step": 34548 + }, + { + "epoch": 7.54, + "learning_rate": 1.2312650207559537e-06, + "loss": 0.4754, + "step": 34549 + }, + { + "epoch": 7.54, + "learning_rate": 1.2311557788944726e-06, + "loss": 0.9653, + "step": 34550 + }, + { + "epoch": 7.54, + "learning_rate": 1.2310465370329912e-06, + "loss": 0.926, + "step": 34551 + }, + { + "epoch": 7.54, + "learning_rate": 1.2309372951715098e-06, + "loss": 0.1284, + "step": 34552 + }, + { + "epoch": 7.54, + "learning_rate": 1.2308280533100284e-06, + "loss": 2.3169, + "step": 34553 + }, + { + "epoch": 7.54, + "learning_rate": 1.2307188114485472e-06, + "loss": 1.6652, + "step": 34554 + }, + { + "epoch": 7.54, + "learning_rate": 1.2306095695870658e-06, + "loss": 1.968, + "step": 34555 + }, + { + "epoch": 7.54, + "learning_rate": 1.2305003277255844e-06, + "loss": 0.1238, + "step": 34556 + }, + { + "epoch": 7.54, + "learning_rate": 1.2303910858641033e-06, + "loss": 1.9768, + "step": 34557 + }, + { + "epoch": 7.54, + "learning_rate": 1.2302818440026219e-06, + "loss": 0.8208, + "step": 34558 + }, + { + "epoch": 7.54, + "learning_rate": 1.2301726021411405e-06, + "loss": 1.0779, + "step": 34559 + }, + { + "epoch": 7.54, + "learning_rate": 1.2300633602796593e-06, + "loss": 0.8546, + "step": 34560 + }, + { + "epoch": 7.54, + "learning_rate": 1.229954118418178e-06, + "loss": 1.9132, + "step": 34561 + }, + { + "epoch": 7.54, + "learning_rate": 1.2298448765566965e-06, + "loss": 2.0856, + "step": 34562 + }, + { + "epoch": 7.54, + "learning_rate": 1.2297356346952154e-06, + "loss": 1.4186, + "step": 34563 + }, + { + "epoch": 7.54, + "learning_rate": 1.229626392833734e-06, + "loss": 0.8899, + "step": 34564 + }, + { + "epoch": 7.54, + "learning_rate": 1.2295171509722528e-06, + "loss": 1.4463, + "step": 34565 + }, + { + "epoch": 7.54, + "learning_rate": 1.2294079091107714e-06, + "loss": 1.769, + "step": 34566 + }, + { + "epoch": 7.54, + "learning_rate": 1.22929866724929e-06, + "loss": 2.2754, + "step": 34567 + }, + { + "epoch": 7.54, + "learning_rate": 1.2291894253878086e-06, + "loss": 1.0898, + "step": 34568 + }, + { + "epoch": 7.54, + "learning_rate": 1.2290801835263272e-06, + "loss": 0.7369, + "step": 34569 + }, + { + "epoch": 7.54, + "learning_rate": 1.228970941664846e-06, + "loss": 2.2795, + "step": 34570 + }, + { + "epoch": 7.54, + "learning_rate": 1.2288616998033649e-06, + "loss": 1.2016, + "step": 34571 + }, + { + "epoch": 7.55, + "learning_rate": 1.2287524579418835e-06, + "loss": 1.6598, + "step": 34572 + }, + { + "epoch": 7.55, + "learning_rate": 1.228643216080402e-06, + "loss": 1.0902, + "step": 34573 + }, + { + "epoch": 7.55, + "learning_rate": 1.2285339742189207e-06, + "loss": 0.8228, + "step": 34574 + }, + { + "epoch": 7.55, + "learning_rate": 1.2284247323574393e-06, + "loss": 2.249, + "step": 34575 + }, + { + "epoch": 7.55, + "learning_rate": 1.2283154904959581e-06, + "loss": 1.2266, + "step": 34576 + }, + { + "epoch": 7.55, + "learning_rate": 1.228206248634477e-06, + "loss": 0.7674, + "step": 34577 + }, + { + "epoch": 7.55, + "learning_rate": 1.2280970067729956e-06, + "loss": 0.4625, + "step": 34578 + }, + { + "epoch": 7.55, + "learning_rate": 1.2279877649115142e-06, + "loss": 0.8361, + "step": 34579 + }, + { + "epoch": 7.55, + "learning_rate": 1.2278785230500328e-06, + "loss": 0.9163, + "step": 34580 + }, + { + "epoch": 7.55, + "learning_rate": 1.2277692811885514e-06, + "loss": 1.17, + "step": 34581 + }, + { + "epoch": 7.55, + "learning_rate": 1.2276600393270702e-06, + "loss": 1.0355, + "step": 34582 + }, + { + "epoch": 7.55, + "learning_rate": 1.227550797465589e-06, + "loss": 1.0898, + "step": 34583 + }, + { + "epoch": 7.55, + "learning_rate": 1.2274415556041077e-06, + "loss": 0.8454, + "step": 34584 + }, + { + "epoch": 7.55, + "learning_rate": 1.2273323137426263e-06, + "loss": 1.4257, + "step": 34585 + }, + { + "epoch": 7.55, + "learning_rate": 1.2272230718811449e-06, + "loss": 0.7803, + "step": 34586 + }, + { + "epoch": 7.55, + "learning_rate": 1.2271138300196637e-06, + "loss": 0.6841, + "step": 34587 + }, + { + "epoch": 7.55, + "learning_rate": 1.2270045881581823e-06, + "loss": 0.2585, + "step": 34588 + }, + { + "epoch": 7.55, + "learning_rate": 1.226895346296701e-06, + "loss": 0.9645, + "step": 34589 + }, + { + "epoch": 7.55, + "learning_rate": 1.2267861044352198e-06, + "loss": 0.3082, + "step": 34590 + }, + { + "epoch": 7.55, + "learning_rate": 1.2266768625737384e-06, + "loss": 0.481, + "step": 34591 + }, + { + "epoch": 7.55, + "learning_rate": 1.226567620712257e-06, + "loss": 0.8344, + "step": 34592 + }, + { + "epoch": 7.55, + "learning_rate": 1.2264583788507758e-06, + "loss": 1.3245, + "step": 34593 + }, + { + "epoch": 7.55, + "learning_rate": 1.2263491369892944e-06, + "loss": 0.2305, + "step": 34594 + }, + { + "epoch": 7.55, + "learning_rate": 1.226239895127813e-06, + "loss": 2.8524, + "step": 34595 + }, + { + "epoch": 7.55, + "learning_rate": 1.2261306532663318e-06, + "loss": 1.0788, + "step": 34596 + }, + { + "epoch": 7.55, + "learning_rate": 1.2260214114048505e-06, + "loss": 1.3576, + "step": 34597 + }, + { + "epoch": 7.55, + "learning_rate": 1.225912169543369e-06, + "loss": 0.96, + "step": 34598 + }, + { + "epoch": 7.55, + "learning_rate": 1.225802927681888e-06, + "loss": 0.6578, + "step": 34599 + }, + { + "epoch": 7.55, + "learning_rate": 1.2256936858204065e-06, + "loss": 0.2807, + "step": 34600 + }, + { + "epoch": 7.55, + "learning_rate": 1.2255844439589251e-06, + "loss": 1.5405, + "step": 34601 + }, + { + "epoch": 7.55, + "learning_rate": 1.2254752020974437e-06, + "loss": 0.6945, + "step": 34602 + }, + { + "epoch": 7.55, + "learning_rate": 1.2253659602359626e-06, + "loss": 1.9932, + "step": 34603 + }, + { + "epoch": 7.55, + "learning_rate": 1.2252567183744812e-06, + "loss": 0.7423, + "step": 34604 + }, + { + "epoch": 7.55, + "learning_rate": 1.225147476513e-06, + "loss": 1.2483, + "step": 34605 + }, + { + "epoch": 7.55, + "learning_rate": 1.2250382346515186e-06, + "loss": 0.9868, + "step": 34606 + }, + { + "epoch": 7.55, + "learning_rate": 1.2249289927900372e-06, + "loss": 0.3958, + "step": 34607 + }, + { + "epoch": 7.55, + "learning_rate": 1.2248197509285558e-06, + "loss": 0.343, + "step": 34608 + }, + { + "epoch": 7.55, + "learning_rate": 1.2247105090670746e-06, + "loss": 0.7176, + "step": 34609 + }, + { + "epoch": 7.55, + "learning_rate": 1.2246012672055933e-06, + "loss": 0.5239, + "step": 34610 + }, + { + "epoch": 7.55, + "learning_rate": 1.224492025344112e-06, + "loss": 2.2623, + "step": 34611 + }, + { + "epoch": 7.55, + "learning_rate": 1.2243827834826307e-06, + "loss": 0.9854, + "step": 34612 + }, + { + "epoch": 7.55, + "learning_rate": 1.2242735416211493e-06, + "loss": 0.6408, + "step": 34613 + }, + { + "epoch": 7.55, + "learning_rate": 1.224164299759668e-06, + "loss": 0.7146, + "step": 34614 + }, + { + "epoch": 7.55, + "learning_rate": 1.2240550578981867e-06, + "loss": 2.2375, + "step": 34615 + }, + { + "epoch": 7.55, + "learning_rate": 1.2239458160367053e-06, + "loss": 1.2714, + "step": 34616 + }, + { + "epoch": 7.55, + "learning_rate": 1.2238365741752242e-06, + "loss": 0.684, + "step": 34617 + }, + { + "epoch": 7.56, + "learning_rate": 1.2237273323137428e-06, + "loss": 0.1089, + "step": 34618 + }, + { + "epoch": 7.56, + "learning_rate": 1.2236180904522614e-06, + "loss": 0.7583, + "step": 34619 + }, + { + "epoch": 7.56, + "learning_rate": 1.22350884859078e-06, + "loss": 0.3765, + "step": 34620 + }, + { + "epoch": 7.56, + "learning_rate": 1.2233996067292988e-06, + "loss": 1.6191, + "step": 34621 + }, + { + "epoch": 7.56, + "learning_rate": 1.2232903648678174e-06, + "loss": 0.1521, + "step": 34622 + }, + { + "epoch": 7.56, + "learning_rate": 1.2231811230063363e-06, + "loss": 0.6026, + "step": 34623 + }, + { + "epoch": 7.56, + "learning_rate": 1.2230718811448549e-06, + "loss": 0.8705, + "step": 34624 + }, + { + "epoch": 7.56, + "learning_rate": 1.2229626392833735e-06, + "loss": 1.229, + "step": 34625 + }, + { + "epoch": 7.56, + "learning_rate": 1.222853397421892e-06, + "loss": 1.1432, + "step": 34626 + }, + { + "epoch": 7.56, + "learning_rate": 1.222744155560411e-06, + "loss": 0.0632, + "step": 34627 + }, + { + "epoch": 7.56, + "learning_rate": 1.2226349136989295e-06, + "loss": 0.4316, + "step": 34628 + }, + { + "epoch": 7.56, + "learning_rate": 1.2225256718374483e-06, + "loss": 1.1513, + "step": 34629 + }, + { + "epoch": 7.56, + "learning_rate": 1.222416429975967e-06, + "loss": 4.2604, + "step": 34630 + }, + { + "epoch": 7.56, + "learning_rate": 1.2223071881144856e-06, + "loss": 1.5064, + "step": 34631 + }, + { + "epoch": 7.56, + "learning_rate": 1.2221979462530042e-06, + "loss": 0.8739, + "step": 34632 + }, + { + "epoch": 7.56, + "learning_rate": 1.222088704391523e-06, + "loss": 0.8955, + "step": 34633 + }, + { + "epoch": 7.56, + "learning_rate": 1.2219794625300416e-06, + "loss": 0.8392, + "step": 34634 + }, + { + "epoch": 7.56, + "learning_rate": 1.2218702206685602e-06, + "loss": 1.3511, + "step": 34635 + }, + { + "epoch": 7.56, + "learning_rate": 1.221760978807079e-06, + "loss": 2.1187, + "step": 34636 + }, + { + "epoch": 7.56, + "learning_rate": 1.2216517369455977e-06, + "loss": 1.7396, + "step": 34637 + }, + { + "epoch": 7.56, + "learning_rate": 1.2215424950841163e-06, + "loss": 1.2205, + "step": 34638 + }, + { + "epoch": 7.56, + "learning_rate": 1.221433253222635e-06, + "loss": 0.2914, + "step": 34639 + }, + { + "epoch": 7.56, + "learning_rate": 1.2213240113611537e-06, + "loss": 1.1225, + "step": 34640 + }, + { + "epoch": 7.56, + "learning_rate": 1.2212147694996723e-06, + "loss": 1.1947, + "step": 34641 + }, + { + "epoch": 7.56, + "learning_rate": 1.2211055276381911e-06, + "loss": 1.9496, + "step": 34642 + }, + { + "epoch": 7.56, + "learning_rate": 1.2209962857767097e-06, + "loss": 1.0155, + "step": 34643 + }, + { + "epoch": 7.56, + "learning_rate": 1.2208870439152284e-06, + "loss": 0.3698, + "step": 34644 + }, + { + "epoch": 7.56, + "learning_rate": 1.2207778020537472e-06, + "loss": 0.6976, + "step": 34645 + }, + { + "epoch": 7.56, + "learning_rate": 1.2206685601922658e-06, + "loss": 1.67, + "step": 34646 + }, + { + "epoch": 7.56, + "learning_rate": 1.2205593183307844e-06, + "loss": 0.3459, + "step": 34647 + }, + { + "epoch": 7.56, + "learning_rate": 1.220450076469303e-06, + "loss": 0.0517, + "step": 34648 + }, + { + "epoch": 7.56, + "learning_rate": 1.2203408346078218e-06, + "loss": 0.807, + "step": 34649 + }, + { + "epoch": 7.56, + "learning_rate": 1.2202315927463404e-06, + "loss": 0.2915, + "step": 34650 + }, + { + "epoch": 7.56, + "learning_rate": 1.2201223508848593e-06, + "loss": 2.6064, + "step": 34651 + }, + { + "epoch": 7.56, + "learning_rate": 1.2200131090233779e-06, + "loss": 2.3478, + "step": 34652 + }, + { + "epoch": 7.56, + "learning_rate": 1.2199038671618965e-06, + "loss": 0.6033, + "step": 34653 + }, + { + "epoch": 7.56, + "learning_rate": 1.219794625300415e-06, + "loss": 1.4208, + "step": 34654 + }, + { + "epoch": 7.56, + "learning_rate": 1.219685383438934e-06, + "loss": 0.0909, + "step": 34655 + }, + { + "epoch": 7.56, + "learning_rate": 1.2195761415774525e-06, + "loss": 2.801, + "step": 34656 + }, + { + "epoch": 7.56, + "learning_rate": 1.2194668997159714e-06, + "loss": 2.8811, + "step": 34657 + }, + { + "epoch": 7.56, + "learning_rate": 1.21935765785449e-06, + "loss": 0.1926, + "step": 34658 + }, + { + "epoch": 7.56, + "learning_rate": 1.2192484159930086e-06, + "loss": 1.0705, + "step": 34659 + }, + { + "epoch": 7.56, + "learning_rate": 1.2191391741315272e-06, + "loss": 2.666, + "step": 34660 + }, + { + "epoch": 7.56, + "learning_rate": 1.219029932270046e-06, + "loss": 1.4955, + "step": 34661 + }, + { + "epoch": 7.56, + "learning_rate": 1.2189206904085646e-06, + "loss": 0.1083, + "step": 34662 + }, + { + "epoch": 7.57, + "learning_rate": 1.2188114485470834e-06, + "loss": 1.5963, + "step": 34663 + }, + { + "epoch": 7.57, + "learning_rate": 1.218702206685602e-06, + "loss": 1.8963, + "step": 34664 + }, + { + "epoch": 7.57, + "learning_rate": 1.2185929648241207e-06, + "loss": 0.4983, + "step": 34665 + }, + { + "epoch": 7.57, + "learning_rate": 1.2184837229626393e-06, + "loss": 0.8955, + "step": 34666 + }, + { + "epoch": 7.57, + "learning_rate": 1.218374481101158e-06, + "loss": 1.0677, + "step": 34667 + }, + { + "epoch": 7.57, + "learning_rate": 1.2182652392396767e-06, + "loss": 3.3217, + "step": 34668 + }, + { + "epoch": 7.57, + "learning_rate": 1.2181559973781955e-06, + "loss": 3.0995, + "step": 34669 + }, + { + "epoch": 7.57, + "learning_rate": 1.2180467555167141e-06, + "loss": 0.5725, + "step": 34670 + }, + { + "epoch": 7.57, + "learning_rate": 1.2179375136552328e-06, + "loss": 3.4057, + "step": 34671 + }, + { + "epoch": 7.57, + "learning_rate": 1.2178282717937514e-06, + "loss": 1.2862, + "step": 34672 + }, + { + "epoch": 7.57, + "learning_rate": 1.2177190299322702e-06, + "loss": 0.7386, + "step": 34673 + }, + { + "epoch": 7.57, + "learning_rate": 1.2176097880707888e-06, + "loss": 0.1158, + "step": 34674 + }, + { + "epoch": 7.57, + "learning_rate": 1.2175005462093074e-06, + "loss": 2.3786, + "step": 34675 + }, + { + "epoch": 7.57, + "learning_rate": 1.2173913043478262e-06, + "loss": 1.5497, + "step": 34676 + }, + { + "epoch": 7.57, + "learning_rate": 1.2172820624863448e-06, + "loss": 1.1676, + "step": 34677 + }, + { + "epoch": 7.57, + "learning_rate": 1.2171728206248635e-06, + "loss": 0.7565, + "step": 34678 + }, + { + "epoch": 7.57, + "learning_rate": 1.2170635787633823e-06, + "loss": 0.9831, + "step": 34679 + }, + { + "epoch": 7.57, + "learning_rate": 1.2169543369019009e-06, + "loss": 1.7717, + "step": 34680 + }, + { + "epoch": 7.57, + "learning_rate": 1.2168450950404195e-06, + "loss": 2.1473, + "step": 34681 + }, + { + "epoch": 7.57, + "learning_rate": 1.2167358531789383e-06, + "loss": 0.1385, + "step": 34682 + }, + { + "epoch": 7.57, + "learning_rate": 1.216626611317457e-06, + "loss": 2.6867, + "step": 34683 + }, + { + "epoch": 7.57, + "learning_rate": 1.2165173694559755e-06, + "loss": 1.6362, + "step": 34684 + }, + { + "epoch": 7.57, + "learning_rate": 1.2164081275944944e-06, + "loss": 2.0186, + "step": 34685 + }, + { + "epoch": 7.57, + "learning_rate": 1.216298885733013e-06, + "loss": 0.8113, + "step": 34686 + }, + { + "epoch": 7.57, + "learning_rate": 1.2161896438715316e-06, + "loss": 2.8178, + "step": 34687 + }, + { + "epoch": 7.57, + "learning_rate": 1.2160804020100502e-06, + "loss": 1.172, + "step": 34688 + }, + { + "epoch": 7.57, + "learning_rate": 1.215971160148569e-06, + "loss": 1.3205, + "step": 34689 + }, + { + "epoch": 7.57, + "learning_rate": 1.2158619182870878e-06, + "loss": 0.9129, + "step": 34690 + }, + { + "epoch": 7.57, + "learning_rate": 1.2157526764256065e-06, + "loss": 0.5921, + "step": 34691 + }, + { + "epoch": 7.57, + "learning_rate": 1.215643434564125e-06, + "loss": 1.0306, + "step": 34692 + }, + { + "epoch": 7.57, + "learning_rate": 1.2155341927026437e-06, + "loss": 3.3788, + "step": 34693 + }, + { + "epoch": 7.57, + "learning_rate": 1.2154249508411623e-06, + "loss": 0.7532, + "step": 34694 + }, + { + "epoch": 7.57, + "learning_rate": 1.2153157089796811e-06, + "loss": 1.4443, + "step": 34695 + }, + { + "epoch": 7.57, + "learning_rate": 1.2152064671182e-06, + "loss": 0.265, + "step": 34696 + }, + { + "epoch": 7.57, + "learning_rate": 1.2150972252567185e-06, + "loss": 1.4059, + "step": 34697 + }, + { + "epoch": 7.57, + "learning_rate": 1.2149879833952372e-06, + "loss": 0.5368, + "step": 34698 + }, + { + "epoch": 7.57, + "learning_rate": 1.2148787415337558e-06, + "loss": 0.5894, + "step": 34699 + }, + { + "epoch": 7.57, + "learning_rate": 1.2147694996722744e-06, + "loss": 1.3396, + "step": 34700 + }, + { + "epoch": 7.57, + "learning_rate": 1.2146602578107932e-06, + "loss": 1.074, + "step": 34701 + }, + { + "epoch": 7.57, + "learning_rate": 1.214551015949312e-06, + "loss": 0.6011, + "step": 34702 + }, + { + "epoch": 7.57, + "learning_rate": 1.2144417740878306e-06, + "loss": 0.2065, + "step": 34703 + }, + { + "epoch": 7.57, + "learning_rate": 1.2143325322263492e-06, + "loss": 0.1338, + "step": 34704 + }, + { + "epoch": 7.57, + "learning_rate": 1.2142232903648679e-06, + "loss": 2.3433, + "step": 34705 + }, + { + "epoch": 7.57, + "learning_rate": 1.2141140485033865e-06, + "loss": 0.7942, + "step": 34706 + }, + { + "epoch": 7.57, + "learning_rate": 1.2140048066419053e-06, + "loss": 2.0124, + "step": 34707 + }, + { + "epoch": 7.57, + "learning_rate": 1.213895564780424e-06, + "loss": 0.815, + "step": 34708 + }, + { + "epoch": 7.58, + "learning_rate": 1.2137863229189427e-06, + "loss": 0.9183, + "step": 34709 + }, + { + "epoch": 7.58, + "learning_rate": 1.2136770810574613e-06, + "loss": 1.6436, + "step": 34710 + }, + { + "epoch": 7.58, + "learning_rate": 1.21356783919598e-06, + "loss": 1.159, + "step": 34711 + }, + { + "epoch": 7.58, + "learning_rate": 1.2134585973344988e-06, + "loss": 0.6298, + "step": 34712 + }, + { + "epoch": 7.58, + "learning_rate": 1.2133493554730174e-06, + "loss": 1.4282, + "step": 34713 + }, + { + "epoch": 7.58, + "learning_rate": 1.213240113611536e-06, + "loss": 0.7969, + "step": 34714 + }, + { + "epoch": 7.58, + "learning_rate": 1.2131308717500548e-06, + "loss": 2.8802, + "step": 34715 + }, + { + "epoch": 7.58, + "learning_rate": 1.2130216298885734e-06, + "loss": 2.901, + "step": 34716 + }, + { + "epoch": 7.58, + "learning_rate": 1.212912388027092e-06, + "loss": 0.1348, + "step": 34717 + }, + { + "epoch": 7.58, + "learning_rate": 1.2128031461656109e-06, + "loss": 0.7228, + "step": 34718 + }, + { + "epoch": 7.58, + "learning_rate": 1.2126939043041295e-06, + "loss": 1.1625, + "step": 34719 + }, + { + "epoch": 7.58, + "learning_rate": 1.212584662442648e-06, + "loss": 0.7659, + "step": 34720 + }, + { + "epoch": 7.58, + "learning_rate": 1.2124754205811667e-06, + "loss": 1.5341, + "step": 34721 + }, + { + "epoch": 7.58, + "learning_rate": 1.2123661787196855e-06, + "loss": 2.3805, + "step": 34722 + }, + { + "epoch": 7.58, + "learning_rate": 1.2122569368582041e-06, + "loss": 0.9488, + "step": 34723 + }, + { + "epoch": 7.58, + "learning_rate": 1.212147694996723e-06, + "loss": 1.4528, + "step": 34724 + }, + { + "epoch": 7.58, + "learning_rate": 1.2120384531352416e-06, + "loss": 0.6286, + "step": 34725 + }, + { + "epoch": 7.58, + "learning_rate": 1.2119292112737602e-06, + "loss": 0.372, + "step": 34726 + }, + { + "epoch": 7.58, + "learning_rate": 1.2118199694122788e-06, + "loss": 0.7444, + "step": 34727 + }, + { + "epoch": 7.58, + "learning_rate": 1.2117107275507976e-06, + "loss": 0.7372, + "step": 34728 + }, + { + "epoch": 7.58, + "learning_rate": 1.2116014856893162e-06, + "loss": 1.3283, + "step": 34729 + }, + { + "epoch": 7.58, + "learning_rate": 1.211492243827835e-06, + "loss": 0.1979, + "step": 34730 + }, + { + "epoch": 7.58, + "learning_rate": 1.2113830019663537e-06, + "loss": 1.1026, + "step": 34731 + }, + { + "epoch": 7.58, + "learning_rate": 1.2112737601048723e-06, + "loss": 2.2934, + "step": 34732 + }, + { + "epoch": 7.58, + "learning_rate": 1.2111645182433909e-06, + "loss": 2.4251, + "step": 34733 + }, + { + "epoch": 7.58, + "learning_rate": 1.2110552763819095e-06, + "loss": 0.4906, + "step": 34734 + }, + { + "epoch": 7.58, + "learning_rate": 1.2109460345204283e-06, + "loss": 2.0886, + "step": 34735 + }, + { + "epoch": 7.58, + "learning_rate": 1.2108367926589471e-06, + "loss": 1.8252, + "step": 34736 + }, + { + "epoch": 7.58, + "learning_rate": 1.2107275507974657e-06, + "loss": 0.7995, + "step": 34737 + }, + { + "epoch": 7.58, + "learning_rate": 1.2106183089359844e-06, + "loss": 0.1416, + "step": 34738 + }, + { + "epoch": 7.58, + "learning_rate": 1.210509067074503e-06, + "loss": 1.82, + "step": 34739 + }, + { + "epoch": 7.58, + "learning_rate": 1.2103998252130218e-06, + "loss": 3.6387, + "step": 34740 + }, + { + "epoch": 7.58, + "learning_rate": 1.2102905833515404e-06, + "loss": 1.3292, + "step": 34741 + }, + { + "epoch": 7.58, + "learning_rate": 1.2101813414900592e-06, + "loss": 0.6376, + "step": 34742 + }, + { + "epoch": 7.58, + "learning_rate": 1.2100720996285778e-06, + "loss": 1.2561, + "step": 34743 + }, + { + "epoch": 7.58, + "learning_rate": 1.2099628577670964e-06, + "loss": 0.4888, + "step": 34744 + }, + { + "epoch": 7.58, + "learning_rate": 1.209853615905615e-06, + "loss": 1.197, + "step": 34745 + }, + { + "epoch": 7.58, + "learning_rate": 1.2097443740441339e-06, + "loss": 0.8178, + "step": 34746 + }, + { + "epoch": 7.58, + "learning_rate": 1.2096351321826525e-06, + "loss": 0.9442, + "step": 34747 + }, + { + "epoch": 7.58, + "learning_rate": 1.2095258903211713e-06, + "loss": 0.6383, + "step": 34748 + }, + { + "epoch": 7.58, + "learning_rate": 1.20941664845969e-06, + "loss": 1.0722, + "step": 34749 + }, + { + "epoch": 7.58, + "learning_rate": 1.2093074065982085e-06, + "loss": 1.01, + "step": 34750 + }, + { + "epoch": 7.58, + "learning_rate": 1.2091981647367271e-06, + "loss": 1.6407, + "step": 34751 + }, + { + "epoch": 7.58, + "learning_rate": 1.209088922875246e-06, + "loss": 2.2152, + "step": 34752 + }, + { + "epoch": 7.58, + "learning_rate": 1.2089796810137646e-06, + "loss": 1.6347, + "step": 34753 + }, + { + "epoch": 7.58, + "learning_rate": 1.2088704391522832e-06, + "loss": 0.9597, + "step": 34754 + }, + { + "epoch": 7.59, + "learning_rate": 1.208761197290802e-06, + "loss": 2.4421, + "step": 34755 + }, + { + "epoch": 7.59, + "learning_rate": 1.2086519554293206e-06, + "loss": 1.897, + "step": 34756 + }, + { + "epoch": 7.59, + "learning_rate": 1.2085427135678392e-06, + "loss": 0.4035, + "step": 34757 + }, + { + "epoch": 7.59, + "learning_rate": 1.208433471706358e-06, + "loss": 1.1138, + "step": 34758 + }, + { + "epoch": 7.59, + "learning_rate": 1.2083242298448767e-06, + "loss": 0.7322, + "step": 34759 + }, + { + "epoch": 7.59, + "learning_rate": 1.2082149879833953e-06, + "loss": 0.8457, + "step": 34760 + }, + { + "epoch": 7.59, + "learning_rate": 1.2081057461219139e-06, + "loss": 1.0117, + "step": 34761 + }, + { + "epoch": 7.59, + "learning_rate": 1.2079965042604327e-06, + "loss": 1.2043, + "step": 34762 + }, + { + "epoch": 7.59, + "learning_rate": 1.2078872623989513e-06, + "loss": 1.5956, + "step": 34763 + }, + { + "epoch": 7.59, + "learning_rate": 1.2077780205374701e-06, + "loss": 1.5617, + "step": 34764 + }, + { + "epoch": 7.59, + "learning_rate": 1.2076687786759888e-06, + "loss": 1.633, + "step": 34765 + }, + { + "epoch": 7.59, + "learning_rate": 1.2075595368145074e-06, + "loss": 2.0447, + "step": 34766 + }, + { + "epoch": 7.59, + "learning_rate": 1.207450294953026e-06, + "loss": 0.5601, + "step": 34767 + }, + { + "epoch": 7.59, + "learning_rate": 1.2073410530915448e-06, + "loss": 4.0088, + "step": 34768 + }, + { + "epoch": 7.59, + "learning_rate": 1.2072318112300634e-06, + "loss": 2.7262, + "step": 34769 + }, + { + "epoch": 7.59, + "learning_rate": 1.2071225693685822e-06, + "loss": 2.6343, + "step": 34770 + }, + { + "epoch": 7.59, + "learning_rate": 1.2070133275071008e-06, + "loss": 1.7196, + "step": 34771 + }, + { + "epoch": 7.59, + "learning_rate": 1.2069040856456195e-06, + "loss": 1.31, + "step": 34772 + }, + { + "epoch": 7.59, + "learning_rate": 1.206794843784138e-06, + "loss": 0.6358, + "step": 34773 + }, + { + "epoch": 7.59, + "learning_rate": 1.2066856019226569e-06, + "loss": 1.4033, + "step": 34774 + }, + { + "epoch": 7.59, + "learning_rate": 1.2065763600611755e-06, + "loss": 2.1231, + "step": 34775 + }, + { + "epoch": 7.59, + "learning_rate": 1.2064671181996943e-06, + "loss": 0.6567, + "step": 34776 + }, + { + "epoch": 7.59, + "learning_rate": 1.206357876338213e-06, + "loss": 0.4594, + "step": 34777 + }, + { + "epoch": 7.59, + "learning_rate": 1.2062486344767315e-06, + "loss": 1.401, + "step": 34778 + }, + { + "epoch": 7.59, + "learning_rate": 1.2061393926152502e-06, + "loss": 0.5377, + "step": 34779 + }, + { + "epoch": 7.59, + "learning_rate": 1.206030150753769e-06, + "loss": 1.8608, + "step": 34780 + }, + { + "epoch": 7.59, + "learning_rate": 1.2059209088922876e-06, + "loss": 0.4708, + "step": 34781 + }, + { + "epoch": 7.59, + "learning_rate": 1.2058116670308064e-06, + "loss": 0.4997, + "step": 34782 + }, + { + "epoch": 7.59, + "learning_rate": 1.205702425169325e-06, + "loss": 0.3479, + "step": 34783 + }, + { + "epoch": 7.59, + "learning_rate": 1.2055931833078436e-06, + "loss": 1.351, + "step": 34784 + }, + { + "epoch": 7.59, + "learning_rate": 1.2054839414463622e-06, + "loss": 1.9016, + "step": 34785 + }, + { + "epoch": 7.59, + "learning_rate": 1.205374699584881e-06, + "loss": 0.8893, + "step": 34786 + }, + { + "epoch": 7.59, + "learning_rate": 1.2052654577233997e-06, + "loss": 1.5058, + "step": 34787 + }, + { + "epoch": 7.59, + "learning_rate": 1.2051562158619185e-06, + "loss": 0.1459, + "step": 34788 + }, + { + "epoch": 7.59, + "learning_rate": 1.2050469740004371e-06, + "loss": 1.2816, + "step": 34789 + }, + { + "epoch": 7.59, + "learning_rate": 1.2049377321389557e-06, + "loss": 2.1842, + "step": 34790 + }, + { + "epoch": 7.59, + "learning_rate": 1.2048284902774743e-06, + "loss": 1.6189, + "step": 34791 + }, + { + "epoch": 7.59, + "learning_rate": 1.2047192484159932e-06, + "loss": 0.7821, + "step": 34792 + }, + { + "epoch": 7.59, + "learning_rate": 1.2046100065545118e-06, + "loss": 0.2944, + "step": 34793 + }, + { + "epoch": 7.59, + "learning_rate": 1.2045007646930304e-06, + "loss": 0.9568, + "step": 34794 + }, + { + "epoch": 7.59, + "learning_rate": 1.2043915228315492e-06, + "loss": 1.0419, + "step": 34795 + }, + { + "epoch": 7.59, + "learning_rate": 1.2042822809700678e-06, + "loss": 0.8736, + "step": 34796 + }, + { + "epoch": 7.59, + "learning_rate": 1.2041730391085864e-06, + "loss": 2.2903, + "step": 34797 + }, + { + "epoch": 7.59, + "learning_rate": 1.2040637972471052e-06, + "loss": 0.9085, + "step": 34798 + }, + { + "epoch": 7.59, + "learning_rate": 1.2039545553856239e-06, + "loss": 0.6563, + "step": 34799 + }, + { + "epoch": 7.59, + "learning_rate": 1.2038453135241425e-06, + "loss": 1.4725, + "step": 34800 + }, + { + "epoch": 7.6, + "learning_rate": 1.2037360716626613e-06, + "loss": 1.1126, + "step": 34801 + }, + { + "epoch": 7.6, + "learning_rate": 1.20362682980118e-06, + "loss": 0.2755, + "step": 34802 + }, + { + "epoch": 7.6, + "learning_rate": 1.2035175879396985e-06, + "loss": 1.3005, + "step": 34803 + }, + { + "epoch": 7.6, + "learning_rate": 1.2034083460782173e-06, + "loss": 0.2339, + "step": 34804 + }, + { + "epoch": 7.6, + "learning_rate": 1.203299104216736e-06, + "loss": 1.2671, + "step": 34805 + }, + { + "epoch": 7.6, + "learning_rate": 1.2031898623552546e-06, + "loss": 3.0707, + "step": 34806 + }, + { + "epoch": 7.6, + "learning_rate": 1.2030806204937732e-06, + "loss": 0.84, + "step": 34807 + }, + { + "epoch": 7.6, + "learning_rate": 1.202971378632292e-06, + "loss": 1.4527, + "step": 34808 + }, + { + "epoch": 7.6, + "learning_rate": 1.2028621367708106e-06, + "loss": 2.0289, + "step": 34809 + }, + { + "epoch": 7.6, + "learning_rate": 1.2027528949093294e-06, + "loss": 0.9035, + "step": 34810 + }, + { + "epoch": 7.6, + "learning_rate": 1.202643653047848e-06, + "loss": 0.9654, + "step": 34811 + }, + { + "epoch": 7.6, + "learning_rate": 1.2025344111863666e-06, + "loss": 0.3567, + "step": 34812 + }, + { + "epoch": 7.6, + "learning_rate": 1.2024251693248853e-06, + "loss": 0.8895, + "step": 34813 + }, + { + "epoch": 7.6, + "learning_rate": 1.202315927463404e-06, + "loss": 2.0269, + "step": 34814 + }, + { + "epoch": 7.6, + "learning_rate": 1.202206685601923e-06, + "loss": 0.7654, + "step": 34815 + }, + { + "epoch": 7.6, + "learning_rate": 1.2020974437404415e-06, + "loss": 0.9053, + "step": 34816 + }, + { + "epoch": 7.6, + "learning_rate": 1.2019882018789601e-06, + "loss": 0.2993, + "step": 34817 + }, + { + "epoch": 7.6, + "learning_rate": 1.2018789600174787e-06, + "loss": 1.3134, + "step": 34818 + }, + { + "epoch": 7.6, + "learning_rate": 1.2017697181559973e-06, + "loss": 0.5497, + "step": 34819 + }, + { + "epoch": 7.6, + "learning_rate": 1.2016604762945162e-06, + "loss": 0.2909, + "step": 34820 + }, + { + "epoch": 7.6, + "learning_rate": 1.201551234433035e-06, + "loss": 2.2092, + "step": 34821 + }, + { + "epoch": 7.6, + "learning_rate": 1.2014419925715536e-06, + "loss": 2.0669, + "step": 34822 + }, + { + "epoch": 7.6, + "learning_rate": 1.2013327507100722e-06, + "loss": 0.4893, + "step": 34823 + }, + { + "epoch": 7.6, + "learning_rate": 1.2012235088485908e-06, + "loss": 0.4782, + "step": 34824 + }, + { + "epoch": 7.6, + "learning_rate": 1.2011142669871094e-06, + "loss": 1.4869, + "step": 34825 + }, + { + "epoch": 7.6, + "learning_rate": 1.2010050251256283e-06, + "loss": 0.8945, + "step": 34826 + }, + { + "epoch": 7.6, + "learning_rate": 1.2008957832641469e-06, + "loss": 0.5175, + "step": 34827 + }, + { + "epoch": 7.6, + "learning_rate": 1.2007865414026657e-06, + "loss": 0.3472, + "step": 34828 + }, + { + "epoch": 7.6, + "learning_rate": 1.2006772995411843e-06, + "loss": 0.9073, + "step": 34829 + }, + { + "epoch": 7.6, + "learning_rate": 1.200568057679703e-06, + "loss": 0.6391, + "step": 34830 + }, + { + "epoch": 7.6, + "learning_rate": 1.2004588158182215e-06, + "loss": 0.7939, + "step": 34831 + }, + { + "epoch": 7.6, + "learning_rate": 1.2003495739567403e-06, + "loss": 2.009, + "step": 34832 + }, + { + "epoch": 7.6, + "learning_rate": 1.200240332095259e-06, + "loss": 0.4202, + "step": 34833 + }, + { + "epoch": 7.6, + "learning_rate": 1.2001310902337778e-06, + "loss": 0.2124, + "step": 34834 + }, + { + "epoch": 7.6, + "learning_rate": 1.2000218483722964e-06, + "loss": 1.1274, + "step": 34835 + }, + { + "epoch": 7.6, + "learning_rate": 1.199912606510815e-06, + "loss": 1.15, + "step": 34836 + }, + { + "epoch": 7.6, + "learning_rate": 1.1998033646493338e-06, + "loss": 1.0261, + "step": 34837 + }, + { + "epoch": 7.6, + "learning_rate": 1.1996941227878524e-06, + "loss": 1.5902, + "step": 34838 + }, + { + "epoch": 7.6, + "learning_rate": 1.199584880926371e-06, + "loss": 2.7367, + "step": 34839 + }, + { + "epoch": 7.6, + "learning_rate": 1.1994756390648897e-06, + "loss": 0.2997, + "step": 34840 + }, + { + "epoch": 7.6, + "learning_rate": 1.1993663972034085e-06, + "loss": 2.5124, + "step": 34841 + }, + { + "epoch": 7.6, + "learning_rate": 1.199257155341927e-06, + "loss": 0.7446, + "step": 34842 + }, + { + "epoch": 7.6, + "learning_rate": 1.199147913480446e-06, + "loss": 1.7776, + "step": 34843 + }, + { + "epoch": 7.6, + "learning_rate": 1.1990386716189645e-06, + "loss": 1.1959, + "step": 34844 + }, + { + "epoch": 7.6, + "learning_rate": 1.1989294297574831e-06, + "loss": 1.6397, + "step": 34845 + }, + { + "epoch": 7.6, + "learning_rate": 1.1988201878960017e-06, + "loss": 1.4769, + "step": 34846 + }, + { + "epoch": 7.61, + "learning_rate": 1.1987109460345206e-06, + "loss": 0.984, + "step": 34847 + }, + { + "epoch": 7.61, + "learning_rate": 1.1986017041730392e-06, + "loss": 1.8387, + "step": 34848 + }, + { + "epoch": 7.61, + "learning_rate": 1.198492462311558e-06, + "loss": 0.8257, + "step": 34849 + }, + { + "epoch": 7.61, + "learning_rate": 1.1983832204500766e-06, + "loss": 0.8929, + "step": 34850 + }, + { + "epoch": 7.61, + "learning_rate": 1.1982739785885952e-06, + "loss": 0.7561, + "step": 34851 + }, + { + "epoch": 7.61, + "learning_rate": 1.1981647367271138e-06, + "loss": 1.4522, + "step": 34852 + }, + { + "epoch": 7.61, + "learning_rate": 1.1980554948656324e-06, + "loss": 1.8016, + "step": 34853 + }, + { + "epoch": 7.61, + "learning_rate": 1.1979462530041513e-06, + "loss": 0.9334, + "step": 34854 + }, + { + "epoch": 7.61, + "learning_rate": 1.19783701114267e-06, + "loss": 2.0981, + "step": 34855 + }, + { + "epoch": 7.61, + "learning_rate": 1.1977277692811887e-06, + "loss": 1.0382, + "step": 34856 + }, + { + "epoch": 7.61, + "learning_rate": 1.1976185274197073e-06, + "loss": 0.6666, + "step": 34857 + }, + { + "epoch": 7.61, + "learning_rate": 1.197509285558226e-06, + "loss": 0.5635, + "step": 34858 + }, + { + "epoch": 7.61, + "learning_rate": 1.1974000436967448e-06, + "loss": 0.6709, + "step": 34859 + }, + { + "epoch": 7.61, + "learning_rate": 1.1972908018352634e-06, + "loss": 1.1694, + "step": 34860 + }, + { + "epoch": 7.61, + "learning_rate": 1.1971815599737822e-06, + "loss": 2.9543, + "step": 34861 + }, + { + "epoch": 7.61, + "learning_rate": 1.1970723181123008e-06, + "loss": 0.6842, + "step": 34862 + }, + { + "epoch": 7.61, + "learning_rate": 1.1969630762508194e-06, + "loss": 0.1641, + "step": 34863 + }, + { + "epoch": 7.61, + "learning_rate": 1.196853834389338e-06, + "loss": 2.6244, + "step": 34864 + }, + { + "epoch": 7.61, + "learning_rate": 1.1967445925278568e-06, + "loss": 0.4669, + "step": 34865 + }, + { + "epoch": 7.61, + "learning_rate": 1.1966353506663755e-06, + "loss": 0.6845, + "step": 34866 + }, + { + "epoch": 7.61, + "learning_rate": 1.1965261088048943e-06, + "loss": 1.4319, + "step": 34867 + }, + { + "epoch": 7.61, + "learning_rate": 1.1964168669434129e-06, + "loss": 0.9164, + "step": 34868 + }, + { + "epoch": 7.61, + "learning_rate": 1.1963076250819315e-06, + "loss": 2.3615, + "step": 34869 + }, + { + "epoch": 7.61, + "learning_rate": 1.1961983832204501e-06, + "loss": 1.2246, + "step": 34870 + }, + { + "epoch": 7.61, + "learning_rate": 1.196089141358969e-06, + "loss": 1.395, + "step": 34871 + }, + { + "epoch": 7.61, + "learning_rate": 1.1959798994974875e-06, + "loss": 2.7912, + "step": 34872 + }, + { + "epoch": 7.61, + "learning_rate": 1.1958706576360062e-06, + "loss": 0.9621, + "step": 34873 + }, + { + "epoch": 7.61, + "learning_rate": 1.195761415774525e-06, + "loss": 1.4321, + "step": 34874 + }, + { + "epoch": 7.61, + "learning_rate": 1.1956521739130436e-06, + "loss": 2.319, + "step": 34875 + }, + { + "epoch": 7.61, + "learning_rate": 1.1955429320515622e-06, + "loss": 1.2316, + "step": 34876 + }, + { + "epoch": 7.61, + "learning_rate": 1.195433690190081e-06, + "loss": 0.9338, + "step": 34877 + }, + { + "epoch": 7.61, + "learning_rate": 1.1953244483285996e-06, + "loss": 0.8896, + "step": 34878 + }, + { + "epoch": 7.61, + "learning_rate": 1.1952152064671182e-06, + "loss": 0.175, + "step": 34879 + }, + { + "epoch": 7.61, + "learning_rate": 1.1951059646056369e-06, + "loss": 1.9859, + "step": 34880 + }, + { + "epoch": 7.61, + "learning_rate": 1.1949967227441557e-06, + "loss": 1.5879, + "step": 34881 + }, + { + "epoch": 7.61, + "learning_rate": 1.1948874808826743e-06, + "loss": 0.5606, + "step": 34882 + }, + { + "epoch": 7.61, + "learning_rate": 1.1947782390211931e-06, + "loss": 1.4755, + "step": 34883 + }, + { + "epoch": 7.61, + "learning_rate": 1.1946689971597117e-06, + "loss": 1.3929, + "step": 34884 + }, + { + "epoch": 7.61, + "learning_rate": 1.1945597552982303e-06, + "loss": 0.5609, + "step": 34885 + }, + { + "epoch": 7.61, + "learning_rate": 1.194450513436749e-06, + "loss": 0.778, + "step": 34886 + }, + { + "epoch": 7.61, + "learning_rate": 1.1943412715752678e-06, + "loss": 1.644, + "step": 34887 + }, + { + "epoch": 7.61, + "learning_rate": 1.1942320297137864e-06, + "loss": 0.3162, + "step": 34888 + }, + { + "epoch": 7.61, + "learning_rate": 1.1941227878523052e-06, + "loss": 2.0929, + "step": 34889 + }, + { + "epoch": 7.61, + "learning_rate": 1.1940135459908238e-06, + "loss": 1.5883, + "step": 34890 + }, + { + "epoch": 7.61, + "learning_rate": 1.1939043041293424e-06, + "loss": 1.0817, + "step": 34891 + }, + { + "epoch": 7.62, + "learning_rate": 1.193795062267861e-06, + "loss": 1.6508, + "step": 34892 + }, + { + "epoch": 7.62, + "learning_rate": 1.1936858204063799e-06, + "loss": 1.1965, + "step": 34893 + }, + { + "epoch": 7.62, + "learning_rate": 1.1935765785448985e-06, + "loss": 0.0598, + "step": 34894 + }, + { + "epoch": 7.62, + "learning_rate": 1.1934673366834173e-06, + "loss": 0.0671, + "step": 34895 + }, + { + "epoch": 7.62, + "learning_rate": 1.193358094821936e-06, + "loss": 0.0744, + "step": 34896 + }, + { + "epoch": 7.62, + "learning_rate": 1.1932488529604545e-06, + "loss": 1.6185, + "step": 34897 + }, + { + "epoch": 7.62, + "learning_rate": 1.1931396110989731e-06, + "loss": 1.0541, + "step": 34898 + }, + { + "epoch": 7.62, + "learning_rate": 1.193030369237492e-06, + "loss": 3.0377, + "step": 34899 + }, + { + "epoch": 7.62, + "learning_rate": 1.1929211273760106e-06, + "loss": 1.4183, + "step": 34900 + }, + { + "epoch": 7.62, + "learning_rate": 1.1928118855145294e-06, + "loss": 1.0084, + "step": 34901 + }, + { + "epoch": 7.62, + "learning_rate": 1.192702643653048e-06, + "loss": 1.0336, + "step": 34902 + }, + { + "epoch": 7.62, + "learning_rate": 1.1925934017915666e-06, + "loss": 2.4726, + "step": 34903 + }, + { + "epoch": 7.62, + "learning_rate": 1.1924841599300852e-06, + "loss": 0.7752, + "step": 34904 + }, + { + "epoch": 7.62, + "learning_rate": 1.192374918068604e-06, + "loss": 2.4845, + "step": 34905 + }, + { + "epoch": 7.62, + "learning_rate": 1.1922656762071226e-06, + "loss": 2.18, + "step": 34906 + }, + { + "epoch": 7.62, + "learning_rate": 1.1921564343456415e-06, + "loss": 1.9794, + "step": 34907 + }, + { + "epoch": 7.62, + "learning_rate": 1.19204719248416e-06, + "loss": 2.6443, + "step": 34908 + }, + { + "epoch": 7.62, + "learning_rate": 1.1919379506226787e-06, + "loss": 1.6959, + "step": 34909 + }, + { + "epoch": 7.62, + "learning_rate": 1.1918287087611973e-06, + "loss": 2.7038, + "step": 34910 + }, + { + "epoch": 7.62, + "learning_rate": 1.1917194668997161e-06, + "loss": 1.644, + "step": 34911 + }, + { + "epoch": 7.62, + "learning_rate": 1.1916102250382347e-06, + "loss": 1.2069, + "step": 34912 + }, + { + "epoch": 7.62, + "learning_rate": 1.1915009831767533e-06, + "loss": 2.221, + "step": 34913 + }, + { + "epoch": 7.62, + "learning_rate": 1.1913917413152722e-06, + "loss": 1.1627, + "step": 34914 + }, + { + "epoch": 7.62, + "learning_rate": 1.1912824994537908e-06, + "loss": 0.5165, + "step": 34915 + }, + { + "epoch": 7.62, + "learning_rate": 1.1911732575923094e-06, + "loss": 1.1, + "step": 34916 + }, + { + "epoch": 7.62, + "learning_rate": 1.1910640157308282e-06, + "loss": 1.8535, + "step": 34917 + }, + { + "epoch": 7.62, + "learning_rate": 1.1909547738693468e-06, + "loss": 2.1086, + "step": 34918 + }, + { + "epoch": 7.62, + "learning_rate": 1.1908455320078654e-06, + "loss": 0.814, + "step": 34919 + }, + { + "epoch": 7.62, + "learning_rate": 1.1907362901463843e-06, + "loss": 2.2747, + "step": 34920 + }, + { + "epoch": 7.62, + "learning_rate": 1.1906270482849029e-06, + "loss": 0.9637, + "step": 34921 + }, + { + "epoch": 7.62, + "learning_rate": 1.1905178064234215e-06, + "loss": 0.7275, + "step": 34922 + }, + { + "epoch": 7.62, + "learning_rate": 1.1904085645619403e-06, + "loss": 1.8338, + "step": 34923 + }, + { + "epoch": 7.62, + "learning_rate": 1.190299322700459e-06, + "loss": 0.7534, + "step": 34924 + }, + { + "epoch": 7.62, + "learning_rate": 1.1901900808389775e-06, + "loss": 0.8676, + "step": 34925 + }, + { + "epoch": 7.62, + "learning_rate": 1.1900808389774961e-06, + "loss": 0.5612, + "step": 34926 + }, + { + "epoch": 7.62, + "learning_rate": 1.189971597116015e-06, + "loss": 1.13, + "step": 34927 + }, + { + "epoch": 7.62, + "learning_rate": 1.1898623552545336e-06, + "loss": 0.5655, + "step": 34928 + }, + { + "epoch": 7.62, + "learning_rate": 1.1897531133930524e-06, + "loss": 0.8448, + "step": 34929 + }, + { + "epoch": 7.62, + "learning_rate": 1.189643871531571e-06, + "loss": 0.3491, + "step": 34930 + }, + { + "epoch": 7.62, + "learning_rate": 1.1895346296700896e-06, + "loss": 1.6963, + "step": 34931 + }, + { + "epoch": 7.62, + "learning_rate": 1.1894253878086082e-06, + "loss": 0.0754, + "step": 34932 + }, + { + "epoch": 7.62, + "learning_rate": 1.189316145947127e-06, + "loss": 0.9582, + "step": 34933 + }, + { + "epoch": 7.62, + "learning_rate": 1.1892069040856459e-06, + "loss": 1.089, + "step": 34934 + }, + { + "epoch": 7.62, + "learning_rate": 1.1890976622241645e-06, + "loss": 1.1314, + "step": 34935 + }, + { + "epoch": 7.62, + "learning_rate": 1.188988420362683e-06, + "loss": 0.9083, + "step": 34936 + }, + { + "epoch": 7.62, + "learning_rate": 1.1888791785012017e-06, + "loss": 3.2404, + "step": 34937 + }, + { + "epoch": 7.63, + "learning_rate": 1.1887699366397203e-06, + "loss": 0.4123, + "step": 34938 + }, + { + "epoch": 7.63, + "learning_rate": 1.1886606947782391e-06, + "loss": 0.267, + "step": 34939 + }, + { + "epoch": 7.63, + "learning_rate": 1.188551452916758e-06, + "loss": 1.1657, + "step": 34940 + }, + { + "epoch": 7.63, + "learning_rate": 1.1884422110552766e-06, + "loss": 1.3689, + "step": 34941 + }, + { + "epoch": 7.63, + "learning_rate": 1.1883329691937952e-06, + "loss": 1.0034, + "step": 34942 + }, + { + "epoch": 7.63, + "learning_rate": 1.1882237273323138e-06, + "loss": 0.7475, + "step": 34943 + }, + { + "epoch": 7.63, + "learning_rate": 1.1881144854708324e-06, + "loss": 1.2208, + "step": 34944 + }, + { + "epoch": 7.63, + "learning_rate": 1.1880052436093512e-06, + "loss": 0.9805, + "step": 34945 + }, + { + "epoch": 7.63, + "learning_rate": 1.1878960017478698e-06, + "loss": 2.4546, + "step": 34946 + }, + { + "epoch": 7.63, + "learning_rate": 1.1877867598863887e-06, + "loss": 1.4865, + "step": 34947 + }, + { + "epoch": 7.63, + "learning_rate": 1.1876775180249073e-06, + "loss": 1.8337, + "step": 34948 + }, + { + "epoch": 7.63, + "learning_rate": 1.1875682761634259e-06, + "loss": 1.5101, + "step": 34949 + }, + { + "epoch": 7.63, + "learning_rate": 1.1874590343019445e-06, + "loss": 1.5096, + "step": 34950 + }, + { + "epoch": 7.63, + "learning_rate": 1.1873497924404633e-06, + "loss": 2.241, + "step": 34951 + }, + { + "epoch": 7.63, + "learning_rate": 1.187240550578982e-06, + "loss": 1.5634, + "step": 34952 + }, + { + "epoch": 7.63, + "learning_rate": 1.1871313087175007e-06, + "loss": 1.7095, + "step": 34953 + }, + { + "epoch": 7.63, + "learning_rate": 1.1870220668560194e-06, + "loss": 0.658, + "step": 34954 + }, + { + "epoch": 7.63, + "learning_rate": 1.186912824994538e-06, + "loss": 1.9955, + "step": 34955 + }, + { + "epoch": 7.63, + "learning_rate": 1.1868035831330566e-06, + "loss": 0.606, + "step": 34956 + }, + { + "epoch": 7.63, + "learning_rate": 1.1866943412715754e-06, + "loss": 1.8258, + "step": 34957 + }, + { + "epoch": 7.63, + "learning_rate": 1.186585099410094e-06, + "loss": 0.3109, + "step": 34958 + }, + { + "epoch": 7.63, + "learning_rate": 1.1864758575486126e-06, + "loss": 0.8768, + "step": 34959 + }, + { + "epoch": 7.63, + "learning_rate": 1.1863666156871314e-06, + "loss": 0.705, + "step": 34960 + }, + { + "epoch": 7.63, + "learning_rate": 1.18625737382565e-06, + "loss": 1.8651, + "step": 34961 + }, + { + "epoch": 7.63, + "learning_rate": 1.1861481319641689e-06, + "loss": 1.0303, + "step": 34962 + }, + { + "epoch": 7.63, + "learning_rate": 1.1860388901026875e-06, + "loss": 1.6572, + "step": 34963 + }, + { + "epoch": 7.63, + "learning_rate": 1.185929648241206e-06, + "loss": 1.6341, + "step": 34964 + }, + { + "epoch": 7.63, + "learning_rate": 1.1858204063797247e-06, + "loss": 1.2641, + "step": 34965 + }, + { + "epoch": 7.63, + "learning_rate": 1.1857111645182435e-06, + "loss": 0.8661, + "step": 34966 + }, + { + "epoch": 7.63, + "learning_rate": 1.1856019226567621e-06, + "loss": 2.0408, + "step": 34967 + }, + { + "epoch": 7.63, + "learning_rate": 1.185492680795281e-06, + "loss": 1.185, + "step": 34968 + }, + { + "epoch": 7.63, + "learning_rate": 1.1853834389337996e-06, + "loss": 0.4039, + "step": 34969 + }, + { + "epoch": 7.63, + "learning_rate": 1.1852741970723182e-06, + "loss": 0.7041, + "step": 34970 + }, + { + "epoch": 7.63, + "learning_rate": 1.1851649552108368e-06, + "loss": 0.9361, + "step": 34971 + }, + { + "epoch": 7.63, + "learning_rate": 1.1850557133493554e-06, + "loss": 1.0161, + "step": 34972 + }, + { + "epoch": 7.63, + "learning_rate": 1.1849464714878742e-06, + "loss": 1.441, + "step": 34973 + }, + { + "epoch": 7.63, + "learning_rate": 1.184837229626393e-06, + "loss": 0.772, + "step": 34974 + }, + { + "epoch": 7.63, + "learning_rate": 1.1847279877649117e-06, + "loss": 1.2436, + "step": 34975 + }, + { + "epoch": 7.63, + "learning_rate": 1.1846187459034303e-06, + "loss": 1.467, + "step": 34976 + }, + { + "epoch": 7.63, + "learning_rate": 1.1845095040419489e-06, + "loss": 1.6914, + "step": 34977 + }, + { + "epoch": 7.63, + "learning_rate": 1.1844002621804675e-06, + "loss": 0.9733, + "step": 34978 + }, + { + "epoch": 7.63, + "learning_rate": 1.1842910203189863e-06, + "loss": 0.7357, + "step": 34979 + }, + { + "epoch": 7.63, + "learning_rate": 1.1841817784575052e-06, + "loss": 1.3601, + "step": 34980 + }, + { + "epoch": 7.63, + "learning_rate": 1.1840725365960238e-06, + "loss": 1.3113, + "step": 34981 + }, + { + "epoch": 7.63, + "learning_rate": 1.1839632947345424e-06, + "loss": 0.3975, + "step": 34982 + }, + { + "epoch": 7.63, + "learning_rate": 1.183854052873061e-06, + "loss": 1.6976, + "step": 34983 + }, + { + "epoch": 7.64, + "learning_rate": 1.1837448110115798e-06, + "loss": 1.1574, + "step": 34984 + }, + { + "epoch": 7.64, + "learning_rate": 1.1836355691500984e-06, + "loss": 2.6508, + "step": 34985 + }, + { + "epoch": 7.64, + "learning_rate": 1.1835263272886172e-06, + "loss": 1.2272, + "step": 34986 + }, + { + "epoch": 7.64, + "learning_rate": 1.1834170854271359e-06, + "loss": 1.4222, + "step": 34987 + }, + { + "epoch": 7.64, + "learning_rate": 1.1833078435656545e-06, + "loss": 1.0815, + "step": 34988 + }, + { + "epoch": 7.64, + "learning_rate": 1.183198601704173e-06, + "loss": 1.0957, + "step": 34989 + }, + { + "epoch": 7.64, + "learning_rate": 1.183089359842692e-06, + "loss": 0.2471, + "step": 34990 + }, + { + "epoch": 7.64, + "learning_rate": 1.1829801179812105e-06, + "loss": 1.7347, + "step": 34991 + }, + { + "epoch": 7.64, + "learning_rate": 1.1828708761197291e-06, + "loss": 1.7748, + "step": 34992 + }, + { + "epoch": 7.64, + "learning_rate": 1.182761634258248e-06, + "loss": 0.1818, + "step": 34993 + }, + { + "epoch": 7.64, + "learning_rate": 1.1826523923967666e-06, + "loss": 1.1824, + "step": 34994 + }, + { + "epoch": 7.64, + "learning_rate": 1.1825431505352852e-06, + "loss": 1.4883, + "step": 34995 + }, + { + "epoch": 7.64, + "learning_rate": 1.182433908673804e-06, + "loss": 1.0206, + "step": 34996 + }, + { + "epoch": 7.64, + "learning_rate": 1.1823246668123226e-06, + "loss": 1.3572, + "step": 34997 + }, + { + "epoch": 7.64, + "learning_rate": 1.1822154249508412e-06, + "loss": 1.7077, + "step": 34998 + }, + { + "epoch": 7.64, + "learning_rate": 1.1821061830893598e-06, + "loss": 1.3747, + "step": 34999 + }, + { + "epoch": 7.64, + "learning_rate": 1.1819969412278786e-06, + "loss": 0.2755, + "step": 35000 + }, + { + "epoch": 7.64, + "learning_rate": 1.1818876993663973e-06, + "loss": 0.5681, + "step": 35001 + }, + { + "epoch": 7.64, + "learning_rate": 1.181778457504916e-06, + "loss": 2.1678, + "step": 35002 + }, + { + "epoch": 7.64, + "learning_rate": 1.1816692156434347e-06, + "loss": 2.1252, + "step": 35003 + }, + { + "epoch": 7.64, + "learning_rate": 1.1815599737819533e-06, + "loss": 1.4633, + "step": 35004 + }, + { + "epoch": 7.64, + "learning_rate": 1.181450731920472e-06, + "loss": 2.4104, + "step": 35005 + }, + { + "epoch": 7.64, + "learning_rate": 1.1813414900589907e-06, + "loss": 1.0841, + "step": 35006 + }, + { + "epoch": 7.64, + "learning_rate": 1.1812322481975093e-06, + "loss": 1.7526, + "step": 35007 + }, + { + "epoch": 7.64, + "learning_rate": 1.1811230063360282e-06, + "loss": 1.4878, + "step": 35008 + }, + { + "epoch": 7.64, + "learning_rate": 1.1810137644745468e-06, + "loss": 1.2333, + "step": 35009 + }, + { + "epoch": 7.64, + "learning_rate": 1.1809045226130654e-06, + "loss": 1.9158, + "step": 35010 + }, + { + "epoch": 7.64, + "learning_rate": 1.180795280751584e-06, + "loss": 1.5313, + "step": 35011 + }, + { + "epoch": 7.64, + "learning_rate": 1.1806860388901028e-06, + "loss": 1.3147, + "step": 35012 + }, + { + "epoch": 7.64, + "learning_rate": 1.1805767970286214e-06, + "loss": 1.0292, + "step": 35013 + }, + { + "epoch": 7.64, + "learning_rate": 1.1804675551671403e-06, + "loss": 1.5596, + "step": 35014 + }, + { + "epoch": 7.64, + "learning_rate": 1.1803583133056589e-06, + "loss": 0.9622, + "step": 35015 + }, + { + "epoch": 7.64, + "learning_rate": 1.1802490714441775e-06, + "loss": 1.0385, + "step": 35016 + }, + { + "epoch": 7.64, + "learning_rate": 1.180139829582696e-06, + "loss": 1.5434, + "step": 35017 + }, + { + "epoch": 7.64, + "learning_rate": 1.180030587721215e-06, + "loss": 1.9174, + "step": 35018 + }, + { + "epoch": 7.64, + "learning_rate": 1.1799213458597335e-06, + "loss": 0.7829, + "step": 35019 + }, + { + "epoch": 7.64, + "learning_rate": 1.1798121039982523e-06, + "loss": 0.6697, + "step": 35020 + }, + { + "epoch": 7.64, + "learning_rate": 1.179702862136771e-06, + "loss": 1.0537, + "step": 35021 + }, + { + "epoch": 7.64, + "learning_rate": 1.1795936202752896e-06, + "loss": 2.5874, + "step": 35022 + }, + { + "epoch": 7.64, + "learning_rate": 1.1794843784138082e-06, + "loss": 1.5998, + "step": 35023 + }, + { + "epoch": 7.64, + "learning_rate": 1.179375136552327e-06, + "loss": 1.2708, + "step": 35024 + }, + { + "epoch": 7.64, + "learning_rate": 1.1792658946908456e-06, + "loss": 0.7304, + "step": 35025 + }, + { + "epoch": 7.64, + "learning_rate": 1.1791566528293644e-06, + "loss": 1.0213, + "step": 35026 + }, + { + "epoch": 7.64, + "learning_rate": 1.179047410967883e-06, + "loss": 1.4355, + "step": 35027 + }, + { + "epoch": 7.64, + "learning_rate": 1.1789381691064017e-06, + "loss": 0.2185, + "step": 35028 + }, + { + "epoch": 7.64, + "learning_rate": 1.1788289272449203e-06, + "loss": 0.4567, + "step": 35029 + }, + { + "epoch": 7.65, + "learning_rate": 1.178719685383439e-06, + "loss": 1.6521, + "step": 35030 + }, + { + "epoch": 7.65, + "learning_rate": 1.1786104435219577e-06, + "loss": 0.7457, + "step": 35031 + }, + { + "epoch": 7.65, + "learning_rate": 1.1785012016604763e-06, + "loss": 0.8985, + "step": 35032 + }, + { + "epoch": 7.65, + "learning_rate": 1.1783919597989951e-06, + "loss": 1.8656, + "step": 35033 + }, + { + "epoch": 7.65, + "learning_rate": 1.1782827179375137e-06, + "loss": 1.046, + "step": 35034 + }, + { + "epoch": 7.65, + "learning_rate": 1.1781734760760324e-06, + "loss": 1.9742, + "step": 35035 + }, + { + "epoch": 7.65, + "learning_rate": 1.1780642342145512e-06, + "loss": 2.1389, + "step": 35036 + }, + { + "epoch": 7.65, + "learning_rate": 1.1779549923530698e-06, + "loss": 2.2718, + "step": 35037 + }, + { + "epoch": 7.65, + "learning_rate": 1.1778457504915884e-06, + "loss": 1.253, + "step": 35038 + }, + { + "epoch": 7.65, + "learning_rate": 1.1777365086301072e-06, + "loss": 0.9095, + "step": 35039 + }, + { + "epoch": 7.65, + "learning_rate": 1.1776272667686258e-06, + "loss": 0.8142, + "step": 35040 + }, + { + "epoch": 7.65, + "learning_rate": 1.1775180249071444e-06, + "loss": 2.4621, + "step": 35041 + }, + { + "epoch": 7.65, + "learning_rate": 1.1774087830456633e-06, + "loss": 0.3986, + "step": 35042 + }, + { + "epoch": 7.65, + "learning_rate": 1.1772995411841819e-06, + "loss": 2.7523, + "step": 35043 + }, + { + "epoch": 7.65, + "learning_rate": 1.1771902993227005e-06, + "loss": 1.6605, + "step": 35044 + }, + { + "epoch": 7.65, + "learning_rate": 1.177081057461219e-06, + "loss": 1.8932, + "step": 35045 + }, + { + "epoch": 7.65, + "learning_rate": 1.176971815599738e-06, + "loss": 0.9501, + "step": 35046 + }, + { + "epoch": 7.65, + "learning_rate": 1.1768625737382565e-06, + "loss": 1.2919, + "step": 35047 + }, + { + "epoch": 7.65, + "learning_rate": 1.1767533318767754e-06, + "loss": 0.9636, + "step": 35048 + }, + { + "epoch": 7.65, + "learning_rate": 1.176644090015294e-06, + "loss": 3.0208, + "step": 35049 + }, + { + "epoch": 7.65, + "learning_rate": 1.1765348481538126e-06, + "loss": 1.0423, + "step": 35050 + }, + { + "epoch": 7.65, + "learning_rate": 1.1764256062923312e-06, + "loss": 1.0392, + "step": 35051 + }, + { + "epoch": 7.65, + "learning_rate": 1.17631636443085e-06, + "loss": 2.3983, + "step": 35052 + }, + { + "epoch": 7.65, + "learning_rate": 1.1762071225693686e-06, + "loss": 1.2018, + "step": 35053 + }, + { + "epoch": 7.65, + "learning_rate": 1.1760978807078874e-06, + "loss": 2.7996, + "step": 35054 + }, + { + "epoch": 7.65, + "learning_rate": 1.175988638846406e-06, + "loss": 1.4105, + "step": 35055 + }, + { + "epoch": 7.65, + "learning_rate": 1.1758793969849247e-06, + "loss": 0.7622, + "step": 35056 + }, + { + "epoch": 7.65, + "learning_rate": 1.1757701551234433e-06, + "loss": 0.1905, + "step": 35057 + }, + { + "epoch": 7.65, + "learning_rate": 1.175660913261962e-06, + "loss": 1.1371, + "step": 35058 + }, + { + "epoch": 7.65, + "learning_rate": 1.175551671400481e-06, + "loss": 0.7647, + "step": 35059 + }, + { + "epoch": 7.65, + "learning_rate": 1.1754424295389995e-06, + "loss": 0.8082, + "step": 35060 + }, + { + "epoch": 7.65, + "learning_rate": 1.1753331876775181e-06, + "loss": 0.6711, + "step": 35061 + }, + { + "epoch": 7.65, + "learning_rate": 1.1752239458160368e-06, + "loss": 0.3765, + "step": 35062 + }, + { + "epoch": 7.65, + "learning_rate": 1.1751147039545554e-06, + "loss": 0.8417, + "step": 35063 + }, + { + "epoch": 7.65, + "learning_rate": 1.1750054620930742e-06, + "loss": 2.3853, + "step": 35064 + }, + { + "epoch": 7.65, + "learning_rate": 1.1748962202315928e-06, + "loss": 1.0084, + "step": 35065 + }, + { + "epoch": 7.65, + "learning_rate": 1.1747869783701116e-06, + "loss": 0.9163, + "step": 35066 + }, + { + "epoch": 7.65, + "learning_rate": 1.1746777365086302e-06, + "loss": 0.8171, + "step": 35067 + }, + { + "epoch": 7.65, + "learning_rate": 1.1745684946471488e-06, + "loss": 1.4711, + "step": 35068 + }, + { + "epoch": 7.65, + "learning_rate": 1.1744592527856675e-06, + "loss": 0.943, + "step": 35069 + }, + { + "epoch": 7.65, + "learning_rate": 1.1743500109241863e-06, + "loss": 2.5367, + "step": 35070 + }, + { + "epoch": 7.65, + "learning_rate": 1.1742407690627049e-06, + "loss": 0.6991, + "step": 35071 + }, + { + "epoch": 7.65, + "learning_rate": 1.1741315272012237e-06, + "loss": 0.8506, + "step": 35072 + }, + { + "epoch": 7.65, + "learning_rate": 1.1740222853397423e-06, + "loss": 2.6716, + "step": 35073 + }, + { + "epoch": 7.65, + "learning_rate": 1.173913043478261e-06, + "loss": 0.4685, + "step": 35074 + }, + { + "epoch": 7.65, + "learning_rate": 1.1738038016167795e-06, + "loss": 0.2878, + "step": 35075 + }, + { + "epoch": 7.66, + "learning_rate": 1.1736945597552984e-06, + "loss": 0.5299, + "step": 35076 + }, + { + "epoch": 7.66, + "learning_rate": 1.173585317893817e-06, + "loss": 0.6915, + "step": 35077 + }, + { + "epoch": 7.66, + "learning_rate": 1.1734760760323356e-06, + "loss": 1.5235, + "step": 35078 + }, + { + "epoch": 7.66, + "learning_rate": 1.1733668341708544e-06, + "loss": 0.7726, + "step": 35079 + }, + { + "epoch": 7.66, + "learning_rate": 1.173257592309373e-06, + "loss": 1.0256, + "step": 35080 + }, + { + "epoch": 7.66, + "learning_rate": 1.1731483504478918e-06, + "loss": 0.5894, + "step": 35081 + }, + { + "epoch": 7.66, + "learning_rate": 1.1730391085864105e-06, + "loss": 0.4389, + "step": 35082 + }, + { + "epoch": 7.66, + "learning_rate": 1.172929866724929e-06, + "loss": 1.5082, + "step": 35083 + }, + { + "epoch": 7.66, + "learning_rate": 1.1728206248634477e-06, + "loss": 1.1919, + "step": 35084 + }, + { + "epoch": 7.66, + "learning_rate": 1.1727113830019663e-06, + "loss": 0.8087, + "step": 35085 + }, + { + "epoch": 7.66, + "learning_rate": 1.1726021411404851e-06, + "loss": 0.9507, + "step": 35086 + }, + { + "epoch": 7.66, + "learning_rate": 1.172492899279004e-06, + "loss": 0.1442, + "step": 35087 + }, + { + "epoch": 7.66, + "learning_rate": 1.1723836574175225e-06, + "loss": 0.7584, + "step": 35088 + }, + { + "epoch": 7.66, + "learning_rate": 1.1722744155560412e-06, + "loss": 0.0907, + "step": 35089 + }, + { + "epoch": 7.66, + "learning_rate": 1.1721651736945598e-06, + "loss": 1.8913, + "step": 35090 + }, + { + "epoch": 7.66, + "learning_rate": 1.1720559318330784e-06, + "loss": 0.1357, + "step": 35091 + }, + { + "epoch": 7.66, + "learning_rate": 1.1719466899715972e-06, + "loss": 2.109, + "step": 35092 + }, + { + "epoch": 7.66, + "learning_rate": 1.171837448110116e-06, + "loss": 0.5179, + "step": 35093 + }, + { + "epoch": 7.66, + "learning_rate": 1.1717282062486346e-06, + "loss": 1.6976, + "step": 35094 + }, + { + "epoch": 7.66, + "learning_rate": 1.1716189643871532e-06, + "loss": 1.3704, + "step": 35095 + }, + { + "epoch": 7.66, + "learning_rate": 1.1715097225256719e-06, + "loss": 0.932, + "step": 35096 + }, + { + "epoch": 7.66, + "learning_rate": 1.1714004806641905e-06, + "loss": 0.7226, + "step": 35097 + }, + { + "epoch": 7.66, + "learning_rate": 1.1712912388027093e-06, + "loss": 1.272, + "step": 35098 + }, + { + "epoch": 7.66, + "learning_rate": 1.1711819969412281e-06, + "loss": 2.6582, + "step": 35099 + }, + { + "epoch": 7.66, + "learning_rate": 1.1710727550797467e-06, + "loss": 0.1585, + "step": 35100 + }, + { + "epoch": 7.66, + "learning_rate": 1.1709635132182653e-06, + "loss": 1.2893, + "step": 35101 + }, + { + "epoch": 7.66, + "learning_rate": 1.170854271356784e-06, + "loss": 0.9063, + "step": 35102 + }, + { + "epoch": 7.66, + "learning_rate": 1.1707450294953026e-06, + "loss": 0.9455, + "step": 35103 + }, + { + "epoch": 7.66, + "learning_rate": 1.1706357876338214e-06, + "loss": 2.6501, + "step": 35104 + }, + { + "epoch": 7.66, + "learning_rate": 1.1705265457723402e-06, + "loss": 0.4508, + "step": 35105 + }, + { + "epoch": 7.66, + "learning_rate": 1.1704173039108588e-06, + "loss": 1.3429, + "step": 35106 + }, + { + "epoch": 7.66, + "learning_rate": 1.1703080620493774e-06, + "loss": 1.1966, + "step": 35107 + }, + { + "epoch": 7.66, + "learning_rate": 1.170198820187896e-06, + "loss": 2.3558, + "step": 35108 + }, + { + "epoch": 7.66, + "learning_rate": 1.1700895783264149e-06, + "loss": 0.4841, + "step": 35109 + }, + { + "epoch": 7.66, + "learning_rate": 1.1699803364649335e-06, + "loss": 0.6418, + "step": 35110 + }, + { + "epoch": 7.66, + "learning_rate": 1.169871094603452e-06, + "loss": 0.7704, + "step": 35111 + }, + { + "epoch": 7.66, + "learning_rate": 1.169761852741971e-06, + "loss": 1.3658, + "step": 35112 + }, + { + "epoch": 7.66, + "learning_rate": 1.1696526108804895e-06, + "loss": 1.3439, + "step": 35113 + }, + { + "epoch": 7.66, + "learning_rate": 1.1695433690190081e-06, + "loss": 2.2203, + "step": 35114 + }, + { + "epoch": 7.66, + "learning_rate": 1.169434127157527e-06, + "loss": 0.6152, + "step": 35115 + }, + { + "epoch": 7.66, + "learning_rate": 1.1693248852960456e-06, + "loss": 0.1373, + "step": 35116 + }, + { + "epoch": 7.66, + "learning_rate": 1.1692156434345642e-06, + "loss": 2.1878, + "step": 35117 + }, + { + "epoch": 7.66, + "learning_rate": 1.1691064015730828e-06, + "loss": 1.9963, + "step": 35118 + }, + { + "epoch": 7.66, + "learning_rate": 1.1689971597116016e-06, + "loss": 1.0484, + "step": 35119 + }, + { + "epoch": 7.66, + "learning_rate": 1.1688879178501202e-06, + "loss": 0.7334, + "step": 35120 + }, + { + "epoch": 7.66, + "learning_rate": 1.168778675988639e-06, + "loss": 1.3633, + "step": 35121 + }, + { + "epoch": 7.67, + "learning_rate": 1.1686694341271577e-06, + "loss": 1.1007, + "step": 35122 + }, + { + "epoch": 7.67, + "learning_rate": 1.1685601922656763e-06, + "loss": 1.2834, + "step": 35123 + }, + { + "epoch": 7.67, + "learning_rate": 1.1684509504041949e-06, + "loss": 0.8091, + "step": 35124 + }, + { + "epoch": 7.67, + "learning_rate": 1.1683417085427137e-06, + "loss": 0.2922, + "step": 35125 + }, + { + "epoch": 7.67, + "learning_rate": 1.1682324666812323e-06, + "loss": 0.9115, + "step": 35126 + }, + { + "epoch": 7.67, + "learning_rate": 1.1681232248197511e-06, + "loss": 1.9277, + "step": 35127 + }, + { + "epoch": 7.67, + "learning_rate": 1.1680139829582697e-06, + "loss": 1.585, + "step": 35128 + }, + { + "epoch": 7.67, + "learning_rate": 1.1679047410967884e-06, + "loss": 1.9521, + "step": 35129 + }, + { + "epoch": 7.67, + "learning_rate": 1.167795499235307e-06, + "loss": 0.7424, + "step": 35130 + }, + { + "epoch": 7.67, + "learning_rate": 1.1676862573738258e-06, + "loss": 2.6298, + "step": 35131 + }, + { + "epoch": 7.67, + "learning_rate": 1.1675770155123444e-06, + "loss": 0.9507, + "step": 35132 + }, + { + "epoch": 7.67, + "learning_rate": 1.1674677736508632e-06, + "loss": 1.3842, + "step": 35133 + }, + { + "epoch": 7.67, + "learning_rate": 1.1673585317893818e-06, + "loss": 0.6353, + "step": 35134 + }, + { + "epoch": 7.67, + "learning_rate": 1.1672492899279004e-06, + "loss": 1.7737, + "step": 35135 + }, + { + "epoch": 7.67, + "learning_rate": 1.167140048066419e-06, + "loss": 0.6711, + "step": 35136 + }, + { + "epoch": 7.67, + "learning_rate": 1.1670308062049379e-06, + "loss": 1.1582, + "step": 35137 + }, + { + "epoch": 7.67, + "learning_rate": 1.1669215643434565e-06, + "loss": 0.5837, + "step": 35138 + }, + { + "epoch": 7.67, + "learning_rate": 1.1668123224819753e-06, + "loss": 0.6872, + "step": 35139 + }, + { + "epoch": 7.67, + "learning_rate": 1.166703080620494e-06, + "loss": 2.6619, + "step": 35140 + }, + { + "epoch": 7.67, + "learning_rate": 1.1665938387590125e-06, + "loss": 1.4178, + "step": 35141 + }, + { + "epoch": 7.67, + "learning_rate": 1.1664845968975311e-06, + "loss": 0.7549, + "step": 35142 + }, + { + "epoch": 7.67, + "learning_rate": 1.16637535503605e-06, + "loss": 1.542, + "step": 35143 + }, + { + "epoch": 7.67, + "learning_rate": 1.1662661131745686e-06, + "loss": 1.7181, + "step": 35144 + }, + { + "epoch": 7.67, + "learning_rate": 1.1661568713130874e-06, + "loss": 1.0526, + "step": 35145 + }, + { + "epoch": 7.67, + "learning_rate": 1.166047629451606e-06, + "loss": 0.9883, + "step": 35146 + }, + { + "epoch": 7.67, + "learning_rate": 1.1659383875901246e-06, + "loss": 0.3105, + "step": 35147 + }, + { + "epoch": 7.67, + "learning_rate": 1.1658291457286432e-06, + "loss": 2.2963, + "step": 35148 + }, + { + "epoch": 7.67, + "learning_rate": 1.165719903867162e-06, + "loss": 0.1802, + "step": 35149 + }, + { + "epoch": 7.67, + "learning_rate": 1.1656106620056807e-06, + "loss": 0.5501, + "step": 35150 + }, + { + "epoch": 7.67, + "learning_rate": 1.1655014201441993e-06, + "loss": 1.0447, + "step": 35151 + }, + { + "epoch": 7.67, + "learning_rate": 1.165392178282718e-06, + "loss": 1.2013, + "step": 35152 + }, + { + "epoch": 7.67, + "learning_rate": 1.1652829364212367e-06, + "loss": 2.6371, + "step": 35153 + }, + { + "epoch": 7.67, + "learning_rate": 1.1651736945597553e-06, + "loss": 2.6016, + "step": 35154 + }, + { + "epoch": 7.67, + "learning_rate": 1.1650644526982741e-06, + "loss": 1.0255, + "step": 35155 + }, + { + "epoch": 7.67, + "learning_rate": 1.1649552108367928e-06, + "loss": 0.652, + "step": 35156 + }, + { + "epoch": 7.67, + "learning_rate": 1.1648459689753114e-06, + "loss": 1.7735, + "step": 35157 + }, + { + "epoch": 7.67, + "learning_rate": 1.1647367271138302e-06, + "loss": 2.0094, + "step": 35158 + }, + { + "epoch": 7.67, + "learning_rate": 1.1646274852523488e-06, + "loss": 1.7371, + "step": 35159 + }, + { + "epoch": 7.67, + "learning_rate": 1.1645182433908674e-06, + "loss": 0.62, + "step": 35160 + }, + { + "epoch": 7.67, + "learning_rate": 1.1644090015293862e-06, + "loss": 0.5609, + "step": 35161 + }, + { + "epoch": 7.67, + "learning_rate": 1.1642997596679048e-06, + "loss": 0.5014, + "step": 35162 + }, + { + "epoch": 7.67, + "learning_rate": 1.1641905178064235e-06, + "loss": 3.5252, + "step": 35163 + }, + { + "epoch": 7.67, + "learning_rate": 1.164081275944942e-06, + "loss": 0.8835, + "step": 35164 + }, + { + "epoch": 7.67, + "learning_rate": 1.1639720340834609e-06, + "loss": 1.9009, + "step": 35165 + }, + { + "epoch": 7.67, + "learning_rate": 1.1638627922219795e-06, + "loss": 1.5232, + "step": 35166 + }, + { + "epoch": 7.68, + "learning_rate": 1.1637535503604983e-06, + "loss": 2.2114, + "step": 35167 + }, + { + "epoch": 7.68, + "learning_rate": 1.163644308499017e-06, + "loss": 0.1779, + "step": 35168 + }, + { + "epoch": 7.68, + "learning_rate": 1.1635350666375355e-06, + "loss": 1.8119, + "step": 35169 + }, + { + "epoch": 7.68, + "learning_rate": 1.1634258247760542e-06, + "loss": 2.1702, + "step": 35170 + }, + { + "epoch": 7.68, + "learning_rate": 1.163316582914573e-06, + "loss": 1.275, + "step": 35171 + }, + { + "epoch": 7.68, + "learning_rate": 1.1632073410530916e-06, + "loss": 0.8226, + "step": 35172 + }, + { + "epoch": 7.68, + "learning_rate": 1.1630980991916104e-06, + "loss": 0.7703, + "step": 35173 + }, + { + "epoch": 7.68, + "learning_rate": 1.162988857330129e-06, + "loss": 1.842, + "step": 35174 + }, + { + "epoch": 7.68, + "learning_rate": 1.1628796154686476e-06, + "loss": 0.6818, + "step": 35175 + }, + { + "epoch": 7.68, + "learning_rate": 1.1627703736071662e-06, + "loss": 1.3123, + "step": 35176 + }, + { + "epoch": 7.68, + "learning_rate": 1.162661131745685e-06, + "loss": 2.0223, + "step": 35177 + }, + { + "epoch": 7.68, + "learning_rate": 1.1625518898842037e-06, + "loss": 0.7888, + "step": 35178 + }, + { + "epoch": 7.68, + "learning_rate": 1.1624426480227225e-06, + "loss": 0.7785, + "step": 35179 + }, + { + "epoch": 7.68, + "learning_rate": 1.1623334061612411e-06, + "loss": 1.2074, + "step": 35180 + }, + { + "epoch": 7.68, + "learning_rate": 1.1622241642997597e-06, + "loss": 1.883, + "step": 35181 + }, + { + "epoch": 7.68, + "learning_rate": 1.1621149224382783e-06, + "loss": 0.3423, + "step": 35182 + }, + { + "epoch": 7.68, + "learning_rate": 1.1620056805767972e-06, + "loss": 1.8905, + "step": 35183 + }, + { + "epoch": 7.68, + "learning_rate": 1.1618964387153158e-06, + "loss": 2.5803, + "step": 35184 + }, + { + "epoch": 7.68, + "learning_rate": 1.1617871968538346e-06, + "loss": 0.4766, + "step": 35185 + }, + { + "epoch": 7.68, + "learning_rate": 1.1616779549923532e-06, + "loss": 1.7309, + "step": 35186 + }, + { + "epoch": 7.68, + "learning_rate": 1.1615687131308718e-06, + "loss": 0.2621, + "step": 35187 + }, + { + "epoch": 7.68, + "learning_rate": 1.1614594712693904e-06, + "loss": 0.6254, + "step": 35188 + }, + { + "epoch": 7.68, + "learning_rate": 1.1613502294079092e-06, + "loss": 0.6959, + "step": 35189 + }, + { + "epoch": 7.68, + "learning_rate": 1.1612409875464279e-06, + "loss": 1.4023, + "step": 35190 + }, + { + "epoch": 7.68, + "learning_rate": 1.1611317456849467e-06, + "loss": 1.367, + "step": 35191 + }, + { + "epoch": 7.68, + "learning_rate": 1.1610225038234653e-06, + "loss": 2.093, + "step": 35192 + }, + { + "epoch": 7.68, + "learning_rate": 1.160913261961984e-06, + "loss": 1.6841, + "step": 35193 + }, + { + "epoch": 7.68, + "learning_rate": 1.1608040201005025e-06, + "loss": 0.7084, + "step": 35194 + }, + { + "epoch": 7.68, + "learning_rate": 1.1606947782390213e-06, + "loss": 0.6792, + "step": 35195 + }, + { + "epoch": 7.68, + "learning_rate": 1.16058553637754e-06, + "loss": 0.8649, + "step": 35196 + }, + { + "epoch": 7.68, + "learning_rate": 1.1604762945160586e-06, + "loss": 1.1946, + "step": 35197 + }, + { + "epoch": 7.68, + "learning_rate": 1.1603670526545774e-06, + "loss": 1.5972, + "step": 35198 + }, + { + "epoch": 7.68, + "learning_rate": 1.160257810793096e-06, + "loss": 0.3956, + "step": 35199 + }, + { + "epoch": 7.68, + "learning_rate": 1.1601485689316146e-06, + "loss": 1.2733, + "step": 35200 + }, + { + "epoch": 7.68, + "learning_rate": 1.1600393270701334e-06, + "loss": 1.1259, + "step": 35201 + }, + { + "epoch": 7.68, + "learning_rate": 1.159930085208652e-06, + "loss": 1.2187, + "step": 35202 + }, + { + "epoch": 7.68, + "learning_rate": 1.1598208433471706e-06, + "loss": 0.6146, + "step": 35203 + }, + { + "epoch": 7.68, + "learning_rate": 1.1597116014856893e-06, + "loss": 1.8322, + "step": 35204 + }, + { + "epoch": 7.68, + "learning_rate": 1.159602359624208e-06, + "loss": 1.6327, + "step": 35205 + }, + { + "epoch": 7.68, + "learning_rate": 1.159493117762727e-06, + "loss": 0.3863, + "step": 35206 + }, + { + "epoch": 7.68, + "learning_rate": 1.1593838759012455e-06, + "loss": 0.7764, + "step": 35207 + }, + { + "epoch": 7.68, + "learning_rate": 1.1592746340397641e-06, + "loss": 1.0894, + "step": 35208 + }, + { + "epoch": 7.68, + "learning_rate": 1.1591653921782827e-06, + "loss": 1.9802, + "step": 35209 + }, + { + "epoch": 7.68, + "learning_rate": 1.1590561503168013e-06, + "loss": 0.2589, + "step": 35210 + }, + { + "epoch": 7.68, + "learning_rate": 1.1589469084553202e-06, + "loss": 1.1568, + "step": 35211 + }, + { + "epoch": 7.68, + "learning_rate": 1.158837666593839e-06, + "loss": 0.7671, + "step": 35212 + }, + { + "epoch": 7.69, + "learning_rate": 1.1587284247323576e-06, + "loss": 1.1574, + "step": 35213 + }, + { + "epoch": 7.69, + "learning_rate": 1.1586191828708762e-06, + "loss": 0.2888, + "step": 35214 + }, + { + "epoch": 7.69, + "learning_rate": 1.1585099410093948e-06, + "loss": 0.7214, + "step": 35215 + }, + { + "epoch": 7.69, + "learning_rate": 1.1584006991479134e-06, + "loss": 1.591, + "step": 35216 + }, + { + "epoch": 7.69, + "learning_rate": 1.1582914572864323e-06, + "loss": 0.8576, + "step": 35217 + }, + { + "epoch": 7.69, + "learning_rate": 1.158182215424951e-06, + "loss": 0.3601, + "step": 35218 + }, + { + "epoch": 7.69, + "learning_rate": 1.1580729735634697e-06, + "loss": 1.7821, + "step": 35219 + }, + { + "epoch": 7.69, + "learning_rate": 1.1579637317019883e-06, + "loss": 0.8137, + "step": 35220 + }, + { + "epoch": 7.69, + "learning_rate": 1.157854489840507e-06, + "loss": 0.7051, + "step": 35221 + }, + { + "epoch": 7.69, + "learning_rate": 1.1577452479790255e-06, + "loss": 0.6496, + "step": 35222 + }, + { + "epoch": 7.69, + "learning_rate": 1.1576360061175443e-06, + "loss": 1.9621, + "step": 35223 + }, + { + "epoch": 7.69, + "learning_rate": 1.1575267642560632e-06, + "loss": 0.7659, + "step": 35224 + }, + { + "epoch": 7.69, + "learning_rate": 1.1574175223945818e-06, + "loss": 2.7052, + "step": 35225 + }, + { + "epoch": 7.69, + "learning_rate": 1.1573082805331004e-06, + "loss": 1.0527, + "step": 35226 + }, + { + "epoch": 7.69, + "learning_rate": 1.157199038671619e-06, + "loss": 0.92, + "step": 35227 + }, + { + "epoch": 7.69, + "learning_rate": 1.1570897968101378e-06, + "loss": 1.4645, + "step": 35228 + }, + { + "epoch": 7.69, + "learning_rate": 1.1569805549486564e-06, + "loss": 0.7085, + "step": 35229 + }, + { + "epoch": 7.69, + "learning_rate": 1.156871313087175e-06, + "loss": 1.9889, + "step": 35230 + }, + { + "epoch": 7.69, + "learning_rate": 1.1567620712256939e-06, + "loss": 1.2241, + "step": 35231 + }, + { + "epoch": 7.69, + "learning_rate": 1.1566528293642125e-06, + "loss": 1.6341, + "step": 35232 + }, + { + "epoch": 7.69, + "learning_rate": 1.156543587502731e-06, + "loss": 0.6976, + "step": 35233 + }, + { + "epoch": 7.69, + "learning_rate": 1.15643434564125e-06, + "loss": 1.2971, + "step": 35234 + }, + { + "epoch": 7.69, + "learning_rate": 1.1563251037797685e-06, + "loss": 0.4564, + "step": 35235 + }, + { + "epoch": 7.69, + "learning_rate": 1.1562158619182871e-06, + "loss": 0.0907, + "step": 35236 + }, + { + "epoch": 7.69, + "learning_rate": 1.1561066200568058e-06, + "loss": 0.6361, + "step": 35237 + }, + { + "epoch": 7.69, + "learning_rate": 1.1559973781953246e-06, + "loss": 0.2302, + "step": 35238 + }, + { + "epoch": 7.69, + "learning_rate": 1.1558881363338432e-06, + "loss": 0.9282, + "step": 35239 + }, + { + "epoch": 7.69, + "learning_rate": 1.155778894472362e-06, + "loss": 0.7158, + "step": 35240 + }, + { + "epoch": 7.69, + "learning_rate": 1.1556696526108806e-06, + "loss": 0.9931, + "step": 35241 + }, + { + "epoch": 7.69, + "learning_rate": 1.1555604107493992e-06, + "loss": 1.3181, + "step": 35242 + }, + { + "epoch": 7.69, + "learning_rate": 1.1554511688879178e-06, + "loss": 2.0678, + "step": 35243 + }, + { + "epoch": 7.69, + "learning_rate": 1.1553419270264367e-06, + "loss": 0.5572, + "step": 35244 + }, + { + "epoch": 7.69, + "learning_rate": 1.1552326851649553e-06, + "loss": 0.831, + "step": 35245 + }, + { + "epoch": 7.69, + "learning_rate": 1.155123443303474e-06, + "loss": 1.9504, + "step": 35246 + }, + { + "epoch": 7.69, + "learning_rate": 1.1550142014419927e-06, + "loss": 1.1533, + "step": 35247 + }, + { + "epoch": 7.69, + "learning_rate": 1.1549049595805113e-06, + "loss": 1.8217, + "step": 35248 + }, + { + "epoch": 7.69, + "learning_rate": 1.15479571771903e-06, + "loss": 1.447, + "step": 35249 + }, + { + "epoch": 7.69, + "learning_rate": 1.1546864758575485e-06, + "loss": 0.9471, + "step": 35250 + }, + { + "epoch": 7.69, + "learning_rate": 1.1545772339960674e-06, + "loss": 0.397, + "step": 35251 + }, + { + "epoch": 7.69, + "learning_rate": 1.1544679921345862e-06, + "loss": 0.7079, + "step": 35252 + }, + { + "epoch": 7.69, + "learning_rate": 1.1543587502731048e-06, + "loss": 0.6741, + "step": 35253 + }, + { + "epoch": 7.69, + "learning_rate": 1.1542495084116234e-06, + "loss": 0.8348, + "step": 35254 + }, + { + "epoch": 7.69, + "learning_rate": 1.154140266550142e-06, + "loss": 1.0003, + "step": 35255 + }, + { + "epoch": 7.69, + "learning_rate": 1.1540310246886608e-06, + "loss": 0.9535, + "step": 35256 + }, + { + "epoch": 7.69, + "learning_rate": 1.1539217828271795e-06, + "loss": 0.555, + "step": 35257 + }, + { + "epoch": 7.69, + "learning_rate": 1.1538125409656983e-06, + "loss": 0.4209, + "step": 35258 + }, + { + "epoch": 7.7, + "learning_rate": 1.1537032991042169e-06, + "loss": 2.0353, + "step": 35259 + }, + { + "epoch": 7.7, + "learning_rate": 1.1535940572427355e-06, + "loss": 0.4462, + "step": 35260 + }, + { + "epoch": 7.7, + "learning_rate": 1.1534848153812541e-06, + "loss": 0.6892, + "step": 35261 + }, + { + "epoch": 7.7, + "learning_rate": 1.153375573519773e-06, + "loss": 1.169, + "step": 35262 + }, + { + "epoch": 7.7, + "learning_rate": 1.1532663316582915e-06, + "loss": 0.0632, + "step": 35263 + }, + { + "epoch": 7.7, + "learning_rate": 1.1531570897968104e-06, + "loss": 2.4165, + "step": 35264 + }, + { + "epoch": 7.7, + "learning_rate": 1.153047847935329e-06, + "loss": 1.0116, + "step": 35265 + }, + { + "epoch": 7.7, + "learning_rate": 1.1529386060738476e-06, + "loss": 2.4768, + "step": 35266 + }, + { + "epoch": 7.7, + "learning_rate": 1.1528293642123662e-06, + "loss": 0.5959, + "step": 35267 + }, + { + "epoch": 7.7, + "learning_rate": 1.152720122350885e-06, + "loss": 0.4482, + "step": 35268 + }, + { + "epoch": 7.7, + "learning_rate": 1.1526108804894036e-06, + "loss": 0.7574, + "step": 35269 + }, + { + "epoch": 7.7, + "learning_rate": 1.1525016386279222e-06, + "loss": 0.4315, + "step": 35270 + }, + { + "epoch": 7.7, + "learning_rate": 1.152392396766441e-06, + "loss": 0.7269, + "step": 35271 + }, + { + "epoch": 7.7, + "learning_rate": 1.1522831549049597e-06, + "loss": 1.0433, + "step": 35272 + }, + { + "epoch": 7.7, + "learning_rate": 1.1521739130434783e-06, + "loss": 0.9585, + "step": 35273 + }, + { + "epoch": 7.7, + "learning_rate": 1.1520646711819971e-06, + "loss": 2.3956, + "step": 35274 + }, + { + "epoch": 7.7, + "learning_rate": 1.1519554293205157e-06, + "loss": 1.8651, + "step": 35275 + }, + { + "epoch": 7.7, + "learning_rate": 1.1518461874590343e-06, + "loss": 1.0861, + "step": 35276 + }, + { + "epoch": 7.7, + "learning_rate": 1.1517369455975532e-06, + "loss": 0.2241, + "step": 35277 + }, + { + "epoch": 7.7, + "learning_rate": 1.1516277037360718e-06, + "loss": 1.1576, + "step": 35278 + }, + { + "epoch": 7.7, + "learning_rate": 1.1515184618745904e-06, + "loss": 1.9001, + "step": 35279 + }, + { + "epoch": 7.7, + "learning_rate": 1.1514092200131092e-06, + "loss": 0.9322, + "step": 35280 + }, + { + "epoch": 7.7, + "learning_rate": 1.1512999781516278e-06, + "loss": 1.9549, + "step": 35281 + }, + { + "epoch": 7.7, + "learning_rate": 1.1511907362901464e-06, + "loss": 1.0986, + "step": 35282 + }, + { + "epoch": 7.7, + "learning_rate": 1.151081494428665e-06, + "loss": 1.9853, + "step": 35283 + }, + { + "epoch": 7.7, + "learning_rate": 1.1509722525671839e-06, + "loss": 0.8205, + "step": 35284 + }, + { + "epoch": 7.7, + "learning_rate": 1.1508630107057025e-06, + "loss": 0.5012, + "step": 35285 + }, + { + "epoch": 7.7, + "learning_rate": 1.1507537688442213e-06, + "loss": 0.9799, + "step": 35286 + }, + { + "epoch": 7.7, + "learning_rate": 1.15064452698274e-06, + "loss": 1.9245, + "step": 35287 + }, + { + "epoch": 7.7, + "learning_rate": 1.1505352851212585e-06, + "loss": 2.1549, + "step": 35288 + }, + { + "epoch": 7.7, + "learning_rate": 1.1504260432597771e-06, + "loss": 0.7499, + "step": 35289 + }, + { + "epoch": 7.7, + "learning_rate": 1.150316801398296e-06, + "loss": 1.4663, + "step": 35290 + }, + { + "epoch": 7.7, + "learning_rate": 1.1502075595368146e-06, + "loss": 1.1476, + "step": 35291 + }, + { + "epoch": 7.7, + "learning_rate": 1.1500983176753334e-06, + "loss": 0.5753, + "step": 35292 + }, + { + "epoch": 7.7, + "learning_rate": 1.149989075813852e-06, + "loss": 2.4165, + "step": 35293 + }, + { + "epoch": 7.7, + "learning_rate": 1.1498798339523706e-06, + "loss": 0.5773, + "step": 35294 + }, + { + "epoch": 7.7, + "learning_rate": 1.1497705920908892e-06, + "loss": 0.7452, + "step": 35295 + }, + { + "epoch": 7.7, + "learning_rate": 1.149661350229408e-06, + "loss": 1.0282, + "step": 35296 + }, + { + "epoch": 7.7, + "learning_rate": 1.1495521083679266e-06, + "loss": 0.5891, + "step": 35297 + }, + { + "epoch": 7.7, + "learning_rate": 1.1494428665064455e-06, + "loss": 0.56, + "step": 35298 + }, + { + "epoch": 7.7, + "learning_rate": 1.149333624644964e-06, + "loss": 1.4516, + "step": 35299 + }, + { + "epoch": 7.7, + "learning_rate": 1.1492243827834827e-06, + "loss": 0.1256, + "step": 35300 + }, + { + "epoch": 7.7, + "learning_rate": 1.1491151409220013e-06, + "loss": 1.9466, + "step": 35301 + }, + { + "epoch": 7.7, + "learning_rate": 1.1490058990605201e-06, + "loss": 1.971, + "step": 35302 + }, + { + "epoch": 7.7, + "learning_rate": 1.1488966571990387e-06, + "loss": 0.4608, + "step": 35303 + }, + { + "epoch": 7.7, + "learning_rate": 1.1487874153375576e-06, + "loss": 0.9188, + "step": 35304 + }, + { + "epoch": 7.71, + "learning_rate": 1.1486781734760762e-06, + "loss": 0.0577, + "step": 35305 + }, + { + "epoch": 7.71, + "learning_rate": 1.1485689316145948e-06, + "loss": 1.7544, + "step": 35306 + }, + { + "epoch": 7.71, + "learning_rate": 1.1484596897531134e-06, + "loss": 2.5087, + "step": 35307 + }, + { + "epoch": 7.71, + "learning_rate": 1.1483504478916322e-06, + "loss": 0.5578, + "step": 35308 + }, + { + "epoch": 7.71, + "learning_rate": 1.1482412060301508e-06, + "loss": 0.1934, + "step": 35309 + }, + { + "epoch": 7.71, + "learning_rate": 1.1481319641686696e-06, + "loss": 0.6493, + "step": 35310 + }, + { + "epoch": 7.71, + "learning_rate": 1.1480227223071883e-06, + "loss": 0.738, + "step": 35311 + }, + { + "epoch": 7.71, + "learning_rate": 1.1479134804457069e-06, + "loss": 0.5348, + "step": 35312 + }, + { + "epoch": 7.71, + "learning_rate": 1.1478042385842255e-06, + "loss": 2.387, + "step": 35313 + }, + { + "epoch": 7.71, + "learning_rate": 1.1476949967227443e-06, + "loss": 1.92, + "step": 35314 + }, + { + "epoch": 7.71, + "learning_rate": 1.147585754861263e-06, + "loss": 0.3027, + "step": 35315 + }, + { + "epoch": 7.71, + "learning_rate": 1.1474765129997815e-06, + "loss": 1.0961, + "step": 35316 + }, + { + "epoch": 7.71, + "learning_rate": 1.1473672711383003e-06, + "loss": 0.2481, + "step": 35317 + }, + { + "epoch": 7.71, + "learning_rate": 1.147258029276819e-06, + "loss": 1.306, + "step": 35318 + }, + { + "epoch": 7.71, + "learning_rate": 1.1471487874153376e-06, + "loss": 1.1979, + "step": 35319 + }, + { + "epoch": 7.71, + "learning_rate": 1.1470395455538564e-06, + "loss": 1.9635, + "step": 35320 + }, + { + "epoch": 7.71, + "learning_rate": 1.146930303692375e-06, + "loss": 1.0987, + "step": 35321 + }, + { + "epoch": 7.71, + "learning_rate": 1.1468210618308936e-06, + "loss": 1.4548, + "step": 35322 + }, + { + "epoch": 7.71, + "learning_rate": 1.1467118199694122e-06, + "loss": 2.6031, + "step": 35323 + }, + { + "epoch": 7.71, + "learning_rate": 1.146602578107931e-06, + "loss": 0.1061, + "step": 35324 + }, + { + "epoch": 7.71, + "learning_rate": 1.1464933362464497e-06, + "loss": 0.5036, + "step": 35325 + }, + { + "epoch": 7.71, + "learning_rate": 1.1463840943849685e-06, + "loss": 1.9775, + "step": 35326 + }, + { + "epoch": 7.71, + "learning_rate": 1.146274852523487e-06, + "loss": 2.0791, + "step": 35327 + }, + { + "epoch": 7.71, + "learning_rate": 1.1461656106620057e-06, + "loss": 2.3238, + "step": 35328 + }, + { + "epoch": 7.71, + "learning_rate": 1.1460563688005243e-06, + "loss": 1.6866, + "step": 35329 + }, + { + "epoch": 7.71, + "learning_rate": 1.1459471269390431e-06, + "loss": 0.7445, + "step": 35330 + }, + { + "epoch": 7.71, + "learning_rate": 1.145837885077562e-06, + "loss": 0.3142, + "step": 35331 + }, + { + "epoch": 7.71, + "learning_rate": 1.1457286432160806e-06, + "loss": 0.8896, + "step": 35332 + }, + { + "epoch": 7.71, + "learning_rate": 1.1456194013545992e-06, + "loss": 2.3462, + "step": 35333 + }, + { + "epoch": 7.71, + "learning_rate": 1.1455101594931178e-06, + "loss": 1.3646, + "step": 35334 + }, + { + "epoch": 7.71, + "learning_rate": 1.1454009176316364e-06, + "loss": 1.7124, + "step": 35335 + }, + { + "epoch": 7.71, + "learning_rate": 1.1452916757701552e-06, + "loss": 1.6211, + "step": 35336 + }, + { + "epoch": 7.71, + "learning_rate": 1.145182433908674e-06, + "loss": 1.6217, + "step": 35337 + }, + { + "epoch": 7.71, + "learning_rate": 1.1450731920471927e-06, + "loss": 1.6174, + "step": 35338 + }, + { + "epoch": 7.71, + "learning_rate": 1.1449639501857113e-06, + "loss": 1.0351, + "step": 35339 + }, + { + "epoch": 7.71, + "learning_rate": 1.1448547083242299e-06, + "loss": 0.8326, + "step": 35340 + }, + { + "epoch": 7.71, + "learning_rate": 1.1447454664627485e-06, + "loss": 1.2555, + "step": 35341 + }, + { + "epoch": 7.71, + "learning_rate": 1.1446362246012673e-06, + "loss": 0.5633, + "step": 35342 + }, + { + "epoch": 7.71, + "learning_rate": 1.1445269827397861e-06, + "loss": 2.0529, + "step": 35343 + }, + { + "epoch": 7.71, + "learning_rate": 1.1444177408783047e-06, + "loss": 2.3171, + "step": 35344 + }, + { + "epoch": 7.71, + "learning_rate": 1.1443084990168234e-06, + "loss": 1.3608, + "step": 35345 + }, + { + "epoch": 7.71, + "learning_rate": 1.144199257155342e-06, + "loss": 1.1076, + "step": 35346 + }, + { + "epoch": 7.71, + "learning_rate": 1.1440900152938606e-06, + "loss": 1.2478, + "step": 35347 + }, + { + "epoch": 7.71, + "learning_rate": 1.1439807734323794e-06, + "loss": 0.098, + "step": 35348 + }, + { + "epoch": 7.71, + "learning_rate": 1.143871531570898e-06, + "loss": 0.9685, + "step": 35349 + }, + { + "epoch": 7.71, + "learning_rate": 1.1437622897094168e-06, + "loss": 0.4786, + "step": 35350 + }, + { + "epoch": 7.72, + "learning_rate": 1.1436530478479354e-06, + "loss": 2.4903, + "step": 35351 + }, + { + "epoch": 7.72, + "learning_rate": 1.143543805986454e-06, + "loss": 2.4798, + "step": 35352 + }, + { + "epoch": 7.72, + "learning_rate": 1.1434345641249729e-06, + "loss": 1.4898, + "step": 35353 + }, + { + "epoch": 7.72, + "learning_rate": 1.1433253222634915e-06, + "loss": 0.8556, + "step": 35354 + }, + { + "epoch": 7.72, + "learning_rate": 1.14321608040201e-06, + "loss": 0.8446, + "step": 35355 + }, + { + "epoch": 7.72, + "learning_rate": 1.1431068385405287e-06, + "loss": 1.68, + "step": 35356 + }, + { + "epoch": 7.72, + "learning_rate": 1.1429975966790475e-06, + "loss": 0.6838, + "step": 35357 + }, + { + "epoch": 7.72, + "learning_rate": 1.1428883548175662e-06, + "loss": 0.693, + "step": 35358 + }, + { + "epoch": 7.72, + "learning_rate": 1.142779112956085e-06, + "loss": 1.1391, + "step": 35359 + }, + { + "epoch": 7.72, + "learning_rate": 1.1426698710946036e-06, + "loss": 1.4517, + "step": 35360 + }, + { + "epoch": 7.72, + "learning_rate": 1.1425606292331222e-06, + "loss": 0.4338, + "step": 35361 + }, + { + "epoch": 7.72, + "learning_rate": 1.1424513873716408e-06, + "loss": 0.4161, + "step": 35362 + }, + { + "epoch": 7.72, + "learning_rate": 1.1423421455101596e-06, + "loss": 0.7376, + "step": 35363 + }, + { + "epoch": 7.72, + "learning_rate": 1.1422329036486782e-06, + "loss": 1.3002, + "step": 35364 + }, + { + "epoch": 7.72, + "learning_rate": 1.142123661787197e-06, + "loss": 1.6981, + "step": 35365 + }, + { + "epoch": 7.72, + "learning_rate": 1.1420144199257157e-06, + "loss": 0.4293, + "step": 35366 + }, + { + "epoch": 7.72, + "learning_rate": 1.1419051780642343e-06, + "loss": 1.7905, + "step": 35367 + }, + { + "epoch": 7.72, + "learning_rate": 1.141795936202753e-06, + "loss": 1.2589, + "step": 35368 + }, + { + "epoch": 7.72, + "learning_rate": 1.1416866943412715e-06, + "loss": 0.5421, + "step": 35369 + }, + { + "epoch": 7.72, + "learning_rate": 1.1415774524797903e-06, + "loss": 1.855, + "step": 35370 + }, + { + "epoch": 7.72, + "learning_rate": 1.1414682106183092e-06, + "loss": 1.1454, + "step": 35371 + }, + { + "epoch": 7.72, + "learning_rate": 1.1413589687568278e-06, + "loss": 0.6444, + "step": 35372 + }, + { + "epoch": 7.72, + "learning_rate": 1.1412497268953464e-06, + "loss": 1.2228, + "step": 35373 + }, + { + "epoch": 7.72, + "learning_rate": 1.141140485033865e-06, + "loss": 1.4281, + "step": 35374 + }, + { + "epoch": 7.72, + "learning_rate": 1.1410312431723836e-06, + "loss": 1.9758, + "step": 35375 + }, + { + "epoch": 7.72, + "learning_rate": 1.1409220013109024e-06, + "loss": 1.0687, + "step": 35376 + }, + { + "epoch": 7.72, + "learning_rate": 1.1408127594494212e-06, + "loss": 1.8681, + "step": 35377 + }, + { + "epoch": 7.72, + "learning_rate": 1.1407035175879399e-06, + "loss": 1.755, + "step": 35378 + }, + { + "epoch": 7.72, + "learning_rate": 1.1405942757264585e-06, + "loss": 0.8954, + "step": 35379 + }, + { + "epoch": 7.72, + "learning_rate": 1.140485033864977e-06, + "loss": 0.3112, + "step": 35380 + }, + { + "epoch": 7.72, + "learning_rate": 1.140375792003496e-06, + "loss": 0.8596, + "step": 35381 + }, + { + "epoch": 7.72, + "learning_rate": 1.1402665501420145e-06, + "loss": 2.3978, + "step": 35382 + }, + { + "epoch": 7.72, + "learning_rate": 1.1401573082805333e-06, + "loss": 0.5453, + "step": 35383 + }, + { + "epoch": 7.72, + "learning_rate": 1.140048066419052e-06, + "loss": 1.3014, + "step": 35384 + }, + { + "epoch": 7.72, + "learning_rate": 1.1399388245575706e-06, + "loss": 0.3713, + "step": 35385 + }, + { + "epoch": 7.72, + "learning_rate": 1.1398295826960892e-06, + "loss": 1.0336, + "step": 35386 + }, + { + "epoch": 7.72, + "learning_rate": 1.139720340834608e-06, + "loss": 0.3237, + "step": 35387 + }, + { + "epoch": 7.72, + "learning_rate": 1.1396110989731266e-06, + "loss": 1.5743, + "step": 35388 + }, + { + "epoch": 7.72, + "learning_rate": 1.1395018571116452e-06, + "loss": 0.5444, + "step": 35389 + }, + { + "epoch": 7.72, + "learning_rate": 1.139392615250164e-06, + "loss": 1.0565, + "step": 35390 + }, + { + "epoch": 7.72, + "learning_rate": 1.1392833733886826e-06, + "loss": 0.6322, + "step": 35391 + }, + { + "epoch": 7.72, + "learning_rate": 1.1391741315272013e-06, + "loss": 0.4127, + "step": 35392 + }, + { + "epoch": 7.72, + "learning_rate": 1.13906488966572e-06, + "loss": 0.3123, + "step": 35393 + }, + { + "epoch": 7.72, + "learning_rate": 1.1389556478042387e-06, + "loss": 1.434, + "step": 35394 + }, + { + "epoch": 7.72, + "learning_rate": 1.1388464059427573e-06, + "loss": 1.4981, + "step": 35395 + }, + { + "epoch": 7.73, + "learning_rate": 1.1387371640812761e-06, + "loss": 0.9778, + "step": 35396 + }, + { + "epoch": 7.73, + "learning_rate": 1.1386279222197947e-06, + "loss": 0.7865, + "step": 35397 + }, + { + "epoch": 7.73, + "learning_rate": 1.1385186803583133e-06, + "loss": 0.2485, + "step": 35398 + }, + { + "epoch": 7.73, + "learning_rate": 1.1384094384968322e-06, + "loss": 0.3339, + "step": 35399 + }, + { + "epoch": 7.73, + "learning_rate": 1.1383001966353508e-06, + "loss": 2.2065, + "step": 35400 + }, + { + "epoch": 7.73, + "learning_rate": 1.1381909547738694e-06, + "loss": 0.5122, + "step": 35401 + }, + { + "epoch": 7.73, + "learning_rate": 1.138081712912388e-06, + "loss": 2.9219, + "step": 35402 + }, + { + "epoch": 7.73, + "learning_rate": 1.1379724710509068e-06, + "loss": 0.199, + "step": 35403 + }, + { + "epoch": 7.73, + "learning_rate": 1.1378632291894254e-06, + "loss": 1.0304, + "step": 35404 + }, + { + "epoch": 7.73, + "learning_rate": 1.1377539873279443e-06, + "loss": 0.5987, + "step": 35405 + }, + { + "epoch": 7.73, + "learning_rate": 1.1376447454664629e-06, + "loss": 1.2885, + "step": 35406 + }, + { + "epoch": 7.73, + "learning_rate": 1.1375355036049815e-06, + "loss": 1.1826, + "step": 35407 + }, + { + "epoch": 7.73, + "learning_rate": 1.1374262617435e-06, + "loss": 1.5996, + "step": 35408 + }, + { + "epoch": 7.73, + "learning_rate": 1.137317019882019e-06, + "loss": 1.1117, + "step": 35409 + }, + { + "epoch": 7.73, + "learning_rate": 1.1372077780205375e-06, + "loss": 3.0894, + "step": 35410 + }, + { + "epoch": 7.73, + "learning_rate": 1.1370985361590563e-06, + "loss": 0.5718, + "step": 35411 + }, + { + "epoch": 7.73, + "learning_rate": 1.136989294297575e-06, + "loss": 2.4528, + "step": 35412 + }, + { + "epoch": 7.73, + "learning_rate": 1.1368800524360936e-06, + "loss": 0.3309, + "step": 35413 + }, + { + "epoch": 7.73, + "learning_rate": 1.1367708105746122e-06, + "loss": 0.4481, + "step": 35414 + }, + { + "epoch": 7.73, + "learning_rate": 1.136661568713131e-06, + "loss": 0.7289, + "step": 35415 + }, + { + "epoch": 7.73, + "learning_rate": 1.1365523268516496e-06, + "loss": 2.0454, + "step": 35416 + }, + { + "epoch": 7.73, + "learning_rate": 1.1364430849901684e-06, + "loss": 1.9194, + "step": 35417 + }, + { + "epoch": 7.73, + "learning_rate": 1.136333843128687e-06, + "loss": 1.668, + "step": 35418 + }, + { + "epoch": 7.73, + "learning_rate": 1.1362246012672057e-06, + "loss": 1.3063, + "step": 35419 + }, + { + "epoch": 7.73, + "learning_rate": 1.1361153594057243e-06, + "loss": 1.591, + "step": 35420 + }, + { + "epoch": 7.73, + "learning_rate": 1.136006117544243e-06, + "loss": 1.8742, + "step": 35421 + }, + { + "epoch": 7.73, + "learning_rate": 1.1358968756827617e-06, + "loss": 1.5289, + "step": 35422 + }, + { + "epoch": 7.73, + "learning_rate": 1.1357876338212805e-06, + "loss": 1.5285, + "step": 35423 + }, + { + "epoch": 7.73, + "learning_rate": 1.1356783919597991e-06, + "loss": 0.9029, + "step": 35424 + }, + { + "epoch": 7.73, + "learning_rate": 1.1355691500983177e-06, + "loss": 1.4217, + "step": 35425 + }, + { + "epoch": 7.73, + "learning_rate": 1.1354599082368364e-06, + "loss": 1.6938, + "step": 35426 + }, + { + "epoch": 7.73, + "learning_rate": 1.1353506663753552e-06, + "loss": 0.1994, + "step": 35427 + }, + { + "epoch": 7.73, + "learning_rate": 1.1352414245138738e-06, + "loss": 3.2153, + "step": 35428 + }, + { + "epoch": 7.73, + "learning_rate": 1.1351321826523926e-06, + "loss": 2.2453, + "step": 35429 + }, + { + "epoch": 7.73, + "learning_rate": 1.1350229407909112e-06, + "loss": 0.2996, + "step": 35430 + }, + { + "epoch": 7.73, + "learning_rate": 1.1349136989294298e-06, + "loss": 1.533, + "step": 35431 + }, + { + "epoch": 7.73, + "learning_rate": 1.1348044570679484e-06, + "loss": 1.6983, + "step": 35432 + }, + { + "epoch": 7.73, + "learning_rate": 1.1346952152064673e-06, + "loss": 1.6439, + "step": 35433 + }, + { + "epoch": 7.73, + "learning_rate": 1.1345859733449859e-06, + "loss": 1.9059, + "step": 35434 + }, + { + "epoch": 7.73, + "learning_rate": 1.1344767314835045e-06, + "loss": 1.6271, + "step": 35435 + }, + { + "epoch": 7.73, + "learning_rate": 1.1343674896220233e-06, + "loss": 1.7403, + "step": 35436 + }, + { + "epoch": 7.73, + "learning_rate": 1.134258247760542e-06, + "loss": 0.3834, + "step": 35437 + }, + { + "epoch": 7.73, + "learning_rate": 1.1341490058990605e-06, + "loss": 1.2229, + "step": 35438 + }, + { + "epoch": 7.73, + "learning_rate": 1.1340397640375794e-06, + "loss": 2.2254, + "step": 35439 + }, + { + "epoch": 7.73, + "learning_rate": 1.133930522176098e-06, + "loss": 1.494, + "step": 35440 + }, + { + "epoch": 7.73, + "learning_rate": 1.1338212803146166e-06, + "loss": 1.8175, + "step": 35441 + }, + { + "epoch": 7.74, + "learning_rate": 1.1337120384531352e-06, + "loss": 1.8101, + "step": 35442 + }, + { + "epoch": 7.74, + "learning_rate": 1.133602796591654e-06, + "loss": 1.3221, + "step": 35443 + }, + { + "epoch": 7.74, + "learning_rate": 1.1334935547301726e-06, + "loss": 0.6725, + "step": 35444 + }, + { + "epoch": 7.74, + "learning_rate": 1.1333843128686914e-06, + "loss": 0.9923, + "step": 35445 + }, + { + "epoch": 7.74, + "learning_rate": 1.13327507100721e-06, + "loss": 1.4172, + "step": 35446 + }, + { + "epoch": 7.74, + "learning_rate": 1.1331658291457287e-06, + "loss": 1.2787, + "step": 35447 + }, + { + "epoch": 7.74, + "learning_rate": 1.1330565872842473e-06, + "loss": 0.9177, + "step": 35448 + }, + { + "epoch": 7.74, + "learning_rate": 1.132947345422766e-06, + "loss": 0.468, + "step": 35449 + }, + { + "epoch": 7.74, + "learning_rate": 1.1328381035612847e-06, + "loss": 0.2347, + "step": 35450 + }, + { + "epoch": 7.74, + "learning_rate": 1.1327288616998035e-06, + "loss": 1.2494, + "step": 35451 + }, + { + "epoch": 7.74, + "learning_rate": 1.1326196198383221e-06, + "loss": 1.504, + "step": 35452 + }, + { + "epoch": 7.74, + "learning_rate": 1.1325103779768408e-06, + "loss": 1.6771, + "step": 35453 + }, + { + "epoch": 7.74, + "learning_rate": 1.1324011361153594e-06, + "loss": 1.0939, + "step": 35454 + }, + { + "epoch": 7.74, + "learning_rate": 1.1322918942538782e-06, + "loss": 0.8348, + "step": 35455 + }, + { + "epoch": 7.74, + "learning_rate": 1.132182652392397e-06, + "loss": 0.2174, + "step": 35456 + }, + { + "epoch": 7.74, + "learning_rate": 1.1320734105309156e-06, + "loss": 0.1822, + "step": 35457 + }, + { + "epoch": 7.74, + "learning_rate": 1.1319641686694342e-06, + "loss": 1.4657, + "step": 35458 + }, + { + "epoch": 7.74, + "learning_rate": 1.1318549268079528e-06, + "loss": 1.3035, + "step": 35459 + }, + { + "epoch": 7.74, + "learning_rate": 1.1317456849464715e-06, + "loss": 2.5152, + "step": 35460 + }, + { + "epoch": 7.74, + "learning_rate": 1.1316364430849903e-06, + "loss": 1.264, + "step": 35461 + }, + { + "epoch": 7.74, + "learning_rate": 1.131527201223509e-06, + "loss": 1.3682, + "step": 35462 + }, + { + "epoch": 7.74, + "learning_rate": 1.1314179593620277e-06, + "loss": 1.8485, + "step": 35463 + }, + { + "epoch": 7.74, + "learning_rate": 1.1313087175005463e-06, + "loss": 3.5153, + "step": 35464 + }, + { + "epoch": 7.74, + "learning_rate": 1.131199475639065e-06, + "loss": 0.9505, + "step": 35465 + }, + { + "epoch": 7.74, + "learning_rate": 1.1310902337775835e-06, + "loss": 2.7555, + "step": 35466 + }, + { + "epoch": 7.74, + "learning_rate": 1.1309809919161024e-06, + "loss": 1.801, + "step": 35467 + }, + { + "epoch": 7.74, + "learning_rate": 1.130871750054621e-06, + "loss": 1.9838, + "step": 35468 + }, + { + "epoch": 7.74, + "learning_rate": 1.1307625081931398e-06, + "loss": 1.2228, + "step": 35469 + }, + { + "epoch": 7.74, + "learning_rate": 1.1306532663316584e-06, + "loss": 1.0114, + "step": 35470 + }, + { + "epoch": 7.74, + "learning_rate": 1.130544024470177e-06, + "loss": 0.8233, + "step": 35471 + }, + { + "epoch": 7.74, + "learning_rate": 1.1304347826086956e-06, + "loss": 1.4088, + "step": 35472 + }, + { + "epoch": 7.74, + "learning_rate": 1.1303255407472145e-06, + "loss": 1.2604, + "step": 35473 + }, + { + "epoch": 7.74, + "learning_rate": 1.130216298885733e-06, + "loss": 0.3141, + "step": 35474 + }, + { + "epoch": 7.74, + "learning_rate": 1.1301070570242517e-06, + "loss": 0.7321, + "step": 35475 + }, + { + "epoch": 7.74, + "learning_rate": 1.1299978151627705e-06, + "loss": 1.2082, + "step": 35476 + }, + { + "epoch": 7.74, + "learning_rate": 1.1298885733012891e-06, + "loss": 0.3229, + "step": 35477 + }, + { + "epoch": 7.74, + "learning_rate": 1.129779331439808e-06, + "loss": 1.3816, + "step": 35478 + }, + { + "epoch": 7.74, + "learning_rate": 1.1296700895783266e-06, + "loss": 0.6819, + "step": 35479 + }, + { + "epoch": 7.74, + "learning_rate": 1.1295608477168452e-06, + "loss": 1.2434, + "step": 35480 + }, + { + "epoch": 7.74, + "learning_rate": 1.1294516058553638e-06, + "loss": 1.3982, + "step": 35481 + }, + { + "epoch": 7.74, + "learning_rate": 1.1293423639938826e-06, + "loss": 1.575, + "step": 35482 + }, + { + "epoch": 7.74, + "learning_rate": 1.1292331221324012e-06, + "loss": 1.759, + "step": 35483 + }, + { + "epoch": 7.74, + "learning_rate": 1.12912388027092e-06, + "loss": 1.1188, + "step": 35484 + }, + { + "epoch": 7.74, + "learning_rate": 1.1290146384094386e-06, + "loss": 1.8987, + "step": 35485 + }, + { + "epoch": 7.74, + "learning_rate": 1.1289053965479573e-06, + "loss": 1.6684, + "step": 35486 + }, + { + "epoch": 7.74, + "learning_rate": 1.1287961546864759e-06, + "loss": 1.4147, + "step": 35487 + }, + { + "epoch": 7.75, + "learning_rate": 1.1286869128249945e-06, + "loss": 1.1714, + "step": 35488 + }, + { + "epoch": 7.75, + "learning_rate": 1.1285776709635133e-06, + "loss": 0.8304, + "step": 35489 + }, + { + "epoch": 7.75, + "learning_rate": 1.1284684291020321e-06, + "loss": 0.9464, + "step": 35490 + }, + { + "epoch": 7.75, + "learning_rate": 1.1283591872405507e-06, + "loss": 1.6325, + "step": 35491 + }, + { + "epoch": 7.75, + "learning_rate": 1.1282499453790693e-06, + "loss": 1.5055, + "step": 35492 + }, + { + "epoch": 7.75, + "learning_rate": 1.128140703517588e-06, + "loss": 0.4479, + "step": 35493 + }, + { + "epoch": 7.75, + "learning_rate": 1.1280314616561066e-06, + "loss": 0.8052, + "step": 35494 + }, + { + "epoch": 7.75, + "learning_rate": 1.1279222197946254e-06, + "loss": 0.9121, + "step": 35495 + }, + { + "epoch": 7.75, + "learning_rate": 1.1278129779331442e-06, + "loss": 1.3388, + "step": 35496 + }, + { + "epoch": 7.75, + "learning_rate": 1.1277037360716628e-06, + "loss": 1.6193, + "step": 35497 + }, + { + "epoch": 7.75, + "learning_rate": 1.1275944942101814e-06, + "loss": 1.2444, + "step": 35498 + }, + { + "epoch": 7.75, + "learning_rate": 1.1274852523487e-06, + "loss": 1.1273, + "step": 35499 + }, + { + "epoch": 7.75, + "learning_rate": 1.1273760104872189e-06, + "loss": 1.6979, + "step": 35500 + }, + { + "epoch": 7.75, + "learning_rate": 1.1272667686257375e-06, + "loss": 1.6805, + "step": 35501 + }, + { + "epoch": 7.75, + "learning_rate": 1.1271575267642563e-06, + "loss": 0.4678, + "step": 35502 + }, + { + "epoch": 7.75, + "learning_rate": 1.127048284902775e-06, + "loss": 1.0658, + "step": 35503 + }, + { + "epoch": 7.75, + "learning_rate": 1.1269390430412935e-06, + "loss": 2.0921, + "step": 35504 + }, + { + "epoch": 7.75, + "learning_rate": 1.1268298011798121e-06, + "loss": 2.0889, + "step": 35505 + }, + { + "epoch": 7.75, + "learning_rate": 1.126720559318331e-06, + "loss": 0.7534, + "step": 35506 + }, + { + "epoch": 7.75, + "learning_rate": 1.1266113174568496e-06, + "loss": 2.1157, + "step": 35507 + }, + { + "epoch": 7.75, + "learning_rate": 1.1265020755953682e-06, + "loss": 2.0908, + "step": 35508 + }, + { + "epoch": 7.75, + "learning_rate": 1.126392833733887e-06, + "loss": 1.3237, + "step": 35509 + }, + { + "epoch": 7.75, + "learning_rate": 1.1262835918724056e-06, + "loss": 2.0336, + "step": 35510 + }, + { + "epoch": 7.75, + "learning_rate": 1.1261743500109242e-06, + "loss": 0.3439, + "step": 35511 + }, + { + "epoch": 7.75, + "learning_rate": 1.126065108149443e-06, + "loss": 1.1465, + "step": 35512 + }, + { + "epoch": 7.75, + "learning_rate": 1.1259558662879617e-06, + "loss": 2.3461, + "step": 35513 + }, + { + "epoch": 7.75, + "learning_rate": 1.1258466244264803e-06, + "loss": 0.6174, + "step": 35514 + }, + { + "epoch": 7.75, + "learning_rate": 1.125737382564999e-06, + "loss": 2.2382, + "step": 35515 + }, + { + "epoch": 7.75, + "learning_rate": 1.1256281407035177e-06, + "loss": 1.9021, + "step": 35516 + }, + { + "epoch": 7.75, + "learning_rate": 1.1255188988420363e-06, + "loss": 0.0843, + "step": 35517 + }, + { + "epoch": 7.75, + "learning_rate": 1.1254096569805551e-06, + "loss": 0.4973, + "step": 35518 + }, + { + "epoch": 7.75, + "learning_rate": 1.1253004151190737e-06, + "loss": 2.3848, + "step": 35519 + }, + { + "epoch": 7.75, + "learning_rate": 1.1251911732575924e-06, + "loss": 0.5889, + "step": 35520 + }, + { + "epoch": 7.75, + "learning_rate": 1.125081931396111e-06, + "loss": 0.7224, + "step": 35521 + }, + { + "epoch": 7.75, + "learning_rate": 1.1249726895346298e-06, + "loss": 1.0478, + "step": 35522 + }, + { + "epoch": 7.75, + "learning_rate": 1.1248634476731484e-06, + "loss": 1.0855, + "step": 35523 + }, + { + "epoch": 7.75, + "learning_rate": 1.1247542058116672e-06, + "loss": 0.6111, + "step": 35524 + }, + { + "epoch": 7.75, + "learning_rate": 1.1246449639501858e-06, + "loss": 0.6824, + "step": 35525 + }, + { + "epoch": 7.75, + "learning_rate": 1.1245357220887044e-06, + "loss": 1.1933, + "step": 35526 + }, + { + "epoch": 7.75, + "learning_rate": 1.124426480227223e-06, + "loss": 0.7557, + "step": 35527 + }, + { + "epoch": 7.75, + "learning_rate": 1.1243172383657419e-06, + "loss": 0.8857, + "step": 35528 + }, + { + "epoch": 7.75, + "learning_rate": 1.1242079965042605e-06, + "loss": 1.7535, + "step": 35529 + }, + { + "epoch": 7.75, + "learning_rate": 1.1240987546427793e-06, + "loss": 0.7507, + "step": 35530 + }, + { + "epoch": 7.75, + "learning_rate": 1.123989512781298e-06, + "loss": 0.7378, + "step": 35531 + }, + { + "epoch": 7.75, + "learning_rate": 1.1238802709198165e-06, + "loss": 1.4717, + "step": 35532 + }, + { + "epoch": 7.75, + "learning_rate": 1.1237710290583351e-06, + "loss": 2.0925, + "step": 35533 + }, + { + "epoch": 7.76, + "learning_rate": 1.123661787196854e-06, + "loss": 0.1917, + "step": 35534 + }, + { + "epoch": 7.76, + "learning_rate": 1.1235525453353726e-06, + "loss": 0.3624, + "step": 35535 + }, + { + "epoch": 7.76, + "learning_rate": 1.1234433034738914e-06, + "loss": 1.9944, + "step": 35536 + }, + { + "epoch": 7.76, + "learning_rate": 1.12333406161241e-06, + "loss": 1.3437, + "step": 35537 + }, + { + "epoch": 7.76, + "learning_rate": 1.1232248197509286e-06, + "loss": 1.7637, + "step": 35538 + }, + { + "epoch": 7.76, + "learning_rate": 1.1231155778894472e-06, + "loss": 1.6409, + "step": 35539 + }, + { + "epoch": 7.76, + "learning_rate": 1.123006336027966e-06, + "loss": 1.332, + "step": 35540 + }, + { + "epoch": 7.76, + "learning_rate": 1.1228970941664847e-06, + "loss": 2.7378, + "step": 35541 + }, + { + "epoch": 7.76, + "learning_rate": 1.1227878523050035e-06, + "loss": 1.084, + "step": 35542 + }, + { + "epoch": 7.76, + "learning_rate": 1.122678610443522e-06, + "loss": 1.2037, + "step": 35543 + }, + { + "epoch": 7.76, + "learning_rate": 1.1225693685820407e-06, + "loss": 0.8895, + "step": 35544 + }, + { + "epoch": 7.76, + "learning_rate": 1.1224601267205593e-06, + "loss": 0.9483, + "step": 35545 + }, + { + "epoch": 7.76, + "learning_rate": 1.1223508848590781e-06, + "loss": 0.5959, + "step": 35546 + }, + { + "epoch": 7.76, + "learning_rate": 1.1222416429975968e-06, + "loss": 1.0182, + "step": 35547 + }, + { + "epoch": 7.76, + "learning_rate": 1.1221324011361156e-06, + "loss": 0.7768, + "step": 35548 + }, + { + "epoch": 7.76, + "learning_rate": 1.1220231592746342e-06, + "loss": 1.8415, + "step": 35549 + }, + { + "epoch": 7.76, + "learning_rate": 1.1219139174131528e-06, + "loss": 0.6424, + "step": 35550 + }, + { + "epoch": 7.76, + "learning_rate": 1.1218046755516714e-06, + "loss": 0.9719, + "step": 35551 + }, + { + "epoch": 7.76, + "learning_rate": 1.1216954336901902e-06, + "loss": 0.6039, + "step": 35552 + }, + { + "epoch": 7.76, + "learning_rate": 1.1215861918287088e-06, + "loss": 1.9637, + "step": 35553 + }, + { + "epoch": 7.76, + "learning_rate": 1.1214769499672275e-06, + "loss": 1.7547, + "step": 35554 + }, + { + "epoch": 7.76, + "learning_rate": 1.1213677081057463e-06, + "loss": 1.4538, + "step": 35555 + }, + { + "epoch": 7.76, + "learning_rate": 1.1212584662442649e-06, + "loss": 0.6722, + "step": 35556 + }, + { + "epoch": 7.76, + "learning_rate": 1.1211492243827835e-06, + "loss": 1.5995, + "step": 35557 + }, + { + "epoch": 7.76, + "learning_rate": 1.1210399825213023e-06, + "loss": 1.7757, + "step": 35558 + }, + { + "epoch": 7.76, + "learning_rate": 1.120930740659821e-06, + "loss": 0.1187, + "step": 35559 + }, + { + "epoch": 7.76, + "learning_rate": 1.1208214987983395e-06, + "loss": 1.011, + "step": 35560 + }, + { + "epoch": 7.76, + "learning_rate": 1.1207122569368582e-06, + "loss": 1.3411, + "step": 35561 + }, + { + "epoch": 7.76, + "learning_rate": 1.120603015075377e-06, + "loss": 1.9056, + "step": 35562 + }, + { + "epoch": 7.76, + "learning_rate": 1.1204937732138956e-06, + "loss": 0.8261, + "step": 35563 + }, + { + "epoch": 7.76, + "learning_rate": 1.1203845313524144e-06, + "loss": 1.8744, + "step": 35564 + }, + { + "epoch": 7.76, + "learning_rate": 1.120275289490933e-06, + "loss": 0.7172, + "step": 35565 + }, + { + "epoch": 7.76, + "learning_rate": 1.1201660476294516e-06, + "loss": 1.8399, + "step": 35566 + }, + { + "epoch": 7.76, + "learning_rate": 1.1200568057679702e-06, + "loss": 0.0888, + "step": 35567 + }, + { + "epoch": 7.76, + "learning_rate": 1.119947563906489e-06, + "loss": 1.2584, + "step": 35568 + }, + { + "epoch": 7.76, + "learning_rate": 1.1198383220450077e-06, + "loss": 0.31, + "step": 35569 + }, + { + "epoch": 7.76, + "learning_rate": 1.1197290801835265e-06, + "loss": 1.1119, + "step": 35570 + }, + { + "epoch": 7.76, + "learning_rate": 1.1196198383220451e-06, + "loss": 0.6402, + "step": 35571 + }, + { + "epoch": 7.76, + "learning_rate": 1.1195105964605637e-06, + "loss": 0.0948, + "step": 35572 + }, + { + "epoch": 7.76, + "learning_rate": 1.1194013545990823e-06, + "loss": 2.0399, + "step": 35573 + }, + { + "epoch": 7.76, + "learning_rate": 1.1192921127376012e-06, + "loss": 1.4685, + "step": 35574 + }, + { + "epoch": 7.76, + "learning_rate": 1.11918287087612e-06, + "loss": 4.0452, + "step": 35575 + }, + { + "epoch": 7.76, + "learning_rate": 1.1190736290146386e-06, + "loss": 1.977, + "step": 35576 + }, + { + "epoch": 7.76, + "learning_rate": 1.1189643871531572e-06, + "loss": 3.0316, + "step": 35577 + }, + { + "epoch": 7.76, + "learning_rate": 1.1188551452916758e-06, + "loss": 0.3127, + "step": 35578 + }, + { + "epoch": 7.76, + "learning_rate": 1.1187459034301944e-06, + "loss": 1.081, + "step": 35579 + }, + { + "epoch": 7.77, + "learning_rate": 1.1186366615687132e-06, + "loss": 1.4129, + "step": 35580 + }, + { + "epoch": 7.77, + "learning_rate": 1.118527419707232e-06, + "loss": 2.4308, + "step": 35581 + }, + { + "epoch": 7.77, + "learning_rate": 1.1184181778457507e-06, + "loss": 3.185, + "step": 35582 + }, + { + "epoch": 7.77, + "learning_rate": 1.1183089359842693e-06, + "loss": 1.2207, + "step": 35583 + }, + { + "epoch": 7.77, + "learning_rate": 1.118199694122788e-06, + "loss": 2.0897, + "step": 35584 + }, + { + "epoch": 7.77, + "learning_rate": 1.1180904522613065e-06, + "loss": 0.3408, + "step": 35585 + }, + { + "epoch": 7.77, + "learning_rate": 1.1179812103998253e-06, + "loss": 1.4689, + "step": 35586 + }, + { + "epoch": 7.77, + "learning_rate": 1.117871968538344e-06, + "loss": 0.2713, + "step": 35587 + }, + { + "epoch": 7.77, + "learning_rate": 1.1177627266768628e-06, + "loss": 1.2469, + "step": 35588 + }, + { + "epoch": 7.77, + "learning_rate": 1.1176534848153814e-06, + "loss": 0.351, + "step": 35589 + }, + { + "epoch": 7.77, + "learning_rate": 1.1175442429539e-06, + "loss": 0.5683, + "step": 35590 + }, + { + "epoch": 7.77, + "learning_rate": 1.1174350010924186e-06, + "loss": 0.9439, + "step": 35591 + }, + { + "epoch": 7.77, + "learning_rate": 1.1173257592309374e-06, + "loss": 0.4418, + "step": 35592 + }, + { + "epoch": 7.77, + "learning_rate": 1.117216517369456e-06, + "loss": 2.0277, + "step": 35593 + }, + { + "epoch": 7.77, + "learning_rate": 1.1171072755079746e-06, + "loss": 0.6073, + "step": 35594 + }, + { + "epoch": 7.77, + "learning_rate": 1.1169980336464935e-06, + "loss": 0.6965, + "step": 35595 + }, + { + "epoch": 7.77, + "learning_rate": 1.116888791785012e-06, + "loss": 0.726, + "step": 35596 + }, + { + "epoch": 7.77, + "learning_rate": 1.1167795499235307e-06, + "loss": 0.6681, + "step": 35597 + }, + { + "epoch": 7.77, + "learning_rate": 1.1166703080620495e-06, + "loss": 1.0078, + "step": 35598 + }, + { + "epoch": 7.77, + "learning_rate": 1.1165610662005681e-06, + "loss": 0.5046, + "step": 35599 + }, + { + "epoch": 7.77, + "learning_rate": 1.1164518243390867e-06, + "loss": 1.3382, + "step": 35600 + }, + { + "epoch": 7.77, + "learning_rate": 1.1163425824776056e-06, + "loss": 1.2272, + "step": 35601 + }, + { + "epoch": 7.77, + "learning_rate": 1.1162333406161242e-06, + "loss": 1.9695, + "step": 35602 + }, + { + "epoch": 7.77, + "learning_rate": 1.116124098754643e-06, + "loss": 0.4521, + "step": 35603 + }, + { + "epoch": 7.77, + "learning_rate": 1.1160148568931616e-06, + "loss": 0.8715, + "step": 35604 + }, + { + "epoch": 7.77, + "learning_rate": 1.1159056150316802e-06, + "loss": 0.6065, + "step": 35605 + }, + { + "epoch": 7.77, + "learning_rate": 1.1157963731701988e-06, + "loss": 0.7336, + "step": 35606 + }, + { + "epoch": 7.77, + "learning_rate": 1.1156871313087174e-06, + "loss": 1.1764, + "step": 35607 + }, + { + "epoch": 7.77, + "learning_rate": 1.1155778894472363e-06, + "loss": 0.7577, + "step": 35608 + }, + { + "epoch": 7.77, + "learning_rate": 1.115468647585755e-06, + "loss": 1.7075, + "step": 35609 + }, + { + "epoch": 7.77, + "learning_rate": 1.1153594057242737e-06, + "loss": 0.7838, + "step": 35610 + }, + { + "epoch": 7.77, + "learning_rate": 1.1152501638627923e-06, + "loss": 0.3957, + "step": 35611 + }, + { + "epoch": 7.77, + "learning_rate": 1.115140922001311e-06, + "loss": 0.8408, + "step": 35612 + }, + { + "epoch": 7.77, + "learning_rate": 1.1150316801398295e-06, + "loss": 0.3873, + "step": 35613 + }, + { + "epoch": 7.77, + "learning_rate": 1.1149224382783484e-06, + "loss": 0.883, + "step": 35614 + }, + { + "epoch": 7.77, + "learning_rate": 1.1148131964168672e-06, + "loss": 0.8293, + "step": 35615 + }, + { + "epoch": 7.77, + "learning_rate": 1.1147039545553858e-06, + "loss": 0.5231, + "step": 35616 + }, + { + "epoch": 7.77, + "learning_rate": 1.1145947126939044e-06, + "loss": 1.3275, + "step": 35617 + }, + { + "epoch": 7.77, + "learning_rate": 1.114485470832423e-06, + "loss": 1.2095, + "step": 35618 + }, + { + "epoch": 7.77, + "learning_rate": 1.1143762289709416e-06, + "loss": 1.5526, + "step": 35619 + }, + { + "epoch": 7.77, + "learning_rate": 1.1142669871094604e-06, + "loss": 0.7548, + "step": 35620 + }, + { + "epoch": 7.77, + "learning_rate": 1.1141577452479793e-06, + "loss": 1.4289, + "step": 35621 + }, + { + "epoch": 7.77, + "learning_rate": 1.1140485033864979e-06, + "loss": 0.5868, + "step": 35622 + }, + { + "epoch": 7.77, + "learning_rate": 1.1139392615250165e-06, + "loss": 1.9133, + "step": 35623 + }, + { + "epoch": 7.77, + "learning_rate": 1.113830019663535e-06, + "loss": 0.1905, + "step": 35624 + }, + { + "epoch": 7.77, + "learning_rate": 1.113720777802054e-06, + "loss": 1.3997, + "step": 35625 + }, + { + "epoch": 7.78, + "learning_rate": 1.1136115359405725e-06, + "loss": 1.4602, + "step": 35626 + }, + { + "epoch": 7.78, + "learning_rate": 1.1135022940790911e-06, + "loss": 1.7723, + "step": 35627 + }, + { + "epoch": 7.78, + "learning_rate": 1.11339305221761e-06, + "loss": 0.4963, + "step": 35628 + }, + { + "epoch": 7.78, + "learning_rate": 1.1132838103561286e-06, + "loss": 1.8215, + "step": 35629 + }, + { + "epoch": 7.78, + "learning_rate": 1.1131745684946472e-06, + "loss": 0.2166, + "step": 35630 + }, + { + "epoch": 7.78, + "learning_rate": 1.113065326633166e-06, + "loss": 1.3415, + "step": 35631 + }, + { + "epoch": 7.78, + "learning_rate": 1.1129560847716846e-06, + "loss": 2.5688, + "step": 35632 + }, + { + "epoch": 7.78, + "learning_rate": 1.1128468429102032e-06, + "loss": 0.1733, + "step": 35633 + }, + { + "epoch": 7.78, + "learning_rate": 1.112737601048722e-06, + "loss": 2.2225, + "step": 35634 + }, + { + "epoch": 7.78, + "learning_rate": 1.1126283591872407e-06, + "loss": 0.637, + "step": 35635 + }, + { + "epoch": 7.78, + "learning_rate": 1.1125191173257593e-06, + "loss": 2.1177, + "step": 35636 + }, + { + "epoch": 7.78, + "learning_rate": 1.112409875464278e-06, + "loss": 0.6322, + "step": 35637 + }, + { + "epoch": 7.78, + "learning_rate": 1.1123006336027967e-06, + "loss": 1.8098, + "step": 35638 + }, + { + "epoch": 7.78, + "learning_rate": 1.1121913917413153e-06, + "loss": 1.3592, + "step": 35639 + }, + { + "epoch": 7.78, + "learning_rate": 1.112082149879834e-06, + "loss": 0.9218, + "step": 35640 + }, + { + "epoch": 7.78, + "learning_rate": 1.1119729080183528e-06, + "loss": 0.9148, + "step": 35641 + }, + { + "epoch": 7.78, + "learning_rate": 1.1118636661568714e-06, + "loss": 1.8556, + "step": 35642 + }, + { + "epoch": 7.78, + "learning_rate": 1.1117544242953902e-06, + "loss": 0.7473, + "step": 35643 + }, + { + "epoch": 7.78, + "learning_rate": 1.1116451824339088e-06, + "loss": 1.6515, + "step": 35644 + }, + { + "epoch": 7.78, + "learning_rate": 1.1115359405724274e-06, + "loss": 1.035, + "step": 35645 + }, + { + "epoch": 7.78, + "learning_rate": 1.111426698710946e-06, + "loss": 0.9653, + "step": 35646 + }, + { + "epoch": 7.78, + "learning_rate": 1.1113174568494648e-06, + "loss": 0.0488, + "step": 35647 + }, + { + "epoch": 7.78, + "learning_rate": 1.1112082149879835e-06, + "loss": 1.6696, + "step": 35648 + }, + { + "epoch": 7.78, + "learning_rate": 1.1110989731265023e-06, + "loss": 0.4543, + "step": 35649 + }, + { + "epoch": 7.78, + "learning_rate": 1.1109897312650209e-06, + "loss": 0.0929, + "step": 35650 + }, + { + "epoch": 7.78, + "learning_rate": 1.1108804894035395e-06, + "loss": 1.6019, + "step": 35651 + }, + { + "epoch": 7.78, + "learning_rate": 1.1107712475420581e-06, + "loss": 0.7134, + "step": 35652 + }, + { + "epoch": 7.78, + "learning_rate": 1.110662005680577e-06, + "loss": 0.9083, + "step": 35653 + }, + { + "epoch": 7.78, + "learning_rate": 1.1105527638190955e-06, + "loss": 1.8492, + "step": 35654 + }, + { + "epoch": 7.78, + "learning_rate": 1.1104435219576144e-06, + "loss": 1.3135, + "step": 35655 + }, + { + "epoch": 7.78, + "learning_rate": 1.110334280096133e-06, + "loss": 0.9387, + "step": 35656 + }, + { + "epoch": 7.78, + "learning_rate": 1.1102250382346516e-06, + "loss": 0.5868, + "step": 35657 + }, + { + "epoch": 7.78, + "learning_rate": 1.1101157963731702e-06, + "loss": 2.1305, + "step": 35658 + }, + { + "epoch": 7.78, + "learning_rate": 1.110006554511689e-06, + "loss": 1.8094, + "step": 35659 + }, + { + "epoch": 7.78, + "learning_rate": 1.1098973126502076e-06, + "loss": 0.606, + "step": 35660 + }, + { + "epoch": 7.78, + "learning_rate": 1.1097880707887265e-06, + "loss": 2.5703, + "step": 35661 + }, + { + "epoch": 7.78, + "learning_rate": 1.109678828927245e-06, + "loss": 1.4099, + "step": 35662 + }, + { + "epoch": 7.78, + "learning_rate": 1.1095695870657637e-06, + "loss": 3.2955, + "step": 35663 + }, + { + "epoch": 7.78, + "learning_rate": 1.1094603452042823e-06, + "loss": 0.8908, + "step": 35664 + }, + { + "epoch": 7.78, + "learning_rate": 1.1093511033428011e-06, + "loss": 3.1168, + "step": 35665 + }, + { + "epoch": 7.78, + "learning_rate": 1.1092418614813197e-06, + "loss": 0.3995, + "step": 35666 + }, + { + "epoch": 7.78, + "learning_rate": 1.1091326196198385e-06, + "loss": 2.01, + "step": 35667 + }, + { + "epoch": 7.78, + "learning_rate": 1.1090233777583572e-06, + "loss": 0.3949, + "step": 35668 + }, + { + "epoch": 7.78, + "learning_rate": 1.1089141358968758e-06, + "loss": 0.713, + "step": 35669 + }, + { + "epoch": 7.78, + "learning_rate": 1.1088048940353944e-06, + "loss": 0.6785, + "step": 35670 + }, + { + "epoch": 7.79, + "learning_rate": 1.1086956521739132e-06, + "loss": 0.4939, + "step": 35671 + }, + { + "epoch": 7.79, + "learning_rate": 1.1085864103124318e-06, + "loss": 0.486, + "step": 35672 + }, + { + "epoch": 7.79, + "learning_rate": 1.1084771684509504e-06, + "loss": 0.1917, + "step": 35673 + }, + { + "epoch": 7.79, + "learning_rate": 1.1083679265894692e-06, + "loss": 2.235, + "step": 35674 + }, + { + "epoch": 7.79, + "learning_rate": 1.1082586847279879e-06, + "loss": 1.1431, + "step": 35675 + }, + { + "epoch": 7.79, + "learning_rate": 1.1081494428665065e-06, + "loss": 0.4441, + "step": 35676 + }, + { + "epoch": 7.79, + "learning_rate": 1.1080402010050253e-06, + "loss": 1.0493, + "step": 35677 + }, + { + "epoch": 7.79, + "learning_rate": 1.107930959143544e-06, + "loss": 0.6474, + "step": 35678 + }, + { + "epoch": 7.79, + "learning_rate": 1.1078217172820625e-06, + "loss": 0.1158, + "step": 35679 + }, + { + "epoch": 7.79, + "learning_rate": 1.1077124754205811e-06, + "loss": 1.9731, + "step": 35680 + }, + { + "epoch": 7.79, + "learning_rate": 1.1076032335591e-06, + "loss": 1.1633, + "step": 35681 + }, + { + "epoch": 7.79, + "learning_rate": 1.1074939916976186e-06, + "loss": 0.1411, + "step": 35682 + }, + { + "epoch": 7.79, + "learning_rate": 1.1073847498361374e-06, + "loss": 2.0983, + "step": 35683 + }, + { + "epoch": 7.79, + "learning_rate": 1.107275507974656e-06, + "loss": 1.2324, + "step": 35684 + }, + { + "epoch": 7.79, + "learning_rate": 1.1071662661131746e-06, + "loss": 1.1041, + "step": 35685 + }, + { + "epoch": 7.79, + "learning_rate": 1.1070570242516932e-06, + "loss": 0.6586, + "step": 35686 + }, + { + "epoch": 7.79, + "learning_rate": 1.106947782390212e-06, + "loss": 1.2326, + "step": 35687 + }, + { + "epoch": 7.79, + "learning_rate": 1.1068385405287306e-06, + "loss": 1.6047, + "step": 35688 + }, + { + "epoch": 7.79, + "learning_rate": 1.1067292986672495e-06, + "loss": 1.0756, + "step": 35689 + }, + { + "epoch": 7.79, + "learning_rate": 1.106620056805768e-06, + "loss": 0.2152, + "step": 35690 + }, + { + "epoch": 7.79, + "learning_rate": 1.1065108149442867e-06, + "loss": 0.325, + "step": 35691 + }, + { + "epoch": 7.79, + "learning_rate": 1.1064015730828053e-06, + "loss": 1.0741, + "step": 35692 + }, + { + "epoch": 7.79, + "learning_rate": 1.1062923312213241e-06, + "loss": 1.625, + "step": 35693 + }, + { + "epoch": 7.79, + "learning_rate": 1.1061830893598427e-06, + "loss": 0.5403, + "step": 35694 + }, + { + "epoch": 7.79, + "learning_rate": 1.1060738474983616e-06, + "loss": 1.887, + "step": 35695 + }, + { + "epoch": 7.79, + "learning_rate": 1.1059646056368802e-06, + "loss": 2.6662, + "step": 35696 + }, + { + "epoch": 7.79, + "learning_rate": 1.1058553637753988e-06, + "loss": 0.6925, + "step": 35697 + }, + { + "epoch": 7.79, + "learning_rate": 1.1057461219139174e-06, + "loss": 0.459, + "step": 35698 + }, + { + "epoch": 7.79, + "learning_rate": 1.1056368800524362e-06, + "loss": 0.3411, + "step": 35699 + }, + { + "epoch": 7.79, + "learning_rate": 1.105527638190955e-06, + "loss": 0.4584, + "step": 35700 + }, + { + "epoch": 7.79, + "learning_rate": 1.1054183963294736e-06, + "loss": 1.7607, + "step": 35701 + }, + { + "epoch": 7.79, + "learning_rate": 1.1053091544679923e-06, + "loss": 1.9981, + "step": 35702 + }, + { + "epoch": 7.79, + "learning_rate": 1.1051999126065109e-06, + "loss": 1.6057, + "step": 35703 + }, + { + "epoch": 7.79, + "learning_rate": 1.1050906707450295e-06, + "loss": 0.5269, + "step": 35704 + }, + { + "epoch": 7.79, + "learning_rate": 1.1049814288835483e-06, + "loss": 0.2796, + "step": 35705 + }, + { + "epoch": 7.79, + "learning_rate": 1.104872187022067e-06, + "loss": 2.2342, + "step": 35706 + }, + { + "epoch": 7.79, + "learning_rate": 1.1047629451605857e-06, + "loss": 1.1814, + "step": 35707 + }, + { + "epoch": 7.79, + "learning_rate": 1.1046537032991043e-06, + "loss": 0.6206, + "step": 35708 + }, + { + "epoch": 7.79, + "learning_rate": 1.104544461437623e-06, + "loss": 1.3089, + "step": 35709 + }, + { + "epoch": 7.79, + "learning_rate": 1.1044352195761416e-06, + "loss": 0.3601, + "step": 35710 + }, + { + "epoch": 7.79, + "learning_rate": 1.1043259777146604e-06, + "loss": 3.3586, + "step": 35711 + }, + { + "epoch": 7.79, + "learning_rate": 1.104216735853179e-06, + "loss": 0.6449, + "step": 35712 + }, + { + "epoch": 7.79, + "learning_rate": 1.1041074939916976e-06, + "loss": 0.1426, + "step": 35713 + }, + { + "epoch": 7.79, + "learning_rate": 1.1039982521302164e-06, + "loss": 0.3782, + "step": 35714 + }, + { + "epoch": 7.79, + "learning_rate": 1.103889010268735e-06, + "loss": 0.9867, + "step": 35715 + }, + { + "epoch": 7.79, + "learning_rate": 1.1037797684072537e-06, + "loss": 1.8215, + "step": 35716 + }, + { + "epoch": 7.8, + "learning_rate": 1.1036705265457725e-06, + "loss": 2.7725, + "step": 35717 + }, + { + "epoch": 7.8, + "learning_rate": 1.103561284684291e-06, + "loss": 1.0497, + "step": 35718 + }, + { + "epoch": 7.8, + "learning_rate": 1.1034520428228097e-06, + "loss": 2.4543, + "step": 35719 + }, + { + "epoch": 7.8, + "learning_rate": 1.1033428009613285e-06, + "loss": 1.6767, + "step": 35720 + }, + { + "epoch": 7.8, + "learning_rate": 1.1032335590998471e-06, + "loss": 1.3732, + "step": 35721 + }, + { + "epoch": 7.8, + "learning_rate": 1.103124317238366e-06, + "loss": 0.4332, + "step": 35722 + }, + { + "epoch": 7.8, + "learning_rate": 1.1030150753768846e-06, + "loss": 0.4908, + "step": 35723 + }, + { + "epoch": 7.8, + "learning_rate": 1.1029058335154032e-06, + "loss": 3.26, + "step": 35724 + }, + { + "epoch": 7.8, + "learning_rate": 1.1027965916539218e-06, + "loss": 2.7825, + "step": 35725 + }, + { + "epoch": 7.8, + "learning_rate": 1.1026873497924404e-06, + "loss": 0.378, + "step": 35726 + }, + { + "epoch": 7.8, + "learning_rate": 1.1025781079309592e-06, + "loss": 1.6457, + "step": 35727 + }, + { + "epoch": 7.8, + "learning_rate": 1.102468866069478e-06, + "loss": 0.7961, + "step": 35728 + }, + { + "epoch": 7.8, + "learning_rate": 1.1023596242079967e-06, + "loss": 1.4598, + "step": 35729 + }, + { + "epoch": 7.8, + "learning_rate": 1.1022503823465153e-06, + "loss": 1.4487, + "step": 35730 + }, + { + "epoch": 7.8, + "learning_rate": 1.1021411404850339e-06, + "loss": 0.2283, + "step": 35731 + }, + { + "epoch": 7.8, + "learning_rate": 1.1020318986235525e-06, + "loss": 1.376, + "step": 35732 + }, + { + "epoch": 7.8, + "learning_rate": 1.1019226567620713e-06, + "loss": 0.325, + "step": 35733 + }, + { + "epoch": 7.8, + "learning_rate": 1.1018134149005901e-06, + "loss": 1.033, + "step": 35734 + }, + { + "epoch": 7.8, + "learning_rate": 1.1017041730391088e-06, + "loss": 1.7311, + "step": 35735 + }, + { + "epoch": 7.8, + "learning_rate": 1.1015949311776274e-06, + "loss": 1.2457, + "step": 35736 + }, + { + "epoch": 7.8, + "learning_rate": 1.101485689316146e-06, + "loss": 1.4298, + "step": 35737 + }, + { + "epoch": 7.8, + "learning_rate": 1.1013764474546646e-06, + "loss": 1.1971, + "step": 35738 + }, + { + "epoch": 7.8, + "learning_rate": 1.1012672055931834e-06, + "loss": 0.63, + "step": 35739 + }, + { + "epoch": 7.8, + "learning_rate": 1.1011579637317022e-06, + "loss": 0.1476, + "step": 35740 + }, + { + "epoch": 7.8, + "learning_rate": 1.1010487218702208e-06, + "loss": 0.1689, + "step": 35741 + }, + { + "epoch": 7.8, + "learning_rate": 1.1009394800087395e-06, + "loss": 0.8469, + "step": 35742 + }, + { + "epoch": 7.8, + "learning_rate": 1.100830238147258e-06, + "loss": 0.0604, + "step": 35743 + }, + { + "epoch": 7.8, + "learning_rate": 1.1007209962857767e-06, + "loss": 0.7044, + "step": 35744 + }, + { + "epoch": 7.8, + "learning_rate": 1.1006117544242955e-06, + "loss": 1.0007, + "step": 35745 + }, + { + "epoch": 7.8, + "learning_rate": 1.1005025125628141e-06, + "loss": 0.5875, + "step": 35746 + }, + { + "epoch": 7.8, + "learning_rate": 1.100393270701333e-06, + "loss": 1.5782, + "step": 35747 + }, + { + "epoch": 7.8, + "learning_rate": 1.1002840288398515e-06, + "loss": 1.1849, + "step": 35748 + }, + { + "epoch": 7.8, + "learning_rate": 1.1001747869783702e-06, + "loss": 0.9541, + "step": 35749 + }, + { + "epoch": 7.8, + "learning_rate": 1.100065545116889e-06, + "loss": 0.0776, + "step": 35750 + }, + { + "epoch": 7.8, + "learning_rate": 1.0999563032554076e-06, + "loss": 0.6356, + "step": 35751 + }, + { + "epoch": 7.8, + "learning_rate": 1.0998470613939262e-06, + "loss": 0.1871, + "step": 35752 + }, + { + "epoch": 7.8, + "learning_rate": 1.099737819532445e-06, + "loss": 2.6396, + "step": 35753 + }, + { + "epoch": 7.8, + "learning_rate": 1.0996285776709636e-06, + "loss": 0.8687, + "step": 35754 + }, + { + "epoch": 7.8, + "learning_rate": 1.0995193358094822e-06, + "loss": 1.0636, + "step": 35755 + }, + { + "epoch": 7.8, + "learning_rate": 1.099410093948001e-06, + "loss": 1.7405, + "step": 35756 + }, + { + "epoch": 7.8, + "learning_rate": 1.0993008520865197e-06, + "loss": 1.2428, + "step": 35757 + }, + { + "epoch": 7.8, + "learning_rate": 1.0991916102250383e-06, + "loss": 0.514, + "step": 35758 + }, + { + "epoch": 7.8, + "learning_rate": 1.099082368363557e-06, + "loss": 1.0662, + "step": 35759 + }, + { + "epoch": 7.8, + "learning_rate": 1.0989731265020757e-06, + "loss": 0.3156, + "step": 35760 + }, + { + "epoch": 7.8, + "learning_rate": 1.0988638846405943e-06, + "loss": 2.4563, + "step": 35761 + }, + { + "epoch": 7.8, + "learning_rate": 1.0987546427791132e-06, + "loss": 0.874, + "step": 35762 + }, + { + "epoch": 7.81, + "learning_rate": 1.0986454009176318e-06, + "loss": 1.1988, + "step": 35763 + }, + { + "epoch": 7.81, + "learning_rate": 1.0985361590561504e-06, + "loss": 0.7677, + "step": 35764 + }, + { + "epoch": 7.81, + "learning_rate": 1.098426917194669e-06, + "loss": 0.5573, + "step": 35765 + }, + { + "epoch": 7.81, + "learning_rate": 1.0983176753331876e-06, + "loss": 1.5643, + "step": 35766 + }, + { + "epoch": 7.81, + "learning_rate": 1.0982084334717064e-06, + "loss": 2.0766, + "step": 35767 + }, + { + "epoch": 7.81, + "learning_rate": 1.0980991916102252e-06, + "loss": 0.2023, + "step": 35768 + }, + { + "epoch": 7.81, + "learning_rate": 1.0979899497487439e-06, + "loss": 1.8031, + "step": 35769 + }, + { + "epoch": 7.81, + "learning_rate": 1.0978807078872625e-06, + "loss": 1.6494, + "step": 35770 + }, + { + "epoch": 7.81, + "learning_rate": 1.097771466025781e-06, + "loss": 0.2831, + "step": 35771 + }, + { + "epoch": 7.81, + "learning_rate": 1.0976622241643e-06, + "loss": 0.7201, + "step": 35772 + }, + { + "epoch": 7.81, + "learning_rate": 1.0975529823028185e-06, + "loss": 1.8108, + "step": 35773 + }, + { + "epoch": 7.81, + "learning_rate": 1.0974437404413373e-06, + "loss": 1.2857, + "step": 35774 + }, + { + "epoch": 7.81, + "learning_rate": 1.097334498579856e-06, + "loss": 0.6977, + "step": 35775 + }, + { + "epoch": 7.81, + "learning_rate": 1.0972252567183746e-06, + "loss": 2.8392, + "step": 35776 + }, + { + "epoch": 7.81, + "learning_rate": 1.0971160148568932e-06, + "loss": 2.64, + "step": 35777 + }, + { + "epoch": 7.81, + "learning_rate": 1.097006772995412e-06, + "loss": 2.0133, + "step": 35778 + }, + { + "epoch": 7.81, + "learning_rate": 1.0968975311339306e-06, + "loss": 0.5124, + "step": 35779 + }, + { + "epoch": 7.81, + "learning_rate": 1.0967882892724494e-06, + "loss": 0.9685, + "step": 35780 + }, + { + "epoch": 7.81, + "learning_rate": 1.096679047410968e-06, + "loss": 0.3879, + "step": 35781 + }, + { + "epoch": 7.81, + "learning_rate": 1.0965698055494866e-06, + "loss": 0.5508, + "step": 35782 + }, + { + "epoch": 7.81, + "learning_rate": 1.0964605636880053e-06, + "loss": 1.8835, + "step": 35783 + }, + { + "epoch": 7.81, + "learning_rate": 1.096351321826524e-06, + "loss": 0.8007, + "step": 35784 + }, + { + "epoch": 7.81, + "learning_rate": 1.0962420799650427e-06, + "loss": 1.7862, + "step": 35785 + }, + { + "epoch": 7.81, + "learning_rate": 1.0961328381035615e-06, + "loss": 2.4927, + "step": 35786 + }, + { + "epoch": 7.81, + "learning_rate": 1.0960235962420801e-06, + "loss": 2.4767, + "step": 35787 + }, + { + "epoch": 7.81, + "learning_rate": 1.0959143543805987e-06, + "loss": 1.0221, + "step": 35788 + }, + { + "epoch": 7.81, + "learning_rate": 1.0958051125191173e-06, + "loss": 1.2367, + "step": 35789 + }, + { + "epoch": 7.81, + "learning_rate": 1.0956958706576362e-06, + "loss": 1.3119, + "step": 35790 + }, + { + "epoch": 7.81, + "learning_rate": 1.0955866287961548e-06, + "loss": 0.4107, + "step": 35791 + }, + { + "epoch": 7.81, + "learning_rate": 1.0954773869346734e-06, + "loss": 0.4452, + "step": 35792 + }, + { + "epoch": 7.81, + "learning_rate": 1.0953681450731922e-06, + "loss": 1.9586, + "step": 35793 + }, + { + "epoch": 7.81, + "learning_rate": 1.0952589032117108e-06, + "loss": 1.8398, + "step": 35794 + }, + { + "epoch": 7.81, + "learning_rate": 1.0951496613502294e-06, + "loss": 0.6409, + "step": 35795 + }, + { + "epoch": 7.81, + "learning_rate": 1.0950404194887483e-06, + "loss": 3.0691, + "step": 35796 + }, + { + "epoch": 7.81, + "learning_rate": 1.0949311776272669e-06, + "loss": 0.5932, + "step": 35797 + }, + { + "epoch": 7.81, + "learning_rate": 1.0948219357657855e-06, + "loss": 0.1746, + "step": 35798 + }, + { + "epoch": 7.81, + "learning_rate": 1.094712693904304e-06, + "loss": 1.0239, + "step": 35799 + }, + { + "epoch": 7.81, + "learning_rate": 1.094603452042823e-06, + "loss": 0.7919, + "step": 35800 + }, + { + "epoch": 7.81, + "learning_rate": 1.0944942101813415e-06, + "loss": 0.9011, + "step": 35801 + }, + { + "epoch": 7.81, + "learning_rate": 1.0943849683198603e-06, + "loss": 1.3925, + "step": 35802 + }, + { + "epoch": 7.81, + "learning_rate": 1.094275726458379e-06, + "loss": 0.7776, + "step": 35803 + }, + { + "epoch": 7.81, + "learning_rate": 1.0941664845968976e-06, + "loss": 1.4782, + "step": 35804 + }, + { + "epoch": 7.81, + "learning_rate": 1.0940572427354162e-06, + "loss": 0.7764, + "step": 35805 + }, + { + "epoch": 7.81, + "learning_rate": 1.093948000873935e-06, + "loss": 1.0059, + "step": 35806 + }, + { + "epoch": 7.81, + "learning_rate": 1.0938387590124536e-06, + "loss": 0.2349, + "step": 35807 + }, + { + "epoch": 7.81, + "learning_rate": 1.0937295171509724e-06, + "loss": 1.4424, + "step": 35808 + }, + { + "epoch": 7.82, + "learning_rate": 1.093620275289491e-06, + "loss": 1.6521, + "step": 35809 + }, + { + "epoch": 7.82, + "learning_rate": 1.0935110334280097e-06, + "loss": 0.4862, + "step": 35810 + }, + { + "epoch": 7.82, + "learning_rate": 1.0934017915665283e-06, + "loss": 1.4568, + "step": 35811 + }, + { + "epoch": 7.82, + "learning_rate": 1.093292549705047e-06, + "loss": 0.9853, + "step": 35812 + }, + { + "epoch": 7.82, + "learning_rate": 1.0931833078435657e-06, + "loss": 1.6957, + "step": 35813 + }, + { + "epoch": 7.82, + "learning_rate": 1.0930740659820845e-06, + "loss": 2.9669, + "step": 35814 + }, + { + "epoch": 7.82, + "learning_rate": 1.0929648241206031e-06, + "loss": 0.9595, + "step": 35815 + }, + { + "epoch": 7.82, + "learning_rate": 1.0928555822591217e-06, + "loss": 0.4664, + "step": 35816 + }, + { + "epoch": 7.82, + "learning_rate": 1.0927463403976404e-06, + "loss": 1.8065, + "step": 35817 + }, + { + "epoch": 7.82, + "learning_rate": 1.0926370985361592e-06, + "loss": 3.2801, + "step": 35818 + }, + { + "epoch": 7.82, + "learning_rate": 1.0925278566746778e-06, + "loss": 0.5345, + "step": 35819 + }, + { + "epoch": 7.82, + "learning_rate": 1.0924186148131966e-06, + "loss": 1.5329, + "step": 35820 + }, + { + "epoch": 7.82, + "learning_rate": 1.0923093729517152e-06, + "loss": 0.5873, + "step": 35821 + }, + { + "epoch": 7.82, + "learning_rate": 1.0922001310902338e-06, + "loss": 1.1931, + "step": 35822 + }, + { + "epoch": 7.82, + "learning_rate": 1.0920908892287524e-06, + "loss": 1.2286, + "step": 35823 + }, + { + "epoch": 7.82, + "learning_rate": 1.0919816473672713e-06, + "loss": 0.5977, + "step": 35824 + }, + { + "epoch": 7.82, + "learning_rate": 1.0918724055057899e-06, + "loss": 1.4379, + "step": 35825 + }, + { + "epoch": 7.82, + "learning_rate": 1.0917631636443087e-06, + "loss": 0.2346, + "step": 35826 + }, + { + "epoch": 7.82, + "learning_rate": 1.0916539217828273e-06, + "loss": 2.0878, + "step": 35827 + }, + { + "epoch": 7.82, + "learning_rate": 1.091544679921346e-06, + "loss": 2.2881, + "step": 35828 + }, + { + "epoch": 7.82, + "learning_rate": 1.0914354380598645e-06, + "loss": 1.0903, + "step": 35829 + }, + { + "epoch": 7.82, + "learning_rate": 1.0913261961983834e-06, + "loss": 1.2364, + "step": 35830 + }, + { + "epoch": 7.82, + "learning_rate": 1.091216954336902e-06, + "loss": 2.2793, + "step": 35831 + }, + { + "epoch": 7.82, + "learning_rate": 1.0911077124754206e-06, + "loss": 0.2518, + "step": 35832 + }, + { + "epoch": 7.82, + "learning_rate": 1.0909984706139394e-06, + "loss": 1.5134, + "step": 35833 + }, + { + "epoch": 7.82, + "learning_rate": 1.090889228752458e-06, + "loss": 1.2319, + "step": 35834 + }, + { + "epoch": 7.82, + "learning_rate": 1.0907799868909766e-06, + "loss": 1.6455, + "step": 35835 + }, + { + "epoch": 7.82, + "learning_rate": 1.0906707450294954e-06, + "loss": 1.0052, + "step": 35836 + }, + { + "epoch": 7.82, + "learning_rate": 1.090561503168014e-06, + "loss": 1.9708, + "step": 35837 + }, + { + "epoch": 7.82, + "learning_rate": 1.0904522613065327e-06, + "loss": 1.2149, + "step": 35838 + }, + { + "epoch": 7.82, + "learning_rate": 1.0903430194450515e-06, + "loss": 0.9078, + "step": 35839 + }, + { + "epoch": 7.82, + "learning_rate": 1.09023377758357e-06, + "loss": 1.1479, + "step": 35840 + }, + { + "epoch": 7.82, + "learning_rate": 1.0901245357220887e-06, + "loss": 0.572, + "step": 35841 + }, + { + "epoch": 7.82, + "learning_rate": 1.0900152938606075e-06, + "loss": 1.5757, + "step": 35842 + }, + { + "epoch": 7.82, + "learning_rate": 1.0899060519991261e-06, + "loss": 0.1406, + "step": 35843 + }, + { + "epoch": 7.82, + "learning_rate": 1.0897968101376448e-06, + "loss": 1.1517, + "step": 35844 + }, + { + "epoch": 7.82, + "learning_rate": 1.0896875682761634e-06, + "loss": 0.7187, + "step": 35845 + }, + { + "epoch": 7.82, + "learning_rate": 1.0895783264146822e-06, + "loss": 0.4269, + "step": 35846 + }, + { + "epoch": 7.82, + "learning_rate": 1.089469084553201e-06, + "loss": 1.5805, + "step": 35847 + }, + { + "epoch": 7.82, + "learning_rate": 1.0893598426917196e-06, + "loss": 0.3689, + "step": 35848 + }, + { + "epoch": 7.82, + "learning_rate": 1.0892506008302382e-06, + "loss": 1.1591, + "step": 35849 + }, + { + "epoch": 7.82, + "learning_rate": 1.0891413589687568e-06, + "loss": 0.9801, + "step": 35850 + }, + { + "epoch": 7.82, + "learning_rate": 1.0890321171072755e-06, + "loss": 0.7456, + "step": 35851 + }, + { + "epoch": 7.82, + "learning_rate": 1.0889228752457943e-06, + "loss": 4.1201, + "step": 35852 + }, + { + "epoch": 7.82, + "learning_rate": 1.088813633384313e-06, + "loss": 1.095, + "step": 35853 + }, + { + "epoch": 7.82, + "learning_rate": 1.0887043915228317e-06, + "loss": 0.7162, + "step": 35854 + }, + { + "epoch": 7.83, + "learning_rate": 1.0885951496613503e-06, + "loss": 0.0734, + "step": 35855 + }, + { + "epoch": 7.83, + "learning_rate": 1.088485907799869e-06, + "loss": 1.6958, + "step": 35856 + }, + { + "epoch": 7.83, + "learning_rate": 1.0883766659383875e-06, + "loss": 0.1754, + "step": 35857 + }, + { + "epoch": 7.83, + "learning_rate": 1.0882674240769064e-06, + "loss": 0.3143, + "step": 35858 + }, + { + "epoch": 7.83, + "learning_rate": 1.0881581822154252e-06, + "loss": 0.7166, + "step": 35859 + }, + { + "epoch": 7.83, + "learning_rate": 1.0880489403539438e-06, + "loss": 2.1647, + "step": 35860 + }, + { + "epoch": 7.83, + "learning_rate": 1.0879396984924624e-06, + "loss": 2.9155, + "step": 35861 + }, + { + "epoch": 7.83, + "learning_rate": 1.087830456630981e-06, + "loss": 0.7012, + "step": 35862 + }, + { + "epoch": 7.83, + "learning_rate": 1.0877212147694996e-06, + "loss": 0.5545, + "step": 35863 + }, + { + "epoch": 7.83, + "learning_rate": 1.0876119729080185e-06, + "loss": 0.8755, + "step": 35864 + }, + { + "epoch": 7.83, + "learning_rate": 1.087502731046537e-06, + "loss": 1.3962, + "step": 35865 + }, + { + "epoch": 7.83, + "learning_rate": 1.087393489185056e-06, + "loss": 2.1937, + "step": 35866 + }, + { + "epoch": 7.83, + "learning_rate": 1.0872842473235745e-06, + "loss": 1.7198, + "step": 35867 + }, + { + "epoch": 7.83, + "learning_rate": 1.0871750054620931e-06, + "loss": 1.7279, + "step": 35868 + }, + { + "epoch": 7.83, + "learning_rate": 1.0870657636006117e-06, + "loss": 1.2405, + "step": 35869 + }, + { + "epoch": 7.83, + "learning_rate": 1.0869565217391306e-06, + "loss": 1.4887, + "step": 35870 + }, + { + "epoch": 7.83, + "learning_rate": 1.0868472798776492e-06, + "loss": 0.3594, + "step": 35871 + }, + { + "epoch": 7.83, + "learning_rate": 1.086738038016168e-06, + "loss": 0.9789, + "step": 35872 + }, + { + "epoch": 7.83, + "learning_rate": 1.0866287961546866e-06, + "loss": 1.9342, + "step": 35873 + }, + { + "epoch": 7.83, + "learning_rate": 1.0865195542932052e-06, + "loss": 1.2562, + "step": 35874 + }, + { + "epoch": 7.83, + "learning_rate": 1.086410312431724e-06, + "loss": 3.1394, + "step": 35875 + }, + { + "epoch": 7.83, + "learning_rate": 1.0863010705702426e-06, + "loss": 0.564, + "step": 35876 + }, + { + "epoch": 7.83, + "learning_rate": 1.0861918287087613e-06, + "loss": 0.961, + "step": 35877 + }, + { + "epoch": 7.83, + "learning_rate": 1.0860825868472799e-06, + "loss": 1.0943, + "step": 35878 + }, + { + "epoch": 7.83, + "learning_rate": 1.0859733449857987e-06, + "loss": 1.2188, + "step": 35879 + }, + { + "epoch": 7.83, + "learning_rate": 1.0858641031243173e-06, + "loss": 1.6047, + "step": 35880 + }, + { + "epoch": 7.83, + "learning_rate": 1.0857548612628361e-06, + "loss": 0.0837, + "step": 35881 + }, + { + "epoch": 7.83, + "learning_rate": 1.0856456194013547e-06, + "loss": 0.4864, + "step": 35882 + }, + { + "epoch": 7.83, + "learning_rate": 1.0855363775398733e-06, + "loss": 1.9229, + "step": 35883 + }, + { + "epoch": 7.83, + "learning_rate": 1.085427135678392e-06, + "loss": 1.7262, + "step": 35884 + }, + { + "epoch": 7.83, + "learning_rate": 1.0853178938169106e-06, + "loss": 0.5583, + "step": 35885 + }, + { + "epoch": 7.83, + "learning_rate": 1.0852086519554294e-06, + "loss": 0.9757, + "step": 35886 + }, + { + "epoch": 7.83, + "learning_rate": 1.0850994100939482e-06, + "loss": 0.732, + "step": 35887 + }, + { + "epoch": 7.83, + "learning_rate": 1.0849901682324668e-06, + "loss": 1.2998, + "step": 35888 + }, + { + "epoch": 7.83, + "learning_rate": 1.0848809263709854e-06, + "loss": 0.6534, + "step": 35889 + }, + { + "epoch": 7.83, + "learning_rate": 1.084771684509504e-06, + "loss": 0.2107, + "step": 35890 + }, + { + "epoch": 7.83, + "learning_rate": 1.0846624426480227e-06, + "loss": 0.0723, + "step": 35891 + }, + { + "epoch": 7.83, + "learning_rate": 1.0845532007865415e-06, + "loss": 0.6488, + "step": 35892 + }, + { + "epoch": 7.83, + "learning_rate": 1.0844439589250603e-06, + "loss": 0.767, + "step": 35893 + }, + { + "epoch": 7.83, + "learning_rate": 1.084334717063579e-06, + "loss": 0.4565, + "step": 35894 + }, + { + "epoch": 7.83, + "learning_rate": 1.0842254752020975e-06, + "loss": 0.0547, + "step": 35895 + }, + { + "epoch": 7.83, + "learning_rate": 1.0841162333406161e-06, + "loss": 1.6396, + "step": 35896 + }, + { + "epoch": 7.83, + "learning_rate": 1.084006991479135e-06, + "loss": 0.8073, + "step": 35897 + }, + { + "epoch": 7.83, + "learning_rate": 1.0838977496176536e-06, + "loss": 1.1677, + "step": 35898 + }, + { + "epoch": 7.83, + "learning_rate": 1.0837885077561724e-06, + "loss": 0.9854, + "step": 35899 + }, + { + "epoch": 7.84, + "learning_rate": 1.083679265894691e-06, + "loss": 0.75, + "step": 35900 + }, + { + "epoch": 7.84, + "learning_rate": 1.0835700240332096e-06, + "loss": 1.5618, + "step": 35901 + }, + { + "epoch": 7.84, + "learning_rate": 1.0834607821717282e-06, + "loss": 0.9111, + "step": 35902 + }, + { + "epoch": 7.84, + "learning_rate": 1.083351540310247e-06, + "loss": 0.6322, + "step": 35903 + }, + { + "epoch": 7.84, + "learning_rate": 1.0832422984487657e-06, + "loss": 0.4817, + "step": 35904 + }, + { + "epoch": 7.84, + "learning_rate": 1.0831330565872845e-06, + "loss": 0.0909, + "step": 35905 + }, + { + "epoch": 7.84, + "learning_rate": 1.083023814725803e-06, + "loss": 1.5005, + "step": 35906 + }, + { + "epoch": 7.84, + "learning_rate": 1.0829145728643217e-06, + "loss": 2.3508, + "step": 35907 + }, + { + "epoch": 7.84, + "learning_rate": 1.0828053310028403e-06, + "loss": 0.594, + "step": 35908 + }, + { + "epoch": 7.84, + "learning_rate": 1.0826960891413591e-06, + "loss": 0.7245, + "step": 35909 + }, + { + "epoch": 7.84, + "learning_rate": 1.0825868472798777e-06, + "loss": 1.0223, + "step": 35910 + }, + { + "epoch": 7.84, + "learning_rate": 1.0824776054183964e-06, + "loss": 1.7951, + "step": 35911 + }, + { + "epoch": 7.84, + "learning_rate": 1.0823683635569152e-06, + "loss": 0.6133, + "step": 35912 + }, + { + "epoch": 7.84, + "learning_rate": 1.0822591216954338e-06, + "loss": 0.2973, + "step": 35913 + }, + { + "epoch": 7.84, + "learning_rate": 1.0821498798339524e-06, + "loss": 1.7228, + "step": 35914 + }, + { + "epoch": 7.84, + "learning_rate": 1.0820406379724712e-06, + "loss": 1.5261, + "step": 35915 + }, + { + "epoch": 7.84, + "learning_rate": 1.0819313961109898e-06, + "loss": 1.2966, + "step": 35916 + }, + { + "epoch": 7.84, + "learning_rate": 1.0818221542495084e-06, + "loss": 2.0256, + "step": 35917 + }, + { + "epoch": 7.84, + "learning_rate": 1.081712912388027e-06, + "loss": 0.5803, + "step": 35918 + }, + { + "epoch": 7.84, + "learning_rate": 1.0816036705265459e-06, + "loss": 0.395, + "step": 35919 + }, + { + "epoch": 7.84, + "learning_rate": 1.0814944286650645e-06, + "loss": 0.405, + "step": 35920 + }, + { + "epoch": 7.84, + "learning_rate": 1.0813851868035833e-06, + "loss": 1.1913, + "step": 35921 + }, + { + "epoch": 7.84, + "learning_rate": 1.081275944942102e-06, + "loss": 1.8358, + "step": 35922 + }, + { + "epoch": 7.84, + "learning_rate": 1.0811667030806205e-06, + "loss": 1.2066, + "step": 35923 + }, + { + "epoch": 7.84, + "learning_rate": 1.0810574612191391e-06, + "loss": 1.8122, + "step": 35924 + }, + { + "epoch": 7.84, + "learning_rate": 1.080948219357658e-06, + "loss": 0.7759, + "step": 35925 + }, + { + "epoch": 7.84, + "learning_rate": 1.0808389774961766e-06, + "loss": 0.7396, + "step": 35926 + }, + { + "epoch": 7.84, + "learning_rate": 1.0807297356346954e-06, + "loss": 0.9063, + "step": 35927 + }, + { + "epoch": 7.84, + "learning_rate": 1.080620493773214e-06, + "loss": 1.5067, + "step": 35928 + }, + { + "epoch": 7.84, + "learning_rate": 1.0805112519117326e-06, + "loss": 0.4148, + "step": 35929 + }, + { + "epoch": 7.84, + "learning_rate": 1.0804020100502512e-06, + "loss": 2.362, + "step": 35930 + }, + { + "epoch": 7.84, + "learning_rate": 1.08029276818877e-06, + "loss": 2.1834, + "step": 35931 + }, + { + "epoch": 7.84, + "learning_rate": 1.0801835263272887e-06, + "loss": 0.3074, + "step": 35932 + }, + { + "epoch": 7.84, + "learning_rate": 1.0800742844658075e-06, + "loss": 1.5418, + "step": 35933 + }, + { + "epoch": 7.84, + "learning_rate": 1.079965042604326e-06, + "loss": 0.993, + "step": 35934 + }, + { + "epoch": 7.84, + "learning_rate": 1.0798558007428447e-06, + "loss": 0.0808, + "step": 35935 + }, + { + "epoch": 7.84, + "learning_rate": 1.0797465588813633e-06, + "loss": 1.8948, + "step": 35936 + }, + { + "epoch": 7.84, + "learning_rate": 1.0796373170198821e-06, + "loss": 0.8699, + "step": 35937 + }, + { + "epoch": 7.84, + "learning_rate": 1.0795280751584008e-06, + "loss": 1.8262, + "step": 35938 + }, + { + "epoch": 7.84, + "learning_rate": 1.0794188332969196e-06, + "loss": 1.4451, + "step": 35939 + }, + { + "epoch": 7.84, + "learning_rate": 1.0793095914354382e-06, + "loss": 0.8213, + "step": 35940 + }, + { + "epoch": 7.84, + "learning_rate": 1.0792003495739568e-06, + "loss": 1.4576, + "step": 35941 + }, + { + "epoch": 7.84, + "learning_rate": 1.0790911077124754e-06, + "loss": 0.2299, + "step": 35942 + }, + { + "epoch": 7.84, + "learning_rate": 1.0789818658509942e-06, + "loss": 0.4366, + "step": 35943 + }, + { + "epoch": 7.84, + "learning_rate": 1.0788726239895128e-06, + "loss": 0.4009, + "step": 35944 + }, + { + "epoch": 7.84, + "learning_rate": 1.0787633821280317e-06, + "loss": 2.4095, + "step": 35945 + }, + { + "epoch": 7.85, + "learning_rate": 1.0786541402665503e-06, + "loss": 2.0039, + "step": 35946 + }, + { + "epoch": 7.85, + "learning_rate": 1.0785448984050689e-06, + "loss": 0.6631, + "step": 35947 + }, + { + "epoch": 7.85, + "learning_rate": 1.0784356565435875e-06, + "loss": 1.7433, + "step": 35948 + }, + { + "epoch": 7.85, + "learning_rate": 1.0783264146821063e-06, + "loss": 1.1673, + "step": 35949 + }, + { + "epoch": 7.85, + "learning_rate": 1.078217172820625e-06, + "loss": 0.2083, + "step": 35950 + }, + { + "epoch": 7.85, + "learning_rate": 1.0781079309591435e-06, + "loss": 0.6673, + "step": 35951 + }, + { + "epoch": 7.85, + "learning_rate": 1.0779986890976624e-06, + "loss": 1.1068, + "step": 35952 + }, + { + "epoch": 7.85, + "learning_rate": 1.077889447236181e-06, + "loss": 1.2163, + "step": 35953 + }, + { + "epoch": 7.85, + "learning_rate": 1.0777802053746996e-06, + "loss": 0.2194, + "step": 35954 + }, + { + "epoch": 7.85, + "learning_rate": 1.0776709635132184e-06, + "loss": 1.5655, + "step": 35955 + }, + { + "epoch": 7.85, + "learning_rate": 1.077561721651737e-06, + "loss": 0.8142, + "step": 35956 + }, + { + "epoch": 7.85, + "learning_rate": 1.0774524797902556e-06, + "loss": 0.5965, + "step": 35957 + }, + { + "epoch": 7.85, + "learning_rate": 1.0773432379287745e-06, + "loss": 1.0944, + "step": 35958 + }, + { + "epoch": 7.85, + "learning_rate": 1.077233996067293e-06, + "loss": 0.7454, + "step": 35959 + }, + { + "epoch": 7.85, + "learning_rate": 1.0771247542058117e-06, + "loss": 0.575, + "step": 35960 + }, + { + "epoch": 7.85, + "learning_rate": 1.0770155123443305e-06, + "loss": 0.771, + "step": 35961 + }, + { + "epoch": 7.85, + "learning_rate": 1.0769062704828491e-06, + "loss": 0.4866, + "step": 35962 + }, + { + "epoch": 7.85, + "learning_rate": 1.0767970286213677e-06, + "loss": 0.6692, + "step": 35963 + }, + { + "epoch": 7.85, + "learning_rate": 1.0766877867598863e-06, + "loss": 1.6687, + "step": 35964 + }, + { + "epoch": 7.85, + "learning_rate": 1.0765785448984052e-06, + "loss": 0.0866, + "step": 35965 + }, + { + "epoch": 7.85, + "learning_rate": 1.0764693030369238e-06, + "loss": 0.8508, + "step": 35966 + }, + { + "epoch": 7.85, + "learning_rate": 1.0763600611754426e-06, + "loss": 1.0938, + "step": 35967 + }, + { + "epoch": 7.85, + "learning_rate": 1.0762508193139612e-06, + "loss": 0.6727, + "step": 35968 + }, + { + "epoch": 7.85, + "learning_rate": 1.0761415774524798e-06, + "loss": 0.4759, + "step": 35969 + }, + { + "epoch": 7.85, + "learning_rate": 1.0760323355909984e-06, + "loss": 0.2341, + "step": 35970 + }, + { + "epoch": 7.85, + "learning_rate": 1.0759230937295172e-06, + "loss": 0.161, + "step": 35971 + }, + { + "epoch": 7.85, + "learning_rate": 1.075813851868036e-06, + "loss": 0.5847, + "step": 35972 + }, + { + "epoch": 7.85, + "learning_rate": 1.0757046100065547e-06, + "loss": 1.1877, + "step": 35973 + }, + { + "epoch": 7.85, + "learning_rate": 1.0755953681450733e-06, + "loss": 1.0225, + "step": 35974 + }, + { + "epoch": 7.85, + "learning_rate": 1.075486126283592e-06, + "loss": 1.4381, + "step": 35975 + }, + { + "epoch": 7.85, + "learning_rate": 1.0753768844221105e-06, + "loss": 0.2353, + "step": 35976 + }, + { + "epoch": 7.85, + "learning_rate": 1.0752676425606293e-06, + "loss": 1.5905, + "step": 35977 + }, + { + "epoch": 7.85, + "learning_rate": 1.0751584006991482e-06, + "loss": 0.4992, + "step": 35978 + }, + { + "epoch": 7.85, + "learning_rate": 1.0750491588376668e-06, + "loss": 3.8706, + "step": 35979 + }, + { + "epoch": 7.85, + "learning_rate": 1.0749399169761854e-06, + "loss": 1.3873, + "step": 35980 + }, + { + "epoch": 7.85, + "learning_rate": 1.074830675114704e-06, + "loss": 0.2106, + "step": 35981 + }, + { + "epoch": 7.85, + "learning_rate": 1.0747214332532226e-06, + "loss": 2.0047, + "step": 35982 + }, + { + "epoch": 7.85, + "learning_rate": 1.0746121913917414e-06, + "loss": 0.1149, + "step": 35983 + }, + { + "epoch": 7.85, + "learning_rate": 1.07450294953026e-06, + "loss": 2.2855, + "step": 35984 + }, + { + "epoch": 7.85, + "learning_rate": 1.0743937076687789e-06, + "loss": 3.6603, + "step": 35985 + }, + { + "epoch": 7.85, + "learning_rate": 1.0742844658072975e-06, + "loss": 0.8837, + "step": 35986 + }, + { + "epoch": 7.85, + "learning_rate": 1.074175223945816e-06, + "loss": 0.8908, + "step": 35987 + }, + { + "epoch": 7.85, + "learning_rate": 1.0740659820843347e-06, + "loss": 1.7207, + "step": 35988 + }, + { + "epoch": 7.85, + "learning_rate": 1.0739567402228535e-06, + "loss": 3.0065, + "step": 35989 + }, + { + "epoch": 7.85, + "learning_rate": 1.0738474983613721e-06, + "loss": 1.2205, + "step": 35990 + }, + { + "epoch": 7.85, + "learning_rate": 1.073738256499891e-06, + "loss": 1.1091, + "step": 35991 + }, + { + "epoch": 7.86, + "learning_rate": 1.0736290146384096e-06, + "loss": 0.4039, + "step": 35992 + }, + { + "epoch": 7.86, + "learning_rate": 1.0735197727769282e-06, + "loss": 2.7076, + "step": 35993 + }, + { + "epoch": 7.86, + "learning_rate": 1.073410530915447e-06, + "loss": 1.7723, + "step": 35994 + }, + { + "epoch": 7.86, + "learning_rate": 1.0733012890539656e-06, + "loss": 0.0918, + "step": 35995 + }, + { + "epoch": 7.86, + "learning_rate": 1.0731920471924842e-06, + "loss": 2.0567, + "step": 35996 + }, + { + "epoch": 7.86, + "learning_rate": 1.0730828053310028e-06, + "loss": 1.8215, + "step": 35997 + }, + { + "epoch": 7.86, + "learning_rate": 1.0729735634695217e-06, + "loss": 1.739, + "step": 35998 + }, + { + "epoch": 7.86, + "learning_rate": 1.0728643216080403e-06, + "loss": 0.6658, + "step": 35999 + }, + { + "epoch": 7.86, + "learning_rate": 1.072755079746559e-06, + "loss": 0.8878, + "step": 36000 + }, + { + "epoch": 7.86, + "learning_rate": 1.0726458378850777e-06, + "loss": 1.8442, + "step": 36001 + }, + { + "epoch": 7.86, + "learning_rate": 1.0725365960235963e-06, + "loss": 0.6605, + "step": 36002 + }, + { + "epoch": 7.86, + "learning_rate": 1.072427354162115e-06, + "loss": 0.6902, + "step": 36003 + }, + { + "epoch": 7.86, + "learning_rate": 1.0723181123006335e-06, + "loss": 0.9842, + "step": 36004 + }, + { + "epoch": 7.86, + "learning_rate": 1.0722088704391524e-06, + "loss": 0.3786, + "step": 36005 + }, + { + "epoch": 7.86, + "learning_rate": 1.0720996285776712e-06, + "loss": 0.2185, + "step": 36006 + }, + { + "epoch": 7.86, + "learning_rate": 1.0719903867161898e-06, + "loss": 1.337, + "step": 36007 + }, + { + "epoch": 7.86, + "learning_rate": 1.0718811448547084e-06, + "loss": 1.1226, + "step": 36008 + }, + { + "epoch": 7.86, + "learning_rate": 1.071771902993227e-06, + "loss": 3.7378, + "step": 36009 + }, + { + "epoch": 7.86, + "learning_rate": 1.0716626611317456e-06, + "loss": 1.8363, + "step": 36010 + }, + { + "epoch": 7.86, + "learning_rate": 1.0715534192702644e-06, + "loss": 0.2764, + "step": 36011 + }, + { + "epoch": 7.86, + "learning_rate": 1.0714441774087833e-06, + "loss": 0.7108, + "step": 36012 + }, + { + "epoch": 7.86, + "learning_rate": 1.0713349355473019e-06, + "loss": 0.4333, + "step": 36013 + }, + { + "epoch": 7.86, + "learning_rate": 1.0712256936858205e-06, + "loss": 0.6007, + "step": 36014 + }, + { + "epoch": 7.86, + "learning_rate": 1.071116451824339e-06, + "loss": 0.4089, + "step": 36015 + }, + { + "epoch": 7.86, + "learning_rate": 1.0710072099628577e-06, + "loss": 1.122, + "step": 36016 + }, + { + "epoch": 7.86, + "learning_rate": 1.0708979681013765e-06, + "loss": 0.9521, + "step": 36017 + }, + { + "epoch": 7.86, + "learning_rate": 1.0707887262398954e-06, + "loss": 1.1033, + "step": 36018 + }, + { + "epoch": 7.86, + "learning_rate": 1.070679484378414e-06, + "loss": 0.2732, + "step": 36019 + }, + { + "epoch": 7.86, + "learning_rate": 1.0705702425169326e-06, + "loss": 0.1968, + "step": 36020 + }, + { + "epoch": 7.86, + "learning_rate": 1.0704610006554512e-06, + "loss": 1.0771, + "step": 36021 + }, + { + "epoch": 7.86, + "learning_rate": 1.07035175879397e-06, + "loss": 0.7035, + "step": 36022 + }, + { + "epoch": 7.86, + "learning_rate": 1.0702425169324886e-06, + "loss": 0.0424, + "step": 36023 + }, + { + "epoch": 7.86, + "learning_rate": 1.0701332750710074e-06, + "loss": 0.0706, + "step": 36024 + }, + { + "epoch": 7.86, + "learning_rate": 1.070024033209526e-06, + "loss": 0.9992, + "step": 36025 + }, + { + "epoch": 7.86, + "learning_rate": 1.0699147913480447e-06, + "loss": 0.6994, + "step": 36026 + }, + { + "epoch": 7.86, + "learning_rate": 1.0698055494865633e-06, + "loss": 1.1354, + "step": 36027 + }, + { + "epoch": 7.86, + "learning_rate": 1.069696307625082e-06, + "loss": 0.2349, + "step": 36028 + }, + { + "epoch": 7.86, + "learning_rate": 1.0695870657636007e-06, + "loss": 1.9485, + "step": 36029 + }, + { + "epoch": 7.86, + "learning_rate": 1.0694778239021193e-06, + "loss": 1.9068, + "step": 36030 + }, + { + "epoch": 7.86, + "learning_rate": 1.0693685820406381e-06, + "loss": 0.7559, + "step": 36031 + }, + { + "epoch": 7.86, + "learning_rate": 1.0692593401791568e-06, + "loss": 1.0086, + "step": 36032 + }, + { + "epoch": 7.86, + "learning_rate": 1.0691500983176754e-06, + "loss": 0.5285, + "step": 36033 + }, + { + "epoch": 7.86, + "learning_rate": 1.0690408564561942e-06, + "loss": 0.8899, + "step": 36034 + }, + { + "epoch": 7.86, + "learning_rate": 1.0689316145947128e-06, + "loss": 1.9317, + "step": 36035 + }, + { + "epoch": 7.86, + "learning_rate": 1.0688223727332314e-06, + "loss": 0.7144, + "step": 36036 + }, + { + "epoch": 7.86, + "learning_rate": 1.06871313087175e-06, + "loss": 0.2305, + "step": 36037 + }, + { + "epoch": 7.87, + "learning_rate": 1.0686038890102688e-06, + "loss": 0.5523, + "step": 36038 + }, + { + "epoch": 7.87, + "learning_rate": 1.0684946471487875e-06, + "loss": 0.8207, + "step": 36039 + }, + { + "epoch": 7.87, + "learning_rate": 1.0683854052873063e-06, + "loss": 0.7583, + "step": 36040 + }, + { + "epoch": 7.87, + "learning_rate": 1.0682761634258249e-06, + "loss": 0.1101, + "step": 36041 + }, + { + "epoch": 7.87, + "learning_rate": 1.0681669215643435e-06, + "loss": 1.9626, + "step": 36042 + }, + { + "epoch": 7.87, + "learning_rate": 1.0680576797028621e-06, + "loss": 1.0246, + "step": 36043 + }, + { + "epoch": 7.87, + "learning_rate": 1.067948437841381e-06, + "loss": 0.8762, + "step": 36044 + }, + { + "epoch": 7.87, + "learning_rate": 1.0678391959798995e-06, + "loss": 1.3029, + "step": 36045 + }, + { + "epoch": 7.87, + "learning_rate": 1.0677299541184184e-06, + "loss": 0.905, + "step": 36046 + }, + { + "epoch": 7.87, + "learning_rate": 1.067620712256937e-06, + "loss": 1.3774, + "step": 36047 + }, + { + "epoch": 7.87, + "learning_rate": 1.0675114703954556e-06, + "loss": 0.7009, + "step": 36048 + }, + { + "epoch": 7.87, + "learning_rate": 1.0674022285339742e-06, + "loss": 0.9081, + "step": 36049 + }, + { + "epoch": 7.87, + "learning_rate": 1.067292986672493e-06, + "loss": 2.3241, + "step": 36050 + }, + { + "epoch": 7.87, + "learning_rate": 1.0671837448110116e-06, + "loss": 1.6211, + "step": 36051 + }, + { + "epoch": 7.87, + "learning_rate": 1.0670745029495305e-06, + "loss": 1.1299, + "step": 36052 + }, + { + "epoch": 7.87, + "learning_rate": 1.066965261088049e-06, + "loss": 1.9183, + "step": 36053 + }, + { + "epoch": 7.87, + "learning_rate": 1.0668560192265677e-06, + "loss": 0.2984, + "step": 36054 + }, + { + "epoch": 7.87, + "learning_rate": 1.0667467773650863e-06, + "loss": 0.6133, + "step": 36055 + }, + { + "epoch": 7.87, + "learning_rate": 1.0666375355036051e-06, + "loss": 0.5138, + "step": 36056 + }, + { + "epoch": 7.87, + "learning_rate": 1.0665282936421237e-06, + "loss": 0.8722, + "step": 36057 + }, + { + "epoch": 7.87, + "learning_rate": 1.0664190517806425e-06, + "loss": 2.1767, + "step": 36058 + }, + { + "epoch": 7.87, + "learning_rate": 1.0663098099191612e-06, + "loss": 1.0443, + "step": 36059 + }, + { + "epoch": 7.87, + "learning_rate": 1.0662005680576798e-06, + "loss": 1.2247, + "step": 36060 + }, + { + "epoch": 7.87, + "learning_rate": 1.0660913261961984e-06, + "loss": 1.5331, + "step": 36061 + }, + { + "epoch": 7.87, + "learning_rate": 1.0659820843347172e-06, + "loss": 1.5274, + "step": 36062 + }, + { + "epoch": 7.87, + "learning_rate": 1.0658728424732358e-06, + "loss": 1.9202, + "step": 36063 + }, + { + "epoch": 7.87, + "learning_rate": 1.0657636006117546e-06, + "loss": 2.8143, + "step": 36064 + }, + { + "epoch": 7.87, + "learning_rate": 1.0656543587502732e-06, + "loss": 1.538, + "step": 36065 + }, + { + "epoch": 7.87, + "learning_rate": 1.0655451168887919e-06, + "loss": 0.5003, + "step": 36066 + }, + { + "epoch": 7.87, + "learning_rate": 1.0654358750273105e-06, + "loss": 0.6778, + "step": 36067 + }, + { + "epoch": 7.87, + "learning_rate": 1.0653266331658293e-06, + "loss": 0.8072, + "step": 36068 + }, + { + "epoch": 7.87, + "learning_rate": 1.065217391304348e-06, + "loss": 0.3653, + "step": 36069 + }, + { + "epoch": 7.87, + "learning_rate": 1.0651081494428665e-06, + "loss": 2.4641, + "step": 36070 + }, + { + "epoch": 7.87, + "learning_rate": 1.0649989075813853e-06, + "loss": 2.2472, + "step": 36071 + }, + { + "epoch": 7.87, + "learning_rate": 1.064889665719904e-06, + "loss": 0.9061, + "step": 36072 + }, + { + "epoch": 7.87, + "learning_rate": 1.0647804238584226e-06, + "loss": 0.7196, + "step": 36073 + }, + { + "epoch": 7.87, + "learning_rate": 1.0646711819969414e-06, + "loss": 0.7433, + "step": 36074 + }, + { + "epoch": 7.87, + "learning_rate": 1.06456194013546e-06, + "loss": 1.0422, + "step": 36075 + }, + { + "epoch": 7.87, + "learning_rate": 1.0644526982739786e-06, + "loss": 2.5453, + "step": 36076 + }, + { + "epoch": 7.87, + "learning_rate": 1.0643434564124974e-06, + "loss": 0.6733, + "step": 36077 + }, + { + "epoch": 7.87, + "learning_rate": 1.064234214551016e-06, + "loss": 0.8301, + "step": 36078 + }, + { + "epoch": 7.87, + "learning_rate": 1.0641249726895346e-06, + "loss": 0.8394, + "step": 36079 + }, + { + "epoch": 7.87, + "learning_rate": 1.0640157308280535e-06, + "loss": 0.9576, + "step": 36080 + }, + { + "epoch": 7.87, + "learning_rate": 1.063906488966572e-06, + "loss": 1.4149, + "step": 36081 + }, + { + "epoch": 7.87, + "learning_rate": 1.0637972471050907e-06, + "loss": 1.4511, + "step": 36082 + }, + { + "epoch": 7.87, + "learning_rate": 1.0636880052436093e-06, + "loss": 0.5428, + "step": 36083 + }, + { + "epoch": 7.88, + "learning_rate": 1.0635787633821281e-06, + "loss": 1.9413, + "step": 36084 + }, + { + "epoch": 7.88, + "learning_rate": 1.0634695215206467e-06, + "loss": 0.6367, + "step": 36085 + }, + { + "epoch": 7.88, + "learning_rate": 1.0633602796591656e-06, + "loss": 0.8404, + "step": 36086 + }, + { + "epoch": 7.88, + "learning_rate": 1.0632510377976842e-06, + "loss": 0.9001, + "step": 36087 + }, + { + "epoch": 7.88, + "learning_rate": 1.0631417959362028e-06, + "loss": 1.003, + "step": 36088 + }, + { + "epoch": 7.88, + "learning_rate": 1.0630325540747214e-06, + "loss": 0.1263, + "step": 36089 + }, + { + "epoch": 7.88, + "learning_rate": 1.0629233122132402e-06, + "loss": 2.4812, + "step": 36090 + }, + { + "epoch": 7.88, + "learning_rate": 1.0628140703517588e-06, + "loss": 3.2249, + "step": 36091 + }, + { + "epoch": 7.88, + "learning_rate": 1.0627048284902776e-06, + "loss": 1.1153, + "step": 36092 + }, + { + "epoch": 7.88, + "learning_rate": 1.0625955866287963e-06, + "loss": 0.0911, + "step": 36093 + }, + { + "epoch": 7.88, + "learning_rate": 1.0624863447673149e-06, + "loss": 0.3976, + "step": 36094 + }, + { + "epoch": 7.88, + "learning_rate": 1.0623771029058335e-06, + "loss": 1.3401, + "step": 36095 + }, + { + "epoch": 7.88, + "learning_rate": 1.0622678610443523e-06, + "loss": 0.9834, + "step": 36096 + }, + { + "epoch": 7.88, + "learning_rate": 1.0621586191828711e-06, + "loss": 0.891, + "step": 36097 + }, + { + "epoch": 7.88, + "learning_rate": 1.0620493773213897e-06, + "loss": 2.3647, + "step": 36098 + }, + { + "epoch": 7.88, + "learning_rate": 1.0619401354599083e-06, + "loss": 0.2684, + "step": 36099 + }, + { + "epoch": 7.88, + "learning_rate": 1.061830893598427e-06, + "loss": 2.2201, + "step": 36100 + }, + { + "epoch": 7.88, + "learning_rate": 1.0617216517369456e-06, + "loss": 0.0709, + "step": 36101 + }, + { + "epoch": 7.88, + "learning_rate": 1.0616124098754644e-06, + "loss": 0.1632, + "step": 36102 + }, + { + "epoch": 7.88, + "learning_rate": 1.061503168013983e-06, + "loss": 1.2316, + "step": 36103 + }, + { + "epoch": 7.88, + "learning_rate": 1.0613939261525018e-06, + "loss": 0.9598, + "step": 36104 + }, + { + "epoch": 7.88, + "learning_rate": 1.0612846842910204e-06, + "loss": 0.3224, + "step": 36105 + }, + { + "epoch": 7.88, + "learning_rate": 1.061175442429539e-06, + "loss": 1.981, + "step": 36106 + }, + { + "epoch": 7.88, + "learning_rate": 1.0610662005680577e-06, + "loss": 1.2156, + "step": 36107 + }, + { + "epoch": 7.88, + "learning_rate": 1.0609569587065765e-06, + "loss": 1.0453, + "step": 36108 + }, + { + "epoch": 7.88, + "learning_rate": 1.060847716845095e-06, + "loss": 0.7671, + "step": 36109 + }, + { + "epoch": 7.88, + "learning_rate": 1.060738474983614e-06, + "loss": 2.5271, + "step": 36110 + }, + { + "epoch": 7.88, + "learning_rate": 1.0606292331221325e-06, + "loss": 0.3024, + "step": 36111 + }, + { + "epoch": 7.88, + "learning_rate": 1.0605199912606511e-06, + "loss": 0.4057, + "step": 36112 + }, + { + "epoch": 7.88, + "learning_rate": 1.0604107493991698e-06, + "loss": 0.3742, + "step": 36113 + }, + { + "epoch": 7.88, + "learning_rate": 1.0603015075376886e-06, + "loss": 1.7164, + "step": 36114 + }, + { + "epoch": 7.88, + "learning_rate": 1.0601922656762072e-06, + "loss": 1.0273, + "step": 36115 + }, + { + "epoch": 7.88, + "learning_rate": 1.0600830238147258e-06, + "loss": 0.9476, + "step": 36116 + }, + { + "epoch": 7.88, + "learning_rate": 1.0599737819532446e-06, + "loss": 1.5685, + "step": 36117 + }, + { + "epoch": 7.88, + "learning_rate": 1.0598645400917632e-06, + "loss": 2.238, + "step": 36118 + }, + { + "epoch": 7.88, + "learning_rate": 1.059755298230282e-06, + "loss": 0.1618, + "step": 36119 + }, + { + "epoch": 7.88, + "learning_rate": 1.0596460563688007e-06, + "loss": 0.0738, + "step": 36120 + }, + { + "epoch": 7.88, + "learning_rate": 1.0595368145073193e-06, + "loss": 0.0672, + "step": 36121 + }, + { + "epoch": 7.88, + "learning_rate": 1.0594275726458379e-06, + "loss": 0.0775, + "step": 36122 + }, + { + "epoch": 7.88, + "learning_rate": 1.0593183307843565e-06, + "loss": 0.7983, + "step": 36123 + }, + { + "epoch": 7.88, + "learning_rate": 1.0592090889228753e-06, + "loss": 1.966, + "step": 36124 + }, + { + "epoch": 7.88, + "learning_rate": 1.0590998470613941e-06, + "loss": 0.2958, + "step": 36125 + }, + { + "epoch": 7.88, + "learning_rate": 1.0589906051999128e-06, + "loss": 0.5065, + "step": 36126 + }, + { + "epoch": 7.88, + "learning_rate": 1.0588813633384314e-06, + "loss": 1.5594, + "step": 36127 + }, + { + "epoch": 7.88, + "learning_rate": 1.05877212147695e-06, + "loss": 0.095, + "step": 36128 + }, + { + "epoch": 7.88, + "learning_rate": 1.0586628796154686e-06, + "loss": 2.9356, + "step": 36129 + }, + { + "epoch": 7.89, + "learning_rate": 1.0585536377539874e-06, + "loss": 0.7523, + "step": 36130 + }, + { + "epoch": 7.89, + "learning_rate": 1.0584443958925062e-06, + "loss": 2.2803, + "step": 36131 + }, + { + "epoch": 7.89, + "learning_rate": 1.0583351540310248e-06, + "loss": 2.2706, + "step": 36132 + }, + { + "epoch": 7.89, + "learning_rate": 1.0582259121695435e-06, + "loss": 0.2496, + "step": 36133 + }, + { + "epoch": 7.89, + "learning_rate": 1.058116670308062e-06, + "loss": 1.6287, + "step": 36134 + }, + { + "epoch": 7.89, + "learning_rate": 1.0580074284465807e-06, + "loss": 2.5248, + "step": 36135 + }, + { + "epoch": 7.89, + "learning_rate": 1.0578981865850995e-06, + "loss": 0.4142, + "step": 36136 + }, + { + "epoch": 7.89, + "learning_rate": 1.0577889447236183e-06, + "loss": 1.1144, + "step": 36137 + }, + { + "epoch": 7.89, + "learning_rate": 1.057679702862137e-06, + "loss": 3.2842, + "step": 36138 + }, + { + "epoch": 7.89, + "learning_rate": 1.0575704610006555e-06, + "loss": 2.9582, + "step": 36139 + }, + { + "epoch": 7.89, + "learning_rate": 1.0574612191391742e-06, + "loss": 1.5549, + "step": 36140 + }, + { + "epoch": 7.89, + "learning_rate": 1.057351977277693e-06, + "loss": 0.3095, + "step": 36141 + }, + { + "epoch": 7.89, + "learning_rate": 1.0572427354162116e-06, + "loss": 0.8796, + "step": 36142 + }, + { + "epoch": 7.89, + "learning_rate": 1.0571334935547304e-06, + "loss": 1.2452, + "step": 36143 + }, + { + "epoch": 7.89, + "learning_rate": 1.057024251693249e-06, + "loss": 1.7266, + "step": 36144 + }, + { + "epoch": 7.89, + "learning_rate": 1.0569150098317676e-06, + "loss": 0.1574, + "step": 36145 + }, + { + "epoch": 7.89, + "learning_rate": 1.0568057679702862e-06, + "loss": 0.4063, + "step": 36146 + }, + { + "epoch": 7.89, + "learning_rate": 1.056696526108805e-06, + "loss": 0.8662, + "step": 36147 + }, + { + "epoch": 7.89, + "learning_rate": 1.0565872842473237e-06, + "loss": 0.5664, + "step": 36148 + }, + { + "epoch": 7.89, + "learning_rate": 1.0564780423858423e-06, + "loss": 1.859, + "step": 36149 + }, + { + "epoch": 7.89, + "learning_rate": 1.0563688005243611e-06, + "loss": 0.9576, + "step": 36150 + }, + { + "epoch": 7.89, + "learning_rate": 1.0562595586628797e-06, + "loss": 2.9101, + "step": 36151 + }, + { + "epoch": 7.89, + "learning_rate": 1.0561503168013983e-06, + "loss": 0.4016, + "step": 36152 + }, + { + "epoch": 7.89, + "learning_rate": 1.0560410749399172e-06, + "loss": 1.5993, + "step": 36153 + }, + { + "epoch": 7.89, + "learning_rate": 1.0559318330784358e-06, + "loss": 0.9651, + "step": 36154 + }, + { + "epoch": 7.89, + "learning_rate": 1.0558225912169544e-06, + "loss": 0.4324, + "step": 36155 + }, + { + "epoch": 7.89, + "learning_rate": 1.055713349355473e-06, + "loss": 2.0221, + "step": 36156 + }, + { + "epoch": 7.89, + "learning_rate": 1.0556041074939918e-06, + "loss": 0.2432, + "step": 36157 + }, + { + "epoch": 7.89, + "learning_rate": 1.0554948656325104e-06, + "loss": 0.9138, + "step": 36158 + }, + { + "epoch": 7.89, + "learning_rate": 1.0553856237710292e-06, + "loss": 0.8183, + "step": 36159 + }, + { + "epoch": 7.89, + "learning_rate": 1.0552763819095479e-06, + "loss": 0.5644, + "step": 36160 + }, + { + "epoch": 7.89, + "learning_rate": 1.0551671400480665e-06, + "loss": 0.7004, + "step": 36161 + }, + { + "epoch": 7.89, + "learning_rate": 1.055057898186585e-06, + "loss": 0.8034, + "step": 36162 + }, + { + "epoch": 7.89, + "learning_rate": 1.054948656325104e-06, + "loss": 0.6004, + "step": 36163 + }, + { + "epoch": 7.89, + "learning_rate": 1.0548394144636225e-06, + "loss": 1.3818, + "step": 36164 + }, + { + "epoch": 7.89, + "learning_rate": 1.0547301726021413e-06, + "loss": 1.6113, + "step": 36165 + }, + { + "epoch": 7.89, + "learning_rate": 1.05462093074066e-06, + "loss": 1.0775, + "step": 36166 + }, + { + "epoch": 7.89, + "learning_rate": 1.0545116888791786e-06, + "loss": 1.8371, + "step": 36167 + }, + { + "epoch": 7.89, + "learning_rate": 1.0544024470176972e-06, + "loss": 1.9137, + "step": 36168 + }, + { + "epoch": 7.89, + "learning_rate": 1.054293205156216e-06, + "loss": 2.2609, + "step": 36169 + }, + { + "epoch": 7.89, + "learning_rate": 1.0541839632947346e-06, + "loss": 2.2866, + "step": 36170 + }, + { + "epoch": 7.89, + "learning_rate": 1.0540747214332534e-06, + "loss": 1.0482, + "step": 36171 + }, + { + "epoch": 7.89, + "learning_rate": 1.053965479571772e-06, + "loss": 0.2216, + "step": 36172 + }, + { + "epoch": 7.89, + "learning_rate": 1.0538562377102906e-06, + "loss": 1.5369, + "step": 36173 + }, + { + "epoch": 7.89, + "learning_rate": 1.0537469958488093e-06, + "loss": 1.3536, + "step": 36174 + }, + { + "epoch": 7.9, + "learning_rate": 1.053637753987328e-06, + "loss": 0.8636, + "step": 36175 + }, + { + "epoch": 7.9, + "learning_rate": 1.0535285121258467e-06, + "loss": 0.9407, + "step": 36176 + }, + { + "epoch": 7.9, + "learning_rate": 1.0534192702643655e-06, + "loss": 1.7659, + "step": 36177 + }, + { + "epoch": 7.9, + "learning_rate": 1.0533100284028841e-06, + "loss": 1.7365, + "step": 36178 + }, + { + "epoch": 7.9, + "learning_rate": 1.0532007865414027e-06, + "loss": 1.1864, + "step": 36179 + }, + { + "epoch": 7.9, + "learning_rate": 1.0530915446799213e-06, + "loss": 2.5407, + "step": 36180 + }, + { + "epoch": 7.9, + "learning_rate": 1.0529823028184402e-06, + "loss": 0.7733, + "step": 36181 + }, + { + "epoch": 7.9, + "learning_rate": 1.0528730609569588e-06, + "loss": 1.6999, + "step": 36182 + }, + { + "epoch": 7.9, + "learning_rate": 1.0527638190954776e-06, + "loss": 0.4956, + "step": 36183 + }, + { + "epoch": 7.9, + "learning_rate": 1.0526545772339962e-06, + "loss": 1.2879, + "step": 36184 + }, + { + "epoch": 7.9, + "learning_rate": 1.0525453353725148e-06, + "loss": 0.8464, + "step": 36185 + }, + { + "epoch": 7.9, + "learning_rate": 1.0524360935110334e-06, + "loss": 2.3293, + "step": 36186 + }, + { + "epoch": 7.9, + "learning_rate": 1.0523268516495523e-06, + "loss": 0.659, + "step": 36187 + }, + { + "epoch": 7.9, + "learning_rate": 1.0522176097880709e-06, + "loss": 0.1526, + "step": 36188 + }, + { + "epoch": 7.9, + "learning_rate": 1.0521083679265895e-06, + "loss": 2.0945, + "step": 36189 + }, + { + "epoch": 7.9, + "learning_rate": 1.0519991260651083e-06, + "loss": 0.5089, + "step": 36190 + }, + { + "epoch": 7.9, + "learning_rate": 1.051889884203627e-06, + "loss": 0.6915, + "step": 36191 + }, + { + "epoch": 7.9, + "learning_rate": 1.0517806423421455e-06, + "loss": 2.1861, + "step": 36192 + }, + { + "epoch": 7.9, + "learning_rate": 1.0516714004806643e-06, + "loss": 1.2475, + "step": 36193 + }, + { + "epoch": 7.9, + "learning_rate": 1.051562158619183e-06, + "loss": 0.9078, + "step": 36194 + }, + { + "epoch": 7.9, + "learning_rate": 1.0514529167577016e-06, + "loss": 0.7384, + "step": 36195 + }, + { + "epoch": 7.9, + "learning_rate": 1.0513436748962204e-06, + "loss": 1.9675, + "step": 36196 + }, + { + "epoch": 7.9, + "learning_rate": 1.051234433034739e-06, + "loss": 0.3049, + "step": 36197 + }, + { + "epoch": 7.9, + "learning_rate": 1.0511251911732576e-06, + "loss": 0.4947, + "step": 36198 + }, + { + "epoch": 7.9, + "learning_rate": 1.0510159493117764e-06, + "loss": 1.2443, + "step": 36199 + }, + { + "epoch": 7.9, + "learning_rate": 1.050906707450295e-06, + "loss": 1.4941, + "step": 36200 + }, + { + "epoch": 7.9, + "learning_rate": 1.0507974655888137e-06, + "loss": 0.5671, + "step": 36201 + }, + { + "epoch": 7.9, + "learning_rate": 1.0506882237273323e-06, + "loss": 0.4847, + "step": 36202 + }, + { + "epoch": 7.9, + "learning_rate": 1.050578981865851e-06, + "loss": 1.714, + "step": 36203 + }, + { + "epoch": 7.9, + "learning_rate": 1.0504697400043697e-06, + "loss": 0.3205, + "step": 36204 + }, + { + "epoch": 7.9, + "learning_rate": 1.0503604981428885e-06, + "loss": 0.0337, + "step": 36205 + }, + { + "epoch": 7.9, + "learning_rate": 1.0502512562814071e-06, + "loss": 0.5396, + "step": 36206 + }, + { + "epoch": 7.9, + "learning_rate": 1.0501420144199257e-06, + "loss": 0.9315, + "step": 36207 + }, + { + "epoch": 7.9, + "learning_rate": 1.0500327725584444e-06, + "loss": 0.675, + "step": 36208 + }, + { + "epoch": 7.9, + "learning_rate": 1.0499235306969632e-06, + "loss": 2.1604, + "step": 36209 + }, + { + "epoch": 7.9, + "learning_rate": 1.0498142888354818e-06, + "loss": 1.1457, + "step": 36210 + }, + { + "epoch": 7.9, + "learning_rate": 1.0497050469740006e-06, + "loss": 2.3813, + "step": 36211 + }, + { + "epoch": 7.9, + "learning_rate": 1.0495958051125192e-06, + "loss": 0.2837, + "step": 36212 + }, + { + "epoch": 7.9, + "learning_rate": 1.0494865632510378e-06, + "loss": 1.3189, + "step": 36213 + }, + { + "epoch": 7.9, + "learning_rate": 1.0493773213895564e-06, + "loss": 0.1794, + "step": 36214 + }, + { + "epoch": 7.9, + "learning_rate": 1.0492680795280753e-06, + "loss": 0.1028, + "step": 36215 + }, + { + "epoch": 7.9, + "learning_rate": 1.049158837666594e-06, + "loss": 0.5947, + "step": 36216 + }, + { + "epoch": 7.9, + "learning_rate": 1.0490495958051127e-06, + "loss": 1.3993, + "step": 36217 + }, + { + "epoch": 7.9, + "learning_rate": 1.0489403539436313e-06, + "loss": 1.0948, + "step": 36218 + }, + { + "epoch": 7.9, + "learning_rate": 1.04883111208215e-06, + "loss": 0.8879, + "step": 36219 + }, + { + "epoch": 7.9, + "learning_rate": 1.0487218702206685e-06, + "loss": 0.8693, + "step": 36220 + }, + { + "epoch": 7.91, + "learning_rate": 1.0486126283591874e-06, + "loss": 1.7082, + "step": 36221 + }, + { + "epoch": 7.91, + "learning_rate": 1.048503386497706e-06, + "loss": 0.7404, + "step": 36222 + }, + { + "epoch": 7.91, + "learning_rate": 1.0483941446362248e-06, + "loss": 0.3994, + "step": 36223 + }, + { + "epoch": 7.91, + "learning_rate": 1.0482849027747434e-06, + "loss": 3.1763, + "step": 36224 + }, + { + "epoch": 7.91, + "learning_rate": 1.048175660913262e-06, + "loss": 1.9126, + "step": 36225 + }, + { + "epoch": 7.91, + "learning_rate": 1.0480664190517806e-06, + "loss": 0.7429, + "step": 36226 + }, + { + "epoch": 7.91, + "learning_rate": 1.0479571771902994e-06, + "loss": 2.2904, + "step": 36227 + }, + { + "epoch": 7.91, + "learning_rate": 1.047847935328818e-06, + "loss": 0.1622, + "step": 36228 + }, + { + "epoch": 7.91, + "learning_rate": 1.0477386934673369e-06, + "loss": 4.8377, + "step": 36229 + }, + { + "epoch": 7.91, + "learning_rate": 1.0476294516058555e-06, + "loss": 0.2188, + "step": 36230 + }, + { + "epoch": 7.91, + "learning_rate": 1.047520209744374e-06, + "loss": 0.5491, + "step": 36231 + }, + { + "epoch": 7.91, + "learning_rate": 1.0474109678828927e-06, + "loss": 1.3906, + "step": 36232 + }, + { + "epoch": 7.91, + "learning_rate": 1.0473017260214115e-06, + "loss": 0.7164, + "step": 36233 + }, + { + "epoch": 7.91, + "learning_rate": 1.0471924841599302e-06, + "loss": 0.3531, + "step": 36234 + }, + { + "epoch": 7.91, + "learning_rate": 1.0470832422984488e-06, + "loss": 0.4049, + "step": 36235 + }, + { + "epoch": 7.91, + "learning_rate": 1.0469740004369676e-06, + "loss": 0.6053, + "step": 36236 + }, + { + "epoch": 7.91, + "learning_rate": 1.0468647585754862e-06, + "loss": 1.3166, + "step": 36237 + }, + { + "epoch": 7.91, + "learning_rate": 1.0467555167140048e-06, + "loss": 1.1648, + "step": 36238 + }, + { + "epoch": 7.91, + "learning_rate": 1.0466462748525236e-06, + "loss": 1.6398, + "step": 36239 + }, + { + "epoch": 7.91, + "learning_rate": 1.0465370329910422e-06, + "loss": 2.1714, + "step": 36240 + }, + { + "epoch": 7.91, + "learning_rate": 1.0464277911295609e-06, + "loss": 1.2126, + "step": 36241 + }, + { + "epoch": 7.91, + "learning_rate": 1.0463185492680795e-06, + "loss": 0.8314, + "step": 36242 + }, + { + "epoch": 7.91, + "learning_rate": 1.0462093074065983e-06, + "loss": 1.2559, + "step": 36243 + }, + { + "epoch": 7.91, + "learning_rate": 1.0461000655451171e-06, + "loss": 1.9587, + "step": 36244 + }, + { + "epoch": 7.91, + "learning_rate": 1.0459908236836357e-06, + "loss": 1.398, + "step": 36245 + }, + { + "epoch": 7.91, + "learning_rate": 1.0458815818221543e-06, + "loss": 1.0905, + "step": 36246 + }, + { + "epoch": 7.91, + "learning_rate": 1.045772339960673e-06, + "loss": 1.2858, + "step": 36247 + }, + { + "epoch": 7.91, + "learning_rate": 1.0456630980991916e-06, + "loss": 1.2158, + "step": 36248 + }, + { + "epoch": 7.91, + "learning_rate": 1.0455538562377104e-06, + "loss": 3.0122, + "step": 36249 + }, + { + "epoch": 7.91, + "learning_rate": 1.0454446143762292e-06, + "loss": 1.2589, + "step": 36250 + }, + { + "epoch": 7.91, + "learning_rate": 1.0453353725147478e-06, + "loss": 1.2435, + "step": 36251 + }, + { + "epoch": 7.91, + "learning_rate": 1.0452261306532664e-06, + "loss": 0.0773, + "step": 36252 + }, + { + "epoch": 7.91, + "learning_rate": 1.045116888791785e-06, + "loss": 3.2609, + "step": 36253 + }, + { + "epoch": 7.91, + "learning_rate": 1.0450076469303036e-06, + "loss": 1.1994, + "step": 36254 + }, + { + "epoch": 7.91, + "learning_rate": 1.0448984050688225e-06, + "loss": 1.8448, + "step": 36255 + }, + { + "epoch": 7.91, + "learning_rate": 1.0447891632073413e-06, + "loss": 0.6327, + "step": 36256 + }, + { + "epoch": 7.91, + "learning_rate": 1.04467992134586e-06, + "loss": 0.7311, + "step": 36257 + }, + { + "epoch": 7.91, + "learning_rate": 1.0445706794843785e-06, + "loss": 0.4857, + "step": 36258 + }, + { + "epoch": 7.91, + "learning_rate": 1.0444614376228971e-06, + "loss": 0.3456, + "step": 36259 + }, + { + "epoch": 7.91, + "learning_rate": 1.0443521957614157e-06, + "loss": 1.2093, + "step": 36260 + }, + { + "epoch": 7.91, + "learning_rate": 1.0442429538999346e-06, + "loss": 1.4897, + "step": 36261 + }, + { + "epoch": 7.91, + "learning_rate": 1.0441337120384534e-06, + "loss": 0.2559, + "step": 36262 + }, + { + "epoch": 7.91, + "learning_rate": 1.044024470176972e-06, + "loss": 0.0648, + "step": 36263 + }, + { + "epoch": 7.91, + "learning_rate": 1.0439152283154906e-06, + "loss": 1.9401, + "step": 36264 + }, + { + "epoch": 7.91, + "learning_rate": 1.0438059864540092e-06, + "loss": 0.4518, + "step": 36265 + }, + { + "epoch": 7.91, + "learning_rate": 1.043696744592528e-06, + "loss": 0.7129, + "step": 36266 + }, + { + "epoch": 7.92, + "learning_rate": 1.0435875027310466e-06, + "loss": 1.0409, + "step": 36267 + }, + { + "epoch": 7.92, + "learning_rate": 1.0434782608695653e-06, + "loss": 1.4152, + "step": 36268 + }, + { + "epoch": 7.92, + "learning_rate": 1.043369019008084e-06, + "loss": 1.1034, + "step": 36269 + }, + { + "epoch": 7.92, + "learning_rate": 1.0432597771466027e-06, + "loss": 0.827, + "step": 36270 + }, + { + "epoch": 7.92, + "learning_rate": 1.0431505352851213e-06, + "loss": 1.4079, + "step": 36271 + }, + { + "epoch": 7.92, + "learning_rate": 1.0430412934236401e-06, + "loss": 1.9403, + "step": 36272 + }, + { + "epoch": 7.92, + "learning_rate": 1.0429320515621587e-06, + "loss": 0.2408, + "step": 36273 + }, + { + "epoch": 7.92, + "learning_rate": 1.0428228097006773e-06, + "loss": 0.3105, + "step": 36274 + }, + { + "epoch": 7.92, + "learning_rate": 1.042713567839196e-06, + "loss": 0.9025, + "step": 36275 + }, + { + "epoch": 7.92, + "learning_rate": 1.0426043259777148e-06, + "loss": 1.8613, + "step": 36276 + }, + { + "epoch": 7.92, + "learning_rate": 1.0424950841162334e-06, + "loss": 0.9157, + "step": 36277 + }, + { + "epoch": 7.92, + "learning_rate": 1.0423858422547522e-06, + "loss": 1.4899, + "step": 36278 + }, + { + "epoch": 7.92, + "learning_rate": 1.0422766003932708e-06, + "loss": 0.6296, + "step": 36279 + }, + { + "epoch": 7.92, + "learning_rate": 1.0421673585317894e-06, + "loss": 1.7748, + "step": 36280 + }, + { + "epoch": 7.92, + "learning_rate": 1.042058116670308e-06, + "loss": 0.9471, + "step": 36281 + }, + { + "epoch": 7.92, + "learning_rate": 1.0419488748088269e-06, + "loss": 0.1706, + "step": 36282 + }, + { + "epoch": 7.92, + "learning_rate": 1.0418396329473455e-06, + "loss": 1.5076, + "step": 36283 + }, + { + "epoch": 7.92, + "learning_rate": 1.0417303910858643e-06, + "loss": 1.7148, + "step": 36284 + }, + { + "epoch": 7.92, + "learning_rate": 1.041621149224383e-06, + "loss": 1.3539, + "step": 36285 + }, + { + "epoch": 7.92, + "learning_rate": 1.0415119073629015e-06, + "loss": 1.7252, + "step": 36286 + }, + { + "epoch": 7.92, + "learning_rate": 1.0414026655014201e-06, + "loss": 1.0805, + "step": 36287 + }, + { + "epoch": 7.92, + "learning_rate": 1.041293423639939e-06, + "loss": 1.303, + "step": 36288 + }, + { + "epoch": 7.92, + "learning_rate": 1.0411841817784576e-06, + "loss": 1.0919, + "step": 36289 + }, + { + "epoch": 7.92, + "learning_rate": 1.0410749399169764e-06, + "loss": 0.4199, + "step": 36290 + }, + { + "epoch": 7.92, + "learning_rate": 1.040965698055495e-06, + "loss": 0.378, + "step": 36291 + }, + { + "epoch": 7.92, + "learning_rate": 1.0408564561940136e-06, + "loss": 1.22, + "step": 36292 + }, + { + "epoch": 7.92, + "learning_rate": 1.0407472143325322e-06, + "loss": 0.6328, + "step": 36293 + }, + { + "epoch": 7.92, + "learning_rate": 1.040637972471051e-06, + "loss": 3.5737, + "step": 36294 + }, + { + "epoch": 7.92, + "learning_rate": 1.0405287306095697e-06, + "loss": 1.0158, + "step": 36295 + }, + { + "epoch": 7.92, + "learning_rate": 1.0404194887480885e-06, + "loss": 0.4884, + "step": 36296 + }, + { + "epoch": 7.92, + "learning_rate": 1.040310246886607e-06, + "loss": 1.2626, + "step": 36297 + }, + { + "epoch": 7.92, + "learning_rate": 1.0402010050251257e-06, + "loss": 0.3945, + "step": 36298 + }, + { + "epoch": 7.92, + "learning_rate": 1.0400917631636443e-06, + "loss": 2.3737, + "step": 36299 + }, + { + "epoch": 7.92, + "learning_rate": 1.0399825213021631e-06, + "loss": 1.2816, + "step": 36300 + }, + { + "epoch": 7.92, + "learning_rate": 1.0398732794406817e-06, + "loss": 0.8335, + "step": 36301 + }, + { + "epoch": 7.92, + "learning_rate": 1.0397640375792006e-06, + "loss": 3.3155, + "step": 36302 + }, + { + "epoch": 7.92, + "learning_rate": 1.0396547957177192e-06, + "loss": 0.5422, + "step": 36303 + }, + { + "epoch": 7.92, + "learning_rate": 1.0395455538562378e-06, + "loss": 1.9112, + "step": 36304 + }, + { + "epoch": 7.92, + "learning_rate": 1.0394363119947564e-06, + "loss": 0.5929, + "step": 36305 + }, + { + "epoch": 7.92, + "learning_rate": 1.0393270701332752e-06, + "loss": 0.4616, + "step": 36306 + }, + { + "epoch": 7.92, + "learning_rate": 1.0392178282717938e-06, + "loss": 1.2896, + "step": 36307 + }, + { + "epoch": 7.92, + "learning_rate": 1.0391085864103124e-06, + "loss": 0.7966, + "step": 36308 + }, + { + "epoch": 7.92, + "learning_rate": 1.0389993445488313e-06, + "loss": 1.1991, + "step": 36309 + }, + { + "epoch": 7.92, + "learning_rate": 1.0388901026873499e-06, + "loss": 1.7814, + "step": 36310 + }, + { + "epoch": 7.92, + "learning_rate": 1.0387808608258685e-06, + "loss": 1.4099, + "step": 36311 + }, + { + "epoch": 7.92, + "learning_rate": 1.0386716189643873e-06, + "loss": 2.0994, + "step": 36312 + }, + { + "epoch": 7.93, + "learning_rate": 1.038562377102906e-06, + "loss": 0.3039, + "step": 36313 + }, + { + "epoch": 7.93, + "learning_rate": 1.0384531352414245e-06, + "loss": 1.0711, + "step": 36314 + }, + { + "epoch": 7.93, + "learning_rate": 1.0383438933799434e-06, + "loss": 1.3666, + "step": 36315 + }, + { + "epoch": 7.93, + "learning_rate": 1.038234651518462e-06, + "loss": 1.4946, + "step": 36316 + }, + { + "epoch": 7.93, + "learning_rate": 1.0381254096569806e-06, + "loss": 0.1894, + "step": 36317 + }, + { + "epoch": 7.93, + "learning_rate": 1.0380161677954994e-06, + "loss": 1.8729, + "step": 36318 + }, + { + "epoch": 7.93, + "learning_rate": 1.037906925934018e-06, + "loss": 0.0518, + "step": 36319 + }, + { + "epoch": 7.93, + "learning_rate": 1.0377976840725366e-06, + "loss": 1.3637, + "step": 36320 + }, + { + "epoch": 7.93, + "learning_rate": 1.0376884422110552e-06, + "loss": 3.1155, + "step": 36321 + }, + { + "epoch": 7.93, + "learning_rate": 1.037579200349574e-06, + "loss": 1.0679, + "step": 36322 + }, + { + "epoch": 7.93, + "learning_rate": 1.0374699584880927e-06, + "loss": 3.0488, + "step": 36323 + }, + { + "epoch": 7.93, + "learning_rate": 1.0373607166266115e-06, + "loss": 2.7148, + "step": 36324 + }, + { + "epoch": 7.93, + "learning_rate": 1.03725147476513e-06, + "loss": 0.3437, + "step": 36325 + }, + { + "epoch": 7.93, + "learning_rate": 1.0371422329036487e-06, + "loss": 1.3922, + "step": 36326 + }, + { + "epoch": 7.93, + "learning_rate": 1.0370329910421673e-06, + "loss": 1.7026, + "step": 36327 + }, + { + "epoch": 7.93, + "learning_rate": 1.0369237491806861e-06, + "loss": 0.6118, + "step": 36328 + }, + { + "epoch": 7.93, + "learning_rate": 1.0368145073192048e-06, + "loss": 0.3897, + "step": 36329 + }, + { + "epoch": 7.93, + "learning_rate": 1.0367052654577236e-06, + "loss": 1.0836, + "step": 36330 + }, + { + "epoch": 7.93, + "learning_rate": 1.0365960235962422e-06, + "loss": 1.4345, + "step": 36331 + }, + { + "epoch": 7.93, + "learning_rate": 1.0364867817347608e-06, + "loss": 1.3426, + "step": 36332 + }, + { + "epoch": 7.93, + "learning_rate": 1.0363775398732794e-06, + "loss": 2.207, + "step": 36333 + }, + { + "epoch": 7.93, + "learning_rate": 1.0362682980117982e-06, + "loss": 0.8185, + "step": 36334 + }, + { + "epoch": 7.93, + "learning_rate": 1.0361590561503168e-06, + "loss": 1.5082, + "step": 36335 + }, + { + "epoch": 7.93, + "learning_rate": 1.0360498142888357e-06, + "loss": 0.9324, + "step": 36336 + }, + { + "epoch": 7.93, + "learning_rate": 1.0359405724273543e-06, + "loss": 1.7936, + "step": 36337 + }, + { + "epoch": 7.93, + "learning_rate": 1.0358313305658729e-06, + "loss": 2.1415, + "step": 36338 + }, + { + "epoch": 7.93, + "learning_rate": 1.0357220887043915e-06, + "loss": 0.8174, + "step": 36339 + }, + { + "epoch": 7.93, + "learning_rate": 1.0356128468429103e-06, + "loss": 1.6714, + "step": 36340 + }, + { + "epoch": 7.93, + "learning_rate": 1.035503604981429e-06, + "loss": 0.423, + "step": 36341 + }, + { + "epoch": 7.93, + "learning_rate": 1.0353943631199478e-06, + "loss": 0.4675, + "step": 36342 + }, + { + "epoch": 7.93, + "learning_rate": 1.0352851212584664e-06, + "loss": 1.028, + "step": 36343 + }, + { + "epoch": 7.93, + "learning_rate": 1.035175879396985e-06, + "loss": 0.9841, + "step": 36344 + }, + { + "epoch": 7.93, + "learning_rate": 1.0350666375355036e-06, + "loss": 0.8032, + "step": 36345 + }, + { + "epoch": 7.93, + "learning_rate": 1.0349573956740224e-06, + "loss": 1.1422, + "step": 36346 + }, + { + "epoch": 7.93, + "learning_rate": 1.034848153812541e-06, + "loss": 0.5864, + "step": 36347 + }, + { + "epoch": 7.93, + "learning_rate": 1.0347389119510598e-06, + "loss": 0.076, + "step": 36348 + }, + { + "epoch": 7.93, + "learning_rate": 1.0346296700895785e-06, + "loss": 2.3652, + "step": 36349 + }, + { + "epoch": 7.93, + "learning_rate": 1.034520428228097e-06, + "loss": 1.4837, + "step": 36350 + }, + { + "epoch": 7.93, + "learning_rate": 1.0344111863666157e-06, + "loss": 1.1779, + "step": 36351 + }, + { + "epoch": 7.93, + "learning_rate": 1.0343019445051345e-06, + "loss": 0.267, + "step": 36352 + }, + { + "epoch": 7.93, + "learning_rate": 1.0341927026436531e-06, + "loss": 1.2102, + "step": 36353 + }, + { + "epoch": 7.93, + "learning_rate": 1.0340834607821717e-06, + "loss": 0.1115, + "step": 36354 + }, + { + "epoch": 7.93, + "learning_rate": 1.0339742189206905e-06, + "loss": 1.8523, + "step": 36355 + }, + { + "epoch": 7.93, + "learning_rate": 1.0338649770592092e-06, + "loss": 1.1262, + "step": 36356 + }, + { + "epoch": 7.93, + "learning_rate": 1.0337557351977278e-06, + "loss": 2.0871, + "step": 36357 + }, + { + "epoch": 7.93, + "learning_rate": 1.0336464933362466e-06, + "loss": 1.1466, + "step": 36358 + }, + { + "epoch": 7.94, + "learning_rate": 1.0335372514747652e-06, + "loss": 1.4951, + "step": 36359 + }, + { + "epoch": 7.94, + "learning_rate": 1.0334280096132838e-06, + "loss": 2.6229, + "step": 36360 + }, + { + "epoch": 7.94, + "learning_rate": 1.0333187677518024e-06, + "loss": 1.3514, + "step": 36361 + }, + { + "epoch": 7.94, + "learning_rate": 1.0332095258903213e-06, + "loss": 0.8245, + "step": 36362 + }, + { + "epoch": 7.94, + "learning_rate": 1.0331002840288399e-06, + "loss": 1.694, + "step": 36363 + }, + { + "epoch": 7.94, + "learning_rate": 1.0329910421673587e-06, + "loss": 0.8808, + "step": 36364 + }, + { + "epoch": 7.94, + "learning_rate": 1.0328818003058773e-06, + "loss": 0.8575, + "step": 36365 + }, + { + "epoch": 7.94, + "learning_rate": 1.032772558444396e-06, + "loss": 1.4973, + "step": 36366 + }, + { + "epoch": 7.94, + "learning_rate": 1.0326633165829145e-06, + "loss": 2.1549, + "step": 36367 + }, + { + "epoch": 7.94, + "learning_rate": 1.0325540747214333e-06, + "loss": 0.1875, + "step": 36368 + }, + { + "epoch": 7.94, + "learning_rate": 1.0324448328599522e-06, + "loss": 2.3806, + "step": 36369 + }, + { + "epoch": 7.94, + "learning_rate": 1.0323355909984708e-06, + "loss": 0.075, + "step": 36370 + }, + { + "epoch": 7.94, + "learning_rate": 1.0322263491369894e-06, + "loss": 0.244, + "step": 36371 + }, + { + "epoch": 7.94, + "learning_rate": 1.032117107275508e-06, + "loss": 2.5165, + "step": 36372 + }, + { + "epoch": 7.94, + "learning_rate": 1.0320078654140266e-06, + "loss": 1.3505, + "step": 36373 + }, + { + "epoch": 7.94, + "learning_rate": 1.0318986235525454e-06, + "loss": 1.4803, + "step": 36374 + }, + { + "epoch": 7.94, + "learning_rate": 1.0317893816910643e-06, + "loss": 0.9873, + "step": 36375 + }, + { + "epoch": 7.94, + "learning_rate": 1.0316801398295829e-06, + "loss": 0.323, + "step": 36376 + }, + { + "epoch": 7.94, + "learning_rate": 1.0315708979681015e-06, + "loss": 1.1812, + "step": 36377 + }, + { + "epoch": 7.94, + "learning_rate": 1.03146165610662e-06, + "loss": 0.8708, + "step": 36378 + }, + { + "epoch": 7.94, + "learning_rate": 1.0313524142451387e-06, + "loss": 1.1021, + "step": 36379 + }, + { + "epoch": 7.94, + "learning_rate": 1.0312431723836575e-06, + "loss": 0.5961, + "step": 36380 + }, + { + "epoch": 7.94, + "learning_rate": 1.0311339305221761e-06, + "loss": 0.4484, + "step": 36381 + }, + { + "epoch": 7.94, + "learning_rate": 1.031024688660695e-06, + "loss": 0.9821, + "step": 36382 + }, + { + "epoch": 7.94, + "learning_rate": 1.0309154467992136e-06, + "loss": 0.7068, + "step": 36383 + }, + { + "epoch": 7.94, + "learning_rate": 1.0308062049377322e-06, + "loss": 0.8998, + "step": 36384 + }, + { + "epoch": 7.94, + "learning_rate": 1.0306969630762508e-06, + "loss": 1.1788, + "step": 36385 + }, + { + "epoch": 7.94, + "learning_rate": 1.0305877212147696e-06, + "loss": 1.4873, + "step": 36386 + }, + { + "epoch": 7.94, + "learning_rate": 1.0304784793532882e-06, + "loss": 1.5598, + "step": 36387 + }, + { + "epoch": 7.94, + "learning_rate": 1.030369237491807e-06, + "loss": 0.7912, + "step": 36388 + }, + { + "epoch": 7.94, + "learning_rate": 1.0302599956303257e-06, + "loss": 0.7937, + "step": 36389 + }, + { + "epoch": 7.94, + "learning_rate": 1.0301507537688443e-06, + "loss": 1.0441, + "step": 36390 + }, + { + "epoch": 7.94, + "learning_rate": 1.030041511907363e-06, + "loss": 1.5336, + "step": 36391 + }, + { + "epoch": 7.94, + "learning_rate": 1.0299322700458817e-06, + "loss": 1.7174, + "step": 36392 + }, + { + "epoch": 7.94, + "learning_rate": 1.0298230281844003e-06, + "loss": 1.0543, + "step": 36393 + }, + { + "epoch": 7.94, + "learning_rate": 1.029713786322919e-06, + "loss": 2.9033, + "step": 36394 + }, + { + "epoch": 7.94, + "learning_rate": 1.0296045444614377e-06, + "loss": 2.1495, + "step": 36395 + }, + { + "epoch": 7.94, + "learning_rate": 1.0294953025999564e-06, + "loss": 2.0621, + "step": 36396 + }, + { + "epoch": 7.94, + "learning_rate": 1.0293860607384752e-06, + "loss": 0.8864, + "step": 36397 + }, + { + "epoch": 7.94, + "learning_rate": 1.0292768188769938e-06, + "loss": 1.7101, + "step": 36398 + }, + { + "epoch": 7.94, + "learning_rate": 1.0291675770155124e-06, + "loss": 1.0978, + "step": 36399 + }, + { + "epoch": 7.94, + "learning_rate": 1.029058335154031e-06, + "loss": 0.7436, + "step": 36400 + }, + { + "epoch": 7.94, + "learning_rate": 1.0289490932925498e-06, + "loss": 0.3483, + "step": 36401 + }, + { + "epoch": 7.94, + "learning_rate": 1.0288398514310684e-06, + "loss": 0.6792, + "step": 36402 + }, + { + "epoch": 7.94, + "learning_rate": 1.0287306095695873e-06, + "loss": 1.3324, + "step": 36403 + }, + { + "epoch": 7.95, + "learning_rate": 1.0286213677081059e-06, + "loss": 0.7983, + "step": 36404 + }, + { + "epoch": 7.95, + "learning_rate": 1.0285121258466245e-06, + "loss": 2.2532, + "step": 36405 + }, + { + "epoch": 7.95, + "learning_rate": 1.028402883985143e-06, + "loss": 0.5982, + "step": 36406 + }, + { + "epoch": 7.95, + "learning_rate": 1.0282936421236617e-06, + "loss": 0.4585, + "step": 36407 + }, + { + "epoch": 7.95, + "learning_rate": 1.0281844002621805e-06, + "loss": 0.6857, + "step": 36408 + }, + { + "epoch": 7.95, + "learning_rate": 1.0280751584006994e-06, + "loss": 0.1462, + "step": 36409 + }, + { + "epoch": 7.95, + "learning_rate": 1.027965916539218e-06, + "loss": 0.6559, + "step": 36410 + }, + { + "epoch": 7.95, + "learning_rate": 1.0278566746777366e-06, + "loss": 0.3109, + "step": 36411 + }, + { + "epoch": 7.95, + "learning_rate": 1.0277474328162552e-06, + "loss": 0.7216, + "step": 36412 + }, + { + "epoch": 7.95, + "learning_rate": 1.027638190954774e-06, + "loss": 0.7775, + "step": 36413 + }, + { + "epoch": 7.95, + "learning_rate": 1.0275289490932926e-06, + "loss": 1.0741, + "step": 36414 + }, + { + "epoch": 7.95, + "learning_rate": 1.0274197072318114e-06, + "loss": 0.7355, + "step": 36415 + }, + { + "epoch": 7.95, + "learning_rate": 1.02731046537033e-06, + "loss": 1.1309, + "step": 36416 + }, + { + "epoch": 7.95, + "learning_rate": 1.0272012235088487e-06, + "loss": 1.5255, + "step": 36417 + }, + { + "epoch": 7.95, + "learning_rate": 1.0270919816473673e-06, + "loss": 1.165, + "step": 36418 + }, + { + "epoch": 7.95, + "learning_rate": 1.026982739785886e-06, + "loss": 1.9989, + "step": 36419 + }, + { + "epoch": 7.95, + "learning_rate": 1.0268734979244047e-06, + "loss": 1.672, + "step": 36420 + }, + { + "epoch": 7.95, + "learning_rate": 1.0267642560629235e-06, + "loss": 0.9577, + "step": 36421 + }, + { + "epoch": 7.95, + "learning_rate": 1.0266550142014421e-06, + "loss": 0.2764, + "step": 36422 + }, + { + "epoch": 7.95, + "learning_rate": 1.0265457723399608e-06, + "loss": 2.4185, + "step": 36423 + }, + { + "epoch": 7.95, + "learning_rate": 1.0264365304784794e-06, + "loss": 1.3691, + "step": 36424 + }, + { + "epoch": 7.95, + "learning_rate": 1.0263272886169982e-06, + "loss": 0.6914, + "step": 36425 + }, + { + "epoch": 7.95, + "learning_rate": 1.0262180467555168e-06, + "loss": 0.9025, + "step": 36426 + }, + { + "epoch": 7.95, + "learning_rate": 1.0261088048940354e-06, + "loss": 1.2718, + "step": 36427 + }, + { + "epoch": 7.95, + "learning_rate": 1.0259995630325542e-06, + "loss": 2.7843, + "step": 36428 + }, + { + "epoch": 7.95, + "learning_rate": 1.0258903211710728e-06, + "loss": 1.5289, + "step": 36429 + }, + { + "epoch": 7.95, + "learning_rate": 1.0257810793095915e-06, + "loss": 1.3617, + "step": 36430 + }, + { + "epoch": 7.95, + "learning_rate": 1.0256718374481103e-06, + "loss": 0.7974, + "step": 36431 + }, + { + "epoch": 7.95, + "learning_rate": 1.0255625955866289e-06, + "loss": 0.0984, + "step": 36432 + }, + { + "epoch": 7.95, + "learning_rate": 1.0254533537251475e-06, + "loss": 1.7171, + "step": 36433 + }, + { + "epoch": 7.95, + "learning_rate": 1.0253441118636663e-06, + "loss": 0.8315, + "step": 36434 + }, + { + "epoch": 7.95, + "learning_rate": 1.025234870002185e-06, + "loss": 2.7034, + "step": 36435 + }, + { + "epoch": 7.95, + "learning_rate": 1.0251256281407035e-06, + "loss": 0.4846, + "step": 36436 + }, + { + "epoch": 7.95, + "learning_rate": 1.0250163862792224e-06, + "loss": 0.6028, + "step": 36437 + }, + { + "epoch": 7.95, + "learning_rate": 1.024907144417741e-06, + "loss": 0.2252, + "step": 36438 + }, + { + "epoch": 7.95, + "learning_rate": 1.0247979025562596e-06, + "loss": 1.9884, + "step": 36439 + }, + { + "epoch": 7.95, + "learning_rate": 1.0246886606947782e-06, + "loss": 0.6227, + "step": 36440 + }, + { + "epoch": 7.95, + "learning_rate": 1.024579418833297e-06, + "loss": 0.5254, + "step": 36441 + }, + { + "epoch": 7.95, + "learning_rate": 1.0244701769718156e-06, + "loss": 1.5213, + "step": 36442 + }, + { + "epoch": 7.95, + "learning_rate": 1.0243609351103345e-06, + "loss": 1.0726, + "step": 36443 + }, + { + "epoch": 7.95, + "learning_rate": 1.024251693248853e-06, + "loss": 1.3285, + "step": 36444 + }, + { + "epoch": 7.95, + "learning_rate": 1.0241424513873717e-06, + "loss": 2.7834, + "step": 36445 + }, + { + "epoch": 7.95, + "learning_rate": 1.0240332095258903e-06, + "loss": 1.5695, + "step": 36446 + }, + { + "epoch": 7.95, + "learning_rate": 1.0239239676644091e-06, + "loss": 1.1053, + "step": 36447 + }, + { + "epoch": 7.95, + "learning_rate": 1.0238147258029277e-06, + "loss": 0.0987, + "step": 36448 + }, + { + "epoch": 7.95, + "learning_rate": 1.0237054839414465e-06, + "loss": 2.0492, + "step": 36449 + }, + { + "epoch": 7.96, + "learning_rate": 1.0235962420799652e-06, + "loss": 1.7469, + "step": 36450 + }, + { + "epoch": 7.96, + "learning_rate": 1.0234870002184838e-06, + "loss": 2.0645, + "step": 36451 + }, + { + "epoch": 7.96, + "learning_rate": 1.0233777583570024e-06, + "loss": 1.0447, + "step": 36452 + }, + { + "epoch": 7.96, + "learning_rate": 1.0232685164955212e-06, + "loss": 1.0887, + "step": 36453 + }, + { + "epoch": 7.96, + "learning_rate": 1.0231592746340398e-06, + "loss": 0.6252, + "step": 36454 + }, + { + "epoch": 7.96, + "learning_rate": 1.0230500327725586e-06, + "loss": 0.6356, + "step": 36455 + }, + { + "epoch": 7.96, + "learning_rate": 1.0229407909110772e-06, + "loss": 3.6338, + "step": 36456 + }, + { + "epoch": 7.96, + "learning_rate": 1.0228315490495959e-06, + "loss": 0.2128, + "step": 36457 + }, + { + "epoch": 7.96, + "learning_rate": 1.0227223071881145e-06, + "loss": 0.2017, + "step": 36458 + }, + { + "epoch": 7.96, + "learning_rate": 1.0226130653266333e-06, + "loss": 0.4374, + "step": 36459 + }, + { + "epoch": 7.96, + "learning_rate": 1.022503823465152e-06, + "loss": 1.3784, + "step": 36460 + }, + { + "epoch": 7.96, + "learning_rate": 1.0223945816036707e-06, + "loss": 1.347, + "step": 36461 + }, + { + "epoch": 7.96, + "learning_rate": 1.0222853397421893e-06, + "loss": 1.9801, + "step": 36462 + }, + { + "epoch": 7.96, + "learning_rate": 1.022176097880708e-06, + "loss": 1.6763, + "step": 36463 + }, + { + "epoch": 7.96, + "learning_rate": 1.0220668560192266e-06, + "loss": 0.4298, + "step": 36464 + }, + { + "epoch": 7.96, + "learning_rate": 1.0219576141577454e-06, + "loss": 0.9693, + "step": 36465 + }, + { + "epoch": 7.96, + "learning_rate": 1.021848372296264e-06, + "loss": 0.8985, + "step": 36466 + }, + { + "epoch": 7.96, + "learning_rate": 1.0217391304347828e-06, + "loss": 0.7551, + "step": 36467 + }, + { + "epoch": 7.96, + "learning_rate": 1.0216298885733014e-06, + "loss": 1.2484, + "step": 36468 + }, + { + "epoch": 7.96, + "learning_rate": 1.02152064671182e-06, + "loss": 1.0854, + "step": 36469 + }, + { + "epoch": 7.96, + "learning_rate": 1.0214114048503386e-06, + "loss": 0.7024, + "step": 36470 + }, + { + "epoch": 7.96, + "learning_rate": 1.0213021629888575e-06, + "loss": 1.3892, + "step": 36471 + }, + { + "epoch": 7.96, + "learning_rate": 1.021192921127376e-06, + "loss": 1.0411, + "step": 36472 + }, + { + "epoch": 7.96, + "learning_rate": 1.0210836792658947e-06, + "loss": 1.7423, + "step": 36473 + }, + { + "epoch": 7.96, + "learning_rate": 1.0209744374044135e-06, + "loss": 2.0304, + "step": 36474 + }, + { + "epoch": 7.96, + "learning_rate": 1.0208651955429321e-06, + "loss": 0.8653, + "step": 36475 + }, + { + "epoch": 7.96, + "learning_rate": 1.0207559536814507e-06, + "loss": 1.9932, + "step": 36476 + }, + { + "epoch": 7.96, + "learning_rate": 1.0206467118199696e-06, + "loss": 0.9432, + "step": 36477 + }, + { + "epoch": 7.96, + "learning_rate": 1.0205374699584882e-06, + "loss": 2.4175, + "step": 36478 + }, + { + "epoch": 7.96, + "learning_rate": 1.0204282280970068e-06, + "loss": 1.2088, + "step": 36479 + }, + { + "epoch": 7.96, + "learning_rate": 1.0203189862355254e-06, + "loss": 2.4611, + "step": 36480 + }, + { + "epoch": 7.96, + "learning_rate": 1.0202097443740442e-06, + "loss": 0.6942, + "step": 36481 + }, + { + "epoch": 7.96, + "learning_rate": 1.0201005025125628e-06, + "loss": 1.1129, + "step": 36482 + }, + { + "epoch": 7.96, + "learning_rate": 1.0199912606510817e-06, + "loss": 0.7402, + "step": 36483 + }, + { + "epoch": 7.96, + "learning_rate": 1.0198820187896003e-06, + "loss": 0.6093, + "step": 36484 + }, + { + "epoch": 7.96, + "learning_rate": 1.0197727769281189e-06, + "loss": 2.2208, + "step": 36485 + }, + { + "epoch": 7.96, + "learning_rate": 1.0196635350666375e-06, + "loss": 0.2857, + "step": 36486 + }, + { + "epoch": 7.96, + "learning_rate": 1.0195542932051563e-06, + "loss": 2.3944, + "step": 36487 + }, + { + "epoch": 7.96, + "learning_rate": 1.0194450513436751e-06, + "loss": 2.4693, + "step": 36488 + }, + { + "epoch": 7.96, + "learning_rate": 1.0193358094821937e-06, + "loss": 0.1213, + "step": 36489 + }, + { + "epoch": 7.96, + "learning_rate": 1.0192265676207124e-06, + "loss": 1.6959, + "step": 36490 + }, + { + "epoch": 7.96, + "learning_rate": 1.019117325759231e-06, + "loss": 0.4597, + "step": 36491 + }, + { + "epoch": 7.96, + "learning_rate": 1.0190080838977496e-06, + "loss": 0.784, + "step": 36492 + }, + { + "epoch": 7.96, + "learning_rate": 1.0188988420362684e-06, + "loss": 1.45, + "step": 36493 + }, + { + "epoch": 7.96, + "learning_rate": 1.0187896001747872e-06, + "loss": 1.2502, + "step": 36494 + }, + { + "epoch": 7.96, + "learning_rate": 1.0186803583133058e-06, + "loss": 0.954, + "step": 36495 + }, + { + "epoch": 7.97, + "learning_rate": 1.0185711164518244e-06, + "loss": 0.0613, + "step": 36496 + }, + { + "epoch": 7.97, + "learning_rate": 1.018461874590343e-06, + "loss": 0.2823, + "step": 36497 + }, + { + "epoch": 7.97, + "learning_rate": 1.0183526327288617e-06, + "loss": 1.7198, + "step": 36498 + }, + { + "epoch": 7.97, + "learning_rate": 1.0182433908673805e-06, + "loss": 1.6985, + "step": 36499 + }, + { + "epoch": 7.97, + "learning_rate": 1.018134149005899e-06, + "loss": 1.0649, + "step": 36500 + }, + { + "epoch": 7.97, + "learning_rate": 1.018024907144418e-06, + "loss": 0.7593, + "step": 36501 + }, + { + "epoch": 7.97, + "learning_rate": 1.0179156652829365e-06, + "loss": 1.29, + "step": 36502 + }, + { + "epoch": 7.97, + "learning_rate": 1.0178064234214551e-06, + "loss": 0.9178, + "step": 36503 + }, + { + "epoch": 7.97, + "learning_rate": 1.0176971815599738e-06, + "loss": 1.6585, + "step": 36504 + }, + { + "epoch": 7.97, + "learning_rate": 1.0175879396984926e-06, + "loss": 1.2992, + "step": 36505 + }, + { + "epoch": 7.97, + "learning_rate": 1.0174786978370112e-06, + "loss": 1.7873, + "step": 36506 + }, + { + "epoch": 7.97, + "learning_rate": 1.01736945597553e-06, + "loss": 1.33, + "step": 36507 + }, + { + "epoch": 7.97, + "learning_rate": 1.0172602141140486e-06, + "loss": 0.2306, + "step": 36508 + }, + { + "epoch": 7.97, + "learning_rate": 1.0171509722525672e-06, + "loss": 1.2233, + "step": 36509 + }, + { + "epoch": 7.97, + "learning_rate": 1.0170417303910858e-06, + "loss": 0.7071, + "step": 36510 + }, + { + "epoch": 7.97, + "learning_rate": 1.0169324885296047e-06, + "loss": 1.7505, + "step": 36511 + }, + { + "epoch": 7.97, + "learning_rate": 1.0168232466681233e-06, + "loss": 2.6474, + "step": 36512 + }, + { + "epoch": 7.97, + "learning_rate": 1.0167140048066419e-06, + "loss": 2.3205, + "step": 36513 + }, + { + "epoch": 7.97, + "learning_rate": 1.0166047629451607e-06, + "loss": 1.1556, + "step": 36514 + }, + { + "epoch": 7.97, + "learning_rate": 1.0164955210836793e-06, + "loss": 0.9799, + "step": 36515 + }, + { + "epoch": 7.97, + "learning_rate": 1.0163862792221981e-06, + "loss": 1.1196, + "step": 36516 + }, + { + "epoch": 7.97, + "learning_rate": 1.0162770373607168e-06, + "loss": 1.926, + "step": 36517 + }, + { + "epoch": 7.97, + "learning_rate": 1.0161677954992354e-06, + "loss": 0.7682, + "step": 36518 + }, + { + "epoch": 7.97, + "learning_rate": 1.016058553637754e-06, + "loss": 2.1686, + "step": 36519 + }, + { + "epoch": 7.97, + "learning_rate": 1.0159493117762728e-06, + "loss": 0.611, + "step": 36520 + }, + { + "epoch": 7.97, + "learning_rate": 1.0158400699147914e-06, + "loss": 1.9303, + "step": 36521 + }, + { + "epoch": 7.97, + "learning_rate": 1.0157308280533102e-06, + "loss": 0.6662, + "step": 36522 + }, + { + "epoch": 7.97, + "learning_rate": 1.0156215861918288e-06, + "loss": 1.0396, + "step": 36523 + }, + { + "epoch": 7.97, + "learning_rate": 1.0155123443303475e-06, + "loss": 1.2556, + "step": 36524 + }, + { + "epoch": 7.97, + "learning_rate": 1.015403102468866e-06, + "loss": 1.1515, + "step": 36525 + }, + { + "epoch": 7.97, + "learning_rate": 1.0152938606073847e-06, + "loss": 1.034, + "step": 36526 + }, + { + "epoch": 7.97, + "learning_rate": 1.0151846187459035e-06, + "loss": 0.5783, + "step": 36527 + }, + { + "epoch": 7.97, + "learning_rate": 1.0150753768844223e-06, + "loss": 0.8567, + "step": 36528 + }, + { + "epoch": 7.97, + "learning_rate": 1.014966135022941e-06, + "loss": 1.2815, + "step": 36529 + }, + { + "epoch": 7.97, + "learning_rate": 1.0148568931614595e-06, + "loss": 0.2438, + "step": 36530 + }, + { + "epoch": 7.97, + "learning_rate": 1.0147476512999782e-06, + "loss": 1.8881, + "step": 36531 + }, + { + "epoch": 7.97, + "learning_rate": 1.0146384094384968e-06, + "loss": 2.1258, + "step": 36532 + }, + { + "epoch": 7.97, + "learning_rate": 1.0145291675770156e-06, + "loss": 0.6845, + "step": 36533 + }, + { + "epoch": 7.97, + "learning_rate": 1.0144199257155344e-06, + "loss": 1.156, + "step": 36534 + }, + { + "epoch": 7.97, + "learning_rate": 1.014310683854053e-06, + "loss": 0.8503, + "step": 36535 + }, + { + "epoch": 7.97, + "learning_rate": 1.0142014419925716e-06, + "loss": 1.8914, + "step": 36536 + }, + { + "epoch": 7.97, + "learning_rate": 1.0140922001310902e-06, + "loss": 1.1693, + "step": 36537 + }, + { + "epoch": 7.97, + "learning_rate": 1.013982958269609e-06, + "loss": 1.1913, + "step": 36538 + }, + { + "epoch": 7.97, + "learning_rate": 1.0138737164081277e-06, + "loss": 1.8982, + "step": 36539 + }, + { + "epoch": 7.97, + "learning_rate": 1.0137644745466465e-06, + "loss": 0.9841, + "step": 36540 + }, + { + "epoch": 7.97, + "learning_rate": 1.0136552326851651e-06, + "loss": 0.2212, + "step": 36541 + }, + { + "epoch": 7.98, + "learning_rate": 1.0135459908236837e-06, + "loss": 1.2558, + "step": 36542 + }, + { + "epoch": 7.98, + "learning_rate": 1.0134367489622023e-06, + "loss": 1.9253, + "step": 36543 + }, + { + "epoch": 7.98, + "learning_rate": 1.0133275071007212e-06, + "loss": 0.5582, + "step": 36544 + }, + { + "epoch": 7.98, + "learning_rate": 1.0132182652392398e-06, + "loss": 0.9339, + "step": 36545 + }, + { + "epoch": 7.98, + "learning_rate": 1.0131090233777584e-06, + "loss": 0.545, + "step": 36546 + }, + { + "epoch": 7.98, + "learning_rate": 1.0129997815162772e-06, + "loss": 0.8181, + "step": 36547 + }, + { + "epoch": 7.98, + "learning_rate": 1.0128905396547958e-06, + "loss": 1.1704, + "step": 36548 + }, + { + "epoch": 7.98, + "learning_rate": 1.0127812977933144e-06, + "loss": 0.6557, + "step": 36549 + }, + { + "epoch": 7.98, + "learning_rate": 1.0126720559318332e-06, + "loss": 0.1251, + "step": 36550 + }, + { + "epoch": 7.98, + "learning_rate": 1.0125628140703519e-06, + "loss": 1.3874, + "step": 36551 + }, + { + "epoch": 7.98, + "learning_rate": 1.0124535722088705e-06, + "loss": 0.5568, + "step": 36552 + }, + { + "epoch": 7.98, + "learning_rate": 1.0123443303473893e-06, + "loss": 0.5994, + "step": 36553 + }, + { + "epoch": 7.98, + "learning_rate": 1.012235088485908e-06, + "loss": 1.1642, + "step": 36554 + }, + { + "epoch": 7.98, + "learning_rate": 1.0121258466244265e-06, + "loss": 0.549, + "step": 36555 + }, + { + "epoch": 7.98, + "learning_rate": 1.0120166047629453e-06, + "loss": 2.3709, + "step": 36556 + }, + { + "epoch": 7.98, + "learning_rate": 1.011907362901464e-06, + "loss": 1.3695, + "step": 36557 + }, + { + "epoch": 7.98, + "learning_rate": 1.0117981210399826e-06, + "loss": 0.1935, + "step": 36558 + }, + { + "epoch": 7.98, + "learning_rate": 1.0116888791785012e-06, + "loss": 0.4826, + "step": 36559 + }, + { + "epoch": 7.98, + "learning_rate": 1.01157963731702e-06, + "loss": 0.0832, + "step": 36560 + }, + { + "epoch": 7.98, + "learning_rate": 1.0114703954555386e-06, + "loss": 1.3749, + "step": 36561 + }, + { + "epoch": 7.98, + "learning_rate": 1.0113611535940574e-06, + "loss": 2.1054, + "step": 36562 + }, + { + "epoch": 7.98, + "learning_rate": 1.011251911732576e-06, + "loss": 1.3501, + "step": 36563 + }, + { + "epoch": 7.98, + "learning_rate": 1.0111426698710946e-06, + "loss": 1.305, + "step": 36564 + }, + { + "epoch": 7.98, + "learning_rate": 1.0110334280096133e-06, + "loss": 2.5608, + "step": 36565 + }, + { + "epoch": 7.98, + "learning_rate": 1.010924186148132e-06, + "loss": 0.2026, + "step": 36566 + }, + { + "epoch": 7.98, + "learning_rate": 1.0108149442866507e-06, + "loss": 1.0442, + "step": 36567 + }, + { + "epoch": 7.98, + "learning_rate": 1.0107057024251695e-06, + "loss": 0.9641, + "step": 36568 + }, + { + "epoch": 7.98, + "learning_rate": 1.0105964605636881e-06, + "loss": 1.7652, + "step": 36569 + }, + { + "epoch": 7.98, + "learning_rate": 1.0104872187022067e-06, + "loss": 0.9141, + "step": 36570 + }, + { + "epoch": 7.98, + "learning_rate": 1.0103779768407253e-06, + "loss": 1.3182, + "step": 36571 + }, + { + "epoch": 7.98, + "learning_rate": 1.0102687349792442e-06, + "loss": 1.253, + "step": 36572 + }, + { + "epoch": 7.98, + "learning_rate": 1.0101594931177628e-06, + "loss": 0.7504, + "step": 36573 + }, + { + "epoch": 7.98, + "learning_rate": 1.0100502512562816e-06, + "loss": 2.0597, + "step": 36574 + }, + { + "epoch": 7.98, + "learning_rate": 1.0099410093948002e-06, + "loss": 0.1948, + "step": 36575 + }, + { + "epoch": 7.98, + "learning_rate": 1.0098317675333188e-06, + "loss": 0.5693, + "step": 36576 + }, + { + "epoch": 7.98, + "learning_rate": 1.0097225256718374e-06, + "loss": 1.8423, + "step": 36577 + }, + { + "epoch": 7.98, + "learning_rate": 1.0096132838103563e-06, + "loss": 1.2292, + "step": 36578 + }, + { + "epoch": 7.98, + "learning_rate": 1.0095040419488749e-06, + "loss": 0.9951, + "step": 36579 + }, + { + "epoch": 7.98, + "learning_rate": 1.0093948000873937e-06, + "loss": 3.8481, + "step": 36580 + }, + { + "epoch": 7.98, + "learning_rate": 1.0092855582259123e-06, + "loss": 2.3105, + "step": 36581 + }, + { + "epoch": 7.98, + "learning_rate": 1.009176316364431e-06, + "loss": 0.3431, + "step": 36582 + }, + { + "epoch": 7.98, + "learning_rate": 1.0090670745029495e-06, + "loss": 0.3132, + "step": 36583 + }, + { + "epoch": 7.98, + "learning_rate": 1.0089578326414683e-06, + "loss": 3.1194, + "step": 36584 + }, + { + "epoch": 7.98, + "learning_rate": 1.008848590779987e-06, + "loss": 1.7968, + "step": 36585 + }, + { + "epoch": 7.98, + "learning_rate": 1.0087393489185058e-06, + "loss": 0.7436, + "step": 36586 + }, + { + "epoch": 7.98, + "learning_rate": 1.0086301070570244e-06, + "loss": 0.1288, + "step": 36587 + }, + { + "epoch": 7.99, + "learning_rate": 1.008520865195543e-06, + "loss": 1.4684, + "step": 36588 + }, + { + "epoch": 7.99, + "learning_rate": 1.0084116233340616e-06, + "loss": 1.2419, + "step": 36589 + }, + { + "epoch": 7.99, + "learning_rate": 1.0083023814725804e-06, + "loss": 1.005, + "step": 36590 + }, + { + "epoch": 7.99, + "learning_rate": 1.008193139611099e-06, + "loss": 1.7396, + "step": 36591 + }, + { + "epoch": 7.99, + "learning_rate": 1.0080838977496177e-06, + "loss": 1.5226, + "step": 36592 + }, + { + "epoch": 7.99, + "learning_rate": 1.0079746558881365e-06, + "loss": 3.9301, + "step": 36593 + }, + { + "epoch": 7.99, + "learning_rate": 1.007865414026655e-06, + "loss": 1.7556, + "step": 36594 + }, + { + "epoch": 7.99, + "learning_rate": 1.0077561721651737e-06, + "loss": 1.0428, + "step": 36595 + }, + { + "epoch": 7.99, + "learning_rate": 1.0076469303036925e-06, + "loss": 0.466, + "step": 36596 + }, + { + "epoch": 7.99, + "learning_rate": 1.0075376884422111e-06, + "loss": 0.6007, + "step": 36597 + }, + { + "epoch": 7.99, + "learning_rate": 1.0074284465807297e-06, + "loss": 0.5736, + "step": 36598 + }, + { + "epoch": 7.99, + "learning_rate": 1.0073192047192484e-06, + "loss": 2.2987, + "step": 36599 + }, + { + "epoch": 7.99, + "learning_rate": 1.0072099628577672e-06, + "loss": 1.4263, + "step": 36600 + }, + { + "epoch": 7.99, + "learning_rate": 1.0071007209962858e-06, + "loss": 1.3176, + "step": 36601 + }, + { + "epoch": 7.99, + "learning_rate": 1.0069914791348046e-06, + "loss": 0.2354, + "step": 36602 + }, + { + "epoch": 7.99, + "learning_rate": 1.0068822372733232e-06, + "loss": 2.433, + "step": 36603 + }, + { + "epoch": 7.99, + "learning_rate": 1.0067729954118418e-06, + "loss": 1.9735, + "step": 36604 + }, + { + "epoch": 7.99, + "learning_rate": 1.0066637535503604e-06, + "loss": 1.7285, + "step": 36605 + }, + { + "epoch": 7.99, + "learning_rate": 1.0065545116888793e-06, + "loss": 2.3429, + "step": 36606 + }, + { + "epoch": 7.99, + "learning_rate": 1.0064452698273979e-06, + "loss": 0.7694, + "step": 36607 + }, + { + "epoch": 7.99, + "learning_rate": 1.0063360279659167e-06, + "loss": 1.1884, + "step": 36608 + }, + { + "epoch": 7.99, + "learning_rate": 1.0062267861044353e-06, + "loss": 0.8944, + "step": 36609 + }, + { + "epoch": 7.99, + "learning_rate": 1.006117544242954e-06, + "loss": 1.0479, + "step": 36610 + }, + { + "epoch": 7.99, + "learning_rate": 1.0060083023814725e-06, + "loss": 0.8692, + "step": 36611 + }, + { + "epoch": 7.99, + "learning_rate": 1.0058990605199914e-06, + "loss": 0.6981, + "step": 36612 + }, + { + "epoch": 7.99, + "learning_rate": 1.0057898186585102e-06, + "loss": 1.368, + "step": 36613 + }, + { + "epoch": 7.99, + "learning_rate": 1.0056805767970288e-06, + "loss": 1.6562, + "step": 36614 + }, + { + "epoch": 7.99, + "learning_rate": 1.0055713349355474e-06, + "loss": 1.156, + "step": 36615 + }, + { + "epoch": 7.99, + "learning_rate": 1.005462093074066e-06, + "loss": 0.5797, + "step": 36616 + }, + { + "epoch": 7.99, + "learning_rate": 1.0053528512125846e-06, + "loss": 3.6538, + "step": 36617 + }, + { + "epoch": 7.99, + "learning_rate": 1.0052436093511035e-06, + "loss": 1.2004, + "step": 36618 + }, + { + "epoch": 7.99, + "learning_rate": 1.005134367489622e-06, + "loss": 3.3451, + "step": 36619 + }, + { + "epoch": 7.99, + "learning_rate": 1.0050251256281409e-06, + "loss": 1.6581, + "step": 36620 + }, + { + "epoch": 7.99, + "learning_rate": 1.0049158837666595e-06, + "loss": 0.6256, + "step": 36621 + }, + { + "epoch": 7.99, + "learning_rate": 1.0048066419051781e-06, + "loss": 2.1311, + "step": 36622 + }, + { + "epoch": 7.99, + "learning_rate": 1.0046974000436967e-06, + "loss": 1.2225, + "step": 36623 + }, + { + "epoch": 7.99, + "learning_rate": 1.0045881581822155e-06, + "loss": 1.885, + "step": 36624 + }, + { + "epoch": 7.99, + "learning_rate": 1.0044789163207342e-06, + "loss": 0.3914, + "step": 36625 + }, + { + "epoch": 7.99, + "learning_rate": 1.004369674459253e-06, + "loss": 1.4012, + "step": 36626 + }, + { + "epoch": 7.99, + "learning_rate": 1.0042604325977716e-06, + "loss": 1.5059, + "step": 36627 + }, + { + "epoch": 7.99, + "learning_rate": 1.0041511907362902e-06, + "loss": 2.2865, + "step": 36628 + }, + { + "epoch": 7.99, + "learning_rate": 1.0040419488748088e-06, + "loss": 1.448, + "step": 36629 + }, + { + "epoch": 7.99, + "learning_rate": 1.0039327070133276e-06, + "loss": 0.8558, + "step": 36630 + }, + { + "epoch": 7.99, + "learning_rate": 1.0038234651518462e-06, + "loss": 1.3188, + "step": 36631 + }, + { + "epoch": 7.99, + "learning_rate": 1.0037142232903649e-06, + "loss": 1.378, + "step": 36632 + }, + { + "epoch": 7.99, + "learning_rate": 1.0036049814288837e-06, + "loss": 1.1208, + "step": 36633 + }, + { + "epoch": 8.0, + "learning_rate": 1.0034957395674023e-06, + "loss": 0.8239, + "step": 36634 + }, + { + "epoch": 8.0, + "learning_rate": 1.0033864977059211e-06, + "loss": 0.6148, + "step": 36635 + }, + { + "epoch": 8.0, + "learning_rate": 1.0032772558444397e-06, + "loss": 2.6826, + "step": 36636 + }, + { + "epoch": 8.0, + "learning_rate": 1.0031680139829583e-06, + "loss": 1.414, + "step": 36637 + }, + { + "epoch": 8.0, + "learning_rate": 1.003058772121477e-06, + "loss": 2.861, + "step": 36638 + }, + { + "epoch": 8.0, + "learning_rate": 1.0029495302599958e-06, + "loss": 0.1425, + "step": 36639 + }, + { + "epoch": 8.0, + "learning_rate": 1.0028402883985144e-06, + "loss": 0.5933, + "step": 36640 + }, + { + "epoch": 8.0, + "learning_rate": 1.0027310465370332e-06, + "loss": 1.8478, + "step": 36641 + }, + { + "epoch": 8.0, + "learning_rate": 1.0026218046755518e-06, + "loss": 1.1375, + "step": 36642 + }, + { + "epoch": 8.0, + "learning_rate": 1.0025125628140704e-06, + "loss": 0.4059, + "step": 36643 + }, + { + "epoch": 8.0, + "learning_rate": 1.002403320952589e-06, + "loss": 0.8121, + "step": 36644 + }, + { + "epoch": 8.0, + "learning_rate": 1.0022940790911076e-06, + "loss": 1.3439, + "step": 36645 + }, + { + "epoch": 8.0, + "learning_rate": 1.0021848372296265e-06, + "loss": 1.1662, + "step": 36646 + }, + { + "epoch": 8.0, + "learning_rate": 1.0020755953681453e-06, + "loss": 2.6199, + "step": 36647 + }, + { + "epoch": 8.0, + "learning_rate": 1.001966353506664e-06, + "loss": 0.9293, + "step": 36648 + }, + { + "epoch": 8.0, + "learning_rate": 1.0018571116451825e-06, + "loss": 0.8591, + "step": 36649 + }, + { + "epoch": 8.0, + "learning_rate": 1.0017478697837011e-06, + "loss": 2.3054, + "step": 36650 + }, + { + "epoch": 8.0, + "learning_rate": 1.0016386279222197e-06, + "loss": 2.2652, + "step": 36651 + }, + { + "epoch": 8.0, + "learning_rate": 1.0015293860607386e-06, + "loss": 3.7314, + "step": 36652 + }, + { + "epoch": 8.0, + "learning_rate": 1.0014201441992574e-06, + "loss": 0.6269, + "step": 36653 + }, + { + "epoch": 8.0, + "learning_rate": 1.001310902337776e-06, + "loss": 0.5963, + "step": 36654 + }, + { + "epoch": 8.0, + "learning_rate": 1.0012016604762946e-06, + "loss": 0.2649, + "step": 36655 + }, + { + "epoch": 8.0, + "learning_rate": 1.0010924186148132e-06, + "loss": 1.0434, + "step": 36656 + }, + { + "epoch": 8.0, + "eval_accuracy": 0.7057283142389525, + "eval_loss": 1.2027839422225952, + "eval_runtime": 335.5274, + "eval_samples_per_second": 18.21, + "eval_steps_per_second": 4.554, + "step": 36656 + }, + { + "epoch": 8.0, + "learning_rate": 1.0009831767533318e-06, + "loss": 0.3065, + "step": 36657 + }, + { + "epoch": 8.0, + "learning_rate": 1.0008739348918506e-06, + "loss": 1.0029, + "step": 36658 + }, + { + "epoch": 8.0, + "learning_rate": 1.0007646930303695e-06, + "loss": 1.7875, + "step": 36659 + }, + { + "epoch": 8.0, + "learning_rate": 1.000655451168888e-06, + "loss": 0.6833, + "step": 36660 + }, + { + "epoch": 8.0, + "learning_rate": 1.0005462093074067e-06, + "loss": 0.8862, + "step": 36661 + }, + { + "epoch": 8.0, + "learning_rate": 1.0004369674459253e-06, + "loss": 1.3874, + "step": 36662 + }, + { + "epoch": 8.0, + "learning_rate": 1.0003277255844441e-06, + "loss": 1.3613, + "step": 36663 + }, + { + "epoch": 8.0, + "learning_rate": 1.0002184837229627e-06, + "loss": 0.9592, + "step": 36664 + }, + { + "epoch": 8.0, + "learning_rate": 1.0001092418614813e-06, + "loss": 0.8496, + "step": 36665 + }, + { + "epoch": 8.0, + "learning_rate": 1.0000000000000002e-06, + "loss": 1.0902, + "step": 36666 + }, + { + "epoch": 8.0, + "learning_rate": 9.998907581385188e-07, + "loss": 1.2195, + "step": 36667 + }, + { + "epoch": 8.0, + "learning_rate": 9.997815162770374e-07, + "loss": 0.4358, + "step": 36668 + }, + { + "epoch": 8.0, + "learning_rate": 9.996722744155562e-07, + "loss": 2.4073, + "step": 36669 + }, + { + "epoch": 8.0, + "learning_rate": 9.995630325540748e-07, + "loss": 1.8401, + "step": 36670 + }, + { + "epoch": 8.0, + "learning_rate": 9.994537906925934e-07, + "loss": 0.0688, + "step": 36671 + }, + { + "epoch": 8.0, + "learning_rate": 9.993445488311123e-07, + "loss": 2.1502, + "step": 36672 + }, + { + "epoch": 8.0, + "learning_rate": 9.992353069696309e-07, + "loss": 0.243, + "step": 36673 + }, + { + "epoch": 8.0, + "learning_rate": 9.991260651081495e-07, + "loss": 2.3129, + "step": 36674 + }, + { + "epoch": 8.0, + "learning_rate": 9.990168232466683e-07, + "loss": 1.8732, + "step": 36675 + }, + { + "epoch": 8.0, + "learning_rate": 9.98907581385187e-07, + "loss": 1.2246, + "step": 36676 + }, + { + "epoch": 8.0, + "learning_rate": 9.987983395237055e-07, + "loss": 1.5727, + "step": 36677 + }, + { + "epoch": 8.0, + "learning_rate": 9.986890976622241e-07, + "loss": 3.4511, + "step": 36678 + }, + { + "epoch": 8.01, + "learning_rate": 9.98579855800743e-07, + "loss": 1.8105, + "step": 36679 + }, + { + "epoch": 8.01, + "learning_rate": 9.984706139392616e-07, + "loss": 0.4465, + "step": 36680 + }, + { + "epoch": 8.01, + "learning_rate": 9.983613720777804e-07, + "loss": 3.2677, + "step": 36681 + }, + { + "epoch": 8.01, + "learning_rate": 9.98252130216299e-07, + "loss": 0.8794, + "step": 36682 + }, + { + "epoch": 8.01, + "learning_rate": 9.981428883548176e-07, + "loss": 0.3622, + "step": 36683 + }, + { + "epoch": 8.01, + "learning_rate": 9.980336464933362e-07, + "loss": 1.6419, + "step": 36684 + }, + { + "epoch": 8.01, + "learning_rate": 9.97924404631855e-07, + "loss": 1.3226, + "step": 36685 + }, + { + "epoch": 8.01, + "learning_rate": 9.978151627703737e-07, + "loss": 0.6022, + "step": 36686 + }, + { + "epoch": 8.01, + "learning_rate": 9.977059209088925e-07, + "loss": 1.2343, + "step": 36687 + }, + { + "epoch": 8.01, + "learning_rate": 9.97596679047411e-07, + "loss": 0.7226, + "step": 36688 + }, + { + "epoch": 8.01, + "learning_rate": 9.974874371859297e-07, + "loss": 1.5393, + "step": 36689 + }, + { + "epoch": 8.01, + "learning_rate": 9.973781953244483e-07, + "loss": 3.5624, + "step": 36690 + }, + { + "epoch": 8.01, + "learning_rate": 9.972689534629671e-07, + "loss": 1.2812, + "step": 36691 + }, + { + "epoch": 8.01, + "learning_rate": 9.971597116014857e-07, + "loss": 1.5727, + "step": 36692 + }, + { + "epoch": 8.01, + "learning_rate": 9.970504697400046e-07, + "loss": 3.0992, + "step": 36693 + }, + { + "epoch": 8.01, + "learning_rate": 9.969412278785232e-07, + "loss": 0.6177, + "step": 36694 + }, + { + "epoch": 8.01, + "learning_rate": 9.968319860170418e-07, + "loss": 1.9997, + "step": 36695 + }, + { + "epoch": 8.01, + "learning_rate": 9.967227441555604e-07, + "loss": 0.2931, + "step": 36696 + }, + { + "epoch": 8.01, + "learning_rate": 9.966135022940792e-07, + "loss": 1.2311, + "step": 36697 + }, + { + "epoch": 8.01, + "learning_rate": 9.965042604325978e-07, + "loss": 2.7447, + "step": 36698 + }, + { + "epoch": 8.01, + "learning_rate": 9.963950185711167e-07, + "loss": 0.3025, + "step": 36699 + }, + { + "epoch": 8.01, + "learning_rate": 9.962857767096353e-07, + "loss": 0.1048, + "step": 36700 + }, + { + "epoch": 8.01, + "learning_rate": 9.961765348481539e-07, + "loss": 0.7133, + "step": 36701 + }, + { + "epoch": 8.01, + "learning_rate": 9.960672929866725e-07, + "loss": 0.1835, + "step": 36702 + }, + { + "epoch": 8.01, + "learning_rate": 9.959580511251913e-07, + "loss": 1.1751, + "step": 36703 + }, + { + "epoch": 8.01, + "learning_rate": 9.9584880926371e-07, + "loss": 0.4916, + "step": 36704 + }, + { + "epoch": 8.01, + "learning_rate": 9.957395674022285e-07, + "loss": 1.2114, + "step": 36705 + }, + { + "epoch": 8.01, + "learning_rate": 9.956303255407474e-07, + "loss": 0.3472, + "step": 36706 + }, + { + "epoch": 8.01, + "learning_rate": 9.95521083679266e-07, + "loss": 0.7835, + "step": 36707 + }, + { + "epoch": 8.01, + "learning_rate": 9.954118418177846e-07, + "loss": 0.967, + "step": 36708 + }, + { + "epoch": 8.01, + "learning_rate": 9.953025999563034e-07, + "loss": 2.4913, + "step": 36709 + }, + { + "epoch": 8.01, + "learning_rate": 9.95193358094822e-07, + "loss": 1.9273, + "step": 36710 + }, + { + "epoch": 8.01, + "learning_rate": 9.950841162333406e-07, + "loss": 1.0121, + "step": 36711 + }, + { + "epoch": 8.01, + "learning_rate": 9.949748743718594e-07, + "loss": 0.1028, + "step": 36712 + }, + { + "epoch": 8.01, + "learning_rate": 9.94865632510378e-07, + "loss": 1.1442, + "step": 36713 + }, + { + "epoch": 8.01, + "learning_rate": 9.947563906488967e-07, + "loss": 0.6927, + "step": 36714 + }, + { + "epoch": 8.01, + "learning_rate": 9.946471487874155e-07, + "loss": 0.688, + "step": 36715 + }, + { + "epoch": 8.01, + "learning_rate": 9.94537906925934e-07, + "loss": 1.1303, + "step": 36716 + }, + { + "epoch": 8.01, + "learning_rate": 9.944286650644527e-07, + "loss": 1.2588, + "step": 36717 + }, + { + "epoch": 8.01, + "learning_rate": 9.943194232029713e-07, + "loss": 1.1551, + "step": 36718 + }, + { + "epoch": 8.01, + "learning_rate": 9.942101813414901e-07, + "loss": 1.57, + "step": 36719 + }, + { + "epoch": 8.01, + "learning_rate": 9.941009394800088e-07, + "loss": 0.7727, + "step": 36720 + }, + { + "epoch": 8.01, + "learning_rate": 9.939916976185276e-07, + "loss": 0.8594, + "step": 36721 + }, + { + "epoch": 8.01, + "learning_rate": 9.938824557570462e-07, + "loss": 1.3121, + "step": 36722 + }, + { + "epoch": 8.01, + "learning_rate": 9.937732138955648e-07, + "loss": 2.1204, + "step": 36723 + }, + { + "epoch": 8.01, + "learning_rate": 9.936639720340834e-07, + "loss": 1.9003, + "step": 36724 + }, + { + "epoch": 8.02, + "learning_rate": 9.935547301726022e-07, + "loss": 1.2684, + "step": 36725 + }, + { + "epoch": 8.02, + "learning_rate": 9.934454883111208e-07, + "loss": 1.0407, + "step": 36726 + }, + { + "epoch": 8.02, + "learning_rate": 9.933362464496397e-07, + "loss": 1.2043, + "step": 36727 + }, + { + "epoch": 8.02, + "learning_rate": 9.932270045881583e-07, + "loss": 0.7546, + "step": 36728 + }, + { + "epoch": 8.02, + "learning_rate": 9.931177627266769e-07, + "loss": 0.9298, + "step": 36729 + }, + { + "epoch": 8.02, + "learning_rate": 9.930085208651955e-07, + "loss": 0.546, + "step": 36730 + }, + { + "epoch": 8.02, + "learning_rate": 9.928992790037143e-07, + "loss": 1.2581, + "step": 36731 + }, + { + "epoch": 8.02, + "learning_rate": 9.92790037142233e-07, + "loss": 1.0389, + "step": 36732 + }, + { + "epoch": 8.02, + "learning_rate": 9.926807952807518e-07, + "loss": 0.9637, + "step": 36733 + }, + { + "epoch": 8.02, + "learning_rate": 9.925715534192704e-07, + "loss": 1.0014, + "step": 36734 + }, + { + "epoch": 8.02, + "learning_rate": 9.92462311557789e-07, + "loss": 1.1861, + "step": 36735 + }, + { + "epoch": 8.02, + "learning_rate": 9.923530696963076e-07, + "loss": 1.5243, + "step": 36736 + }, + { + "epoch": 8.02, + "learning_rate": 9.922438278348264e-07, + "loss": 0.6839, + "step": 36737 + }, + { + "epoch": 8.02, + "learning_rate": 9.92134585973345e-07, + "loss": 0.4825, + "step": 36738 + }, + { + "epoch": 8.02, + "learning_rate": 9.920253441118639e-07, + "loss": 0.8855, + "step": 36739 + }, + { + "epoch": 8.02, + "learning_rate": 9.919161022503825e-07, + "loss": 0.1888, + "step": 36740 + }, + { + "epoch": 8.02, + "learning_rate": 9.91806860388901e-07, + "loss": 0.9676, + "step": 36741 + }, + { + "epoch": 8.02, + "learning_rate": 9.916976185274197e-07, + "loss": 2.1048, + "step": 36742 + }, + { + "epoch": 8.02, + "learning_rate": 9.915883766659385e-07, + "loss": 1.2049, + "step": 36743 + }, + { + "epoch": 8.02, + "learning_rate": 9.914791348044571e-07, + "loss": 0.264, + "step": 36744 + }, + { + "epoch": 8.02, + "learning_rate": 9.91369892942976e-07, + "loss": 1.0659, + "step": 36745 + }, + { + "epoch": 8.02, + "learning_rate": 9.912606510814946e-07, + "loss": 0.6862, + "step": 36746 + }, + { + "epoch": 8.02, + "learning_rate": 9.911514092200132e-07, + "loss": 1.1141, + "step": 36747 + }, + { + "epoch": 8.02, + "learning_rate": 9.910421673585318e-07, + "loss": 0.3094, + "step": 36748 + }, + { + "epoch": 8.02, + "learning_rate": 9.909329254970506e-07, + "loss": 0.9678, + "step": 36749 + }, + { + "epoch": 8.02, + "learning_rate": 9.908236836355692e-07, + "loss": 0.9836, + "step": 36750 + }, + { + "epoch": 8.02, + "learning_rate": 9.907144417740878e-07, + "loss": 1.6881, + "step": 36751 + }, + { + "epoch": 8.02, + "learning_rate": 9.906051999126066e-07, + "loss": 1.5546, + "step": 36752 + }, + { + "epoch": 8.02, + "learning_rate": 9.904959580511253e-07, + "loss": 0.8766, + "step": 36753 + }, + { + "epoch": 8.02, + "learning_rate": 9.903867161896439e-07, + "loss": 0.4848, + "step": 36754 + }, + { + "epoch": 8.02, + "learning_rate": 9.902774743281627e-07, + "loss": 1.6929, + "step": 36755 + }, + { + "epoch": 8.02, + "learning_rate": 9.901682324666813e-07, + "loss": 1.0803, + "step": 36756 + }, + { + "epoch": 8.02, + "learning_rate": 9.900589906052e-07, + "loss": 1.3052, + "step": 36757 + }, + { + "epoch": 8.02, + "learning_rate": 9.899497487437187e-07, + "loss": 0.3697, + "step": 36758 + }, + { + "epoch": 8.02, + "learning_rate": 9.898405068822373e-07, + "loss": 1.1927, + "step": 36759 + }, + { + "epoch": 8.02, + "learning_rate": 9.897312650207562e-07, + "loss": 1.0897, + "step": 36760 + }, + { + "epoch": 8.02, + "learning_rate": 9.896220231592748e-07, + "loss": 1.7234, + "step": 36761 + }, + { + "epoch": 8.02, + "learning_rate": 9.895127812977934e-07, + "loss": 0.757, + "step": 36762 + }, + { + "epoch": 8.02, + "learning_rate": 9.89403539436312e-07, + "loss": 1.1417, + "step": 36763 + }, + { + "epoch": 8.02, + "learning_rate": 9.892942975748306e-07, + "loss": 1.3002, + "step": 36764 + }, + { + "epoch": 8.02, + "learning_rate": 9.891850557133494e-07, + "loss": 0.0738, + "step": 36765 + }, + { + "epoch": 8.02, + "learning_rate": 9.890758138518683e-07, + "loss": 3.3726, + "step": 36766 + }, + { + "epoch": 8.02, + "learning_rate": 9.889665719903869e-07, + "loss": 1.4453, + "step": 36767 + }, + { + "epoch": 8.02, + "learning_rate": 9.888573301289055e-07, + "loss": 0.8003, + "step": 36768 + }, + { + "epoch": 8.02, + "learning_rate": 9.88748088267424e-07, + "loss": 1.3262, + "step": 36769 + }, + { + "epoch": 8.02, + "learning_rate": 9.886388464059427e-07, + "loss": 1.0341, + "step": 36770 + }, + { + "epoch": 8.03, + "learning_rate": 9.885296045444615e-07, + "loss": 1.8783, + "step": 36771 + }, + { + "epoch": 8.03, + "learning_rate": 9.884203626829803e-07, + "loss": 0.1291, + "step": 36772 + }, + { + "epoch": 8.03, + "learning_rate": 9.88311120821499e-07, + "loss": 1.1653, + "step": 36773 + }, + { + "epoch": 8.03, + "learning_rate": 9.882018789600176e-07, + "loss": 1.4138, + "step": 36774 + }, + { + "epoch": 8.03, + "learning_rate": 9.880926370985362e-07, + "loss": 0.6122, + "step": 36775 + }, + { + "epoch": 8.03, + "learning_rate": 9.879833952370548e-07, + "loss": 0.8215, + "step": 36776 + }, + { + "epoch": 8.03, + "learning_rate": 9.878741533755736e-07, + "loss": 3.2506, + "step": 36777 + }, + { + "epoch": 8.03, + "learning_rate": 9.877649115140924e-07, + "loss": 0.692, + "step": 36778 + }, + { + "epoch": 8.03, + "learning_rate": 9.87655669652611e-07, + "loss": 1.6593, + "step": 36779 + }, + { + "epoch": 8.03, + "learning_rate": 9.875464277911297e-07, + "loss": 1.9655, + "step": 36780 + }, + { + "epoch": 8.03, + "learning_rate": 9.874371859296483e-07, + "loss": 0.1805, + "step": 36781 + }, + { + "epoch": 8.03, + "learning_rate": 9.87327944068167e-07, + "loss": 0.5431, + "step": 36782 + }, + { + "epoch": 8.03, + "learning_rate": 9.872187022066857e-07, + "loss": 0.9448, + "step": 36783 + }, + { + "epoch": 8.03, + "learning_rate": 9.871094603452043e-07, + "loss": 1.0439, + "step": 36784 + }, + { + "epoch": 8.03, + "learning_rate": 9.870002184837231e-07, + "loss": 0.9704, + "step": 36785 + }, + { + "epoch": 8.03, + "learning_rate": 9.868909766222417e-07, + "loss": 1.1364, + "step": 36786 + }, + { + "epoch": 8.03, + "learning_rate": 9.867817347607604e-07, + "loss": 1.3021, + "step": 36787 + }, + { + "epoch": 8.03, + "learning_rate": 9.866724928992792e-07, + "loss": 0.7249, + "step": 36788 + }, + { + "epoch": 8.03, + "learning_rate": 9.865632510377978e-07, + "loss": 0.9925, + "step": 36789 + }, + { + "epoch": 8.03, + "learning_rate": 9.864540091763164e-07, + "loss": 2.0264, + "step": 36790 + }, + { + "epoch": 8.03, + "learning_rate": 9.863447673148352e-07, + "loss": 0.4227, + "step": 36791 + }, + { + "epoch": 8.03, + "learning_rate": 9.862355254533538e-07, + "loss": 0.34, + "step": 36792 + }, + { + "epoch": 8.03, + "learning_rate": 9.861262835918724e-07, + "loss": 2.1752, + "step": 36793 + }, + { + "epoch": 8.03, + "learning_rate": 9.860170417303913e-07, + "loss": 0.0831, + "step": 36794 + }, + { + "epoch": 8.03, + "learning_rate": 9.859077998689099e-07, + "loss": 0.7954, + "step": 36795 + }, + { + "epoch": 8.03, + "learning_rate": 9.857985580074285e-07, + "loss": 1.0336, + "step": 36796 + }, + { + "epoch": 8.03, + "learning_rate": 9.85689316145947e-07, + "loss": 2.5694, + "step": 36797 + }, + { + "epoch": 8.03, + "learning_rate": 9.85580074284466e-07, + "loss": 1.6982, + "step": 36798 + }, + { + "epoch": 8.03, + "learning_rate": 9.854708324229845e-07, + "loss": 1.5534, + "step": 36799 + }, + { + "epoch": 8.03, + "learning_rate": 9.853615905615034e-07, + "loss": 0.9999, + "step": 36800 + }, + { + "epoch": 8.03, + "learning_rate": 9.85252348700022e-07, + "loss": 0.9614, + "step": 36801 + }, + { + "epoch": 8.03, + "learning_rate": 9.851431068385406e-07, + "loss": 0.227, + "step": 36802 + }, + { + "epoch": 8.03, + "learning_rate": 9.850338649770592e-07, + "loss": 0.6894, + "step": 36803 + }, + { + "epoch": 8.03, + "learning_rate": 9.849246231155778e-07, + "loss": 0.9507, + "step": 36804 + }, + { + "epoch": 8.03, + "learning_rate": 9.848153812540966e-07, + "loss": 0.7014, + "step": 36805 + }, + { + "epoch": 8.03, + "learning_rate": 9.847061393926154e-07, + "loss": 0.097, + "step": 36806 + }, + { + "epoch": 8.03, + "learning_rate": 9.84596897531134e-07, + "loss": 0.2535, + "step": 36807 + }, + { + "epoch": 8.03, + "learning_rate": 9.844876556696527e-07, + "loss": 0.7043, + "step": 36808 + }, + { + "epoch": 8.03, + "learning_rate": 9.843784138081713e-07, + "loss": 0.4628, + "step": 36809 + }, + { + "epoch": 8.03, + "learning_rate": 9.8426917194669e-07, + "loss": 1.0452, + "step": 36810 + }, + { + "epoch": 8.03, + "learning_rate": 9.841599300852087e-07, + "loss": 1.7064, + "step": 36811 + }, + { + "epoch": 8.03, + "learning_rate": 9.840506882237275e-07, + "loss": 0.351, + "step": 36812 + }, + { + "epoch": 8.03, + "learning_rate": 9.839414463622461e-07, + "loss": 0.4445, + "step": 36813 + }, + { + "epoch": 8.03, + "learning_rate": 9.838322045007648e-07, + "loss": 0.3826, + "step": 36814 + }, + { + "epoch": 8.03, + "learning_rate": 9.837229626392834e-07, + "loss": 0.9026, + "step": 36815 + }, + { + "epoch": 8.03, + "learning_rate": 9.836137207778022e-07, + "loss": 1.6633, + "step": 36816 + }, + { + "epoch": 8.04, + "learning_rate": 9.835044789163208e-07, + "loss": 0.4023, + "step": 36817 + }, + { + "epoch": 8.04, + "learning_rate": 9.833952370548396e-07, + "loss": 0.6239, + "step": 36818 + }, + { + "epoch": 8.04, + "learning_rate": 9.832859951933582e-07, + "loss": 0.7094, + "step": 36819 + }, + { + "epoch": 8.04, + "learning_rate": 9.831767533318768e-07, + "loss": 2.535, + "step": 36820 + }, + { + "epoch": 8.04, + "learning_rate": 9.830675114703955e-07, + "loss": 1.7617, + "step": 36821 + }, + { + "epoch": 8.04, + "learning_rate": 9.829582696089143e-07, + "loss": 0.7528, + "step": 36822 + }, + { + "epoch": 8.04, + "learning_rate": 9.828490277474329e-07, + "loss": 2.0106, + "step": 36823 + }, + { + "epoch": 8.04, + "learning_rate": 9.827397858859515e-07, + "loss": 0.4681, + "step": 36824 + }, + { + "epoch": 8.04, + "learning_rate": 9.826305440244703e-07, + "loss": 1.4568, + "step": 36825 + }, + { + "epoch": 8.04, + "learning_rate": 9.82521302162989e-07, + "loss": 0.1586, + "step": 36826 + }, + { + "epoch": 8.04, + "learning_rate": 9.824120603015075e-07, + "loss": 0.9568, + "step": 36827 + }, + { + "epoch": 8.04, + "learning_rate": 9.823028184400264e-07, + "loss": 0.6937, + "step": 36828 + }, + { + "epoch": 8.04, + "learning_rate": 9.82193576578545e-07, + "loss": 1.7088, + "step": 36829 + }, + { + "epoch": 8.04, + "learning_rate": 9.820843347170636e-07, + "loss": 0.9611, + "step": 36830 + }, + { + "epoch": 8.04, + "learning_rate": 9.819750928555824e-07, + "loss": 0.1919, + "step": 36831 + }, + { + "epoch": 8.04, + "learning_rate": 9.81865850994101e-07, + "loss": 1.5052, + "step": 36832 + }, + { + "epoch": 8.04, + "learning_rate": 9.817566091326196e-07, + "loss": 0.9684, + "step": 36833 + }, + { + "epoch": 8.04, + "learning_rate": 9.816473672711385e-07, + "loss": 0.5121, + "step": 36834 + }, + { + "epoch": 8.04, + "learning_rate": 9.81538125409657e-07, + "loss": 0.9581, + "step": 36835 + }, + { + "epoch": 8.04, + "learning_rate": 9.814288835481757e-07, + "loss": 0.1355, + "step": 36836 + }, + { + "epoch": 8.04, + "learning_rate": 9.813196416866943e-07, + "loss": 1.5378, + "step": 36837 + }, + { + "epoch": 8.04, + "learning_rate": 9.812103998252131e-07, + "loss": 1.7878, + "step": 36838 + }, + { + "epoch": 8.04, + "learning_rate": 9.811011579637317e-07, + "loss": 1.1728, + "step": 36839 + }, + { + "epoch": 8.04, + "learning_rate": 9.809919161022505e-07, + "loss": 1.3113, + "step": 36840 + }, + { + "epoch": 8.04, + "learning_rate": 9.808826742407692e-07, + "loss": 0.5177, + "step": 36841 + }, + { + "epoch": 8.04, + "learning_rate": 9.807734323792878e-07, + "loss": 2.523, + "step": 36842 + }, + { + "epoch": 8.04, + "learning_rate": 9.806641905178064e-07, + "loss": 0.6285, + "step": 36843 + }, + { + "epoch": 8.04, + "learning_rate": 9.805549486563252e-07, + "loss": 1.7646, + "step": 36844 + }, + { + "epoch": 8.04, + "learning_rate": 9.804457067948438e-07, + "loss": 0.3524, + "step": 36845 + }, + { + "epoch": 8.04, + "learning_rate": 9.803364649333626e-07, + "loss": 1.0838, + "step": 36846 + }, + { + "epoch": 8.04, + "learning_rate": 9.802272230718812e-07, + "loss": 2.1839, + "step": 36847 + }, + { + "epoch": 8.04, + "learning_rate": 9.801179812103999e-07, + "loss": 1.7078, + "step": 36848 + }, + { + "epoch": 8.04, + "learning_rate": 9.800087393489185e-07, + "loss": 0.5848, + "step": 36849 + }, + { + "epoch": 8.04, + "learning_rate": 9.798994974874373e-07, + "loss": 2.0061, + "step": 36850 + }, + { + "epoch": 8.04, + "learning_rate": 9.79790255625956e-07, + "loss": 2.2386, + "step": 36851 + }, + { + "epoch": 8.04, + "learning_rate": 9.796810137644747e-07, + "loss": 0.6186, + "step": 36852 + }, + { + "epoch": 8.04, + "learning_rate": 9.795717719029933e-07, + "loss": 0.0379, + "step": 36853 + }, + { + "epoch": 8.04, + "learning_rate": 9.79462530041512e-07, + "loss": 0.7125, + "step": 36854 + }, + { + "epoch": 8.04, + "learning_rate": 9.793532881800306e-07, + "loss": 0.7879, + "step": 36855 + }, + { + "epoch": 8.04, + "learning_rate": 9.792440463185494e-07, + "loss": 2.7385, + "step": 36856 + }, + { + "epoch": 8.04, + "learning_rate": 9.79134804457068e-07, + "loss": 0.1028, + "step": 36857 + }, + { + "epoch": 8.04, + "learning_rate": 9.790255625955868e-07, + "loss": 0.4564, + "step": 36858 + }, + { + "epoch": 8.04, + "learning_rate": 9.789163207341054e-07, + "loss": 2.6146, + "step": 36859 + }, + { + "epoch": 8.04, + "learning_rate": 9.78807078872624e-07, + "loss": 0.862, + "step": 36860 + }, + { + "epoch": 8.04, + "learning_rate": 9.786978370111426e-07, + "loss": 1.847, + "step": 36861 + }, + { + "epoch": 8.04, + "learning_rate": 9.785885951496615e-07, + "loss": 0.1021, + "step": 36862 + }, + { + "epoch": 8.05, + "learning_rate": 9.7847935328818e-07, + "loss": 1.1797, + "step": 36863 + }, + { + "epoch": 8.05, + "learning_rate": 9.78370111426699e-07, + "loss": 2.2014, + "step": 36864 + }, + { + "epoch": 8.05, + "learning_rate": 9.782608695652175e-07, + "loss": 0.7867, + "step": 36865 + }, + { + "epoch": 8.05, + "learning_rate": 9.781516277037361e-07, + "loss": 1.3086, + "step": 36866 + }, + { + "epoch": 8.05, + "learning_rate": 9.780423858422547e-07, + "loss": 0.6166, + "step": 36867 + }, + { + "epoch": 8.05, + "learning_rate": 9.779331439807736e-07, + "loss": 0.2016, + "step": 36868 + }, + { + "epoch": 8.05, + "learning_rate": 9.778239021192922e-07, + "loss": 0.9028, + "step": 36869 + }, + { + "epoch": 8.05, + "learning_rate": 9.777146602578108e-07, + "loss": 1.2766, + "step": 36870 + }, + { + "epoch": 8.05, + "learning_rate": 9.776054183963296e-07, + "loss": 0.6375, + "step": 36871 + }, + { + "epoch": 8.05, + "learning_rate": 9.774961765348482e-07, + "loss": 2.0768, + "step": 36872 + }, + { + "epoch": 8.05, + "learning_rate": 9.773869346733668e-07, + "loss": 0.9824, + "step": 36873 + }, + { + "epoch": 8.05, + "learning_rate": 9.772776928118857e-07, + "loss": 0.5241, + "step": 36874 + }, + { + "epoch": 8.05, + "learning_rate": 9.771684509504043e-07, + "loss": 0.2202, + "step": 36875 + }, + { + "epoch": 8.05, + "learning_rate": 9.770592090889229e-07, + "loss": 0.8004, + "step": 36876 + }, + { + "epoch": 8.05, + "learning_rate": 9.769499672274417e-07, + "loss": 0.8463, + "step": 36877 + }, + { + "epoch": 8.05, + "learning_rate": 9.768407253659603e-07, + "loss": 2.0779, + "step": 36878 + }, + { + "epoch": 8.05, + "learning_rate": 9.76731483504479e-07, + "loss": 1.1148, + "step": 36879 + }, + { + "epoch": 8.05, + "learning_rate": 9.766222416429977e-07, + "loss": 1.0798, + "step": 36880 + }, + { + "epoch": 8.05, + "learning_rate": 9.765129997815164e-07, + "loss": 0.6638, + "step": 36881 + }, + { + "epoch": 8.05, + "learning_rate": 9.76403757920035e-07, + "loss": 0.3781, + "step": 36882 + }, + { + "epoch": 8.05, + "learning_rate": 9.762945160585536e-07, + "loss": 1.6649, + "step": 36883 + }, + { + "epoch": 8.05, + "learning_rate": 9.761852741970724e-07, + "loss": 2.1055, + "step": 36884 + }, + { + "epoch": 8.05, + "learning_rate": 9.760760323355912e-07, + "loss": 0.0911, + "step": 36885 + }, + { + "epoch": 8.05, + "learning_rate": 9.759667904741098e-07, + "loss": 0.4826, + "step": 36886 + }, + { + "epoch": 8.05, + "learning_rate": 9.758575486126284e-07, + "loss": 0.8318, + "step": 36887 + }, + { + "epoch": 8.05, + "learning_rate": 9.75748306751147e-07, + "loss": 1.8074, + "step": 36888 + }, + { + "epoch": 8.05, + "learning_rate": 9.756390648896657e-07, + "loss": 1.0607, + "step": 36889 + }, + { + "epoch": 8.05, + "learning_rate": 9.755298230281845e-07, + "loss": 0.6482, + "step": 36890 + }, + { + "epoch": 8.05, + "learning_rate": 9.754205811667033e-07, + "loss": 0.9754, + "step": 36891 + }, + { + "epoch": 8.05, + "learning_rate": 9.75311339305222e-07, + "loss": 0.6258, + "step": 36892 + }, + { + "epoch": 8.05, + "learning_rate": 9.752020974437405e-07, + "loss": 1.2253, + "step": 36893 + }, + { + "epoch": 8.05, + "learning_rate": 9.750928555822591e-07, + "loss": 1.9203, + "step": 36894 + }, + { + "epoch": 8.05, + "learning_rate": 9.749836137207778e-07, + "loss": 0.7196, + "step": 36895 + }, + { + "epoch": 8.05, + "learning_rate": 9.748743718592966e-07, + "loss": 1.6252, + "step": 36896 + }, + { + "epoch": 8.05, + "learning_rate": 9.747651299978154e-07, + "loss": 1.0291, + "step": 36897 + }, + { + "epoch": 8.05, + "learning_rate": 9.74655888136334e-07, + "loss": 2.877, + "step": 36898 + }, + { + "epoch": 8.05, + "learning_rate": 9.745466462748526e-07, + "loss": 2.095, + "step": 36899 + }, + { + "epoch": 8.05, + "learning_rate": 9.744374044133712e-07, + "loss": 1.017, + "step": 36900 + }, + { + "epoch": 8.05, + "learning_rate": 9.743281625518898e-07, + "loss": 0.4962, + "step": 36901 + }, + { + "epoch": 8.05, + "learning_rate": 9.742189206904087e-07, + "loss": 1.2111, + "step": 36902 + }, + { + "epoch": 8.05, + "learning_rate": 9.741096788289273e-07, + "loss": 0.2158, + "step": 36903 + }, + { + "epoch": 8.05, + "learning_rate": 9.74000436967446e-07, + "loss": 1.7289, + "step": 36904 + }, + { + "epoch": 8.05, + "learning_rate": 9.738911951059647e-07, + "loss": 1.238, + "step": 36905 + }, + { + "epoch": 8.05, + "learning_rate": 9.737819532444833e-07, + "loss": 1.4547, + "step": 36906 + }, + { + "epoch": 8.05, + "learning_rate": 9.736727113830021e-07, + "loss": 0.4764, + "step": 36907 + }, + { + "epoch": 8.05, + "learning_rate": 9.735634695215208e-07, + "loss": 0.0887, + "step": 36908 + }, + { + "epoch": 8.06, + "learning_rate": 9.734542276600394e-07, + "loss": 0.1728, + "step": 36909 + }, + { + "epoch": 8.06, + "learning_rate": 9.733449857985582e-07, + "loss": 1.298, + "step": 36910 + }, + { + "epoch": 8.06, + "learning_rate": 9.732357439370768e-07, + "loss": 2.7652, + "step": 36911 + }, + { + "epoch": 8.06, + "learning_rate": 9.731265020755954e-07, + "loss": 2.1614, + "step": 36912 + }, + { + "epoch": 8.06, + "learning_rate": 9.730172602141142e-07, + "loss": 2.3021, + "step": 36913 + }, + { + "epoch": 8.06, + "learning_rate": 9.729080183526328e-07, + "loss": 1.6841, + "step": 36914 + }, + { + "epoch": 8.06, + "learning_rate": 9.727987764911515e-07, + "loss": 1.8619, + "step": 36915 + }, + { + "epoch": 8.06, + "learning_rate": 9.7268953462967e-07, + "loss": 1.2097, + "step": 36916 + }, + { + "epoch": 8.06, + "learning_rate": 9.725802927681889e-07, + "loss": 0.7762, + "step": 36917 + }, + { + "epoch": 8.06, + "learning_rate": 9.724710509067075e-07, + "loss": 1.8359, + "step": 36918 + }, + { + "epoch": 8.06, + "learning_rate": 9.723618090452263e-07, + "loss": 1.1975, + "step": 36919 + }, + { + "epoch": 8.06, + "learning_rate": 9.72252567183745e-07, + "loss": 2.8344, + "step": 36920 + }, + { + "epoch": 8.06, + "learning_rate": 9.721433253222635e-07, + "loss": 1.0238, + "step": 36921 + }, + { + "epoch": 8.06, + "learning_rate": 9.720340834607822e-07, + "loss": 3.4026, + "step": 36922 + }, + { + "epoch": 8.06, + "learning_rate": 9.719248415993008e-07, + "loss": 0.5311, + "step": 36923 + }, + { + "epoch": 8.06, + "learning_rate": 9.718155997378196e-07, + "loss": 0.6143, + "step": 36924 + }, + { + "epoch": 8.06, + "learning_rate": 9.717063578763384e-07, + "loss": 0.9043, + "step": 36925 + }, + { + "epoch": 8.06, + "learning_rate": 9.71597116014857e-07, + "loss": 2.722, + "step": 36926 + }, + { + "epoch": 8.06, + "learning_rate": 9.714878741533756e-07, + "loss": 0.3475, + "step": 36927 + }, + { + "epoch": 8.06, + "learning_rate": 9.713786322918942e-07, + "loss": 0.0722, + "step": 36928 + }, + { + "epoch": 8.06, + "learning_rate": 9.71269390430413e-07, + "loss": 1.7207, + "step": 36929 + }, + { + "epoch": 8.06, + "learning_rate": 9.711601485689317e-07, + "loss": 0.2749, + "step": 36930 + }, + { + "epoch": 8.06, + "learning_rate": 9.710509067074505e-07, + "loss": 0.6429, + "step": 36931 + }, + { + "epoch": 8.06, + "learning_rate": 9.709416648459691e-07, + "loss": 1.1251, + "step": 36932 + }, + { + "epoch": 8.06, + "learning_rate": 9.708324229844877e-07, + "loss": 1.1079, + "step": 36933 + }, + { + "epoch": 8.06, + "learning_rate": 9.707231811230063e-07, + "loss": 3.4175, + "step": 36934 + }, + { + "epoch": 8.06, + "learning_rate": 9.706139392615252e-07, + "loss": 0.9002, + "step": 36935 + }, + { + "epoch": 8.06, + "learning_rate": 9.705046974000438e-07, + "loss": 0.6158, + "step": 36936 + }, + { + "epoch": 8.06, + "learning_rate": 9.703954555385626e-07, + "loss": 1.2398, + "step": 36937 + }, + { + "epoch": 8.06, + "learning_rate": 9.702862136770812e-07, + "loss": 0.5224, + "step": 36938 + }, + { + "epoch": 8.06, + "learning_rate": 9.701769718155998e-07, + "loss": 2.0616, + "step": 36939 + }, + { + "epoch": 8.06, + "learning_rate": 9.700677299541184e-07, + "loss": 0.2187, + "step": 36940 + }, + { + "epoch": 8.06, + "learning_rate": 9.699584880926372e-07, + "loss": 0.3854, + "step": 36941 + }, + { + "epoch": 8.06, + "learning_rate": 9.698492462311559e-07, + "loss": 0.7042, + "step": 36942 + }, + { + "epoch": 8.06, + "learning_rate": 9.697400043696745e-07, + "loss": 1.5974, + "step": 36943 + }, + { + "epoch": 8.06, + "learning_rate": 9.696307625081933e-07, + "loss": 0.4358, + "step": 36944 + }, + { + "epoch": 8.06, + "learning_rate": 9.69521520646712e-07, + "loss": 0.5026, + "step": 36945 + }, + { + "epoch": 8.06, + "learning_rate": 9.694122787852305e-07, + "loss": 0.5058, + "step": 36946 + }, + { + "epoch": 8.06, + "learning_rate": 9.693030369237493e-07, + "loss": 1.39, + "step": 36947 + }, + { + "epoch": 8.06, + "learning_rate": 9.69193795062268e-07, + "loss": 1.0984, + "step": 36948 + }, + { + "epoch": 8.06, + "learning_rate": 9.690845532007866e-07, + "loss": 1.6612, + "step": 36949 + }, + { + "epoch": 8.06, + "learning_rate": 9.689753113393054e-07, + "loss": 1.4415, + "step": 36950 + }, + { + "epoch": 8.06, + "learning_rate": 9.68866069477824e-07, + "loss": 2.0032, + "step": 36951 + }, + { + "epoch": 8.06, + "learning_rate": 9.687568276163426e-07, + "loss": 2.6847, + "step": 36952 + }, + { + "epoch": 8.06, + "learning_rate": 9.686475857548614e-07, + "loss": 0.4375, + "step": 36953 + }, + { + "epoch": 8.07, + "learning_rate": 9.6853834389338e-07, + "loss": 1.5188, + "step": 36954 + }, + { + "epoch": 8.07, + "learning_rate": 9.684291020318986e-07, + "loss": 1.5786, + "step": 36955 + }, + { + "epoch": 8.07, + "learning_rate": 9.683198601704173e-07, + "loss": 1.8136, + "step": 36956 + }, + { + "epoch": 8.07, + "learning_rate": 9.68210618308936e-07, + "loss": 1.1614, + "step": 36957 + }, + { + "epoch": 8.07, + "learning_rate": 9.681013764474547e-07, + "loss": 0.8966, + "step": 36958 + }, + { + "epoch": 8.07, + "learning_rate": 9.679921345859735e-07, + "loss": 1.1777, + "step": 36959 + }, + { + "epoch": 8.07, + "learning_rate": 9.678828927244921e-07, + "loss": 1.8158, + "step": 36960 + }, + { + "epoch": 8.07, + "learning_rate": 9.677736508630107e-07, + "loss": 0.9135, + "step": 36961 + }, + { + "epoch": 8.07, + "learning_rate": 9.676644090015293e-07, + "loss": 1.3953, + "step": 36962 + }, + { + "epoch": 8.07, + "learning_rate": 9.675551671400482e-07, + "loss": 1.0118, + "step": 36963 + }, + { + "epoch": 8.07, + "learning_rate": 9.674459252785668e-07, + "loss": 0.7433, + "step": 36964 + }, + { + "epoch": 8.07, + "learning_rate": 9.673366834170856e-07, + "loss": 0.9041, + "step": 36965 + }, + { + "epoch": 8.07, + "learning_rate": 9.672274415556042e-07, + "loss": 1.0752, + "step": 36966 + }, + { + "epoch": 8.07, + "learning_rate": 9.671181996941228e-07, + "loss": 0.3839, + "step": 36967 + }, + { + "epoch": 8.07, + "learning_rate": 9.670089578326414e-07, + "loss": 2.784, + "step": 36968 + }, + { + "epoch": 8.07, + "learning_rate": 9.668997159711603e-07, + "loss": 0.1193, + "step": 36969 + }, + { + "epoch": 8.07, + "learning_rate": 9.667904741096789e-07, + "loss": 1.168, + "step": 36970 + }, + { + "epoch": 8.07, + "learning_rate": 9.666812322481977e-07, + "loss": 1.1024, + "step": 36971 + }, + { + "epoch": 8.07, + "learning_rate": 9.665719903867163e-07, + "loss": 0.9954, + "step": 36972 + }, + { + "epoch": 8.07, + "learning_rate": 9.66462748525235e-07, + "loss": 0.1139, + "step": 36973 + }, + { + "epoch": 8.07, + "learning_rate": 9.663535066637535e-07, + "loss": 0.6092, + "step": 36974 + }, + { + "epoch": 8.07, + "learning_rate": 9.662442648022723e-07, + "loss": 0.1199, + "step": 36975 + }, + { + "epoch": 8.07, + "learning_rate": 9.66135022940791e-07, + "loss": 0.0998, + "step": 36976 + }, + { + "epoch": 8.07, + "learning_rate": 9.660257810793098e-07, + "loss": 2.0369, + "step": 36977 + }, + { + "epoch": 8.07, + "learning_rate": 9.659165392178284e-07, + "loss": 0.5202, + "step": 36978 + }, + { + "epoch": 8.07, + "learning_rate": 9.65807297356347e-07, + "loss": 1.1309, + "step": 36979 + }, + { + "epoch": 8.07, + "learning_rate": 9.656980554948656e-07, + "loss": 1.5922, + "step": 36980 + }, + { + "epoch": 8.07, + "learning_rate": 9.655888136333844e-07, + "loss": 1.0502, + "step": 36981 + }, + { + "epoch": 8.07, + "learning_rate": 9.65479571771903e-07, + "loss": 1.4241, + "step": 36982 + }, + { + "epoch": 8.07, + "learning_rate": 9.653703299104219e-07, + "loss": 1.1069, + "step": 36983 + }, + { + "epoch": 8.07, + "learning_rate": 9.652610880489405e-07, + "loss": 0.3838, + "step": 36984 + }, + { + "epoch": 8.07, + "learning_rate": 9.65151846187459e-07, + "loss": 2.6274, + "step": 36985 + }, + { + "epoch": 8.07, + "learning_rate": 9.650426043259777e-07, + "loss": 1.2321, + "step": 36986 + }, + { + "epoch": 8.07, + "learning_rate": 9.649333624644965e-07, + "loss": 0.5768, + "step": 36987 + }, + { + "epoch": 8.07, + "learning_rate": 9.648241206030151e-07, + "loss": 3.3408, + "step": 36988 + }, + { + "epoch": 8.07, + "learning_rate": 9.647148787415338e-07, + "loss": 0.6206, + "step": 36989 + }, + { + "epoch": 8.07, + "learning_rate": 9.646056368800526e-07, + "loss": 1.1541, + "step": 36990 + }, + { + "epoch": 8.07, + "learning_rate": 9.644963950185712e-07, + "loss": 2.0706, + "step": 36991 + }, + { + "epoch": 8.07, + "learning_rate": 9.643871531570898e-07, + "loss": 0.756, + "step": 36992 + }, + { + "epoch": 8.07, + "learning_rate": 9.642779112956086e-07, + "loss": 0.2956, + "step": 36993 + }, + { + "epoch": 8.07, + "learning_rate": 9.641686694341272e-07, + "loss": 1.0689, + "step": 36994 + }, + { + "epoch": 8.07, + "learning_rate": 9.640594275726458e-07, + "loss": 1.0245, + "step": 36995 + }, + { + "epoch": 8.07, + "learning_rate": 9.639501857111647e-07, + "loss": 2.7601, + "step": 36996 + }, + { + "epoch": 8.07, + "learning_rate": 9.638409438496833e-07, + "loss": 0.8247, + "step": 36997 + }, + { + "epoch": 8.07, + "learning_rate": 9.637317019882019e-07, + "loss": 0.6634, + "step": 36998 + }, + { + "epoch": 8.07, + "learning_rate": 9.636224601267207e-07, + "loss": 0.9083, + "step": 36999 + }, + { + "epoch": 8.08, + "learning_rate": 9.635132182652393e-07, + "loss": 1.0536, + "step": 37000 + }, + { + "epoch": 8.08, + "learning_rate": 9.63403976403758e-07, + "loss": 1.3012, + "step": 37001 + }, + { + "epoch": 8.08, + "learning_rate": 9.632947345422765e-07, + "loss": 2.2845, + "step": 37002 + }, + { + "epoch": 8.08, + "learning_rate": 9.631854926807954e-07, + "loss": 0.4348, + "step": 37003 + }, + { + "epoch": 8.08, + "learning_rate": 9.63076250819314e-07, + "loss": 3.2175, + "step": 37004 + }, + { + "epoch": 8.08, + "learning_rate": 9.629670089578328e-07, + "loss": 0.8482, + "step": 37005 + }, + { + "epoch": 8.08, + "learning_rate": 9.628577670963514e-07, + "loss": 1.5164, + "step": 37006 + }, + { + "epoch": 8.08, + "learning_rate": 9.6274852523487e-07, + "loss": 0.706, + "step": 37007 + }, + { + "epoch": 8.08, + "learning_rate": 9.626392833733886e-07, + "loss": 0.5197, + "step": 37008 + }, + { + "epoch": 8.08, + "learning_rate": 9.625300415119075e-07, + "loss": 0.1391, + "step": 37009 + }, + { + "epoch": 8.08, + "learning_rate": 9.624207996504263e-07, + "loss": 0.273, + "step": 37010 + }, + { + "epoch": 8.08, + "learning_rate": 9.623115577889449e-07, + "loss": 0.456, + "step": 37011 + }, + { + "epoch": 8.08, + "learning_rate": 9.622023159274635e-07, + "loss": 2.7428, + "step": 37012 + }, + { + "epoch": 8.08, + "learning_rate": 9.620930740659821e-07, + "loss": 0.8827, + "step": 37013 + }, + { + "epoch": 8.08, + "learning_rate": 9.619838322045007e-07, + "loss": 2.0847, + "step": 37014 + }, + { + "epoch": 8.08, + "learning_rate": 9.618745903430195e-07, + "loss": 0.6048, + "step": 37015 + }, + { + "epoch": 8.08, + "learning_rate": 9.617653484815384e-07, + "loss": 0.3659, + "step": 37016 + }, + { + "epoch": 8.08, + "learning_rate": 9.61656106620057e-07, + "loss": 0.8055, + "step": 37017 + }, + { + "epoch": 8.08, + "learning_rate": 9.615468647585756e-07, + "loss": 0.1062, + "step": 37018 + }, + { + "epoch": 8.08, + "learning_rate": 9.614376228970942e-07, + "loss": 0.772, + "step": 37019 + }, + { + "epoch": 8.08, + "learning_rate": 9.613283810356128e-07, + "loss": 0.3371, + "step": 37020 + }, + { + "epoch": 8.08, + "learning_rate": 9.612191391741316e-07, + "loss": 0.5534, + "step": 37021 + }, + { + "epoch": 8.08, + "learning_rate": 9.611098973126502e-07, + "loss": 0.8378, + "step": 37022 + }, + { + "epoch": 8.08, + "learning_rate": 9.61000655451169e-07, + "loss": 1.3087, + "step": 37023 + }, + { + "epoch": 8.08, + "learning_rate": 9.608914135896877e-07, + "loss": 1.9319, + "step": 37024 + }, + { + "epoch": 8.08, + "learning_rate": 9.607821717282063e-07, + "loss": 2.2262, + "step": 37025 + }, + { + "epoch": 8.08, + "learning_rate": 9.60672929866725e-07, + "loss": 1.5096, + "step": 37026 + }, + { + "epoch": 8.08, + "learning_rate": 9.605636880052437e-07, + "loss": 0.858, + "step": 37027 + }, + { + "epoch": 8.08, + "learning_rate": 9.604544461437623e-07, + "loss": 1.5272, + "step": 37028 + }, + { + "epoch": 8.08, + "learning_rate": 9.60345204282281e-07, + "loss": 0.1479, + "step": 37029 + }, + { + "epoch": 8.08, + "learning_rate": 9.602359624207998e-07, + "loss": 0.0722, + "step": 37030 + }, + { + "epoch": 8.08, + "learning_rate": 9.601267205593184e-07, + "loss": 0.5116, + "step": 37031 + }, + { + "epoch": 8.08, + "learning_rate": 9.600174786978372e-07, + "loss": 0.6923, + "step": 37032 + }, + { + "epoch": 8.08, + "learning_rate": 9.599082368363558e-07, + "loss": 0.7514, + "step": 37033 + }, + { + "epoch": 8.08, + "learning_rate": 9.597989949748744e-07, + "loss": 1.2262, + "step": 37034 + }, + { + "epoch": 8.08, + "learning_rate": 9.59689753113393e-07, + "loss": 0.5773, + "step": 37035 + }, + { + "epoch": 8.08, + "learning_rate": 9.595805112519119e-07, + "loss": 1.2542, + "step": 37036 + }, + { + "epoch": 8.08, + "learning_rate": 9.594712693904305e-07, + "loss": 1.0257, + "step": 37037 + }, + { + "epoch": 8.08, + "learning_rate": 9.593620275289493e-07, + "loss": 1.6325, + "step": 37038 + }, + { + "epoch": 8.08, + "learning_rate": 9.59252785667468e-07, + "loss": 1.1124, + "step": 37039 + }, + { + "epoch": 8.08, + "learning_rate": 9.591435438059865e-07, + "loss": 2.6309, + "step": 37040 + }, + { + "epoch": 8.08, + "learning_rate": 9.590343019445051e-07, + "loss": 1.5556, + "step": 37041 + }, + { + "epoch": 8.08, + "learning_rate": 9.589250600830237e-07, + "loss": 1.1949, + "step": 37042 + }, + { + "epoch": 8.08, + "learning_rate": 9.588158182215426e-07, + "loss": 2.3356, + "step": 37043 + }, + { + "epoch": 8.08, + "learning_rate": 9.587065763600614e-07, + "loss": 0.5821, + "step": 37044 + }, + { + "epoch": 8.08, + "learning_rate": 9.5859733449858e-07, + "loss": 0.2416, + "step": 37045 + }, + { + "epoch": 8.09, + "learning_rate": 9.584880926370986e-07, + "loss": 0.9471, + "step": 37046 + }, + { + "epoch": 8.09, + "learning_rate": 9.583788507756172e-07, + "loss": 0.1883, + "step": 37047 + }, + { + "epoch": 8.09, + "learning_rate": 9.582696089141358e-07, + "loss": 1.463, + "step": 37048 + }, + { + "epoch": 8.09, + "learning_rate": 9.581603670526546e-07, + "loss": 1.4026, + "step": 37049 + }, + { + "epoch": 8.09, + "learning_rate": 9.580511251911735e-07, + "loss": 0.8365, + "step": 37050 + }, + { + "epoch": 8.09, + "learning_rate": 9.57941883329692e-07, + "loss": 0.2548, + "step": 37051 + }, + { + "epoch": 8.09, + "learning_rate": 9.578326414682107e-07, + "loss": 2.2048, + "step": 37052 + }, + { + "epoch": 8.09, + "learning_rate": 9.577233996067293e-07, + "loss": 1.1406, + "step": 37053 + }, + { + "epoch": 8.09, + "learning_rate": 9.576141577452481e-07, + "loss": 0.7158, + "step": 37054 + }, + { + "epoch": 8.09, + "learning_rate": 9.575049158837667e-07, + "loss": 2.6559, + "step": 37055 + }, + { + "epoch": 8.09, + "learning_rate": 9.573956740222856e-07, + "loss": 1.9272, + "step": 37056 + }, + { + "epoch": 8.09, + "learning_rate": 9.572864321608042e-07, + "loss": 0.5194, + "step": 37057 + }, + { + "epoch": 8.09, + "learning_rate": 9.571771902993228e-07, + "loss": 1.0248, + "step": 37058 + }, + { + "epoch": 8.09, + "learning_rate": 9.570679484378414e-07, + "loss": 2.0629, + "step": 37059 + }, + { + "epoch": 8.09, + "learning_rate": 9.569587065763602e-07, + "loss": 0.846, + "step": 37060 + }, + { + "epoch": 8.09, + "learning_rate": 9.568494647148788e-07, + "loss": 0.2461, + "step": 37061 + }, + { + "epoch": 8.09, + "learning_rate": 9.567402228533974e-07, + "loss": 1.9107, + "step": 37062 + }, + { + "epoch": 8.09, + "learning_rate": 9.566309809919163e-07, + "loss": 0.5203, + "step": 37063 + }, + { + "epoch": 8.09, + "learning_rate": 9.565217391304349e-07, + "loss": 1.2016, + "step": 37064 + }, + { + "epoch": 8.09, + "learning_rate": 9.564124972689535e-07, + "loss": 1.0952, + "step": 37065 + }, + { + "epoch": 8.09, + "learning_rate": 9.563032554074723e-07, + "loss": 3.3379, + "step": 37066 + }, + { + "epoch": 8.09, + "learning_rate": 9.56194013545991e-07, + "loss": 0.7809, + "step": 37067 + }, + { + "epoch": 8.09, + "learning_rate": 9.560847716845095e-07, + "loss": 0.7313, + "step": 37068 + }, + { + "epoch": 8.09, + "learning_rate": 9.559755298230283e-07, + "loss": 0.7987, + "step": 37069 + }, + { + "epoch": 8.09, + "learning_rate": 9.55866287961547e-07, + "loss": 1.1423, + "step": 37070 + }, + { + "epoch": 8.09, + "learning_rate": 9.557570461000656e-07, + "loss": 1.3159, + "step": 37071 + }, + { + "epoch": 8.09, + "learning_rate": 9.556478042385844e-07, + "loss": 1.2145, + "step": 37072 + }, + { + "epoch": 8.09, + "learning_rate": 9.55538562377103e-07, + "loss": 1.5008, + "step": 37073 + }, + { + "epoch": 8.09, + "learning_rate": 9.554293205156216e-07, + "loss": 2.6022, + "step": 37074 + }, + { + "epoch": 8.09, + "learning_rate": 9.553200786541402e-07, + "loss": 1.1254, + "step": 37075 + }, + { + "epoch": 8.09, + "learning_rate": 9.55210836792659e-07, + "loss": 3.1625, + "step": 37076 + }, + { + "epoch": 8.09, + "learning_rate": 9.551015949311777e-07, + "loss": 1.0861, + "step": 37077 + }, + { + "epoch": 8.09, + "learning_rate": 9.549923530696965e-07, + "loss": 0.6602, + "step": 37078 + }, + { + "epoch": 8.09, + "learning_rate": 9.54883111208215e-07, + "loss": 0.9338, + "step": 37079 + }, + { + "epoch": 8.09, + "learning_rate": 9.547738693467337e-07, + "loss": 1.4494, + "step": 37080 + }, + { + "epoch": 8.09, + "learning_rate": 9.546646274852523e-07, + "loss": 2.0561, + "step": 37081 + }, + { + "epoch": 8.09, + "learning_rate": 9.545553856237711e-07, + "loss": 1.1453, + "step": 37082 + }, + { + "epoch": 8.09, + "learning_rate": 9.544461437622897e-07, + "loss": 2.0602, + "step": 37083 + }, + { + "epoch": 8.09, + "learning_rate": 9.543369019008086e-07, + "loss": 1.1192, + "step": 37084 + }, + { + "epoch": 8.09, + "learning_rate": 9.542276600393272e-07, + "loss": 0.7391, + "step": 37085 + }, + { + "epoch": 8.09, + "learning_rate": 9.541184181778458e-07, + "loss": 1.8396, + "step": 37086 + }, + { + "epoch": 8.09, + "learning_rate": 9.540091763163644e-07, + "loss": 0.5525, + "step": 37087 + }, + { + "epoch": 8.09, + "learning_rate": 9.538999344548832e-07, + "loss": 0.6505, + "step": 37088 + }, + { + "epoch": 8.09, + "learning_rate": 9.537906925934018e-07, + "loss": 1.0482, + "step": 37089 + }, + { + "epoch": 8.09, + "learning_rate": 9.536814507319207e-07, + "loss": 1.4327, + "step": 37090 + }, + { + "epoch": 8.09, + "learning_rate": 9.535722088704393e-07, + "loss": 1.2918, + "step": 37091 + }, + { + "epoch": 8.1, + "learning_rate": 9.534629670089579e-07, + "loss": 0.7378, + "step": 37092 + }, + { + "epoch": 8.1, + "learning_rate": 9.533537251474766e-07, + "loss": 1.2413, + "step": 37093 + }, + { + "epoch": 8.1, + "learning_rate": 9.532444832859952e-07, + "loss": 1.5677, + "step": 37094 + }, + { + "epoch": 8.1, + "learning_rate": 9.531352414245139e-07, + "loss": 0.1348, + "step": 37095 + }, + { + "epoch": 8.1, + "learning_rate": 9.530259995630326e-07, + "loss": 0.5881, + "step": 37096 + }, + { + "epoch": 8.1, + "learning_rate": 9.529167577015514e-07, + "loss": 0.9697, + "step": 37097 + }, + { + "epoch": 8.1, + "learning_rate": 9.5280751584007e-07, + "loss": 0.9215, + "step": 37098 + }, + { + "epoch": 8.1, + "learning_rate": 9.526982739785887e-07, + "loss": 1.9454, + "step": 37099 + }, + { + "epoch": 8.1, + "learning_rate": 9.525890321171073e-07, + "loss": 0.9595, + "step": 37100 + }, + { + "epoch": 8.1, + "learning_rate": 9.52479790255626e-07, + "loss": 1.939, + "step": 37101 + }, + { + "epoch": 8.1, + "learning_rate": 9.523705483941447e-07, + "loss": 1.3793, + "step": 37102 + }, + { + "epoch": 8.1, + "learning_rate": 9.522613065326634e-07, + "loss": 2.3878, + "step": 37103 + }, + { + "epoch": 8.1, + "learning_rate": 9.521520646711821e-07, + "loss": 2.2042, + "step": 37104 + }, + { + "epoch": 8.1, + "learning_rate": 9.520428228097008e-07, + "loss": 0.9258, + "step": 37105 + }, + { + "epoch": 8.1, + "learning_rate": 9.519335809482194e-07, + "loss": 2.0692, + "step": 37106 + }, + { + "epoch": 8.1, + "learning_rate": 9.518243390867381e-07, + "loss": 2.1932, + "step": 37107 + }, + { + "epoch": 8.1, + "learning_rate": 9.517150972252567e-07, + "loss": 0.0583, + "step": 37108 + }, + { + "epoch": 8.1, + "learning_rate": 9.516058553637755e-07, + "loss": 0.6735, + "step": 37109 + }, + { + "epoch": 8.1, + "learning_rate": 9.514966135022942e-07, + "loss": 1.597, + "step": 37110 + }, + { + "epoch": 8.1, + "learning_rate": 9.513873716408129e-07, + "loss": 1.0529, + "step": 37111 + }, + { + "epoch": 8.1, + "learning_rate": 9.512781297793315e-07, + "loss": 0.3166, + "step": 37112 + }, + { + "epoch": 8.1, + "learning_rate": 9.511688879178502e-07, + "loss": 0.9671, + "step": 37113 + }, + { + "epoch": 8.1, + "learning_rate": 9.510596460563688e-07, + "loss": 1.5885, + "step": 37114 + }, + { + "epoch": 8.1, + "learning_rate": 9.509504041948876e-07, + "loss": 2.2647, + "step": 37115 + }, + { + "epoch": 8.1, + "learning_rate": 9.508411623334062e-07, + "loss": 0.7653, + "step": 37116 + }, + { + "epoch": 8.1, + "learning_rate": 9.50731920471925e-07, + "loss": 0.8207, + "step": 37117 + }, + { + "epoch": 8.1, + "learning_rate": 9.506226786104436e-07, + "loss": 0.5196, + "step": 37118 + }, + { + "epoch": 8.1, + "learning_rate": 9.505134367489623e-07, + "loss": 1.2706, + "step": 37119 + }, + { + "epoch": 8.1, + "learning_rate": 9.504041948874809e-07, + "loss": 1.7911, + "step": 37120 + }, + { + "epoch": 8.1, + "learning_rate": 9.502949530259996e-07, + "loss": 1.0327, + "step": 37121 + }, + { + "epoch": 8.1, + "learning_rate": 9.501857111645183e-07, + "loss": 0.4944, + "step": 37122 + }, + { + "epoch": 8.1, + "learning_rate": 9.50076469303037e-07, + "loss": 0.8096, + "step": 37123 + }, + { + "epoch": 8.1, + "learning_rate": 9.499672274415557e-07, + "loss": 1.3773, + "step": 37124 + }, + { + "epoch": 8.1, + "learning_rate": 9.498579855800744e-07, + "loss": 0.5227, + "step": 37125 + }, + { + "epoch": 8.1, + "learning_rate": 9.49748743718593e-07, + "loss": 1.7942, + "step": 37126 + }, + { + "epoch": 8.1, + "learning_rate": 9.496395018571117e-07, + "loss": 1.2851, + "step": 37127 + }, + { + "epoch": 8.1, + "learning_rate": 9.495302599956303e-07, + "loss": 1.3943, + "step": 37128 + }, + { + "epoch": 8.1, + "learning_rate": 9.494210181341491e-07, + "loss": 0.4015, + "step": 37129 + }, + { + "epoch": 8.1, + "learning_rate": 9.493117762726677e-07, + "loss": 1.7483, + "step": 37130 + }, + { + "epoch": 8.1, + "learning_rate": 9.492025344111865e-07, + "loss": 1.4115, + "step": 37131 + }, + { + "epoch": 8.1, + "learning_rate": 9.490932925497051e-07, + "loss": 0.275, + "step": 37132 + }, + { + "epoch": 8.1, + "learning_rate": 9.489840506882238e-07, + "loss": 0.8856, + "step": 37133 + }, + { + "epoch": 8.1, + "learning_rate": 9.488748088267424e-07, + "loss": 1.4855, + "step": 37134 + }, + { + "epoch": 8.1, + "learning_rate": 9.487655669652612e-07, + "loss": 1.991, + "step": 37135 + }, + { + "epoch": 8.1, + "learning_rate": 9.486563251037799e-07, + "loss": 0.5904, + "step": 37136 + }, + { + "epoch": 8.1, + "learning_rate": 9.485470832422986e-07, + "loss": 0.8039, + "step": 37137 + }, + { + "epoch": 8.11, + "learning_rate": 9.484378413808172e-07, + "loss": 2.487, + "step": 37138 + }, + { + "epoch": 8.11, + "learning_rate": 9.483285995193359e-07, + "loss": 0.7971, + "step": 37139 + }, + { + "epoch": 8.11, + "learning_rate": 9.482193576578545e-07, + "loss": 1.4835, + "step": 37140 + }, + { + "epoch": 8.11, + "learning_rate": 9.481101157963732e-07, + "loss": 0.9785, + "step": 37141 + }, + { + "epoch": 8.11, + "learning_rate": 9.48000873934892e-07, + "loss": 0.0853, + "step": 37142 + }, + { + "epoch": 8.11, + "learning_rate": 9.478916320734106e-07, + "loss": 1.2961, + "step": 37143 + }, + { + "epoch": 8.11, + "learning_rate": 9.477823902119293e-07, + "loss": 1.6068, + "step": 37144 + }, + { + "epoch": 8.11, + "learning_rate": 9.47673148350448e-07, + "loss": 2.3703, + "step": 37145 + }, + { + "epoch": 8.11, + "learning_rate": 9.475639064889666e-07, + "loss": 1.8155, + "step": 37146 + }, + { + "epoch": 8.11, + "learning_rate": 9.474546646274853e-07, + "loss": 1.3539, + "step": 37147 + }, + { + "epoch": 8.11, + "learning_rate": 9.473454227660039e-07, + "loss": 0.8998, + "step": 37148 + }, + { + "epoch": 8.11, + "learning_rate": 9.472361809045227e-07, + "loss": 1.2648, + "step": 37149 + }, + { + "epoch": 8.11, + "learning_rate": 9.471269390430414e-07, + "loss": 0.6843, + "step": 37150 + }, + { + "epoch": 8.11, + "learning_rate": 9.470176971815601e-07, + "loss": 0.4889, + "step": 37151 + }, + { + "epoch": 8.11, + "learning_rate": 9.469084553200787e-07, + "loss": 0.178, + "step": 37152 + }, + { + "epoch": 8.11, + "learning_rate": 9.467992134585974e-07, + "loss": 1.5488, + "step": 37153 + }, + { + "epoch": 8.11, + "learning_rate": 9.46689971597116e-07, + "loss": 0.2631, + "step": 37154 + }, + { + "epoch": 8.11, + "learning_rate": 9.465807297356348e-07, + "loss": 0.1391, + "step": 37155 + }, + { + "epoch": 8.11, + "learning_rate": 9.464714878741535e-07, + "loss": 4.3348, + "step": 37156 + }, + { + "epoch": 8.11, + "learning_rate": 9.463622460126721e-07, + "loss": 2.3391, + "step": 37157 + }, + { + "epoch": 8.11, + "learning_rate": 9.462530041511909e-07, + "loss": 1.1552, + "step": 37158 + }, + { + "epoch": 8.11, + "learning_rate": 9.461437622897095e-07, + "loss": 0.1406, + "step": 37159 + }, + { + "epoch": 8.11, + "learning_rate": 9.460345204282281e-07, + "loss": 0.8375, + "step": 37160 + }, + { + "epoch": 8.11, + "learning_rate": 9.459252785667468e-07, + "loss": 1.0032, + "step": 37161 + }, + { + "epoch": 8.11, + "learning_rate": 9.458160367052656e-07, + "loss": 0.5653, + "step": 37162 + }, + { + "epoch": 8.11, + "learning_rate": 9.457067948437842e-07, + "loss": 1.7083, + "step": 37163 + }, + { + "epoch": 8.11, + "learning_rate": 9.45597552982303e-07, + "loss": 0.4774, + "step": 37164 + }, + { + "epoch": 8.11, + "learning_rate": 9.454883111208216e-07, + "loss": 0.074, + "step": 37165 + }, + { + "epoch": 8.11, + "learning_rate": 9.453790692593402e-07, + "loss": 1.8267, + "step": 37166 + }, + { + "epoch": 8.11, + "learning_rate": 9.452698273978589e-07, + "loss": 1.6962, + "step": 37167 + }, + { + "epoch": 8.11, + "learning_rate": 9.451605855363777e-07, + "loss": 0.8771, + "step": 37168 + }, + { + "epoch": 8.11, + "learning_rate": 9.450513436748963e-07, + "loss": 1.1566, + "step": 37169 + }, + { + "epoch": 8.11, + "learning_rate": 9.44942101813415e-07, + "loss": 0.9185, + "step": 37170 + }, + { + "epoch": 8.11, + "learning_rate": 9.448328599519337e-07, + "loss": 0.5984, + "step": 37171 + }, + { + "epoch": 8.11, + "learning_rate": 9.447236180904524e-07, + "loss": 1.7684, + "step": 37172 + }, + { + "epoch": 8.11, + "learning_rate": 9.44614376228971e-07, + "loss": 0.8873, + "step": 37173 + }, + { + "epoch": 8.11, + "learning_rate": 9.445051343674896e-07, + "loss": 1.1556, + "step": 37174 + }, + { + "epoch": 8.11, + "learning_rate": 9.443958925060084e-07, + "loss": 2.378, + "step": 37175 + }, + { + "epoch": 8.11, + "learning_rate": 9.442866506445271e-07, + "loss": 0.8214, + "step": 37176 + }, + { + "epoch": 8.11, + "learning_rate": 9.441774087830457e-07, + "loss": 0.8547, + "step": 37177 + }, + { + "epoch": 8.11, + "learning_rate": 9.440681669215645e-07, + "loss": 0.181, + "step": 37178 + }, + { + "epoch": 8.11, + "learning_rate": 9.439589250600831e-07, + "loss": 1.3345, + "step": 37179 + }, + { + "epoch": 8.11, + "learning_rate": 9.438496831986017e-07, + "loss": 0.6862, + "step": 37180 + }, + { + "epoch": 8.11, + "learning_rate": 9.437404413371204e-07, + "loss": 1.0649, + "step": 37181 + }, + { + "epoch": 8.11, + "learning_rate": 9.436311994756392e-07, + "loss": 0.1058, + "step": 37182 + }, + { + "epoch": 8.12, + "learning_rate": 9.435219576141578e-07, + "loss": 1.4931, + "step": 37183 + }, + { + "epoch": 8.12, + "learning_rate": 9.434127157526766e-07, + "loss": 0.7007, + "step": 37184 + }, + { + "epoch": 8.12, + "learning_rate": 9.433034738911952e-07, + "loss": 2.3212, + "step": 37185 + }, + { + "epoch": 8.12, + "learning_rate": 9.431942320297139e-07, + "loss": 0.2834, + "step": 37186 + }, + { + "epoch": 8.12, + "learning_rate": 9.430849901682325e-07, + "loss": 0.9501, + "step": 37187 + }, + { + "epoch": 8.12, + "learning_rate": 9.429757483067513e-07, + "loss": 0.4603, + "step": 37188 + }, + { + "epoch": 8.12, + "learning_rate": 9.428665064452699e-07, + "loss": 2.039, + "step": 37189 + }, + { + "epoch": 8.12, + "learning_rate": 9.427572645837886e-07, + "loss": 0.7128, + "step": 37190 + }, + { + "epoch": 8.12, + "learning_rate": 9.426480227223073e-07, + "loss": 0.8333, + "step": 37191 + }, + { + "epoch": 8.12, + "learning_rate": 9.42538780860826e-07, + "loss": 0.6936, + "step": 37192 + }, + { + "epoch": 8.12, + "learning_rate": 9.424295389993446e-07, + "loss": 1.5285, + "step": 37193 + }, + { + "epoch": 8.12, + "learning_rate": 9.423202971378633e-07, + "loss": 1.2176, + "step": 37194 + }, + { + "epoch": 8.12, + "learning_rate": 9.42211055276382e-07, + "loss": 1.0933, + "step": 37195 + }, + { + "epoch": 8.12, + "learning_rate": 9.421018134149007e-07, + "loss": 1.6463, + "step": 37196 + }, + { + "epoch": 8.12, + "learning_rate": 9.419925715534193e-07, + "loss": 1.0548, + "step": 37197 + }, + { + "epoch": 8.12, + "learning_rate": 9.418833296919381e-07, + "loss": 1.8997, + "step": 37198 + }, + { + "epoch": 8.12, + "learning_rate": 9.417740878304567e-07, + "loss": 1.651, + "step": 37199 + }, + { + "epoch": 8.12, + "learning_rate": 9.416648459689754e-07, + "loss": 3.0033, + "step": 37200 + }, + { + "epoch": 8.12, + "learning_rate": 9.415556041074941e-07, + "loss": 1.9953, + "step": 37201 + }, + { + "epoch": 8.12, + "learning_rate": 9.414463622460128e-07, + "loss": 1.3262, + "step": 37202 + }, + { + "epoch": 8.12, + "learning_rate": 9.413371203845314e-07, + "loss": 0.8645, + "step": 37203 + }, + { + "epoch": 8.12, + "learning_rate": 9.412278785230501e-07, + "loss": 1.2724, + "step": 37204 + }, + { + "epoch": 8.12, + "learning_rate": 9.411186366615688e-07, + "loss": 1.3379, + "step": 37205 + }, + { + "epoch": 8.12, + "learning_rate": 9.410093948000875e-07, + "loss": 0.3623, + "step": 37206 + }, + { + "epoch": 8.12, + "learning_rate": 9.409001529386061e-07, + "loss": 1.3806, + "step": 37207 + }, + { + "epoch": 8.12, + "learning_rate": 9.407909110771249e-07, + "loss": 2.4268, + "step": 37208 + }, + { + "epoch": 8.12, + "learning_rate": 9.406816692156435e-07, + "loss": 1.4899, + "step": 37209 + }, + { + "epoch": 8.12, + "learning_rate": 9.405724273541622e-07, + "loss": 1.6274, + "step": 37210 + }, + { + "epoch": 8.12, + "learning_rate": 9.404631854926808e-07, + "loss": 0.7581, + "step": 37211 + }, + { + "epoch": 8.12, + "learning_rate": 9.403539436311996e-07, + "loss": 1.2868, + "step": 37212 + }, + { + "epoch": 8.12, + "learning_rate": 9.402447017697182e-07, + "loss": 3.1423, + "step": 37213 + }, + { + "epoch": 8.12, + "learning_rate": 9.401354599082369e-07, + "loss": 1.4892, + "step": 37214 + }, + { + "epoch": 8.12, + "learning_rate": 9.400262180467556e-07, + "loss": 1.0262, + "step": 37215 + }, + { + "epoch": 8.12, + "learning_rate": 9.399169761852743e-07, + "loss": 1.7171, + "step": 37216 + }, + { + "epoch": 8.12, + "learning_rate": 9.398077343237929e-07, + "loss": 1.4602, + "step": 37217 + }, + { + "epoch": 8.12, + "learning_rate": 9.396984924623117e-07, + "loss": 1.2257, + "step": 37218 + }, + { + "epoch": 8.12, + "learning_rate": 9.395892506008303e-07, + "loss": 0.4816, + "step": 37219 + }, + { + "epoch": 8.12, + "learning_rate": 9.39480008739349e-07, + "loss": 0.2804, + "step": 37220 + }, + { + "epoch": 8.12, + "learning_rate": 9.393707668778677e-07, + "loss": 1.177, + "step": 37221 + }, + { + "epoch": 8.12, + "learning_rate": 9.392615250163864e-07, + "loss": 1.0623, + "step": 37222 + }, + { + "epoch": 8.12, + "learning_rate": 9.39152283154905e-07, + "loss": 0.714, + "step": 37223 + }, + { + "epoch": 8.12, + "learning_rate": 9.390430412934237e-07, + "loss": 0.8271, + "step": 37224 + }, + { + "epoch": 8.12, + "learning_rate": 9.389337994319424e-07, + "loss": 2.7074, + "step": 37225 + }, + { + "epoch": 8.12, + "learning_rate": 9.388245575704611e-07, + "loss": 0.6437, + "step": 37226 + }, + { + "epoch": 8.12, + "learning_rate": 9.387153157089797e-07, + "loss": 2.4444, + "step": 37227 + }, + { + "epoch": 8.12, + "learning_rate": 9.386060738474985e-07, + "loss": 1.6224, + "step": 37228 + }, + { + "epoch": 8.13, + "learning_rate": 9.384968319860171e-07, + "loss": 0.1058, + "step": 37229 + }, + { + "epoch": 8.13, + "learning_rate": 9.383875901245358e-07, + "loss": 0.7096, + "step": 37230 + }, + { + "epoch": 8.13, + "learning_rate": 9.382783482630544e-07, + "loss": 0.8352, + "step": 37231 + }, + { + "epoch": 8.13, + "learning_rate": 9.381691064015732e-07, + "loss": 1.868, + "step": 37232 + }, + { + "epoch": 8.13, + "learning_rate": 9.380598645400918e-07, + "loss": 1.3338, + "step": 37233 + }, + { + "epoch": 8.13, + "learning_rate": 9.379506226786106e-07, + "loss": 2.8444, + "step": 37234 + }, + { + "epoch": 8.13, + "learning_rate": 9.378413808171292e-07, + "loss": 1.9086, + "step": 37235 + }, + { + "epoch": 8.13, + "learning_rate": 9.377321389556479e-07, + "loss": 1.1897, + "step": 37236 + }, + { + "epoch": 8.13, + "learning_rate": 9.376228970941665e-07, + "loss": 0.9859, + "step": 37237 + }, + { + "epoch": 8.13, + "learning_rate": 9.375136552326853e-07, + "loss": 1.1225, + "step": 37238 + }, + { + "epoch": 8.13, + "learning_rate": 9.374044133712039e-07, + "loss": 0.8465, + "step": 37239 + }, + { + "epoch": 8.13, + "learning_rate": 9.372951715097226e-07, + "loss": 0.7084, + "step": 37240 + }, + { + "epoch": 8.13, + "learning_rate": 9.371859296482413e-07, + "loss": 2.1683, + "step": 37241 + }, + { + "epoch": 8.13, + "learning_rate": 9.3707668778676e-07, + "loss": 1.7825, + "step": 37242 + }, + { + "epoch": 8.13, + "learning_rate": 9.369674459252786e-07, + "loss": 0.4562, + "step": 37243 + }, + { + "epoch": 8.13, + "learning_rate": 9.368582040637973e-07, + "loss": 0.5769, + "step": 37244 + }, + { + "epoch": 8.13, + "learning_rate": 9.36748962202316e-07, + "loss": 0.6468, + "step": 37245 + }, + { + "epoch": 8.13, + "learning_rate": 9.366397203408347e-07, + "loss": 1.0606, + "step": 37246 + }, + { + "epoch": 8.13, + "learning_rate": 9.365304784793533e-07, + "loss": 0.1968, + "step": 37247 + }, + { + "epoch": 8.13, + "learning_rate": 9.364212366178721e-07, + "loss": 2.2245, + "step": 37248 + }, + { + "epoch": 8.13, + "learning_rate": 9.363119947563907e-07, + "loss": 3.8662, + "step": 37249 + }, + { + "epoch": 8.13, + "learning_rate": 9.362027528949094e-07, + "loss": 1.6696, + "step": 37250 + }, + { + "epoch": 8.13, + "learning_rate": 9.36093511033428e-07, + "loss": 0.3512, + "step": 37251 + }, + { + "epoch": 8.13, + "learning_rate": 9.359842691719468e-07, + "loss": 0.8111, + "step": 37252 + }, + { + "epoch": 8.13, + "learning_rate": 9.358750273104654e-07, + "loss": 0.7285, + "step": 37253 + }, + { + "epoch": 8.13, + "learning_rate": 9.357657854489842e-07, + "loss": 0.8393, + "step": 37254 + }, + { + "epoch": 8.13, + "learning_rate": 9.356565435875028e-07, + "loss": 0.6261, + "step": 37255 + }, + { + "epoch": 8.13, + "learning_rate": 9.355473017260215e-07, + "loss": 2.1236, + "step": 37256 + }, + { + "epoch": 8.13, + "learning_rate": 9.354380598645401e-07, + "loss": 1.0351, + "step": 37257 + }, + { + "epoch": 8.13, + "learning_rate": 9.353288180030588e-07, + "loss": 0.9074, + "step": 37258 + }, + { + "epoch": 8.13, + "learning_rate": 9.352195761415775e-07, + "loss": 1.0218, + "step": 37259 + }, + { + "epoch": 8.13, + "learning_rate": 9.351103342800962e-07, + "loss": 1.668, + "step": 37260 + }, + { + "epoch": 8.13, + "learning_rate": 9.35001092418615e-07, + "loss": 1.0124, + "step": 37261 + }, + { + "epoch": 8.13, + "learning_rate": 9.348918505571336e-07, + "loss": 0.8371, + "step": 37262 + }, + { + "epoch": 8.13, + "learning_rate": 9.347826086956522e-07, + "loss": 2.2201, + "step": 37263 + }, + { + "epoch": 8.13, + "learning_rate": 9.346733668341709e-07, + "loss": 2.5107, + "step": 37264 + }, + { + "epoch": 8.13, + "learning_rate": 9.345641249726895e-07, + "loss": 0.808, + "step": 37265 + }, + { + "epoch": 8.13, + "learning_rate": 9.344548831112083e-07, + "loss": 0.4757, + "step": 37266 + }, + { + "epoch": 8.13, + "learning_rate": 9.343456412497269e-07, + "loss": 0.8196, + "step": 37267 + }, + { + "epoch": 8.13, + "learning_rate": 9.342363993882457e-07, + "loss": 2.2797, + "step": 37268 + }, + { + "epoch": 8.13, + "learning_rate": 9.341271575267643e-07, + "loss": 1.0361, + "step": 37269 + }, + { + "epoch": 8.13, + "learning_rate": 9.34017915665283e-07, + "loss": 0.0903, + "step": 37270 + }, + { + "epoch": 8.13, + "learning_rate": 9.339086738038016e-07, + "loss": 1.1746, + "step": 37271 + }, + { + "epoch": 8.13, + "learning_rate": 9.337994319423204e-07, + "loss": 0.4959, + "step": 37272 + }, + { + "epoch": 8.13, + "learning_rate": 9.33690190080839e-07, + "loss": 0.7533, + "step": 37273 + }, + { + "epoch": 8.13, + "learning_rate": 9.335809482193578e-07, + "loss": 0.8452, + "step": 37274 + }, + { + "epoch": 8.14, + "learning_rate": 9.334717063578765e-07, + "loss": 1.6134, + "step": 37275 + }, + { + "epoch": 8.14, + "learning_rate": 9.333624644963951e-07, + "loss": 0.9151, + "step": 37276 + }, + { + "epoch": 8.14, + "learning_rate": 9.332532226349137e-07, + "loss": 1.5057, + "step": 37277 + }, + { + "epoch": 8.14, + "learning_rate": 9.331439807734324e-07, + "loss": 1.6719, + "step": 37278 + }, + { + "epoch": 8.14, + "learning_rate": 9.330347389119511e-07, + "loss": 1.0911, + "step": 37279 + }, + { + "epoch": 8.14, + "learning_rate": 9.329254970504698e-07, + "loss": 0.8521, + "step": 37280 + }, + { + "epoch": 8.14, + "learning_rate": 9.328162551889886e-07, + "loss": 0.8427, + "step": 37281 + }, + { + "epoch": 8.14, + "learning_rate": 9.327070133275072e-07, + "loss": 0.5739, + "step": 37282 + }, + { + "epoch": 8.14, + "learning_rate": 9.325977714660259e-07, + "loss": 1.1716, + "step": 37283 + }, + { + "epoch": 8.14, + "learning_rate": 9.324885296045445e-07, + "loss": 1.3501, + "step": 37284 + }, + { + "epoch": 8.14, + "learning_rate": 9.323792877430631e-07, + "loss": 1.4065, + "step": 37285 + }, + { + "epoch": 8.14, + "learning_rate": 9.322700458815819e-07, + "loss": 1.083, + "step": 37286 + }, + { + "epoch": 8.14, + "learning_rate": 9.321608040201007e-07, + "loss": 0.8247, + "step": 37287 + }, + { + "epoch": 8.14, + "learning_rate": 9.320515621586193e-07, + "loss": 1.3829, + "step": 37288 + }, + { + "epoch": 8.14, + "learning_rate": 9.31942320297138e-07, + "loss": 0.4263, + "step": 37289 + }, + { + "epoch": 8.14, + "learning_rate": 9.318330784356566e-07, + "loss": 0.3582, + "step": 37290 + }, + { + "epoch": 8.14, + "learning_rate": 9.317238365741752e-07, + "loss": 1.0939, + "step": 37291 + }, + { + "epoch": 8.14, + "learning_rate": 9.31614594712694e-07, + "loss": 2.6454, + "step": 37292 + }, + { + "epoch": 8.14, + "learning_rate": 9.315053528512126e-07, + "loss": 1.4091, + "step": 37293 + }, + { + "epoch": 8.14, + "learning_rate": 9.313961109897314e-07, + "loss": 1.1448, + "step": 37294 + }, + { + "epoch": 8.14, + "learning_rate": 9.312868691282501e-07, + "loss": 0.5294, + "step": 37295 + }, + { + "epoch": 8.14, + "learning_rate": 9.311776272667687e-07, + "loss": 1.2055, + "step": 37296 + }, + { + "epoch": 8.14, + "learning_rate": 9.310683854052874e-07, + "loss": 0.7674, + "step": 37297 + }, + { + "epoch": 8.14, + "learning_rate": 9.30959143543806e-07, + "loss": 0.2143, + "step": 37298 + }, + { + "epoch": 8.14, + "learning_rate": 9.308499016823246e-07, + "loss": 0.8249, + "step": 37299 + }, + { + "epoch": 8.14, + "learning_rate": 9.307406598208434e-07, + "loss": 0.9884, + "step": 37300 + }, + { + "epoch": 8.14, + "learning_rate": 9.306314179593622e-07, + "loss": 3.4877, + "step": 37301 + }, + { + "epoch": 8.14, + "learning_rate": 9.305221760978808e-07, + "loss": 0.2386, + "step": 37302 + }, + { + "epoch": 8.14, + "learning_rate": 9.304129342363995e-07, + "loss": 1.3328, + "step": 37303 + }, + { + "epoch": 8.14, + "learning_rate": 9.303036923749181e-07, + "loss": 0.1935, + "step": 37304 + }, + { + "epoch": 8.14, + "learning_rate": 9.301944505134367e-07, + "loss": 0.7673, + "step": 37305 + }, + { + "epoch": 8.14, + "learning_rate": 9.300852086519555e-07, + "loss": 0.5266, + "step": 37306 + }, + { + "epoch": 8.14, + "learning_rate": 9.299759667904743e-07, + "loss": 1.7187, + "step": 37307 + }, + { + "epoch": 8.14, + "learning_rate": 9.298667249289929e-07, + "loss": 0.9309, + "step": 37308 + }, + { + "epoch": 8.14, + "learning_rate": 9.297574830675116e-07, + "loss": 1.3492, + "step": 37309 + }, + { + "epoch": 8.14, + "learning_rate": 9.296482412060302e-07, + "loss": 0.7172, + "step": 37310 + }, + { + "epoch": 8.14, + "learning_rate": 9.295389993445489e-07, + "loss": 1.7838, + "step": 37311 + }, + { + "epoch": 8.14, + "learning_rate": 9.294297574830675e-07, + "loss": 0.6114, + "step": 37312 + }, + { + "epoch": 8.14, + "learning_rate": 9.293205156215862e-07, + "loss": 2.3665, + "step": 37313 + }, + { + "epoch": 8.14, + "learning_rate": 9.29211273760105e-07, + "loss": 1.4451, + "step": 37314 + }, + { + "epoch": 8.14, + "learning_rate": 9.291020318986237e-07, + "loss": 0.5785, + "step": 37315 + }, + { + "epoch": 8.14, + "learning_rate": 9.289927900371423e-07, + "loss": 1.4946, + "step": 37316 + }, + { + "epoch": 8.14, + "learning_rate": 9.28883548175661e-07, + "loss": 2.2607, + "step": 37317 + }, + { + "epoch": 8.14, + "learning_rate": 9.287743063141796e-07, + "loss": 0.1356, + "step": 37318 + }, + { + "epoch": 8.14, + "learning_rate": 9.286650644526984e-07, + "loss": 0.176, + "step": 37319 + }, + { + "epoch": 8.14, + "learning_rate": 9.285558225912171e-07, + "loss": 1.2925, + "step": 37320 + }, + { + "epoch": 8.15, + "learning_rate": 9.284465807297358e-07, + "loss": 0.2051, + "step": 37321 + }, + { + "epoch": 8.15, + "learning_rate": 9.283373388682544e-07, + "loss": 1.4166, + "step": 37322 + }, + { + "epoch": 8.15, + "learning_rate": 9.282280970067731e-07, + "loss": 0.8039, + "step": 37323 + }, + { + "epoch": 8.15, + "learning_rate": 9.281188551452917e-07, + "loss": 0.7335, + "step": 37324 + }, + { + "epoch": 8.15, + "learning_rate": 9.280096132838104e-07, + "loss": 0.6867, + "step": 37325 + }, + { + "epoch": 8.15, + "learning_rate": 9.27900371422329e-07, + "loss": 1.3888, + "step": 37326 + }, + { + "epoch": 8.15, + "learning_rate": 9.277911295608479e-07, + "loss": 1.3643, + "step": 37327 + }, + { + "epoch": 8.15, + "learning_rate": 9.276818876993665e-07, + "loss": 0.0939, + "step": 37328 + }, + { + "epoch": 8.15, + "learning_rate": 9.275726458378852e-07, + "loss": 0.9316, + "step": 37329 + }, + { + "epoch": 8.15, + "learning_rate": 9.274634039764038e-07, + "loss": 1.2453, + "step": 37330 + }, + { + "epoch": 8.15, + "learning_rate": 9.273541621149225e-07, + "loss": 2.5485, + "step": 37331 + }, + { + "epoch": 8.15, + "learning_rate": 9.272449202534411e-07, + "loss": 0.3534, + "step": 37332 + }, + { + "epoch": 8.15, + "learning_rate": 9.271356783919599e-07, + "loss": 1.3211, + "step": 37333 + }, + { + "epoch": 8.15, + "learning_rate": 9.270264365304786e-07, + "loss": 1.9239, + "step": 37334 + }, + { + "epoch": 8.15, + "learning_rate": 9.269171946689973e-07, + "loss": 0.0703, + "step": 37335 + }, + { + "epoch": 8.15, + "learning_rate": 9.268079528075159e-07, + "loss": 0.8891, + "step": 37336 + }, + { + "epoch": 8.15, + "learning_rate": 9.266987109460346e-07, + "loss": 0.7923, + "step": 37337 + }, + { + "epoch": 8.15, + "learning_rate": 9.265894690845532e-07, + "loss": 1.5777, + "step": 37338 + }, + { + "epoch": 8.15, + "learning_rate": 9.26480227223072e-07, + "loss": 1.4843, + "step": 37339 + }, + { + "epoch": 8.15, + "learning_rate": 9.263709853615907e-07, + "loss": 0.8815, + "step": 37340 + }, + { + "epoch": 8.15, + "learning_rate": 9.262617435001094e-07, + "loss": 1.2903, + "step": 37341 + }, + { + "epoch": 8.15, + "learning_rate": 9.26152501638628e-07, + "loss": 0.9193, + "step": 37342 + }, + { + "epoch": 8.15, + "learning_rate": 9.260432597771467e-07, + "loss": 0.2747, + "step": 37343 + }, + { + "epoch": 8.15, + "learning_rate": 9.259340179156653e-07, + "loss": 1.0214, + "step": 37344 + }, + { + "epoch": 8.15, + "learning_rate": 9.25824776054184e-07, + "loss": 0.9946, + "step": 37345 + }, + { + "epoch": 8.15, + "learning_rate": 9.257155341927026e-07, + "loss": 1.3035, + "step": 37346 + }, + { + "epoch": 8.15, + "learning_rate": 9.256062923312215e-07, + "loss": 0.2714, + "step": 37347 + }, + { + "epoch": 8.15, + "learning_rate": 9.254970504697401e-07, + "loss": 2.0404, + "step": 37348 + }, + { + "epoch": 8.15, + "learning_rate": 9.253878086082588e-07, + "loss": 0.8757, + "step": 37349 + }, + { + "epoch": 8.15, + "learning_rate": 9.252785667467774e-07, + "loss": 0.7034, + "step": 37350 + }, + { + "epoch": 8.15, + "learning_rate": 9.251693248852961e-07, + "loss": 0.2713, + "step": 37351 + }, + { + "epoch": 8.15, + "learning_rate": 9.250600830238147e-07, + "loss": 1.5567, + "step": 37352 + }, + { + "epoch": 8.15, + "learning_rate": 9.249508411623335e-07, + "loss": 1.6316, + "step": 37353 + }, + { + "epoch": 8.15, + "learning_rate": 9.248415993008522e-07, + "loss": 0.7698, + "step": 37354 + }, + { + "epoch": 8.15, + "learning_rate": 9.247323574393709e-07, + "loss": 2.2436, + "step": 37355 + }, + { + "epoch": 8.15, + "learning_rate": 9.246231155778895e-07, + "loss": 1.4751, + "step": 37356 + }, + { + "epoch": 8.15, + "learning_rate": 9.245138737164082e-07, + "loss": 2.2791, + "step": 37357 + }, + { + "epoch": 8.15, + "learning_rate": 9.244046318549268e-07, + "loss": 0.108, + "step": 37358 + }, + { + "epoch": 8.15, + "learning_rate": 9.242953899934455e-07, + "loss": 1.1715, + "step": 37359 + }, + { + "epoch": 8.15, + "learning_rate": 9.241861481319643e-07, + "loss": 1.3351, + "step": 37360 + }, + { + "epoch": 8.15, + "learning_rate": 9.24076906270483e-07, + "loss": 0.914, + "step": 37361 + }, + { + "epoch": 8.15, + "learning_rate": 9.239676644090016e-07, + "loss": 0.9484, + "step": 37362 + }, + { + "epoch": 8.15, + "learning_rate": 9.238584225475203e-07, + "loss": 0.6058, + "step": 37363 + }, + { + "epoch": 8.15, + "learning_rate": 9.237491806860389e-07, + "loss": 1.2662, + "step": 37364 + }, + { + "epoch": 8.15, + "learning_rate": 9.236399388245576e-07, + "loss": 0.795, + "step": 37365 + }, + { + "epoch": 8.15, + "learning_rate": 9.235306969630762e-07, + "loss": 0.1074, + "step": 37366 + }, + { + "epoch": 8.16, + "learning_rate": 9.234214551015951e-07, + "loss": 0.9645, + "step": 37367 + }, + { + "epoch": 8.16, + "learning_rate": 9.233122132401137e-07, + "loss": 0.9594, + "step": 37368 + }, + { + "epoch": 8.16, + "learning_rate": 9.232029713786324e-07, + "loss": 1.2976, + "step": 37369 + }, + { + "epoch": 8.16, + "learning_rate": 9.23093729517151e-07, + "loss": 0.5012, + "step": 37370 + }, + { + "epoch": 8.16, + "learning_rate": 9.229844876556697e-07, + "loss": 1.3933, + "step": 37371 + }, + { + "epoch": 8.16, + "learning_rate": 9.228752457941883e-07, + "loss": 1.9346, + "step": 37372 + }, + { + "epoch": 8.16, + "learning_rate": 9.227660039327072e-07, + "loss": 1.171, + "step": 37373 + }, + { + "epoch": 8.16, + "learning_rate": 9.226567620712258e-07, + "loss": 0.8264, + "step": 37374 + }, + { + "epoch": 8.16, + "learning_rate": 9.225475202097445e-07, + "loss": 1.9736, + "step": 37375 + }, + { + "epoch": 8.16, + "learning_rate": 9.224382783482631e-07, + "loss": 0.4379, + "step": 37376 + }, + { + "epoch": 8.16, + "learning_rate": 9.223290364867818e-07, + "loss": 1.423, + "step": 37377 + }, + { + "epoch": 8.16, + "learning_rate": 9.222197946253004e-07, + "loss": 2.2087, + "step": 37378 + }, + { + "epoch": 8.16, + "learning_rate": 9.221105527638191e-07, + "loss": 2.3793, + "step": 37379 + }, + { + "epoch": 8.16, + "learning_rate": 9.220013109023379e-07, + "loss": 0.561, + "step": 37380 + }, + { + "epoch": 8.16, + "learning_rate": 9.218920690408566e-07, + "loss": 0.7428, + "step": 37381 + }, + { + "epoch": 8.16, + "learning_rate": 9.217828271793752e-07, + "loss": 1.3789, + "step": 37382 + }, + { + "epoch": 8.16, + "learning_rate": 9.216735853178939e-07, + "loss": 1.3188, + "step": 37383 + }, + { + "epoch": 8.16, + "learning_rate": 9.215643434564125e-07, + "loss": 1.8681, + "step": 37384 + }, + { + "epoch": 8.16, + "learning_rate": 9.214551015949312e-07, + "loss": 0.8691, + "step": 37385 + }, + { + "epoch": 8.16, + "learning_rate": 9.213458597334498e-07, + "loss": 1.2339, + "step": 37386 + }, + { + "epoch": 8.16, + "learning_rate": 9.212366178719687e-07, + "loss": 1.3744, + "step": 37387 + }, + { + "epoch": 8.16, + "learning_rate": 9.211273760104873e-07, + "loss": 0.1924, + "step": 37388 + }, + { + "epoch": 8.16, + "learning_rate": 9.21018134149006e-07, + "loss": 1.0813, + "step": 37389 + }, + { + "epoch": 8.16, + "learning_rate": 9.209088922875246e-07, + "loss": 0.1686, + "step": 37390 + }, + { + "epoch": 8.16, + "learning_rate": 9.207996504260433e-07, + "loss": 0.6556, + "step": 37391 + }, + { + "epoch": 8.16, + "learning_rate": 9.206904085645619e-07, + "loss": 2.1574, + "step": 37392 + }, + { + "epoch": 8.16, + "learning_rate": 9.205811667030808e-07, + "loss": 0.8272, + "step": 37393 + }, + { + "epoch": 8.16, + "learning_rate": 9.204719248415995e-07, + "loss": 0.6613, + "step": 37394 + }, + { + "epoch": 8.16, + "learning_rate": 9.203626829801181e-07, + "loss": 0.5784, + "step": 37395 + }, + { + "epoch": 8.16, + "learning_rate": 9.202534411186367e-07, + "loss": 0.1097, + "step": 37396 + }, + { + "epoch": 8.16, + "learning_rate": 9.201441992571554e-07, + "loss": 2.0752, + "step": 37397 + }, + { + "epoch": 8.16, + "learning_rate": 9.20034957395674e-07, + "loss": 2.2838, + "step": 37398 + }, + { + "epoch": 8.16, + "learning_rate": 9.199257155341927e-07, + "loss": 0.9977, + "step": 37399 + }, + { + "epoch": 8.16, + "learning_rate": 9.198164736727116e-07, + "loss": 1.1959, + "step": 37400 + }, + { + "epoch": 8.16, + "learning_rate": 9.197072318112302e-07, + "loss": 0.6095, + "step": 37401 + }, + { + "epoch": 8.16, + "learning_rate": 9.195979899497488e-07, + "loss": 0.123, + "step": 37402 + }, + { + "epoch": 8.16, + "learning_rate": 9.194887480882675e-07, + "loss": 0.5941, + "step": 37403 + }, + { + "epoch": 8.16, + "learning_rate": 9.193795062267861e-07, + "loss": 1.1444, + "step": 37404 + }, + { + "epoch": 8.16, + "learning_rate": 9.192702643653048e-07, + "loss": 0.9546, + "step": 37405 + }, + { + "epoch": 8.16, + "learning_rate": 9.191610225038236e-07, + "loss": 2.212, + "step": 37406 + }, + { + "epoch": 8.16, + "learning_rate": 9.190517806423423e-07, + "loss": 0.5189, + "step": 37407 + }, + { + "epoch": 8.16, + "learning_rate": 9.18942538780861e-07, + "loss": 1.0172, + "step": 37408 + }, + { + "epoch": 8.16, + "learning_rate": 9.188332969193796e-07, + "loss": 2.609, + "step": 37409 + }, + { + "epoch": 8.16, + "learning_rate": 9.187240550578982e-07, + "loss": 1.2258, + "step": 37410 + }, + { + "epoch": 8.16, + "learning_rate": 9.186148131964169e-07, + "loss": 1.667, + "step": 37411 + }, + { + "epoch": 8.16, + "learning_rate": 9.185055713349355e-07, + "loss": 0.8612, + "step": 37412 + }, + { + "epoch": 8.17, + "learning_rate": 9.183963294734543e-07, + "loss": 0.5686, + "step": 37413 + }, + { + "epoch": 8.17, + "learning_rate": 9.182870876119731e-07, + "loss": 1.9097, + "step": 37414 + }, + { + "epoch": 8.17, + "learning_rate": 9.181778457504917e-07, + "loss": 1.1524, + "step": 37415 + }, + { + "epoch": 8.17, + "learning_rate": 9.180686038890103e-07, + "loss": 2.4999, + "step": 37416 + }, + { + "epoch": 8.17, + "learning_rate": 9.17959362027529e-07, + "loss": 2.6641, + "step": 37417 + }, + { + "epoch": 8.17, + "learning_rate": 9.178501201660476e-07, + "loss": 1.7015, + "step": 37418 + }, + { + "epoch": 8.17, + "learning_rate": 9.177408783045663e-07, + "loss": 2.6395, + "step": 37419 + }, + { + "epoch": 8.17, + "learning_rate": 9.176316364430852e-07, + "loss": 0.8997, + "step": 37420 + }, + { + "epoch": 8.17, + "learning_rate": 9.175223945816038e-07, + "loss": 1.1378, + "step": 37421 + }, + { + "epoch": 8.17, + "learning_rate": 9.174131527201225e-07, + "loss": 1.4972, + "step": 37422 + }, + { + "epoch": 8.17, + "learning_rate": 9.173039108586411e-07, + "loss": 1.4366, + "step": 37423 + }, + { + "epoch": 8.17, + "learning_rate": 9.171946689971597e-07, + "loss": 0.5416, + "step": 37424 + }, + { + "epoch": 8.17, + "learning_rate": 9.170854271356784e-07, + "loss": 0.1498, + "step": 37425 + }, + { + "epoch": 8.17, + "learning_rate": 9.169761852741972e-07, + "loss": 1.009, + "step": 37426 + }, + { + "epoch": 8.17, + "learning_rate": 9.168669434127159e-07, + "loss": 1.3756, + "step": 37427 + }, + { + "epoch": 8.17, + "learning_rate": 9.167577015512346e-07, + "loss": 3.4237, + "step": 37428 + }, + { + "epoch": 8.17, + "learning_rate": 9.166484596897532e-07, + "loss": 1.2345, + "step": 37429 + }, + { + "epoch": 8.17, + "learning_rate": 9.165392178282719e-07, + "loss": 0.6494, + "step": 37430 + }, + { + "epoch": 8.17, + "learning_rate": 9.164299759667905e-07, + "loss": 1.3227, + "step": 37431 + }, + { + "epoch": 8.17, + "learning_rate": 9.163207341053091e-07, + "loss": 0.0809, + "step": 37432 + }, + { + "epoch": 8.17, + "learning_rate": 9.162114922438279e-07, + "loss": 1.2851, + "step": 37433 + }, + { + "epoch": 8.17, + "learning_rate": 9.161022503823467e-07, + "loss": 0.1821, + "step": 37434 + }, + { + "epoch": 8.17, + "learning_rate": 9.159930085208653e-07, + "loss": 0.7045, + "step": 37435 + }, + { + "epoch": 8.17, + "learning_rate": 9.15883766659384e-07, + "loss": 1.256, + "step": 37436 + }, + { + "epoch": 8.17, + "learning_rate": 9.157745247979026e-07, + "loss": 0.7391, + "step": 37437 + }, + { + "epoch": 8.17, + "learning_rate": 9.156652829364212e-07, + "loss": 0.7204, + "step": 37438 + }, + { + "epoch": 8.17, + "learning_rate": 9.1555604107494e-07, + "loss": 0.9848, + "step": 37439 + }, + { + "epoch": 8.17, + "learning_rate": 9.154467992134588e-07, + "loss": 0.6995, + "step": 37440 + }, + { + "epoch": 8.17, + "learning_rate": 9.153375573519774e-07, + "loss": 0.0972, + "step": 37441 + }, + { + "epoch": 8.17, + "learning_rate": 9.152283154904961e-07, + "loss": 0.8443, + "step": 37442 + }, + { + "epoch": 8.17, + "learning_rate": 9.151190736290147e-07, + "loss": 1.2092, + "step": 37443 + }, + { + "epoch": 8.17, + "learning_rate": 9.150098317675334e-07, + "loss": 1.0712, + "step": 37444 + }, + { + "epoch": 8.17, + "learning_rate": 9.14900589906052e-07, + "loss": 0.4825, + "step": 37445 + }, + { + "epoch": 8.17, + "learning_rate": 9.147913480445708e-07, + "loss": 1.5218, + "step": 37446 + }, + { + "epoch": 8.17, + "learning_rate": 9.146821061830895e-07, + "loss": 2.0632, + "step": 37447 + }, + { + "epoch": 8.17, + "learning_rate": 9.145728643216082e-07, + "loss": 1.1324, + "step": 37448 + }, + { + "epoch": 8.17, + "learning_rate": 9.144636224601268e-07, + "loss": 0.3008, + "step": 37449 + }, + { + "epoch": 8.17, + "learning_rate": 9.143543805986455e-07, + "loss": 1.6488, + "step": 37450 + }, + { + "epoch": 8.17, + "learning_rate": 9.142451387371641e-07, + "loss": 0.9772, + "step": 37451 + }, + { + "epoch": 8.17, + "learning_rate": 9.141358968756827e-07, + "loss": 0.77, + "step": 37452 + }, + { + "epoch": 8.17, + "learning_rate": 9.140266550142015e-07, + "loss": 1.9876, + "step": 37453 + }, + { + "epoch": 8.17, + "learning_rate": 9.139174131527203e-07, + "loss": 0.2215, + "step": 37454 + }, + { + "epoch": 8.17, + "learning_rate": 9.138081712912389e-07, + "loss": 1.5406, + "step": 37455 + }, + { + "epoch": 8.17, + "learning_rate": 9.136989294297576e-07, + "loss": 0.5704, + "step": 37456 + }, + { + "epoch": 8.17, + "learning_rate": 9.135896875682762e-07, + "loss": 0.4906, + "step": 37457 + }, + { + "epoch": 8.18, + "learning_rate": 9.134804457067949e-07, + "loss": 1.6337, + "step": 37458 + }, + { + "epoch": 8.18, + "learning_rate": 9.133712038453136e-07, + "loss": 0.5493, + "step": 37459 + }, + { + "epoch": 8.18, + "learning_rate": 9.132619619838323e-07, + "loss": 1.6432, + "step": 37460 + }, + { + "epoch": 8.18, + "learning_rate": 9.13152720122351e-07, + "loss": 1.0711, + "step": 37461 + }, + { + "epoch": 8.18, + "learning_rate": 9.130434782608697e-07, + "loss": 1.1979, + "step": 37462 + }, + { + "epoch": 8.18, + "learning_rate": 9.129342363993883e-07, + "loss": 1.2819, + "step": 37463 + }, + { + "epoch": 8.18, + "learning_rate": 9.12824994537907e-07, + "loss": 0.1435, + "step": 37464 + }, + { + "epoch": 8.18, + "learning_rate": 9.127157526764256e-07, + "loss": 0.8026, + "step": 37465 + }, + { + "epoch": 8.18, + "learning_rate": 9.126065108149444e-07, + "loss": 1.1224, + "step": 37466 + }, + { + "epoch": 8.18, + "learning_rate": 9.12497268953463e-07, + "loss": 0.6475, + "step": 37467 + }, + { + "epoch": 8.18, + "learning_rate": 9.123880270919818e-07, + "loss": 2.7696, + "step": 37468 + }, + { + "epoch": 8.18, + "learning_rate": 9.122787852305004e-07, + "loss": 0.3382, + "step": 37469 + }, + { + "epoch": 8.18, + "learning_rate": 9.121695433690191e-07, + "loss": 2.181, + "step": 37470 + }, + { + "epoch": 8.18, + "learning_rate": 9.120603015075377e-07, + "loss": 0.9813, + "step": 37471 + }, + { + "epoch": 8.18, + "learning_rate": 9.119510596460564e-07, + "loss": 1.2876, + "step": 37472 + }, + { + "epoch": 8.18, + "learning_rate": 9.118418177845751e-07, + "loss": 1.2045, + "step": 37473 + }, + { + "epoch": 8.18, + "learning_rate": 9.117325759230939e-07, + "loss": 0.0894, + "step": 37474 + }, + { + "epoch": 8.18, + "learning_rate": 9.116233340616125e-07, + "loss": 0.6892, + "step": 37475 + }, + { + "epoch": 8.18, + "learning_rate": 9.115140922001312e-07, + "loss": 1.1907, + "step": 37476 + }, + { + "epoch": 8.18, + "learning_rate": 9.114048503386498e-07, + "loss": 0.3992, + "step": 37477 + }, + { + "epoch": 8.18, + "learning_rate": 9.112956084771685e-07, + "loss": 1.4582, + "step": 37478 + }, + { + "epoch": 8.18, + "learning_rate": 9.111863666156872e-07, + "loss": 2.2235, + "step": 37479 + }, + { + "epoch": 8.18, + "learning_rate": 9.110771247542059e-07, + "loss": 0.5628, + "step": 37480 + }, + { + "epoch": 8.18, + "learning_rate": 9.109678828927246e-07, + "loss": 0.4222, + "step": 37481 + }, + { + "epoch": 8.18, + "learning_rate": 9.108586410312433e-07, + "loss": 0.3109, + "step": 37482 + }, + { + "epoch": 8.18, + "learning_rate": 9.107493991697619e-07, + "loss": 0.6776, + "step": 37483 + }, + { + "epoch": 8.18, + "learning_rate": 9.106401573082806e-07, + "loss": 0.911, + "step": 37484 + }, + { + "epoch": 8.18, + "learning_rate": 9.105309154467992e-07, + "loss": 2.9847, + "step": 37485 + }, + { + "epoch": 8.18, + "learning_rate": 9.10421673585318e-07, + "loss": 0.4841, + "step": 37486 + }, + { + "epoch": 8.18, + "learning_rate": 9.103124317238366e-07, + "loss": 1.621, + "step": 37487 + }, + { + "epoch": 8.18, + "learning_rate": 9.102031898623554e-07, + "loss": 0.5671, + "step": 37488 + }, + { + "epoch": 8.18, + "learning_rate": 9.10093948000874e-07, + "loss": 0.7601, + "step": 37489 + }, + { + "epoch": 8.18, + "learning_rate": 9.099847061393927e-07, + "loss": 0.9272, + "step": 37490 + }, + { + "epoch": 8.18, + "learning_rate": 9.098754642779113e-07, + "loss": 0.7774, + "step": 37491 + }, + { + "epoch": 8.18, + "learning_rate": 9.097662224164301e-07, + "loss": 1.4989, + "step": 37492 + }, + { + "epoch": 8.18, + "learning_rate": 9.096569805549487e-07, + "loss": 1.3603, + "step": 37493 + }, + { + "epoch": 8.18, + "learning_rate": 9.095477386934675e-07, + "loss": 0.816, + "step": 37494 + }, + { + "epoch": 8.18, + "learning_rate": 9.094384968319861e-07, + "loss": 1.7394, + "step": 37495 + }, + { + "epoch": 8.18, + "learning_rate": 9.093292549705048e-07, + "loss": 0.7108, + "step": 37496 + }, + { + "epoch": 8.18, + "learning_rate": 9.092200131090234e-07, + "loss": 0.8646, + "step": 37497 + }, + { + "epoch": 8.18, + "learning_rate": 9.091107712475421e-07, + "loss": 1.0532, + "step": 37498 + }, + { + "epoch": 8.18, + "learning_rate": 9.090015293860608e-07, + "loss": 1.9867, + "step": 37499 + }, + { + "epoch": 8.18, + "learning_rate": 9.088922875245795e-07, + "loss": 1.8784, + "step": 37500 + }, + { + "epoch": 8.18, + "learning_rate": 9.087830456630982e-07, + "loss": 1.4242, + "step": 37501 + }, + { + "epoch": 8.18, + "learning_rate": 9.086738038016169e-07, + "loss": 1.2225, + "step": 37502 + }, + { + "epoch": 8.18, + "learning_rate": 9.085645619401355e-07, + "loss": 1.4113, + "step": 37503 + }, + { + "epoch": 8.19, + "learning_rate": 9.084553200786542e-07, + "loss": 0.9882, + "step": 37504 + }, + { + "epoch": 8.19, + "learning_rate": 9.083460782171728e-07, + "loss": 0.0991, + "step": 37505 + }, + { + "epoch": 8.19, + "learning_rate": 9.082368363556916e-07, + "loss": 0.3275, + "step": 37506 + }, + { + "epoch": 8.19, + "learning_rate": 9.081275944942102e-07, + "loss": 0.5626, + "step": 37507 + }, + { + "epoch": 8.19, + "learning_rate": 9.08018352632729e-07, + "loss": 1.6209, + "step": 37508 + }, + { + "epoch": 8.19, + "learning_rate": 9.079091107712476e-07, + "loss": 1.3523, + "step": 37509 + }, + { + "epoch": 8.19, + "learning_rate": 9.077998689097663e-07, + "loss": 1.5125, + "step": 37510 + }, + { + "epoch": 8.19, + "learning_rate": 9.076906270482849e-07, + "loss": 0.1081, + "step": 37511 + }, + { + "epoch": 8.19, + "learning_rate": 9.075813851868037e-07, + "loss": 1.7044, + "step": 37512 + }, + { + "epoch": 8.19, + "learning_rate": 9.074721433253223e-07, + "loss": 0.1326, + "step": 37513 + }, + { + "epoch": 8.19, + "learning_rate": 9.07362901463841e-07, + "loss": 2.1085, + "step": 37514 + }, + { + "epoch": 8.19, + "learning_rate": 9.072536596023597e-07, + "loss": 0.4329, + "step": 37515 + }, + { + "epoch": 8.19, + "learning_rate": 9.071444177408784e-07, + "loss": 1.2773, + "step": 37516 + }, + { + "epoch": 8.19, + "learning_rate": 9.07035175879397e-07, + "loss": 0.9716, + "step": 37517 + }, + { + "epoch": 8.19, + "learning_rate": 9.069259340179157e-07, + "loss": 1.4107, + "step": 37518 + }, + { + "epoch": 8.19, + "learning_rate": 9.068166921564345e-07, + "loss": 2.0209, + "step": 37519 + }, + { + "epoch": 8.19, + "learning_rate": 9.067074502949531e-07, + "loss": 0.0483, + "step": 37520 + }, + { + "epoch": 8.19, + "learning_rate": 9.065982084334717e-07, + "loss": 1.0076, + "step": 37521 + }, + { + "epoch": 8.19, + "learning_rate": 9.064889665719905e-07, + "loss": 0.9132, + "step": 37522 + }, + { + "epoch": 8.19, + "learning_rate": 9.063797247105091e-07, + "loss": 1.0662, + "step": 37523 + }, + { + "epoch": 8.19, + "learning_rate": 9.062704828490278e-07, + "loss": 1.1205, + "step": 37524 + }, + { + "epoch": 8.19, + "learning_rate": 9.061612409875466e-07, + "loss": 0.9219, + "step": 37525 + }, + { + "epoch": 8.19, + "learning_rate": 9.060519991260652e-07, + "loss": 1.1041, + "step": 37526 + }, + { + "epoch": 8.19, + "learning_rate": 9.059427572645838e-07, + "loss": 0.7297, + "step": 37527 + }, + { + "epoch": 8.19, + "learning_rate": 9.058335154031026e-07, + "loss": 0.5775, + "step": 37528 + }, + { + "epoch": 8.19, + "learning_rate": 9.057242735416212e-07, + "loss": 1.7019, + "step": 37529 + }, + { + "epoch": 8.19, + "learning_rate": 9.056150316801399e-07, + "loss": 1.2499, + "step": 37530 + }, + { + "epoch": 8.19, + "learning_rate": 9.055057898186585e-07, + "loss": 1.1301, + "step": 37531 + }, + { + "epoch": 8.19, + "learning_rate": 9.053965479571773e-07, + "loss": 1.7667, + "step": 37532 + }, + { + "epoch": 8.19, + "learning_rate": 9.05287306095696e-07, + "loss": 2.4657, + "step": 37533 + }, + { + "epoch": 8.19, + "learning_rate": 9.051780642342146e-07, + "loss": 2.4533, + "step": 37534 + }, + { + "epoch": 8.19, + "learning_rate": 9.050688223727333e-07, + "loss": 1.5398, + "step": 37535 + }, + { + "epoch": 8.19, + "learning_rate": 9.04959580511252e-07, + "loss": 2.7364, + "step": 37536 + }, + { + "epoch": 8.19, + "learning_rate": 9.048503386497706e-07, + "loss": 0.7241, + "step": 37537 + }, + { + "epoch": 8.19, + "learning_rate": 9.047410967882893e-07, + "loss": 0.8127, + "step": 37538 + }, + { + "epoch": 8.19, + "learning_rate": 9.046318549268081e-07, + "loss": 0.451, + "step": 37539 + }, + { + "epoch": 8.19, + "learning_rate": 9.045226130653267e-07, + "loss": 0.8488, + "step": 37540 + }, + { + "epoch": 8.19, + "learning_rate": 9.044133712038454e-07, + "loss": 0.6046, + "step": 37541 + }, + { + "epoch": 8.19, + "learning_rate": 9.043041293423641e-07, + "loss": 0.3789, + "step": 37542 + }, + { + "epoch": 8.19, + "learning_rate": 9.041948874808827e-07, + "loss": 0.7623, + "step": 37543 + }, + { + "epoch": 8.19, + "learning_rate": 9.040856456194014e-07, + "loss": 1.5948, + "step": 37544 + }, + { + "epoch": 8.19, + "learning_rate": 9.039764037579202e-07, + "loss": 0.1037, + "step": 37545 + }, + { + "epoch": 8.19, + "learning_rate": 9.038671618964388e-07, + "loss": 2.6579, + "step": 37546 + }, + { + "epoch": 8.19, + "learning_rate": 9.037579200349575e-07, + "loss": 2.0954, + "step": 37547 + }, + { + "epoch": 8.19, + "learning_rate": 9.036486781734761e-07, + "loss": 1.4074, + "step": 37548 + }, + { + "epoch": 8.19, + "learning_rate": 9.035394363119948e-07, + "loss": 1.6796, + "step": 37549 + }, + { + "epoch": 8.2, + "learning_rate": 9.034301944505135e-07, + "loss": 0.1989, + "step": 37550 + }, + { + "epoch": 8.2, + "learning_rate": 9.033209525890321e-07, + "loss": 1.7234, + "step": 37551 + }, + { + "epoch": 8.2, + "learning_rate": 9.032117107275509e-07, + "loss": 0.2872, + "step": 37552 + }, + { + "epoch": 8.2, + "learning_rate": 9.031024688660696e-07, + "loss": 0.5837, + "step": 37553 + }, + { + "epoch": 8.2, + "learning_rate": 9.029932270045882e-07, + "loss": 0.8213, + "step": 37554 + }, + { + "epoch": 8.2, + "learning_rate": 9.02883985143107e-07, + "loss": 0.413, + "step": 37555 + }, + { + "epoch": 8.2, + "learning_rate": 9.027747432816256e-07, + "loss": 3.0559, + "step": 37556 + }, + { + "epoch": 8.2, + "learning_rate": 9.026655014201442e-07, + "loss": 0.8068, + "step": 37557 + }, + { + "epoch": 8.2, + "learning_rate": 9.02556259558663e-07, + "loss": 1.3422, + "step": 37558 + }, + { + "epoch": 8.2, + "learning_rate": 9.024470176971817e-07, + "loss": 2.1567, + "step": 37559 + }, + { + "epoch": 8.2, + "learning_rate": 9.023377758357003e-07, + "loss": 0.8716, + "step": 37560 + }, + { + "epoch": 8.2, + "learning_rate": 9.02228533974219e-07, + "loss": 1.379, + "step": 37561 + }, + { + "epoch": 8.2, + "learning_rate": 9.021192921127377e-07, + "loss": 4.0169, + "step": 37562 + }, + { + "epoch": 8.2, + "learning_rate": 9.020100502512563e-07, + "loss": 1.3065, + "step": 37563 + }, + { + "epoch": 8.2, + "learning_rate": 9.01900808389775e-07, + "loss": 0.3889, + "step": 37564 + }, + { + "epoch": 8.2, + "learning_rate": 9.017915665282938e-07, + "loss": 0.3004, + "step": 37565 + }, + { + "epoch": 8.2, + "learning_rate": 9.016823246668124e-07, + "loss": 0.87, + "step": 37566 + }, + { + "epoch": 8.2, + "learning_rate": 9.015730828053311e-07, + "loss": 1.6227, + "step": 37567 + }, + { + "epoch": 8.2, + "learning_rate": 9.014638409438497e-07, + "loss": 1.5913, + "step": 37568 + }, + { + "epoch": 8.2, + "learning_rate": 9.013545990823685e-07, + "loss": 2.0646, + "step": 37569 + }, + { + "epoch": 8.2, + "learning_rate": 9.012453572208871e-07, + "loss": 0.4012, + "step": 37570 + }, + { + "epoch": 8.2, + "learning_rate": 9.011361153594057e-07, + "loss": 0.9019, + "step": 37571 + }, + { + "epoch": 8.2, + "learning_rate": 9.010268734979245e-07, + "loss": 1.1213, + "step": 37572 + }, + { + "epoch": 8.2, + "learning_rate": 9.009176316364432e-07, + "loss": 1.4521, + "step": 37573 + }, + { + "epoch": 8.2, + "learning_rate": 9.008083897749618e-07, + "loss": 2.6729, + "step": 37574 + }, + { + "epoch": 8.2, + "learning_rate": 9.006991479134806e-07, + "loss": 0.6341, + "step": 37575 + }, + { + "epoch": 8.2, + "learning_rate": 9.005899060519992e-07, + "loss": 0.9178, + "step": 37576 + }, + { + "epoch": 8.2, + "learning_rate": 9.004806641905179e-07, + "loss": 0.4429, + "step": 37577 + }, + { + "epoch": 8.2, + "learning_rate": 9.003714223290366e-07, + "loss": 0.7625, + "step": 37578 + }, + { + "epoch": 8.2, + "learning_rate": 9.002621804675553e-07, + "loss": 1.2047, + "step": 37579 + }, + { + "epoch": 8.2, + "learning_rate": 9.001529386060739e-07, + "loss": 0.5914, + "step": 37580 + }, + { + "epoch": 8.2, + "learning_rate": 9.000436967445926e-07, + "loss": 1.3592, + "step": 37581 + }, + { + "epoch": 8.2, + "learning_rate": 8.999344548831113e-07, + "loss": 0.7306, + "step": 37582 + }, + { + "epoch": 8.2, + "learning_rate": 8.9982521302163e-07, + "loss": 1.6083, + "step": 37583 + }, + { + "epoch": 8.2, + "learning_rate": 8.997159711601486e-07, + "loss": 1.5955, + "step": 37584 + }, + { + "epoch": 8.2, + "learning_rate": 8.996067292986674e-07, + "loss": 1.1415, + "step": 37585 + }, + { + "epoch": 8.2, + "learning_rate": 8.99497487437186e-07, + "loss": 0.1436, + "step": 37586 + }, + { + "epoch": 8.2, + "learning_rate": 8.993882455757047e-07, + "loss": 0.7875, + "step": 37587 + }, + { + "epoch": 8.2, + "learning_rate": 8.992790037142233e-07, + "loss": 0.4391, + "step": 37588 + }, + { + "epoch": 8.2, + "learning_rate": 8.991697618527421e-07, + "loss": 0.488, + "step": 37589 + }, + { + "epoch": 8.2, + "learning_rate": 8.990605199912607e-07, + "loss": 0.0882, + "step": 37590 + }, + { + "epoch": 8.2, + "learning_rate": 8.989512781297794e-07, + "loss": 0.0693, + "step": 37591 + }, + { + "epoch": 8.2, + "learning_rate": 8.988420362682981e-07, + "loss": 0.6036, + "step": 37592 + }, + { + "epoch": 8.2, + "learning_rate": 8.987327944068168e-07, + "loss": 0.8887, + "step": 37593 + }, + { + "epoch": 8.2, + "learning_rate": 8.986235525453354e-07, + "loss": 2.1291, + "step": 37594 + }, + { + "epoch": 8.2, + "learning_rate": 8.985143106838541e-07, + "loss": 0.3384, + "step": 37595 + }, + { + "epoch": 8.21, + "learning_rate": 8.984050688223728e-07, + "loss": 0.4324, + "step": 37596 + }, + { + "epoch": 8.21, + "learning_rate": 8.982958269608915e-07, + "loss": 0.0688, + "step": 37597 + }, + { + "epoch": 8.21, + "learning_rate": 8.981865850994102e-07, + "loss": 1.9248, + "step": 37598 + }, + { + "epoch": 8.21, + "learning_rate": 8.980773432379289e-07, + "loss": 0.6859, + "step": 37599 + }, + { + "epoch": 8.21, + "learning_rate": 8.979681013764475e-07, + "loss": 1.0727, + "step": 37600 + }, + { + "epoch": 8.21, + "learning_rate": 8.978588595149662e-07, + "loss": 1.5092, + "step": 37601 + }, + { + "epoch": 8.21, + "learning_rate": 8.977496176534848e-07, + "loss": 0.7093, + "step": 37602 + }, + { + "epoch": 8.21, + "learning_rate": 8.976403757920036e-07, + "loss": 0.0421, + "step": 37603 + }, + { + "epoch": 8.21, + "learning_rate": 8.975311339305222e-07, + "loss": 0.9718, + "step": 37604 + }, + { + "epoch": 8.21, + "learning_rate": 8.97421892069041e-07, + "loss": 1.5047, + "step": 37605 + }, + { + "epoch": 8.21, + "learning_rate": 8.973126502075596e-07, + "loss": 1.4904, + "step": 37606 + }, + { + "epoch": 8.21, + "learning_rate": 8.972034083460783e-07, + "loss": 0.7872, + "step": 37607 + }, + { + "epoch": 8.21, + "learning_rate": 8.970941664845969e-07, + "loss": 1.4588, + "step": 37608 + }, + { + "epoch": 8.21, + "learning_rate": 8.969849246231157e-07, + "loss": 1.646, + "step": 37609 + }, + { + "epoch": 8.21, + "learning_rate": 8.968756827616343e-07, + "loss": 0.9403, + "step": 37610 + }, + { + "epoch": 8.21, + "learning_rate": 8.967664409001531e-07, + "loss": 2.1835, + "step": 37611 + }, + { + "epoch": 8.21, + "learning_rate": 8.966571990386717e-07, + "loss": 2.4781, + "step": 37612 + }, + { + "epoch": 8.21, + "learning_rate": 8.965479571771904e-07, + "loss": 0.859, + "step": 37613 + }, + { + "epoch": 8.21, + "learning_rate": 8.96438715315709e-07, + "loss": 0.2382, + "step": 37614 + }, + { + "epoch": 8.21, + "learning_rate": 8.963294734542277e-07, + "loss": 1.0552, + "step": 37615 + }, + { + "epoch": 8.21, + "learning_rate": 8.962202315927464e-07, + "loss": 0.0993, + "step": 37616 + }, + { + "epoch": 8.21, + "learning_rate": 8.961109897312651e-07, + "loss": 1.1374, + "step": 37617 + }, + { + "epoch": 8.21, + "learning_rate": 8.960017478697838e-07, + "loss": 0.1568, + "step": 37618 + }, + { + "epoch": 8.21, + "learning_rate": 8.958925060083025e-07, + "loss": 1.6484, + "step": 37619 + }, + { + "epoch": 8.21, + "learning_rate": 8.957832641468211e-07, + "loss": 0.1851, + "step": 37620 + }, + { + "epoch": 8.21, + "learning_rate": 8.956740222853398e-07, + "loss": 1.0841, + "step": 37621 + }, + { + "epoch": 8.21, + "learning_rate": 8.955647804238584e-07, + "loss": 1.4297, + "step": 37622 + }, + { + "epoch": 8.21, + "learning_rate": 8.954555385623772e-07, + "loss": 0.0833, + "step": 37623 + }, + { + "epoch": 8.21, + "learning_rate": 8.953462967008958e-07, + "loss": 0.2117, + "step": 37624 + }, + { + "epoch": 8.21, + "learning_rate": 8.952370548394146e-07, + "loss": 1.9226, + "step": 37625 + }, + { + "epoch": 8.21, + "learning_rate": 8.951278129779332e-07, + "loss": 1.4769, + "step": 37626 + }, + { + "epoch": 8.21, + "learning_rate": 8.950185711164519e-07, + "loss": 0.6611, + "step": 37627 + }, + { + "epoch": 8.21, + "learning_rate": 8.949093292549705e-07, + "loss": 1.289, + "step": 37628 + }, + { + "epoch": 8.21, + "learning_rate": 8.948000873934893e-07, + "loss": 0.8406, + "step": 37629 + }, + { + "epoch": 8.21, + "learning_rate": 8.946908455320079e-07, + "loss": 0.7189, + "step": 37630 + }, + { + "epoch": 8.21, + "learning_rate": 8.945816036705267e-07, + "loss": 1.7776, + "step": 37631 + }, + { + "epoch": 8.21, + "learning_rate": 8.944723618090453e-07, + "loss": 0.1438, + "step": 37632 + }, + { + "epoch": 8.21, + "learning_rate": 8.94363119947564e-07, + "loss": 1.0733, + "step": 37633 + }, + { + "epoch": 8.21, + "learning_rate": 8.942538780860826e-07, + "loss": 0.5259, + "step": 37634 + }, + { + "epoch": 8.21, + "learning_rate": 8.941446362246013e-07, + "loss": 1.4684, + "step": 37635 + }, + { + "epoch": 8.21, + "learning_rate": 8.9403539436312e-07, + "loss": 1.455, + "step": 37636 + }, + { + "epoch": 8.21, + "learning_rate": 8.939261525016387e-07, + "loss": 1.7204, + "step": 37637 + }, + { + "epoch": 8.21, + "learning_rate": 8.938169106401574e-07, + "loss": 0.1604, + "step": 37638 + }, + { + "epoch": 8.21, + "learning_rate": 8.937076687786761e-07, + "loss": 0.1828, + "step": 37639 + }, + { + "epoch": 8.21, + "learning_rate": 8.935984269171947e-07, + "loss": 2.2359, + "step": 37640 + }, + { + "epoch": 8.21, + "learning_rate": 8.934891850557134e-07, + "loss": 0.9231, + "step": 37641 + }, + { + "epoch": 8.22, + "learning_rate": 8.93379943194232e-07, + "loss": 1.4896, + "step": 37642 + }, + { + "epoch": 8.22, + "learning_rate": 8.932707013327508e-07, + "loss": 2.1037, + "step": 37643 + }, + { + "epoch": 8.22, + "learning_rate": 8.931614594712696e-07, + "loss": 1.4291, + "step": 37644 + }, + { + "epoch": 8.22, + "learning_rate": 8.930522176097882e-07, + "loss": 0.9677, + "step": 37645 + }, + { + "epoch": 8.22, + "learning_rate": 8.929429757483068e-07, + "loss": 1.782, + "step": 37646 + }, + { + "epoch": 8.22, + "learning_rate": 8.928337338868255e-07, + "loss": 1.2647, + "step": 37647 + }, + { + "epoch": 8.22, + "learning_rate": 8.927244920253441e-07, + "loss": 1.5681, + "step": 37648 + }, + { + "epoch": 8.22, + "learning_rate": 8.926152501638628e-07, + "loss": 2.0378, + "step": 37649 + }, + { + "epoch": 8.22, + "learning_rate": 8.925060083023815e-07, + "loss": 3.2821, + "step": 37650 + }, + { + "epoch": 8.22, + "learning_rate": 8.923967664409003e-07, + "loss": 0.4095, + "step": 37651 + }, + { + "epoch": 8.22, + "learning_rate": 8.92287524579419e-07, + "loss": 0.4896, + "step": 37652 + }, + { + "epoch": 8.22, + "learning_rate": 8.921782827179376e-07, + "loss": 0.0526, + "step": 37653 + }, + { + "epoch": 8.22, + "learning_rate": 8.920690408564562e-07, + "loss": 0.596, + "step": 37654 + }, + { + "epoch": 8.22, + "learning_rate": 8.919597989949749e-07, + "loss": 2.2685, + "step": 37655 + }, + { + "epoch": 8.22, + "learning_rate": 8.918505571334935e-07, + "loss": 0.1807, + "step": 37656 + }, + { + "epoch": 8.22, + "learning_rate": 8.917413152720123e-07, + "loss": 1.5003, + "step": 37657 + }, + { + "epoch": 8.22, + "learning_rate": 8.916320734105311e-07, + "loss": 1.1269, + "step": 37658 + }, + { + "epoch": 8.22, + "learning_rate": 8.915228315490497e-07, + "loss": 1.0409, + "step": 37659 + }, + { + "epoch": 8.22, + "learning_rate": 8.914135896875683e-07, + "loss": 0.1787, + "step": 37660 + }, + { + "epoch": 8.22, + "learning_rate": 8.91304347826087e-07, + "loss": 0.9115, + "step": 37661 + }, + { + "epoch": 8.22, + "learning_rate": 8.911951059646056e-07, + "loss": 0.8775, + "step": 37662 + }, + { + "epoch": 8.22, + "learning_rate": 8.910858641031244e-07, + "loss": 0.637, + "step": 37663 + }, + { + "epoch": 8.22, + "learning_rate": 8.909766222416432e-07, + "loss": 1.0145, + "step": 37664 + }, + { + "epoch": 8.22, + "learning_rate": 8.908673803801618e-07, + "loss": 1.6411, + "step": 37665 + }, + { + "epoch": 8.22, + "learning_rate": 8.907581385186805e-07, + "loss": 1.2866, + "step": 37666 + }, + { + "epoch": 8.22, + "learning_rate": 8.906488966571991e-07, + "loss": 1.3177, + "step": 37667 + }, + { + "epoch": 8.22, + "learning_rate": 8.905396547957177e-07, + "loss": 1.2843, + "step": 37668 + }, + { + "epoch": 8.22, + "learning_rate": 8.904304129342364e-07, + "loss": 1.5933, + "step": 37669 + }, + { + "epoch": 8.22, + "learning_rate": 8.903211710727551e-07, + "loss": 1.5468, + "step": 37670 + }, + { + "epoch": 8.22, + "learning_rate": 8.902119292112739e-07, + "loss": 1.1175, + "step": 37671 + }, + { + "epoch": 8.22, + "learning_rate": 8.901026873497926e-07, + "loss": 3.5326, + "step": 37672 + }, + { + "epoch": 8.22, + "learning_rate": 8.899934454883112e-07, + "loss": 2.7055, + "step": 37673 + }, + { + "epoch": 8.22, + "learning_rate": 8.898842036268298e-07, + "loss": 0.1843, + "step": 37674 + }, + { + "epoch": 8.22, + "learning_rate": 8.897749617653485e-07, + "loss": 1.1811, + "step": 37675 + }, + { + "epoch": 8.22, + "learning_rate": 8.896657199038671e-07, + "loss": 1.5604, + "step": 37676 + }, + { + "epoch": 8.22, + "learning_rate": 8.895564780423859e-07, + "loss": 0.2325, + "step": 37677 + }, + { + "epoch": 8.22, + "learning_rate": 8.894472361809047e-07, + "loss": 0.7882, + "step": 37678 + }, + { + "epoch": 8.22, + "learning_rate": 8.893379943194233e-07, + "loss": 1.3206, + "step": 37679 + }, + { + "epoch": 8.22, + "learning_rate": 8.89228752457942e-07, + "loss": 0.8542, + "step": 37680 + }, + { + "epoch": 8.22, + "learning_rate": 8.891195105964606e-07, + "loss": 0.0769, + "step": 37681 + }, + { + "epoch": 8.22, + "learning_rate": 8.890102687349792e-07, + "loss": 0.7942, + "step": 37682 + }, + { + "epoch": 8.22, + "learning_rate": 8.88901026873498e-07, + "loss": 2.4169, + "step": 37683 + }, + { + "epoch": 8.22, + "learning_rate": 8.887917850120168e-07, + "loss": 0.2752, + "step": 37684 + }, + { + "epoch": 8.22, + "learning_rate": 8.886825431505354e-07, + "loss": 1.2893, + "step": 37685 + }, + { + "epoch": 8.22, + "learning_rate": 8.885733012890541e-07, + "loss": 4.4693, + "step": 37686 + }, + { + "epoch": 8.23, + "learning_rate": 8.884640594275727e-07, + "loss": 1.4328, + "step": 37687 + }, + { + "epoch": 8.23, + "learning_rate": 8.883548175660914e-07, + "loss": 0.1177, + "step": 37688 + }, + { + "epoch": 8.23, + "learning_rate": 8.8824557570461e-07, + "loss": 0.5893, + "step": 37689 + }, + { + "epoch": 8.23, + "learning_rate": 8.881363338431287e-07, + "loss": 0.9296, + "step": 37690 + }, + { + "epoch": 8.23, + "learning_rate": 8.880270919816475e-07, + "loss": 1.022, + "step": 37691 + }, + { + "epoch": 8.23, + "learning_rate": 8.879178501201662e-07, + "loss": 1.9035, + "step": 37692 + }, + { + "epoch": 8.23, + "learning_rate": 8.878086082586848e-07, + "loss": 0.3486, + "step": 37693 + }, + { + "epoch": 8.23, + "learning_rate": 8.876993663972035e-07, + "loss": 0.3831, + "step": 37694 + }, + { + "epoch": 8.23, + "learning_rate": 8.875901245357221e-07, + "loss": 1.0821, + "step": 37695 + }, + { + "epoch": 8.23, + "learning_rate": 8.874808826742407e-07, + "loss": 0.804, + "step": 37696 + }, + { + "epoch": 8.23, + "learning_rate": 8.873716408127596e-07, + "loss": 0.1434, + "step": 37697 + }, + { + "epoch": 8.23, + "learning_rate": 8.872623989512783e-07, + "loss": 2.3044, + "step": 37698 + }, + { + "epoch": 8.23, + "learning_rate": 8.871531570897969e-07, + "loss": 2.0825, + "step": 37699 + }, + { + "epoch": 8.23, + "learning_rate": 8.870439152283156e-07, + "loss": 0.158, + "step": 37700 + }, + { + "epoch": 8.23, + "learning_rate": 8.869346733668342e-07, + "loss": 0.8266, + "step": 37701 + }, + { + "epoch": 8.23, + "learning_rate": 8.868254315053529e-07, + "loss": 1.2832, + "step": 37702 + }, + { + "epoch": 8.23, + "learning_rate": 8.867161896438715e-07, + "loss": 0.5054, + "step": 37703 + }, + { + "epoch": 8.23, + "learning_rate": 8.866069477823904e-07, + "loss": 1.7612, + "step": 37704 + }, + { + "epoch": 8.23, + "learning_rate": 8.86497705920909e-07, + "loss": 0.7446, + "step": 37705 + }, + { + "epoch": 8.23, + "learning_rate": 8.863884640594277e-07, + "loss": 1.6204, + "step": 37706 + }, + { + "epoch": 8.23, + "learning_rate": 8.862792221979463e-07, + "loss": 0.19, + "step": 37707 + }, + { + "epoch": 8.23, + "learning_rate": 8.86169980336465e-07, + "loss": 2.6322, + "step": 37708 + }, + { + "epoch": 8.23, + "learning_rate": 8.860607384749836e-07, + "loss": 0.5528, + "step": 37709 + }, + { + "epoch": 8.23, + "learning_rate": 8.859514966135022e-07, + "loss": 1.0442, + "step": 37710 + }, + { + "epoch": 8.23, + "learning_rate": 8.858422547520211e-07, + "loss": 1.9951, + "step": 37711 + }, + { + "epoch": 8.23, + "learning_rate": 8.857330128905398e-07, + "loss": 1.3891, + "step": 37712 + }, + { + "epoch": 8.23, + "learning_rate": 8.856237710290584e-07, + "loss": 1.0632, + "step": 37713 + }, + { + "epoch": 8.23, + "learning_rate": 8.855145291675771e-07, + "loss": 1.0368, + "step": 37714 + }, + { + "epoch": 8.23, + "learning_rate": 8.854052873060957e-07, + "loss": 1.2927, + "step": 37715 + }, + { + "epoch": 8.23, + "learning_rate": 8.852960454446144e-07, + "loss": 0.488, + "step": 37716 + }, + { + "epoch": 8.23, + "learning_rate": 8.851868035831332e-07, + "loss": 1.323, + "step": 37717 + }, + { + "epoch": 8.23, + "learning_rate": 8.850775617216519e-07, + "loss": 0.7823, + "step": 37718 + }, + { + "epoch": 8.23, + "learning_rate": 8.849683198601705e-07, + "loss": 2.227, + "step": 37719 + }, + { + "epoch": 8.23, + "learning_rate": 8.848590779986892e-07, + "loss": 1.1566, + "step": 37720 + }, + { + "epoch": 8.23, + "learning_rate": 8.847498361372078e-07, + "loss": 1.0308, + "step": 37721 + }, + { + "epoch": 8.23, + "learning_rate": 8.846405942757265e-07, + "loss": 0.6453, + "step": 37722 + }, + { + "epoch": 8.23, + "learning_rate": 8.845313524142451e-07, + "loss": 0.8367, + "step": 37723 + }, + { + "epoch": 8.23, + "learning_rate": 8.84422110552764e-07, + "loss": 1.0359, + "step": 37724 + }, + { + "epoch": 8.23, + "learning_rate": 8.843128686912826e-07, + "loss": 1.4269, + "step": 37725 + }, + { + "epoch": 8.23, + "learning_rate": 8.842036268298013e-07, + "loss": 1.7608, + "step": 37726 + }, + { + "epoch": 8.23, + "learning_rate": 8.840943849683199e-07, + "loss": 0.9412, + "step": 37727 + }, + { + "epoch": 8.23, + "learning_rate": 8.839851431068386e-07, + "loss": 0.3703, + "step": 37728 + }, + { + "epoch": 8.23, + "learning_rate": 8.838759012453572e-07, + "loss": 1.0501, + "step": 37729 + }, + { + "epoch": 8.23, + "learning_rate": 8.837666593838761e-07, + "loss": 0.2247, + "step": 37730 + }, + { + "epoch": 8.23, + "learning_rate": 8.836574175223947e-07, + "loss": 1.3427, + "step": 37731 + }, + { + "epoch": 8.23, + "learning_rate": 8.835481756609134e-07, + "loss": 1.3527, + "step": 37732 + }, + { + "epoch": 8.24, + "learning_rate": 8.83438933799432e-07, + "loss": 1.3184, + "step": 37733 + }, + { + "epoch": 8.24, + "learning_rate": 8.833296919379507e-07, + "loss": 1.3058, + "step": 37734 + }, + { + "epoch": 8.24, + "learning_rate": 8.832204500764693e-07, + "loss": 1.3063, + "step": 37735 + }, + { + "epoch": 8.24, + "learning_rate": 8.83111208214988e-07, + "loss": 1.257, + "step": 37736 + }, + { + "epoch": 8.24, + "learning_rate": 8.830019663535068e-07, + "loss": 1.7874, + "step": 37737 + }, + { + "epoch": 8.24, + "learning_rate": 8.828927244920255e-07, + "loss": 1.0295, + "step": 37738 + }, + { + "epoch": 8.24, + "learning_rate": 8.827834826305441e-07, + "loss": 0.8181, + "step": 37739 + }, + { + "epoch": 8.24, + "learning_rate": 8.826742407690628e-07, + "loss": 0.5259, + "step": 37740 + }, + { + "epoch": 8.24, + "learning_rate": 8.825649989075814e-07, + "loss": 0.631, + "step": 37741 + }, + { + "epoch": 8.24, + "learning_rate": 8.824557570461001e-07, + "loss": 2.3129, + "step": 37742 + }, + { + "epoch": 8.24, + "learning_rate": 8.823465151846187e-07, + "loss": 0.5359, + "step": 37743 + }, + { + "epoch": 8.24, + "learning_rate": 8.822372733231376e-07, + "loss": 0.8005, + "step": 37744 + }, + { + "epoch": 8.24, + "learning_rate": 8.821280314616562e-07, + "loss": 0.4282, + "step": 37745 + }, + { + "epoch": 8.24, + "learning_rate": 8.820187896001749e-07, + "loss": 1.4466, + "step": 37746 + }, + { + "epoch": 8.24, + "learning_rate": 8.819095477386935e-07, + "loss": 0.3894, + "step": 37747 + }, + { + "epoch": 8.24, + "learning_rate": 8.818003058772122e-07, + "loss": 0.2541, + "step": 37748 + }, + { + "epoch": 8.24, + "learning_rate": 8.816910640157308e-07, + "loss": 0.7817, + "step": 37749 + }, + { + "epoch": 8.24, + "learning_rate": 8.815818221542497e-07, + "loss": 0.7911, + "step": 37750 + }, + { + "epoch": 8.24, + "learning_rate": 8.814725802927683e-07, + "loss": 0.1833, + "step": 37751 + }, + { + "epoch": 8.24, + "learning_rate": 8.81363338431287e-07, + "loss": 1.2252, + "step": 37752 + }, + { + "epoch": 8.24, + "learning_rate": 8.812540965698056e-07, + "loss": 0.7951, + "step": 37753 + }, + { + "epoch": 8.24, + "learning_rate": 8.811448547083243e-07, + "loss": 0.8566, + "step": 37754 + }, + { + "epoch": 8.24, + "learning_rate": 8.810356128468429e-07, + "loss": 2.341, + "step": 37755 + }, + { + "epoch": 8.24, + "learning_rate": 8.809263709853616e-07, + "loss": 1.8314, + "step": 37756 + }, + { + "epoch": 8.24, + "learning_rate": 8.808171291238804e-07, + "loss": 0.8446, + "step": 37757 + }, + { + "epoch": 8.24, + "learning_rate": 8.807078872623991e-07, + "loss": 1.6928, + "step": 37758 + }, + { + "epoch": 8.24, + "learning_rate": 8.805986454009177e-07, + "loss": 1.1854, + "step": 37759 + }, + { + "epoch": 8.24, + "learning_rate": 8.804894035394364e-07, + "loss": 1.4947, + "step": 37760 + }, + { + "epoch": 8.24, + "learning_rate": 8.80380161677955e-07, + "loss": 0.6793, + "step": 37761 + }, + { + "epoch": 8.24, + "learning_rate": 8.802709198164737e-07, + "loss": 0.102, + "step": 37762 + }, + { + "epoch": 8.24, + "learning_rate": 8.801616779549924e-07, + "loss": 0.9863, + "step": 37763 + }, + { + "epoch": 8.24, + "learning_rate": 8.800524360935112e-07, + "loss": 0.0797, + "step": 37764 + }, + { + "epoch": 8.24, + "learning_rate": 8.799431942320298e-07, + "loss": 0.716, + "step": 37765 + }, + { + "epoch": 8.24, + "learning_rate": 8.798339523705485e-07, + "loss": 0.7957, + "step": 37766 + }, + { + "epoch": 8.24, + "learning_rate": 8.797247105090671e-07, + "loss": 0.323, + "step": 37767 + }, + { + "epoch": 8.24, + "learning_rate": 8.796154686475858e-07, + "loss": 2.1945, + "step": 37768 + }, + { + "epoch": 8.24, + "learning_rate": 8.795062267861044e-07, + "loss": 0.3763, + "step": 37769 + }, + { + "epoch": 8.24, + "learning_rate": 8.793969849246232e-07, + "loss": 1.4571, + "step": 37770 + }, + { + "epoch": 8.24, + "learning_rate": 8.792877430631419e-07, + "loss": 1.6886, + "step": 37771 + }, + { + "epoch": 8.24, + "learning_rate": 8.791785012016606e-07, + "loss": 1.2784, + "step": 37772 + }, + { + "epoch": 8.24, + "learning_rate": 8.790692593401792e-07, + "loss": 1.1896, + "step": 37773 + }, + { + "epoch": 8.24, + "learning_rate": 8.789600174786979e-07, + "loss": 0.2275, + "step": 37774 + }, + { + "epoch": 8.24, + "learning_rate": 8.788507756172165e-07, + "loss": 2.7801, + "step": 37775 + }, + { + "epoch": 8.24, + "learning_rate": 8.787415337557352e-07, + "loss": 0.7183, + "step": 37776 + }, + { + "epoch": 8.24, + "learning_rate": 8.786322918942541e-07, + "loss": 1.8679, + "step": 37777 + }, + { + "epoch": 8.24, + "learning_rate": 8.785230500327727e-07, + "loss": 0.134, + "step": 37778 + }, + { + "epoch": 8.25, + "learning_rate": 8.784138081712913e-07, + "loss": 0.1166, + "step": 37779 + }, + { + "epoch": 8.25, + "learning_rate": 8.7830456630981e-07, + "loss": 2.5423, + "step": 37780 + }, + { + "epoch": 8.25, + "learning_rate": 8.781953244483286e-07, + "loss": 1.2025, + "step": 37781 + }, + { + "epoch": 8.25, + "learning_rate": 8.780860825868473e-07, + "loss": 1.1946, + "step": 37782 + }, + { + "epoch": 8.25, + "learning_rate": 8.779768407253661e-07, + "loss": 0.081, + "step": 37783 + }, + { + "epoch": 8.25, + "learning_rate": 8.778675988638848e-07, + "loss": 0.7996, + "step": 37784 + }, + { + "epoch": 8.25, + "learning_rate": 8.777583570024034e-07, + "loss": 1.773, + "step": 37785 + }, + { + "epoch": 8.25, + "learning_rate": 8.776491151409221e-07, + "loss": 1.6222, + "step": 37786 + }, + { + "epoch": 8.25, + "learning_rate": 8.775398732794407e-07, + "loss": 1.233, + "step": 37787 + }, + { + "epoch": 8.25, + "learning_rate": 8.774306314179594e-07, + "loss": 1.8024, + "step": 37788 + }, + { + "epoch": 8.25, + "learning_rate": 8.77321389556478e-07, + "loss": 0.8452, + "step": 37789 + }, + { + "epoch": 8.25, + "learning_rate": 8.772121476949968e-07, + "loss": 1.0386, + "step": 37790 + }, + { + "epoch": 8.25, + "learning_rate": 8.771029058335156e-07, + "loss": 1.8212, + "step": 37791 + }, + { + "epoch": 8.25, + "learning_rate": 8.769936639720342e-07, + "loss": 0.9384, + "step": 37792 + }, + { + "epoch": 8.25, + "learning_rate": 8.768844221105528e-07, + "loss": 0.3593, + "step": 37793 + }, + { + "epoch": 8.25, + "learning_rate": 8.767751802490715e-07, + "loss": 2.2447, + "step": 37794 + }, + { + "epoch": 8.25, + "learning_rate": 8.766659383875901e-07, + "loss": 1.4649, + "step": 37795 + }, + { + "epoch": 8.25, + "learning_rate": 8.765566965261088e-07, + "loss": 0.9132, + "step": 37796 + }, + { + "epoch": 8.25, + "learning_rate": 8.764474546646277e-07, + "loss": 2.031, + "step": 37797 + }, + { + "epoch": 8.25, + "learning_rate": 8.763382128031463e-07, + "loss": 0.1233, + "step": 37798 + }, + { + "epoch": 8.25, + "learning_rate": 8.762289709416649e-07, + "loss": 0.8902, + "step": 37799 + }, + { + "epoch": 8.25, + "learning_rate": 8.761197290801836e-07, + "loss": 1.8462, + "step": 37800 + }, + { + "epoch": 8.25, + "learning_rate": 8.760104872187022e-07, + "loss": 0.9687, + "step": 37801 + }, + { + "epoch": 8.25, + "learning_rate": 8.759012453572209e-07, + "loss": 0.6676, + "step": 37802 + }, + { + "epoch": 8.25, + "learning_rate": 8.757920034957397e-07, + "loss": 1.3834, + "step": 37803 + }, + { + "epoch": 8.25, + "learning_rate": 8.756827616342584e-07, + "loss": 2.1897, + "step": 37804 + }, + { + "epoch": 8.25, + "learning_rate": 8.755735197727771e-07, + "loss": 1.1222, + "step": 37805 + }, + { + "epoch": 8.25, + "learning_rate": 8.754642779112957e-07, + "loss": 2.5432, + "step": 37806 + }, + { + "epoch": 8.25, + "learning_rate": 8.753550360498143e-07, + "loss": 0.7734, + "step": 37807 + }, + { + "epoch": 8.25, + "learning_rate": 8.75245794188333e-07, + "loss": 1.3474, + "step": 37808 + }, + { + "epoch": 8.25, + "learning_rate": 8.751365523268516e-07, + "loss": 1.0908, + "step": 37809 + }, + { + "epoch": 8.25, + "learning_rate": 8.750273104653704e-07, + "loss": 2.0944, + "step": 37810 + }, + { + "epoch": 8.25, + "learning_rate": 8.749180686038892e-07, + "loss": 1.3274, + "step": 37811 + }, + { + "epoch": 8.25, + "learning_rate": 8.748088267424078e-07, + "loss": 1.0655, + "step": 37812 + }, + { + "epoch": 8.25, + "learning_rate": 8.746995848809265e-07, + "loss": 1.2191, + "step": 37813 + }, + { + "epoch": 8.25, + "learning_rate": 8.745903430194451e-07, + "loss": 1.9433, + "step": 37814 + }, + { + "epoch": 8.25, + "learning_rate": 8.744811011579637e-07, + "loss": 0.5303, + "step": 37815 + }, + { + "epoch": 8.25, + "learning_rate": 8.743718592964825e-07, + "loss": 1.4235, + "step": 37816 + }, + { + "epoch": 8.25, + "learning_rate": 8.742626174350012e-07, + "loss": 2.6041, + "step": 37817 + }, + { + "epoch": 8.25, + "learning_rate": 8.741533755735199e-07, + "loss": 0.9821, + "step": 37818 + }, + { + "epoch": 8.25, + "learning_rate": 8.740441337120386e-07, + "loss": 3.2037, + "step": 37819 + }, + { + "epoch": 8.25, + "learning_rate": 8.739348918505572e-07, + "loss": 1.1974, + "step": 37820 + }, + { + "epoch": 8.25, + "learning_rate": 8.738256499890758e-07, + "loss": 1.5535, + "step": 37821 + }, + { + "epoch": 8.25, + "learning_rate": 8.737164081275945e-07, + "loss": 0.7873, + "step": 37822 + }, + { + "epoch": 8.25, + "learning_rate": 8.736071662661133e-07, + "loss": 0.725, + "step": 37823 + }, + { + "epoch": 8.25, + "learning_rate": 8.734979244046319e-07, + "loss": 0.7614, + "step": 37824 + }, + { + "epoch": 8.26, + "learning_rate": 8.733886825431507e-07, + "loss": 1.2153, + "step": 37825 + }, + { + "epoch": 8.26, + "learning_rate": 8.732794406816693e-07, + "loss": 0.6334, + "step": 37826 + }, + { + "epoch": 8.26, + "learning_rate": 8.73170198820188e-07, + "loss": 1.8232, + "step": 37827 + }, + { + "epoch": 8.26, + "learning_rate": 8.730609569587066e-07, + "loss": 0.9623, + "step": 37828 + }, + { + "epoch": 8.26, + "learning_rate": 8.729517150972252e-07, + "loss": 0.6329, + "step": 37829 + }, + { + "epoch": 8.26, + "learning_rate": 8.72842473235744e-07, + "loss": 1.2073, + "step": 37830 + }, + { + "epoch": 8.26, + "learning_rate": 8.727332313742628e-07, + "loss": 1.0958, + "step": 37831 + }, + { + "epoch": 8.26, + "learning_rate": 8.726239895127814e-07, + "loss": 0.598, + "step": 37832 + }, + { + "epoch": 8.26, + "learning_rate": 8.725147476513001e-07, + "loss": 1.5727, + "step": 37833 + }, + { + "epoch": 8.26, + "learning_rate": 8.724055057898187e-07, + "loss": 1.2068, + "step": 37834 + }, + { + "epoch": 8.26, + "learning_rate": 8.722962639283373e-07, + "loss": 1.722, + "step": 37835 + }, + { + "epoch": 8.26, + "learning_rate": 8.721870220668561e-07, + "loss": 1.6525, + "step": 37836 + }, + { + "epoch": 8.26, + "learning_rate": 8.720777802053748e-07, + "loss": 1.138, + "step": 37837 + }, + { + "epoch": 8.26, + "learning_rate": 8.719685383438935e-07, + "loss": 0.1722, + "step": 37838 + }, + { + "epoch": 8.26, + "learning_rate": 8.718592964824122e-07, + "loss": 1.5725, + "step": 37839 + }, + { + "epoch": 8.26, + "learning_rate": 8.717500546209308e-07, + "loss": 2.7499, + "step": 37840 + }, + { + "epoch": 8.26, + "learning_rate": 8.716408127594495e-07, + "loss": 1.8889, + "step": 37841 + }, + { + "epoch": 8.26, + "learning_rate": 8.715315708979681e-07, + "loss": 1.2635, + "step": 37842 + }, + { + "epoch": 8.26, + "learning_rate": 8.714223290364869e-07, + "loss": 2.1496, + "step": 37843 + }, + { + "epoch": 8.26, + "learning_rate": 8.713130871750055e-07, + "loss": 0.4902, + "step": 37844 + }, + { + "epoch": 8.26, + "learning_rate": 8.712038453135243e-07, + "loss": 0.0882, + "step": 37845 + }, + { + "epoch": 8.26, + "learning_rate": 8.710946034520429e-07, + "loss": 0.5233, + "step": 37846 + }, + { + "epoch": 8.26, + "learning_rate": 8.709853615905616e-07, + "loss": 0.9587, + "step": 37847 + }, + { + "epoch": 8.26, + "learning_rate": 8.708761197290802e-07, + "loss": 1.156, + "step": 37848 + }, + { + "epoch": 8.26, + "learning_rate": 8.70766877867599e-07, + "loss": 1.7463, + "step": 37849 + }, + { + "epoch": 8.26, + "learning_rate": 8.706576360061176e-07, + "loss": 2.3968, + "step": 37850 + }, + { + "epoch": 8.26, + "learning_rate": 8.705483941446363e-07, + "loss": 0.3266, + "step": 37851 + }, + { + "epoch": 8.26, + "learning_rate": 8.70439152283155e-07, + "loss": 1.0127, + "step": 37852 + }, + { + "epoch": 8.26, + "learning_rate": 8.703299104216737e-07, + "loss": 0.9109, + "step": 37853 + }, + { + "epoch": 8.26, + "learning_rate": 8.702206685601923e-07, + "loss": 0.3696, + "step": 37854 + }, + { + "epoch": 8.26, + "learning_rate": 8.70111426698711e-07, + "loss": 1.4701, + "step": 37855 + }, + { + "epoch": 8.26, + "learning_rate": 8.700021848372297e-07, + "loss": 0.875, + "step": 37856 + }, + { + "epoch": 8.26, + "learning_rate": 8.698929429757484e-07, + "loss": 1.9205, + "step": 37857 + }, + { + "epoch": 8.26, + "learning_rate": 8.69783701114267e-07, + "loss": 2.3756, + "step": 37858 + }, + { + "epoch": 8.26, + "learning_rate": 8.696744592527858e-07, + "loss": 1.132, + "step": 37859 + }, + { + "epoch": 8.26, + "learning_rate": 8.695652173913044e-07, + "loss": 0.1484, + "step": 37860 + }, + { + "epoch": 8.26, + "learning_rate": 8.694559755298231e-07, + "loss": 1.0523, + "step": 37861 + }, + { + "epoch": 8.26, + "learning_rate": 8.693467336683417e-07, + "loss": 1.1902, + "step": 37862 + }, + { + "epoch": 8.26, + "learning_rate": 8.692374918068605e-07, + "loss": 1.2772, + "step": 37863 + }, + { + "epoch": 8.26, + "learning_rate": 8.691282499453791e-07, + "loss": 0.7961, + "step": 37864 + }, + { + "epoch": 8.26, + "learning_rate": 8.690190080838979e-07, + "loss": 0.8741, + "step": 37865 + }, + { + "epoch": 8.26, + "learning_rate": 8.689097662224165e-07, + "loss": 2.588, + "step": 37866 + }, + { + "epoch": 8.26, + "learning_rate": 8.688005243609352e-07, + "loss": 1.7471, + "step": 37867 + }, + { + "epoch": 8.26, + "learning_rate": 8.686912824994538e-07, + "loss": 0.2084, + "step": 37868 + }, + { + "epoch": 8.26, + "learning_rate": 8.685820406379726e-07, + "loss": 0.3766, + "step": 37869 + }, + { + "epoch": 8.26, + "learning_rate": 8.684727987764912e-07, + "loss": 2.6422, + "step": 37870 + }, + { + "epoch": 8.27, + "learning_rate": 8.683635569150099e-07, + "loss": 1.8106, + "step": 37871 + }, + { + "epoch": 8.27, + "learning_rate": 8.682543150535286e-07, + "loss": 0.6146, + "step": 37872 + }, + { + "epoch": 8.27, + "learning_rate": 8.681450731920473e-07, + "loss": 0.0623, + "step": 37873 + }, + { + "epoch": 8.27, + "learning_rate": 8.680358313305659e-07, + "loss": 0.6755, + "step": 37874 + }, + { + "epoch": 8.27, + "learning_rate": 8.679265894690846e-07, + "loss": 1.6655, + "step": 37875 + }, + { + "epoch": 8.27, + "learning_rate": 8.678173476076033e-07, + "loss": 1.3041, + "step": 37876 + }, + { + "epoch": 8.27, + "learning_rate": 8.67708105746122e-07, + "loss": 0.7638, + "step": 37877 + }, + { + "epoch": 8.27, + "learning_rate": 8.675988638846406e-07, + "loss": 0.5911, + "step": 37878 + }, + { + "epoch": 8.27, + "learning_rate": 8.674896220231594e-07, + "loss": 2.5415, + "step": 37879 + }, + { + "epoch": 8.27, + "learning_rate": 8.67380380161678e-07, + "loss": 1.8984, + "step": 37880 + }, + { + "epoch": 8.27, + "learning_rate": 8.672711383001967e-07, + "loss": 1.3871, + "step": 37881 + }, + { + "epoch": 8.27, + "learning_rate": 8.671618964387154e-07, + "loss": 0.4447, + "step": 37882 + }, + { + "epoch": 8.27, + "learning_rate": 8.670526545772341e-07, + "loss": 0.5245, + "step": 37883 + }, + { + "epoch": 8.27, + "learning_rate": 8.669434127157527e-07, + "loss": 0.8269, + "step": 37884 + }, + { + "epoch": 8.27, + "learning_rate": 8.668341708542715e-07, + "loss": 1.844, + "step": 37885 + }, + { + "epoch": 8.27, + "learning_rate": 8.667249289927901e-07, + "loss": 1.5253, + "step": 37886 + }, + { + "epoch": 8.27, + "learning_rate": 8.666156871313088e-07, + "loss": 1.6466, + "step": 37887 + }, + { + "epoch": 8.27, + "learning_rate": 8.665064452698274e-07, + "loss": 0.95, + "step": 37888 + }, + { + "epoch": 8.27, + "learning_rate": 8.663972034083462e-07, + "loss": 0.9936, + "step": 37889 + }, + { + "epoch": 8.27, + "learning_rate": 8.662879615468648e-07, + "loss": 1.6536, + "step": 37890 + }, + { + "epoch": 8.27, + "learning_rate": 8.661787196853835e-07, + "loss": 2.6397, + "step": 37891 + }, + { + "epoch": 8.27, + "learning_rate": 8.660694778239022e-07, + "loss": 0.3791, + "step": 37892 + }, + { + "epoch": 8.27, + "learning_rate": 8.659602359624209e-07, + "loss": 0.65, + "step": 37893 + }, + { + "epoch": 8.27, + "learning_rate": 8.658509941009395e-07, + "loss": 0.9738, + "step": 37894 + }, + { + "epoch": 8.27, + "learning_rate": 8.657417522394582e-07, + "loss": 0.9304, + "step": 37895 + }, + { + "epoch": 8.27, + "learning_rate": 8.656325103779769e-07, + "loss": 0.6716, + "step": 37896 + }, + { + "epoch": 8.27, + "learning_rate": 8.655232685164956e-07, + "loss": 0.7038, + "step": 37897 + }, + { + "epoch": 8.27, + "learning_rate": 8.654140266550142e-07, + "loss": 1.0869, + "step": 37898 + }, + { + "epoch": 8.27, + "learning_rate": 8.65304784793533e-07, + "loss": 0.1655, + "step": 37899 + }, + { + "epoch": 8.27, + "learning_rate": 8.651955429320516e-07, + "loss": 0.6975, + "step": 37900 + }, + { + "epoch": 8.27, + "learning_rate": 8.650863010705703e-07, + "loss": 1.9922, + "step": 37901 + }, + { + "epoch": 8.27, + "learning_rate": 8.649770592090891e-07, + "loss": 1.7971, + "step": 37902 + }, + { + "epoch": 8.27, + "learning_rate": 8.648678173476077e-07, + "loss": 0.8556, + "step": 37903 + }, + { + "epoch": 8.27, + "learning_rate": 8.647585754861263e-07, + "loss": 0.423, + "step": 37904 + }, + { + "epoch": 8.27, + "learning_rate": 8.64649333624645e-07, + "loss": 0.6985, + "step": 37905 + }, + { + "epoch": 8.27, + "learning_rate": 8.645400917631637e-07, + "loss": 1.5828, + "step": 37906 + }, + { + "epoch": 8.27, + "learning_rate": 8.644308499016824e-07, + "loss": 1.4072, + "step": 37907 + }, + { + "epoch": 8.27, + "learning_rate": 8.64321608040201e-07, + "loss": 1.954, + "step": 37908 + }, + { + "epoch": 8.27, + "learning_rate": 8.642123661787198e-07, + "loss": 1.6492, + "step": 37909 + }, + { + "epoch": 8.27, + "learning_rate": 8.641031243172384e-07, + "loss": 1.9108, + "step": 37910 + }, + { + "epoch": 8.27, + "learning_rate": 8.639938824557571e-07, + "loss": 1.9602, + "step": 37911 + }, + { + "epoch": 8.27, + "learning_rate": 8.638846405942757e-07, + "loss": 1.1689, + "step": 37912 + }, + { + "epoch": 8.27, + "learning_rate": 8.637753987327945e-07, + "loss": 0.1779, + "step": 37913 + }, + { + "epoch": 8.27, + "learning_rate": 8.636661568713131e-07, + "loss": 0.4125, + "step": 37914 + }, + { + "epoch": 8.27, + "learning_rate": 8.635569150098318e-07, + "loss": 1.1735, + "step": 37915 + }, + { + "epoch": 8.27, + "learning_rate": 8.634476731483506e-07, + "loss": 2.2355, + "step": 37916 + }, + { + "epoch": 8.28, + "learning_rate": 8.633384312868692e-07, + "loss": 0.6679, + "step": 37917 + }, + { + "epoch": 8.28, + "learning_rate": 8.632291894253878e-07, + "loss": 0.6705, + "step": 37918 + }, + { + "epoch": 8.28, + "learning_rate": 8.631199475639066e-07, + "loss": 2.3513, + "step": 37919 + }, + { + "epoch": 8.28, + "learning_rate": 8.630107057024252e-07, + "loss": 0.3177, + "step": 37920 + }, + { + "epoch": 8.28, + "learning_rate": 8.629014638409439e-07, + "loss": 0.791, + "step": 37921 + }, + { + "epoch": 8.28, + "learning_rate": 8.627922219794627e-07, + "loss": 2.1377, + "step": 37922 + }, + { + "epoch": 8.28, + "learning_rate": 8.626829801179813e-07, + "loss": 1.1558, + "step": 37923 + }, + { + "epoch": 8.28, + "learning_rate": 8.625737382565e-07, + "loss": 0.234, + "step": 37924 + }, + { + "epoch": 8.28, + "learning_rate": 8.624644963950186e-07, + "loss": 1.61, + "step": 37925 + }, + { + "epoch": 8.28, + "learning_rate": 8.623552545335373e-07, + "loss": 1.6106, + "step": 37926 + }, + { + "epoch": 8.28, + "learning_rate": 8.62246012672056e-07, + "loss": 0.7842, + "step": 37927 + }, + { + "epoch": 8.28, + "learning_rate": 8.621367708105746e-07, + "loss": 1.4435, + "step": 37928 + }, + { + "epoch": 8.28, + "learning_rate": 8.620275289490934e-07, + "loss": 1.9949, + "step": 37929 + }, + { + "epoch": 8.28, + "learning_rate": 8.619182870876121e-07, + "loss": 0.5792, + "step": 37930 + }, + { + "epoch": 8.28, + "learning_rate": 8.618090452261307e-07, + "loss": 0.2237, + "step": 37931 + }, + { + "epoch": 8.28, + "learning_rate": 8.616998033646493e-07, + "loss": 0.4338, + "step": 37932 + }, + { + "epoch": 8.28, + "learning_rate": 8.615905615031681e-07, + "loss": 0.1085, + "step": 37933 + }, + { + "epoch": 8.28, + "learning_rate": 8.614813196416867e-07, + "loss": 1.0581, + "step": 37934 + }, + { + "epoch": 8.28, + "learning_rate": 8.613720777802055e-07, + "loss": 1.2998, + "step": 37935 + }, + { + "epoch": 8.28, + "learning_rate": 8.612628359187242e-07, + "loss": 0.6398, + "step": 37936 + }, + { + "epoch": 8.28, + "learning_rate": 8.611535940572428e-07, + "loss": 0.717, + "step": 37937 + }, + { + "epoch": 8.28, + "learning_rate": 8.610443521957615e-07, + "loss": 1.1319, + "step": 37938 + }, + { + "epoch": 8.28, + "learning_rate": 8.609351103342802e-07, + "loss": 0.7831, + "step": 37939 + }, + { + "epoch": 8.28, + "learning_rate": 8.608258684727988e-07, + "loss": 1.0526, + "step": 37940 + }, + { + "epoch": 8.28, + "learning_rate": 8.607166266113175e-07, + "loss": 0.3562, + "step": 37941 + }, + { + "epoch": 8.28, + "learning_rate": 8.606073847498363e-07, + "loss": 1.921, + "step": 37942 + }, + { + "epoch": 8.28, + "learning_rate": 8.604981428883549e-07, + "loss": 0.7219, + "step": 37943 + }, + { + "epoch": 8.28, + "learning_rate": 8.603889010268736e-07, + "loss": 0.09, + "step": 37944 + }, + { + "epoch": 8.28, + "learning_rate": 8.602796591653922e-07, + "loss": 2.804, + "step": 37945 + }, + { + "epoch": 8.28, + "learning_rate": 8.601704173039109e-07, + "loss": 1.6535, + "step": 37946 + }, + { + "epoch": 8.28, + "learning_rate": 8.600611754424296e-07, + "loss": 0.601, + "step": 37947 + }, + { + "epoch": 8.28, + "learning_rate": 8.599519335809482e-07, + "loss": 0.8527, + "step": 37948 + }, + { + "epoch": 8.28, + "learning_rate": 8.59842691719467e-07, + "loss": 0.9258, + "step": 37949 + }, + { + "epoch": 8.28, + "learning_rate": 8.597334498579857e-07, + "loss": 0.9547, + "step": 37950 + }, + { + "epoch": 8.28, + "learning_rate": 8.596242079965043e-07, + "loss": 0.3282, + "step": 37951 + }, + { + "epoch": 8.28, + "learning_rate": 8.59514966135023e-07, + "loss": 0.4538, + "step": 37952 + }, + { + "epoch": 8.28, + "learning_rate": 8.594057242735417e-07, + "loss": 1.7901, + "step": 37953 + }, + { + "epoch": 8.28, + "learning_rate": 8.592964824120603e-07, + "loss": 1.0815, + "step": 37954 + }, + { + "epoch": 8.28, + "learning_rate": 8.591872405505791e-07, + "loss": 0.871, + "step": 37955 + }, + { + "epoch": 8.28, + "learning_rate": 8.590779986890978e-07, + "loss": 0.7146, + "step": 37956 + }, + { + "epoch": 8.28, + "learning_rate": 8.589687568276164e-07, + "loss": 0.6924, + "step": 37957 + }, + { + "epoch": 8.28, + "learning_rate": 8.588595149661351e-07, + "loss": 1.265, + "step": 37958 + }, + { + "epoch": 8.28, + "learning_rate": 8.587502731046537e-07, + "loss": 1.1019, + "step": 37959 + }, + { + "epoch": 8.28, + "learning_rate": 8.586410312431725e-07, + "loss": 0.873, + "step": 37960 + }, + { + "epoch": 8.28, + "learning_rate": 8.585317893816911e-07, + "loss": 2.1416, + "step": 37961 + }, + { + "epoch": 8.29, + "learning_rate": 8.584225475202099e-07, + "loss": 0.8538, + "step": 37962 + }, + { + "epoch": 8.29, + "learning_rate": 8.583133056587285e-07, + "loss": 0.687, + "step": 37963 + }, + { + "epoch": 8.29, + "learning_rate": 8.582040637972472e-07, + "loss": 3.515, + "step": 37964 + }, + { + "epoch": 8.29, + "learning_rate": 8.580948219357658e-07, + "loss": 1.9048, + "step": 37965 + }, + { + "epoch": 8.29, + "learning_rate": 8.579855800742846e-07, + "loss": 1.4883, + "step": 37966 + }, + { + "epoch": 8.29, + "learning_rate": 8.578763382128032e-07, + "loss": 0.1183, + "step": 37967 + }, + { + "epoch": 8.29, + "learning_rate": 8.57767096351322e-07, + "loss": 0.2006, + "step": 37968 + }, + { + "epoch": 8.29, + "learning_rate": 8.576578544898406e-07, + "loss": 0.9393, + "step": 37969 + }, + { + "epoch": 8.29, + "learning_rate": 8.575486126283593e-07, + "loss": 0.4764, + "step": 37970 + }, + { + "epoch": 8.29, + "learning_rate": 8.574393707668779e-07, + "loss": 1.4299, + "step": 37971 + }, + { + "epoch": 8.29, + "learning_rate": 8.573301289053966e-07, + "loss": 1.0689, + "step": 37972 + }, + { + "epoch": 8.29, + "learning_rate": 8.572208870439153e-07, + "loss": 1.4593, + "step": 37973 + }, + { + "epoch": 8.29, + "learning_rate": 8.57111645182434e-07, + "loss": 1.68, + "step": 37974 + }, + { + "epoch": 8.29, + "learning_rate": 8.570024033209527e-07, + "loss": 0.4506, + "step": 37975 + }, + { + "epoch": 8.29, + "learning_rate": 8.568931614594714e-07, + "loss": 0.8034, + "step": 37976 + }, + { + "epoch": 8.29, + "learning_rate": 8.5678391959799e-07, + "loss": 1.1786, + "step": 37977 + }, + { + "epoch": 8.29, + "learning_rate": 8.566746777365087e-07, + "loss": 1.8089, + "step": 37978 + }, + { + "epoch": 8.29, + "learning_rate": 8.565654358750273e-07, + "loss": 0.6338, + "step": 37979 + }, + { + "epoch": 8.29, + "learning_rate": 8.564561940135461e-07, + "loss": 1.5346, + "step": 37980 + }, + { + "epoch": 8.29, + "learning_rate": 8.563469521520647e-07, + "loss": 2.4665, + "step": 37981 + }, + { + "epoch": 8.29, + "learning_rate": 8.562377102905835e-07, + "loss": 2.1757, + "step": 37982 + }, + { + "epoch": 8.29, + "learning_rate": 8.561284684291021e-07, + "loss": 1.114, + "step": 37983 + }, + { + "epoch": 8.29, + "learning_rate": 8.560192265676208e-07, + "loss": 1.2591, + "step": 37984 + }, + { + "epoch": 8.29, + "learning_rate": 8.559099847061394e-07, + "loss": 2.7487, + "step": 37985 + }, + { + "epoch": 8.29, + "learning_rate": 8.558007428446581e-07, + "loss": 0.7662, + "step": 37986 + }, + { + "epoch": 8.29, + "learning_rate": 8.556915009831768e-07, + "loss": 0.6989, + "step": 37987 + }, + { + "epoch": 8.29, + "learning_rate": 8.555822591216956e-07, + "loss": 2.0263, + "step": 37988 + }, + { + "epoch": 8.29, + "learning_rate": 8.554730172602142e-07, + "loss": 0.2541, + "step": 37989 + }, + { + "epoch": 8.29, + "learning_rate": 8.553637753987329e-07, + "loss": 0.8449, + "step": 37990 + }, + { + "epoch": 8.29, + "learning_rate": 8.552545335372515e-07, + "loss": 0.9362, + "step": 37991 + }, + { + "epoch": 8.29, + "learning_rate": 8.551452916757702e-07, + "loss": 0.7948, + "step": 37992 + }, + { + "epoch": 8.29, + "learning_rate": 8.550360498142889e-07, + "loss": 0.9644, + "step": 37993 + }, + { + "epoch": 8.29, + "learning_rate": 8.549268079528076e-07, + "loss": 2.153, + "step": 37994 + }, + { + "epoch": 8.29, + "learning_rate": 8.548175660913263e-07, + "loss": 0.4975, + "step": 37995 + }, + { + "epoch": 8.29, + "learning_rate": 8.54708324229845e-07, + "loss": 1.5191, + "step": 37996 + }, + { + "epoch": 8.29, + "learning_rate": 8.545990823683636e-07, + "loss": 1.0083, + "step": 37997 + }, + { + "epoch": 8.29, + "learning_rate": 8.544898405068823e-07, + "loss": 1.0471, + "step": 37998 + }, + { + "epoch": 8.29, + "learning_rate": 8.543805986454009e-07, + "loss": 1.5121, + "step": 37999 + }, + { + "epoch": 8.29, + "learning_rate": 8.542713567839197e-07, + "loss": 0.4701, + "step": 38000 + }, + { + "epoch": 8.29, + "learning_rate": 8.541621149224383e-07, + "loss": 1.688, + "step": 38001 + }, + { + "epoch": 8.29, + "learning_rate": 8.540528730609571e-07, + "loss": 0.5855, + "step": 38002 + }, + { + "epoch": 8.29, + "learning_rate": 8.539436311994757e-07, + "loss": 0.1482, + "step": 38003 + }, + { + "epoch": 8.29, + "learning_rate": 8.538343893379944e-07, + "loss": 0.297, + "step": 38004 + }, + { + "epoch": 8.29, + "learning_rate": 8.53725147476513e-07, + "loss": 0.4991, + "step": 38005 + }, + { + "epoch": 8.29, + "learning_rate": 8.536159056150317e-07, + "loss": 0.3406, + "step": 38006 + }, + { + "epoch": 8.29, + "learning_rate": 8.535066637535504e-07, + "loss": 0.3419, + "step": 38007 + }, + { + "epoch": 8.3, + "learning_rate": 8.533974218920692e-07, + "loss": 0.2455, + "step": 38008 + }, + { + "epoch": 8.3, + "learning_rate": 8.532881800305878e-07, + "loss": 0.966, + "step": 38009 + }, + { + "epoch": 8.3, + "learning_rate": 8.531789381691065e-07, + "loss": 0.0903, + "step": 38010 + }, + { + "epoch": 8.3, + "learning_rate": 8.530696963076251e-07, + "loss": 0.2845, + "step": 38011 + }, + { + "epoch": 8.3, + "learning_rate": 8.529604544461438e-07, + "loss": 2.0257, + "step": 38012 + }, + { + "epoch": 8.3, + "learning_rate": 8.528512125846624e-07, + "loss": 0.6975, + "step": 38013 + }, + { + "epoch": 8.3, + "learning_rate": 8.527419707231812e-07, + "loss": 1.1946, + "step": 38014 + }, + { + "epoch": 8.3, + "learning_rate": 8.526327288616999e-07, + "loss": 1.1861, + "step": 38015 + }, + { + "epoch": 8.3, + "learning_rate": 8.525234870002186e-07, + "loss": 0.6407, + "step": 38016 + }, + { + "epoch": 8.3, + "learning_rate": 8.524142451387372e-07, + "loss": 0.4543, + "step": 38017 + }, + { + "epoch": 8.3, + "learning_rate": 8.523050032772559e-07, + "loss": 1.5765, + "step": 38018 + }, + { + "epoch": 8.3, + "learning_rate": 8.521957614157745e-07, + "loss": 1.1479, + "step": 38019 + }, + { + "epoch": 8.3, + "learning_rate": 8.520865195542933e-07, + "loss": 1.2881, + "step": 38020 + }, + { + "epoch": 8.3, + "learning_rate": 8.51977277692812e-07, + "loss": 0.7938, + "step": 38021 + }, + { + "epoch": 8.3, + "learning_rate": 8.518680358313307e-07, + "loss": 0.5683, + "step": 38022 + }, + { + "epoch": 8.3, + "learning_rate": 8.517587939698493e-07, + "loss": 0.2165, + "step": 38023 + }, + { + "epoch": 8.3, + "learning_rate": 8.51649552108368e-07, + "loss": 0.2365, + "step": 38024 + }, + { + "epoch": 8.3, + "learning_rate": 8.515403102468866e-07, + "loss": 0.7805, + "step": 38025 + }, + { + "epoch": 8.3, + "learning_rate": 8.514310683854053e-07, + "loss": 1.205, + "step": 38026 + }, + { + "epoch": 8.3, + "learning_rate": 8.51321826523924e-07, + "loss": 1.146, + "step": 38027 + }, + { + "epoch": 8.3, + "learning_rate": 8.512125846624428e-07, + "loss": 1.8221, + "step": 38028 + }, + { + "epoch": 8.3, + "learning_rate": 8.511033428009614e-07, + "loss": 2.0272, + "step": 38029 + }, + { + "epoch": 8.3, + "learning_rate": 8.509941009394801e-07, + "loss": 0.8788, + "step": 38030 + }, + { + "epoch": 8.3, + "learning_rate": 8.508848590779987e-07, + "loss": 1.2093, + "step": 38031 + }, + { + "epoch": 8.3, + "learning_rate": 8.507756172165174e-07, + "loss": 1.7005, + "step": 38032 + }, + { + "epoch": 8.3, + "learning_rate": 8.50666375355036e-07, + "loss": 0.6957, + "step": 38033 + }, + { + "epoch": 8.3, + "learning_rate": 8.505571334935548e-07, + "loss": 1.6157, + "step": 38034 + }, + { + "epoch": 8.3, + "learning_rate": 8.504478916320736e-07, + "loss": 0.8034, + "step": 38035 + }, + { + "epoch": 8.3, + "learning_rate": 8.503386497705922e-07, + "loss": 1.0412, + "step": 38036 + }, + { + "epoch": 8.3, + "learning_rate": 8.502294079091108e-07, + "loss": 0.1157, + "step": 38037 + }, + { + "epoch": 8.3, + "learning_rate": 8.501201660476295e-07, + "loss": 1.5864, + "step": 38038 + }, + { + "epoch": 8.3, + "learning_rate": 8.500109241861481e-07, + "loss": 1.2806, + "step": 38039 + }, + { + "epoch": 8.3, + "learning_rate": 8.499016823246668e-07, + "loss": 1.6302, + "step": 38040 + }, + { + "epoch": 8.3, + "learning_rate": 8.497924404631857e-07, + "loss": 1.653, + "step": 38041 + }, + { + "epoch": 8.3, + "learning_rate": 8.496831986017043e-07, + "loss": 1.107, + "step": 38042 + }, + { + "epoch": 8.3, + "learning_rate": 8.495739567402229e-07, + "loss": 0.0991, + "step": 38043 + }, + { + "epoch": 8.3, + "learning_rate": 8.494647148787416e-07, + "loss": 1.2797, + "step": 38044 + }, + { + "epoch": 8.3, + "learning_rate": 8.493554730172602e-07, + "loss": 0.5444, + "step": 38045 + }, + { + "epoch": 8.3, + "learning_rate": 8.492462311557789e-07, + "loss": 0.3922, + "step": 38046 + }, + { + "epoch": 8.3, + "learning_rate": 8.491369892942975e-07, + "loss": 2.1001, + "step": 38047 + }, + { + "epoch": 8.3, + "learning_rate": 8.490277474328164e-07, + "loss": 1.4739, + "step": 38048 + }, + { + "epoch": 8.3, + "learning_rate": 8.489185055713351e-07, + "loss": 0.3365, + "step": 38049 + }, + { + "epoch": 8.3, + "learning_rate": 8.488092637098537e-07, + "loss": 5.3125, + "step": 38050 + }, + { + "epoch": 8.3, + "learning_rate": 8.487000218483723e-07, + "loss": 0.6163, + "step": 38051 + }, + { + "epoch": 8.3, + "learning_rate": 8.48590779986891e-07, + "loss": 1.2348, + "step": 38052 + }, + { + "epoch": 8.3, + "learning_rate": 8.484815381254096e-07, + "loss": 2.1797, + "step": 38053 + }, + { + "epoch": 8.31, + "learning_rate": 8.483722962639285e-07, + "loss": 2.4087, + "step": 38054 + }, + { + "epoch": 8.31, + "learning_rate": 8.482630544024472e-07, + "loss": 1.006, + "step": 38055 + }, + { + "epoch": 8.31, + "learning_rate": 8.481538125409658e-07, + "loss": 0.1596, + "step": 38056 + }, + { + "epoch": 8.31, + "learning_rate": 8.480445706794844e-07, + "loss": 0.5618, + "step": 38057 + }, + { + "epoch": 8.31, + "learning_rate": 8.479353288180031e-07, + "loss": 0.265, + "step": 38058 + }, + { + "epoch": 8.31, + "learning_rate": 8.478260869565217e-07, + "loss": 2.6211, + "step": 38059 + }, + { + "epoch": 8.31, + "learning_rate": 8.477168450950404e-07, + "loss": 1.0504, + "step": 38060 + }, + { + "epoch": 8.31, + "learning_rate": 8.476076032335593e-07, + "loss": 0.3812, + "step": 38061 + }, + { + "epoch": 8.31, + "learning_rate": 8.474983613720779e-07, + "loss": 0.3976, + "step": 38062 + }, + { + "epoch": 8.31, + "learning_rate": 8.473891195105966e-07, + "loss": 1.5732, + "step": 38063 + }, + { + "epoch": 8.31, + "learning_rate": 8.472798776491152e-07, + "loss": 1.2863, + "step": 38064 + }, + { + "epoch": 8.31, + "learning_rate": 8.471706357876338e-07, + "loss": 0.6979, + "step": 38065 + }, + { + "epoch": 8.31, + "learning_rate": 8.470613939261525e-07, + "loss": 0.618, + "step": 38066 + }, + { + "epoch": 8.31, + "learning_rate": 8.469521520646711e-07, + "loss": 2.0095, + "step": 38067 + }, + { + "epoch": 8.31, + "learning_rate": 8.4684291020319e-07, + "loss": 1.5269, + "step": 38068 + }, + { + "epoch": 8.31, + "learning_rate": 8.467336683417087e-07, + "loss": 1.9749, + "step": 38069 + }, + { + "epoch": 8.31, + "learning_rate": 8.466244264802273e-07, + "loss": 0.844, + "step": 38070 + }, + { + "epoch": 8.31, + "learning_rate": 8.46515184618746e-07, + "loss": 0.7805, + "step": 38071 + }, + { + "epoch": 8.31, + "learning_rate": 8.464059427572646e-07, + "loss": 0.0712, + "step": 38072 + }, + { + "epoch": 8.31, + "learning_rate": 8.462967008957832e-07, + "loss": 1.1376, + "step": 38073 + }, + { + "epoch": 8.31, + "learning_rate": 8.461874590343021e-07, + "loss": 1.6044, + "step": 38074 + }, + { + "epoch": 8.31, + "learning_rate": 8.460782171728208e-07, + "loss": 1.4643, + "step": 38075 + }, + { + "epoch": 8.31, + "learning_rate": 8.459689753113394e-07, + "loss": 0.6541, + "step": 38076 + }, + { + "epoch": 8.31, + "learning_rate": 8.458597334498581e-07, + "loss": 1.4781, + "step": 38077 + }, + { + "epoch": 8.31, + "learning_rate": 8.457504915883767e-07, + "loss": 1.7564, + "step": 38078 + }, + { + "epoch": 8.31, + "learning_rate": 8.456412497268953e-07, + "loss": 0.6182, + "step": 38079 + }, + { + "epoch": 8.31, + "learning_rate": 8.45532007865414e-07, + "loss": 0.5853, + "step": 38080 + }, + { + "epoch": 8.31, + "learning_rate": 8.454227660039329e-07, + "loss": 0.7095, + "step": 38081 + }, + { + "epoch": 8.31, + "learning_rate": 8.453135241424515e-07, + "loss": 1.3295, + "step": 38082 + }, + { + "epoch": 8.31, + "learning_rate": 8.452042822809702e-07, + "loss": 0.067, + "step": 38083 + }, + { + "epoch": 8.31, + "learning_rate": 8.450950404194888e-07, + "loss": 0.741, + "step": 38084 + }, + { + "epoch": 8.31, + "learning_rate": 8.449857985580075e-07, + "loss": 1.0515, + "step": 38085 + }, + { + "epoch": 8.31, + "learning_rate": 8.448765566965261e-07, + "loss": 0.4974, + "step": 38086 + }, + { + "epoch": 8.31, + "learning_rate": 8.44767314835045e-07, + "loss": 1.2208, + "step": 38087 + }, + { + "epoch": 8.31, + "learning_rate": 8.446580729735636e-07, + "loss": 0.928, + "step": 38088 + }, + { + "epoch": 8.31, + "learning_rate": 8.445488311120823e-07, + "loss": 1.2148, + "step": 38089 + }, + { + "epoch": 8.31, + "learning_rate": 8.444395892506009e-07, + "loss": 1.1363, + "step": 38090 + }, + { + "epoch": 8.31, + "learning_rate": 8.443303473891196e-07, + "loss": 1.6268, + "step": 38091 + }, + { + "epoch": 8.31, + "learning_rate": 8.442211055276382e-07, + "loss": 1.7376, + "step": 38092 + }, + { + "epoch": 8.31, + "learning_rate": 8.441118636661568e-07, + "loss": 0.7463, + "step": 38093 + }, + { + "epoch": 8.31, + "learning_rate": 8.440026218046757e-07, + "loss": 0.1902, + "step": 38094 + }, + { + "epoch": 8.31, + "learning_rate": 8.438933799431944e-07, + "loss": 1.0603, + "step": 38095 + }, + { + "epoch": 8.31, + "learning_rate": 8.43784138081713e-07, + "loss": 1.1566, + "step": 38096 + }, + { + "epoch": 8.31, + "learning_rate": 8.436748962202317e-07, + "loss": 1.3256, + "step": 38097 + }, + { + "epoch": 8.31, + "learning_rate": 8.435656543587503e-07, + "loss": 1.338, + "step": 38098 + }, + { + "epoch": 8.31, + "learning_rate": 8.43456412497269e-07, + "loss": 1.8228, + "step": 38099 + }, + { + "epoch": 8.32, + "learning_rate": 8.433471706357876e-07, + "loss": 0.3706, + "step": 38100 + }, + { + "epoch": 8.32, + "learning_rate": 8.432379287743065e-07, + "loss": 0.9091, + "step": 38101 + }, + { + "epoch": 8.32, + "learning_rate": 8.431286869128251e-07, + "loss": 1.2702, + "step": 38102 + }, + { + "epoch": 8.32, + "learning_rate": 8.430194450513438e-07, + "loss": 0.7688, + "step": 38103 + }, + { + "epoch": 8.32, + "learning_rate": 8.429102031898624e-07, + "loss": 1.0526, + "step": 38104 + }, + { + "epoch": 8.32, + "learning_rate": 8.428009613283811e-07, + "loss": 0.898, + "step": 38105 + }, + { + "epoch": 8.32, + "learning_rate": 8.426917194668997e-07, + "loss": 2.6116, + "step": 38106 + }, + { + "epoch": 8.32, + "learning_rate": 8.425824776054185e-07, + "loss": 1.1726, + "step": 38107 + }, + { + "epoch": 8.32, + "learning_rate": 8.424732357439372e-07, + "loss": 0.1355, + "step": 38108 + }, + { + "epoch": 8.32, + "learning_rate": 8.423639938824559e-07, + "loss": 2.3256, + "step": 38109 + }, + { + "epoch": 8.32, + "learning_rate": 8.422547520209745e-07, + "loss": 0.3233, + "step": 38110 + }, + { + "epoch": 8.32, + "learning_rate": 8.421455101594932e-07, + "loss": 1.274, + "step": 38111 + }, + { + "epoch": 8.32, + "learning_rate": 8.420362682980118e-07, + "loss": 2.5765, + "step": 38112 + }, + { + "epoch": 8.32, + "learning_rate": 8.419270264365305e-07, + "loss": 0.6208, + "step": 38113 + }, + { + "epoch": 8.32, + "learning_rate": 8.418177845750493e-07, + "loss": 0.8308, + "step": 38114 + }, + { + "epoch": 8.32, + "learning_rate": 8.41708542713568e-07, + "loss": 1.5605, + "step": 38115 + }, + { + "epoch": 8.32, + "learning_rate": 8.415993008520866e-07, + "loss": 0.2166, + "step": 38116 + }, + { + "epoch": 8.32, + "learning_rate": 8.414900589906053e-07, + "loss": 2.4081, + "step": 38117 + }, + { + "epoch": 8.32, + "learning_rate": 8.413808171291239e-07, + "loss": 1.6779, + "step": 38118 + }, + { + "epoch": 8.32, + "learning_rate": 8.412715752676426e-07, + "loss": 1.8385, + "step": 38119 + }, + { + "epoch": 8.32, + "learning_rate": 8.411623334061612e-07, + "loss": 0.6961, + "step": 38120 + }, + { + "epoch": 8.32, + "learning_rate": 8.410530915446801e-07, + "loss": 1.299, + "step": 38121 + }, + { + "epoch": 8.32, + "learning_rate": 8.409438496831987e-07, + "loss": 2.316, + "step": 38122 + }, + { + "epoch": 8.32, + "learning_rate": 8.408346078217174e-07, + "loss": 2.1572, + "step": 38123 + }, + { + "epoch": 8.32, + "learning_rate": 8.40725365960236e-07, + "loss": 0.541, + "step": 38124 + }, + { + "epoch": 8.32, + "learning_rate": 8.406161240987547e-07, + "loss": 0.181, + "step": 38125 + }, + { + "epoch": 8.32, + "learning_rate": 8.405068822372733e-07, + "loss": 1.0895, + "step": 38126 + }, + { + "epoch": 8.32, + "learning_rate": 8.403976403757921e-07, + "loss": 0.1618, + "step": 38127 + }, + { + "epoch": 8.32, + "learning_rate": 8.402883985143108e-07, + "loss": 1.3001, + "step": 38128 + }, + { + "epoch": 8.32, + "learning_rate": 8.401791566528295e-07, + "loss": 1.4848, + "step": 38129 + }, + { + "epoch": 8.32, + "learning_rate": 8.400699147913481e-07, + "loss": 0.7184, + "step": 38130 + }, + { + "epoch": 8.32, + "learning_rate": 8.399606729298668e-07, + "loss": 0.4115, + "step": 38131 + }, + { + "epoch": 8.32, + "learning_rate": 8.398514310683854e-07, + "loss": 1.5425, + "step": 38132 + }, + { + "epoch": 8.32, + "learning_rate": 8.397421892069041e-07, + "loss": 0.8463, + "step": 38133 + }, + { + "epoch": 8.32, + "learning_rate": 8.396329473454228e-07, + "loss": 1.9657, + "step": 38134 + }, + { + "epoch": 8.32, + "learning_rate": 8.395237054839416e-07, + "loss": 1.3739, + "step": 38135 + }, + { + "epoch": 8.32, + "learning_rate": 8.394144636224602e-07, + "loss": 0.2874, + "step": 38136 + }, + { + "epoch": 8.32, + "learning_rate": 8.393052217609789e-07, + "loss": 0.9536, + "step": 38137 + }, + { + "epoch": 8.32, + "learning_rate": 8.391959798994975e-07, + "loss": 0.8168, + "step": 38138 + }, + { + "epoch": 8.32, + "learning_rate": 8.390867380380162e-07, + "loss": 0.7459, + "step": 38139 + }, + { + "epoch": 8.32, + "learning_rate": 8.389774961765349e-07, + "loss": 0.3578, + "step": 38140 + }, + { + "epoch": 8.32, + "learning_rate": 8.388682543150537e-07, + "loss": 1.8817, + "step": 38141 + }, + { + "epoch": 8.32, + "learning_rate": 8.387590124535723e-07, + "loss": 0.0932, + "step": 38142 + }, + { + "epoch": 8.32, + "learning_rate": 8.38649770592091e-07, + "loss": 1.2593, + "step": 38143 + }, + { + "epoch": 8.32, + "learning_rate": 8.385405287306096e-07, + "loss": 0.5557, + "step": 38144 + }, + { + "epoch": 8.32, + "learning_rate": 8.384312868691283e-07, + "loss": 2.3028, + "step": 38145 + }, + { + "epoch": 8.33, + "learning_rate": 8.383220450076469e-07, + "loss": 0.5356, + "step": 38146 + }, + { + "epoch": 8.33, + "learning_rate": 8.382128031461657e-07, + "loss": 0.289, + "step": 38147 + }, + { + "epoch": 8.33, + "learning_rate": 8.381035612846844e-07, + "loss": 0.9736, + "step": 38148 + }, + { + "epoch": 8.33, + "learning_rate": 8.379943194232031e-07, + "loss": 1.3222, + "step": 38149 + }, + { + "epoch": 8.33, + "learning_rate": 8.378850775617217e-07, + "loss": 1.8647, + "step": 38150 + }, + { + "epoch": 8.33, + "learning_rate": 8.377758357002404e-07, + "loss": 0.6687, + "step": 38151 + }, + { + "epoch": 8.33, + "learning_rate": 8.37666593838759e-07, + "loss": 1.6529, + "step": 38152 + }, + { + "epoch": 8.33, + "learning_rate": 8.375573519772777e-07, + "loss": 0.7752, + "step": 38153 + }, + { + "epoch": 8.33, + "learning_rate": 8.374481101157964e-07, + "loss": 1.3075, + "step": 38154 + }, + { + "epoch": 8.33, + "learning_rate": 8.373388682543152e-07, + "loss": 0.1303, + "step": 38155 + }, + { + "epoch": 8.33, + "learning_rate": 8.372296263928338e-07, + "loss": 1.4314, + "step": 38156 + }, + { + "epoch": 8.33, + "learning_rate": 8.371203845313525e-07, + "loss": 0.1073, + "step": 38157 + }, + { + "epoch": 8.33, + "learning_rate": 8.370111426698711e-07, + "loss": 1.8513, + "step": 38158 + }, + { + "epoch": 8.33, + "learning_rate": 8.369019008083898e-07, + "loss": 2.2696, + "step": 38159 + }, + { + "epoch": 8.33, + "learning_rate": 8.367926589469086e-07, + "loss": 1.9409, + "step": 38160 + }, + { + "epoch": 8.33, + "learning_rate": 8.366834170854272e-07, + "loss": 0.0682, + "step": 38161 + }, + { + "epoch": 8.33, + "learning_rate": 8.365741752239459e-07, + "loss": 0.0886, + "step": 38162 + }, + { + "epoch": 8.33, + "learning_rate": 8.364649333624646e-07, + "loss": 0.5749, + "step": 38163 + }, + { + "epoch": 8.33, + "learning_rate": 8.363556915009832e-07, + "loss": 0.1817, + "step": 38164 + }, + { + "epoch": 8.33, + "learning_rate": 8.362464496395019e-07, + "loss": 3.0974, + "step": 38165 + }, + { + "epoch": 8.33, + "learning_rate": 8.361372077780205e-07, + "loss": 1.2203, + "step": 38166 + }, + { + "epoch": 8.33, + "learning_rate": 8.360279659165393e-07, + "loss": 1.1475, + "step": 38167 + }, + { + "epoch": 8.33, + "learning_rate": 8.35918724055058e-07, + "loss": 3.684, + "step": 38168 + }, + { + "epoch": 8.33, + "learning_rate": 8.358094821935767e-07, + "loss": 0.8417, + "step": 38169 + }, + { + "epoch": 8.33, + "learning_rate": 8.357002403320953e-07, + "loss": 1.3966, + "step": 38170 + }, + { + "epoch": 8.33, + "learning_rate": 8.35590998470614e-07, + "loss": 3.7322, + "step": 38171 + }, + { + "epoch": 8.33, + "learning_rate": 8.354817566091326e-07, + "loss": 1.5134, + "step": 38172 + }, + { + "epoch": 8.33, + "learning_rate": 8.353725147476514e-07, + "loss": 0.7012, + "step": 38173 + }, + { + "epoch": 8.33, + "learning_rate": 8.352632728861701e-07, + "loss": 0.5327, + "step": 38174 + }, + { + "epoch": 8.33, + "learning_rate": 8.351540310246888e-07, + "loss": 1.661, + "step": 38175 + }, + { + "epoch": 8.33, + "learning_rate": 8.350447891632074e-07, + "loss": 0.1656, + "step": 38176 + }, + { + "epoch": 8.33, + "learning_rate": 8.349355473017261e-07, + "loss": 0.7962, + "step": 38177 + }, + { + "epoch": 8.33, + "learning_rate": 8.348263054402447e-07, + "loss": 0.8111, + "step": 38178 + }, + { + "epoch": 8.33, + "learning_rate": 8.347170635787634e-07, + "loss": 0.8945, + "step": 38179 + }, + { + "epoch": 8.33, + "learning_rate": 8.346078217172822e-07, + "loss": 0.1822, + "step": 38180 + }, + { + "epoch": 8.33, + "learning_rate": 8.344985798558008e-07, + "loss": 0.1293, + "step": 38181 + }, + { + "epoch": 8.33, + "learning_rate": 8.343893379943196e-07, + "loss": 0.6796, + "step": 38182 + }, + { + "epoch": 8.33, + "learning_rate": 8.342800961328382e-07, + "loss": 0.892, + "step": 38183 + }, + { + "epoch": 8.33, + "learning_rate": 8.341708542713568e-07, + "loss": 0.108, + "step": 38184 + }, + { + "epoch": 8.33, + "learning_rate": 8.340616124098755e-07, + "loss": 1.0914, + "step": 38185 + }, + { + "epoch": 8.33, + "learning_rate": 8.339523705483941e-07, + "loss": 1.5105, + "step": 38186 + }, + { + "epoch": 8.33, + "learning_rate": 8.338431286869129e-07, + "loss": 1.0654, + "step": 38187 + }, + { + "epoch": 8.33, + "learning_rate": 8.337338868254317e-07, + "loss": 1.4065, + "step": 38188 + }, + { + "epoch": 8.33, + "learning_rate": 8.336246449639503e-07, + "loss": 0.0717, + "step": 38189 + }, + { + "epoch": 8.33, + "learning_rate": 8.335154031024689e-07, + "loss": 1.3342, + "step": 38190 + }, + { + "epoch": 8.34, + "learning_rate": 8.334061612409876e-07, + "loss": 1.1388, + "step": 38191 + }, + { + "epoch": 8.34, + "learning_rate": 8.332969193795062e-07, + "loss": 0.7348, + "step": 38192 + }, + { + "epoch": 8.34, + "learning_rate": 8.33187677518025e-07, + "loss": 0.7404, + "step": 38193 + }, + { + "epoch": 8.34, + "learning_rate": 8.330784356565437e-07, + "loss": 0.9368, + "step": 38194 + }, + { + "epoch": 8.34, + "learning_rate": 8.329691937950624e-07, + "loss": 1.707, + "step": 38195 + }, + { + "epoch": 8.34, + "learning_rate": 8.328599519335811e-07, + "loss": 0.8921, + "step": 38196 + }, + { + "epoch": 8.34, + "learning_rate": 8.327507100720997e-07, + "loss": 1.0751, + "step": 38197 + }, + { + "epoch": 8.34, + "learning_rate": 8.326414682106183e-07, + "loss": 2.0481, + "step": 38198 + }, + { + "epoch": 8.34, + "learning_rate": 8.32532226349137e-07, + "loss": 1.2733, + "step": 38199 + }, + { + "epoch": 8.34, + "learning_rate": 8.324229844876558e-07, + "loss": 2.7427, + "step": 38200 + }, + { + "epoch": 8.34, + "learning_rate": 8.323137426261744e-07, + "loss": 1.8575, + "step": 38201 + }, + { + "epoch": 8.34, + "learning_rate": 8.322045007646932e-07, + "loss": 0.5643, + "step": 38202 + }, + { + "epoch": 8.34, + "learning_rate": 8.320952589032118e-07, + "loss": 1.0531, + "step": 38203 + }, + { + "epoch": 8.34, + "learning_rate": 8.319860170417304e-07, + "loss": 0.9895, + "step": 38204 + }, + { + "epoch": 8.34, + "learning_rate": 8.318767751802491e-07, + "loss": 1.7971, + "step": 38205 + }, + { + "epoch": 8.34, + "learning_rate": 8.317675333187679e-07, + "loss": 1.8757, + "step": 38206 + }, + { + "epoch": 8.34, + "learning_rate": 8.316582914572865e-07, + "loss": 0.8171, + "step": 38207 + }, + { + "epoch": 8.34, + "learning_rate": 8.315490495958052e-07, + "loss": 1.196, + "step": 38208 + }, + { + "epoch": 8.34, + "learning_rate": 8.314398077343239e-07, + "loss": 0.963, + "step": 38209 + }, + { + "epoch": 8.34, + "learning_rate": 8.313305658728426e-07, + "loss": 0.975, + "step": 38210 + }, + { + "epoch": 8.34, + "learning_rate": 8.312213240113612e-07, + "loss": 0.8372, + "step": 38211 + }, + { + "epoch": 8.34, + "learning_rate": 8.311120821498798e-07, + "loss": 1.8453, + "step": 38212 + }, + { + "epoch": 8.34, + "learning_rate": 8.310028402883986e-07, + "loss": 0.6384, + "step": 38213 + }, + { + "epoch": 8.34, + "learning_rate": 8.308935984269173e-07, + "loss": 1.319, + "step": 38214 + }, + { + "epoch": 8.34, + "learning_rate": 8.30784356565436e-07, + "loss": 1.2026, + "step": 38215 + }, + { + "epoch": 8.34, + "learning_rate": 8.306751147039547e-07, + "loss": 1.1386, + "step": 38216 + }, + { + "epoch": 8.34, + "learning_rate": 8.305658728424733e-07, + "loss": 1.6972, + "step": 38217 + }, + { + "epoch": 8.34, + "learning_rate": 8.30456630980992e-07, + "loss": 0.8296, + "step": 38218 + }, + { + "epoch": 8.34, + "learning_rate": 8.303473891195106e-07, + "loss": 1.433, + "step": 38219 + }, + { + "epoch": 8.34, + "learning_rate": 8.302381472580294e-07, + "loss": 1.625, + "step": 38220 + }, + { + "epoch": 8.34, + "learning_rate": 8.30128905396548e-07, + "loss": 0.4686, + "step": 38221 + }, + { + "epoch": 8.34, + "learning_rate": 8.300196635350668e-07, + "loss": 0.7786, + "step": 38222 + }, + { + "epoch": 8.34, + "learning_rate": 8.299104216735854e-07, + "loss": 1.0398, + "step": 38223 + }, + { + "epoch": 8.34, + "learning_rate": 8.298011798121041e-07, + "loss": 0.67, + "step": 38224 + }, + { + "epoch": 8.34, + "learning_rate": 8.296919379506227e-07, + "loss": 1.4622, + "step": 38225 + }, + { + "epoch": 8.34, + "learning_rate": 8.295826960891415e-07, + "loss": 2.0685, + "step": 38226 + }, + { + "epoch": 8.34, + "learning_rate": 8.294734542276601e-07, + "loss": 1.7537, + "step": 38227 + }, + { + "epoch": 8.34, + "learning_rate": 8.293642123661788e-07, + "loss": 2.6014, + "step": 38228 + }, + { + "epoch": 8.34, + "learning_rate": 8.292549705046975e-07, + "loss": 0.8411, + "step": 38229 + }, + { + "epoch": 8.34, + "learning_rate": 8.291457286432162e-07, + "loss": 1.4716, + "step": 38230 + }, + { + "epoch": 8.34, + "learning_rate": 8.290364867817348e-07, + "loss": 1.1699, + "step": 38231 + }, + { + "epoch": 8.34, + "learning_rate": 8.289272449202535e-07, + "loss": 1.0701, + "step": 38232 + }, + { + "epoch": 8.34, + "learning_rate": 8.288180030587722e-07, + "loss": 2.4296, + "step": 38233 + }, + { + "epoch": 8.34, + "learning_rate": 8.287087611972909e-07, + "loss": 1.849, + "step": 38234 + }, + { + "epoch": 8.34, + "learning_rate": 8.285995193358095e-07, + "loss": 1.1814, + "step": 38235 + }, + { + "epoch": 8.34, + "learning_rate": 8.284902774743283e-07, + "loss": 1.5096, + "step": 38236 + }, + { + "epoch": 8.35, + "learning_rate": 8.283810356128469e-07, + "loss": 0.8274, + "step": 38237 + }, + { + "epoch": 8.35, + "learning_rate": 8.282717937513656e-07, + "loss": 1.8718, + "step": 38238 + }, + { + "epoch": 8.35, + "learning_rate": 8.281625518898842e-07, + "loss": 1.8336, + "step": 38239 + }, + { + "epoch": 8.35, + "learning_rate": 8.28053310028403e-07, + "loss": 0.5968, + "step": 38240 + }, + { + "epoch": 8.35, + "learning_rate": 8.279440681669216e-07, + "loss": 1.4264, + "step": 38241 + }, + { + "epoch": 8.35, + "learning_rate": 8.278348263054404e-07, + "loss": 2.0116, + "step": 38242 + }, + { + "epoch": 8.35, + "learning_rate": 8.27725584443959e-07, + "loss": 0.0374, + "step": 38243 + }, + { + "epoch": 8.35, + "learning_rate": 8.276163425824777e-07, + "loss": 1.5535, + "step": 38244 + }, + { + "epoch": 8.35, + "learning_rate": 8.275071007209963e-07, + "loss": 0.6071, + "step": 38245 + }, + { + "epoch": 8.35, + "learning_rate": 8.273978588595151e-07, + "loss": 0.6869, + "step": 38246 + }, + { + "epoch": 8.35, + "learning_rate": 8.272886169980337e-07, + "loss": 0.2474, + "step": 38247 + }, + { + "epoch": 8.35, + "learning_rate": 8.271793751365524e-07, + "loss": 0.7235, + "step": 38248 + }, + { + "epoch": 8.35, + "learning_rate": 8.27070133275071e-07, + "loss": 0.4212, + "step": 38249 + }, + { + "epoch": 8.35, + "learning_rate": 8.269608914135898e-07, + "loss": 1.6587, + "step": 38250 + }, + { + "epoch": 8.35, + "learning_rate": 8.268516495521084e-07, + "loss": 0.465, + "step": 38251 + }, + { + "epoch": 8.35, + "learning_rate": 8.267424076906271e-07, + "loss": 0.119, + "step": 38252 + }, + { + "epoch": 8.35, + "learning_rate": 8.266331658291458e-07, + "loss": 0.6052, + "step": 38253 + }, + { + "epoch": 8.35, + "learning_rate": 8.265239239676645e-07, + "loss": 1.1074, + "step": 38254 + }, + { + "epoch": 8.35, + "learning_rate": 8.264146821061831e-07, + "loss": 0.646, + "step": 38255 + }, + { + "epoch": 8.35, + "learning_rate": 8.263054402447019e-07, + "loss": 0.8508, + "step": 38256 + }, + { + "epoch": 8.35, + "learning_rate": 8.261961983832205e-07, + "loss": 1.2055, + "step": 38257 + }, + { + "epoch": 8.35, + "learning_rate": 8.260869565217392e-07, + "loss": 1.4609, + "step": 38258 + }, + { + "epoch": 8.35, + "learning_rate": 8.259777146602579e-07, + "loss": 0.6986, + "step": 38259 + }, + { + "epoch": 8.35, + "learning_rate": 8.258684727987766e-07, + "loss": 1.7123, + "step": 38260 + }, + { + "epoch": 8.35, + "learning_rate": 8.257592309372952e-07, + "loss": 0.2896, + "step": 38261 + }, + { + "epoch": 8.35, + "learning_rate": 8.256499890758139e-07, + "loss": 1.6318, + "step": 38262 + }, + { + "epoch": 8.35, + "learning_rate": 8.255407472143326e-07, + "loss": 0.9369, + "step": 38263 + }, + { + "epoch": 8.35, + "learning_rate": 8.254315053528513e-07, + "loss": 1.1204, + "step": 38264 + }, + { + "epoch": 8.35, + "learning_rate": 8.253222634913699e-07, + "loss": 0.5779, + "step": 38265 + }, + { + "epoch": 8.35, + "learning_rate": 8.252130216298887e-07, + "loss": 2.7305, + "step": 38266 + }, + { + "epoch": 8.35, + "learning_rate": 8.251037797684073e-07, + "loss": 1.1541, + "step": 38267 + }, + { + "epoch": 8.35, + "learning_rate": 8.24994537906926e-07, + "loss": 0.5415, + "step": 38268 + }, + { + "epoch": 8.35, + "learning_rate": 8.248852960454446e-07, + "loss": 1.4884, + "step": 38269 + }, + { + "epoch": 8.35, + "learning_rate": 8.247760541839634e-07, + "loss": 0.5971, + "step": 38270 + }, + { + "epoch": 8.35, + "learning_rate": 8.24666812322482e-07, + "loss": 1.351, + "step": 38271 + }, + { + "epoch": 8.35, + "learning_rate": 8.245575704610007e-07, + "loss": 1.1579, + "step": 38272 + }, + { + "epoch": 8.35, + "learning_rate": 8.244483285995194e-07, + "loss": 0.6713, + "step": 38273 + }, + { + "epoch": 8.35, + "learning_rate": 8.243390867380381e-07, + "loss": 0.0806, + "step": 38274 + }, + { + "epoch": 8.35, + "learning_rate": 8.242298448765567e-07, + "loss": 0.798, + "step": 38275 + }, + { + "epoch": 8.35, + "learning_rate": 8.241206030150755e-07, + "loss": 1.7506, + "step": 38276 + }, + { + "epoch": 8.35, + "learning_rate": 8.240113611535941e-07, + "loss": 1.933, + "step": 38277 + }, + { + "epoch": 8.35, + "learning_rate": 8.239021192921128e-07, + "loss": 0.9032, + "step": 38278 + }, + { + "epoch": 8.35, + "learning_rate": 8.237928774306315e-07, + "loss": 0.2712, + "step": 38279 + }, + { + "epoch": 8.35, + "learning_rate": 8.236836355691502e-07, + "loss": 1.9807, + "step": 38280 + }, + { + "epoch": 8.35, + "learning_rate": 8.235743937076688e-07, + "loss": 1.2041, + "step": 38281 + }, + { + "epoch": 8.35, + "learning_rate": 8.234651518461875e-07, + "loss": 1.7138, + "step": 38282 + }, + { + "epoch": 8.36, + "learning_rate": 8.233559099847062e-07, + "loss": 1.3101, + "step": 38283 + }, + { + "epoch": 8.36, + "learning_rate": 8.232466681232249e-07, + "loss": 0.9891, + "step": 38284 + }, + { + "epoch": 8.36, + "learning_rate": 8.231374262617435e-07, + "loss": 1.4007, + "step": 38285 + }, + { + "epoch": 8.36, + "learning_rate": 8.230281844002623e-07, + "loss": 2.8902, + "step": 38286 + }, + { + "epoch": 8.36, + "learning_rate": 8.229189425387809e-07, + "loss": 0.6417, + "step": 38287 + }, + { + "epoch": 8.36, + "learning_rate": 8.228097006772996e-07, + "loss": 1.2578, + "step": 38288 + }, + { + "epoch": 8.36, + "learning_rate": 8.227004588158182e-07, + "loss": 1.7252, + "step": 38289 + }, + { + "epoch": 8.36, + "learning_rate": 8.22591216954337e-07, + "loss": 1.0149, + "step": 38290 + }, + { + "epoch": 8.36, + "learning_rate": 8.224819750928556e-07, + "loss": 0.4276, + "step": 38291 + }, + { + "epoch": 8.36, + "learning_rate": 8.223727332313744e-07, + "loss": 0.8843, + "step": 38292 + }, + { + "epoch": 8.36, + "learning_rate": 8.22263491369893e-07, + "loss": 0.7005, + "step": 38293 + }, + { + "epoch": 8.36, + "learning_rate": 8.221542495084117e-07, + "loss": 0.7819, + "step": 38294 + }, + { + "epoch": 8.36, + "learning_rate": 8.220450076469303e-07, + "loss": 3.4055, + "step": 38295 + }, + { + "epoch": 8.36, + "learning_rate": 8.21935765785449e-07, + "loss": 1.1227, + "step": 38296 + }, + { + "epoch": 8.36, + "learning_rate": 8.218265239239677e-07, + "loss": 0.6068, + "step": 38297 + }, + { + "epoch": 8.36, + "learning_rate": 8.217172820624864e-07, + "loss": 1.4292, + "step": 38298 + }, + { + "epoch": 8.36, + "learning_rate": 8.216080402010052e-07, + "loss": 0.2842, + "step": 38299 + }, + { + "epoch": 8.36, + "learning_rate": 8.214987983395238e-07, + "loss": 2.073, + "step": 38300 + }, + { + "epoch": 8.36, + "learning_rate": 8.213895564780424e-07, + "loss": 1.8085, + "step": 38301 + }, + { + "epoch": 8.36, + "learning_rate": 8.212803146165611e-07, + "loss": 0.6974, + "step": 38302 + }, + { + "epoch": 8.36, + "learning_rate": 8.211710727550798e-07, + "loss": 1.5513, + "step": 38303 + }, + { + "epoch": 8.36, + "learning_rate": 8.210618308935985e-07, + "loss": 1.9724, + "step": 38304 + }, + { + "epoch": 8.36, + "learning_rate": 8.209525890321171e-07, + "loss": 1.4994, + "step": 38305 + }, + { + "epoch": 8.36, + "learning_rate": 8.208433471706359e-07, + "loss": 0.475, + "step": 38306 + }, + { + "epoch": 8.36, + "learning_rate": 8.207341053091546e-07, + "loss": 0.4647, + "step": 38307 + }, + { + "epoch": 8.36, + "learning_rate": 8.206248634476732e-07, + "loss": 0.3907, + "step": 38308 + }, + { + "epoch": 8.36, + "learning_rate": 8.205156215861918e-07, + "loss": 0.1891, + "step": 38309 + }, + { + "epoch": 8.36, + "learning_rate": 8.204063797247106e-07, + "loss": 1.8169, + "step": 38310 + }, + { + "epoch": 8.36, + "learning_rate": 8.202971378632292e-07, + "loss": 2.077, + "step": 38311 + }, + { + "epoch": 8.36, + "learning_rate": 8.20187896001748e-07, + "loss": 1.1823, + "step": 38312 + }, + { + "epoch": 8.36, + "learning_rate": 8.200786541402667e-07, + "loss": 0.7362, + "step": 38313 + }, + { + "epoch": 8.36, + "learning_rate": 8.199694122787853e-07, + "loss": 1.0398, + "step": 38314 + }, + { + "epoch": 8.36, + "learning_rate": 8.198601704173039e-07, + "loss": 0.9261, + "step": 38315 + }, + { + "epoch": 8.36, + "learning_rate": 8.197509285558226e-07, + "loss": 1.1574, + "step": 38316 + }, + { + "epoch": 8.36, + "learning_rate": 8.196416866943413e-07, + "loss": 2.4775, + "step": 38317 + }, + { + "epoch": 8.36, + "learning_rate": 8.1953244483286e-07, + "loss": 1.3941, + "step": 38318 + }, + { + "epoch": 8.36, + "learning_rate": 8.194232029713788e-07, + "loss": 0.4995, + "step": 38319 + }, + { + "epoch": 8.36, + "learning_rate": 8.193139611098974e-07, + "loss": 0.1569, + "step": 38320 + }, + { + "epoch": 8.36, + "learning_rate": 8.192047192484161e-07, + "loss": 0.4393, + "step": 38321 + }, + { + "epoch": 8.36, + "learning_rate": 8.190954773869347e-07, + "loss": 0.3997, + "step": 38322 + }, + { + "epoch": 8.36, + "learning_rate": 8.189862355254533e-07, + "loss": 2.6865, + "step": 38323 + }, + { + "epoch": 8.36, + "learning_rate": 8.188769936639721e-07, + "loss": 0.3226, + "step": 38324 + }, + { + "epoch": 8.36, + "learning_rate": 8.187677518024907e-07, + "loss": 0.1253, + "step": 38325 + }, + { + "epoch": 8.36, + "learning_rate": 8.186585099410095e-07, + "loss": 0.6636, + "step": 38326 + }, + { + "epoch": 8.36, + "learning_rate": 8.185492680795282e-07, + "loss": 0.2692, + "step": 38327 + }, + { + "epoch": 8.36, + "learning_rate": 8.184400262180468e-07, + "loss": 1.4201, + "step": 38328 + }, + { + "epoch": 8.37, + "learning_rate": 8.183307843565654e-07, + "loss": 0.7037, + "step": 38329 + }, + { + "epoch": 8.37, + "learning_rate": 8.182215424950842e-07, + "loss": 1.7981, + "step": 38330 + }, + { + "epoch": 8.37, + "learning_rate": 8.181123006336028e-07, + "loss": 1.8759, + "step": 38331 + }, + { + "epoch": 8.37, + "learning_rate": 8.180030587721216e-07, + "loss": 1.1826, + "step": 38332 + }, + { + "epoch": 8.37, + "learning_rate": 8.178938169106403e-07, + "loss": 0.4409, + "step": 38333 + }, + { + "epoch": 8.37, + "learning_rate": 8.177845750491589e-07, + "loss": 1.2755, + "step": 38334 + }, + { + "epoch": 8.37, + "learning_rate": 8.176753331876776e-07, + "loss": 1.6943, + "step": 38335 + }, + { + "epoch": 8.37, + "learning_rate": 8.175660913261962e-07, + "loss": 1.79, + "step": 38336 + }, + { + "epoch": 8.37, + "learning_rate": 8.174568494647149e-07, + "loss": 0.7421, + "step": 38337 + }, + { + "epoch": 8.37, + "learning_rate": 8.173476076032336e-07, + "loss": 0.116, + "step": 38338 + }, + { + "epoch": 8.37, + "learning_rate": 8.172383657417524e-07, + "loss": 1.283, + "step": 38339 + }, + { + "epoch": 8.37, + "learning_rate": 8.17129123880271e-07, + "loss": 0.7584, + "step": 38340 + }, + { + "epoch": 8.37, + "learning_rate": 8.170198820187897e-07, + "loss": 2.3266, + "step": 38341 + }, + { + "epoch": 8.37, + "learning_rate": 8.169106401573083e-07, + "loss": 1.6515, + "step": 38342 + }, + { + "epoch": 8.37, + "learning_rate": 8.16801398295827e-07, + "loss": 1.3381, + "step": 38343 + }, + { + "epoch": 8.37, + "learning_rate": 8.166921564343457e-07, + "loss": 2.2995, + "step": 38344 + }, + { + "epoch": 8.37, + "learning_rate": 8.165829145728645e-07, + "loss": 0.9577, + "step": 38345 + }, + { + "epoch": 8.37, + "learning_rate": 8.164736727113831e-07, + "loss": 1.0861, + "step": 38346 + }, + { + "epoch": 8.37, + "learning_rate": 8.163644308499018e-07, + "loss": 0.5064, + "step": 38347 + }, + { + "epoch": 8.37, + "learning_rate": 8.162551889884204e-07, + "loss": 0.8965, + "step": 38348 + }, + { + "epoch": 8.37, + "learning_rate": 8.161459471269391e-07, + "loss": 1.1069, + "step": 38349 + }, + { + "epoch": 8.37, + "learning_rate": 8.160367052654577e-07, + "loss": 1.1954, + "step": 38350 + }, + { + "epoch": 8.37, + "learning_rate": 8.159274634039764e-07, + "loss": 0.7372, + "step": 38351 + }, + { + "epoch": 8.37, + "learning_rate": 8.158182215424952e-07, + "loss": 2.7561, + "step": 38352 + }, + { + "epoch": 8.37, + "learning_rate": 8.157089796810139e-07, + "loss": 0.9934, + "step": 38353 + }, + { + "epoch": 8.37, + "learning_rate": 8.155997378195325e-07, + "loss": 1.7696, + "step": 38354 + }, + { + "epoch": 8.37, + "learning_rate": 8.154904959580512e-07, + "loss": 0.6339, + "step": 38355 + }, + { + "epoch": 8.37, + "learning_rate": 8.153812540965698e-07, + "loss": 0.8859, + "step": 38356 + }, + { + "epoch": 8.37, + "learning_rate": 8.152720122350886e-07, + "loss": 1.0517, + "step": 38357 + }, + { + "epoch": 8.37, + "learning_rate": 8.151627703736072e-07, + "loss": 1.1159, + "step": 38358 + }, + { + "epoch": 8.37, + "learning_rate": 8.15053528512126e-07, + "loss": 0.5239, + "step": 38359 + }, + { + "epoch": 8.37, + "learning_rate": 8.149442866506446e-07, + "loss": 0.1143, + "step": 38360 + }, + { + "epoch": 8.37, + "learning_rate": 8.148350447891633e-07, + "loss": 0.2369, + "step": 38361 + }, + { + "epoch": 8.37, + "learning_rate": 8.147258029276819e-07, + "loss": 1.127, + "step": 38362 + }, + { + "epoch": 8.37, + "learning_rate": 8.146165610662006e-07, + "loss": 1.1885, + "step": 38363 + }, + { + "epoch": 8.37, + "learning_rate": 8.145073192047193e-07, + "loss": 1.2812, + "step": 38364 + }, + { + "epoch": 8.37, + "learning_rate": 8.143980773432381e-07, + "loss": 1.2965, + "step": 38365 + }, + { + "epoch": 8.37, + "learning_rate": 8.142888354817567e-07, + "loss": 1.4492, + "step": 38366 + }, + { + "epoch": 8.37, + "learning_rate": 8.141795936202754e-07, + "loss": 0.3362, + "step": 38367 + }, + { + "epoch": 8.37, + "learning_rate": 8.14070351758794e-07, + "loss": 1.1688, + "step": 38368 + }, + { + "epoch": 8.37, + "learning_rate": 8.139611098973127e-07, + "loss": 1.1966, + "step": 38369 + }, + { + "epoch": 8.37, + "learning_rate": 8.138518680358313e-07, + "loss": 0.8293, + "step": 38370 + }, + { + "epoch": 8.37, + "learning_rate": 8.137426261743501e-07, + "loss": 0.2505, + "step": 38371 + }, + { + "epoch": 8.37, + "learning_rate": 8.136333843128688e-07, + "loss": 1.552, + "step": 38372 + }, + { + "epoch": 8.37, + "learning_rate": 8.135241424513875e-07, + "loss": 0.276, + "step": 38373 + }, + { + "epoch": 8.37, + "learning_rate": 8.134149005899061e-07, + "loss": 0.5593, + "step": 38374 + }, + { + "epoch": 8.38, + "learning_rate": 8.133056587284248e-07, + "loss": 0.6472, + "step": 38375 + }, + { + "epoch": 8.38, + "learning_rate": 8.131964168669434e-07, + "loss": 1.4263, + "step": 38376 + }, + { + "epoch": 8.38, + "learning_rate": 8.130871750054622e-07, + "loss": 2.3277, + "step": 38377 + }, + { + "epoch": 8.38, + "learning_rate": 8.129779331439809e-07, + "loss": 0.8496, + "step": 38378 + }, + { + "epoch": 8.38, + "learning_rate": 8.128686912824996e-07, + "loss": 2.7958, + "step": 38379 + }, + { + "epoch": 8.38, + "learning_rate": 8.127594494210182e-07, + "loss": 0.0433, + "step": 38380 + }, + { + "epoch": 8.38, + "learning_rate": 8.126502075595369e-07, + "loss": 0.2533, + "step": 38381 + }, + { + "epoch": 8.38, + "learning_rate": 8.125409656980555e-07, + "loss": 0.7344, + "step": 38382 + }, + { + "epoch": 8.38, + "learning_rate": 8.124317238365742e-07, + "loss": 0.7271, + "step": 38383 + }, + { + "epoch": 8.38, + "learning_rate": 8.123224819750929e-07, + "loss": 1.644, + "step": 38384 + }, + { + "epoch": 8.38, + "learning_rate": 8.122132401136117e-07, + "loss": 2.5255, + "step": 38385 + }, + { + "epoch": 8.38, + "learning_rate": 8.121039982521303e-07, + "loss": 1.8395, + "step": 38386 + }, + { + "epoch": 8.38, + "learning_rate": 8.11994756390649e-07, + "loss": 1.0147, + "step": 38387 + }, + { + "epoch": 8.38, + "learning_rate": 8.118855145291676e-07, + "loss": 1.3948, + "step": 38388 + }, + { + "epoch": 8.38, + "learning_rate": 8.117762726676863e-07, + "loss": 0.7874, + "step": 38389 + }, + { + "epoch": 8.38, + "learning_rate": 8.116670308062049e-07, + "loss": 1.6019, + "step": 38390 + }, + { + "epoch": 8.38, + "learning_rate": 8.115577889447237e-07, + "loss": 0.648, + "step": 38391 + }, + { + "epoch": 8.38, + "learning_rate": 8.114485470832424e-07, + "loss": 0.5509, + "step": 38392 + }, + { + "epoch": 8.38, + "learning_rate": 8.113393052217611e-07, + "loss": 0.952, + "step": 38393 + }, + { + "epoch": 8.38, + "learning_rate": 8.112300633602797e-07, + "loss": 0.2759, + "step": 38394 + }, + { + "epoch": 8.38, + "learning_rate": 8.111208214987984e-07, + "loss": 1.6641, + "step": 38395 + }, + { + "epoch": 8.38, + "learning_rate": 8.11011579637317e-07, + "loss": 0.9631, + "step": 38396 + }, + { + "epoch": 8.38, + "learning_rate": 8.109023377758357e-07, + "loss": 0.745, + "step": 38397 + }, + { + "epoch": 8.38, + "learning_rate": 8.107930959143545e-07, + "loss": 1.0955, + "step": 38398 + }, + { + "epoch": 8.38, + "learning_rate": 8.106838540528732e-07, + "loss": 1.5503, + "step": 38399 + }, + { + "epoch": 8.38, + "learning_rate": 8.105746121913918e-07, + "loss": 0.4931, + "step": 38400 + }, + { + "epoch": 8.38, + "learning_rate": 8.104653703299105e-07, + "loss": 0.7711, + "step": 38401 + }, + { + "epoch": 8.38, + "learning_rate": 8.103561284684291e-07, + "loss": 2.125, + "step": 38402 + }, + { + "epoch": 8.38, + "learning_rate": 8.102468866069478e-07, + "loss": 2.0703, + "step": 38403 + }, + { + "epoch": 8.38, + "learning_rate": 8.101376447454664e-07, + "loss": 2.7445, + "step": 38404 + }, + { + "epoch": 8.38, + "learning_rate": 8.100284028839853e-07, + "loss": 0.6871, + "step": 38405 + }, + { + "epoch": 8.38, + "learning_rate": 8.099191610225039e-07, + "loss": 0.1338, + "step": 38406 + }, + { + "epoch": 8.38, + "learning_rate": 8.098099191610226e-07, + "loss": 0.4002, + "step": 38407 + }, + { + "epoch": 8.38, + "learning_rate": 8.097006772995412e-07, + "loss": 1.0113, + "step": 38408 + }, + { + "epoch": 8.38, + "learning_rate": 8.095914354380599e-07, + "loss": 2.1241, + "step": 38409 + }, + { + "epoch": 8.38, + "learning_rate": 8.094821935765785e-07, + "loss": 1.7926, + "step": 38410 + }, + { + "epoch": 8.38, + "learning_rate": 8.093729517150974e-07, + "loss": 2.0676, + "step": 38411 + }, + { + "epoch": 8.38, + "learning_rate": 8.09263709853616e-07, + "loss": 0.6577, + "step": 38412 + }, + { + "epoch": 8.38, + "learning_rate": 8.091544679921347e-07, + "loss": 1.403, + "step": 38413 + }, + { + "epoch": 8.38, + "learning_rate": 8.090452261306533e-07, + "loss": 1.1042, + "step": 38414 + }, + { + "epoch": 8.38, + "learning_rate": 8.08935984269172e-07, + "loss": 0.2887, + "step": 38415 + }, + { + "epoch": 8.38, + "learning_rate": 8.088267424076906e-07, + "loss": 1.9893, + "step": 38416 + }, + { + "epoch": 8.38, + "learning_rate": 8.087175005462093e-07, + "loss": 1.6355, + "step": 38417 + }, + { + "epoch": 8.38, + "learning_rate": 8.086082586847282e-07, + "loss": 1.8242, + "step": 38418 + }, + { + "epoch": 8.38, + "learning_rate": 8.084990168232468e-07, + "loss": 1.1645, + "step": 38419 + }, + { + "epoch": 8.38, + "learning_rate": 8.083897749617654e-07, + "loss": 1.1859, + "step": 38420 + }, + { + "epoch": 8.39, + "learning_rate": 8.082805331002841e-07, + "loss": 0.9388, + "step": 38421 + }, + { + "epoch": 8.39, + "learning_rate": 8.081712912388027e-07, + "loss": 1.167, + "step": 38422 + }, + { + "epoch": 8.39, + "learning_rate": 8.080620493773214e-07, + "loss": 2.593, + "step": 38423 + }, + { + "epoch": 8.39, + "learning_rate": 8.0795280751584e-07, + "loss": 0.6759, + "step": 38424 + }, + { + "epoch": 8.39, + "learning_rate": 8.078435656543589e-07, + "loss": 1.7636, + "step": 38425 + }, + { + "epoch": 8.39, + "learning_rate": 8.077343237928775e-07, + "loss": 1.6957, + "step": 38426 + }, + { + "epoch": 8.39, + "learning_rate": 8.076250819313962e-07, + "loss": 1.4015, + "step": 38427 + }, + { + "epoch": 8.39, + "learning_rate": 8.075158400699148e-07, + "loss": 1.6746, + "step": 38428 + }, + { + "epoch": 8.39, + "learning_rate": 8.074065982084335e-07, + "loss": 0.4792, + "step": 38429 + }, + { + "epoch": 8.39, + "learning_rate": 8.072973563469521e-07, + "loss": 0.9317, + "step": 38430 + }, + { + "epoch": 8.39, + "learning_rate": 8.07188114485471e-07, + "loss": 1.121, + "step": 38431 + }, + { + "epoch": 8.39, + "learning_rate": 8.070788726239897e-07, + "loss": 0.127, + "step": 38432 + }, + { + "epoch": 8.39, + "learning_rate": 8.069696307625083e-07, + "loss": 0.7523, + "step": 38433 + }, + { + "epoch": 8.39, + "learning_rate": 8.068603889010269e-07, + "loss": 1.1215, + "step": 38434 + }, + { + "epoch": 8.39, + "learning_rate": 8.067511470395456e-07, + "loss": 0.6446, + "step": 38435 + }, + { + "epoch": 8.39, + "learning_rate": 8.066419051780642e-07, + "loss": 0.7026, + "step": 38436 + }, + { + "epoch": 8.39, + "learning_rate": 8.065326633165829e-07, + "loss": 0.7271, + "step": 38437 + }, + { + "epoch": 8.39, + "learning_rate": 8.064234214551018e-07, + "loss": 0.2403, + "step": 38438 + }, + { + "epoch": 8.39, + "learning_rate": 8.063141795936204e-07, + "loss": 1.0286, + "step": 38439 + }, + { + "epoch": 8.39, + "learning_rate": 8.06204937732139e-07, + "loss": 0.7946, + "step": 38440 + }, + { + "epoch": 8.39, + "learning_rate": 8.060956958706577e-07, + "loss": 0.9899, + "step": 38441 + }, + { + "epoch": 8.39, + "learning_rate": 8.059864540091763e-07, + "loss": 1.3591, + "step": 38442 + }, + { + "epoch": 8.39, + "learning_rate": 8.05877212147695e-07, + "loss": 0.8033, + "step": 38443 + }, + { + "epoch": 8.39, + "learning_rate": 8.057679702862136e-07, + "loss": 1.8089, + "step": 38444 + }, + { + "epoch": 8.39, + "learning_rate": 8.056587284247325e-07, + "loss": 2.1134, + "step": 38445 + }, + { + "epoch": 8.39, + "learning_rate": 8.055494865632512e-07, + "loss": 2.595, + "step": 38446 + }, + { + "epoch": 8.39, + "learning_rate": 8.054402447017698e-07, + "loss": 2.1573, + "step": 38447 + }, + { + "epoch": 8.39, + "learning_rate": 8.053310028402884e-07, + "loss": 0.1738, + "step": 38448 + }, + { + "epoch": 8.39, + "learning_rate": 8.052217609788071e-07, + "loss": 0.6851, + "step": 38449 + }, + { + "epoch": 8.39, + "learning_rate": 8.051125191173257e-07, + "loss": 0.6015, + "step": 38450 + }, + { + "epoch": 8.39, + "learning_rate": 8.050032772558446e-07, + "loss": 1.858, + "step": 38451 + }, + { + "epoch": 8.39, + "learning_rate": 8.048940353943633e-07, + "loss": 0.9501, + "step": 38452 + }, + { + "epoch": 8.39, + "learning_rate": 8.047847935328819e-07, + "loss": 1.1571, + "step": 38453 + }, + { + "epoch": 8.39, + "learning_rate": 8.046755516714006e-07, + "loss": 1.4728, + "step": 38454 + }, + { + "epoch": 8.39, + "learning_rate": 8.045663098099192e-07, + "loss": 0.2623, + "step": 38455 + }, + { + "epoch": 8.39, + "learning_rate": 8.044570679484378e-07, + "loss": 1.2546, + "step": 38456 + }, + { + "epoch": 8.39, + "learning_rate": 8.043478260869565e-07, + "loss": 1.9028, + "step": 38457 + }, + { + "epoch": 8.39, + "learning_rate": 8.042385842254754e-07, + "loss": 1.0034, + "step": 38458 + }, + { + "epoch": 8.39, + "learning_rate": 8.04129342363994e-07, + "loss": 1.1959, + "step": 38459 + }, + { + "epoch": 8.39, + "learning_rate": 8.040201005025127e-07, + "loss": 0.9239, + "step": 38460 + }, + { + "epoch": 8.39, + "learning_rate": 8.039108586410313e-07, + "loss": 0.7861, + "step": 38461 + }, + { + "epoch": 8.39, + "learning_rate": 8.038016167795499e-07, + "loss": 0.8069, + "step": 38462 + }, + { + "epoch": 8.39, + "learning_rate": 8.036923749180686e-07, + "loss": 1.1817, + "step": 38463 + }, + { + "epoch": 8.39, + "learning_rate": 8.035831330565874e-07, + "loss": 2.0779, + "step": 38464 + }, + { + "epoch": 8.39, + "learning_rate": 8.034738911951061e-07, + "loss": 0.0561, + "step": 38465 + }, + { + "epoch": 8.4, + "learning_rate": 8.033646493336248e-07, + "loss": 0.2557, + "step": 38466 + }, + { + "epoch": 8.4, + "learning_rate": 8.032554074721434e-07, + "loss": 0.5601, + "step": 38467 + }, + { + "epoch": 8.4, + "learning_rate": 8.031461656106621e-07, + "loss": 0.5346, + "step": 38468 + }, + { + "epoch": 8.4, + "learning_rate": 8.030369237491807e-07, + "loss": 0.5253, + "step": 38469 + }, + { + "epoch": 8.4, + "learning_rate": 8.029276818876993e-07, + "loss": 1.0376, + "step": 38470 + }, + { + "epoch": 8.4, + "learning_rate": 8.028184400262181e-07, + "loss": 2.5843, + "step": 38471 + }, + { + "epoch": 8.4, + "learning_rate": 8.027091981647369e-07, + "loss": 0.6268, + "step": 38472 + }, + { + "epoch": 8.4, + "learning_rate": 8.025999563032555e-07, + "loss": 0.7392, + "step": 38473 + }, + { + "epoch": 8.4, + "learning_rate": 8.024907144417742e-07, + "loss": 0.6616, + "step": 38474 + }, + { + "epoch": 8.4, + "learning_rate": 8.023814725802928e-07, + "loss": 1.6152, + "step": 38475 + }, + { + "epoch": 8.4, + "learning_rate": 8.022722307188114e-07, + "loss": 0.4019, + "step": 38476 + }, + { + "epoch": 8.4, + "learning_rate": 8.021629888573301e-07, + "loss": 0.0629, + "step": 38477 + }, + { + "epoch": 8.4, + "learning_rate": 8.02053746995849e-07, + "loss": 1.0883, + "step": 38478 + }, + { + "epoch": 8.4, + "learning_rate": 8.019445051343676e-07, + "loss": 0.9524, + "step": 38479 + }, + { + "epoch": 8.4, + "learning_rate": 8.018352632728863e-07, + "loss": 1.024, + "step": 38480 + }, + { + "epoch": 8.4, + "learning_rate": 8.017260214114049e-07, + "loss": 0.3793, + "step": 38481 + }, + { + "epoch": 8.4, + "learning_rate": 8.016167795499236e-07, + "loss": 1.5228, + "step": 38482 + }, + { + "epoch": 8.4, + "learning_rate": 8.015075376884422e-07, + "loss": 0.9368, + "step": 38483 + }, + { + "epoch": 8.4, + "learning_rate": 8.01398295826961e-07, + "loss": 1.3752, + "step": 38484 + }, + { + "epoch": 8.4, + "learning_rate": 8.012890539654797e-07, + "loss": 2.7327, + "step": 38485 + }, + { + "epoch": 8.4, + "learning_rate": 8.011798121039984e-07, + "loss": 1.9967, + "step": 38486 + }, + { + "epoch": 8.4, + "learning_rate": 8.01070570242517e-07, + "loss": 0.9716, + "step": 38487 + }, + { + "epoch": 8.4, + "learning_rate": 8.009613283810357e-07, + "loss": 0.6781, + "step": 38488 + }, + { + "epoch": 8.4, + "learning_rate": 8.008520865195543e-07, + "loss": 0.1637, + "step": 38489 + }, + { + "epoch": 8.4, + "learning_rate": 8.00742844658073e-07, + "loss": 1.4509, + "step": 38490 + }, + { + "epoch": 8.4, + "learning_rate": 8.006336027965917e-07, + "loss": 1.3189, + "step": 38491 + }, + { + "epoch": 8.4, + "learning_rate": 8.005243609351105e-07, + "loss": 1.1021, + "step": 38492 + }, + { + "epoch": 8.4, + "learning_rate": 8.004151190736291e-07, + "loss": 0.9087, + "step": 38493 + }, + { + "epoch": 8.4, + "learning_rate": 8.003058772121478e-07, + "loss": 1.5053, + "step": 38494 + }, + { + "epoch": 8.4, + "learning_rate": 8.001966353506664e-07, + "loss": 0.5997, + "step": 38495 + }, + { + "epoch": 8.4, + "learning_rate": 8.000873934891851e-07, + "loss": 0.6828, + "step": 38496 + }, + { + "epoch": 8.4, + "learning_rate": 7.999781516277038e-07, + "loss": 0.5372, + "step": 38497 + }, + { + "epoch": 8.4, + "learning_rate": 7.998689097662226e-07, + "loss": 0.505, + "step": 38498 + }, + { + "epoch": 8.4, + "learning_rate": 7.997596679047412e-07, + "loss": 1.2716, + "step": 38499 + }, + { + "epoch": 8.4, + "learning_rate": 7.996504260432599e-07, + "loss": 3.1051, + "step": 38500 + }, + { + "epoch": 8.4, + "learning_rate": 7.995411841817785e-07, + "loss": 1.1223, + "step": 38501 + }, + { + "epoch": 8.4, + "learning_rate": 7.994319423202972e-07, + "loss": 0.4179, + "step": 38502 + }, + { + "epoch": 8.4, + "learning_rate": 7.993227004588158e-07, + "loss": 0.9829, + "step": 38503 + }, + { + "epoch": 8.4, + "learning_rate": 7.992134585973346e-07, + "loss": 0.5957, + "step": 38504 + }, + { + "epoch": 8.4, + "learning_rate": 7.991042167358533e-07, + "loss": 1.1312, + "step": 38505 + }, + { + "epoch": 8.4, + "learning_rate": 7.98994974874372e-07, + "loss": 0.44, + "step": 38506 + }, + { + "epoch": 8.4, + "learning_rate": 7.988857330128906e-07, + "loss": 1.3047, + "step": 38507 + }, + { + "epoch": 8.4, + "learning_rate": 7.987764911514093e-07, + "loss": 1.6407, + "step": 38508 + }, + { + "epoch": 8.4, + "learning_rate": 7.986672492899279e-07, + "loss": 1.4964, + "step": 38509 + }, + { + "epoch": 8.4, + "learning_rate": 7.985580074284466e-07, + "loss": 1.6461, + "step": 38510 + }, + { + "epoch": 8.4, + "learning_rate": 7.984487655669653e-07, + "loss": 1.9167, + "step": 38511 + }, + { + "epoch": 8.41, + "learning_rate": 7.983395237054841e-07, + "loss": 1.4644, + "step": 38512 + }, + { + "epoch": 8.41, + "learning_rate": 7.982302818440027e-07, + "loss": 1.5974, + "step": 38513 + }, + { + "epoch": 8.41, + "learning_rate": 7.981210399825214e-07, + "loss": 0.7618, + "step": 38514 + }, + { + "epoch": 8.41, + "learning_rate": 7.9801179812104e-07, + "loss": 0.4391, + "step": 38515 + }, + { + "epoch": 8.41, + "learning_rate": 7.979025562595587e-07, + "loss": 1.7591, + "step": 38516 + }, + { + "epoch": 8.41, + "learning_rate": 7.977933143980774e-07, + "loss": 1.4229, + "step": 38517 + }, + { + "epoch": 8.41, + "learning_rate": 7.976840725365961e-07, + "loss": 1.1453, + "step": 38518 + }, + { + "epoch": 8.41, + "learning_rate": 7.975748306751148e-07, + "loss": 1.4498, + "step": 38519 + }, + { + "epoch": 8.41, + "learning_rate": 7.974655888136335e-07, + "loss": 0.8093, + "step": 38520 + }, + { + "epoch": 8.41, + "learning_rate": 7.973563469521521e-07, + "loss": 1.2241, + "step": 38521 + }, + { + "epoch": 8.41, + "learning_rate": 7.972471050906708e-07, + "loss": 1.1231, + "step": 38522 + }, + { + "epoch": 8.41, + "learning_rate": 7.971378632291894e-07, + "loss": 2.0635, + "step": 38523 + }, + { + "epoch": 8.41, + "learning_rate": 7.970286213677082e-07, + "loss": 0.1471, + "step": 38524 + }, + { + "epoch": 8.41, + "learning_rate": 7.969193795062268e-07, + "loss": 1.1448, + "step": 38525 + }, + { + "epoch": 8.41, + "learning_rate": 7.968101376447456e-07, + "loss": 0.1184, + "step": 38526 + }, + { + "epoch": 8.41, + "learning_rate": 7.967008957832642e-07, + "loss": 0.7554, + "step": 38527 + }, + { + "epoch": 8.41, + "learning_rate": 7.965916539217829e-07, + "loss": 0.1796, + "step": 38528 + }, + { + "epoch": 8.41, + "learning_rate": 7.964824120603015e-07, + "loss": 1.1674, + "step": 38529 + }, + { + "epoch": 8.41, + "learning_rate": 7.963731701988203e-07, + "loss": 1.2127, + "step": 38530 + }, + { + "epoch": 8.41, + "learning_rate": 7.962639283373389e-07, + "loss": 2.1027, + "step": 38531 + }, + { + "epoch": 8.41, + "learning_rate": 7.961546864758577e-07, + "loss": 1.2852, + "step": 38532 + }, + { + "epoch": 8.41, + "learning_rate": 7.960454446143763e-07, + "loss": 0.7003, + "step": 38533 + }, + { + "epoch": 8.41, + "learning_rate": 7.95936202752895e-07, + "loss": 1.3425, + "step": 38534 + }, + { + "epoch": 8.41, + "learning_rate": 7.958269608914136e-07, + "loss": 0.2338, + "step": 38535 + }, + { + "epoch": 8.41, + "learning_rate": 7.957177190299323e-07, + "loss": 0.9424, + "step": 38536 + }, + { + "epoch": 8.41, + "learning_rate": 7.95608477168451e-07, + "loss": 1.9796, + "step": 38537 + }, + { + "epoch": 8.41, + "learning_rate": 7.954992353069697e-07, + "loss": 0.6585, + "step": 38538 + }, + { + "epoch": 8.41, + "learning_rate": 7.953899934454884e-07, + "loss": 0.2831, + "step": 38539 + }, + { + "epoch": 8.41, + "learning_rate": 7.952807515840071e-07, + "loss": 1.7235, + "step": 38540 + }, + { + "epoch": 8.41, + "learning_rate": 7.951715097225257e-07, + "loss": 0.8227, + "step": 38541 + }, + { + "epoch": 8.41, + "learning_rate": 7.950622678610444e-07, + "loss": 1.3796, + "step": 38542 + }, + { + "epoch": 8.41, + "learning_rate": 7.94953025999563e-07, + "loss": 0.3823, + "step": 38543 + }, + { + "epoch": 8.41, + "learning_rate": 7.948437841380818e-07, + "loss": 0.8482, + "step": 38544 + }, + { + "epoch": 8.41, + "learning_rate": 7.947345422766004e-07, + "loss": 0.4415, + "step": 38545 + }, + { + "epoch": 8.41, + "learning_rate": 7.946253004151192e-07, + "loss": 0.7697, + "step": 38546 + }, + { + "epoch": 8.41, + "learning_rate": 7.945160585536378e-07, + "loss": 0.7972, + "step": 38547 + }, + { + "epoch": 8.41, + "learning_rate": 7.944068166921565e-07, + "loss": 1.8374, + "step": 38548 + }, + { + "epoch": 8.41, + "learning_rate": 7.942975748306751e-07, + "loss": 0.9429, + "step": 38549 + }, + { + "epoch": 8.41, + "learning_rate": 7.941883329691939e-07, + "loss": 1.0535, + "step": 38550 + }, + { + "epoch": 8.41, + "learning_rate": 7.940790911077125e-07, + "loss": 0.5574, + "step": 38551 + }, + { + "epoch": 8.41, + "learning_rate": 7.939698492462313e-07, + "loss": 2.1737, + "step": 38552 + }, + { + "epoch": 8.41, + "learning_rate": 7.938606073847499e-07, + "loss": 0.0988, + "step": 38553 + }, + { + "epoch": 8.41, + "learning_rate": 7.937513655232686e-07, + "loss": 0.1132, + "step": 38554 + }, + { + "epoch": 8.41, + "learning_rate": 7.936421236617872e-07, + "loss": 2.2557, + "step": 38555 + }, + { + "epoch": 8.41, + "learning_rate": 7.935328818003059e-07, + "loss": 0.6283, + "step": 38556 + }, + { + "epoch": 8.41, + "learning_rate": 7.934236399388247e-07, + "loss": 0.1799, + "step": 38557 + }, + { + "epoch": 8.42, + "learning_rate": 7.933143980773433e-07, + "loss": 0.62, + "step": 38558 + }, + { + "epoch": 8.42, + "learning_rate": 7.93205156215862e-07, + "loss": 1.3662, + "step": 38559 + }, + { + "epoch": 8.42, + "learning_rate": 7.930959143543807e-07, + "loss": 0.3439, + "step": 38560 + }, + { + "epoch": 8.42, + "learning_rate": 7.929866724928993e-07, + "loss": 1.8826, + "step": 38561 + }, + { + "epoch": 8.42, + "learning_rate": 7.92877430631418e-07, + "loss": 0.6142, + "step": 38562 + }, + { + "epoch": 8.42, + "learning_rate": 7.927681887699366e-07, + "loss": 1.971, + "step": 38563 + }, + { + "epoch": 8.42, + "learning_rate": 7.926589469084554e-07, + "loss": 2.4014, + "step": 38564 + }, + { + "epoch": 8.42, + "learning_rate": 7.925497050469741e-07, + "loss": 1.9413, + "step": 38565 + }, + { + "epoch": 8.42, + "learning_rate": 7.924404631854928e-07, + "loss": 2.7003, + "step": 38566 + }, + { + "epoch": 8.42, + "learning_rate": 7.923312213240114e-07, + "loss": 0.877, + "step": 38567 + }, + { + "epoch": 8.42, + "learning_rate": 7.922219794625301e-07, + "loss": 1.6432, + "step": 38568 + }, + { + "epoch": 8.42, + "learning_rate": 7.921127376010487e-07, + "loss": 0.1114, + "step": 38569 + }, + { + "epoch": 8.42, + "learning_rate": 7.920034957395675e-07, + "loss": 1.9111, + "step": 38570 + }, + { + "epoch": 8.42, + "learning_rate": 7.918942538780862e-07, + "loss": 1.2814, + "step": 38571 + }, + { + "epoch": 8.42, + "learning_rate": 7.917850120166048e-07, + "loss": 0.2767, + "step": 38572 + }, + { + "epoch": 8.42, + "learning_rate": 7.916757701551235e-07, + "loss": 0.8436, + "step": 38573 + }, + { + "epoch": 8.42, + "learning_rate": 7.915665282936422e-07, + "loss": 0.9062, + "step": 38574 + }, + { + "epoch": 8.42, + "learning_rate": 7.914572864321608e-07, + "loss": 1.3375, + "step": 38575 + }, + { + "epoch": 8.42, + "learning_rate": 7.913480445706795e-07, + "loss": 1.8991, + "step": 38576 + }, + { + "epoch": 8.42, + "learning_rate": 7.912388027091983e-07, + "loss": 2.0254, + "step": 38577 + }, + { + "epoch": 8.42, + "learning_rate": 7.911295608477169e-07, + "loss": 1.0001, + "step": 38578 + }, + { + "epoch": 8.42, + "learning_rate": 7.910203189862357e-07, + "loss": 0.7599, + "step": 38579 + }, + { + "epoch": 8.42, + "learning_rate": 7.909110771247543e-07, + "loss": 0.8295, + "step": 38580 + }, + { + "epoch": 8.42, + "learning_rate": 7.908018352632729e-07, + "loss": 2.7254, + "step": 38581 + }, + { + "epoch": 8.42, + "learning_rate": 7.906925934017916e-07, + "loss": 2.0787, + "step": 38582 + }, + { + "epoch": 8.42, + "learning_rate": 7.905833515403104e-07, + "loss": 1.4923, + "step": 38583 + }, + { + "epoch": 8.42, + "learning_rate": 7.90474109678829e-07, + "loss": 1.0763, + "step": 38584 + }, + { + "epoch": 8.42, + "learning_rate": 7.903648678173477e-07, + "loss": 1.2191, + "step": 38585 + }, + { + "epoch": 8.42, + "learning_rate": 7.902556259558664e-07, + "loss": 1.045, + "step": 38586 + }, + { + "epoch": 8.42, + "learning_rate": 7.90146384094385e-07, + "loss": 0.9688, + "step": 38587 + }, + { + "epoch": 8.42, + "learning_rate": 7.900371422329037e-07, + "loss": 0.8406, + "step": 38588 + }, + { + "epoch": 8.42, + "learning_rate": 7.899279003714223e-07, + "loss": 0.8493, + "step": 38589 + }, + { + "epoch": 8.42, + "learning_rate": 7.898186585099411e-07, + "loss": 1.701, + "step": 38590 + }, + { + "epoch": 8.42, + "learning_rate": 7.897094166484598e-07, + "loss": 0.1776, + "step": 38591 + }, + { + "epoch": 8.42, + "learning_rate": 7.896001747869784e-07, + "loss": 1.0096, + "step": 38592 + }, + { + "epoch": 8.42, + "learning_rate": 7.894909329254972e-07, + "loss": 2.6175, + "step": 38593 + }, + { + "epoch": 8.42, + "learning_rate": 7.893816910640158e-07, + "loss": 0.3451, + "step": 38594 + }, + { + "epoch": 8.42, + "learning_rate": 7.892724492025344e-07, + "loss": 1.3724, + "step": 38595 + }, + { + "epoch": 8.42, + "learning_rate": 7.891632073410531e-07, + "loss": 0.8269, + "step": 38596 + }, + { + "epoch": 8.42, + "learning_rate": 7.890539654795719e-07, + "loss": 1.3324, + "step": 38597 + }, + { + "epoch": 8.42, + "learning_rate": 7.889447236180905e-07, + "loss": 4.2262, + "step": 38598 + }, + { + "epoch": 8.42, + "learning_rate": 7.888354817566092e-07, + "loss": 1.0001, + "step": 38599 + }, + { + "epoch": 8.42, + "learning_rate": 7.887262398951279e-07, + "loss": 1.3665, + "step": 38600 + }, + { + "epoch": 8.42, + "learning_rate": 7.886169980336466e-07, + "loss": 1.3058, + "step": 38601 + }, + { + "epoch": 8.42, + "learning_rate": 7.885077561721652e-07, + "loss": 0.7698, + "step": 38602 + }, + { + "epoch": 8.42, + "learning_rate": 7.88398514310684e-07, + "loss": 0.832, + "step": 38603 + }, + { + "epoch": 8.43, + "learning_rate": 7.882892724492026e-07, + "loss": 2.1779, + "step": 38604 + }, + { + "epoch": 8.43, + "learning_rate": 7.881800305877213e-07, + "loss": 1.8352, + "step": 38605 + }, + { + "epoch": 8.43, + "learning_rate": 7.8807078872624e-07, + "loss": 0.4937, + "step": 38606 + }, + { + "epoch": 8.43, + "learning_rate": 7.879615468647587e-07, + "loss": 1.8048, + "step": 38607 + }, + { + "epoch": 8.43, + "learning_rate": 7.878523050032773e-07, + "loss": 1.9916, + "step": 38608 + }, + { + "epoch": 8.43, + "learning_rate": 7.877430631417959e-07, + "loss": 0.3612, + "step": 38609 + }, + { + "epoch": 8.43, + "learning_rate": 7.876338212803147e-07, + "loss": 0.8176, + "step": 38610 + }, + { + "epoch": 8.43, + "learning_rate": 7.875245794188334e-07, + "loss": 0.8462, + "step": 38611 + }, + { + "epoch": 8.43, + "learning_rate": 7.87415337557352e-07, + "loss": 0.4511, + "step": 38612 + }, + { + "epoch": 8.43, + "learning_rate": 7.873060956958708e-07, + "loss": 0.5667, + "step": 38613 + }, + { + "epoch": 8.43, + "learning_rate": 7.871968538343894e-07, + "loss": 0.5211, + "step": 38614 + }, + { + "epoch": 8.43, + "learning_rate": 7.870876119729081e-07, + "loss": 0.9009, + "step": 38615 + }, + { + "epoch": 8.43, + "learning_rate": 7.869783701114268e-07, + "loss": 1.5609, + "step": 38616 + }, + { + "epoch": 8.43, + "learning_rate": 7.868691282499455e-07, + "loss": 0.6838, + "step": 38617 + }, + { + "epoch": 8.43, + "learning_rate": 7.867598863884641e-07, + "loss": 2.1651, + "step": 38618 + }, + { + "epoch": 8.43, + "learning_rate": 7.866506445269828e-07, + "loss": 0.38, + "step": 38619 + }, + { + "epoch": 8.43, + "learning_rate": 7.865414026655015e-07, + "loss": 0.5946, + "step": 38620 + }, + { + "epoch": 8.43, + "learning_rate": 7.864321608040202e-07, + "loss": 0.2444, + "step": 38621 + }, + { + "epoch": 8.43, + "learning_rate": 7.863229189425388e-07, + "loss": 1.2954, + "step": 38622 + }, + { + "epoch": 8.43, + "learning_rate": 7.862136770810576e-07, + "loss": 1.186, + "step": 38623 + }, + { + "epoch": 8.43, + "learning_rate": 7.861044352195762e-07, + "loss": 1.0182, + "step": 38624 + }, + { + "epoch": 8.43, + "learning_rate": 7.859951933580949e-07, + "loss": 0.6709, + "step": 38625 + }, + { + "epoch": 8.43, + "learning_rate": 7.858859514966135e-07, + "loss": 0.0558, + "step": 38626 + }, + { + "epoch": 8.43, + "learning_rate": 7.857767096351323e-07, + "loss": 0.5886, + "step": 38627 + }, + { + "epoch": 8.43, + "learning_rate": 7.856674677736509e-07, + "loss": 0.6487, + "step": 38628 + }, + { + "epoch": 8.43, + "learning_rate": 7.855582259121696e-07, + "loss": 1.2522, + "step": 38629 + }, + { + "epoch": 8.43, + "learning_rate": 7.854489840506883e-07, + "loss": 0.961, + "step": 38630 + }, + { + "epoch": 8.43, + "learning_rate": 7.85339742189207e-07, + "loss": 1.2563, + "step": 38631 + }, + { + "epoch": 8.43, + "learning_rate": 7.852305003277256e-07, + "loss": 1.8557, + "step": 38632 + }, + { + "epoch": 8.43, + "learning_rate": 7.851212584662444e-07, + "loss": 1.5425, + "step": 38633 + }, + { + "epoch": 8.43, + "learning_rate": 7.85012016604763e-07, + "loss": 1.2059, + "step": 38634 + }, + { + "epoch": 8.43, + "learning_rate": 7.849027747432817e-07, + "loss": 2.8842, + "step": 38635 + }, + { + "epoch": 8.43, + "learning_rate": 7.847935328818004e-07, + "loss": 1.4036, + "step": 38636 + }, + { + "epoch": 8.43, + "learning_rate": 7.846842910203191e-07, + "loss": 0.9038, + "step": 38637 + }, + { + "epoch": 8.43, + "learning_rate": 7.845750491588377e-07, + "loss": 1.1414, + "step": 38638 + }, + { + "epoch": 8.43, + "learning_rate": 7.844658072973564e-07, + "loss": 0.6062, + "step": 38639 + }, + { + "epoch": 8.43, + "learning_rate": 7.84356565435875e-07, + "loss": 0.2344, + "step": 38640 + }, + { + "epoch": 8.43, + "learning_rate": 7.842473235743938e-07, + "loss": 0.6304, + "step": 38641 + }, + { + "epoch": 8.43, + "learning_rate": 7.841380817129124e-07, + "loss": 1.7292, + "step": 38642 + }, + { + "epoch": 8.43, + "learning_rate": 7.840288398514312e-07, + "loss": 0.9877, + "step": 38643 + }, + { + "epoch": 8.43, + "learning_rate": 7.839195979899498e-07, + "loss": 0.8908, + "step": 38644 + }, + { + "epoch": 8.43, + "learning_rate": 7.838103561284685e-07, + "loss": 1.3624, + "step": 38645 + }, + { + "epoch": 8.43, + "learning_rate": 7.837011142669871e-07, + "loss": 0.1674, + "step": 38646 + }, + { + "epoch": 8.43, + "learning_rate": 7.835918724055059e-07, + "loss": 0.2206, + "step": 38647 + }, + { + "epoch": 8.43, + "learning_rate": 7.834826305440245e-07, + "loss": 1.0234, + "step": 38648 + }, + { + "epoch": 8.43, + "learning_rate": 7.833733886825432e-07, + "loss": 0.8039, + "step": 38649 + }, + { + "epoch": 8.44, + "learning_rate": 7.832641468210619e-07, + "loss": 1.4098, + "step": 38650 + }, + { + "epoch": 8.44, + "learning_rate": 7.831549049595806e-07, + "loss": 0.4489, + "step": 38651 + }, + { + "epoch": 8.44, + "learning_rate": 7.830456630980992e-07, + "loss": 1.5868, + "step": 38652 + }, + { + "epoch": 8.44, + "learning_rate": 7.82936421236618e-07, + "loss": 2.4743, + "step": 38653 + }, + { + "epoch": 8.44, + "learning_rate": 7.828271793751366e-07, + "loss": 0.2632, + "step": 38654 + }, + { + "epoch": 8.44, + "learning_rate": 7.827179375136553e-07, + "loss": 0.2112, + "step": 38655 + }, + { + "epoch": 8.44, + "learning_rate": 7.82608695652174e-07, + "loss": 0.2749, + "step": 38656 + }, + { + "epoch": 8.44, + "learning_rate": 7.824994537906927e-07, + "loss": 1.5611, + "step": 38657 + }, + { + "epoch": 8.44, + "learning_rate": 7.823902119292113e-07, + "loss": 1.5936, + "step": 38658 + }, + { + "epoch": 8.44, + "learning_rate": 7.8228097006773e-07, + "loss": 2.8591, + "step": 38659 + }, + { + "epoch": 8.44, + "learning_rate": 7.821717282062486e-07, + "loss": 0.4528, + "step": 38660 + }, + { + "epoch": 8.44, + "learning_rate": 7.820624863447674e-07, + "loss": 0.3929, + "step": 38661 + }, + { + "epoch": 8.44, + "learning_rate": 7.81953244483286e-07, + "loss": 1.5761, + "step": 38662 + }, + { + "epoch": 8.44, + "learning_rate": 7.818440026218048e-07, + "loss": 1.1064, + "step": 38663 + }, + { + "epoch": 8.44, + "learning_rate": 7.817347607603234e-07, + "loss": 0.8957, + "step": 38664 + }, + { + "epoch": 8.44, + "learning_rate": 7.816255188988421e-07, + "loss": 1.1194, + "step": 38665 + }, + { + "epoch": 8.44, + "learning_rate": 7.815162770373607e-07, + "loss": 2.2468, + "step": 38666 + }, + { + "epoch": 8.44, + "learning_rate": 7.814070351758795e-07, + "loss": 2.6859, + "step": 38667 + }, + { + "epoch": 8.44, + "learning_rate": 7.812977933143981e-07, + "loss": 2.1257, + "step": 38668 + }, + { + "epoch": 8.44, + "learning_rate": 7.811885514529169e-07, + "loss": 1.6253, + "step": 38669 + }, + { + "epoch": 8.44, + "learning_rate": 7.810793095914355e-07, + "loss": 1.4052, + "step": 38670 + }, + { + "epoch": 8.44, + "learning_rate": 7.809700677299542e-07, + "loss": 0.7034, + "step": 38671 + }, + { + "epoch": 8.44, + "learning_rate": 7.808608258684728e-07, + "loss": 1.6539, + "step": 38672 + }, + { + "epoch": 8.44, + "learning_rate": 7.807515840069915e-07, + "loss": 0.1004, + "step": 38673 + }, + { + "epoch": 8.44, + "learning_rate": 7.806423421455102e-07, + "loss": 0.607, + "step": 38674 + }, + { + "epoch": 8.44, + "learning_rate": 7.805331002840289e-07, + "loss": 2.3561, + "step": 38675 + }, + { + "epoch": 8.44, + "learning_rate": 7.804238584225477e-07, + "loss": 1.0335, + "step": 38676 + }, + { + "epoch": 8.44, + "learning_rate": 7.803146165610663e-07, + "loss": 1.7832, + "step": 38677 + }, + { + "epoch": 8.44, + "learning_rate": 7.802053746995849e-07, + "loss": 2.0618, + "step": 38678 + }, + { + "epoch": 8.44, + "learning_rate": 7.800961328381036e-07, + "loss": 0.8247, + "step": 38679 + }, + { + "epoch": 8.44, + "learning_rate": 7.799868909766222e-07, + "loss": 2.5585, + "step": 38680 + }, + { + "epoch": 8.44, + "learning_rate": 7.79877649115141e-07, + "loss": 2.4139, + "step": 38681 + }, + { + "epoch": 8.44, + "learning_rate": 7.797684072536596e-07, + "loss": 0.7762, + "step": 38682 + }, + { + "epoch": 8.44, + "learning_rate": 7.796591653921784e-07, + "loss": 0.2315, + "step": 38683 + }, + { + "epoch": 8.44, + "learning_rate": 7.79549923530697e-07, + "loss": 0.4174, + "step": 38684 + }, + { + "epoch": 8.44, + "learning_rate": 7.794406816692157e-07, + "loss": 1.2345, + "step": 38685 + }, + { + "epoch": 8.44, + "learning_rate": 7.793314398077343e-07, + "loss": 0.7177, + "step": 38686 + }, + { + "epoch": 8.44, + "learning_rate": 7.79222197946253e-07, + "loss": 0.9964, + "step": 38687 + }, + { + "epoch": 8.44, + "learning_rate": 7.791129560847717e-07, + "loss": 0.8356, + "step": 38688 + }, + { + "epoch": 8.44, + "learning_rate": 7.790037142232905e-07, + "loss": 1.7444, + "step": 38689 + }, + { + "epoch": 8.44, + "learning_rate": 7.788944723618092e-07, + "loss": 1.062, + "step": 38690 + }, + { + "epoch": 8.44, + "learning_rate": 7.787852305003278e-07, + "loss": 2.6229, + "step": 38691 + }, + { + "epoch": 8.44, + "learning_rate": 7.786759886388464e-07, + "loss": 0.2328, + "step": 38692 + }, + { + "epoch": 8.44, + "learning_rate": 7.785667467773651e-07, + "loss": 0.9756, + "step": 38693 + }, + { + "epoch": 8.44, + "learning_rate": 7.784575049158838e-07, + "loss": 0.834, + "step": 38694 + }, + { + "epoch": 8.45, + "learning_rate": 7.783482630544025e-07, + "loss": 1.6726, + "step": 38695 + }, + { + "epoch": 8.45, + "learning_rate": 7.782390211929213e-07, + "loss": 1.2644, + "step": 38696 + }, + { + "epoch": 8.45, + "learning_rate": 7.781297793314399e-07, + "loss": 0.8729, + "step": 38697 + }, + { + "epoch": 8.45, + "learning_rate": 7.780205374699585e-07, + "loss": 1.6498, + "step": 38698 + }, + { + "epoch": 8.45, + "learning_rate": 7.779112956084772e-07, + "loss": 1.736, + "step": 38699 + }, + { + "epoch": 8.45, + "learning_rate": 7.778020537469958e-07, + "loss": 0.4283, + "step": 38700 + }, + { + "epoch": 8.45, + "learning_rate": 7.776928118855146e-07, + "loss": 0.9759, + "step": 38701 + }, + { + "epoch": 8.45, + "learning_rate": 7.775835700240334e-07, + "loss": 1.3201, + "step": 38702 + }, + { + "epoch": 8.45, + "learning_rate": 7.77474328162552e-07, + "loss": 1.6604, + "step": 38703 + }, + { + "epoch": 8.45, + "learning_rate": 7.773650863010707e-07, + "loss": 0.6537, + "step": 38704 + }, + { + "epoch": 8.45, + "learning_rate": 7.772558444395893e-07, + "loss": 0.4814, + "step": 38705 + }, + { + "epoch": 8.45, + "learning_rate": 7.771466025781079e-07, + "loss": 0.5824, + "step": 38706 + }, + { + "epoch": 8.45, + "learning_rate": 7.770373607166266e-07, + "loss": 1.5676, + "step": 38707 + }, + { + "epoch": 8.45, + "learning_rate": 7.769281188551453e-07, + "loss": 0.9525, + "step": 38708 + }, + { + "epoch": 8.45, + "learning_rate": 7.768188769936641e-07, + "loss": 1.7065, + "step": 38709 + }, + { + "epoch": 8.45, + "learning_rate": 7.767096351321828e-07, + "loss": 1.5393, + "step": 38710 + }, + { + "epoch": 8.45, + "learning_rate": 7.766003932707014e-07, + "loss": 1.6799, + "step": 38711 + }, + { + "epoch": 8.45, + "learning_rate": 7.764911514092201e-07, + "loss": 0.5656, + "step": 38712 + }, + { + "epoch": 8.45, + "learning_rate": 7.763819095477387e-07, + "loss": 1.2328, + "step": 38713 + }, + { + "epoch": 8.45, + "learning_rate": 7.762726676862573e-07, + "loss": 0.6149, + "step": 38714 + }, + { + "epoch": 8.45, + "learning_rate": 7.761634258247761e-07, + "loss": 1.4896, + "step": 38715 + }, + { + "epoch": 8.45, + "learning_rate": 7.760541839632949e-07, + "loss": 3.3962, + "step": 38716 + }, + { + "epoch": 8.45, + "learning_rate": 7.759449421018135e-07, + "loss": 1.0493, + "step": 38717 + }, + { + "epoch": 8.45, + "learning_rate": 7.758357002403322e-07, + "loss": 0.2325, + "step": 38718 + }, + { + "epoch": 8.45, + "learning_rate": 7.757264583788508e-07, + "loss": 0.0779, + "step": 38719 + }, + { + "epoch": 8.45, + "learning_rate": 7.756172165173694e-07, + "loss": 0.8057, + "step": 38720 + }, + { + "epoch": 8.45, + "learning_rate": 7.755079746558882e-07, + "loss": 0.2266, + "step": 38721 + }, + { + "epoch": 8.45, + "learning_rate": 7.75398732794407e-07, + "loss": 2.8813, + "step": 38722 + }, + { + "epoch": 8.45, + "learning_rate": 7.752894909329256e-07, + "loss": 0.3609, + "step": 38723 + }, + { + "epoch": 8.45, + "learning_rate": 7.751802490714443e-07, + "loss": 0.9911, + "step": 38724 + }, + { + "epoch": 8.45, + "learning_rate": 7.750710072099629e-07, + "loss": 0.4946, + "step": 38725 + }, + { + "epoch": 8.45, + "learning_rate": 7.749617653484816e-07, + "loss": 1.3325, + "step": 38726 + }, + { + "epoch": 8.45, + "learning_rate": 7.748525234870002e-07, + "loss": 1.2444, + "step": 38727 + }, + { + "epoch": 8.45, + "learning_rate": 7.747432816255189e-07, + "loss": 1.9638, + "step": 38728 + }, + { + "epoch": 8.45, + "learning_rate": 7.746340397640377e-07, + "loss": 0.2145, + "step": 38729 + }, + { + "epoch": 8.45, + "learning_rate": 7.745247979025564e-07, + "loss": 1.4878, + "step": 38730 + }, + { + "epoch": 8.45, + "learning_rate": 7.74415556041075e-07, + "loss": 1.1695, + "step": 38731 + }, + { + "epoch": 8.45, + "learning_rate": 7.743063141795937e-07, + "loss": 2.9925, + "step": 38732 + }, + { + "epoch": 8.45, + "learning_rate": 7.741970723181123e-07, + "loss": 2.9483, + "step": 38733 + }, + { + "epoch": 8.45, + "learning_rate": 7.740878304566309e-07, + "loss": 0.9631, + "step": 38734 + }, + { + "epoch": 8.45, + "learning_rate": 7.739785885951498e-07, + "loss": 0.7904, + "step": 38735 + }, + { + "epoch": 8.45, + "learning_rate": 7.738693467336685e-07, + "loss": 1.3254, + "step": 38736 + }, + { + "epoch": 8.45, + "learning_rate": 7.737601048721871e-07, + "loss": 1.3556, + "step": 38737 + }, + { + "epoch": 8.45, + "learning_rate": 7.736508630107058e-07, + "loss": 1.1012, + "step": 38738 + }, + { + "epoch": 8.45, + "learning_rate": 7.735416211492244e-07, + "loss": 0.1475, + "step": 38739 + }, + { + "epoch": 8.45, + "learning_rate": 7.734323792877431e-07, + "loss": 1.5385, + "step": 38740 + }, + { + "epoch": 8.46, + "learning_rate": 7.733231374262618e-07, + "loss": 0.5952, + "step": 38741 + }, + { + "epoch": 8.46, + "learning_rate": 7.732138955647806e-07, + "loss": 1.4647, + "step": 38742 + }, + { + "epoch": 8.46, + "learning_rate": 7.731046537032992e-07, + "loss": 1.5656, + "step": 38743 + }, + { + "epoch": 8.46, + "learning_rate": 7.729954118418179e-07, + "loss": 0.9113, + "step": 38744 + }, + { + "epoch": 8.46, + "learning_rate": 7.728861699803365e-07, + "loss": 0.7344, + "step": 38745 + }, + { + "epoch": 8.46, + "learning_rate": 7.727769281188552e-07, + "loss": 1.7959, + "step": 38746 + }, + { + "epoch": 8.46, + "learning_rate": 7.726676862573738e-07, + "loss": 0.9573, + "step": 38747 + }, + { + "epoch": 8.46, + "learning_rate": 7.725584443958926e-07, + "loss": 1.0134, + "step": 38748 + }, + { + "epoch": 8.46, + "learning_rate": 7.724492025344113e-07, + "loss": 0.8374, + "step": 38749 + }, + { + "epoch": 8.46, + "learning_rate": 7.7233996067293e-07, + "loss": 0.7993, + "step": 38750 + }, + { + "epoch": 8.46, + "learning_rate": 7.722307188114486e-07, + "loss": 1.3334, + "step": 38751 + }, + { + "epoch": 8.46, + "learning_rate": 7.721214769499673e-07, + "loss": 1.6562, + "step": 38752 + }, + { + "epoch": 8.46, + "learning_rate": 7.720122350884859e-07, + "loss": 1.2525, + "step": 38753 + }, + { + "epoch": 8.46, + "learning_rate": 7.719029932270046e-07, + "loss": 2.9934, + "step": 38754 + }, + { + "epoch": 8.46, + "learning_rate": 7.717937513655234e-07, + "loss": 0.7477, + "step": 38755 + }, + { + "epoch": 8.46, + "learning_rate": 7.716845095040421e-07, + "loss": 0.6878, + "step": 38756 + }, + { + "epoch": 8.46, + "learning_rate": 7.715752676425607e-07, + "loss": 0.7021, + "step": 38757 + }, + { + "epoch": 8.46, + "learning_rate": 7.714660257810794e-07, + "loss": 0.8696, + "step": 38758 + }, + { + "epoch": 8.46, + "learning_rate": 7.71356783919598e-07, + "loss": 2.1038, + "step": 38759 + }, + { + "epoch": 8.46, + "learning_rate": 7.712475420581167e-07, + "loss": 1.0742, + "step": 38760 + }, + { + "epoch": 8.46, + "learning_rate": 7.711383001966353e-07, + "loss": 2.632, + "step": 38761 + }, + { + "epoch": 8.46, + "learning_rate": 7.710290583351542e-07, + "loss": 1.6721, + "step": 38762 + }, + { + "epoch": 8.46, + "learning_rate": 7.709198164736728e-07, + "loss": 1.8726, + "step": 38763 + }, + { + "epoch": 8.46, + "learning_rate": 7.708105746121915e-07, + "loss": 1.0738, + "step": 38764 + }, + { + "epoch": 8.46, + "learning_rate": 7.707013327507101e-07, + "loss": 1.0012, + "step": 38765 + }, + { + "epoch": 8.46, + "learning_rate": 7.705920908892288e-07, + "loss": 1.7613, + "step": 38766 + }, + { + "epoch": 8.46, + "learning_rate": 7.704828490277474e-07, + "loss": 0.3429, + "step": 38767 + }, + { + "epoch": 8.46, + "learning_rate": 7.703736071662662e-07, + "loss": 2.3383, + "step": 38768 + }, + { + "epoch": 8.46, + "learning_rate": 7.702643653047849e-07, + "loss": 1.5636, + "step": 38769 + }, + { + "epoch": 8.46, + "learning_rate": 7.701551234433036e-07, + "loss": 1.4127, + "step": 38770 + }, + { + "epoch": 8.46, + "learning_rate": 7.700458815818222e-07, + "loss": 0.5679, + "step": 38771 + }, + { + "epoch": 8.46, + "learning_rate": 7.699366397203409e-07, + "loss": 1.5114, + "step": 38772 + }, + { + "epoch": 8.46, + "learning_rate": 7.698273978588595e-07, + "loss": 1.5104, + "step": 38773 + }, + { + "epoch": 8.46, + "learning_rate": 7.697181559973782e-07, + "loss": 0.2192, + "step": 38774 + }, + { + "epoch": 8.46, + "learning_rate": 7.69608914135897e-07, + "loss": 1.0168, + "step": 38775 + }, + { + "epoch": 8.46, + "learning_rate": 7.694996722744157e-07, + "loss": 1.0898, + "step": 38776 + }, + { + "epoch": 8.46, + "learning_rate": 7.693904304129343e-07, + "loss": 0.694, + "step": 38777 + }, + { + "epoch": 8.46, + "learning_rate": 7.69281188551453e-07, + "loss": 1.1518, + "step": 38778 + }, + { + "epoch": 8.46, + "learning_rate": 7.691719466899716e-07, + "loss": 0.3566, + "step": 38779 + }, + { + "epoch": 8.46, + "learning_rate": 7.690627048284903e-07, + "loss": 2.351, + "step": 38780 + }, + { + "epoch": 8.46, + "learning_rate": 7.689534629670089e-07, + "loss": 0.2744, + "step": 38781 + }, + { + "epoch": 8.46, + "learning_rate": 7.688442211055278e-07, + "loss": 1.7681, + "step": 38782 + }, + { + "epoch": 8.46, + "learning_rate": 7.687349792440464e-07, + "loss": 0.6247, + "step": 38783 + }, + { + "epoch": 8.46, + "learning_rate": 7.686257373825651e-07, + "loss": 1.543, + "step": 38784 + }, + { + "epoch": 8.46, + "learning_rate": 7.685164955210837e-07, + "loss": 0.7788, + "step": 38785 + }, + { + "epoch": 8.46, + "learning_rate": 7.684072536596024e-07, + "loss": 0.4379, + "step": 38786 + }, + { + "epoch": 8.47, + "learning_rate": 7.68298011798121e-07, + "loss": 0.7098, + "step": 38787 + }, + { + "epoch": 8.47, + "learning_rate": 7.681887699366399e-07, + "loss": 0.4787, + "step": 38788 + }, + { + "epoch": 8.47, + "learning_rate": 7.680795280751585e-07, + "loss": 1.1199, + "step": 38789 + }, + { + "epoch": 8.47, + "learning_rate": 7.679702862136772e-07, + "loss": 0.613, + "step": 38790 + }, + { + "epoch": 8.47, + "learning_rate": 7.678610443521958e-07, + "loss": 0.8068, + "step": 38791 + }, + { + "epoch": 8.47, + "learning_rate": 7.677518024907145e-07, + "loss": 0.8732, + "step": 38792 + }, + { + "epoch": 8.47, + "learning_rate": 7.676425606292331e-07, + "loss": 0.8614, + "step": 38793 + }, + { + "epoch": 8.47, + "learning_rate": 7.675333187677518e-07, + "loss": 2.8723, + "step": 38794 + }, + { + "epoch": 8.47, + "learning_rate": 7.674240769062706e-07, + "loss": 2.5375, + "step": 38795 + }, + { + "epoch": 8.47, + "learning_rate": 7.673148350447893e-07, + "loss": 0.4806, + "step": 38796 + }, + { + "epoch": 8.47, + "learning_rate": 7.672055931833079e-07, + "loss": 1.7417, + "step": 38797 + }, + { + "epoch": 8.47, + "learning_rate": 7.670963513218266e-07, + "loss": 2.3045, + "step": 38798 + }, + { + "epoch": 8.47, + "learning_rate": 7.669871094603452e-07, + "loss": 1.9364, + "step": 38799 + }, + { + "epoch": 8.47, + "learning_rate": 7.668778675988639e-07, + "loss": 0.4692, + "step": 38800 + }, + { + "epoch": 8.47, + "learning_rate": 7.667686257373825e-07, + "loss": 0.8188, + "step": 38801 + }, + { + "epoch": 8.47, + "learning_rate": 7.666593838759014e-07, + "loss": 2.1165, + "step": 38802 + }, + { + "epoch": 8.47, + "learning_rate": 7.6655014201442e-07, + "loss": 1.4184, + "step": 38803 + }, + { + "epoch": 8.47, + "learning_rate": 7.664409001529387e-07, + "loss": 0.8954, + "step": 38804 + }, + { + "epoch": 8.47, + "learning_rate": 7.663316582914573e-07, + "loss": 0.6816, + "step": 38805 + }, + { + "epoch": 8.47, + "learning_rate": 7.66222416429976e-07, + "loss": 0.7653, + "step": 38806 + }, + { + "epoch": 8.47, + "learning_rate": 7.661131745684946e-07, + "loss": 0.8167, + "step": 38807 + }, + { + "epoch": 8.47, + "learning_rate": 7.660039327070135e-07, + "loss": 0.6167, + "step": 38808 + }, + { + "epoch": 8.47, + "learning_rate": 7.658946908455321e-07, + "loss": 0.7497, + "step": 38809 + }, + { + "epoch": 8.47, + "learning_rate": 7.657854489840508e-07, + "loss": 0.5617, + "step": 38810 + }, + { + "epoch": 8.47, + "learning_rate": 7.656762071225694e-07, + "loss": 0.7378, + "step": 38811 + }, + { + "epoch": 8.47, + "learning_rate": 7.655669652610881e-07, + "loss": 0.0464, + "step": 38812 + }, + { + "epoch": 8.47, + "learning_rate": 7.654577233996067e-07, + "loss": 1.6277, + "step": 38813 + }, + { + "epoch": 8.47, + "learning_rate": 7.653484815381254e-07, + "loss": 0.8422, + "step": 38814 + }, + { + "epoch": 8.47, + "learning_rate": 7.652392396766443e-07, + "loss": 0.0849, + "step": 38815 + }, + { + "epoch": 8.47, + "learning_rate": 7.651299978151629e-07, + "loss": 3.3399, + "step": 38816 + }, + { + "epoch": 8.47, + "learning_rate": 7.650207559536815e-07, + "loss": 0.871, + "step": 38817 + }, + { + "epoch": 8.47, + "learning_rate": 7.649115140922002e-07, + "loss": 0.3991, + "step": 38818 + }, + { + "epoch": 8.47, + "learning_rate": 7.648022722307188e-07, + "loss": 1.4151, + "step": 38819 + }, + { + "epoch": 8.47, + "learning_rate": 7.646930303692375e-07, + "loss": 0.1585, + "step": 38820 + }, + { + "epoch": 8.47, + "learning_rate": 7.645837885077563e-07, + "loss": 2.2904, + "step": 38821 + }, + { + "epoch": 8.47, + "learning_rate": 7.64474546646275e-07, + "loss": 1.654, + "step": 38822 + }, + { + "epoch": 8.47, + "learning_rate": 7.643653047847936e-07, + "loss": 1.4603, + "step": 38823 + }, + { + "epoch": 8.47, + "learning_rate": 7.642560629233123e-07, + "loss": 2.0474, + "step": 38824 + }, + { + "epoch": 8.47, + "learning_rate": 7.641468210618309e-07, + "loss": 0.6483, + "step": 38825 + }, + { + "epoch": 8.47, + "learning_rate": 7.640375792003496e-07, + "loss": 2.2631, + "step": 38826 + }, + { + "epoch": 8.47, + "learning_rate": 7.639283373388682e-07, + "loss": 0.5613, + "step": 38827 + }, + { + "epoch": 8.47, + "learning_rate": 7.63819095477387e-07, + "loss": 0.2104, + "step": 38828 + }, + { + "epoch": 8.47, + "learning_rate": 7.637098536159058e-07, + "loss": 1.0266, + "step": 38829 + }, + { + "epoch": 8.47, + "learning_rate": 7.636006117544244e-07, + "loss": 0.0996, + "step": 38830 + }, + { + "epoch": 8.47, + "learning_rate": 7.63491369892943e-07, + "loss": 0.6533, + "step": 38831 + }, + { + "epoch": 8.47, + "learning_rate": 7.633821280314617e-07, + "loss": 0.6584, + "step": 38832 + }, + { + "epoch": 8.48, + "learning_rate": 7.632728861699803e-07, + "loss": 0.9305, + "step": 38833 + }, + { + "epoch": 8.48, + "learning_rate": 7.63163644308499e-07, + "loss": 1.0803, + "step": 38834 + }, + { + "epoch": 8.48, + "learning_rate": 7.630544024470179e-07, + "loss": 1.8967, + "step": 38835 + }, + { + "epoch": 8.48, + "learning_rate": 7.629451605855365e-07, + "loss": 1.7781, + "step": 38836 + }, + { + "epoch": 8.48, + "learning_rate": 7.628359187240552e-07, + "loss": 1.2679, + "step": 38837 + }, + { + "epoch": 8.48, + "learning_rate": 7.627266768625738e-07, + "loss": 3.0116, + "step": 38838 + }, + { + "epoch": 8.48, + "learning_rate": 7.626174350010924e-07, + "loss": 1.8682, + "step": 38839 + }, + { + "epoch": 8.48, + "learning_rate": 7.625081931396111e-07, + "loss": 1.0914, + "step": 38840 + }, + { + "epoch": 8.48, + "learning_rate": 7.623989512781299e-07, + "loss": 0.3834, + "step": 38841 + }, + { + "epoch": 8.48, + "learning_rate": 7.622897094166486e-07, + "loss": 0.3041, + "step": 38842 + }, + { + "epoch": 8.48, + "learning_rate": 7.621804675551673e-07, + "loss": 0.2868, + "step": 38843 + }, + { + "epoch": 8.48, + "learning_rate": 7.620712256936859e-07, + "loss": 1.0614, + "step": 38844 + }, + { + "epoch": 8.48, + "learning_rate": 7.619619838322045e-07, + "loss": 0.244, + "step": 38845 + }, + { + "epoch": 8.48, + "learning_rate": 7.618527419707232e-07, + "loss": 0.146, + "step": 38846 + }, + { + "epoch": 8.48, + "learning_rate": 7.617435001092418e-07, + "loss": 0.8715, + "step": 38847 + }, + { + "epoch": 8.48, + "learning_rate": 7.616342582477606e-07, + "loss": 1.9025, + "step": 38848 + }, + { + "epoch": 8.48, + "learning_rate": 7.615250163862794e-07, + "loss": 0.5658, + "step": 38849 + }, + { + "epoch": 8.48, + "learning_rate": 7.61415774524798e-07, + "loss": 1.1571, + "step": 38850 + }, + { + "epoch": 8.48, + "learning_rate": 7.613065326633167e-07, + "loss": 0.9694, + "step": 38851 + }, + { + "epoch": 8.48, + "learning_rate": 7.611972908018353e-07, + "loss": 0.5539, + "step": 38852 + }, + { + "epoch": 8.48, + "learning_rate": 7.610880489403539e-07, + "loss": 0.7181, + "step": 38853 + }, + { + "epoch": 8.48, + "learning_rate": 7.609788070788727e-07, + "loss": 0.8337, + "step": 38854 + }, + { + "epoch": 8.48, + "learning_rate": 7.608695652173914e-07, + "loss": 1.2375, + "step": 38855 + }, + { + "epoch": 8.48, + "learning_rate": 7.607603233559101e-07, + "loss": 1.512, + "step": 38856 + }, + { + "epoch": 8.48, + "learning_rate": 7.606510814944288e-07, + "loss": 0.6167, + "step": 38857 + }, + { + "epoch": 8.48, + "learning_rate": 7.605418396329474e-07, + "loss": 1.2671, + "step": 38858 + }, + { + "epoch": 8.48, + "learning_rate": 7.604325977714661e-07, + "loss": 0.4238, + "step": 38859 + }, + { + "epoch": 8.48, + "learning_rate": 7.603233559099847e-07, + "loss": 1.8171, + "step": 38860 + }, + { + "epoch": 8.48, + "learning_rate": 7.602141140485035e-07, + "loss": 1.8556, + "step": 38861 + }, + { + "epoch": 8.48, + "learning_rate": 7.601048721870221e-07, + "loss": 1.5217, + "step": 38862 + }, + { + "epoch": 8.48, + "learning_rate": 7.599956303255409e-07, + "loss": 0.8245, + "step": 38863 + }, + { + "epoch": 8.48, + "learning_rate": 7.598863884640595e-07, + "loss": 1.3538, + "step": 38864 + }, + { + "epoch": 8.48, + "learning_rate": 7.597771466025782e-07, + "loss": 0.6609, + "step": 38865 + }, + { + "epoch": 8.48, + "learning_rate": 7.596679047410968e-07, + "loss": 0.2342, + "step": 38866 + }, + { + "epoch": 8.48, + "learning_rate": 7.595586628796154e-07, + "loss": 2.0341, + "step": 38867 + }, + { + "epoch": 8.48, + "learning_rate": 7.594494210181342e-07, + "loss": 2.2262, + "step": 38868 + }, + { + "epoch": 8.48, + "learning_rate": 7.59340179156653e-07, + "loss": 1.8439, + "step": 38869 + }, + { + "epoch": 8.48, + "learning_rate": 7.592309372951716e-07, + "loss": 1.6174, + "step": 38870 + }, + { + "epoch": 8.48, + "learning_rate": 7.591216954336903e-07, + "loss": 2.2454, + "step": 38871 + }, + { + "epoch": 8.48, + "learning_rate": 7.590124535722089e-07, + "loss": 1.3471, + "step": 38872 + }, + { + "epoch": 8.48, + "learning_rate": 7.589032117107276e-07, + "loss": 0.7581, + "step": 38873 + }, + { + "epoch": 8.48, + "learning_rate": 7.587939698492463e-07, + "loss": 1.3409, + "step": 38874 + }, + { + "epoch": 8.48, + "learning_rate": 7.58684727987765e-07, + "loss": 2.5721, + "step": 38875 + }, + { + "epoch": 8.48, + "learning_rate": 7.585754861262837e-07, + "loss": 0.5744, + "step": 38876 + }, + { + "epoch": 8.48, + "learning_rate": 7.584662442648024e-07, + "loss": 1.1057, + "step": 38877 + }, + { + "epoch": 8.48, + "learning_rate": 7.58357002403321e-07, + "loss": 1.8277, + "step": 38878 + }, + { + "epoch": 8.49, + "learning_rate": 7.582477605418397e-07, + "loss": 0.9117, + "step": 38879 + }, + { + "epoch": 8.49, + "learning_rate": 7.581385186803583e-07, + "loss": 0.9453, + "step": 38880 + }, + { + "epoch": 8.49, + "learning_rate": 7.580292768188771e-07, + "loss": 1.2459, + "step": 38881 + }, + { + "epoch": 8.49, + "learning_rate": 7.579200349573957e-07, + "loss": 1.1529, + "step": 38882 + }, + { + "epoch": 8.49, + "learning_rate": 7.578107930959145e-07, + "loss": 1.7246, + "step": 38883 + }, + { + "epoch": 8.49, + "learning_rate": 7.577015512344331e-07, + "loss": 1.3719, + "step": 38884 + }, + { + "epoch": 8.49, + "learning_rate": 7.575923093729518e-07, + "loss": 1.2374, + "step": 38885 + }, + { + "epoch": 8.49, + "learning_rate": 7.574830675114704e-07, + "loss": 0.2639, + "step": 38886 + }, + { + "epoch": 8.49, + "learning_rate": 7.573738256499891e-07, + "loss": 2.0807, + "step": 38887 + }, + { + "epoch": 8.49, + "learning_rate": 7.572645837885078e-07, + "loss": 1.2004, + "step": 38888 + }, + { + "epoch": 8.49, + "learning_rate": 7.571553419270266e-07, + "loss": 0.7507, + "step": 38889 + }, + { + "epoch": 8.49, + "learning_rate": 7.570461000655452e-07, + "loss": 0.7685, + "step": 38890 + }, + { + "epoch": 8.49, + "learning_rate": 7.569368582040639e-07, + "loss": 1.0386, + "step": 38891 + }, + { + "epoch": 8.49, + "learning_rate": 7.568276163425825e-07, + "loss": 0.6261, + "step": 38892 + }, + { + "epoch": 8.49, + "learning_rate": 7.567183744811012e-07, + "loss": 2.2152, + "step": 38893 + }, + { + "epoch": 8.49, + "learning_rate": 7.566091326196199e-07, + "loss": 1.4481, + "step": 38894 + }, + { + "epoch": 8.49, + "learning_rate": 7.564998907581386e-07, + "loss": 0.9115, + "step": 38895 + }, + { + "epoch": 8.49, + "learning_rate": 7.563906488966573e-07, + "loss": 1.4795, + "step": 38896 + }, + { + "epoch": 8.49, + "learning_rate": 7.56281407035176e-07, + "loss": 0.5301, + "step": 38897 + }, + { + "epoch": 8.49, + "learning_rate": 7.561721651736946e-07, + "loss": 3.1286, + "step": 38898 + }, + { + "epoch": 8.49, + "learning_rate": 7.560629233122133e-07, + "loss": 1.6765, + "step": 38899 + }, + { + "epoch": 8.49, + "learning_rate": 7.559536814507319e-07, + "loss": 2.3687, + "step": 38900 + }, + { + "epoch": 8.49, + "learning_rate": 7.558444395892507e-07, + "loss": 0.1115, + "step": 38901 + }, + { + "epoch": 8.49, + "learning_rate": 7.557351977277693e-07, + "loss": 0.5472, + "step": 38902 + }, + { + "epoch": 8.49, + "learning_rate": 7.556259558662881e-07, + "loss": 1.8965, + "step": 38903 + }, + { + "epoch": 8.49, + "learning_rate": 7.555167140048067e-07, + "loss": 0.085, + "step": 38904 + }, + { + "epoch": 8.49, + "learning_rate": 7.554074721433254e-07, + "loss": 0.7659, + "step": 38905 + }, + { + "epoch": 8.49, + "learning_rate": 7.55298230281844e-07, + "loss": 1.4513, + "step": 38906 + }, + { + "epoch": 8.49, + "learning_rate": 7.551889884203628e-07, + "loss": 1.5546, + "step": 38907 + }, + { + "epoch": 8.49, + "learning_rate": 7.550797465588814e-07, + "loss": 0.4211, + "step": 38908 + }, + { + "epoch": 8.49, + "learning_rate": 7.549705046974001e-07, + "loss": 0.3589, + "step": 38909 + }, + { + "epoch": 8.49, + "learning_rate": 7.548612628359188e-07, + "loss": 1.0357, + "step": 38910 + }, + { + "epoch": 8.49, + "learning_rate": 7.547520209744375e-07, + "loss": 0.8218, + "step": 38911 + }, + { + "epoch": 8.49, + "learning_rate": 7.546427791129561e-07, + "loss": 0.3906, + "step": 38912 + }, + { + "epoch": 8.49, + "learning_rate": 7.545335372514748e-07, + "loss": 2.2625, + "step": 38913 + }, + { + "epoch": 8.49, + "learning_rate": 7.544242953899935e-07, + "loss": 1.2555, + "step": 38914 + }, + { + "epoch": 8.49, + "learning_rate": 7.543150535285122e-07, + "loss": 1.2445, + "step": 38915 + }, + { + "epoch": 8.49, + "learning_rate": 7.542058116670308e-07, + "loss": 2.0518, + "step": 38916 + }, + { + "epoch": 8.49, + "learning_rate": 7.540965698055496e-07, + "loss": 1.2812, + "step": 38917 + }, + { + "epoch": 8.49, + "learning_rate": 7.539873279440682e-07, + "loss": 1.1518, + "step": 38918 + }, + { + "epoch": 8.49, + "learning_rate": 7.538780860825869e-07, + "loss": 0.6151, + "step": 38919 + }, + { + "epoch": 8.49, + "learning_rate": 7.537688442211055e-07, + "loss": 1.1452, + "step": 38920 + }, + { + "epoch": 8.49, + "learning_rate": 7.536596023596243e-07, + "loss": 0.238, + "step": 38921 + }, + { + "epoch": 8.49, + "learning_rate": 7.535503604981429e-07, + "loss": 0.7043, + "step": 38922 + }, + { + "epoch": 8.49, + "learning_rate": 7.534411186366617e-07, + "loss": 0.7806, + "step": 38923 + }, + { + "epoch": 8.49, + "learning_rate": 7.533318767751803e-07, + "loss": 0.5332, + "step": 38924 + }, + { + "epoch": 8.5, + "learning_rate": 7.53222634913699e-07, + "loss": 1.243, + "step": 38925 + }, + { + "epoch": 8.5, + "learning_rate": 7.531133930522176e-07, + "loss": 0.4942, + "step": 38926 + }, + { + "epoch": 8.5, + "learning_rate": 7.530041511907364e-07, + "loss": 0.161, + "step": 38927 + }, + { + "epoch": 8.5, + "learning_rate": 7.52894909329255e-07, + "loss": 1.1998, + "step": 38928 + }, + { + "epoch": 8.5, + "learning_rate": 7.527856674677737e-07, + "loss": 0.7143, + "step": 38929 + }, + { + "epoch": 8.5, + "learning_rate": 7.526764256062924e-07, + "loss": 1.4981, + "step": 38930 + }, + { + "epoch": 8.5, + "learning_rate": 7.525671837448111e-07, + "loss": 0.1432, + "step": 38931 + }, + { + "epoch": 8.5, + "learning_rate": 7.524579418833297e-07, + "loss": 1.3276, + "step": 38932 + }, + { + "epoch": 8.5, + "learning_rate": 7.523487000218484e-07, + "loss": 1.0136, + "step": 38933 + }, + { + "epoch": 8.5, + "learning_rate": 7.522394581603671e-07, + "loss": 1.8157, + "step": 38934 + }, + { + "epoch": 8.5, + "learning_rate": 7.521302162988858e-07, + "loss": 1.3658, + "step": 38935 + }, + { + "epoch": 8.5, + "learning_rate": 7.520209744374044e-07, + "loss": 0.2985, + "step": 38936 + }, + { + "epoch": 8.5, + "learning_rate": 7.519117325759232e-07, + "loss": 0.8412, + "step": 38937 + }, + { + "epoch": 8.5, + "learning_rate": 7.518024907144418e-07, + "loss": 1.1626, + "step": 38938 + }, + { + "epoch": 8.5, + "learning_rate": 7.516932488529605e-07, + "loss": 0.7556, + "step": 38939 + }, + { + "epoch": 8.5, + "learning_rate": 7.515840069914793e-07, + "loss": 2.1483, + "step": 38940 + }, + { + "epoch": 8.5, + "learning_rate": 7.514747651299979e-07, + "loss": 0.5766, + "step": 38941 + }, + { + "epoch": 8.5, + "learning_rate": 7.513655232685165e-07, + "loss": 0.2355, + "step": 38942 + }, + { + "epoch": 8.5, + "learning_rate": 7.512562814070353e-07, + "loss": 1.532, + "step": 38943 + }, + { + "epoch": 8.5, + "learning_rate": 7.511470395455539e-07, + "loss": 2.8346, + "step": 38944 + }, + { + "epoch": 8.5, + "learning_rate": 7.510377976840726e-07, + "loss": 0.8204, + "step": 38945 + }, + { + "epoch": 8.5, + "learning_rate": 7.509285558225912e-07, + "loss": 2.2632, + "step": 38946 + }, + { + "epoch": 8.5, + "learning_rate": 7.5081931396111e-07, + "loss": 0.892, + "step": 38947 + }, + { + "epoch": 8.5, + "learning_rate": 7.507100720996287e-07, + "loss": 0.9314, + "step": 38948 + }, + { + "epoch": 8.5, + "learning_rate": 7.506008302381473e-07, + "loss": 0.5506, + "step": 38949 + }, + { + "epoch": 8.5, + "learning_rate": 7.50491588376666e-07, + "loss": 1.4105, + "step": 38950 + }, + { + "epoch": 8.5, + "learning_rate": 7.503823465151847e-07, + "loss": 1.9316, + "step": 38951 + }, + { + "epoch": 8.5, + "learning_rate": 7.502731046537033e-07, + "loss": 1.0818, + "step": 38952 + }, + { + "epoch": 8.5, + "learning_rate": 7.50163862792222e-07, + "loss": 0.8529, + "step": 38953 + }, + { + "epoch": 8.5, + "learning_rate": 7.500546209307408e-07, + "loss": 1.1329, + "step": 38954 + }, + { + "epoch": 8.5, + "learning_rate": 7.499453790692594e-07, + "loss": 0.7557, + "step": 38955 + }, + { + "epoch": 8.5, + "learning_rate": 7.49836137207778e-07, + "loss": 1.2934, + "step": 38956 + }, + { + "epoch": 8.5, + "learning_rate": 7.497268953462968e-07, + "loss": 1.635, + "step": 38957 + }, + { + "epoch": 8.5, + "learning_rate": 7.496176534848154e-07, + "loss": 1.0688, + "step": 38958 + }, + { + "epoch": 8.5, + "learning_rate": 7.495084116233341e-07, + "loss": 1.1269, + "step": 38959 + }, + { + "epoch": 8.5, + "learning_rate": 7.493991697618529e-07, + "loss": 0.7515, + "step": 38960 + }, + { + "epoch": 8.5, + "learning_rate": 7.492899279003715e-07, + "loss": 1.1137, + "step": 38961 + }, + { + "epoch": 8.5, + "learning_rate": 7.491806860388902e-07, + "loss": 1.2224, + "step": 38962 + }, + { + "epoch": 8.5, + "learning_rate": 7.490714441774088e-07, + "loss": 0.5854, + "step": 38963 + }, + { + "epoch": 8.5, + "learning_rate": 7.489622023159275e-07, + "loss": 0.0624, + "step": 38964 + }, + { + "epoch": 8.5, + "learning_rate": 7.488529604544462e-07, + "loss": 0.7831, + "step": 38965 + }, + { + "epoch": 8.5, + "learning_rate": 7.487437185929648e-07, + "loss": 1.5474, + "step": 38966 + }, + { + "epoch": 8.5, + "learning_rate": 7.486344767314836e-07, + "loss": 0.6778, + "step": 38967 + }, + { + "epoch": 8.5, + "learning_rate": 7.485252348700023e-07, + "loss": 1.2844, + "step": 38968 + }, + { + "epoch": 8.5, + "learning_rate": 7.484159930085209e-07, + "loss": 1.2313, + "step": 38969 + }, + { + "epoch": 8.51, + "learning_rate": 7.483067511470395e-07, + "loss": 1.9232, + "step": 38970 + }, + { + "epoch": 8.51, + "learning_rate": 7.481975092855583e-07, + "loss": 1.1798, + "step": 38971 + }, + { + "epoch": 8.51, + "learning_rate": 7.480882674240769e-07, + "loss": 1.1294, + "step": 38972 + }, + { + "epoch": 8.51, + "learning_rate": 7.479790255625956e-07, + "loss": 0.1836, + "step": 38973 + }, + { + "epoch": 8.51, + "learning_rate": 7.478697837011144e-07, + "loss": 0.5662, + "step": 38974 + }, + { + "epoch": 8.51, + "learning_rate": 7.47760541839633e-07, + "loss": 0.9797, + "step": 38975 + }, + { + "epoch": 8.51, + "learning_rate": 7.476512999781517e-07, + "loss": 0.8436, + "step": 38976 + }, + { + "epoch": 8.51, + "learning_rate": 7.475420581166704e-07, + "loss": 0.6417, + "step": 38977 + }, + { + "epoch": 8.51, + "learning_rate": 7.47432816255189e-07, + "loss": 1.2499, + "step": 38978 + }, + { + "epoch": 8.51, + "learning_rate": 7.473235743937077e-07, + "loss": 0.8793, + "step": 38979 + }, + { + "epoch": 8.51, + "learning_rate": 7.472143325322265e-07, + "loss": 0.6503, + "step": 38980 + }, + { + "epoch": 8.51, + "learning_rate": 7.471050906707451e-07, + "loss": 1.8008, + "step": 38981 + }, + { + "epoch": 8.51, + "learning_rate": 7.469958488092638e-07, + "loss": 0.1555, + "step": 38982 + }, + { + "epoch": 8.51, + "learning_rate": 7.468866069477824e-07, + "loss": 2.8988, + "step": 38983 + }, + { + "epoch": 8.51, + "learning_rate": 7.467773650863012e-07, + "loss": 1.3379, + "step": 38984 + }, + { + "epoch": 8.51, + "learning_rate": 7.466681232248198e-07, + "loss": 1.3376, + "step": 38985 + }, + { + "epoch": 8.51, + "learning_rate": 7.465588813633384e-07, + "loss": 0.2485, + "step": 38986 + }, + { + "epoch": 8.51, + "learning_rate": 7.464496395018572e-07, + "loss": 1.3782, + "step": 38987 + }, + { + "epoch": 8.51, + "learning_rate": 7.463403976403759e-07, + "loss": 0.5795, + "step": 38988 + }, + { + "epoch": 8.51, + "learning_rate": 7.462311557788945e-07, + "loss": 0.6702, + "step": 38989 + }, + { + "epoch": 8.51, + "learning_rate": 7.461219139174133e-07, + "loss": 1.9109, + "step": 38990 + }, + { + "epoch": 8.51, + "learning_rate": 7.460126720559319e-07, + "loss": 1.3747, + "step": 38991 + }, + { + "epoch": 8.51, + "learning_rate": 7.459034301944505e-07, + "loss": 1.4645, + "step": 38992 + }, + { + "epoch": 8.51, + "learning_rate": 7.457941883329693e-07, + "loss": 0.9394, + "step": 38993 + }, + { + "epoch": 8.51, + "learning_rate": 7.45684946471488e-07, + "loss": 1.6808, + "step": 38994 + }, + { + "epoch": 8.51, + "learning_rate": 7.455757046100066e-07, + "loss": 0.6404, + "step": 38995 + }, + { + "epoch": 8.51, + "learning_rate": 7.454664627485253e-07, + "loss": 1.0602, + "step": 38996 + }, + { + "epoch": 8.51, + "learning_rate": 7.45357220887044e-07, + "loss": 0.861, + "step": 38997 + }, + { + "epoch": 8.51, + "learning_rate": 7.452479790255627e-07, + "loss": 0.7005, + "step": 38998 + }, + { + "epoch": 8.51, + "learning_rate": 7.451387371640813e-07, + "loss": 0.942, + "step": 38999 + }, + { + "epoch": 8.51, + "learning_rate": 7.450294953026001e-07, + "loss": 2.4435, + "step": 39000 + }, + { + "epoch": 8.51, + "learning_rate": 7.449202534411187e-07, + "loss": 0.7484, + "step": 39001 + }, + { + "epoch": 8.51, + "learning_rate": 7.448110115796374e-07, + "loss": 0.9606, + "step": 39002 + }, + { + "epoch": 8.51, + "learning_rate": 7.44701769718156e-07, + "loss": 1.1524, + "step": 39003 + }, + { + "epoch": 8.51, + "learning_rate": 7.445925278566748e-07, + "loss": 1.2292, + "step": 39004 + }, + { + "epoch": 8.51, + "learning_rate": 7.444832859951934e-07, + "loss": 1.5893, + "step": 39005 + }, + { + "epoch": 8.51, + "learning_rate": 7.44374044133712e-07, + "loss": 1.168, + "step": 39006 + }, + { + "epoch": 8.51, + "learning_rate": 7.442648022722308e-07, + "loss": 0.6678, + "step": 39007 + }, + { + "epoch": 8.51, + "learning_rate": 7.441555604107495e-07, + "loss": 1.5436, + "step": 39008 + }, + { + "epoch": 8.51, + "learning_rate": 7.440463185492681e-07, + "loss": 1.1116, + "step": 39009 + }, + { + "epoch": 8.51, + "learning_rate": 7.439370766877868e-07, + "loss": 0.7606, + "step": 39010 + }, + { + "epoch": 8.51, + "learning_rate": 7.438278348263055e-07, + "loss": 0.8632, + "step": 39011 + }, + { + "epoch": 8.51, + "learning_rate": 7.437185929648242e-07, + "loss": 0.7072, + "step": 39012 + }, + { + "epoch": 8.51, + "learning_rate": 7.436093511033429e-07, + "loss": 0.1958, + "step": 39013 + }, + { + "epoch": 8.51, + "learning_rate": 7.435001092418616e-07, + "loss": 1.2818, + "step": 39014 + }, + { + "epoch": 8.51, + "learning_rate": 7.433908673803802e-07, + "loss": 0.9833, + "step": 39015 + }, + { + "epoch": 8.52, + "learning_rate": 7.432816255188989e-07, + "loss": 1.2588, + "step": 39016 + }, + { + "epoch": 8.52, + "learning_rate": 7.431723836574175e-07, + "loss": 0.1117, + "step": 39017 + }, + { + "epoch": 8.52, + "learning_rate": 7.430631417959363e-07, + "loss": 1.3902, + "step": 39018 + }, + { + "epoch": 8.52, + "learning_rate": 7.429538999344549e-07, + "loss": 1.3885, + "step": 39019 + }, + { + "epoch": 8.52, + "learning_rate": 7.428446580729737e-07, + "loss": 1.7682, + "step": 39020 + }, + { + "epoch": 8.52, + "learning_rate": 7.427354162114923e-07, + "loss": 1.8067, + "step": 39021 + }, + { + "epoch": 8.52, + "learning_rate": 7.42626174350011e-07, + "loss": 0.0403, + "step": 39022 + }, + { + "epoch": 8.52, + "learning_rate": 7.425169324885296e-07, + "loss": 2.5988, + "step": 39023 + }, + { + "epoch": 8.52, + "learning_rate": 7.424076906270484e-07, + "loss": 2.0827, + "step": 39024 + }, + { + "epoch": 8.52, + "learning_rate": 7.42298448765567e-07, + "loss": 0.5807, + "step": 39025 + }, + { + "epoch": 8.52, + "learning_rate": 7.421892069040858e-07, + "loss": 0.6786, + "step": 39026 + }, + { + "epoch": 8.52, + "learning_rate": 7.420799650426044e-07, + "loss": 0.6715, + "step": 39027 + }, + { + "epoch": 8.52, + "learning_rate": 7.419707231811231e-07, + "loss": 0.6712, + "step": 39028 + }, + { + "epoch": 8.52, + "learning_rate": 7.418614813196417e-07, + "loss": 0.7502, + "step": 39029 + }, + { + "epoch": 8.52, + "learning_rate": 7.417522394581604e-07, + "loss": 0.5725, + "step": 39030 + }, + { + "epoch": 8.52, + "learning_rate": 7.416429975966791e-07, + "loss": 1.7486, + "step": 39031 + }, + { + "epoch": 8.52, + "learning_rate": 7.415337557351978e-07, + "loss": 1.7812, + "step": 39032 + }, + { + "epoch": 8.52, + "learning_rate": 7.414245138737165e-07, + "loss": 1.5203, + "step": 39033 + }, + { + "epoch": 8.52, + "learning_rate": 7.413152720122352e-07, + "loss": 1.6157, + "step": 39034 + }, + { + "epoch": 8.52, + "learning_rate": 7.412060301507538e-07, + "loss": 2.2077, + "step": 39035 + }, + { + "epoch": 8.52, + "learning_rate": 7.410967882892725e-07, + "loss": 0.274, + "step": 39036 + }, + { + "epoch": 8.52, + "learning_rate": 7.409875464277911e-07, + "loss": 2.353, + "step": 39037 + }, + { + "epoch": 8.52, + "learning_rate": 7.408783045663099e-07, + "loss": 0.2046, + "step": 39038 + }, + { + "epoch": 8.52, + "learning_rate": 7.407690627048285e-07, + "loss": 1.4959, + "step": 39039 + }, + { + "epoch": 8.52, + "learning_rate": 7.406598208433473e-07, + "loss": 0.731, + "step": 39040 + }, + { + "epoch": 8.52, + "learning_rate": 7.405505789818659e-07, + "loss": 1.9663, + "step": 39041 + }, + { + "epoch": 8.52, + "learning_rate": 7.404413371203846e-07, + "loss": 0.3758, + "step": 39042 + }, + { + "epoch": 8.52, + "learning_rate": 7.403320952589032e-07, + "loss": 0.2644, + "step": 39043 + }, + { + "epoch": 8.52, + "learning_rate": 7.40222853397422e-07, + "loss": 0.8492, + "step": 39044 + }, + { + "epoch": 8.52, + "learning_rate": 7.401136115359406e-07, + "loss": 1.2333, + "step": 39045 + }, + { + "epoch": 8.52, + "learning_rate": 7.400043696744594e-07, + "loss": 0.826, + "step": 39046 + }, + { + "epoch": 8.52, + "learning_rate": 7.39895127812978e-07, + "loss": 0.2119, + "step": 39047 + }, + { + "epoch": 8.52, + "learning_rate": 7.397858859514967e-07, + "loss": 0.7992, + "step": 39048 + }, + { + "epoch": 8.52, + "learning_rate": 7.396766440900153e-07, + "loss": 1.3503, + "step": 39049 + }, + { + "epoch": 8.52, + "learning_rate": 7.39567402228534e-07, + "loss": 0.1673, + "step": 39050 + }, + { + "epoch": 8.52, + "learning_rate": 7.394581603670526e-07, + "loss": 0.959, + "step": 39051 + }, + { + "epoch": 8.52, + "learning_rate": 7.393489185055714e-07, + "loss": 1.4074, + "step": 39052 + }, + { + "epoch": 8.52, + "learning_rate": 7.392396766440901e-07, + "loss": 0.7925, + "step": 39053 + }, + { + "epoch": 8.52, + "learning_rate": 7.391304347826088e-07, + "loss": 1.1294, + "step": 39054 + }, + { + "epoch": 8.52, + "learning_rate": 7.390211929211274e-07, + "loss": 1.0987, + "step": 39055 + }, + { + "epoch": 8.52, + "learning_rate": 7.389119510596461e-07, + "loss": 0.4399, + "step": 39056 + }, + { + "epoch": 8.52, + "learning_rate": 7.388027091981647e-07, + "loss": 0.6915, + "step": 39057 + }, + { + "epoch": 8.52, + "learning_rate": 7.386934673366835e-07, + "loss": 1.8681, + "step": 39058 + }, + { + "epoch": 8.52, + "learning_rate": 7.385842254752023e-07, + "loss": 2.0774, + "step": 39059 + }, + { + "epoch": 8.52, + "learning_rate": 7.384749836137209e-07, + "loss": 0.2008, + "step": 39060 + }, + { + "epoch": 8.52, + "learning_rate": 7.383657417522395e-07, + "loss": 1.5396, + "step": 39061 + }, + { + "epoch": 8.53, + "learning_rate": 7.382564998907582e-07, + "loss": 0.2668, + "step": 39062 + }, + { + "epoch": 8.53, + "learning_rate": 7.381472580292768e-07, + "loss": 1.8076, + "step": 39063 + }, + { + "epoch": 8.53, + "learning_rate": 7.380380161677955e-07, + "loss": 0.7404, + "step": 39064 + }, + { + "epoch": 8.53, + "learning_rate": 7.379287743063142e-07, + "loss": 0.4038, + "step": 39065 + }, + { + "epoch": 8.53, + "learning_rate": 7.37819532444833e-07, + "loss": 1.1632, + "step": 39066 + }, + { + "epoch": 8.53, + "learning_rate": 7.377102905833516e-07, + "loss": 1.346, + "step": 39067 + }, + { + "epoch": 8.53, + "learning_rate": 7.376010487218703e-07, + "loss": 0.7673, + "step": 39068 + }, + { + "epoch": 8.53, + "learning_rate": 7.374918068603889e-07, + "loss": 0.2747, + "step": 39069 + }, + { + "epoch": 8.53, + "learning_rate": 7.373825649989076e-07, + "loss": 0.5216, + "step": 39070 + }, + { + "epoch": 8.53, + "learning_rate": 7.372733231374262e-07, + "loss": 0.9961, + "step": 39071 + }, + { + "epoch": 8.53, + "learning_rate": 7.37164081275945e-07, + "loss": 1.5184, + "step": 39072 + }, + { + "epoch": 8.53, + "learning_rate": 7.370548394144638e-07, + "loss": 0.7907, + "step": 39073 + }, + { + "epoch": 8.53, + "learning_rate": 7.369455975529824e-07, + "loss": 1.0752, + "step": 39074 + }, + { + "epoch": 8.53, + "learning_rate": 7.36836355691501e-07, + "loss": 1.1187, + "step": 39075 + }, + { + "epoch": 8.53, + "learning_rate": 7.367271138300197e-07, + "loss": 0.2426, + "step": 39076 + }, + { + "epoch": 8.53, + "learning_rate": 7.366178719685383e-07, + "loss": 2.1751, + "step": 39077 + }, + { + "epoch": 8.53, + "learning_rate": 7.36508630107057e-07, + "loss": 1.2281, + "step": 39078 + }, + { + "epoch": 8.53, + "learning_rate": 7.363993882455759e-07, + "loss": 0.5901, + "step": 39079 + }, + { + "epoch": 8.53, + "learning_rate": 7.362901463840945e-07, + "loss": 0.401, + "step": 39080 + }, + { + "epoch": 8.53, + "learning_rate": 7.361809045226131e-07, + "loss": 1.2545, + "step": 39081 + }, + { + "epoch": 8.53, + "learning_rate": 7.360716626611318e-07, + "loss": 0.8133, + "step": 39082 + }, + { + "epoch": 8.53, + "learning_rate": 7.359624207996504e-07, + "loss": 1.3351, + "step": 39083 + }, + { + "epoch": 8.53, + "learning_rate": 7.358531789381691e-07, + "loss": 3.037, + "step": 39084 + }, + { + "epoch": 8.53, + "learning_rate": 7.357439370766878e-07, + "loss": 0.4528, + "step": 39085 + }, + { + "epoch": 8.53, + "learning_rate": 7.356346952152066e-07, + "loss": 0.1105, + "step": 39086 + }, + { + "epoch": 8.53, + "learning_rate": 7.355254533537253e-07, + "loss": 0.4446, + "step": 39087 + }, + { + "epoch": 8.53, + "learning_rate": 7.354162114922439e-07, + "loss": 2.4622, + "step": 39088 + }, + { + "epoch": 8.53, + "learning_rate": 7.353069696307625e-07, + "loss": 1.2149, + "step": 39089 + }, + { + "epoch": 8.53, + "learning_rate": 7.351977277692812e-07, + "loss": 0.6886, + "step": 39090 + }, + { + "epoch": 8.53, + "learning_rate": 7.350884859077998e-07, + "loss": 1.4592, + "step": 39091 + }, + { + "epoch": 8.53, + "learning_rate": 7.349792440463186e-07, + "loss": 0.5533, + "step": 39092 + }, + { + "epoch": 8.53, + "learning_rate": 7.348700021848374e-07, + "loss": 1.196, + "step": 39093 + }, + { + "epoch": 8.53, + "learning_rate": 7.34760760323356e-07, + "loss": 0.3793, + "step": 39094 + }, + { + "epoch": 8.53, + "learning_rate": 7.346515184618747e-07, + "loss": 3.0881, + "step": 39095 + }, + { + "epoch": 8.53, + "learning_rate": 7.345422766003933e-07, + "loss": 2.051, + "step": 39096 + }, + { + "epoch": 8.53, + "learning_rate": 7.344330347389119e-07, + "loss": 2.9139, + "step": 39097 + }, + { + "epoch": 8.53, + "learning_rate": 7.343237928774306e-07, + "loss": 0.5438, + "step": 39098 + }, + { + "epoch": 8.53, + "learning_rate": 7.342145510159495e-07, + "loss": 0.827, + "step": 39099 + }, + { + "epoch": 8.53, + "learning_rate": 7.341053091544681e-07, + "loss": 0.5614, + "step": 39100 + }, + { + "epoch": 8.53, + "learning_rate": 7.339960672929868e-07, + "loss": 1.2021, + "step": 39101 + }, + { + "epoch": 8.53, + "learning_rate": 7.338868254315054e-07, + "loss": 0.5401, + "step": 39102 + }, + { + "epoch": 8.53, + "learning_rate": 7.33777583570024e-07, + "loss": 1.1168, + "step": 39103 + }, + { + "epoch": 8.53, + "learning_rate": 7.336683417085427e-07, + "loss": 1.5162, + "step": 39104 + }, + { + "epoch": 8.53, + "learning_rate": 7.335590998470613e-07, + "loss": 0.2865, + "step": 39105 + }, + { + "epoch": 8.53, + "learning_rate": 7.334498579855802e-07, + "loss": 0.0778, + "step": 39106 + }, + { + "epoch": 8.53, + "learning_rate": 7.333406161240989e-07, + "loss": 2.7024, + "step": 39107 + }, + { + "epoch": 8.54, + "learning_rate": 7.332313742626175e-07, + "loss": 3.0008, + "step": 39108 + }, + { + "epoch": 8.54, + "learning_rate": 7.331221324011362e-07, + "loss": 1.078, + "step": 39109 + }, + { + "epoch": 8.54, + "learning_rate": 7.330128905396548e-07, + "loss": 0.6604, + "step": 39110 + }, + { + "epoch": 8.54, + "learning_rate": 7.329036486781734e-07, + "loss": 0.9804, + "step": 39111 + }, + { + "epoch": 8.54, + "learning_rate": 7.327944068166923e-07, + "loss": 1.5352, + "step": 39112 + }, + { + "epoch": 8.54, + "learning_rate": 7.32685164955211e-07, + "loss": 1.1389, + "step": 39113 + }, + { + "epoch": 8.54, + "learning_rate": 7.325759230937296e-07, + "loss": 1.5592, + "step": 39114 + }, + { + "epoch": 8.54, + "learning_rate": 7.324666812322483e-07, + "loss": 1.6509, + "step": 39115 + }, + { + "epoch": 8.54, + "learning_rate": 7.323574393707669e-07, + "loss": 0.575, + "step": 39116 + }, + { + "epoch": 8.54, + "learning_rate": 7.322481975092855e-07, + "loss": 0.7549, + "step": 39117 + }, + { + "epoch": 8.54, + "learning_rate": 7.321389556478042e-07, + "loss": 0.9237, + "step": 39118 + }, + { + "epoch": 8.54, + "learning_rate": 7.320297137863231e-07, + "loss": 1.6126, + "step": 39119 + }, + { + "epoch": 8.54, + "learning_rate": 7.319204719248417e-07, + "loss": 1.4112, + "step": 39120 + }, + { + "epoch": 8.54, + "learning_rate": 7.318112300633604e-07, + "loss": 0.7218, + "step": 39121 + }, + { + "epoch": 8.54, + "learning_rate": 7.31701988201879e-07, + "loss": 1.325, + "step": 39122 + }, + { + "epoch": 8.54, + "learning_rate": 7.315927463403977e-07, + "loss": 0.0793, + "step": 39123 + }, + { + "epoch": 8.54, + "learning_rate": 7.314835044789163e-07, + "loss": 1.3117, + "step": 39124 + }, + { + "epoch": 8.54, + "learning_rate": 7.313742626174349e-07, + "loss": 1.3768, + "step": 39125 + }, + { + "epoch": 8.54, + "learning_rate": 7.312650207559538e-07, + "loss": 1.4239, + "step": 39126 + }, + { + "epoch": 8.54, + "learning_rate": 7.311557788944725e-07, + "loss": 0.7794, + "step": 39127 + }, + { + "epoch": 8.54, + "learning_rate": 7.310465370329911e-07, + "loss": 0.9895, + "step": 39128 + }, + { + "epoch": 8.54, + "learning_rate": 7.309372951715098e-07, + "loss": 1.3147, + "step": 39129 + }, + { + "epoch": 8.54, + "learning_rate": 7.308280533100284e-07, + "loss": 1.5575, + "step": 39130 + }, + { + "epoch": 8.54, + "learning_rate": 7.307188114485471e-07, + "loss": 0.6426, + "step": 39131 + }, + { + "epoch": 8.54, + "learning_rate": 7.306095695870659e-07, + "loss": 1.4706, + "step": 39132 + }, + { + "epoch": 8.54, + "learning_rate": 7.305003277255846e-07, + "loss": 0.097, + "step": 39133 + }, + { + "epoch": 8.54, + "learning_rate": 7.303910858641032e-07, + "loss": 0.1754, + "step": 39134 + }, + { + "epoch": 8.54, + "learning_rate": 7.302818440026219e-07, + "loss": 1.7193, + "step": 39135 + }, + { + "epoch": 8.54, + "learning_rate": 7.301726021411405e-07, + "loss": 0.1262, + "step": 39136 + }, + { + "epoch": 8.54, + "learning_rate": 7.300633602796592e-07, + "loss": 0.5153, + "step": 39137 + }, + { + "epoch": 8.54, + "learning_rate": 7.299541184181778e-07, + "loss": 2.0011, + "step": 39138 + }, + { + "epoch": 8.54, + "learning_rate": 7.298448765566967e-07, + "loss": 1.2076, + "step": 39139 + }, + { + "epoch": 8.54, + "learning_rate": 7.297356346952153e-07, + "loss": 0.176, + "step": 39140 + }, + { + "epoch": 8.54, + "learning_rate": 7.29626392833734e-07, + "loss": 0.5615, + "step": 39141 + }, + { + "epoch": 8.54, + "learning_rate": 7.295171509722526e-07, + "loss": 1.1183, + "step": 39142 + }, + { + "epoch": 8.54, + "learning_rate": 7.294079091107713e-07, + "loss": 0.7671, + "step": 39143 + }, + { + "epoch": 8.54, + "learning_rate": 7.292986672492899e-07, + "loss": 0.5504, + "step": 39144 + }, + { + "epoch": 8.54, + "learning_rate": 7.291894253878088e-07, + "loss": 0.2567, + "step": 39145 + }, + { + "epoch": 8.54, + "learning_rate": 7.290801835263274e-07, + "loss": 0.6493, + "step": 39146 + }, + { + "epoch": 8.54, + "learning_rate": 7.289709416648461e-07, + "loss": 0.9184, + "step": 39147 + }, + { + "epoch": 8.54, + "learning_rate": 7.288616998033647e-07, + "loss": 1.5823, + "step": 39148 + }, + { + "epoch": 8.54, + "learning_rate": 7.287524579418834e-07, + "loss": 0.4899, + "step": 39149 + }, + { + "epoch": 8.54, + "learning_rate": 7.28643216080402e-07, + "loss": 0.8297, + "step": 39150 + }, + { + "epoch": 8.54, + "learning_rate": 7.285339742189207e-07, + "loss": 1.825, + "step": 39151 + }, + { + "epoch": 8.54, + "learning_rate": 7.284247323574395e-07, + "loss": 1.3489, + "step": 39152 + }, + { + "epoch": 8.54, + "learning_rate": 7.283154904959582e-07, + "loss": 1.3084, + "step": 39153 + }, + { + "epoch": 8.55, + "learning_rate": 7.282062486344768e-07, + "loss": 0.3429, + "step": 39154 + }, + { + "epoch": 8.55, + "learning_rate": 7.280970067729955e-07, + "loss": 1.3414, + "step": 39155 + }, + { + "epoch": 8.55, + "learning_rate": 7.279877649115141e-07, + "loss": 1.5366, + "step": 39156 + }, + { + "epoch": 8.55, + "learning_rate": 7.278785230500328e-07, + "loss": 0.5021, + "step": 39157 + }, + { + "epoch": 8.55, + "learning_rate": 7.277692811885514e-07, + "loss": 0.8516, + "step": 39158 + }, + { + "epoch": 8.55, + "learning_rate": 7.276600393270703e-07, + "loss": 1.4038, + "step": 39159 + }, + { + "epoch": 8.55, + "learning_rate": 7.275507974655889e-07, + "loss": 0.6193, + "step": 39160 + }, + { + "epoch": 8.55, + "learning_rate": 7.274415556041076e-07, + "loss": 2.0997, + "step": 39161 + }, + { + "epoch": 8.55, + "learning_rate": 7.273323137426262e-07, + "loss": 0.6848, + "step": 39162 + }, + { + "epoch": 8.55, + "learning_rate": 7.272230718811449e-07, + "loss": 0.8414, + "step": 39163 + }, + { + "epoch": 8.55, + "learning_rate": 7.271138300196635e-07, + "loss": 0.5039, + "step": 39164 + }, + { + "epoch": 8.55, + "learning_rate": 7.270045881581823e-07, + "loss": 1.6457, + "step": 39165 + }, + { + "epoch": 8.55, + "learning_rate": 7.26895346296701e-07, + "loss": 2.3128, + "step": 39166 + }, + { + "epoch": 8.55, + "learning_rate": 7.267861044352197e-07, + "loss": 0.505, + "step": 39167 + }, + { + "epoch": 8.55, + "learning_rate": 7.266768625737383e-07, + "loss": 2.027, + "step": 39168 + }, + { + "epoch": 8.55, + "learning_rate": 7.26567620712257e-07, + "loss": 1.2065, + "step": 39169 + }, + { + "epoch": 8.55, + "learning_rate": 7.264583788507756e-07, + "loss": 0.6755, + "step": 39170 + }, + { + "epoch": 8.55, + "learning_rate": 7.263491369892943e-07, + "loss": 0.1743, + "step": 39171 + }, + { + "epoch": 8.55, + "learning_rate": 7.26239895127813e-07, + "loss": 0.5014, + "step": 39172 + }, + { + "epoch": 8.55, + "learning_rate": 7.261306532663318e-07, + "loss": 0.868, + "step": 39173 + }, + { + "epoch": 8.55, + "learning_rate": 7.260214114048504e-07, + "loss": 1.178, + "step": 39174 + }, + { + "epoch": 8.55, + "learning_rate": 7.259121695433691e-07, + "loss": 0.5023, + "step": 39175 + }, + { + "epoch": 8.55, + "learning_rate": 7.258029276818877e-07, + "loss": 0.7812, + "step": 39176 + }, + { + "epoch": 8.55, + "learning_rate": 7.256936858204064e-07, + "loss": 1.6363, + "step": 39177 + }, + { + "epoch": 8.55, + "learning_rate": 7.255844439589251e-07, + "loss": 0.87, + "step": 39178 + }, + { + "epoch": 8.55, + "learning_rate": 7.254752020974439e-07, + "loss": 0.6203, + "step": 39179 + }, + { + "epoch": 8.55, + "learning_rate": 7.253659602359625e-07, + "loss": 1.5669, + "step": 39180 + }, + { + "epoch": 8.55, + "learning_rate": 7.252567183744812e-07, + "loss": 1.0866, + "step": 39181 + }, + { + "epoch": 8.55, + "learning_rate": 7.251474765129998e-07, + "loss": 0.5415, + "step": 39182 + }, + { + "epoch": 8.55, + "learning_rate": 7.250382346515185e-07, + "loss": 3.2899, + "step": 39183 + }, + { + "epoch": 8.55, + "learning_rate": 7.249289927900371e-07, + "loss": 0.3189, + "step": 39184 + }, + { + "epoch": 8.55, + "learning_rate": 7.248197509285559e-07, + "loss": 0.4941, + "step": 39185 + }, + { + "epoch": 8.55, + "learning_rate": 7.247105090670746e-07, + "loss": 1.9069, + "step": 39186 + }, + { + "epoch": 8.55, + "learning_rate": 7.246012672055933e-07, + "loss": 0.7428, + "step": 39187 + }, + { + "epoch": 8.55, + "learning_rate": 7.244920253441119e-07, + "loss": 0.9006, + "step": 39188 + }, + { + "epoch": 8.55, + "learning_rate": 7.243827834826306e-07, + "loss": 1.3754, + "step": 39189 + }, + { + "epoch": 8.55, + "learning_rate": 7.242735416211492e-07, + "loss": 1.5111, + "step": 39190 + }, + { + "epoch": 8.55, + "learning_rate": 7.241642997596679e-07, + "loss": 1.1526, + "step": 39191 + }, + { + "epoch": 8.55, + "learning_rate": 7.240550578981866e-07, + "loss": 0.8014, + "step": 39192 + }, + { + "epoch": 8.55, + "learning_rate": 7.239458160367054e-07, + "loss": 2.9519, + "step": 39193 + }, + { + "epoch": 8.55, + "learning_rate": 7.23836574175224e-07, + "loss": 0.5464, + "step": 39194 + }, + { + "epoch": 8.55, + "learning_rate": 7.237273323137427e-07, + "loss": 3.1893, + "step": 39195 + }, + { + "epoch": 8.55, + "learning_rate": 7.236180904522613e-07, + "loss": 1.5865, + "step": 39196 + }, + { + "epoch": 8.55, + "learning_rate": 7.2350884859078e-07, + "loss": 2.6543, + "step": 39197 + }, + { + "epoch": 8.55, + "learning_rate": 7.233996067292988e-07, + "loss": 1.7381, + "step": 39198 + }, + { + "epoch": 8.55, + "learning_rate": 7.232903648678175e-07, + "loss": 1.1753, + "step": 39199 + }, + { + "epoch": 8.56, + "learning_rate": 7.231811230063361e-07, + "loss": 2.026, + "step": 39200 + }, + { + "epoch": 8.56, + "learning_rate": 7.230718811448548e-07, + "loss": 1.08, + "step": 39201 + }, + { + "epoch": 8.56, + "learning_rate": 7.229626392833734e-07, + "loss": 0.4771, + "step": 39202 + }, + { + "epoch": 8.56, + "learning_rate": 7.228533974218921e-07, + "loss": 0.9278, + "step": 39203 + }, + { + "epoch": 8.56, + "learning_rate": 7.227441555604107e-07, + "loss": 1.7336, + "step": 39204 + }, + { + "epoch": 8.56, + "learning_rate": 7.226349136989295e-07, + "loss": 1.0756, + "step": 39205 + }, + { + "epoch": 8.56, + "learning_rate": 7.225256718374483e-07, + "loss": 0.3749, + "step": 39206 + }, + { + "epoch": 8.56, + "learning_rate": 7.224164299759669e-07, + "loss": 0.6726, + "step": 39207 + }, + { + "epoch": 8.56, + "learning_rate": 7.223071881144855e-07, + "loss": 1.1361, + "step": 39208 + }, + { + "epoch": 8.56, + "learning_rate": 7.221979462530042e-07, + "loss": 0.2336, + "step": 39209 + }, + { + "epoch": 8.56, + "learning_rate": 7.220887043915228e-07, + "loss": 1.9313, + "step": 39210 + }, + { + "epoch": 8.56, + "learning_rate": 7.219794625300415e-07, + "loss": 1.2869, + "step": 39211 + }, + { + "epoch": 8.56, + "learning_rate": 7.218702206685603e-07, + "loss": 0.8959, + "step": 39212 + }, + { + "epoch": 8.56, + "learning_rate": 7.21760978807079e-07, + "loss": 0.751, + "step": 39213 + }, + { + "epoch": 8.56, + "learning_rate": 7.216517369455976e-07, + "loss": 1.2105, + "step": 39214 + }, + { + "epoch": 8.56, + "learning_rate": 7.215424950841163e-07, + "loss": 0.5221, + "step": 39215 + }, + { + "epoch": 8.56, + "learning_rate": 7.214332532226349e-07, + "loss": 1.5417, + "step": 39216 + }, + { + "epoch": 8.56, + "learning_rate": 7.213240113611536e-07, + "loss": 0.987, + "step": 39217 + }, + { + "epoch": 8.56, + "learning_rate": 7.212147694996724e-07, + "loss": 1.3702, + "step": 39218 + }, + { + "epoch": 8.56, + "learning_rate": 7.21105527638191e-07, + "loss": 0.7775, + "step": 39219 + }, + { + "epoch": 8.56, + "learning_rate": 7.209962857767098e-07, + "loss": 2.7779, + "step": 39220 + }, + { + "epoch": 8.56, + "learning_rate": 7.208870439152284e-07, + "loss": 1.3952, + "step": 39221 + }, + { + "epoch": 8.56, + "learning_rate": 7.20777802053747e-07, + "loss": 1.5837, + "step": 39222 + }, + { + "epoch": 8.56, + "learning_rate": 7.206685601922657e-07, + "loss": 0.8797, + "step": 39223 + }, + { + "epoch": 8.56, + "learning_rate": 7.205593183307843e-07, + "loss": 1.1036, + "step": 39224 + }, + { + "epoch": 8.56, + "learning_rate": 7.204500764693031e-07, + "loss": 1.6373, + "step": 39225 + }, + { + "epoch": 8.56, + "learning_rate": 7.203408346078219e-07, + "loss": 0.4898, + "step": 39226 + }, + { + "epoch": 8.56, + "learning_rate": 7.202315927463405e-07, + "loss": 0.6604, + "step": 39227 + }, + { + "epoch": 8.56, + "learning_rate": 7.201223508848591e-07, + "loss": 0.5756, + "step": 39228 + }, + { + "epoch": 8.56, + "learning_rate": 7.200131090233778e-07, + "loss": 1.1501, + "step": 39229 + }, + { + "epoch": 8.56, + "learning_rate": 7.199038671618964e-07, + "loss": 1.3228, + "step": 39230 + }, + { + "epoch": 8.56, + "learning_rate": 7.197946253004152e-07, + "loss": 2.9288, + "step": 39231 + }, + { + "epoch": 8.56, + "learning_rate": 7.196853834389339e-07, + "loss": 1.6857, + "step": 39232 + }, + { + "epoch": 8.56, + "learning_rate": 7.195761415774526e-07, + "loss": 1.287, + "step": 39233 + }, + { + "epoch": 8.56, + "learning_rate": 7.194668997159713e-07, + "loss": 0.3311, + "step": 39234 + }, + { + "epoch": 8.56, + "learning_rate": 7.193576578544899e-07, + "loss": 1.2464, + "step": 39235 + }, + { + "epoch": 8.56, + "learning_rate": 7.192484159930085e-07, + "loss": 1.1287, + "step": 39236 + }, + { + "epoch": 8.56, + "learning_rate": 7.191391741315272e-07, + "loss": 1.3422, + "step": 39237 + }, + { + "epoch": 8.56, + "learning_rate": 7.19029932270046e-07, + "loss": 0.2485, + "step": 39238 + }, + { + "epoch": 8.56, + "learning_rate": 7.189206904085646e-07, + "loss": 3.0234, + "step": 39239 + }, + { + "epoch": 8.56, + "learning_rate": 7.188114485470834e-07, + "loss": 2.8382, + "step": 39240 + }, + { + "epoch": 8.56, + "learning_rate": 7.18702206685602e-07, + "loss": 0.2536, + "step": 39241 + }, + { + "epoch": 8.56, + "learning_rate": 7.185929648241207e-07, + "loss": 2.4089, + "step": 39242 + }, + { + "epoch": 8.56, + "learning_rate": 7.184837229626393e-07, + "loss": 1.2548, + "step": 39243 + }, + { + "epoch": 8.56, + "learning_rate": 7.183744811011579e-07, + "loss": 0.7007, + "step": 39244 + }, + { + "epoch": 8.57, + "learning_rate": 7.182652392396767e-07, + "loss": 0.8268, + "step": 39245 + }, + { + "epoch": 8.57, + "learning_rate": 7.181559973781955e-07, + "loss": 0.6396, + "step": 39246 + }, + { + "epoch": 8.57, + "learning_rate": 7.180467555167141e-07, + "loss": 1.452, + "step": 39247 + }, + { + "epoch": 8.57, + "learning_rate": 7.179375136552328e-07, + "loss": 3.1085, + "step": 39248 + }, + { + "epoch": 8.57, + "learning_rate": 7.178282717937514e-07, + "loss": 0.2211, + "step": 39249 + }, + { + "epoch": 8.57, + "learning_rate": 7.1771902993227e-07, + "loss": 0.0788, + "step": 39250 + }, + { + "epoch": 8.57, + "learning_rate": 7.176097880707888e-07, + "loss": 3.0588, + "step": 39251 + }, + { + "epoch": 8.57, + "learning_rate": 7.175005462093075e-07, + "loss": 1.3301, + "step": 39252 + }, + { + "epoch": 8.57, + "learning_rate": 7.173913043478262e-07, + "loss": 2.1316, + "step": 39253 + }, + { + "epoch": 8.57, + "learning_rate": 7.172820624863449e-07, + "loss": 0.5014, + "step": 39254 + }, + { + "epoch": 8.57, + "learning_rate": 7.171728206248635e-07, + "loss": 0.8942, + "step": 39255 + }, + { + "epoch": 8.57, + "learning_rate": 7.170635787633822e-07, + "loss": 0.2296, + "step": 39256 + }, + { + "epoch": 8.57, + "learning_rate": 7.169543369019008e-07, + "loss": 1.291, + "step": 39257 + }, + { + "epoch": 8.57, + "learning_rate": 7.168450950404196e-07, + "loss": 1.8212, + "step": 39258 + }, + { + "epoch": 8.57, + "learning_rate": 7.167358531789382e-07, + "loss": 0.3422, + "step": 39259 + }, + { + "epoch": 8.57, + "learning_rate": 7.16626611317457e-07, + "loss": 2.9117, + "step": 39260 + }, + { + "epoch": 8.57, + "learning_rate": 7.165173694559756e-07, + "loss": 0.604, + "step": 39261 + }, + { + "epoch": 8.57, + "learning_rate": 7.164081275944943e-07, + "loss": 3.5778, + "step": 39262 + }, + { + "epoch": 8.57, + "learning_rate": 7.162988857330129e-07, + "loss": 0.1809, + "step": 39263 + }, + { + "epoch": 8.57, + "learning_rate": 7.161896438715317e-07, + "loss": 0.1581, + "step": 39264 + }, + { + "epoch": 8.57, + "learning_rate": 7.160804020100503e-07, + "loss": 0.059, + "step": 39265 + }, + { + "epoch": 8.57, + "learning_rate": 7.15971160148569e-07, + "loss": 0.4131, + "step": 39266 + }, + { + "epoch": 8.57, + "learning_rate": 7.158619182870877e-07, + "loss": 0.7943, + "step": 39267 + }, + { + "epoch": 8.57, + "learning_rate": 7.157526764256064e-07, + "loss": 1.7934, + "step": 39268 + }, + { + "epoch": 8.57, + "learning_rate": 7.15643434564125e-07, + "loss": 1.5658, + "step": 39269 + }, + { + "epoch": 8.57, + "learning_rate": 7.155341927026437e-07, + "loss": 0.9575, + "step": 39270 + }, + { + "epoch": 8.57, + "learning_rate": 7.154249508411624e-07, + "loss": 1.7473, + "step": 39271 + }, + { + "epoch": 8.57, + "learning_rate": 7.153157089796811e-07, + "loss": 2.1126, + "step": 39272 + }, + { + "epoch": 8.57, + "learning_rate": 7.152064671181997e-07, + "loss": 0.0348, + "step": 39273 + }, + { + "epoch": 8.57, + "learning_rate": 7.150972252567185e-07, + "loss": 2.0335, + "step": 39274 + }, + { + "epoch": 8.57, + "learning_rate": 7.149879833952371e-07, + "loss": 2.5417, + "step": 39275 + }, + { + "epoch": 8.57, + "learning_rate": 7.148787415337558e-07, + "loss": 0.3699, + "step": 39276 + }, + { + "epoch": 8.57, + "learning_rate": 7.147694996722744e-07, + "loss": 1.5461, + "step": 39277 + }, + { + "epoch": 8.57, + "learning_rate": 7.146602578107932e-07, + "loss": 0.8813, + "step": 39278 + }, + { + "epoch": 8.57, + "learning_rate": 7.145510159493118e-07, + "loss": 1.3395, + "step": 39279 + }, + { + "epoch": 8.57, + "learning_rate": 7.144417740878306e-07, + "loss": 0.2671, + "step": 39280 + }, + { + "epoch": 8.57, + "learning_rate": 7.143325322263492e-07, + "loss": 1.7836, + "step": 39281 + }, + { + "epoch": 8.57, + "learning_rate": 7.142232903648679e-07, + "loss": 3.3529, + "step": 39282 + }, + { + "epoch": 8.57, + "learning_rate": 7.141140485033865e-07, + "loss": 1.5891, + "step": 39283 + }, + { + "epoch": 8.57, + "learning_rate": 7.140048066419053e-07, + "loss": 0.8739, + "step": 39284 + }, + { + "epoch": 8.57, + "learning_rate": 7.138955647804239e-07, + "loss": 2.1498, + "step": 39285 + }, + { + "epoch": 8.57, + "learning_rate": 7.137863229189426e-07, + "loss": 3.6119, + "step": 39286 + }, + { + "epoch": 8.57, + "learning_rate": 7.136770810574613e-07, + "loss": 0.8445, + "step": 39287 + }, + { + "epoch": 8.57, + "learning_rate": 7.1356783919598e-07, + "loss": 0.21, + "step": 39288 + }, + { + "epoch": 8.57, + "learning_rate": 7.134585973344986e-07, + "loss": 1.1488, + "step": 39289 + }, + { + "epoch": 8.57, + "learning_rate": 7.133493554730173e-07, + "loss": 3.0307, + "step": 39290 + }, + { + "epoch": 8.58, + "learning_rate": 7.13240113611536e-07, + "loss": 0.5018, + "step": 39291 + }, + { + "epoch": 8.58, + "learning_rate": 7.131308717500547e-07, + "loss": 2.1199, + "step": 39292 + }, + { + "epoch": 8.58, + "learning_rate": 7.130216298885733e-07, + "loss": 0.2226, + "step": 39293 + }, + { + "epoch": 8.58, + "learning_rate": 7.129123880270921e-07, + "loss": 0.3868, + "step": 39294 + }, + { + "epoch": 8.58, + "learning_rate": 7.128031461656107e-07, + "loss": 0.14, + "step": 39295 + }, + { + "epoch": 8.58, + "learning_rate": 7.126939043041294e-07, + "loss": 1.2688, + "step": 39296 + }, + { + "epoch": 8.58, + "learning_rate": 7.12584662442648e-07, + "loss": 2.0929, + "step": 39297 + }, + { + "epoch": 8.58, + "learning_rate": 7.124754205811668e-07, + "loss": 0.985, + "step": 39298 + }, + { + "epoch": 8.58, + "learning_rate": 7.123661787196854e-07, + "loss": 1.2317, + "step": 39299 + }, + { + "epoch": 8.58, + "learning_rate": 7.122569368582041e-07, + "loss": 2.392, + "step": 39300 + }, + { + "epoch": 8.58, + "learning_rate": 7.121476949967228e-07, + "loss": 1.7957, + "step": 39301 + }, + { + "epoch": 8.58, + "learning_rate": 7.120384531352415e-07, + "loss": 1.0672, + "step": 39302 + }, + { + "epoch": 8.58, + "learning_rate": 7.119292112737601e-07, + "loss": 0.5873, + "step": 39303 + }, + { + "epoch": 8.58, + "learning_rate": 7.118199694122789e-07, + "loss": 1.6027, + "step": 39304 + }, + { + "epoch": 8.58, + "learning_rate": 7.117107275507975e-07, + "loss": 1.7697, + "step": 39305 + }, + { + "epoch": 8.58, + "learning_rate": 7.116014856893162e-07, + "loss": 0.1304, + "step": 39306 + }, + { + "epoch": 8.58, + "learning_rate": 7.114922438278349e-07, + "loss": 0.426, + "step": 39307 + }, + { + "epoch": 8.58, + "learning_rate": 7.113830019663536e-07, + "loss": 0.153, + "step": 39308 + }, + { + "epoch": 8.58, + "learning_rate": 7.112737601048722e-07, + "loss": 1.9456, + "step": 39309 + }, + { + "epoch": 8.58, + "learning_rate": 7.111645182433909e-07, + "loss": 0.5625, + "step": 39310 + }, + { + "epoch": 8.58, + "learning_rate": 7.110552763819096e-07, + "loss": 1.0193, + "step": 39311 + }, + { + "epoch": 8.58, + "learning_rate": 7.109460345204283e-07, + "loss": 1.3204, + "step": 39312 + }, + { + "epoch": 8.58, + "learning_rate": 7.108367926589469e-07, + "loss": 0.5106, + "step": 39313 + }, + { + "epoch": 8.58, + "learning_rate": 7.107275507974657e-07, + "loss": 0.9273, + "step": 39314 + }, + { + "epoch": 8.58, + "learning_rate": 7.106183089359843e-07, + "loss": 1.3855, + "step": 39315 + }, + { + "epoch": 8.58, + "learning_rate": 7.10509067074503e-07, + "loss": 4.1792, + "step": 39316 + }, + { + "epoch": 8.58, + "learning_rate": 7.103998252130218e-07, + "loss": 1.2262, + "step": 39317 + }, + { + "epoch": 8.58, + "learning_rate": 7.102905833515404e-07, + "loss": 2.1394, + "step": 39318 + }, + { + "epoch": 8.58, + "learning_rate": 7.10181341490059e-07, + "loss": 1.4946, + "step": 39319 + }, + { + "epoch": 8.58, + "learning_rate": 7.100720996285777e-07, + "loss": 2.6708, + "step": 39320 + }, + { + "epoch": 8.58, + "learning_rate": 7.099628577670964e-07, + "loss": 1.8948, + "step": 39321 + }, + { + "epoch": 8.58, + "learning_rate": 7.098536159056151e-07, + "loss": 3.4609, + "step": 39322 + }, + { + "epoch": 8.58, + "learning_rate": 7.097443740441337e-07, + "loss": 1.9793, + "step": 39323 + }, + { + "epoch": 8.58, + "learning_rate": 7.096351321826525e-07, + "loss": 1.2803, + "step": 39324 + }, + { + "epoch": 8.58, + "learning_rate": 7.095258903211711e-07, + "loss": 1.4099, + "step": 39325 + }, + { + "epoch": 8.58, + "learning_rate": 7.094166484596898e-07, + "loss": 2.1451, + "step": 39326 + }, + { + "epoch": 8.58, + "learning_rate": 7.093074065982084e-07, + "loss": 0.0672, + "step": 39327 + }, + { + "epoch": 8.58, + "learning_rate": 7.091981647367272e-07, + "loss": 1.1219, + "step": 39328 + }, + { + "epoch": 8.58, + "learning_rate": 7.090889228752458e-07, + "loss": 1.5761, + "step": 39329 + }, + { + "epoch": 8.58, + "learning_rate": 7.089796810137645e-07, + "loss": 0.4805, + "step": 39330 + }, + { + "epoch": 8.58, + "learning_rate": 7.088704391522833e-07, + "loss": 2.2383, + "step": 39331 + }, + { + "epoch": 8.58, + "learning_rate": 7.087611972908019e-07, + "loss": 0.8782, + "step": 39332 + }, + { + "epoch": 8.58, + "learning_rate": 7.086519554293205e-07, + "loss": 1.5222, + "step": 39333 + }, + { + "epoch": 8.58, + "learning_rate": 7.085427135678393e-07, + "loss": 1.6181, + "step": 39334 + }, + { + "epoch": 8.58, + "learning_rate": 7.084334717063579e-07, + "loss": 0.3855, + "step": 39335 + }, + { + "epoch": 8.58, + "learning_rate": 7.083242298448766e-07, + "loss": 0.6745, + "step": 39336 + }, + { + "epoch": 8.59, + "learning_rate": 7.082149879833954e-07, + "loss": 0.9463, + "step": 39337 + }, + { + "epoch": 8.59, + "learning_rate": 7.08105746121914e-07, + "loss": 2.2169, + "step": 39338 + }, + { + "epoch": 8.59, + "learning_rate": 7.079965042604326e-07, + "loss": 1.0119, + "step": 39339 + }, + { + "epoch": 8.59, + "learning_rate": 7.078872623989513e-07, + "loss": 1.1837, + "step": 39340 + }, + { + "epoch": 8.59, + "learning_rate": 7.0777802053747e-07, + "loss": 2.3419, + "step": 39341 + }, + { + "epoch": 8.59, + "learning_rate": 7.076687786759887e-07, + "loss": 0.126, + "step": 39342 + }, + { + "epoch": 8.59, + "learning_rate": 7.075595368145073e-07, + "loss": 0.8636, + "step": 39343 + }, + { + "epoch": 8.59, + "learning_rate": 7.074502949530261e-07, + "loss": 1.4715, + "step": 39344 + }, + { + "epoch": 8.59, + "learning_rate": 7.073410530915448e-07, + "loss": 1.1832, + "step": 39345 + }, + { + "epoch": 8.59, + "learning_rate": 7.072318112300634e-07, + "loss": 2.7277, + "step": 39346 + }, + { + "epoch": 8.59, + "learning_rate": 7.07122569368582e-07, + "loss": 1.9167, + "step": 39347 + }, + { + "epoch": 8.59, + "learning_rate": 7.070133275071008e-07, + "loss": 2.2001, + "step": 39348 + }, + { + "epoch": 8.59, + "learning_rate": 7.069040856456194e-07, + "loss": 2.13, + "step": 39349 + }, + { + "epoch": 8.59, + "learning_rate": 7.067948437841382e-07, + "loss": 0.4884, + "step": 39350 + }, + { + "epoch": 8.59, + "learning_rate": 7.066856019226569e-07, + "loss": 3.8755, + "step": 39351 + }, + { + "epoch": 8.59, + "learning_rate": 7.065763600611755e-07, + "loss": 2.7173, + "step": 39352 + }, + { + "epoch": 8.59, + "learning_rate": 7.064671181996942e-07, + "loss": 0.8737, + "step": 39353 + }, + { + "epoch": 8.59, + "learning_rate": 7.063578763382128e-07, + "loss": 2.6541, + "step": 39354 + }, + { + "epoch": 8.59, + "learning_rate": 7.062486344767315e-07, + "loss": 2.84, + "step": 39355 + }, + { + "epoch": 8.59, + "learning_rate": 7.061393926152502e-07, + "loss": 2.1917, + "step": 39356 + }, + { + "epoch": 8.59, + "learning_rate": 7.06030150753769e-07, + "loss": 3.2154, + "step": 39357 + }, + { + "epoch": 8.59, + "learning_rate": 7.059209088922876e-07, + "loss": 1.6674, + "step": 39358 + }, + { + "epoch": 8.59, + "learning_rate": 7.058116670308063e-07, + "loss": 0.424, + "step": 39359 + }, + { + "epoch": 8.59, + "learning_rate": 7.057024251693249e-07, + "loss": 0.8724, + "step": 39360 + }, + { + "epoch": 8.59, + "learning_rate": 7.055931833078435e-07, + "loss": 1.6519, + "step": 39361 + }, + { + "epoch": 8.59, + "learning_rate": 7.054839414463623e-07, + "loss": 0.2341, + "step": 39362 + }, + { + "epoch": 8.59, + "learning_rate": 7.053746995848809e-07, + "loss": 1.8294, + "step": 39363 + }, + { + "epoch": 8.59, + "learning_rate": 7.052654577233997e-07, + "loss": 2.0246, + "step": 39364 + }, + { + "epoch": 8.59, + "learning_rate": 7.051562158619184e-07, + "loss": 1.7404, + "step": 39365 + }, + { + "epoch": 8.59, + "learning_rate": 7.05046974000437e-07, + "loss": 1.4574, + "step": 39366 + }, + { + "epoch": 8.59, + "learning_rate": 7.049377321389557e-07, + "loss": 0.4291, + "step": 39367 + }, + { + "epoch": 8.59, + "learning_rate": 7.048284902774744e-07, + "loss": 0.6425, + "step": 39368 + }, + { + "epoch": 8.59, + "learning_rate": 7.04719248415993e-07, + "loss": 1.2117, + "step": 39369 + }, + { + "epoch": 8.59, + "learning_rate": 7.046100065545118e-07, + "loss": 1.6594, + "step": 39370 + }, + { + "epoch": 8.59, + "learning_rate": 7.045007646930305e-07, + "loss": 1.707, + "step": 39371 + }, + { + "epoch": 8.59, + "learning_rate": 7.043915228315491e-07, + "loss": 0.646, + "step": 39372 + }, + { + "epoch": 8.59, + "learning_rate": 7.042822809700678e-07, + "loss": 0.7495, + "step": 39373 + }, + { + "epoch": 8.59, + "learning_rate": 7.041730391085864e-07, + "loss": 2.5211, + "step": 39374 + }, + { + "epoch": 8.59, + "learning_rate": 7.040637972471051e-07, + "loss": 1.7309, + "step": 39375 + }, + { + "epoch": 8.59, + "learning_rate": 7.039545553856238e-07, + "loss": 0.1045, + "step": 39376 + }, + { + "epoch": 8.59, + "learning_rate": 7.038453135241426e-07, + "loss": 0.8195, + "step": 39377 + }, + { + "epoch": 8.59, + "learning_rate": 7.037360716626612e-07, + "loss": 0.8481, + "step": 39378 + }, + { + "epoch": 8.59, + "learning_rate": 7.036268298011799e-07, + "loss": 3.3673, + "step": 39379 + }, + { + "epoch": 8.59, + "learning_rate": 7.035175879396985e-07, + "loss": 0.7912, + "step": 39380 + }, + { + "epoch": 8.59, + "learning_rate": 7.034083460782173e-07, + "loss": 1.077, + "step": 39381 + }, + { + "epoch": 8.59, + "learning_rate": 7.032991042167359e-07, + "loss": 0.3827, + "step": 39382 + }, + { + "epoch": 8.6, + "learning_rate": 7.031898623552547e-07, + "loss": 1.02, + "step": 39383 + }, + { + "epoch": 8.6, + "learning_rate": 7.030806204937733e-07, + "loss": 1.9401, + "step": 39384 + }, + { + "epoch": 8.6, + "learning_rate": 7.02971378632292e-07, + "loss": 2.7948, + "step": 39385 + }, + { + "epoch": 8.6, + "learning_rate": 7.028621367708106e-07, + "loss": 1.3911, + "step": 39386 + }, + { + "epoch": 8.6, + "learning_rate": 7.027528949093293e-07, + "loss": 0.8575, + "step": 39387 + }, + { + "epoch": 8.6, + "learning_rate": 7.02643653047848e-07, + "loss": 2.7661, + "step": 39388 + }, + { + "epoch": 8.6, + "learning_rate": 7.025344111863667e-07, + "loss": 1.8936, + "step": 39389 + }, + { + "epoch": 8.6, + "learning_rate": 7.024251693248854e-07, + "loss": 1.5778, + "step": 39390 + }, + { + "epoch": 8.6, + "learning_rate": 7.023159274634041e-07, + "loss": 1.5011, + "step": 39391 + }, + { + "epoch": 8.6, + "learning_rate": 7.022066856019227e-07, + "loss": 2.9981, + "step": 39392 + }, + { + "epoch": 8.6, + "learning_rate": 7.020974437404414e-07, + "loss": 1.2819, + "step": 39393 + }, + { + "epoch": 8.6, + "learning_rate": 7.0198820187896e-07, + "loss": 0.5015, + "step": 39394 + }, + { + "epoch": 8.6, + "learning_rate": 7.018789600174788e-07, + "loss": 1.6215, + "step": 39395 + }, + { + "epoch": 8.6, + "learning_rate": 7.017697181559974e-07, + "loss": 0.1985, + "step": 39396 + }, + { + "epoch": 8.6, + "learning_rate": 7.016604762945162e-07, + "loss": 1.0228, + "step": 39397 + }, + { + "epoch": 8.6, + "learning_rate": 7.015512344330348e-07, + "loss": 1.5286, + "step": 39398 + }, + { + "epoch": 8.6, + "learning_rate": 7.014419925715535e-07, + "loss": 0.8813, + "step": 39399 + }, + { + "epoch": 8.6, + "learning_rate": 7.013327507100721e-07, + "loss": 1.0693, + "step": 39400 + }, + { + "epoch": 8.6, + "learning_rate": 7.012235088485908e-07, + "loss": 0.441, + "step": 39401 + }, + { + "epoch": 8.6, + "learning_rate": 7.011142669871095e-07, + "loss": 0.1614, + "step": 39402 + }, + { + "epoch": 8.6, + "learning_rate": 7.010050251256283e-07, + "loss": 0.378, + "step": 39403 + }, + { + "epoch": 8.6, + "learning_rate": 7.008957832641469e-07, + "loss": 0.0787, + "step": 39404 + }, + { + "epoch": 8.6, + "learning_rate": 7.007865414026656e-07, + "loss": 0.1608, + "step": 39405 + }, + { + "epoch": 8.6, + "learning_rate": 7.006772995411842e-07, + "loss": 1.3956, + "step": 39406 + }, + { + "epoch": 8.6, + "learning_rate": 7.005680576797029e-07, + "loss": 0.6259, + "step": 39407 + }, + { + "epoch": 8.6, + "learning_rate": 7.004588158182215e-07, + "loss": 1.2474, + "step": 39408 + }, + { + "epoch": 8.6, + "learning_rate": 7.003495739567403e-07, + "loss": 1.6775, + "step": 39409 + }, + { + "epoch": 8.6, + "learning_rate": 7.00240332095259e-07, + "loss": 1.1915, + "step": 39410 + }, + { + "epoch": 8.6, + "learning_rate": 7.001310902337777e-07, + "loss": 0.886, + "step": 39411 + }, + { + "epoch": 8.6, + "learning_rate": 7.000218483722963e-07, + "loss": 1.5364, + "step": 39412 + }, + { + "epoch": 8.6, + "learning_rate": 6.99912606510815e-07, + "loss": 1.214, + "step": 39413 + }, + { + "epoch": 8.6, + "learning_rate": 6.998033646493336e-07, + "loss": 0.5706, + "step": 39414 + }, + { + "epoch": 8.6, + "learning_rate": 6.996941227878524e-07, + "loss": 0.5012, + "step": 39415 + }, + { + "epoch": 8.6, + "learning_rate": 6.99584880926371e-07, + "loss": 1.3223, + "step": 39416 + }, + { + "epoch": 8.6, + "learning_rate": 6.994756390648898e-07, + "loss": 3.0464, + "step": 39417 + }, + { + "epoch": 8.6, + "learning_rate": 6.993663972034084e-07, + "loss": 0.6139, + "step": 39418 + }, + { + "epoch": 8.6, + "learning_rate": 6.992571553419271e-07, + "loss": 1.6942, + "step": 39419 + }, + { + "epoch": 8.6, + "learning_rate": 6.991479134804457e-07, + "loss": 0.7738, + "step": 39420 + }, + { + "epoch": 8.6, + "learning_rate": 6.990386716189644e-07, + "loss": 1.2447, + "step": 39421 + }, + { + "epoch": 8.6, + "learning_rate": 6.989294297574831e-07, + "loss": 0.4061, + "step": 39422 + }, + { + "epoch": 8.6, + "learning_rate": 6.988201878960019e-07, + "loss": 1.4952, + "step": 39423 + }, + { + "epoch": 8.6, + "learning_rate": 6.987109460345205e-07, + "loss": 2.0688, + "step": 39424 + }, + { + "epoch": 8.6, + "learning_rate": 6.986017041730392e-07, + "loss": 1.3444, + "step": 39425 + }, + { + "epoch": 8.6, + "learning_rate": 6.984924623115578e-07, + "loss": 3.5762, + "step": 39426 + }, + { + "epoch": 8.6, + "learning_rate": 6.983832204500765e-07, + "loss": 1.7207, + "step": 39427 + }, + { + "epoch": 8.6, + "learning_rate": 6.982739785885951e-07, + "loss": 0.9045, + "step": 39428 + }, + { + "epoch": 8.61, + "learning_rate": 6.981647367271139e-07, + "loss": 1.0913, + "step": 39429 + }, + { + "epoch": 8.61, + "learning_rate": 6.980554948656326e-07, + "loss": 0.9519, + "step": 39430 + }, + { + "epoch": 8.61, + "learning_rate": 6.979462530041513e-07, + "loss": 2.6755, + "step": 39431 + }, + { + "epoch": 8.61, + "learning_rate": 6.978370111426699e-07, + "loss": 0.575, + "step": 39432 + }, + { + "epoch": 8.61, + "learning_rate": 6.977277692811886e-07, + "loss": 0.5139, + "step": 39433 + }, + { + "epoch": 8.61, + "learning_rate": 6.976185274197072e-07, + "loss": 1.4157, + "step": 39434 + }, + { + "epoch": 8.61, + "learning_rate": 6.97509285558226e-07, + "loss": 0.9043, + "step": 39435 + }, + { + "epoch": 8.61, + "learning_rate": 6.974000436967447e-07, + "loss": 0.9917, + "step": 39436 + }, + { + "epoch": 8.61, + "learning_rate": 6.972908018352634e-07, + "loss": 0.3411, + "step": 39437 + }, + { + "epoch": 8.61, + "learning_rate": 6.97181559973782e-07, + "loss": 0.4286, + "step": 39438 + }, + { + "epoch": 8.61, + "learning_rate": 6.970723181123007e-07, + "loss": 0.7888, + "step": 39439 + }, + { + "epoch": 8.61, + "learning_rate": 6.969630762508193e-07, + "loss": 0.5274, + "step": 39440 + }, + { + "epoch": 8.61, + "learning_rate": 6.96853834389338e-07, + "loss": 0.8021, + "step": 39441 + }, + { + "epoch": 8.61, + "learning_rate": 6.967445925278567e-07, + "loss": 1.9945, + "step": 39442 + }, + { + "epoch": 8.61, + "learning_rate": 6.966353506663755e-07, + "loss": 0.7669, + "step": 39443 + }, + { + "epoch": 8.61, + "learning_rate": 6.965261088048941e-07, + "loss": 0.8282, + "step": 39444 + }, + { + "epoch": 8.61, + "learning_rate": 6.964168669434128e-07, + "loss": 0.098, + "step": 39445 + }, + { + "epoch": 8.61, + "learning_rate": 6.963076250819314e-07, + "loss": 0.1137, + "step": 39446 + }, + { + "epoch": 8.61, + "learning_rate": 6.961983832204501e-07, + "loss": 0.9115, + "step": 39447 + }, + { + "epoch": 8.61, + "learning_rate": 6.960891413589687e-07, + "loss": 3.0638, + "step": 39448 + }, + { + "epoch": 8.61, + "learning_rate": 6.959798994974875e-07, + "loss": 0.5579, + "step": 39449 + }, + { + "epoch": 8.61, + "learning_rate": 6.958706576360062e-07, + "loss": 1.1869, + "step": 39450 + }, + { + "epoch": 8.61, + "learning_rate": 6.957614157745249e-07, + "loss": 1.1363, + "step": 39451 + }, + { + "epoch": 8.61, + "learning_rate": 6.956521739130435e-07, + "loss": 0.6507, + "step": 39452 + }, + { + "epoch": 8.61, + "learning_rate": 6.955429320515622e-07, + "loss": 0.3129, + "step": 39453 + }, + { + "epoch": 8.61, + "learning_rate": 6.954336901900808e-07, + "loss": 1.5962, + "step": 39454 + }, + { + "epoch": 8.61, + "learning_rate": 6.953244483285995e-07, + "loss": 1.268, + "step": 39455 + }, + { + "epoch": 8.61, + "learning_rate": 6.952152064671184e-07, + "loss": 1.1037, + "step": 39456 + }, + { + "epoch": 8.61, + "learning_rate": 6.95105964605637e-07, + "loss": 0.1956, + "step": 39457 + }, + { + "epoch": 8.61, + "learning_rate": 6.949967227441556e-07, + "loss": 2.8046, + "step": 39458 + }, + { + "epoch": 8.61, + "learning_rate": 6.948874808826743e-07, + "loss": 1.1542, + "step": 39459 + }, + { + "epoch": 8.61, + "learning_rate": 6.947782390211929e-07, + "loss": 1.49, + "step": 39460 + }, + { + "epoch": 8.61, + "learning_rate": 6.946689971597116e-07, + "loss": 1.9264, + "step": 39461 + }, + { + "epoch": 8.61, + "learning_rate": 6.945597552982302e-07, + "loss": 1.4847, + "step": 39462 + }, + { + "epoch": 8.61, + "learning_rate": 6.944505134367491e-07, + "loss": 0.1312, + "step": 39463 + }, + { + "epoch": 8.61, + "learning_rate": 6.943412715752677e-07, + "loss": 0.2293, + "step": 39464 + }, + { + "epoch": 8.61, + "learning_rate": 6.942320297137864e-07, + "loss": 1.1978, + "step": 39465 + }, + { + "epoch": 8.61, + "learning_rate": 6.94122787852305e-07, + "loss": 1.2228, + "step": 39466 + }, + { + "epoch": 8.61, + "learning_rate": 6.940135459908237e-07, + "loss": 1.4428, + "step": 39467 + }, + { + "epoch": 8.61, + "learning_rate": 6.939043041293423e-07, + "loss": 0.152, + "step": 39468 + }, + { + "epoch": 8.61, + "learning_rate": 6.937950622678612e-07, + "loss": 1.3591, + "step": 39469 + }, + { + "epoch": 8.61, + "learning_rate": 6.936858204063799e-07, + "loss": 0.6787, + "step": 39470 + }, + { + "epoch": 8.61, + "learning_rate": 6.935765785448985e-07, + "loss": 0.4539, + "step": 39471 + }, + { + "epoch": 8.61, + "learning_rate": 6.934673366834171e-07, + "loss": 0.9108, + "step": 39472 + }, + { + "epoch": 8.61, + "learning_rate": 6.933580948219358e-07, + "loss": 0.3233, + "step": 39473 + }, + { + "epoch": 8.62, + "learning_rate": 6.932488529604544e-07, + "loss": 0.7358, + "step": 39474 + }, + { + "epoch": 8.62, + "learning_rate": 6.931396110989731e-07, + "loss": 2.9147, + "step": 39475 + }, + { + "epoch": 8.62, + "learning_rate": 6.93030369237492e-07, + "loss": 1.0475, + "step": 39476 + }, + { + "epoch": 8.62, + "learning_rate": 6.929211273760106e-07, + "loss": 0.6591, + "step": 39477 + }, + { + "epoch": 8.62, + "learning_rate": 6.928118855145293e-07, + "loss": 0.5485, + "step": 39478 + }, + { + "epoch": 8.62, + "learning_rate": 6.927026436530479e-07, + "loss": 0.5947, + "step": 39479 + }, + { + "epoch": 8.62, + "learning_rate": 6.925934017915665e-07, + "loss": 0.9751, + "step": 39480 + }, + { + "epoch": 8.62, + "learning_rate": 6.924841599300852e-07, + "loss": 1.6791, + "step": 39481 + }, + { + "epoch": 8.62, + "learning_rate": 6.923749180686038e-07, + "loss": 0.6881, + "step": 39482 + }, + { + "epoch": 8.62, + "learning_rate": 6.922656762071227e-07, + "loss": 1.554, + "step": 39483 + }, + { + "epoch": 8.62, + "learning_rate": 6.921564343456414e-07, + "loss": 0.6575, + "step": 39484 + }, + { + "epoch": 8.62, + "learning_rate": 6.9204719248416e-07, + "loss": 0.569, + "step": 39485 + }, + { + "epoch": 8.62, + "learning_rate": 6.919379506226786e-07, + "loss": 1.3086, + "step": 39486 + }, + { + "epoch": 8.62, + "learning_rate": 6.918287087611973e-07, + "loss": 0.9787, + "step": 39487 + }, + { + "epoch": 8.62, + "learning_rate": 6.917194668997159e-07, + "loss": 0.2212, + "step": 39488 + }, + { + "epoch": 8.62, + "learning_rate": 6.916102250382348e-07, + "loss": 0.6759, + "step": 39489 + }, + { + "epoch": 8.62, + "learning_rate": 6.915009831767535e-07, + "loss": 0.4891, + "step": 39490 + }, + { + "epoch": 8.62, + "learning_rate": 6.913917413152721e-07, + "loss": 0.6073, + "step": 39491 + }, + { + "epoch": 8.62, + "learning_rate": 6.912824994537908e-07, + "loss": 0.5683, + "step": 39492 + }, + { + "epoch": 8.62, + "learning_rate": 6.911732575923094e-07, + "loss": 2.2284, + "step": 39493 + }, + { + "epoch": 8.62, + "learning_rate": 6.91064015730828e-07, + "loss": 0.688, + "step": 39494 + }, + { + "epoch": 8.62, + "learning_rate": 6.909547738693467e-07, + "loss": 0.0782, + "step": 39495 + }, + { + "epoch": 8.62, + "learning_rate": 6.908455320078656e-07, + "loss": 0.6653, + "step": 39496 + }, + { + "epoch": 8.62, + "learning_rate": 6.907362901463842e-07, + "loss": 0.4121, + "step": 39497 + }, + { + "epoch": 8.62, + "learning_rate": 6.906270482849029e-07, + "loss": 2.4189, + "step": 39498 + }, + { + "epoch": 8.62, + "learning_rate": 6.905178064234215e-07, + "loss": 0.6179, + "step": 39499 + }, + { + "epoch": 8.62, + "learning_rate": 6.904085645619401e-07, + "loss": 1.4684, + "step": 39500 + }, + { + "epoch": 8.62, + "learning_rate": 6.902993227004588e-07, + "loss": 1.1847, + "step": 39501 + }, + { + "epoch": 8.62, + "learning_rate": 6.901900808389777e-07, + "loss": 1.0127, + "step": 39502 + }, + { + "epoch": 8.62, + "learning_rate": 6.900808389774963e-07, + "loss": 3.7029, + "step": 39503 + }, + { + "epoch": 8.62, + "learning_rate": 6.89971597116015e-07, + "loss": 0.2498, + "step": 39504 + }, + { + "epoch": 8.62, + "learning_rate": 6.898623552545336e-07, + "loss": 0.7009, + "step": 39505 + }, + { + "epoch": 8.62, + "learning_rate": 6.897531133930523e-07, + "loss": 0.2968, + "step": 39506 + }, + { + "epoch": 8.62, + "learning_rate": 6.896438715315709e-07, + "loss": 1.0569, + "step": 39507 + }, + { + "epoch": 8.62, + "learning_rate": 6.895346296700895e-07, + "loss": 1.982, + "step": 39508 + }, + { + "epoch": 8.62, + "learning_rate": 6.894253878086084e-07, + "loss": 3.325, + "step": 39509 + }, + { + "epoch": 8.62, + "learning_rate": 6.893161459471271e-07, + "loss": 0.078, + "step": 39510 + }, + { + "epoch": 8.62, + "learning_rate": 6.892069040856457e-07, + "loss": 2.6575, + "step": 39511 + }, + { + "epoch": 8.62, + "learning_rate": 6.890976622241644e-07, + "loss": 0.7464, + "step": 39512 + }, + { + "epoch": 8.62, + "learning_rate": 6.88988420362683e-07, + "loss": 1.3892, + "step": 39513 + }, + { + "epoch": 8.62, + "learning_rate": 6.888791785012017e-07, + "loss": 2.1894, + "step": 39514 + }, + { + "epoch": 8.62, + "learning_rate": 6.887699366397203e-07, + "loss": 1.7818, + "step": 39515 + }, + { + "epoch": 8.62, + "learning_rate": 6.886606947782392e-07, + "loss": 1.0931, + "step": 39516 + }, + { + "epoch": 8.62, + "learning_rate": 6.885514529167578e-07, + "loss": 2.8924, + "step": 39517 + }, + { + "epoch": 8.62, + "learning_rate": 6.884422110552765e-07, + "loss": 1.874, + "step": 39518 + }, + { + "epoch": 8.62, + "learning_rate": 6.883329691937951e-07, + "loss": 1.5861, + "step": 39519 + }, + { + "epoch": 8.63, + "learning_rate": 6.882237273323138e-07, + "loss": 1.2649, + "step": 39520 + }, + { + "epoch": 8.63, + "learning_rate": 6.881144854708324e-07, + "loss": 1.0578, + "step": 39521 + }, + { + "epoch": 8.63, + "learning_rate": 6.880052436093512e-07, + "loss": 0.45, + "step": 39522 + }, + { + "epoch": 8.63, + "learning_rate": 6.878960017478699e-07, + "loss": 0.8072, + "step": 39523 + }, + { + "epoch": 8.63, + "learning_rate": 6.877867598863886e-07, + "loss": 0.8376, + "step": 39524 + }, + { + "epoch": 8.63, + "learning_rate": 6.876775180249072e-07, + "loss": 0.7422, + "step": 39525 + }, + { + "epoch": 8.63, + "learning_rate": 6.875682761634259e-07, + "loss": 0.521, + "step": 39526 + }, + { + "epoch": 8.63, + "learning_rate": 6.874590343019445e-07, + "loss": 0.7874, + "step": 39527 + }, + { + "epoch": 8.63, + "learning_rate": 6.873497924404632e-07, + "loss": 1.0957, + "step": 39528 + }, + { + "epoch": 8.63, + "learning_rate": 6.87240550578982e-07, + "loss": 1.108, + "step": 39529 + }, + { + "epoch": 8.63, + "learning_rate": 6.871313087175007e-07, + "loss": 0.5756, + "step": 39530 + }, + { + "epoch": 8.63, + "learning_rate": 6.870220668560193e-07, + "loss": 0.6991, + "step": 39531 + }, + { + "epoch": 8.63, + "learning_rate": 6.86912824994538e-07, + "loss": 1.8712, + "step": 39532 + }, + { + "epoch": 8.63, + "learning_rate": 6.868035831330566e-07, + "loss": 1.1347, + "step": 39533 + }, + { + "epoch": 8.63, + "learning_rate": 6.866943412715753e-07, + "loss": 0.792, + "step": 39534 + }, + { + "epoch": 8.63, + "learning_rate": 6.865850994100939e-07, + "loss": 1.5113, + "step": 39535 + }, + { + "epoch": 8.63, + "learning_rate": 6.864758575486128e-07, + "loss": 0.6008, + "step": 39536 + }, + { + "epoch": 8.63, + "learning_rate": 6.863666156871314e-07, + "loss": 2.2968, + "step": 39537 + }, + { + "epoch": 8.63, + "learning_rate": 6.862573738256501e-07, + "loss": 0.6203, + "step": 39538 + }, + { + "epoch": 8.63, + "learning_rate": 6.861481319641687e-07, + "loss": 0.1756, + "step": 39539 + }, + { + "epoch": 8.63, + "learning_rate": 6.860388901026874e-07, + "loss": 1.6893, + "step": 39540 + }, + { + "epoch": 8.63, + "learning_rate": 6.85929648241206e-07, + "loss": 3.1041, + "step": 39541 + }, + { + "epoch": 8.63, + "learning_rate": 6.858204063797248e-07, + "loss": 0.9929, + "step": 39542 + }, + { + "epoch": 8.63, + "learning_rate": 6.857111645182435e-07, + "loss": 0.6101, + "step": 39543 + }, + { + "epoch": 8.63, + "learning_rate": 6.856019226567622e-07, + "loss": 1.9223, + "step": 39544 + }, + { + "epoch": 8.63, + "learning_rate": 6.854926807952808e-07, + "loss": 0.7949, + "step": 39545 + }, + { + "epoch": 8.63, + "learning_rate": 6.853834389337995e-07, + "loss": 0.2817, + "step": 39546 + }, + { + "epoch": 8.63, + "learning_rate": 6.852741970723181e-07, + "loss": 0.8621, + "step": 39547 + }, + { + "epoch": 8.63, + "learning_rate": 6.851649552108368e-07, + "loss": 0.4846, + "step": 39548 + }, + { + "epoch": 8.63, + "learning_rate": 6.850557133493555e-07, + "loss": 0.4677, + "step": 39549 + }, + { + "epoch": 8.63, + "learning_rate": 6.849464714878743e-07, + "loss": 0.9754, + "step": 39550 + }, + { + "epoch": 8.63, + "learning_rate": 6.848372296263929e-07, + "loss": 0.5342, + "step": 39551 + }, + { + "epoch": 8.63, + "learning_rate": 6.847279877649116e-07, + "loss": 0.8487, + "step": 39552 + }, + { + "epoch": 8.63, + "learning_rate": 6.846187459034302e-07, + "loss": 1.4904, + "step": 39553 + }, + { + "epoch": 8.63, + "learning_rate": 6.845095040419489e-07, + "loss": 0.6202, + "step": 39554 + }, + { + "epoch": 8.63, + "learning_rate": 6.844002621804676e-07, + "loss": 0.2891, + "step": 39555 + }, + { + "epoch": 8.63, + "learning_rate": 6.842910203189864e-07, + "loss": 1.3091, + "step": 39556 + }, + { + "epoch": 8.63, + "learning_rate": 6.84181778457505e-07, + "loss": 2.027, + "step": 39557 + }, + { + "epoch": 8.63, + "learning_rate": 6.840725365960237e-07, + "loss": 1.6822, + "step": 39558 + }, + { + "epoch": 8.63, + "learning_rate": 6.839632947345423e-07, + "loss": 0.2121, + "step": 39559 + }, + { + "epoch": 8.63, + "learning_rate": 6.83854052873061e-07, + "loss": 1.5051, + "step": 39560 + }, + { + "epoch": 8.63, + "learning_rate": 6.837448110115796e-07, + "loss": 0.6689, + "step": 39561 + }, + { + "epoch": 8.63, + "learning_rate": 6.836355691500984e-07, + "loss": 0.6916, + "step": 39562 + }, + { + "epoch": 8.63, + "learning_rate": 6.83526327288617e-07, + "loss": 2.4874, + "step": 39563 + }, + { + "epoch": 8.63, + "learning_rate": 6.834170854271358e-07, + "loss": 0.3853, + "step": 39564 + }, + { + "epoch": 8.63, + "learning_rate": 6.833078435656544e-07, + "loss": 0.2442, + "step": 39565 + }, + { + "epoch": 8.64, + "learning_rate": 6.831986017041731e-07, + "loss": 1.6091, + "step": 39566 + }, + { + "epoch": 8.64, + "learning_rate": 6.830893598426917e-07, + "loss": 1.4105, + "step": 39567 + }, + { + "epoch": 8.64, + "learning_rate": 6.829801179812104e-07, + "loss": 2.5409, + "step": 39568 + }, + { + "epoch": 8.64, + "learning_rate": 6.828708761197291e-07, + "loss": 0.1263, + "step": 39569 + }, + { + "epoch": 8.64, + "learning_rate": 6.827616342582479e-07, + "loss": 2.1425, + "step": 39570 + }, + { + "epoch": 8.64, + "learning_rate": 6.826523923967665e-07, + "loss": 0.7863, + "step": 39571 + }, + { + "epoch": 8.64, + "learning_rate": 6.825431505352852e-07, + "loss": 1.2364, + "step": 39572 + }, + { + "epoch": 8.64, + "learning_rate": 6.824339086738038e-07, + "loss": 1.4699, + "step": 39573 + }, + { + "epoch": 8.64, + "learning_rate": 6.823246668123225e-07, + "loss": 0.9447, + "step": 39574 + }, + { + "epoch": 8.64, + "learning_rate": 6.822154249508412e-07, + "loss": 0.9661, + "step": 39575 + }, + { + "epoch": 8.64, + "learning_rate": 6.821061830893599e-07, + "loss": 1.8417, + "step": 39576 + }, + { + "epoch": 8.64, + "learning_rate": 6.819969412278786e-07, + "loss": 1.2814, + "step": 39577 + }, + { + "epoch": 8.64, + "learning_rate": 6.818876993663973e-07, + "loss": 1.6968, + "step": 39578 + }, + { + "epoch": 8.64, + "learning_rate": 6.817784575049159e-07, + "loss": 0.335, + "step": 39579 + }, + { + "epoch": 8.64, + "learning_rate": 6.816692156434346e-07, + "loss": 0.5341, + "step": 39580 + }, + { + "epoch": 8.64, + "learning_rate": 6.815599737819532e-07, + "loss": 2.0526, + "step": 39581 + }, + { + "epoch": 8.64, + "learning_rate": 6.81450731920472e-07, + "loss": 0.1043, + "step": 39582 + }, + { + "epoch": 8.64, + "learning_rate": 6.813414900589906e-07, + "loss": 1.174, + "step": 39583 + }, + { + "epoch": 8.64, + "learning_rate": 6.812322481975094e-07, + "loss": 1.039, + "step": 39584 + }, + { + "epoch": 8.64, + "learning_rate": 6.81123006336028e-07, + "loss": 0.9734, + "step": 39585 + }, + { + "epoch": 8.64, + "learning_rate": 6.810137644745467e-07, + "loss": 3.2024, + "step": 39586 + }, + { + "epoch": 8.64, + "learning_rate": 6.809045226130653e-07, + "loss": 1.9954, + "step": 39587 + }, + { + "epoch": 8.64, + "learning_rate": 6.807952807515841e-07, + "loss": 0.3203, + "step": 39588 + }, + { + "epoch": 8.64, + "learning_rate": 6.806860388901028e-07, + "loss": 1.0764, + "step": 39589 + }, + { + "epoch": 8.64, + "learning_rate": 6.805767970286215e-07, + "loss": 1.1405, + "step": 39590 + }, + { + "epoch": 8.64, + "learning_rate": 6.804675551671401e-07, + "loss": 0.8702, + "step": 39591 + }, + { + "epoch": 8.64, + "learning_rate": 6.803583133056588e-07, + "loss": 0.3901, + "step": 39592 + }, + { + "epoch": 8.64, + "learning_rate": 6.802490714441774e-07, + "loss": 2.2421, + "step": 39593 + }, + { + "epoch": 8.64, + "learning_rate": 6.801398295826961e-07, + "loss": 1.5249, + "step": 39594 + }, + { + "epoch": 8.64, + "learning_rate": 6.800305877212149e-07, + "loss": 0.8505, + "step": 39595 + }, + { + "epoch": 8.64, + "learning_rate": 6.799213458597335e-07, + "loss": 1.0761, + "step": 39596 + }, + { + "epoch": 8.64, + "learning_rate": 6.798121039982522e-07, + "loss": 0.7931, + "step": 39597 + }, + { + "epoch": 8.64, + "learning_rate": 6.797028621367709e-07, + "loss": 0.6238, + "step": 39598 + }, + { + "epoch": 8.64, + "learning_rate": 6.795936202752895e-07, + "loss": 0.6402, + "step": 39599 + }, + { + "epoch": 8.64, + "learning_rate": 6.794843784138082e-07, + "loss": 0.4739, + "step": 39600 + }, + { + "epoch": 8.64, + "learning_rate": 6.793751365523268e-07, + "loss": 0.1717, + "step": 39601 + }, + { + "epoch": 8.64, + "learning_rate": 6.792658946908456e-07, + "loss": 1.1369, + "step": 39602 + }, + { + "epoch": 8.64, + "learning_rate": 6.791566528293643e-07, + "loss": 0.8845, + "step": 39603 + }, + { + "epoch": 8.64, + "learning_rate": 6.79047410967883e-07, + "loss": 0.835, + "step": 39604 + }, + { + "epoch": 8.64, + "learning_rate": 6.789381691064016e-07, + "loss": 0.9403, + "step": 39605 + }, + { + "epoch": 8.64, + "learning_rate": 6.788289272449203e-07, + "loss": 0.377, + "step": 39606 + }, + { + "epoch": 8.64, + "learning_rate": 6.787196853834389e-07, + "loss": 2.1618, + "step": 39607 + }, + { + "epoch": 8.64, + "learning_rate": 6.786104435219577e-07, + "loss": 0.1884, + "step": 39608 + }, + { + "epoch": 8.64, + "learning_rate": 6.785012016604764e-07, + "loss": 0.6834, + "step": 39609 + }, + { + "epoch": 8.64, + "learning_rate": 6.78391959798995e-07, + "loss": 0.8748, + "step": 39610 + }, + { + "epoch": 8.64, + "learning_rate": 6.782827179375137e-07, + "loss": 1.1371, + "step": 39611 + }, + { + "epoch": 8.65, + "learning_rate": 6.781734760760324e-07, + "loss": 0.8817, + "step": 39612 + }, + { + "epoch": 8.65, + "learning_rate": 6.78064234214551e-07, + "loss": 2.2275, + "step": 39613 + }, + { + "epoch": 8.65, + "learning_rate": 6.779549923530697e-07, + "loss": 1.6079, + "step": 39614 + }, + { + "epoch": 8.65, + "learning_rate": 6.778457504915885e-07, + "loss": 0.6928, + "step": 39615 + }, + { + "epoch": 8.65, + "learning_rate": 6.777365086301071e-07, + "loss": 2.2021, + "step": 39616 + }, + { + "epoch": 8.65, + "learning_rate": 6.776272667686259e-07, + "loss": 1.1942, + "step": 39617 + }, + { + "epoch": 8.65, + "learning_rate": 6.775180249071445e-07, + "loss": 1.1141, + "step": 39618 + }, + { + "epoch": 8.65, + "learning_rate": 6.774087830456631e-07, + "loss": 0.7287, + "step": 39619 + }, + { + "epoch": 8.65, + "learning_rate": 6.772995411841818e-07, + "loss": 0.7026, + "step": 39620 + }, + { + "epoch": 8.65, + "learning_rate": 6.771902993227004e-07, + "loss": 0.7198, + "step": 39621 + }, + { + "epoch": 8.65, + "learning_rate": 6.770810574612192e-07, + "loss": 1.8252, + "step": 39622 + }, + { + "epoch": 8.65, + "learning_rate": 6.769718155997379e-07, + "loss": 0.5287, + "step": 39623 + }, + { + "epoch": 8.65, + "learning_rate": 6.768625737382566e-07, + "loss": 0.8963, + "step": 39624 + }, + { + "epoch": 8.65, + "learning_rate": 6.767533318767753e-07, + "loss": 1.7334, + "step": 39625 + }, + { + "epoch": 8.65, + "learning_rate": 6.766440900152939e-07, + "loss": 0.9557, + "step": 39626 + }, + { + "epoch": 8.65, + "learning_rate": 6.765348481538125e-07, + "loss": 2.0372, + "step": 39627 + }, + { + "epoch": 8.65, + "learning_rate": 6.764256062923313e-07, + "loss": 1.36, + "step": 39628 + }, + { + "epoch": 8.65, + "learning_rate": 6.7631636443085e-07, + "loss": 1.1098, + "step": 39629 + }, + { + "epoch": 8.65, + "learning_rate": 6.762071225693686e-07, + "loss": 1.5268, + "step": 39630 + }, + { + "epoch": 8.65, + "learning_rate": 6.760978807078874e-07, + "loss": 1.3015, + "step": 39631 + }, + { + "epoch": 8.65, + "learning_rate": 6.75988638846406e-07, + "loss": 0.4958, + "step": 39632 + }, + { + "epoch": 8.65, + "learning_rate": 6.758793969849246e-07, + "loss": 1.4565, + "step": 39633 + }, + { + "epoch": 8.65, + "learning_rate": 6.757701551234433e-07, + "loss": 0.3157, + "step": 39634 + }, + { + "epoch": 8.65, + "learning_rate": 6.756609132619621e-07, + "loss": 0.1427, + "step": 39635 + }, + { + "epoch": 8.65, + "learning_rate": 6.755516714004807e-07, + "loss": 3.1832, + "step": 39636 + }, + { + "epoch": 8.65, + "learning_rate": 6.754424295389995e-07, + "loss": 4.0181, + "step": 39637 + }, + { + "epoch": 8.65, + "learning_rate": 6.753331876775181e-07, + "loss": 2.1186, + "step": 39638 + }, + { + "epoch": 8.65, + "learning_rate": 6.752239458160368e-07, + "loss": 0.6575, + "step": 39639 + }, + { + "epoch": 8.65, + "learning_rate": 6.751147039545554e-07, + "loss": 1.0881, + "step": 39640 + }, + { + "epoch": 8.65, + "learning_rate": 6.750054620930742e-07, + "loss": 1.3051, + "step": 39641 + }, + { + "epoch": 8.65, + "learning_rate": 6.748962202315928e-07, + "loss": 0.2188, + "step": 39642 + }, + { + "epoch": 8.65, + "learning_rate": 6.747869783701115e-07, + "loss": 1.932, + "step": 39643 + }, + { + "epoch": 8.65, + "learning_rate": 6.746777365086302e-07, + "loss": 0.6878, + "step": 39644 + }, + { + "epoch": 8.65, + "learning_rate": 6.745684946471489e-07, + "loss": 0.7051, + "step": 39645 + }, + { + "epoch": 8.65, + "learning_rate": 6.744592527856675e-07, + "loss": 0.9856, + "step": 39646 + }, + { + "epoch": 8.65, + "learning_rate": 6.743500109241861e-07, + "loss": 1.291, + "step": 39647 + }, + { + "epoch": 8.65, + "learning_rate": 6.742407690627049e-07, + "loss": 1.4012, + "step": 39648 + }, + { + "epoch": 8.65, + "learning_rate": 6.741315272012236e-07, + "loss": 2.0627, + "step": 39649 + }, + { + "epoch": 8.65, + "learning_rate": 6.740222853397422e-07, + "loss": 0.657, + "step": 39650 + }, + { + "epoch": 8.65, + "learning_rate": 6.73913043478261e-07, + "loss": 0.6198, + "step": 39651 + }, + { + "epoch": 8.65, + "learning_rate": 6.738038016167796e-07, + "loss": 1.4339, + "step": 39652 + }, + { + "epoch": 8.65, + "learning_rate": 6.736945597552983e-07, + "loss": 0.1283, + "step": 39653 + }, + { + "epoch": 8.65, + "learning_rate": 6.735853178938169e-07, + "loss": 1.3185, + "step": 39654 + }, + { + "epoch": 8.65, + "learning_rate": 6.734760760323357e-07, + "loss": 0.4353, + "step": 39655 + }, + { + "epoch": 8.65, + "learning_rate": 6.733668341708543e-07, + "loss": 1.6663, + "step": 39656 + }, + { + "epoch": 8.65, + "learning_rate": 6.73257592309373e-07, + "loss": 1.813, + "step": 39657 + }, + { + "epoch": 8.66, + "learning_rate": 6.731483504478917e-07, + "loss": 0.617, + "step": 39658 + }, + { + "epoch": 8.66, + "learning_rate": 6.730391085864104e-07, + "loss": 0.7853, + "step": 39659 + }, + { + "epoch": 8.66, + "learning_rate": 6.72929866724929e-07, + "loss": 0.7645, + "step": 39660 + }, + { + "epoch": 8.66, + "learning_rate": 6.728206248634478e-07, + "loss": 0.8555, + "step": 39661 + }, + { + "epoch": 8.66, + "learning_rate": 6.727113830019664e-07, + "loss": 2.2691, + "step": 39662 + }, + { + "epoch": 8.66, + "learning_rate": 6.726021411404851e-07, + "loss": 1.0071, + "step": 39663 + }, + { + "epoch": 8.66, + "learning_rate": 6.724928992790037e-07, + "loss": 0.8711, + "step": 39664 + }, + { + "epoch": 8.66, + "learning_rate": 6.723836574175225e-07, + "loss": 1.5195, + "step": 39665 + }, + { + "epoch": 8.66, + "learning_rate": 6.722744155560411e-07, + "loss": 2.2489, + "step": 39666 + }, + { + "epoch": 8.66, + "learning_rate": 6.721651736945598e-07, + "loss": 2.1209, + "step": 39667 + }, + { + "epoch": 8.66, + "learning_rate": 6.720559318330785e-07, + "loss": 0.6483, + "step": 39668 + }, + { + "epoch": 8.66, + "learning_rate": 6.719466899715972e-07, + "loss": 2.2722, + "step": 39669 + }, + { + "epoch": 8.66, + "learning_rate": 6.718374481101158e-07, + "loss": 0.8553, + "step": 39670 + }, + { + "epoch": 8.66, + "learning_rate": 6.717282062486346e-07, + "loss": 0.6457, + "step": 39671 + }, + { + "epoch": 8.66, + "learning_rate": 6.716189643871532e-07, + "loss": 0.7579, + "step": 39672 + }, + { + "epoch": 8.66, + "learning_rate": 6.715097225256719e-07, + "loss": 0.4079, + "step": 39673 + }, + { + "epoch": 8.66, + "learning_rate": 6.714004806641906e-07, + "loss": 0.2418, + "step": 39674 + }, + { + "epoch": 8.66, + "learning_rate": 6.712912388027093e-07, + "loss": 0.7006, + "step": 39675 + }, + { + "epoch": 8.66, + "learning_rate": 6.711819969412279e-07, + "loss": 0.9824, + "step": 39676 + }, + { + "epoch": 8.66, + "learning_rate": 6.710727550797466e-07, + "loss": 1.3535, + "step": 39677 + }, + { + "epoch": 8.66, + "learning_rate": 6.709635132182653e-07, + "loss": 1.8, + "step": 39678 + }, + { + "epoch": 8.66, + "learning_rate": 6.70854271356784e-07, + "loss": 1.6543, + "step": 39679 + }, + { + "epoch": 8.66, + "learning_rate": 6.707450294953026e-07, + "loss": 2.3402, + "step": 39680 + }, + { + "epoch": 8.66, + "learning_rate": 6.706357876338214e-07, + "loss": 0.9378, + "step": 39681 + }, + { + "epoch": 8.66, + "learning_rate": 6.7052654577234e-07, + "loss": 1.6163, + "step": 39682 + }, + { + "epoch": 8.66, + "learning_rate": 6.704173039108587e-07, + "loss": 0.5827, + "step": 39683 + }, + { + "epoch": 8.66, + "learning_rate": 6.703080620493773e-07, + "loss": 4.0282, + "step": 39684 + }, + { + "epoch": 8.66, + "learning_rate": 6.701988201878961e-07, + "loss": 2.3151, + "step": 39685 + }, + { + "epoch": 8.66, + "learning_rate": 6.700895783264147e-07, + "loss": 1.7044, + "step": 39686 + }, + { + "epoch": 8.66, + "learning_rate": 6.699803364649334e-07, + "loss": 0.3304, + "step": 39687 + }, + { + "epoch": 8.66, + "learning_rate": 6.698710946034521e-07, + "loss": 0.9191, + "step": 39688 + }, + { + "epoch": 8.66, + "learning_rate": 6.697618527419708e-07, + "loss": 0.6311, + "step": 39689 + }, + { + "epoch": 8.66, + "learning_rate": 6.696526108804894e-07, + "loss": 0.1363, + "step": 39690 + }, + { + "epoch": 8.66, + "learning_rate": 6.695433690190082e-07, + "loss": 0.9421, + "step": 39691 + }, + { + "epoch": 8.66, + "learning_rate": 6.694341271575268e-07, + "loss": 1.1875, + "step": 39692 + }, + { + "epoch": 8.66, + "learning_rate": 6.693248852960455e-07, + "loss": 0.4676, + "step": 39693 + }, + { + "epoch": 8.66, + "learning_rate": 6.692156434345642e-07, + "loss": 0.7541, + "step": 39694 + }, + { + "epoch": 8.66, + "learning_rate": 6.691064015730829e-07, + "loss": 1.0576, + "step": 39695 + }, + { + "epoch": 8.66, + "learning_rate": 6.689971597116015e-07, + "loss": 0.6037, + "step": 39696 + }, + { + "epoch": 8.66, + "learning_rate": 6.688879178501202e-07, + "loss": 0.2537, + "step": 39697 + }, + { + "epoch": 8.66, + "learning_rate": 6.687786759886389e-07, + "loss": 1.4085, + "step": 39698 + }, + { + "epoch": 8.66, + "learning_rate": 6.686694341271576e-07, + "loss": 2.7398, + "step": 39699 + }, + { + "epoch": 8.66, + "learning_rate": 6.685601922656762e-07, + "loss": 1.7077, + "step": 39700 + }, + { + "epoch": 8.66, + "learning_rate": 6.68450950404195e-07, + "loss": 2.1695, + "step": 39701 + }, + { + "epoch": 8.66, + "learning_rate": 6.683417085427136e-07, + "loss": 0.9771, + "step": 39702 + }, + { + "epoch": 8.66, + "learning_rate": 6.682324666812323e-07, + "loss": 0.8238, + "step": 39703 + }, + { + "epoch": 8.67, + "learning_rate": 6.681232248197509e-07, + "loss": 0.6372, + "step": 39704 + }, + { + "epoch": 8.67, + "learning_rate": 6.680139829582697e-07, + "loss": 3.1678, + "step": 39705 + }, + { + "epoch": 8.67, + "learning_rate": 6.679047410967883e-07, + "loss": 2.2895, + "step": 39706 + }, + { + "epoch": 8.67, + "learning_rate": 6.677954992353071e-07, + "loss": 2.5949, + "step": 39707 + }, + { + "epoch": 8.67, + "learning_rate": 6.676862573738257e-07, + "loss": 0.5671, + "step": 39708 + }, + { + "epoch": 8.67, + "learning_rate": 6.675770155123444e-07, + "loss": 0.5144, + "step": 39709 + }, + { + "epoch": 8.67, + "learning_rate": 6.67467773650863e-07, + "loss": 0.736, + "step": 39710 + }, + { + "epoch": 8.67, + "learning_rate": 6.673585317893817e-07, + "loss": 2.8901, + "step": 39711 + }, + { + "epoch": 8.67, + "learning_rate": 6.672492899279004e-07, + "loss": 0.8083, + "step": 39712 + }, + { + "epoch": 8.67, + "learning_rate": 6.671400480664191e-07, + "loss": 1.2873, + "step": 39713 + }, + { + "epoch": 8.67, + "learning_rate": 6.670308062049379e-07, + "loss": 1.7002, + "step": 39714 + }, + { + "epoch": 8.67, + "learning_rate": 6.669215643434565e-07, + "loss": 1.4045, + "step": 39715 + }, + { + "epoch": 8.67, + "learning_rate": 6.668123224819751e-07, + "loss": 1.1894, + "step": 39716 + }, + { + "epoch": 8.67, + "learning_rate": 6.667030806204938e-07, + "loss": 0.8996, + "step": 39717 + }, + { + "epoch": 8.67, + "learning_rate": 6.665938387590124e-07, + "loss": 1.1492, + "step": 39718 + }, + { + "epoch": 8.67, + "learning_rate": 6.664845968975312e-07, + "loss": 2.2859, + "step": 39719 + }, + { + "epoch": 8.67, + "learning_rate": 6.663753550360498e-07, + "loss": 0.7787, + "step": 39720 + }, + { + "epoch": 8.67, + "learning_rate": 6.662661131745686e-07, + "loss": 3.3527, + "step": 39721 + }, + { + "epoch": 8.67, + "learning_rate": 6.661568713130872e-07, + "loss": 0.1293, + "step": 39722 + }, + { + "epoch": 8.67, + "learning_rate": 6.660476294516059e-07, + "loss": 0.0678, + "step": 39723 + }, + { + "epoch": 8.67, + "learning_rate": 6.659383875901245e-07, + "loss": 0.9776, + "step": 39724 + }, + { + "epoch": 8.67, + "learning_rate": 6.658291457286433e-07, + "loss": 1.8984, + "step": 39725 + }, + { + "epoch": 8.67, + "learning_rate": 6.657199038671619e-07, + "loss": 0.4197, + "step": 39726 + }, + { + "epoch": 8.67, + "learning_rate": 6.656106620056807e-07, + "loss": 2.086, + "step": 39727 + }, + { + "epoch": 8.67, + "learning_rate": 6.655014201441994e-07, + "loss": 0.9736, + "step": 39728 + }, + { + "epoch": 8.67, + "learning_rate": 6.65392178282718e-07, + "loss": 1.7352, + "step": 39729 + }, + { + "epoch": 8.67, + "learning_rate": 6.652829364212366e-07, + "loss": 0.4817, + "step": 39730 + }, + { + "epoch": 8.67, + "learning_rate": 6.651736945597553e-07, + "loss": 1.6153, + "step": 39731 + }, + { + "epoch": 8.67, + "learning_rate": 6.65064452698274e-07, + "loss": 0.8503, + "step": 39732 + }, + { + "epoch": 8.67, + "learning_rate": 6.649552108367927e-07, + "loss": 1.2987, + "step": 39733 + }, + { + "epoch": 8.67, + "learning_rate": 6.648459689753115e-07, + "loss": 1.6614, + "step": 39734 + }, + { + "epoch": 8.67, + "learning_rate": 6.647367271138301e-07, + "loss": 2.6882, + "step": 39735 + }, + { + "epoch": 8.67, + "learning_rate": 6.646274852523488e-07, + "loss": 1.1286, + "step": 39736 + }, + { + "epoch": 8.67, + "learning_rate": 6.645182433908674e-07, + "loss": 0.6529, + "step": 39737 + }, + { + "epoch": 8.67, + "learning_rate": 6.64409001529386e-07, + "loss": 1.1042, + "step": 39738 + }, + { + "epoch": 8.67, + "learning_rate": 6.642997596679048e-07, + "loss": 1.4089, + "step": 39739 + }, + { + "epoch": 8.67, + "learning_rate": 6.641905178064234e-07, + "loss": 1.0014, + "step": 39740 + }, + { + "epoch": 8.67, + "learning_rate": 6.640812759449422e-07, + "loss": 0.0895, + "step": 39741 + }, + { + "epoch": 8.67, + "learning_rate": 6.639720340834609e-07, + "loss": 1.542, + "step": 39742 + }, + { + "epoch": 8.67, + "learning_rate": 6.638627922219795e-07, + "loss": 1.0595, + "step": 39743 + }, + { + "epoch": 8.67, + "learning_rate": 6.637535503604981e-07, + "loss": 0.3221, + "step": 39744 + }, + { + "epoch": 8.67, + "learning_rate": 6.636443084990169e-07, + "loss": 1.6757, + "step": 39745 + }, + { + "epoch": 8.67, + "learning_rate": 6.635350666375355e-07, + "loss": 0.6499, + "step": 39746 + }, + { + "epoch": 8.67, + "learning_rate": 6.634258247760543e-07, + "loss": 0.7931, + "step": 39747 + }, + { + "epoch": 8.67, + "learning_rate": 6.63316582914573e-07, + "loss": 2.8347, + "step": 39748 + }, + { + "epoch": 8.68, + "learning_rate": 6.632073410530916e-07, + "loss": 0.5017, + "step": 39749 + }, + { + "epoch": 8.68, + "learning_rate": 6.630980991916103e-07, + "loss": 0.9925, + "step": 39750 + }, + { + "epoch": 8.68, + "learning_rate": 6.629888573301289e-07, + "loss": 1.6416, + "step": 39751 + }, + { + "epoch": 8.68, + "learning_rate": 6.628796154686476e-07, + "loss": 0.3242, + "step": 39752 + }, + { + "epoch": 8.68, + "learning_rate": 6.627703736071663e-07, + "loss": 0.4341, + "step": 39753 + }, + { + "epoch": 8.68, + "learning_rate": 6.626611317456851e-07, + "loss": 1.2454, + "step": 39754 + }, + { + "epoch": 8.68, + "learning_rate": 6.625518898842037e-07, + "loss": 0.1743, + "step": 39755 + }, + { + "epoch": 8.68, + "learning_rate": 6.624426480227224e-07, + "loss": 0.4431, + "step": 39756 + }, + { + "epoch": 8.68, + "learning_rate": 6.62333406161241e-07, + "loss": 0.6779, + "step": 39757 + }, + { + "epoch": 8.68, + "learning_rate": 6.622241642997596e-07, + "loss": 1.565, + "step": 39758 + }, + { + "epoch": 8.68, + "learning_rate": 6.621149224382784e-07, + "loss": 1.1905, + "step": 39759 + }, + { + "epoch": 8.68, + "learning_rate": 6.620056805767972e-07, + "loss": 1.0108, + "step": 39760 + }, + { + "epoch": 8.68, + "learning_rate": 6.618964387153158e-07, + "loss": 1.3736, + "step": 39761 + }, + { + "epoch": 8.68, + "learning_rate": 6.617871968538345e-07, + "loss": 0.5459, + "step": 39762 + }, + { + "epoch": 8.68, + "learning_rate": 6.616779549923531e-07, + "loss": 0.4869, + "step": 39763 + }, + { + "epoch": 8.68, + "learning_rate": 6.615687131308718e-07, + "loss": 2.0339, + "step": 39764 + }, + { + "epoch": 8.68, + "learning_rate": 6.614594712693904e-07, + "loss": 1.2624, + "step": 39765 + }, + { + "epoch": 8.68, + "learning_rate": 6.613502294079091e-07, + "loss": 0.5929, + "step": 39766 + }, + { + "epoch": 8.68, + "learning_rate": 6.612409875464279e-07, + "loss": 1.8096, + "step": 39767 + }, + { + "epoch": 8.68, + "learning_rate": 6.611317456849466e-07, + "loss": 2.3641, + "step": 39768 + }, + { + "epoch": 8.68, + "learning_rate": 6.610225038234652e-07, + "loss": 0.8885, + "step": 39769 + }, + { + "epoch": 8.68, + "learning_rate": 6.609132619619839e-07, + "loss": 0.5971, + "step": 39770 + }, + { + "epoch": 8.68, + "learning_rate": 6.608040201005025e-07, + "loss": 2.8497, + "step": 39771 + }, + { + "epoch": 8.68, + "learning_rate": 6.606947782390213e-07, + "loss": 1.425, + "step": 39772 + }, + { + "epoch": 8.68, + "learning_rate": 6.605855363775399e-07, + "loss": 1.8253, + "step": 39773 + }, + { + "epoch": 8.68, + "learning_rate": 6.604762945160587e-07, + "loss": 1.5678, + "step": 39774 + }, + { + "epoch": 8.68, + "learning_rate": 6.603670526545773e-07, + "loss": 1.0171, + "step": 39775 + }, + { + "epoch": 8.68, + "learning_rate": 6.60257810793096e-07, + "loss": 0.3615, + "step": 39776 + }, + { + "epoch": 8.68, + "learning_rate": 6.601485689316146e-07, + "loss": 0.9483, + "step": 39777 + }, + { + "epoch": 8.68, + "learning_rate": 6.600393270701333e-07, + "loss": 1.1402, + "step": 39778 + }, + { + "epoch": 8.68, + "learning_rate": 6.59930085208652e-07, + "loss": 1.143, + "step": 39779 + }, + { + "epoch": 8.68, + "learning_rate": 6.598208433471708e-07, + "loss": 1.1324, + "step": 39780 + }, + { + "epoch": 8.68, + "learning_rate": 6.597116014856894e-07, + "loss": 0.6162, + "step": 39781 + }, + { + "epoch": 8.68, + "learning_rate": 6.596023596242081e-07, + "loss": 2.1845, + "step": 39782 + }, + { + "epoch": 8.68, + "learning_rate": 6.594931177627267e-07, + "loss": 1.3088, + "step": 39783 + }, + { + "epoch": 8.68, + "learning_rate": 6.593838759012454e-07, + "loss": 0.8153, + "step": 39784 + }, + { + "epoch": 8.68, + "learning_rate": 6.59274634039764e-07, + "loss": 2.7962, + "step": 39785 + }, + { + "epoch": 8.68, + "learning_rate": 6.591653921782828e-07, + "loss": 1.1491, + "step": 39786 + }, + { + "epoch": 8.68, + "learning_rate": 6.590561503168015e-07, + "loss": 0.3649, + "step": 39787 + }, + { + "epoch": 8.68, + "learning_rate": 6.589469084553202e-07, + "loss": 1.6106, + "step": 39788 + }, + { + "epoch": 8.68, + "learning_rate": 6.588376665938388e-07, + "loss": 0.6001, + "step": 39789 + }, + { + "epoch": 8.68, + "learning_rate": 6.587284247323575e-07, + "loss": 1.0471, + "step": 39790 + }, + { + "epoch": 8.68, + "learning_rate": 6.586191828708761e-07, + "loss": 1.5289, + "step": 39791 + }, + { + "epoch": 8.68, + "learning_rate": 6.585099410093948e-07, + "loss": 1.1104, + "step": 39792 + }, + { + "epoch": 8.68, + "learning_rate": 6.584006991479136e-07, + "loss": 1.5513, + "step": 39793 + }, + { + "epoch": 8.68, + "learning_rate": 6.582914572864323e-07, + "loss": 1.5961, + "step": 39794 + }, + { + "epoch": 8.69, + "learning_rate": 6.581822154249509e-07, + "loss": 0.9205, + "step": 39795 + }, + { + "epoch": 8.69, + "learning_rate": 6.580729735634696e-07, + "loss": 0.7083, + "step": 39796 + }, + { + "epoch": 8.69, + "learning_rate": 6.579637317019882e-07, + "loss": 0.713, + "step": 39797 + }, + { + "epoch": 8.69, + "learning_rate": 6.578544898405069e-07, + "loss": 0.8742, + "step": 39798 + }, + { + "epoch": 8.69, + "learning_rate": 6.577452479790255e-07, + "loss": 1.8522, + "step": 39799 + }, + { + "epoch": 8.69, + "learning_rate": 6.576360061175444e-07, + "loss": 0.0946, + "step": 39800 + }, + { + "epoch": 8.69, + "learning_rate": 6.57526764256063e-07, + "loss": 1.8014, + "step": 39801 + }, + { + "epoch": 8.69, + "learning_rate": 6.574175223945817e-07, + "loss": 0.9373, + "step": 39802 + }, + { + "epoch": 8.69, + "learning_rate": 6.573082805331003e-07, + "loss": 1.8722, + "step": 39803 + }, + { + "epoch": 8.69, + "learning_rate": 6.57199038671619e-07, + "loss": 1.1414, + "step": 39804 + }, + { + "epoch": 8.69, + "learning_rate": 6.570897968101376e-07, + "loss": 0.9805, + "step": 39805 + }, + { + "epoch": 8.69, + "learning_rate": 6.569805549486564e-07, + "loss": 0.6271, + "step": 39806 + }, + { + "epoch": 8.69, + "learning_rate": 6.568713130871751e-07, + "loss": 0.9254, + "step": 39807 + }, + { + "epoch": 8.69, + "learning_rate": 6.567620712256938e-07, + "loss": 1.8676, + "step": 39808 + }, + { + "epoch": 8.69, + "learning_rate": 6.566528293642124e-07, + "loss": 1.2142, + "step": 39809 + }, + { + "epoch": 8.69, + "learning_rate": 6.565435875027311e-07, + "loss": 0.8459, + "step": 39810 + }, + { + "epoch": 8.69, + "learning_rate": 6.564343456412497e-07, + "loss": 0.0772, + "step": 39811 + }, + { + "epoch": 8.69, + "learning_rate": 6.563251037797684e-07, + "loss": 2.072, + "step": 39812 + }, + { + "epoch": 8.69, + "learning_rate": 6.562158619182872e-07, + "loss": 0.8347, + "step": 39813 + }, + { + "epoch": 8.69, + "learning_rate": 6.561066200568059e-07, + "loss": 2.4119, + "step": 39814 + }, + { + "epoch": 8.69, + "learning_rate": 6.559973781953245e-07, + "loss": 0.9833, + "step": 39815 + }, + { + "epoch": 8.69, + "learning_rate": 6.558881363338432e-07, + "loss": 1.5539, + "step": 39816 + }, + { + "epoch": 8.69, + "learning_rate": 6.557788944723618e-07, + "loss": 1.0101, + "step": 39817 + }, + { + "epoch": 8.69, + "learning_rate": 6.556696526108805e-07, + "loss": 1.7548, + "step": 39818 + }, + { + "epoch": 8.69, + "learning_rate": 6.555604107493991e-07, + "loss": 0.8572, + "step": 39819 + }, + { + "epoch": 8.69, + "learning_rate": 6.55451168887918e-07, + "loss": 1.617, + "step": 39820 + }, + { + "epoch": 8.69, + "learning_rate": 6.553419270264366e-07, + "loss": 1.1512, + "step": 39821 + }, + { + "epoch": 8.69, + "learning_rate": 6.552326851649553e-07, + "loss": 2.7674, + "step": 39822 + }, + { + "epoch": 8.69, + "learning_rate": 6.551234433034739e-07, + "loss": 2.4041, + "step": 39823 + }, + { + "epoch": 8.69, + "learning_rate": 6.550142014419926e-07, + "loss": 1.1893, + "step": 39824 + }, + { + "epoch": 8.69, + "learning_rate": 6.549049595805112e-07, + "loss": 2.6987, + "step": 39825 + }, + { + "epoch": 8.69, + "learning_rate": 6.547957177190301e-07, + "loss": 0.68, + "step": 39826 + }, + { + "epoch": 8.69, + "learning_rate": 6.546864758575487e-07, + "loss": 0.9009, + "step": 39827 + }, + { + "epoch": 8.69, + "learning_rate": 6.545772339960674e-07, + "loss": 1.413, + "step": 39828 + }, + { + "epoch": 8.69, + "learning_rate": 6.54467992134586e-07, + "loss": 0.5058, + "step": 39829 + }, + { + "epoch": 8.69, + "learning_rate": 6.543587502731047e-07, + "loss": 0.8831, + "step": 39830 + }, + { + "epoch": 8.69, + "learning_rate": 6.542495084116233e-07, + "loss": 2.5617, + "step": 39831 + }, + { + "epoch": 8.69, + "learning_rate": 6.54140266550142e-07, + "loss": 2.0463, + "step": 39832 + }, + { + "epoch": 8.69, + "learning_rate": 6.540310246886608e-07, + "loss": 2.0646, + "step": 39833 + }, + { + "epoch": 8.69, + "learning_rate": 6.539217828271795e-07, + "loss": 1.7541, + "step": 39834 + }, + { + "epoch": 8.69, + "learning_rate": 6.538125409656981e-07, + "loss": 1.1017, + "step": 39835 + }, + { + "epoch": 8.69, + "learning_rate": 6.537032991042168e-07, + "loss": 1.0867, + "step": 39836 + }, + { + "epoch": 8.69, + "learning_rate": 6.535940572427354e-07, + "loss": 1.1003, + "step": 39837 + }, + { + "epoch": 8.69, + "learning_rate": 6.534848153812541e-07, + "loss": 1.5888, + "step": 39838 + }, + { + "epoch": 8.69, + "learning_rate": 6.533755735197727e-07, + "loss": 0.4723, + "step": 39839 + }, + { + "epoch": 8.69, + "learning_rate": 6.532663316582916e-07, + "loss": 1.4415, + "step": 39840 + }, + { + "epoch": 8.7, + "learning_rate": 6.531570897968102e-07, + "loss": 0.4793, + "step": 39841 + }, + { + "epoch": 8.7, + "learning_rate": 6.530478479353289e-07, + "loss": 1.6999, + "step": 39842 + }, + { + "epoch": 8.7, + "learning_rate": 6.529386060738475e-07, + "loss": 2.5508, + "step": 39843 + }, + { + "epoch": 8.7, + "learning_rate": 6.528293642123662e-07, + "loss": 2.0938, + "step": 39844 + }, + { + "epoch": 8.7, + "learning_rate": 6.527201223508848e-07, + "loss": 0.5242, + "step": 39845 + }, + { + "epoch": 8.7, + "learning_rate": 6.526108804894037e-07, + "loss": 0.3334, + "step": 39846 + }, + { + "epoch": 8.7, + "learning_rate": 6.525016386279224e-07, + "loss": 1.1561, + "step": 39847 + }, + { + "epoch": 8.7, + "learning_rate": 6.52392396766441e-07, + "loss": 0.2262, + "step": 39848 + }, + { + "epoch": 8.7, + "learning_rate": 6.522831549049596e-07, + "loss": 1.8161, + "step": 39849 + }, + { + "epoch": 8.7, + "learning_rate": 6.521739130434783e-07, + "loss": 1.6787, + "step": 39850 + }, + { + "epoch": 8.7, + "learning_rate": 6.520646711819969e-07, + "loss": 1.1623, + "step": 39851 + }, + { + "epoch": 8.7, + "learning_rate": 6.519554293205156e-07, + "loss": 0.0982, + "step": 39852 + }, + { + "epoch": 8.7, + "learning_rate": 6.518461874590345e-07, + "loss": 0.2356, + "step": 39853 + }, + { + "epoch": 8.7, + "learning_rate": 6.517369455975531e-07, + "loss": 0.8751, + "step": 39854 + }, + { + "epoch": 8.7, + "learning_rate": 6.516277037360717e-07, + "loss": 0.5642, + "step": 39855 + }, + { + "epoch": 8.7, + "learning_rate": 6.515184618745904e-07, + "loss": 1.7036, + "step": 39856 + }, + { + "epoch": 8.7, + "learning_rate": 6.51409220013109e-07, + "loss": 2.9446, + "step": 39857 + }, + { + "epoch": 8.7, + "learning_rate": 6.512999781516277e-07, + "loss": 1.4462, + "step": 39858 + }, + { + "epoch": 8.7, + "learning_rate": 6.511907362901463e-07, + "loss": 1.7285, + "step": 39859 + }, + { + "epoch": 8.7, + "learning_rate": 6.510814944286652e-07, + "loss": 1.4933, + "step": 39860 + }, + { + "epoch": 8.7, + "learning_rate": 6.509722525671839e-07, + "loss": 0.1906, + "step": 39861 + }, + { + "epoch": 8.7, + "learning_rate": 6.508630107057025e-07, + "loss": 1.1601, + "step": 39862 + }, + { + "epoch": 8.7, + "learning_rate": 6.507537688442211e-07, + "loss": 0.7267, + "step": 39863 + }, + { + "epoch": 8.7, + "learning_rate": 6.506445269827398e-07, + "loss": 0.602, + "step": 39864 + }, + { + "epoch": 8.7, + "learning_rate": 6.505352851212584e-07, + "loss": 2.1007, + "step": 39865 + }, + { + "epoch": 8.7, + "learning_rate": 6.504260432597773e-07, + "loss": 0.4765, + "step": 39866 + }, + { + "epoch": 8.7, + "learning_rate": 6.50316801398296e-07, + "loss": 0.6483, + "step": 39867 + }, + { + "epoch": 8.7, + "learning_rate": 6.502075595368146e-07, + "loss": 0.6134, + "step": 39868 + }, + { + "epoch": 8.7, + "learning_rate": 6.500983176753332e-07, + "loss": 0.8298, + "step": 39869 + }, + { + "epoch": 8.7, + "learning_rate": 6.499890758138519e-07, + "loss": 1.1578, + "step": 39870 + }, + { + "epoch": 8.7, + "learning_rate": 6.498798339523705e-07, + "loss": 0.3617, + "step": 39871 + }, + { + "epoch": 8.7, + "learning_rate": 6.497705920908892e-07, + "loss": 0.559, + "step": 39872 + }, + { + "epoch": 8.7, + "learning_rate": 6.496613502294081e-07, + "loss": 1.2466, + "step": 39873 + }, + { + "epoch": 8.7, + "learning_rate": 6.495521083679267e-07, + "loss": 0.9608, + "step": 39874 + }, + { + "epoch": 8.7, + "learning_rate": 6.494428665064454e-07, + "loss": 1.4021, + "step": 39875 + }, + { + "epoch": 8.7, + "learning_rate": 6.49333624644964e-07, + "loss": 0.2148, + "step": 39876 + }, + { + "epoch": 8.7, + "learning_rate": 6.492243827834826e-07, + "loss": 0.2008, + "step": 39877 + }, + { + "epoch": 8.7, + "learning_rate": 6.491151409220013e-07, + "loss": 0.6693, + "step": 39878 + }, + { + "epoch": 8.7, + "learning_rate": 6.490058990605201e-07, + "loss": 0.9185, + "step": 39879 + }, + { + "epoch": 8.7, + "learning_rate": 6.488966571990388e-07, + "loss": 1.8004, + "step": 39880 + }, + { + "epoch": 8.7, + "learning_rate": 6.487874153375575e-07, + "loss": 0.6388, + "step": 39881 + }, + { + "epoch": 8.7, + "learning_rate": 6.486781734760761e-07, + "loss": 0.943, + "step": 39882 + }, + { + "epoch": 8.7, + "learning_rate": 6.485689316145948e-07, + "loss": 1.8283, + "step": 39883 + }, + { + "epoch": 8.7, + "learning_rate": 6.484596897531134e-07, + "loss": 0.4785, + "step": 39884 + }, + { + "epoch": 8.7, + "learning_rate": 6.48350447891632e-07, + "loss": 0.9264, + "step": 39885 + }, + { + "epoch": 8.7, + "learning_rate": 6.482412060301508e-07, + "loss": 1.8041, + "step": 39886 + }, + { + "epoch": 8.71, + "learning_rate": 6.481319641686696e-07, + "loss": 0.5008, + "step": 39887 + }, + { + "epoch": 8.71, + "learning_rate": 6.480227223071882e-07, + "loss": 0.6761, + "step": 39888 + }, + { + "epoch": 8.71, + "learning_rate": 6.479134804457069e-07, + "loss": 0.1431, + "step": 39889 + }, + { + "epoch": 8.71, + "learning_rate": 6.478042385842255e-07, + "loss": 1.668, + "step": 39890 + }, + { + "epoch": 8.71, + "learning_rate": 6.476949967227441e-07, + "loss": 1.0876, + "step": 39891 + }, + { + "epoch": 8.71, + "learning_rate": 6.475857548612628e-07, + "loss": 0.5201, + "step": 39892 + }, + { + "epoch": 8.71, + "learning_rate": 6.474765129997817e-07, + "loss": 0.4727, + "step": 39893 + }, + { + "epoch": 8.71, + "learning_rate": 6.473672711383003e-07, + "loss": 0.806, + "step": 39894 + }, + { + "epoch": 8.71, + "learning_rate": 6.47258029276819e-07, + "loss": 1.1356, + "step": 39895 + }, + { + "epoch": 8.71, + "learning_rate": 6.471487874153376e-07, + "loss": 0.4019, + "step": 39896 + }, + { + "epoch": 8.71, + "learning_rate": 6.470395455538563e-07, + "loss": 1.5941, + "step": 39897 + }, + { + "epoch": 8.71, + "learning_rate": 6.469303036923749e-07, + "loss": 0.9357, + "step": 39898 + }, + { + "epoch": 8.71, + "learning_rate": 6.468210618308937e-07, + "loss": 0.4458, + "step": 39899 + }, + { + "epoch": 8.71, + "learning_rate": 6.467118199694124e-07, + "loss": 0.1964, + "step": 39900 + }, + { + "epoch": 8.71, + "learning_rate": 6.466025781079311e-07, + "loss": 1.1, + "step": 39901 + }, + { + "epoch": 8.71, + "learning_rate": 6.464933362464497e-07, + "loss": 0.1354, + "step": 39902 + }, + { + "epoch": 8.71, + "learning_rate": 6.463840943849684e-07, + "loss": 0.1546, + "step": 39903 + }, + { + "epoch": 8.71, + "learning_rate": 6.46274852523487e-07, + "loss": 0.376, + "step": 39904 + }, + { + "epoch": 8.71, + "learning_rate": 6.461656106620056e-07, + "loss": 1.9598, + "step": 39905 + }, + { + "epoch": 8.71, + "learning_rate": 6.460563688005244e-07, + "loss": 0.9925, + "step": 39906 + }, + { + "epoch": 8.71, + "learning_rate": 6.459471269390432e-07, + "loss": 0.3729, + "step": 39907 + }, + { + "epoch": 8.71, + "learning_rate": 6.458378850775618e-07, + "loss": 2.6533, + "step": 39908 + }, + { + "epoch": 8.71, + "learning_rate": 6.457286432160805e-07, + "loss": 1.71, + "step": 39909 + }, + { + "epoch": 8.71, + "learning_rate": 6.456194013545991e-07, + "loss": 1.7537, + "step": 39910 + }, + { + "epoch": 8.71, + "learning_rate": 6.455101594931178e-07, + "loss": 3.096, + "step": 39911 + }, + { + "epoch": 8.71, + "learning_rate": 6.454009176316365e-07, + "loss": 0.3968, + "step": 39912 + }, + { + "epoch": 8.71, + "learning_rate": 6.452916757701552e-07, + "loss": 1.1965, + "step": 39913 + }, + { + "epoch": 8.71, + "learning_rate": 6.451824339086739e-07, + "loss": 0.6749, + "step": 39914 + }, + { + "epoch": 8.71, + "learning_rate": 6.450731920471926e-07, + "loss": 0.8426, + "step": 39915 + }, + { + "epoch": 8.71, + "learning_rate": 6.449639501857112e-07, + "loss": 0.5764, + "step": 39916 + }, + { + "epoch": 8.71, + "learning_rate": 6.448547083242299e-07, + "loss": 0.7211, + "step": 39917 + }, + { + "epoch": 8.71, + "learning_rate": 6.447454664627485e-07, + "loss": 0.9275, + "step": 39918 + }, + { + "epoch": 8.71, + "learning_rate": 6.446362246012673e-07, + "loss": 0.7015, + "step": 39919 + }, + { + "epoch": 8.71, + "learning_rate": 6.44526982739786e-07, + "loss": 1.2574, + "step": 39920 + }, + { + "epoch": 8.71, + "learning_rate": 6.444177408783047e-07, + "loss": 1.1182, + "step": 39921 + }, + { + "epoch": 8.71, + "learning_rate": 6.443084990168233e-07, + "loss": 0.8408, + "step": 39922 + }, + { + "epoch": 8.71, + "learning_rate": 6.44199257155342e-07, + "loss": 1.1899, + "step": 39923 + }, + { + "epoch": 8.71, + "learning_rate": 6.440900152938606e-07, + "loss": 1.2094, + "step": 39924 + }, + { + "epoch": 8.71, + "learning_rate": 6.439807734323793e-07, + "loss": 1.0653, + "step": 39925 + }, + { + "epoch": 8.71, + "learning_rate": 6.43871531570898e-07, + "loss": 1.401, + "step": 39926 + }, + { + "epoch": 8.71, + "learning_rate": 6.437622897094168e-07, + "loss": 0.2218, + "step": 39927 + }, + { + "epoch": 8.71, + "learning_rate": 6.436530478479354e-07, + "loss": 0.5338, + "step": 39928 + }, + { + "epoch": 8.71, + "learning_rate": 6.435438059864541e-07, + "loss": 1.2118, + "step": 39929 + }, + { + "epoch": 8.71, + "learning_rate": 6.434345641249727e-07, + "loss": 1.1229, + "step": 39930 + }, + { + "epoch": 8.71, + "learning_rate": 6.433253222634914e-07, + "loss": 1.1114, + "step": 39931 + }, + { + "epoch": 8.71, + "learning_rate": 6.432160804020101e-07, + "loss": 0.0409, + "step": 39932 + }, + { + "epoch": 8.72, + "learning_rate": 6.431068385405288e-07, + "loss": 0.6582, + "step": 39933 + }, + { + "epoch": 8.72, + "learning_rate": 6.429975966790475e-07, + "loss": 0.1565, + "step": 39934 + }, + { + "epoch": 8.72, + "learning_rate": 6.428883548175662e-07, + "loss": 0.9247, + "step": 39935 + }, + { + "epoch": 8.72, + "learning_rate": 6.427791129560848e-07, + "loss": 0.944, + "step": 39936 + }, + { + "epoch": 8.72, + "learning_rate": 6.426698710946035e-07, + "loss": 1.0833, + "step": 39937 + }, + { + "epoch": 8.72, + "learning_rate": 6.425606292331221e-07, + "loss": 0.852, + "step": 39938 + }, + { + "epoch": 8.72, + "learning_rate": 6.424513873716409e-07, + "loss": 1.7031, + "step": 39939 + }, + { + "epoch": 8.72, + "learning_rate": 6.423421455101595e-07, + "loss": 1.3983, + "step": 39940 + }, + { + "epoch": 8.72, + "learning_rate": 6.422329036486783e-07, + "loss": 1.1898, + "step": 39941 + }, + { + "epoch": 8.72, + "learning_rate": 6.421236617871969e-07, + "loss": 0.7006, + "step": 39942 + }, + { + "epoch": 8.72, + "learning_rate": 6.420144199257156e-07, + "loss": 0.9575, + "step": 39943 + }, + { + "epoch": 8.72, + "learning_rate": 6.419051780642342e-07, + "loss": 2.3187, + "step": 39944 + }, + { + "epoch": 8.72, + "learning_rate": 6.417959362027529e-07, + "loss": 1.6095, + "step": 39945 + }, + { + "epoch": 8.72, + "learning_rate": 6.416866943412716e-07, + "loss": 1.5203, + "step": 39946 + }, + { + "epoch": 8.72, + "learning_rate": 6.415774524797904e-07, + "loss": 0.6884, + "step": 39947 + }, + { + "epoch": 8.72, + "learning_rate": 6.41468210618309e-07, + "loss": 0.6984, + "step": 39948 + }, + { + "epoch": 8.72, + "learning_rate": 6.413589687568277e-07, + "loss": 0.6057, + "step": 39949 + }, + { + "epoch": 8.72, + "learning_rate": 6.412497268953463e-07, + "loss": 0.5176, + "step": 39950 + }, + { + "epoch": 8.72, + "learning_rate": 6.41140485033865e-07, + "loss": 0.9554, + "step": 39951 + }, + { + "epoch": 8.72, + "learning_rate": 6.410312431723837e-07, + "loss": 1.7007, + "step": 39952 + }, + { + "epoch": 8.72, + "learning_rate": 6.409220013109024e-07, + "loss": 2.165, + "step": 39953 + }, + { + "epoch": 8.72, + "learning_rate": 6.40812759449421e-07, + "loss": 0.7376, + "step": 39954 + }, + { + "epoch": 8.72, + "learning_rate": 6.407035175879398e-07, + "loss": 1.0458, + "step": 39955 + }, + { + "epoch": 8.72, + "learning_rate": 6.405942757264584e-07, + "loss": 1.5627, + "step": 39956 + }, + { + "epoch": 8.72, + "learning_rate": 6.404850338649771e-07, + "loss": 1.4958, + "step": 39957 + }, + { + "epoch": 8.72, + "learning_rate": 6.403757920034957e-07, + "loss": 1.6244, + "step": 39958 + }, + { + "epoch": 8.72, + "learning_rate": 6.402665501420145e-07, + "loss": 1.0837, + "step": 39959 + }, + { + "epoch": 8.72, + "learning_rate": 6.401573082805331e-07, + "loss": 2.1182, + "step": 39960 + }, + { + "epoch": 8.72, + "learning_rate": 6.400480664190519e-07, + "loss": 1.1931, + "step": 39961 + }, + { + "epoch": 8.72, + "learning_rate": 6.399388245575705e-07, + "loss": 2.2784, + "step": 39962 + }, + { + "epoch": 8.72, + "learning_rate": 6.398295826960892e-07, + "loss": 0.1527, + "step": 39963 + }, + { + "epoch": 8.72, + "learning_rate": 6.397203408346078e-07, + "loss": 1.7657, + "step": 39964 + }, + { + "epoch": 8.72, + "learning_rate": 6.396110989731266e-07, + "loss": 0.1214, + "step": 39965 + }, + { + "epoch": 8.72, + "learning_rate": 6.395018571116452e-07, + "loss": 2.6152, + "step": 39966 + }, + { + "epoch": 8.72, + "learning_rate": 6.39392615250164e-07, + "loss": 1.0631, + "step": 39967 + }, + { + "epoch": 8.72, + "learning_rate": 6.392833733886826e-07, + "loss": 0.7621, + "step": 39968 + }, + { + "epoch": 8.72, + "learning_rate": 6.391741315272013e-07, + "loss": 0.9762, + "step": 39969 + }, + { + "epoch": 8.72, + "learning_rate": 6.390648896657199e-07, + "loss": 1.6103, + "step": 39970 + }, + { + "epoch": 8.72, + "learning_rate": 6.389556478042386e-07, + "loss": 1.5682, + "step": 39971 + }, + { + "epoch": 8.72, + "learning_rate": 6.388464059427574e-07, + "loss": 0.7342, + "step": 39972 + }, + { + "epoch": 8.72, + "learning_rate": 6.38737164081276e-07, + "loss": 1.3566, + "step": 39973 + }, + { + "epoch": 8.72, + "learning_rate": 6.386279222197946e-07, + "loss": 2.9344, + "step": 39974 + }, + { + "epoch": 8.72, + "learning_rate": 6.385186803583134e-07, + "loss": 3.0926, + "step": 39975 + }, + { + "epoch": 8.72, + "learning_rate": 6.38409438496832e-07, + "loss": 2.1678, + "step": 39976 + }, + { + "epoch": 8.72, + "learning_rate": 6.383001966353507e-07, + "loss": 1.1653, + "step": 39977 + }, + { + "epoch": 8.73, + "learning_rate": 6.381909547738693e-07, + "loss": 1.1162, + "step": 39978 + }, + { + "epoch": 8.73, + "learning_rate": 6.380817129123881e-07, + "loss": 0.1994, + "step": 39979 + }, + { + "epoch": 8.73, + "learning_rate": 6.379724710509067e-07, + "loss": 1.7215, + "step": 39980 + }, + { + "epoch": 8.73, + "learning_rate": 6.378632291894255e-07, + "loss": 1.9998, + "step": 39981 + }, + { + "epoch": 8.73, + "learning_rate": 6.377539873279441e-07, + "loss": 2.6758, + "step": 39982 + }, + { + "epoch": 8.73, + "learning_rate": 6.376447454664628e-07, + "loss": 3.8746, + "step": 39983 + }, + { + "epoch": 8.73, + "learning_rate": 6.375355036049814e-07, + "loss": 0.6394, + "step": 39984 + }, + { + "epoch": 8.73, + "learning_rate": 6.374262617435002e-07, + "loss": 1.2, + "step": 39985 + }, + { + "epoch": 8.73, + "learning_rate": 6.373170198820189e-07, + "loss": 1.1795, + "step": 39986 + }, + { + "epoch": 8.73, + "learning_rate": 6.372077780205375e-07, + "loss": 1.1591, + "step": 39987 + }, + { + "epoch": 8.73, + "learning_rate": 6.370985361590562e-07, + "loss": 1.387, + "step": 39988 + }, + { + "epoch": 8.73, + "learning_rate": 6.369892942975749e-07, + "loss": 2.4821, + "step": 39989 + }, + { + "epoch": 8.73, + "learning_rate": 6.368800524360935e-07, + "loss": 0.2324, + "step": 39990 + }, + { + "epoch": 8.73, + "learning_rate": 6.367708105746122e-07, + "loss": 0.8019, + "step": 39991 + }, + { + "epoch": 8.73, + "learning_rate": 6.36661568713131e-07, + "loss": 0.8862, + "step": 39992 + }, + { + "epoch": 8.73, + "learning_rate": 6.365523268516496e-07, + "loss": 0.2129, + "step": 39993 + }, + { + "epoch": 8.73, + "learning_rate": 6.364430849901682e-07, + "loss": 0.9762, + "step": 39994 + }, + { + "epoch": 8.73, + "learning_rate": 6.36333843128687e-07, + "loss": 0.2424, + "step": 39995 + }, + { + "epoch": 8.73, + "learning_rate": 6.362246012672056e-07, + "loss": 2.9799, + "step": 39996 + }, + { + "epoch": 8.73, + "learning_rate": 6.361153594057243e-07, + "loss": 0.3205, + "step": 39997 + }, + { + "epoch": 8.73, + "learning_rate": 6.360061175442431e-07, + "loss": 1.8976, + "step": 39998 + }, + { + "epoch": 8.73, + "learning_rate": 6.358968756827617e-07, + "loss": 0.7895, + "step": 39999 + }, + { + "epoch": 8.73, + "learning_rate": 6.357876338212804e-07, + "loss": 0.9748, + "step": 40000 + }, + { + "epoch": 8.73, + "learning_rate": 6.35678391959799e-07, + "loss": 0.5761, + "step": 40001 + }, + { + "epoch": 8.73, + "learning_rate": 6.355691500983177e-07, + "loss": 0.681, + "step": 40002 + }, + { + "epoch": 8.73, + "learning_rate": 6.354599082368364e-07, + "loss": 0.2805, + "step": 40003 + }, + { + "epoch": 8.73, + "learning_rate": 6.35350666375355e-07, + "loss": 2.4544, + "step": 40004 + }, + { + "epoch": 8.73, + "learning_rate": 6.352414245138738e-07, + "loss": 1.5716, + "step": 40005 + }, + { + "epoch": 8.73, + "learning_rate": 6.351321826523925e-07, + "loss": 1.1153, + "step": 40006 + }, + { + "epoch": 8.73, + "learning_rate": 6.350229407909111e-07, + "loss": 1.1508, + "step": 40007 + }, + { + "epoch": 8.73, + "learning_rate": 6.349136989294299e-07, + "loss": 1.1659, + "step": 40008 + }, + { + "epoch": 8.73, + "learning_rate": 6.348044570679485e-07, + "loss": 1.0186, + "step": 40009 + }, + { + "epoch": 8.73, + "learning_rate": 6.346952152064671e-07, + "loss": 1.757, + "step": 40010 + }, + { + "epoch": 8.73, + "learning_rate": 6.345859733449858e-07, + "loss": 2.0665, + "step": 40011 + }, + { + "epoch": 8.73, + "learning_rate": 6.344767314835046e-07, + "loss": 0.937, + "step": 40012 + }, + { + "epoch": 8.73, + "learning_rate": 6.343674896220232e-07, + "loss": 1.4166, + "step": 40013 + }, + { + "epoch": 8.73, + "learning_rate": 6.342582477605419e-07, + "loss": 1.2859, + "step": 40014 + }, + { + "epoch": 8.73, + "learning_rate": 6.341490058990606e-07, + "loss": 1.9944, + "step": 40015 + }, + { + "epoch": 8.73, + "learning_rate": 6.340397640375792e-07, + "loss": 0.9325, + "step": 40016 + }, + { + "epoch": 8.73, + "learning_rate": 6.339305221760979e-07, + "loss": 1.7986, + "step": 40017 + }, + { + "epoch": 8.73, + "learning_rate": 6.338212803146167e-07, + "loss": 0.7745, + "step": 40018 + }, + { + "epoch": 8.73, + "learning_rate": 6.337120384531353e-07, + "loss": 0.6455, + "step": 40019 + }, + { + "epoch": 8.73, + "learning_rate": 6.33602796591654e-07, + "loss": 0.9375, + "step": 40020 + }, + { + "epoch": 8.73, + "learning_rate": 6.334935547301726e-07, + "loss": 1.2149, + "step": 40021 + }, + { + "epoch": 8.73, + "learning_rate": 6.333843128686914e-07, + "loss": 1.462, + "step": 40022 + }, + { + "epoch": 8.73, + "learning_rate": 6.3327507100721e-07, + "loss": 1.9139, + "step": 40023 + }, + { + "epoch": 8.74, + "learning_rate": 6.331658291457286e-07, + "loss": 1.8684, + "step": 40024 + }, + { + "epoch": 8.74, + "learning_rate": 6.330565872842474e-07, + "loss": 1.2389, + "step": 40025 + }, + { + "epoch": 8.74, + "learning_rate": 6.329473454227661e-07, + "loss": 0.3266, + "step": 40026 + }, + { + "epoch": 8.74, + "learning_rate": 6.328381035612847e-07, + "loss": 0.5208, + "step": 40027 + }, + { + "epoch": 8.74, + "learning_rate": 6.327288616998035e-07, + "loss": 1.3858, + "step": 40028 + }, + { + "epoch": 8.74, + "learning_rate": 6.326196198383221e-07, + "loss": 2.4547, + "step": 40029 + }, + { + "epoch": 8.74, + "learning_rate": 6.325103779768407e-07, + "loss": 0.1943, + "step": 40030 + }, + { + "epoch": 8.74, + "learning_rate": 6.324011361153595e-07, + "loss": 0.5648, + "step": 40031 + }, + { + "epoch": 8.74, + "learning_rate": 6.322918942538782e-07, + "loss": 2.0641, + "step": 40032 + }, + { + "epoch": 8.74, + "learning_rate": 6.321826523923968e-07, + "loss": 1.3241, + "step": 40033 + }, + { + "epoch": 8.74, + "learning_rate": 6.320734105309155e-07, + "loss": 1.5826, + "step": 40034 + }, + { + "epoch": 8.74, + "learning_rate": 6.319641686694342e-07, + "loss": 1.8914, + "step": 40035 + }, + { + "epoch": 8.74, + "learning_rate": 6.318549268079529e-07, + "loss": 1.3863, + "step": 40036 + }, + { + "epoch": 8.74, + "learning_rate": 6.317456849464715e-07, + "loss": 0.2062, + "step": 40037 + }, + { + "epoch": 8.74, + "learning_rate": 6.316364430849903e-07, + "loss": 1.2524, + "step": 40038 + }, + { + "epoch": 8.74, + "learning_rate": 6.315272012235089e-07, + "loss": 0.1243, + "step": 40039 + }, + { + "epoch": 8.74, + "learning_rate": 6.314179593620276e-07, + "loss": 0.183, + "step": 40040 + }, + { + "epoch": 8.74, + "learning_rate": 6.313087175005462e-07, + "loss": 2.1211, + "step": 40041 + }, + { + "epoch": 8.74, + "learning_rate": 6.31199475639065e-07, + "loss": 0.112, + "step": 40042 + }, + { + "epoch": 8.74, + "learning_rate": 6.310902337775836e-07, + "loss": 2.2667, + "step": 40043 + }, + { + "epoch": 8.74, + "learning_rate": 6.309809919161023e-07, + "loss": 0.7124, + "step": 40044 + }, + { + "epoch": 8.74, + "learning_rate": 6.30871750054621e-07, + "loss": 1.4408, + "step": 40045 + }, + { + "epoch": 8.74, + "learning_rate": 6.307625081931397e-07, + "loss": 2.0824, + "step": 40046 + }, + { + "epoch": 8.74, + "learning_rate": 6.306532663316583e-07, + "loss": 0.6419, + "step": 40047 + }, + { + "epoch": 8.74, + "learning_rate": 6.30544024470177e-07, + "loss": 0.5223, + "step": 40048 + }, + { + "epoch": 8.74, + "learning_rate": 6.304347826086957e-07, + "loss": 1.4407, + "step": 40049 + }, + { + "epoch": 8.74, + "learning_rate": 6.303255407472144e-07, + "loss": 1.3828, + "step": 40050 + }, + { + "epoch": 8.74, + "learning_rate": 6.302162988857331e-07, + "loss": 2.486, + "step": 40051 + }, + { + "epoch": 8.74, + "learning_rate": 6.301070570242518e-07, + "loss": 1.1447, + "step": 40052 + }, + { + "epoch": 8.74, + "learning_rate": 6.299978151627704e-07, + "loss": 0.1582, + "step": 40053 + }, + { + "epoch": 8.74, + "learning_rate": 6.298885733012891e-07, + "loss": 0.7616, + "step": 40054 + }, + { + "epoch": 8.74, + "learning_rate": 6.297793314398077e-07, + "loss": 1.6801, + "step": 40055 + }, + { + "epoch": 8.74, + "learning_rate": 6.296700895783265e-07, + "loss": 0.6949, + "step": 40056 + }, + { + "epoch": 8.74, + "learning_rate": 6.295608477168451e-07, + "loss": 1.219, + "step": 40057 + }, + { + "epoch": 8.74, + "learning_rate": 6.294516058553639e-07, + "loss": 1.0045, + "step": 40058 + }, + { + "epoch": 8.74, + "learning_rate": 6.293423639938825e-07, + "loss": 1.0224, + "step": 40059 + }, + { + "epoch": 8.74, + "learning_rate": 6.292331221324012e-07, + "loss": 0.4852, + "step": 40060 + }, + { + "epoch": 8.74, + "learning_rate": 6.291238802709198e-07, + "loss": 1.2373, + "step": 40061 + }, + { + "epoch": 8.74, + "learning_rate": 6.290146384094386e-07, + "loss": 1.5297, + "step": 40062 + }, + { + "epoch": 8.74, + "learning_rate": 6.289053965479572e-07, + "loss": 0.7608, + "step": 40063 + }, + { + "epoch": 8.74, + "learning_rate": 6.287961546864759e-07, + "loss": 0.4641, + "step": 40064 + }, + { + "epoch": 8.74, + "learning_rate": 6.286869128249946e-07, + "loss": 0.2887, + "step": 40065 + }, + { + "epoch": 8.74, + "learning_rate": 6.285776709635133e-07, + "loss": 0.4596, + "step": 40066 + }, + { + "epoch": 8.74, + "learning_rate": 6.284684291020319e-07, + "loss": 0.156, + "step": 40067 + }, + { + "epoch": 8.74, + "learning_rate": 6.283591872405506e-07, + "loss": 1.732, + "step": 40068 + }, + { + "epoch": 8.74, + "learning_rate": 6.282499453790693e-07, + "loss": 0.1711, + "step": 40069 + }, + { + "epoch": 8.75, + "learning_rate": 6.28140703517588e-07, + "loss": 2.0597, + "step": 40070 + }, + { + "epoch": 8.75, + "learning_rate": 6.280314616561067e-07, + "loss": 0.2463, + "step": 40071 + }, + { + "epoch": 8.75, + "learning_rate": 6.279222197946254e-07, + "loss": 0.6543, + "step": 40072 + }, + { + "epoch": 8.75, + "learning_rate": 6.27812977933144e-07, + "loss": 1.2334, + "step": 40073 + }, + { + "epoch": 8.75, + "learning_rate": 6.277037360716627e-07, + "loss": 0.993, + "step": 40074 + }, + { + "epoch": 8.75, + "learning_rate": 6.275944942101813e-07, + "loss": 1.6339, + "step": 40075 + }, + { + "epoch": 8.75, + "learning_rate": 6.274852523487001e-07, + "loss": 2.5625, + "step": 40076 + }, + { + "epoch": 8.75, + "learning_rate": 6.273760104872187e-07, + "loss": 0.7864, + "step": 40077 + }, + { + "epoch": 8.75, + "learning_rate": 6.272667686257375e-07, + "loss": 1.0724, + "step": 40078 + }, + { + "epoch": 8.75, + "learning_rate": 6.271575267642561e-07, + "loss": 1.8638, + "step": 40079 + }, + { + "epoch": 8.75, + "learning_rate": 6.270482849027748e-07, + "loss": 0.3379, + "step": 40080 + }, + { + "epoch": 8.75, + "learning_rate": 6.269390430412934e-07, + "loss": 0.5063, + "step": 40081 + }, + { + "epoch": 8.75, + "learning_rate": 6.268298011798122e-07, + "loss": 0.7442, + "step": 40082 + }, + { + "epoch": 8.75, + "learning_rate": 6.267205593183308e-07, + "loss": 1.1664, + "step": 40083 + }, + { + "epoch": 8.75, + "learning_rate": 6.266113174568496e-07, + "loss": 0.3348, + "step": 40084 + }, + { + "epoch": 8.75, + "learning_rate": 6.265020755953682e-07, + "loss": 1.2722, + "step": 40085 + }, + { + "epoch": 8.75, + "learning_rate": 6.263928337338869e-07, + "loss": 2.7829, + "step": 40086 + }, + { + "epoch": 8.75, + "learning_rate": 6.262835918724055e-07, + "loss": 2.5868, + "step": 40087 + }, + { + "epoch": 8.75, + "learning_rate": 6.261743500109242e-07, + "loss": 1.0443, + "step": 40088 + }, + { + "epoch": 8.75, + "learning_rate": 6.260651081494429e-07, + "loss": 0.4295, + "step": 40089 + }, + { + "epoch": 8.75, + "learning_rate": 6.259558662879616e-07, + "loss": 2.7711, + "step": 40090 + }, + { + "epoch": 8.75, + "learning_rate": 6.258466244264803e-07, + "loss": 2.0334, + "step": 40091 + }, + { + "epoch": 8.75, + "learning_rate": 6.25737382564999e-07, + "loss": 0.7403, + "step": 40092 + }, + { + "epoch": 8.75, + "learning_rate": 6.256281407035176e-07, + "loss": 1.5354, + "step": 40093 + }, + { + "epoch": 8.75, + "learning_rate": 6.255188988420363e-07, + "loss": 0.7427, + "step": 40094 + }, + { + "epoch": 8.75, + "learning_rate": 6.254096569805549e-07, + "loss": 2.7232, + "step": 40095 + }, + { + "epoch": 8.75, + "learning_rate": 6.253004151190737e-07, + "loss": 1.8925, + "step": 40096 + }, + { + "epoch": 8.75, + "learning_rate": 6.251911732575923e-07, + "loss": 2.7928, + "step": 40097 + }, + { + "epoch": 8.75, + "learning_rate": 6.250819313961111e-07, + "loss": 3.3756, + "step": 40098 + }, + { + "epoch": 8.75, + "learning_rate": 6.249726895346297e-07, + "loss": 1.1359, + "step": 40099 + }, + { + "epoch": 8.75, + "learning_rate": 6.248634476731484e-07, + "loss": 2.004, + "step": 40100 + }, + { + "epoch": 8.75, + "learning_rate": 6.24754205811667e-07, + "loss": 0.9062, + "step": 40101 + }, + { + "epoch": 8.75, + "learning_rate": 6.246449639501857e-07, + "loss": 0.5765, + "step": 40102 + }, + { + "epoch": 8.75, + "learning_rate": 6.245357220887045e-07, + "loss": 0.0354, + "step": 40103 + }, + { + "epoch": 8.75, + "learning_rate": 6.244264802272231e-07, + "loss": 1.2345, + "step": 40104 + }, + { + "epoch": 8.75, + "learning_rate": 6.243172383657418e-07, + "loss": 0.7058, + "step": 40105 + }, + { + "epoch": 8.75, + "learning_rate": 6.242079965042605e-07, + "loss": 0.8362, + "step": 40106 + }, + { + "epoch": 8.75, + "learning_rate": 6.240987546427791e-07, + "loss": 0.6397, + "step": 40107 + }, + { + "epoch": 8.75, + "learning_rate": 6.239895127812978e-07, + "loss": 2.1817, + "step": 40108 + }, + { + "epoch": 8.75, + "learning_rate": 6.238802709198166e-07, + "loss": 1.3046, + "step": 40109 + }, + { + "epoch": 8.75, + "learning_rate": 6.237710290583352e-07, + "loss": 0.434, + "step": 40110 + }, + { + "epoch": 8.75, + "learning_rate": 6.236617871968539e-07, + "loss": 1.1732, + "step": 40111 + }, + { + "epoch": 8.75, + "learning_rate": 6.235525453353726e-07, + "loss": 1.5967, + "step": 40112 + }, + { + "epoch": 8.75, + "learning_rate": 6.234433034738912e-07, + "loss": 0.8198, + "step": 40113 + }, + { + "epoch": 8.75, + "learning_rate": 6.233340616124099e-07, + "loss": 1.0934, + "step": 40114 + }, + { + "epoch": 8.75, + "learning_rate": 6.232248197509285e-07, + "loss": 0.4593, + "step": 40115 + }, + { + "epoch": 8.76, + "learning_rate": 6.231155778894473e-07, + "loss": 1.0445, + "step": 40116 + }, + { + "epoch": 8.76, + "learning_rate": 6.23006336027966e-07, + "loss": 1.687, + "step": 40117 + }, + { + "epoch": 8.76, + "learning_rate": 6.228970941664846e-07, + "loss": 0.7566, + "step": 40118 + }, + { + "epoch": 8.76, + "learning_rate": 6.227878523050034e-07, + "loss": 0.6026, + "step": 40119 + }, + { + "epoch": 8.76, + "learning_rate": 6.22678610443522e-07, + "loss": 1.0914, + "step": 40120 + }, + { + "epoch": 8.76, + "learning_rate": 6.225693685820406e-07, + "loss": 0.4586, + "step": 40121 + }, + { + "epoch": 8.76, + "learning_rate": 6.224601267205595e-07, + "loss": 0.9304, + "step": 40122 + }, + { + "epoch": 8.76, + "learning_rate": 6.223508848590781e-07, + "loss": 1.0228, + "step": 40123 + }, + { + "epoch": 8.76, + "learning_rate": 6.222416429975967e-07, + "loss": 1.6536, + "step": 40124 + }, + { + "epoch": 8.76, + "learning_rate": 6.221324011361154e-07, + "loss": 1.2174, + "step": 40125 + }, + { + "epoch": 8.76, + "learning_rate": 6.220231592746341e-07, + "loss": 0.6957, + "step": 40126 + }, + { + "epoch": 8.76, + "learning_rate": 6.219139174131527e-07, + "loss": 0.457, + "step": 40127 + }, + { + "epoch": 8.76, + "learning_rate": 6.218046755516714e-07, + "loss": 0.2168, + "step": 40128 + }, + { + "epoch": 8.76, + "learning_rate": 6.216954336901902e-07, + "loss": 1.2937, + "step": 40129 + }, + { + "epoch": 8.76, + "learning_rate": 6.215861918287089e-07, + "loss": 0.8174, + "step": 40130 + }, + { + "epoch": 8.76, + "learning_rate": 6.214769499672275e-07, + "loss": 2.5198, + "step": 40131 + }, + { + "epoch": 8.76, + "learning_rate": 6.213677081057462e-07, + "loss": 0.6723, + "step": 40132 + }, + { + "epoch": 8.76, + "learning_rate": 6.212584662442649e-07, + "loss": 0.6221, + "step": 40133 + }, + { + "epoch": 8.76, + "learning_rate": 6.211492243827835e-07, + "loss": 0.3142, + "step": 40134 + }, + { + "epoch": 8.76, + "learning_rate": 6.210399825213021e-07, + "loss": 1.9205, + "step": 40135 + }, + { + "epoch": 8.76, + "learning_rate": 6.20930740659821e-07, + "loss": 2.0797, + "step": 40136 + }, + { + "epoch": 8.76, + "learning_rate": 6.208214987983396e-07, + "loss": 2.4186, + "step": 40137 + }, + { + "epoch": 8.76, + "learning_rate": 6.207122569368582e-07, + "loss": 1.7467, + "step": 40138 + }, + { + "epoch": 8.76, + "learning_rate": 6.20603015075377e-07, + "loss": 1.1713, + "step": 40139 + }, + { + "epoch": 8.76, + "learning_rate": 6.204937732138956e-07, + "loss": 0.9242, + "step": 40140 + }, + { + "epoch": 8.76, + "learning_rate": 6.203845313524142e-07, + "loss": 1.6837, + "step": 40141 + }, + { + "epoch": 8.76, + "learning_rate": 6.20275289490933e-07, + "loss": 0.0455, + "step": 40142 + }, + { + "epoch": 8.76, + "learning_rate": 6.201660476294517e-07, + "loss": 0.9375, + "step": 40143 + }, + { + "epoch": 8.76, + "learning_rate": 6.200568057679704e-07, + "loss": 1.2428, + "step": 40144 + }, + { + "epoch": 8.76, + "learning_rate": 6.19947563906489e-07, + "loss": 1.9142, + "step": 40145 + }, + { + "epoch": 8.76, + "learning_rate": 6.198383220450077e-07, + "loss": 1.4757, + "step": 40146 + }, + { + "epoch": 8.76, + "learning_rate": 6.197290801835264e-07, + "loss": 1.2528, + "step": 40147 + }, + { + "epoch": 8.76, + "learning_rate": 6.19619838322045e-07, + "loss": 1.0053, + "step": 40148 + }, + { + "epoch": 8.76, + "learning_rate": 6.195105964605637e-07, + "loss": 0.7699, + "step": 40149 + }, + { + "epoch": 8.76, + "learning_rate": 6.194013545990825e-07, + "loss": 1.1237, + "step": 40150 + }, + { + "epoch": 8.76, + "learning_rate": 6.192921127376011e-07, + "loss": 2.0767, + "step": 40151 + }, + { + "epoch": 8.76, + "learning_rate": 6.191828708761198e-07, + "loss": 0.4661, + "step": 40152 + }, + { + "epoch": 8.76, + "learning_rate": 6.190736290146385e-07, + "loss": 1.066, + "step": 40153 + }, + { + "epoch": 8.76, + "learning_rate": 6.189643871531571e-07, + "loss": 1.0505, + "step": 40154 + }, + { + "epoch": 8.76, + "learning_rate": 6.188551452916758e-07, + "loss": 2.9231, + "step": 40155 + }, + { + "epoch": 8.76, + "learning_rate": 6.187459034301946e-07, + "loss": 1.2357, + "step": 40156 + }, + { + "epoch": 8.76, + "learning_rate": 6.186366615687132e-07, + "loss": 1.0161, + "step": 40157 + }, + { + "epoch": 8.76, + "learning_rate": 6.185274197072319e-07, + "loss": 2.9334, + "step": 40158 + }, + { + "epoch": 8.76, + "learning_rate": 6.184181778457506e-07, + "loss": 0.376, + "step": 40159 + }, + { + "epoch": 8.76, + "learning_rate": 6.183089359842692e-07, + "loss": 0.1051, + "step": 40160 + }, + { + "epoch": 8.76, + "learning_rate": 6.181996941227879e-07, + "loss": 0.3971, + "step": 40161 + }, + { + "epoch": 8.77, + "learning_rate": 6.180904522613066e-07, + "loss": 2.3252, + "step": 40162 + }, + { + "epoch": 8.77, + "learning_rate": 6.179812103998253e-07, + "loss": 0.461, + "step": 40163 + }, + { + "epoch": 8.77, + "learning_rate": 6.17871968538344e-07, + "loss": 0.5424, + "step": 40164 + }, + { + "epoch": 8.77, + "learning_rate": 6.177627266768627e-07, + "loss": 2.6563, + "step": 40165 + }, + { + "epoch": 8.77, + "learning_rate": 6.176534848153813e-07, + "loss": 1.6298, + "step": 40166 + }, + { + "epoch": 8.77, + "learning_rate": 6.175442429539e-07, + "loss": 0.9122, + "step": 40167 + }, + { + "epoch": 8.77, + "learning_rate": 6.174350010924186e-07, + "loss": 1.7551, + "step": 40168 + }, + { + "epoch": 8.77, + "learning_rate": 6.173257592309373e-07, + "loss": 0.1078, + "step": 40169 + }, + { + "epoch": 8.77, + "learning_rate": 6.172165173694561e-07, + "loss": 0.3564, + "step": 40170 + }, + { + "epoch": 8.77, + "learning_rate": 6.171072755079747e-07, + "loss": 0.1729, + "step": 40171 + }, + { + "epoch": 8.77, + "learning_rate": 6.169980336464934e-07, + "loss": 1.0903, + "step": 40172 + }, + { + "epoch": 8.77, + "learning_rate": 6.168887917850121e-07, + "loss": 2.0245, + "step": 40173 + }, + { + "epoch": 8.77, + "learning_rate": 6.167795499235307e-07, + "loss": 0.5611, + "step": 40174 + }, + { + "epoch": 8.77, + "learning_rate": 6.166703080620494e-07, + "loss": 1.8853, + "step": 40175 + }, + { + "epoch": 8.77, + "learning_rate": 6.165610662005681e-07, + "loss": 1.4813, + "step": 40176 + }, + { + "epoch": 8.77, + "learning_rate": 6.164518243390868e-07, + "loss": 1.876, + "step": 40177 + }, + { + "epoch": 8.77, + "learning_rate": 6.163425824776055e-07, + "loss": 2.2357, + "step": 40178 + }, + { + "epoch": 8.77, + "learning_rate": 6.162333406161242e-07, + "loss": 0.8676, + "step": 40179 + }, + { + "epoch": 8.77, + "learning_rate": 6.161240987546428e-07, + "loss": 0.519, + "step": 40180 + }, + { + "epoch": 8.77, + "learning_rate": 6.160148568931615e-07, + "loss": 0.5591, + "step": 40181 + }, + { + "epoch": 8.77, + "learning_rate": 6.159056150316802e-07, + "loss": 0.5967, + "step": 40182 + }, + { + "epoch": 8.77, + "learning_rate": 6.157963731701989e-07, + "loss": 0.2141, + "step": 40183 + }, + { + "epoch": 8.77, + "learning_rate": 6.156871313087176e-07, + "loss": 0.9856, + "step": 40184 + }, + { + "epoch": 8.77, + "learning_rate": 6.155778894472363e-07, + "loss": 0.6182, + "step": 40185 + }, + { + "epoch": 8.77, + "learning_rate": 6.154686475857549e-07, + "loss": 0.8726, + "step": 40186 + }, + { + "epoch": 8.77, + "learning_rate": 6.153594057242736e-07, + "loss": 0.8017, + "step": 40187 + }, + { + "epoch": 8.77, + "learning_rate": 6.152501638627922e-07, + "loss": 0.0732, + "step": 40188 + }, + { + "epoch": 8.77, + "learning_rate": 6.151409220013109e-07, + "loss": 0.6271, + "step": 40189 + }, + { + "epoch": 8.77, + "learning_rate": 6.150316801398297e-07, + "loss": 1.2117, + "step": 40190 + }, + { + "epoch": 8.77, + "learning_rate": 6.149224382783483e-07, + "loss": 0.91, + "step": 40191 + }, + { + "epoch": 8.77, + "learning_rate": 6.14813196416867e-07, + "loss": 0.7817, + "step": 40192 + }, + { + "epoch": 8.77, + "learning_rate": 6.147039545553857e-07, + "loss": 2.1006, + "step": 40193 + }, + { + "epoch": 8.77, + "learning_rate": 6.145947126939043e-07, + "loss": 0.6375, + "step": 40194 + }, + { + "epoch": 8.77, + "learning_rate": 6.14485470832423e-07, + "loss": 0.0433, + "step": 40195 + }, + { + "epoch": 8.77, + "learning_rate": 6.143762289709417e-07, + "loss": 1.8566, + "step": 40196 + }, + { + "epoch": 8.77, + "learning_rate": 6.142669871094604e-07, + "loss": 0.3489, + "step": 40197 + }, + { + "epoch": 8.77, + "learning_rate": 6.141577452479791e-07, + "loss": 1.2161, + "step": 40198 + }, + { + "epoch": 8.77, + "learning_rate": 6.140485033864978e-07, + "loss": 1.9378, + "step": 40199 + }, + { + "epoch": 8.77, + "learning_rate": 6.139392615250164e-07, + "loss": 0.7468, + "step": 40200 + }, + { + "epoch": 8.77, + "learning_rate": 6.138300196635351e-07, + "loss": 1.9203, + "step": 40201 + }, + { + "epoch": 8.77, + "learning_rate": 6.137207778020538e-07, + "loss": 1.5583, + "step": 40202 + }, + { + "epoch": 8.77, + "learning_rate": 6.136115359405724e-07, + "loss": 1.1132, + "step": 40203 + }, + { + "epoch": 8.77, + "learning_rate": 6.135022940790912e-07, + "loss": 1.3065, + "step": 40204 + }, + { + "epoch": 8.77, + "learning_rate": 6.133930522176099e-07, + "loss": 0.6645, + "step": 40205 + }, + { + "epoch": 8.77, + "learning_rate": 6.132838103561285e-07, + "loss": 0.5749, + "step": 40206 + }, + { + "epoch": 8.77, + "learning_rate": 6.131745684946472e-07, + "loss": 1.3101, + "step": 40207 + }, + { + "epoch": 8.78, + "learning_rate": 6.130653266331659e-07, + "loss": 0.7626, + "step": 40208 + }, + { + "epoch": 8.78, + "learning_rate": 6.129560847716845e-07, + "loss": 0.6528, + "step": 40209 + }, + { + "epoch": 8.78, + "learning_rate": 6.128468429102033e-07, + "loss": 1.806, + "step": 40210 + }, + { + "epoch": 8.78, + "learning_rate": 6.127376010487219e-07, + "loss": 1.0298, + "step": 40211 + }, + { + "epoch": 8.78, + "learning_rate": 6.126283591872406e-07, + "loss": 1.12, + "step": 40212 + }, + { + "epoch": 8.78, + "learning_rate": 6.125191173257593e-07, + "loss": 0.1256, + "step": 40213 + }, + { + "epoch": 8.78, + "learning_rate": 6.124098754642779e-07, + "loss": 2.596, + "step": 40214 + }, + { + "epoch": 8.78, + "learning_rate": 6.123006336027966e-07, + "loss": 1.0104, + "step": 40215 + }, + { + "epoch": 8.78, + "learning_rate": 6.121913917413153e-07, + "loss": 0.1087, + "step": 40216 + }, + { + "epoch": 8.78, + "learning_rate": 6.12082149879834e-07, + "loss": 1.5888, + "step": 40217 + }, + { + "epoch": 8.78, + "learning_rate": 6.119729080183527e-07, + "loss": 0.3707, + "step": 40218 + }, + { + "epoch": 8.78, + "learning_rate": 6.118636661568714e-07, + "loss": 2.1286, + "step": 40219 + }, + { + "epoch": 8.78, + "learning_rate": 6.1175442429539e-07, + "loss": 0.9398, + "step": 40220 + }, + { + "epoch": 8.78, + "learning_rate": 6.116451824339087e-07, + "loss": 2.2037, + "step": 40221 + }, + { + "epoch": 8.78, + "learning_rate": 6.115359405724274e-07, + "loss": 1.2504, + "step": 40222 + }, + { + "epoch": 8.78, + "learning_rate": 6.11426698710946e-07, + "loss": 0.8252, + "step": 40223 + }, + { + "epoch": 8.78, + "learning_rate": 6.113174568494648e-07, + "loss": 0.9006, + "step": 40224 + }, + { + "epoch": 8.78, + "learning_rate": 6.112082149879835e-07, + "loss": 0.6777, + "step": 40225 + }, + { + "epoch": 8.78, + "learning_rate": 6.110989731265021e-07, + "loss": 0.6742, + "step": 40226 + }, + { + "epoch": 8.78, + "learning_rate": 6.109897312650208e-07, + "loss": 0.5064, + "step": 40227 + }, + { + "epoch": 8.78, + "learning_rate": 6.108804894035395e-07, + "loss": 1.4134, + "step": 40228 + }, + { + "epoch": 8.78, + "learning_rate": 6.107712475420581e-07, + "loss": 1.7566, + "step": 40229 + }, + { + "epoch": 8.78, + "learning_rate": 6.106620056805768e-07, + "loss": 0.0945, + "step": 40230 + }, + { + "epoch": 8.78, + "learning_rate": 6.105527638190956e-07, + "loss": 0.5375, + "step": 40231 + }, + { + "epoch": 8.78, + "learning_rate": 6.104435219576142e-07, + "loss": 0.7496, + "step": 40232 + }, + { + "epoch": 8.78, + "learning_rate": 6.103342800961329e-07, + "loss": 0.9328, + "step": 40233 + }, + { + "epoch": 8.78, + "learning_rate": 6.102250382346515e-07, + "loss": 3.5845, + "step": 40234 + }, + { + "epoch": 8.78, + "learning_rate": 6.101157963731702e-07, + "loss": 0.0734, + "step": 40235 + }, + { + "epoch": 8.78, + "learning_rate": 6.100065545116889e-07, + "loss": 1.2809, + "step": 40236 + }, + { + "epoch": 8.78, + "learning_rate": 6.098973126502075e-07, + "loss": 0.6864, + "step": 40237 + }, + { + "epoch": 8.78, + "learning_rate": 6.097880707887263e-07, + "loss": 0.4477, + "step": 40238 + }, + { + "epoch": 8.78, + "learning_rate": 6.09678828927245e-07, + "loss": 0.5886, + "step": 40239 + }, + { + "epoch": 8.78, + "learning_rate": 6.095695870657636e-07, + "loss": 0.7393, + "step": 40240 + }, + { + "epoch": 8.78, + "learning_rate": 6.094603452042823e-07, + "loss": 1.6931, + "step": 40241 + }, + { + "epoch": 8.78, + "learning_rate": 6.09351103342801e-07, + "loss": 1.0181, + "step": 40242 + }, + { + "epoch": 8.78, + "learning_rate": 6.092418614813196e-07, + "loss": 3.3398, + "step": 40243 + }, + { + "epoch": 8.78, + "learning_rate": 6.091326196198384e-07, + "loss": 1.1541, + "step": 40244 + }, + { + "epoch": 8.78, + "learning_rate": 6.090233777583571e-07, + "loss": 2.8708, + "step": 40245 + }, + { + "epoch": 8.78, + "learning_rate": 6.089141358968757e-07, + "loss": 0.5067, + "step": 40246 + }, + { + "epoch": 8.78, + "learning_rate": 6.088048940353944e-07, + "loss": 1.041, + "step": 40247 + }, + { + "epoch": 8.78, + "learning_rate": 6.086956521739131e-07, + "loss": 1.0036, + "step": 40248 + }, + { + "epoch": 8.78, + "learning_rate": 6.085864103124317e-07, + "loss": 1.2335, + "step": 40249 + }, + { + "epoch": 8.78, + "learning_rate": 6.084771684509504e-07, + "loss": 1.3212, + "step": 40250 + }, + { + "epoch": 8.78, + "learning_rate": 6.083679265894692e-07, + "loss": 1.0577, + "step": 40251 + }, + { + "epoch": 8.78, + "learning_rate": 6.082586847279878e-07, + "loss": 1.6168, + "step": 40252 + }, + { + "epoch": 8.79, + "learning_rate": 6.081494428665065e-07, + "loss": 0.7712, + "step": 40253 + }, + { + "epoch": 8.79, + "learning_rate": 6.080402010050251e-07, + "loss": 0.5493, + "step": 40254 + }, + { + "epoch": 8.79, + "learning_rate": 6.079309591435439e-07, + "loss": 0.2109, + "step": 40255 + }, + { + "epoch": 8.79, + "learning_rate": 6.078217172820625e-07, + "loss": 1.7684, + "step": 40256 + }, + { + "epoch": 8.79, + "learning_rate": 6.077124754205811e-07, + "loss": 0.5407, + "step": 40257 + }, + { + "epoch": 8.79, + "learning_rate": 6.076032335591e-07, + "loss": 0.5746, + "step": 40258 + }, + { + "epoch": 8.79, + "learning_rate": 6.074939916976186e-07, + "loss": 1.8246, + "step": 40259 + }, + { + "epoch": 8.79, + "learning_rate": 6.073847498361372e-07, + "loss": 2.1118, + "step": 40260 + }, + { + "epoch": 8.79, + "learning_rate": 6.07275507974656e-07, + "loss": 1.1894, + "step": 40261 + }, + { + "epoch": 8.79, + "learning_rate": 6.071662661131746e-07, + "loss": 0.8129, + "step": 40262 + }, + { + "epoch": 8.79, + "learning_rate": 6.070570242516932e-07, + "loss": 1.1592, + "step": 40263 + }, + { + "epoch": 8.79, + "learning_rate": 6.06947782390212e-07, + "loss": 0.922, + "step": 40264 + }, + { + "epoch": 8.79, + "learning_rate": 6.068385405287307e-07, + "loss": 0.2637, + "step": 40265 + }, + { + "epoch": 8.79, + "learning_rate": 6.067292986672494e-07, + "loss": 2.6797, + "step": 40266 + }, + { + "epoch": 8.79, + "learning_rate": 6.06620056805768e-07, + "loss": 0.9001, + "step": 40267 + }, + { + "epoch": 8.79, + "learning_rate": 6.065108149442867e-07, + "loss": 1.7534, + "step": 40268 + }, + { + "epoch": 8.79, + "learning_rate": 6.064015730828054e-07, + "loss": 0.7762, + "step": 40269 + }, + { + "epoch": 8.79, + "learning_rate": 6.06292331221324e-07, + "loss": 0.8838, + "step": 40270 + }, + { + "epoch": 8.79, + "learning_rate": 6.061830893598428e-07, + "loss": 1.8498, + "step": 40271 + }, + { + "epoch": 8.79, + "learning_rate": 6.060738474983615e-07, + "loss": 1.4008, + "step": 40272 + }, + { + "epoch": 8.79, + "learning_rate": 6.059646056368801e-07, + "loss": 2.4448, + "step": 40273 + }, + { + "epoch": 8.79, + "learning_rate": 6.058553637753988e-07, + "loss": 0.2182, + "step": 40274 + }, + { + "epoch": 8.79, + "learning_rate": 6.057461219139175e-07, + "loss": 1.3635, + "step": 40275 + }, + { + "epoch": 8.79, + "learning_rate": 6.056368800524361e-07, + "loss": 0.9663, + "step": 40276 + }, + { + "epoch": 8.79, + "learning_rate": 6.055276381909547e-07, + "loss": 1.9337, + "step": 40277 + }, + { + "epoch": 8.79, + "learning_rate": 6.054183963294736e-07, + "loss": 1.254, + "step": 40278 + }, + { + "epoch": 8.79, + "learning_rate": 6.053091544679922e-07, + "loss": 0.6648, + "step": 40279 + }, + { + "epoch": 8.79, + "learning_rate": 6.051999126065109e-07, + "loss": 0.7721, + "step": 40280 + }, + { + "epoch": 8.79, + "learning_rate": 6.050906707450296e-07, + "loss": 1.8648, + "step": 40281 + }, + { + "epoch": 8.79, + "learning_rate": 6.049814288835482e-07, + "loss": 1.4197, + "step": 40282 + }, + { + "epoch": 8.79, + "learning_rate": 6.048721870220669e-07, + "loss": 2.0317, + "step": 40283 + }, + { + "epoch": 8.79, + "learning_rate": 6.047629451605857e-07, + "loss": 2.5578, + "step": 40284 + }, + { + "epoch": 8.79, + "learning_rate": 6.046537032991043e-07, + "loss": 1.6258, + "step": 40285 + }, + { + "epoch": 8.79, + "learning_rate": 6.04544461437623e-07, + "loss": 2.2376, + "step": 40286 + }, + { + "epoch": 8.79, + "learning_rate": 6.044352195761416e-07, + "loss": 0.9356, + "step": 40287 + }, + { + "epoch": 8.79, + "learning_rate": 6.043259777146603e-07, + "loss": 1.2407, + "step": 40288 + }, + { + "epoch": 8.79, + "learning_rate": 6.04216735853179e-07, + "loss": 1.1622, + "step": 40289 + }, + { + "epoch": 8.79, + "learning_rate": 6.041074939916976e-07, + "loss": 0.9257, + "step": 40290 + }, + { + "epoch": 8.79, + "learning_rate": 6.039982521302164e-07, + "loss": 2.0342, + "step": 40291 + }, + { + "epoch": 8.79, + "learning_rate": 6.038890102687351e-07, + "loss": 1.4629, + "step": 40292 + }, + { + "epoch": 8.79, + "learning_rate": 6.037797684072537e-07, + "loss": 0.6357, + "step": 40293 + }, + { + "epoch": 8.79, + "learning_rate": 6.036705265457724e-07, + "loss": 0.8417, + "step": 40294 + }, + { + "epoch": 8.79, + "learning_rate": 6.035612846842911e-07, + "loss": 0.0386, + "step": 40295 + }, + { + "epoch": 8.79, + "learning_rate": 6.034520428228097e-07, + "loss": 0.8289, + "step": 40296 + }, + { + "epoch": 8.79, + "learning_rate": 6.033428009613284e-07, + "loss": 1.2076, + "step": 40297 + }, + { + "epoch": 8.79, + "learning_rate": 6.032335590998472e-07, + "loss": 0.8216, + "step": 40298 + }, + { + "epoch": 8.8, + "learning_rate": 6.031243172383658e-07, + "loss": 1.7567, + "step": 40299 + }, + { + "epoch": 8.8, + "learning_rate": 6.030150753768845e-07, + "loss": 0.334, + "step": 40300 + }, + { + "epoch": 8.8, + "learning_rate": 6.029058335154032e-07, + "loss": 0.9794, + "step": 40301 + }, + { + "epoch": 8.8, + "learning_rate": 6.027965916539218e-07, + "loss": 0.8536, + "step": 40302 + }, + { + "epoch": 8.8, + "learning_rate": 6.026873497924405e-07, + "loss": 2.5191, + "step": 40303 + }, + { + "epoch": 8.8, + "learning_rate": 6.025781079309593e-07, + "loss": 1.0083, + "step": 40304 + }, + { + "epoch": 8.8, + "learning_rate": 6.024688660694779e-07, + "loss": 0.9587, + "step": 40305 + }, + { + "epoch": 8.8, + "learning_rate": 6.023596242079966e-07, + "loss": 0.9948, + "step": 40306 + }, + { + "epoch": 8.8, + "learning_rate": 6.022503823465152e-07, + "loss": 1.6929, + "step": 40307 + }, + { + "epoch": 8.8, + "learning_rate": 6.021411404850339e-07, + "loss": 2.0006, + "step": 40308 + }, + { + "epoch": 8.8, + "learning_rate": 6.020318986235526e-07, + "loss": 1.0113, + "step": 40309 + }, + { + "epoch": 8.8, + "learning_rate": 6.019226567620712e-07, + "loss": 1.4715, + "step": 40310 + }, + { + "epoch": 8.8, + "learning_rate": 6.0181341490059e-07, + "loss": 1.3422, + "step": 40311 + }, + { + "epoch": 8.8, + "learning_rate": 6.017041730391087e-07, + "loss": 1.9063, + "step": 40312 + }, + { + "epoch": 8.8, + "learning_rate": 6.015949311776273e-07, + "loss": 0.5332, + "step": 40313 + }, + { + "epoch": 8.8, + "learning_rate": 6.01485689316146e-07, + "loss": 1.1958, + "step": 40314 + }, + { + "epoch": 8.8, + "learning_rate": 6.013764474546647e-07, + "loss": 0.7387, + "step": 40315 + }, + { + "epoch": 8.8, + "learning_rate": 6.012672055931833e-07, + "loss": 1.7663, + "step": 40316 + }, + { + "epoch": 8.8, + "learning_rate": 6.01157963731702e-07, + "loss": 1.1328, + "step": 40317 + }, + { + "epoch": 8.8, + "learning_rate": 6.010487218702208e-07, + "loss": 2.167, + "step": 40318 + }, + { + "epoch": 8.8, + "learning_rate": 6.009394800087394e-07, + "loss": 1.5819, + "step": 40319 + }, + { + "epoch": 8.8, + "learning_rate": 6.008302381472581e-07, + "loss": 1.1337, + "step": 40320 + }, + { + "epoch": 8.8, + "learning_rate": 6.007209962857768e-07, + "loss": 1.4113, + "step": 40321 + }, + { + "epoch": 8.8, + "learning_rate": 6.006117544242954e-07, + "loss": 1.3903, + "step": 40322 + }, + { + "epoch": 8.8, + "learning_rate": 6.005025125628141e-07, + "loss": 0.8462, + "step": 40323 + }, + { + "epoch": 8.8, + "learning_rate": 6.003932707013328e-07, + "loss": 1.2377, + "step": 40324 + }, + { + "epoch": 8.8, + "learning_rate": 6.002840288398515e-07, + "loss": 1.3732, + "step": 40325 + }, + { + "epoch": 8.8, + "learning_rate": 6.001747869783702e-07, + "loss": 1.2858, + "step": 40326 + }, + { + "epoch": 8.8, + "learning_rate": 6.000655451168889e-07, + "loss": 0.6112, + "step": 40327 + }, + { + "epoch": 8.8, + "learning_rate": 5.999563032554075e-07, + "loss": 1.2011, + "step": 40328 + }, + { + "epoch": 8.8, + "learning_rate": 5.998470613939262e-07, + "loss": 0.215, + "step": 40329 + }, + { + "epoch": 8.8, + "learning_rate": 5.997378195324448e-07, + "loss": 0.4796, + "step": 40330 + }, + { + "epoch": 8.8, + "learning_rate": 5.996285776709635e-07, + "loss": 0.071, + "step": 40331 + }, + { + "epoch": 8.8, + "learning_rate": 5.995193358094823e-07, + "loss": 1.9122, + "step": 40332 + }, + { + "epoch": 8.8, + "learning_rate": 5.994100939480009e-07, + "loss": 3.1918, + "step": 40333 + }, + { + "epoch": 8.8, + "learning_rate": 5.993008520865196e-07, + "loss": 1.3564, + "step": 40334 + }, + { + "epoch": 8.8, + "learning_rate": 5.991916102250383e-07, + "loss": 0.8806, + "step": 40335 + }, + { + "epoch": 8.8, + "learning_rate": 5.990823683635569e-07, + "loss": 0.7994, + "step": 40336 + }, + { + "epoch": 8.8, + "learning_rate": 5.989731265020756e-07, + "loss": 1.9904, + "step": 40337 + }, + { + "epoch": 8.8, + "learning_rate": 5.988638846405944e-07, + "loss": 1.2241, + "step": 40338 + }, + { + "epoch": 8.8, + "learning_rate": 5.98754642779113e-07, + "loss": 1.3603, + "step": 40339 + }, + { + "epoch": 8.8, + "learning_rate": 5.986454009176317e-07, + "loss": 0.7873, + "step": 40340 + }, + { + "epoch": 8.8, + "learning_rate": 5.985361590561504e-07, + "loss": 0.8554, + "step": 40341 + }, + { + "epoch": 8.8, + "learning_rate": 5.98426917194669e-07, + "loss": 1.2845, + "step": 40342 + }, + { + "epoch": 8.8, + "learning_rate": 5.983176753331877e-07, + "loss": 1.5305, + "step": 40343 + }, + { + "epoch": 8.8, + "learning_rate": 5.982084334717064e-07, + "loss": 0.966, + "step": 40344 + }, + { + "epoch": 8.81, + "learning_rate": 5.980991916102251e-07, + "loss": 1.3446, + "step": 40345 + }, + { + "epoch": 8.81, + "learning_rate": 5.979899497487438e-07, + "loss": 2.033, + "step": 40346 + }, + { + "epoch": 8.81, + "learning_rate": 5.978807078872625e-07, + "loss": 1.0705, + "step": 40347 + }, + { + "epoch": 8.81, + "learning_rate": 5.977714660257811e-07, + "loss": 0.422, + "step": 40348 + }, + { + "epoch": 8.81, + "learning_rate": 5.976622241642998e-07, + "loss": 1.0439, + "step": 40349 + }, + { + "epoch": 8.81, + "learning_rate": 5.975529823028184e-07, + "loss": 0.457, + "step": 40350 + }, + { + "epoch": 8.81, + "learning_rate": 5.974437404413371e-07, + "loss": 0.3467, + "step": 40351 + }, + { + "epoch": 8.81, + "learning_rate": 5.973344985798559e-07, + "loss": 1.0614, + "step": 40352 + }, + { + "epoch": 8.81, + "learning_rate": 5.972252567183745e-07, + "loss": 0.7603, + "step": 40353 + }, + { + "epoch": 8.81, + "learning_rate": 5.971160148568932e-07, + "loss": 1.6178, + "step": 40354 + }, + { + "epoch": 8.81, + "learning_rate": 5.970067729954119e-07, + "loss": 1.2798, + "step": 40355 + }, + { + "epoch": 8.81, + "learning_rate": 5.968975311339305e-07, + "loss": 1.0797, + "step": 40356 + }, + { + "epoch": 8.81, + "learning_rate": 5.967882892724492e-07, + "loss": 0.1726, + "step": 40357 + }, + { + "epoch": 8.81, + "learning_rate": 5.96679047410968e-07, + "loss": 0.8916, + "step": 40358 + }, + { + "epoch": 8.81, + "learning_rate": 5.965698055494866e-07, + "loss": 1.0873, + "step": 40359 + }, + { + "epoch": 8.81, + "learning_rate": 5.964605636880053e-07, + "loss": 1.6412, + "step": 40360 + }, + { + "epoch": 8.81, + "learning_rate": 5.96351321826524e-07, + "loss": 1.2929, + "step": 40361 + }, + { + "epoch": 8.81, + "learning_rate": 5.962420799650426e-07, + "loss": 0.1057, + "step": 40362 + }, + { + "epoch": 8.81, + "learning_rate": 5.961328381035613e-07, + "loss": 0.6369, + "step": 40363 + }, + { + "epoch": 8.81, + "learning_rate": 5.9602359624208e-07, + "loss": 1.0583, + "step": 40364 + }, + { + "epoch": 8.81, + "learning_rate": 5.959143543805986e-07, + "loss": 0.2042, + "step": 40365 + }, + { + "epoch": 8.81, + "learning_rate": 5.958051125191174e-07, + "loss": 0.5821, + "step": 40366 + }, + { + "epoch": 8.81, + "learning_rate": 5.956958706576361e-07, + "loss": 0.2475, + "step": 40367 + }, + { + "epoch": 8.81, + "learning_rate": 5.955866287961547e-07, + "loss": 0.4246, + "step": 40368 + }, + { + "epoch": 8.81, + "learning_rate": 5.954773869346734e-07, + "loss": 0.8535, + "step": 40369 + }, + { + "epoch": 8.81, + "learning_rate": 5.953681450731921e-07, + "loss": 0.8665, + "step": 40370 + }, + { + "epoch": 8.81, + "learning_rate": 5.952589032117107e-07, + "loss": 2.6202, + "step": 40371 + }, + { + "epoch": 8.81, + "learning_rate": 5.951496613502295e-07, + "loss": 0.9924, + "step": 40372 + }, + { + "epoch": 8.81, + "learning_rate": 5.950404194887481e-07, + "loss": 1.4208, + "step": 40373 + }, + { + "epoch": 8.81, + "learning_rate": 5.949311776272668e-07, + "loss": 1.8616, + "step": 40374 + }, + { + "epoch": 8.81, + "learning_rate": 5.948219357657855e-07, + "loss": 0.7311, + "step": 40375 + }, + { + "epoch": 8.81, + "learning_rate": 5.947126939043041e-07, + "loss": 0.4067, + "step": 40376 + }, + { + "epoch": 8.81, + "learning_rate": 5.946034520428229e-07, + "loss": 0.632, + "step": 40377 + }, + { + "epoch": 8.81, + "learning_rate": 5.944942101813415e-07, + "loss": 2.1069, + "step": 40378 + }, + { + "epoch": 8.81, + "learning_rate": 5.943849683198602e-07, + "loss": 1.2934, + "step": 40379 + }, + { + "epoch": 8.81, + "learning_rate": 5.94275726458379e-07, + "loss": 2.2299, + "step": 40380 + }, + { + "epoch": 8.81, + "learning_rate": 5.941664845968976e-07, + "loss": 2.4902, + "step": 40381 + }, + { + "epoch": 8.81, + "learning_rate": 5.940572427354162e-07, + "loss": 1.5144, + "step": 40382 + }, + { + "epoch": 8.81, + "learning_rate": 5.939480008739349e-07, + "loss": 1.1231, + "step": 40383 + }, + { + "epoch": 8.81, + "learning_rate": 5.938387590124536e-07, + "loss": 1.2603, + "step": 40384 + }, + { + "epoch": 8.81, + "learning_rate": 5.937295171509722e-07, + "loss": 0.9693, + "step": 40385 + }, + { + "epoch": 8.81, + "learning_rate": 5.93620275289491e-07, + "loss": 1.1777, + "step": 40386 + }, + { + "epoch": 8.81, + "learning_rate": 5.935110334280097e-07, + "loss": 1.7034, + "step": 40387 + }, + { + "epoch": 8.81, + "learning_rate": 5.934017915665283e-07, + "loss": 0.7644, + "step": 40388 + }, + { + "epoch": 8.81, + "learning_rate": 5.93292549705047e-07, + "loss": 0.4006, + "step": 40389 + }, + { + "epoch": 8.81, + "learning_rate": 5.931833078435657e-07, + "loss": 1.8928, + "step": 40390 + }, + { + "epoch": 8.82, + "learning_rate": 5.930740659820844e-07, + "loss": 1.231, + "step": 40391 + }, + { + "epoch": 8.82, + "learning_rate": 5.92964824120603e-07, + "loss": 0.5216, + "step": 40392 + }, + { + "epoch": 8.82, + "learning_rate": 5.928555822591218e-07, + "loss": 1.0951, + "step": 40393 + }, + { + "epoch": 8.82, + "learning_rate": 5.927463403976405e-07, + "loss": 1.0596, + "step": 40394 + }, + { + "epoch": 8.82, + "learning_rate": 5.926370985361591e-07, + "loss": 4.3037, + "step": 40395 + }, + { + "epoch": 8.82, + "learning_rate": 5.925278566746777e-07, + "loss": 1.1222, + "step": 40396 + }, + { + "epoch": 8.82, + "learning_rate": 5.924186148131965e-07, + "loss": 0.3148, + "step": 40397 + }, + { + "epoch": 8.82, + "learning_rate": 5.923093729517151e-07, + "loss": 0.318, + "step": 40398 + }, + { + "epoch": 8.82, + "learning_rate": 5.922001310902338e-07, + "loss": 0.2591, + "step": 40399 + }, + { + "epoch": 8.82, + "learning_rate": 5.920908892287526e-07, + "loss": 0.7606, + "step": 40400 + }, + { + "epoch": 8.82, + "learning_rate": 5.919816473672712e-07, + "loss": 0.5579, + "step": 40401 + }, + { + "epoch": 8.82, + "learning_rate": 5.918724055057899e-07, + "loss": 0.5169, + "step": 40402 + }, + { + "epoch": 8.82, + "learning_rate": 5.917631636443086e-07, + "loss": 0.6679, + "step": 40403 + }, + { + "epoch": 8.82, + "learning_rate": 5.916539217828272e-07, + "loss": 1.2954, + "step": 40404 + }, + { + "epoch": 8.82, + "learning_rate": 5.91544679921346e-07, + "loss": 1.2054, + "step": 40405 + }, + { + "epoch": 8.82, + "learning_rate": 5.914354380598646e-07, + "loss": 1.3073, + "step": 40406 + }, + { + "epoch": 8.82, + "learning_rate": 5.913261961983833e-07, + "loss": 0.7509, + "step": 40407 + }, + { + "epoch": 8.82, + "learning_rate": 5.91216954336902e-07, + "loss": 1.6018, + "step": 40408 + }, + { + "epoch": 8.82, + "learning_rate": 5.911077124754206e-07, + "loss": 0.2217, + "step": 40409 + }, + { + "epoch": 8.82, + "learning_rate": 5.909984706139393e-07, + "loss": 1.3692, + "step": 40410 + }, + { + "epoch": 8.82, + "learning_rate": 5.90889228752458e-07, + "loss": 1.8709, + "step": 40411 + }, + { + "epoch": 8.82, + "learning_rate": 5.907799868909766e-07, + "loss": 0.1897, + "step": 40412 + }, + { + "epoch": 8.82, + "learning_rate": 5.906707450294954e-07, + "loss": 0.5277, + "step": 40413 + }, + { + "epoch": 8.82, + "learning_rate": 5.905615031680141e-07, + "loss": 1.2692, + "step": 40414 + }, + { + "epoch": 8.82, + "learning_rate": 5.904522613065327e-07, + "loss": 1.6254, + "step": 40415 + }, + { + "epoch": 8.82, + "learning_rate": 5.903430194450514e-07, + "loss": 1.1948, + "step": 40416 + }, + { + "epoch": 8.82, + "learning_rate": 5.902337775835701e-07, + "loss": 0.6429, + "step": 40417 + }, + { + "epoch": 8.82, + "learning_rate": 5.901245357220887e-07, + "loss": 0.9394, + "step": 40418 + }, + { + "epoch": 8.82, + "learning_rate": 5.900152938606075e-07, + "loss": 3.2364, + "step": 40419 + }, + { + "epoch": 8.82, + "learning_rate": 5.899060519991262e-07, + "loss": 1.0899, + "step": 40420 + }, + { + "epoch": 8.82, + "learning_rate": 5.897968101376448e-07, + "loss": 0.4034, + "step": 40421 + }, + { + "epoch": 8.82, + "learning_rate": 5.896875682761635e-07, + "loss": 0.9064, + "step": 40422 + }, + { + "epoch": 8.82, + "learning_rate": 5.895783264146822e-07, + "loss": 0.6272, + "step": 40423 + }, + { + "epoch": 8.82, + "learning_rate": 5.894690845532008e-07, + "loss": 1.0116, + "step": 40424 + }, + { + "epoch": 8.82, + "learning_rate": 5.893598426917195e-07, + "loss": 0.7847, + "step": 40425 + }, + { + "epoch": 8.82, + "learning_rate": 5.892506008302382e-07, + "loss": 0.2246, + "step": 40426 + }, + { + "epoch": 8.82, + "learning_rate": 5.891413589687569e-07, + "loss": 1.1573, + "step": 40427 + }, + { + "epoch": 8.82, + "learning_rate": 5.890321171072756e-07, + "loss": 1.0849, + "step": 40428 + }, + { + "epoch": 8.82, + "learning_rate": 5.889228752457942e-07, + "loss": 0.957, + "step": 40429 + }, + { + "epoch": 8.82, + "learning_rate": 5.888136333843129e-07, + "loss": 3.082, + "step": 40430 + }, + { + "epoch": 8.82, + "learning_rate": 5.887043915228316e-07, + "loss": 1.1427, + "step": 40431 + }, + { + "epoch": 8.82, + "learning_rate": 5.885951496613502e-07, + "loss": 0.5629, + "step": 40432 + }, + { + "epoch": 8.82, + "learning_rate": 5.88485907799869e-07, + "loss": 0.6374, + "step": 40433 + }, + { + "epoch": 8.82, + "learning_rate": 5.883766659383877e-07, + "loss": 0.5014, + "step": 40434 + }, + { + "epoch": 8.82, + "learning_rate": 5.882674240769063e-07, + "loss": 0.5124, + "step": 40435 + }, + { + "epoch": 8.82, + "learning_rate": 5.88158182215425e-07, + "loss": 1.9568, + "step": 40436 + }, + { + "epoch": 8.83, + "learning_rate": 5.880489403539437e-07, + "loss": 2.0327, + "step": 40437 + }, + { + "epoch": 8.83, + "learning_rate": 5.879396984924623e-07, + "loss": 0.1569, + "step": 40438 + }, + { + "epoch": 8.83, + "learning_rate": 5.87830456630981e-07, + "loss": 1.5637, + "step": 40439 + }, + { + "epoch": 8.83, + "learning_rate": 5.877212147694998e-07, + "loss": 1.0614, + "step": 40440 + }, + { + "epoch": 8.83, + "learning_rate": 5.876119729080184e-07, + "loss": 0.5664, + "step": 40441 + }, + { + "epoch": 8.83, + "learning_rate": 5.875027310465371e-07, + "loss": 1.1501, + "step": 40442 + }, + { + "epoch": 8.83, + "learning_rate": 5.873934891850558e-07, + "loss": 0.8615, + "step": 40443 + }, + { + "epoch": 8.83, + "learning_rate": 5.872842473235744e-07, + "loss": 0.6443, + "step": 40444 + }, + { + "epoch": 8.83, + "learning_rate": 5.871750054620931e-07, + "loss": 0.8557, + "step": 40445 + }, + { + "epoch": 8.83, + "learning_rate": 5.870657636006119e-07, + "loss": 1.5674, + "step": 40446 + }, + { + "epoch": 8.83, + "learning_rate": 5.869565217391305e-07, + "loss": 1.3442, + "step": 40447 + }, + { + "epoch": 8.83, + "learning_rate": 5.868472798776492e-07, + "loss": 0.9781, + "step": 40448 + }, + { + "epoch": 8.83, + "learning_rate": 5.867380380161678e-07, + "loss": 0.6343, + "step": 40449 + }, + { + "epoch": 8.83, + "learning_rate": 5.866287961546865e-07, + "loss": 1.9138, + "step": 40450 + }, + { + "epoch": 8.83, + "learning_rate": 5.865195542932052e-07, + "loss": 0.2368, + "step": 40451 + }, + { + "epoch": 8.83, + "learning_rate": 5.864103124317238e-07, + "loss": 0.9865, + "step": 40452 + }, + { + "epoch": 8.83, + "learning_rate": 5.863010705702426e-07, + "loss": 0.5055, + "step": 40453 + }, + { + "epoch": 8.83, + "learning_rate": 5.861918287087613e-07, + "loss": 1.2138, + "step": 40454 + }, + { + "epoch": 8.83, + "learning_rate": 5.860825868472799e-07, + "loss": 0.5198, + "step": 40455 + }, + { + "epoch": 8.83, + "learning_rate": 5.859733449857986e-07, + "loss": 1.308, + "step": 40456 + }, + { + "epoch": 8.83, + "learning_rate": 5.858641031243173e-07, + "loss": 0.8723, + "step": 40457 + }, + { + "epoch": 8.83, + "learning_rate": 5.857548612628359e-07, + "loss": 2.0182, + "step": 40458 + }, + { + "epoch": 8.83, + "learning_rate": 5.856456194013546e-07, + "loss": 0.6524, + "step": 40459 + }, + { + "epoch": 8.83, + "learning_rate": 5.855363775398734e-07, + "loss": 1.1195, + "step": 40460 + }, + { + "epoch": 8.83, + "learning_rate": 5.85427135678392e-07, + "loss": 0.1829, + "step": 40461 + }, + { + "epoch": 8.83, + "learning_rate": 5.853178938169107e-07, + "loss": 0.5495, + "step": 40462 + }, + { + "epoch": 8.83, + "learning_rate": 5.852086519554294e-07, + "loss": 1.6371, + "step": 40463 + }, + { + "epoch": 8.83, + "learning_rate": 5.85099410093948e-07, + "loss": 2.6368, + "step": 40464 + }, + { + "epoch": 8.83, + "learning_rate": 5.849901682324667e-07, + "loss": 1.3658, + "step": 40465 + }, + { + "epoch": 8.83, + "learning_rate": 5.848809263709855e-07, + "loss": 0.146, + "step": 40466 + }, + { + "epoch": 8.83, + "learning_rate": 5.847716845095041e-07, + "loss": 0.7597, + "step": 40467 + }, + { + "epoch": 8.83, + "learning_rate": 5.846624426480228e-07, + "loss": 0.0505, + "step": 40468 + }, + { + "epoch": 8.83, + "learning_rate": 5.845532007865414e-07, + "loss": 0.5479, + "step": 40469 + }, + { + "epoch": 8.83, + "learning_rate": 5.844439589250601e-07, + "loss": 1.3584, + "step": 40470 + }, + { + "epoch": 8.83, + "learning_rate": 5.843347170635788e-07, + "loss": 1.4396, + "step": 40471 + }, + { + "epoch": 8.83, + "learning_rate": 5.842254752020974e-07, + "loss": 1.014, + "step": 40472 + }, + { + "epoch": 8.83, + "learning_rate": 5.841162333406162e-07, + "loss": 0.7221, + "step": 40473 + }, + { + "epoch": 8.83, + "learning_rate": 5.840069914791349e-07, + "loss": 0.105, + "step": 40474 + }, + { + "epoch": 8.83, + "learning_rate": 5.838977496176535e-07, + "loss": 1.8154, + "step": 40475 + }, + { + "epoch": 8.83, + "learning_rate": 5.837885077561722e-07, + "loss": 1.7111, + "step": 40476 + }, + { + "epoch": 8.83, + "learning_rate": 5.836792658946909e-07, + "loss": 1.2684, + "step": 40477 + }, + { + "epoch": 8.83, + "learning_rate": 5.835700240332095e-07, + "loss": 1.452, + "step": 40478 + }, + { + "epoch": 8.83, + "learning_rate": 5.834607821717282e-07, + "loss": 1.6355, + "step": 40479 + }, + { + "epoch": 8.83, + "learning_rate": 5.83351540310247e-07, + "loss": 1.3685, + "step": 40480 + }, + { + "epoch": 8.83, + "learning_rate": 5.832422984487656e-07, + "loss": 2.0521, + "step": 40481 + }, + { + "epoch": 8.84, + "learning_rate": 5.831330565872843e-07, + "loss": 0.2065, + "step": 40482 + }, + { + "epoch": 8.84, + "learning_rate": 5.83023814725803e-07, + "loss": 0.9617, + "step": 40483 + }, + { + "epoch": 8.84, + "learning_rate": 5.829145728643216e-07, + "loss": 1.4911, + "step": 40484 + }, + { + "epoch": 8.84, + "learning_rate": 5.828053310028403e-07, + "loss": 0.5088, + "step": 40485 + }, + { + "epoch": 8.84, + "learning_rate": 5.82696089141359e-07, + "loss": 0.5841, + "step": 40486 + }, + { + "epoch": 8.84, + "learning_rate": 5.825868472798777e-07, + "loss": 0.8863, + "step": 40487 + }, + { + "epoch": 8.84, + "learning_rate": 5.824776054183964e-07, + "loss": 0.4631, + "step": 40488 + }, + { + "epoch": 8.84, + "learning_rate": 5.823683635569151e-07, + "loss": 0.3748, + "step": 40489 + }, + { + "epoch": 8.84, + "learning_rate": 5.822591216954337e-07, + "loss": 0.3593, + "step": 40490 + }, + { + "epoch": 8.84, + "learning_rate": 5.821498798339524e-07, + "loss": 0.6274, + "step": 40491 + }, + { + "epoch": 8.84, + "learning_rate": 5.82040637972471e-07, + "loss": 1.3831, + "step": 40492 + }, + { + "epoch": 8.84, + "learning_rate": 5.819313961109897e-07, + "loss": 3.954, + "step": 40493 + }, + { + "epoch": 8.84, + "learning_rate": 5.818221542495085e-07, + "loss": 1.3183, + "step": 40494 + }, + { + "epoch": 8.84, + "learning_rate": 5.817129123880271e-07, + "loss": 1.1479, + "step": 40495 + }, + { + "epoch": 8.84, + "learning_rate": 5.816036705265458e-07, + "loss": 1.2526, + "step": 40496 + }, + { + "epoch": 8.84, + "learning_rate": 5.814944286650645e-07, + "loss": 0.6469, + "step": 40497 + }, + { + "epoch": 8.84, + "learning_rate": 5.813851868035831e-07, + "loss": 0.8928, + "step": 40498 + }, + { + "epoch": 8.84, + "learning_rate": 5.812759449421018e-07, + "loss": 1.0605, + "step": 40499 + }, + { + "epoch": 8.84, + "learning_rate": 5.811667030806206e-07, + "loss": 1.2106, + "step": 40500 + }, + { + "epoch": 8.84, + "learning_rate": 5.810574612191392e-07, + "loss": 0.8811, + "step": 40501 + }, + { + "epoch": 8.84, + "learning_rate": 5.809482193576579e-07, + "loss": 1.2024, + "step": 40502 + }, + { + "epoch": 8.84, + "learning_rate": 5.808389774961766e-07, + "loss": 0.7659, + "step": 40503 + }, + { + "epoch": 8.84, + "learning_rate": 5.807297356346952e-07, + "loss": 0.9248, + "step": 40504 + }, + { + "epoch": 8.84, + "learning_rate": 5.806204937732139e-07, + "loss": 0.3681, + "step": 40505 + }, + { + "epoch": 8.84, + "learning_rate": 5.805112519117326e-07, + "loss": 2.019, + "step": 40506 + }, + { + "epoch": 8.84, + "learning_rate": 5.804020100502513e-07, + "loss": 0.1418, + "step": 40507 + }, + { + "epoch": 8.84, + "learning_rate": 5.8029276818877e-07, + "loss": 1.0055, + "step": 40508 + }, + { + "epoch": 8.84, + "learning_rate": 5.801835263272887e-07, + "loss": 1.2606, + "step": 40509 + }, + { + "epoch": 8.84, + "learning_rate": 5.800742844658073e-07, + "loss": 0.8239, + "step": 40510 + }, + { + "epoch": 8.84, + "learning_rate": 5.79965042604326e-07, + "loss": 2.2053, + "step": 40511 + }, + { + "epoch": 8.84, + "learning_rate": 5.798558007428446e-07, + "loss": 0.467, + "step": 40512 + }, + { + "epoch": 8.84, + "learning_rate": 5.797465588813635e-07, + "loss": 0.0684, + "step": 40513 + }, + { + "epoch": 8.84, + "learning_rate": 5.796373170198821e-07, + "loss": 1.2713, + "step": 40514 + }, + { + "epoch": 8.84, + "learning_rate": 5.795280751584007e-07, + "loss": 0.3282, + "step": 40515 + }, + { + "epoch": 8.84, + "learning_rate": 5.794188332969195e-07, + "loss": 0.1435, + "step": 40516 + }, + { + "epoch": 8.84, + "learning_rate": 5.793095914354381e-07, + "loss": 0.9999, + "step": 40517 + }, + { + "epoch": 8.84, + "learning_rate": 5.792003495739567e-07, + "loss": 0.7909, + "step": 40518 + }, + { + "epoch": 8.84, + "learning_rate": 5.790911077124755e-07, + "loss": 0.4982, + "step": 40519 + }, + { + "epoch": 8.84, + "learning_rate": 5.789818658509942e-07, + "loss": 0.7748, + "step": 40520 + }, + { + "epoch": 8.84, + "learning_rate": 5.788726239895128e-07, + "loss": 0.5791, + "step": 40521 + }, + { + "epoch": 8.84, + "learning_rate": 5.787633821280316e-07, + "loss": 0.3918, + "step": 40522 + }, + { + "epoch": 8.84, + "learning_rate": 5.786541402665502e-07, + "loss": 2.8483, + "step": 40523 + }, + { + "epoch": 8.84, + "learning_rate": 5.785448984050689e-07, + "loss": 1.8804, + "step": 40524 + }, + { + "epoch": 8.84, + "learning_rate": 5.784356565435875e-07, + "loss": 1.0684, + "step": 40525 + }, + { + "epoch": 8.84, + "learning_rate": 5.783264146821062e-07, + "loss": 1.6726, + "step": 40526 + }, + { + "epoch": 8.84, + "learning_rate": 5.78217172820625e-07, + "loss": 1.6044, + "step": 40527 + }, + { + "epoch": 8.85, + "learning_rate": 5.781079309591436e-07, + "loss": 0.8029, + "step": 40528 + }, + { + "epoch": 8.85, + "learning_rate": 5.779986890976623e-07, + "loss": 1.8031, + "step": 40529 + }, + { + "epoch": 8.85, + "learning_rate": 5.77889447236181e-07, + "loss": 0.9952, + "step": 40530 + }, + { + "epoch": 8.85, + "learning_rate": 5.777802053746996e-07, + "loss": 0.6916, + "step": 40531 + }, + { + "epoch": 8.85, + "learning_rate": 5.776709635132183e-07, + "loss": 0.3725, + "step": 40532 + }, + { + "epoch": 8.85, + "learning_rate": 5.77561721651737e-07, + "loss": 1.0066, + "step": 40533 + }, + { + "epoch": 8.85, + "learning_rate": 5.774524797902557e-07, + "loss": 1.7871, + "step": 40534 + }, + { + "epoch": 8.85, + "learning_rate": 5.773432379287743e-07, + "loss": 1.6278, + "step": 40535 + }, + { + "epoch": 8.85, + "learning_rate": 5.772339960672931e-07, + "loss": 0.7145, + "step": 40536 + }, + { + "epoch": 8.85, + "learning_rate": 5.771247542058117e-07, + "loss": 0.1989, + "step": 40537 + }, + { + "epoch": 8.85, + "learning_rate": 5.770155123443304e-07, + "loss": 0.9623, + "step": 40538 + }, + { + "epoch": 8.85, + "learning_rate": 5.769062704828491e-07, + "loss": 0.6375, + "step": 40539 + }, + { + "epoch": 8.85, + "learning_rate": 5.767970286213677e-07, + "loss": 0.6654, + "step": 40540 + }, + { + "epoch": 8.85, + "learning_rate": 5.766877867598865e-07, + "loss": 1.1828, + "step": 40541 + }, + { + "epoch": 8.85, + "learning_rate": 5.765785448984052e-07, + "loss": 0.1876, + "step": 40542 + }, + { + "epoch": 8.85, + "learning_rate": 5.764693030369238e-07, + "loss": 0.8304, + "step": 40543 + }, + { + "epoch": 8.85, + "learning_rate": 5.763600611754425e-07, + "loss": 1.5906, + "step": 40544 + }, + { + "epoch": 8.85, + "learning_rate": 5.762508193139611e-07, + "loss": 1.239, + "step": 40545 + }, + { + "epoch": 8.85, + "learning_rate": 5.761415774524798e-07, + "loss": 1.1581, + "step": 40546 + }, + { + "epoch": 8.85, + "learning_rate": 5.760323355909986e-07, + "loss": 0.6938, + "step": 40547 + }, + { + "epoch": 8.85, + "learning_rate": 5.759230937295172e-07, + "loss": 2.1541, + "step": 40548 + }, + { + "epoch": 8.85, + "learning_rate": 5.758138518680359e-07, + "loss": 0.3408, + "step": 40549 + }, + { + "epoch": 8.85, + "learning_rate": 5.757046100065546e-07, + "loss": 1.6614, + "step": 40550 + }, + { + "epoch": 8.85, + "learning_rate": 5.755953681450732e-07, + "loss": 1.1402, + "step": 40551 + }, + { + "epoch": 8.85, + "learning_rate": 5.754861262835919e-07, + "loss": 1.8065, + "step": 40552 + }, + { + "epoch": 8.85, + "learning_rate": 5.753768844221106e-07, + "loss": 1.1802, + "step": 40553 + }, + { + "epoch": 8.85, + "learning_rate": 5.752676425606293e-07, + "loss": 1.4186, + "step": 40554 + }, + { + "epoch": 8.85, + "learning_rate": 5.75158400699148e-07, + "loss": 1.4281, + "step": 40555 + }, + { + "epoch": 8.85, + "learning_rate": 5.750491588376667e-07, + "loss": 0.5059, + "step": 40556 + }, + { + "epoch": 8.85, + "learning_rate": 5.749399169761853e-07, + "loss": 2.0661, + "step": 40557 + }, + { + "epoch": 8.85, + "learning_rate": 5.74830675114704e-07, + "loss": 1.6766, + "step": 40558 + }, + { + "epoch": 8.85, + "learning_rate": 5.747214332532227e-07, + "loss": 1.1079, + "step": 40559 + }, + { + "epoch": 8.85, + "learning_rate": 5.746121913917413e-07, + "loss": 1.3418, + "step": 40560 + }, + { + "epoch": 8.85, + "learning_rate": 5.745029495302601e-07, + "loss": 2.1256, + "step": 40561 + }, + { + "epoch": 8.85, + "learning_rate": 5.743937076687788e-07, + "loss": 0.137, + "step": 40562 + }, + { + "epoch": 8.85, + "learning_rate": 5.742844658072974e-07, + "loss": 1.0666, + "step": 40563 + }, + { + "epoch": 8.85, + "learning_rate": 5.741752239458161e-07, + "loss": 1.0363, + "step": 40564 + }, + { + "epoch": 8.85, + "learning_rate": 5.740659820843348e-07, + "loss": 0.9938, + "step": 40565 + }, + { + "epoch": 8.85, + "learning_rate": 5.739567402228534e-07, + "loss": 0.8505, + "step": 40566 + }, + { + "epoch": 8.85, + "learning_rate": 5.738474983613722e-07, + "loss": 0.6785, + "step": 40567 + }, + { + "epoch": 8.85, + "learning_rate": 5.737382564998908e-07, + "loss": 0.6735, + "step": 40568 + }, + { + "epoch": 8.85, + "learning_rate": 5.736290146384095e-07, + "loss": 2.112, + "step": 40569 + }, + { + "epoch": 8.85, + "learning_rate": 5.735197727769282e-07, + "loss": 1.9977, + "step": 40570 + }, + { + "epoch": 8.85, + "learning_rate": 5.734105309154468e-07, + "loss": 0.4955, + "step": 40571 + }, + { + "epoch": 8.85, + "learning_rate": 5.733012890539655e-07, + "loss": 0.0682, + "step": 40572 + }, + { + "epoch": 8.85, + "learning_rate": 5.731920471924842e-07, + "loss": 1.5106, + "step": 40573 + }, + { + "epoch": 8.86, + "learning_rate": 5.730828053310029e-07, + "loss": 0.7827, + "step": 40574 + }, + { + "epoch": 8.86, + "learning_rate": 5.729735634695216e-07, + "loss": 0.2551, + "step": 40575 + }, + { + "epoch": 8.86, + "learning_rate": 5.728643216080403e-07, + "loss": 1.8045, + "step": 40576 + }, + { + "epoch": 8.86, + "learning_rate": 5.727550797465589e-07, + "loss": 2.1488, + "step": 40577 + }, + { + "epoch": 8.86, + "learning_rate": 5.726458378850776e-07, + "loss": 0.5311, + "step": 40578 + }, + { + "epoch": 8.86, + "learning_rate": 5.725365960235963e-07, + "loss": 0.7802, + "step": 40579 + }, + { + "epoch": 8.86, + "learning_rate": 5.724273541621149e-07, + "loss": 0.7226, + "step": 40580 + }, + { + "epoch": 8.86, + "learning_rate": 5.723181123006337e-07, + "loss": 1.6282, + "step": 40581 + }, + { + "epoch": 8.86, + "learning_rate": 5.722088704391524e-07, + "loss": 1.1628, + "step": 40582 + }, + { + "epoch": 8.86, + "learning_rate": 5.72099628577671e-07, + "loss": 0.7098, + "step": 40583 + }, + { + "epoch": 8.86, + "learning_rate": 5.719903867161897e-07, + "loss": 0.4575, + "step": 40584 + }, + { + "epoch": 8.86, + "learning_rate": 5.718811448547084e-07, + "loss": 0.6769, + "step": 40585 + }, + { + "epoch": 8.86, + "learning_rate": 5.71771902993227e-07, + "loss": 1.0047, + "step": 40586 + }, + { + "epoch": 8.86, + "learning_rate": 5.716626611317457e-07, + "loss": 2.4252, + "step": 40587 + }, + { + "epoch": 8.86, + "learning_rate": 5.715534192702644e-07, + "loss": 1.3748, + "step": 40588 + }, + { + "epoch": 8.86, + "learning_rate": 5.714441774087831e-07, + "loss": 0.9426, + "step": 40589 + }, + { + "epoch": 8.86, + "learning_rate": 5.713349355473018e-07, + "loss": 1.5539, + "step": 40590 + }, + { + "epoch": 8.86, + "learning_rate": 5.712256936858204e-07, + "loss": 0.7404, + "step": 40591 + }, + { + "epoch": 8.86, + "learning_rate": 5.711164518243391e-07, + "loss": 0.1356, + "step": 40592 + }, + { + "epoch": 8.86, + "learning_rate": 5.710072099628578e-07, + "loss": 1.0184, + "step": 40593 + }, + { + "epoch": 8.86, + "learning_rate": 5.708979681013764e-07, + "loss": 2.4368, + "step": 40594 + }, + { + "epoch": 8.86, + "learning_rate": 5.707887262398952e-07, + "loss": 1.3165, + "step": 40595 + }, + { + "epoch": 8.86, + "learning_rate": 5.706794843784139e-07, + "loss": 1.0477, + "step": 40596 + }, + { + "epoch": 8.86, + "learning_rate": 5.705702425169325e-07, + "loss": 1.26, + "step": 40597 + }, + { + "epoch": 8.86, + "learning_rate": 5.704610006554512e-07, + "loss": 0.6336, + "step": 40598 + }, + { + "epoch": 8.86, + "learning_rate": 5.703517587939699e-07, + "loss": 2.0841, + "step": 40599 + }, + { + "epoch": 8.86, + "learning_rate": 5.702425169324885e-07, + "loss": 0.9333, + "step": 40600 + }, + { + "epoch": 8.86, + "learning_rate": 5.701332750710073e-07, + "loss": 2.1476, + "step": 40601 + }, + { + "epoch": 8.86, + "learning_rate": 5.70024033209526e-07, + "loss": 0.7127, + "step": 40602 + }, + { + "epoch": 8.86, + "learning_rate": 5.699147913480446e-07, + "loss": 1.5958, + "step": 40603 + }, + { + "epoch": 8.86, + "learning_rate": 5.698055494865633e-07, + "loss": 2.6161, + "step": 40604 + }, + { + "epoch": 8.86, + "learning_rate": 5.69696307625082e-07, + "loss": 0.6394, + "step": 40605 + }, + { + "epoch": 8.86, + "learning_rate": 5.695870657636006e-07, + "loss": 3.3989, + "step": 40606 + }, + { + "epoch": 8.86, + "learning_rate": 5.694778239021193e-07, + "loss": 1.2603, + "step": 40607 + }, + { + "epoch": 8.86, + "learning_rate": 5.693685820406381e-07, + "loss": 1.1607, + "step": 40608 + }, + { + "epoch": 8.86, + "learning_rate": 5.692593401791567e-07, + "loss": 2.0748, + "step": 40609 + }, + { + "epoch": 8.86, + "learning_rate": 5.691500983176754e-07, + "loss": 1.7279, + "step": 40610 + }, + { + "epoch": 8.86, + "learning_rate": 5.69040856456194e-07, + "loss": 0.6849, + "step": 40611 + }, + { + "epoch": 8.86, + "learning_rate": 5.689316145947127e-07, + "loss": 1.056, + "step": 40612 + }, + { + "epoch": 8.86, + "learning_rate": 5.688223727332314e-07, + "loss": 0.4269, + "step": 40613 + }, + { + "epoch": 8.86, + "learning_rate": 5.6871313087175e-07, + "loss": 1.03, + "step": 40614 + }, + { + "epoch": 8.86, + "learning_rate": 5.686038890102688e-07, + "loss": 0.4356, + "step": 40615 + }, + { + "epoch": 8.86, + "learning_rate": 5.684946471487875e-07, + "loss": 0.723, + "step": 40616 + }, + { + "epoch": 8.86, + "learning_rate": 5.683854052873061e-07, + "loss": 1.004, + "step": 40617 + }, + { + "epoch": 8.86, + "learning_rate": 5.682761634258248e-07, + "loss": 1.562, + "step": 40618 + }, + { + "epoch": 8.86, + "learning_rate": 5.681669215643435e-07, + "loss": 1.2087, + "step": 40619 + }, + { + "epoch": 8.87, + "learning_rate": 5.680576797028621e-07, + "loss": 0.6286, + "step": 40620 + }, + { + "epoch": 8.87, + "learning_rate": 5.679484378413809e-07, + "loss": 0.5784, + "step": 40621 + }, + { + "epoch": 8.87, + "learning_rate": 5.678391959798996e-07, + "loss": 0.9135, + "step": 40622 + }, + { + "epoch": 8.87, + "learning_rate": 5.677299541184182e-07, + "loss": 0.7424, + "step": 40623 + }, + { + "epoch": 8.87, + "learning_rate": 5.676207122569369e-07, + "loss": 1.0495, + "step": 40624 + }, + { + "epoch": 8.87, + "learning_rate": 5.675114703954556e-07, + "loss": 0.6437, + "step": 40625 + }, + { + "epoch": 8.87, + "learning_rate": 5.674022285339742e-07, + "loss": 3.7725, + "step": 40626 + }, + { + "epoch": 8.87, + "learning_rate": 5.672929866724929e-07, + "loss": 0.8449, + "step": 40627 + }, + { + "epoch": 8.87, + "learning_rate": 5.671837448110117e-07, + "loss": 0.6378, + "step": 40628 + }, + { + "epoch": 8.87, + "learning_rate": 5.670745029495303e-07, + "loss": 0.5172, + "step": 40629 + }, + { + "epoch": 8.87, + "learning_rate": 5.66965261088049e-07, + "loss": 0.5065, + "step": 40630 + }, + { + "epoch": 8.87, + "learning_rate": 5.668560192265676e-07, + "loss": 1.0582, + "step": 40631 + }, + { + "epoch": 8.87, + "learning_rate": 5.667467773650863e-07, + "loss": 0.8722, + "step": 40632 + }, + { + "epoch": 8.87, + "learning_rate": 5.66637535503605e-07, + "loss": 0.4662, + "step": 40633 + }, + { + "epoch": 8.87, + "learning_rate": 5.665282936421236e-07, + "loss": 1.3614, + "step": 40634 + }, + { + "epoch": 8.87, + "learning_rate": 5.664190517806424e-07, + "loss": 0.2942, + "step": 40635 + }, + { + "epoch": 8.87, + "learning_rate": 5.663098099191611e-07, + "loss": 2.8421, + "step": 40636 + }, + { + "epoch": 8.87, + "learning_rate": 5.662005680576797e-07, + "loss": 1.1671, + "step": 40637 + }, + { + "epoch": 8.87, + "learning_rate": 5.660913261961985e-07, + "loss": 0.8318, + "step": 40638 + }, + { + "epoch": 8.87, + "learning_rate": 5.659820843347171e-07, + "loss": 0.481, + "step": 40639 + }, + { + "epoch": 8.87, + "learning_rate": 5.658728424732357e-07, + "loss": 1.2519, + "step": 40640 + }, + { + "epoch": 8.87, + "learning_rate": 5.657636006117546e-07, + "loss": 0.9558, + "step": 40641 + }, + { + "epoch": 8.87, + "learning_rate": 5.656543587502732e-07, + "loss": 1.914, + "step": 40642 + }, + { + "epoch": 8.87, + "learning_rate": 5.655451168887918e-07, + "loss": 0.8929, + "step": 40643 + }, + { + "epoch": 8.87, + "learning_rate": 5.654358750273105e-07, + "loss": 2.5139, + "step": 40644 + }, + { + "epoch": 8.87, + "learning_rate": 5.653266331658292e-07, + "loss": 0.6082, + "step": 40645 + }, + { + "epoch": 8.87, + "learning_rate": 5.652173913043478e-07, + "loss": 2.3919, + "step": 40646 + }, + { + "epoch": 8.87, + "learning_rate": 5.651081494428665e-07, + "loss": 1.2805, + "step": 40647 + }, + { + "epoch": 8.87, + "learning_rate": 5.649989075813853e-07, + "loss": 4.0082, + "step": 40648 + }, + { + "epoch": 8.87, + "learning_rate": 5.64889665719904e-07, + "loss": 0.743, + "step": 40649 + }, + { + "epoch": 8.87, + "learning_rate": 5.647804238584226e-07, + "loss": 0.7387, + "step": 40650 + }, + { + "epoch": 8.87, + "learning_rate": 5.646711819969413e-07, + "loss": 1.3262, + "step": 40651 + }, + { + "epoch": 8.87, + "learning_rate": 5.6456194013546e-07, + "loss": 0.5354, + "step": 40652 + }, + { + "epoch": 8.87, + "learning_rate": 5.644526982739786e-07, + "loss": 0.3564, + "step": 40653 + }, + { + "epoch": 8.87, + "learning_rate": 5.643434564124972e-07, + "loss": 1.0517, + "step": 40654 + }, + { + "epoch": 8.87, + "learning_rate": 5.642342145510161e-07, + "loss": 1.8058, + "step": 40655 + }, + { + "epoch": 8.87, + "learning_rate": 5.641249726895347e-07, + "loss": 1.1843, + "step": 40656 + }, + { + "epoch": 8.87, + "learning_rate": 5.640157308280533e-07, + "loss": 0.3985, + "step": 40657 + }, + { + "epoch": 8.87, + "learning_rate": 5.639064889665721e-07, + "loss": 1.9815, + "step": 40658 + }, + { + "epoch": 8.87, + "learning_rate": 5.637972471050907e-07, + "loss": 1.7222, + "step": 40659 + }, + { + "epoch": 8.87, + "learning_rate": 5.636880052436094e-07, + "loss": 0.5635, + "step": 40660 + }, + { + "epoch": 8.87, + "learning_rate": 5.635787633821281e-07, + "loss": 1.0495, + "step": 40661 + }, + { + "epoch": 8.87, + "learning_rate": 5.634695215206468e-07, + "loss": 1.3634, + "step": 40662 + }, + { + "epoch": 8.87, + "learning_rate": 5.633602796591655e-07, + "loss": 0.6434, + "step": 40663 + }, + { + "epoch": 8.87, + "learning_rate": 5.632510377976841e-07, + "loss": 0.9691, + "step": 40664 + }, + { + "epoch": 8.87, + "learning_rate": 5.631417959362028e-07, + "loss": 1.3854, + "step": 40665 + }, + { + "epoch": 8.88, + "learning_rate": 5.630325540747215e-07, + "loss": 0.4899, + "step": 40666 + }, + { + "epoch": 8.88, + "learning_rate": 5.629233122132401e-07, + "loss": 0.7007, + "step": 40667 + }, + { + "epoch": 8.88, + "learning_rate": 5.628140703517588e-07, + "loss": 0.1956, + "step": 40668 + }, + { + "epoch": 8.88, + "learning_rate": 5.627048284902776e-07, + "loss": 0.3826, + "step": 40669 + }, + { + "epoch": 8.88, + "learning_rate": 5.625955866287962e-07, + "loss": 1.716, + "step": 40670 + }, + { + "epoch": 8.88, + "learning_rate": 5.624863447673149e-07, + "loss": 1.1686, + "step": 40671 + }, + { + "epoch": 8.88, + "learning_rate": 5.623771029058336e-07, + "loss": 0.8997, + "step": 40672 + }, + { + "epoch": 8.88, + "learning_rate": 5.622678610443522e-07, + "loss": 1.014, + "step": 40673 + }, + { + "epoch": 8.88, + "learning_rate": 5.621586191828709e-07, + "loss": 1.8767, + "step": 40674 + }, + { + "epoch": 8.88, + "learning_rate": 5.620493773213897e-07, + "loss": 0.2514, + "step": 40675 + }, + { + "epoch": 8.88, + "learning_rate": 5.619401354599083e-07, + "loss": 0.7775, + "step": 40676 + }, + { + "epoch": 8.88, + "learning_rate": 5.61830893598427e-07, + "loss": 0.6714, + "step": 40677 + }, + { + "epoch": 8.88, + "learning_rate": 5.617216517369457e-07, + "loss": 0.5363, + "step": 40678 + }, + { + "epoch": 8.88, + "learning_rate": 5.616124098754643e-07, + "loss": 1.851, + "step": 40679 + }, + { + "epoch": 8.88, + "learning_rate": 5.61503168013983e-07, + "loss": 1.0684, + "step": 40680 + }, + { + "epoch": 8.88, + "learning_rate": 5.613939261525017e-07, + "loss": 0.5117, + "step": 40681 + }, + { + "epoch": 8.88, + "learning_rate": 5.612846842910204e-07, + "loss": 2.4023, + "step": 40682 + }, + { + "epoch": 8.88, + "learning_rate": 5.611754424295391e-07, + "loss": 0.7102, + "step": 40683 + }, + { + "epoch": 8.88, + "learning_rate": 5.610662005680578e-07, + "loss": 1.5823, + "step": 40684 + }, + { + "epoch": 8.88, + "learning_rate": 5.609569587065764e-07, + "loss": 0.1483, + "step": 40685 + }, + { + "epoch": 8.88, + "learning_rate": 5.608477168450951e-07, + "loss": 0.0983, + "step": 40686 + }, + { + "epoch": 8.88, + "learning_rate": 5.607384749836137e-07, + "loss": 1.9477, + "step": 40687 + }, + { + "epoch": 8.88, + "learning_rate": 5.606292331221324e-07, + "loss": 0.1423, + "step": 40688 + }, + { + "epoch": 8.88, + "learning_rate": 5.605199912606512e-07, + "loss": 0.6193, + "step": 40689 + }, + { + "epoch": 8.88, + "learning_rate": 5.604107493991698e-07, + "loss": 1.3271, + "step": 40690 + }, + { + "epoch": 8.88, + "learning_rate": 5.603015075376885e-07, + "loss": 0.9237, + "step": 40691 + }, + { + "epoch": 8.88, + "learning_rate": 5.601922656762072e-07, + "loss": 1.2326, + "step": 40692 + }, + { + "epoch": 8.88, + "learning_rate": 5.600830238147258e-07, + "loss": 1.9882, + "step": 40693 + }, + { + "epoch": 8.88, + "learning_rate": 5.599737819532445e-07, + "loss": 0.0936, + "step": 40694 + }, + { + "epoch": 8.88, + "learning_rate": 5.598645400917633e-07, + "loss": 0.5044, + "step": 40695 + }, + { + "epoch": 8.88, + "learning_rate": 5.597552982302819e-07, + "loss": 1.4012, + "step": 40696 + }, + { + "epoch": 8.88, + "learning_rate": 5.596460563688006e-07, + "loss": 0.6707, + "step": 40697 + }, + { + "epoch": 8.88, + "learning_rate": 5.595368145073193e-07, + "loss": 1.489, + "step": 40698 + }, + { + "epoch": 8.88, + "learning_rate": 5.594275726458379e-07, + "loss": 0.8902, + "step": 40699 + }, + { + "epoch": 8.88, + "learning_rate": 5.593183307843566e-07, + "loss": 1.0741, + "step": 40700 + }, + { + "epoch": 8.88, + "learning_rate": 5.592090889228753e-07, + "loss": 0.9651, + "step": 40701 + }, + { + "epoch": 8.88, + "learning_rate": 5.59099847061394e-07, + "loss": 0.3693, + "step": 40702 + }, + { + "epoch": 8.88, + "learning_rate": 5.589906051999127e-07, + "loss": 0.6768, + "step": 40703 + }, + { + "epoch": 8.88, + "learning_rate": 5.588813633384314e-07, + "loss": 1.6654, + "step": 40704 + }, + { + "epoch": 8.88, + "learning_rate": 5.5877212147695e-07, + "loss": 1.1433, + "step": 40705 + }, + { + "epoch": 8.88, + "learning_rate": 5.586628796154687e-07, + "loss": 2.5245, + "step": 40706 + }, + { + "epoch": 8.88, + "learning_rate": 5.585536377539873e-07, + "loss": 0.585, + "step": 40707 + }, + { + "epoch": 8.88, + "learning_rate": 5.58444395892506e-07, + "loss": 0.5362, + "step": 40708 + }, + { + "epoch": 8.88, + "learning_rate": 5.583351540310248e-07, + "loss": 0.5143, + "step": 40709 + }, + { + "epoch": 8.88, + "learning_rate": 5.582259121695434e-07, + "loss": 0.6437, + "step": 40710 + }, + { + "epoch": 8.88, + "learning_rate": 5.581166703080621e-07, + "loss": 0.6806, + "step": 40711 + }, + { + "epoch": 8.89, + "learning_rate": 5.580074284465808e-07, + "loss": 0.86, + "step": 40712 + }, + { + "epoch": 8.89, + "learning_rate": 5.578981865850994e-07, + "loss": 2.1027, + "step": 40713 + }, + { + "epoch": 8.89, + "learning_rate": 5.577889447236181e-07, + "loss": 1.7168, + "step": 40714 + }, + { + "epoch": 8.89, + "learning_rate": 5.576797028621368e-07, + "loss": 1.3932, + "step": 40715 + }, + { + "epoch": 8.89, + "learning_rate": 5.575704610006555e-07, + "loss": 0.4262, + "step": 40716 + }, + { + "epoch": 8.89, + "learning_rate": 5.574612191391742e-07, + "loss": 1.5201, + "step": 40717 + }, + { + "epoch": 8.89, + "learning_rate": 5.573519772776929e-07, + "loss": 1.5921, + "step": 40718 + }, + { + "epoch": 8.89, + "learning_rate": 5.572427354162115e-07, + "loss": 1.1227, + "step": 40719 + }, + { + "epoch": 8.89, + "learning_rate": 5.571334935547302e-07, + "loss": 0.5208, + "step": 40720 + }, + { + "epoch": 8.89, + "learning_rate": 5.570242516932489e-07, + "loss": 1.708, + "step": 40721 + }, + { + "epoch": 8.89, + "learning_rate": 5.569150098317675e-07, + "loss": 0.7294, + "step": 40722 + }, + { + "epoch": 8.89, + "learning_rate": 5.568057679702863e-07, + "loss": 2.8214, + "step": 40723 + }, + { + "epoch": 8.89, + "learning_rate": 5.56696526108805e-07, + "loss": 1.2903, + "step": 40724 + }, + { + "epoch": 8.89, + "learning_rate": 5.565872842473236e-07, + "loss": 1.6684, + "step": 40725 + }, + { + "epoch": 8.89, + "learning_rate": 5.564780423858423e-07, + "loss": 0.771, + "step": 40726 + }, + { + "epoch": 8.89, + "learning_rate": 5.56368800524361e-07, + "loss": 0.8447, + "step": 40727 + }, + { + "epoch": 8.89, + "learning_rate": 5.562595586628796e-07, + "loss": 0.7379, + "step": 40728 + }, + { + "epoch": 8.89, + "learning_rate": 5.561503168013984e-07, + "loss": 3.4497, + "step": 40729 + }, + { + "epoch": 8.89, + "learning_rate": 5.56041074939917e-07, + "loss": 0.4834, + "step": 40730 + }, + { + "epoch": 8.89, + "learning_rate": 5.559318330784357e-07, + "loss": 0.4457, + "step": 40731 + }, + { + "epoch": 8.89, + "learning_rate": 5.558225912169544e-07, + "loss": 1.3033, + "step": 40732 + }, + { + "epoch": 8.89, + "learning_rate": 5.55713349355473e-07, + "loss": 0.2591, + "step": 40733 + }, + { + "epoch": 8.89, + "learning_rate": 5.556041074939917e-07, + "loss": 1.7126, + "step": 40734 + }, + { + "epoch": 8.89, + "learning_rate": 5.554948656325104e-07, + "loss": 0.4595, + "step": 40735 + }, + { + "epoch": 8.89, + "learning_rate": 5.553856237710291e-07, + "loss": 0.6836, + "step": 40736 + }, + { + "epoch": 8.89, + "learning_rate": 5.552763819095478e-07, + "loss": 1.8576, + "step": 40737 + }, + { + "epoch": 8.89, + "learning_rate": 5.551671400480665e-07, + "loss": 0.6177, + "step": 40738 + }, + { + "epoch": 8.89, + "learning_rate": 5.550578981865851e-07, + "loss": 0.4665, + "step": 40739 + }, + { + "epoch": 8.89, + "learning_rate": 5.549486563251038e-07, + "loss": 0.963, + "step": 40740 + }, + { + "epoch": 8.89, + "learning_rate": 5.548394144636225e-07, + "loss": 0.0355, + "step": 40741 + }, + { + "epoch": 8.89, + "learning_rate": 5.547301726021411e-07, + "loss": 0.2784, + "step": 40742 + }, + { + "epoch": 8.89, + "learning_rate": 5.546209307406599e-07, + "loss": 1.2649, + "step": 40743 + }, + { + "epoch": 8.89, + "learning_rate": 5.545116888791786e-07, + "loss": 1.0573, + "step": 40744 + }, + { + "epoch": 8.89, + "learning_rate": 5.544024470176972e-07, + "loss": 0.8881, + "step": 40745 + }, + { + "epoch": 8.89, + "learning_rate": 5.542932051562159e-07, + "loss": 1.1394, + "step": 40746 + }, + { + "epoch": 8.89, + "learning_rate": 5.541839632947346e-07, + "loss": 2.1944, + "step": 40747 + }, + { + "epoch": 8.89, + "learning_rate": 5.540747214332532e-07, + "loss": 1.1395, + "step": 40748 + }, + { + "epoch": 8.89, + "learning_rate": 5.53965479571772e-07, + "loss": 0.7418, + "step": 40749 + }, + { + "epoch": 8.89, + "learning_rate": 5.538562377102906e-07, + "loss": 1.5026, + "step": 40750 + }, + { + "epoch": 8.89, + "learning_rate": 5.537469958488093e-07, + "loss": 1.3097, + "step": 40751 + }, + { + "epoch": 8.89, + "learning_rate": 5.53637753987328e-07, + "loss": 0.07, + "step": 40752 + }, + { + "epoch": 8.89, + "learning_rate": 5.535285121258466e-07, + "loss": 0.4399, + "step": 40753 + }, + { + "epoch": 8.89, + "learning_rate": 5.534192702643653e-07, + "loss": 1.2134, + "step": 40754 + }, + { + "epoch": 8.89, + "learning_rate": 5.53310028402884e-07, + "loss": 0.1071, + "step": 40755 + }, + { + "epoch": 8.89, + "learning_rate": 5.532007865414027e-07, + "loss": 1.1802, + "step": 40756 + }, + { + "epoch": 8.9, + "learning_rate": 5.530915446799214e-07, + "loss": 1.5082, + "step": 40757 + }, + { + "epoch": 8.9, + "learning_rate": 5.529823028184401e-07, + "loss": 2.0922, + "step": 40758 + }, + { + "epoch": 8.9, + "learning_rate": 5.528730609569587e-07, + "loss": 1.0778, + "step": 40759 + }, + { + "epoch": 8.9, + "learning_rate": 5.527638190954775e-07, + "loss": 1.7002, + "step": 40760 + }, + { + "epoch": 8.9, + "learning_rate": 5.526545772339961e-07, + "loss": 2.3296, + "step": 40761 + }, + { + "epoch": 8.9, + "learning_rate": 5.525453353725147e-07, + "loss": 0.7305, + "step": 40762 + }, + { + "epoch": 8.9, + "learning_rate": 5.524360935110335e-07, + "loss": 0.0551, + "step": 40763 + }, + { + "epoch": 8.9, + "learning_rate": 5.523268516495522e-07, + "loss": 1.2045, + "step": 40764 + }, + { + "epoch": 8.9, + "learning_rate": 5.522176097880708e-07, + "loss": 0.9553, + "step": 40765 + }, + { + "epoch": 8.9, + "learning_rate": 5.521083679265895e-07, + "loss": 1.0826, + "step": 40766 + }, + { + "epoch": 8.9, + "learning_rate": 5.519991260651082e-07, + "loss": 1.04, + "step": 40767 + }, + { + "epoch": 8.9, + "learning_rate": 5.518898842036268e-07, + "loss": 2.3947, + "step": 40768 + }, + { + "epoch": 8.9, + "learning_rate": 5.517806423421455e-07, + "loss": 0.0935, + "step": 40769 + }, + { + "epoch": 8.9, + "learning_rate": 5.516714004806643e-07, + "loss": 0.5405, + "step": 40770 + }, + { + "epoch": 8.9, + "learning_rate": 5.51562158619183e-07, + "loss": 3.0347, + "step": 40771 + }, + { + "epoch": 8.9, + "learning_rate": 5.514529167577016e-07, + "loss": 1.1973, + "step": 40772 + }, + { + "epoch": 8.9, + "learning_rate": 5.513436748962202e-07, + "loss": 0.7694, + "step": 40773 + }, + { + "epoch": 8.9, + "learning_rate": 5.51234433034739e-07, + "loss": 0.5148, + "step": 40774 + }, + { + "epoch": 8.9, + "learning_rate": 5.511251911732576e-07, + "loss": 0.699, + "step": 40775 + }, + { + "epoch": 8.9, + "learning_rate": 5.510159493117762e-07, + "loss": 0.1246, + "step": 40776 + }, + { + "epoch": 8.9, + "learning_rate": 5.509067074502951e-07, + "loss": 1.4569, + "step": 40777 + }, + { + "epoch": 8.9, + "learning_rate": 5.507974655888137e-07, + "loss": 0.6633, + "step": 40778 + }, + { + "epoch": 8.9, + "learning_rate": 5.506882237273323e-07, + "loss": 0.3566, + "step": 40779 + }, + { + "epoch": 8.9, + "learning_rate": 5.505789818658511e-07, + "loss": 0.7702, + "step": 40780 + }, + { + "epoch": 8.9, + "learning_rate": 5.504697400043697e-07, + "loss": 1.1005, + "step": 40781 + }, + { + "epoch": 8.9, + "learning_rate": 5.503604981428883e-07, + "loss": 1.265, + "step": 40782 + }, + { + "epoch": 8.9, + "learning_rate": 5.502512562814071e-07, + "loss": 1.095, + "step": 40783 + }, + { + "epoch": 8.9, + "learning_rate": 5.501420144199258e-07, + "loss": 0.2618, + "step": 40784 + }, + { + "epoch": 8.9, + "learning_rate": 5.500327725584445e-07, + "loss": 0.5275, + "step": 40785 + }, + { + "epoch": 8.9, + "learning_rate": 5.499235306969631e-07, + "loss": 1.0264, + "step": 40786 + }, + { + "epoch": 8.9, + "learning_rate": 5.498142888354818e-07, + "loss": 0.8344, + "step": 40787 + }, + { + "epoch": 8.9, + "learning_rate": 5.497050469740005e-07, + "loss": 2.1507, + "step": 40788 + }, + { + "epoch": 8.9, + "learning_rate": 5.495958051125191e-07, + "loss": 0.7237, + "step": 40789 + }, + { + "epoch": 8.9, + "learning_rate": 5.494865632510379e-07, + "loss": 0.7852, + "step": 40790 + }, + { + "epoch": 8.9, + "learning_rate": 5.493773213895566e-07, + "loss": 0.6768, + "step": 40791 + }, + { + "epoch": 8.9, + "learning_rate": 5.492680795280752e-07, + "loss": 1.1696, + "step": 40792 + }, + { + "epoch": 8.9, + "learning_rate": 5.491588376665938e-07, + "loss": 0.2327, + "step": 40793 + }, + { + "epoch": 8.9, + "learning_rate": 5.490495958051126e-07, + "loss": 0.6635, + "step": 40794 + }, + { + "epoch": 8.9, + "learning_rate": 5.489403539436312e-07, + "loss": 1.1931, + "step": 40795 + }, + { + "epoch": 8.9, + "learning_rate": 5.4883111208215e-07, + "loss": 0.0592, + "step": 40796 + }, + { + "epoch": 8.9, + "learning_rate": 5.487218702206687e-07, + "loss": 0.7391, + "step": 40797 + }, + { + "epoch": 8.9, + "learning_rate": 5.486126283591873e-07, + "loss": 1.3123, + "step": 40798 + }, + { + "epoch": 8.9, + "learning_rate": 5.48503386497706e-07, + "loss": 1.4888, + "step": 40799 + }, + { + "epoch": 8.9, + "learning_rate": 5.483941446362247e-07, + "loss": 0.7059, + "step": 40800 + }, + { + "epoch": 8.9, + "learning_rate": 5.482849027747433e-07, + "loss": 0.6311, + "step": 40801 + }, + { + "epoch": 8.9, + "learning_rate": 5.48175660913262e-07, + "loss": 1.5405, + "step": 40802 + }, + { + "epoch": 8.91, + "learning_rate": 5.480664190517808e-07, + "loss": 0.9509, + "step": 40803 + }, + { + "epoch": 8.91, + "learning_rate": 5.479571771902994e-07, + "loss": 0.2081, + "step": 40804 + }, + { + "epoch": 8.91, + "learning_rate": 5.478479353288181e-07, + "loss": 0.5219, + "step": 40805 + }, + { + "epoch": 8.91, + "learning_rate": 5.477386934673367e-07, + "loss": 0.5098, + "step": 40806 + }, + { + "epoch": 8.91, + "learning_rate": 5.476294516058554e-07, + "loss": 2.49, + "step": 40807 + }, + { + "epoch": 8.91, + "learning_rate": 5.475202097443741e-07, + "loss": 1.451, + "step": 40808 + }, + { + "epoch": 8.91, + "learning_rate": 5.474109678828927e-07, + "loss": 1.693, + "step": 40809 + }, + { + "epoch": 8.91, + "learning_rate": 5.473017260214115e-07, + "loss": 0.7077, + "step": 40810 + }, + { + "epoch": 8.91, + "learning_rate": 5.471924841599302e-07, + "loss": 0.4151, + "step": 40811 + }, + { + "epoch": 8.91, + "learning_rate": 5.470832422984488e-07, + "loss": 0.1797, + "step": 40812 + }, + { + "epoch": 8.91, + "learning_rate": 5.469740004369675e-07, + "loss": 1.5283, + "step": 40813 + }, + { + "epoch": 8.91, + "learning_rate": 5.468647585754862e-07, + "loss": 1.4922, + "step": 40814 + }, + { + "epoch": 8.91, + "learning_rate": 5.467555167140048e-07, + "loss": 0.475, + "step": 40815 + }, + { + "epoch": 8.91, + "learning_rate": 5.466462748525235e-07, + "loss": 0.969, + "step": 40816 + }, + { + "epoch": 8.91, + "learning_rate": 5.465370329910423e-07, + "loss": 1.4173, + "step": 40817 + }, + { + "epoch": 8.91, + "learning_rate": 5.464277911295609e-07, + "loss": 1.1323, + "step": 40818 + }, + { + "epoch": 8.91, + "learning_rate": 5.463185492680796e-07, + "loss": 1.6391, + "step": 40819 + }, + { + "epoch": 8.91, + "learning_rate": 5.462093074065983e-07, + "loss": 0.6356, + "step": 40820 + }, + { + "epoch": 8.91, + "learning_rate": 5.461000655451169e-07, + "loss": 0.9289, + "step": 40821 + }, + { + "epoch": 8.91, + "learning_rate": 5.459908236836356e-07, + "loss": 0.9709, + "step": 40822 + }, + { + "epoch": 8.91, + "learning_rate": 5.458815818221544e-07, + "loss": 0.8038, + "step": 40823 + }, + { + "epoch": 8.91, + "learning_rate": 5.45772339960673e-07, + "loss": 0.3651, + "step": 40824 + }, + { + "epoch": 8.91, + "learning_rate": 5.456630980991917e-07, + "loss": 0.4388, + "step": 40825 + }, + { + "epoch": 8.91, + "learning_rate": 5.455538562377103e-07, + "loss": 0.3328, + "step": 40826 + }, + { + "epoch": 8.91, + "learning_rate": 5.45444614376229e-07, + "loss": 0.8519, + "step": 40827 + }, + { + "epoch": 8.91, + "learning_rate": 5.453353725147477e-07, + "loss": 0.7846, + "step": 40828 + }, + { + "epoch": 8.91, + "learning_rate": 5.452261306532663e-07, + "loss": 0.1489, + "step": 40829 + }, + { + "epoch": 8.91, + "learning_rate": 5.45116888791785e-07, + "loss": 0.7994, + "step": 40830 + }, + { + "epoch": 8.91, + "learning_rate": 5.450076469303038e-07, + "loss": 1.0725, + "step": 40831 + }, + { + "epoch": 8.91, + "learning_rate": 5.448984050688224e-07, + "loss": 1.059, + "step": 40832 + }, + { + "epoch": 8.91, + "learning_rate": 5.447891632073411e-07, + "loss": 1.5194, + "step": 40833 + }, + { + "epoch": 8.91, + "learning_rate": 5.446799213458598e-07, + "loss": 0.942, + "step": 40834 + }, + { + "epoch": 8.91, + "learning_rate": 5.445706794843784e-07, + "loss": 0.8231, + "step": 40835 + }, + { + "epoch": 8.91, + "learning_rate": 5.444614376228971e-07, + "loss": 0.3419, + "step": 40836 + }, + { + "epoch": 8.91, + "learning_rate": 5.443521957614159e-07, + "loss": 0.8144, + "step": 40837 + }, + { + "epoch": 8.91, + "learning_rate": 5.442429538999345e-07, + "loss": 0.0896, + "step": 40838 + }, + { + "epoch": 8.91, + "learning_rate": 5.441337120384532e-07, + "loss": 0.6994, + "step": 40839 + }, + { + "epoch": 8.91, + "learning_rate": 5.440244701769719e-07, + "loss": 0.8959, + "step": 40840 + }, + { + "epoch": 8.91, + "learning_rate": 5.439152283154905e-07, + "loss": 1.2059, + "step": 40841 + }, + { + "epoch": 8.91, + "learning_rate": 5.438059864540092e-07, + "loss": 1.5954, + "step": 40842 + }, + { + "epoch": 8.91, + "learning_rate": 5.43696744592528e-07, + "loss": 2.5445, + "step": 40843 + }, + { + "epoch": 8.91, + "learning_rate": 5.435875027310466e-07, + "loss": 1.7954, + "step": 40844 + }, + { + "epoch": 8.91, + "learning_rate": 5.434782608695653e-07, + "loss": 2.6229, + "step": 40845 + }, + { + "epoch": 8.91, + "learning_rate": 5.43369019008084e-07, + "loss": 0.575, + "step": 40846 + }, + { + "epoch": 8.91, + "learning_rate": 5.432597771466026e-07, + "loss": 0.6791, + "step": 40847 + }, + { + "epoch": 8.91, + "learning_rate": 5.431505352851213e-07, + "loss": 2.537, + "step": 40848 + }, + { + "epoch": 8.92, + "learning_rate": 5.430412934236399e-07, + "loss": 1.4967, + "step": 40849 + }, + { + "epoch": 8.92, + "learning_rate": 5.429320515621586e-07, + "loss": 1.1493, + "step": 40850 + }, + { + "epoch": 8.92, + "learning_rate": 5.428228097006774e-07, + "loss": 1.8193, + "step": 40851 + }, + { + "epoch": 8.92, + "learning_rate": 5.42713567839196e-07, + "loss": 1.0607, + "step": 40852 + }, + { + "epoch": 8.92, + "learning_rate": 5.426043259777147e-07, + "loss": 0.3897, + "step": 40853 + }, + { + "epoch": 8.92, + "learning_rate": 5.424950841162334e-07, + "loss": 0.4304, + "step": 40854 + }, + { + "epoch": 8.92, + "learning_rate": 5.42385842254752e-07, + "loss": 1.726, + "step": 40855 + }, + { + "epoch": 8.92, + "learning_rate": 5.422766003932707e-07, + "loss": 1.7024, + "step": 40856 + }, + { + "epoch": 8.92, + "learning_rate": 5.421673585317895e-07, + "loss": 1.5685, + "step": 40857 + }, + { + "epoch": 8.92, + "learning_rate": 5.420581166703081e-07, + "loss": 0.9382, + "step": 40858 + }, + { + "epoch": 8.92, + "learning_rate": 5.419488748088268e-07, + "loss": 1.8788, + "step": 40859 + }, + { + "epoch": 8.92, + "learning_rate": 5.418396329473455e-07, + "loss": 0.7164, + "step": 40860 + }, + { + "epoch": 8.92, + "learning_rate": 5.417303910858641e-07, + "loss": 0.3671, + "step": 40861 + }, + { + "epoch": 8.92, + "learning_rate": 5.416211492243828e-07, + "loss": 1.1843, + "step": 40862 + }, + { + "epoch": 8.92, + "learning_rate": 5.415119073629015e-07, + "loss": 0.8351, + "step": 40863 + }, + { + "epoch": 8.92, + "learning_rate": 5.414026655014202e-07, + "loss": 0.3991, + "step": 40864 + }, + { + "epoch": 8.92, + "learning_rate": 5.412934236399389e-07, + "loss": 1.0429, + "step": 40865 + }, + { + "epoch": 8.92, + "learning_rate": 5.411841817784576e-07, + "loss": 0.9772, + "step": 40866 + }, + { + "epoch": 8.92, + "learning_rate": 5.410749399169762e-07, + "loss": 1.1621, + "step": 40867 + }, + { + "epoch": 8.92, + "learning_rate": 5.409656980554949e-07, + "loss": 2.3142, + "step": 40868 + }, + { + "epoch": 8.92, + "learning_rate": 5.408564561940135e-07, + "loss": 0.1635, + "step": 40869 + }, + { + "epoch": 8.92, + "learning_rate": 5.407472143325322e-07, + "loss": 1.422, + "step": 40870 + }, + { + "epoch": 8.92, + "learning_rate": 5.40637972471051e-07, + "loss": 2.7264, + "step": 40871 + }, + { + "epoch": 8.92, + "learning_rate": 5.405287306095696e-07, + "loss": 1.4703, + "step": 40872 + }, + { + "epoch": 8.92, + "learning_rate": 5.404194887480883e-07, + "loss": 0.5731, + "step": 40873 + }, + { + "epoch": 8.92, + "learning_rate": 5.40310246886607e-07, + "loss": 0.9316, + "step": 40874 + }, + { + "epoch": 8.92, + "learning_rate": 5.402010050251256e-07, + "loss": 0.8227, + "step": 40875 + }, + { + "epoch": 8.92, + "learning_rate": 5.400917631636443e-07, + "loss": 0.9568, + "step": 40876 + }, + { + "epoch": 8.92, + "learning_rate": 5.39982521302163e-07, + "loss": 1.9307, + "step": 40877 + }, + { + "epoch": 8.92, + "learning_rate": 5.398732794406817e-07, + "loss": 0.9977, + "step": 40878 + }, + { + "epoch": 8.92, + "learning_rate": 5.397640375792004e-07, + "loss": 1.4011, + "step": 40879 + }, + { + "epoch": 8.92, + "learning_rate": 5.396547957177191e-07, + "loss": 2.12, + "step": 40880 + }, + { + "epoch": 8.92, + "learning_rate": 5.395455538562377e-07, + "loss": 0.0854, + "step": 40881 + }, + { + "epoch": 8.92, + "learning_rate": 5.394363119947564e-07, + "loss": 1.6098, + "step": 40882 + }, + { + "epoch": 8.92, + "learning_rate": 5.393270701332751e-07, + "loss": 1.2357, + "step": 40883 + }, + { + "epoch": 8.92, + "learning_rate": 5.392178282717938e-07, + "loss": 0.271, + "step": 40884 + }, + { + "epoch": 8.92, + "learning_rate": 5.391085864103125e-07, + "loss": 1.0235, + "step": 40885 + }, + { + "epoch": 8.92, + "learning_rate": 5.389993445488312e-07, + "loss": 1.1753, + "step": 40886 + }, + { + "epoch": 8.92, + "learning_rate": 5.388901026873498e-07, + "loss": 1.5952, + "step": 40887 + }, + { + "epoch": 8.92, + "learning_rate": 5.387808608258685e-07, + "loss": 0.7586, + "step": 40888 + }, + { + "epoch": 8.92, + "learning_rate": 5.386716189643872e-07, + "loss": 1.0506, + "step": 40889 + }, + { + "epoch": 8.92, + "learning_rate": 5.385623771029058e-07, + "loss": 0.3245, + "step": 40890 + }, + { + "epoch": 8.92, + "learning_rate": 5.384531352414246e-07, + "loss": 0.7835, + "step": 40891 + }, + { + "epoch": 8.92, + "learning_rate": 5.383438933799432e-07, + "loss": 0.903, + "step": 40892 + }, + { + "epoch": 8.92, + "learning_rate": 5.382346515184619e-07, + "loss": 1.5481, + "step": 40893 + }, + { + "epoch": 8.92, + "learning_rate": 5.381254096569806e-07, + "loss": 0.8846, + "step": 40894 + }, + { + "epoch": 8.93, + "learning_rate": 5.380161677954992e-07, + "loss": 0.999, + "step": 40895 + }, + { + "epoch": 8.93, + "learning_rate": 5.37906925934018e-07, + "loss": 1.1988, + "step": 40896 + }, + { + "epoch": 8.93, + "learning_rate": 5.377976840725366e-07, + "loss": 0.8989, + "step": 40897 + }, + { + "epoch": 8.93, + "learning_rate": 5.376884422110553e-07, + "loss": 1.687, + "step": 40898 + }, + { + "epoch": 8.93, + "learning_rate": 5.375792003495741e-07, + "loss": 1.271, + "step": 40899 + }, + { + "epoch": 8.93, + "learning_rate": 5.374699584880927e-07, + "loss": 1.3178, + "step": 40900 + }, + { + "epoch": 8.93, + "learning_rate": 5.373607166266113e-07, + "loss": 1.1977, + "step": 40901 + }, + { + "epoch": 8.93, + "learning_rate": 5.3725147476513e-07, + "loss": 0.5423, + "step": 40902 + }, + { + "epoch": 8.93, + "learning_rate": 5.371422329036487e-07, + "loss": 1.1431, + "step": 40903 + }, + { + "epoch": 8.93, + "learning_rate": 5.370329910421673e-07, + "loss": 1.0657, + "step": 40904 + }, + { + "epoch": 8.93, + "learning_rate": 5.369237491806861e-07, + "loss": 0.4233, + "step": 40905 + }, + { + "epoch": 8.93, + "learning_rate": 5.368145073192048e-07, + "loss": 0.8696, + "step": 40906 + }, + { + "epoch": 8.93, + "learning_rate": 5.367052654577235e-07, + "loss": 0.186, + "step": 40907 + }, + { + "epoch": 8.93, + "learning_rate": 5.365960235962421e-07, + "loss": 1.4439, + "step": 40908 + }, + { + "epoch": 8.93, + "learning_rate": 5.364867817347608e-07, + "loss": 0.2106, + "step": 40909 + }, + { + "epoch": 8.93, + "learning_rate": 5.363775398732795e-07, + "loss": 0.7556, + "step": 40910 + }, + { + "epoch": 8.93, + "learning_rate": 5.362682980117982e-07, + "loss": 0.4288, + "step": 40911 + }, + { + "epoch": 8.93, + "learning_rate": 5.361590561503168e-07, + "loss": 0.4546, + "step": 40912 + }, + { + "epoch": 8.93, + "learning_rate": 5.360498142888356e-07, + "loss": 1.0289, + "step": 40913 + }, + { + "epoch": 8.93, + "learning_rate": 5.359405724273542e-07, + "loss": 0.9853, + "step": 40914 + }, + { + "epoch": 8.93, + "learning_rate": 5.358313305658728e-07, + "loss": 1.7624, + "step": 40915 + }, + { + "epoch": 8.93, + "learning_rate": 5.357220887043916e-07, + "loss": 2.4047, + "step": 40916 + }, + { + "epoch": 8.93, + "learning_rate": 5.356128468429102e-07, + "loss": 3.0878, + "step": 40917 + }, + { + "epoch": 8.93, + "learning_rate": 5.355036049814289e-07, + "loss": 0.8114, + "step": 40918 + }, + { + "epoch": 8.93, + "learning_rate": 5.353943631199477e-07, + "loss": 1.0414, + "step": 40919 + }, + { + "epoch": 8.93, + "learning_rate": 5.352851212584663e-07, + "loss": 0.1679, + "step": 40920 + }, + { + "epoch": 8.93, + "learning_rate": 5.35175879396985e-07, + "loss": 1.9896, + "step": 40921 + }, + { + "epoch": 8.93, + "learning_rate": 5.350666375355037e-07, + "loss": 1.7103, + "step": 40922 + }, + { + "epoch": 8.93, + "learning_rate": 5.349573956740223e-07, + "loss": 0.8505, + "step": 40923 + }, + { + "epoch": 8.93, + "learning_rate": 5.34848153812541e-07, + "loss": 1.4901, + "step": 40924 + }, + { + "epoch": 8.93, + "learning_rate": 5.347389119510597e-07, + "loss": 0.2941, + "step": 40925 + }, + { + "epoch": 8.93, + "learning_rate": 5.346296700895784e-07, + "loss": 0.1372, + "step": 40926 + }, + { + "epoch": 8.93, + "learning_rate": 5.345204282280971e-07, + "loss": 0.5613, + "step": 40927 + }, + { + "epoch": 8.93, + "learning_rate": 5.344111863666157e-07, + "loss": 1.8071, + "step": 40928 + }, + { + "epoch": 8.93, + "learning_rate": 5.343019445051344e-07, + "loss": 0.376, + "step": 40929 + }, + { + "epoch": 8.93, + "learning_rate": 5.341927026436531e-07, + "loss": 0.2545, + "step": 40930 + }, + { + "epoch": 8.93, + "learning_rate": 5.340834607821717e-07, + "loss": 0.8688, + "step": 40931 + }, + { + "epoch": 8.93, + "learning_rate": 5.339742189206905e-07, + "loss": 0.7473, + "step": 40932 + }, + { + "epoch": 8.93, + "learning_rate": 5.338649770592092e-07, + "loss": 0.6206, + "step": 40933 + }, + { + "epoch": 8.93, + "learning_rate": 5.337557351977278e-07, + "loss": 1.7498, + "step": 40934 + }, + { + "epoch": 8.93, + "learning_rate": 5.336464933362465e-07, + "loss": 0.2383, + "step": 40935 + }, + { + "epoch": 8.93, + "learning_rate": 5.335372514747652e-07, + "loss": 2.0632, + "step": 40936 + }, + { + "epoch": 8.93, + "learning_rate": 5.334280096132838e-07, + "loss": 1.4827, + "step": 40937 + }, + { + "epoch": 8.93, + "learning_rate": 5.333187677518026e-07, + "loss": 1.985, + "step": 40938 + }, + { + "epoch": 8.93, + "learning_rate": 5.332095258903213e-07, + "loss": 0.7074, + "step": 40939 + }, + { + "epoch": 8.93, + "learning_rate": 5.331002840288399e-07, + "loss": 0.6343, + "step": 40940 + }, + { + "epoch": 8.94, + "learning_rate": 5.329910421673586e-07, + "loss": 0.8064, + "step": 40941 + }, + { + "epoch": 8.94, + "learning_rate": 5.328818003058773e-07, + "loss": 0.9057, + "step": 40942 + }, + { + "epoch": 8.94, + "learning_rate": 5.327725584443959e-07, + "loss": 1.8734, + "step": 40943 + }, + { + "epoch": 8.94, + "learning_rate": 5.326633165829146e-07, + "loss": 0.8758, + "step": 40944 + }, + { + "epoch": 8.94, + "learning_rate": 5.325540747214333e-07, + "loss": 0.5278, + "step": 40945 + }, + { + "epoch": 8.94, + "learning_rate": 5.32444832859952e-07, + "loss": 1.8122, + "step": 40946 + }, + { + "epoch": 8.94, + "learning_rate": 5.323355909984707e-07, + "loss": 2.023, + "step": 40947 + }, + { + "epoch": 8.94, + "learning_rate": 5.322263491369893e-07, + "loss": 1.6983, + "step": 40948 + }, + { + "epoch": 8.94, + "learning_rate": 5.32117107275508e-07, + "loss": 0.4091, + "step": 40949 + }, + { + "epoch": 8.94, + "learning_rate": 5.320078654140267e-07, + "loss": 1.0, + "step": 40950 + }, + { + "epoch": 8.94, + "learning_rate": 5.318986235525453e-07, + "loss": 0.2635, + "step": 40951 + }, + { + "epoch": 8.94, + "learning_rate": 5.317893816910641e-07, + "loss": 1.2906, + "step": 40952 + }, + { + "epoch": 8.94, + "learning_rate": 5.316801398295828e-07, + "loss": 0.9535, + "step": 40953 + }, + { + "epoch": 8.94, + "learning_rate": 5.315708979681014e-07, + "loss": 2.6671, + "step": 40954 + }, + { + "epoch": 8.94, + "learning_rate": 5.314616561066201e-07, + "loss": 0.5595, + "step": 40955 + }, + { + "epoch": 8.94, + "learning_rate": 5.313524142451388e-07, + "loss": 0.1816, + "step": 40956 + }, + { + "epoch": 8.94, + "learning_rate": 5.312431723836574e-07, + "loss": 0.1607, + "step": 40957 + }, + { + "epoch": 8.94, + "learning_rate": 5.311339305221762e-07, + "loss": 0.6998, + "step": 40958 + }, + { + "epoch": 8.94, + "learning_rate": 5.310246886606949e-07, + "loss": 1.5089, + "step": 40959 + }, + { + "epoch": 8.94, + "learning_rate": 5.309154467992135e-07, + "loss": 1.2974, + "step": 40960 + }, + { + "epoch": 8.94, + "learning_rate": 5.308062049377322e-07, + "loss": 1.7482, + "step": 40961 + }, + { + "epoch": 8.94, + "learning_rate": 5.306969630762509e-07, + "loss": 0.3164, + "step": 40962 + }, + { + "epoch": 8.94, + "learning_rate": 5.305877212147695e-07, + "loss": 0.848, + "step": 40963 + }, + { + "epoch": 8.94, + "learning_rate": 5.304784793532882e-07, + "loss": 1.4383, + "step": 40964 + }, + { + "epoch": 8.94, + "learning_rate": 5.30369237491807e-07, + "loss": 0.5639, + "step": 40965 + }, + { + "epoch": 8.94, + "learning_rate": 5.302599956303256e-07, + "loss": 0.6681, + "step": 40966 + }, + { + "epoch": 8.94, + "learning_rate": 5.301507537688443e-07, + "loss": 2.1166, + "step": 40967 + }, + { + "epoch": 8.94, + "learning_rate": 5.300415119073629e-07, + "loss": 0.0497, + "step": 40968 + }, + { + "epoch": 8.94, + "learning_rate": 5.299322700458816e-07, + "loss": 1.5521, + "step": 40969 + }, + { + "epoch": 8.94, + "learning_rate": 5.298230281844003e-07, + "loss": 1.9059, + "step": 40970 + }, + { + "epoch": 8.94, + "learning_rate": 5.297137863229189e-07, + "loss": 1.215, + "step": 40971 + }, + { + "epoch": 8.94, + "learning_rate": 5.296045444614377e-07, + "loss": 1.8979, + "step": 40972 + }, + { + "epoch": 8.94, + "learning_rate": 5.294953025999564e-07, + "loss": 1.4844, + "step": 40973 + }, + { + "epoch": 8.94, + "learning_rate": 5.29386060738475e-07, + "loss": 0.168, + "step": 40974 + }, + { + "epoch": 8.94, + "learning_rate": 5.292768188769937e-07, + "loss": 1.0628, + "step": 40975 + }, + { + "epoch": 8.94, + "learning_rate": 5.291675770155124e-07, + "loss": 1.1415, + "step": 40976 + }, + { + "epoch": 8.94, + "learning_rate": 5.29058335154031e-07, + "loss": 0.4154, + "step": 40977 + }, + { + "epoch": 8.94, + "learning_rate": 5.289490932925497e-07, + "loss": 0.0777, + "step": 40978 + }, + { + "epoch": 8.94, + "learning_rate": 5.288398514310685e-07, + "loss": 1.1421, + "step": 40979 + }, + { + "epoch": 8.94, + "learning_rate": 5.287306095695871e-07, + "loss": 0.8618, + "step": 40980 + }, + { + "epoch": 8.94, + "learning_rate": 5.286213677081058e-07, + "loss": 0.542, + "step": 40981 + }, + { + "epoch": 8.94, + "learning_rate": 5.285121258466245e-07, + "loss": 0.8311, + "step": 40982 + }, + { + "epoch": 8.94, + "learning_rate": 5.284028839851431e-07, + "loss": 2.2946, + "step": 40983 + }, + { + "epoch": 8.94, + "learning_rate": 5.282936421236618e-07, + "loss": 1.2526, + "step": 40984 + }, + { + "epoch": 8.94, + "learning_rate": 5.281844002621806e-07, + "loss": 0.8082, + "step": 40985 + }, + { + "epoch": 8.95, + "learning_rate": 5.280751584006992e-07, + "loss": 0.5957, + "step": 40986 + }, + { + "epoch": 8.95, + "learning_rate": 5.279659165392179e-07, + "loss": 0.1569, + "step": 40987 + }, + { + "epoch": 8.95, + "learning_rate": 5.278566746777365e-07, + "loss": 0.8549, + "step": 40988 + }, + { + "epoch": 8.95, + "learning_rate": 5.277474328162552e-07, + "loss": 1.2333, + "step": 40989 + }, + { + "epoch": 8.95, + "learning_rate": 5.276381909547739e-07, + "loss": 0.3842, + "step": 40990 + }, + { + "epoch": 8.95, + "learning_rate": 5.275289490932925e-07, + "loss": 1.1963, + "step": 40991 + }, + { + "epoch": 8.95, + "learning_rate": 5.274197072318113e-07, + "loss": 2.5188, + "step": 40992 + }, + { + "epoch": 8.95, + "learning_rate": 5.2731046537033e-07, + "loss": 0.9584, + "step": 40993 + }, + { + "epoch": 8.95, + "learning_rate": 5.272012235088486e-07, + "loss": 0.9658, + "step": 40994 + }, + { + "epoch": 8.95, + "learning_rate": 5.270919816473673e-07, + "loss": 2.8616, + "step": 40995 + }, + { + "epoch": 8.95, + "learning_rate": 5.26982739785886e-07, + "loss": 1.3699, + "step": 40996 + }, + { + "epoch": 8.95, + "learning_rate": 5.268734979244046e-07, + "loss": 0.5801, + "step": 40997 + }, + { + "epoch": 8.95, + "learning_rate": 5.267642560629233e-07, + "loss": 1.62, + "step": 40998 + }, + { + "epoch": 8.95, + "learning_rate": 5.266550142014421e-07, + "loss": 0.7546, + "step": 40999 + }, + { + "epoch": 8.95, + "learning_rate": 5.265457723399607e-07, + "loss": 2.1887, + "step": 41000 + }, + { + "epoch": 8.95, + "learning_rate": 5.264365304784794e-07, + "loss": 0.7137, + "step": 41001 + }, + { + "epoch": 8.95, + "learning_rate": 5.263272886169981e-07, + "loss": 1.85, + "step": 41002 + }, + { + "epoch": 8.95, + "learning_rate": 5.262180467555167e-07, + "loss": 1.1927, + "step": 41003 + }, + { + "epoch": 8.95, + "learning_rate": 5.261088048940354e-07, + "loss": 1.6459, + "step": 41004 + }, + { + "epoch": 8.95, + "learning_rate": 5.259995630325542e-07, + "loss": 2.1213, + "step": 41005 + }, + { + "epoch": 8.95, + "learning_rate": 5.258903211710728e-07, + "loss": 0.8722, + "step": 41006 + }, + { + "epoch": 8.95, + "learning_rate": 5.257810793095915e-07, + "loss": 1.4522, + "step": 41007 + }, + { + "epoch": 8.95, + "learning_rate": 5.256718374481102e-07, + "loss": 1.4096, + "step": 41008 + }, + { + "epoch": 8.95, + "learning_rate": 5.255625955866288e-07, + "loss": 0.5032, + "step": 41009 + }, + { + "epoch": 8.95, + "learning_rate": 5.254533537251475e-07, + "loss": 0.9356, + "step": 41010 + }, + { + "epoch": 8.95, + "learning_rate": 5.253441118636661e-07, + "loss": 1.7953, + "step": 41011 + }, + { + "epoch": 8.95, + "learning_rate": 5.252348700021849e-07, + "loss": 2.2031, + "step": 41012 + }, + { + "epoch": 8.95, + "learning_rate": 5.251256281407036e-07, + "loss": 0.5335, + "step": 41013 + }, + { + "epoch": 8.95, + "learning_rate": 5.250163862792222e-07, + "loss": 0.9657, + "step": 41014 + }, + { + "epoch": 8.95, + "learning_rate": 5.249071444177409e-07, + "loss": 0.5184, + "step": 41015 + }, + { + "epoch": 8.95, + "learning_rate": 5.247979025562596e-07, + "loss": 0.7074, + "step": 41016 + }, + { + "epoch": 8.95, + "learning_rate": 5.246886606947782e-07, + "loss": 1.0686, + "step": 41017 + }, + { + "epoch": 8.95, + "learning_rate": 5.24579418833297e-07, + "loss": 2.1703, + "step": 41018 + }, + { + "epoch": 8.95, + "learning_rate": 5.244701769718157e-07, + "loss": 1.6557, + "step": 41019 + }, + { + "epoch": 8.95, + "learning_rate": 5.243609351103343e-07, + "loss": 1.5203, + "step": 41020 + }, + { + "epoch": 8.95, + "learning_rate": 5.24251693248853e-07, + "loss": 1.6754, + "step": 41021 + }, + { + "epoch": 8.95, + "learning_rate": 5.241424513873717e-07, + "loss": 1.3654, + "step": 41022 + }, + { + "epoch": 8.95, + "learning_rate": 5.240332095258903e-07, + "loss": 2.166, + "step": 41023 + }, + { + "epoch": 8.95, + "learning_rate": 5.23923967664409e-07, + "loss": 0.0732, + "step": 41024 + }, + { + "epoch": 8.95, + "learning_rate": 5.238147258029277e-07, + "loss": 2.1405, + "step": 41025 + }, + { + "epoch": 8.95, + "learning_rate": 5.237054839414464e-07, + "loss": 0.7691, + "step": 41026 + }, + { + "epoch": 8.95, + "learning_rate": 5.235962420799651e-07, + "loss": 0.7006, + "step": 41027 + }, + { + "epoch": 8.95, + "learning_rate": 5.234870002184838e-07, + "loss": 1.8189, + "step": 41028 + }, + { + "epoch": 8.95, + "learning_rate": 5.233777583570024e-07, + "loss": 1.8661, + "step": 41029 + }, + { + "epoch": 8.95, + "learning_rate": 5.232685164955211e-07, + "loss": 0.493, + "step": 41030 + }, + { + "epoch": 8.95, + "learning_rate": 5.231592746340397e-07, + "loss": 1.1767, + "step": 41031 + }, + { + "epoch": 8.96, + "learning_rate": 5.230500327725586e-07, + "loss": 0.1226, + "step": 41032 + }, + { + "epoch": 8.96, + "learning_rate": 5.229407909110772e-07, + "loss": 0.0919, + "step": 41033 + }, + { + "epoch": 8.96, + "learning_rate": 5.228315490495958e-07, + "loss": 0.7725, + "step": 41034 + }, + { + "epoch": 8.96, + "learning_rate": 5.227223071881146e-07, + "loss": 1.6079, + "step": 41035 + }, + { + "epoch": 8.96, + "learning_rate": 5.226130653266332e-07, + "loss": 1.0244, + "step": 41036 + }, + { + "epoch": 8.96, + "learning_rate": 5.225038234651518e-07, + "loss": 1.8805, + "step": 41037 + }, + { + "epoch": 8.96, + "learning_rate": 5.223945816036706e-07, + "loss": 0.6839, + "step": 41038 + }, + { + "epoch": 8.96, + "learning_rate": 5.222853397421893e-07, + "loss": 1.659, + "step": 41039 + }, + { + "epoch": 8.96, + "learning_rate": 5.221760978807079e-07, + "loss": 1.1782, + "step": 41040 + }, + { + "epoch": 8.96, + "learning_rate": 5.220668560192267e-07, + "loss": 1.1812, + "step": 41041 + }, + { + "epoch": 8.96, + "learning_rate": 5.219576141577453e-07, + "loss": 0.1659, + "step": 41042 + }, + { + "epoch": 8.96, + "learning_rate": 5.21848372296264e-07, + "loss": 2.1927, + "step": 41043 + }, + { + "epoch": 8.96, + "learning_rate": 5.217391304347826e-07, + "loss": 1.3654, + "step": 41044 + }, + { + "epoch": 8.96, + "learning_rate": 5.216298885733013e-07, + "loss": 0.6063, + "step": 41045 + }, + { + "epoch": 8.96, + "learning_rate": 5.215206467118201e-07, + "loss": 4.2821, + "step": 41046 + }, + { + "epoch": 8.96, + "learning_rate": 5.214114048503387e-07, + "loss": 2.5508, + "step": 41047 + }, + { + "epoch": 8.96, + "learning_rate": 5.213021629888574e-07, + "loss": 0.6784, + "step": 41048 + }, + { + "epoch": 8.96, + "learning_rate": 5.211929211273761e-07, + "loss": 2.4658, + "step": 41049 + }, + { + "epoch": 8.96, + "learning_rate": 5.210836792658947e-07, + "loss": 0.4373, + "step": 41050 + }, + { + "epoch": 8.96, + "learning_rate": 5.209744374044134e-07, + "loss": 0.1338, + "step": 41051 + }, + { + "epoch": 8.96, + "learning_rate": 5.208651955429321e-07, + "loss": 2.5265, + "step": 41052 + }, + { + "epoch": 8.96, + "learning_rate": 5.207559536814508e-07, + "loss": 0.7243, + "step": 41053 + }, + { + "epoch": 8.96, + "learning_rate": 5.206467118199695e-07, + "loss": 0.238, + "step": 41054 + }, + { + "epoch": 8.96, + "learning_rate": 5.205374699584882e-07, + "loss": 0.3948, + "step": 41055 + }, + { + "epoch": 8.96, + "learning_rate": 5.204282280970068e-07, + "loss": 0.3055, + "step": 41056 + }, + { + "epoch": 8.96, + "learning_rate": 5.203189862355255e-07, + "loss": 1.918, + "step": 41057 + }, + { + "epoch": 8.96, + "learning_rate": 5.202097443740442e-07, + "loss": 0.2456, + "step": 41058 + }, + { + "epoch": 8.96, + "learning_rate": 5.201005025125629e-07, + "loss": 2.5687, + "step": 41059 + }, + { + "epoch": 8.96, + "learning_rate": 5.199912606510816e-07, + "loss": 1.085, + "step": 41060 + }, + { + "epoch": 8.96, + "learning_rate": 5.198820187896003e-07, + "loss": 0.6582, + "step": 41061 + }, + { + "epoch": 8.96, + "learning_rate": 5.197727769281189e-07, + "loss": 1.0787, + "step": 41062 + }, + { + "epoch": 8.96, + "learning_rate": 5.196635350666376e-07, + "loss": 1.0539, + "step": 41063 + }, + { + "epoch": 8.96, + "learning_rate": 5.195542932051562e-07, + "loss": 1.8717, + "step": 41064 + }, + { + "epoch": 8.96, + "learning_rate": 5.194450513436749e-07, + "loss": 0.2674, + "step": 41065 + }, + { + "epoch": 8.96, + "learning_rate": 5.193358094821937e-07, + "loss": 3.157, + "step": 41066 + }, + { + "epoch": 8.96, + "learning_rate": 5.192265676207123e-07, + "loss": 0.6173, + "step": 41067 + }, + { + "epoch": 8.96, + "learning_rate": 5.19117325759231e-07, + "loss": 1.1345, + "step": 41068 + }, + { + "epoch": 8.96, + "learning_rate": 5.190080838977497e-07, + "loss": 0.3792, + "step": 41069 + }, + { + "epoch": 8.96, + "learning_rate": 5.188988420362683e-07, + "loss": 1.4109, + "step": 41070 + }, + { + "epoch": 8.96, + "learning_rate": 5.18789600174787e-07, + "loss": 0.9035, + "step": 41071 + }, + { + "epoch": 8.96, + "learning_rate": 5.186803583133057e-07, + "loss": 1.4389, + "step": 41072 + }, + { + "epoch": 8.96, + "learning_rate": 5.185711164518244e-07, + "loss": 0.9118, + "step": 41073 + }, + { + "epoch": 8.96, + "learning_rate": 5.184618745903431e-07, + "loss": 0.8031, + "step": 41074 + }, + { + "epoch": 8.96, + "learning_rate": 5.183526327288618e-07, + "loss": 1.0212, + "step": 41075 + }, + { + "epoch": 8.96, + "learning_rate": 5.182433908673804e-07, + "loss": 1.2654, + "step": 41076 + }, + { + "epoch": 8.96, + "learning_rate": 5.181341490058991e-07, + "loss": 1.4296, + "step": 41077 + }, + { + "epoch": 8.97, + "learning_rate": 5.180249071444178e-07, + "loss": 0.987, + "step": 41078 + }, + { + "epoch": 8.97, + "learning_rate": 5.179156652829364e-07, + "loss": 1.716, + "step": 41079 + }, + { + "epoch": 8.97, + "learning_rate": 5.178064234214552e-07, + "loss": 1.273, + "step": 41080 + }, + { + "epoch": 8.97, + "learning_rate": 5.176971815599739e-07, + "loss": 0.9947, + "step": 41081 + }, + { + "epoch": 8.97, + "learning_rate": 5.175879396984925e-07, + "loss": 0.8489, + "step": 41082 + }, + { + "epoch": 8.97, + "learning_rate": 5.174786978370112e-07, + "loss": 1.7212, + "step": 41083 + }, + { + "epoch": 8.97, + "learning_rate": 5.173694559755299e-07, + "loss": 0.52, + "step": 41084 + }, + { + "epoch": 8.97, + "learning_rate": 5.172602141140485e-07, + "loss": 0.6885, + "step": 41085 + }, + { + "epoch": 8.97, + "learning_rate": 5.171509722525673e-07, + "loss": 3.175, + "step": 41086 + }, + { + "epoch": 8.97, + "learning_rate": 5.170417303910859e-07, + "loss": 1.135, + "step": 41087 + }, + { + "epoch": 8.97, + "learning_rate": 5.169324885296046e-07, + "loss": 0.4693, + "step": 41088 + }, + { + "epoch": 8.97, + "learning_rate": 5.168232466681233e-07, + "loss": 0.6974, + "step": 41089 + }, + { + "epoch": 8.97, + "learning_rate": 5.167140048066419e-07, + "loss": 0.6409, + "step": 41090 + }, + { + "epoch": 8.97, + "learning_rate": 5.166047629451606e-07, + "loss": 1.3638, + "step": 41091 + }, + { + "epoch": 8.97, + "learning_rate": 5.164955210836793e-07, + "loss": 0.9361, + "step": 41092 + }, + { + "epoch": 8.97, + "learning_rate": 5.16386279222198e-07, + "loss": 1.0767, + "step": 41093 + }, + { + "epoch": 8.97, + "learning_rate": 5.162770373607167e-07, + "loss": 1.6627, + "step": 41094 + }, + { + "epoch": 8.97, + "learning_rate": 5.161677954992354e-07, + "loss": 0.2333, + "step": 41095 + }, + { + "epoch": 8.97, + "learning_rate": 5.16058553637754e-07, + "loss": 3.0326, + "step": 41096 + }, + { + "epoch": 8.97, + "learning_rate": 5.159493117762727e-07, + "loss": 0.0838, + "step": 41097 + }, + { + "epoch": 8.97, + "learning_rate": 5.158400699147914e-07, + "loss": 0.9345, + "step": 41098 + }, + { + "epoch": 8.97, + "learning_rate": 5.1573082805331e-07, + "loss": 1.5117, + "step": 41099 + }, + { + "epoch": 8.97, + "learning_rate": 5.156215861918288e-07, + "loss": 1.0134, + "step": 41100 + }, + { + "epoch": 8.97, + "learning_rate": 5.155123443303475e-07, + "loss": 0.3834, + "step": 41101 + }, + { + "epoch": 8.97, + "learning_rate": 5.154031024688661e-07, + "loss": 1.5685, + "step": 41102 + }, + { + "epoch": 8.97, + "learning_rate": 5.152938606073848e-07, + "loss": 0.8645, + "step": 41103 + }, + { + "epoch": 8.97, + "learning_rate": 5.151846187459035e-07, + "loss": 0.0534, + "step": 41104 + }, + { + "epoch": 8.97, + "learning_rate": 5.150753768844221e-07, + "loss": 2.5904, + "step": 41105 + }, + { + "epoch": 8.97, + "learning_rate": 5.149661350229408e-07, + "loss": 2.4037, + "step": 41106 + }, + { + "epoch": 8.97, + "learning_rate": 5.148568931614595e-07, + "loss": 1.6589, + "step": 41107 + }, + { + "epoch": 8.97, + "learning_rate": 5.147476512999782e-07, + "loss": 1.5069, + "step": 41108 + }, + { + "epoch": 8.97, + "learning_rate": 5.146384094384969e-07, + "loss": 0.4579, + "step": 41109 + }, + { + "epoch": 8.97, + "learning_rate": 5.145291675770155e-07, + "loss": 0.2923, + "step": 41110 + }, + { + "epoch": 8.97, + "learning_rate": 5.144199257155342e-07, + "loss": 1.509, + "step": 41111 + }, + { + "epoch": 8.97, + "learning_rate": 5.143106838540529e-07, + "loss": 0.3025, + "step": 41112 + }, + { + "epoch": 8.97, + "learning_rate": 5.142014419925715e-07, + "loss": 0.5397, + "step": 41113 + }, + { + "epoch": 8.97, + "learning_rate": 5.140922001310903e-07, + "loss": 1.878, + "step": 41114 + }, + { + "epoch": 8.97, + "learning_rate": 5.13982958269609e-07, + "loss": 0.2799, + "step": 41115 + }, + { + "epoch": 8.97, + "learning_rate": 5.138737164081276e-07, + "loss": 1.2628, + "step": 41116 + }, + { + "epoch": 8.97, + "learning_rate": 5.137644745466463e-07, + "loss": 0.1381, + "step": 41117 + }, + { + "epoch": 8.97, + "learning_rate": 5.13655232685165e-07, + "loss": 1.2107, + "step": 41118 + }, + { + "epoch": 8.97, + "learning_rate": 5.135459908236836e-07, + "loss": 0.2658, + "step": 41119 + }, + { + "epoch": 8.97, + "learning_rate": 5.134367489622024e-07, + "loss": 0.9823, + "step": 41120 + }, + { + "epoch": 8.97, + "learning_rate": 5.133275071007211e-07, + "loss": 0.6407, + "step": 41121 + }, + { + "epoch": 8.97, + "learning_rate": 5.132182652392397e-07, + "loss": 2.1252, + "step": 41122 + }, + { + "epoch": 8.97, + "learning_rate": 5.131090233777584e-07, + "loss": 0.0859, + "step": 41123 + }, + { + "epoch": 8.98, + "learning_rate": 5.129997815162771e-07, + "loss": 1.6612, + "step": 41124 + }, + { + "epoch": 8.98, + "learning_rate": 5.128905396547957e-07, + "loss": 0.8656, + "step": 41125 + }, + { + "epoch": 8.98, + "learning_rate": 5.127812977933144e-07, + "loss": 0.157, + "step": 41126 + }, + { + "epoch": 8.98, + "learning_rate": 5.126720559318332e-07, + "loss": 0.5571, + "step": 41127 + }, + { + "epoch": 8.98, + "learning_rate": 5.125628140703518e-07, + "loss": 0.2208, + "step": 41128 + }, + { + "epoch": 8.98, + "learning_rate": 5.124535722088705e-07, + "loss": 0.3028, + "step": 41129 + }, + { + "epoch": 8.98, + "learning_rate": 5.123443303473891e-07, + "loss": 0.2837, + "step": 41130 + }, + { + "epoch": 8.98, + "learning_rate": 5.122350884859078e-07, + "loss": 2.2681, + "step": 41131 + }, + { + "epoch": 8.98, + "learning_rate": 5.121258466244265e-07, + "loss": 0.7333, + "step": 41132 + }, + { + "epoch": 8.98, + "learning_rate": 5.120166047629451e-07, + "loss": 1.5031, + "step": 41133 + }, + { + "epoch": 8.98, + "learning_rate": 5.119073629014639e-07, + "loss": 2.4342, + "step": 41134 + }, + { + "epoch": 8.98, + "learning_rate": 5.117981210399826e-07, + "loss": 0.1194, + "step": 41135 + }, + { + "epoch": 8.98, + "learning_rate": 5.116888791785012e-07, + "loss": 1.717, + "step": 41136 + }, + { + "epoch": 8.98, + "learning_rate": 5.115796373170199e-07, + "loss": 0.7904, + "step": 41137 + }, + { + "epoch": 8.98, + "learning_rate": 5.114703954555386e-07, + "loss": 0.8824, + "step": 41138 + }, + { + "epoch": 8.98, + "learning_rate": 5.113611535940572e-07, + "loss": 0.7971, + "step": 41139 + }, + { + "epoch": 8.98, + "learning_rate": 5.11251911732576e-07, + "loss": 1.2731, + "step": 41140 + }, + { + "epoch": 8.98, + "learning_rate": 5.111426698710947e-07, + "loss": 1.0348, + "step": 41141 + }, + { + "epoch": 8.98, + "learning_rate": 5.110334280096133e-07, + "loss": 1.0752, + "step": 41142 + }, + { + "epoch": 8.98, + "learning_rate": 5.10924186148132e-07, + "loss": 1.5198, + "step": 41143 + }, + { + "epoch": 8.98, + "learning_rate": 5.108149442866507e-07, + "loss": 0.8535, + "step": 41144 + }, + { + "epoch": 8.98, + "learning_rate": 5.107057024251693e-07, + "loss": 2.645, + "step": 41145 + }, + { + "epoch": 8.98, + "learning_rate": 5.10596460563688e-07, + "loss": 1.1839, + "step": 41146 + }, + { + "epoch": 8.98, + "learning_rate": 5.104872187022068e-07, + "loss": 0.9501, + "step": 41147 + }, + { + "epoch": 8.98, + "learning_rate": 5.103779768407254e-07, + "loss": 0.6625, + "step": 41148 + }, + { + "epoch": 8.98, + "learning_rate": 5.102687349792441e-07, + "loss": 1.4872, + "step": 41149 + }, + { + "epoch": 8.98, + "learning_rate": 5.101594931177627e-07, + "loss": 0.1352, + "step": 41150 + }, + { + "epoch": 8.98, + "learning_rate": 5.100502512562814e-07, + "loss": 0.5728, + "step": 41151 + }, + { + "epoch": 8.98, + "learning_rate": 5.099410093948001e-07, + "loss": 2.0185, + "step": 41152 + }, + { + "epoch": 8.98, + "learning_rate": 5.098317675333187e-07, + "loss": 0.1891, + "step": 41153 + }, + { + "epoch": 8.98, + "learning_rate": 5.097225256718376e-07, + "loss": 1.3173, + "step": 41154 + }, + { + "epoch": 8.98, + "learning_rate": 5.096132838103562e-07, + "loss": 1.0988, + "step": 41155 + }, + { + "epoch": 8.98, + "learning_rate": 5.095040419488748e-07, + "loss": 2.1201, + "step": 41156 + }, + { + "epoch": 8.98, + "learning_rate": 5.093948000873936e-07, + "loss": 1.5601, + "step": 41157 + }, + { + "epoch": 8.98, + "learning_rate": 5.092855582259122e-07, + "loss": 2.2271, + "step": 41158 + }, + { + "epoch": 8.98, + "learning_rate": 5.091763163644308e-07, + "loss": 0.7132, + "step": 41159 + }, + { + "epoch": 8.98, + "learning_rate": 5.090670745029495e-07, + "loss": 0.2137, + "step": 41160 + }, + { + "epoch": 8.98, + "learning_rate": 5.089578326414683e-07, + "loss": 0.888, + "step": 41161 + }, + { + "epoch": 8.98, + "learning_rate": 5.088485907799869e-07, + "loss": 0.8328, + "step": 41162 + }, + { + "epoch": 8.98, + "learning_rate": 5.087393489185056e-07, + "loss": 0.1293, + "step": 41163 + }, + { + "epoch": 8.98, + "learning_rate": 5.086301070570243e-07, + "loss": 1.3081, + "step": 41164 + }, + { + "epoch": 8.98, + "learning_rate": 5.085208651955429e-07, + "loss": 1.5949, + "step": 41165 + }, + { + "epoch": 8.98, + "learning_rate": 5.084116233340616e-07, + "loss": 3.1702, + "step": 41166 + }, + { + "epoch": 8.98, + "learning_rate": 5.083023814725804e-07, + "loss": 1.2553, + "step": 41167 + }, + { + "epoch": 8.98, + "learning_rate": 5.081931396110991e-07, + "loss": 1.1056, + "step": 41168 + }, + { + "epoch": 8.98, + "learning_rate": 5.080838977496177e-07, + "loss": 0.4653, + "step": 41169 + }, + { + "epoch": 8.99, + "learning_rate": 5.079746558881364e-07, + "loss": 0.3611, + "step": 41170 + }, + { + "epoch": 8.99, + "learning_rate": 5.078654140266551e-07, + "loss": 0.5097, + "step": 41171 + }, + { + "epoch": 8.99, + "learning_rate": 5.077561721651737e-07, + "loss": 1.6855, + "step": 41172 + }, + { + "epoch": 8.99, + "learning_rate": 5.076469303036923e-07, + "loss": 0.0258, + "step": 41173 + }, + { + "epoch": 8.99, + "learning_rate": 5.075376884422112e-07, + "loss": 3.4169, + "step": 41174 + }, + { + "epoch": 8.99, + "learning_rate": 5.074284465807298e-07, + "loss": 3.2657, + "step": 41175 + }, + { + "epoch": 8.99, + "learning_rate": 5.073192047192484e-07, + "loss": 1.1175, + "step": 41176 + }, + { + "epoch": 8.99, + "learning_rate": 5.072099628577672e-07, + "loss": 1.1388, + "step": 41177 + }, + { + "epoch": 8.99, + "learning_rate": 5.071007209962858e-07, + "loss": 1.1983, + "step": 41178 + }, + { + "epoch": 8.99, + "learning_rate": 5.069914791348045e-07, + "loss": 0.4035, + "step": 41179 + }, + { + "epoch": 8.99, + "learning_rate": 5.068822372733232e-07, + "loss": 0.6656, + "step": 41180 + }, + { + "epoch": 8.99, + "learning_rate": 5.067729954118419e-07, + "loss": 1.0605, + "step": 41181 + }, + { + "epoch": 8.99, + "learning_rate": 5.066637535503606e-07, + "loss": 1.1872, + "step": 41182 + }, + { + "epoch": 8.99, + "learning_rate": 5.065545116888792e-07, + "loss": 0.3241, + "step": 41183 + }, + { + "epoch": 8.99, + "learning_rate": 5.064452698273979e-07, + "loss": 1.3704, + "step": 41184 + }, + { + "epoch": 8.99, + "learning_rate": 5.063360279659166e-07, + "loss": 1.9896, + "step": 41185 + }, + { + "epoch": 8.99, + "learning_rate": 5.062267861044352e-07, + "loss": 1.7711, + "step": 41186 + }, + { + "epoch": 8.99, + "learning_rate": 5.06117544242954e-07, + "loss": 1.3469, + "step": 41187 + }, + { + "epoch": 8.99, + "learning_rate": 5.060083023814727e-07, + "loss": 0.6197, + "step": 41188 + }, + { + "epoch": 8.99, + "learning_rate": 5.058990605199913e-07, + "loss": 0.8056, + "step": 41189 + }, + { + "epoch": 8.99, + "learning_rate": 5.0578981865851e-07, + "loss": 0.7825, + "step": 41190 + }, + { + "epoch": 8.99, + "learning_rate": 5.056805767970287e-07, + "loss": 1.1937, + "step": 41191 + }, + { + "epoch": 8.99, + "learning_rate": 5.055713349355473e-07, + "loss": 1.0614, + "step": 41192 + }, + { + "epoch": 8.99, + "learning_rate": 5.05462093074066e-07, + "loss": 0.882, + "step": 41193 + }, + { + "epoch": 8.99, + "learning_rate": 5.053528512125848e-07, + "loss": 1.0434, + "step": 41194 + }, + { + "epoch": 8.99, + "learning_rate": 5.052436093511034e-07, + "loss": 0.1413, + "step": 41195 + }, + { + "epoch": 8.99, + "learning_rate": 5.051343674896221e-07, + "loss": 0.4537, + "step": 41196 + }, + { + "epoch": 8.99, + "learning_rate": 5.050251256281408e-07, + "loss": 0.3716, + "step": 41197 + }, + { + "epoch": 8.99, + "learning_rate": 5.049158837666594e-07, + "loss": 0.1456, + "step": 41198 + }, + { + "epoch": 8.99, + "learning_rate": 5.048066419051781e-07, + "loss": 0.7303, + "step": 41199 + }, + { + "epoch": 8.99, + "learning_rate": 5.046974000436968e-07, + "loss": 0.9873, + "step": 41200 + }, + { + "epoch": 8.99, + "learning_rate": 5.045881581822155e-07, + "loss": 3.0508, + "step": 41201 + }, + { + "epoch": 8.99, + "learning_rate": 5.044789163207342e-07, + "loss": 1.6571, + "step": 41202 + }, + { + "epoch": 8.99, + "learning_rate": 5.043696744592529e-07, + "loss": 2.1387, + "step": 41203 + }, + { + "epoch": 8.99, + "learning_rate": 5.042604325977715e-07, + "loss": 2.1141, + "step": 41204 + }, + { + "epoch": 8.99, + "learning_rate": 5.041511907362902e-07, + "loss": 1.502, + "step": 41205 + }, + { + "epoch": 8.99, + "learning_rate": 5.040419488748088e-07, + "loss": 2.0722, + "step": 41206 + }, + { + "epoch": 8.99, + "learning_rate": 5.039327070133275e-07, + "loss": 0.5943, + "step": 41207 + }, + { + "epoch": 8.99, + "learning_rate": 5.038234651518463e-07, + "loss": 1.5955, + "step": 41208 + }, + { + "epoch": 8.99, + "learning_rate": 5.037142232903649e-07, + "loss": 0.6625, + "step": 41209 + }, + { + "epoch": 8.99, + "learning_rate": 5.036049814288836e-07, + "loss": 1.2013, + "step": 41210 + }, + { + "epoch": 8.99, + "learning_rate": 5.034957395674023e-07, + "loss": 1.2014, + "step": 41211 + }, + { + "epoch": 8.99, + "learning_rate": 5.033864977059209e-07, + "loss": 0.7314, + "step": 41212 + }, + { + "epoch": 8.99, + "learning_rate": 5.032772558444396e-07, + "loss": 1.3869, + "step": 41213 + }, + { + "epoch": 8.99, + "learning_rate": 5.031680139829584e-07, + "loss": 1.3763, + "step": 41214 + }, + { + "epoch": 8.99, + "learning_rate": 5.03058772121477e-07, + "loss": 1.1633, + "step": 41215 + }, + { + "epoch": 9.0, + "learning_rate": 5.029495302599957e-07, + "loss": 0.8706, + "step": 41216 + }, + { + "epoch": 9.0, + "learning_rate": 5.028402883985144e-07, + "loss": 0.0785, + "step": 41217 + }, + { + "epoch": 9.0, + "learning_rate": 5.02731046537033e-07, + "loss": 1.2764, + "step": 41218 + }, + { + "epoch": 9.0, + "learning_rate": 5.026218046755517e-07, + "loss": 0.4457, + "step": 41219 + }, + { + "epoch": 9.0, + "learning_rate": 5.025125628140704e-07, + "loss": 0.7786, + "step": 41220 + }, + { + "epoch": 9.0, + "learning_rate": 5.024033209525891e-07, + "loss": 3.4797, + "step": 41221 + }, + { + "epoch": 9.0, + "learning_rate": 5.022940790911078e-07, + "loss": 1.5282, + "step": 41222 + }, + { + "epoch": 9.0, + "learning_rate": 5.021848372296265e-07, + "loss": 0.6999, + "step": 41223 + }, + { + "epoch": 9.0, + "learning_rate": 5.020755953681451e-07, + "loss": 0.9124, + "step": 41224 + }, + { + "epoch": 9.0, + "learning_rate": 5.019663535066638e-07, + "loss": 0.1058, + "step": 41225 + }, + { + "epoch": 9.0, + "learning_rate": 5.018571116451824e-07, + "loss": 0.1455, + "step": 41226 + }, + { + "epoch": 9.0, + "learning_rate": 5.017478697837011e-07, + "loss": 0.6936, + "step": 41227 + }, + { + "epoch": 9.0, + "learning_rate": 5.016386279222199e-07, + "loss": 1.5211, + "step": 41228 + }, + { + "epoch": 9.0, + "learning_rate": 5.015293860607385e-07, + "loss": 1.6776, + "step": 41229 + }, + { + "epoch": 9.0, + "learning_rate": 5.014201441992572e-07, + "loss": 1.4964, + "step": 41230 + }, + { + "epoch": 9.0, + "learning_rate": 5.013109023377759e-07, + "loss": 0.2514, + "step": 41231 + }, + { + "epoch": 9.0, + "learning_rate": 5.012016604762945e-07, + "loss": 0.7635, + "step": 41232 + }, + { + "epoch": 9.0, + "learning_rate": 5.010924186148132e-07, + "loss": 1.7985, + "step": 41233 + }, + { + "epoch": 9.0, + "learning_rate": 5.00983176753332e-07, + "loss": 1.8427, + "step": 41234 + }, + { + "epoch": 9.0, + "learning_rate": 5.008739348918506e-07, + "loss": 0.3265, + "step": 41235 + }, + { + "epoch": 9.0, + "learning_rate": 5.007646930303693e-07, + "loss": 0.1403, + "step": 41236 + }, + { + "epoch": 9.0, + "learning_rate": 5.00655451168888e-07, + "loss": 1.4519, + "step": 41237 + }, + { + "epoch": 9.0, + "learning_rate": 5.005462093074066e-07, + "loss": 0.1009, + "step": 41238 + }, + { + "epoch": 9.0, + "eval_accuracy": 0.7165302782324059, + "eval_loss": 1.136122703552246, + "eval_runtime": 334.4001, + "eval_samples_per_second": 18.272, + "eval_steps_per_second": 4.569, + "step": 41238 + }, + { + "epoch": 9.0, + "learning_rate": 5.004369674459253e-07, + "loss": 0.1323, + "step": 41239 + }, + { + "epoch": 9.0, + "learning_rate": 5.00327725584444e-07, + "loss": 2.2203, + "step": 41240 + }, + { + "epoch": 9.0, + "learning_rate": 5.002184837229626e-07, + "loss": 0.1517, + "step": 41241 + }, + { + "epoch": 9.0, + "learning_rate": 5.001092418614814e-07, + "loss": 1.8506, + "step": 41242 + }, + { + "epoch": 9.0, + "learning_rate": 5.000000000000001e-07, + "loss": 0.6698, + "step": 41243 + }, + { + "epoch": 9.0, + "learning_rate": 4.998907581385187e-07, + "loss": 0.5183, + "step": 41244 + }, + { + "epoch": 9.0, + "learning_rate": 4.997815162770374e-07, + "loss": 0.5453, + "step": 41245 + }, + { + "epoch": 9.0, + "learning_rate": 4.996722744155561e-07, + "loss": 0.5667, + "step": 41246 + }, + { + "epoch": 9.0, + "learning_rate": 4.995630325540747e-07, + "loss": 1.5025, + "step": 41247 + }, + { + "epoch": 9.0, + "learning_rate": 4.994537906925935e-07, + "loss": 1.4273, + "step": 41248 + }, + { + "epoch": 9.0, + "learning_rate": 4.993445488311121e-07, + "loss": 1.5271, + "step": 41249 + }, + { + "epoch": 9.0, + "learning_rate": 4.992353069696308e-07, + "loss": 2.4165, + "step": 41250 + }, + { + "epoch": 9.0, + "learning_rate": 4.991260651081495e-07, + "loss": 1.0538, + "step": 41251 + }, + { + "epoch": 9.0, + "learning_rate": 4.990168232466681e-07, + "loss": 0.6844, + "step": 41252 + }, + { + "epoch": 9.0, + "learning_rate": 4.989075813851868e-07, + "loss": 2.2213, + "step": 41253 + }, + { + "epoch": 9.0, + "learning_rate": 4.987983395237055e-07, + "loss": 1.8032, + "step": 41254 + }, + { + "epoch": 9.0, + "learning_rate": 4.986890976622242e-07, + "loss": 0.5027, + "step": 41255 + }, + { + "epoch": 9.0, + "learning_rate": 4.985798558007429e-07, + "loss": 1.0537, + "step": 41256 + }, + { + "epoch": 9.0, + "learning_rate": 4.984706139392616e-07, + "loss": 1.7171, + "step": 41257 + }, + { + "epoch": 9.0, + "learning_rate": 4.983613720777802e-07, + "loss": 1.795, + "step": 41258 + }, + { + "epoch": 9.0, + "learning_rate": 4.982521302162989e-07, + "loss": 0.275, + "step": 41259 + }, + { + "epoch": 9.0, + "learning_rate": 4.981428883548176e-07, + "loss": 1.3183, + "step": 41260 + }, + { + "epoch": 9.01, + "learning_rate": 4.980336464933362e-07, + "loss": 0.6371, + "step": 41261 + }, + { + "epoch": 9.01, + "learning_rate": 4.97924404631855e-07, + "loss": 1.4231, + "step": 41262 + }, + { + "epoch": 9.01, + "learning_rate": 4.978151627703737e-07, + "loss": 0.6121, + "step": 41263 + }, + { + "epoch": 9.01, + "learning_rate": 4.977059209088923e-07, + "loss": 0.3282, + "step": 41264 + }, + { + "epoch": 9.01, + "learning_rate": 4.97596679047411e-07, + "loss": 1.7199, + "step": 41265 + }, + { + "epoch": 9.01, + "learning_rate": 4.974874371859297e-07, + "loss": 1.029, + "step": 41266 + }, + { + "epoch": 9.01, + "learning_rate": 4.973781953244483e-07, + "loss": 0.7571, + "step": 41267 + }, + { + "epoch": 9.01, + "learning_rate": 4.97268953462967e-07, + "loss": 1.8706, + "step": 41268 + }, + { + "epoch": 9.01, + "learning_rate": 4.971597116014857e-07, + "loss": 2.5822, + "step": 41269 + }, + { + "epoch": 9.01, + "learning_rate": 4.970504697400044e-07, + "loss": 0.2093, + "step": 41270 + }, + { + "epoch": 9.01, + "learning_rate": 4.969412278785231e-07, + "loss": 0.5016, + "step": 41271 + }, + { + "epoch": 9.01, + "learning_rate": 4.968319860170417e-07, + "loss": 1.0345, + "step": 41272 + }, + { + "epoch": 9.01, + "learning_rate": 4.967227441555604e-07, + "loss": 0.1947, + "step": 41273 + }, + { + "epoch": 9.01, + "learning_rate": 4.966135022940791e-07, + "loss": 0.5725, + "step": 41274 + }, + { + "epoch": 9.01, + "learning_rate": 4.965042604325978e-07, + "loss": 1.5495, + "step": 41275 + }, + { + "epoch": 9.01, + "learning_rate": 4.963950185711165e-07, + "loss": 0.992, + "step": 41276 + }, + { + "epoch": 9.01, + "learning_rate": 4.962857767096352e-07, + "loss": 1.8856, + "step": 41277 + }, + { + "epoch": 9.01, + "learning_rate": 4.961765348481538e-07, + "loss": 0.0511, + "step": 41278 + }, + { + "epoch": 9.01, + "learning_rate": 4.960672929866725e-07, + "loss": 0.4431, + "step": 41279 + }, + { + "epoch": 9.01, + "learning_rate": 4.959580511251912e-07, + "loss": 1.3192, + "step": 41280 + }, + { + "epoch": 9.01, + "learning_rate": 4.958488092637098e-07, + "loss": 1.1734, + "step": 41281 + }, + { + "epoch": 9.01, + "learning_rate": 4.957395674022286e-07, + "loss": 1.8338, + "step": 41282 + }, + { + "epoch": 9.01, + "learning_rate": 4.956303255407473e-07, + "loss": 0.6974, + "step": 41283 + }, + { + "epoch": 9.01, + "learning_rate": 4.955210836792659e-07, + "loss": 0.1379, + "step": 41284 + }, + { + "epoch": 9.01, + "learning_rate": 4.954118418177846e-07, + "loss": 0.6668, + "step": 41285 + }, + { + "epoch": 9.01, + "learning_rate": 4.953025999563033e-07, + "loss": 0.5934, + "step": 41286 + }, + { + "epoch": 9.01, + "learning_rate": 4.951933580948219e-07, + "loss": 1.2736, + "step": 41287 + }, + { + "epoch": 9.01, + "learning_rate": 4.950841162333406e-07, + "loss": 1.3701, + "step": 41288 + }, + { + "epoch": 9.01, + "learning_rate": 4.949748743718594e-07, + "loss": 0.1891, + "step": 41289 + }, + { + "epoch": 9.01, + "learning_rate": 4.948656325103781e-07, + "loss": 0.9887, + "step": 41290 + }, + { + "epoch": 9.01, + "learning_rate": 4.947563906488967e-07, + "loss": 1.5456, + "step": 41291 + }, + { + "epoch": 9.01, + "learning_rate": 4.946471487874153e-07, + "loss": 0.1599, + "step": 41292 + }, + { + "epoch": 9.01, + "learning_rate": 4.945379069259341e-07, + "loss": 0.1288, + "step": 41293 + }, + { + "epoch": 9.01, + "learning_rate": 4.944286650644527e-07, + "loss": 1.0572, + "step": 41294 + }, + { + "epoch": 9.01, + "learning_rate": 4.943194232029713e-07, + "loss": 0.2523, + "step": 41295 + }, + { + "epoch": 9.01, + "learning_rate": 4.942101813414902e-07, + "loss": 0.2429, + "step": 41296 + }, + { + "epoch": 9.01, + "learning_rate": 4.941009394800088e-07, + "loss": 0.3536, + "step": 41297 + }, + { + "epoch": 9.01, + "learning_rate": 4.939916976185274e-07, + "loss": 1.942, + "step": 41298 + }, + { + "epoch": 9.01, + "learning_rate": 4.938824557570462e-07, + "loss": 1.2807, + "step": 41299 + }, + { + "epoch": 9.01, + "learning_rate": 4.937732138955648e-07, + "loss": 0.7034, + "step": 41300 + }, + { + "epoch": 9.01, + "learning_rate": 4.936639720340835e-07, + "loss": 1.9727, + "step": 41301 + }, + { + "epoch": 9.01, + "learning_rate": 4.935547301726022e-07, + "loss": 0.6677, + "step": 41302 + }, + { + "epoch": 9.01, + "learning_rate": 4.934454883111209e-07, + "loss": 0.4358, + "step": 41303 + }, + { + "epoch": 9.01, + "learning_rate": 4.933362464496396e-07, + "loss": 1.257, + "step": 41304 + }, + { + "epoch": 9.01, + "learning_rate": 4.932270045881582e-07, + "loss": 1.8199, + "step": 41305 + }, + { + "epoch": 9.01, + "learning_rate": 4.931177627266769e-07, + "loss": 1.5566, + "step": 41306 + }, + { + "epoch": 9.02, + "learning_rate": 4.930085208651956e-07, + "loss": 1.6975, + "step": 41307 + }, + { + "epoch": 9.02, + "learning_rate": 4.928992790037142e-07, + "loss": 1.0819, + "step": 41308 + }, + { + "epoch": 9.02, + "learning_rate": 4.92790037142233e-07, + "loss": 0.6321, + "step": 41309 + }, + { + "epoch": 9.02, + "learning_rate": 4.926807952807517e-07, + "loss": 1.5413, + "step": 41310 + }, + { + "epoch": 9.02, + "learning_rate": 4.925715534192703e-07, + "loss": 1.2676, + "step": 41311 + }, + { + "epoch": 9.02, + "learning_rate": 4.924623115577889e-07, + "loss": 0.0739, + "step": 41312 + }, + { + "epoch": 9.02, + "learning_rate": 4.923530696963077e-07, + "loss": 1.1519, + "step": 41313 + }, + { + "epoch": 9.02, + "learning_rate": 4.922438278348263e-07, + "loss": 1.5733, + "step": 41314 + }, + { + "epoch": 9.02, + "learning_rate": 4.92134585973345e-07, + "loss": 0.1815, + "step": 41315 + }, + { + "epoch": 9.02, + "learning_rate": 4.920253441118638e-07, + "loss": 0.5846, + "step": 41316 + }, + { + "epoch": 9.02, + "learning_rate": 4.919161022503824e-07, + "loss": 0.9556, + "step": 41317 + }, + { + "epoch": 9.02, + "learning_rate": 4.918068603889011e-07, + "loss": 1.5101, + "step": 41318 + }, + { + "epoch": 9.02, + "learning_rate": 4.916976185274198e-07, + "loss": 0.7086, + "step": 41319 + }, + { + "epoch": 9.02, + "learning_rate": 4.915883766659384e-07, + "loss": 0.5948, + "step": 41320 + }, + { + "epoch": 9.02, + "learning_rate": 4.914791348044571e-07, + "loss": 1.92, + "step": 41321 + }, + { + "epoch": 9.02, + "learning_rate": 4.913698929429758e-07, + "loss": 2.4054, + "step": 41322 + }, + { + "epoch": 9.02, + "learning_rate": 4.912606510814945e-07, + "loss": 0.2276, + "step": 41323 + }, + { + "epoch": 9.02, + "learning_rate": 4.911514092200132e-07, + "loss": 1.2706, + "step": 41324 + }, + { + "epoch": 9.02, + "learning_rate": 4.910421673585318e-07, + "loss": 1.1378, + "step": 41325 + }, + { + "epoch": 9.02, + "learning_rate": 4.909329254970505e-07, + "loss": 0.2384, + "step": 41326 + }, + { + "epoch": 9.02, + "learning_rate": 4.908236836355692e-07, + "loss": 2.2742, + "step": 41327 + }, + { + "epoch": 9.02, + "learning_rate": 4.907144417740878e-07, + "loss": 0.1057, + "step": 41328 + }, + { + "epoch": 9.02, + "learning_rate": 4.906051999126066e-07, + "loss": 0.5114, + "step": 41329 + }, + { + "epoch": 9.02, + "learning_rate": 4.904959580511253e-07, + "loss": 1.6447, + "step": 41330 + }, + { + "epoch": 9.02, + "learning_rate": 4.903867161896439e-07, + "loss": 1.4027, + "step": 41331 + }, + { + "epoch": 9.02, + "learning_rate": 4.902774743281626e-07, + "loss": 2.5219, + "step": 41332 + }, + { + "epoch": 9.02, + "learning_rate": 4.901682324666813e-07, + "loss": 1.523, + "step": 41333 + }, + { + "epoch": 9.02, + "learning_rate": 4.900589906051999e-07, + "loss": 0.6388, + "step": 41334 + }, + { + "epoch": 9.02, + "learning_rate": 4.899497487437186e-07, + "loss": 0.8261, + "step": 41335 + }, + { + "epoch": 9.02, + "learning_rate": 4.898405068822374e-07, + "loss": 1.1786, + "step": 41336 + }, + { + "epoch": 9.02, + "learning_rate": 4.89731265020756e-07, + "loss": 2.6377, + "step": 41337 + }, + { + "epoch": 9.02, + "learning_rate": 4.896220231592747e-07, + "loss": 1.4099, + "step": 41338 + }, + { + "epoch": 9.02, + "learning_rate": 4.895127812977934e-07, + "loss": 0.518, + "step": 41339 + }, + { + "epoch": 9.02, + "learning_rate": 4.89403539436312e-07, + "loss": 1.3615, + "step": 41340 + }, + { + "epoch": 9.02, + "learning_rate": 4.892942975748307e-07, + "loss": 1.0059, + "step": 41341 + }, + { + "epoch": 9.02, + "learning_rate": 4.891850557133495e-07, + "loss": 0.0775, + "step": 41342 + }, + { + "epoch": 9.02, + "learning_rate": 4.890758138518681e-07, + "loss": 1.6365, + "step": 41343 + }, + { + "epoch": 9.02, + "learning_rate": 4.889665719903868e-07, + "loss": 1.2636, + "step": 41344 + }, + { + "epoch": 9.02, + "learning_rate": 4.888573301289054e-07, + "loss": 1.5144, + "step": 41345 + }, + { + "epoch": 9.02, + "learning_rate": 4.887480882674241e-07, + "loss": 0.4161, + "step": 41346 + }, + { + "epoch": 9.02, + "learning_rate": 4.886388464059428e-07, + "loss": 0.1505, + "step": 41347 + }, + { + "epoch": 9.02, + "learning_rate": 4.885296045444614e-07, + "loss": 0.6018, + "step": 41348 + }, + { + "epoch": 9.02, + "learning_rate": 4.884203626829802e-07, + "loss": 0.3737, + "step": 41349 + }, + { + "epoch": 9.02, + "learning_rate": 4.883111208214989e-07, + "loss": 0.6698, + "step": 41350 + }, + { + "epoch": 9.02, + "learning_rate": 4.882018789600175e-07, + "loss": 0.8153, + "step": 41351 + }, + { + "epoch": 9.02, + "learning_rate": 4.880926370985362e-07, + "loss": 1.3539, + "step": 41352 + }, + { + "epoch": 9.03, + "learning_rate": 4.879833952370549e-07, + "loss": 1.0996, + "step": 41353 + }, + { + "epoch": 9.03, + "learning_rate": 4.878741533755735e-07, + "loss": 1.6464, + "step": 41354 + }, + { + "epoch": 9.03, + "learning_rate": 4.877649115140922e-07, + "loss": 1.2811, + "step": 41355 + }, + { + "epoch": 9.03, + "learning_rate": 4.87655669652611e-07, + "loss": 1.4183, + "step": 41356 + }, + { + "epoch": 9.03, + "learning_rate": 4.875464277911296e-07, + "loss": 0.0379, + "step": 41357 + }, + { + "epoch": 9.03, + "learning_rate": 4.874371859296483e-07, + "loss": 2.1632, + "step": 41358 + }, + { + "epoch": 9.03, + "learning_rate": 4.87327944068167e-07, + "loss": 1.1927, + "step": 41359 + }, + { + "epoch": 9.03, + "learning_rate": 4.872187022066856e-07, + "loss": 1.5773, + "step": 41360 + }, + { + "epoch": 9.03, + "learning_rate": 4.871094603452043e-07, + "loss": 1.0669, + "step": 41361 + }, + { + "epoch": 9.03, + "learning_rate": 4.87000218483723e-07, + "loss": 1.1463, + "step": 41362 + }, + { + "epoch": 9.03, + "learning_rate": 4.868909766222417e-07, + "loss": 1.1475, + "step": 41363 + }, + { + "epoch": 9.03, + "learning_rate": 4.867817347607604e-07, + "loss": 0.7258, + "step": 41364 + }, + { + "epoch": 9.03, + "learning_rate": 4.866724928992791e-07, + "loss": 0.8632, + "step": 41365 + }, + { + "epoch": 9.03, + "learning_rate": 4.865632510377977e-07, + "loss": 0.3408, + "step": 41366 + }, + { + "epoch": 9.03, + "learning_rate": 4.864540091763164e-07, + "loss": 0.8423, + "step": 41367 + }, + { + "epoch": 9.03, + "learning_rate": 4.86344767314835e-07, + "loss": 0.1224, + "step": 41368 + }, + { + "epoch": 9.03, + "learning_rate": 4.862355254533537e-07, + "loss": 1.9016, + "step": 41369 + }, + { + "epoch": 9.03, + "learning_rate": 4.861262835918725e-07, + "loss": 1.4895, + "step": 41370 + }, + { + "epoch": 9.03, + "learning_rate": 4.860170417303911e-07, + "loss": 1.7119, + "step": 41371 + }, + { + "epoch": 9.03, + "learning_rate": 4.859077998689098e-07, + "loss": 0.4933, + "step": 41372 + }, + { + "epoch": 9.03, + "learning_rate": 4.857985580074285e-07, + "loss": 0.7109, + "step": 41373 + }, + { + "epoch": 9.03, + "learning_rate": 4.856893161459471e-07, + "loss": 1.7316, + "step": 41374 + }, + { + "epoch": 9.03, + "learning_rate": 4.855800742844658e-07, + "loss": 0.3144, + "step": 41375 + }, + { + "epoch": 9.03, + "learning_rate": 4.854708324229846e-07, + "loss": 0.7358, + "step": 41376 + }, + { + "epoch": 9.03, + "learning_rate": 4.853615905615032e-07, + "loss": 1.8345, + "step": 41377 + }, + { + "epoch": 9.03, + "learning_rate": 4.852523487000219e-07, + "loss": 0.7205, + "step": 41378 + }, + { + "epoch": 9.03, + "learning_rate": 4.851431068385406e-07, + "loss": 0.5395, + "step": 41379 + }, + { + "epoch": 9.03, + "learning_rate": 4.850338649770592e-07, + "loss": 1.3179, + "step": 41380 + }, + { + "epoch": 9.03, + "learning_rate": 4.849246231155779e-07, + "loss": 1.5367, + "step": 41381 + }, + { + "epoch": 9.03, + "learning_rate": 4.848153812540966e-07, + "loss": 1.708, + "step": 41382 + }, + { + "epoch": 9.03, + "learning_rate": 4.847061393926153e-07, + "loss": 0.7843, + "step": 41383 + }, + { + "epoch": 9.03, + "learning_rate": 4.84596897531134e-07, + "loss": 0.7486, + "step": 41384 + }, + { + "epoch": 9.03, + "learning_rate": 4.844876556696527e-07, + "loss": 0.0429, + "step": 41385 + }, + { + "epoch": 9.03, + "learning_rate": 4.843784138081713e-07, + "loss": 0.9862, + "step": 41386 + }, + { + "epoch": 9.03, + "learning_rate": 4.8426917194669e-07, + "loss": 2.0164, + "step": 41387 + }, + { + "epoch": 9.03, + "learning_rate": 4.841599300852086e-07, + "loss": 1.9385, + "step": 41388 + }, + { + "epoch": 9.03, + "learning_rate": 4.840506882237273e-07, + "loss": 0.4996, + "step": 41389 + }, + { + "epoch": 9.03, + "learning_rate": 4.839414463622461e-07, + "loss": 1.4217, + "step": 41390 + }, + { + "epoch": 9.03, + "learning_rate": 4.838322045007647e-07, + "loss": 0.1824, + "step": 41391 + }, + { + "epoch": 9.03, + "learning_rate": 4.837229626392834e-07, + "loss": 0.6739, + "step": 41392 + }, + { + "epoch": 9.03, + "learning_rate": 4.836137207778021e-07, + "loss": 1.0756, + "step": 41393 + }, + { + "epoch": 9.03, + "learning_rate": 4.835044789163207e-07, + "loss": 0.6818, + "step": 41394 + }, + { + "epoch": 9.03, + "learning_rate": 4.833952370548394e-07, + "loss": 1.8815, + "step": 41395 + }, + { + "epoch": 9.03, + "learning_rate": 4.832859951933582e-07, + "loss": 0.7621, + "step": 41396 + }, + { + "epoch": 9.03, + "learning_rate": 4.831767533318768e-07, + "loss": 0.7545, + "step": 41397 + }, + { + "epoch": 9.03, + "learning_rate": 4.830675114703955e-07, + "loss": 0.1444, + "step": 41398 + }, + { + "epoch": 9.04, + "learning_rate": 4.829582696089142e-07, + "loss": 2.064, + "step": 41399 + }, + { + "epoch": 9.04, + "learning_rate": 4.828490277474328e-07, + "loss": 2.0697, + "step": 41400 + }, + { + "epoch": 9.04, + "learning_rate": 4.827397858859515e-07, + "loss": 1.7997, + "step": 41401 + }, + { + "epoch": 9.04, + "learning_rate": 4.826305440244702e-07, + "loss": 0.3819, + "step": 41402 + }, + { + "epoch": 9.04, + "learning_rate": 4.825213021629889e-07, + "loss": 1.4805, + "step": 41403 + }, + { + "epoch": 9.04, + "learning_rate": 4.824120603015076e-07, + "loss": 0.8857, + "step": 41404 + }, + { + "epoch": 9.04, + "learning_rate": 4.823028184400263e-07, + "loss": 2.1994, + "step": 41405 + }, + { + "epoch": 9.04, + "learning_rate": 4.821935765785449e-07, + "loss": 0.9807, + "step": 41406 + }, + { + "epoch": 9.04, + "learning_rate": 4.820843347170636e-07, + "loss": 2.2831, + "step": 41407 + }, + { + "epoch": 9.04, + "learning_rate": 4.819750928555823e-07, + "loss": 1.3655, + "step": 41408 + }, + { + "epoch": 9.04, + "learning_rate": 4.818658509941009e-07, + "loss": 0.3285, + "step": 41409 + }, + { + "epoch": 9.04, + "learning_rate": 4.817566091326197e-07, + "loss": 1.1359, + "step": 41410 + }, + { + "epoch": 9.04, + "learning_rate": 4.816473672711383e-07, + "loss": 1.1517, + "step": 41411 + }, + { + "epoch": 9.04, + "learning_rate": 4.81538125409657e-07, + "loss": 0.9878, + "step": 41412 + }, + { + "epoch": 9.04, + "learning_rate": 4.814288835481757e-07, + "loss": 1.0659, + "step": 41413 + }, + { + "epoch": 9.04, + "learning_rate": 4.813196416866943e-07, + "loss": 0.4546, + "step": 41414 + }, + { + "epoch": 9.04, + "learning_rate": 4.812103998252131e-07, + "loss": 0.9593, + "step": 41415 + }, + { + "epoch": 9.04, + "learning_rate": 4.811011579637317e-07, + "loss": 2.7388, + "step": 41416 + }, + { + "epoch": 9.04, + "learning_rate": 4.809919161022504e-07, + "loss": 1.7945, + "step": 41417 + }, + { + "epoch": 9.04, + "learning_rate": 4.808826742407692e-07, + "loss": 0.3137, + "step": 41418 + }, + { + "epoch": 9.04, + "learning_rate": 4.807734323792878e-07, + "loss": 1.149, + "step": 41419 + }, + { + "epoch": 9.04, + "learning_rate": 4.806641905178064e-07, + "loss": 1.4942, + "step": 41420 + }, + { + "epoch": 9.04, + "learning_rate": 4.805549486563251e-07, + "loss": 2.3102, + "step": 41421 + }, + { + "epoch": 9.04, + "learning_rate": 4.804457067948438e-07, + "loss": 1.0362, + "step": 41422 + }, + { + "epoch": 9.04, + "learning_rate": 4.803364649333624e-07, + "loss": 1.189, + "step": 41423 + }, + { + "epoch": 9.04, + "learning_rate": 4.802272230718812e-07, + "loss": 0.6151, + "step": 41424 + }, + { + "epoch": 9.04, + "learning_rate": 4.801179812103999e-07, + "loss": 1.7417, + "step": 41425 + }, + { + "epoch": 9.04, + "learning_rate": 4.800087393489186e-07, + "loss": 1.1217, + "step": 41426 + }, + { + "epoch": 9.04, + "learning_rate": 4.798994974874372e-07, + "loss": 1.4377, + "step": 41427 + }, + { + "epoch": 9.04, + "learning_rate": 4.797902556259559e-07, + "loss": 1.5505, + "step": 41428 + }, + { + "epoch": 9.04, + "learning_rate": 4.796810137644746e-07, + "loss": 1.0941, + "step": 41429 + }, + { + "epoch": 9.04, + "learning_rate": 4.795717719029933e-07, + "loss": 1.8792, + "step": 41430 + }, + { + "epoch": 9.04, + "learning_rate": 4.794625300415119e-07, + "loss": 1.1658, + "step": 41431 + }, + { + "epoch": 9.04, + "learning_rate": 4.793532881800307e-07, + "loss": 1.2108, + "step": 41432 + }, + { + "epoch": 9.04, + "learning_rate": 4.792440463185493e-07, + "loss": 1.6695, + "step": 41433 + }, + { + "epoch": 9.04, + "learning_rate": 4.791348044570679e-07, + "loss": 2.0175, + "step": 41434 + }, + { + "epoch": 9.04, + "learning_rate": 4.790255625955867e-07, + "loss": 0.7881, + "step": 41435 + }, + { + "epoch": 9.04, + "learning_rate": 4.789163207341053e-07, + "loss": 1.5781, + "step": 41436 + }, + { + "epoch": 9.04, + "learning_rate": 4.788070788726241e-07, + "loss": 2.8242, + "step": 41437 + }, + { + "epoch": 9.04, + "learning_rate": 4.786978370111428e-07, + "loss": 0.5127, + "step": 41438 + }, + { + "epoch": 9.04, + "learning_rate": 4.785885951496614e-07, + "loss": 1.1449, + "step": 41439 + }, + { + "epoch": 9.04, + "learning_rate": 4.784793532881801e-07, + "loss": 1.4901, + "step": 41440 + }, + { + "epoch": 9.04, + "learning_rate": 4.783701114266987e-07, + "loss": 1.6558, + "step": 41441 + }, + { + "epoch": 9.04, + "learning_rate": 4.782608695652174e-07, + "loss": 0.6671, + "step": 41442 + }, + { + "epoch": 9.04, + "learning_rate": 4.781516277037362e-07, + "loss": 0.5345, + "step": 41443 + }, + { + "epoch": 9.04, + "learning_rate": 4.780423858422548e-07, + "loss": 0.512, + "step": 41444 + }, + { + "epoch": 9.05, + "learning_rate": 4.779331439807735e-07, + "loss": 0.1728, + "step": 41445 + }, + { + "epoch": 9.05, + "learning_rate": 4.778239021192922e-07, + "loss": 0.9585, + "step": 41446 + }, + { + "epoch": 9.05, + "learning_rate": 4.777146602578108e-07, + "loss": 0.6965, + "step": 41447 + }, + { + "epoch": 9.05, + "learning_rate": 4.776054183963295e-07, + "loss": 1.0501, + "step": 41448 + }, + { + "epoch": 9.05, + "learning_rate": 4.774961765348482e-07, + "loss": 1.5873, + "step": 41449 + }, + { + "epoch": 9.05, + "learning_rate": 4.773869346733669e-07, + "loss": 0.2873, + "step": 41450 + }, + { + "epoch": 9.05, + "learning_rate": 4.772776928118856e-07, + "loss": 0.4013, + "step": 41451 + }, + { + "epoch": 9.05, + "learning_rate": 4.771684509504043e-07, + "loss": 1.1977, + "step": 41452 + }, + { + "epoch": 9.05, + "learning_rate": 4.770592090889229e-07, + "loss": 0.7923, + "step": 41453 + }, + { + "epoch": 9.05, + "learning_rate": 4.769499672274416e-07, + "loss": 0.926, + "step": 41454 + }, + { + "epoch": 9.05, + "learning_rate": 4.768407253659603e-07, + "loss": 0.5959, + "step": 41455 + }, + { + "epoch": 9.05, + "learning_rate": 4.7673148350447894e-07, + "loss": 1.4551, + "step": 41456 + }, + { + "epoch": 9.05, + "learning_rate": 4.766222416429976e-07, + "loss": 0.8818, + "step": 41457 + }, + { + "epoch": 9.05, + "learning_rate": 4.765129997815163e-07, + "loss": 0.8925, + "step": 41458 + }, + { + "epoch": 9.05, + "learning_rate": 4.76403757920035e-07, + "loss": 0.6817, + "step": 41459 + }, + { + "epoch": 9.05, + "learning_rate": 4.7629451605855365e-07, + "loss": 1.5567, + "step": 41460 + }, + { + "epoch": 9.05, + "learning_rate": 4.7618527419707237e-07, + "loss": 1.8042, + "step": 41461 + }, + { + "epoch": 9.05, + "learning_rate": 4.7607603233559103e-07, + "loss": 0.9349, + "step": 41462 + }, + { + "epoch": 9.05, + "learning_rate": 4.759667904741097e-07, + "loss": 0.9146, + "step": 41463 + }, + { + "epoch": 9.05, + "learning_rate": 4.7585754861262836e-07, + "loss": 2.8786, + "step": 41464 + }, + { + "epoch": 9.05, + "learning_rate": 4.757483067511471e-07, + "loss": 0.5488, + "step": 41465 + }, + { + "epoch": 9.05, + "learning_rate": 4.7563906488966574e-07, + "loss": 0.9251, + "step": 41466 + }, + { + "epoch": 9.05, + "learning_rate": 4.755298230281844e-07, + "loss": 1.6876, + "step": 41467 + }, + { + "epoch": 9.05, + "learning_rate": 4.754205811667031e-07, + "loss": 1.3337, + "step": 41468 + }, + { + "epoch": 9.05, + "learning_rate": 4.753113393052218e-07, + "loss": 0.7767, + "step": 41469 + }, + { + "epoch": 9.05, + "learning_rate": 4.7520209744374045e-07, + "loss": 1.8085, + "step": 41470 + }, + { + "epoch": 9.05, + "learning_rate": 4.7509285558225916e-07, + "loss": 0.5364, + "step": 41471 + }, + { + "epoch": 9.05, + "learning_rate": 4.7498361372077783e-07, + "loss": 0.2559, + "step": 41472 + }, + { + "epoch": 9.05, + "learning_rate": 4.748743718592965e-07, + "loss": 1.9863, + "step": 41473 + }, + { + "epoch": 9.05, + "learning_rate": 4.7476512999781516e-07, + "loss": 1.3385, + "step": 41474 + }, + { + "epoch": 9.05, + "learning_rate": 4.7465588813633387e-07, + "loss": 1.771, + "step": 41475 + }, + { + "epoch": 9.05, + "learning_rate": 4.7454664627485254e-07, + "loss": 1.3887, + "step": 41476 + }, + { + "epoch": 9.05, + "learning_rate": 4.744374044133712e-07, + "loss": 0.7351, + "step": 41477 + }, + { + "epoch": 9.05, + "learning_rate": 4.7432816255188997e-07, + "loss": 2.3022, + "step": 41478 + }, + { + "epoch": 9.05, + "learning_rate": 4.742189206904086e-07, + "loss": 1.5882, + "step": 41479 + }, + { + "epoch": 9.05, + "learning_rate": 4.7410967882892725e-07, + "loss": 1.1017, + "step": 41480 + }, + { + "epoch": 9.05, + "learning_rate": 4.74000436967446e-07, + "loss": 1.2067, + "step": 41481 + }, + { + "epoch": 9.05, + "learning_rate": 4.7389119510596463e-07, + "loss": 2.1665, + "step": 41482 + }, + { + "epoch": 9.05, + "learning_rate": 4.737819532444833e-07, + "loss": 2.4568, + "step": 41483 + }, + { + "epoch": 9.05, + "learning_rate": 4.7367271138300195e-07, + "loss": 0.1885, + "step": 41484 + }, + { + "epoch": 9.05, + "learning_rate": 4.735634695215207e-07, + "loss": 0.9663, + "step": 41485 + }, + { + "epoch": 9.05, + "learning_rate": 4.7345422766003934e-07, + "loss": 0.9451, + "step": 41486 + }, + { + "epoch": 9.05, + "learning_rate": 4.73344985798558e-07, + "loss": 1.5352, + "step": 41487 + }, + { + "epoch": 9.05, + "learning_rate": 4.7323574393707677e-07, + "loss": 1.335, + "step": 41488 + }, + { + "epoch": 9.05, + "learning_rate": 4.7312650207559543e-07, + "loss": 0.5574, + "step": 41489 + }, + { + "epoch": 9.05, + "learning_rate": 4.7301726021411404e-07, + "loss": 1.4423, + "step": 41490 + }, + { + "epoch": 9.06, + "learning_rate": 4.729080183526328e-07, + "loss": 0.3576, + "step": 41491 + }, + { + "epoch": 9.06, + "learning_rate": 4.727987764911515e-07, + "loss": 1.2358, + "step": 41492 + }, + { + "epoch": 9.06, + "learning_rate": 4.726895346296701e-07, + "loss": 1.7583, + "step": 41493 + }, + { + "epoch": 9.06, + "learning_rate": 4.7258029276818886e-07, + "loss": 1.9282, + "step": 41494 + }, + { + "epoch": 9.06, + "learning_rate": 4.724710509067075e-07, + "loss": 0.0468, + "step": 41495 + }, + { + "epoch": 9.06, + "learning_rate": 4.723618090452262e-07, + "loss": 0.874, + "step": 41496 + }, + { + "epoch": 9.06, + "learning_rate": 4.722525671837448e-07, + "loss": 2.812, + "step": 41497 + }, + { + "epoch": 9.06, + "learning_rate": 4.7214332532226357e-07, + "loss": 1.5697, + "step": 41498 + }, + { + "epoch": 9.06, + "learning_rate": 4.7203408346078223e-07, + "loss": 1.2648, + "step": 41499 + }, + { + "epoch": 9.06, + "learning_rate": 4.7192484159930084e-07, + "loss": 0.6661, + "step": 41500 + }, + { + "epoch": 9.06, + "learning_rate": 4.718155997378196e-07, + "loss": 2.8066, + "step": 41501 + }, + { + "epoch": 9.06, + "learning_rate": 4.717063578763383e-07, + "loss": 0.6259, + "step": 41502 + }, + { + "epoch": 9.06, + "learning_rate": 4.7159711601485694e-07, + "loss": 2.2199, + "step": 41503 + }, + { + "epoch": 9.06, + "learning_rate": 4.7148787415337566e-07, + "loss": 1.7085, + "step": 41504 + }, + { + "epoch": 9.06, + "learning_rate": 4.713786322918943e-07, + "loss": 0.2267, + "step": 41505 + }, + { + "epoch": 9.06, + "learning_rate": 4.71269390430413e-07, + "loss": 2.1128, + "step": 41506 + }, + { + "epoch": 9.06, + "learning_rate": 4.7116014856893165e-07, + "loss": 0.1432, + "step": 41507 + }, + { + "epoch": 9.06, + "learning_rate": 4.7105090670745037e-07, + "loss": 1.7678, + "step": 41508 + }, + { + "epoch": 9.06, + "learning_rate": 4.7094166484596903e-07, + "loss": 2.4164, + "step": 41509 + }, + { + "epoch": 9.06, + "learning_rate": 4.708324229844877e-07, + "loss": 2.2084, + "step": 41510 + }, + { + "epoch": 9.06, + "learning_rate": 4.707231811230064e-07, + "loss": 1.6025, + "step": 41511 + }, + { + "epoch": 9.06, + "learning_rate": 4.706139392615251e-07, + "loss": 0.3681, + "step": 41512 + }, + { + "epoch": 9.06, + "learning_rate": 4.7050469740004374e-07, + "loss": 1.8563, + "step": 41513 + }, + { + "epoch": 9.06, + "learning_rate": 4.7039545553856245e-07, + "loss": 1.2979, + "step": 41514 + }, + { + "epoch": 9.06, + "learning_rate": 4.702862136770811e-07, + "loss": 1.2661, + "step": 41515 + }, + { + "epoch": 9.06, + "learning_rate": 4.701769718155998e-07, + "loss": 1.0031, + "step": 41516 + }, + { + "epoch": 9.06, + "learning_rate": 4.7006772995411845e-07, + "loss": 0.6239, + "step": 41517 + }, + { + "epoch": 9.06, + "learning_rate": 4.6995848809263716e-07, + "loss": 0.287, + "step": 41518 + }, + { + "epoch": 9.06, + "learning_rate": 4.6984924623115583e-07, + "loss": 0.698, + "step": 41519 + }, + { + "epoch": 9.06, + "learning_rate": 4.697400043696745e-07, + "loss": 0.3793, + "step": 41520 + }, + { + "epoch": 9.06, + "learning_rate": 4.696307625081932e-07, + "loss": 1.3299, + "step": 41521 + }, + { + "epoch": 9.06, + "learning_rate": 4.6952152064671187e-07, + "loss": 1.123, + "step": 41522 + }, + { + "epoch": 9.06, + "learning_rate": 4.6941227878523054e-07, + "loss": 1.1977, + "step": 41523 + }, + { + "epoch": 9.06, + "learning_rate": 4.6930303692374925e-07, + "loss": 1.3742, + "step": 41524 + }, + { + "epoch": 9.06, + "learning_rate": 4.691937950622679e-07, + "loss": 1.2204, + "step": 41525 + }, + { + "epoch": 9.06, + "learning_rate": 4.690845532007866e-07, + "loss": 0.995, + "step": 41526 + }, + { + "epoch": 9.06, + "learning_rate": 4.689753113393053e-07, + "loss": 2.1437, + "step": 41527 + }, + { + "epoch": 9.06, + "learning_rate": 4.6886606947782396e-07, + "loss": 0.7674, + "step": 41528 + }, + { + "epoch": 9.06, + "learning_rate": 4.687568276163426e-07, + "loss": 1.252, + "step": 41529 + }, + { + "epoch": 9.06, + "learning_rate": 4.686475857548613e-07, + "loss": 0.7348, + "step": 41530 + }, + { + "epoch": 9.06, + "learning_rate": 4.6853834389338e-07, + "loss": 0.2897, + "step": 41531 + }, + { + "epoch": 9.06, + "learning_rate": 4.6842910203189867e-07, + "loss": 1.7658, + "step": 41532 + }, + { + "epoch": 9.06, + "learning_rate": 4.6831986017041733e-07, + "loss": 1.4301, + "step": 41533 + }, + { + "epoch": 9.06, + "learning_rate": 4.6821061830893605e-07, + "loss": 0.5838, + "step": 41534 + }, + { + "epoch": 9.06, + "learning_rate": 4.681013764474547e-07, + "loss": 0.6343, + "step": 41535 + }, + { + "epoch": 9.07, + "learning_rate": 4.679921345859734e-07, + "loss": 1.9, + "step": 41536 + }, + { + "epoch": 9.07, + "learning_rate": 4.678828927244921e-07, + "loss": 0.8189, + "step": 41537 + }, + { + "epoch": 9.07, + "learning_rate": 4.6777365086301076e-07, + "loss": 1.2856, + "step": 41538 + }, + { + "epoch": 9.07, + "learning_rate": 4.676644090015294e-07, + "loss": 0.4508, + "step": 41539 + }, + { + "epoch": 9.07, + "learning_rate": 4.675551671400481e-07, + "loss": 2.2471, + "step": 41540 + }, + { + "epoch": 9.07, + "learning_rate": 4.674459252785668e-07, + "loss": 0.0354, + "step": 41541 + }, + { + "epoch": 9.07, + "learning_rate": 4.6733668341708547e-07, + "loss": 0.5581, + "step": 41542 + }, + { + "epoch": 9.07, + "learning_rate": 4.6722744155560413e-07, + "loss": 1.3734, + "step": 41543 + }, + { + "epoch": 9.07, + "learning_rate": 4.6711819969412285e-07, + "loss": 0.6887, + "step": 41544 + }, + { + "epoch": 9.07, + "learning_rate": 4.670089578326415e-07, + "loss": 0.168, + "step": 41545 + }, + { + "epoch": 9.07, + "learning_rate": 4.668997159711602e-07, + "loss": 1.3717, + "step": 41546 + }, + { + "epoch": 9.07, + "learning_rate": 4.667904741096789e-07, + "loss": 1.002, + "step": 41547 + }, + { + "epoch": 9.07, + "learning_rate": 4.6668123224819756e-07, + "loss": 1.3445, + "step": 41548 + }, + { + "epoch": 9.07, + "learning_rate": 4.665719903867162e-07, + "loss": 0.8764, + "step": 41549 + }, + { + "epoch": 9.07, + "learning_rate": 4.664627485252349e-07, + "loss": 0.1529, + "step": 41550 + }, + { + "epoch": 9.07, + "learning_rate": 4.663535066637536e-07, + "loss": 0.6265, + "step": 41551 + }, + { + "epoch": 9.07, + "learning_rate": 4.6624426480227227e-07, + "loss": 1.1136, + "step": 41552 + }, + { + "epoch": 9.07, + "learning_rate": 4.6613502294079093e-07, + "loss": 0.6044, + "step": 41553 + }, + { + "epoch": 9.07, + "learning_rate": 4.6602578107930965e-07, + "loss": 1.315, + "step": 41554 + }, + { + "epoch": 9.07, + "learning_rate": 4.659165392178283e-07, + "loss": 1.3581, + "step": 41555 + }, + { + "epoch": 9.07, + "learning_rate": 4.65807297356347e-07, + "loss": 1.03, + "step": 41556 + }, + { + "epoch": 9.07, + "learning_rate": 4.656980554948657e-07, + "loss": 2.2056, + "step": 41557 + }, + { + "epoch": 9.07, + "learning_rate": 4.6558881363338436e-07, + "loss": 1.4144, + "step": 41558 + }, + { + "epoch": 9.07, + "learning_rate": 4.65479571771903e-07, + "loss": 0.1701, + "step": 41559 + }, + { + "epoch": 9.07, + "learning_rate": 4.653703299104217e-07, + "loss": 2.143, + "step": 41560 + }, + { + "epoch": 9.07, + "learning_rate": 4.652610880489404e-07, + "loss": 0.3726, + "step": 41561 + }, + { + "epoch": 9.07, + "learning_rate": 4.6515184618745906e-07, + "loss": 1.6945, + "step": 41562 + }, + { + "epoch": 9.07, + "learning_rate": 4.6504260432597773e-07, + "loss": 0.567, + "step": 41563 + }, + { + "epoch": 9.07, + "learning_rate": 4.6493336246449644e-07, + "loss": 0.727, + "step": 41564 + }, + { + "epoch": 9.07, + "learning_rate": 4.648241206030151e-07, + "loss": 0.6034, + "step": 41565 + }, + { + "epoch": 9.07, + "learning_rate": 4.6471487874153377e-07, + "loss": 1.2718, + "step": 41566 + }, + { + "epoch": 9.07, + "learning_rate": 4.646056368800525e-07, + "loss": 2.5789, + "step": 41567 + }, + { + "epoch": 9.07, + "learning_rate": 4.6449639501857115e-07, + "loss": 0.7908, + "step": 41568 + }, + { + "epoch": 9.07, + "learning_rate": 4.643871531570898e-07, + "loss": 1.7232, + "step": 41569 + }, + { + "epoch": 9.07, + "learning_rate": 4.6427791129560853e-07, + "loss": 0.7558, + "step": 41570 + }, + { + "epoch": 9.07, + "learning_rate": 4.641686694341272e-07, + "loss": 1.3235, + "step": 41571 + }, + { + "epoch": 9.07, + "learning_rate": 4.6405942757264586e-07, + "loss": 3.5295, + "step": 41572 + }, + { + "epoch": 9.07, + "learning_rate": 4.639501857111645e-07, + "loss": 0.7194, + "step": 41573 + }, + { + "epoch": 9.07, + "learning_rate": 4.6384094384968324e-07, + "loss": 1.8963, + "step": 41574 + }, + { + "epoch": 9.07, + "learning_rate": 4.637317019882019e-07, + "loss": 1.2168, + "step": 41575 + }, + { + "epoch": 9.07, + "learning_rate": 4.6362246012672057e-07, + "loss": 0.6873, + "step": 41576 + }, + { + "epoch": 9.07, + "learning_rate": 4.635132182652393e-07, + "loss": 1.9012, + "step": 41577 + }, + { + "epoch": 9.07, + "learning_rate": 4.6340397640375795e-07, + "loss": 2.1772, + "step": 41578 + }, + { + "epoch": 9.07, + "learning_rate": 4.632947345422766e-07, + "loss": 0.9331, + "step": 41579 + }, + { + "epoch": 9.07, + "learning_rate": 4.6318549268079533e-07, + "loss": 1.0334, + "step": 41580 + }, + { + "epoch": 9.07, + "learning_rate": 4.63076250819314e-07, + "loss": 0.906, + "step": 41581 + }, + { + "epoch": 9.08, + "learning_rate": 4.6296700895783266e-07, + "loss": 1.3414, + "step": 41582 + }, + { + "epoch": 9.08, + "learning_rate": 4.628577670963513e-07, + "loss": 0.4797, + "step": 41583 + }, + { + "epoch": 9.08, + "learning_rate": 4.6274852523487004e-07, + "loss": 1.9956, + "step": 41584 + }, + { + "epoch": 9.08, + "learning_rate": 4.626392833733887e-07, + "loss": 2.0046, + "step": 41585 + }, + { + "epoch": 9.08, + "learning_rate": 4.6253004151190737e-07, + "loss": 0.4759, + "step": 41586 + }, + { + "epoch": 9.08, + "learning_rate": 4.624207996504261e-07, + "loss": 0.402, + "step": 41587 + }, + { + "epoch": 9.08, + "learning_rate": 4.6231155778894475e-07, + "loss": 2.0197, + "step": 41588 + }, + { + "epoch": 9.08, + "learning_rate": 4.622023159274634e-07, + "loss": 0.9652, + "step": 41589 + }, + { + "epoch": 9.08, + "learning_rate": 4.6209307406598213e-07, + "loss": 1.0612, + "step": 41590 + }, + { + "epoch": 9.08, + "learning_rate": 4.619838322045008e-07, + "loss": 0.972, + "step": 41591 + }, + { + "epoch": 9.08, + "learning_rate": 4.6187459034301946e-07, + "loss": 1.3955, + "step": 41592 + }, + { + "epoch": 9.08, + "learning_rate": 4.617653484815381e-07, + "loss": 1.9188, + "step": 41593 + }, + { + "epoch": 9.08, + "learning_rate": 4.6165610662005684e-07, + "loss": 1.5782, + "step": 41594 + }, + { + "epoch": 9.08, + "learning_rate": 4.615468647585755e-07, + "loss": 0.6778, + "step": 41595 + }, + { + "epoch": 9.08, + "learning_rate": 4.6143762289709417e-07, + "loss": 0.4428, + "step": 41596 + }, + { + "epoch": 9.08, + "learning_rate": 4.613283810356129e-07, + "loss": 0.0887, + "step": 41597 + }, + { + "epoch": 9.08, + "learning_rate": 4.6121913917413155e-07, + "loss": 0.4482, + "step": 41598 + }, + { + "epoch": 9.08, + "learning_rate": 4.611098973126502e-07, + "loss": 2.3802, + "step": 41599 + }, + { + "epoch": 9.08, + "learning_rate": 4.6100065545116893e-07, + "loss": 0.5505, + "step": 41600 + }, + { + "epoch": 9.08, + "learning_rate": 4.608914135896876e-07, + "loss": 0.0376, + "step": 41601 + }, + { + "epoch": 9.08, + "learning_rate": 4.6078217172820626e-07, + "loss": 2.1236, + "step": 41602 + }, + { + "epoch": 9.08, + "learning_rate": 4.606729298667249e-07, + "loss": 1.7654, + "step": 41603 + }, + { + "epoch": 9.08, + "learning_rate": 4.6056368800524364e-07, + "loss": 1.3017, + "step": 41604 + }, + { + "epoch": 9.08, + "learning_rate": 4.604544461437623e-07, + "loss": 0.1612, + "step": 41605 + }, + { + "epoch": 9.08, + "learning_rate": 4.6034520428228097e-07, + "loss": 1.4146, + "step": 41606 + }, + { + "epoch": 9.08, + "learning_rate": 4.6023596242079973e-07, + "loss": 1.1088, + "step": 41607 + }, + { + "epoch": 9.08, + "learning_rate": 4.6012672055931835e-07, + "loss": 1.1112, + "step": 41608 + }, + { + "epoch": 9.08, + "learning_rate": 4.60017478697837e-07, + "loss": 2.0646, + "step": 41609 + }, + { + "epoch": 9.08, + "learning_rate": 4.599082368363558e-07, + "loss": 1.6101, + "step": 41610 + }, + { + "epoch": 9.08, + "learning_rate": 4.597989949748744e-07, + "loss": 1.825, + "step": 41611 + }, + { + "epoch": 9.08, + "learning_rate": 4.5968975311339305e-07, + "loss": 0.1106, + "step": 41612 + }, + { + "epoch": 9.08, + "learning_rate": 4.595805112519118e-07, + "loss": 0.0677, + "step": 41613 + }, + { + "epoch": 9.08, + "learning_rate": 4.594712693904305e-07, + "loss": 1.3983, + "step": 41614 + }, + { + "epoch": 9.08, + "learning_rate": 4.593620275289491e-07, + "loss": 2.7076, + "step": 41615 + }, + { + "epoch": 9.08, + "learning_rate": 4.5925278566746776e-07, + "loss": 4.749, + "step": 41616 + }, + { + "epoch": 9.08, + "learning_rate": 4.5914354380598653e-07, + "loss": 1.1271, + "step": 41617 + }, + { + "epoch": 9.08, + "learning_rate": 4.5903430194450514e-07, + "loss": 0.8863, + "step": 41618 + }, + { + "epoch": 9.08, + "learning_rate": 4.589250600830238e-07, + "loss": 0.5095, + "step": 41619 + }, + { + "epoch": 9.08, + "learning_rate": 4.588158182215426e-07, + "loss": 0.6055, + "step": 41620 + }, + { + "epoch": 9.08, + "learning_rate": 4.5870657636006124e-07, + "loss": 1.9858, + "step": 41621 + }, + { + "epoch": 9.08, + "learning_rate": 4.5859733449857985e-07, + "loss": 0.9616, + "step": 41622 + }, + { + "epoch": 9.08, + "learning_rate": 4.584880926370986e-07, + "loss": 1.436, + "step": 41623 + }, + { + "epoch": 9.08, + "learning_rate": 4.583788507756173e-07, + "loss": 1.3404, + "step": 41624 + }, + { + "epoch": 9.08, + "learning_rate": 4.5826960891413595e-07, + "loss": 0.3673, + "step": 41625 + }, + { + "epoch": 9.08, + "learning_rate": 4.5816036705265456e-07, + "loss": 0.5094, + "step": 41626 + }, + { + "epoch": 9.08, + "learning_rate": 4.5805112519117333e-07, + "loss": 0.4769, + "step": 41627 + }, + { + "epoch": 9.09, + "learning_rate": 4.57941883329692e-07, + "loss": 1.3175, + "step": 41628 + }, + { + "epoch": 9.09, + "learning_rate": 4.578326414682106e-07, + "loss": 1.3182, + "step": 41629 + }, + { + "epoch": 9.09, + "learning_rate": 4.577233996067294e-07, + "loss": 0.7899, + "step": 41630 + }, + { + "epoch": 9.09, + "learning_rate": 4.5761415774524804e-07, + "loss": 0.1308, + "step": 41631 + }, + { + "epoch": 9.09, + "learning_rate": 4.575049158837667e-07, + "loss": 0.8769, + "step": 41632 + }, + { + "epoch": 9.09, + "learning_rate": 4.573956740222854e-07, + "loss": 0.4123, + "step": 41633 + }, + { + "epoch": 9.09, + "learning_rate": 4.572864321608041e-07, + "loss": 0.3037, + "step": 41634 + }, + { + "epoch": 9.09, + "learning_rate": 4.5717719029932275e-07, + "loss": 0.4247, + "step": 41635 + }, + { + "epoch": 9.09, + "learning_rate": 4.5706794843784136e-07, + "loss": 1.5186, + "step": 41636 + }, + { + "epoch": 9.09, + "learning_rate": 4.5695870657636013e-07, + "loss": 1.6667, + "step": 41637 + }, + { + "epoch": 9.09, + "learning_rate": 4.568494647148788e-07, + "loss": 1.1797, + "step": 41638 + }, + { + "epoch": 9.09, + "learning_rate": 4.5674022285339746e-07, + "loss": 2.8522, + "step": 41639 + }, + { + "epoch": 9.09, + "learning_rate": 4.566309809919162e-07, + "loss": 0.3138, + "step": 41640 + }, + { + "epoch": 9.09, + "learning_rate": 4.5652173913043484e-07, + "loss": 0.3921, + "step": 41641 + }, + { + "epoch": 9.09, + "learning_rate": 4.564124972689535e-07, + "loss": 0.2562, + "step": 41642 + }, + { + "epoch": 9.09, + "learning_rate": 4.563032554074722e-07, + "loss": 4.5453, + "step": 41643 + }, + { + "epoch": 9.09, + "learning_rate": 4.561940135459909e-07, + "loss": 0.6296, + "step": 41644 + }, + { + "epoch": 9.09, + "learning_rate": 4.5608477168450955e-07, + "loss": 0.3659, + "step": 41645 + }, + { + "epoch": 9.09, + "learning_rate": 4.559755298230282e-07, + "loss": 0.7665, + "step": 41646 + }, + { + "epoch": 9.09, + "learning_rate": 4.5586628796154693e-07, + "loss": 0.1581, + "step": 41647 + }, + { + "epoch": 9.09, + "learning_rate": 4.557570461000656e-07, + "loss": 0.1394, + "step": 41648 + }, + { + "epoch": 9.09, + "learning_rate": 4.5564780423858426e-07, + "loss": 0.8717, + "step": 41649 + }, + { + "epoch": 9.09, + "learning_rate": 4.5553856237710297e-07, + "loss": 0.9149, + "step": 41650 + }, + { + "epoch": 9.09, + "learning_rate": 4.5542932051562164e-07, + "loss": 1.2828, + "step": 41651 + }, + { + "epoch": 9.09, + "learning_rate": 4.553200786541403e-07, + "loss": 1.4378, + "step": 41652 + }, + { + "epoch": 9.09, + "learning_rate": 4.55210836792659e-07, + "loss": 1.5464, + "step": 41653 + }, + { + "epoch": 9.09, + "learning_rate": 4.551015949311777e-07, + "loss": 0.9945, + "step": 41654 + }, + { + "epoch": 9.09, + "learning_rate": 4.5499235306969634e-07, + "loss": 1.2046, + "step": 41655 + }, + { + "epoch": 9.09, + "learning_rate": 4.5488311120821506e-07, + "loss": 0.6594, + "step": 41656 + }, + { + "epoch": 9.09, + "learning_rate": 4.547738693467337e-07, + "loss": 1.725, + "step": 41657 + }, + { + "epoch": 9.09, + "learning_rate": 4.546646274852524e-07, + "loss": 0.1378, + "step": 41658 + }, + { + "epoch": 9.09, + "learning_rate": 4.5455538562377105e-07, + "loss": 2.2806, + "step": 41659 + }, + { + "epoch": 9.09, + "learning_rate": 4.5444614376228977e-07, + "loss": 0.6438, + "step": 41660 + }, + { + "epoch": 9.09, + "learning_rate": 4.5433690190080843e-07, + "loss": 0.2801, + "step": 41661 + }, + { + "epoch": 9.09, + "learning_rate": 4.542276600393271e-07, + "loss": 1.1781, + "step": 41662 + }, + { + "epoch": 9.09, + "learning_rate": 4.541184181778458e-07, + "loss": 1.0287, + "step": 41663 + }, + { + "epoch": 9.09, + "learning_rate": 4.540091763163645e-07, + "loss": 1.5482, + "step": 41664 + }, + { + "epoch": 9.09, + "learning_rate": 4.5389993445488314e-07, + "loss": 0.6565, + "step": 41665 + }, + { + "epoch": 9.09, + "learning_rate": 4.5379069259340186e-07, + "loss": 0.4691, + "step": 41666 + }, + { + "epoch": 9.09, + "learning_rate": 4.536814507319205e-07, + "loss": 0.6803, + "step": 41667 + }, + { + "epoch": 9.09, + "learning_rate": 4.535722088704392e-07, + "loss": 0.4974, + "step": 41668 + }, + { + "epoch": 9.09, + "learning_rate": 4.5346296700895785e-07, + "loss": 1.2036, + "step": 41669 + }, + { + "epoch": 9.09, + "learning_rate": 4.5335372514747657e-07, + "loss": 2.9379, + "step": 41670 + }, + { + "epoch": 9.09, + "learning_rate": 4.5324448328599523e-07, + "loss": 1.2017, + "step": 41671 + }, + { + "epoch": 9.09, + "learning_rate": 4.531352414245139e-07, + "loss": 0.4393, + "step": 41672 + }, + { + "epoch": 9.09, + "learning_rate": 4.530259995630326e-07, + "loss": 0.9339, + "step": 41673 + }, + { + "epoch": 9.1, + "learning_rate": 4.529167577015513e-07, + "loss": 0.7157, + "step": 41674 + }, + { + "epoch": 9.1, + "learning_rate": 4.5280751584006994e-07, + "loss": 1.8532, + "step": 41675 + }, + { + "epoch": 9.1, + "learning_rate": 4.5269827397858866e-07, + "loss": 0.2207, + "step": 41676 + }, + { + "epoch": 9.1, + "learning_rate": 4.525890321171073e-07, + "loss": 0.6435, + "step": 41677 + }, + { + "epoch": 9.1, + "learning_rate": 4.52479790255626e-07, + "loss": 1.1052, + "step": 41678 + }, + { + "epoch": 9.1, + "learning_rate": 4.5237054839414465e-07, + "loss": 1.3175, + "step": 41679 + }, + { + "epoch": 9.1, + "learning_rate": 4.5226130653266337e-07, + "loss": 1.9151, + "step": 41680 + }, + { + "epoch": 9.1, + "learning_rate": 4.5215206467118203e-07, + "loss": 0.2483, + "step": 41681 + }, + { + "epoch": 9.1, + "learning_rate": 4.520428228097007e-07, + "loss": 0.7156, + "step": 41682 + }, + { + "epoch": 9.1, + "learning_rate": 4.519335809482194e-07, + "loss": 1.1243, + "step": 41683 + }, + { + "epoch": 9.1, + "learning_rate": 4.518243390867381e-07, + "loss": 1.4056, + "step": 41684 + }, + { + "epoch": 9.1, + "learning_rate": 4.5171509722525674e-07, + "loss": 1.6204, + "step": 41685 + }, + { + "epoch": 9.1, + "learning_rate": 4.5160585536377546e-07, + "loss": 0.0994, + "step": 41686 + }, + { + "epoch": 9.1, + "learning_rate": 4.514966135022941e-07, + "loss": 1.548, + "step": 41687 + }, + { + "epoch": 9.1, + "learning_rate": 4.513873716408128e-07, + "loss": 2.1076, + "step": 41688 + }, + { + "epoch": 9.1, + "learning_rate": 4.512781297793315e-07, + "loss": 1.2142, + "step": 41689 + }, + { + "epoch": 9.1, + "learning_rate": 4.5116888791785016e-07, + "loss": 1.3954, + "step": 41690 + }, + { + "epoch": 9.1, + "learning_rate": 4.5105964605636883e-07, + "loss": 1.8934, + "step": 41691 + }, + { + "epoch": 9.1, + "learning_rate": 4.509504041948875e-07, + "loss": 1.6699, + "step": 41692 + }, + { + "epoch": 9.1, + "learning_rate": 4.508411623334062e-07, + "loss": 1.7799, + "step": 41693 + }, + { + "epoch": 9.1, + "learning_rate": 4.5073192047192487e-07, + "loss": 1.6293, + "step": 41694 + }, + { + "epoch": 9.1, + "learning_rate": 4.5062267861044354e-07, + "loss": 0.9834, + "step": 41695 + }, + { + "epoch": 9.1, + "learning_rate": 4.5051343674896225e-07, + "loss": 0.6865, + "step": 41696 + }, + { + "epoch": 9.1, + "learning_rate": 4.504041948874809e-07, + "loss": 1.0273, + "step": 41697 + }, + { + "epoch": 9.1, + "learning_rate": 4.502949530259996e-07, + "loss": 0.8467, + "step": 41698 + }, + { + "epoch": 9.1, + "learning_rate": 4.501857111645183e-07, + "loss": 0.4805, + "step": 41699 + }, + { + "epoch": 9.1, + "learning_rate": 4.5007646930303696e-07, + "loss": 1.2124, + "step": 41700 + }, + { + "epoch": 9.1, + "learning_rate": 4.499672274415556e-07, + "loss": 0.7413, + "step": 41701 + }, + { + "epoch": 9.1, + "learning_rate": 4.498579855800743e-07, + "loss": 0.9072, + "step": 41702 + }, + { + "epoch": 9.1, + "learning_rate": 4.49748743718593e-07, + "loss": 0.2182, + "step": 41703 + }, + { + "epoch": 9.1, + "learning_rate": 4.4963950185711167e-07, + "loss": 0.0416, + "step": 41704 + }, + { + "epoch": 9.1, + "learning_rate": 4.4953025999563034e-07, + "loss": 0.4984, + "step": 41705 + }, + { + "epoch": 9.1, + "learning_rate": 4.4942101813414905e-07, + "loss": 1.4198, + "step": 41706 + }, + { + "epoch": 9.1, + "learning_rate": 4.493117762726677e-07, + "loss": 1.6147, + "step": 41707 + }, + { + "epoch": 9.1, + "learning_rate": 4.492025344111864e-07, + "loss": 0.6844, + "step": 41708 + }, + { + "epoch": 9.1, + "learning_rate": 4.490932925497051e-07, + "loss": 0.1706, + "step": 41709 + }, + { + "epoch": 9.1, + "learning_rate": 4.4898405068822376e-07, + "loss": 0.7216, + "step": 41710 + }, + { + "epoch": 9.1, + "learning_rate": 4.488748088267424e-07, + "loss": 2.0549, + "step": 41711 + }, + { + "epoch": 9.1, + "learning_rate": 4.487655669652611e-07, + "loss": 1.3738, + "step": 41712 + }, + { + "epoch": 9.1, + "learning_rate": 4.486563251037798e-07, + "loss": 0.9626, + "step": 41713 + }, + { + "epoch": 9.1, + "learning_rate": 4.4854708324229847e-07, + "loss": 1.5417, + "step": 41714 + }, + { + "epoch": 9.1, + "learning_rate": 4.4843784138081713e-07, + "loss": 1.3391, + "step": 41715 + }, + { + "epoch": 9.1, + "learning_rate": 4.4832859951933585e-07, + "loss": 0.2078, + "step": 41716 + }, + { + "epoch": 9.1, + "learning_rate": 4.482193576578545e-07, + "loss": 0.7808, + "step": 41717 + }, + { + "epoch": 9.1, + "learning_rate": 4.481101157963732e-07, + "loss": 1.3332, + "step": 41718 + }, + { + "epoch": 9.1, + "learning_rate": 4.480008739348919e-07, + "loss": 0.298, + "step": 41719 + }, + { + "epoch": 9.11, + "learning_rate": 4.4789163207341056e-07, + "loss": 1.6137, + "step": 41720 + }, + { + "epoch": 9.11, + "learning_rate": 4.477823902119292e-07, + "loss": 1.4859, + "step": 41721 + }, + { + "epoch": 9.11, + "learning_rate": 4.476731483504479e-07, + "loss": 1.6929, + "step": 41722 + }, + { + "epoch": 9.11, + "learning_rate": 4.475639064889666e-07, + "loss": 1.6833, + "step": 41723 + }, + { + "epoch": 9.11, + "learning_rate": 4.4745466462748527e-07, + "loss": 2.2204, + "step": 41724 + }, + { + "epoch": 9.11, + "learning_rate": 4.4734542276600393e-07, + "loss": 1.4661, + "step": 41725 + }, + { + "epoch": 9.11, + "learning_rate": 4.4723618090452265e-07, + "loss": 0.1108, + "step": 41726 + }, + { + "epoch": 9.11, + "learning_rate": 4.471269390430413e-07, + "loss": 2.0052, + "step": 41727 + }, + { + "epoch": 9.11, + "learning_rate": 4.4701769718156e-07, + "loss": 0.4438, + "step": 41728 + }, + { + "epoch": 9.11, + "learning_rate": 4.469084553200787e-07, + "loss": 0.3633, + "step": 41729 + }, + { + "epoch": 9.11, + "learning_rate": 4.4679921345859736e-07, + "loss": 1.284, + "step": 41730 + }, + { + "epoch": 9.11, + "learning_rate": 4.46689971597116e-07, + "loss": 1.8139, + "step": 41731 + }, + { + "epoch": 9.11, + "learning_rate": 4.465807297356348e-07, + "loss": 1.44, + "step": 41732 + }, + { + "epoch": 9.11, + "learning_rate": 4.464714878741534e-07, + "loss": 2.6026, + "step": 41733 + }, + { + "epoch": 9.11, + "learning_rate": 4.4636224601267207e-07, + "loss": 2.2871, + "step": 41734 + }, + { + "epoch": 9.11, + "learning_rate": 4.4625300415119073e-07, + "loss": 0.0952, + "step": 41735 + }, + { + "epoch": 9.11, + "learning_rate": 4.461437622897095e-07, + "loss": 1.5808, + "step": 41736 + }, + { + "epoch": 9.11, + "learning_rate": 4.460345204282281e-07, + "loss": 0.6401, + "step": 41737 + }, + { + "epoch": 9.11, + "learning_rate": 4.459252785667468e-07, + "loss": 2.3268, + "step": 41738 + }, + { + "epoch": 9.11, + "learning_rate": 4.4581603670526554e-07, + "loss": 0.7766, + "step": 41739 + }, + { + "epoch": 9.11, + "learning_rate": 4.4570679484378415e-07, + "loss": 1.1425, + "step": 41740 + }, + { + "epoch": 9.11, + "learning_rate": 4.455975529823028e-07, + "loss": 0.7255, + "step": 41741 + }, + { + "epoch": 9.11, + "learning_rate": 4.454883111208216e-07, + "loss": 1.3636, + "step": 41742 + }, + { + "epoch": 9.11, + "learning_rate": 4.4537906925934025e-07, + "loss": 1.5384, + "step": 41743 + }, + { + "epoch": 9.11, + "learning_rate": 4.4526982739785886e-07, + "loss": 0.4746, + "step": 41744 + }, + { + "epoch": 9.11, + "learning_rate": 4.4516058553637753e-07, + "loss": 1.0808, + "step": 41745 + }, + { + "epoch": 9.11, + "learning_rate": 4.450513436748963e-07, + "loss": 0.7811, + "step": 41746 + }, + { + "epoch": 9.11, + "learning_rate": 4.449421018134149e-07, + "loss": 1.192, + "step": 41747 + }, + { + "epoch": 9.11, + "learning_rate": 4.4483285995193357e-07, + "loss": 2.0117, + "step": 41748 + }, + { + "epoch": 9.11, + "learning_rate": 4.4472361809045234e-07, + "loss": 2.2544, + "step": 41749 + }, + { + "epoch": 9.11, + "learning_rate": 4.44614376228971e-07, + "loss": 1.0743, + "step": 41750 + }, + { + "epoch": 9.11, + "learning_rate": 4.445051343674896e-07, + "loss": 0.1077, + "step": 41751 + }, + { + "epoch": 9.11, + "learning_rate": 4.443958925060084e-07, + "loss": 1.7369, + "step": 41752 + }, + { + "epoch": 9.11, + "learning_rate": 4.4428665064452705e-07, + "loss": 0.3679, + "step": 41753 + }, + { + "epoch": 9.11, + "learning_rate": 4.441774087830457e-07, + "loss": 0.0596, + "step": 41754 + }, + { + "epoch": 9.11, + "learning_rate": 4.440681669215643e-07, + "loss": 1.1382, + "step": 41755 + }, + { + "epoch": 9.11, + "learning_rate": 4.439589250600831e-07, + "loss": 1.3432, + "step": 41756 + }, + { + "epoch": 9.11, + "learning_rate": 4.4384968319860176e-07, + "loss": 0.9824, + "step": 41757 + }, + { + "epoch": 9.11, + "learning_rate": 4.4374044133712037e-07, + "loss": 0.8298, + "step": 41758 + }, + { + "epoch": 9.11, + "learning_rate": 4.4363119947563914e-07, + "loss": 1.3695, + "step": 41759 + }, + { + "epoch": 9.11, + "learning_rate": 4.435219576141578e-07, + "loss": 0.3057, + "step": 41760 + }, + { + "epoch": 9.11, + "learning_rate": 4.4341271575267647e-07, + "loss": 0.7438, + "step": 41761 + }, + { + "epoch": 9.11, + "learning_rate": 4.433034738911952e-07, + "loss": 1.9996, + "step": 41762 + }, + { + "epoch": 9.11, + "learning_rate": 4.4319423202971385e-07, + "loss": 0.3497, + "step": 41763 + }, + { + "epoch": 9.11, + "learning_rate": 4.430849901682325e-07, + "loss": 2.0254, + "step": 41764 + }, + { + "epoch": 9.12, + "learning_rate": 4.429757483067511e-07, + "loss": 0.0556, + "step": 41765 + }, + { + "epoch": 9.12, + "learning_rate": 4.428665064452699e-07, + "loss": 1.1162, + "step": 41766 + }, + { + "epoch": 9.12, + "learning_rate": 4.4275726458378856e-07, + "loss": 1.5767, + "step": 41767 + }, + { + "epoch": 9.12, + "learning_rate": 4.426480227223072e-07, + "loss": 2.6473, + "step": 41768 + }, + { + "epoch": 9.12, + "learning_rate": 4.4253878086082594e-07, + "loss": 0.4483, + "step": 41769 + }, + { + "epoch": 9.12, + "learning_rate": 4.424295389993446e-07, + "loss": 0.9614, + "step": 41770 + }, + { + "epoch": 9.12, + "learning_rate": 4.4232029713786327e-07, + "loss": 0.6257, + "step": 41771 + }, + { + "epoch": 9.12, + "learning_rate": 4.42211055276382e-07, + "loss": 1.5119, + "step": 41772 + }, + { + "epoch": 9.12, + "learning_rate": 4.4210181341490065e-07, + "loss": 1.0849, + "step": 41773 + }, + { + "epoch": 9.12, + "learning_rate": 4.419925715534193e-07, + "loss": 0.8643, + "step": 41774 + }, + { + "epoch": 9.12, + "learning_rate": 4.4188332969193803e-07, + "loss": 0.0438, + "step": 41775 + }, + { + "epoch": 9.12, + "learning_rate": 4.417740878304567e-07, + "loss": 0.0396, + "step": 41776 + }, + { + "epoch": 9.12, + "learning_rate": 4.4166484596897536e-07, + "loss": 0.8205, + "step": 41777 + }, + { + "epoch": 9.12, + "learning_rate": 4.41555604107494e-07, + "loss": 0.3304, + "step": 41778 + }, + { + "epoch": 9.12, + "learning_rate": 4.4144636224601274e-07, + "loss": 0.9536, + "step": 41779 + }, + { + "epoch": 9.12, + "learning_rate": 4.413371203845314e-07, + "loss": 3.6212, + "step": 41780 + }, + { + "epoch": 9.12, + "learning_rate": 4.4122787852305006e-07, + "loss": 0.9626, + "step": 41781 + }, + { + "epoch": 9.12, + "learning_rate": 4.411186366615688e-07, + "loss": 1.7713, + "step": 41782 + }, + { + "epoch": 9.12, + "learning_rate": 4.4100939480008744e-07, + "loss": 1.4207, + "step": 41783 + }, + { + "epoch": 9.12, + "learning_rate": 4.409001529386061e-07, + "loss": 0.6228, + "step": 41784 + }, + { + "epoch": 9.12, + "learning_rate": 4.407909110771248e-07, + "loss": 0.6991, + "step": 41785 + }, + { + "epoch": 9.12, + "learning_rate": 4.406816692156435e-07, + "loss": 1.9021, + "step": 41786 + }, + { + "epoch": 9.12, + "learning_rate": 4.4057242735416215e-07, + "loss": 1.5913, + "step": 41787 + }, + { + "epoch": 9.12, + "learning_rate": 4.404631854926808e-07, + "loss": 1.3736, + "step": 41788 + }, + { + "epoch": 9.12, + "learning_rate": 4.4035394363119953e-07, + "loss": 0.9574, + "step": 41789 + }, + { + "epoch": 9.12, + "learning_rate": 4.402447017697182e-07, + "loss": 1.4236, + "step": 41790 + }, + { + "epoch": 9.12, + "learning_rate": 4.4013545990823686e-07, + "loss": 0.3335, + "step": 41791 + }, + { + "epoch": 9.12, + "learning_rate": 4.400262180467556e-07, + "loss": 1.7697, + "step": 41792 + }, + { + "epoch": 9.12, + "learning_rate": 4.3991697618527424e-07, + "loss": 0.101, + "step": 41793 + }, + { + "epoch": 9.12, + "learning_rate": 4.398077343237929e-07, + "loss": 1.1073, + "step": 41794 + }, + { + "epoch": 9.12, + "learning_rate": 4.396984924623116e-07, + "loss": 0.6367, + "step": 41795 + }, + { + "epoch": 9.12, + "learning_rate": 4.395892506008303e-07, + "loss": 0.6899, + "step": 41796 + }, + { + "epoch": 9.12, + "learning_rate": 4.3948000873934895e-07, + "loss": 0.9954, + "step": 41797 + }, + { + "epoch": 9.12, + "learning_rate": 4.393707668778676e-07, + "loss": 1.0031, + "step": 41798 + }, + { + "epoch": 9.12, + "learning_rate": 4.3926152501638633e-07, + "loss": 1.2045, + "step": 41799 + }, + { + "epoch": 9.12, + "learning_rate": 4.39152283154905e-07, + "loss": 2.4091, + "step": 41800 + }, + { + "epoch": 9.12, + "learning_rate": 4.3904304129342366e-07, + "loss": 0.9836, + "step": 41801 + }, + { + "epoch": 9.12, + "learning_rate": 4.389337994319424e-07, + "loss": 2.4222, + "step": 41802 + }, + { + "epoch": 9.12, + "learning_rate": 4.3882455757046104e-07, + "loss": 1.0503, + "step": 41803 + }, + { + "epoch": 9.12, + "learning_rate": 4.387153157089797e-07, + "loss": 1.2319, + "step": 41804 + }, + { + "epoch": 9.12, + "learning_rate": 4.386060738474984e-07, + "loss": 1.5461, + "step": 41805 + }, + { + "epoch": 9.12, + "learning_rate": 4.384968319860171e-07, + "loss": 1.1896, + "step": 41806 + }, + { + "epoch": 9.12, + "learning_rate": 4.3838759012453575e-07, + "loss": 0.661, + "step": 41807 + }, + { + "epoch": 9.12, + "learning_rate": 4.382783482630544e-07, + "loss": 0.5011, + "step": 41808 + }, + { + "epoch": 9.12, + "learning_rate": 4.3816910640157313e-07, + "loss": 0.5414, + "step": 41809 + }, + { + "epoch": 9.12, + "learning_rate": 4.380598645400918e-07, + "loss": 1.465, + "step": 41810 + }, + { + "epoch": 9.13, + "learning_rate": 4.3795062267861046e-07, + "loss": 1.3889, + "step": 41811 + }, + { + "epoch": 9.13, + "learning_rate": 4.378413808171292e-07, + "loss": 0.868, + "step": 41812 + }, + { + "epoch": 9.13, + "learning_rate": 4.3773213895564784e-07, + "loss": 0.4212, + "step": 41813 + }, + { + "epoch": 9.13, + "learning_rate": 4.376228970941665e-07, + "loss": 0.3955, + "step": 41814 + }, + { + "epoch": 9.13, + "learning_rate": 4.375136552326852e-07, + "loss": 1.634, + "step": 41815 + }, + { + "epoch": 9.13, + "learning_rate": 4.374044133712039e-07, + "loss": 0.5234, + "step": 41816 + }, + { + "epoch": 9.13, + "learning_rate": 4.3729517150972255e-07, + "loss": 2.048, + "step": 41817 + }, + { + "epoch": 9.13, + "learning_rate": 4.3718592964824126e-07, + "loss": 0.9818, + "step": 41818 + }, + { + "epoch": 9.13, + "learning_rate": 4.3707668778675993e-07, + "loss": 1.7905, + "step": 41819 + }, + { + "epoch": 9.13, + "learning_rate": 4.369674459252786e-07, + "loss": 0.5547, + "step": 41820 + }, + { + "epoch": 9.13, + "learning_rate": 4.3685820406379726e-07, + "loss": 1.3299, + "step": 41821 + }, + { + "epoch": 9.13, + "learning_rate": 4.3674896220231597e-07, + "loss": 1.3596, + "step": 41822 + }, + { + "epoch": 9.13, + "learning_rate": 4.3663972034083464e-07, + "loss": 1.1223, + "step": 41823 + }, + { + "epoch": 9.13, + "learning_rate": 4.365304784793533e-07, + "loss": 0.4234, + "step": 41824 + }, + { + "epoch": 9.13, + "learning_rate": 4.36421236617872e-07, + "loss": 0.63, + "step": 41825 + }, + { + "epoch": 9.13, + "learning_rate": 4.363119947563907e-07, + "loss": 2.2259, + "step": 41826 + }, + { + "epoch": 9.13, + "learning_rate": 4.3620275289490935e-07, + "loss": 0.6537, + "step": 41827 + }, + { + "epoch": 9.13, + "learning_rate": 4.3609351103342806e-07, + "loss": 1.2482, + "step": 41828 + }, + { + "epoch": 9.13, + "learning_rate": 4.3598426917194673e-07, + "loss": 0.206, + "step": 41829 + }, + { + "epoch": 9.13, + "learning_rate": 4.358750273104654e-07, + "loss": 1.5183, + "step": 41830 + }, + { + "epoch": 9.13, + "learning_rate": 4.3576578544898405e-07, + "loss": 1.2323, + "step": 41831 + }, + { + "epoch": 9.13, + "learning_rate": 4.3565654358750277e-07, + "loss": 1.7014, + "step": 41832 + }, + { + "epoch": 9.13, + "learning_rate": 4.3554730172602144e-07, + "loss": 0.9231, + "step": 41833 + }, + { + "epoch": 9.13, + "learning_rate": 4.354380598645401e-07, + "loss": 0.4533, + "step": 41834 + }, + { + "epoch": 9.13, + "learning_rate": 4.353288180030588e-07, + "loss": 1.7484, + "step": 41835 + }, + { + "epoch": 9.13, + "learning_rate": 4.352195761415775e-07, + "loss": 0.0604, + "step": 41836 + }, + { + "epoch": 9.13, + "learning_rate": 4.3511033428009614e-07, + "loss": 1.5337, + "step": 41837 + }, + { + "epoch": 9.13, + "learning_rate": 4.3500109241861486e-07, + "loss": 1.2377, + "step": 41838 + }, + { + "epoch": 9.13, + "learning_rate": 4.348918505571335e-07, + "loss": 1.417, + "step": 41839 + }, + { + "epoch": 9.13, + "learning_rate": 4.347826086956522e-07, + "loss": 1.4287, + "step": 41840 + }, + { + "epoch": 9.13, + "learning_rate": 4.3467336683417085e-07, + "loss": 1.1425, + "step": 41841 + }, + { + "epoch": 9.13, + "learning_rate": 4.3456412497268957e-07, + "loss": 1.2357, + "step": 41842 + }, + { + "epoch": 9.13, + "learning_rate": 4.3445488311120823e-07, + "loss": 0.7082, + "step": 41843 + }, + { + "epoch": 9.13, + "learning_rate": 4.343456412497269e-07, + "loss": 0.455, + "step": 41844 + }, + { + "epoch": 9.13, + "learning_rate": 4.342363993882456e-07, + "loss": 0.3035, + "step": 41845 + }, + { + "epoch": 9.13, + "learning_rate": 4.341271575267643e-07, + "loss": 1.8886, + "step": 41846 + }, + { + "epoch": 9.13, + "learning_rate": 4.3401791566528294e-07, + "loss": 0.4044, + "step": 41847 + }, + { + "epoch": 9.13, + "learning_rate": 4.3390867380380166e-07, + "loss": 2.7027, + "step": 41848 + }, + { + "epoch": 9.13, + "learning_rate": 4.337994319423203e-07, + "loss": 0.509, + "step": 41849 + }, + { + "epoch": 9.13, + "learning_rate": 4.33690190080839e-07, + "loss": 0.3278, + "step": 41850 + }, + { + "epoch": 9.13, + "learning_rate": 4.335809482193577e-07, + "loss": 1.2836, + "step": 41851 + }, + { + "epoch": 9.13, + "learning_rate": 4.3347170635787637e-07, + "loss": 0.891, + "step": 41852 + }, + { + "epoch": 9.13, + "learning_rate": 4.3336246449639503e-07, + "loss": 2.2119, + "step": 41853 + }, + { + "epoch": 9.13, + "learning_rate": 4.332532226349137e-07, + "loss": 2.1018, + "step": 41854 + }, + { + "epoch": 9.13, + "learning_rate": 4.331439807734324e-07, + "loss": 1.4917, + "step": 41855 + }, + { + "epoch": 9.13, + "learning_rate": 4.330347389119511e-07, + "loss": 3.9643, + "step": 41856 + }, + { + "epoch": 9.14, + "learning_rate": 4.3292549705046974e-07, + "loss": 1.4113, + "step": 41857 + }, + { + "epoch": 9.14, + "learning_rate": 4.3281625518898846e-07, + "loss": 1.716, + "step": 41858 + }, + { + "epoch": 9.14, + "learning_rate": 4.327070133275071e-07, + "loss": 1.5231, + "step": 41859 + }, + { + "epoch": 9.14, + "learning_rate": 4.325977714660258e-07, + "loss": 1.3712, + "step": 41860 + }, + { + "epoch": 9.14, + "learning_rate": 4.3248852960454455e-07, + "loss": 0.464, + "step": 41861 + }, + { + "epoch": 9.14, + "learning_rate": 4.3237928774306317e-07, + "loss": 0.6835, + "step": 41862 + }, + { + "epoch": 9.14, + "learning_rate": 4.3227004588158183e-07, + "loss": 3.7651, + "step": 41863 + }, + { + "epoch": 9.14, + "learning_rate": 4.321608040201005e-07, + "loss": 0.7203, + "step": 41864 + }, + { + "epoch": 9.14, + "learning_rate": 4.320515621586192e-07, + "loss": 0.148, + "step": 41865 + }, + { + "epoch": 9.14, + "learning_rate": 4.319423202971379e-07, + "loss": 1.1818, + "step": 41866 + }, + { + "epoch": 9.14, + "learning_rate": 4.3183307843565654e-07, + "loss": 0.687, + "step": 41867 + }, + { + "epoch": 9.14, + "learning_rate": 4.317238365741753e-07, + "loss": 0.7099, + "step": 41868 + }, + { + "epoch": 9.14, + "learning_rate": 4.316145947126939e-07, + "loss": 0.4384, + "step": 41869 + }, + { + "epoch": 9.14, + "learning_rate": 4.315053528512126e-07, + "loss": 0.0722, + "step": 41870 + }, + { + "epoch": 9.14, + "learning_rate": 4.3139611098973135e-07, + "loss": 0.8798, + "step": 41871 + }, + { + "epoch": 9.14, + "learning_rate": 4.3128686912825e-07, + "loss": 0.2669, + "step": 41872 + }, + { + "epoch": 9.14, + "learning_rate": 4.3117762726676863e-07, + "loss": 0.5463, + "step": 41873 + }, + { + "epoch": 9.14, + "learning_rate": 4.310683854052873e-07, + "loss": 1.5604, + "step": 41874 + }, + { + "epoch": 9.14, + "learning_rate": 4.3095914354380606e-07, + "loss": 0.6092, + "step": 41875 + }, + { + "epoch": 9.14, + "learning_rate": 4.3084990168232467e-07, + "loss": 2.2331, + "step": 41876 + }, + { + "epoch": 9.14, + "learning_rate": 4.3074065982084334e-07, + "loss": 0.9734, + "step": 41877 + }, + { + "epoch": 9.14, + "learning_rate": 4.306314179593621e-07, + "loss": 0.1582, + "step": 41878 + }, + { + "epoch": 9.14, + "learning_rate": 4.3052217609788077e-07, + "loss": 1.1995, + "step": 41879 + }, + { + "epoch": 9.14, + "learning_rate": 4.304129342363994e-07, + "loss": 1.9788, + "step": 41880 + }, + { + "epoch": 9.14, + "learning_rate": 4.3030369237491815e-07, + "loss": 2.2697, + "step": 41881 + }, + { + "epoch": 9.14, + "learning_rate": 4.301944505134368e-07, + "loss": 2.8325, + "step": 41882 + }, + { + "epoch": 9.14, + "learning_rate": 4.300852086519554e-07, + "loss": 0.366, + "step": 41883 + }, + { + "epoch": 9.14, + "learning_rate": 4.299759667904741e-07, + "loss": 0.3651, + "step": 41884 + }, + { + "epoch": 9.14, + "learning_rate": 4.2986672492899286e-07, + "loss": 1.2208, + "step": 41885 + }, + { + "epoch": 9.14, + "learning_rate": 4.297574830675115e-07, + "loss": 1.897, + "step": 41886 + }, + { + "epoch": 9.14, + "learning_rate": 4.2964824120603013e-07, + "loss": 0.7569, + "step": 41887 + }, + { + "epoch": 9.14, + "learning_rate": 4.295389993445489e-07, + "loss": 0.3445, + "step": 41888 + }, + { + "epoch": 9.14, + "learning_rate": 4.2942975748306757e-07, + "loss": 1.342, + "step": 41889 + }, + { + "epoch": 9.14, + "learning_rate": 4.2932051562158623e-07, + "loss": 2.9728, + "step": 41890 + }, + { + "epoch": 9.14, + "learning_rate": 4.2921127376010495e-07, + "loss": 2.3066, + "step": 41891 + }, + { + "epoch": 9.14, + "learning_rate": 4.291020318986236e-07, + "loss": 0.8712, + "step": 41892 + }, + { + "epoch": 9.14, + "learning_rate": 4.289927900371423e-07, + "loss": 0.6432, + "step": 41893 + }, + { + "epoch": 9.14, + "learning_rate": 4.28883548175661e-07, + "loss": 1.3119, + "step": 41894 + }, + { + "epoch": 9.14, + "learning_rate": 4.2877430631417966e-07, + "loss": 1.2356, + "step": 41895 + }, + { + "epoch": 9.14, + "learning_rate": 4.286650644526983e-07, + "loss": 0.3477, + "step": 41896 + }, + { + "epoch": 9.14, + "learning_rate": 4.28555822591217e-07, + "loss": 0.7182, + "step": 41897 + }, + { + "epoch": 9.14, + "learning_rate": 4.284465807297357e-07, + "loss": 0.2471, + "step": 41898 + }, + { + "epoch": 9.14, + "learning_rate": 4.2833733886825437e-07, + "loss": 1.8631, + "step": 41899 + }, + { + "epoch": 9.14, + "learning_rate": 4.2822809700677303e-07, + "loss": 1.5637, + "step": 41900 + }, + { + "epoch": 9.14, + "learning_rate": 4.2811885514529175e-07, + "loss": 1.6787, + "step": 41901 + }, + { + "epoch": 9.14, + "learning_rate": 4.280096132838104e-07, + "loss": 1.1356, + "step": 41902 + }, + { + "epoch": 9.15, + "learning_rate": 4.279003714223291e-07, + "loss": 0.9728, + "step": 41903 + }, + { + "epoch": 9.15, + "learning_rate": 4.277911295608478e-07, + "loss": 0.6987, + "step": 41904 + }, + { + "epoch": 9.15, + "learning_rate": 4.2768188769936646e-07, + "loss": 1.1647, + "step": 41905 + }, + { + "epoch": 9.15, + "learning_rate": 4.275726458378851e-07, + "loss": 1.0248, + "step": 41906 + }, + { + "epoch": 9.15, + "learning_rate": 4.274634039764038e-07, + "loss": 0.5674, + "step": 41907 + }, + { + "epoch": 9.15, + "learning_rate": 4.273541621149225e-07, + "loss": 1.679, + "step": 41908 + }, + { + "epoch": 9.15, + "learning_rate": 4.2724492025344116e-07, + "loss": 0.5884, + "step": 41909 + }, + { + "epoch": 9.15, + "learning_rate": 4.2713567839195983e-07, + "loss": 0.9464, + "step": 41910 + }, + { + "epoch": 9.15, + "learning_rate": 4.2702643653047855e-07, + "loss": 1.4915, + "step": 41911 + }, + { + "epoch": 9.15, + "learning_rate": 4.269171946689972e-07, + "loss": 0.3716, + "step": 41912 + }, + { + "epoch": 9.15, + "learning_rate": 4.2680795280751587e-07, + "loss": 2.5631, + "step": 41913 + }, + { + "epoch": 9.15, + "learning_rate": 4.266987109460346e-07, + "loss": 1.2477, + "step": 41914 + }, + { + "epoch": 9.15, + "learning_rate": 4.2658946908455325e-07, + "loss": 0.4444, + "step": 41915 + }, + { + "epoch": 9.15, + "learning_rate": 4.264802272230719e-07, + "loss": 0.9409, + "step": 41916 + }, + { + "epoch": 9.15, + "learning_rate": 4.263709853615906e-07, + "loss": 1.8305, + "step": 41917 + }, + { + "epoch": 9.15, + "learning_rate": 4.262617435001093e-07, + "loss": 1.4607, + "step": 41918 + }, + { + "epoch": 9.15, + "learning_rate": 4.2615250163862796e-07, + "loss": 2.3605, + "step": 41919 + }, + { + "epoch": 9.15, + "learning_rate": 4.260432597771466e-07, + "loss": 0.7223, + "step": 41920 + }, + { + "epoch": 9.15, + "learning_rate": 4.2593401791566534e-07, + "loss": 1.067, + "step": 41921 + }, + { + "epoch": 9.15, + "learning_rate": 4.25824776054184e-07, + "loss": 1.3067, + "step": 41922 + }, + { + "epoch": 9.15, + "learning_rate": 4.2571553419270267e-07, + "loss": 0.0605, + "step": 41923 + }, + { + "epoch": 9.15, + "learning_rate": 4.256062923312214e-07, + "loss": 3.3955, + "step": 41924 + }, + { + "epoch": 9.15, + "learning_rate": 4.2549705046974005e-07, + "loss": 1.3712, + "step": 41925 + }, + { + "epoch": 9.15, + "learning_rate": 4.253878086082587e-07, + "loss": 1.2542, + "step": 41926 + }, + { + "epoch": 9.15, + "learning_rate": 4.252785667467774e-07, + "loss": 1.0565, + "step": 41927 + }, + { + "epoch": 9.15, + "learning_rate": 4.251693248852961e-07, + "loss": 0.7262, + "step": 41928 + }, + { + "epoch": 9.15, + "learning_rate": 4.2506008302381476e-07, + "loss": 1.1039, + "step": 41929 + }, + { + "epoch": 9.15, + "learning_rate": 4.249508411623334e-07, + "loss": 0.7167, + "step": 41930 + }, + { + "epoch": 9.15, + "learning_rate": 4.2484159930085214e-07, + "loss": 0.55, + "step": 41931 + }, + { + "epoch": 9.15, + "learning_rate": 4.247323574393708e-07, + "loss": 0.1206, + "step": 41932 + }, + { + "epoch": 9.15, + "learning_rate": 4.2462311557788947e-07, + "loss": 0.6382, + "step": 41933 + }, + { + "epoch": 9.15, + "learning_rate": 4.245138737164082e-07, + "loss": 0.1971, + "step": 41934 + }, + { + "epoch": 9.15, + "learning_rate": 4.2440463185492685e-07, + "loss": 0.7274, + "step": 41935 + }, + { + "epoch": 9.15, + "learning_rate": 4.242953899934455e-07, + "loss": 0.8563, + "step": 41936 + }, + { + "epoch": 9.15, + "learning_rate": 4.2418614813196423e-07, + "loss": 0.0905, + "step": 41937 + }, + { + "epoch": 9.15, + "learning_rate": 4.240769062704829e-07, + "loss": 0.7002, + "step": 41938 + }, + { + "epoch": 9.15, + "learning_rate": 4.2396766440900156e-07, + "loss": 0.2823, + "step": 41939 + }, + { + "epoch": 9.15, + "learning_rate": 4.238584225475202e-07, + "loss": 1.6345, + "step": 41940 + }, + { + "epoch": 9.15, + "learning_rate": 4.2374918068603894e-07, + "loss": 0.1577, + "step": 41941 + }, + { + "epoch": 9.15, + "learning_rate": 4.236399388245576e-07, + "loss": 0.7323, + "step": 41942 + }, + { + "epoch": 9.15, + "learning_rate": 4.2353069696307627e-07, + "loss": 2.4401, + "step": 41943 + }, + { + "epoch": 9.15, + "learning_rate": 4.23421455101595e-07, + "loss": 2.7268, + "step": 41944 + }, + { + "epoch": 9.15, + "learning_rate": 4.2331221324011365e-07, + "loss": 0.0594, + "step": 41945 + }, + { + "epoch": 9.15, + "learning_rate": 4.232029713786323e-07, + "loss": 0.7989, + "step": 41946 + }, + { + "epoch": 9.15, + "learning_rate": 4.2309372951715103e-07, + "loss": 0.9097, + "step": 41947 + }, + { + "epoch": 9.15, + "learning_rate": 4.229844876556697e-07, + "loss": 0.4044, + "step": 41948 + }, + { + "epoch": 9.16, + "learning_rate": 4.2287524579418836e-07, + "loss": 2.0974, + "step": 41949 + }, + { + "epoch": 9.16, + "learning_rate": 4.22766003932707e-07, + "loss": 0.8957, + "step": 41950 + }, + { + "epoch": 9.16, + "learning_rate": 4.2265676207122574e-07, + "loss": 1.8924, + "step": 41951 + }, + { + "epoch": 9.16, + "learning_rate": 4.225475202097444e-07, + "loss": 0.8018, + "step": 41952 + }, + { + "epoch": 9.16, + "learning_rate": 4.2243827834826307e-07, + "loss": 1.7812, + "step": 41953 + }, + { + "epoch": 9.16, + "learning_rate": 4.223290364867818e-07, + "loss": 0.7718, + "step": 41954 + }, + { + "epoch": 9.16, + "learning_rate": 4.2221979462530045e-07, + "loss": 1.0578, + "step": 41955 + }, + { + "epoch": 9.16, + "learning_rate": 4.221105527638191e-07, + "loss": 0.3538, + "step": 41956 + }, + { + "epoch": 9.16, + "learning_rate": 4.2200131090233783e-07, + "loss": 1.329, + "step": 41957 + }, + { + "epoch": 9.16, + "learning_rate": 4.218920690408565e-07, + "loss": 1.4598, + "step": 41958 + }, + { + "epoch": 9.16, + "learning_rate": 4.2178282717937515e-07, + "loss": 1.0847, + "step": 41959 + }, + { + "epoch": 9.16, + "learning_rate": 4.216735853178938e-07, + "loss": 1.5976, + "step": 41960 + }, + { + "epoch": 9.16, + "learning_rate": 4.2156434345641254e-07, + "loss": 2.1602, + "step": 41961 + }, + { + "epoch": 9.16, + "learning_rate": 4.214551015949312e-07, + "loss": 0.9261, + "step": 41962 + }, + { + "epoch": 9.16, + "learning_rate": 4.2134585973344986e-07, + "loss": 0.7855, + "step": 41963 + }, + { + "epoch": 9.16, + "learning_rate": 4.212366178719686e-07, + "loss": 0.8979, + "step": 41964 + }, + { + "epoch": 9.16, + "learning_rate": 4.2112737601048724e-07, + "loss": 0.7204, + "step": 41965 + }, + { + "epoch": 9.16, + "learning_rate": 4.210181341490059e-07, + "loss": 0.6648, + "step": 41966 + }, + { + "epoch": 9.16, + "learning_rate": 4.209088922875246e-07, + "loss": 0.1552, + "step": 41967 + }, + { + "epoch": 9.16, + "learning_rate": 4.207996504260433e-07, + "loss": 1.0415, + "step": 41968 + }, + { + "epoch": 9.16, + "learning_rate": 4.2069040856456195e-07, + "loss": 0.8105, + "step": 41969 + }, + { + "epoch": 9.16, + "learning_rate": 4.205811667030806e-07, + "loss": 0.6415, + "step": 41970 + }, + { + "epoch": 9.16, + "learning_rate": 4.2047192484159933e-07, + "loss": 1.0886, + "step": 41971 + }, + { + "epoch": 9.16, + "learning_rate": 4.20362682980118e-07, + "loss": 1.127, + "step": 41972 + }, + { + "epoch": 9.16, + "learning_rate": 4.2025344111863666e-07, + "loss": 0.2367, + "step": 41973 + }, + { + "epoch": 9.16, + "learning_rate": 4.201441992571554e-07, + "loss": 1.0618, + "step": 41974 + }, + { + "epoch": 9.16, + "learning_rate": 4.2003495739567404e-07, + "loss": 0.8346, + "step": 41975 + }, + { + "epoch": 9.16, + "learning_rate": 4.199257155341927e-07, + "loss": 1.2337, + "step": 41976 + }, + { + "epoch": 9.16, + "learning_rate": 4.198164736727114e-07, + "loss": 0.7178, + "step": 41977 + }, + { + "epoch": 9.16, + "learning_rate": 4.197072318112301e-07, + "loss": 1.3788, + "step": 41978 + }, + { + "epoch": 9.16, + "learning_rate": 4.1959798994974875e-07, + "loss": 1.7615, + "step": 41979 + }, + { + "epoch": 9.16, + "learning_rate": 4.1948874808826747e-07, + "loss": 0.4267, + "step": 41980 + }, + { + "epoch": 9.16, + "learning_rate": 4.1937950622678613e-07, + "loss": 0.1508, + "step": 41981 + }, + { + "epoch": 9.16, + "learning_rate": 4.192702643653048e-07, + "loss": 0.6365, + "step": 41982 + }, + { + "epoch": 9.16, + "learning_rate": 4.1916102250382346e-07, + "loss": 1.9537, + "step": 41983 + }, + { + "epoch": 9.16, + "learning_rate": 4.190517806423422e-07, + "loss": 1.5629, + "step": 41984 + }, + { + "epoch": 9.16, + "learning_rate": 4.1894253878086084e-07, + "loss": 1.4455, + "step": 41985 + }, + { + "epoch": 9.16, + "learning_rate": 4.188332969193795e-07, + "loss": 0.7584, + "step": 41986 + }, + { + "epoch": 9.16, + "learning_rate": 4.187240550578982e-07, + "loss": 1.6771, + "step": 41987 + }, + { + "epoch": 9.16, + "learning_rate": 4.186148131964169e-07, + "loss": 0.7061, + "step": 41988 + }, + { + "epoch": 9.16, + "learning_rate": 4.1850557133493555e-07, + "loss": 0.837, + "step": 41989 + }, + { + "epoch": 9.16, + "learning_rate": 4.183963294734543e-07, + "loss": 0.9875, + "step": 41990 + }, + { + "epoch": 9.16, + "learning_rate": 4.1828708761197293e-07, + "loss": 1.2919, + "step": 41991 + }, + { + "epoch": 9.16, + "learning_rate": 4.181778457504916e-07, + "loss": 1.078, + "step": 41992 + }, + { + "epoch": 9.16, + "learning_rate": 4.1806860388901026e-07, + "loss": 0.0723, + "step": 41993 + }, + { + "epoch": 9.16, + "learning_rate": 4.17959362027529e-07, + "loss": 0.8296, + "step": 41994 + }, + { + "epoch": 9.17, + "learning_rate": 4.1785012016604764e-07, + "loss": 1.5799, + "step": 41995 + }, + { + "epoch": 9.17, + "learning_rate": 4.177408783045663e-07, + "loss": 0.5079, + "step": 41996 + }, + { + "epoch": 9.17, + "learning_rate": 4.1763163644308507e-07, + "loss": 1.7689, + "step": 41997 + }, + { + "epoch": 9.17, + "learning_rate": 4.175223945816037e-07, + "loss": 1.392, + "step": 41998 + }, + { + "epoch": 9.17, + "learning_rate": 4.1741315272012235e-07, + "loss": 1.9472, + "step": 41999 + }, + { + "epoch": 9.17, + "learning_rate": 4.173039108586411e-07, + "loss": 0.3932, + "step": 42000 + }, + { + "epoch": 9.17, + "learning_rate": 4.171946689971598e-07, + "loss": 2.0422, + "step": 42001 + }, + { + "epoch": 9.17, + "learning_rate": 4.170854271356784e-07, + "loss": 1.2637, + "step": 42002 + }, + { + "epoch": 9.17, + "learning_rate": 4.1697618527419706e-07, + "loss": 1.0008, + "step": 42003 + }, + { + "epoch": 9.17, + "learning_rate": 4.168669434127158e-07, + "loss": 0.4153, + "step": 42004 + }, + { + "epoch": 9.17, + "learning_rate": 4.1675770155123444e-07, + "loss": 1.3392, + "step": 42005 + }, + { + "epoch": 9.17, + "learning_rate": 4.166484596897531e-07, + "loss": 1.0476, + "step": 42006 + }, + { + "epoch": 9.17, + "learning_rate": 4.1653921782827187e-07, + "loss": 0.4622, + "step": 42007 + }, + { + "epoch": 9.17, + "learning_rate": 4.1642997596679053e-07, + "loss": 0.531, + "step": 42008 + }, + { + "epoch": 9.17, + "learning_rate": 4.1632073410530915e-07, + "loss": 1.1289, + "step": 42009 + }, + { + "epoch": 9.17, + "learning_rate": 4.162114922438279e-07, + "loss": 1.419, + "step": 42010 + }, + { + "epoch": 9.17, + "learning_rate": 4.161022503823466e-07, + "loss": 0.3054, + "step": 42011 + }, + { + "epoch": 9.17, + "learning_rate": 4.159930085208652e-07, + "loss": 0.8576, + "step": 42012 + }, + { + "epoch": 9.17, + "learning_rate": 4.1588376665938396e-07, + "loss": 1.2109, + "step": 42013 + }, + { + "epoch": 9.17, + "learning_rate": 4.157745247979026e-07, + "loss": 0.0682, + "step": 42014 + }, + { + "epoch": 9.17, + "learning_rate": 4.156652829364213e-07, + "loss": 0.6198, + "step": 42015 + }, + { + "epoch": 9.17, + "learning_rate": 4.155560410749399e-07, + "loss": 1.104, + "step": 42016 + }, + { + "epoch": 9.17, + "learning_rate": 4.1544679921345867e-07, + "loss": 0.8478, + "step": 42017 + }, + { + "epoch": 9.17, + "learning_rate": 4.1533755735197733e-07, + "loss": 0.2696, + "step": 42018 + }, + { + "epoch": 9.17, + "learning_rate": 4.15228315490496e-07, + "loss": 2.419, + "step": 42019 + }, + { + "epoch": 9.17, + "learning_rate": 4.151190736290147e-07, + "loss": 1.7201, + "step": 42020 + }, + { + "epoch": 9.17, + "learning_rate": 4.150098317675334e-07, + "loss": 1.1874, + "step": 42021 + }, + { + "epoch": 9.17, + "learning_rate": 4.1490058990605204e-07, + "loss": 0.9938, + "step": 42022 + }, + { + "epoch": 9.17, + "learning_rate": 4.1479134804457076e-07, + "loss": 1.0802, + "step": 42023 + }, + { + "epoch": 9.17, + "learning_rate": 4.146821061830894e-07, + "loss": 2.4936, + "step": 42024 + }, + { + "epoch": 9.17, + "learning_rate": 4.145728643216081e-07, + "loss": 0.3904, + "step": 42025 + }, + { + "epoch": 9.17, + "learning_rate": 4.1446362246012675e-07, + "loss": 0.6899, + "step": 42026 + }, + { + "epoch": 9.17, + "learning_rate": 4.1435438059864547e-07, + "loss": 0.9381, + "step": 42027 + }, + { + "epoch": 9.17, + "learning_rate": 4.1424513873716413e-07, + "loss": 0.6091, + "step": 42028 + }, + { + "epoch": 9.17, + "learning_rate": 4.141358968756828e-07, + "loss": 1.5534, + "step": 42029 + }, + { + "epoch": 9.17, + "learning_rate": 4.140266550142015e-07, + "loss": 0.039, + "step": 42030 + }, + { + "epoch": 9.17, + "learning_rate": 4.139174131527202e-07, + "loss": 1.7124, + "step": 42031 + }, + { + "epoch": 9.17, + "learning_rate": 4.1380817129123884e-07, + "loss": 1.8596, + "step": 42032 + }, + { + "epoch": 9.17, + "learning_rate": 4.1369892942975756e-07, + "loss": 0.55, + "step": 42033 + }, + { + "epoch": 9.17, + "learning_rate": 4.135896875682762e-07, + "loss": 0.5798, + "step": 42034 + }, + { + "epoch": 9.17, + "learning_rate": 4.134804457067949e-07, + "loss": 0.7665, + "step": 42035 + }, + { + "epoch": 9.17, + "learning_rate": 4.1337120384531355e-07, + "loss": 0.0596, + "step": 42036 + }, + { + "epoch": 9.17, + "learning_rate": 4.1326196198383226e-07, + "loss": 1.9001, + "step": 42037 + }, + { + "epoch": 9.17, + "learning_rate": 4.1315272012235093e-07, + "loss": 0.6132, + "step": 42038 + }, + { + "epoch": 9.17, + "learning_rate": 4.130434782608696e-07, + "loss": 0.5615, + "step": 42039 + }, + { + "epoch": 9.18, + "learning_rate": 4.129342363993883e-07, + "loss": 0.2245, + "step": 42040 + }, + { + "epoch": 9.18, + "learning_rate": 4.1282499453790697e-07, + "loss": 0.8478, + "step": 42041 + }, + { + "epoch": 9.18, + "learning_rate": 4.1271575267642564e-07, + "loss": 1.0168, + "step": 42042 + }, + { + "epoch": 9.18, + "learning_rate": 4.1260651081494435e-07, + "loss": 1.2941, + "step": 42043 + }, + { + "epoch": 9.18, + "learning_rate": 4.12497268953463e-07, + "loss": 1.3863, + "step": 42044 + }, + { + "epoch": 9.18, + "learning_rate": 4.123880270919817e-07, + "loss": 0.3423, + "step": 42045 + }, + { + "epoch": 9.18, + "learning_rate": 4.1227878523050035e-07, + "loss": 2.3301, + "step": 42046 + }, + { + "epoch": 9.18, + "learning_rate": 4.1216954336901906e-07, + "loss": 0.5155, + "step": 42047 + }, + { + "epoch": 9.18, + "learning_rate": 4.1206030150753773e-07, + "loss": 0.7574, + "step": 42048 + }, + { + "epoch": 9.18, + "learning_rate": 4.119510596460564e-07, + "loss": 2.3069, + "step": 42049 + }, + { + "epoch": 9.18, + "learning_rate": 4.118418177845751e-07, + "loss": 0.612, + "step": 42050 + }, + { + "epoch": 9.18, + "learning_rate": 4.1173257592309377e-07, + "loss": 0.3332, + "step": 42051 + }, + { + "epoch": 9.18, + "learning_rate": 4.1162333406161244e-07, + "loss": 1.437, + "step": 42052 + }, + { + "epoch": 9.18, + "learning_rate": 4.1151409220013115e-07, + "loss": 0.9177, + "step": 42053 + }, + { + "epoch": 9.18, + "learning_rate": 4.114048503386498e-07, + "loss": 1.7955, + "step": 42054 + }, + { + "epoch": 9.18, + "learning_rate": 4.112956084771685e-07, + "loss": 0.6975, + "step": 42055 + }, + { + "epoch": 9.18, + "learning_rate": 4.111863666156872e-07, + "loss": 4.1427, + "step": 42056 + }, + { + "epoch": 9.18, + "learning_rate": 4.1107712475420586e-07, + "loss": 1.8568, + "step": 42057 + }, + { + "epoch": 9.18, + "learning_rate": 4.109678828927245e-07, + "loss": 0.5473, + "step": 42058 + }, + { + "epoch": 9.18, + "learning_rate": 4.108586410312432e-07, + "loss": 2.0944, + "step": 42059 + }, + { + "epoch": 9.18, + "learning_rate": 4.107493991697619e-07, + "loss": 0.692, + "step": 42060 + }, + { + "epoch": 9.18, + "learning_rate": 4.1064015730828057e-07, + "loss": 1.7862, + "step": 42061 + }, + { + "epoch": 9.18, + "learning_rate": 4.1053091544679923e-07, + "loss": 0.5019, + "step": 42062 + }, + { + "epoch": 9.18, + "learning_rate": 4.1042167358531795e-07, + "loss": 1.0207, + "step": 42063 + }, + { + "epoch": 9.18, + "learning_rate": 4.103124317238366e-07, + "loss": 0.0644, + "step": 42064 + }, + { + "epoch": 9.18, + "learning_rate": 4.102031898623553e-07, + "loss": 1.8356, + "step": 42065 + }, + { + "epoch": 9.18, + "learning_rate": 4.10093948000874e-07, + "loss": 0.9165, + "step": 42066 + }, + { + "epoch": 9.18, + "learning_rate": 4.0998470613939266e-07, + "loss": 0.6714, + "step": 42067 + }, + { + "epoch": 9.18, + "learning_rate": 4.098754642779113e-07, + "loss": 1.4684, + "step": 42068 + }, + { + "epoch": 9.18, + "learning_rate": 4.0976622241643e-07, + "loss": 1.1667, + "step": 42069 + }, + { + "epoch": 9.18, + "learning_rate": 4.096569805549487e-07, + "loss": 1.522, + "step": 42070 + }, + { + "epoch": 9.18, + "learning_rate": 4.0954773869346737e-07, + "loss": 1.6549, + "step": 42071 + }, + { + "epoch": 9.18, + "learning_rate": 4.0943849683198603e-07, + "loss": 1.8676, + "step": 42072 + }, + { + "epoch": 9.18, + "learning_rate": 4.0932925497050475e-07, + "loss": 0.169, + "step": 42073 + }, + { + "epoch": 9.18, + "learning_rate": 4.092200131090234e-07, + "loss": 2.607, + "step": 42074 + }, + { + "epoch": 9.18, + "learning_rate": 4.091107712475421e-07, + "loss": 2.3522, + "step": 42075 + }, + { + "epoch": 9.18, + "learning_rate": 4.090015293860608e-07, + "loss": 0.2335, + "step": 42076 + }, + { + "epoch": 9.18, + "learning_rate": 4.0889228752457946e-07, + "loss": 0.2074, + "step": 42077 + }, + { + "epoch": 9.18, + "learning_rate": 4.087830456630981e-07, + "loss": 1.5215, + "step": 42078 + }, + { + "epoch": 9.18, + "learning_rate": 4.086738038016168e-07, + "loss": 0.8786, + "step": 42079 + }, + { + "epoch": 9.18, + "learning_rate": 4.085645619401355e-07, + "loss": 0.9541, + "step": 42080 + }, + { + "epoch": 9.18, + "learning_rate": 4.0845532007865417e-07, + "loss": 0.4104, + "step": 42081 + }, + { + "epoch": 9.18, + "learning_rate": 4.0834607821717283e-07, + "loss": 0.6199, + "step": 42082 + }, + { + "epoch": 9.18, + "learning_rate": 4.0823683635569155e-07, + "loss": 0.7399, + "step": 42083 + }, + { + "epoch": 9.18, + "learning_rate": 4.081275944942102e-07, + "loss": 1.008, + "step": 42084 + }, + { + "epoch": 9.18, + "learning_rate": 4.080183526327289e-07, + "loss": 1.6046, + "step": 42085 + }, + { + "epoch": 9.19, + "learning_rate": 4.079091107712476e-07, + "loss": 0.4336, + "step": 42086 + }, + { + "epoch": 9.19, + "learning_rate": 4.0779986890976626e-07, + "loss": 1.3797, + "step": 42087 + }, + { + "epoch": 9.19, + "learning_rate": 4.076906270482849e-07, + "loss": 0.4054, + "step": 42088 + }, + { + "epoch": 9.19, + "learning_rate": 4.075813851868036e-07, + "loss": 0.4545, + "step": 42089 + }, + { + "epoch": 9.19, + "learning_rate": 4.074721433253223e-07, + "loss": 0.4504, + "step": 42090 + }, + { + "epoch": 9.19, + "learning_rate": 4.0736290146384096e-07, + "loss": 0.222, + "step": 42091 + }, + { + "epoch": 9.19, + "learning_rate": 4.0725365960235963e-07, + "loss": 2.4069, + "step": 42092 + }, + { + "epoch": 9.19, + "learning_rate": 4.0714441774087834e-07, + "loss": 0.6006, + "step": 42093 + }, + { + "epoch": 9.19, + "learning_rate": 4.07035175879397e-07, + "loss": 1.2182, + "step": 42094 + }, + { + "epoch": 9.19, + "learning_rate": 4.0692593401791567e-07, + "loss": 0.5105, + "step": 42095 + }, + { + "epoch": 9.19, + "learning_rate": 4.068166921564344e-07, + "loss": 1.9867, + "step": 42096 + }, + { + "epoch": 9.19, + "learning_rate": 4.0670745029495305e-07, + "loss": 0.466, + "step": 42097 + }, + { + "epoch": 9.19, + "learning_rate": 4.065982084334717e-07, + "loss": 0.9792, + "step": 42098 + }, + { + "epoch": 9.19, + "learning_rate": 4.0648896657199043e-07, + "loss": 1.2093, + "step": 42099 + }, + { + "epoch": 9.19, + "learning_rate": 4.063797247105091e-07, + "loss": 0.3378, + "step": 42100 + }, + { + "epoch": 9.19, + "learning_rate": 4.0627048284902776e-07, + "loss": 2.2209, + "step": 42101 + }, + { + "epoch": 9.19, + "learning_rate": 4.061612409875464e-07, + "loss": 2.2286, + "step": 42102 + }, + { + "epoch": 9.19, + "learning_rate": 4.0605199912606514e-07, + "loss": 1.3475, + "step": 42103 + }, + { + "epoch": 9.19, + "learning_rate": 4.059427572645838e-07, + "loss": 0.2278, + "step": 42104 + }, + { + "epoch": 9.19, + "learning_rate": 4.0583351540310247e-07, + "loss": 0.4704, + "step": 42105 + }, + { + "epoch": 9.19, + "learning_rate": 4.057242735416212e-07, + "loss": 0.8436, + "step": 42106 + }, + { + "epoch": 9.19, + "learning_rate": 4.0561503168013985e-07, + "loss": 0.996, + "step": 42107 + }, + { + "epoch": 9.19, + "learning_rate": 4.055057898186585e-07, + "loss": 0.911, + "step": 42108 + }, + { + "epoch": 9.19, + "learning_rate": 4.0539654795717723e-07, + "loss": 2.8284, + "step": 42109 + }, + { + "epoch": 9.19, + "learning_rate": 4.052873060956959e-07, + "loss": 1.5777, + "step": 42110 + }, + { + "epoch": 9.19, + "learning_rate": 4.0517806423421456e-07, + "loss": 1.3946, + "step": 42111 + }, + { + "epoch": 9.19, + "learning_rate": 4.050688223727332e-07, + "loss": 1.4344, + "step": 42112 + }, + { + "epoch": 9.19, + "learning_rate": 4.0495958051125194e-07, + "loss": 1.1098, + "step": 42113 + }, + { + "epoch": 9.19, + "learning_rate": 4.048503386497706e-07, + "loss": 0.9959, + "step": 42114 + }, + { + "epoch": 9.19, + "learning_rate": 4.0474109678828927e-07, + "loss": 0.5978, + "step": 42115 + }, + { + "epoch": 9.19, + "learning_rate": 4.04631854926808e-07, + "loss": 1.5645, + "step": 42116 + }, + { + "epoch": 9.19, + "learning_rate": 4.0452261306532665e-07, + "loss": 1.0753, + "step": 42117 + }, + { + "epoch": 9.19, + "learning_rate": 4.044133712038453e-07, + "loss": 1.4068, + "step": 42118 + }, + { + "epoch": 9.19, + "learning_rate": 4.043041293423641e-07, + "loss": 1.5147, + "step": 42119 + }, + { + "epoch": 9.19, + "learning_rate": 4.041948874808827e-07, + "loss": 1.0222, + "step": 42120 + }, + { + "epoch": 9.19, + "learning_rate": 4.0408564561940136e-07, + "loss": 1.1124, + "step": 42121 + }, + { + "epoch": 9.19, + "learning_rate": 4.0397640375792e-07, + "loss": 0.9184, + "step": 42122 + }, + { + "epoch": 9.19, + "learning_rate": 4.0386716189643874e-07, + "loss": 0.9213, + "step": 42123 + }, + { + "epoch": 9.19, + "learning_rate": 4.037579200349574e-07, + "loss": 0.3907, + "step": 42124 + }, + { + "epoch": 9.19, + "learning_rate": 4.0364867817347607e-07, + "loss": 2.0886, + "step": 42125 + }, + { + "epoch": 9.19, + "learning_rate": 4.0353943631199484e-07, + "loss": 0.4263, + "step": 42126 + }, + { + "epoch": 9.19, + "learning_rate": 4.0343019445051345e-07, + "loss": 0.0779, + "step": 42127 + }, + { + "epoch": 9.19, + "learning_rate": 4.033209525890321e-07, + "loss": 1.2487, + "step": 42128 + }, + { + "epoch": 9.19, + "learning_rate": 4.032117107275509e-07, + "loss": 2.0648, + "step": 42129 + }, + { + "epoch": 9.19, + "learning_rate": 4.031024688660695e-07, + "loss": 0.756, + "step": 42130 + }, + { + "epoch": 9.19, + "learning_rate": 4.0299322700458816e-07, + "loss": 1.4157, + "step": 42131 + }, + { + "epoch": 9.2, + "learning_rate": 4.028839851431068e-07, + "loss": 1.8224, + "step": 42132 + }, + { + "epoch": 9.2, + "learning_rate": 4.027747432816256e-07, + "loss": 3.5227, + "step": 42133 + }, + { + "epoch": 9.2, + "learning_rate": 4.026655014201442e-07, + "loss": 0.5206, + "step": 42134 + }, + { + "epoch": 9.2, + "learning_rate": 4.0255625955866286e-07, + "loss": 1.9597, + "step": 42135 + }, + { + "epoch": 9.2, + "learning_rate": 4.0244701769718163e-07, + "loss": 0.4024, + "step": 42136 + }, + { + "epoch": 9.2, + "learning_rate": 4.023377758357003e-07, + "loss": 0.9442, + "step": 42137 + }, + { + "epoch": 9.2, + "learning_rate": 4.022285339742189e-07, + "loss": 0.625, + "step": 42138 + }, + { + "epoch": 9.2, + "learning_rate": 4.021192921127377e-07, + "loss": 0.0748, + "step": 42139 + }, + { + "epoch": 9.2, + "learning_rate": 4.0201005025125634e-07, + "loss": 0.8867, + "step": 42140 + }, + { + "epoch": 9.2, + "learning_rate": 4.0190080838977495e-07, + "loss": 0.9628, + "step": 42141 + }, + { + "epoch": 9.2, + "learning_rate": 4.017915665282937e-07, + "loss": 1.2844, + "step": 42142 + }, + { + "epoch": 9.2, + "learning_rate": 4.016823246668124e-07, + "loss": 1.3257, + "step": 42143 + }, + { + "epoch": 9.2, + "learning_rate": 4.0157308280533105e-07, + "loss": 1.2132, + "step": 42144 + }, + { + "epoch": 9.2, + "learning_rate": 4.0146384094384966e-07, + "loss": 1.1968, + "step": 42145 + }, + { + "epoch": 9.2, + "learning_rate": 4.0135459908236843e-07, + "loss": 2.9733, + "step": 42146 + }, + { + "epoch": 9.2, + "learning_rate": 4.012453572208871e-07, + "loss": 1.1369, + "step": 42147 + }, + { + "epoch": 9.2, + "learning_rate": 4.011361153594057e-07, + "loss": 0.7559, + "step": 42148 + }, + { + "epoch": 9.2, + "learning_rate": 4.010268734979245e-07, + "loss": 2.0322, + "step": 42149 + }, + { + "epoch": 9.2, + "learning_rate": 4.0091763163644314e-07, + "loss": 0.7814, + "step": 42150 + }, + { + "epoch": 9.2, + "learning_rate": 4.008083897749618e-07, + "loss": 1.5613, + "step": 42151 + }, + { + "epoch": 9.2, + "learning_rate": 4.006991479134805e-07, + "loss": 2.0365, + "step": 42152 + }, + { + "epoch": 9.2, + "learning_rate": 4.005899060519992e-07, + "loss": 0.2223, + "step": 42153 + }, + { + "epoch": 9.2, + "learning_rate": 4.0048066419051785e-07, + "loss": 1.4136, + "step": 42154 + }, + { + "epoch": 9.2, + "learning_rate": 4.003714223290365e-07, + "loss": 2.3471, + "step": 42155 + }, + { + "epoch": 9.2, + "learning_rate": 4.0026218046755523e-07, + "loss": 0.0198, + "step": 42156 + }, + { + "epoch": 9.2, + "learning_rate": 4.001529386060739e-07, + "loss": 0.8841, + "step": 42157 + }, + { + "epoch": 9.2, + "learning_rate": 4.0004369674459256e-07, + "loss": 1.7102, + "step": 42158 + }, + { + "epoch": 9.2, + "learning_rate": 3.999344548831113e-07, + "loss": 0.7825, + "step": 42159 + }, + { + "epoch": 9.2, + "learning_rate": 3.9982521302162994e-07, + "loss": 1.4327, + "step": 42160 + }, + { + "epoch": 9.2, + "learning_rate": 3.997159711601486e-07, + "loss": 0.723, + "step": 42161 + }, + { + "epoch": 9.2, + "learning_rate": 3.996067292986673e-07, + "loss": 0.6447, + "step": 42162 + }, + { + "epoch": 9.2, + "learning_rate": 3.99497487437186e-07, + "loss": 1.7319, + "step": 42163 + }, + { + "epoch": 9.2, + "learning_rate": 3.9938824557570465e-07, + "loss": 1.2751, + "step": 42164 + }, + { + "epoch": 9.2, + "learning_rate": 3.992790037142233e-07, + "loss": 1.6322, + "step": 42165 + }, + { + "epoch": 9.2, + "learning_rate": 3.9916976185274203e-07, + "loss": 1.8313, + "step": 42166 + }, + { + "epoch": 9.2, + "learning_rate": 3.990605199912607e-07, + "loss": 3.0486, + "step": 42167 + }, + { + "epoch": 9.2, + "learning_rate": 3.9895127812977936e-07, + "loss": 3.554, + "step": 42168 + }, + { + "epoch": 9.2, + "learning_rate": 3.988420362682981e-07, + "loss": 0.9317, + "step": 42169 + }, + { + "epoch": 9.2, + "learning_rate": 3.9873279440681674e-07, + "loss": 1.8057, + "step": 42170 + }, + { + "epoch": 9.2, + "learning_rate": 3.986235525453354e-07, + "loss": 1.1205, + "step": 42171 + }, + { + "epoch": 9.2, + "learning_rate": 3.985143106838541e-07, + "loss": 1.031, + "step": 42172 + }, + { + "epoch": 9.2, + "learning_rate": 3.984050688223728e-07, + "loss": 0.3676, + "step": 42173 + }, + { + "epoch": 9.2, + "learning_rate": 3.9829582696089145e-07, + "loss": 1.6428, + "step": 42174 + }, + { + "epoch": 9.2, + "learning_rate": 3.9818658509941016e-07, + "loss": 0.2613, + "step": 42175 + }, + { + "epoch": 9.2, + "learning_rate": 3.9807734323792883e-07, + "loss": 0.5259, + "step": 42176 + }, + { + "epoch": 9.2, + "learning_rate": 3.979681013764475e-07, + "loss": 1.0971, + "step": 42177 + }, + { + "epoch": 9.21, + "learning_rate": 3.9785885951496615e-07, + "loss": 1.5197, + "step": 42178 + }, + { + "epoch": 9.21, + "learning_rate": 3.9774961765348487e-07, + "loss": 1.024, + "step": 42179 + }, + { + "epoch": 9.21, + "learning_rate": 3.9764037579200354e-07, + "loss": 0.9748, + "step": 42180 + }, + { + "epoch": 9.21, + "learning_rate": 3.975311339305222e-07, + "loss": 1.8773, + "step": 42181 + }, + { + "epoch": 9.21, + "learning_rate": 3.974218920690409e-07, + "loss": 0.2125, + "step": 42182 + }, + { + "epoch": 9.21, + "learning_rate": 3.973126502075596e-07, + "loss": 1.0519, + "step": 42183 + }, + { + "epoch": 9.21, + "learning_rate": 3.9720340834607824e-07, + "loss": 0.479, + "step": 42184 + }, + { + "epoch": 9.21, + "learning_rate": 3.9709416648459696e-07, + "loss": 0.4276, + "step": 42185 + }, + { + "epoch": 9.21, + "learning_rate": 3.969849246231156e-07, + "loss": 1.7809, + "step": 42186 + }, + { + "epoch": 9.21, + "learning_rate": 3.968756827616343e-07, + "loss": 1.8143, + "step": 42187 + }, + { + "epoch": 9.21, + "learning_rate": 3.9676644090015295e-07, + "loss": 1.3363, + "step": 42188 + }, + { + "epoch": 9.21, + "learning_rate": 3.9665719903867167e-07, + "loss": 1.0702, + "step": 42189 + }, + { + "epoch": 9.21, + "learning_rate": 3.9654795717719033e-07, + "loss": 1.0619, + "step": 42190 + }, + { + "epoch": 9.21, + "learning_rate": 3.96438715315709e-07, + "loss": 0.7143, + "step": 42191 + }, + { + "epoch": 9.21, + "learning_rate": 3.963294734542277e-07, + "loss": 1.5711, + "step": 42192 + }, + { + "epoch": 9.21, + "learning_rate": 3.962202315927464e-07, + "loss": 1.3682, + "step": 42193 + }, + { + "epoch": 9.21, + "learning_rate": 3.9611098973126504e-07, + "loss": 2.0164, + "step": 42194 + }, + { + "epoch": 9.21, + "learning_rate": 3.9600174786978376e-07, + "loss": 2.0124, + "step": 42195 + }, + { + "epoch": 9.21, + "learning_rate": 3.958925060083024e-07, + "loss": 0.59, + "step": 42196 + }, + { + "epoch": 9.21, + "learning_rate": 3.957832641468211e-07, + "loss": 0.2267, + "step": 42197 + }, + { + "epoch": 9.21, + "learning_rate": 3.9567402228533975e-07, + "loss": 2.0018, + "step": 42198 + }, + { + "epoch": 9.21, + "learning_rate": 3.9556478042385847e-07, + "loss": 2.4676, + "step": 42199 + }, + { + "epoch": 9.21, + "learning_rate": 3.9545553856237713e-07, + "loss": 0.4425, + "step": 42200 + }, + { + "epoch": 9.21, + "learning_rate": 3.953462967008958e-07, + "loss": 1.0438, + "step": 42201 + }, + { + "epoch": 9.21, + "learning_rate": 3.952370548394145e-07, + "loss": 0.4213, + "step": 42202 + }, + { + "epoch": 9.21, + "learning_rate": 3.951278129779332e-07, + "loss": 2.4946, + "step": 42203 + }, + { + "epoch": 9.21, + "learning_rate": 3.9501857111645184e-07, + "loss": 0.6315, + "step": 42204 + }, + { + "epoch": 9.21, + "learning_rate": 3.9490932925497056e-07, + "loss": 0.1245, + "step": 42205 + }, + { + "epoch": 9.21, + "learning_rate": 3.948000873934892e-07, + "loss": 2.4627, + "step": 42206 + }, + { + "epoch": 9.21, + "learning_rate": 3.946908455320079e-07, + "loss": 0.2396, + "step": 42207 + }, + { + "epoch": 9.21, + "learning_rate": 3.9458160367052655e-07, + "loss": 1.1308, + "step": 42208 + }, + { + "epoch": 9.21, + "learning_rate": 3.9447236180904527e-07, + "loss": 1.0533, + "step": 42209 + }, + { + "epoch": 9.21, + "learning_rate": 3.9436311994756393e-07, + "loss": 1.4499, + "step": 42210 + }, + { + "epoch": 9.21, + "learning_rate": 3.942538780860826e-07, + "loss": 1.7135, + "step": 42211 + }, + { + "epoch": 9.21, + "learning_rate": 3.941446362246013e-07, + "loss": 0.3682, + "step": 42212 + }, + { + "epoch": 9.21, + "learning_rate": 3.9403539436312e-07, + "loss": 1.0992, + "step": 42213 + }, + { + "epoch": 9.21, + "learning_rate": 3.9392615250163864e-07, + "loss": 0.2005, + "step": 42214 + }, + { + "epoch": 9.21, + "learning_rate": 3.9381691064015736e-07, + "loss": 1.1723, + "step": 42215 + }, + { + "epoch": 9.21, + "learning_rate": 3.93707668778676e-07, + "loss": 1.3392, + "step": 42216 + }, + { + "epoch": 9.21, + "learning_rate": 3.935984269171947e-07, + "loss": 1.1996, + "step": 42217 + }, + { + "epoch": 9.21, + "learning_rate": 3.934891850557134e-07, + "loss": 1.3711, + "step": 42218 + }, + { + "epoch": 9.21, + "learning_rate": 3.9337994319423206e-07, + "loss": 0.3967, + "step": 42219 + }, + { + "epoch": 9.21, + "learning_rate": 3.9327070133275073e-07, + "loss": 1.1484, + "step": 42220 + }, + { + "epoch": 9.21, + "learning_rate": 3.931614594712694e-07, + "loss": 1.1481, + "step": 42221 + }, + { + "epoch": 9.21, + "learning_rate": 3.930522176097881e-07, + "loss": 0.344, + "step": 42222 + }, + { + "epoch": 9.21, + "learning_rate": 3.9294297574830677e-07, + "loss": 1.2752, + "step": 42223 + }, + { + "epoch": 9.22, + "learning_rate": 3.9283373388682544e-07, + "loss": 0.6008, + "step": 42224 + }, + { + "epoch": 9.22, + "learning_rate": 3.9272449202534415e-07, + "loss": 0.428, + "step": 42225 + }, + { + "epoch": 9.22, + "learning_rate": 3.926152501638628e-07, + "loss": 0.6338, + "step": 42226 + }, + { + "epoch": 9.22, + "learning_rate": 3.925060083023815e-07, + "loss": 0.7336, + "step": 42227 + }, + { + "epoch": 9.22, + "learning_rate": 3.923967664409002e-07, + "loss": 1.4499, + "step": 42228 + }, + { + "epoch": 9.22, + "learning_rate": 3.9228752457941886e-07, + "loss": 1.2804, + "step": 42229 + }, + { + "epoch": 9.22, + "learning_rate": 3.921782827179375e-07, + "loss": 0.6457, + "step": 42230 + }, + { + "epoch": 9.22, + "learning_rate": 3.920690408564562e-07, + "loss": 1.7519, + "step": 42231 + }, + { + "epoch": 9.22, + "learning_rate": 3.919597989949749e-07, + "loss": 1.0512, + "step": 42232 + }, + { + "epoch": 9.22, + "learning_rate": 3.9185055713349357e-07, + "loss": 1.6091, + "step": 42233 + }, + { + "epoch": 9.22, + "learning_rate": 3.9174131527201223e-07, + "loss": 2.0101, + "step": 42234 + }, + { + "epoch": 9.22, + "learning_rate": 3.9163207341053095e-07, + "loss": 3.3489, + "step": 42235 + }, + { + "epoch": 9.22, + "learning_rate": 3.915228315490496e-07, + "loss": 0.4164, + "step": 42236 + }, + { + "epoch": 9.22, + "learning_rate": 3.914135896875683e-07, + "loss": 1.8251, + "step": 42237 + }, + { + "epoch": 9.22, + "learning_rate": 3.91304347826087e-07, + "loss": 0.1598, + "step": 42238 + }, + { + "epoch": 9.22, + "learning_rate": 3.9119510596460566e-07, + "loss": 2.8219, + "step": 42239 + }, + { + "epoch": 9.22, + "learning_rate": 3.910858641031243e-07, + "loss": 1.8958, + "step": 42240 + }, + { + "epoch": 9.22, + "learning_rate": 3.90976622241643e-07, + "loss": 0.8989, + "step": 42241 + }, + { + "epoch": 9.22, + "learning_rate": 3.908673803801617e-07, + "loss": 0.4573, + "step": 42242 + }, + { + "epoch": 9.22, + "learning_rate": 3.9075813851868037e-07, + "loss": 0.5686, + "step": 42243 + }, + { + "epoch": 9.22, + "learning_rate": 3.9064889665719903e-07, + "loss": 1.5487, + "step": 42244 + }, + { + "epoch": 9.22, + "learning_rate": 3.9053965479571775e-07, + "loss": 1.3475, + "step": 42245 + }, + { + "epoch": 9.22, + "learning_rate": 3.904304129342364e-07, + "loss": 0.6218, + "step": 42246 + }, + { + "epoch": 9.22, + "learning_rate": 3.903211710727551e-07, + "loss": 1.479, + "step": 42247 + }, + { + "epoch": 9.22, + "learning_rate": 3.9021192921127385e-07, + "loss": 0.5626, + "step": 42248 + }, + { + "epoch": 9.22, + "learning_rate": 3.9010268734979246e-07, + "loss": 0.523, + "step": 42249 + }, + { + "epoch": 9.22, + "learning_rate": 3.899934454883111e-07, + "loss": 0.5377, + "step": 42250 + }, + { + "epoch": 9.22, + "learning_rate": 3.898842036268298e-07, + "loss": 0.7345, + "step": 42251 + }, + { + "epoch": 9.22, + "learning_rate": 3.897749617653485e-07, + "loss": 0.4275, + "step": 42252 + }, + { + "epoch": 9.22, + "learning_rate": 3.8966571990386717e-07, + "loss": 0.4603, + "step": 42253 + }, + { + "epoch": 9.22, + "learning_rate": 3.8955647804238583e-07, + "loss": 1.1964, + "step": 42254 + }, + { + "epoch": 9.22, + "learning_rate": 3.894472361809046e-07, + "loss": 0.6875, + "step": 42255 + }, + { + "epoch": 9.22, + "learning_rate": 3.893379943194232e-07, + "loss": 0.2429, + "step": 42256 + }, + { + "epoch": 9.22, + "learning_rate": 3.892287524579419e-07, + "loss": 1.4089, + "step": 42257 + }, + { + "epoch": 9.22, + "learning_rate": 3.8911951059646065e-07, + "loss": 2.4779, + "step": 42258 + }, + { + "epoch": 9.22, + "learning_rate": 3.8901026873497926e-07, + "loss": 2.1696, + "step": 42259 + }, + { + "epoch": 9.22, + "learning_rate": 3.889010268734979e-07, + "loss": 1.1048, + "step": 42260 + }, + { + "epoch": 9.22, + "learning_rate": 3.887917850120167e-07, + "loss": 1.4373, + "step": 42261 + }, + { + "epoch": 9.22, + "learning_rate": 3.8868254315053535e-07, + "loss": 1.3113, + "step": 42262 + }, + { + "epoch": 9.22, + "learning_rate": 3.8857330128905397e-07, + "loss": 0.4504, + "step": 42263 + }, + { + "epoch": 9.22, + "learning_rate": 3.8846405942757263e-07, + "loss": 0.1284, + "step": 42264 + }, + { + "epoch": 9.22, + "learning_rate": 3.883548175660914e-07, + "loss": 1.164, + "step": 42265 + }, + { + "epoch": 9.22, + "learning_rate": 3.8824557570461006e-07, + "loss": 0.0396, + "step": 42266 + }, + { + "epoch": 9.22, + "learning_rate": 3.881363338431287e-07, + "loss": 1.5171, + "step": 42267 + }, + { + "epoch": 9.22, + "learning_rate": 3.8802709198164744e-07, + "loss": 2.0104, + "step": 42268 + }, + { + "epoch": 9.23, + "learning_rate": 3.879178501201661e-07, + "loss": 2.4116, + "step": 42269 + }, + { + "epoch": 9.23, + "learning_rate": 3.878086082586847e-07, + "loss": 1.5611, + "step": 42270 + }, + { + "epoch": 9.23, + "learning_rate": 3.876993663972035e-07, + "loss": 0.7661, + "step": 42271 + }, + { + "epoch": 9.23, + "learning_rate": 3.8759012453572215e-07, + "loss": 1.9253, + "step": 42272 + }, + { + "epoch": 9.23, + "learning_rate": 3.874808826742408e-07, + "loss": 2.0776, + "step": 42273 + }, + { + "epoch": 9.23, + "learning_rate": 3.8737164081275943e-07, + "loss": 3.1549, + "step": 42274 + }, + { + "epoch": 9.23, + "learning_rate": 3.872623989512782e-07, + "loss": 1.207, + "step": 42275 + }, + { + "epoch": 9.23, + "learning_rate": 3.8715315708979686e-07, + "loss": 1.5825, + "step": 42276 + }, + { + "epoch": 9.23, + "learning_rate": 3.8704391522831547e-07, + "loss": 0.5198, + "step": 42277 + }, + { + "epoch": 9.23, + "learning_rate": 3.8693467336683424e-07, + "loss": 0.6124, + "step": 42278 + }, + { + "epoch": 9.23, + "learning_rate": 3.868254315053529e-07, + "loss": 2.4169, + "step": 42279 + }, + { + "epoch": 9.23, + "learning_rate": 3.8671618964387157e-07, + "loss": 0.7698, + "step": 42280 + }, + { + "epoch": 9.23, + "learning_rate": 3.866069477823903e-07, + "loss": 0.1927, + "step": 42281 + }, + { + "epoch": 9.23, + "learning_rate": 3.8649770592090895e-07, + "loss": 2.1907, + "step": 42282 + }, + { + "epoch": 9.23, + "learning_rate": 3.863884640594276e-07, + "loss": 0.612, + "step": 42283 + }, + { + "epoch": 9.23, + "learning_rate": 3.862792221979463e-07, + "loss": 1.1499, + "step": 42284 + }, + { + "epoch": 9.23, + "learning_rate": 3.86169980336465e-07, + "loss": 0.1101, + "step": 42285 + }, + { + "epoch": 9.23, + "learning_rate": 3.8606073847498366e-07, + "loss": 1.0837, + "step": 42286 + }, + { + "epoch": 9.23, + "learning_rate": 3.859514966135023e-07, + "loss": 0.383, + "step": 42287 + }, + { + "epoch": 9.23, + "learning_rate": 3.8584225475202104e-07, + "loss": 2.0188, + "step": 42288 + }, + { + "epoch": 9.23, + "learning_rate": 3.857330128905397e-07, + "loss": 2.0196, + "step": 42289 + }, + { + "epoch": 9.23, + "learning_rate": 3.8562377102905837e-07, + "loss": 0.343, + "step": 42290 + }, + { + "epoch": 9.23, + "learning_rate": 3.855145291675771e-07, + "loss": 1.0973, + "step": 42291 + }, + { + "epoch": 9.23, + "learning_rate": 3.8540528730609575e-07, + "loss": 0.088, + "step": 42292 + }, + { + "epoch": 9.23, + "learning_rate": 3.852960454446144e-07, + "loss": 1.1567, + "step": 42293 + }, + { + "epoch": 9.23, + "learning_rate": 3.851868035831331e-07, + "loss": 0.8163, + "step": 42294 + }, + { + "epoch": 9.23, + "learning_rate": 3.850775617216518e-07, + "loss": 0.9346, + "step": 42295 + }, + { + "epoch": 9.23, + "learning_rate": 3.8496831986017046e-07, + "loss": 0.1609, + "step": 42296 + }, + { + "epoch": 9.23, + "learning_rate": 3.848590779986891e-07, + "loss": 2.0159, + "step": 42297 + }, + { + "epoch": 9.23, + "learning_rate": 3.8474983613720784e-07, + "loss": 2.1294, + "step": 42298 + }, + { + "epoch": 9.23, + "learning_rate": 3.846405942757265e-07, + "loss": 0.5819, + "step": 42299 + }, + { + "epoch": 9.23, + "learning_rate": 3.8453135241424517e-07, + "loss": 1.1457, + "step": 42300 + }, + { + "epoch": 9.23, + "learning_rate": 3.844221105527639e-07, + "loss": 0.3844, + "step": 42301 + }, + { + "epoch": 9.23, + "learning_rate": 3.8431286869128255e-07, + "loss": 0.6898, + "step": 42302 + }, + { + "epoch": 9.23, + "learning_rate": 3.842036268298012e-07, + "loss": 2.5772, + "step": 42303 + }, + { + "epoch": 9.23, + "learning_rate": 3.8409438496831993e-07, + "loss": 1.213, + "step": 42304 + }, + { + "epoch": 9.23, + "learning_rate": 3.839851431068386e-07, + "loss": 1.3047, + "step": 42305 + }, + { + "epoch": 9.23, + "learning_rate": 3.8387590124535726e-07, + "loss": 1.2113, + "step": 42306 + }, + { + "epoch": 9.23, + "learning_rate": 3.837666593838759e-07, + "loss": 1.8927, + "step": 42307 + }, + { + "epoch": 9.23, + "learning_rate": 3.8365741752239464e-07, + "loss": 0.5643, + "step": 42308 + }, + { + "epoch": 9.23, + "learning_rate": 3.835481756609133e-07, + "loss": 0.6796, + "step": 42309 + }, + { + "epoch": 9.23, + "learning_rate": 3.8343893379943196e-07, + "loss": 0.1724, + "step": 42310 + }, + { + "epoch": 9.23, + "learning_rate": 3.833296919379507e-07, + "loss": 0.4311, + "step": 42311 + }, + { + "epoch": 9.23, + "learning_rate": 3.8322045007646934e-07, + "loss": 0.0493, + "step": 42312 + }, + { + "epoch": 9.23, + "learning_rate": 3.83111208214988e-07, + "loss": 0.1698, + "step": 42313 + }, + { + "epoch": 9.23, + "learning_rate": 3.830019663535067e-07, + "loss": 2.225, + "step": 42314 + }, + { + "epoch": 9.24, + "learning_rate": 3.828927244920254e-07, + "loss": 0.6821, + "step": 42315 + }, + { + "epoch": 9.24, + "learning_rate": 3.8278348263054405e-07, + "loss": 1.3861, + "step": 42316 + }, + { + "epoch": 9.24, + "learning_rate": 3.826742407690627e-07, + "loss": 0.8297, + "step": 42317 + }, + { + "epoch": 9.24, + "learning_rate": 3.8256499890758143e-07, + "loss": 1.5156, + "step": 42318 + }, + { + "epoch": 9.24, + "learning_rate": 3.824557570461001e-07, + "loss": 2.944, + "step": 42319 + }, + { + "epoch": 9.24, + "learning_rate": 3.8234651518461876e-07, + "loss": 0.4646, + "step": 42320 + }, + { + "epoch": 9.24, + "learning_rate": 3.822372733231375e-07, + "loss": 2.0663, + "step": 42321 + }, + { + "epoch": 9.24, + "learning_rate": 3.8212803146165614e-07, + "loss": 0.9728, + "step": 42322 + }, + { + "epoch": 9.24, + "learning_rate": 3.820187896001748e-07, + "loss": 1.6679, + "step": 42323 + }, + { + "epoch": 9.24, + "learning_rate": 3.819095477386935e-07, + "loss": 3.3352, + "step": 42324 + }, + { + "epoch": 9.24, + "learning_rate": 3.818003058772122e-07, + "loss": 1.8543, + "step": 42325 + }, + { + "epoch": 9.24, + "learning_rate": 3.8169106401573085e-07, + "loss": 0.9052, + "step": 42326 + }, + { + "epoch": 9.24, + "learning_rate": 3.815818221542495e-07, + "loss": 0.8484, + "step": 42327 + }, + { + "epoch": 9.24, + "learning_rate": 3.8147258029276823e-07, + "loss": 1.562, + "step": 42328 + }, + { + "epoch": 9.24, + "learning_rate": 3.813633384312869e-07, + "loss": 1.2124, + "step": 42329 + }, + { + "epoch": 9.24, + "learning_rate": 3.8125409656980556e-07, + "loss": 1.9536, + "step": 42330 + }, + { + "epoch": 9.24, + "learning_rate": 3.811448547083243e-07, + "loss": 0.8783, + "step": 42331 + }, + { + "epoch": 9.24, + "learning_rate": 3.8103561284684294e-07, + "loss": 1.4079, + "step": 42332 + }, + { + "epoch": 9.24, + "learning_rate": 3.809263709853616e-07, + "loss": 2.2636, + "step": 42333 + }, + { + "epoch": 9.24, + "learning_rate": 3.808171291238803e-07, + "loss": 0.4107, + "step": 42334 + }, + { + "epoch": 9.24, + "learning_rate": 3.80707887262399e-07, + "loss": 0.7083, + "step": 42335 + }, + { + "epoch": 9.24, + "learning_rate": 3.8059864540091765e-07, + "loss": 2.0753, + "step": 42336 + }, + { + "epoch": 9.24, + "learning_rate": 3.8048940353943637e-07, + "loss": 1.7079, + "step": 42337 + }, + { + "epoch": 9.24, + "learning_rate": 3.8038016167795503e-07, + "loss": 0.8266, + "step": 42338 + }, + { + "epoch": 9.24, + "learning_rate": 3.802709198164737e-07, + "loss": 1.4636, + "step": 42339 + }, + { + "epoch": 9.24, + "learning_rate": 3.8016167795499236e-07, + "loss": 1.0719, + "step": 42340 + }, + { + "epoch": 9.24, + "learning_rate": 3.800524360935111e-07, + "loss": 1.766, + "step": 42341 + }, + { + "epoch": 9.24, + "learning_rate": 3.7994319423202974e-07, + "loss": 0.4767, + "step": 42342 + }, + { + "epoch": 9.24, + "learning_rate": 3.798339523705484e-07, + "loss": 0.7038, + "step": 42343 + }, + { + "epoch": 9.24, + "learning_rate": 3.797247105090671e-07, + "loss": 0.5634, + "step": 42344 + }, + { + "epoch": 9.24, + "learning_rate": 3.796154686475858e-07, + "loss": 0.4651, + "step": 42345 + }, + { + "epoch": 9.24, + "learning_rate": 3.7950622678610445e-07, + "loss": 1.0789, + "step": 42346 + }, + { + "epoch": 9.24, + "learning_rate": 3.7939698492462316e-07, + "loss": 1.0783, + "step": 42347 + }, + { + "epoch": 9.24, + "learning_rate": 3.7928774306314183e-07, + "loss": 1.5437, + "step": 42348 + }, + { + "epoch": 9.24, + "learning_rate": 3.791785012016605e-07, + "loss": 0.8469, + "step": 42349 + }, + { + "epoch": 9.24, + "learning_rate": 3.7906925934017916e-07, + "loss": 1.2222, + "step": 42350 + }, + { + "epoch": 9.24, + "learning_rate": 3.7896001747869787e-07, + "loss": 0.5939, + "step": 42351 + }, + { + "epoch": 9.24, + "learning_rate": 3.7885077561721654e-07, + "loss": 0.9777, + "step": 42352 + }, + { + "epoch": 9.24, + "learning_rate": 3.787415337557352e-07, + "loss": 0.108, + "step": 42353 + }, + { + "epoch": 9.24, + "learning_rate": 3.786322918942539e-07, + "loss": 0.6222, + "step": 42354 + }, + { + "epoch": 9.24, + "learning_rate": 3.785230500327726e-07, + "loss": 0.6814, + "step": 42355 + }, + { + "epoch": 9.24, + "learning_rate": 3.7841380817129125e-07, + "loss": 1.0766, + "step": 42356 + }, + { + "epoch": 9.24, + "learning_rate": 3.7830456630980996e-07, + "loss": 1.4108, + "step": 42357 + }, + { + "epoch": 9.24, + "learning_rate": 3.781953244483286e-07, + "loss": 2.3947, + "step": 42358 + }, + { + "epoch": 9.24, + "learning_rate": 3.780860825868473e-07, + "loss": 1.634, + "step": 42359 + }, + { + "epoch": 9.24, + "learning_rate": 3.7797684072536595e-07, + "loss": 0.5277, + "step": 42360 + }, + { + "epoch": 9.25, + "learning_rate": 3.7786759886388467e-07, + "loss": 0.8375, + "step": 42361 + }, + { + "epoch": 9.25, + "learning_rate": 3.7775835700240334e-07, + "loss": 0.7496, + "step": 42362 + }, + { + "epoch": 9.25, + "learning_rate": 3.77649115140922e-07, + "loss": 2.2619, + "step": 42363 + }, + { + "epoch": 9.25, + "learning_rate": 3.775398732794407e-07, + "loss": 1.3359, + "step": 42364 + }, + { + "epoch": 9.25, + "learning_rate": 3.774306314179594e-07, + "loss": 1.2876, + "step": 42365 + }, + { + "epoch": 9.25, + "learning_rate": 3.7732138955647804e-07, + "loss": 0.5952, + "step": 42366 + }, + { + "epoch": 9.25, + "learning_rate": 3.7721214769499676e-07, + "loss": 1.6153, + "step": 42367 + }, + { + "epoch": 9.25, + "learning_rate": 3.771029058335154e-07, + "loss": 1.0039, + "step": 42368 + }, + { + "epoch": 9.25, + "learning_rate": 3.769936639720341e-07, + "loss": 0.1795, + "step": 42369 + }, + { + "epoch": 9.25, + "learning_rate": 3.7688442211055275e-07, + "loss": 0.2773, + "step": 42370 + }, + { + "epoch": 9.25, + "learning_rate": 3.7677518024907147e-07, + "loss": 0.5431, + "step": 42371 + }, + { + "epoch": 9.25, + "learning_rate": 3.7666593838759013e-07, + "loss": 0.9086, + "step": 42372 + }, + { + "epoch": 9.25, + "learning_rate": 3.765566965261088e-07, + "loss": 0.3975, + "step": 42373 + }, + { + "epoch": 9.25, + "learning_rate": 3.764474546646275e-07, + "loss": 1.1948, + "step": 42374 + }, + { + "epoch": 9.25, + "learning_rate": 3.763382128031462e-07, + "loss": 0.1624, + "step": 42375 + }, + { + "epoch": 9.25, + "learning_rate": 3.7622897094166484e-07, + "loss": 1.2139, + "step": 42376 + }, + { + "epoch": 9.25, + "learning_rate": 3.7611972908018356e-07, + "loss": 0.3111, + "step": 42377 + }, + { + "epoch": 9.25, + "learning_rate": 3.760104872187022e-07, + "loss": 0.229, + "step": 42378 + }, + { + "epoch": 9.25, + "learning_rate": 3.759012453572209e-07, + "loss": 0.2191, + "step": 42379 + }, + { + "epoch": 9.25, + "learning_rate": 3.7579200349573966e-07, + "loss": 0.9514, + "step": 42380 + }, + { + "epoch": 9.25, + "learning_rate": 3.7568276163425827e-07, + "loss": 0.1888, + "step": 42381 + }, + { + "epoch": 9.25, + "learning_rate": 3.7557351977277693e-07, + "loss": 2.8368, + "step": 42382 + }, + { + "epoch": 9.25, + "learning_rate": 3.754642779112956e-07, + "loss": 2.0872, + "step": 42383 + }, + { + "epoch": 9.25, + "learning_rate": 3.7535503604981436e-07, + "loss": 1.0563, + "step": 42384 + }, + { + "epoch": 9.25, + "learning_rate": 3.75245794188333e-07, + "loss": 0.9518, + "step": 42385 + }, + { + "epoch": 9.25, + "learning_rate": 3.7513655232685164e-07, + "loss": 2.6312, + "step": 42386 + }, + { + "epoch": 9.25, + "learning_rate": 3.750273104653704e-07, + "loss": 1.3841, + "step": 42387 + }, + { + "epoch": 9.25, + "learning_rate": 3.74918068603889e-07, + "loss": 0.8375, + "step": 42388 + }, + { + "epoch": 9.25, + "learning_rate": 3.748088267424077e-07, + "loss": 0.5373, + "step": 42389 + }, + { + "epoch": 9.25, + "learning_rate": 3.7469958488092645e-07, + "loss": 0.6715, + "step": 42390 + }, + { + "epoch": 9.25, + "learning_rate": 3.745903430194451e-07, + "loss": 0.3678, + "step": 42391 + }, + { + "epoch": 9.25, + "learning_rate": 3.7448110115796373e-07, + "loss": 2.8835, + "step": 42392 + }, + { + "epoch": 9.25, + "learning_rate": 3.743718592964824e-07, + "loss": 1.0383, + "step": 42393 + }, + { + "epoch": 9.25, + "learning_rate": 3.7426261743500116e-07, + "loss": 0.8791, + "step": 42394 + }, + { + "epoch": 9.25, + "learning_rate": 3.741533755735198e-07, + "loss": 0.8537, + "step": 42395 + }, + { + "epoch": 9.25, + "learning_rate": 3.7404413371203844e-07, + "loss": 0.6719, + "step": 42396 + }, + { + "epoch": 9.25, + "learning_rate": 3.739348918505572e-07, + "loss": 1.3893, + "step": 42397 + }, + { + "epoch": 9.25, + "learning_rate": 3.7382564998907587e-07, + "loss": 2.1046, + "step": 42398 + }, + { + "epoch": 9.25, + "learning_rate": 3.737164081275945e-07, + "loss": 0.1812, + "step": 42399 + }, + { + "epoch": 9.25, + "learning_rate": 3.7360716626611325e-07, + "loss": 0.3451, + "step": 42400 + }, + { + "epoch": 9.25, + "learning_rate": 3.734979244046319e-07, + "loss": 0.8193, + "step": 42401 + }, + { + "epoch": 9.25, + "learning_rate": 3.733886825431506e-07, + "loss": 0.819, + "step": 42402 + }, + { + "epoch": 9.25, + "learning_rate": 3.732794406816692e-07, + "loss": 1.2837, + "step": 42403 + }, + { + "epoch": 9.25, + "learning_rate": 3.7317019882018796e-07, + "loss": 1.1332, + "step": 42404 + }, + { + "epoch": 9.25, + "learning_rate": 3.730609569587066e-07, + "loss": 1.1412, + "step": 42405 + }, + { + "epoch": 9.25, + "learning_rate": 3.7295171509722524e-07, + "loss": 1.2433, + "step": 42406 + }, + { + "epoch": 9.26, + "learning_rate": 3.72842473235744e-07, + "loss": 1.2408, + "step": 42407 + }, + { + "epoch": 9.26, + "learning_rate": 3.7273323137426267e-07, + "loss": 0.4109, + "step": 42408 + }, + { + "epoch": 9.26, + "learning_rate": 3.7262398951278133e-07, + "loss": 0.6828, + "step": 42409 + }, + { + "epoch": 9.26, + "learning_rate": 3.7251474765130005e-07, + "loss": 2.2669, + "step": 42410 + }, + { + "epoch": 9.26, + "learning_rate": 3.724055057898187e-07, + "loss": 1.5999, + "step": 42411 + }, + { + "epoch": 9.26, + "learning_rate": 3.722962639283374e-07, + "loss": 3.2895, + "step": 42412 + }, + { + "epoch": 9.26, + "learning_rate": 3.72187022066856e-07, + "loss": 1.2179, + "step": 42413 + }, + { + "epoch": 9.26, + "learning_rate": 3.7207778020537476e-07, + "loss": 0.1246, + "step": 42414 + }, + { + "epoch": 9.26, + "learning_rate": 3.719685383438934e-07, + "loss": 1.1705, + "step": 42415 + }, + { + "epoch": 9.26, + "learning_rate": 3.718592964824121e-07, + "loss": 1.059, + "step": 42416 + }, + { + "epoch": 9.26, + "learning_rate": 3.717500546209308e-07, + "loss": 0.6763, + "step": 42417 + }, + { + "epoch": 9.26, + "learning_rate": 3.7164081275944947e-07, + "loss": 1.9304, + "step": 42418 + }, + { + "epoch": 9.26, + "learning_rate": 3.7153157089796813e-07, + "loss": 0.9622, + "step": 42419 + }, + { + "epoch": 9.26, + "learning_rate": 3.7142232903648685e-07, + "loss": 0.6175, + "step": 42420 + }, + { + "epoch": 9.26, + "learning_rate": 3.713130871750055e-07, + "loss": 0.7986, + "step": 42421 + }, + { + "epoch": 9.26, + "learning_rate": 3.712038453135242e-07, + "loss": 1.1899, + "step": 42422 + }, + { + "epoch": 9.26, + "learning_rate": 3.710946034520429e-07, + "loss": 0.4417, + "step": 42423 + }, + { + "epoch": 9.26, + "learning_rate": 3.7098536159056156e-07, + "loss": 0.969, + "step": 42424 + }, + { + "epoch": 9.26, + "learning_rate": 3.708761197290802e-07, + "loss": 1.4234, + "step": 42425 + }, + { + "epoch": 9.26, + "learning_rate": 3.707668778675989e-07, + "loss": 1.3714, + "step": 42426 + }, + { + "epoch": 9.26, + "learning_rate": 3.706576360061176e-07, + "loss": 0.6083, + "step": 42427 + }, + { + "epoch": 9.26, + "learning_rate": 3.7054839414463627e-07, + "loss": 2.9638, + "step": 42428 + }, + { + "epoch": 9.26, + "learning_rate": 3.7043915228315493e-07, + "loss": 2.8681, + "step": 42429 + }, + { + "epoch": 9.26, + "learning_rate": 3.7032991042167365e-07, + "loss": 0.7811, + "step": 42430 + }, + { + "epoch": 9.26, + "learning_rate": 3.702206685601923e-07, + "loss": 1.1709, + "step": 42431 + }, + { + "epoch": 9.26, + "learning_rate": 3.70111426698711e-07, + "loss": 1.0219, + "step": 42432 + }, + { + "epoch": 9.26, + "learning_rate": 3.700021848372297e-07, + "loss": 1.1894, + "step": 42433 + }, + { + "epoch": 9.26, + "learning_rate": 3.6989294297574836e-07, + "loss": 1.0239, + "step": 42434 + }, + { + "epoch": 9.26, + "learning_rate": 3.69783701114267e-07, + "loss": 0.0433, + "step": 42435 + }, + { + "epoch": 9.26, + "learning_rate": 3.696744592527857e-07, + "loss": 0.9653, + "step": 42436 + }, + { + "epoch": 9.26, + "learning_rate": 3.695652173913044e-07, + "loss": 1.1107, + "step": 42437 + }, + { + "epoch": 9.26, + "learning_rate": 3.6945597552982306e-07, + "loss": 0.4194, + "step": 42438 + }, + { + "epoch": 9.26, + "learning_rate": 3.6934673366834173e-07, + "loss": 1.7355, + "step": 42439 + }, + { + "epoch": 9.26, + "learning_rate": 3.6923749180686044e-07, + "loss": 1.0701, + "step": 42440 + }, + { + "epoch": 9.26, + "learning_rate": 3.691282499453791e-07, + "loss": 1.128, + "step": 42441 + }, + { + "epoch": 9.26, + "learning_rate": 3.6901900808389777e-07, + "loss": 0.5002, + "step": 42442 + }, + { + "epoch": 9.26, + "learning_rate": 3.689097662224165e-07, + "loss": 0.6916, + "step": 42443 + }, + { + "epoch": 9.26, + "learning_rate": 3.6880052436093515e-07, + "loss": 0.7175, + "step": 42444 + }, + { + "epoch": 9.26, + "learning_rate": 3.686912824994538e-07, + "loss": 2.4963, + "step": 42445 + }, + { + "epoch": 9.26, + "learning_rate": 3.685820406379725e-07, + "loss": 0.5991, + "step": 42446 + }, + { + "epoch": 9.26, + "learning_rate": 3.684727987764912e-07, + "loss": 2.6665, + "step": 42447 + }, + { + "epoch": 9.26, + "learning_rate": 3.6836355691500986e-07, + "loss": 0.6347, + "step": 42448 + }, + { + "epoch": 9.26, + "learning_rate": 3.682543150535285e-07, + "loss": 1.3955, + "step": 42449 + }, + { + "epoch": 9.26, + "learning_rate": 3.6814507319204724e-07, + "loss": 0.3912, + "step": 42450 + }, + { + "epoch": 9.26, + "learning_rate": 3.680358313305659e-07, + "loss": 1.1817, + "step": 42451 + }, + { + "epoch": 9.26, + "learning_rate": 3.6792658946908457e-07, + "loss": 1.2265, + "step": 42452 + }, + { + "epoch": 9.27, + "learning_rate": 3.678173476076033e-07, + "loss": 0.4463, + "step": 42453 + }, + { + "epoch": 9.27, + "learning_rate": 3.6770810574612195e-07, + "loss": 0.6667, + "step": 42454 + }, + { + "epoch": 9.27, + "learning_rate": 3.675988638846406e-07, + "loss": 0.3001, + "step": 42455 + }, + { + "epoch": 9.27, + "learning_rate": 3.674896220231593e-07, + "loss": 0.6838, + "step": 42456 + }, + { + "epoch": 9.27, + "learning_rate": 3.67380380161678e-07, + "loss": 1.3873, + "step": 42457 + }, + { + "epoch": 9.27, + "learning_rate": 3.6727113830019666e-07, + "loss": 0.5246, + "step": 42458 + }, + { + "epoch": 9.27, + "learning_rate": 3.671618964387153e-07, + "loss": 0.4507, + "step": 42459 + }, + { + "epoch": 9.27, + "learning_rate": 3.6705265457723404e-07, + "loss": 2.1232, + "step": 42460 + }, + { + "epoch": 9.27, + "learning_rate": 3.669434127157527e-07, + "loss": 0.5064, + "step": 42461 + }, + { + "epoch": 9.27, + "learning_rate": 3.6683417085427137e-07, + "loss": 0.8614, + "step": 42462 + }, + { + "epoch": 9.27, + "learning_rate": 3.667249289927901e-07, + "loss": 0.6024, + "step": 42463 + }, + { + "epoch": 9.27, + "learning_rate": 3.6661568713130875e-07, + "loss": 1.9571, + "step": 42464 + }, + { + "epoch": 9.27, + "learning_rate": 3.665064452698274e-07, + "loss": 1.5023, + "step": 42465 + }, + { + "epoch": 9.27, + "learning_rate": 3.6639720340834613e-07, + "loss": 1.5117, + "step": 42466 + }, + { + "epoch": 9.27, + "learning_rate": 3.662879615468648e-07, + "loss": 1.2712, + "step": 42467 + }, + { + "epoch": 9.27, + "learning_rate": 3.6617871968538346e-07, + "loss": 0.5407, + "step": 42468 + }, + { + "epoch": 9.27, + "learning_rate": 3.660694778239021e-07, + "loss": 0.2295, + "step": 42469 + }, + { + "epoch": 9.27, + "learning_rate": 3.6596023596242084e-07, + "loss": 1.8323, + "step": 42470 + }, + { + "epoch": 9.27, + "learning_rate": 3.658509941009395e-07, + "loss": 1.668, + "step": 42471 + }, + { + "epoch": 9.27, + "learning_rate": 3.6574175223945817e-07, + "loss": 0.5166, + "step": 42472 + }, + { + "epoch": 9.27, + "learning_rate": 3.656325103779769e-07, + "loss": 1.4745, + "step": 42473 + }, + { + "epoch": 9.27, + "learning_rate": 3.6552326851649555e-07, + "loss": 0.3606, + "step": 42474 + }, + { + "epoch": 9.27, + "learning_rate": 3.654140266550142e-07, + "loss": 0.9156, + "step": 42475 + }, + { + "epoch": 9.27, + "learning_rate": 3.6530478479353293e-07, + "loss": 2.4291, + "step": 42476 + }, + { + "epoch": 9.27, + "learning_rate": 3.651955429320516e-07, + "loss": 2.6441, + "step": 42477 + }, + { + "epoch": 9.27, + "learning_rate": 3.6508630107057026e-07, + "loss": 2.0211, + "step": 42478 + }, + { + "epoch": 9.27, + "learning_rate": 3.649770592090889e-07, + "loss": 0.9636, + "step": 42479 + }, + { + "epoch": 9.27, + "learning_rate": 3.6486781734760764e-07, + "loss": 0.9913, + "step": 42480 + }, + { + "epoch": 9.27, + "learning_rate": 3.647585754861263e-07, + "loss": 1.0492, + "step": 42481 + }, + { + "epoch": 9.27, + "learning_rate": 3.6464933362464497e-07, + "loss": 0.4968, + "step": 42482 + }, + { + "epoch": 9.27, + "learning_rate": 3.645400917631637e-07, + "loss": 0.7437, + "step": 42483 + }, + { + "epoch": 9.27, + "learning_rate": 3.6443084990168235e-07, + "loss": 1.2305, + "step": 42484 + }, + { + "epoch": 9.27, + "learning_rate": 3.64321608040201e-07, + "loss": 1.3282, + "step": 42485 + }, + { + "epoch": 9.27, + "learning_rate": 3.6421236617871973e-07, + "loss": 0.894, + "step": 42486 + }, + { + "epoch": 9.27, + "learning_rate": 3.641031243172384e-07, + "loss": 0.6489, + "step": 42487 + }, + { + "epoch": 9.27, + "learning_rate": 3.6399388245575705e-07, + "loss": 1.5332, + "step": 42488 + }, + { + "epoch": 9.27, + "learning_rate": 3.638846405942757e-07, + "loss": 1.5897, + "step": 42489 + }, + { + "epoch": 9.27, + "learning_rate": 3.6377539873279444e-07, + "loss": 1.3239, + "step": 42490 + }, + { + "epoch": 9.27, + "learning_rate": 3.636661568713131e-07, + "loss": 2.0289, + "step": 42491 + }, + { + "epoch": 9.27, + "learning_rate": 3.6355691500983176e-07, + "loss": 0.7773, + "step": 42492 + }, + { + "epoch": 9.27, + "learning_rate": 3.634476731483505e-07, + "loss": 2.0838, + "step": 42493 + }, + { + "epoch": 9.27, + "learning_rate": 3.6333843128686914e-07, + "loss": 1.6111, + "step": 42494 + }, + { + "epoch": 9.27, + "learning_rate": 3.632291894253878e-07, + "loss": 1.261, + "step": 42495 + }, + { + "epoch": 9.27, + "learning_rate": 3.631199475639065e-07, + "loss": 1.3068, + "step": 42496 + }, + { + "epoch": 9.27, + "learning_rate": 3.630107057024252e-07, + "loss": 3.3794, + "step": 42497 + }, + { + "epoch": 9.27, + "learning_rate": 3.6290146384094385e-07, + "loss": 1.6327, + "step": 42498 + }, + { + "epoch": 9.28, + "learning_rate": 3.6279222197946257e-07, + "loss": 1.917, + "step": 42499 + }, + { + "epoch": 9.28, + "learning_rate": 3.6268298011798123e-07, + "loss": 3.0218, + "step": 42500 + }, + { + "epoch": 9.28, + "learning_rate": 3.625737382564999e-07, + "loss": 1.0752, + "step": 42501 + }, + { + "epoch": 9.28, + "learning_rate": 3.6246449639501856e-07, + "loss": 0.2944, + "step": 42502 + }, + { + "epoch": 9.28, + "learning_rate": 3.623552545335373e-07, + "loss": 0.5587, + "step": 42503 + }, + { + "epoch": 9.28, + "learning_rate": 3.6224601267205594e-07, + "loss": 0.9379, + "step": 42504 + }, + { + "epoch": 9.28, + "learning_rate": 3.621367708105746e-07, + "loss": 1.9602, + "step": 42505 + }, + { + "epoch": 9.28, + "learning_rate": 3.620275289490933e-07, + "loss": 0.6597, + "step": 42506 + }, + { + "epoch": 9.28, + "learning_rate": 3.61918287087612e-07, + "loss": 1.1334, + "step": 42507 + }, + { + "epoch": 9.28, + "learning_rate": 3.6180904522613065e-07, + "loss": 1.2859, + "step": 42508 + }, + { + "epoch": 9.28, + "learning_rate": 3.616998033646494e-07, + "loss": 0.2246, + "step": 42509 + }, + { + "epoch": 9.28, + "learning_rate": 3.6159056150316803e-07, + "loss": 0.9151, + "step": 42510 + }, + { + "epoch": 9.28, + "learning_rate": 3.614813196416867e-07, + "loss": 2.1251, + "step": 42511 + }, + { + "epoch": 9.28, + "learning_rate": 3.6137207778020536e-07, + "loss": 0.597, + "step": 42512 + }, + { + "epoch": 9.28, + "learning_rate": 3.6126283591872413e-07, + "loss": 1.408, + "step": 42513 + }, + { + "epoch": 9.28, + "learning_rate": 3.6115359405724274e-07, + "loss": 3.3057, + "step": 42514 + }, + { + "epoch": 9.28, + "learning_rate": 3.610443521957614e-07, + "loss": 0.5837, + "step": 42515 + }, + { + "epoch": 9.28, + "learning_rate": 3.609351103342802e-07, + "loss": 0.5556, + "step": 42516 + }, + { + "epoch": 9.28, + "learning_rate": 3.608258684727988e-07, + "loss": 1.7387, + "step": 42517 + }, + { + "epoch": 9.28, + "learning_rate": 3.6071662661131745e-07, + "loss": 1.4027, + "step": 42518 + }, + { + "epoch": 9.28, + "learning_rate": 3.606073847498362e-07, + "loss": 1.4571, + "step": 42519 + }, + { + "epoch": 9.28, + "learning_rate": 3.604981428883549e-07, + "loss": 0.1735, + "step": 42520 + }, + { + "epoch": 9.28, + "learning_rate": 3.603889010268735e-07, + "loss": 0.9815, + "step": 42521 + }, + { + "epoch": 9.28, + "learning_rate": 3.6027965916539216e-07, + "loss": 2.3555, + "step": 42522 + }, + { + "epoch": 9.28, + "learning_rate": 3.6017041730391093e-07, + "loss": 0.8902, + "step": 42523 + }, + { + "epoch": 9.28, + "learning_rate": 3.6006117544242954e-07, + "loss": 1.287, + "step": 42524 + }, + { + "epoch": 9.28, + "learning_rate": 3.599519335809482e-07, + "loss": 2.5812, + "step": 42525 + }, + { + "epoch": 9.28, + "learning_rate": 3.5984269171946697e-07, + "loss": 0.8633, + "step": 42526 + }, + { + "epoch": 9.28, + "learning_rate": 3.5973344985798564e-07, + "loss": 1.4196, + "step": 42527 + }, + { + "epoch": 9.28, + "learning_rate": 3.5962420799650425e-07, + "loss": 0.63, + "step": 42528 + }, + { + "epoch": 9.28, + "learning_rate": 3.59514966135023e-07, + "loss": 0.0756, + "step": 42529 + }, + { + "epoch": 9.28, + "learning_rate": 3.594057242735417e-07, + "loss": 1.749, + "step": 42530 + }, + { + "epoch": 9.28, + "learning_rate": 3.5929648241206034e-07, + "loss": 0.3111, + "step": 42531 + }, + { + "epoch": 9.28, + "learning_rate": 3.5918724055057896e-07, + "loss": 0.3968, + "step": 42532 + }, + { + "epoch": 9.28, + "learning_rate": 3.590779986890977e-07, + "loss": 0.4597, + "step": 42533 + }, + { + "epoch": 9.28, + "learning_rate": 3.589687568276164e-07, + "loss": 1.2588, + "step": 42534 + }, + { + "epoch": 9.28, + "learning_rate": 3.58859514966135e-07, + "loss": 0.5653, + "step": 42535 + }, + { + "epoch": 9.28, + "learning_rate": 3.5875027310465377e-07, + "loss": 1.1799, + "step": 42536 + }, + { + "epoch": 9.28, + "learning_rate": 3.5864103124317243e-07, + "loss": 0.0619, + "step": 42537 + }, + { + "epoch": 9.28, + "learning_rate": 3.585317893816911e-07, + "loss": 0.1607, + "step": 42538 + }, + { + "epoch": 9.28, + "learning_rate": 3.584225475202098e-07, + "loss": 0.9249, + "step": 42539 + }, + { + "epoch": 9.28, + "learning_rate": 3.583133056587285e-07, + "loss": 2.2771, + "step": 42540 + }, + { + "epoch": 9.28, + "learning_rate": 3.5820406379724714e-07, + "loss": 0.3593, + "step": 42541 + }, + { + "epoch": 9.28, + "learning_rate": 3.5809482193576586e-07, + "loss": 1.6996, + "step": 42542 + }, + { + "epoch": 9.28, + "learning_rate": 3.579855800742845e-07, + "loss": 1.0418, + "step": 42543 + }, + { + "epoch": 9.29, + "learning_rate": 3.578763382128032e-07, + "loss": 0.4525, + "step": 42544 + }, + { + "epoch": 9.29, + "learning_rate": 3.5776709635132185e-07, + "loss": 1.2894, + "step": 42545 + }, + { + "epoch": 9.29, + "learning_rate": 3.5765785448984057e-07, + "loss": 1.2011, + "step": 42546 + }, + { + "epoch": 9.29, + "learning_rate": 3.5754861262835923e-07, + "loss": 1.7509, + "step": 42547 + }, + { + "epoch": 9.29, + "learning_rate": 3.574393707668779e-07, + "loss": 1.9307, + "step": 42548 + }, + { + "epoch": 9.29, + "learning_rate": 3.573301289053966e-07, + "loss": 1.1141, + "step": 42549 + }, + { + "epoch": 9.29, + "learning_rate": 3.572208870439153e-07, + "loss": 0.552, + "step": 42550 + }, + { + "epoch": 9.29, + "learning_rate": 3.5711164518243394e-07, + "loss": 1.9933, + "step": 42551 + }, + { + "epoch": 9.29, + "learning_rate": 3.5700240332095266e-07, + "loss": 0.6375, + "step": 42552 + }, + { + "epoch": 9.29, + "learning_rate": 3.568931614594713e-07, + "loss": 0.0886, + "step": 42553 + }, + { + "epoch": 9.29, + "learning_rate": 3.5678391959799e-07, + "loss": 3.2001, + "step": 42554 + }, + { + "epoch": 9.29, + "learning_rate": 3.5667467773650865e-07, + "loss": 0.1395, + "step": 42555 + }, + { + "epoch": 9.29, + "learning_rate": 3.5656543587502737e-07, + "loss": 1.0685, + "step": 42556 + }, + { + "epoch": 9.29, + "learning_rate": 3.5645619401354603e-07, + "loss": 1.6898, + "step": 42557 + }, + { + "epoch": 9.29, + "learning_rate": 3.563469521520647e-07, + "loss": 0.6947, + "step": 42558 + }, + { + "epoch": 9.29, + "learning_rate": 3.562377102905834e-07, + "loss": 0.6257, + "step": 42559 + }, + { + "epoch": 9.29, + "learning_rate": 3.561284684291021e-07, + "loss": 1.5712, + "step": 42560 + }, + { + "epoch": 9.29, + "learning_rate": 3.5601922656762074e-07, + "loss": 2.9378, + "step": 42561 + }, + { + "epoch": 9.29, + "learning_rate": 3.5590998470613946e-07, + "loss": 1.1302, + "step": 42562 + }, + { + "epoch": 9.29, + "learning_rate": 3.558007428446581e-07, + "loss": 0.6492, + "step": 42563 + }, + { + "epoch": 9.29, + "learning_rate": 3.556915009831768e-07, + "loss": 1.4216, + "step": 42564 + }, + { + "epoch": 9.29, + "learning_rate": 3.5558225912169545e-07, + "loss": 1.5072, + "step": 42565 + }, + { + "epoch": 9.29, + "learning_rate": 3.5547301726021416e-07, + "loss": 0.061, + "step": 42566 + }, + { + "epoch": 9.29, + "learning_rate": 3.5536377539873283e-07, + "loss": 0.6211, + "step": 42567 + }, + { + "epoch": 9.29, + "learning_rate": 3.552545335372515e-07, + "loss": 0.6586, + "step": 42568 + }, + { + "epoch": 9.29, + "learning_rate": 3.551452916757702e-07, + "loss": 1.1564, + "step": 42569 + }, + { + "epoch": 9.29, + "learning_rate": 3.5503604981428887e-07, + "loss": 1.3051, + "step": 42570 + }, + { + "epoch": 9.29, + "learning_rate": 3.5492680795280754e-07, + "loss": 2.8854, + "step": 42571 + }, + { + "epoch": 9.29, + "learning_rate": 3.5481756609132625e-07, + "loss": 2.3368, + "step": 42572 + }, + { + "epoch": 9.29, + "learning_rate": 3.547083242298449e-07, + "loss": 1.8309, + "step": 42573 + }, + { + "epoch": 9.29, + "learning_rate": 3.545990823683636e-07, + "loss": 1.6757, + "step": 42574 + }, + { + "epoch": 9.29, + "learning_rate": 3.5448984050688225e-07, + "loss": 4.1627, + "step": 42575 + }, + { + "epoch": 9.29, + "learning_rate": 3.5438059864540096e-07, + "loss": 1.3858, + "step": 42576 + }, + { + "epoch": 9.29, + "learning_rate": 3.542713567839196e-07, + "loss": 1.6723, + "step": 42577 + }, + { + "epoch": 9.29, + "learning_rate": 3.541621149224383e-07, + "loss": 1.6374, + "step": 42578 + }, + { + "epoch": 9.29, + "learning_rate": 3.54052873060957e-07, + "loss": 0.3019, + "step": 42579 + }, + { + "epoch": 9.29, + "learning_rate": 3.5394363119947567e-07, + "loss": 1.0558, + "step": 42580 + }, + { + "epoch": 9.29, + "learning_rate": 3.5383438933799434e-07, + "loss": 0.9114, + "step": 42581 + }, + { + "epoch": 9.29, + "learning_rate": 3.5372514747651305e-07, + "loss": 0.5876, + "step": 42582 + }, + { + "epoch": 9.29, + "learning_rate": 3.536159056150317e-07, + "loss": 0.2099, + "step": 42583 + }, + { + "epoch": 9.29, + "learning_rate": 3.535066637535504e-07, + "loss": 1.8634, + "step": 42584 + }, + { + "epoch": 9.29, + "learning_rate": 3.533974218920691e-07, + "loss": 0.9776, + "step": 42585 + }, + { + "epoch": 9.29, + "learning_rate": 3.5328818003058776e-07, + "loss": 0.9573, + "step": 42586 + }, + { + "epoch": 9.29, + "learning_rate": 3.531789381691064e-07, + "loss": 0.9332, + "step": 42587 + }, + { + "epoch": 9.29, + "learning_rate": 3.530696963076251e-07, + "loss": 0.8741, + "step": 42588 + }, + { + "epoch": 9.29, + "learning_rate": 3.529604544461438e-07, + "loss": 0.5648, + "step": 42589 + }, + { + "epoch": 9.3, + "learning_rate": 3.5285121258466247e-07, + "loss": 1.9844, + "step": 42590 + }, + { + "epoch": 9.3, + "learning_rate": 3.5274197072318113e-07, + "loss": 0.5922, + "step": 42591 + }, + { + "epoch": 9.3, + "learning_rate": 3.5263272886169985e-07, + "loss": 1.3154, + "step": 42592 + }, + { + "epoch": 9.3, + "learning_rate": 3.525234870002185e-07, + "loss": 0.1929, + "step": 42593 + }, + { + "epoch": 9.3, + "learning_rate": 3.524142451387372e-07, + "loss": 2.0679, + "step": 42594 + }, + { + "epoch": 9.3, + "learning_rate": 3.523050032772559e-07, + "loss": 1.0987, + "step": 42595 + }, + { + "epoch": 9.3, + "learning_rate": 3.5219576141577456e-07, + "loss": 0.953, + "step": 42596 + }, + { + "epoch": 9.3, + "learning_rate": 3.520865195542932e-07, + "loss": 1.4007, + "step": 42597 + }, + { + "epoch": 9.3, + "learning_rate": 3.519772776928119e-07, + "loss": 0.9089, + "step": 42598 + }, + { + "epoch": 9.3, + "learning_rate": 3.518680358313306e-07, + "loss": 0.7841, + "step": 42599 + }, + { + "epoch": 9.3, + "learning_rate": 3.5175879396984927e-07, + "loss": 0.4807, + "step": 42600 + }, + { + "epoch": 9.3, + "learning_rate": 3.5164955210836793e-07, + "loss": 1.9851, + "step": 42601 + }, + { + "epoch": 9.3, + "learning_rate": 3.5154031024688665e-07, + "loss": 0.8853, + "step": 42602 + }, + { + "epoch": 9.3, + "learning_rate": 3.514310683854053e-07, + "loss": 1.1928, + "step": 42603 + }, + { + "epoch": 9.3, + "learning_rate": 3.51321826523924e-07, + "loss": 0.3293, + "step": 42604 + }, + { + "epoch": 9.3, + "learning_rate": 3.512125846624427e-07, + "loss": 1.4556, + "step": 42605 + }, + { + "epoch": 9.3, + "learning_rate": 3.5110334280096136e-07, + "loss": 0.1292, + "step": 42606 + }, + { + "epoch": 9.3, + "learning_rate": 3.5099410093948e-07, + "loss": 0.0375, + "step": 42607 + }, + { + "epoch": 9.3, + "learning_rate": 3.508848590779987e-07, + "loss": 0.8798, + "step": 42608 + }, + { + "epoch": 9.3, + "learning_rate": 3.507756172165174e-07, + "loss": 0.7068, + "step": 42609 + }, + { + "epoch": 9.3, + "learning_rate": 3.5066637535503607e-07, + "loss": 1.2942, + "step": 42610 + }, + { + "epoch": 9.3, + "learning_rate": 3.5055713349355473e-07, + "loss": 1.3557, + "step": 42611 + }, + { + "epoch": 9.3, + "learning_rate": 3.5044789163207345e-07, + "loss": 1.8257, + "step": 42612 + }, + { + "epoch": 9.3, + "learning_rate": 3.503386497705921e-07, + "loss": 0.6704, + "step": 42613 + }, + { + "epoch": 9.3, + "learning_rate": 3.502294079091108e-07, + "loss": 2.0936, + "step": 42614 + }, + { + "epoch": 9.3, + "learning_rate": 3.501201660476295e-07, + "loss": 0.9729, + "step": 42615 + }, + { + "epoch": 9.3, + "learning_rate": 3.5001092418614815e-07, + "loss": 1.1263, + "step": 42616 + }, + { + "epoch": 9.3, + "learning_rate": 3.499016823246668e-07, + "loss": 1.4561, + "step": 42617 + }, + { + "epoch": 9.3, + "learning_rate": 3.497924404631855e-07, + "loss": 0.3707, + "step": 42618 + }, + { + "epoch": 9.3, + "learning_rate": 3.496831986017042e-07, + "loss": 2.3577, + "step": 42619 + }, + { + "epoch": 9.3, + "learning_rate": 3.4957395674022286e-07, + "loss": 0.8451, + "step": 42620 + }, + { + "epoch": 9.3, + "learning_rate": 3.4946471487874153e-07, + "loss": 1.7288, + "step": 42621 + }, + { + "epoch": 9.3, + "learning_rate": 3.4935547301726024e-07, + "loss": 0.5337, + "step": 42622 + }, + { + "epoch": 9.3, + "learning_rate": 3.492462311557789e-07, + "loss": 0.4657, + "step": 42623 + }, + { + "epoch": 9.3, + "learning_rate": 3.4913698929429757e-07, + "loss": 0.133, + "step": 42624 + }, + { + "epoch": 9.3, + "learning_rate": 3.490277474328163e-07, + "loss": 0.5885, + "step": 42625 + }, + { + "epoch": 9.3, + "learning_rate": 3.4891850557133495e-07, + "loss": 1.0389, + "step": 42626 + }, + { + "epoch": 9.3, + "learning_rate": 3.488092637098536e-07, + "loss": 0.5301, + "step": 42627 + }, + { + "epoch": 9.3, + "learning_rate": 3.4870002184837233e-07, + "loss": 1.3106, + "step": 42628 + }, + { + "epoch": 9.3, + "learning_rate": 3.48590779986891e-07, + "loss": 1.1947, + "step": 42629 + }, + { + "epoch": 9.3, + "learning_rate": 3.4848153812540966e-07, + "loss": 0.0487, + "step": 42630 + }, + { + "epoch": 9.3, + "learning_rate": 3.483722962639283e-07, + "loss": 0.0286, + "step": 42631 + }, + { + "epoch": 9.3, + "learning_rate": 3.4826305440244704e-07, + "loss": 0.0952, + "step": 42632 + }, + { + "epoch": 9.3, + "learning_rate": 3.481538125409657e-07, + "loss": 0.3756, + "step": 42633 + }, + { + "epoch": 9.3, + "learning_rate": 3.4804457067948437e-07, + "loss": 0.6226, + "step": 42634 + }, + { + "epoch": 9.3, + "learning_rate": 3.479353288180031e-07, + "loss": 1.1931, + "step": 42635 + }, + { + "epoch": 9.31, + "learning_rate": 3.4782608695652175e-07, + "loss": 0.2353, + "step": 42636 + }, + { + "epoch": 9.31, + "learning_rate": 3.477168450950404e-07, + "loss": 1.5474, + "step": 42637 + }, + { + "epoch": 9.31, + "learning_rate": 3.476076032335592e-07, + "loss": 0.407, + "step": 42638 + }, + { + "epoch": 9.31, + "learning_rate": 3.474983613720778e-07, + "loss": 0.7716, + "step": 42639 + }, + { + "epoch": 9.31, + "learning_rate": 3.4738911951059646e-07, + "loss": 0.1785, + "step": 42640 + }, + { + "epoch": 9.31, + "learning_rate": 3.472798776491151e-07, + "loss": 3.0054, + "step": 42641 + }, + { + "epoch": 9.31, + "learning_rate": 3.4717063578763384e-07, + "loss": 0.872, + "step": 42642 + }, + { + "epoch": 9.31, + "learning_rate": 3.470613939261525e-07, + "loss": 1.5376, + "step": 42643 + }, + { + "epoch": 9.31, + "learning_rate": 3.4695215206467117e-07, + "loss": 0.9269, + "step": 42644 + }, + { + "epoch": 9.31, + "learning_rate": 3.4684291020318994e-07, + "loss": 1.0437, + "step": 42645 + }, + { + "epoch": 9.31, + "learning_rate": 3.4673366834170855e-07, + "loss": 0.6417, + "step": 42646 + }, + { + "epoch": 9.31, + "learning_rate": 3.466244264802272e-07, + "loss": 1.4538, + "step": 42647 + }, + { + "epoch": 9.31, + "learning_rate": 3.46515184618746e-07, + "loss": 0.6829, + "step": 42648 + }, + { + "epoch": 9.31, + "learning_rate": 3.4640594275726465e-07, + "loss": 0.8104, + "step": 42649 + }, + { + "epoch": 9.31, + "learning_rate": 3.4629670089578326e-07, + "loss": 0.8967, + "step": 42650 + }, + { + "epoch": 9.31, + "learning_rate": 3.461874590343019e-07, + "loss": 0.6313, + "step": 42651 + }, + { + "epoch": 9.31, + "learning_rate": 3.460782171728207e-07, + "loss": 0.0545, + "step": 42652 + }, + { + "epoch": 9.31, + "learning_rate": 3.459689753113393e-07, + "loss": 0.158, + "step": 42653 + }, + { + "epoch": 9.31, + "learning_rate": 3.4585973344985797e-07, + "loss": 0.4066, + "step": 42654 + }, + { + "epoch": 9.31, + "learning_rate": 3.4575049158837674e-07, + "loss": 0.8549, + "step": 42655 + }, + { + "epoch": 9.31, + "learning_rate": 3.456412497268954e-07, + "loss": 1.598, + "step": 42656 + }, + { + "epoch": 9.31, + "learning_rate": 3.45532007865414e-07, + "loss": 1.1826, + "step": 42657 + }, + { + "epoch": 9.31, + "learning_rate": 3.454227660039328e-07, + "loss": 2.2346, + "step": 42658 + }, + { + "epoch": 9.31, + "learning_rate": 3.4531352414245144e-07, + "loss": 0.6675, + "step": 42659 + }, + { + "epoch": 9.31, + "learning_rate": 3.4520428228097006e-07, + "loss": 0.0489, + "step": 42660 + }, + { + "epoch": 9.31, + "learning_rate": 3.450950404194888e-07, + "loss": 0.3004, + "step": 42661 + }, + { + "epoch": 9.31, + "learning_rate": 3.449857985580075e-07, + "loss": 0.1099, + "step": 42662 + }, + { + "epoch": 9.31, + "learning_rate": 3.4487655669652615e-07, + "loss": 1.3681, + "step": 42663 + }, + { + "epoch": 9.31, + "learning_rate": 3.4476731483504476e-07, + "loss": 2.2865, + "step": 42664 + }, + { + "epoch": 9.31, + "learning_rate": 3.4465807297356353e-07, + "loss": 0.1369, + "step": 42665 + }, + { + "epoch": 9.31, + "learning_rate": 3.445488311120822e-07, + "loss": 1.8077, + "step": 42666 + }, + { + "epoch": 9.31, + "learning_rate": 3.4443958925060086e-07, + "loss": 1.5438, + "step": 42667 + }, + { + "epoch": 9.31, + "learning_rate": 3.443303473891196e-07, + "loss": 0.5913, + "step": 42668 + }, + { + "epoch": 9.31, + "learning_rate": 3.4422110552763824e-07, + "loss": 0.6433, + "step": 42669 + }, + { + "epoch": 9.31, + "learning_rate": 3.441118636661569e-07, + "loss": 1.2926, + "step": 42670 + }, + { + "epoch": 9.31, + "learning_rate": 3.440026218046756e-07, + "loss": 0.0943, + "step": 42671 + }, + { + "epoch": 9.31, + "learning_rate": 3.438933799431943e-07, + "loss": 0.1726, + "step": 42672 + }, + { + "epoch": 9.31, + "learning_rate": 3.4378413808171295e-07, + "loss": 2.3471, + "step": 42673 + }, + { + "epoch": 9.31, + "learning_rate": 3.436748962202316e-07, + "loss": 0.5473, + "step": 42674 + }, + { + "epoch": 9.31, + "learning_rate": 3.4356565435875033e-07, + "loss": 0.7851, + "step": 42675 + }, + { + "epoch": 9.31, + "learning_rate": 3.43456412497269e-07, + "loss": 0.4991, + "step": 42676 + }, + { + "epoch": 9.31, + "learning_rate": 3.4334717063578766e-07, + "loss": 1.7764, + "step": 42677 + }, + { + "epoch": 9.31, + "learning_rate": 3.432379287743064e-07, + "loss": 1.4862, + "step": 42678 + }, + { + "epoch": 9.31, + "learning_rate": 3.4312868691282504e-07, + "loss": 0.6358, + "step": 42679 + }, + { + "epoch": 9.31, + "learning_rate": 3.430194450513437e-07, + "loss": 0.7369, + "step": 42680 + }, + { + "epoch": 9.31, + "learning_rate": 3.429102031898624e-07, + "loss": 0.0582, + "step": 42681 + }, + { + "epoch": 9.32, + "learning_rate": 3.428009613283811e-07, + "loss": 1.8369, + "step": 42682 + }, + { + "epoch": 9.32, + "learning_rate": 3.4269171946689975e-07, + "loss": 0.3353, + "step": 42683 + }, + { + "epoch": 9.32, + "learning_rate": 3.425824776054184e-07, + "loss": 2.5533, + "step": 42684 + }, + { + "epoch": 9.32, + "learning_rate": 3.4247323574393713e-07, + "loss": 0.7653, + "step": 42685 + }, + { + "epoch": 9.32, + "learning_rate": 3.423639938824558e-07, + "loss": 0.6259, + "step": 42686 + }, + { + "epoch": 9.32, + "learning_rate": 3.4225475202097446e-07, + "loss": 0.7747, + "step": 42687 + }, + { + "epoch": 9.32, + "learning_rate": 3.421455101594932e-07, + "loss": 0.9468, + "step": 42688 + }, + { + "epoch": 9.32, + "learning_rate": 3.4203626829801184e-07, + "loss": 0.8544, + "step": 42689 + }, + { + "epoch": 9.32, + "learning_rate": 3.419270264365305e-07, + "loss": 0.77, + "step": 42690 + }, + { + "epoch": 9.32, + "learning_rate": 3.418177845750492e-07, + "loss": 0.9813, + "step": 42691 + }, + { + "epoch": 9.32, + "learning_rate": 3.417085427135679e-07, + "loss": 0.4389, + "step": 42692 + }, + { + "epoch": 9.32, + "learning_rate": 3.4159930085208655e-07, + "loss": 0.6903, + "step": 42693 + }, + { + "epoch": 9.32, + "learning_rate": 3.414900589906052e-07, + "loss": 1.8842, + "step": 42694 + }, + { + "epoch": 9.32, + "learning_rate": 3.4138081712912393e-07, + "loss": 3.596, + "step": 42695 + }, + { + "epoch": 9.32, + "learning_rate": 3.412715752676426e-07, + "loss": 0.4006, + "step": 42696 + }, + { + "epoch": 9.32, + "learning_rate": 3.4116233340616126e-07, + "loss": 3.1559, + "step": 42697 + }, + { + "epoch": 9.32, + "learning_rate": 3.4105309154467997e-07, + "loss": 1.2535, + "step": 42698 + }, + { + "epoch": 9.32, + "learning_rate": 3.4094384968319864e-07, + "loss": 0.589, + "step": 42699 + }, + { + "epoch": 9.32, + "learning_rate": 3.408346078217173e-07, + "loss": 1.1534, + "step": 42700 + }, + { + "epoch": 9.32, + "learning_rate": 3.40725365960236e-07, + "loss": 1.7422, + "step": 42701 + }, + { + "epoch": 9.32, + "learning_rate": 3.406161240987547e-07, + "loss": 1.4152, + "step": 42702 + }, + { + "epoch": 9.32, + "learning_rate": 3.4050688223727335e-07, + "loss": 1.199, + "step": 42703 + }, + { + "epoch": 9.32, + "learning_rate": 3.4039764037579206e-07, + "loss": 0.8273, + "step": 42704 + }, + { + "epoch": 9.32, + "learning_rate": 3.4028839851431073e-07, + "loss": 0.4191, + "step": 42705 + }, + { + "epoch": 9.32, + "learning_rate": 3.401791566528294e-07, + "loss": 1.1186, + "step": 42706 + }, + { + "epoch": 9.32, + "learning_rate": 3.4006991479134805e-07, + "loss": 0.0918, + "step": 42707 + }, + { + "epoch": 9.32, + "learning_rate": 3.3996067292986677e-07, + "loss": 1.6183, + "step": 42708 + }, + { + "epoch": 9.32, + "learning_rate": 3.3985143106838544e-07, + "loss": 0.6126, + "step": 42709 + }, + { + "epoch": 9.32, + "learning_rate": 3.397421892069041e-07, + "loss": 0.5153, + "step": 42710 + }, + { + "epoch": 9.32, + "learning_rate": 3.396329473454228e-07, + "loss": 1.1372, + "step": 42711 + }, + { + "epoch": 9.32, + "learning_rate": 3.395237054839415e-07, + "loss": 0.8309, + "step": 42712 + }, + { + "epoch": 9.32, + "learning_rate": 3.3941446362246014e-07, + "loss": 0.0889, + "step": 42713 + }, + { + "epoch": 9.32, + "learning_rate": 3.3930522176097886e-07, + "loss": 0.078, + "step": 42714 + }, + { + "epoch": 9.32, + "learning_rate": 3.391959798994975e-07, + "loss": 0.7626, + "step": 42715 + }, + { + "epoch": 9.32, + "learning_rate": 3.390867380380162e-07, + "loss": 0.5604, + "step": 42716 + }, + { + "epoch": 9.32, + "learning_rate": 3.3897749617653485e-07, + "loss": 0.3849, + "step": 42717 + }, + { + "epoch": 9.32, + "learning_rate": 3.3886825431505357e-07, + "loss": 1.5808, + "step": 42718 + }, + { + "epoch": 9.32, + "learning_rate": 3.3875901245357223e-07, + "loss": 2.8938, + "step": 42719 + }, + { + "epoch": 9.32, + "learning_rate": 3.386497705920909e-07, + "loss": 0.4355, + "step": 42720 + }, + { + "epoch": 9.32, + "learning_rate": 3.385405287306096e-07, + "loss": 0.4339, + "step": 42721 + }, + { + "epoch": 9.32, + "learning_rate": 3.384312868691283e-07, + "loss": 0.8434, + "step": 42722 + }, + { + "epoch": 9.32, + "learning_rate": 3.3832204500764694e-07, + "loss": 1.8346, + "step": 42723 + }, + { + "epoch": 9.32, + "learning_rate": 3.3821280314616566e-07, + "loss": 1.2118, + "step": 42724 + }, + { + "epoch": 9.32, + "learning_rate": 3.381035612846843e-07, + "loss": 0.9893, + "step": 42725 + }, + { + "epoch": 9.32, + "learning_rate": 3.37994319423203e-07, + "loss": 0.7942, + "step": 42726 + }, + { + "epoch": 9.32, + "learning_rate": 3.3788507756172165e-07, + "loss": 0.8655, + "step": 42727 + }, + { + "epoch": 9.33, + "learning_rate": 3.3777583570024037e-07, + "loss": 1.237, + "step": 42728 + }, + { + "epoch": 9.33, + "learning_rate": 3.3766659383875903e-07, + "loss": 1.2258, + "step": 42729 + }, + { + "epoch": 9.33, + "learning_rate": 3.375573519772777e-07, + "loss": 0.9028, + "step": 42730 + }, + { + "epoch": 9.33, + "learning_rate": 3.374481101157964e-07, + "loss": 0.4688, + "step": 42731 + }, + { + "epoch": 9.33, + "learning_rate": 3.373388682543151e-07, + "loss": 1.2401, + "step": 42732 + }, + { + "epoch": 9.33, + "learning_rate": 3.3722962639283374e-07, + "loss": 0.7189, + "step": 42733 + }, + { + "epoch": 9.33, + "learning_rate": 3.3712038453135246e-07, + "loss": 1.0471, + "step": 42734 + }, + { + "epoch": 9.33, + "learning_rate": 3.370111426698711e-07, + "loss": 0.3042, + "step": 42735 + }, + { + "epoch": 9.33, + "learning_rate": 3.369019008083898e-07, + "loss": 1.0108, + "step": 42736 + }, + { + "epoch": 9.33, + "learning_rate": 3.3679265894690845e-07, + "loss": 1.6003, + "step": 42737 + }, + { + "epoch": 9.33, + "learning_rate": 3.3668341708542717e-07, + "loss": 2.0947, + "step": 42738 + }, + { + "epoch": 9.33, + "learning_rate": 3.3657417522394583e-07, + "loss": 0.1535, + "step": 42739 + }, + { + "epoch": 9.33, + "learning_rate": 3.364649333624645e-07, + "loss": 1.8472, + "step": 42740 + }, + { + "epoch": 9.33, + "learning_rate": 3.363556915009832e-07, + "loss": 1.0849, + "step": 42741 + }, + { + "epoch": 9.33, + "learning_rate": 3.362464496395019e-07, + "loss": 0.603, + "step": 42742 + }, + { + "epoch": 9.33, + "learning_rate": 3.3613720777802054e-07, + "loss": 2.7472, + "step": 42743 + }, + { + "epoch": 9.33, + "learning_rate": 3.3602796591653926e-07, + "loss": 0.0793, + "step": 42744 + }, + { + "epoch": 9.33, + "learning_rate": 3.359187240550579e-07, + "loss": 2.0498, + "step": 42745 + }, + { + "epoch": 9.33, + "learning_rate": 3.358094821935766e-07, + "loss": 1.5696, + "step": 42746 + }, + { + "epoch": 9.33, + "learning_rate": 3.357002403320953e-07, + "loss": 1.4825, + "step": 42747 + }, + { + "epoch": 9.33, + "learning_rate": 3.3559099847061396e-07, + "loss": 2.0137, + "step": 42748 + }, + { + "epoch": 9.33, + "learning_rate": 3.3548175660913263e-07, + "loss": 1.066, + "step": 42749 + }, + { + "epoch": 9.33, + "learning_rate": 3.353725147476513e-07, + "loss": 1.7339, + "step": 42750 + }, + { + "epoch": 9.33, + "learning_rate": 3.3526327288617e-07, + "loss": 0.1085, + "step": 42751 + }, + { + "epoch": 9.33, + "learning_rate": 3.3515403102468867e-07, + "loss": 0.5664, + "step": 42752 + }, + { + "epoch": 9.33, + "learning_rate": 3.3504478916320734e-07, + "loss": 0.5238, + "step": 42753 + }, + { + "epoch": 9.33, + "learning_rate": 3.3493554730172605e-07, + "loss": 0.3278, + "step": 42754 + }, + { + "epoch": 9.33, + "learning_rate": 3.348263054402447e-07, + "loss": 0.0491, + "step": 42755 + }, + { + "epoch": 9.33, + "learning_rate": 3.347170635787634e-07, + "loss": 1.1835, + "step": 42756 + }, + { + "epoch": 9.33, + "learning_rate": 3.346078217172821e-07, + "loss": 0.0813, + "step": 42757 + }, + { + "epoch": 9.33, + "learning_rate": 3.3449857985580076e-07, + "loss": 1.1148, + "step": 42758 + }, + { + "epoch": 9.33, + "learning_rate": 3.343893379943194e-07, + "loss": 0.9335, + "step": 42759 + }, + { + "epoch": 9.33, + "learning_rate": 3.342800961328381e-07, + "loss": 1.0694, + "step": 42760 + }, + { + "epoch": 9.33, + "learning_rate": 3.341708542713568e-07, + "loss": 0.974, + "step": 42761 + }, + { + "epoch": 9.33, + "learning_rate": 3.3406161240987547e-07, + "loss": 0.953, + "step": 42762 + }, + { + "epoch": 9.33, + "learning_rate": 3.3395237054839413e-07, + "loss": 1.0597, + "step": 42763 + }, + { + "epoch": 9.33, + "learning_rate": 3.3384312868691285e-07, + "loss": 0.5617, + "step": 42764 + }, + { + "epoch": 9.33, + "learning_rate": 3.337338868254315e-07, + "loss": 0.4974, + "step": 42765 + }, + { + "epoch": 9.33, + "learning_rate": 3.336246449639502e-07, + "loss": 0.5673, + "step": 42766 + }, + { + "epoch": 9.33, + "learning_rate": 3.3351540310246895e-07, + "loss": 1.0132, + "step": 42767 + }, + { + "epoch": 9.33, + "learning_rate": 3.3340616124098756e-07, + "loss": 0.7062, + "step": 42768 + }, + { + "epoch": 9.33, + "learning_rate": 3.332969193795062e-07, + "loss": 1.0308, + "step": 42769 + }, + { + "epoch": 9.33, + "learning_rate": 3.331876775180249e-07, + "loss": 2.1104, + "step": 42770 + }, + { + "epoch": 9.33, + "learning_rate": 3.330784356565436e-07, + "loss": 0.6037, + "step": 42771 + }, + { + "epoch": 9.33, + "learning_rate": 3.3296919379506227e-07, + "loss": 0.0903, + "step": 42772 + }, + { + "epoch": 9.34, + "learning_rate": 3.3285995193358093e-07, + "loss": 1.7867, + "step": 42773 + }, + { + "epoch": 9.34, + "learning_rate": 3.327507100720997e-07, + "loss": 1.6961, + "step": 42774 + }, + { + "epoch": 9.34, + "learning_rate": 3.326414682106183e-07, + "loss": 1.1245, + "step": 42775 + }, + { + "epoch": 9.34, + "learning_rate": 3.32532226349137e-07, + "loss": 1.4368, + "step": 42776 + }, + { + "epoch": 9.34, + "learning_rate": 3.3242298448765575e-07, + "loss": 0.0859, + "step": 42777 + }, + { + "epoch": 9.34, + "learning_rate": 3.323137426261744e-07, + "loss": 1.304, + "step": 42778 + }, + { + "epoch": 9.34, + "learning_rate": 3.32204500764693e-07, + "loss": 2.0879, + "step": 42779 + }, + { + "epoch": 9.34, + "learning_rate": 3.320952589032117e-07, + "loss": 0.7699, + "step": 42780 + }, + { + "epoch": 9.34, + "learning_rate": 3.3198601704173046e-07, + "loss": 1.5878, + "step": 42781 + }, + { + "epoch": 9.34, + "learning_rate": 3.3187677518024907e-07, + "loss": 2.0844, + "step": 42782 + }, + { + "epoch": 9.34, + "learning_rate": 3.3176753331876773e-07, + "loss": 1.4484, + "step": 42783 + }, + { + "epoch": 9.34, + "learning_rate": 3.316582914572865e-07, + "loss": 1.4361, + "step": 42784 + }, + { + "epoch": 9.34, + "learning_rate": 3.3154904959580516e-07, + "loss": 0.8152, + "step": 42785 + }, + { + "epoch": 9.34, + "learning_rate": 3.314398077343238e-07, + "loss": 0.7904, + "step": 42786 + }, + { + "epoch": 9.34, + "learning_rate": 3.3133056587284255e-07, + "loss": 0.2965, + "step": 42787 + }, + { + "epoch": 9.34, + "learning_rate": 3.312213240113612e-07, + "loss": 1.9359, + "step": 42788 + }, + { + "epoch": 9.34, + "learning_rate": 3.311120821498798e-07, + "loss": 0.8761, + "step": 42789 + }, + { + "epoch": 9.34, + "learning_rate": 3.310028402883986e-07, + "loss": 0.0921, + "step": 42790 + }, + { + "epoch": 9.34, + "learning_rate": 3.3089359842691725e-07, + "loss": 1.9332, + "step": 42791 + }, + { + "epoch": 9.34, + "learning_rate": 3.307843565654359e-07, + "loss": 1.0852, + "step": 42792 + }, + { + "epoch": 9.34, + "learning_rate": 3.3067511470395453e-07, + "loss": 0.7969, + "step": 42793 + }, + { + "epoch": 9.34, + "learning_rate": 3.305658728424733e-07, + "loss": 0.6413, + "step": 42794 + }, + { + "epoch": 9.34, + "learning_rate": 3.3045663098099196e-07, + "loss": 2.1281, + "step": 42795 + }, + { + "epoch": 9.34, + "learning_rate": 3.303473891195106e-07, + "loss": 1.2604, + "step": 42796 + }, + { + "epoch": 9.34, + "learning_rate": 3.3023814725802934e-07, + "loss": 3.8603, + "step": 42797 + }, + { + "epoch": 9.34, + "learning_rate": 3.30128905396548e-07, + "loss": 0.314, + "step": 42798 + }, + { + "epoch": 9.34, + "learning_rate": 3.3001966353506667e-07, + "loss": 0.8915, + "step": 42799 + }, + { + "epoch": 9.34, + "learning_rate": 3.299104216735854e-07, + "loss": 0.254, + "step": 42800 + }, + { + "epoch": 9.34, + "learning_rate": 3.2980117981210405e-07, + "loss": 0.341, + "step": 42801 + }, + { + "epoch": 9.34, + "learning_rate": 3.296919379506227e-07, + "loss": 1.9657, + "step": 42802 + }, + { + "epoch": 9.34, + "learning_rate": 3.295826960891414e-07, + "loss": 1.2213, + "step": 42803 + }, + { + "epoch": 9.34, + "learning_rate": 3.294734542276601e-07, + "loss": 1.2194, + "step": 42804 + }, + { + "epoch": 9.34, + "learning_rate": 3.2936421236617876e-07, + "loss": 2.0776, + "step": 42805 + }, + { + "epoch": 9.34, + "learning_rate": 3.292549705046974e-07, + "loss": 1.5185, + "step": 42806 + }, + { + "epoch": 9.34, + "learning_rate": 3.2914572864321614e-07, + "loss": 1.0283, + "step": 42807 + }, + { + "epoch": 9.34, + "learning_rate": 3.290364867817348e-07, + "loss": 0.8194, + "step": 42808 + }, + { + "epoch": 9.34, + "learning_rate": 3.2892724492025347e-07, + "loss": 1.2454, + "step": 42809 + }, + { + "epoch": 9.34, + "learning_rate": 3.288180030587722e-07, + "loss": 1.0479, + "step": 42810 + }, + { + "epoch": 9.34, + "learning_rate": 3.2870876119729085e-07, + "loss": 1.63, + "step": 42811 + }, + { + "epoch": 9.34, + "learning_rate": 3.285995193358095e-07, + "loss": 1.0426, + "step": 42812 + }, + { + "epoch": 9.34, + "learning_rate": 3.284902774743282e-07, + "loss": 1.8097, + "step": 42813 + }, + { + "epoch": 9.34, + "learning_rate": 3.283810356128469e-07, + "loss": 0.9194, + "step": 42814 + }, + { + "epoch": 9.34, + "learning_rate": 3.2827179375136556e-07, + "loss": 0.5064, + "step": 42815 + }, + { + "epoch": 9.34, + "learning_rate": 3.281625518898842e-07, + "loss": 2.3071, + "step": 42816 + }, + { + "epoch": 9.34, + "learning_rate": 3.2805331002840294e-07, + "loss": 3.2464, + "step": 42817 + }, + { + "epoch": 9.34, + "learning_rate": 3.279440681669216e-07, + "loss": 0.54, + "step": 42818 + }, + { + "epoch": 9.35, + "learning_rate": 3.2783482630544027e-07, + "loss": 1.4208, + "step": 42819 + }, + { + "epoch": 9.35, + "learning_rate": 3.27725584443959e-07, + "loss": 2.5938, + "step": 42820 + }, + { + "epoch": 9.35, + "learning_rate": 3.2761634258247765e-07, + "loss": 0.5605, + "step": 42821 + }, + { + "epoch": 9.35, + "learning_rate": 3.275071007209963e-07, + "loss": 0.0725, + "step": 42822 + }, + { + "epoch": 9.35, + "learning_rate": 3.2739785885951503e-07, + "loss": 1.4435, + "step": 42823 + }, + { + "epoch": 9.35, + "learning_rate": 3.272886169980337e-07, + "loss": 1.9978, + "step": 42824 + }, + { + "epoch": 9.35, + "learning_rate": 3.2717937513655236e-07, + "loss": 0.9872, + "step": 42825 + }, + { + "epoch": 9.35, + "learning_rate": 3.27070133275071e-07, + "loss": 1.6707, + "step": 42826 + }, + { + "epoch": 9.35, + "learning_rate": 3.2696089141358974e-07, + "loss": 2.7221, + "step": 42827 + }, + { + "epoch": 9.35, + "learning_rate": 3.268516495521084e-07, + "loss": 1.4308, + "step": 42828 + }, + { + "epoch": 9.35, + "learning_rate": 3.2674240769062707e-07, + "loss": 1.2885, + "step": 42829 + }, + { + "epoch": 9.35, + "learning_rate": 3.266331658291458e-07, + "loss": 1.405, + "step": 42830 + }, + { + "epoch": 9.35, + "learning_rate": 3.2652392396766445e-07, + "loss": 0.867, + "step": 42831 + }, + { + "epoch": 9.35, + "learning_rate": 3.264146821061831e-07, + "loss": 2.1432, + "step": 42832 + }, + { + "epoch": 9.35, + "learning_rate": 3.2630544024470183e-07, + "loss": 0.6754, + "step": 42833 + }, + { + "epoch": 9.35, + "learning_rate": 3.261961983832205e-07, + "loss": 0.7432, + "step": 42834 + }, + { + "epoch": 9.35, + "learning_rate": 3.2608695652173915e-07, + "loss": 0.9149, + "step": 42835 + }, + { + "epoch": 9.35, + "learning_rate": 3.259777146602578e-07, + "loss": 3.1138, + "step": 42836 + }, + { + "epoch": 9.35, + "learning_rate": 3.2586847279877654e-07, + "loss": 0.6672, + "step": 42837 + }, + { + "epoch": 9.35, + "learning_rate": 3.257592309372952e-07, + "loss": 1.0579, + "step": 42838 + }, + { + "epoch": 9.35, + "learning_rate": 3.2564998907581386e-07, + "loss": 0.4421, + "step": 42839 + }, + { + "epoch": 9.35, + "learning_rate": 3.255407472143326e-07, + "loss": 1.1653, + "step": 42840 + }, + { + "epoch": 9.35, + "learning_rate": 3.2543150535285124e-07, + "loss": 0.9739, + "step": 42841 + }, + { + "epoch": 9.35, + "learning_rate": 3.253222634913699e-07, + "loss": 0.8984, + "step": 42842 + }, + { + "epoch": 9.35, + "learning_rate": 3.252130216298886e-07, + "loss": 1.6433, + "step": 42843 + }, + { + "epoch": 9.35, + "learning_rate": 3.251037797684073e-07, + "loss": 0.2595, + "step": 42844 + }, + { + "epoch": 9.35, + "learning_rate": 3.2499453790692595e-07, + "loss": 0.0353, + "step": 42845 + }, + { + "epoch": 9.35, + "learning_rate": 3.248852960454446e-07, + "loss": 1.5777, + "step": 42846 + }, + { + "epoch": 9.35, + "learning_rate": 3.2477605418396333e-07, + "loss": 1.9459, + "step": 42847 + }, + { + "epoch": 9.35, + "learning_rate": 3.24666812322482e-07, + "loss": 0.752, + "step": 42848 + }, + { + "epoch": 9.35, + "learning_rate": 3.2455757046100066e-07, + "loss": 1.493, + "step": 42849 + }, + { + "epoch": 9.35, + "learning_rate": 3.244483285995194e-07, + "loss": 1.2112, + "step": 42850 + }, + { + "epoch": 9.35, + "learning_rate": 3.2433908673803804e-07, + "loss": 0.3486, + "step": 42851 + }, + { + "epoch": 9.35, + "learning_rate": 3.242298448765567e-07, + "loss": 0.7602, + "step": 42852 + }, + { + "epoch": 9.35, + "learning_rate": 3.241206030150754e-07, + "loss": 1.0232, + "step": 42853 + }, + { + "epoch": 9.35, + "learning_rate": 3.240113611535941e-07, + "loss": 0.3595, + "step": 42854 + }, + { + "epoch": 9.35, + "learning_rate": 3.2390211929211275e-07, + "loss": 0.9896, + "step": 42855 + }, + { + "epoch": 9.35, + "learning_rate": 3.237928774306314e-07, + "loss": 0.3193, + "step": 42856 + }, + { + "epoch": 9.35, + "learning_rate": 3.2368363556915013e-07, + "loss": 1.2288, + "step": 42857 + }, + { + "epoch": 9.35, + "learning_rate": 3.235743937076688e-07, + "loss": 0.5711, + "step": 42858 + }, + { + "epoch": 9.35, + "learning_rate": 3.2346515184618746e-07, + "loss": 1.1344, + "step": 42859 + }, + { + "epoch": 9.35, + "learning_rate": 3.233559099847062e-07, + "loss": 0.7668, + "step": 42860 + }, + { + "epoch": 9.35, + "learning_rate": 3.2324666812322484e-07, + "loss": 1.0155, + "step": 42861 + }, + { + "epoch": 9.35, + "learning_rate": 3.231374262617435e-07, + "loss": 0.7251, + "step": 42862 + }, + { + "epoch": 9.35, + "learning_rate": 3.230281844002622e-07, + "loss": 1.0918, + "step": 42863 + }, + { + "epoch": 9.35, + "learning_rate": 3.229189425387809e-07, + "loss": 1.403, + "step": 42864 + }, + { + "epoch": 9.36, + "learning_rate": 3.2280970067729955e-07, + "loss": 1.669, + "step": 42865 + }, + { + "epoch": 9.36, + "learning_rate": 3.2270045881581827e-07, + "loss": 0.6972, + "step": 42866 + }, + { + "epoch": 9.36, + "learning_rate": 3.2259121695433693e-07, + "loss": 1.1112, + "step": 42867 + }, + { + "epoch": 9.36, + "learning_rate": 3.224819750928556e-07, + "loss": 0.469, + "step": 42868 + }, + { + "epoch": 9.36, + "learning_rate": 3.2237273323137426e-07, + "loss": 0.6166, + "step": 42869 + }, + { + "epoch": 9.36, + "learning_rate": 3.22263491369893e-07, + "loss": 1.0031, + "step": 42870 + }, + { + "epoch": 9.36, + "learning_rate": 3.2215424950841164e-07, + "loss": 1.1617, + "step": 42871 + }, + { + "epoch": 9.36, + "learning_rate": 3.220450076469303e-07, + "loss": 0.5512, + "step": 42872 + }, + { + "epoch": 9.36, + "learning_rate": 3.21935765785449e-07, + "loss": 0.5123, + "step": 42873 + }, + { + "epoch": 9.36, + "learning_rate": 3.218265239239677e-07, + "loss": 0.7623, + "step": 42874 + }, + { + "epoch": 9.36, + "learning_rate": 3.2171728206248635e-07, + "loss": 2.2899, + "step": 42875 + }, + { + "epoch": 9.36, + "learning_rate": 3.2160804020100506e-07, + "loss": 0.7086, + "step": 42876 + }, + { + "epoch": 9.36, + "learning_rate": 3.2149879833952373e-07, + "loss": 2.6264, + "step": 42877 + }, + { + "epoch": 9.36, + "learning_rate": 3.213895564780424e-07, + "loss": 1.0597, + "step": 42878 + }, + { + "epoch": 9.36, + "learning_rate": 3.2128031461656106e-07, + "loss": 1.1864, + "step": 42879 + }, + { + "epoch": 9.36, + "learning_rate": 3.2117107275507977e-07, + "loss": 0.8457, + "step": 42880 + }, + { + "epoch": 9.36, + "learning_rate": 3.2106183089359844e-07, + "loss": 0.7478, + "step": 42881 + }, + { + "epoch": 9.36, + "learning_rate": 3.209525890321171e-07, + "loss": 1.2174, + "step": 42882 + }, + { + "epoch": 9.36, + "learning_rate": 3.208433471706358e-07, + "loss": 0.8834, + "step": 42883 + }, + { + "epoch": 9.36, + "learning_rate": 3.207341053091545e-07, + "loss": 1.8411, + "step": 42884 + }, + { + "epoch": 9.36, + "learning_rate": 3.2062486344767315e-07, + "loss": 0.6702, + "step": 42885 + }, + { + "epoch": 9.36, + "learning_rate": 3.2051562158619186e-07, + "loss": 0.7043, + "step": 42886 + }, + { + "epoch": 9.36, + "learning_rate": 3.204063797247105e-07, + "loss": 1.3072, + "step": 42887 + }, + { + "epoch": 9.36, + "learning_rate": 3.202971378632292e-07, + "loss": 0.934, + "step": 42888 + }, + { + "epoch": 9.36, + "learning_rate": 3.2018789600174785e-07, + "loss": 1.3973, + "step": 42889 + }, + { + "epoch": 9.36, + "learning_rate": 3.2007865414026657e-07, + "loss": 0.3655, + "step": 42890 + }, + { + "epoch": 9.36, + "learning_rate": 3.1996941227878523e-07, + "loss": 0.5924, + "step": 42891 + }, + { + "epoch": 9.36, + "learning_rate": 3.198601704173039e-07, + "loss": 1.6838, + "step": 42892 + }, + { + "epoch": 9.36, + "learning_rate": 3.197509285558226e-07, + "loss": 1.4107, + "step": 42893 + }, + { + "epoch": 9.36, + "learning_rate": 3.196416866943413e-07, + "loss": 2.3273, + "step": 42894 + }, + { + "epoch": 9.36, + "learning_rate": 3.1953244483285994e-07, + "loss": 1.0155, + "step": 42895 + }, + { + "epoch": 9.36, + "learning_rate": 3.194232029713787e-07, + "loss": 1.0674, + "step": 42896 + }, + { + "epoch": 9.36, + "learning_rate": 3.193139611098973e-07, + "loss": 0.9094, + "step": 42897 + }, + { + "epoch": 9.36, + "learning_rate": 3.19204719248416e-07, + "loss": 0.805, + "step": 42898 + }, + { + "epoch": 9.36, + "learning_rate": 3.1909547738693465e-07, + "loss": 2.4379, + "step": 42899 + }, + { + "epoch": 9.36, + "learning_rate": 3.1898623552545337e-07, + "loss": 2.755, + "step": 42900 + }, + { + "epoch": 9.36, + "learning_rate": 3.1887699366397203e-07, + "loss": 0.2924, + "step": 42901 + }, + { + "epoch": 9.36, + "learning_rate": 3.187677518024907e-07, + "loss": 1.5847, + "step": 42902 + }, + { + "epoch": 9.36, + "learning_rate": 3.1865850994100947e-07, + "loss": 1.9861, + "step": 42903 + }, + { + "epoch": 9.36, + "learning_rate": 3.185492680795281e-07, + "loss": 0.828, + "step": 42904 + }, + { + "epoch": 9.36, + "learning_rate": 3.1844002621804674e-07, + "loss": 1.7416, + "step": 42905 + }, + { + "epoch": 9.36, + "learning_rate": 3.183307843565655e-07, + "loss": 0.8318, + "step": 42906 + }, + { + "epoch": 9.36, + "learning_rate": 3.182215424950841e-07, + "loss": 1.166, + "step": 42907 + }, + { + "epoch": 9.36, + "learning_rate": 3.181123006336028e-07, + "loss": 0.7649, + "step": 42908 + }, + { + "epoch": 9.36, + "learning_rate": 3.1800305877212156e-07, + "loss": 2.9009, + "step": 42909 + }, + { + "epoch": 9.36, + "learning_rate": 3.178938169106402e-07, + "loss": 0.0983, + "step": 42910 + }, + { + "epoch": 9.37, + "learning_rate": 3.1778457504915883e-07, + "loss": 3.1808, + "step": 42911 + }, + { + "epoch": 9.37, + "learning_rate": 3.176753331876775e-07, + "loss": 2.142, + "step": 42912 + }, + { + "epoch": 9.37, + "learning_rate": 3.1756609132619626e-07, + "loss": 1.2661, + "step": 42913 + }, + { + "epoch": 9.37, + "learning_rate": 3.1745684946471493e-07, + "loss": 1.858, + "step": 42914 + }, + { + "epoch": 9.37, + "learning_rate": 3.1734760760323354e-07, + "loss": 0.7255, + "step": 42915 + }, + { + "epoch": 9.37, + "learning_rate": 3.172383657417523e-07, + "loss": 0.7285, + "step": 42916 + }, + { + "epoch": 9.37, + "learning_rate": 3.1712912388027097e-07, + "loss": 0.638, + "step": 42917 + }, + { + "epoch": 9.37, + "learning_rate": 3.170198820187896e-07, + "loss": 1.0115, + "step": 42918 + }, + { + "epoch": 9.37, + "learning_rate": 3.1691064015730835e-07, + "loss": 0.1173, + "step": 42919 + }, + { + "epoch": 9.37, + "learning_rate": 3.16801398295827e-07, + "loss": 1.3864, + "step": 42920 + }, + { + "epoch": 9.37, + "learning_rate": 3.166921564343457e-07, + "loss": 0.1252, + "step": 42921 + }, + { + "epoch": 9.37, + "learning_rate": 3.165829145728643e-07, + "loss": 0.5071, + "step": 42922 + }, + { + "epoch": 9.37, + "learning_rate": 3.1647367271138306e-07, + "loss": 1.1058, + "step": 42923 + }, + { + "epoch": 9.37, + "learning_rate": 3.1636443084990173e-07, + "loss": 2.0654, + "step": 42924 + }, + { + "epoch": 9.37, + "learning_rate": 3.1625518898842034e-07, + "loss": 0.1857, + "step": 42925 + }, + { + "epoch": 9.37, + "learning_rate": 3.161459471269391e-07, + "loss": 0.6812, + "step": 42926 + }, + { + "epoch": 9.37, + "learning_rate": 3.1603670526545777e-07, + "loss": 0.8682, + "step": 42927 + }, + { + "epoch": 9.37, + "learning_rate": 3.1592746340397644e-07, + "loss": 0.6308, + "step": 42928 + }, + { + "epoch": 9.37, + "learning_rate": 3.1581822154249515e-07, + "loss": 0.7207, + "step": 42929 + }, + { + "epoch": 9.37, + "learning_rate": 3.157089796810138e-07, + "loss": 0.4002, + "step": 42930 + }, + { + "epoch": 9.37, + "learning_rate": 3.155997378195325e-07, + "loss": 0.6527, + "step": 42931 + }, + { + "epoch": 9.37, + "learning_rate": 3.1549049595805114e-07, + "loss": 0.939, + "step": 42932 + }, + { + "epoch": 9.37, + "learning_rate": 3.1538125409656986e-07, + "loss": 0.6136, + "step": 42933 + }, + { + "epoch": 9.37, + "learning_rate": 3.152720122350885e-07, + "loss": 0.1615, + "step": 42934 + }, + { + "epoch": 9.37, + "learning_rate": 3.151627703736072e-07, + "loss": 1.5859, + "step": 42935 + }, + { + "epoch": 9.37, + "learning_rate": 3.150535285121259e-07, + "loss": 0.5324, + "step": 42936 + }, + { + "epoch": 9.37, + "learning_rate": 3.1494428665064457e-07, + "loss": 0.1426, + "step": 42937 + }, + { + "epoch": 9.37, + "learning_rate": 3.1483504478916323e-07, + "loss": 0.0788, + "step": 42938 + }, + { + "epoch": 9.37, + "learning_rate": 3.1472580292768195e-07, + "loss": 1.1379, + "step": 42939 + }, + { + "epoch": 9.37, + "learning_rate": 3.146165610662006e-07, + "loss": 0.3923, + "step": 42940 + }, + { + "epoch": 9.37, + "learning_rate": 3.145073192047193e-07, + "loss": 0.605, + "step": 42941 + }, + { + "epoch": 9.37, + "learning_rate": 3.1439807734323794e-07, + "loss": 1.2334, + "step": 42942 + }, + { + "epoch": 9.37, + "learning_rate": 3.1428883548175666e-07, + "loss": 1.4103, + "step": 42943 + }, + { + "epoch": 9.37, + "learning_rate": 3.141795936202753e-07, + "loss": 1.1177, + "step": 42944 + }, + { + "epoch": 9.37, + "learning_rate": 3.14070351758794e-07, + "loss": 1.1682, + "step": 42945 + }, + { + "epoch": 9.37, + "learning_rate": 3.139611098973127e-07, + "loss": 5.5011, + "step": 42946 + }, + { + "epoch": 9.37, + "learning_rate": 3.1385186803583137e-07, + "loss": 1.5071, + "step": 42947 + }, + { + "epoch": 9.37, + "learning_rate": 3.1374262617435003e-07, + "loss": 0.3578, + "step": 42948 + }, + { + "epoch": 9.37, + "learning_rate": 3.1363338431286875e-07, + "loss": 1.4187, + "step": 42949 + }, + { + "epoch": 9.37, + "learning_rate": 3.135241424513874e-07, + "loss": 0.6511, + "step": 42950 + }, + { + "epoch": 9.37, + "learning_rate": 3.134149005899061e-07, + "loss": 1.1071, + "step": 42951 + }, + { + "epoch": 9.37, + "learning_rate": 3.133056587284248e-07, + "loss": 1.4664, + "step": 42952 + }, + { + "epoch": 9.37, + "learning_rate": 3.1319641686694346e-07, + "loss": 0.6419, + "step": 42953 + }, + { + "epoch": 9.37, + "learning_rate": 3.130871750054621e-07, + "loss": 0.4827, + "step": 42954 + }, + { + "epoch": 9.37, + "learning_rate": 3.129779331439808e-07, + "loss": 0.2079, + "step": 42955 + }, + { + "epoch": 9.37, + "learning_rate": 3.128686912824995e-07, + "loss": 1.4623, + "step": 42956 + }, + { + "epoch": 9.38, + "learning_rate": 3.1275944942101817e-07, + "loss": 1.2293, + "step": 42957 + }, + { + "epoch": 9.38, + "learning_rate": 3.1265020755953683e-07, + "loss": 0.5222, + "step": 42958 + }, + { + "epoch": 9.38, + "learning_rate": 3.1254096569805555e-07, + "loss": 1.3628, + "step": 42959 + }, + { + "epoch": 9.38, + "learning_rate": 3.124317238365742e-07, + "loss": 0.1703, + "step": 42960 + }, + { + "epoch": 9.38, + "learning_rate": 3.123224819750929e-07, + "loss": 0.8206, + "step": 42961 + }, + { + "epoch": 9.38, + "learning_rate": 3.1221324011361154e-07, + "loss": 0.367, + "step": 42962 + }, + { + "epoch": 9.38, + "learning_rate": 3.1210399825213026e-07, + "loss": 0.555, + "step": 42963 + }, + { + "epoch": 9.38, + "learning_rate": 3.119947563906489e-07, + "loss": 0.6507, + "step": 42964 + }, + { + "epoch": 9.38, + "learning_rate": 3.118855145291676e-07, + "loss": 0.8103, + "step": 42965 + }, + { + "epoch": 9.38, + "learning_rate": 3.117762726676863e-07, + "loss": 1.6358, + "step": 42966 + }, + { + "epoch": 9.38, + "learning_rate": 3.1166703080620496e-07, + "loss": 1.1886, + "step": 42967 + }, + { + "epoch": 9.38, + "learning_rate": 3.1155778894472363e-07, + "loss": 0.1071, + "step": 42968 + }, + { + "epoch": 9.38, + "learning_rate": 3.114485470832423e-07, + "loss": 0.9291, + "step": 42969 + }, + { + "epoch": 9.38, + "learning_rate": 3.11339305221761e-07, + "loss": 0.8668, + "step": 42970 + }, + { + "epoch": 9.38, + "learning_rate": 3.112300633602797e-07, + "loss": 0.9237, + "step": 42971 + }, + { + "epoch": 9.38, + "learning_rate": 3.1112082149879834e-07, + "loss": 0.7977, + "step": 42972 + }, + { + "epoch": 9.38, + "learning_rate": 3.1101157963731705e-07, + "loss": 1.6591, + "step": 42973 + }, + { + "epoch": 9.38, + "learning_rate": 3.109023377758357e-07, + "loss": 1.3589, + "step": 42974 + }, + { + "epoch": 9.38, + "learning_rate": 3.1079309591435443e-07, + "loss": 1.5538, + "step": 42975 + }, + { + "epoch": 9.38, + "learning_rate": 3.106838540528731e-07, + "loss": 0.6982, + "step": 42976 + }, + { + "epoch": 9.38, + "learning_rate": 3.1057461219139176e-07, + "loss": 1.2959, + "step": 42977 + }, + { + "epoch": 9.38, + "learning_rate": 3.104653703299105e-07, + "loss": 0.761, + "step": 42978 + }, + { + "epoch": 9.38, + "learning_rate": 3.103561284684291e-07, + "loss": 2.5817, + "step": 42979 + }, + { + "epoch": 9.38, + "learning_rate": 3.102468866069478e-07, + "loss": 0.4281, + "step": 42980 + }, + { + "epoch": 9.38, + "learning_rate": 3.101376447454665e-07, + "loss": 0.1504, + "step": 42981 + }, + { + "epoch": 9.38, + "learning_rate": 3.100284028839852e-07, + "loss": 0.473, + "step": 42982 + }, + { + "epoch": 9.38, + "learning_rate": 3.0991916102250385e-07, + "loss": 1.1065, + "step": 42983 + }, + { + "epoch": 9.38, + "learning_rate": 3.098099191610225e-07, + "loss": 1.9822, + "step": 42984 + }, + { + "epoch": 9.38, + "learning_rate": 3.0970067729954123e-07, + "loss": 1.0888, + "step": 42985 + }, + { + "epoch": 9.38, + "learning_rate": 3.095914354380599e-07, + "loss": 0.876, + "step": 42986 + }, + { + "epoch": 9.38, + "learning_rate": 3.0948219357657856e-07, + "loss": 1.1037, + "step": 42987 + }, + { + "epoch": 9.38, + "learning_rate": 3.093729517150973e-07, + "loss": 1.0291, + "step": 42988 + }, + { + "epoch": 9.38, + "learning_rate": 3.0926370985361594e-07, + "loss": 1.1212, + "step": 42989 + }, + { + "epoch": 9.38, + "learning_rate": 3.091544679921346e-07, + "loss": 1.016, + "step": 42990 + }, + { + "epoch": 9.38, + "learning_rate": 3.090452261306533e-07, + "loss": 0.4903, + "step": 42991 + }, + { + "epoch": 9.38, + "learning_rate": 3.08935984269172e-07, + "loss": 1.6404, + "step": 42992 + }, + { + "epoch": 9.38, + "learning_rate": 3.0882674240769065e-07, + "loss": 0.3237, + "step": 42993 + }, + { + "epoch": 9.38, + "learning_rate": 3.087175005462093e-07, + "loss": 0.3458, + "step": 42994 + }, + { + "epoch": 9.38, + "learning_rate": 3.0860825868472803e-07, + "loss": 0.7215, + "step": 42995 + }, + { + "epoch": 9.38, + "learning_rate": 3.084990168232467e-07, + "loss": 0.6606, + "step": 42996 + }, + { + "epoch": 9.38, + "learning_rate": 3.0838977496176536e-07, + "loss": 2.1652, + "step": 42997 + }, + { + "epoch": 9.38, + "learning_rate": 3.082805331002841e-07, + "loss": 3.5793, + "step": 42998 + }, + { + "epoch": 9.38, + "learning_rate": 3.0817129123880274e-07, + "loss": 0.1759, + "step": 42999 + }, + { + "epoch": 9.38, + "learning_rate": 3.080620493773214e-07, + "loss": 1.5735, + "step": 43000 + }, + { + "epoch": 9.38, + "learning_rate": 3.079528075158401e-07, + "loss": 0.3489, + "step": 43001 + }, + { + "epoch": 9.38, + "learning_rate": 3.078435656543588e-07, + "loss": 2.6893, + "step": 43002 + }, + { + "epoch": 9.39, + "learning_rate": 3.0773432379287745e-07, + "loss": 0.8635, + "step": 43003 + }, + { + "epoch": 9.39, + "learning_rate": 3.076250819313961e-07, + "loss": 0.6124, + "step": 43004 + }, + { + "epoch": 9.39, + "learning_rate": 3.0751584006991483e-07, + "loss": 1.2832, + "step": 43005 + }, + { + "epoch": 9.39, + "learning_rate": 3.074065982084335e-07, + "loss": 1.6752, + "step": 43006 + }, + { + "epoch": 9.39, + "learning_rate": 3.0729735634695216e-07, + "loss": 0.5236, + "step": 43007 + }, + { + "epoch": 9.39, + "learning_rate": 3.0718811448547087e-07, + "loss": 0.8812, + "step": 43008 + }, + { + "epoch": 9.39, + "learning_rate": 3.0707887262398954e-07, + "loss": 2.0536, + "step": 43009 + }, + { + "epoch": 9.39, + "learning_rate": 3.069696307625082e-07, + "loss": 0.8031, + "step": 43010 + }, + { + "epoch": 9.39, + "learning_rate": 3.068603889010269e-07, + "loss": 0.9363, + "step": 43011 + }, + { + "epoch": 9.39, + "learning_rate": 3.067511470395456e-07, + "loss": 0.047, + "step": 43012 + }, + { + "epoch": 9.39, + "learning_rate": 3.0664190517806425e-07, + "loss": 0.5417, + "step": 43013 + }, + { + "epoch": 9.39, + "learning_rate": 3.0653266331658296e-07, + "loss": 1.9762, + "step": 43014 + }, + { + "epoch": 9.39, + "learning_rate": 3.064234214551016e-07, + "loss": 0.445, + "step": 43015 + }, + { + "epoch": 9.39, + "learning_rate": 3.063141795936203e-07, + "loss": 0.1654, + "step": 43016 + }, + { + "epoch": 9.39, + "learning_rate": 3.0620493773213895e-07, + "loss": 1.1692, + "step": 43017 + }, + { + "epoch": 9.39, + "learning_rate": 3.0609569587065767e-07, + "loss": 0.105, + "step": 43018 + }, + { + "epoch": 9.39, + "learning_rate": 3.0598645400917633e-07, + "loss": 1.0522, + "step": 43019 + }, + { + "epoch": 9.39, + "learning_rate": 3.05877212147695e-07, + "loss": 0.8395, + "step": 43020 + }, + { + "epoch": 9.39, + "learning_rate": 3.057679702862137e-07, + "loss": 2.1683, + "step": 43021 + }, + { + "epoch": 9.39, + "learning_rate": 3.056587284247324e-07, + "loss": 3.5706, + "step": 43022 + }, + { + "epoch": 9.39, + "learning_rate": 3.0554948656325104e-07, + "loss": 1.3113, + "step": 43023 + }, + { + "epoch": 9.39, + "learning_rate": 3.0544024470176976e-07, + "loss": 1.1215, + "step": 43024 + }, + { + "epoch": 9.39, + "learning_rate": 3.053310028402884e-07, + "loss": 0.8537, + "step": 43025 + }, + { + "epoch": 9.39, + "learning_rate": 3.052217609788071e-07, + "loss": 0.8702, + "step": 43026 + }, + { + "epoch": 9.39, + "learning_rate": 3.0511251911732575e-07, + "loss": 1.2189, + "step": 43027 + }, + { + "epoch": 9.39, + "learning_rate": 3.0500327725584447e-07, + "loss": 1.6872, + "step": 43028 + }, + { + "epoch": 9.39, + "learning_rate": 3.0489403539436313e-07, + "loss": 0.8451, + "step": 43029 + }, + { + "epoch": 9.39, + "learning_rate": 3.047847935328818e-07, + "loss": 1.0268, + "step": 43030 + }, + { + "epoch": 9.39, + "learning_rate": 3.046755516714005e-07, + "loss": 0.6488, + "step": 43031 + }, + { + "epoch": 9.39, + "learning_rate": 3.045663098099192e-07, + "loss": 2.282, + "step": 43032 + }, + { + "epoch": 9.39, + "learning_rate": 3.0445706794843784e-07, + "loss": 1.7017, + "step": 43033 + }, + { + "epoch": 9.39, + "learning_rate": 3.0434782608695656e-07, + "loss": 1.1723, + "step": 43034 + }, + { + "epoch": 9.39, + "learning_rate": 3.042385842254752e-07, + "loss": 2.2406, + "step": 43035 + }, + { + "epoch": 9.39, + "learning_rate": 3.041293423639939e-07, + "loss": 0.3079, + "step": 43036 + }, + { + "epoch": 9.39, + "learning_rate": 3.0402010050251255e-07, + "loss": 1.6178, + "step": 43037 + }, + { + "epoch": 9.39, + "learning_rate": 3.0391085864103127e-07, + "loss": 0.7574, + "step": 43038 + }, + { + "epoch": 9.39, + "learning_rate": 3.0380161677955e-07, + "loss": 0.9638, + "step": 43039 + }, + { + "epoch": 9.39, + "learning_rate": 3.036923749180686e-07, + "loss": 0.1101, + "step": 43040 + }, + { + "epoch": 9.39, + "learning_rate": 3.035831330565873e-07, + "loss": 1.6425, + "step": 43041 + }, + { + "epoch": 9.39, + "learning_rate": 3.03473891195106e-07, + "loss": 0.742, + "step": 43042 + }, + { + "epoch": 9.39, + "learning_rate": 3.033646493336247e-07, + "loss": 0.751, + "step": 43043 + }, + { + "epoch": 9.39, + "learning_rate": 3.0325540747214336e-07, + "loss": 1.1094, + "step": 43044 + }, + { + "epoch": 9.39, + "learning_rate": 3.03146165610662e-07, + "loss": 1.5774, + "step": 43045 + }, + { + "epoch": 9.39, + "learning_rate": 3.0303692374918074e-07, + "loss": 1.3994, + "step": 43046 + }, + { + "epoch": 9.39, + "learning_rate": 3.029276818876994e-07, + "loss": 0.9354, + "step": 43047 + }, + { + "epoch": 9.4, + "learning_rate": 3.0281844002621807e-07, + "loss": 0.3686, + "step": 43048 + }, + { + "epoch": 9.4, + "learning_rate": 3.027091981647368e-07, + "loss": 0.8533, + "step": 43049 + }, + { + "epoch": 9.4, + "learning_rate": 3.0259995630325545e-07, + "loss": 0.7639, + "step": 43050 + }, + { + "epoch": 9.4, + "learning_rate": 3.024907144417741e-07, + "loss": 0.9002, + "step": 43051 + }, + { + "epoch": 9.4, + "learning_rate": 3.0238147258029283e-07, + "loss": 2.1759, + "step": 43052 + }, + { + "epoch": 9.4, + "learning_rate": 3.022722307188115e-07, + "loss": 0.1096, + "step": 43053 + }, + { + "epoch": 9.4, + "learning_rate": 3.0216298885733015e-07, + "loss": 0.7416, + "step": 43054 + }, + { + "epoch": 9.4, + "learning_rate": 3.020537469958488e-07, + "loss": 1.3639, + "step": 43055 + }, + { + "epoch": 9.4, + "learning_rate": 3.0194450513436754e-07, + "loss": 0.8051, + "step": 43056 + }, + { + "epoch": 9.4, + "learning_rate": 3.018352632728862e-07, + "loss": 0.8088, + "step": 43057 + }, + { + "epoch": 9.4, + "learning_rate": 3.0172602141140486e-07, + "loss": 1.905, + "step": 43058 + }, + { + "epoch": 9.4, + "learning_rate": 3.016167795499236e-07, + "loss": 0.5378, + "step": 43059 + }, + { + "epoch": 9.4, + "learning_rate": 3.0150753768844224e-07, + "loss": 2.9885, + "step": 43060 + }, + { + "epoch": 9.4, + "learning_rate": 3.013982958269609e-07, + "loss": 0.6382, + "step": 43061 + }, + { + "epoch": 9.4, + "learning_rate": 3.012890539654796e-07, + "loss": 1.3813, + "step": 43062 + }, + { + "epoch": 9.4, + "learning_rate": 3.011798121039983e-07, + "loss": 0.3715, + "step": 43063 + }, + { + "epoch": 9.4, + "learning_rate": 3.0107057024251695e-07, + "loss": 1.2493, + "step": 43064 + }, + { + "epoch": 9.4, + "learning_rate": 3.009613283810356e-07, + "loss": 0.5789, + "step": 43065 + }, + { + "epoch": 9.4, + "learning_rate": 3.0085208651955433e-07, + "loss": 3.2284, + "step": 43066 + }, + { + "epoch": 9.4, + "learning_rate": 3.00742844658073e-07, + "loss": 0.8103, + "step": 43067 + }, + { + "epoch": 9.4, + "learning_rate": 3.0063360279659166e-07, + "loss": 0.3125, + "step": 43068 + }, + { + "epoch": 9.4, + "learning_rate": 3.005243609351104e-07, + "loss": 0.9209, + "step": 43069 + }, + { + "epoch": 9.4, + "learning_rate": 3.0041511907362904e-07, + "loss": 1.872, + "step": 43070 + }, + { + "epoch": 9.4, + "learning_rate": 3.003058772121477e-07, + "loss": 0.0444, + "step": 43071 + }, + { + "epoch": 9.4, + "learning_rate": 3.001966353506664e-07, + "loss": 0.657, + "step": 43072 + }, + { + "epoch": 9.4, + "learning_rate": 3.000873934891851e-07, + "loss": 0.4221, + "step": 43073 + }, + { + "epoch": 9.4, + "learning_rate": 2.9997815162770375e-07, + "loss": 1.7937, + "step": 43074 + }, + { + "epoch": 9.4, + "learning_rate": 2.998689097662224e-07, + "loss": 1.5272, + "step": 43075 + }, + { + "epoch": 9.4, + "learning_rate": 2.9975966790474113e-07, + "loss": 0.6906, + "step": 43076 + }, + { + "epoch": 9.4, + "learning_rate": 2.996504260432598e-07, + "loss": 1.764, + "step": 43077 + }, + { + "epoch": 9.4, + "learning_rate": 2.9954118418177846e-07, + "loss": 0.6788, + "step": 43078 + }, + { + "epoch": 9.4, + "learning_rate": 2.994319423202972e-07, + "loss": 0.5073, + "step": 43079 + }, + { + "epoch": 9.4, + "learning_rate": 2.9932270045881584e-07, + "loss": 1.8382, + "step": 43080 + }, + { + "epoch": 9.4, + "learning_rate": 2.992134585973345e-07, + "loss": 1.6541, + "step": 43081 + }, + { + "epoch": 9.4, + "learning_rate": 2.991042167358532e-07, + "loss": 1.2194, + "step": 43082 + }, + { + "epoch": 9.4, + "learning_rate": 2.989949748743719e-07, + "loss": 1.4282, + "step": 43083 + }, + { + "epoch": 9.4, + "learning_rate": 2.9888573301289055e-07, + "loss": 1.4811, + "step": 43084 + }, + { + "epoch": 9.4, + "learning_rate": 2.987764911514092e-07, + "loss": 1.8171, + "step": 43085 + }, + { + "epoch": 9.4, + "learning_rate": 2.9866724928992793e-07, + "loss": 0.4653, + "step": 43086 + }, + { + "epoch": 9.4, + "learning_rate": 2.985580074284466e-07, + "loss": 0.0402, + "step": 43087 + }, + { + "epoch": 9.4, + "learning_rate": 2.9844876556696526e-07, + "loss": 0.2389, + "step": 43088 + }, + { + "epoch": 9.4, + "learning_rate": 2.98339523705484e-07, + "loss": 1.3554, + "step": 43089 + }, + { + "epoch": 9.4, + "learning_rate": 2.9823028184400264e-07, + "loss": 1.4551, + "step": 43090 + }, + { + "epoch": 9.4, + "learning_rate": 2.981210399825213e-07, + "loss": 1.345, + "step": 43091 + }, + { + "epoch": 9.4, + "learning_rate": 2.9801179812104e-07, + "loss": 0.0659, + "step": 43092 + }, + { + "epoch": 9.4, + "learning_rate": 2.979025562595587e-07, + "loss": 0.8214, + "step": 43093 + }, + { + "epoch": 9.41, + "learning_rate": 2.9779331439807735e-07, + "loss": 0.6331, + "step": 43094 + }, + { + "epoch": 9.41, + "learning_rate": 2.9768407253659606e-07, + "loss": 1.1339, + "step": 43095 + }, + { + "epoch": 9.41, + "learning_rate": 2.9757483067511473e-07, + "loss": 0.3013, + "step": 43096 + }, + { + "epoch": 9.41, + "learning_rate": 2.974655888136334e-07, + "loss": 1.1082, + "step": 43097 + }, + { + "epoch": 9.41, + "learning_rate": 2.9735634695215206e-07, + "loss": 0.1948, + "step": 43098 + }, + { + "epoch": 9.41, + "learning_rate": 2.9724710509067077e-07, + "loss": 1.7397, + "step": 43099 + }, + { + "epoch": 9.41, + "learning_rate": 2.971378632291895e-07, + "loss": 0.7973, + "step": 43100 + }, + { + "epoch": 9.41, + "learning_rate": 2.970286213677081e-07, + "loss": 0.8979, + "step": 43101 + }, + { + "epoch": 9.41, + "learning_rate": 2.969193795062268e-07, + "loss": 0.2475, + "step": 43102 + }, + { + "epoch": 9.41, + "learning_rate": 2.968101376447455e-07, + "loss": 3.1715, + "step": 43103 + }, + { + "epoch": 9.41, + "learning_rate": 2.9670089578326415e-07, + "loss": 0.0659, + "step": 43104 + }, + { + "epoch": 9.41, + "learning_rate": 2.9659165392178286e-07, + "loss": 0.7873, + "step": 43105 + }, + { + "epoch": 9.41, + "learning_rate": 2.964824120603015e-07, + "loss": 0.4572, + "step": 43106 + }, + { + "epoch": 9.41, + "learning_rate": 2.9637317019882024e-07, + "loss": 0.6549, + "step": 43107 + }, + { + "epoch": 9.41, + "learning_rate": 2.9626392833733885e-07, + "loss": 1.5824, + "step": 43108 + }, + { + "epoch": 9.41, + "learning_rate": 2.9615468647585757e-07, + "loss": 0.296, + "step": 43109 + }, + { + "epoch": 9.41, + "learning_rate": 2.960454446143763e-07, + "loss": 1.3689, + "step": 43110 + }, + { + "epoch": 9.41, + "learning_rate": 2.9593620275289495e-07, + "loss": 2.3638, + "step": 43111 + }, + { + "epoch": 9.41, + "learning_rate": 2.958269608914136e-07, + "loss": 0.1178, + "step": 43112 + }, + { + "epoch": 9.41, + "learning_rate": 2.957177190299323e-07, + "loss": 0.3616, + "step": 43113 + }, + { + "epoch": 9.41, + "learning_rate": 2.95608477168451e-07, + "loss": 2.8708, + "step": 43114 + }, + { + "epoch": 9.41, + "learning_rate": 2.9549923530696966e-07, + "loss": 0.8132, + "step": 43115 + }, + { + "epoch": 9.41, + "learning_rate": 2.953899934454883e-07, + "loss": 0.1686, + "step": 43116 + }, + { + "epoch": 9.41, + "learning_rate": 2.9528075158400704e-07, + "loss": 0.6303, + "step": 43117 + }, + { + "epoch": 9.41, + "learning_rate": 2.951715097225257e-07, + "loss": 0.4283, + "step": 43118 + }, + { + "epoch": 9.41, + "learning_rate": 2.9506226786104437e-07, + "loss": 1.7656, + "step": 43119 + }, + { + "epoch": 9.41, + "learning_rate": 2.949530259995631e-07, + "loss": 1.5617, + "step": 43120 + }, + { + "epoch": 9.41, + "learning_rate": 2.9484378413808175e-07, + "loss": 0.3365, + "step": 43121 + }, + { + "epoch": 9.41, + "learning_rate": 2.947345422766004e-07, + "loss": 0.6535, + "step": 43122 + }, + { + "epoch": 9.41, + "learning_rate": 2.946253004151191e-07, + "loss": 0.4419, + "step": 43123 + }, + { + "epoch": 9.41, + "learning_rate": 2.945160585536378e-07, + "loss": 0.8669, + "step": 43124 + }, + { + "epoch": 9.41, + "learning_rate": 2.9440681669215646e-07, + "loss": 2.7251, + "step": 43125 + }, + { + "epoch": 9.41, + "learning_rate": 2.942975748306751e-07, + "loss": 1.1463, + "step": 43126 + }, + { + "epoch": 9.41, + "learning_rate": 2.9418833296919384e-07, + "loss": 2.1907, + "step": 43127 + }, + { + "epoch": 9.41, + "learning_rate": 2.940790911077125e-07, + "loss": 0.7807, + "step": 43128 + }, + { + "epoch": 9.41, + "learning_rate": 2.9396984924623117e-07, + "loss": 0.3126, + "step": 43129 + }, + { + "epoch": 9.41, + "learning_rate": 2.938606073847499e-07, + "loss": 1.4291, + "step": 43130 + }, + { + "epoch": 9.41, + "learning_rate": 2.9375136552326855e-07, + "loss": 1.8272, + "step": 43131 + }, + { + "epoch": 9.41, + "learning_rate": 2.936421236617872e-07, + "loss": 0.064, + "step": 43132 + }, + { + "epoch": 9.41, + "learning_rate": 2.9353288180030593e-07, + "loss": 1.6922, + "step": 43133 + }, + { + "epoch": 9.41, + "learning_rate": 2.934236399388246e-07, + "loss": 1.7584, + "step": 43134 + }, + { + "epoch": 9.41, + "learning_rate": 2.9331439807734326e-07, + "loss": 0.7145, + "step": 43135 + }, + { + "epoch": 9.41, + "learning_rate": 2.932051562158619e-07, + "loss": 1.0391, + "step": 43136 + }, + { + "epoch": 9.41, + "learning_rate": 2.9309591435438064e-07, + "loss": 1.4873, + "step": 43137 + }, + { + "epoch": 9.41, + "learning_rate": 2.929866724928993e-07, + "loss": 1.3192, + "step": 43138 + }, + { + "epoch": 9.41, + "learning_rate": 2.9287743063141797e-07, + "loss": 0.1054, + "step": 43139 + }, + { + "epoch": 9.42, + "learning_rate": 2.927681887699367e-07, + "loss": 0.6577, + "step": 43140 + }, + { + "epoch": 9.42, + "learning_rate": 2.9265894690845535e-07, + "loss": 1.101, + "step": 43141 + }, + { + "epoch": 9.42, + "learning_rate": 2.92549705046974e-07, + "loss": 0.5188, + "step": 43142 + }, + { + "epoch": 9.42, + "learning_rate": 2.9244046318549273e-07, + "loss": 0.9497, + "step": 43143 + }, + { + "epoch": 9.42, + "learning_rate": 2.923312213240114e-07, + "loss": 0.4799, + "step": 43144 + }, + { + "epoch": 9.42, + "learning_rate": 2.9222197946253005e-07, + "loss": 2.0178, + "step": 43145 + }, + { + "epoch": 9.42, + "learning_rate": 2.921127376010487e-07, + "loss": 1.2029, + "step": 43146 + }, + { + "epoch": 9.42, + "learning_rate": 2.9200349573956744e-07, + "loss": 0.9782, + "step": 43147 + }, + { + "epoch": 9.42, + "learning_rate": 2.918942538780861e-07, + "loss": 0.7712, + "step": 43148 + }, + { + "epoch": 9.42, + "learning_rate": 2.9178501201660476e-07, + "loss": 0.2603, + "step": 43149 + }, + { + "epoch": 9.42, + "learning_rate": 2.916757701551235e-07, + "loss": 1.161, + "step": 43150 + }, + { + "epoch": 9.42, + "learning_rate": 2.9156652829364214e-07, + "loss": 0.0518, + "step": 43151 + }, + { + "epoch": 9.42, + "learning_rate": 2.914572864321608e-07, + "loss": 1.9408, + "step": 43152 + }, + { + "epoch": 9.42, + "learning_rate": 2.913480445706795e-07, + "loss": 0.5748, + "step": 43153 + }, + { + "epoch": 9.42, + "learning_rate": 2.912388027091982e-07, + "loss": 0.2329, + "step": 43154 + }, + { + "epoch": 9.42, + "learning_rate": 2.9112956084771685e-07, + "loss": 0.8481, + "step": 43155 + }, + { + "epoch": 9.42, + "learning_rate": 2.910203189862355e-07, + "loss": 2.274, + "step": 43156 + }, + { + "epoch": 9.42, + "learning_rate": 2.9091107712475423e-07, + "loss": 0.6016, + "step": 43157 + }, + { + "epoch": 9.42, + "learning_rate": 2.908018352632729e-07, + "loss": 1.2839, + "step": 43158 + }, + { + "epoch": 9.42, + "learning_rate": 2.9069259340179156e-07, + "loss": 2.5615, + "step": 43159 + }, + { + "epoch": 9.42, + "learning_rate": 2.905833515403103e-07, + "loss": 1.1267, + "step": 43160 + }, + { + "epoch": 9.42, + "learning_rate": 2.9047410967882894e-07, + "loss": 0.8176, + "step": 43161 + }, + { + "epoch": 9.42, + "learning_rate": 2.903648678173476e-07, + "loss": 0.1883, + "step": 43162 + }, + { + "epoch": 9.42, + "learning_rate": 2.902556259558663e-07, + "loss": 0.3539, + "step": 43163 + }, + { + "epoch": 9.42, + "learning_rate": 2.90146384094385e-07, + "loss": 1.2309, + "step": 43164 + }, + { + "epoch": 9.42, + "learning_rate": 2.9003714223290365e-07, + "loss": 1.4396, + "step": 43165 + }, + { + "epoch": 9.42, + "learning_rate": 2.899279003714223e-07, + "loss": 0.4622, + "step": 43166 + }, + { + "epoch": 9.42, + "learning_rate": 2.8981865850994103e-07, + "loss": 0.5464, + "step": 43167 + }, + { + "epoch": 9.42, + "learning_rate": 2.8970941664845975e-07, + "loss": 1.4397, + "step": 43168 + }, + { + "epoch": 9.42, + "learning_rate": 2.8960017478697836e-07, + "loss": 1.2646, + "step": 43169 + }, + { + "epoch": 9.42, + "learning_rate": 2.894909329254971e-07, + "loss": 0.1398, + "step": 43170 + }, + { + "epoch": 9.42, + "learning_rate": 2.893816910640158e-07, + "loss": 0.7448, + "step": 43171 + }, + { + "epoch": 9.42, + "learning_rate": 2.8927244920253446e-07, + "loss": 1.8811, + "step": 43172 + }, + { + "epoch": 9.42, + "learning_rate": 2.891632073410531e-07, + "loss": 1.5089, + "step": 43173 + }, + { + "epoch": 9.42, + "learning_rate": 2.890539654795718e-07, + "loss": 0.7089, + "step": 43174 + }, + { + "epoch": 9.42, + "learning_rate": 2.889447236180905e-07, + "loss": 0.4706, + "step": 43175 + }, + { + "epoch": 9.42, + "learning_rate": 2.8883548175660917e-07, + "loss": 0.422, + "step": 43176 + }, + { + "epoch": 9.42, + "learning_rate": 2.8872623989512783e-07, + "loss": 1.13, + "step": 43177 + }, + { + "epoch": 9.42, + "learning_rate": 2.8861699803364655e-07, + "loss": 0.0381, + "step": 43178 + }, + { + "epoch": 9.42, + "learning_rate": 2.885077561721652e-07, + "loss": 0.7145, + "step": 43179 + }, + { + "epoch": 9.42, + "learning_rate": 2.883985143106839e-07, + "loss": 0.9729, + "step": 43180 + }, + { + "epoch": 9.42, + "learning_rate": 2.882892724492026e-07, + "loss": 1.1123, + "step": 43181 + }, + { + "epoch": 9.42, + "learning_rate": 2.8818003058772126e-07, + "loss": 1.4007, + "step": 43182 + }, + { + "epoch": 9.42, + "learning_rate": 2.880707887262399e-07, + "loss": 1.0197, + "step": 43183 + }, + { + "epoch": 9.42, + "learning_rate": 2.879615468647586e-07, + "loss": 2.7756, + "step": 43184 + }, + { + "epoch": 9.42, + "learning_rate": 2.878523050032773e-07, + "loss": 0.5709, + "step": 43185 + }, + { + "epoch": 9.43, + "learning_rate": 2.8774306314179596e-07, + "loss": 0.109, + "step": 43186 + }, + { + "epoch": 9.43, + "learning_rate": 2.8763382128031463e-07, + "loss": 2.4426, + "step": 43187 + }, + { + "epoch": 9.43, + "learning_rate": 2.8752457941883334e-07, + "loss": 2.3545, + "step": 43188 + }, + { + "epoch": 9.43, + "learning_rate": 2.87415337557352e-07, + "loss": 0.4173, + "step": 43189 + }, + { + "epoch": 9.43, + "learning_rate": 2.8730609569587067e-07, + "loss": 1.6081, + "step": 43190 + }, + { + "epoch": 9.43, + "learning_rate": 2.871968538343894e-07, + "loss": 0.6085, + "step": 43191 + }, + { + "epoch": 9.43, + "learning_rate": 2.8708761197290805e-07, + "loss": 0.5624, + "step": 43192 + }, + { + "epoch": 9.43, + "learning_rate": 2.869783701114267e-07, + "loss": 0.8222, + "step": 43193 + }, + { + "epoch": 9.43, + "learning_rate": 2.868691282499454e-07, + "loss": 1.9575, + "step": 43194 + }, + { + "epoch": 9.43, + "learning_rate": 2.867598863884641e-07, + "loss": 0.842, + "step": 43195 + }, + { + "epoch": 9.43, + "learning_rate": 2.8665064452698276e-07, + "loss": 1.3813, + "step": 43196 + }, + { + "epoch": 9.43, + "learning_rate": 2.865414026655014e-07, + "loss": 0.081, + "step": 43197 + }, + { + "epoch": 9.43, + "learning_rate": 2.8643216080402014e-07, + "loss": 1.1378, + "step": 43198 + }, + { + "epoch": 9.43, + "learning_rate": 2.863229189425388e-07, + "loss": 1.5408, + "step": 43199 + }, + { + "epoch": 9.43, + "learning_rate": 2.8621367708105747e-07, + "loss": 2.3023, + "step": 43200 + }, + { + "epoch": 9.43, + "learning_rate": 2.861044352195762e-07, + "loss": 1.4935, + "step": 43201 + }, + { + "epoch": 9.43, + "learning_rate": 2.8599519335809485e-07, + "loss": 2.4124, + "step": 43202 + }, + { + "epoch": 9.43, + "learning_rate": 2.858859514966135e-07, + "loss": 0.6887, + "step": 43203 + }, + { + "epoch": 9.43, + "learning_rate": 2.857767096351322e-07, + "loss": 1.6628, + "step": 43204 + }, + { + "epoch": 9.43, + "learning_rate": 2.856674677736509e-07, + "loss": 0.7831, + "step": 43205 + }, + { + "epoch": 9.43, + "learning_rate": 2.8555822591216956e-07, + "loss": 1.0079, + "step": 43206 + }, + { + "epoch": 9.43, + "learning_rate": 2.854489840506882e-07, + "loss": 0.0428, + "step": 43207 + }, + { + "epoch": 9.43, + "learning_rate": 2.8533974218920694e-07, + "loss": 1.4385, + "step": 43208 + }, + { + "epoch": 9.43, + "learning_rate": 2.852305003277256e-07, + "loss": 0.0467, + "step": 43209 + }, + { + "epoch": 9.43, + "learning_rate": 2.8512125846624427e-07, + "loss": 0.3679, + "step": 43210 + }, + { + "epoch": 9.43, + "learning_rate": 2.85012016604763e-07, + "loss": 0.7376, + "step": 43211 + }, + { + "epoch": 9.43, + "learning_rate": 2.8490277474328165e-07, + "loss": 0.6333, + "step": 43212 + }, + { + "epoch": 9.43, + "learning_rate": 2.847935328818003e-07, + "loss": 0.2834, + "step": 43213 + }, + { + "epoch": 9.43, + "learning_rate": 2.8468429102031903e-07, + "loss": 0.8477, + "step": 43214 + }, + { + "epoch": 9.43, + "learning_rate": 2.845750491588377e-07, + "loss": 0.9735, + "step": 43215 + }, + { + "epoch": 9.43, + "learning_rate": 2.8446580729735636e-07, + "loss": 0.9048, + "step": 43216 + }, + { + "epoch": 9.43, + "learning_rate": 2.84356565435875e-07, + "loss": 0.827, + "step": 43217 + }, + { + "epoch": 9.43, + "learning_rate": 2.8424732357439374e-07, + "loss": 1.0112, + "step": 43218 + }, + { + "epoch": 9.43, + "learning_rate": 2.841380817129124e-07, + "loss": 1.5468, + "step": 43219 + }, + { + "epoch": 9.43, + "learning_rate": 2.8402883985143107e-07, + "loss": 1.8836, + "step": 43220 + }, + { + "epoch": 9.43, + "learning_rate": 2.839195979899498e-07, + "loss": 2.228, + "step": 43221 + }, + { + "epoch": 9.43, + "learning_rate": 2.8381035612846845e-07, + "loss": 3.1171, + "step": 43222 + }, + { + "epoch": 9.43, + "learning_rate": 2.837011142669871e-07, + "loss": 1.2433, + "step": 43223 + }, + { + "epoch": 9.43, + "learning_rate": 2.8359187240550583e-07, + "loss": 2.355, + "step": 43224 + }, + { + "epoch": 9.43, + "learning_rate": 2.834826305440245e-07, + "loss": 0.6546, + "step": 43225 + }, + { + "epoch": 9.43, + "learning_rate": 2.8337338868254316e-07, + "loss": 1.7226, + "step": 43226 + }, + { + "epoch": 9.43, + "learning_rate": 2.832641468210618e-07, + "loss": 0.3841, + "step": 43227 + }, + { + "epoch": 9.43, + "learning_rate": 2.8315490495958054e-07, + "loss": 1.6328, + "step": 43228 + }, + { + "epoch": 9.43, + "learning_rate": 2.8304566309809925e-07, + "loss": 0.6591, + "step": 43229 + }, + { + "epoch": 9.43, + "learning_rate": 2.8293642123661786e-07, + "loss": 1.523, + "step": 43230 + }, + { + "epoch": 9.43, + "learning_rate": 2.828271793751366e-07, + "loss": 2.7898, + "step": 43231 + }, + { + "epoch": 9.44, + "learning_rate": 2.8271793751365525e-07, + "loss": 2.0204, + "step": 43232 + }, + { + "epoch": 9.44, + "learning_rate": 2.826086956521739e-07, + "loss": 0.7675, + "step": 43233 + }, + { + "epoch": 9.44, + "learning_rate": 2.824994537906926e-07, + "loss": 0.7373, + "step": 43234 + }, + { + "epoch": 9.44, + "learning_rate": 2.823902119292113e-07, + "loss": 0.0969, + "step": 43235 + }, + { + "epoch": 9.44, + "learning_rate": 2.8228097006773e-07, + "loss": 1.1853, + "step": 43236 + }, + { + "epoch": 9.44, + "learning_rate": 2.821717282062486e-07, + "loss": 0.3492, + "step": 43237 + }, + { + "epoch": 9.44, + "learning_rate": 2.8206248634476733e-07, + "loss": 0.8788, + "step": 43238 + }, + { + "epoch": 9.44, + "learning_rate": 2.8195324448328605e-07, + "loss": 0.2396, + "step": 43239 + }, + { + "epoch": 9.44, + "learning_rate": 2.818440026218047e-07, + "loss": 0.7159, + "step": 43240 + }, + { + "epoch": 9.44, + "learning_rate": 2.817347607603234e-07, + "loss": 0.0384, + "step": 43241 + }, + { + "epoch": 9.44, + "learning_rate": 2.8162551889884204e-07, + "loss": 0.9641, + "step": 43242 + }, + { + "epoch": 9.44, + "learning_rate": 2.8151627703736076e-07, + "loss": 0.7839, + "step": 43243 + }, + { + "epoch": 9.44, + "learning_rate": 2.814070351758794e-07, + "loss": 1.6042, + "step": 43244 + }, + { + "epoch": 9.44, + "learning_rate": 2.812977933143981e-07, + "loss": 0.3448, + "step": 43245 + }, + { + "epoch": 9.44, + "learning_rate": 2.811885514529168e-07, + "loss": 1.2994, + "step": 43246 + }, + { + "epoch": 9.44, + "learning_rate": 2.8107930959143547e-07, + "loss": 2.4415, + "step": 43247 + }, + { + "epoch": 9.44, + "learning_rate": 2.8097006772995413e-07, + "loss": 1.8619, + "step": 43248 + }, + { + "epoch": 9.44, + "learning_rate": 2.8086082586847285e-07, + "loss": 1.7626, + "step": 43249 + }, + { + "epoch": 9.44, + "learning_rate": 2.807515840069915e-07, + "loss": 0.8183, + "step": 43250 + }, + { + "epoch": 9.44, + "learning_rate": 2.806423421455102e-07, + "loss": 0.0631, + "step": 43251 + }, + { + "epoch": 9.44, + "learning_rate": 2.805331002840289e-07, + "loss": 0.4663, + "step": 43252 + }, + { + "epoch": 9.44, + "learning_rate": 2.8042385842254756e-07, + "loss": 1.8064, + "step": 43253 + }, + { + "epoch": 9.44, + "learning_rate": 2.803146165610662e-07, + "loss": 0.9901, + "step": 43254 + }, + { + "epoch": 9.44, + "learning_rate": 2.802053746995849e-07, + "loss": 1.0469, + "step": 43255 + }, + { + "epoch": 9.44, + "learning_rate": 2.800961328381036e-07, + "loss": 0.3029, + "step": 43256 + }, + { + "epoch": 9.44, + "learning_rate": 2.7998689097662227e-07, + "loss": 1.1677, + "step": 43257 + }, + { + "epoch": 9.44, + "learning_rate": 2.7987764911514093e-07, + "loss": 0.9642, + "step": 43258 + }, + { + "epoch": 9.44, + "learning_rate": 2.7976840725365965e-07, + "loss": 0.4165, + "step": 43259 + }, + { + "epoch": 9.44, + "learning_rate": 2.796591653921783e-07, + "loss": 1.0457, + "step": 43260 + }, + { + "epoch": 9.44, + "learning_rate": 2.79549923530697e-07, + "loss": 1.1811, + "step": 43261 + }, + { + "epoch": 9.44, + "learning_rate": 2.794406816692157e-07, + "loss": 0.7267, + "step": 43262 + }, + { + "epoch": 9.44, + "learning_rate": 2.7933143980773436e-07, + "loss": 0.2853, + "step": 43263 + }, + { + "epoch": 9.44, + "learning_rate": 2.79222197946253e-07, + "loss": 0.7754, + "step": 43264 + }, + { + "epoch": 9.44, + "learning_rate": 2.791129560847717e-07, + "loss": 0.505, + "step": 43265 + }, + { + "epoch": 9.44, + "learning_rate": 2.790037142232904e-07, + "loss": 0.4124, + "step": 43266 + }, + { + "epoch": 9.44, + "learning_rate": 2.7889447236180907e-07, + "loss": 1.0811, + "step": 43267 + }, + { + "epoch": 9.44, + "learning_rate": 2.7878523050032773e-07, + "loss": 3.4164, + "step": 43268 + }, + { + "epoch": 9.44, + "learning_rate": 2.7867598863884645e-07, + "loss": 1.8418, + "step": 43269 + }, + { + "epoch": 9.44, + "learning_rate": 2.785667467773651e-07, + "loss": 0.9413, + "step": 43270 + }, + { + "epoch": 9.44, + "learning_rate": 2.784575049158838e-07, + "loss": 1.7345, + "step": 43271 + }, + { + "epoch": 9.44, + "learning_rate": 2.783482630544025e-07, + "loss": 1.3494, + "step": 43272 + }, + { + "epoch": 9.44, + "learning_rate": 2.7823902119292115e-07, + "loss": 1.5458, + "step": 43273 + }, + { + "epoch": 9.44, + "learning_rate": 2.781297793314398e-07, + "loss": 1.4413, + "step": 43274 + }, + { + "epoch": 9.44, + "learning_rate": 2.780205374699585e-07, + "loss": 1.4114, + "step": 43275 + }, + { + "epoch": 9.44, + "learning_rate": 2.779112956084772e-07, + "loss": 0.0608, + "step": 43276 + }, + { + "epoch": 9.45, + "learning_rate": 2.7780205374699586e-07, + "loss": 0.5547, + "step": 43277 + }, + { + "epoch": 9.45, + "learning_rate": 2.7769281188551453e-07, + "loss": 0.5295, + "step": 43278 + }, + { + "epoch": 9.45, + "learning_rate": 2.7758357002403324e-07, + "loss": 0.291, + "step": 43279 + }, + { + "epoch": 9.45, + "learning_rate": 2.774743281625519e-07, + "loss": 0.3699, + "step": 43280 + }, + { + "epoch": 9.45, + "learning_rate": 2.7736508630107057e-07, + "loss": 0.839, + "step": 43281 + }, + { + "epoch": 9.45, + "learning_rate": 2.772558444395893e-07, + "loss": 0.6686, + "step": 43282 + }, + { + "epoch": 9.45, + "learning_rate": 2.7714660257810795e-07, + "loss": 1.8183, + "step": 43283 + }, + { + "epoch": 9.45, + "learning_rate": 2.770373607166266e-07, + "loss": 2.7171, + "step": 43284 + }, + { + "epoch": 9.45, + "learning_rate": 2.769281188551453e-07, + "loss": 0.2851, + "step": 43285 + }, + { + "epoch": 9.45, + "learning_rate": 2.76818876993664e-07, + "loss": 2.1397, + "step": 43286 + }, + { + "epoch": 9.45, + "learning_rate": 2.7670963513218266e-07, + "loss": 1.3023, + "step": 43287 + }, + { + "epoch": 9.45, + "learning_rate": 2.766003932707013e-07, + "loss": 3.1643, + "step": 43288 + }, + { + "epoch": 9.45, + "learning_rate": 2.7649115140922004e-07, + "loss": 1.8654, + "step": 43289 + }, + { + "epoch": 9.45, + "learning_rate": 2.7638190954773876e-07, + "loss": 2.6387, + "step": 43290 + }, + { + "epoch": 9.45, + "learning_rate": 2.7627266768625737e-07, + "loss": 0.6288, + "step": 43291 + }, + { + "epoch": 9.45, + "learning_rate": 2.761634258247761e-07, + "loss": 2.0113, + "step": 43292 + }, + { + "epoch": 9.45, + "learning_rate": 2.7605418396329475e-07, + "loss": 0.0974, + "step": 43293 + }, + { + "epoch": 9.45, + "learning_rate": 2.759449421018134e-07, + "loss": 2.4025, + "step": 43294 + }, + { + "epoch": 9.45, + "learning_rate": 2.7583570024033213e-07, + "loss": 0.068, + "step": 43295 + }, + { + "epoch": 9.45, + "learning_rate": 2.757264583788508e-07, + "loss": 1.6646, + "step": 43296 + }, + { + "epoch": 9.45, + "learning_rate": 2.756172165173695e-07, + "loss": 0.2304, + "step": 43297 + }, + { + "epoch": 9.45, + "learning_rate": 2.755079746558881e-07, + "loss": 1.9156, + "step": 43298 + }, + { + "epoch": 9.45, + "learning_rate": 2.7539873279440684e-07, + "loss": 0.5281, + "step": 43299 + }, + { + "epoch": 9.45, + "learning_rate": 2.7528949093292556e-07, + "loss": 0.6964, + "step": 43300 + }, + { + "epoch": 9.45, + "learning_rate": 2.7518024907144417e-07, + "loss": 0.9184, + "step": 43301 + }, + { + "epoch": 9.45, + "learning_rate": 2.750710072099629e-07, + "loss": 1.227, + "step": 43302 + }, + { + "epoch": 9.45, + "learning_rate": 2.7496176534848155e-07, + "loss": 0.7796, + "step": 43303 + }, + { + "epoch": 9.45, + "learning_rate": 2.7485252348700027e-07, + "loss": 3.8491, + "step": 43304 + }, + { + "epoch": 9.45, + "learning_rate": 2.7474328162551893e-07, + "loss": 1.7497, + "step": 43305 + }, + { + "epoch": 9.45, + "learning_rate": 2.746340397640376e-07, + "loss": 0.3259, + "step": 43306 + }, + { + "epoch": 9.45, + "learning_rate": 2.745247979025563e-07, + "loss": 1.441, + "step": 43307 + }, + { + "epoch": 9.45, + "learning_rate": 2.74415556041075e-07, + "loss": 1.6305, + "step": 43308 + }, + { + "epoch": 9.45, + "learning_rate": 2.7430631417959364e-07, + "loss": 0.8834, + "step": 43309 + }, + { + "epoch": 9.45, + "learning_rate": 2.7419707231811236e-07, + "loss": 0.5507, + "step": 43310 + }, + { + "epoch": 9.45, + "learning_rate": 2.74087830456631e-07, + "loss": 2.7128, + "step": 43311 + }, + { + "epoch": 9.45, + "learning_rate": 2.739785885951497e-07, + "loss": 0.2444, + "step": 43312 + }, + { + "epoch": 9.45, + "learning_rate": 2.7386934673366835e-07, + "loss": 0.1436, + "step": 43313 + }, + { + "epoch": 9.45, + "learning_rate": 2.7376010487218706e-07, + "loss": 2.2895, + "step": 43314 + }, + { + "epoch": 9.45, + "learning_rate": 2.7365086301070573e-07, + "loss": 0.9498, + "step": 43315 + }, + { + "epoch": 9.45, + "learning_rate": 2.735416211492244e-07, + "loss": 0.1451, + "step": 43316 + }, + { + "epoch": 9.45, + "learning_rate": 2.734323792877431e-07, + "loss": 0.1249, + "step": 43317 + }, + { + "epoch": 9.45, + "learning_rate": 2.7332313742626177e-07, + "loss": 0.5646, + "step": 43318 + }, + { + "epoch": 9.45, + "learning_rate": 2.7321389556478044e-07, + "loss": 1.0606, + "step": 43319 + }, + { + "epoch": 9.45, + "learning_rate": 2.7310465370329915e-07, + "loss": 0.1139, + "step": 43320 + }, + { + "epoch": 9.45, + "learning_rate": 2.729954118418178e-07, + "loss": 0.6302, + "step": 43321 + }, + { + "epoch": 9.45, + "learning_rate": 2.728861699803365e-07, + "loss": 0.9587, + "step": 43322 + }, + { + "epoch": 9.46, + "learning_rate": 2.7277692811885515e-07, + "loss": 1.3564, + "step": 43323 + }, + { + "epoch": 9.46, + "learning_rate": 2.7266768625737386e-07, + "loss": 1.2185, + "step": 43324 + }, + { + "epoch": 9.46, + "learning_rate": 2.725584443958925e-07, + "loss": 0.6411, + "step": 43325 + }, + { + "epoch": 9.46, + "learning_rate": 2.724492025344112e-07, + "loss": 1.2897, + "step": 43326 + }, + { + "epoch": 9.46, + "learning_rate": 2.723399606729299e-07, + "loss": 0.455, + "step": 43327 + }, + { + "epoch": 9.46, + "learning_rate": 2.7223071881144857e-07, + "loss": 0.4322, + "step": 43328 + }, + { + "epoch": 9.46, + "learning_rate": 2.7212147694996723e-07, + "loss": 0.6328, + "step": 43329 + }, + { + "epoch": 9.46, + "learning_rate": 2.7201223508848595e-07, + "loss": 1.7506, + "step": 43330 + }, + { + "epoch": 9.46, + "learning_rate": 2.719029932270046e-07, + "loss": 0.7412, + "step": 43331 + }, + { + "epoch": 9.46, + "learning_rate": 2.717937513655233e-07, + "loss": 2.255, + "step": 43332 + }, + { + "epoch": 9.46, + "learning_rate": 2.71684509504042e-07, + "loss": 0.8061, + "step": 43333 + }, + { + "epoch": 9.46, + "learning_rate": 2.7157526764256066e-07, + "loss": 0.9451, + "step": 43334 + }, + { + "epoch": 9.46, + "learning_rate": 2.714660257810793e-07, + "loss": 0.7048, + "step": 43335 + }, + { + "epoch": 9.46, + "learning_rate": 2.71356783919598e-07, + "loss": 0.751, + "step": 43336 + }, + { + "epoch": 9.46, + "learning_rate": 2.712475420581167e-07, + "loss": 2.5075, + "step": 43337 + }, + { + "epoch": 9.46, + "learning_rate": 2.7113830019663537e-07, + "loss": 0.615, + "step": 43338 + }, + { + "epoch": 9.46, + "learning_rate": 2.7102905833515403e-07, + "loss": 0.6663, + "step": 43339 + }, + { + "epoch": 9.46, + "learning_rate": 2.7091981647367275e-07, + "loss": 1.3869, + "step": 43340 + }, + { + "epoch": 9.46, + "learning_rate": 2.708105746121914e-07, + "loss": 1.9083, + "step": 43341 + }, + { + "epoch": 9.46, + "learning_rate": 2.707013327507101e-07, + "loss": 0.7385, + "step": 43342 + }, + { + "epoch": 9.46, + "learning_rate": 2.705920908892288e-07, + "loss": 1.3865, + "step": 43343 + }, + { + "epoch": 9.46, + "learning_rate": 2.7048284902774746e-07, + "loss": 0.7415, + "step": 43344 + }, + { + "epoch": 9.46, + "learning_rate": 2.703736071662661e-07, + "loss": 1.0528, + "step": 43345 + }, + { + "epoch": 9.46, + "learning_rate": 2.702643653047848e-07, + "loss": 1.9727, + "step": 43346 + }, + { + "epoch": 9.46, + "learning_rate": 2.701551234433035e-07, + "loss": 0.5448, + "step": 43347 + }, + { + "epoch": 9.46, + "learning_rate": 2.7004588158182217e-07, + "loss": 1.3036, + "step": 43348 + }, + { + "epoch": 9.46, + "learning_rate": 2.6993663972034083e-07, + "loss": 0.0845, + "step": 43349 + }, + { + "epoch": 9.46, + "learning_rate": 2.6982739785885955e-07, + "loss": 0.8669, + "step": 43350 + }, + { + "epoch": 9.46, + "learning_rate": 2.697181559973782e-07, + "loss": 1.6311, + "step": 43351 + }, + { + "epoch": 9.46, + "learning_rate": 2.696089141358969e-07, + "loss": 0.91, + "step": 43352 + }, + { + "epoch": 9.46, + "learning_rate": 2.694996722744156e-07, + "loss": 0.0834, + "step": 43353 + }, + { + "epoch": 9.46, + "learning_rate": 2.6939043041293426e-07, + "loss": 0.5195, + "step": 43354 + }, + { + "epoch": 9.46, + "learning_rate": 2.692811885514529e-07, + "loss": 0.7519, + "step": 43355 + }, + { + "epoch": 9.46, + "learning_rate": 2.691719466899716e-07, + "loss": 1.4977, + "step": 43356 + }, + { + "epoch": 9.46, + "learning_rate": 2.690627048284903e-07, + "loss": 0.3704, + "step": 43357 + }, + { + "epoch": 9.46, + "learning_rate": 2.68953462967009e-07, + "loss": 0.8846, + "step": 43358 + }, + { + "epoch": 9.46, + "learning_rate": 2.6884422110552763e-07, + "loss": 1.6434, + "step": 43359 + }, + { + "epoch": 9.46, + "learning_rate": 2.6873497924404635e-07, + "loss": 1.3477, + "step": 43360 + }, + { + "epoch": 9.46, + "learning_rate": 2.68625737382565e-07, + "loss": 1.4819, + "step": 43361 + }, + { + "epoch": 9.46, + "learning_rate": 2.685164955210837e-07, + "loss": 1.3054, + "step": 43362 + }, + { + "epoch": 9.46, + "learning_rate": 2.684072536596024e-07, + "loss": 1.576, + "step": 43363 + }, + { + "epoch": 9.46, + "learning_rate": 2.6829801179812105e-07, + "loss": 1.3007, + "step": 43364 + }, + { + "epoch": 9.46, + "learning_rate": 2.6818876993663977e-07, + "loss": 0.116, + "step": 43365 + }, + { + "epoch": 9.46, + "learning_rate": 2.680795280751584e-07, + "loss": 0.3867, + "step": 43366 + }, + { + "epoch": 9.46, + "learning_rate": 2.679702862136771e-07, + "loss": 1.4112, + "step": 43367 + }, + { + "epoch": 9.46, + "learning_rate": 2.678610443521958e-07, + "loss": 1.1973, + "step": 43368 + }, + { + "epoch": 9.47, + "learning_rate": 2.6775180249071443e-07, + "loss": 1.9612, + "step": 43369 + }, + { + "epoch": 9.47, + "learning_rate": 2.6764256062923314e-07, + "loss": 0.3782, + "step": 43370 + }, + { + "epoch": 9.47, + "learning_rate": 2.6753331876775186e-07, + "loss": 0.4287, + "step": 43371 + }, + { + "epoch": 9.47, + "learning_rate": 2.674240769062705e-07, + "loss": 2.2995, + "step": 43372 + }, + { + "epoch": 9.47, + "learning_rate": 2.673148350447892e-07, + "loss": 0.3676, + "step": 43373 + }, + { + "epoch": 9.47, + "learning_rate": 2.6720559318330785e-07, + "loss": 0.3489, + "step": 43374 + }, + { + "epoch": 9.47, + "learning_rate": 2.6709635132182657e-07, + "loss": 0.6861, + "step": 43375 + }, + { + "epoch": 9.47, + "learning_rate": 2.6698710946034523e-07, + "loss": 0.499, + "step": 43376 + }, + { + "epoch": 9.47, + "learning_rate": 2.668778675988639e-07, + "loss": 0.789, + "step": 43377 + }, + { + "epoch": 9.47, + "learning_rate": 2.667686257373826e-07, + "loss": 0.5387, + "step": 43378 + }, + { + "epoch": 9.47, + "learning_rate": 2.666593838759013e-07, + "loss": 4.5925, + "step": 43379 + }, + { + "epoch": 9.47, + "learning_rate": 2.6655014201441994e-07, + "loss": 0.4823, + "step": 43380 + }, + { + "epoch": 9.47, + "learning_rate": 2.6644090015293866e-07, + "loss": 0.68, + "step": 43381 + }, + { + "epoch": 9.47, + "learning_rate": 2.663316582914573e-07, + "loss": 0.2348, + "step": 43382 + }, + { + "epoch": 9.47, + "learning_rate": 2.66222416429976e-07, + "loss": 1.4106, + "step": 43383 + }, + { + "epoch": 9.47, + "learning_rate": 2.6611317456849465e-07, + "loss": 0.7108, + "step": 43384 + }, + { + "epoch": 9.47, + "learning_rate": 2.6600393270701337e-07, + "loss": 1.7774, + "step": 43385 + }, + { + "epoch": 9.47, + "learning_rate": 2.6589469084553203e-07, + "loss": 0.7926, + "step": 43386 + }, + { + "epoch": 9.47, + "learning_rate": 2.657854489840507e-07, + "loss": 2.1014, + "step": 43387 + }, + { + "epoch": 9.47, + "learning_rate": 2.656762071225694e-07, + "loss": 1.3113, + "step": 43388 + }, + { + "epoch": 9.47, + "learning_rate": 2.655669652610881e-07, + "loss": 1.1436, + "step": 43389 + }, + { + "epoch": 9.47, + "learning_rate": 2.6545772339960674e-07, + "loss": 0.1728, + "step": 43390 + }, + { + "epoch": 9.47, + "learning_rate": 2.6534848153812546e-07, + "loss": 3.2124, + "step": 43391 + }, + { + "epoch": 9.47, + "learning_rate": 2.652392396766441e-07, + "loss": 1.1818, + "step": 43392 + }, + { + "epoch": 9.47, + "learning_rate": 2.651299978151628e-07, + "loss": 0.4143, + "step": 43393 + }, + { + "epoch": 9.47, + "learning_rate": 2.6502075595368145e-07, + "loss": 2.5082, + "step": 43394 + }, + { + "epoch": 9.47, + "learning_rate": 2.6491151409220017e-07, + "loss": 0.5727, + "step": 43395 + }, + { + "epoch": 9.47, + "learning_rate": 2.6480227223071883e-07, + "loss": 0.4253, + "step": 43396 + }, + { + "epoch": 9.47, + "learning_rate": 2.646930303692375e-07, + "loss": 0.4069, + "step": 43397 + }, + { + "epoch": 9.47, + "learning_rate": 2.645837885077562e-07, + "loss": 5.6997, + "step": 43398 + }, + { + "epoch": 9.47, + "learning_rate": 2.644745466462749e-07, + "loss": 1.2733, + "step": 43399 + }, + { + "epoch": 9.47, + "learning_rate": 2.6436530478479354e-07, + "loss": 1.8475, + "step": 43400 + }, + { + "epoch": 9.47, + "learning_rate": 2.6425606292331226e-07, + "loss": 1.5705, + "step": 43401 + }, + { + "epoch": 9.47, + "learning_rate": 2.641468210618309e-07, + "loss": 0.5071, + "step": 43402 + }, + { + "epoch": 9.47, + "learning_rate": 2.640375792003496e-07, + "loss": 4.1155, + "step": 43403 + }, + { + "epoch": 9.47, + "learning_rate": 2.6392833733886825e-07, + "loss": 1.0277, + "step": 43404 + }, + { + "epoch": 9.47, + "learning_rate": 2.6381909547738696e-07, + "loss": 1.7826, + "step": 43405 + }, + { + "epoch": 9.47, + "learning_rate": 2.6370985361590563e-07, + "loss": 1.0835, + "step": 43406 + }, + { + "epoch": 9.47, + "learning_rate": 2.636006117544243e-07, + "loss": 1.4787, + "step": 43407 + }, + { + "epoch": 9.47, + "learning_rate": 2.63491369892943e-07, + "loss": 0.6323, + "step": 43408 + }, + { + "epoch": 9.47, + "learning_rate": 2.6338212803146167e-07, + "loss": 1.0111, + "step": 43409 + }, + { + "epoch": 9.47, + "learning_rate": 2.6327288616998034e-07, + "loss": 1.4955, + "step": 43410 + }, + { + "epoch": 9.47, + "learning_rate": 2.6316364430849905e-07, + "loss": 0.4278, + "step": 43411 + }, + { + "epoch": 9.47, + "learning_rate": 2.630544024470177e-07, + "loss": 1.2412, + "step": 43412 + }, + { + "epoch": 9.47, + "learning_rate": 2.629451605855364e-07, + "loss": 0.0633, + "step": 43413 + }, + { + "epoch": 9.47, + "learning_rate": 2.628359187240551e-07, + "loss": 1.1674, + "step": 43414 + }, + { + "epoch": 9.48, + "learning_rate": 2.6272667686257376e-07, + "loss": 0.5923, + "step": 43415 + }, + { + "epoch": 9.48, + "learning_rate": 2.626174350010924e-07, + "loss": 0.7214, + "step": 43416 + }, + { + "epoch": 9.48, + "learning_rate": 2.625081931396111e-07, + "loss": 0.4052, + "step": 43417 + }, + { + "epoch": 9.48, + "learning_rate": 2.623989512781298e-07, + "loss": 0.6213, + "step": 43418 + }, + { + "epoch": 9.48, + "learning_rate": 2.622897094166485e-07, + "loss": 1.8129, + "step": 43419 + }, + { + "epoch": 9.48, + "learning_rate": 2.6218046755516713e-07, + "loss": 1.6454, + "step": 43420 + }, + { + "epoch": 9.48, + "learning_rate": 2.6207122569368585e-07, + "loss": 2.5961, + "step": 43421 + }, + { + "epoch": 9.48, + "learning_rate": 2.619619838322045e-07, + "loss": 0.6394, + "step": 43422 + }, + { + "epoch": 9.48, + "learning_rate": 2.618527419707232e-07, + "loss": 1.5303, + "step": 43423 + }, + { + "epoch": 9.48, + "learning_rate": 2.617435001092419e-07, + "loss": 2.4358, + "step": 43424 + }, + { + "epoch": 9.48, + "learning_rate": 2.6163425824776056e-07, + "loss": 1.0604, + "step": 43425 + }, + { + "epoch": 9.48, + "learning_rate": 2.615250163862793e-07, + "loss": 1.1872, + "step": 43426 + }, + { + "epoch": 9.48, + "learning_rate": 2.614157745247979e-07, + "loss": 1.2324, + "step": 43427 + }, + { + "epoch": 9.48, + "learning_rate": 2.613065326633166e-07, + "loss": 1.837, + "step": 43428 + }, + { + "epoch": 9.48, + "learning_rate": 2.611972908018353e-07, + "loss": 1.4495, + "step": 43429 + }, + { + "epoch": 9.48, + "learning_rate": 2.6108804894035393e-07, + "loss": 2.7841, + "step": 43430 + }, + { + "epoch": 9.48, + "learning_rate": 2.6097880707887265e-07, + "loss": 0.9282, + "step": 43431 + }, + { + "epoch": 9.48, + "learning_rate": 2.608695652173913e-07, + "loss": 1.8443, + "step": 43432 + }, + { + "epoch": 9.48, + "learning_rate": 2.6076032335591003e-07, + "loss": 1.2425, + "step": 43433 + }, + { + "epoch": 9.48, + "learning_rate": 2.606510814944287e-07, + "loss": 0.4183, + "step": 43434 + }, + { + "epoch": 9.48, + "learning_rate": 2.6054183963294736e-07, + "loss": 0.3176, + "step": 43435 + }, + { + "epoch": 9.48, + "learning_rate": 2.604325977714661e-07, + "loss": 1.1966, + "step": 43436 + }, + { + "epoch": 9.48, + "learning_rate": 2.6032335590998474e-07, + "loss": 0.14, + "step": 43437 + }, + { + "epoch": 9.48, + "learning_rate": 2.602141140485034e-07, + "loss": 1.3143, + "step": 43438 + }, + { + "epoch": 9.48, + "learning_rate": 2.601048721870221e-07, + "loss": 1.444, + "step": 43439 + }, + { + "epoch": 9.48, + "learning_rate": 2.599956303255408e-07, + "loss": 0.6577, + "step": 43440 + }, + { + "epoch": 9.48, + "learning_rate": 2.5988638846405945e-07, + "loss": 2.3349, + "step": 43441 + }, + { + "epoch": 9.48, + "learning_rate": 2.597771466025781e-07, + "loss": 1.4842, + "step": 43442 + }, + { + "epoch": 9.48, + "learning_rate": 2.5966790474109683e-07, + "loss": 1.704, + "step": 43443 + }, + { + "epoch": 9.48, + "learning_rate": 2.595586628796155e-07, + "loss": 1.3699, + "step": 43444 + }, + { + "epoch": 9.48, + "learning_rate": 2.5944942101813416e-07, + "loss": 0.5933, + "step": 43445 + }, + { + "epoch": 9.48, + "learning_rate": 2.5934017915665287e-07, + "loss": 2.6664, + "step": 43446 + }, + { + "epoch": 9.48, + "learning_rate": 2.5923093729517154e-07, + "loss": 3.3428, + "step": 43447 + }, + { + "epoch": 9.48, + "learning_rate": 2.591216954336902e-07, + "loss": 0.3623, + "step": 43448 + }, + { + "epoch": 9.48, + "learning_rate": 2.590124535722089e-07, + "loss": 1.2107, + "step": 43449 + }, + { + "epoch": 9.48, + "learning_rate": 2.589032117107276e-07, + "loss": 1.0754, + "step": 43450 + }, + { + "epoch": 9.48, + "learning_rate": 2.5879396984924625e-07, + "loss": 0.2835, + "step": 43451 + }, + { + "epoch": 9.48, + "learning_rate": 2.5868472798776496e-07, + "loss": 1.1479, + "step": 43452 + }, + { + "epoch": 9.48, + "learning_rate": 2.585754861262836e-07, + "loss": 0.1049, + "step": 43453 + }, + { + "epoch": 9.48, + "learning_rate": 2.584662442648023e-07, + "loss": 1.0384, + "step": 43454 + }, + { + "epoch": 9.48, + "learning_rate": 2.5835700240332095e-07, + "loss": 0.7135, + "step": 43455 + }, + { + "epoch": 9.48, + "learning_rate": 2.5824776054183967e-07, + "loss": 0.7656, + "step": 43456 + }, + { + "epoch": 9.48, + "learning_rate": 2.5813851868035833e-07, + "loss": 1.3612, + "step": 43457 + }, + { + "epoch": 9.48, + "learning_rate": 2.58029276818877e-07, + "loss": 1.0563, + "step": 43458 + }, + { + "epoch": 9.48, + "learning_rate": 2.579200349573957e-07, + "loss": 0.9216, + "step": 43459 + }, + { + "epoch": 9.48, + "learning_rate": 2.578107930959144e-07, + "loss": 1.5046, + "step": 43460 + }, + { + "epoch": 9.49, + "learning_rate": 2.5770155123443304e-07, + "loss": 1.7193, + "step": 43461 + }, + { + "epoch": 9.49, + "learning_rate": 2.5759230937295176e-07, + "loss": 1.1193, + "step": 43462 + }, + { + "epoch": 9.49, + "learning_rate": 2.574830675114704e-07, + "loss": 1.1491, + "step": 43463 + }, + { + "epoch": 9.49, + "learning_rate": 2.573738256499891e-07, + "loss": 1.8093, + "step": 43464 + }, + { + "epoch": 9.49, + "learning_rate": 2.5726458378850775e-07, + "loss": 0.9258, + "step": 43465 + }, + { + "epoch": 9.49, + "learning_rate": 2.5715534192702647e-07, + "loss": 1.3365, + "step": 43466 + }, + { + "epoch": 9.49, + "learning_rate": 2.5704610006554513e-07, + "loss": 0.894, + "step": 43467 + }, + { + "epoch": 9.49, + "learning_rate": 2.569368582040638e-07, + "loss": 1.6792, + "step": 43468 + }, + { + "epoch": 9.49, + "learning_rate": 2.568276163425825e-07, + "loss": 1.8631, + "step": 43469 + }, + { + "epoch": 9.49, + "learning_rate": 2.567183744811012e-07, + "loss": 1.0921, + "step": 43470 + }, + { + "epoch": 9.49, + "learning_rate": 2.5660913261961984e-07, + "loss": 0.864, + "step": 43471 + }, + { + "epoch": 9.49, + "learning_rate": 2.5649989075813856e-07, + "loss": 0.3048, + "step": 43472 + }, + { + "epoch": 9.49, + "learning_rate": 2.563906488966572e-07, + "loss": 1.0329, + "step": 43473 + }, + { + "epoch": 9.49, + "learning_rate": 2.562814070351759e-07, + "loss": 0.5932, + "step": 43474 + }, + { + "epoch": 9.49, + "learning_rate": 2.5617216517369455e-07, + "loss": 0.3272, + "step": 43475 + }, + { + "epoch": 9.49, + "learning_rate": 2.5606292331221327e-07, + "loss": 1.1454, + "step": 43476 + }, + { + "epoch": 9.49, + "learning_rate": 2.5595368145073193e-07, + "loss": 2.5594, + "step": 43477 + }, + { + "epoch": 9.49, + "learning_rate": 2.558444395892506e-07, + "loss": 1.8345, + "step": 43478 + }, + { + "epoch": 9.49, + "learning_rate": 2.557351977277693e-07, + "loss": 2.2994, + "step": 43479 + }, + { + "epoch": 9.49, + "learning_rate": 2.55625955866288e-07, + "loss": 1.2267, + "step": 43480 + }, + { + "epoch": 9.49, + "learning_rate": 2.5551671400480664e-07, + "loss": 0.811, + "step": 43481 + }, + { + "epoch": 9.49, + "learning_rate": 2.5540747214332536e-07, + "loss": 0.1074, + "step": 43482 + }, + { + "epoch": 9.49, + "learning_rate": 2.55298230281844e-07, + "loss": 0.0389, + "step": 43483 + }, + { + "epoch": 9.49, + "learning_rate": 2.551889884203627e-07, + "loss": 1.9713, + "step": 43484 + }, + { + "epoch": 9.49, + "learning_rate": 2.5507974655888135e-07, + "loss": 0.2988, + "step": 43485 + }, + { + "epoch": 9.49, + "learning_rate": 2.5497050469740007e-07, + "loss": 1.0864, + "step": 43486 + }, + { + "epoch": 9.49, + "learning_rate": 2.548612628359188e-07, + "loss": 0.3655, + "step": 43487 + }, + { + "epoch": 9.49, + "learning_rate": 2.547520209744374e-07, + "loss": 0.7473, + "step": 43488 + }, + { + "epoch": 9.49, + "learning_rate": 2.546427791129561e-07, + "loss": 0.4812, + "step": 43489 + }, + { + "epoch": 9.49, + "learning_rate": 2.545335372514748e-07, + "loss": 1.2231, + "step": 43490 + }, + { + "epoch": 9.49, + "learning_rate": 2.5442429538999344e-07, + "loss": 0.0921, + "step": 43491 + }, + { + "epoch": 9.49, + "learning_rate": 2.5431505352851215e-07, + "loss": 0.9288, + "step": 43492 + }, + { + "epoch": 9.49, + "learning_rate": 2.542058116670308e-07, + "loss": 1.9477, + "step": 43493 + }, + { + "epoch": 9.49, + "learning_rate": 2.5409656980554954e-07, + "loss": 1.0854, + "step": 43494 + }, + { + "epoch": 9.49, + "learning_rate": 2.539873279440682e-07, + "loss": 0.2141, + "step": 43495 + }, + { + "epoch": 9.49, + "learning_rate": 2.5387808608258686e-07, + "loss": 0.8314, + "step": 43496 + }, + { + "epoch": 9.49, + "learning_rate": 2.537688442211056e-07, + "loss": 0.5265, + "step": 43497 + }, + { + "epoch": 9.49, + "learning_rate": 2.536596023596242e-07, + "loss": 3.8022, + "step": 43498 + }, + { + "epoch": 9.49, + "learning_rate": 2.535503604981429e-07, + "loss": 0.5522, + "step": 43499 + }, + { + "epoch": 9.49, + "learning_rate": 2.534411186366616e-07, + "loss": 0.9155, + "step": 43500 + }, + { + "epoch": 9.49, + "learning_rate": 2.533318767751803e-07, + "loss": 2.0648, + "step": 43501 + }, + { + "epoch": 9.49, + "learning_rate": 2.5322263491369895e-07, + "loss": 0.5097, + "step": 43502 + }, + { + "epoch": 9.49, + "learning_rate": 2.531133930522176e-07, + "loss": 2.1048, + "step": 43503 + }, + { + "epoch": 9.49, + "learning_rate": 2.5300415119073633e-07, + "loss": 1.4352, + "step": 43504 + }, + { + "epoch": 9.49, + "learning_rate": 2.52894909329255e-07, + "loss": 1.2979, + "step": 43505 + }, + { + "epoch": 9.49, + "learning_rate": 2.5278566746777366e-07, + "loss": 2.6744, + "step": 43506 + }, + { + "epoch": 9.5, + "learning_rate": 2.526764256062924e-07, + "loss": 1.5907, + "step": 43507 + }, + { + "epoch": 9.5, + "learning_rate": 2.5256718374481104e-07, + "loss": 2.165, + "step": 43508 + }, + { + "epoch": 9.5, + "learning_rate": 2.524579418833297e-07, + "loss": 0.0482, + "step": 43509 + }, + { + "epoch": 9.5, + "learning_rate": 2.523487000218484e-07, + "loss": 1.0884, + "step": 43510 + }, + { + "epoch": 9.5, + "learning_rate": 2.522394581603671e-07, + "loss": 1.6919, + "step": 43511 + }, + { + "epoch": 9.5, + "learning_rate": 2.5213021629888575e-07, + "loss": 0.6647, + "step": 43512 + }, + { + "epoch": 9.5, + "learning_rate": 2.520209744374044e-07, + "loss": 0.7038, + "step": 43513 + }, + { + "epoch": 9.5, + "learning_rate": 2.5191173257592313e-07, + "loss": 0.5516, + "step": 43514 + }, + { + "epoch": 9.5, + "learning_rate": 2.518024907144418e-07, + "loss": 0.944, + "step": 43515 + }, + { + "epoch": 9.5, + "learning_rate": 2.5169324885296046e-07, + "loss": 1.2398, + "step": 43516 + }, + { + "epoch": 9.5, + "learning_rate": 2.515840069914792e-07, + "loss": 0.8738, + "step": 43517 + }, + { + "epoch": 9.5, + "learning_rate": 2.5147476512999784e-07, + "loss": 2.0747, + "step": 43518 + }, + { + "epoch": 9.5, + "learning_rate": 2.513655232685165e-07, + "loss": 1.5027, + "step": 43519 + }, + { + "epoch": 9.5, + "learning_rate": 2.512562814070352e-07, + "loss": 0.2219, + "step": 43520 + }, + { + "epoch": 9.5, + "learning_rate": 2.511470395455539e-07, + "loss": 0.331, + "step": 43521 + }, + { + "epoch": 9.5, + "learning_rate": 2.5103779768407255e-07, + "loss": 1.8955, + "step": 43522 + }, + { + "epoch": 9.5, + "learning_rate": 2.509285558225912e-07, + "loss": 0.5875, + "step": 43523 + }, + { + "epoch": 9.5, + "learning_rate": 2.5081931396110993e-07, + "loss": 0.0454, + "step": 43524 + }, + { + "epoch": 9.5, + "learning_rate": 2.507100720996286e-07, + "loss": 0.3671, + "step": 43525 + }, + { + "epoch": 9.5, + "learning_rate": 2.5060083023814726e-07, + "loss": 1.2184, + "step": 43526 + }, + { + "epoch": 9.5, + "learning_rate": 2.50491588376666e-07, + "loss": 0.7605, + "step": 43527 + }, + { + "epoch": 9.5, + "learning_rate": 2.5038234651518464e-07, + "loss": 0.6193, + "step": 43528 + }, + { + "epoch": 9.5, + "learning_rate": 2.502731046537033e-07, + "loss": 2.3245, + "step": 43529 + }, + { + "epoch": 9.5, + "learning_rate": 2.50163862792222e-07, + "loss": 0.7374, + "step": 43530 + }, + { + "epoch": 9.5, + "learning_rate": 2.500546209307407e-07, + "loss": 0.6439, + "step": 43531 + }, + { + "epoch": 9.5, + "learning_rate": 2.4994537906925935e-07, + "loss": 1.3045, + "step": 43532 + }, + { + "epoch": 9.5, + "learning_rate": 2.4983613720777806e-07, + "loss": 1.734, + "step": 43533 + }, + { + "epoch": 9.5, + "learning_rate": 2.4972689534629673e-07, + "loss": 0.989, + "step": 43534 + }, + { + "epoch": 9.5, + "learning_rate": 2.496176534848154e-07, + "loss": 0.3412, + "step": 43535 + }, + { + "epoch": 9.5, + "learning_rate": 2.4950841162333406e-07, + "loss": 1.0094, + "step": 43536 + }, + { + "epoch": 9.5, + "learning_rate": 2.4939916976185277e-07, + "loss": 0.1949, + "step": 43537 + }, + { + "epoch": 9.5, + "learning_rate": 2.4928992790037144e-07, + "loss": 1.8171, + "step": 43538 + }, + { + "epoch": 9.5, + "learning_rate": 2.491806860388901e-07, + "loss": 1.3788, + "step": 43539 + }, + { + "epoch": 9.5, + "learning_rate": 2.490714441774088e-07, + "loss": 1.6752, + "step": 43540 + }, + { + "epoch": 9.5, + "learning_rate": 2.489622023159275e-07, + "loss": 2.3597, + "step": 43541 + }, + { + "epoch": 9.5, + "learning_rate": 2.4885296045444615e-07, + "loss": 0.7593, + "step": 43542 + }, + { + "epoch": 9.5, + "learning_rate": 2.4874371859296486e-07, + "loss": 1.9203, + "step": 43543 + }, + { + "epoch": 9.5, + "learning_rate": 2.486344767314835e-07, + "loss": 0.3386, + "step": 43544 + }, + { + "epoch": 9.5, + "learning_rate": 2.485252348700022e-07, + "loss": 0.4859, + "step": 43545 + }, + { + "epoch": 9.5, + "learning_rate": 2.4841599300852085e-07, + "loss": 2.0813, + "step": 43546 + }, + { + "epoch": 9.5, + "learning_rate": 2.4830675114703957e-07, + "loss": 0.3739, + "step": 43547 + }, + { + "epoch": 9.5, + "learning_rate": 2.4819750928555823e-07, + "loss": 1.2374, + "step": 43548 + }, + { + "epoch": 9.5, + "learning_rate": 2.480882674240769e-07, + "loss": 0.2495, + "step": 43549 + }, + { + "epoch": 9.5, + "learning_rate": 2.479790255625956e-07, + "loss": 0.95, + "step": 43550 + }, + { + "epoch": 9.5, + "learning_rate": 2.478697837011143e-07, + "loss": 1.2503, + "step": 43551 + }, + { + "epoch": 9.51, + "learning_rate": 2.4776054183963294e-07, + "loss": 0.3054, + "step": 43552 + }, + { + "epoch": 9.51, + "learning_rate": 2.4765129997815166e-07, + "loss": 0.712, + "step": 43553 + }, + { + "epoch": 9.51, + "learning_rate": 2.475420581166703e-07, + "loss": 1.2514, + "step": 43554 + }, + { + "epoch": 9.51, + "learning_rate": 2.4743281625518904e-07, + "loss": 1.8692, + "step": 43555 + }, + { + "epoch": 9.51, + "learning_rate": 2.4732357439370765e-07, + "loss": 1.8064, + "step": 43556 + }, + { + "epoch": 9.51, + "learning_rate": 2.4721433253222637e-07, + "loss": 1.8573, + "step": 43557 + }, + { + "epoch": 9.51, + "learning_rate": 2.471050906707451e-07, + "loss": 1.5338, + "step": 43558 + }, + { + "epoch": 9.51, + "learning_rate": 2.469958488092637e-07, + "loss": 1.8889, + "step": 43559 + }, + { + "epoch": 9.51, + "learning_rate": 2.468866069477824e-07, + "loss": 1.5446, + "step": 43560 + }, + { + "epoch": 9.51, + "learning_rate": 2.467773650863011e-07, + "loss": 1.3814, + "step": 43561 + }, + { + "epoch": 9.51, + "learning_rate": 2.466681232248198e-07, + "loss": 2.6847, + "step": 43562 + }, + { + "epoch": 9.51, + "learning_rate": 2.4655888136333846e-07, + "loss": 3.6301, + "step": 43563 + }, + { + "epoch": 9.51, + "learning_rate": 2.464496395018571e-07, + "loss": 1.116, + "step": 43564 + }, + { + "epoch": 9.51, + "learning_rate": 2.4634039764037584e-07, + "loss": 1.1072, + "step": 43565 + }, + { + "epoch": 9.51, + "learning_rate": 2.4623115577889445e-07, + "loss": 0.4088, + "step": 43566 + }, + { + "epoch": 9.51, + "learning_rate": 2.4612191391741317e-07, + "loss": 0.23, + "step": 43567 + }, + { + "epoch": 9.51, + "learning_rate": 2.460126720559319e-07, + "loss": 1.0691, + "step": 43568 + }, + { + "epoch": 9.51, + "learning_rate": 2.4590343019445055e-07, + "loss": 0.6161, + "step": 43569 + }, + { + "epoch": 9.51, + "learning_rate": 2.457941883329692e-07, + "loss": 0.9314, + "step": 43570 + }, + { + "epoch": 9.51, + "learning_rate": 2.456849464714879e-07, + "loss": 1.5292, + "step": 43571 + }, + { + "epoch": 9.51, + "learning_rate": 2.455757046100066e-07, + "loss": 1.1409, + "step": 43572 + }, + { + "epoch": 9.51, + "learning_rate": 2.4546646274852526e-07, + "loss": 0.5768, + "step": 43573 + }, + { + "epoch": 9.51, + "learning_rate": 2.453572208870439e-07, + "loss": 1.2459, + "step": 43574 + }, + { + "epoch": 9.51, + "learning_rate": 2.4524797902556264e-07, + "loss": 0.2409, + "step": 43575 + }, + { + "epoch": 9.51, + "learning_rate": 2.451387371640813e-07, + "loss": 0.7199, + "step": 43576 + }, + { + "epoch": 9.51, + "learning_rate": 2.4502949530259997e-07, + "loss": 0.2169, + "step": 43577 + }, + { + "epoch": 9.51, + "learning_rate": 2.449202534411187e-07, + "loss": 1.5791, + "step": 43578 + }, + { + "epoch": 9.51, + "learning_rate": 2.4481101157963735e-07, + "loss": 0.2914, + "step": 43579 + }, + { + "epoch": 9.51, + "learning_rate": 2.44701769718156e-07, + "loss": 1.7365, + "step": 43580 + }, + { + "epoch": 9.51, + "learning_rate": 2.4459252785667473e-07, + "loss": 0.0764, + "step": 43581 + }, + { + "epoch": 9.51, + "learning_rate": 2.444832859951934e-07, + "loss": 0.974, + "step": 43582 + }, + { + "epoch": 9.51, + "learning_rate": 2.4437404413371205e-07, + "loss": 0.5381, + "step": 43583 + }, + { + "epoch": 9.51, + "learning_rate": 2.442648022722307e-07, + "loss": 0.401, + "step": 43584 + }, + { + "epoch": 9.51, + "learning_rate": 2.4415556041074944e-07, + "loss": 1.3547, + "step": 43585 + }, + { + "epoch": 9.51, + "learning_rate": 2.440463185492681e-07, + "loss": 2.2932, + "step": 43586 + }, + { + "epoch": 9.51, + "learning_rate": 2.4393707668778676e-07, + "loss": 1.8998, + "step": 43587 + }, + { + "epoch": 9.51, + "learning_rate": 2.438278348263055e-07, + "loss": 1.4616, + "step": 43588 + }, + { + "epoch": 9.51, + "learning_rate": 2.4371859296482414e-07, + "loss": 0.6071, + "step": 43589 + }, + { + "epoch": 9.51, + "learning_rate": 2.436093511033428e-07, + "loss": 1.222, + "step": 43590 + }, + { + "epoch": 9.51, + "learning_rate": 2.435001092418615e-07, + "loss": 2.1687, + "step": 43591 + }, + { + "epoch": 9.51, + "learning_rate": 2.433908673803802e-07, + "loss": 2.2402, + "step": 43592 + }, + { + "epoch": 9.51, + "learning_rate": 2.4328162551889885e-07, + "loss": 1.4696, + "step": 43593 + }, + { + "epoch": 9.51, + "learning_rate": 2.431723836574175e-07, + "loss": 0.8707, + "step": 43594 + }, + { + "epoch": 9.51, + "learning_rate": 2.4306314179593623e-07, + "loss": 1.4806, + "step": 43595 + }, + { + "epoch": 9.51, + "learning_rate": 2.429538999344549e-07, + "loss": 0.8633, + "step": 43596 + }, + { + "epoch": 9.51, + "learning_rate": 2.4284465807297356e-07, + "loss": 0.4262, + "step": 43597 + }, + { + "epoch": 9.52, + "learning_rate": 2.427354162114923e-07, + "loss": 1.3704, + "step": 43598 + }, + { + "epoch": 9.52, + "learning_rate": 2.4262617435001094e-07, + "loss": 0.4839, + "step": 43599 + }, + { + "epoch": 9.52, + "learning_rate": 2.425169324885296e-07, + "loss": 0.7782, + "step": 43600 + }, + { + "epoch": 9.52, + "learning_rate": 2.424076906270483e-07, + "loss": 1.3091, + "step": 43601 + }, + { + "epoch": 9.52, + "learning_rate": 2.42298448765567e-07, + "loss": 0.7969, + "step": 43602 + }, + { + "epoch": 9.52, + "learning_rate": 2.4218920690408565e-07, + "loss": 0.4116, + "step": 43603 + }, + { + "epoch": 9.52, + "learning_rate": 2.420799650426043e-07, + "loss": 0.3257, + "step": 43604 + }, + { + "epoch": 9.52, + "learning_rate": 2.4197072318112303e-07, + "loss": 1.7582, + "step": 43605 + }, + { + "epoch": 9.52, + "learning_rate": 2.418614813196417e-07, + "loss": 2.6173, + "step": 43606 + }, + { + "epoch": 9.52, + "learning_rate": 2.4175223945816036e-07, + "loss": 0.9968, + "step": 43607 + }, + { + "epoch": 9.52, + "learning_rate": 2.416429975966791e-07, + "loss": 1.6626, + "step": 43608 + }, + { + "epoch": 9.52, + "learning_rate": 2.4153375573519774e-07, + "loss": 0.401, + "step": 43609 + }, + { + "epoch": 9.52, + "learning_rate": 2.414245138737164e-07, + "loss": 2.8192, + "step": 43610 + }, + { + "epoch": 9.52, + "learning_rate": 2.413152720122351e-07, + "loss": 1.1188, + "step": 43611 + }, + { + "epoch": 9.52, + "learning_rate": 2.412060301507538e-07, + "loss": 0.8101, + "step": 43612 + }, + { + "epoch": 9.52, + "learning_rate": 2.4109678828927245e-07, + "loss": 1.598, + "step": 43613 + }, + { + "epoch": 9.52, + "learning_rate": 2.4098754642779117e-07, + "loss": 2.0789, + "step": 43614 + }, + { + "epoch": 9.52, + "learning_rate": 2.4087830456630983e-07, + "loss": 0.0633, + "step": 43615 + }, + { + "epoch": 9.52, + "learning_rate": 2.407690627048285e-07, + "loss": 1.3851, + "step": 43616 + }, + { + "epoch": 9.52, + "learning_rate": 2.4065982084334716e-07, + "loss": 0.0745, + "step": 43617 + }, + { + "epoch": 9.52, + "learning_rate": 2.405505789818659e-07, + "loss": 2.009, + "step": 43618 + }, + { + "epoch": 9.52, + "learning_rate": 2.404413371203846e-07, + "loss": 1.5312, + "step": 43619 + }, + { + "epoch": 9.52, + "learning_rate": 2.403320952589032e-07, + "loss": 0.7834, + "step": 43620 + }, + { + "epoch": 9.52, + "learning_rate": 2.402228533974219e-07, + "loss": 0.5205, + "step": 43621 + }, + { + "epoch": 9.52, + "learning_rate": 2.401136115359406e-07, + "loss": 1.1483, + "step": 43622 + }, + { + "epoch": 9.52, + "learning_rate": 2.400043696744593e-07, + "loss": 1.0627, + "step": 43623 + }, + { + "epoch": 9.52, + "learning_rate": 2.3989512781297796e-07, + "loss": 1.6941, + "step": 43624 + }, + { + "epoch": 9.52, + "learning_rate": 2.3978588595149663e-07, + "loss": 1.5517, + "step": 43625 + }, + { + "epoch": 9.52, + "learning_rate": 2.3967664409001534e-07, + "loss": 0.6955, + "step": 43626 + }, + { + "epoch": 9.52, + "learning_rate": 2.3956740222853396e-07, + "loss": 0.6529, + "step": 43627 + }, + { + "epoch": 9.52, + "learning_rate": 2.3945816036705267e-07, + "loss": 1.0706, + "step": 43628 + }, + { + "epoch": 9.52, + "learning_rate": 2.393489185055714e-07, + "loss": 0.7293, + "step": 43629 + }, + { + "epoch": 9.52, + "learning_rate": 2.3923967664409005e-07, + "loss": 1.4434, + "step": 43630 + }, + { + "epoch": 9.52, + "learning_rate": 2.391304347826087e-07, + "loss": 0.8903, + "step": 43631 + }, + { + "epoch": 9.52, + "learning_rate": 2.390211929211274e-07, + "loss": 2.5087, + "step": 43632 + }, + { + "epoch": 9.52, + "learning_rate": 2.389119510596461e-07, + "loss": 1.464, + "step": 43633 + }, + { + "epoch": 9.52, + "learning_rate": 2.3880270919816476e-07, + "loss": 1.7454, + "step": 43634 + }, + { + "epoch": 9.52, + "learning_rate": 2.386934673366834e-07, + "loss": 0.6107, + "step": 43635 + }, + { + "epoch": 9.52, + "learning_rate": 2.3858422547520214e-07, + "loss": 2.1112, + "step": 43636 + }, + { + "epoch": 9.52, + "learning_rate": 2.384749836137208e-07, + "loss": 1.4261, + "step": 43637 + }, + { + "epoch": 9.52, + "learning_rate": 2.3836574175223947e-07, + "loss": 1.5981, + "step": 43638 + }, + { + "epoch": 9.52, + "learning_rate": 2.3825649989075816e-07, + "loss": 0.5612, + "step": 43639 + }, + { + "epoch": 9.52, + "learning_rate": 2.3814725802927682e-07, + "loss": 0.5444, + "step": 43640 + }, + { + "epoch": 9.52, + "learning_rate": 2.3803801616779552e-07, + "loss": 0.3624, + "step": 43641 + }, + { + "epoch": 9.52, + "learning_rate": 2.3792877430631418e-07, + "loss": 0.0751, + "step": 43642 + }, + { + "epoch": 9.52, + "learning_rate": 2.3781953244483287e-07, + "loss": 1.2629, + "step": 43643 + }, + { + "epoch": 9.53, + "learning_rate": 2.3771029058335156e-07, + "loss": 2.2252, + "step": 43644 + }, + { + "epoch": 9.53, + "learning_rate": 2.3760104872187022e-07, + "loss": 0.4846, + "step": 43645 + }, + { + "epoch": 9.53, + "learning_rate": 2.3749180686038891e-07, + "loss": 0.6392, + "step": 43646 + }, + { + "epoch": 9.53, + "learning_rate": 2.3738256499890758e-07, + "loss": 1.7107, + "step": 43647 + }, + { + "epoch": 9.53, + "learning_rate": 2.3727332313742627e-07, + "loss": 0.411, + "step": 43648 + }, + { + "epoch": 9.53, + "learning_rate": 2.3716408127594499e-07, + "loss": 1.1625, + "step": 43649 + }, + { + "epoch": 9.53, + "learning_rate": 2.3705483941446362e-07, + "loss": 2.5846, + "step": 43650 + }, + { + "epoch": 9.53, + "learning_rate": 2.3694559755298231e-07, + "loss": 1.2924, + "step": 43651 + }, + { + "epoch": 9.53, + "learning_rate": 2.3683635569150098e-07, + "loss": 0.9963, + "step": 43652 + }, + { + "epoch": 9.53, + "learning_rate": 2.3672711383001967e-07, + "loss": 1.4647, + "step": 43653 + }, + { + "epoch": 9.53, + "learning_rate": 2.3661787196853838e-07, + "loss": 1.2437, + "step": 43654 + }, + { + "epoch": 9.53, + "learning_rate": 2.3650863010705702e-07, + "loss": 1.2965, + "step": 43655 + }, + { + "epoch": 9.53, + "learning_rate": 2.3639938824557574e-07, + "loss": 0.2208, + "step": 43656 + }, + { + "epoch": 9.53, + "learning_rate": 2.3629014638409443e-07, + "loss": 0.4336, + "step": 43657 + }, + { + "epoch": 9.53, + "learning_rate": 2.361809045226131e-07, + "loss": 0.1676, + "step": 43658 + }, + { + "epoch": 9.53, + "learning_rate": 2.3607166266113178e-07, + "loss": 0.999, + "step": 43659 + }, + { + "epoch": 9.53, + "learning_rate": 2.3596242079965042e-07, + "loss": 1.1131, + "step": 43660 + }, + { + "epoch": 9.53, + "learning_rate": 2.3585317893816914e-07, + "loss": 0.6503, + "step": 43661 + }, + { + "epoch": 9.53, + "learning_rate": 2.3574393707668783e-07, + "loss": 0.5455, + "step": 43662 + }, + { + "epoch": 9.53, + "learning_rate": 2.356346952152065e-07, + "loss": 0.8239, + "step": 43663 + }, + { + "epoch": 9.53, + "learning_rate": 2.3552545335372518e-07, + "loss": 0.7412, + "step": 43664 + }, + { + "epoch": 9.53, + "learning_rate": 2.3541621149224385e-07, + "loss": 1.1194, + "step": 43665 + }, + { + "epoch": 9.53, + "learning_rate": 2.3530696963076254e-07, + "loss": 1.1136, + "step": 43666 + }, + { + "epoch": 9.53, + "learning_rate": 2.3519772776928123e-07, + "loss": 0.788, + "step": 43667 + }, + { + "epoch": 9.53, + "learning_rate": 2.350884859077999e-07, + "loss": 1.6069, + "step": 43668 + }, + { + "epoch": 9.53, + "learning_rate": 2.3497924404631858e-07, + "loss": 1.7452, + "step": 43669 + }, + { + "epoch": 9.53, + "learning_rate": 2.3487000218483725e-07, + "loss": 0.1126, + "step": 43670 + }, + { + "epoch": 9.53, + "learning_rate": 2.3476076032335594e-07, + "loss": 0.9408, + "step": 43671 + }, + { + "epoch": 9.53, + "learning_rate": 2.3465151846187463e-07, + "loss": 3.0347, + "step": 43672 + }, + { + "epoch": 9.53, + "learning_rate": 2.345422766003933e-07, + "loss": 0.752, + "step": 43673 + }, + { + "epoch": 9.53, + "learning_rate": 2.3443303473891198e-07, + "loss": 0.5193, + "step": 43674 + }, + { + "epoch": 9.53, + "learning_rate": 2.3432379287743064e-07, + "loss": 2.4916, + "step": 43675 + }, + { + "epoch": 9.53, + "learning_rate": 2.3421455101594933e-07, + "loss": 0.5569, + "step": 43676 + }, + { + "epoch": 9.53, + "learning_rate": 2.3410530915446803e-07, + "loss": 2.4498, + "step": 43677 + }, + { + "epoch": 9.53, + "learning_rate": 2.339960672929867e-07, + "loss": 0.9148, + "step": 43678 + }, + { + "epoch": 9.53, + "learning_rate": 2.3388682543150538e-07, + "loss": 1.3535, + "step": 43679 + }, + { + "epoch": 9.53, + "learning_rate": 2.3377758357002404e-07, + "loss": 0.6773, + "step": 43680 + }, + { + "epoch": 9.53, + "learning_rate": 2.3366834170854273e-07, + "loss": 1.2372, + "step": 43681 + }, + { + "epoch": 9.53, + "learning_rate": 2.3355909984706142e-07, + "loss": 0.6779, + "step": 43682 + }, + { + "epoch": 9.53, + "learning_rate": 2.334498579855801e-07, + "loss": 2.5887, + "step": 43683 + }, + { + "epoch": 9.53, + "learning_rate": 2.3334061612409878e-07, + "loss": 1.2017, + "step": 43684 + }, + { + "epoch": 9.53, + "learning_rate": 2.3323137426261744e-07, + "loss": 2.3676, + "step": 43685 + }, + { + "epoch": 9.53, + "learning_rate": 2.3312213240113613e-07, + "loss": 0.5138, + "step": 43686 + }, + { + "epoch": 9.53, + "learning_rate": 2.3301289053965482e-07, + "loss": 2.561, + "step": 43687 + }, + { + "epoch": 9.53, + "learning_rate": 2.329036486781735e-07, + "loss": 0.762, + "step": 43688 + }, + { + "epoch": 9.53, + "learning_rate": 2.3279440681669218e-07, + "loss": 0.6889, + "step": 43689 + }, + { + "epoch": 9.54, + "learning_rate": 2.3268516495521084e-07, + "loss": 2.4565, + "step": 43690 + }, + { + "epoch": 9.54, + "learning_rate": 2.3257592309372953e-07, + "loss": 1.9051, + "step": 43691 + }, + { + "epoch": 9.54, + "learning_rate": 2.3246668123224822e-07, + "loss": 0.6562, + "step": 43692 + }, + { + "epoch": 9.54, + "learning_rate": 2.3235743937076689e-07, + "loss": 1.5732, + "step": 43693 + }, + { + "epoch": 9.54, + "learning_rate": 2.3224819750928558e-07, + "loss": 0.6296, + "step": 43694 + }, + { + "epoch": 9.54, + "learning_rate": 2.3213895564780427e-07, + "loss": 1.4828, + "step": 43695 + }, + { + "epoch": 9.54, + "learning_rate": 2.3202971378632293e-07, + "loss": 0.5158, + "step": 43696 + }, + { + "epoch": 9.54, + "learning_rate": 2.3192047192484162e-07, + "loss": 1.5207, + "step": 43697 + }, + { + "epoch": 9.54, + "learning_rate": 2.3181123006336029e-07, + "loss": 1.0867, + "step": 43698 + }, + { + "epoch": 9.54, + "learning_rate": 2.3170198820187898e-07, + "loss": 0.7258, + "step": 43699 + }, + { + "epoch": 9.54, + "learning_rate": 2.3159274634039767e-07, + "loss": 0.6623, + "step": 43700 + }, + { + "epoch": 9.54, + "learning_rate": 2.3148350447891633e-07, + "loss": 1.6764, + "step": 43701 + }, + { + "epoch": 9.54, + "learning_rate": 2.3137426261743502e-07, + "loss": 0.5336, + "step": 43702 + }, + { + "epoch": 9.54, + "learning_rate": 2.3126502075595368e-07, + "loss": 0.4157, + "step": 43703 + }, + { + "epoch": 9.54, + "learning_rate": 2.3115577889447237e-07, + "loss": 0.7951, + "step": 43704 + }, + { + "epoch": 9.54, + "learning_rate": 2.3104653703299107e-07, + "loss": 1.2069, + "step": 43705 + }, + { + "epoch": 9.54, + "learning_rate": 2.3093729517150973e-07, + "loss": 0.7757, + "step": 43706 + }, + { + "epoch": 9.54, + "learning_rate": 2.3082805331002842e-07, + "loss": 1.0555, + "step": 43707 + }, + { + "epoch": 9.54, + "learning_rate": 2.3071881144854708e-07, + "loss": 1.1295, + "step": 43708 + }, + { + "epoch": 9.54, + "learning_rate": 2.3060956958706577e-07, + "loss": 0.6971, + "step": 43709 + }, + { + "epoch": 9.54, + "learning_rate": 2.3050032772558446e-07, + "loss": 1.8089, + "step": 43710 + }, + { + "epoch": 9.54, + "learning_rate": 2.3039108586410313e-07, + "loss": 0.7505, + "step": 43711 + }, + { + "epoch": 9.54, + "learning_rate": 2.3028184400262182e-07, + "loss": 0.6734, + "step": 43712 + }, + { + "epoch": 9.54, + "learning_rate": 2.3017260214114048e-07, + "loss": 0.3973, + "step": 43713 + }, + { + "epoch": 9.54, + "learning_rate": 2.3006336027965917e-07, + "loss": 1.5864, + "step": 43714 + }, + { + "epoch": 9.54, + "learning_rate": 2.299541184181779e-07, + "loss": 1.5059, + "step": 43715 + }, + { + "epoch": 9.54, + "learning_rate": 2.2984487655669653e-07, + "loss": 0.3789, + "step": 43716 + }, + { + "epoch": 9.54, + "learning_rate": 2.2973563469521524e-07, + "loss": 1.2053, + "step": 43717 + }, + { + "epoch": 9.54, + "learning_rate": 2.2962639283373388e-07, + "loss": 1.3033, + "step": 43718 + }, + { + "epoch": 9.54, + "learning_rate": 2.2951715097225257e-07, + "loss": 0.6707, + "step": 43719 + }, + { + "epoch": 9.54, + "learning_rate": 2.294079091107713e-07, + "loss": 1.697, + "step": 43720 + }, + { + "epoch": 9.54, + "learning_rate": 2.2929866724928993e-07, + "loss": 1.2322, + "step": 43721 + }, + { + "epoch": 9.54, + "learning_rate": 2.2918942538780864e-07, + "loss": 0.2581, + "step": 43722 + }, + { + "epoch": 9.54, + "learning_rate": 2.2908018352632728e-07, + "loss": 0.8976, + "step": 43723 + }, + { + "epoch": 9.54, + "learning_rate": 2.28970941664846e-07, + "loss": 4.1353, + "step": 43724 + }, + { + "epoch": 9.54, + "learning_rate": 2.288616998033647e-07, + "loss": 2.2756, + "step": 43725 + }, + { + "epoch": 9.54, + "learning_rate": 2.2875245794188335e-07, + "loss": 1.5531, + "step": 43726 + }, + { + "epoch": 9.54, + "learning_rate": 2.2864321608040204e-07, + "loss": 1.8127, + "step": 43727 + }, + { + "epoch": 9.54, + "learning_rate": 2.2853397421892068e-07, + "loss": 0.2325, + "step": 43728 + }, + { + "epoch": 9.54, + "learning_rate": 2.284247323574394e-07, + "loss": 0.0272, + "step": 43729 + }, + { + "epoch": 9.54, + "learning_rate": 2.283154904959581e-07, + "loss": 0.1249, + "step": 43730 + }, + { + "epoch": 9.54, + "learning_rate": 2.2820624863447675e-07, + "loss": 0.6695, + "step": 43731 + }, + { + "epoch": 9.54, + "learning_rate": 2.2809700677299544e-07, + "loss": 0.6222, + "step": 43732 + }, + { + "epoch": 9.54, + "learning_rate": 2.279877649115141e-07, + "loss": 0.5425, + "step": 43733 + }, + { + "epoch": 9.54, + "learning_rate": 2.278785230500328e-07, + "loss": 2.2424, + "step": 43734 + }, + { + "epoch": 9.54, + "learning_rate": 2.2776928118855149e-07, + "loss": 0.7674, + "step": 43735 + }, + { + "epoch": 9.55, + "learning_rate": 2.2766003932707015e-07, + "loss": 0.0595, + "step": 43736 + }, + { + "epoch": 9.55, + "learning_rate": 2.2755079746558884e-07, + "loss": 1.1211, + "step": 43737 + }, + { + "epoch": 9.55, + "learning_rate": 2.2744155560410753e-07, + "loss": 0.2188, + "step": 43738 + }, + { + "epoch": 9.55, + "learning_rate": 2.273323137426262e-07, + "loss": 1.8894, + "step": 43739 + }, + { + "epoch": 9.55, + "learning_rate": 2.2722307188114489e-07, + "loss": 0.3168, + "step": 43740 + }, + { + "epoch": 9.55, + "learning_rate": 2.2711383001966355e-07, + "loss": 0.7296, + "step": 43741 + }, + { + "epoch": 9.55, + "learning_rate": 2.2700458815818224e-07, + "loss": 1.4896, + "step": 43742 + }, + { + "epoch": 9.55, + "learning_rate": 2.2689534629670093e-07, + "loss": 0.3658, + "step": 43743 + }, + { + "epoch": 9.55, + "learning_rate": 2.267861044352196e-07, + "loss": 0.4335, + "step": 43744 + }, + { + "epoch": 9.55, + "learning_rate": 2.2667686257373828e-07, + "loss": 0.7495, + "step": 43745 + }, + { + "epoch": 9.55, + "learning_rate": 2.2656762071225695e-07, + "loss": 0.8674, + "step": 43746 + }, + { + "epoch": 9.55, + "learning_rate": 2.2645837885077564e-07, + "loss": 0.5116, + "step": 43747 + }, + { + "epoch": 9.55, + "learning_rate": 2.2634913698929433e-07, + "loss": 1.3069, + "step": 43748 + }, + { + "epoch": 9.55, + "learning_rate": 2.26239895127813e-07, + "loss": 1.2791, + "step": 43749 + }, + { + "epoch": 9.55, + "learning_rate": 2.2613065326633168e-07, + "loss": 0.7406, + "step": 43750 + }, + { + "epoch": 9.55, + "learning_rate": 2.2602141140485035e-07, + "loss": 1.6394, + "step": 43751 + }, + { + "epoch": 9.55, + "learning_rate": 2.2591216954336904e-07, + "loss": 2.1192, + "step": 43752 + }, + { + "epoch": 9.55, + "learning_rate": 2.2580292768188773e-07, + "loss": 0.1498, + "step": 43753 + }, + { + "epoch": 9.55, + "learning_rate": 2.256936858204064e-07, + "loss": 1.6963, + "step": 43754 + }, + { + "epoch": 9.55, + "learning_rate": 2.2558444395892508e-07, + "loss": 1.0039, + "step": 43755 + }, + { + "epoch": 9.55, + "learning_rate": 2.2547520209744375e-07, + "loss": 0.8879, + "step": 43756 + }, + { + "epoch": 9.55, + "learning_rate": 2.2536596023596244e-07, + "loss": 1.798, + "step": 43757 + }, + { + "epoch": 9.55, + "learning_rate": 2.2525671837448113e-07, + "loss": 0.588, + "step": 43758 + }, + { + "epoch": 9.55, + "learning_rate": 2.251474765129998e-07, + "loss": 2.6982, + "step": 43759 + }, + { + "epoch": 9.55, + "learning_rate": 2.2503823465151848e-07, + "loss": 0.6012, + "step": 43760 + }, + { + "epoch": 9.55, + "learning_rate": 2.2492899279003715e-07, + "loss": 0.2537, + "step": 43761 + }, + { + "epoch": 9.55, + "learning_rate": 2.2481975092855584e-07, + "loss": 0.2905, + "step": 43762 + }, + { + "epoch": 9.55, + "learning_rate": 2.2471050906707453e-07, + "loss": 1.6119, + "step": 43763 + }, + { + "epoch": 9.55, + "learning_rate": 2.246012672055932e-07, + "loss": 0.908, + "step": 43764 + }, + { + "epoch": 9.55, + "learning_rate": 2.2449202534411188e-07, + "loss": 0.4676, + "step": 43765 + }, + { + "epoch": 9.55, + "learning_rate": 2.2438278348263054e-07, + "loss": 0.6397, + "step": 43766 + }, + { + "epoch": 9.55, + "learning_rate": 2.2427354162114923e-07, + "loss": 0.4585, + "step": 43767 + }, + { + "epoch": 9.55, + "learning_rate": 2.2416429975966793e-07, + "loss": 0.574, + "step": 43768 + }, + { + "epoch": 9.55, + "learning_rate": 2.240550578981866e-07, + "loss": 3.1516, + "step": 43769 + }, + { + "epoch": 9.55, + "learning_rate": 2.2394581603670528e-07, + "loss": 0.5511, + "step": 43770 + }, + { + "epoch": 9.55, + "learning_rate": 2.2383657417522394e-07, + "loss": 0.6604, + "step": 43771 + }, + { + "epoch": 9.55, + "learning_rate": 2.2372733231374263e-07, + "loss": 0.1195, + "step": 43772 + }, + { + "epoch": 9.55, + "learning_rate": 2.2361809045226132e-07, + "loss": 1.3329, + "step": 43773 + }, + { + "epoch": 9.55, + "learning_rate": 2.2350884859078e-07, + "loss": 0.6884, + "step": 43774 + }, + { + "epoch": 9.55, + "learning_rate": 2.2339960672929868e-07, + "loss": 0.8075, + "step": 43775 + }, + { + "epoch": 9.55, + "learning_rate": 2.232903648678174e-07, + "loss": 2.847, + "step": 43776 + }, + { + "epoch": 9.55, + "learning_rate": 2.2318112300633603e-07, + "loss": 0.7395, + "step": 43777 + }, + { + "epoch": 9.55, + "learning_rate": 2.2307188114485475e-07, + "loss": 0.1397, + "step": 43778 + }, + { + "epoch": 9.55, + "learning_rate": 2.229626392833734e-07, + "loss": 2.2035, + "step": 43779 + }, + { + "epoch": 9.55, + "learning_rate": 2.2285339742189208e-07, + "loss": 1.2195, + "step": 43780 + }, + { + "epoch": 9.55, + "learning_rate": 2.227441555604108e-07, + "loss": 0.6452, + "step": 43781 + }, + { + "epoch": 9.56, + "learning_rate": 2.2263491369892943e-07, + "loss": 0.491, + "step": 43782 + }, + { + "epoch": 9.56, + "learning_rate": 2.2252567183744815e-07, + "loss": 0.8613, + "step": 43783 + }, + { + "epoch": 9.56, + "learning_rate": 2.2241642997596679e-07, + "loss": 0.6159, + "step": 43784 + }, + { + "epoch": 9.56, + "learning_rate": 2.223071881144855e-07, + "loss": 0.9682, + "step": 43785 + }, + { + "epoch": 9.56, + "learning_rate": 2.221979462530042e-07, + "loss": 0.4116, + "step": 43786 + }, + { + "epoch": 9.56, + "learning_rate": 2.2208870439152286e-07, + "loss": 0.2261, + "step": 43787 + }, + { + "epoch": 9.56, + "learning_rate": 2.2197946253004155e-07, + "loss": 0.8983, + "step": 43788 + }, + { + "epoch": 9.56, + "learning_rate": 2.2187022066856019e-07, + "loss": 0.7715, + "step": 43789 + }, + { + "epoch": 9.56, + "learning_rate": 2.217609788070789e-07, + "loss": 0.8765, + "step": 43790 + }, + { + "epoch": 9.56, + "learning_rate": 2.216517369455976e-07, + "loss": 2.8661, + "step": 43791 + }, + { + "epoch": 9.56, + "learning_rate": 2.2154249508411626e-07, + "loss": 0.595, + "step": 43792 + }, + { + "epoch": 9.56, + "learning_rate": 2.2143325322263495e-07, + "loss": 2.3298, + "step": 43793 + }, + { + "epoch": 9.56, + "learning_rate": 2.213240113611536e-07, + "loss": 0.5825, + "step": 43794 + }, + { + "epoch": 9.56, + "learning_rate": 2.212147694996723e-07, + "loss": 1.3992, + "step": 43795 + }, + { + "epoch": 9.56, + "learning_rate": 2.21105527638191e-07, + "loss": 0.2883, + "step": 43796 + }, + { + "epoch": 9.56, + "learning_rate": 2.2099628577670966e-07, + "loss": 1.1497, + "step": 43797 + }, + { + "epoch": 9.56, + "learning_rate": 2.2088704391522835e-07, + "loss": 1.1145, + "step": 43798 + }, + { + "epoch": 9.56, + "learning_rate": 2.20777802053747e-07, + "loss": 0.695, + "step": 43799 + }, + { + "epoch": 9.56, + "learning_rate": 2.206685601922657e-07, + "loss": 1.5476, + "step": 43800 + }, + { + "epoch": 9.56, + "learning_rate": 2.205593183307844e-07, + "loss": 0.574, + "step": 43801 + }, + { + "epoch": 9.56, + "learning_rate": 2.2045007646930305e-07, + "loss": 0.2616, + "step": 43802 + }, + { + "epoch": 9.56, + "learning_rate": 2.2034083460782174e-07, + "loss": 0.1496, + "step": 43803 + }, + { + "epoch": 9.56, + "learning_rate": 2.202315927463404e-07, + "loss": 0.6617, + "step": 43804 + }, + { + "epoch": 9.56, + "learning_rate": 2.201223508848591e-07, + "loss": 0.9567, + "step": 43805 + }, + { + "epoch": 9.56, + "learning_rate": 2.200131090233778e-07, + "loss": 0.6285, + "step": 43806 + }, + { + "epoch": 9.56, + "learning_rate": 2.1990386716189645e-07, + "loss": 1.9584, + "step": 43807 + }, + { + "epoch": 9.56, + "learning_rate": 2.1979462530041514e-07, + "loss": 0.7984, + "step": 43808 + }, + { + "epoch": 9.56, + "learning_rate": 2.196853834389338e-07, + "loss": 0.2572, + "step": 43809 + }, + { + "epoch": 9.56, + "learning_rate": 2.195761415774525e-07, + "loss": 1.1972, + "step": 43810 + }, + { + "epoch": 9.56, + "learning_rate": 2.194668997159712e-07, + "loss": 1.5823, + "step": 43811 + }, + { + "epoch": 9.56, + "learning_rate": 2.1935765785448985e-07, + "loss": 0.0354, + "step": 43812 + }, + { + "epoch": 9.56, + "learning_rate": 2.1924841599300854e-07, + "loss": 0.6325, + "step": 43813 + }, + { + "epoch": 9.56, + "learning_rate": 2.191391741315272e-07, + "loss": 2.7209, + "step": 43814 + }, + { + "epoch": 9.56, + "learning_rate": 2.190299322700459e-07, + "loss": 0.49, + "step": 43815 + }, + { + "epoch": 9.56, + "learning_rate": 2.189206904085646e-07, + "loss": 0.62, + "step": 43816 + }, + { + "epoch": 9.56, + "learning_rate": 2.1881144854708325e-07, + "loss": 0.3004, + "step": 43817 + }, + { + "epoch": 9.56, + "learning_rate": 2.1870220668560194e-07, + "loss": 0.6713, + "step": 43818 + }, + { + "epoch": 9.56, + "learning_rate": 2.1859296482412063e-07, + "loss": 0.3854, + "step": 43819 + }, + { + "epoch": 9.56, + "learning_rate": 2.184837229626393e-07, + "loss": 0.7802, + "step": 43820 + }, + { + "epoch": 9.56, + "learning_rate": 2.1837448110115799e-07, + "loss": 0.1825, + "step": 43821 + }, + { + "epoch": 9.56, + "learning_rate": 2.1826523923967665e-07, + "loss": 1.1082, + "step": 43822 + }, + { + "epoch": 9.56, + "learning_rate": 2.1815599737819534e-07, + "loss": 1.6709, + "step": 43823 + }, + { + "epoch": 9.56, + "learning_rate": 2.1804675551671403e-07, + "loss": 1.1046, + "step": 43824 + }, + { + "epoch": 9.56, + "learning_rate": 2.179375136552327e-07, + "loss": 0.4987, + "step": 43825 + }, + { + "epoch": 9.56, + "learning_rate": 2.1782827179375139e-07, + "loss": 0.4761, + "step": 43826 + }, + { + "epoch": 9.57, + "learning_rate": 2.1771902993227005e-07, + "loss": 1.1107, + "step": 43827 + }, + { + "epoch": 9.57, + "learning_rate": 2.1760978807078874e-07, + "loss": 0.6499, + "step": 43828 + }, + { + "epoch": 9.57, + "learning_rate": 2.1750054620930743e-07, + "loss": 0.2005, + "step": 43829 + }, + { + "epoch": 9.57, + "learning_rate": 2.173913043478261e-07, + "loss": 0.2963, + "step": 43830 + }, + { + "epoch": 9.57, + "learning_rate": 2.1728206248634478e-07, + "loss": 0.1475, + "step": 43831 + }, + { + "epoch": 9.57, + "learning_rate": 2.1717282062486345e-07, + "loss": 0.1862, + "step": 43832 + }, + { + "epoch": 9.57, + "learning_rate": 2.1706357876338214e-07, + "loss": 1.7791, + "step": 43833 + }, + { + "epoch": 9.57, + "learning_rate": 2.1695433690190083e-07, + "loss": 2.938, + "step": 43834 + }, + { + "epoch": 9.57, + "learning_rate": 2.168450950404195e-07, + "loss": 0.647, + "step": 43835 + }, + { + "epoch": 9.57, + "learning_rate": 2.1673585317893818e-07, + "loss": 2.4297, + "step": 43836 + }, + { + "epoch": 9.57, + "learning_rate": 2.1662661131745685e-07, + "loss": 0.8271, + "step": 43837 + }, + { + "epoch": 9.57, + "learning_rate": 2.1651736945597554e-07, + "loss": 0.2076, + "step": 43838 + }, + { + "epoch": 9.57, + "learning_rate": 2.1640812759449423e-07, + "loss": 0.5565, + "step": 43839 + }, + { + "epoch": 9.57, + "learning_rate": 2.162988857330129e-07, + "loss": 0.7878, + "step": 43840 + }, + { + "epoch": 9.57, + "learning_rate": 2.1618964387153158e-07, + "loss": 0.8603, + "step": 43841 + }, + { + "epoch": 9.57, + "learning_rate": 2.1608040201005025e-07, + "loss": 2.2249, + "step": 43842 + }, + { + "epoch": 9.57, + "learning_rate": 2.1597116014856894e-07, + "loss": 0.9599, + "step": 43843 + }, + { + "epoch": 9.57, + "learning_rate": 2.1586191828708765e-07, + "loss": 0.5126, + "step": 43844 + }, + { + "epoch": 9.57, + "learning_rate": 2.157526764256063e-07, + "loss": 0.8771, + "step": 43845 + }, + { + "epoch": 9.57, + "learning_rate": 2.15643434564125e-07, + "loss": 0.5854, + "step": 43846 + }, + { + "epoch": 9.57, + "learning_rate": 2.1553419270264365e-07, + "loss": 1.098, + "step": 43847 + }, + { + "epoch": 9.57, + "learning_rate": 2.1542495084116234e-07, + "loss": 0.6322, + "step": 43848 + }, + { + "epoch": 9.57, + "learning_rate": 2.1531570897968105e-07, + "loss": 1.4967, + "step": 43849 + }, + { + "epoch": 9.57, + "learning_rate": 2.152064671181997e-07, + "loss": 1.2336, + "step": 43850 + }, + { + "epoch": 9.57, + "learning_rate": 2.150972252567184e-07, + "loss": 2.139, + "step": 43851 + }, + { + "epoch": 9.57, + "learning_rate": 2.1498798339523704e-07, + "loss": 2.5552, + "step": 43852 + }, + { + "epoch": 9.57, + "learning_rate": 2.1487874153375576e-07, + "loss": 2.9387, + "step": 43853 + }, + { + "epoch": 9.57, + "learning_rate": 2.1476949967227445e-07, + "loss": 0.7299, + "step": 43854 + }, + { + "epoch": 9.57, + "learning_rate": 2.1466025781079312e-07, + "loss": 1.1253, + "step": 43855 + }, + { + "epoch": 9.57, + "learning_rate": 2.145510159493118e-07, + "loss": 1.6183, + "step": 43856 + }, + { + "epoch": 9.57, + "learning_rate": 2.144417740878305e-07, + "loss": 0.2681, + "step": 43857 + }, + { + "epoch": 9.57, + "learning_rate": 2.1433253222634916e-07, + "loss": 0.526, + "step": 43858 + }, + { + "epoch": 9.57, + "learning_rate": 2.1422329036486785e-07, + "loss": 0.4452, + "step": 43859 + }, + { + "epoch": 9.57, + "learning_rate": 2.1411404850338652e-07, + "loss": 0.807, + "step": 43860 + }, + { + "epoch": 9.57, + "learning_rate": 2.140048066419052e-07, + "loss": 0.656, + "step": 43861 + }, + { + "epoch": 9.57, + "learning_rate": 2.138955647804239e-07, + "loss": 0.5506, + "step": 43862 + }, + { + "epoch": 9.57, + "learning_rate": 2.1378632291894256e-07, + "loss": 1.5172, + "step": 43863 + }, + { + "epoch": 9.57, + "learning_rate": 2.1367708105746125e-07, + "loss": 0.5798, + "step": 43864 + }, + { + "epoch": 9.57, + "learning_rate": 2.1356783919597991e-07, + "loss": 0.0408, + "step": 43865 + }, + { + "epoch": 9.57, + "learning_rate": 2.134585973344986e-07, + "loss": 0.8683, + "step": 43866 + }, + { + "epoch": 9.57, + "learning_rate": 2.133493554730173e-07, + "loss": 1.0942, + "step": 43867 + }, + { + "epoch": 9.57, + "learning_rate": 2.1324011361153596e-07, + "loss": 0.404, + "step": 43868 + }, + { + "epoch": 9.57, + "learning_rate": 2.1313087175005465e-07, + "loss": 0.453, + "step": 43869 + }, + { + "epoch": 9.57, + "learning_rate": 2.130216298885733e-07, + "loss": 1.4336, + "step": 43870 + }, + { + "epoch": 9.57, + "learning_rate": 2.12912388027092e-07, + "loss": 0.3122, + "step": 43871 + }, + { + "epoch": 9.57, + "learning_rate": 2.128031461656107e-07, + "loss": 0.5657, + "step": 43872 + }, + { + "epoch": 9.58, + "learning_rate": 2.1269390430412936e-07, + "loss": 0.9048, + "step": 43873 + }, + { + "epoch": 9.58, + "learning_rate": 2.1258466244264805e-07, + "loss": 0.8478, + "step": 43874 + }, + { + "epoch": 9.58, + "learning_rate": 2.124754205811667e-07, + "loss": 0.3869, + "step": 43875 + }, + { + "epoch": 9.58, + "learning_rate": 2.123661787196854e-07, + "loss": 0.315, + "step": 43876 + }, + { + "epoch": 9.58, + "learning_rate": 2.122569368582041e-07, + "loss": 0.5812, + "step": 43877 + }, + { + "epoch": 9.58, + "learning_rate": 2.1214769499672276e-07, + "loss": 1.4158, + "step": 43878 + }, + { + "epoch": 9.58, + "learning_rate": 2.1203845313524145e-07, + "loss": 0.6675, + "step": 43879 + }, + { + "epoch": 9.58, + "learning_rate": 2.119292112737601e-07, + "loss": 0.4664, + "step": 43880 + }, + { + "epoch": 9.58, + "learning_rate": 2.118199694122788e-07, + "loss": 0.5939, + "step": 43881 + }, + { + "epoch": 9.58, + "learning_rate": 2.117107275507975e-07, + "loss": 0.5133, + "step": 43882 + }, + { + "epoch": 9.58, + "learning_rate": 2.1160148568931616e-07, + "loss": 1.7279, + "step": 43883 + }, + { + "epoch": 9.58, + "learning_rate": 2.1149224382783485e-07, + "loss": 2.1186, + "step": 43884 + }, + { + "epoch": 9.58, + "learning_rate": 2.113830019663535e-07, + "loss": 0.8991, + "step": 43885 + }, + { + "epoch": 9.58, + "learning_rate": 2.112737601048722e-07, + "loss": 1.242, + "step": 43886 + }, + { + "epoch": 9.58, + "learning_rate": 2.111645182433909e-07, + "loss": 0.2115, + "step": 43887 + }, + { + "epoch": 9.58, + "learning_rate": 2.1105527638190956e-07, + "loss": 0.734, + "step": 43888 + }, + { + "epoch": 9.58, + "learning_rate": 2.1094603452042825e-07, + "loss": 0.3708, + "step": 43889 + }, + { + "epoch": 9.58, + "learning_rate": 2.108367926589469e-07, + "loss": 0.9597, + "step": 43890 + }, + { + "epoch": 9.58, + "learning_rate": 2.107275507974656e-07, + "loss": 1.1112, + "step": 43891 + }, + { + "epoch": 9.58, + "learning_rate": 2.106183089359843e-07, + "loss": 0.3728, + "step": 43892 + }, + { + "epoch": 9.58, + "learning_rate": 2.1050906707450295e-07, + "loss": 1.6089, + "step": 43893 + }, + { + "epoch": 9.58, + "learning_rate": 2.1039982521302164e-07, + "loss": 0.9412, + "step": 43894 + }, + { + "epoch": 9.58, + "learning_rate": 2.102905833515403e-07, + "loss": 0.9208, + "step": 43895 + }, + { + "epoch": 9.58, + "learning_rate": 2.10181341490059e-07, + "loss": 0.5892, + "step": 43896 + }, + { + "epoch": 9.58, + "learning_rate": 2.100720996285777e-07, + "loss": 1.2164, + "step": 43897 + }, + { + "epoch": 9.58, + "learning_rate": 2.0996285776709635e-07, + "loss": 2.0247, + "step": 43898 + }, + { + "epoch": 9.58, + "learning_rate": 2.0985361590561504e-07, + "loss": 0.9233, + "step": 43899 + }, + { + "epoch": 9.58, + "learning_rate": 2.0974437404413373e-07, + "loss": 1.0216, + "step": 43900 + }, + { + "epoch": 9.58, + "learning_rate": 2.096351321826524e-07, + "loss": 1.1159, + "step": 43901 + }, + { + "epoch": 9.58, + "learning_rate": 2.095258903211711e-07, + "loss": 0.1573, + "step": 43902 + }, + { + "epoch": 9.58, + "learning_rate": 2.0941664845968975e-07, + "loss": 0.9383, + "step": 43903 + }, + { + "epoch": 9.58, + "learning_rate": 2.0930740659820844e-07, + "loss": 0.8566, + "step": 43904 + }, + { + "epoch": 9.58, + "learning_rate": 2.0919816473672716e-07, + "loss": 0.4395, + "step": 43905 + }, + { + "epoch": 9.58, + "learning_rate": 2.090889228752458e-07, + "loss": 0.3199, + "step": 43906 + }, + { + "epoch": 9.58, + "learning_rate": 2.089796810137645e-07, + "loss": 1.3591, + "step": 43907 + }, + { + "epoch": 9.58, + "learning_rate": 2.0887043915228315e-07, + "loss": 1.2174, + "step": 43908 + }, + { + "epoch": 9.58, + "learning_rate": 2.0876119729080184e-07, + "loss": 0.3126, + "step": 43909 + }, + { + "epoch": 9.58, + "learning_rate": 2.0865195542932056e-07, + "loss": 3.0666, + "step": 43910 + }, + { + "epoch": 9.58, + "learning_rate": 2.085427135678392e-07, + "loss": 0.6561, + "step": 43911 + }, + { + "epoch": 9.58, + "learning_rate": 2.084334717063579e-07, + "loss": 1.7554, + "step": 43912 + }, + { + "epoch": 9.58, + "learning_rate": 2.0832422984487655e-07, + "loss": 1.237, + "step": 43913 + }, + { + "epoch": 9.58, + "learning_rate": 2.0821498798339527e-07, + "loss": 0.7722, + "step": 43914 + }, + { + "epoch": 9.58, + "learning_rate": 2.0810574612191396e-07, + "loss": 0.9959, + "step": 43915 + }, + { + "epoch": 9.58, + "learning_rate": 2.079965042604326e-07, + "loss": 0.7145, + "step": 43916 + }, + { + "epoch": 9.58, + "learning_rate": 2.078872623989513e-07, + "loss": 1.0997, + "step": 43917 + }, + { + "epoch": 9.58, + "learning_rate": 2.0777802053746995e-07, + "loss": 1.7244, + "step": 43918 + }, + { + "epoch": 9.59, + "learning_rate": 2.0766877867598867e-07, + "loss": 0.2834, + "step": 43919 + }, + { + "epoch": 9.59, + "learning_rate": 2.0755953681450736e-07, + "loss": 1.1199, + "step": 43920 + }, + { + "epoch": 9.59, + "learning_rate": 2.0745029495302602e-07, + "loss": 0.5651, + "step": 43921 + }, + { + "epoch": 9.59, + "learning_rate": 2.073410530915447e-07, + "loss": 0.7058, + "step": 43922 + }, + { + "epoch": 9.59, + "learning_rate": 2.0723181123006337e-07, + "loss": 0.5473, + "step": 43923 + }, + { + "epoch": 9.59, + "learning_rate": 2.0712256936858207e-07, + "loss": 2.2218, + "step": 43924 + }, + { + "epoch": 9.59, + "learning_rate": 2.0701332750710076e-07, + "loss": 0.0644, + "step": 43925 + }, + { + "epoch": 9.59, + "learning_rate": 2.0690408564561942e-07, + "loss": 1.1292, + "step": 43926 + }, + { + "epoch": 9.59, + "learning_rate": 2.067948437841381e-07, + "loss": 0.534, + "step": 43927 + }, + { + "epoch": 9.59, + "learning_rate": 2.0668560192265677e-07, + "loss": 0.2128, + "step": 43928 + }, + { + "epoch": 9.59, + "learning_rate": 2.0657636006117546e-07, + "loss": 1.815, + "step": 43929 + }, + { + "epoch": 9.59, + "learning_rate": 2.0646711819969415e-07, + "loss": 0.4121, + "step": 43930 + }, + { + "epoch": 9.59, + "learning_rate": 2.0635787633821282e-07, + "loss": 1.1062, + "step": 43931 + }, + { + "epoch": 9.59, + "learning_rate": 2.062486344767315e-07, + "loss": 1.7842, + "step": 43932 + }, + { + "epoch": 9.59, + "learning_rate": 2.0613939261525017e-07, + "loss": 2.4531, + "step": 43933 + }, + { + "epoch": 9.59, + "learning_rate": 2.0603015075376886e-07, + "loss": 1.9169, + "step": 43934 + }, + { + "epoch": 9.59, + "learning_rate": 2.0592090889228755e-07, + "loss": 0.5163, + "step": 43935 + }, + { + "epoch": 9.59, + "learning_rate": 2.0581166703080622e-07, + "loss": 0.8881, + "step": 43936 + }, + { + "epoch": 9.59, + "learning_rate": 2.057024251693249e-07, + "loss": 0.7469, + "step": 43937 + }, + { + "epoch": 9.59, + "learning_rate": 2.055931833078436e-07, + "loss": 0.5724, + "step": 43938 + }, + { + "epoch": 9.59, + "learning_rate": 2.0548394144636226e-07, + "loss": 1.5009, + "step": 43939 + }, + { + "epoch": 9.59, + "learning_rate": 2.0537469958488095e-07, + "loss": 1.2535, + "step": 43940 + }, + { + "epoch": 9.59, + "learning_rate": 2.0526545772339962e-07, + "loss": 0.2788, + "step": 43941 + }, + { + "epoch": 9.59, + "learning_rate": 2.051562158619183e-07, + "loss": 0.6076, + "step": 43942 + }, + { + "epoch": 9.59, + "learning_rate": 2.05046974000437e-07, + "loss": 0.6048, + "step": 43943 + }, + { + "epoch": 9.59, + "learning_rate": 2.0493773213895566e-07, + "loss": 0.7788, + "step": 43944 + }, + { + "epoch": 9.59, + "learning_rate": 2.0482849027747435e-07, + "loss": 1.1952, + "step": 43945 + }, + { + "epoch": 9.59, + "learning_rate": 2.0471924841599302e-07, + "loss": 0.3195, + "step": 43946 + }, + { + "epoch": 9.59, + "learning_rate": 2.046100065545117e-07, + "loss": 1.498, + "step": 43947 + }, + { + "epoch": 9.59, + "learning_rate": 2.045007646930304e-07, + "loss": 1.9071, + "step": 43948 + }, + { + "epoch": 9.59, + "learning_rate": 2.0439152283154906e-07, + "loss": 1.7371, + "step": 43949 + }, + { + "epoch": 9.59, + "learning_rate": 2.0428228097006775e-07, + "loss": 0.2918, + "step": 43950 + }, + { + "epoch": 9.59, + "learning_rate": 2.0417303910858641e-07, + "loss": 0.6409, + "step": 43951 + }, + { + "epoch": 9.59, + "learning_rate": 2.040637972471051e-07, + "loss": 0.5357, + "step": 43952 + }, + { + "epoch": 9.59, + "learning_rate": 2.039545553856238e-07, + "loss": 2.0613, + "step": 43953 + }, + { + "epoch": 9.59, + "learning_rate": 2.0384531352414246e-07, + "loss": 0.4209, + "step": 43954 + }, + { + "epoch": 9.59, + "learning_rate": 2.0373607166266115e-07, + "loss": 0.6567, + "step": 43955 + }, + { + "epoch": 9.59, + "learning_rate": 2.0362682980117981e-07, + "loss": 1.4249, + "step": 43956 + }, + { + "epoch": 9.59, + "learning_rate": 2.035175879396985e-07, + "loss": 2.1564, + "step": 43957 + }, + { + "epoch": 9.59, + "learning_rate": 2.034083460782172e-07, + "loss": 0.5416, + "step": 43958 + }, + { + "epoch": 9.59, + "learning_rate": 2.0329910421673586e-07, + "loss": 2.1372, + "step": 43959 + }, + { + "epoch": 9.59, + "learning_rate": 2.0318986235525455e-07, + "loss": 1.0475, + "step": 43960 + }, + { + "epoch": 9.59, + "learning_rate": 2.030806204937732e-07, + "loss": 0.5025, + "step": 43961 + }, + { + "epoch": 9.59, + "learning_rate": 2.029713786322919e-07, + "loss": 1.7932, + "step": 43962 + }, + { + "epoch": 9.59, + "learning_rate": 2.028621367708106e-07, + "loss": 0.5968, + "step": 43963 + }, + { + "epoch": 9.59, + "learning_rate": 2.0275289490932926e-07, + "loss": 0.4523, + "step": 43964 + }, + { + "epoch": 9.6, + "learning_rate": 2.0264365304784795e-07, + "loss": 0.7989, + "step": 43965 + }, + { + "epoch": 9.6, + "learning_rate": 2.025344111863666e-07, + "loss": 0.5451, + "step": 43966 + }, + { + "epoch": 9.6, + "learning_rate": 2.024251693248853e-07, + "loss": 1.465, + "step": 43967 + }, + { + "epoch": 9.6, + "learning_rate": 2.02315927463404e-07, + "loss": 0.4737, + "step": 43968 + }, + { + "epoch": 9.6, + "learning_rate": 2.0220668560192266e-07, + "loss": 0.4526, + "step": 43969 + }, + { + "epoch": 9.6, + "learning_rate": 2.0209744374044135e-07, + "loss": 2.841, + "step": 43970 + }, + { + "epoch": 9.6, + "learning_rate": 2.0198820187896e-07, + "loss": 0.7507, + "step": 43971 + }, + { + "epoch": 9.6, + "learning_rate": 2.018789600174787e-07, + "loss": 2.2598, + "step": 43972 + }, + { + "epoch": 9.6, + "learning_rate": 2.0176971815599742e-07, + "loss": 1.0557, + "step": 43973 + }, + { + "epoch": 9.6, + "learning_rate": 2.0166047629451606e-07, + "loss": 0.6467, + "step": 43974 + }, + { + "epoch": 9.6, + "learning_rate": 2.0155123443303475e-07, + "loss": 1.0585, + "step": 43975 + }, + { + "epoch": 9.6, + "learning_rate": 2.014419925715534e-07, + "loss": 0.19, + "step": 43976 + }, + { + "epoch": 9.6, + "learning_rate": 2.013327507100721e-07, + "loss": 0.8831, + "step": 43977 + }, + { + "epoch": 9.6, + "learning_rate": 2.0122350884859082e-07, + "loss": 0.183, + "step": 43978 + }, + { + "epoch": 9.6, + "learning_rate": 2.0111426698710945e-07, + "loss": 1.1166, + "step": 43979 + }, + { + "epoch": 9.6, + "learning_rate": 2.0100502512562817e-07, + "loss": 0.8938, + "step": 43980 + }, + { + "epoch": 9.6, + "learning_rate": 2.0089578326414686e-07, + "loss": 1.4714, + "step": 43981 + }, + { + "epoch": 9.6, + "learning_rate": 2.0078654140266553e-07, + "loss": 0.0397, + "step": 43982 + }, + { + "epoch": 9.6, + "learning_rate": 2.0067729954118422e-07, + "loss": 0.4772, + "step": 43983 + }, + { + "epoch": 9.6, + "learning_rate": 2.0056805767970285e-07, + "loss": 1.0862, + "step": 43984 + }, + { + "epoch": 9.6, + "learning_rate": 2.0045881581822157e-07, + "loss": 1.5379, + "step": 43985 + }, + { + "epoch": 9.6, + "learning_rate": 2.0034957395674026e-07, + "loss": 0.1621, + "step": 43986 + }, + { + "epoch": 9.6, + "learning_rate": 2.0024033209525893e-07, + "loss": 1.1344, + "step": 43987 + }, + { + "epoch": 9.6, + "learning_rate": 2.0013109023377762e-07, + "loss": 0.7147, + "step": 43988 + }, + { + "epoch": 9.6, + "learning_rate": 2.0002184837229628e-07, + "loss": 0.496, + "step": 43989 + }, + { + "epoch": 9.6, + "learning_rate": 1.9991260651081497e-07, + "loss": 0.1795, + "step": 43990 + }, + { + "epoch": 9.6, + "learning_rate": 1.9980336464933366e-07, + "loss": 0.7382, + "step": 43991 + }, + { + "epoch": 9.6, + "learning_rate": 1.9969412278785232e-07, + "loss": 0.8206, + "step": 43992 + }, + { + "epoch": 9.6, + "learning_rate": 1.9958488092637101e-07, + "loss": 0.8216, + "step": 43993 + }, + { + "epoch": 9.6, + "learning_rate": 1.9947563906488968e-07, + "loss": 2.1044, + "step": 43994 + }, + { + "epoch": 9.6, + "learning_rate": 1.9936639720340837e-07, + "loss": 1.3827, + "step": 43995 + }, + { + "epoch": 9.6, + "learning_rate": 1.9925715534192706e-07, + "loss": 1.2925, + "step": 43996 + }, + { + "epoch": 9.6, + "learning_rate": 1.9914791348044572e-07, + "loss": 0.0809, + "step": 43997 + }, + { + "epoch": 9.6, + "learning_rate": 1.9903867161896441e-07, + "loss": 1.5621, + "step": 43998 + }, + { + "epoch": 9.6, + "learning_rate": 1.9892942975748308e-07, + "loss": 1.3777, + "step": 43999 + }, + { + "epoch": 9.6, + "learning_rate": 1.9882018789600177e-07, + "loss": 0.8522, + "step": 44000 + }, + { + "epoch": 9.6, + "learning_rate": 1.9871094603452046e-07, + "loss": 1.8801, + "step": 44001 + }, + { + "epoch": 9.6, + "learning_rate": 1.9860170417303912e-07, + "loss": 2.4022, + "step": 44002 + }, + { + "epoch": 9.6, + "learning_rate": 1.984924623115578e-07, + "loss": 0.9904, + "step": 44003 + }, + { + "epoch": 9.6, + "learning_rate": 1.9838322045007648e-07, + "loss": 0.0855, + "step": 44004 + }, + { + "epoch": 9.6, + "learning_rate": 1.9827397858859517e-07, + "loss": 0.4551, + "step": 44005 + }, + { + "epoch": 9.6, + "learning_rate": 1.9816473672711386e-07, + "loss": 0.6482, + "step": 44006 + }, + { + "epoch": 9.6, + "learning_rate": 1.9805549486563252e-07, + "loss": 0.5029, + "step": 44007 + }, + { + "epoch": 9.6, + "learning_rate": 1.979462530041512e-07, + "loss": 0.1016, + "step": 44008 + }, + { + "epoch": 9.6, + "learning_rate": 1.9783701114266988e-07, + "loss": 0.8368, + "step": 44009 + }, + { + "epoch": 9.6, + "learning_rate": 1.9772776928118857e-07, + "loss": 1.1235, + "step": 44010 + }, + { + "epoch": 9.61, + "learning_rate": 1.9761852741970726e-07, + "loss": 0.269, + "step": 44011 + }, + { + "epoch": 9.61, + "learning_rate": 1.9750928555822592e-07, + "loss": 1.2213, + "step": 44012 + }, + { + "epoch": 9.61, + "learning_rate": 1.974000436967446e-07, + "loss": 0.059, + "step": 44013 + }, + { + "epoch": 9.61, + "learning_rate": 1.9729080183526327e-07, + "loss": 1.8507, + "step": 44014 + }, + { + "epoch": 9.61, + "learning_rate": 1.9718155997378196e-07, + "loss": 0.7606, + "step": 44015 + }, + { + "epoch": 9.61, + "learning_rate": 1.9707231811230066e-07, + "loss": 0.9198, + "step": 44016 + }, + { + "epoch": 9.61, + "learning_rate": 1.9696307625081932e-07, + "loss": 1.3959, + "step": 44017 + }, + { + "epoch": 9.61, + "learning_rate": 1.96853834389338e-07, + "loss": 1.0236, + "step": 44018 + }, + { + "epoch": 9.61, + "learning_rate": 1.967445925278567e-07, + "loss": 0.8036, + "step": 44019 + }, + { + "epoch": 9.61, + "learning_rate": 1.9663535066637536e-07, + "loss": 1.425, + "step": 44020 + }, + { + "epoch": 9.61, + "learning_rate": 1.9652610880489405e-07, + "loss": 1.0396, + "step": 44021 + }, + { + "epoch": 9.61, + "learning_rate": 1.9641686694341272e-07, + "loss": 0.7629, + "step": 44022 + }, + { + "epoch": 9.61, + "learning_rate": 1.963076250819314e-07, + "loss": 0.8086, + "step": 44023 + }, + { + "epoch": 9.61, + "learning_rate": 1.961983832204501e-07, + "loss": 0.9672, + "step": 44024 + }, + { + "epoch": 9.61, + "learning_rate": 1.9608914135896876e-07, + "loss": 3.2305, + "step": 44025 + }, + { + "epoch": 9.61, + "learning_rate": 1.9597989949748745e-07, + "loss": 1.941, + "step": 44026 + }, + { + "epoch": 9.61, + "learning_rate": 1.9587065763600612e-07, + "loss": 0.1186, + "step": 44027 + }, + { + "epoch": 9.61, + "learning_rate": 1.957614157745248e-07, + "loss": 1.3931, + "step": 44028 + }, + { + "epoch": 9.61, + "learning_rate": 1.956521739130435e-07, + "loss": 0.7398, + "step": 44029 + }, + { + "epoch": 9.61, + "learning_rate": 1.9554293205156216e-07, + "loss": 0.2372, + "step": 44030 + }, + { + "epoch": 9.61, + "learning_rate": 1.9543369019008085e-07, + "loss": 2.1159, + "step": 44031 + }, + { + "epoch": 9.61, + "learning_rate": 1.9532444832859952e-07, + "loss": 4.2558, + "step": 44032 + }, + { + "epoch": 9.61, + "learning_rate": 1.952152064671182e-07, + "loss": 1.8952, + "step": 44033 + }, + { + "epoch": 9.61, + "learning_rate": 1.9510596460563692e-07, + "loss": 0.7298, + "step": 44034 + }, + { + "epoch": 9.61, + "learning_rate": 1.9499672274415556e-07, + "loss": 2.0822, + "step": 44035 + }, + { + "epoch": 9.61, + "learning_rate": 1.9488748088267425e-07, + "loss": 0.5143, + "step": 44036 + }, + { + "epoch": 9.61, + "learning_rate": 1.9477823902119292e-07, + "loss": 1.7178, + "step": 44037 + }, + { + "epoch": 9.61, + "learning_rate": 1.946689971597116e-07, + "loss": 1.4559, + "step": 44038 + }, + { + "epoch": 9.61, + "learning_rate": 1.9455975529823032e-07, + "loss": 0.5693, + "step": 44039 + }, + { + "epoch": 9.61, + "learning_rate": 1.9445051343674896e-07, + "loss": 0.1982, + "step": 44040 + }, + { + "epoch": 9.61, + "learning_rate": 1.9434127157526768e-07, + "loss": 0.1999, + "step": 44041 + }, + { + "epoch": 9.61, + "learning_rate": 1.9423202971378631e-07, + "loss": 0.8444, + "step": 44042 + }, + { + "epoch": 9.61, + "learning_rate": 1.9412278785230503e-07, + "loss": 0.4248, + "step": 44043 + }, + { + "epoch": 9.61, + "learning_rate": 1.9401354599082372e-07, + "loss": 0.5814, + "step": 44044 + }, + { + "epoch": 9.61, + "learning_rate": 1.9390430412934236e-07, + "loss": 1.0005, + "step": 44045 + }, + { + "epoch": 9.61, + "learning_rate": 1.9379506226786108e-07, + "loss": 0.106, + "step": 44046 + }, + { + "epoch": 9.61, + "learning_rate": 1.9368582040637971e-07, + "loss": 1.1537, + "step": 44047 + }, + { + "epoch": 9.61, + "learning_rate": 1.9357657854489843e-07, + "loss": 0.3065, + "step": 44048 + }, + { + "epoch": 9.61, + "learning_rate": 1.9346733668341712e-07, + "loss": 0.6914, + "step": 44049 + }, + { + "epoch": 9.61, + "learning_rate": 1.9335809482193578e-07, + "loss": 0.9508, + "step": 44050 + }, + { + "epoch": 9.61, + "learning_rate": 1.9324885296045448e-07, + "loss": 1.0645, + "step": 44051 + }, + { + "epoch": 9.61, + "learning_rate": 1.9313961109897314e-07, + "loss": 1.5629, + "step": 44052 + }, + { + "epoch": 9.61, + "learning_rate": 1.9303036923749183e-07, + "loss": 0.9163, + "step": 44053 + }, + { + "epoch": 9.61, + "learning_rate": 1.9292112737601052e-07, + "loss": 1.3651, + "step": 44054 + }, + { + "epoch": 9.61, + "learning_rate": 1.9281188551452918e-07, + "loss": 0.113, + "step": 44055 + }, + { + "epoch": 9.62, + "learning_rate": 1.9270264365304787e-07, + "loss": 0.8274, + "step": 44056 + }, + { + "epoch": 9.62, + "learning_rate": 1.9259340179156654e-07, + "loss": 2.2863, + "step": 44057 + }, + { + "epoch": 9.62, + "learning_rate": 1.9248415993008523e-07, + "loss": 1.6751, + "step": 44058 + }, + { + "epoch": 9.62, + "learning_rate": 1.9237491806860392e-07, + "loss": 1.6034, + "step": 44059 + }, + { + "epoch": 9.62, + "learning_rate": 1.9226567620712258e-07, + "loss": 0.905, + "step": 44060 + }, + { + "epoch": 9.62, + "learning_rate": 1.9215643434564127e-07, + "loss": 0.6521, + "step": 44061 + }, + { + "epoch": 9.62, + "learning_rate": 1.9204719248415996e-07, + "loss": 0.3079, + "step": 44062 + }, + { + "epoch": 9.62, + "learning_rate": 1.9193795062267863e-07, + "loss": 0.9966, + "step": 44063 + }, + { + "epoch": 9.62, + "learning_rate": 1.9182870876119732e-07, + "loss": 1.1274, + "step": 44064 + }, + { + "epoch": 9.62, + "learning_rate": 1.9171946689971598e-07, + "loss": 0.374, + "step": 44065 + }, + { + "epoch": 9.62, + "learning_rate": 1.9161022503823467e-07, + "loss": 0.3605, + "step": 44066 + }, + { + "epoch": 9.62, + "learning_rate": 1.9150098317675336e-07, + "loss": 0.7913, + "step": 44067 + }, + { + "epoch": 9.62, + "learning_rate": 1.9139174131527203e-07, + "loss": 0.1932, + "step": 44068 + }, + { + "epoch": 9.62, + "learning_rate": 1.9128249945379072e-07, + "loss": 0.7823, + "step": 44069 + }, + { + "epoch": 9.62, + "learning_rate": 1.9117325759230938e-07, + "loss": 2.1061, + "step": 44070 + }, + { + "epoch": 9.62, + "learning_rate": 1.9106401573082807e-07, + "loss": 1.0623, + "step": 44071 + }, + { + "epoch": 9.62, + "learning_rate": 1.9095477386934676e-07, + "loss": 1.1803, + "step": 44072 + }, + { + "epoch": 9.62, + "learning_rate": 1.9084553200786543e-07, + "loss": 1.5648, + "step": 44073 + }, + { + "epoch": 9.62, + "learning_rate": 1.9073629014638412e-07, + "loss": 1.6478, + "step": 44074 + }, + { + "epoch": 9.62, + "learning_rate": 1.9062704828490278e-07, + "loss": 0.3531, + "step": 44075 + }, + { + "epoch": 9.62, + "learning_rate": 1.9051780642342147e-07, + "loss": 0.1844, + "step": 44076 + }, + { + "epoch": 9.62, + "learning_rate": 1.9040856456194016e-07, + "loss": 0.8234, + "step": 44077 + }, + { + "epoch": 9.62, + "learning_rate": 1.9029932270045882e-07, + "loss": 1.5763, + "step": 44078 + }, + { + "epoch": 9.62, + "learning_rate": 1.9019008083897752e-07, + "loss": 0.4452, + "step": 44079 + }, + { + "epoch": 9.62, + "learning_rate": 1.9008083897749618e-07, + "loss": 1.7299, + "step": 44080 + }, + { + "epoch": 9.62, + "learning_rate": 1.8997159711601487e-07, + "loss": 1.3649, + "step": 44081 + }, + { + "epoch": 9.62, + "learning_rate": 1.8986235525453356e-07, + "loss": 1.4933, + "step": 44082 + }, + { + "epoch": 9.62, + "learning_rate": 1.8975311339305222e-07, + "loss": 0.5193, + "step": 44083 + }, + { + "epoch": 9.62, + "learning_rate": 1.8964387153157091e-07, + "loss": 0.419, + "step": 44084 + }, + { + "epoch": 9.62, + "learning_rate": 1.8953462967008958e-07, + "loss": 0.2377, + "step": 44085 + }, + { + "epoch": 9.62, + "learning_rate": 1.8942538780860827e-07, + "loss": 0.6031, + "step": 44086 + }, + { + "epoch": 9.62, + "learning_rate": 1.8931614594712696e-07, + "loss": 1.1687, + "step": 44087 + }, + { + "epoch": 9.62, + "learning_rate": 1.8920690408564562e-07, + "loss": 0.6977, + "step": 44088 + }, + { + "epoch": 9.62, + "learning_rate": 1.890976622241643e-07, + "loss": 1.1695, + "step": 44089 + }, + { + "epoch": 9.62, + "learning_rate": 1.8898842036268298e-07, + "loss": 2.1305, + "step": 44090 + }, + { + "epoch": 9.62, + "learning_rate": 1.8887917850120167e-07, + "loss": 0.5317, + "step": 44091 + }, + { + "epoch": 9.62, + "learning_rate": 1.8876993663972036e-07, + "loss": 1.6953, + "step": 44092 + }, + { + "epoch": 9.62, + "learning_rate": 1.8866069477823902e-07, + "loss": 0.6653, + "step": 44093 + }, + { + "epoch": 9.62, + "learning_rate": 1.885514529167577e-07, + "loss": 1.6875, + "step": 44094 + }, + { + "epoch": 9.62, + "learning_rate": 1.8844221105527638e-07, + "loss": 1.4362, + "step": 44095 + }, + { + "epoch": 9.62, + "learning_rate": 1.8833296919379507e-07, + "loss": 1.3166, + "step": 44096 + }, + { + "epoch": 9.62, + "learning_rate": 1.8822372733231376e-07, + "loss": 0.5086, + "step": 44097 + }, + { + "epoch": 9.62, + "learning_rate": 1.8811448547083242e-07, + "loss": 1.3167, + "step": 44098 + }, + { + "epoch": 9.62, + "learning_rate": 1.880052436093511e-07, + "loss": 0.106, + "step": 44099 + }, + { + "epoch": 9.62, + "learning_rate": 1.8789600174786983e-07, + "loss": 2.3642, + "step": 44100 + }, + { + "epoch": 9.62, + "learning_rate": 1.8778675988638847e-07, + "loss": 0.4052, + "step": 44101 + }, + { + "epoch": 9.63, + "learning_rate": 1.8767751802490718e-07, + "loss": 1.3264, + "step": 44102 + }, + { + "epoch": 9.63, + "learning_rate": 1.8756827616342582e-07, + "loss": 1.411, + "step": 44103 + }, + { + "epoch": 9.63, + "learning_rate": 1.874590343019445e-07, + "loss": 0.8185, + "step": 44104 + }, + { + "epoch": 9.63, + "learning_rate": 1.8734979244046323e-07, + "loss": 0.8123, + "step": 44105 + }, + { + "epoch": 9.63, + "learning_rate": 1.8724055057898186e-07, + "loss": 0.4123, + "step": 44106 + }, + { + "epoch": 9.63, + "learning_rate": 1.8713130871750058e-07, + "loss": 0.7366, + "step": 44107 + }, + { + "epoch": 9.63, + "learning_rate": 1.8702206685601922e-07, + "loss": 0.1919, + "step": 44108 + }, + { + "epoch": 9.63, + "learning_rate": 1.8691282499453794e-07, + "loss": 2.2091, + "step": 44109 + }, + { + "epoch": 9.63, + "learning_rate": 1.8680358313305663e-07, + "loss": 1.7621, + "step": 44110 + }, + { + "epoch": 9.63, + "learning_rate": 1.866943412715753e-07, + "loss": 0.4978, + "step": 44111 + }, + { + "epoch": 9.63, + "learning_rate": 1.8658509941009398e-07, + "loss": 2.2305, + "step": 44112 + }, + { + "epoch": 9.63, + "learning_rate": 1.8647585754861262e-07, + "loss": 1.5949, + "step": 44113 + }, + { + "epoch": 9.63, + "learning_rate": 1.8636661568713133e-07, + "loss": 0.902, + "step": 44114 + }, + { + "epoch": 9.63, + "learning_rate": 1.8625737382565003e-07, + "loss": 1.0391, + "step": 44115 + }, + { + "epoch": 9.63, + "learning_rate": 1.861481319641687e-07, + "loss": 0.6131, + "step": 44116 + }, + { + "epoch": 9.63, + "learning_rate": 1.8603889010268738e-07, + "loss": 0.9772, + "step": 44117 + }, + { + "epoch": 9.63, + "learning_rate": 1.8592964824120604e-07, + "loss": 0.3796, + "step": 44118 + }, + { + "epoch": 9.63, + "learning_rate": 1.8582040637972473e-07, + "loss": 2.7621, + "step": 44119 + }, + { + "epoch": 9.63, + "learning_rate": 1.8571116451824342e-07, + "loss": 0.6973, + "step": 44120 + }, + { + "epoch": 9.63, + "learning_rate": 1.856019226567621e-07, + "loss": 2.5857, + "step": 44121 + }, + { + "epoch": 9.63, + "learning_rate": 1.8549268079528078e-07, + "loss": 0.4379, + "step": 44122 + }, + { + "epoch": 9.63, + "learning_rate": 1.8538343893379944e-07, + "loss": 0.4777, + "step": 44123 + }, + { + "epoch": 9.63, + "learning_rate": 1.8527419707231813e-07, + "loss": 1.8164, + "step": 44124 + }, + { + "epoch": 9.63, + "learning_rate": 1.8516495521083682e-07, + "loss": 0.9581, + "step": 44125 + }, + { + "epoch": 9.63, + "learning_rate": 1.850557133493555e-07, + "loss": 0.4159, + "step": 44126 + }, + { + "epoch": 9.63, + "learning_rate": 1.8494647148787418e-07, + "loss": 1.591, + "step": 44127 + }, + { + "epoch": 9.63, + "learning_rate": 1.8483722962639284e-07, + "loss": 0.9307, + "step": 44128 + }, + { + "epoch": 9.63, + "learning_rate": 1.8472798776491153e-07, + "loss": 0.5324, + "step": 44129 + }, + { + "epoch": 9.63, + "learning_rate": 1.8461874590343022e-07, + "loss": 0.7858, + "step": 44130 + }, + { + "epoch": 9.63, + "learning_rate": 1.8450950404194889e-07, + "loss": 1.3243, + "step": 44131 + }, + { + "epoch": 9.63, + "learning_rate": 1.8440026218046758e-07, + "loss": 1.235, + "step": 44132 + }, + { + "epoch": 9.63, + "learning_rate": 1.8429102031898624e-07, + "loss": 1.8894, + "step": 44133 + }, + { + "epoch": 9.63, + "learning_rate": 1.8418177845750493e-07, + "loss": 1.4637, + "step": 44134 + }, + { + "epoch": 9.63, + "learning_rate": 1.8407253659602362e-07, + "loss": 1.1459, + "step": 44135 + }, + { + "epoch": 9.63, + "learning_rate": 1.8396329473454229e-07, + "loss": 0.6221, + "step": 44136 + }, + { + "epoch": 9.63, + "learning_rate": 1.8385405287306098e-07, + "loss": 1.0986, + "step": 44137 + }, + { + "epoch": 9.63, + "learning_rate": 1.8374481101157964e-07, + "loss": 0.6219, + "step": 44138 + }, + { + "epoch": 9.63, + "learning_rate": 1.8363556915009833e-07, + "loss": 0.5786, + "step": 44139 + }, + { + "epoch": 9.63, + "learning_rate": 1.8352632728861702e-07, + "loss": 0.721, + "step": 44140 + }, + { + "epoch": 9.63, + "learning_rate": 1.8341708542713568e-07, + "loss": 0.7897, + "step": 44141 + }, + { + "epoch": 9.63, + "learning_rate": 1.8330784356565437e-07, + "loss": 0.3761, + "step": 44142 + }, + { + "epoch": 9.63, + "learning_rate": 1.8319860170417307e-07, + "loss": 1.3868, + "step": 44143 + }, + { + "epoch": 9.63, + "learning_rate": 1.8308935984269173e-07, + "loss": 0.4142, + "step": 44144 + }, + { + "epoch": 9.63, + "learning_rate": 1.8298011798121042e-07, + "loss": 3.089, + "step": 44145 + }, + { + "epoch": 9.63, + "learning_rate": 1.8287087611972908e-07, + "loss": 1.9732, + "step": 44146 + }, + { + "epoch": 9.63, + "learning_rate": 1.8276163425824777e-07, + "loss": 0.1273, + "step": 44147 + }, + { + "epoch": 9.64, + "learning_rate": 1.8265239239676646e-07, + "loss": 2.2953, + "step": 44148 + }, + { + "epoch": 9.64, + "learning_rate": 1.8254315053528513e-07, + "loss": 0.5725, + "step": 44149 + }, + { + "epoch": 9.64, + "learning_rate": 1.8243390867380382e-07, + "loss": 2.5765, + "step": 44150 + }, + { + "epoch": 9.64, + "learning_rate": 1.8232466681232248e-07, + "loss": 0.2471, + "step": 44151 + }, + { + "epoch": 9.64, + "learning_rate": 1.8221542495084117e-07, + "loss": 0.387, + "step": 44152 + }, + { + "epoch": 9.64, + "learning_rate": 1.8210618308935986e-07, + "loss": 0.7675, + "step": 44153 + }, + { + "epoch": 9.64, + "learning_rate": 1.8199694122787853e-07, + "loss": 1.1252, + "step": 44154 + }, + { + "epoch": 9.64, + "learning_rate": 1.8188769936639722e-07, + "loss": 0.9312, + "step": 44155 + }, + { + "epoch": 9.64, + "learning_rate": 1.8177845750491588e-07, + "loss": 2.4298, + "step": 44156 + }, + { + "epoch": 9.64, + "learning_rate": 1.8166921564343457e-07, + "loss": 1.9829, + "step": 44157 + }, + { + "epoch": 9.64, + "learning_rate": 1.8155997378195326e-07, + "loss": 0.6412, + "step": 44158 + }, + { + "epoch": 9.64, + "learning_rate": 1.8145073192047193e-07, + "loss": 0.9639, + "step": 44159 + }, + { + "epoch": 9.64, + "learning_rate": 1.8134149005899062e-07, + "loss": 1.2583, + "step": 44160 + }, + { + "epoch": 9.64, + "learning_rate": 1.8123224819750928e-07, + "loss": 0.0816, + "step": 44161 + }, + { + "epoch": 9.64, + "learning_rate": 1.8112300633602797e-07, + "loss": 1.0724, + "step": 44162 + }, + { + "epoch": 9.64, + "learning_rate": 1.8101376447454666e-07, + "loss": 1.3268, + "step": 44163 + }, + { + "epoch": 9.64, + "learning_rate": 1.8090452261306533e-07, + "loss": 0.7755, + "step": 44164 + }, + { + "epoch": 9.64, + "learning_rate": 1.8079528075158402e-07, + "loss": 1.021, + "step": 44165 + }, + { + "epoch": 9.64, + "learning_rate": 1.8068603889010268e-07, + "loss": 1.4991, + "step": 44166 + }, + { + "epoch": 9.64, + "learning_rate": 1.8057679702862137e-07, + "loss": 0.5049, + "step": 44167 + }, + { + "epoch": 9.64, + "learning_rate": 1.804675551671401e-07, + "loss": 0.7682, + "step": 44168 + }, + { + "epoch": 9.64, + "learning_rate": 1.8035831330565872e-07, + "loss": 2.2778, + "step": 44169 + }, + { + "epoch": 9.64, + "learning_rate": 1.8024907144417744e-07, + "loss": 1.7897, + "step": 44170 + }, + { + "epoch": 9.64, + "learning_rate": 1.8013982958269608e-07, + "loss": 0.1931, + "step": 44171 + }, + { + "epoch": 9.64, + "learning_rate": 1.8003058772121477e-07, + "loss": 1.0597, + "step": 44172 + }, + { + "epoch": 9.64, + "learning_rate": 1.7992134585973349e-07, + "loss": 1.5434, + "step": 44173 + }, + { + "epoch": 9.64, + "learning_rate": 1.7981210399825212e-07, + "loss": 0.768, + "step": 44174 + }, + { + "epoch": 9.64, + "learning_rate": 1.7970286213677084e-07, + "loss": 0.9622, + "step": 44175 + }, + { + "epoch": 9.64, + "learning_rate": 1.7959362027528948e-07, + "loss": 0.672, + "step": 44176 + }, + { + "epoch": 9.64, + "learning_rate": 1.794843784138082e-07, + "loss": 1.2091, + "step": 44177 + }, + { + "epoch": 9.64, + "learning_rate": 1.7937513655232689e-07, + "loss": 1.4873, + "step": 44178 + }, + { + "epoch": 9.64, + "learning_rate": 1.7926589469084555e-07, + "loss": 0.689, + "step": 44179 + }, + { + "epoch": 9.64, + "learning_rate": 1.7915665282936424e-07, + "loss": 0.5794, + "step": 44180 + }, + { + "epoch": 9.64, + "learning_rate": 1.7904741096788293e-07, + "loss": 0.3826, + "step": 44181 + }, + { + "epoch": 9.64, + "learning_rate": 1.789381691064016e-07, + "loss": 4.5871, + "step": 44182 + }, + { + "epoch": 9.64, + "learning_rate": 1.7882892724492028e-07, + "loss": 1.0631, + "step": 44183 + }, + { + "epoch": 9.64, + "learning_rate": 1.7871968538343895e-07, + "loss": 1.1519, + "step": 44184 + }, + { + "epoch": 9.64, + "learning_rate": 1.7861044352195764e-07, + "loss": 2.2224, + "step": 44185 + }, + { + "epoch": 9.64, + "learning_rate": 1.7850120166047633e-07, + "loss": 1.2868, + "step": 44186 + }, + { + "epoch": 9.64, + "learning_rate": 1.78391959798995e-07, + "loss": 0.9815, + "step": 44187 + }, + { + "epoch": 9.64, + "learning_rate": 1.7828271793751368e-07, + "loss": 1.4146, + "step": 44188 + }, + { + "epoch": 9.64, + "learning_rate": 1.7817347607603235e-07, + "loss": 1.1513, + "step": 44189 + }, + { + "epoch": 9.64, + "learning_rate": 1.7806423421455104e-07, + "loss": 0.3299, + "step": 44190 + }, + { + "epoch": 9.64, + "learning_rate": 1.7795499235306973e-07, + "loss": 2.8105, + "step": 44191 + }, + { + "epoch": 9.64, + "learning_rate": 1.778457504915884e-07, + "loss": 0.3971, + "step": 44192 + }, + { + "epoch": 9.64, + "learning_rate": 1.7773650863010708e-07, + "loss": 1.1464, + "step": 44193 + }, + { + "epoch": 9.65, + "learning_rate": 1.7762726676862575e-07, + "loss": 0.6497, + "step": 44194 + }, + { + "epoch": 9.65, + "learning_rate": 1.7751802490714444e-07, + "loss": 1.3086, + "step": 44195 + }, + { + "epoch": 9.65, + "learning_rate": 1.7740878304566313e-07, + "loss": 3.2863, + "step": 44196 + }, + { + "epoch": 9.65, + "learning_rate": 1.772995411841818e-07, + "loss": 0.2677, + "step": 44197 + }, + { + "epoch": 9.65, + "learning_rate": 1.7719029932270048e-07, + "loss": 1.8519, + "step": 44198 + }, + { + "epoch": 9.65, + "learning_rate": 1.7708105746121915e-07, + "loss": 1.8574, + "step": 44199 + }, + { + "epoch": 9.65, + "learning_rate": 1.7697181559973784e-07, + "loss": 0.8949, + "step": 44200 + }, + { + "epoch": 9.65, + "learning_rate": 1.7686257373825653e-07, + "loss": 1.0731, + "step": 44201 + }, + { + "epoch": 9.65, + "learning_rate": 1.767533318767752e-07, + "loss": 0.7832, + "step": 44202 + }, + { + "epoch": 9.65, + "learning_rate": 1.7664409001529388e-07, + "loss": 0.1201, + "step": 44203 + }, + { + "epoch": 9.65, + "learning_rate": 1.7653484815381254e-07, + "loss": 0.1367, + "step": 44204 + }, + { + "epoch": 9.65, + "learning_rate": 1.7642560629233123e-07, + "loss": 0.7243, + "step": 44205 + }, + { + "epoch": 9.65, + "learning_rate": 1.7631636443084992e-07, + "loss": 1.3627, + "step": 44206 + }, + { + "epoch": 9.65, + "learning_rate": 1.762071225693686e-07, + "loss": 1.6336, + "step": 44207 + }, + { + "epoch": 9.65, + "learning_rate": 1.7609788070788728e-07, + "loss": 0.867, + "step": 44208 + }, + { + "epoch": 9.65, + "learning_rate": 1.7598863884640594e-07, + "loss": 1.8722, + "step": 44209 + }, + { + "epoch": 9.65, + "learning_rate": 1.7587939698492463e-07, + "loss": 0.4329, + "step": 44210 + }, + { + "epoch": 9.65, + "learning_rate": 1.7577015512344332e-07, + "loss": 0.6768, + "step": 44211 + }, + { + "epoch": 9.65, + "learning_rate": 1.75660913261962e-07, + "loss": 0.5512, + "step": 44212 + }, + { + "epoch": 9.65, + "learning_rate": 1.7555167140048068e-07, + "loss": 1.812, + "step": 44213 + }, + { + "epoch": 9.65, + "learning_rate": 1.7544242953899934e-07, + "loss": 2.616, + "step": 44214 + }, + { + "epoch": 9.65, + "learning_rate": 1.7533318767751803e-07, + "loss": 0.6439, + "step": 44215 + }, + { + "epoch": 9.65, + "learning_rate": 1.7522394581603672e-07, + "loss": 1.978, + "step": 44216 + }, + { + "epoch": 9.65, + "learning_rate": 1.751147039545554e-07, + "loss": 2.8332, + "step": 44217 + }, + { + "epoch": 9.65, + "learning_rate": 1.7500546209307408e-07, + "loss": 0.6265, + "step": 44218 + }, + { + "epoch": 9.65, + "learning_rate": 1.7489622023159274e-07, + "loss": 0.2598, + "step": 44219 + }, + { + "epoch": 9.65, + "learning_rate": 1.7478697837011143e-07, + "loss": 0.9267, + "step": 44220 + }, + { + "epoch": 9.65, + "learning_rate": 1.7467773650863012e-07, + "loss": 1.5307, + "step": 44221 + }, + { + "epoch": 9.65, + "learning_rate": 1.7456849464714879e-07, + "loss": 1.3017, + "step": 44222 + }, + { + "epoch": 9.65, + "learning_rate": 1.7445925278566748e-07, + "loss": 1.845, + "step": 44223 + }, + { + "epoch": 9.65, + "learning_rate": 1.7435001092418617e-07, + "loss": 1.3573, + "step": 44224 + }, + { + "epoch": 9.65, + "learning_rate": 1.7424076906270483e-07, + "loss": 0.588, + "step": 44225 + }, + { + "epoch": 9.65, + "learning_rate": 1.7413152720122352e-07, + "loss": 1.3651, + "step": 44226 + }, + { + "epoch": 9.65, + "learning_rate": 1.7402228533974219e-07, + "loss": 0.0763, + "step": 44227 + }, + { + "epoch": 9.65, + "learning_rate": 1.7391304347826088e-07, + "loss": 2.7556, + "step": 44228 + }, + { + "epoch": 9.65, + "learning_rate": 1.738038016167796e-07, + "loss": 0.5999, + "step": 44229 + }, + { + "epoch": 9.65, + "learning_rate": 1.7369455975529823e-07, + "loss": 1.2748, + "step": 44230 + }, + { + "epoch": 9.65, + "learning_rate": 1.7358531789381692e-07, + "loss": 0.2592, + "step": 44231 + }, + { + "epoch": 9.65, + "learning_rate": 1.7347607603233558e-07, + "loss": 0.8971, + "step": 44232 + }, + { + "epoch": 9.65, + "learning_rate": 1.7336683417085427e-07, + "loss": 0.199, + "step": 44233 + }, + { + "epoch": 9.65, + "learning_rate": 1.73257592309373e-07, + "loss": 0.5034, + "step": 44234 + }, + { + "epoch": 9.65, + "learning_rate": 1.7314835044789163e-07, + "loss": 0.8429, + "step": 44235 + }, + { + "epoch": 9.65, + "learning_rate": 1.7303910858641035e-07, + "loss": 2.0484, + "step": 44236 + }, + { + "epoch": 9.65, + "learning_rate": 1.7292986672492898e-07, + "loss": 1.1691, + "step": 44237 + }, + { + "epoch": 9.65, + "learning_rate": 1.728206248634477e-07, + "loss": 0.2132, + "step": 44238 + }, + { + "epoch": 9.65, + "learning_rate": 1.727113830019664e-07, + "loss": 0.378, + "step": 44239 + }, + { + "epoch": 9.66, + "learning_rate": 1.7260214114048503e-07, + "loss": 1.798, + "step": 44240 + }, + { + "epoch": 9.66, + "learning_rate": 1.7249289927900374e-07, + "loss": 0.9844, + "step": 44241 + }, + { + "epoch": 9.66, + "learning_rate": 1.7238365741752238e-07, + "loss": 1.0971, + "step": 44242 + }, + { + "epoch": 9.66, + "learning_rate": 1.722744155560411e-07, + "loss": 0.4485, + "step": 44243 + }, + { + "epoch": 9.66, + "learning_rate": 1.721651736945598e-07, + "loss": 1.3793, + "step": 44244 + }, + { + "epoch": 9.66, + "learning_rate": 1.7205593183307845e-07, + "loss": 1.1263, + "step": 44245 + }, + { + "epoch": 9.66, + "learning_rate": 1.7194668997159714e-07, + "loss": 0.7194, + "step": 44246 + }, + { + "epoch": 9.66, + "learning_rate": 1.718374481101158e-07, + "loss": 0.9483, + "step": 44247 + }, + { + "epoch": 9.66, + "learning_rate": 1.717282062486345e-07, + "loss": 0.2959, + "step": 44248 + }, + { + "epoch": 9.66, + "learning_rate": 1.716189643871532e-07, + "loss": 0.5389, + "step": 44249 + }, + { + "epoch": 9.66, + "learning_rate": 1.7150972252567185e-07, + "loss": 0.9133, + "step": 44250 + }, + { + "epoch": 9.66, + "learning_rate": 1.7140048066419054e-07, + "loss": 0.5052, + "step": 44251 + }, + { + "epoch": 9.66, + "learning_rate": 1.712912388027092e-07, + "loss": 1.7027, + "step": 44252 + }, + { + "epoch": 9.66, + "learning_rate": 1.711819969412279e-07, + "loss": 1.7667, + "step": 44253 + }, + { + "epoch": 9.66, + "learning_rate": 1.710727550797466e-07, + "loss": 1.3824, + "step": 44254 + }, + { + "epoch": 9.66, + "learning_rate": 1.7096351321826525e-07, + "loss": 1.651, + "step": 44255 + }, + { + "epoch": 9.66, + "learning_rate": 1.7085427135678394e-07, + "loss": 1.2511, + "step": 44256 + }, + { + "epoch": 9.66, + "learning_rate": 1.707450294953026e-07, + "loss": 1.1079, + "step": 44257 + }, + { + "epoch": 9.66, + "learning_rate": 1.706357876338213e-07, + "loss": 2.362, + "step": 44258 + }, + { + "epoch": 9.66, + "learning_rate": 1.7052654577233999e-07, + "loss": 1.6782, + "step": 44259 + }, + { + "epoch": 9.66, + "learning_rate": 1.7041730391085865e-07, + "loss": 0.7784, + "step": 44260 + }, + { + "epoch": 9.66, + "learning_rate": 1.7030806204937734e-07, + "loss": 2.1499, + "step": 44261 + }, + { + "epoch": 9.66, + "learning_rate": 1.7019882018789603e-07, + "loss": 0.6086, + "step": 44262 + }, + { + "epoch": 9.66, + "learning_rate": 1.700895783264147e-07, + "loss": 1.2321, + "step": 44263 + }, + { + "epoch": 9.66, + "learning_rate": 1.6998033646493339e-07, + "loss": 1.5426, + "step": 44264 + }, + { + "epoch": 9.66, + "learning_rate": 1.6987109460345205e-07, + "loss": 1.507, + "step": 44265 + }, + { + "epoch": 9.66, + "learning_rate": 1.6976185274197074e-07, + "loss": 0.7904, + "step": 44266 + }, + { + "epoch": 9.66, + "learning_rate": 1.6965261088048943e-07, + "loss": 0.9443, + "step": 44267 + }, + { + "epoch": 9.66, + "learning_rate": 1.695433690190081e-07, + "loss": 0.9155, + "step": 44268 + }, + { + "epoch": 9.66, + "learning_rate": 1.6943412715752678e-07, + "loss": 1.5063, + "step": 44269 + }, + { + "epoch": 9.66, + "learning_rate": 1.6932488529604545e-07, + "loss": 0.5837, + "step": 44270 + }, + { + "epoch": 9.66, + "learning_rate": 1.6921564343456414e-07, + "loss": 1.5565, + "step": 44271 + }, + { + "epoch": 9.66, + "learning_rate": 1.6910640157308283e-07, + "loss": 0.7469, + "step": 44272 + }, + { + "epoch": 9.66, + "learning_rate": 1.689971597116015e-07, + "loss": 0.5867, + "step": 44273 + }, + { + "epoch": 9.66, + "learning_rate": 1.6888791785012018e-07, + "loss": 1.4998, + "step": 44274 + }, + { + "epoch": 9.66, + "learning_rate": 1.6877867598863885e-07, + "loss": 0.7667, + "step": 44275 + }, + { + "epoch": 9.66, + "learning_rate": 1.6866943412715754e-07, + "loss": 0.566, + "step": 44276 + }, + { + "epoch": 9.66, + "learning_rate": 1.6856019226567623e-07, + "loss": 0.8871, + "step": 44277 + }, + { + "epoch": 9.66, + "learning_rate": 1.684509504041949e-07, + "loss": 0.4958, + "step": 44278 + }, + { + "epoch": 9.66, + "learning_rate": 1.6834170854271358e-07, + "loss": 0.9079, + "step": 44279 + }, + { + "epoch": 9.66, + "learning_rate": 1.6823246668123225e-07, + "loss": 0.1563, + "step": 44280 + }, + { + "epoch": 9.66, + "learning_rate": 1.6812322481975094e-07, + "loss": 1.6624, + "step": 44281 + }, + { + "epoch": 9.66, + "learning_rate": 1.6801398295826963e-07, + "loss": 1.3836, + "step": 44282 + }, + { + "epoch": 9.66, + "learning_rate": 1.679047410967883e-07, + "loss": 0.5758, + "step": 44283 + }, + { + "epoch": 9.66, + "learning_rate": 1.6779549923530698e-07, + "loss": 0.8622, + "step": 44284 + }, + { + "epoch": 9.66, + "learning_rate": 1.6768625737382565e-07, + "loss": 0.1447, + "step": 44285 + }, + { + "epoch": 9.67, + "learning_rate": 1.6757701551234434e-07, + "loss": 0.806, + "step": 44286 + }, + { + "epoch": 9.67, + "learning_rate": 1.6746777365086303e-07, + "loss": 1.6646, + "step": 44287 + }, + { + "epoch": 9.67, + "learning_rate": 1.673585317893817e-07, + "loss": 0.7943, + "step": 44288 + }, + { + "epoch": 9.67, + "learning_rate": 1.6724928992790038e-07, + "loss": 1.4361, + "step": 44289 + }, + { + "epoch": 9.67, + "learning_rate": 1.6714004806641904e-07, + "loss": 0.9754, + "step": 44290 + }, + { + "epoch": 9.67, + "learning_rate": 1.6703080620493774e-07, + "loss": 0.0398, + "step": 44291 + }, + { + "epoch": 9.67, + "learning_rate": 1.6692156434345643e-07, + "loss": 0.2423, + "step": 44292 + }, + { + "epoch": 9.67, + "learning_rate": 1.668123224819751e-07, + "loss": 0.4182, + "step": 44293 + }, + { + "epoch": 9.67, + "learning_rate": 1.6670308062049378e-07, + "loss": 1.5527, + "step": 44294 + }, + { + "epoch": 9.67, + "learning_rate": 1.6659383875901244e-07, + "loss": 0.2697, + "step": 44295 + }, + { + "epoch": 9.67, + "learning_rate": 1.6648459689753113e-07, + "loss": 0.5158, + "step": 44296 + }, + { + "epoch": 9.67, + "learning_rate": 1.6637535503604985e-07, + "loss": 0.7261, + "step": 44297 + }, + { + "epoch": 9.67, + "learning_rate": 1.662661131745685e-07, + "loss": 1.1748, + "step": 44298 + }, + { + "epoch": 9.67, + "learning_rate": 1.661568713130872e-07, + "loss": 0.9188, + "step": 44299 + }, + { + "epoch": 9.67, + "learning_rate": 1.6604762945160584e-07, + "loss": 1.0044, + "step": 44300 + }, + { + "epoch": 9.67, + "learning_rate": 1.6593838759012453e-07, + "loss": 0.6, + "step": 44301 + }, + { + "epoch": 9.67, + "learning_rate": 1.6582914572864325e-07, + "loss": 1.0329, + "step": 44302 + }, + { + "epoch": 9.67, + "learning_rate": 1.657199038671619e-07, + "loss": 0.1304, + "step": 44303 + }, + { + "epoch": 9.67, + "learning_rate": 1.656106620056806e-07, + "loss": 0.582, + "step": 44304 + }, + { + "epoch": 9.67, + "learning_rate": 1.655014201441993e-07, + "loss": 2.0284, + "step": 44305 + }, + { + "epoch": 9.67, + "learning_rate": 1.6539217828271796e-07, + "loss": 0.3443, + "step": 44306 + }, + { + "epoch": 9.67, + "learning_rate": 1.6528293642123665e-07, + "loss": 0.8285, + "step": 44307 + }, + { + "epoch": 9.67, + "learning_rate": 1.651736945597553e-07, + "loss": 0.8155, + "step": 44308 + }, + { + "epoch": 9.67, + "learning_rate": 1.65064452698274e-07, + "loss": 3.8444, + "step": 44309 + }, + { + "epoch": 9.67, + "learning_rate": 1.649552108367927e-07, + "loss": 1.6077, + "step": 44310 + }, + { + "epoch": 9.67, + "learning_rate": 1.6484596897531136e-07, + "loss": 0.8103, + "step": 44311 + }, + { + "epoch": 9.67, + "learning_rate": 1.6473672711383005e-07, + "loss": 0.9466, + "step": 44312 + }, + { + "epoch": 9.67, + "learning_rate": 1.646274852523487e-07, + "loss": 0.9246, + "step": 44313 + }, + { + "epoch": 9.67, + "learning_rate": 1.645182433908674e-07, + "loss": 1.4915, + "step": 44314 + }, + { + "epoch": 9.67, + "learning_rate": 1.644090015293861e-07, + "loss": 1.1769, + "step": 44315 + }, + { + "epoch": 9.67, + "learning_rate": 1.6429975966790476e-07, + "loss": 0.2226, + "step": 44316 + }, + { + "epoch": 9.67, + "learning_rate": 1.6419051780642345e-07, + "loss": 1.3663, + "step": 44317 + }, + { + "epoch": 9.67, + "learning_rate": 1.640812759449421e-07, + "loss": 0.9969, + "step": 44318 + }, + { + "epoch": 9.67, + "learning_rate": 1.639720340834608e-07, + "loss": 1.3455, + "step": 44319 + }, + { + "epoch": 9.67, + "learning_rate": 1.638627922219795e-07, + "loss": 2.4136, + "step": 44320 + }, + { + "epoch": 9.67, + "learning_rate": 1.6375355036049816e-07, + "loss": 1.5999, + "step": 44321 + }, + { + "epoch": 9.67, + "learning_rate": 1.6364430849901685e-07, + "loss": 0.3614, + "step": 44322 + }, + { + "epoch": 9.67, + "learning_rate": 1.635350666375355e-07, + "loss": 0.5509, + "step": 44323 + }, + { + "epoch": 9.67, + "learning_rate": 1.634258247760542e-07, + "loss": 0.7046, + "step": 44324 + }, + { + "epoch": 9.67, + "learning_rate": 1.633165829145729e-07, + "loss": 0.8967, + "step": 44325 + }, + { + "epoch": 9.67, + "learning_rate": 1.6320734105309156e-07, + "loss": 1.0089, + "step": 44326 + }, + { + "epoch": 9.67, + "learning_rate": 1.6309809919161025e-07, + "loss": 0.8041, + "step": 44327 + }, + { + "epoch": 9.67, + "learning_rate": 1.629888573301289e-07, + "loss": 1.2052, + "step": 44328 + }, + { + "epoch": 9.67, + "learning_rate": 1.628796154686476e-07, + "loss": 1.1207, + "step": 44329 + }, + { + "epoch": 9.67, + "learning_rate": 1.627703736071663e-07, + "loss": 0.8095, + "step": 44330 + }, + { + "epoch": 9.68, + "learning_rate": 1.6266113174568495e-07, + "loss": 1.3346, + "step": 44331 + }, + { + "epoch": 9.68, + "learning_rate": 1.6255188988420364e-07, + "loss": 0.766, + "step": 44332 + }, + { + "epoch": 9.68, + "learning_rate": 1.624426480227223e-07, + "loss": 0.7677, + "step": 44333 + }, + { + "epoch": 9.68, + "learning_rate": 1.62333406161241e-07, + "loss": 0.9594, + "step": 44334 + }, + { + "epoch": 9.68, + "learning_rate": 1.622241642997597e-07, + "loss": 0.9288, + "step": 44335 + }, + { + "epoch": 9.68, + "learning_rate": 1.6211492243827835e-07, + "loss": 0.6062, + "step": 44336 + }, + { + "epoch": 9.68, + "learning_rate": 1.6200568057679704e-07, + "loss": 1.1757, + "step": 44337 + }, + { + "epoch": 9.68, + "learning_rate": 1.618964387153157e-07, + "loss": 1.9089, + "step": 44338 + }, + { + "epoch": 9.68, + "learning_rate": 1.617871968538344e-07, + "loss": 1.0959, + "step": 44339 + }, + { + "epoch": 9.68, + "learning_rate": 1.616779549923531e-07, + "loss": 0.757, + "step": 44340 + }, + { + "epoch": 9.68, + "learning_rate": 1.6156871313087175e-07, + "loss": 0.4561, + "step": 44341 + }, + { + "epoch": 9.68, + "learning_rate": 1.6145947126939044e-07, + "loss": 1.1158, + "step": 44342 + }, + { + "epoch": 9.68, + "learning_rate": 1.6135022940790913e-07, + "loss": 1.9093, + "step": 44343 + }, + { + "epoch": 9.68, + "learning_rate": 1.612409875464278e-07, + "loss": 1.0772, + "step": 44344 + }, + { + "epoch": 9.68, + "learning_rate": 1.611317456849465e-07, + "loss": 0.1756, + "step": 44345 + }, + { + "epoch": 9.68, + "learning_rate": 1.6102250382346515e-07, + "loss": 0.5226, + "step": 44346 + }, + { + "epoch": 9.68, + "learning_rate": 1.6091326196198384e-07, + "loss": 0.5586, + "step": 44347 + }, + { + "epoch": 9.68, + "learning_rate": 1.6080402010050253e-07, + "loss": 2.2242, + "step": 44348 + }, + { + "epoch": 9.68, + "learning_rate": 1.606947782390212e-07, + "loss": 0.5202, + "step": 44349 + }, + { + "epoch": 9.68, + "learning_rate": 1.6058553637753989e-07, + "loss": 0.0399, + "step": 44350 + }, + { + "epoch": 9.68, + "learning_rate": 1.6047629451605855e-07, + "loss": 2.8793, + "step": 44351 + }, + { + "epoch": 9.68, + "learning_rate": 1.6036705265457724e-07, + "loss": 0.2035, + "step": 44352 + }, + { + "epoch": 9.68, + "learning_rate": 1.6025781079309593e-07, + "loss": 0.8524, + "step": 44353 + }, + { + "epoch": 9.68, + "learning_rate": 1.601485689316146e-07, + "loss": 0.1349, + "step": 44354 + }, + { + "epoch": 9.68, + "learning_rate": 1.6003932707013329e-07, + "loss": 1.1605, + "step": 44355 + }, + { + "epoch": 9.68, + "learning_rate": 1.5993008520865195e-07, + "loss": 0.15, + "step": 44356 + }, + { + "epoch": 9.68, + "learning_rate": 1.5982084334717064e-07, + "loss": 0.6813, + "step": 44357 + }, + { + "epoch": 9.68, + "learning_rate": 1.5971160148568936e-07, + "loss": 1.1436, + "step": 44358 + }, + { + "epoch": 9.68, + "learning_rate": 1.59602359624208e-07, + "loss": 0.7355, + "step": 44359 + }, + { + "epoch": 9.68, + "learning_rate": 1.5949311776272668e-07, + "loss": 1.3465, + "step": 44360 + }, + { + "epoch": 9.68, + "learning_rate": 1.5938387590124535e-07, + "loss": 1.3854, + "step": 44361 + }, + { + "epoch": 9.68, + "learning_rate": 1.5927463403976404e-07, + "loss": 1.2155, + "step": 44362 + }, + { + "epoch": 9.68, + "learning_rate": 1.5916539217828276e-07, + "loss": 1.0355, + "step": 44363 + }, + { + "epoch": 9.68, + "learning_rate": 1.590561503168014e-07, + "loss": 2.2528, + "step": 44364 + }, + { + "epoch": 9.68, + "learning_rate": 1.589469084553201e-07, + "loss": 0.788, + "step": 44365 + }, + { + "epoch": 9.68, + "learning_rate": 1.5883766659383875e-07, + "loss": 0.6811, + "step": 44366 + }, + { + "epoch": 9.68, + "learning_rate": 1.5872842473235746e-07, + "loss": 1.3092, + "step": 44367 + }, + { + "epoch": 9.68, + "learning_rate": 1.5861918287087615e-07, + "loss": 1.3237, + "step": 44368 + }, + { + "epoch": 9.68, + "learning_rate": 1.585099410093948e-07, + "loss": 1.8301, + "step": 44369 + }, + { + "epoch": 9.68, + "learning_rate": 1.584006991479135e-07, + "loss": 0.7694, + "step": 44370 + }, + { + "epoch": 9.68, + "learning_rate": 1.5829145728643215e-07, + "loss": 0.3239, + "step": 44371 + }, + { + "epoch": 9.68, + "learning_rate": 1.5818221542495086e-07, + "loss": 1.8728, + "step": 44372 + }, + { + "epoch": 9.68, + "learning_rate": 1.5807297356346955e-07, + "loss": 1.1783, + "step": 44373 + }, + { + "epoch": 9.68, + "learning_rate": 1.5796373170198822e-07, + "loss": 1.7823, + "step": 44374 + }, + { + "epoch": 9.68, + "learning_rate": 1.578544898405069e-07, + "loss": 1.7785, + "step": 44375 + }, + { + "epoch": 9.68, + "learning_rate": 1.5774524797902557e-07, + "loss": 0.4834, + "step": 44376 + }, + { + "epoch": 9.69, + "learning_rate": 1.5763600611754426e-07, + "loss": 0.9089, + "step": 44377 + }, + { + "epoch": 9.69, + "learning_rate": 1.5752676425606295e-07, + "loss": 1.1043, + "step": 44378 + }, + { + "epoch": 9.69, + "learning_rate": 1.5741752239458162e-07, + "loss": 2.2424, + "step": 44379 + }, + { + "epoch": 9.69, + "learning_rate": 1.573082805331003e-07, + "loss": 1.0933, + "step": 44380 + }, + { + "epoch": 9.69, + "learning_rate": 1.5719903867161897e-07, + "loss": 0.7166, + "step": 44381 + }, + { + "epoch": 9.69, + "learning_rate": 1.5708979681013766e-07, + "loss": 2.2168, + "step": 44382 + }, + { + "epoch": 9.69, + "learning_rate": 1.5698055494865635e-07, + "loss": 0.6029, + "step": 44383 + }, + { + "epoch": 9.69, + "learning_rate": 1.5687131308717502e-07, + "loss": 0.4379, + "step": 44384 + }, + { + "epoch": 9.69, + "learning_rate": 1.567620712256937e-07, + "loss": 1.3082, + "step": 44385 + }, + { + "epoch": 9.69, + "learning_rate": 1.566528293642124e-07, + "loss": 0.2545, + "step": 44386 + }, + { + "epoch": 9.69, + "learning_rate": 1.5654358750273106e-07, + "loss": 2.0293, + "step": 44387 + }, + { + "epoch": 9.69, + "learning_rate": 1.5643434564124975e-07, + "loss": 1.5011, + "step": 44388 + }, + { + "epoch": 9.69, + "learning_rate": 1.5632510377976841e-07, + "loss": 1.458, + "step": 44389 + }, + { + "epoch": 9.69, + "learning_rate": 1.562158619182871e-07, + "loss": 2.4959, + "step": 44390 + }, + { + "epoch": 9.69, + "learning_rate": 1.5610662005680577e-07, + "loss": 1.9864, + "step": 44391 + }, + { + "epoch": 9.69, + "learning_rate": 1.5599737819532446e-07, + "loss": 0.5247, + "step": 44392 + }, + { + "epoch": 9.69, + "learning_rate": 1.5588813633384315e-07, + "loss": 2.388, + "step": 44393 + }, + { + "epoch": 9.69, + "learning_rate": 1.5577889447236181e-07, + "loss": 0.3814, + "step": 44394 + }, + { + "epoch": 9.69, + "learning_rate": 1.556696526108805e-07, + "loss": 0.6275, + "step": 44395 + }, + { + "epoch": 9.69, + "learning_rate": 1.5556041074939917e-07, + "loss": 0.1481, + "step": 44396 + }, + { + "epoch": 9.69, + "learning_rate": 1.5545116888791786e-07, + "loss": 2.5047, + "step": 44397 + }, + { + "epoch": 9.69, + "learning_rate": 1.5534192702643655e-07, + "loss": 0.9012, + "step": 44398 + }, + { + "epoch": 9.69, + "learning_rate": 1.5523268516495524e-07, + "loss": 0.128, + "step": 44399 + }, + { + "epoch": 9.69, + "learning_rate": 1.551234433034739e-07, + "loss": 1.1347, + "step": 44400 + }, + { + "epoch": 9.69, + "learning_rate": 1.550142014419926e-07, + "loss": 1.1378, + "step": 44401 + }, + { + "epoch": 9.69, + "learning_rate": 1.5490495958051126e-07, + "loss": 0.986, + "step": 44402 + }, + { + "epoch": 9.69, + "learning_rate": 1.5479571771902995e-07, + "loss": 2.6984, + "step": 44403 + }, + { + "epoch": 9.69, + "learning_rate": 1.5468647585754864e-07, + "loss": 0.434, + "step": 44404 + }, + { + "epoch": 9.69, + "learning_rate": 1.545772339960673e-07, + "loss": 0.8248, + "step": 44405 + }, + { + "epoch": 9.69, + "learning_rate": 1.54467992134586e-07, + "loss": 0.071, + "step": 44406 + }, + { + "epoch": 9.69, + "learning_rate": 1.5435875027310466e-07, + "loss": 0.6745, + "step": 44407 + }, + { + "epoch": 9.69, + "learning_rate": 1.5424950841162335e-07, + "loss": 0.9319, + "step": 44408 + }, + { + "epoch": 9.69, + "learning_rate": 1.5414026655014204e-07, + "loss": 1.4766, + "step": 44409 + }, + { + "epoch": 9.69, + "learning_rate": 1.540310246886607e-07, + "loss": 0.754, + "step": 44410 + }, + { + "epoch": 9.69, + "learning_rate": 1.539217828271794e-07, + "loss": 1.8036, + "step": 44411 + }, + { + "epoch": 9.69, + "learning_rate": 1.5381254096569806e-07, + "loss": 0.6063, + "step": 44412 + }, + { + "epoch": 9.69, + "learning_rate": 1.5370329910421675e-07, + "loss": 1.9392, + "step": 44413 + }, + { + "epoch": 9.69, + "learning_rate": 1.5359405724273544e-07, + "loss": 0.5563, + "step": 44414 + }, + { + "epoch": 9.69, + "learning_rate": 1.534848153812541e-07, + "loss": 0.2158, + "step": 44415 + }, + { + "epoch": 9.69, + "learning_rate": 1.533755735197728e-07, + "loss": 1.0802, + "step": 44416 + }, + { + "epoch": 9.69, + "learning_rate": 1.5326633165829148e-07, + "loss": 0.6643, + "step": 44417 + }, + { + "epoch": 9.69, + "learning_rate": 1.5315708979681015e-07, + "loss": 0.6604, + "step": 44418 + }, + { + "epoch": 9.69, + "learning_rate": 1.5304784793532884e-07, + "loss": 0.0801, + "step": 44419 + }, + { + "epoch": 9.69, + "learning_rate": 1.529386060738475e-07, + "loss": 1.8111, + "step": 44420 + }, + { + "epoch": 9.69, + "learning_rate": 1.528293642123662e-07, + "loss": 0.8086, + "step": 44421 + }, + { + "epoch": 9.69, + "learning_rate": 1.5272012235088488e-07, + "loss": 1.8089, + "step": 44422 + }, + { + "epoch": 9.7, + "learning_rate": 1.5261088048940354e-07, + "loss": 0.6974, + "step": 44423 + }, + { + "epoch": 9.7, + "learning_rate": 1.5250163862792223e-07, + "loss": 0.8631, + "step": 44424 + }, + { + "epoch": 9.7, + "learning_rate": 1.523923967664409e-07, + "loss": 1.4273, + "step": 44425 + }, + { + "epoch": 9.7, + "learning_rate": 1.522831549049596e-07, + "loss": 0.4624, + "step": 44426 + }, + { + "epoch": 9.7, + "learning_rate": 1.5217391304347828e-07, + "loss": 0.9984, + "step": 44427 + }, + { + "epoch": 9.7, + "learning_rate": 1.5206467118199694e-07, + "loss": 0.4227, + "step": 44428 + }, + { + "epoch": 9.7, + "learning_rate": 1.5195542932051563e-07, + "loss": 0.3216, + "step": 44429 + }, + { + "epoch": 9.7, + "learning_rate": 1.518461874590343e-07, + "loss": 0.6539, + "step": 44430 + }, + { + "epoch": 9.7, + "learning_rate": 1.51736945597553e-07, + "loss": 0.1466, + "step": 44431 + }, + { + "epoch": 9.7, + "learning_rate": 1.5162770373607168e-07, + "loss": 0.9107, + "step": 44432 + }, + { + "epoch": 9.7, + "learning_rate": 1.5151846187459037e-07, + "loss": 0.379, + "step": 44433 + }, + { + "epoch": 9.7, + "learning_rate": 1.5140922001310903e-07, + "loss": 0.5643, + "step": 44434 + }, + { + "epoch": 9.7, + "learning_rate": 1.5129997815162772e-07, + "loss": 0.4479, + "step": 44435 + }, + { + "epoch": 9.7, + "learning_rate": 1.5119073629014641e-07, + "loss": 1.0278, + "step": 44436 + }, + { + "epoch": 9.7, + "learning_rate": 1.5108149442866508e-07, + "loss": 1.1717, + "step": 44437 + }, + { + "epoch": 9.7, + "learning_rate": 1.5097225256718377e-07, + "loss": 1.122, + "step": 44438 + }, + { + "epoch": 9.7, + "learning_rate": 1.5086301070570243e-07, + "loss": 0.5078, + "step": 44439 + }, + { + "epoch": 9.7, + "learning_rate": 1.5075376884422112e-07, + "loss": 0.6398, + "step": 44440 + }, + { + "epoch": 9.7, + "learning_rate": 1.506445269827398e-07, + "loss": 0.3635, + "step": 44441 + }, + { + "epoch": 9.7, + "learning_rate": 1.5053528512125848e-07, + "loss": 1.0385, + "step": 44442 + }, + { + "epoch": 9.7, + "learning_rate": 1.5042604325977717e-07, + "loss": 1.3085, + "step": 44443 + }, + { + "epoch": 9.7, + "learning_rate": 1.5031680139829583e-07, + "loss": 2.0437, + "step": 44444 + }, + { + "epoch": 9.7, + "learning_rate": 1.5020755953681452e-07, + "loss": 0.7058, + "step": 44445 + }, + { + "epoch": 9.7, + "learning_rate": 1.500983176753332e-07, + "loss": 2.0408, + "step": 44446 + }, + { + "epoch": 9.7, + "learning_rate": 1.4998907581385188e-07, + "loss": 0.3615, + "step": 44447 + }, + { + "epoch": 9.7, + "learning_rate": 1.4987983395237057e-07, + "loss": 1.1292, + "step": 44448 + }, + { + "epoch": 9.7, + "learning_rate": 1.4977059209088923e-07, + "loss": 0.571, + "step": 44449 + }, + { + "epoch": 9.7, + "learning_rate": 1.4966135022940792e-07, + "loss": 0.3617, + "step": 44450 + }, + { + "epoch": 9.7, + "learning_rate": 1.495521083679266e-07, + "loss": 0.175, + "step": 44451 + }, + { + "epoch": 9.7, + "learning_rate": 1.4944286650644527e-07, + "loss": 1.2079, + "step": 44452 + }, + { + "epoch": 9.7, + "learning_rate": 1.4933362464496396e-07, + "loss": 0.138, + "step": 44453 + }, + { + "epoch": 9.7, + "learning_rate": 1.4922438278348263e-07, + "loss": 0.487, + "step": 44454 + }, + { + "epoch": 9.7, + "learning_rate": 1.4911514092200132e-07, + "loss": 0.5618, + "step": 44455 + }, + { + "epoch": 9.7, + "learning_rate": 1.4900589906052e-07, + "loss": 1.912, + "step": 44456 + }, + { + "epoch": 9.7, + "learning_rate": 1.4889665719903867e-07, + "loss": 0.6277, + "step": 44457 + }, + { + "epoch": 9.7, + "learning_rate": 1.4878741533755736e-07, + "loss": 0.5975, + "step": 44458 + }, + { + "epoch": 9.7, + "learning_rate": 1.4867817347607603e-07, + "loss": 0.7508, + "step": 44459 + }, + { + "epoch": 9.7, + "learning_rate": 1.4856893161459474e-07, + "loss": 1.4943, + "step": 44460 + }, + { + "epoch": 9.7, + "learning_rate": 1.484596897531134e-07, + "loss": 1.6734, + "step": 44461 + }, + { + "epoch": 9.7, + "learning_rate": 1.4835044789163207e-07, + "loss": 0.7502, + "step": 44462 + }, + { + "epoch": 9.7, + "learning_rate": 1.4824120603015076e-07, + "loss": 0.7084, + "step": 44463 + }, + { + "epoch": 9.7, + "learning_rate": 1.4813196416866943e-07, + "loss": 1.3063, + "step": 44464 + }, + { + "epoch": 9.7, + "learning_rate": 1.4802272230718814e-07, + "loss": 1.1376, + "step": 44465 + }, + { + "epoch": 9.7, + "learning_rate": 1.479134804457068e-07, + "loss": 0.0859, + "step": 44466 + }, + { + "epoch": 9.7, + "learning_rate": 1.478042385842255e-07, + "loss": 0.0701, + "step": 44467 + }, + { + "epoch": 9.7, + "learning_rate": 1.4769499672274416e-07, + "loss": 2.1865, + "step": 44468 + }, + { + "epoch": 9.71, + "learning_rate": 1.4758575486126285e-07, + "loss": 2.0577, + "step": 44469 + }, + { + "epoch": 9.71, + "learning_rate": 1.4747651299978154e-07, + "loss": 0.0927, + "step": 44470 + }, + { + "epoch": 9.71, + "learning_rate": 1.473672711383002e-07, + "loss": 1.1945, + "step": 44471 + }, + { + "epoch": 9.71, + "learning_rate": 1.472580292768189e-07, + "loss": 2.222, + "step": 44472 + }, + { + "epoch": 9.71, + "learning_rate": 1.4714878741533756e-07, + "loss": 1.7767, + "step": 44473 + }, + { + "epoch": 9.71, + "learning_rate": 1.4703954555385625e-07, + "loss": 1.4018, + "step": 44474 + }, + { + "epoch": 9.71, + "learning_rate": 1.4693030369237494e-07, + "loss": 0.1508, + "step": 44475 + }, + { + "epoch": 9.71, + "learning_rate": 1.468210618308936e-07, + "loss": 1.4175, + "step": 44476 + }, + { + "epoch": 9.71, + "learning_rate": 1.467118199694123e-07, + "loss": 1.8587, + "step": 44477 + }, + { + "epoch": 9.71, + "learning_rate": 1.4660257810793096e-07, + "loss": 0.4842, + "step": 44478 + }, + { + "epoch": 9.71, + "learning_rate": 1.4649333624644965e-07, + "loss": 0.7133, + "step": 44479 + }, + { + "epoch": 9.71, + "learning_rate": 1.4638409438496834e-07, + "loss": 0.749, + "step": 44480 + }, + { + "epoch": 9.71, + "learning_rate": 1.46274852523487e-07, + "loss": 0.7712, + "step": 44481 + }, + { + "epoch": 9.71, + "learning_rate": 1.461656106620057e-07, + "loss": 1.1736, + "step": 44482 + }, + { + "epoch": 9.71, + "learning_rate": 1.4605636880052436e-07, + "loss": 0.4709, + "step": 44483 + }, + { + "epoch": 9.71, + "learning_rate": 1.4594712693904305e-07, + "loss": 0.5494, + "step": 44484 + }, + { + "epoch": 9.71, + "learning_rate": 1.4583788507756174e-07, + "loss": 1.2402, + "step": 44485 + }, + { + "epoch": 9.71, + "learning_rate": 1.457286432160804e-07, + "loss": 0.1283, + "step": 44486 + }, + { + "epoch": 9.71, + "learning_rate": 1.456194013545991e-07, + "loss": 0.7526, + "step": 44487 + }, + { + "epoch": 9.71, + "learning_rate": 1.4551015949311776e-07, + "loss": 1.8363, + "step": 44488 + }, + { + "epoch": 9.71, + "learning_rate": 1.4540091763163645e-07, + "loss": 2.5244, + "step": 44489 + }, + { + "epoch": 9.71, + "learning_rate": 1.4529167577015514e-07, + "loss": 0.665, + "step": 44490 + }, + { + "epoch": 9.71, + "learning_rate": 1.451824339086738e-07, + "loss": 1.1662, + "step": 44491 + }, + { + "epoch": 9.71, + "learning_rate": 1.450731920471925e-07, + "loss": 0.7023, + "step": 44492 + }, + { + "epoch": 9.71, + "learning_rate": 1.4496395018571116e-07, + "loss": 0.5399, + "step": 44493 + }, + { + "epoch": 9.71, + "learning_rate": 1.4485470832422987e-07, + "loss": 1.4788, + "step": 44494 + }, + { + "epoch": 9.71, + "learning_rate": 1.4474546646274854e-07, + "loss": 1.3377, + "step": 44495 + }, + { + "epoch": 9.71, + "learning_rate": 1.4463622460126723e-07, + "loss": 2.0903, + "step": 44496 + }, + { + "epoch": 9.71, + "learning_rate": 1.445269827397859e-07, + "loss": 1.8787, + "step": 44497 + }, + { + "epoch": 9.71, + "learning_rate": 1.4441774087830458e-07, + "loss": 0.6239, + "step": 44498 + }, + { + "epoch": 9.71, + "learning_rate": 1.4430849901682327e-07, + "loss": 0.6303, + "step": 44499 + }, + { + "epoch": 9.71, + "learning_rate": 1.4419925715534194e-07, + "loss": 0.7584, + "step": 44500 + }, + { + "epoch": 9.71, + "learning_rate": 1.4409001529386063e-07, + "loss": 0.0888, + "step": 44501 + }, + { + "epoch": 9.71, + "learning_rate": 1.439807734323793e-07, + "loss": 1.3573, + "step": 44502 + }, + { + "epoch": 9.71, + "learning_rate": 1.4387153157089798e-07, + "loss": 0.3733, + "step": 44503 + }, + { + "epoch": 9.71, + "learning_rate": 1.4376228970941667e-07, + "loss": 0.3566, + "step": 44504 + }, + { + "epoch": 9.71, + "learning_rate": 1.4365304784793534e-07, + "loss": 1.3442, + "step": 44505 + }, + { + "epoch": 9.71, + "learning_rate": 1.4354380598645403e-07, + "loss": 1.7415, + "step": 44506 + }, + { + "epoch": 9.71, + "learning_rate": 1.434345641249727e-07, + "loss": 1.7913, + "step": 44507 + }, + { + "epoch": 9.71, + "learning_rate": 1.4332532226349138e-07, + "loss": 0.7785, + "step": 44508 + }, + { + "epoch": 9.71, + "learning_rate": 1.4321608040201007e-07, + "loss": 0.7581, + "step": 44509 + }, + { + "epoch": 9.71, + "learning_rate": 1.4310683854052874e-07, + "loss": 0.2171, + "step": 44510 + }, + { + "epoch": 9.71, + "learning_rate": 1.4299759667904743e-07, + "loss": 0.5367, + "step": 44511 + }, + { + "epoch": 9.71, + "learning_rate": 1.428883548175661e-07, + "loss": 0.8226, + "step": 44512 + }, + { + "epoch": 9.71, + "learning_rate": 1.4277911295608478e-07, + "loss": 0.1214, + "step": 44513 + }, + { + "epoch": 9.71, + "learning_rate": 1.4266987109460347e-07, + "loss": 1.732, + "step": 44514 + }, + { + "epoch": 9.72, + "learning_rate": 1.4256062923312213e-07, + "loss": 0.0624, + "step": 44515 + }, + { + "epoch": 9.72, + "learning_rate": 1.4245138737164082e-07, + "loss": 1.0453, + "step": 44516 + }, + { + "epoch": 9.72, + "learning_rate": 1.4234214551015952e-07, + "loss": 1.7694, + "step": 44517 + }, + { + "epoch": 9.72, + "learning_rate": 1.4223290364867818e-07, + "loss": 1.2436, + "step": 44518 + }, + { + "epoch": 9.72, + "learning_rate": 1.4212366178719687e-07, + "loss": 1.1818, + "step": 44519 + }, + { + "epoch": 9.72, + "learning_rate": 1.4201441992571553e-07, + "loss": 2.3292, + "step": 44520 + }, + { + "epoch": 9.72, + "learning_rate": 1.4190517806423422e-07, + "loss": 0.5626, + "step": 44521 + }, + { + "epoch": 9.72, + "learning_rate": 1.4179593620275291e-07, + "loss": 1.1612, + "step": 44522 + }, + { + "epoch": 9.72, + "learning_rate": 1.4168669434127158e-07, + "loss": 1.0003, + "step": 44523 + }, + { + "epoch": 9.72, + "learning_rate": 1.4157745247979027e-07, + "loss": 1.1724, + "step": 44524 + }, + { + "epoch": 9.72, + "learning_rate": 1.4146821061830893e-07, + "loss": 1.0556, + "step": 44525 + }, + { + "epoch": 9.72, + "learning_rate": 1.4135896875682762e-07, + "loss": 0.8098, + "step": 44526 + }, + { + "epoch": 9.72, + "learning_rate": 1.412497268953463e-07, + "loss": 0.9566, + "step": 44527 + }, + { + "epoch": 9.72, + "learning_rate": 1.41140485033865e-07, + "loss": 1.8267, + "step": 44528 + }, + { + "epoch": 9.72, + "learning_rate": 1.4103124317238367e-07, + "loss": 0.1118, + "step": 44529 + }, + { + "epoch": 9.72, + "learning_rate": 1.4092200131090236e-07, + "loss": 0.73, + "step": 44530 + }, + { + "epoch": 9.72, + "learning_rate": 1.4081275944942102e-07, + "loss": 1.0463, + "step": 44531 + }, + { + "epoch": 9.72, + "learning_rate": 1.407035175879397e-07, + "loss": 2.3531, + "step": 44532 + }, + { + "epoch": 9.72, + "learning_rate": 1.405942757264584e-07, + "loss": 0.6203, + "step": 44533 + }, + { + "epoch": 9.72, + "learning_rate": 1.4048503386497707e-07, + "loss": 0.6773, + "step": 44534 + }, + { + "epoch": 9.72, + "learning_rate": 1.4037579200349576e-07, + "loss": 0.73, + "step": 44535 + }, + { + "epoch": 9.72, + "learning_rate": 1.4026655014201445e-07, + "loss": 0.7329, + "step": 44536 + }, + { + "epoch": 9.72, + "learning_rate": 1.401573082805331e-07, + "loss": 0.5685, + "step": 44537 + }, + { + "epoch": 9.72, + "learning_rate": 1.400480664190518e-07, + "loss": 1.126, + "step": 44538 + }, + { + "epoch": 9.72, + "learning_rate": 1.3993882455757047e-07, + "loss": 2.1678, + "step": 44539 + }, + { + "epoch": 9.72, + "learning_rate": 1.3982958269608916e-07, + "loss": 2.4345, + "step": 44540 + }, + { + "epoch": 9.72, + "learning_rate": 1.3972034083460785e-07, + "loss": 1.4905, + "step": 44541 + }, + { + "epoch": 9.72, + "learning_rate": 1.396110989731265e-07, + "loss": 0.7599, + "step": 44542 + }, + { + "epoch": 9.72, + "learning_rate": 1.395018571116452e-07, + "loss": 1.7052, + "step": 44543 + }, + { + "epoch": 9.72, + "learning_rate": 1.3939261525016386e-07, + "loss": 2.2781, + "step": 44544 + }, + { + "epoch": 9.72, + "learning_rate": 1.3928337338868256e-07, + "loss": 1.5142, + "step": 44545 + }, + { + "epoch": 9.72, + "learning_rate": 1.3917413152720125e-07, + "loss": 1.4318, + "step": 44546 + }, + { + "epoch": 9.72, + "learning_rate": 1.390648896657199e-07, + "loss": 1.4189, + "step": 44547 + }, + { + "epoch": 9.72, + "learning_rate": 1.389556478042386e-07, + "loss": 0.8197, + "step": 44548 + }, + { + "epoch": 9.72, + "learning_rate": 1.3884640594275726e-07, + "loss": 0.5489, + "step": 44549 + }, + { + "epoch": 9.72, + "learning_rate": 1.3873716408127595e-07, + "loss": 0.6599, + "step": 44550 + }, + { + "epoch": 9.72, + "learning_rate": 1.3862792221979464e-07, + "loss": 0.1309, + "step": 44551 + }, + { + "epoch": 9.72, + "learning_rate": 1.385186803583133e-07, + "loss": 0.4702, + "step": 44552 + }, + { + "epoch": 9.72, + "learning_rate": 1.38409438496832e-07, + "loss": 0.777, + "step": 44553 + }, + { + "epoch": 9.72, + "learning_rate": 1.3830019663535066e-07, + "loss": 0.0943, + "step": 44554 + }, + { + "epoch": 9.72, + "learning_rate": 1.3819095477386938e-07, + "loss": 0.732, + "step": 44555 + }, + { + "epoch": 9.72, + "learning_rate": 1.3808171291238804e-07, + "loss": 0.0439, + "step": 44556 + }, + { + "epoch": 9.72, + "learning_rate": 1.379724710509067e-07, + "loss": 0.2461, + "step": 44557 + }, + { + "epoch": 9.72, + "learning_rate": 1.378632291894254e-07, + "loss": 0.9369, + "step": 44558 + }, + { + "epoch": 9.72, + "learning_rate": 1.3775398732794406e-07, + "loss": 1.209, + "step": 44559 + }, + { + "epoch": 9.73, + "learning_rate": 1.3764474546646278e-07, + "loss": 1.1866, + "step": 44560 + }, + { + "epoch": 9.73, + "learning_rate": 1.3753550360498144e-07, + "loss": 0.4135, + "step": 44561 + }, + { + "epoch": 9.73, + "learning_rate": 1.3742626174350013e-07, + "loss": 0.7386, + "step": 44562 + }, + { + "epoch": 9.73, + "learning_rate": 1.373170198820188e-07, + "loss": 2.4356, + "step": 44563 + }, + { + "epoch": 9.73, + "learning_rate": 1.372077780205375e-07, + "loss": 0.068, + "step": 44564 + }, + { + "epoch": 9.73, + "learning_rate": 1.3709853615905618e-07, + "loss": 0.9562, + "step": 44565 + }, + { + "epoch": 9.73, + "learning_rate": 1.3698929429757484e-07, + "loss": 0.6952, + "step": 44566 + }, + { + "epoch": 9.73, + "learning_rate": 1.3688005243609353e-07, + "loss": 0.9441, + "step": 44567 + }, + { + "epoch": 9.73, + "learning_rate": 1.367708105746122e-07, + "loss": 1.4051, + "step": 44568 + }, + { + "epoch": 9.73, + "learning_rate": 1.3666156871313089e-07, + "loss": 0.606, + "step": 44569 + }, + { + "epoch": 9.73, + "learning_rate": 1.3655232685164958e-07, + "loss": 1.0785, + "step": 44570 + }, + { + "epoch": 9.73, + "learning_rate": 1.3644308499016824e-07, + "loss": 0.2628, + "step": 44571 + }, + { + "epoch": 9.73, + "learning_rate": 1.3633384312868693e-07, + "loss": 0.6455, + "step": 44572 + }, + { + "epoch": 9.73, + "learning_rate": 1.362246012672056e-07, + "loss": 0.5237, + "step": 44573 + }, + { + "epoch": 9.73, + "learning_rate": 1.3611535940572429e-07, + "loss": 1.7252, + "step": 44574 + }, + { + "epoch": 9.73, + "learning_rate": 1.3600611754424298e-07, + "loss": 0.2401, + "step": 44575 + }, + { + "epoch": 9.73, + "learning_rate": 1.3589687568276164e-07, + "loss": 0.5297, + "step": 44576 + }, + { + "epoch": 9.73, + "learning_rate": 1.3578763382128033e-07, + "loss": 1.5957, + "step": 44577 + }, + { + "epoch": 9.73, + "learning_rate": 1.35678391959799e-07, + "loss": 1.3995, + "step": 44578 + }, + { + "epoch": 9.73, + "learning_rate": 1.3556915009831768e-07, + "loss": 1.1661, + "step": 44579 + }, + { + "epoch": 9.73, + "learning_rate": 1.3545990823683637e-07, + "loss": 0.6979, + "step": 44580 + }, + { + "epoch": 9.73, + "learning_rate": 1.3535066637535504e-07, + "loss": 2.813, + "step": 44581 + }, + { + "epoch": 9.73, + "learning_rate": 1.3524142451387373e-07, + "loss": 0.3221, + "step": 44582 + }, + { + "epoch": 9.73, + "learning_rate": 1.351321826523924e-07, + "loss": 2.4534, + "step": 44583 + }, + { + "epoch": 9.73, + "learning_rate": 1.3502294079091108e-07, + "loss": 1.1942, + "step": 44584 + }, + { + "epoch": 9.73, + "learning_rate": 1.3491369892942977e-07, + "loss": 1.2614, + "step": 44585 + }, + { + "epoch": 9.73, + "learning_rate": 1.3480445706794844e-07, + "loss": 2.0632, + "step": 44586 + }, + { + "epoch": 9.73, + "learning_rate": 1.3469521520646713e-07, + "loss": 1.7377, + "step": 44587 + }, + { + "epoch": 9.73, + "learning_rate": 1.345859733449858e-07, + "loss": 1.0933, + "step": 44588 + }, + { + "epoch": 9.73, + "learning_rate": 1.344767314835045e-07, + "loss": 1.871, + "step": 44589 + }, + { + "epoch": 9.73, + "learning_rate": 1.3436748962202317e-07, + "loss": 0.6431, + "step": 44590 + }, + { + "epoch": 9.73, + "learning_rate": 1.3425824776054184e-07, + "loss": 0.957, + "step": 44591 + }, + { + "epoch": 9.73, + "learning_rate": 1.3414900589906053e-07, + "loss": 3.3823, + "step": 44592 + }, + { + "epoch": 9.73, + "learning_rate": 1.340397640375792e-07, + "loss": 1.1776, + "step": 44593 + }, + { + "epoch": 9.73, + "learning_rate": 1.339305221760979e-07, + "loss": 1.4607, + "step": 44594 + }, + { + "epoch": 9.73, + "learning_rate": 1.3382128031461657e-07, + "loss": 0.6829, + "step": 44595 + }, + { + "epoch": 9.73, + "learning_rate": 1.3371203845313526e-07, + "loss": 0.4694, + "step": 44596 + }, + { + "epoch": 9.73, + "learning_rate": 1.3360279659165393e-07, + "loss": 1.0941, + "step": 44597 + }, + { + "epoch": 9.73, + "learning_rate": 1.3349355473017262e-07, + "loss": 1.6791, + "step": 44598 + }, + { + "epoch": 9.73, + "learning_rate": 1.333843128686913e-07, + "loss": 1.5933, + "step": 44599 + }, + { + "epoch": 9.73, + "learning_rate": 1.3327507100720997e-07, + "loss": 0.537, + "step": 44600 + }, + { + "epoch": 9.73, + "learning_rate": 1.3316582914572866e-07, + "loss": 0.1497, + "step": 44601 + }, + { + "epoch": 9.73, + "learning_rate": 1.3305658728424733e-07, + "loss": 0.7184, + "step": 44602 + }, + { + "epoch": 9.73, + "learning_rate": 1.3294734542276602e-07, + "loss": 1.3344, + "step": 44603 + }, + { + "epoch": 9.73, + "learning_rate": 1.328381035612847e-07, + "loss": 0.098, + "step": 44604 + }, + { + "epoch": 9.73, + "learning_rate": 1.3272886169980337e-07, + "loss": 1.5164, + "step": 44605 + }, + { + "epoch": 9.74, + "learning_rate": 1.3261961983832206e-07, + "loss": 0.8866, + "step": 44606 + }, + { + "epoch": 9.74, + "learning_rate": 1.3251037797684072e-07, + "loss": 0.6392, + "step": 44607 + }, + { + "epoch": 9.74, + "learning_rate": 1.3240113611535941e-07, + "loss": 0.8532, + "step": 44608 + }, + { + "epoch": 9.74, + "learning_rate": 1.322918942538781e-07, + "loss": 2.8046, + "step": 44609 + }, + { + "epoch": 9.74, + "learning_rate": 1.3218265239239677e-07, + "loss": 1.4591, + "step": 44610 + }, + { + "epoch": 9.74, + "learning_rate": 1.3207341053091546e-07, + "loss": 0.5476, + "step": 44611 + }, + { + "epoch": 9.74, + "learning_rate": 1.3196416866943412e-07, + "loss": 1.8246, + "step": 44612 + }, + { + "epoch": 9.74, + "learning_rate": 1.3185492680795281e-07, + "loss": 1.2929, + "step": 44613 + }, + { + "epoch": 9.74, + "learning_rate": 1.317456849464715e-07, + "loss": 1.0436, + "step": 44614 + }, + { + "epoch": 9.74, + "learning_rate": 1.3163644308499017e-07, + "loss": 0.5561, + "step": 44615 + }, + { + "epoch": 9.74, + "learning_rate": 1.3152720122350886e-07, + "loss": 1.8324, + "step": 44616 + }, + { + "epoch": 9.74, + "learning_rate": 1.3141795936202755e-07, + "loss": 1.2424, + "step": 44617 + }, + { + "epoch": 9.74, + "learning_rate": 1.313087175005462e-07, + "loss": 3.3869, + "step": 44618 + }, + { + "epoch": 9.74, + "learning_rate": 1.311994756390649e-07, + "loss": 1.6609, + "step": 44619 + }, + { + "epoch": 9.74, + "learning_rate": 1.3109023377758357e-07, + "loss": 0.9719, + "step": 44620 + }, + { + "epoch": 9.74, + "learning_rate": 1.3098099191610226e-07, + "loss": 0.4327, + "step": 44621 + }, + { + "epoch": 9.74, + "learning_rate": 1.3087175005462095e-07, + "loss": 0.7728, + "step": 44622 + }, + { + "epoch": 9.74, + "learning_rate": 1.3076250819313964e-07, + "loss": 0.7278, + "step": 44623 + }, + { + "epoch": 9.74, + "learning_rate": 1.306532663316583e-07, + "loss": 1.1024, + "step": 44624 + }, + { + "epoch": 9.74, + "learning_rate": 1.3054402447017697e-07, + "loss": 0.1934, + "step": 44625 + }, + { + "epoch": 9.74, + "learning_rate": 1.3043478260869566e-07, + "loss": 0.3744, + "step": 44626 + }, + { + "epoch": 9.74, + "learning_rate": 1.3032554074721435e-07, + "loss": 0.8868, + "step": 44627 + }, + { + "epoch": 9.74, + "learning_rate": 1.3021629888573304e-07, + "loss": 2.2949, + "step": 44628 + }, + { + "epoch": 9.74, + "learning_rate": 1.301070570242517e-07, + "loss": 3.426, + "step": 44629 + }, + { + "epoch": 9.74, + "learning_rate": 1.299978151627704e-07, + "loss": 4.1783, + "step": 44630 + }, + { + "epoch": 9.74, + "learning_rate": 1.2988857330128906e-07, + "loss": 0.2039, + "step": 44631 + }, + { + "epoch": 9.74, + "learning_rate": 1.2977933143980775e-07, + "loss": 1.0378, + "step": 44632 + }, + { + "epoch": 9.74, + "learning_rate": 1.2967008957832644e-07, + "loss": 2.1711, + "step": 44633 + }, + { + "epoch": 9.74, + "learning_rate": 1.295608477168451e-07, + "loss": 0.0758, + "step": 44634 + }, + { + "epoch": 9.74, + "learning_rate": 1.294516058553638e-07, + "loss": 0.7716, + "step": 44635 + }, + { + "epoch": 9.74, + "learning_rate": 1.2934236399388248e-07, + "loss": 1.7098, + "step": 44636 + }, + { + "epoch": 9.74, + "learning_rate": 1.2923312213240115e-07, + "loss": 0.3861, + "step": 44637 + }, + { + "epoch": 9.74, + "learning_rate": 1.2912388027091984e-07, + "loss": 0.6887, + "step": 44638 + }, + { + "epoch": 9.74, + "learning_rate": 1.290146384094385e-07, + "loss": 2.0025, + "step": 44639 + }, + { + "epoch": 9.74, + "learning_rate": 1.289053965479572e-07, + "loss": 0.2357, + "step": 44640 + }, + { + "epoch": 9.74, + "learning_rate": 1.2879615468647588e-07, + "loss": 1.3771, + "step": 44641 + }, + { + "epoch": 9.74, + "learning_rate": 1.2868691282499454e-07, + "loss": 0.4131, + "step": 44642 + }, + { + "epoch": 9.74, + "learning_rate": 1.2857767096351323e-07, + "loss": 1.3817, + "step": 44643 + }, + { + "epoch": 9.74, + "learning_rate": 1.284684291020319e-07, + "loss": 1.0589, + "step": 44644 + }, + { + "epoch": 9.74, + "learning_rate": 1.283591872405506e-07, + "loss": 0.3213, + "step": 44645 + }, + { + "epoch": 9.74, + "learning_rate": 1.2824994537906928e-07, + "loss": 0.3811, + "step": 44646 + }, + { + "epoch": 9.74, + "learning_rate": 1.2814070351758794e-07, + "loss": 1.8396, + "step": 44647 + }, + { + "epoch": 9.74, + "learning_rate": 1.2803146165610663e-07, + "loss": 0.2954, + "step": 44648 + }, + { + "epoch": 9.74, + "learning_rate": 1.279222197946253e-07, + "loss": 0.9347, + "step": 44649 + }, + { + "epoch": 9.74, + "learning_rate": 1.27812977933144e-07, + "loss": 2.4567, + "step": 44650 + }, + { + "epoch": 9.74, + "learning_rate": 1.2770373607166268e-07, + "loss": 1.2196, + "step": 44651 + }, + { + "epoch": 9.75, + "learning_rate": 1.2759449421018134e-07, + "loss": 0.9889, + "step": 44652 + }, + { + "epoch": 9.75, + "learning_rate": 1.2748525234870003e-07, + "loss": 1.5545, + "step": 44653 + }, + { + "epoch": 9.75, + "learning_rate": 1.273760104872187e-07, + "loss": 0.9046, + "step": 44654 + }, + { + "epoch": 9.75, + "learning_rate": 1.272667686257374e-07, + "loss": 1.0687, + "step": 44655 + }, + { + "epoch": 9.75, + "learning_rate": 1.2715752676425608e-07, + "loss": 0.2645, + "step": 44656 + }, + { + "epoch": 9.75, + "learning_rate": 1.2704828490277477e-07, + "loss": 0.4871, + "step": 44657 + }, + { + "epoch": 9.75, + "learning_rate": 1.2693904304129343e-07, + "loss": 1.6633, + "step": 44658 + }, + { + "epoch": 9.75, + "learning_rate": 1.268298011798121e-07, + "loss": 0.3782, + "step": 44659 + }, + { + "epoch": 9.75, + "learning_rate": 1.267205593183308e-07, + "loss": 0.4915, + "step": 44660 + }, + { + "epoch": 9.75, + "learning_rate": 1.2661131745684948e-07, + "loss": 1.0584, + "step": 44661 + }, + { + "epoch": 9.75, + "learning_rate": 1.2650207559536817e-07, + "loss": 0.7319, + "step": 44662 + }, + { + "epoch": 9.75, + "learning_rate": 1.2639283373388683e-07, + "loss": 1.2162, + "step": 44663 + }, + { + "epoch": 9.75, + "learning_rate": 1.2628359187240552e-07, + "loss": 0.7432, + "step": 44664 + }, + { + "epoch": 9.75, + "learning_rate": 1.261743500109242e-07, + "loss": 0.991, + "step": 44665 + }, + { + "epoch": 9.75, + "learning_rate": 1.2606510814944288e-07, + "loss": 0.6766, + "step": 44666 + }, + { + "epoch": 9.75, + "learning_rate": 1.2595586628796157e-07, + "loss": 0.6019, + "step": 44667 + }, + { + "epoch": 9.75, + "learning_rate": 1.2584662442648023e-07, + "loss": 0.1702, + "step": 44668 + }, + { + "epoch": 9.75, + "learning_rate": 1.2573738256499892e-07, + "loss": 1.137, + "step": 44669 + }, + { + "epoch": 9.75, + "learning_rate": 1.256281407035176e-07, + "loss": 0.6668, + "step": 44670 + }, + { + "epoch": 9.75, + "learning_rate": 1.2551889884203627e-07, + "loss": 0.9198, + "step": 44671 + }, + { + "epoch": 9.75, + "learning_rate": 1.2540965698055496e-07, + "loss": 1.2901, + "step": 44672 + }, + { + "epoch": 9.75, + "learning_rate": 1.2530041511907363e-07, + "loss": 0.958, + "step": 44673 + }, + { + "epoch": 9.75, + "learning_rate": 1.2519117325759232e-07, + "loss": 2.2498, + "step": 44674 + }, + { + "epoch": 9.75, + "learning_rate": 1.25081931396111e-07, + "loss": 1.2354, + "step": 44675 + }, + { + "epoch": 9.75, + "learning_rate": 1.2497268953462967e-07, + "loss": 0.3969, + "step": 44676 + }, + { + "epoch": 9.75, + "learning_rate": 1.2486344767314836e-07, + "loss": 1.2832, + "step": 44677 + }, + { + "epoch": 9.75, + "learning_rate": 1.2475420581166703e-07, + "loss": 1.299, + "step": 44678 + }, + { + "epoch": 9.75, + "learning_rate": 1.2464496395018572e-07, + "loss": 1.3876, + "step": 44679 + }, + { + "epoch": 9.75, + "learning_rate": 1.245357220887044e-07, + "loss": 0.8429, + "step": 44680 + }, + { + "epoch": 9.75, + "learning_rate": 1.2442648022722307e-07, + "loss": 0.0744, + "step": 44681 + }, + { + "epoch": 9.75, + "learning_rate": 1.2431723836574176e-07, + "loss": 2.287, + "step": 44682 + }, + { + "epoch": 9.75, + "learning_rate": 1.2420799650426043e-07, + "loss": 0.3582, + "step": 44683 + }, + { + "epoch": 9.75, + "learning_rate": 1.2409875464277912e-07, + "loss": 1.5932, + "step": 44684 + }, + { + "epoch": 9.75, + "learning_rate": 1.239895127812978e-07, + "loss": 0.1492, + "step": 44685 + }, + { + "epoch": 9.75, + "learning_rate": 1.2388027091981647e-07, + "loss": 1.0473, + "step": 44686 + }, + { + "epoch": 9.75, + "learning_rate": 1.2377102905833516e-07, + "loss": 0.041, + "step": 44687 + }, + { + "epoch": 9.75, + "learning_rate": 1.2366178719685383e-07, + "loss": 0.6709, + "step": 44688 + }, + { + "epoch": 9.75, + "learning_rate": 1.2355254533537254e-07, + "loss": 0.429, + "step": 44689 + }, + { + "epoch": 9.75, + "learning_rate": 1.234433034738912e-07, + "loss": 2.1906, + "step": 44690 + }, + { + "epoch": 9.75, + "learning_rate": 1.233340616124099e-07, + "loss": 1.0442, + "step": 44691 + }, + { + "epoch": 9.75, + "learning_rate": 1.2322481975092856e-07, + "loss": 1.6597, + "step": 44692 + }, + { + "epoch": 9.75, + "learning_rate": 1.2311557788944723e-07, + "loss": 1.0731, + "step": 44693 + }, + { + "epoch": 9.75, + "learning_rate": 1.2300633602796594e-07, + "loss": 0.1477, + "step": 44694 + }, + { + "epoch": 9.75, + "learning_rate": 1.228970941664846e-07, + "loss": 1.0655, + "step": 44695 + }, + { + "epoch": 9.75, + "learning_rate": 1.227878523050033e-07, + "loss": 0.76, + "step": 44696 + }, + { + "epoch": 9.75, + "learning_rate": 1.2267861044352196e-07, + "loss": 1.8945, + "step": 44697 + }, + { + "epoch": 9.76, + "learning_rate": 1.2256936858204065e-07, + "loss": 0.8127, + "step": 44698 + }, + { + "epoch": 9.76, + "learning_rate": 1.2246012672055934e-07, + "loss": 1.5327, + "step": 44699 + }, + { + "epoch": 9.76, + "learning_rate": 1.22350884859078e-07, + "loss": 0.592, + "step": 44700 + }, + { + "epoch": 9.76, + "learning_rate": 1.222416429975967e-07, + "loss": 3.3288, + "step": 44701 + }, + { + "epoch": 9.76, + "learning_rate": 1.2213240113611536e-07, + "loss": 1.7129, + "step": 44702 + }, + { + "epoch": 9.76, + "learning_rate": 1.2202315927463405e-07, + "loss": 1.5924, + "step": 44703 + }, + { + "epoch": 9.76, + "learning_rate": 1.2191391741315274e-07, + "loss": 0.1622, + "step": 44704 + }, + { + "epoch": 9.76, + "learning_rate": 1.218046755516714e-07, + "loss": 0.9833, + "step": 44705 + }, + { + "epoch": 9.76, + "learning_rate": 1.216954336901901e-07, + "loss": 0.5706, + "step": 44706 + }, + { + "epoch": 9.76, + "learning_rate": 1.2158619182870876e-07, + "loss": 2.3107, + "step": 44707 + }, + { + "epoch": 9.76, + "learning_rate": 1.2147694996722745e-07, + "loss": 1.2346, + "step": 44708 + }, + { + "epoch": 9.76, + "learning_rate": 1.2136770810574614e-07, + "loss": 0.9569, + "step": 44709 + }, + { + "epoch": 9.76, + "learning_rate": 1.212584662442648e-07, + "loss": 0.7914, + "step": 44710 + }, + { + "epoch": 9.76, + "learning_rate": 1.211492243827835e-07, + "loss": 0.1766, + "step": 44711 + }, + { + "epoch": 9.76, + "learning_rate": 1.2103998252130216e-07, + "loss": 1.1413, + "step": 44712 + }, + { + "epoch": 9.76, + "learning_rate": 1.2093074065982085e-07, + "loss": 1.2269, + "step": 44713 + }, + { + "epoch": 9.76, + "learning_rate": 1.2082149879833954e-07, + "loss": 2.258, + "step": 44714 + }, + { + "epoch": 9.76, + "learning_rate": 1.207122569368582e-07, + "loss": 2.169, + "step": 44715 + }, + { + "epoch": 9.76, + "learning_rate": 1.206030150753769e-07, + "loss": 1.8817, + "step": 44716 + }, + { + "epoch": 9.76, + "learning_rate": 1.2049377321389558e-07, + "loss": 0.0477, + "step": 44717 + }, + { + "epoch": 9.76, + "learning_rate": 1.2038453135241425e-07, + "loss": 0.6148, + "step": 44718 + }, + { + "epoch": 9.76, + "learning_rate": 1.2027528949093294e-07, + "loss": 0.5725, + "step": 44719 + }, + { + "epoch": 9.76, + "learning_rate": 1.201660476294516e-07, + "loss": 2.2956, + "step": 44720 + }, + { + "epoch": 9.76, + "learning_rate": 1.200568057679703e-07, + "loss": 2.0437, + "step": 44721 + }, + { + "epoch": 9.76, + "learning_rate": 1.1994756390648898e-07, + "loss": 0.5176, + "step": 44722 + }, + { + "epoch": 9.76, + "learning_rate": 1.1983832204500767e-07, + "loss": 0.1727, + "step": 44723 + }, + { + "epoch": 9.76, + "learning_rate": 1.1972908018352634e-07, + "loss": 3.6256, + "step": 44724 + }, + { + "epoch": 9.76, + "learning_rate": 1.1961983832204503e-07, + "loss": 2.1069, + "step": 44725 + }, + { + "epoch": 9.76, + "learning_rate": 1.195105964605637e-07, + "loss": 0.2122, + "step": 44726 + }, + { + "epoch": 9.76, + "learning_rate": 1.1940135459908238e-07, + "loss": 0.4682, + "step": 44727 + }, + { + "epoch": 9.76, + "learning_rate": 1.1929211273760107e-07, + "loss": 1.5245, + "step": 44728 + }, + { + "epoch": 9.76, + "learning_rate": 1.1918287087611974e-07, + "loss": 1.6897, + "step": 44729 + }, + { + "epoch": 9.76, + "learning_rate": 1.1907362901463841e-07, + "loss": 1.4347, + "step": 44730 + }, + { + "epoch": 9.76, + "learning_rate": 1.1896438715315709e-07, + "loss": 0.0704, + "step": 44731 + }, + { + "epoch": 9.76, + "learning_rate": 1.1885514529167578e-07, + "loss": 0.5127, + "step": 44732 + }, + { + "epoch": 9.76, + "learning_rate": 1.1874590343019446e-07, + "loss": 0.8684, + "step": 44733 + }, + { + "epoch": 9.76, + "learning_rate": 1.1863666156871313e-07, + "loss": 1.163, + "step": 44734 + }, + { + "epoch": 9.76, + "learning_rate": 1.1852741970723181e-07, + "loss": 0.5537, + "step": 44735 + }, + { + "epoch": 9.76, + "learning_rate": 1.1841817784575049e-07, + "loss": 1.5761, + "step": 44736 + }, + { + "epoch": 9.76, + "learning_rate": 1.1830893598426919e-07, + "loss": 0.6217, + "step": 44737 + }, + { + "epoch": 9.76, + "learning_rate": 1.1819969412278787e-07, + "loss": 3.3698, + "step": 44738 + }, + { + "epoch": 9.76, + "learning_rate": 1.1809045226130655e-07, + "loss": 3.364, + "step": 44739 + }, + { + "epoch": 9.76, + "learning_rate": 1.1798121039982521e-07, + "loss": 2.078, + "step": 44740 + }, + { + "epoch": 9.76, + "learning_rate": 1.1787196853834391e-07, + "loss": 1.7219, + "step": 44741 + }, + { + "epoch": 9.76, + "learning_rate": 1.1776272667686259e-07, + "loss": 0.8957, + "step": 44742 + }, + { + "epoch": 9.76, + "learning_rate": 1.1765348481538127e-07, + "loss": 0.7742, + "step": 44743 + }, + { + "epoch": 9.77, + "learning_rate": 1.1754424295389995e-07, + "loss": 0.7663, + "step": 44744 + }, + { + "epoch": 9.77, + "learning_rate": 1.1743500109241862e-07, + "loss": 0.3762, + "step": 44745 + }, + { + "epoch": 9.77, + "learning_rate": 1.1732575923093731e-07, + "loss": 0.3547, + "step": 44746 + }, + { + "epoch": 9.77, + "learning_rate": 1.1721651736945599e-07, + "loss": 0.88, + "step": 44747 + }, + { + "epoch": 9.77, + "learning_rate": 1.1710727550797467e-07, + "loss": 0.6876, + "step": 44748 + }, + { + "epoch": 9.77, + "learning_rate": 1.1699803364649334e-07, + "loss": 1.142, + "step": 44749 + }, + { + "epoch": 9.77, + "learning_rate": 1.1688879178501202e-07, + "loss": 2.8911, + "step": 44750 + }, + { + "epoch": 9.77, + "learning_rate": 1.1677954992353071e-07, + "loss": 0.5704, + "step": 44751 + }, + { + "epoch": 9.77, + "learning_rate": 1.1667030806204939e-07, + "loss": 0.956, + "step": 44752 + }, + { + "epoch": 9.77, + "learning_rate": 1.1656106620056807e-07, + "loss": 2.582, + "step": 44753 + }, + { + "epoch": 9.77, + "learning_rate": 1.1645182433908674e-07, + "loss": 1.3566, + "step": 44754 + }, + { + "epoch": 9.77, + "learning_rate": 1.1634258247760542e-07, + "loss": 2.4029, + "step": 44755 + }, + { + "epoch": 9.77, + "learning_rate": 1.1623334061612411e-07, + "loss": 0.1006, + "step": 44756 + }, + { + "epoch": 9.77, + "learning_rate": 1.1612409875464279e-07, + "loss": 0.5236, + "step": 44757 + }, + { + "epoch": 9.77, + "learning_rate": 1.1601485689316147e-07, + "loss": 0.5383, + "step": 44758 + }, + { + "epoch": 9.77, + "learning_rate": 1.1590561503168014e-07, + "loss": 1.8448, + "step": 44759 + }, + { + "epoch": 9.77, + "learning_rate": 1.1579637317019883e-07, + "loss": 1.5858, + "step": 44760 + }, + { + "epoch": 9.77, + "learning_rate": 1.1568713130871751e-07, + "loss": 0.681, + "step": 44761 + }, + { + "epoch": 9.77, + "learning_rate": 1.1557788944723619e-07, + "loss": 3.5977, + "step": 44762 + }, + { + "epoch": 9.77, + "learning_rate": 1.1546864758575486e-07, + "loss": 0.4361, + "step": 44763 + }, + { + "epoch": 9.77, + "learning_rate": 1.1535940572427354e-07, + "loss": 1.3773, + "step": 44764 + }, + { + "epoch": 9.77, + "learning_rate": 1.1525016386279223e-07, + "loss": 1.919, + "step": 44765 + }, + { + "epoch": 9.77, + "learning_rate": 1.1514092200131091e-07, + "loss": 0.8698, + "step": 44766 + }, + { + "epoch": 9.77, + "learning_rate": 1.1503168013982959e-07, + "loss": 0.5856, + "step": 44767 + }, + { + "epoch": 9.77, + "learning_rate": 1.1492243827834826e-07, + "loss": 0.5199, + "step": 44768 + }, + { + "epoch": 9.77, + "learning_rate": 1.1481319641686694e-07, + "loss": 1.2895, + "step": 44769 + }, + { + "epoch": 9.77, + "learning_rate": 1.1470395455538564e-07, + "loss": 1.1835, + "step": 44770 + }, + { + "epoch": 9.77, + "learning_rate": 1.1459471269390432e-07, + "loss": 1.073, + "step": 44771 + }, + { + "epoch": 9.77, + "learning_rate": 1.14485470832423e-07, + "loss": 0.7448, + "step": 44772 + }, + { + "epoch": 9.77, + "learning_rate": 1.1437622897094168e-07, + "loss": 0.2988, + "step": 44773 + }, + { + "epoch": 9.77, + "learning_rate": 1.1426698710946034e-07, + "loss": 1.8135, + "step": 44774 + }, + { + "epoch": 9.77, + "learning_rate": 1.1415774524797904e-07, + "loss": 0.764, + "step": 44775 + }, + { + "epoch": 9.77, + "learning_rate": 1.1404850338649772e-07, + "loss": 3.1512, + "step": 44776 + }, + { + "epoch": 9.77, + "learning_rate": 1.139392615250164e-07, + "loss": 2.274, + "step": 44777 + }, + { + "epoch": 9.77, + "learning_rate": 1.1383001966353507e-07, + "loss": 0.2922, + "step": 44778 + }, + { + "epoch": 9.77, + "learning_rate": 1.1372077780205377e-07, + "loss": 0.2121, + "step": 44779 + }, + { + "epoch": 9.77, + "learning_rate": 1.1361153594057244e-07, + "loss": 0.7597, + "step": 44780 + }, + { + "epoch": 9.77, + "learning_rate": 1.1350229407909112e-07, + "loss": 1.4131, + "step": 44781 + }, + { + "epoch": 9.77, + "learning_rate": 1.133930522176098e-07, + "loss": 2.6003, + "step": 44782 + }, + { + "epoch": 9.77, + "learning_rate": 1.1328381035612847e-07, + "loss": 0.7641, + "step": 44783 + }, + { + "epoch": 9.77, + "learning_rate": 1.1317456849464716e-07, + "loss": 3.2138, + "step": 44784 + }, + { + "epoch": 9.77, + "learning_rate": 1.1306532663316584e-07, + "loss": 0.5562, + "step": 44785 + }, + { + "epoch": 9.77, + "learning_rate": 1.1295608477168452e-07, + "loss": 0.3095, + "step": 44786 + }, + { + "epoch": 9.77, + "learning_rate": 1.128468429102032e-07, + "loss": 0.7351, + "step": 44787 + }, + { + "epoch": 9.77, + "learning_rate": 1.1273760104872187e-07, + "loss": 0.8612, + "step": 44788 + }, + { + "epoch": 9.77, + "learning_rate": 1.1262835918724056e-07, + "loss": 1.0578, + "step": 44789 + }, + { + "epoch": 9.78, + "learning_rate": 1.1251911732575924e-07, + "loss": 0.0786, + "step": 44790 + }, + { + "epoch": 9.78, + "learning_rate": 1.1240987546427792e-07, + "loss": 2.202, + "step": 44791 + }, + { + "epoch": 9.78, + "learning_rate": 1.123006336027966e-07, + "loss": 1.9675, + "step": 44792 + }, + { + "epoch": 9.78, + "learning_rate": 1.1219139174131527e-07, + "loss": 0.2338, + "step": 44793 + }, + { + "epoch": 9.78, + "learning_rate": 1.1208214987983396e-07, + "loss": 0.4105, + "step": 44794 + }, + { + "epoch": 9.78, + "learning_rate": 1.1197290801835264e-07, + "loss": 1.369, + "step": 44795 + }, + { + "epoch": 9.78, + "learning_rate": 1.1186366615687132e-07, + "loss": 2.3682, + "step": 44796 + }, + { + "epoch": 9.78, + "learning_rate": 1.1175442429539e-07, + "loss": 1.0688, + "step": 44797 + }, + { + "epoch": 9.78, + "learning_rate": 1.116451824339087e-07, + "loss": 3.5224, + "step": 44798 + }, + { + "epoch": 9.78, + "learning_rate": 1.1153594057242737e-07, + "loss": 0.8489, + "step": 44799 + }, + { + "epoch": 9.78, + "learning_rate": 1.1142669871094604e-07, + "loss": 0.3973, + "step": 44800 + }, + { + "epoch": 9.78, + "learning_rate": 1.1131745684946472e-07, + "loss": 0.2346, + "step": 44801 + }, + { + "epoch": 9.78, + "learning_rate": 1.1120821498798339e-07, + "loss": 1.6194, + "step": 44802 + }, + { + "epoch": 9.78, + "learning_rate": 1.110989731265021e-07, + "loss": 0.0641, + "step": 44803 + }, + { + "epoch": 9.78, + "learning_rate": 1.1098973126502077e-07, + "loss": 1.2321, + "step": 44804 + }, + { + "epoch": 9.78, + "learning_rate": 1.1088048940353945e-07, + "loss": 1.177, + "step": 44805 + }, + { + "epoch": 9.78, + "learning_rate": 1.1077124754205813e-07, + "loss": 0.7661, + "step": 44806 + }, + { + "epoch": 9.78, + "learning_rate": 1.106620056805768e-07, + "loss": 1.4416, + "step": 44807 + }, + { + "epoch": 9.78, + "learning_rate": 1.105527638190955e-07, + "loss": 1.3213, + "step": 44808 + }, + { + "epoch": 9.78, + "learning_rate": 1.1044352195761417e-07, + "loss": 0.7619, + "step": 44809 + }, + { + "epoch": 9.78, + "learning_rate": 1.1033428009613285e-07, + "loss": 0.7413, + "step": 44810 + }, + { + "epoch": 9.78, + "learning_rate": 1.1022503823465153e-07, + "loss": 0.8055, + "step": 44811 + }, + { + "epoch": 9.78, + "learning_rate": 1.101157963731702e-07, + "loss": 0.8724, + "step": 44812 + }, + { + "epoch": 9.78, + "learning_rate": 1.100065545116889e-07, + "loss": 1.0925, + "step": 44813 + }, + { + "epoch": 9.78, + "learning_rate": 1.0989731265020757e-07, + "loss": 0.7672, + "step": 44814 + }, + { + "epoch": 9.78, + "learning_rate": 1.0978807078872625e-07, + "loss": 1.8775, + "step": 44815 + }, + { + "epoch": 9.78, + "learning_rate": 1.0967882892724493e-07, + "loss": 0.4341, + "step": 44816 + }, + { + "epoch": 9.78, + "learning_rate": 1.095695870657636e-07, + "loss": 1.4938, + "step": 44817 + }, + { + "epoch": 9.78, + "learning_rate": 1.094603452042823e-07, + "loss": 1.5924, + "step": 44818 + }, + { + "epoch": 9.78, + "learning_rate": 1.0935110334280097e-07, + "loss": 1.155, + "step": 44819 + }, + { + "epoch": 9.78, + "learning_rate": 1.0924186148131965e-07, + "loss": 1.3671, + "step": 44820 + }, + { + "epoch": 9.78, + "learning_rate": 1.0913261961983833e-07, + "loss": 0.0864, + "step": 44821 + }, + { + "epoch": 9.78, + "learning_rate": 1.0902337775835702e-07, + "loss": 0.4491, + "step": 44822 + }, + { + "epoch": 9.78, + "learning_rate": 1.0891413589687569e-07, + "loss": 1.2838, + "step": 44823 + }, + { + "epoch": 9.78, + "learning_rate": 1.0880489403539437e-07, + "loss": 3.4524, + "step": 44824 + }, + { + "epoch": 9.78, + "learning_rate": 1.0869565217391305e-07, + "loss": 0.1519, + "step": 44825 + }, + { + "epoch": 9.78, + "learning_rate": 1.0858641031243172e-07, + "loss": 0.5947, + "step": 44826 + }, + { + "epoch": 9.78, + "learning_rate": 1.0847716845095041e-07, + "loss": 0.2245, + "step": 44827 + }, + { + "epoch": 9.78, + "learning_rate": 1.0836792658946909e-07, + "loss": 1.1801, + "step": 44828 + }, + { + "epoch": 9.78, + "learning_rate": 1.0825868472798777e-07, + "loss": 0.0539, + "step": 44829 + }, + { + "epoch": 9.78, + "learning_rate": 1.0814944286650645e-07, + "loss": 4.3638, + "step": 44830 + }, + { + "epoch": 9.78, + "learning_rate": 1.0804020100502512e-07, + "loss": 0.0938, + "step": 44831 + }, + { + "epoch": 9.78, + "learning_rate": 1.0793095914354383e-07, + "loss": 1.9329, + "step": 44832 + }, + { + "epoch": 9.78, + "learning_rate": 1.078217172820625e-07, + "loss": 2.0359, + "step": 44833 + }, + { + "epoch": 9.78, + "learning_rate": 1.0771247542058117e-07, + "loss": 0.3636, + "step": 44834 + }, + { + "epoch": 9.79, + "learning_rate": 1.0760323355909985e-07, + "loss": 2.3678, + "step": 44835 + }, + { + "epoch": 9.79, + "learning_rate": 1.0749399169761852e-07, + "loss": 2.1296, + "step": 44836 + }, + { + "epoch": 9.79, + "learning_rate": 1.0738474983613723e-07, + "loss": 0.5746, + "step": 44837 + }, + { + "epoch": 9.79, + "learning_rate": 1.072755079746559e-07, + "loss": 0.02, + "step": 44838 + }, + { + "epoch": 9.79, + "learning_rate": 1.0716626611317458e-07, + "loss": 1.7922, + "step": 44839 + }, + { + "epoch": 9.79, + "learning_rate": 1.0705702425169326e-07, + "loss": 0.4852, + "step": 44840 + }, + { + "epoch": 9.79, + "learning_rate": 1.0694778239021195e-07, + "loss": 0.5268, + "step": 44841 + }, + { + "epoch": 9.79, + "learning_rate": 1.0683854052873063e-07, + "loss": 0.5289, + "step": 44842 + }, + { + "epoch": 9.79, + "learning_rate": 1.067292986672493e-07, + "loss": 1.5011, + "step": 44843 + }, + { + "epoch": 9.79, + "learning_rate": 1.0662005680576798e-07, + "loss": 0.8535, + "step": 44844 + }, + { + "epoch": 9.79, + "learning_rate": 1.0651081494428666e-07, + "loss": 1.5805, + "step": 44845 + }, + { + "epoch": 9.79, + "learning_rate": 1.0640157308280535e-07, + "loss": 0.3928, + "step": 44846 + }, + { + "epoch": 9.79, + "learning_rate": 1.0629233122132402e-07, + "loss": 0.2291, + "step": 44847 + }, + { + "epoch": 9.79, + "learning_rate": 1.061830893598427e-07, + "loss": 1.6377, + "step": 44848 + }, + { + "epoch": 9.79, + "learning_rate": 1.0607384749836138e-07, + "loss": 0.1186, + "step": 44849 + }, + { + "epoch": 9.79, + "learning_rate": 1.0596460563688006e-07, + "loss": 0.4953, + "step": 44850 + }, + { + "epoch": 9.79, + "learning_rate": 1.0585536377539875e-07, + "loss": 1.2609, + "step": 44851 + }, + { + "epoch": 9.79, + "learning_rate": 1.0574612191391742e-07, + "loss": 1.281, + "step": 44852 + }, + { + "epoch": 9.79, + "learning_rate": 1.056368800524361e-07, + "loss": 1.5149, + "step": 44853 + }, + { + "epoch": 9.79, + "learning_rate": 1.0552763819095478e-07, + "loss": 1.6516, + "step": 44854 + }, + { + "epoch": 9.79, + "learning_rate": 1.0541839632947345e-07, + "loss": 1.4273, + "step": 44855 + }, + { + "epoch": 9.79, + "learning_rate": 1.0530915446799215e-07, + "loss": 1.3759, + "step": 44856 + }, + { + "epoch": 9.79, + "learning_rate": 1.0519991260651082e-07, + "loss": 1.6593, + "step": 44857 + }, + { + "epoch": 9.79, + "learning_rate": 1.050906707450295e-07, + "loss": 0.2141, + "step": 44858 + }, + { + "epoch": 9.79, + "learning_rate": 1.0498142888354818e-07, + "loss": 2.1782, + "step": 44859 + }, + { + "epoch": 9.79, + "learning_rate": 1.0487218702206687e-07, + "loss": 0.9, + "step": 44860 + }, + { + "epoch": 9.79, + "learning_rate": 1.0476294516058554e-07, + "loss": 1.7146, + "step": 44861 + }, + { + "epoch": 9.79, + "learning_rate": 1.0465370329910422e-07, + "loss": 0.9096, + "step": 44862 + }, + { + "epoch": 9.79, + "learning_rate": 1.045444614376229e-07, + "loss": 1.501, + "step": 44863 + }, + { + "epoch": 9.79, + "learning_rate": 1.0443521957614158e-07, + "loss": 0.9785, + "step": 44864 + }, + { + "epoch": 9.79, + "learning_rate": 1.0432597771466028e-07, + "loss": 0.0735, + "step": 44865 + }, + { + "epoch": 9.79, + "learning_rate": 1.0421673585317896e-07, + "loss": 1.3394, + "step": 44866 + }, + { + "epoch": 9.79, + "learning_rate": 1.0410749399169763e-07, + "loss": 0.8594, + "step": 44867 + }, + { + "epoch": 9.79, + "learning_rate": 1.039982521302163e-07, + "loss": 0.8062, + "step": 44868 + }, + { + "epoch": 9.79, + "learning_rate": 1.0388901026873497e-07, + "loss": 0.7862, + "step": 44869 + }, + { + "epoch": 9.79, + "learning_rate": 1.0377976840725368e-07, + "loss": 1.0992, + "step": 44870 + }, + { + "epoch": 9.79, + "learning_rate": 1.0367052654577236e-07, + "loss": 1.5433, + "step": 44871 + }, + { + "epoch": 9.79, + "learning_rate": 1.0356128468429103e-07, + "loss": 1.8913, + "step": 44872 + }, + { + "epoch": 9.79, + "learning_rate": 1.0345204282280971e-07, + "loss": 0.6969, + "step": 44873 + }, + { + "epoch": 9.79, + "learning_rate": 1.0334280096132839e-07, + "loss": 0.5154, + "step": 44874 + }, + { + "epoch": 9.79, + "learning_rate": 1.0323355909984708e-07, + "loss": 1.9873, + "step": 44875 + }, + { + "epoch": 9.79, + "learning_rate": 1.0312431723836575e-07, + "loss": 1.7393, + "step": 44876 + }, + { + "epoch": 9.79, + "learning_rate": 1.0301507537688443e-07, + "loss": 0.8555, + "step": 44877 + }, + { + "epoch": 9.79, + "learning_rate": 1.0290583351540311e-07, + "loss": 1.7165, + "step": 44878 + }, + { + "epoch": 9.79, + "learning_rate": 1.027965916539218e-07, + "loss": 0.5537, + "step": 44879 + }, + { + "epoch": 9.79, + "learning_rate": 1.0268734979244048e-07, + "loss": 0.5034, + "step": 44880 + }, + { + "epoch": 9.8, + "learning_rate": 1.0257810793095915e-07, + "loss": 1.2773, + "step": 44881 + }, + { + "epoch": 9.8, + "learning_rate": 1.0246886606947783e-07, + "loss": 1.4517, + "step": 44882 + }, + { + "epoch": 9.8, + "learning_rate": 1.0235962420799651e-07, + "loss": 2.2311, + "step": 44883 + }, + { + "epoch": 9.8, + "learning_rate": 1.022503823465152e-07, + "loss": 1.3731, + "step": 44884 + }, + { + "epoch": 9.8, + "learning_rate": 1.0214114048503388e-07, + "loss": 0.1718, + "step": 44885 + }, + { + "epoch": 9.8, + "learning_rate": 1.0203189862355255e-07, + "loss": 0.8267, + "step": 44886 + }, + { + "epoch": 9.8, + "learning_rate": 1.0192265676207123e-07, + "loss": 0.5265, + "step": 44887 + }, + { + "epoch": 9.8, + "learning_rate": 1.0181341490058991e-07, + "loss": 0.8131, + "step": 44888 + }, + { + "epoch": 9.8, + "learning_rate": 1.017041730391086e-07, + "loss": 0.9607, + "step": 44889 + }, + { + "epoch": 9.8, + "learning_rate": 1.0159493117762727e-07, + "loss": 0.1187, + "step": 44890 + }, + { + "epoch": 9.8, + "learning_rate": 1.0148568931614595e-07, + "loss": 1.6932, + "step": 44891 + }, + { + "epoch": 9.8, + "learning_rate": 1.0137644745466463e-07, + "loss": 1.8725, + "step": 44892 + }, + { + "epoch": 9.8, + "learning_rate": 1.012672055931833e-07, + "loss": 3.3286, + "step": 44893 + }, + { + "epoch": 9.8, + "learning_rate": 1.01157963731702e-07, + "loss": 0.6601, + "step": 44894 + }, + { + "epoch": 9.8, + "learning_rate": 1.0104872187022067e-07, + "loss": 0.9829, + "step": 44895 + }, + { + "epoch": 9.8, + "learning_rate": 1.0093948000873935e-07, + "loss": 0.5361, + "step": 44896 + }, + { + "epoch": 9.8, + "learning_rate": 1.0083023814725803e-07, + "loss": 2.2063, + "step": 44897 + }, + { + "epoch": 9.8, + "learning_rate": 1.007209962857767e-07, + "loss": 2.2563, + "step": 44898 + }, + { + "epoch": 9.8, + "learning_rate": 1.0061175442429541e-07, + "loss": 0.1506, + "step": 44899 + }, + { + "epoch": 9.8, + "learning_rate": 1.0050251256281409e-07, + "loss": 1.5491, + "step": 44900 + }, + { + "epoch": 9.8, + "learning_rate": 1.0039327070133276e-07, + "loss": 0.6678, + "step": 44901 + }, + { + "epoch": 9.8, + "learning_rate": 1.0028402883985143e-07, + "loss": 1.565, + "step": 44902 + }, + { + "epoch": 9.8, + "learning_rate": 1.0017478697837013e-07, + "loss": 0.9412, + "step": 44903 + }, + { + "epoch": 9.8, + "learning_rate": 1.0006554511688881e-07, + "loss": 2.1888, + "step": 44904 + }, + { + "epoch": 9.8, + "learning_rate": 9.995630325540748e-08, + "loss": 0.7649, + "step": 44905 + }, + { + "epoch": 9.8, + "learning_rate": 9.984706139392616e-08, + "loss": 0.4715, + "step": 44906 + }, + { + "epoch": 9.8, + "learning_rate": 9.973781953244484e-08, + "loss": 0.4905, + "step": 44907 + }, + { + "epoch": 9.8, + "learning_rate": 9.962857767096353e-08, + "loss": 0.9744, + "step": 44908 + }, + { + "epoch": 9.8, + "learning_rate": 9.951933580948221e-08, + "loss": 0.9938, + "step": 44909 + }, + { + "epoch": 9.8, + "learning_rate": 9.941009394800088e-08, + "loss": 0.5687, + "step": 44910 + }, + { + "epoch": 9.8, + "learning_rate": 9.930085208651956e-08, + "loss": 0.7367, + "step": 44911 + }, + { + "epoch": 9.8, + "learning_rate": 9.919161022503824e-08, + "loss": 1.1469, + "step": 44912 + }, + { + "epoch": 9.8, + "learning_rate": 9.908236836355693e-08, + "loss": 3.5978, + "step": 44913 + }, + { + "epoch": 9.8, + "learning_rate": 9.89731265020756e-08, + "loss": 0.1396, + "step": 44914 + }, + { + "epoch": 9.8, + "learning_rate": 9.886388464059428e-08, + "loss": 2.2587, + "step": 44915 + }, + { + "epoch": 9.8, + "learning_rate": 9.875464277911296e-08, + "loss": 1.2062, + "step": 44916 + }, + { + "epoch": 9.8, + "learning_rate": 9.864540091763164e-08, + "loss": 0.5357, + "step": 44917 + }, + { + "epoch": 9.8, + "learning_rate": 9.853615905615033e-08, + "loss": 0.1918, + "step": 44918 + }, + { + "epoch": 9.8, + "learning_rate": 9.8426917194669e-08, + "loss": 1.8508, + "step": 44919 + }, + { + "epoch": 9.8, + "learning_rate": 9.831767533318768e-08, + "loss": 1.8787, + "step": 44920 + }, + { + "epoch": 9.8, + "learning_rate": 9.820843347170636e-08, + "loss": 1.7127, + "step": 44921 + }, + { + "epoch": 9.8, + "learning_rate": 9.809919161022505e-08, + "loss": 1.9613, + "step": 44922 + }, + { + "epoch": 9.8, + "learning_rate": 9.798994974874373e-08, + "loss": 1.3429, + "step": 44923 + }, + { + "epoch": 9.8, + "learning_rate": 9.78807078872624e-08, + "loss": 0.6596, + "step": 44924 + }, + { + "epoch": 9.8, + "learning_rate": 9.777146602578108e-08, + "loss": 1.8228, + "step": 44925 + }, + { + "epoch": 9.8, + "learning_rate": 9.766222416429976e-08, + "loss": 1.474, + "step": 44926 + }, + { + "epoch": 9.81, + "learning_rate": 9.755298230281846e-08, + "loss": 0.2282, + "step": 44927 + }, + { + "epoch": 9.81, + "learning_rate": 9.744374044133713e-08, + "loss": 0.9313, + "step": 44928 + }, + { + "epoch": 9.81, + "learning_rate": 9.73344985798558e-08, + "loss": 0.1227, + "step": 44929 + }, + { + "epoch": 9.81, + "learning_rate": 9.722525671837448e-08, + "loss": 0.9177, + "step": 44930 + }, + { + "epoch": 9.81, + "learning_rate": 9.711601485689316e-08, + "loss": 0.2233, + "step": 44931 + }, + { + "epoch": 9.81, + "learning_rate": 9.700677299541186e-08, + "loss": 0.8787, + "step": 44932 + }, + { + "epoch": 9.81, + "learning_rate": 9.689753113393054e-08, + "loss": 1.6132, + "step": 44933 + }, + { + "epoch": 9.81, + "learning_rate": 9.678828927244922e-08, + "loss": 1.9658, + "step": 44934 + }, + { + "epoch": 9.81, + "learning_rate": 9.667904741096789e-08, + "loss": 1.8103, + "step": 44935 + }, + { + "epoch": 9.81, + "learning_rate": 9.656980554948657e-08, + "loss": 0.4298, + "step": 44936 + }, + { + "epoch": 9.81, + "learning_rate": 9.646056368800526e-08, + "loss": 1.0289, + "step": 44937 + }, + { + "epoch": 9.81, + "learning_rate": 9.635132182652394e-08, + "loss": 1.4791, + "step": 44938 + }, + { + "epoch": 9.81, + "learning_rate": 9.624207996504261e-08, + "loss": 1.09, + "step": 44939 + }, + { + "epoch": 9.81, + "learning_rate": 9.613283810356129e-08, + "loss": 1.3372, + "step": 44940 + }, + { + "epoch": 9.81, + "learning_rate": 9.602359624207998e-08, + "loss": 0.7881, + "step": 44941 + }, + { + "epoch": 9.81, + "learning_rate": 9.591435438059866e-08, + "loss": 0.8159, + "step": 44942 + }, + { + "epoch": 9.81, + "learning_rate": 9.580511251911734e-08, + "loss": 0.3818, + "step": 44943 + }, + { + "epoch": 9.81, + "learning_rate": 9.569587065763601e-08, + "loss": 1.5089, + "step": 44944 + }, + { + "epoch": 9.81, + "learning_rate": 9.558662879615469e-08, + "loss": 1.5687, + "step": 44945 + }, + { + "epoch": 9.81, + "learning_rate": 9.547738693467338e-08, + "loss": 1.0509, + "step": 44946 + }, + { + "epoch": 9.81, + "learning_rate": 9.536814507319206e-08, + "loss": 0.8717, + "step": 44947 + }, + { + "epoch": 9.81, + "learning_rate": 9.525890321171074e-08, + "loss": 1.4506, + "step": 44948 + }, + { + "epoch": 9.81, + "learning_rate": 9.514966135022941e-08, + "loss": 0.6613, + "step": 44949 + }, + { + "epoch": 9.81, + "learning_rate": 9.504041948874809e-08, + "loss": 2.4828, + "step": 44950 + }, + { + "epoch": 9.81, + "learning_rate": 9.493117762726678e-08, + "loss": 1.124, + "step": 44951 + }, + { + "epoch": 9.81, + "learning_rate": 9.482193576578546e-08, + "loss": 1.4985, + "step": 44952 + }, + { + "epoch": 9.81, + "learning_rate": 9.471269390430413e-08, + "loss": 1.893, + "step": 44953 + }, + { + "epoch": 9.81, + "learning_rate": 9.460345204282281e-08, + "loss": 1.543, + "step": 44954 + }, + { + "epoch": 9.81, + "learning_rate": 9.449421018134149e-08, + "loss": 0.8157, + "step": 44955 + }, + { + "epoch": 9.81, + "learning_rate": 9.438496831986018e-08, + "loss": 1.8602, + "step": 44956 + }, + { + "epoch": 9.81, + "learning_rate": 9.427572645837886e-08, + "loss": 2.0392, + "step": 44957 + }, + { + "epoch": 9.81, + "learning_rate": 9.416648459689753e-08, + "loss": 0.4142, + "step": 44958 + }, + { + "epoch": 9.81, + "learning_rate": 9.405724273541621e-08, + "loss": 1.6089, + "step": 44959 + }, + { + "epoch": 9.81, + "learning_rate": 9.394800087393491e-08, + "loss": 0.0827, + "step": 44960 + }, + { + "epoch": 9.81, + "learning_rate": 9.383875901245359e-08, + "loss": 0.4382, + "step": 44961 + }, + { + "epoch": 9.81, + "learning_rate": 9.372951715097226e-08, + "loss": 0.7708, + "step": 44962 + }, + { + "epoch": 9.81, + "learning_rate": 9.362027528949093e-08, + "loss": 2.2618, + "step": 44963 + }, + { + "epoch": 9.81, + "learning_rate": 9.351103342800961e-08, + "loss": 0.9377, + "step": 44964 + }, + { + "epoch": 9.81, + "learning_rate": 9.340179156652831e-08, + "loss": 1.4146, + "step": 44965 + }, + { + "epoch": 9.81, + "learning_rate": 9.329254970504699e-08, + "loss": 1.7621, + "step": 44966 + }, + { + "epoch": 9.81, + "learning_rate": 9.318330784356567e-08, + "loss": 0.5486, + "step": 44967 + }, + { + "epoch": 9.81, + "learning_rate": 9.307406598208434e-08, + "loss": 1.0414, + "step": 44968 + }, + { + "epoch": 9.81, + "learning_rate": 9.296482412060302e-08, + "loss": 1.1965, + "step": 44969 + }, + { + "epoch": 9.81, + "learning_rate": 9.285558225912171e-08, + "loss": 0.343, + "step": 44970 + }, + { + "epoch": 9.81, + "learning_rate": 9.274634039764039e-08, + "loss": 1.0893, + "step": 44971 + }, + { + "epoch": 9.81, + "learning_rate": 9.263709853615907e-08, + "loss": 2.5269, + "step": 44972 + }, + { + "epoch": 9.82, + "learning_rate": 9.252785667467774e-08, + "loss": 1.867, + "step": 44973 + }, + { + "epoch": 9.82, + "learning_rate": 9.241861481319642e-08, + "loss": 1.6056, + "step": 44974 + }, + { + "epoch": 9.82, + "learning_rate": 9.230937295171511e-08, + "loss": 2.0721, + "step": 44975 + }, + { + "epoch": 9.82, + "learning_rate": 9.220013109023379e-08, + "loss": 1.0156, + "step": 44976 + }, + { + "epoch": 9.82, + "learning_rate": 9.209088922875247e-08, + "loss": 1.6272, + "step": 44977 + }, + { + "epoch": 9.82, + "learning_rate": 9.198164736727114e-08, + "loss": 1.8789, + "step": 44978 + }, + { + "epoch": 9.82, + "learning_rate": 9.187240550578982e-08, + "loss": 0.873, + "step": 44979 + }, + { + "epoch": 9.82, + "learning_rate": 9.176316364430851e-08, + "loss": 0.5531, + "step": 44980 + }, + { + "epoch": 9.82, + "learning_rate": 9.165392178282719e-08, + "loss": 0.4818, + "step": 44981 + }, + { + "epoch": 9.82, + "learning_rate": 9.154467992134586e-08, + "loss": 0.7379, + "step": 44982 + }, + { + "epoch": 9.82, + "learning_rate": 9.143543805986454e-08, + "loss": 1.2072, + "step": 44983 + }, + { + "epoch": 9.82, + "learning_rate": 9.132619619838323e-08, + "loss": 1.391, + "step": 44984 + }, + { + "epoch": 9.82, + "learning_rate": 9.121695433690191e-08, + "loss": 0.4466, + "step": 44985 + }, + { + "epoch": 9.82, + "learning_rate": 9.110771247542059e-08, + "loss": 0.6381, + "step": 44986 + }, + { + "epoch": 9.82, + "learning_rate": 9.099847061393926e-08, + "loss": 0.7329, + "step": 44987 + }, + { + "epoch": 9.82, + "learning_rate": 9.088922875245794e-08, + "loss": 0.9149, + "step": 44988 + }, + { + "epoch": 9.82, + "learning_rate": 9.077998689097663e-08, + "loss": 0.8313, + "step": 44989 + }, + { + "epoch": 9.82, + "learning_rate": 9.067074502949531e-08, + "loss": 0.6172, + "step": 44990 + }, + { + "epoch": 9.82, + "learning_rate": 9.056150316801399e-08, + "loss": 0.94, + "step": 44991 + }, + { + "epoch": 9.82, + "learning_rate": 9.045226130653266e-08, + "loss": 0.5464, + "step": 44992 + }, + { + "epoch": 9.82, + "learning_rate": 9.034301944505134e-08, + "loss": 0.5956, + "step": 44993 + }, + { + "epoch": 9.82, + "learning_rate": 9.023377758357004e-08, + "loss": 0.8206, + "step": 44994 + }, + { + "epoch": 9.82, + "learning_rate": 9.012453572208872e-08, + "loss": 2.153, + "step": 44995 + }, + { + "epoch": 9.82, + "learning_rate": 9.001529386060738e-08, + "loss": 1.5248, + "step": 44996 + }, + { + "epoch": 9.82, + "learning_rate": 8.990605199912606e-08, + "loss": 0.8859, + "step": 44997 + }, + { + "epoch": 9.82, + "learning_rate": 8.979681013764474e-08, + "loss": 0.5369, + "step": 44998 + }, + { + "epoch": 9.82, + "learning_rate": 8.968756827616344e-08, + "loss": 1.7357, + "step": 44999 + }, + { + "epoch": 9.82, + "learning_rate": 8.957832641468212e-08, + "loss": 1.1082, + "step": 45000 + }, + { + "epoch": 9.82, + "learning_rate": 8.94690845532008e-08, + "loss": 0.7786, + "step": 45001 + }, + { + "epoch": 9.82, + "learning_rate": 8.935984269171947e-08, + "loss": 0.6485, + "step": 45002 + }, + { + "epoch": 9.82, + "learning_rate": 8.925060083023816e-08, + "loss": 1.076, + "step": 45003 + }, + { + "epoch": 9.82, + "learning_rate": 8.914135896875684e-08, + "loss": 0.5134, + "step": 45004 + }, + { + "epoch": 9.82, + "learning_rate": 8.903211710727552e-08, + "loss": 1.168, + "step": 45005 + }, + { + "epoch": 9.82, + "learning_rate": 8.89228752457942e-08, + "loss": 1.0773, + "step": 45006 + }, + { + "epoch": 9.82, + "learning_rate": 8.881363338431287e-08, + "loss": 1.6841, + "step": 45007 + }, + { + "epoch": 9.82, + "learning_rate": 8.870439152283156e-08, + "loss": 1.7188, + "step": 45008 + }, + { + "epoch": 9.82, + "learning_rate": 8.859514966135024e-08, + "loss": 0.5791, + "step": 45009 + }, + { + "epoch": 9.82, + "learning_rate": 8.848590779986892e-08, + "loss": 0.2223, + "step": 45010 + }, + { + "epoch": 9.82, + "learning_rate": 8.83766659383876e-08, + "loss": 1.5964, + "step": 45011 + }, + { + "epoch": 9.82, + "learning_rate": 8.826742407690627e-08, + "loss": 2.5886, + "step": 45012 + }, + { + "epoch": 9.82, + "learning_rate": 8.815818221542496e-08, + "loss": 1.0119, + "step": 45013 + }, + { + "epoch": 9.82, + "learning_rate": 8.804894035394364e-08, + "loss": 2.2537, + "step": 45014 + }, + { + "epoch": 9.82, + "learning_rate": 8.793969849246232e-08, + "loss": 0.6546, + "step": 45015 + }, + { + "epoch": 9.82, + "learning_rate": 8.7830456630981e-08, + "loss": 0.2717, + "step": 45016 + }, + { + "epoch": 9.82, + "learning_rate": 8.772121476949967e-08, + "loss": 1.2568, + "step": 45017 + }, + { + "epoch": 9.82, + "learning_rate": 8.761197290801836e-08, + "loss": 1.6236, + "step": 45018 + }, + { + "epoch": 9.83, + "learning_rate": 8.750273104653704e-08, + "loss": 2.1075, + "step": 45019 + }, + { + "epoch": 9.83, + "learning_rate": 8.739348918505572e-08, + "loss": 0.8803, + "step": 45020 + }, + { + "epoch": 9.83, + "learning_rate": 8.728424732357439e-08, + "loss": 2.0359, + "step": 45021 + }, + { + "epoch": 9.83, + "learning_rate": 8.717500546209308e-08, + "loss": 2.659, + "step": 45022 + }, + { + "epoch": 9.83, + "learning_rate": 8.706576360061176e-08, + "loss": 1.8196, + "step": 45023 + }, + { + "epoch": 9.83, + "learning_rate": 8.695652173913044e-08, + "loss": 0.2233, + "step": 45024 + }, + { + "epoch": 9.83, + "learning_rate": 8.684727987764911e-08, + "loss": 0.4003, + "step": 45025 + }, + { + "epoch": 9.83, + "learning_rate": 8.673803801616779e-08, + "loss": 1.5719, + "step": 45026 + }, + { + "epoch": 9.83, + "learning_rate": 8.66287961546865e-08, + "loss": 2.1655, + "step": 45027 + }, + { + "epoch": 9.83, + "learning_rate": 8.651955429320517e-08, + "loss": 0.8986, + "step": 45028 + }, + { + "epoch": 9.83, + "learning_rate": 8.641031243172385e-08, + "loss": 1.3962, + "step": 45029 + }, + { + "epoch": 9.83, + "learning_rate": 8.630107057024251e-08, + "loss": 1.3985, + "step": 45030 + }, + { + "epoch": 9.83, + "learning_rate": 8.619182870876119e-08, + "loss": 2.0654, + "step": 45031 + }, + { + "epoch": 9.83, + "learning_rate": 8.60825868472799e-08, + "loss": 0.2167, + "step": 45032 + }, + { + "epoch": 9.83, + "learning_rate": 8.597334498579857e-08, + "loss": 0.6815, + "step": 45033 + }, + { + "epoch": 9.83, + "learning_rate": 8.586410312431725e-08, + "loss": 1.9924, + "step": 45034 + }, + { + "epoch": 9.83, + "learning_rate": 8.575486126283593e-08, + "loss": 2.3031, + "step": 45035 + }, + { + "epoch": 9.83, + "learning_rate": 8.56456194013546e-08, + "loss": 0.4256, + "step": 45036 + }, + { + "epoch": 9.83, + "learning_rate": 8.55363775398733e-08, + "loss": 1.125, + "step": 45037 + }, + { + "epoch": 9.83, + "learning_rate": 8.542713567839197e-08, + "loss": 2.4583, + "step": 45038 + }, + { + "epoch": 9.83, + "learning_rate": 8.531789381691065e-08, + "loss": 0.9911, + "step": 45039 + }, + { + "epoch": 9.83, + "learning_rate": 8.520865195542933e-08, + "loss": 0.0822, + "step": 45040 + }, + { + "epoch": 9.83, + "learning_rate": 8.509941009394802e-08, + "loss": 2.3888, + "step": 45041 + }, + { + "epoch": 9.83, + "learning_rate": 8.499016823246669e-08, + "loss": 0.154, + "step": 45042 + }, + { + "epoch": 9.83, + "learning_rate": 8.488092637098537e-08, + "loss": 1.6075, + "step": 45043 + }, + { + "epoch": 9.83, + "learning_rate": 8.477168450950405e-08, + "loss": 0.7786, + "step": 45044 + }, + { + "epoch": 9.83, + "learning_rate": 8.466244264802272e-08, + "loss": 0.63, + "step": 45045 + }, + { + "epoch": 9.83, + "learning_rate": 8.455320078654141e-08, + "loss": 1.6834, + "step": 45046 + }, + { + "epoch": 9.83, + "learning_rate": 8.444395892506009e-08, + "loss": 0.7407, + "step": 45047 + }, + { + "epoch": 9.83, + "learning_rate": 8.433471706357877e-08, + "loss": 1.4667, + "step": 45048 + }, + { + "epoch": 9.83, + "learning_rate": 8.422547520209745e-08, + "loss": 1.6144, + "step": 45049 + }, + { + "epoch": 9.83, + "learning_rate": 8.411623334061612e-08, + "loss": 1.0752, + "step": 45050 + }, + { + "epoch": 9.83, + "learning_rate": 8.400699147913481e-08, + "loss": 1.4101, + "step": 45051 + }, + { + "epoch": 9.83, + "learning_rate": 8.389774961765349e-08, + "loss": 3.1863, + "step": 45052 + }, + { + "epoch": 9.83, + "learning_rate": 8.378850775617217e-08, + "loss": 2.8211, + "step": 45053 + }, + { + "epoch": 9.83, + "learning_rate": 8.367926589469085e-08, + "loss": 1.5002, + "step": 45054 + }, + { + "epoch": 9.83, + "learning_rate": 8.357002403320952e-08, + "loss": 1.1546, + "step": 45055 + }, + { + "epoch": 9.83, + "learning_rate": 8.346078217172821e-08, + "loss": 1.3996, + "step": 45056 + }, + { + "epoch": 9.83, + "learning_rate": 8.335154031024689e-08, + "loss": 0.7977, + "step": 45057 + }, + { + "epoch": 9.83, + "learning_rate": 8.324229844876557e-08, + "loss": 1.9548, + "step": 45058 + }, + { + "epoch": 9.83, + "learning_rate": 8.313305658728424e-08, + "loss": 0.7064, + "step": 45059 + }, + { + "epoch": 9.83, + "learning_rate": 8.302381472580292e-08, + "loss": 0.77, + "step": 45060 + }, + { + "epoch": 9.83, + "learning_rate": 8.291457286432163e-08, + "loss": 0.4002, + "step": 45061 + }, + { + "epoch": 9.83, + "learning_rate": 8.28053310028403e-08, + "loss": 1.9333, + "step": 45062 + }, + { + "epoch": 9.83, + "learning_rate": 8.269608914135898e-08, + "loss": 1.4512, + "step": 45063 + }, + { + "epoch": 9.84, + "learning_rate": 8.258684727987766e-08, + "loss": 1.3586, + "step": 45064 + }, + { + "epoch": 9.84, + "learning_rate": 8.247760541839635e-08, + "loss": 0.7305, + "step": 45065 + }, + { + "epoch": 9.84, + "learning_rate": 8.236836355691502e-08, + "loss": 1.0034, + "step": 45066 + }, + { + "epoch": 9.84, + "learning_rate": 8.22591216954337e-08, + "loss": 0.7304, + "step": 45067 + }, + { + "epoch": 9.84, + "learning_rate": 8.214987983395238e-08, + "loss": 0.8991, + "step": 45068 + }, + { + "epoch": 9.84, + "learning_rate": 8.204063797247106e-08, + "loss": 1.5557, + "step": 45069 + }, + { + "epoch": 9.84, + "learning_rate": 8.193139611098975e-08, + "loss": 1.5754, + "step": 45070 + }, + { + "epoch": 9.84, + "learning_rate": 8.182215424950842e-08, + "loss": 1.8911, + "step": 45071 + }, + { + "epoch": 9.84, + "learning_rate": 8.17129123880271e-08, + "loss": 1.6764, + "step": 45072 + }, + { + "epoch": 9.84, + "learning_rate": 8.160367052654578e-08, + "loss": 1.3828, + "step": 45073 + }, + { + "epoch": 9.84, + "learning_rate": 8.149442866506445e-08, + "loss": 0.847, + "step": 45074 + }, + { + "epoch": 9.84, + "learning_rate": 8.138518680358315e-08, + "loss": 0.9114, + "step": 45075 + }, + { + "epoch": 9.84, + "learning_rate": 8.127594494210182e-08, + "loss": 0.04, + "step": 45076 + }, + { + "epoch": 9.84, + "learning_rate": 8.11667030806205e-08, + "loss": 1.5521, + "step": 45077 + }, + { + "epoch": 9.84, + "learning_rate": 8.105746121913918e-08, + "loss": 1.337, + "step": 45078 + }, + { + "epoch": 9.84, + "learning_rate": 8.094821935765785e-08, + "loss": 0.0702, + "step": 45079 + }, + { + "epoch": 9.84, + "learning_rate": 8.083897749617654e-08, + "loss": 1.0193, + "step": 45080 + }, + { + "epoch": 9.84, + "learning_rate": 8.072973563469522e-08, + "loss": 0.4116, + "step": 45081 + }, + { + "epoch": 9.84, + "learning_rate": 8.06204937732139e-08, + "loss": 1.8168, + "step": 45082 + }, + { + "epoch": 9.84, + "learning_rate": 8.051125191173258e-08, + "loss": 1.306, + "step": 45083 + }, + { + "epoch": 9.84, + "learning_rate": 8.040201005025127e-08, + "loss": 1.3802, + "step": 45084 + }, + { + "epoch": 9.84, + "learning_rate": 8.029276818876994e-08, + "loss": 0.7349, + "step": 45085 + }, + { + "epoch": 9.84, + "learning_rate": 8.018352632728862e-08, + "loss": 0.8375, + "step": 45086 + }, + { + "epoch": 9.84, + "learning_rate": 8.00742844658073e-08, + "loss": 1.0308, + "step": 45087 + }, + { + "epoch": 9.84, + "learning_rate": 7.996504260432597e-08, + "loss": 0.079, + "step": 45088 + }, + { + "epoch": 9.84, + "learning_rate": 7.985580074284468e-08, + "loss": 1.4713, + "step": 45089 + }, + { + "epoch": 9.84, + "learning_rate": 7.974655888136334e-08, + "loss": 1.4593, + "step": 45090 + }, + { + "epoch": 9.84, + "learning_rate": 7.963731701988202e-08, + "loss": 0.2166, + "step": 45091 + }, + { + "epoch": 9.84, + "learning_rate": 7.95280751584007e-08, + "loss": 1.2343, + "step": 45092 + }, + { + "epoch": 9.84, + "learning_rate": 7.941883329691937e-08, + "loss": 2.2673, + "step": 45093 + }, + { + "epoch": 9.84, + "learning_rate": 7.930959143543808e-08, + "loss": 0.3077, + "step": 45094 + }, + { + "epoch": 9.84, + "learning_rate": 7.920034957395675e-08, + "loss": 2.6893, + "step": 45095 + }, + { + "epoch": 9.84, + "learning_rate": 7.909110771247543e-08, + "loss": 1.361, + "step": 45096 + }, + { + "epoch": 9.84, + "learning_rate": 7.898186585099411e-08, + "loss": 0.5301, + "step": 45097 + }, + { + "epoch": 9.84, + "learning_rate": 7.887262398951279e-08, + "loss": 0.6679, + "step": 45098 + }, + { + "epoch": 9.84, + "learning_rate": 7.876338212803148e-08, + "loss": 2.276, + "step": 45099 + }, + { + "epoch": 9.84, + "learning_rate": 7.865414026655015e-08, + "loss": 2.3155, + "step": 45100 + }, + { + "epoch": 9.84, + "learning_rate": 7.854489840506883e-08, + "loss": 1.9241, + "step": 45101 + }, + { + "epoch": 9.84, + "learning_rate": 7.843565654358751e-08, + "loss": 0.474, + "step": 45102 + }, + { + "epoch": 9.84, + "learning_rate": 7.83264146821062e-08, + "loss": 0.5949, + "step": 45103 + }, + { + "epoch": 9.84, + "learning_rate": 7.821717282062488e-08, + "loss": 1.3291, + "step": 45104 + }, + { + "epoch": 9.84, + "learning_rate": 7.810793095914355e-08, + "loss": 2.8263, + "step": 45105 + }, + { + "epoch": 9.84, + "learning_rate": 7.799868909766223e-08, + "loss": 0.3417, + "step": 45106 + }, + { + "epoch": 9.84, + "learning_rate": 7.788944723618091e-08, + "loss": 1.2379, + "step": 45107 + }, + { + "epoch": 9.84, + "learning_rate": 7.778020537469958e-08, + "loss": 1.119, + "step": 45108 + }, + { + "epoch": 9.84, + "learning_rate": 7.767096351321827e-08, + "loss": 1.5056, + "step": 45109 + }, + { + "epoch": 9.85, + "learning_rate": 7.756172165173695e-08, + "loss": 2.3192, + "step": 45110 + }, + { + "epoch": 9.85, + "learning_rate": 7.745247979025563e-08, + "loss": 2.2545, + "step": 45111 + }, + { + "epoch": 9.85, + "learning_rate": 7.734323792877432e-08, + "loss": 2.3539, + "step": 45112 + }, + { + "epoch": 9.85, + "learning_rate": 7.7233996067293e-08, + "loss": 2.0949, + "step": 45113 + }, + { + "epoch": 9.85, + "learning_rate": 7.712475420581167e-08, + "loss": 1.7055, + "step": 45114 + }, + { + "epoch": 9.85, + "learning_rate": 7.701551234433035e-08, + "loss": 0.4315, + "step": 45115 + }, + { + "epoch": 9.85, + "learning_rate": 7.690627048284903e-08, + "loss": 0.9295, + "step": 45116 + }, + { + "epoch": 9.85, + "learning_rate": 7.679702862136772e-08, + "loss": 1.1131, + "step": 45117 + }, + { + "epoch": 9.85, + "learning_rate": 7.66877867598864e-08, + "loss": 0.6285, + "step": 45118 + }, + { + "epoch": 9.85, + "learning_rate": 7.657854489840507e-08, + "loss": 0.2876, + "step": 45119 + }, + { + "epoch": 9.85, + "learning_rate": 7.646930303692375e-08, + "loss": 0.5733, + "step": 45120 + }, + { + "epoch": 9.85, + "learning_rate": 7.636006117544244e-08, + "loss": 0.2124, + "step": 45121 + }, + { + "epoch": 9.85, + "learning_rate": 7.625081931396112e-08, + "loss": 1.189, + "step": 45122 + }, + { + "epoch": 9.85, + "learning_rate": 7.61415774524798e-08, + "loss": 1.5601, + "step": 45123 + }, + { + "epoch": 9.85, + "learning_rate": 7.603233559099847e-08, + "loss": 0.6754, + "step": 45124 + }, + { + "epoch": 9.85, + "learning_rate": 7.592309372951715e-08, + "loss": 0.6436, + "step": 45125 + }, + { + "epoch": 9.85, + "learning_rate": 7.581385186803584e-08, + "loss": 0.5006, + "step": 45126 + }, + { + "epoch": 9.85, + "learning_rate": 7.570461000655452e-08, + "loss": 0.741, + "step": 45127 + }, + { + "epoch": 9.85, + "learning_rate": 7.559536814507321e-08, + "loss": 0.9112, + "step": 45128 + }, + { + "epoch": 9.85, + "learning_rate": 7.548612628359188e-08, + "loss": 2.7759, + "step": 45129 + }, + { + "epoch": 9.85, + "learning_rate": 7.537688442211056e-08, + "loss": 2.0702, + "step": 45130 + }, + { + "epoch": 9.85, + "learning_rate": 7.526764256062924e-08, + "loss": 2.4794, + "step": 45131 + }, + { + "epoch": 9.85, + "learning_rate": 7.515840069914792e-08, + "loss": 1.1482, + "step": 45132 + }, + { + "epoch": 9.85, + "learning_rate": 7.50491588376666e-08, + "loss": 0.8055, + "step": 45133 + }, + { + "epoch": 9.85, + "learning_rate": 7.493991697618528e-08, + "loss": 1.5573, + "step": 45134 + }, + { + "epoch": 9.85, + "learning_rate": 7.483067511470396e-08, + "loss": 2.3495, + "step": 45135 + }, + { + "epoch": 9.85, + "learning_rate": 7.472143325322264e-08, + "loss": 0.9038, + "step": 45136 + }, + { + "epoch": 9.85, + "learning_rate": 7.461219139174131e-08, + "loss": 0.1801, + "step": 45137 + }, + { + "epoch": 9.85, + "learning_rate": 7.450294953026e-08, + "loss": 0.5574, + "step": 45138 + }, + { + "epoch": 9.85, + "learning_rate": 7.439370766877868e-08, + "loss": 2.0735, + "step": 45139 + }, + { + "epoch": 9.85, + "learning_rate": 7.428446580729737e-08, + "loss": 1.647, + "step": 45140 + }, + { + "epoch": 9.85, + "learning_rate": 7.417522394581604e-08, + "loss": 0.8159, + "step": 45141 + }, + { + "epoch": 9.85, + "learning_rate": 7.406598208433471e-08, + "loss": 0.1379, + "step": 45142 + }, + { + "epoch": 9.85, + "learning_rate": 7.39567402228534e-08, + "loss": 1.3005, + "step": 45143 + }, + { + "epoch": 9.85, + "learning_rate": 7.384749836137208e-08, + "loss": 0.764, + "step": 45144 + }, + { + "epoch": 9.85, + "learning_rate": 7.373825649989077e-08, + "loss": 1.134, + "step": 45145 + }, + { + "epoch": 9.85, + "learning_rate": 7.362901463840945e-08, + "loss": 1.0774, + "step": 45146 + }, + { + "epoch": 9.85, + "learning_rate": 7.351977277692813e-08, + "loss": 2.1096, + "step": 45147 + }, + { + "epoch": 9.85, + "learning_rate": 7.34105309154468e-08, + "loss": 1.364, + "step": 45148 + }, + { + "epoch": 9.85, + "learning_rate": 7.330128905396548e-08, + "loss": 2.4403, + "step": 45149 + }, + { + "epoch": 9.85, + "learning_rate": 7.319204719248417e-08, + "loss": 0.7095, + "step": 45150 + }, + { + "epoch": 9.85, + "learning_rate": 7.308280533100285e-08, + "loss": 1.1702, + "step": 45151 + }, + { + "epoch": 9.85, + "learning_rate": 7.297356346952152e-08, + "loss": 1.0369, + "step": 45152 + }, + { + "epoch": 9.85, + "learning_rate": 7.28643216080402e-08, + "loss": 2.7208, + "step": 45153 + }, + { + "epoch": 9.85, + "learning_rate": 7.275507974655888e-08, + "loss": 0.5081, + "step": 45154 + }, + { + "epoch": 9.85, + "learning_rate": 7.264583788507757e-08, + "loss": 0.657, + "step": 45155 + }, + { + "epoch": 9.86, + "learning_rate": 7.253659602359625e-08, + "loss": 0.7741, + "step": 45156 + }, + { + "epoch": 9.86, + "learning_rate": 7.242735416211494e-08, + "loss": 1.1332, + "step": 45157 + }, + { + "epoch": 9.86, + "learning_rate": 7.231811230063361e-08, + "loss": 0.394, + "step": 45158 + }, + { + "epoch": 9.86, + "learning_rate": 7.220887043915229e-08, + "loss": 1.2246, + "step": 45159 + }, + { + "epoch": 9.86, + "learning_rate": 7.209962857767097e-08, + "loss": 1.6558, + "step": 45160 + }, + { + "epoch": 9.86, + "learning_rate": 7.199038671618965e-08, + "loss": 3.6343, + "step": 45161 + }, + { + "epoch": 9.86, + "learning_rate": 7.188114485470834e-08, + "loss": 1.5207, + "step": 45162 + }, + { + "epoch": 9.86, + "learning_rate": 7.177190299322701e-08, + "loss": 1.0787, + "step": 45163 + }, + { + "epoch": 9.86, + "learning_rate": 7.166266113174569e-08, + "loss": 2.0708, + "step": 45164 + }, + { + "epoch": 9.86, + "learning_rate": 7.155341927026437e-08, + "loss": 0.377, + "step": 45165 + }, + { + "epoch": 9.86, + "learning_rate": 7.144417740878304e-08, + "loss": 1.0613, + "step": 45166 + }, + { + "epoch": 9.86, + "learning_rate": 7.133493554730174e-08, + "loss": 1.0454, + "step": 45167 + }, + { + "epoch": 9.86, + "learning_rate": 7.122569368582041e-08, + "loss": 1.4478, + "step": 45168 + }, + { + "epoch": 9.86, + "learning_rate": 7.111645182433909e-08, + "loss": 0.804, + "step": 45169 + }, + { + "epoch": 9.86, + "learning_rate": 7.100720996285777e-08, + "loss": 1.0145, + "step": 45170 + }, + { + "epoch": 9.86, + "learning_rate": 7.089796810137646e-08, + "loss": 1.2204, + "step": 45171 + }, + { + "epoch": 9.86, + "learning_rate": 7.078872623989513e-08, + "loss": 0.8019, + "step": 45172 + }, + { + "epoch": 9.86, + "learning_rate": 7.067948437841381e-08, + "loss": 0.2329, + "step": 45173 + }, + { + "epoch": 9.86, + "learning_rate": 7.05702425169325e-08, + "loss": 1.3062, + "step": 45174 + }, + { + "epoch": 9.86, + "learning_rate": 7.046100065545118e-08, + "loss": 2.2091, + "step": 45175 + }, + { + "epoch": 9.86, + "learning_rate": 7.035175879396986e-08, + "loss": 1.7977, + "step": 45176 + }, + { + "epoch": 9.86, + "learning_rate": 7.024251693248853e-08, + "loss": 0.6755, + "step": 45177 + }, + { + "epoch": 9.86, + "learning_rate": 7.013327507100722e-08, + "loss": 1.3111, + "step": 45178 + }, + { + "epoch": 9.86, + "learning_rate": 7.00240332095259e-08, + "loss": 1.5652, + "step": 45179 + }, + { + "epoch": 9.86, + "learning_rate": 6.991479134804458e-08, + "loss": 1.3793, + "step": 45180 + }, + { + "epoch": 9.86, + "learning_rate": 6.980554948656326e-08, + "loss": 1.8358, + "step": 45181 + }, + { + "epoch": 9.86, + "learning_rate": 6.969630762508193e-08, + "loss": 3.3723, + "step": 45182 + }, + { + "epoch": 9.86, + "learning_rate": 6.958706576360062e-08, + "loss": 0.6576, + "step": 45183 + }, + { + "epoch": 9.86, + "learning_rate": 6.94778239021193e-08, + "loss": 2.3147, + "step": 45184 + }, + { + "epoch": 9.86, + "learning_rate": 6.936858204063798e-08, + "loss": 1.014, + "step": 45185 + }, + { + "epoch": 9.86, + "learning_rate": 6.925934017915665e-08, + "loss": 0.7543, + "step": 45186 + }, + { + "epoch": 9.86, + "learning_rate": 6.915009831767533e-08, + "loss": 0.4003, + "step": 45187 + }, + { + "epoch": 9.86, + "learning_rate": 6.904085645619402e-08, + "loss": 0.9695, + "step": 45188 + }, + { + "epoch": 9.86, + "learning_rate": 6.89316145947127e-08, + "loss": 2.4629, + "step": 45189 + }, + { + "epoch": 9.86, + "learning_rate": 6.882237273323139e-08, + "loss": 0.8945, + "step": 45190 + }, + { + "epoch": 9.86, + "learning_rate": 6.871313087175007e-08, + "loss": 1.7399, + "step": 45191 + }, + { + "epoch": 9.86, + "learning_rate": 6.860388901026874e-08, + "loss": 1.1729, + "step": 45192 + }, + { + "epoch": 9.86, + "learning_rate": 6.849464714878742e-08, + "loss": 0.8613, + "step": 45193 + }, + { + "epoch": 9.86, + "learning_rate": 6.83854052873061e-08, + "loss": 0.0611, + "step": 45194 + }, + { + "epoch": 9.86, + "learning_rate": 6.827616342582479e-08, + "loss": 0.0741, + "step": 45195 + }, + { + "epoch": 9.86, + "learning_rate": 6.816692156434347e-08, + "loss": 0.5933, + "step": 45196 + }, + { + "epoch": 9.86, + "learning_rate": 6.805767970286214e-08, + "loss": 0.557, + "step": 45197 + }, + { + "epoch": 9.86, + "learning_rate": 6.794843784138082e-08, + "loss": 0.9798, + "step": 45198 + }, + { + "epoch": 9.86, + "learning_rate": 6.78391959798995e-08, + "loss": 0.5351, + "step": 45199 + }, + { + "epoch": 9.86, + "learning_rate": 6.772995411841819e-08, + "loss": 0.068, + "step": 45200 + }, + { + "epoch": 9.86, + "learning_rate": 6.762071225693686e-08, + "loss": 0.6248, + "step": 45201 + }, + { + "epoch": 9.87, + "learning_rate": 6.751147039545554e-08, + "loss": 0.409, + "step": 45202 + }, + { + "epoch": 9.87, + "learning_rate": 6.740222853397422e-08, + "loss": 0.915, + "step": 45203 + }, + { + "epoch": 9.87, + "learning_rate": 6.72929866724929e-08, + "loss": 2.2736, + "step": 45204 + }, + { + "epoch": 9.87, + "learning_rate": 6.718374481101159e-08, + "loss": 1.1016, + "step": 45205 + }, + { + "epoch": 9.87, + "learning_rate": 6.707450294953026e-08, + "loss": 0.7195, + "step": 45206 + }, + { + "epoch": 9.87, + "learning_rate": 6.696526108804895e-08, + "loss": 0.3869, + "step": 45207 + }, + { + "epoch": 9.87, + "learning_rate": 6.685601922656763e-08, + "loss": 1.3863, + "step": 45208 + }, + { + "epoch": 9.87, + "learning_rate": 6.674677736508631e-08, + "loss": 1.4199, + "step": 45209 + }, + { + "epoch": 9.87, + "learning_rate": 6.663753550360499e-08, + "loss": 0.0673, + "step": 45210 + }, + { + "epoch": 9.87, + "learning_rate": 6.652829364212366e-08, + "loss": 1.1942, + "step": 45211 + }, + { + "epoch": 9.87, + "learning_rate": 6.641905178064235e-08, + "loss": 1.1567, + "step": 45212 + }, + { + "epoch": 9.87, + "learning_rate": 6.630980991916103e-08, + "loss": 1.7557, + "step": 45213 + }, + { + "epoch": 9.87, + "learning_rate": 6.620056805767971e-08, + "loss": 1.1946, + "step": 45214 + }, + { + "epoch": 9.87, + "learning_rate": 6.609132619619838e-08, + "loss": 0.2174, + "step": 45215 + }, + { + "epoch": 9.87, + "learning_rate": 6.598208433471706e-08, + "loss": 0.8425, + "step": 45216 + }, + { + "epoch": 9.87, + "learning_rate": 6.587284247323575e-08, + "loss": 0.3434, + "step": 45217 + }, + { + "epoch": 9.87, + "learning_rate": 6.576360061175443e-08, + "loss": 0.1922, + "step": 45218 + }, + { + "epoch": 9.87, + "learning_rate": 6.56543587502731e-08, + "loss": 1.6825, + "step": 45219 + }, + { + "epoch": 9.87, + "learning_rate": 6.554511688879178e-08, + "loss": 1.952, + "step": 45220 + }, + { + "epoch": 9.87, + "learning_rate": 6.543587502731047e-08, + "loss": 1.2307, + "step": 45221 + }, + { + "epoch": 9.87, + "learning_rate": 6.532663316582915e-08, + "loss": 1.6301, + "step": 45222 + }, + { + "epoch": 9.87, + "learning_rate": 6.521739130434783e-08, + "loss": 0.6519, + "step": 45223 + }, + { + "epoch": 9.87, + "learning_rate": 6.510814944286652e-08, + "loss": 0.3291, + "step": 45224 + }, + { + "epoch": 9.87, + "learning_rate": 6.49989075813852e-08, + "loss": 0.8158, + "step": 45225 + }, + { + "epoch": 9.87, + "learning_rate": 6.488966571990387e-08, + "loss": 1.4146, + "step": 45226 + }, + { + "epoch": 9.87, + "learning_rate": 6.478042385842255e-08, + "loss": 0.4919, + "step": 45227 + }, + { + "epoch": 9.87, + "learning_rate": 6.467118199694124e-08, + "loss": 1.4144, + "step": 45228 + }, + { + "epoch": 9.87, + "learning_rate": 6.456194013545992e-08, + "loss": 2.5127, + "step": 45229 + }, + { + "epoch": 9.87, + "learning_rate": 6.44526982739786e-08, + "loss": 1.009, + "step": 45230 + }, + { + "epoch": 9.87, + "learning_rate": 6.434345641249727e-08, + "loss": 1.8156, + "step": 45231 + }, + { + "epoch": 9.87, + "learning_rate": 6.423421455101595e-08, + "loss": 0.4635, + "step": 45232 + }, + { + "epoch": 9.87, + "learning_rate": 6.412497268953464e-08, + "loss": 0.3171, + "step": 45233 + }, + { + "epoch": 9.87, + "learning_rate": 6.401573082805332e-08, + "loss": 1.9319, + "step": 45234 + }, + { + "epoch": 9.87, + "learning_rate": 6.3906488966572e-08, + "loss": 1.4235, + "step": 45235 + }, + { + "epoch": 9.87, + "learning_rate": 6.379724710509067e-08, + "loss": 1.1261, + "step": 45236 + }, + { + "epoch": 9.87, + "learning_rate": 6.368800524360935e-08, + "loss": 1.6159, + "step": 45237 + }, + { + "epoch": 9.87, + "learning_rate": 6.357876338212804e-08, + "loss": 1.2048, + "step": 45238 + }, + { + "epoch": 9.87, + "learning_rate": 6.346952152064672e-08, + "loss": 0.5211, + "step": 45239 + }, + { + "epoch": 9.87, + "learning_rate": 6.33602796591654e-08, + "loss": 0.3183, + "step": 45240 + }, + { + "epoch": 9.87, + "learning_rate": 6.325103779768408e-08, + "loss": 1.2815, + "step": 45241 + }, + { + "epoch": 9.87, + "learning_rate": 6.314179593620276e-08, + "loss": 0.4036, + "step": 45242 + }, + { + "epoch": 9.87, + "learning_rate": 6.303255407472144e-08, + "loss": 0.0973, + "step": 45243 + }, + { + "epoch": 9.87, + "learning_rate": 6.292331221324011e-08, + "loss": 1.3579, + "step": 45244 + }, + { + "epoch": 9.87, + "learning_rate": 6.28140703517588e-08, + "loss": 0.6359, + "step": 45245 + }, + { + "epoch": 9.87, + "learning_rate": 6.270482849027748e-08, + "loss": 0.2331, + "step": 45246 + }, + { + "epoch": 9.87, + "learning_rate": 6.259558662879616e-08, + "loss": 1.1258, + "step": 45247 + }, + { + "epoch": 9.88, + "learning_rate": 6.248634476731484e-08, + "loss": 0.7403, + "step": 45248 + }, + { + "epoch": 9.88, + "learning_rate": 6.237710290583351e-08, + "loss": 0.6513, + "step": 45249 + }, + { + "epoch": 9.88, + "learning_rate": 6.22678610443522e-08, + "loss": 1.4301, + "step": 45250 + }, + { + "epoch": 9.88, + "learning_rate": 6.215861918287088e-08, + "loss": 1.9775, + "step": 45251 + }, + { + "epoch": 9.88, + "learning_rate": 6.204937732138956e-08, + "loss": 0.7592, + "step": 45252 + }, + { + "epoch": 9.88, + "learning_rate": 6.194013545990824e-08, + "loss": 1.1781, + "step": 45253 + }, + { + "epoch": 9.88, + "learning_rate": 6.183089359842691e-08, + "loss": 0.7486, + "step": 45254 + }, + { + "epoch": 9.88, + "learning_rate": 6.17216517369456e-08, + "loss": 1.1451, + "step": 45255 + }, + { + "epoch": 9.88, + "learning_rate": 6.161240987546428e-08, + "loss": 0.5435, + "step": 45256 + }, + { + "epoch": 9.88, + "learning_rate": 6.150316801398297e-08, + "loss": 0.4274, + "step": 45257 + }, + { + "epoch": 9.88, + "learning_rate": 6.139392615250165e-08, + "loss": 0.9648, + "step": 45258 + }, + { + "epoch": 9.88, + "learning_rate": 6.128468429102033e-08, + "loss": 1.7739, + "step": 45259 + }, + { + "epoch": 9.88, + "learning_rate": 6.1175442429539e-08, + "loss": 0.4907, + "step": 45260 + }, + { + "epoch": 9.88, + "learning_rate": 6.106620056805768e-08, + "loss": 3.234, + "step": 45261 + }, + { + "epoch": 9.88, + "learning_rate": 6.095695870657637e-08, + "loss": 0.9978, + "step": 45262 + }, + { + "epoch": 9.88, + "learning_rate": 6.084771684509505e-08, + "loss": 0.5511, + "step": 45263 + }, + { + "epoch": 9.88, + "learning_rate": 6.073847498361372e-08, + "loss": 1.8023, + "step": 45264 + }, + { + "epoch": 9.88, + "learning_rate": 6.06292331221324e-08, + "loss": 0.5719, + "step": 45265 + }, + { + "epoch": 9.88, + "learning_rate": 6.051999126065108e-08, + "loss": 0.7188, + "step": 45266 + }, + { + "epoch": 9.88, + "learning_rate": 6.041074939916977e-08, + "loss": 0.3368, + "step": 45267 + }, + { + "epoch": 9.88, + "learning_rate": 6.030150753768845e-08, + "loss": 0.1467, + "step": 45268 + }, + { + "epoch": 9.88, + "learning_rate": 6.019226567620712e-08, + "loss": 3.4459, + "step": 45269 + }, + { + "epoch": 9.88, + "learning_rate": 6.00830238147258e-08, + "loss": 1.8493, + "step": 45270 + }, + { + "epoch": 9.88, + "learning_rate": 5.997378195324449e-08, + "loss": 2.7435, + "step": 45271 + }, + { + "epoch": 9.88, + "learning_rate": 5.986454009176317e-08, + "loss": 0.7815, + "step": 45272 + }, + { + "epoch": 9.88, + "learning_rate": 5.975529823028185e-08, + "loss": 0.5472, + "step": 45273 + }, + { + "epoch": 9.88, + "learning_rate": 5.964605636880054e-08, + "loss": 0.9707, + "step": 45274 + }, + { + "epoch": 9.88, + "learning_rate": 5.9536814507319206e-08, + "loss": 1.2252, + "step": 45275 + }, + { + "epoch": 9.88, + "learning_rate": 5.942757264583789e-08, + "loss": 0.4582, + "step": 45276 + }, + { + "epoch": 9.88, + "learning_rate": 5.931833078435657e-08, + "loss": 0.7703, + "step": 45277 + }, + { + "epoch": 9.88, + "learning_rate": 5.9209088922875244e-08, + "loss": 2.0002, + "step": 45278 + }, + { + "epoch": 9.88, + "learning_rate": 5.9099847061393935e-08, + "loss": 0.3655, + "step": 45279 + }, + { + "epoch": 9.88, + "learning_rate": 5.8990605199912605e-08, + "loss": 2.6802, + "step": 45280 + }, + { + "epoch": 9.88, + "learning_rate": 5.8881363338431296e-08, + "loss": 0.9349, + "step": 45281 + }, + { + "epoch": 9.88, + "learning_rate": 5.877212147694997e-08, + "loss": 0.793, + "step": 45282 + }, + { + "epoch": 9.88, + "learning_rate": 5.8662879615468657e-08, + "loss": 1.7596, + "step": 45283 + }, + { + "epoch": 9.88, + "learning_rate": 5.8553637753987334e-08, + "loss": 0.6004, + "step": 45284 + }, + { + "epoch": 9.88, + "learning_rate": 5.844439589250601e-08, + "loss": 0.6631, + "step": 45285 + }, + { + "epoch": 9.88, + "learning_rate": 5.8335154031024695e-08, + "loss": 0.8996, + "step": 45286 + }, + { + "epoch": 9.88, + "learning_rate": 5.822591216954337e-08, + "loss": 1.0387, + "step": 45287 + }, + { + "epoch": 9.88, + "learning_rate": 5.8116670308062056e-08, + "loss": 1.912, + "step": 45288 + }, + { + "epoch": 9.88, + "learning_rate": 5.800742844658073e-08, + "loss": 1.2025, + "step": 45289 + }, + { + "epoch": 9.88, + "learning_rate": 5.7898186585099417e-08, + "loss": 0.8891, + "step": 45290 + }, + { + "epoch": 9.88, + "learning_rate": 5.7788944723618094e-08, + "loss": 1.0252, + "step": 45291 + }, + { + "epoch": 9.88, + "learning_rate": 5.767970286213677e-08, + "loss": 0.8672, + "step": 45292 + }, + { + "epoch": 9.88, + "learning_rate": 5.7570461000655455e-08, + "loss": 1.4838, + "step": 45293 + }, + { + "epoch": 9.89, + "learning_rate": 5.746121913917413e-08, + "loss": 0.769, + "step": 45294 + }, + { + "epoch": 9.89, + "learning_rate": 5.735197727769282e-08, + "loss": 1.5284, + "step": 45295 + }, + { + "epoch": 9.89, + "learning_rate": 5.72427354162115e-08, + "loss": 0.7412, + "step": 45296 + }, + { + "epoch": 9.89, + "learning_rate": 5.713349355473017e-08, + "loss": 1.0638, + "step": 45297 + }, + { + "epoch": 9.89, + "learning_rate": 5.702425169324886e-08, + "loss": 0.5375, + "step": 45298 + }, + { + "epoch": 9.89, + "learning_rate": 5.691500983176754e-08, + "loss": 1.4694, + "step": 45299 + }, + { + "epoch": 9.89, + "learning_rate": 5.680576797028622e-08, + "loss": 1.6001, + "step": 45300 + }, + { + "epoch": 9.89, + "learning_rate": 5.66965261088049e-08, + "loss": 0.4687, + "step": 45301 + }, + { + "epoch": 9.89, + "learning_rate": 5.658728424732358e-08, + "loss": 0.5052, + "step": 45302 + }, + { + "epoch": 9.89, + "learning_rate": 5.647804238584226e-08, + "loss": 1.4796, + "step": 45303 + }, + { + "epoch": 9.89, + "learning_rate": 5.6368800524360937e-08, + "loss": 0.6692, + "step": 45304 + }, + { + "epoch": 9.89, + "learning_rate": 5.625955866287962e-08, + "loss": 0.7054, + "step": 45305 + }, + { + "epoch": 9.89, + "learning_rate": 5.61503168013983e-08, + "loss": 0.8927, + "step": 45306 + }, + { + "epoch": 9.89, + "learning_rate": 5.604107493991698e-08, + "loss": 0.448, + "step": 45307 + }, + { + "epoch": 9.89, + "learning_rate": 5.593183307843566e-08, + "loss": 0.705, + "step": 45308 + }, + { + "epoch": 9.89, + "learning_rate": 5.582259121695435e-08, + "loss": 0.3652, + "step": 45309 + }, + { + "epoch": 9.89, + "learning_rate": 5.571334935547302e-08, + "loss": 1.6549, + "step": 45310 + }, + { + "epoch": 9.89, + "learning_rate": 5.5604107493991697e-08, + "loss": 0.4192, + "step": 45311 + }, + { + "epoch": 9.89, + "learning_rate": 5.549486563251039e-08, + "loss": 3.2803, + "step": 45312 + }, + { + "epoch": 9.89, + "learning_rate": 5.5385623771029064e-08, + "loss": 0.5136, + "step": 45313 + }, + { + "epoch": 9.89, + "learning_rate": 5.527638190954775e-08, + "loss": 0.2972, + "step": 45314 + }, + { + "epoch": 9.89, + "learning_rate": 5.5167140048066425e-08, + "loss": 2.4411, + "step": 45315 + }, + { + "epoch": 9.89, + "learning_rate": 5.50578981865851e-08, + "loss": 1.1787, + "step": 45316 + }, + { + "epoch": 9.89, + "learning_rate": 5.4948656325103786e-08, + "loss": 0.2019, + "step": 45317 + }, + { + "epoch": 9.89, + "learning_rate": 5.483941446362246e-08, + "loss": 0.6688, + "step": 45318 + }, + { + "epoch": 9.89, + "learning_rate": 5.473017260214115e-08, + "loss": 1.8449, + "step": 45319 + }, + { + "epoch": 9.89, + "learning_rate": 5.4620930740659824e-08, + "loss": 0.7022, + "step": 45320 + }, + { + "epoch": 9.89, + "learning_rate": 5.451168887917851e-08, + "loss": 2.2868, + "step": 45321 + }, + { + "epoch": 9.89, + "learning_rate": 5.4402447017697185e-08, + "loss": 0.448, + "step": 45322 + }, + { + "epoch": 9.89, + "learning_rate": 5.429320515621586e-08, + "loss": 0.615, + "step": 45323 + }, + { + "epoch": 9.89, + "learning_rate": 5.4183963294734546e-08, + "loss": 0.3037, + "step": 45324 + }, + { + "epoch": 9.89, + "learning_rate": 5.407472143325322e-08, + "loss": 1.1116, + "step": 45325 + }, + { + "epoch": 9.89, + "learning_rate": 5.3965479571771914e-08, + "loss": 0.6853, + "step": 45326 + }, + { + "epoch": 9.89, + "learning_rate": 5.3856237710290584e-08, + "loss": 1.6693, + "step": 45327 + }, + { + "epoch": 9.89, + "learning_rate": 5.374699584880926e-08, + "loss": 1.8902, + "step": 45328 + }, + { + "epoch": 9.89, + "learning_rate": 5.363775398732795e-08, + "loss": 1.2246, + "step": 45329 + }, + { + "epoch": 9.89, + "learning_rate": 5.352851212584663e-08, + "loss": 0.4999, + "step": 45330 + }, + { + "epoch": 9.89, + "learning_rate": 5.341927026436531e-08, + "loss": 1.5704, + "step": 45331 + }, + { + "epoch": 9.89, + "learning_rate": 5.331002840288399e-08, + "loss": 1.0676, + "step": 45332 + }, + { + "epoch": 9.89, + "learning_rate": 5.3200786541402674e-08, + "loss": 0.2461, + "step": 45333 + }, + { + "epoch": 9.89, + "learning_rate": 5.309154467992135e-08, + "loss": 0.5122, + "step": 45334 + }, + { + "epoch": 9.89, + "learning_rate": 5.298230281844003e-08, + "loss": 0.2798, + "step": 45335 + }, + { + "epoch": 9.89, + "learning_rate": 5.287306095695871e-08, + "loss": 1.864, + "step": 45336 + }, + { + "epoch": 9.89, + "learning_rate": 5.276381909547739e-08, + "loss": 0.9352, + "step": 45337 + }, + { + "epoch": 9.89, + "learning_rate": 5.265457723399607e-08, + "loss": 0.9981, + "step": 45338 + }, + { + "epoch": 9.9, + "learning_rate": 5.254533537251475e-08, + "loss": 0.9758, + "step": 45339 + }, + { + "epoch": 9.9, + "learning_rate": 5.2436093511033433e-08, + "loss": 0.5419, + "step": 45340 + }, + { + "epoch": 9.9, + "learning_rate": 5.232685164955211e-08, + "loss": 2.5982, + "step": 45341 + }, + { + "epoch": 9.9, + "learning_rate": 5.221760978807079e-08, + "loss": 1.119, + "step": 45342 + }, + { + "epoch": 9.9, + "learning_rate": 5.210836792658948e-08, + "loss": 1.434, + "step": 45343 + }, + { + "epoch": 9.9, + "learning_rate": 5.199912606510815e-08, + "loss": 2.2016, + "step": 45344 + }, + { + "epoch": 9.9, + "learning_rate": 5.188988420362684e-08, + "loss": 1.2464, + "step": 45345 + }, + { + "epoch": 9.9, + "learning_rate": 5.1780642342145516e-08, + "loss": 1.0167, + "step": 45346 + }, + { + "epoch": 9.9, + "learning_rate": 5.1671400480664193e-08, + "loss": 1.0582, + "step": 45347 + }, + { + "epoch": 9.9, + "learning_rate": 5.156215861918288e-08, + "loss": 1.1325, + "step": 45348 + }, + { + "epoch": 9.9, + "learning_rate": 5.1452916757701554e-08, + "loss": 2.6404, + "step": 45349 + }, + { + "epoch": 9.9, + "learning_rate": 5.134367489622024e-08, + "loss": 0.7248, + "step": 45350 + }, + { + "epoch": 9.9, + "learning_rate": 5.1234433034738915e-08, + "loss": 0.0886, + "step": 45351 + }, + { + "epoch": 9.9, + "learning_rate": 5.11251911732576e-08, + "loss": 0.8301, + "step": 45352 + }, + { + "epoch": 9.9, + "learning_rate": 5.1015949311776276e-08, + "loss": 0.0915, + "step": 45353 + }, + { + "epoch": 9.9, + "learning_rate": 5.0906707450294953e-08, + "loss": 0.5376, + "step": 45354 + }, + { + "epoch": 9.9, + "learning_rate": 5.079746558881364e-08, + "loss": 0.9046, + "step": 45355 + }, + { + "epoch": 9.9, + "learning_rate": 5.0688223727332314e-08, + "loss": 0.8579, + "step": 45356 + }, + { + "epoch": 9.9, + "learning_rate": 5.0578981865851e-08, + "loss": 0.3249, + "step": 45357 + }, + { + "epoch": 9.9, + "learning_rate": 5.0469740004369675e-08, + "loss": 1.3291, + "step": 45358 + }, + { + "epoch": 9.9, + "learning_rate": 5.036049814288835e-08, + "loss": 0.0745, + "step": 45359 + }, + { + "epoch": 9.9, + "learning_rate": 5.025125628140704e-08, + "loss": 1.7793, + "step": 45360 + }, + { + "epoch": 9.9, + "learning_rate": 5.0142014419925713e-08, + "loss": 0.6508, + "step": 45361 + }, + { + "epoch": 9.9, + "learning_rate": 5.0032772558444404e-08, + "loss": 0.8156, + "step": 45362 + }, + { + "epoch": 9.9, + "learning_rate": 4.992353069696308e-08, + "loss": 2.0597, + "step": 45363 + }, + { + "epoch": 9.9, + "learning_rate": 4.9814288835481765e-08, + "loss": 0.4112, + "step": 45364 + }, + { + "epoch": 9.9, + "learning_rate": 4.970504697400044e-08, + "loss": 0.5412, + "step": 45365 + }, + { + "epoch": 9.9, + "learning_rate": 4.959580511251912e-08, + "loss": 0.7327, + "step": 45366 + }, + { + "epoch": 9.9, + "learning_rate": 4.94865632510378e-08, + "loss": 0.5364, + "step": 45367 + }, + { + "epoch": 9.9, + "learning_rate": 4.937732138955648e-08, + "loss": 1.4593, + "step": 45368 + }, + { + "epoch": 9.9, + "learning_rate": 4.9268079528075164e-08, + "loss": 0.7732, + "step": 45369 + }, + { + "epoch": 9.9, + "learning_rate": 4.915883766659384e-08, + "loss": 3.2343, + "step": 45370 + }, + { + "epoch": 9.9, + "learning_rate": 4.9049595805112525e-08, + "loss": 0.3333, + "step": 45371 + }, + { + "epoch": 9.9, + "learning_rate": 4.89403539436312e-08, + "loss": 0.1468, + "step": 45372 + }, + { + "epoch": 9.9, + "learning_rate": 4.883111208214988e-08, + "loss": 0.7609, + "step": 45373 + }, + { + "epoch": 9.9, + "learning_rate": 4.872187022066856e-08, + "loss": 3.9372, + "step": 45374 + }, + { + "epoch": 9.9, + "learning_rate": 4.861262835918724e-08, + "loss": 0.5427, + "step": 45375 + }, + { + "epoch": 9.9, + "learning_rate": 4.850338649770593e-08, + "loss": 0.6435, + "step": 45376 + }, + { + "epoch": 9.9, + "learning_rate": 4.839414463622461e-08, + "loss": 0.7074, + "step": 45377 + }, + { + "epoch": 9.9, + "learning_rate": 4.8284902774743285e-08, + "loss": 0.0804, + "step": 45378 + }, + { + "epoch": 9.9, + "learning_rate": 4.817566091326197e-08, + "loss": 2.5232, + "step": 45379 + }, + { + "epoch": 9.9, + "learning_rate": 4.8066419051780646e-08, + "loss": 0.5863, + "step": 45380 + }, + { + "epoch": 9.9, + "learning_rate": 4.795717719029933e-08, + "loss": 1.1753, + "step": 45381 + }, + { + "epoch": 9.9, + "learning_rate": 4.7847935328818007e-08, + "loss": 0.5903, + "step": 45382 + }, + { + "epoch": 9.9, + "learning_rate": 4.773869346733669e-08, + "loss": 1.0975, + "step": 45383 + }, + { + "epoch": 9.9, + "learning_rate": 4.762945160585537e-08, + "loss": 0.4087, + "step": 45384 + }, + { + "epoch": 9.91, + "learning_rate": 4.7520209744374045e-08, + "loss": 1.6755, + "step": 45385 + }, + { + "epoch": 9.91, + "learning_rate": 4.741096788289273e-08, + "loss": 0.8975, + "step": 45386 + }, + { + "epoch": 9.91, + "learning_rate": 4.7301726021411406e-08, + "loss": 1.9137, + "step": 45387 + }, + { + "epoch": 9.91, + "learning_rate": 4.719248415993009e-08, + "loss": 1.1074, + "step": 45388 + }, + { + "epoch": 9.91, + "learning_rate": 4.7083242298448767e-08, + "loss": 1.1631, + "step": 45389 + }, + { + "epoch": 9.91, + "learning_rate": 4.697400043696746e-08, + "loss": 1.0637, + "step": 45390 + }, + { + "epoch": 9.91, + "learning_rate": 4.686475857548613e-08, + "loss": 0.1463, + "step": 45391 + }, + { + "epoch": 9.91, + "learning_rate": 4.6755516714004805e-08, + "loss": 1.0283, + "step": 45392 + }, + { + "epoch": 9.91, + "learning_rate": 4.6646274852523495e-08, + "loss": 0.4569, + "step": 45393 + }, + { + "epoch": 9.91, + "learning_rate": 4.653703299104217e-08, + "loss": 2.004, + "step": 45394 + }, + { + "epoch": 9.91, + "learning_rate": 4.6427791129560856e-08, + "loss": 1.8955, + "step": 45395 + }, + { + "epoch": 9.91, + "learning_rate": 4.631854926807953e-08, + "loss": 1.6999, + "step": 45396 + }, + { + "epoch": 9.91, + "learning_rate": 4.620930740659821e-08, + "loss": 2.4977, + "step": 45397 + }, + { + "epoch": 9.91, + "learning_rate": 4.6100065545116894e-08, + "loss": 1.0263, + "step": 45398 + }, + { + "epoch": 9.91, + "learning_rate": 4.599082368363557e-08, + "loss": 0.85, + "step": 45399 + }, + { + "epoch": 9.91, + "learning_rate": 4.5881581822154255e-08, + "loss": 0.8533, + "step": 45400 + }, + { + "epoch": 9.91, + "learning_rate": 4.577233996067293e-08, + "loss": 1.3375, + "step": 45401 + }, + { + "epoch": 9.91, + "learning_rate": 4.5663098099191616e-08, + "loss": 2.8626, + "step": 45402 + }, + { + "epoch": 9.91, + "learning_rate": 4.555385623771029e-08, + "loss": 0.3549, + "step": 45403 + }, + { + "epoch": 9.91, + "learning_rate": 4.544461437622897e-08, + "loss": 0.973, + "step": 45404 + }, + { + "epoch": 9.91, + "learning_rate": 4.5335372514747654e-08, + "loss": 0.5331, + "step": 45405 + }, + { + "epoch": 9.91, + "learning_rate": 4.522613065326633e-08, + "loss": 0.6104, + "step": 45406 + }, + { + "epoch": 9.91, + "learning_rate": 4.511688879178502e-08, + "loss": 0.4715, + "step": 45407 + }, + { + "epoch": 9.91, + "learning_rate": 4.500764693030369e-08, + "loss": 1.2066, + "step": 45408 + }, + { + "epoch": 9.91, + "learning_rate": 4.489840506882237e-08, + "loss": 0.4696, + "step": 45409 + }, + { + "epoch": 9.91, + "learning_rate": 4.478916320734106e-08, + "loss": 1.0602, + "step": 45410 + }, + { + "epoch": 9.91, + "learning_rate": 4.467992134585974e-08, + "loss": 0.1766, + "step": 45411 + }, + { + "epoch": 9.91, + "learning_rate": 4.457067948437842e-08, + "loss": 0.5727, + "step": 45412 + }, + { + "epoch": 9.91, + "learning_rate": 4.44614376228971e-08, + "loss": 0.7975, + "step": 45413 + }, + { + "epoch": 9.91, + "learning_rate": 4.435219576141578e-08, + "loss": 1.1832, + "step": 45414 + }, + { + "epoch": 9.91, + "learning_rate": 4.424295389993446e-08, + "loss": 1.4663, + "step": 45415 + }, + { + "epoch": 9.91, + "learning_rate": 4.4133712038453136e-08, + "loss": 1.5036, + "step": 45416 + }, + { + "epoch": 9.91, + "learning_rate": 4.402447017697182e-08, + "loss": 1.8253, + "step": 45417 + }, + { + "epoch": 9.91, + "learning_rate": 4.39152283154905e-08, + "loss": 2.7157, + "step": 45418 + }, + { + "epoch": 9.91, + "learning_rate": 4.380598645400918e-08, + "loss": 1.1207, + "step": 45419 + }, + { + "epoch": 9.91, + "learning_rate": 4.369674459252786e-08, + "loss": 1.4336, + "step": 45420 + }, + { + "epoch": 9.91, + "learning_rate": 4.358750273104654e-08, + "loss": 1.0448, + "step": 45421 + }, + { + "epoch": 9.91, + "learning_rate": 4.347826086956522e-08, + "loss": 0.8771, + "step": 45422 + }, + { + "epoch": 9.91, + "learning_rate": 4.3369019008083896e-08, + "loss": 2.0685, + "step": 45423 + }, + { + "epoch": 9.91, + "learning_rate": 4.3259777146602586e-08, + "loss": 0.3621, + "step": 45424 + }, + { + "epoch": 9.91, + "learning_rate": 4.315053528512126e-08, + "loss": 0.6312, + "step": 45425 + }, + { + "epoch": 9.91, + "learning_rate": 4.304129342363995e-08, + "loss": 0.878, + "step": 45426 + }, + { + "epoch": 9.91, + "learning_rate": 4.2932051562158625e-08, + "loss": 0.9218, + "step": 45427 + }, + { + "epoch": 9.91, + "learning_rate": 4.28228097006773e-08, + "loss": 2.545, + "step": 45428 + }, + { + "epoch": 9.91, + "learning_rate": 4.2713567839195985e-08, + "loss": 0.0502, + "step": 45429 + }, + { + "epoch": 9.91, + "learning_rate": 4.260432597771466e-08, + "loss": 0.4714, + "step": 45430 + }, + { + "epoch": 9.92, + "learning_rate": 4.2495084116233346e-08, + "loss": 1.551, + "step": 45431 + }, + { + "epoch": 9.92, + "learning_rate": 4.2385842254752024e-08, + "loss": 0.9802, + "step": 45432 + }, + { + "epoch": 9.92, + "learning_rate": 4.227660039327071e-08, + "loss": 1.4336, + "step": 45433 + }, + { + "epoch": 9.92, + "learning_rate": 4.2167358531789385e-08, + "loss": 0.6141, + "step": 45434 + }, + { + "epoch": 9.92, + "learning_rate": 4.205811667030806e-08, + "loss": 1.0785, + "step": 45435 + }, + { + "epoch": 9.92, + "learning_rate": 4.1948874808826745e-08, + "loss": 1.3944, + "step": 45436 + }, + { + "epoch": 9.92, + "learning_rate": 4.183963294734542e-08, + "loss": 2.2406, + "step": 45437 + }, + { + "epoch": 9.92, + "learning_rate": 4.1730391085864106e-08, + "loss": 0.1608, + "step": 45438 + }, + { + "epoch": 9.92, + "learning_rate": 4.1621149224382784e-08, + "loss": 4.0324, + "step": 45439 + }, + { + "epoch": 9.92, + "learning_rate": 4.151190736290146e-08, + "loss": 2.1216, + "step": 45440 + }, + { + "epoch": 9.92, + "learning_rate": 4.140266550142015e-08, + "loss": 1.0932, + "step": 45441 + }, + { + "epoch": 9.92, + "learning_rate": 4.129342363993883e-08, + "loss": 1.7534, + "step": 45442 + }, + { + "epoch": 9.92, + "learning_rate": 4.118418177845751e-08, + "loss": 1.2155, + "step": 45443 + }, + { + "epoch": 9.92, + "learning_rate": 4.107493991697619e-08, + "loss": 0.5482, + "step": 45444 + }, + { + "epoch": 9.92, + "learning_rate": 4.096569805549487e-08, + "loss": 1.0471, + "step": 45445 + }, + { + "epoch": 9.92, + "learning_rate": 4.085645619401355e-08, + "loss": 0.8056, + "step": 45446 + }, + { + "epoch": 9.92, + "learning_rate": 4.074721433253223e-08, + "loss": 2.4291, + "step": 45447 + }, + { + "epoch": 9.92, + "learning_rate": 4.063797247105091e-08, + "loss": 0.1788, + "step": 45448 + }, + { + "epoch": 9.92, + "learning_rate": 4.052873060956959e-08, + "loss": 2.4396, + "step": 45449 + }, + { + "epoch": 9.92, + "learning_rate": 4.041948874808827e-08, + "loss": 0.8847, + "step": 45450 + }, + { + "epoch": 9.92, + "learning_rate": 4.031024688660695e-08, + "loss": 1.555, + "step": 45451 + }, + { + "epoch": 9.92, + "learning_rate": 4.020100502512563e-08, + "loss": 1.0293, + "step": 45452 + }, + { + "epoch": 9.92, + "learning_rate": 4.009176316364431e-08, + "loss": 1.17, + "step": 45453 + }, + { + "epoch": 9.92, + "learning_rate": 3.998252130216299e-08, + "loss": 0.9272, + "step": 45454 + }, + { + "epoch": 9.92, + "learning_rate": 3.987327944068167e-08, + "loss": 1.313, + "step": 45455 + }, + { + "epoch": 9.92, + "learning_rate": 3.976403757920035e-08, + "loss": 1.6015, + "step": 45456 + }, + { + "epoch": 9.92, + "learning_rate": 3.965479571771904e-08, + "loss": 2.2864, + "step": 45457 + }, + { + "epoch": 9.92, + "learning_rate": 3.9545553856237716e-08, + "loss": 1.3044, + "step": 45458 + }, + { + "epoch": 9.92, + "learning_rate": 3.943631199475639e-08, + "loss": 1.8918, + "step": 45459 + }, + { + "epoch": 9.92, + "learning_rate": 3.932707013327508e-08, + "loss": 1.4887, + "step": 45460 + }, + { + "epoch": 9.92, + "learning_rate": 3.9217828271793754e-08, + "loss": 0.0639, + "step": 45461 + }, + { + "epoch": 9.92, + "learning_rate": 3.910858641031244e-08, + "loss": 2.1659, + "step": 45462 + }, + { + "epoch": 9.92, + "learning_rate": 3.8999344548831115e-08, + "loss": 0.2825, + "step": 45463 + }, + { + "epoch": 9.92, + "learning_rate": 3.889010268734979e-08, + "loss": 1.7253, + "step": 45464 + }, + { + "epoch": 9.92, + "learning_rate": 3.8780860825868476e-08, + "loss": 0.8876, + "step": 45465 + }, + { + "epoch": 9.92, + "learning_rate": 3.867161896438716e-08, + "loss": 1.3766, + "step": 45466 + }, + { + "epoch": 9.92, + "learning_rate": 3.856237710290584e-08, + "loss": 0.1348, + "step": 45467 + }, + { + "epoch": 9.92, + "learning_rate": 3.8453135241424514e-08, + "loss": 0.8133, + "step": 45468 + }, + { + "epoch": 9.92, + "learning_rate": 3.83438933799432e-08, + "loss": 1.1301, + "step": 45469 + }, + { + "epoch": 9.92, + "learning_rate": 3.8234651518461875e-08, + "loss": 0.1615, + "step": 45470 + }, + { + "epoch": 9.92, + "learning_rate": 3.812540965698056e-08, + "loss": 1.6637, + "step": 45471 + }, + { + "epoch": 9.92, + "learning_rate": 3.8016167795499236e-08, + "loss": 3.0773, + "step": 45472 + }, + { + "epoch": 9.92, + "learning_rate": 3.790692593401792e-08, + "loss": 0.6596, + "step": 45473 + }, + { + "epoch": 9.92, + "learning_rate": 3.7797684072536603e-08, + "loss": 0.6501, + "step": 45474 + }, + { + "epoch": 9.92, + "learning_rate": 3.768844221105528e-08, + "loss": 0.7683, + "step": 45475 + }, + { + "epoch": 9.92, + "learning_rate": 3.757920034957396e-08, + "loss": 1.3703, + "step": 45476 + }, + { + "epoch": 9.93, + "learning_rate": 3.746995848809264e-08, + "loss": 0.3447, + "step": 45477 + }, + { + "epoch": 9.93, + "learning_rate": 3.736071662661132e-08, + "loss": 1.4307, + "step": 45478 + }, + { + "epoch": 9.93, + "learning_rate": 3.725147476513e-08, + "loss": 0.2338, + "step": 45479 + }, + { + "epoch": 9.93, + "learning_rate": 3.7142232903648686e-08, + "loss": 0.6774, + "step": 45480 + }, + { + "epoch": 9.93, + "learning_rate": 3.703299104216736e-08, + "loss": 0.2302, + "step": 45481 + }, + { + "epoch": 9.93, + "learning_rate": 3.692374918068604e-08, + "loss": 1.7175, + "step": 45482 + }, + { + "epoch": 9.93, + "learning_rate": 3.6814507319204724e-08, + "loss": 0.7286, + "step": 45483 + }, + { + "epoch": 9.93, + "learning_rate": 3.67052654577234e-08, + "loss": 0.5743, + "step": 45484 + }, + { + "epoch": 9.93, + "learning_rate": 3.6596023596242085e-08, + "loss": 0.1371, + "step": 45485 + }, + { + "epoch": 9.93, + "learning_rate": 3.648678173476076e-08, + "loss": 1.1845, + "step": 45486 + }, + { + "epoch": 9.93, + "learning_rate": 3.637753987327944e-08, + "loss": 0.1547, + "step": 45487 + }, + { + "epoch": 9.93, + "learning_rate": 3.6268298011798123e-08, + "loss": 1.5031, + "step": 45488 + }, + { + "epoch": 9.93, + "learning_rate": 3.615905615031681e-08, + "loss": 1.5325, + "step": 45489 + }, + { + "epoch": 9.93, + "learning_rate": 3.6049814288835484e-08, + "loss": 0.7492, + "step": 45490 + }, + { + "epoch": 9.93, + "learning_rate": 3.594057242735417e-08, + "loss": 2.4052, + "step": 45491 + }, + { + "epoch": 9.93, + "learning_rate": 3.5831330565872845e-08, + "loss": 0.1336, + "step": 45492 + }, + { + "epoch": 9.93, + "learning_rate": 3.572208870439152e-08, + "loss": 0.2311, + "step": 45493 + }, + { + "epoch": 9.93, + "learning_rate": 3.5612846842910206e-08, + "loss": 0.1018, + "step": 45494 + }, + { + "epoch": 9.93, + "learning_rate": 3.5503604981428883e-08, + "loss": 1.866, + "step": 45495 + }, + { + "epoch": 9.93, + "learning_rate": 3.539436311994757e-08, + "loss": 0.787, + "step": 45496 + }, + { + "epoch": 9.93, + "learning_rate": 3.528512125846625e-08, + "loss": 2.293, + "step": 45497 + }, + { + "epoch": 9.93, + "learning_rate": 3.517587939698493e-08, + "loss": 1.3762, + "step": 45498 + }, + { + "epoch": 9.93, + "learning_rate": 3.506663753550361e-08, + "loss": 2.1454, + "step": 45499 + }, + { + "epoch": 9.93, + "learning_rate": 3.495739567402229e-08, + "loss": 2.0347, + "step": 45500 + }, + { + "epoch": 9.93, + "learning_rate": 3.4848153812540966e-08, + "loss": 0.6114, + "step": 45501 + }, + { + "epoch": 9.93, + "learning_rate": 3.473891195105965e-08, + "loss": 0.6407, + "step": 45502 + }, + { + "epoch": 9.93, + "learning_rate": 3.462967008957833e-08, + "loss": 1.1416, + "step": 45503 + }, + { + "epoch": 9.93, + "learning_rate": 3.452042822809701e-08, + "loss": 1.2867, + "step": 45504 + }, + { + "epoch": 9.93, + "learning_rate": 3.4411186366615695e-08, + "loss": 0.8551, + "step": 45505 + }, + { + "epoch": 9.93, + "learning_rate": 3.430194450513437e-08, + "loss": 0.8587, + "step": 45506 + }, + { + "epoch": 9.93, + "learning_rate": 3.419270264365305e-08, + "loss": 0.0958, + "step": 45507 + }, + { + "epoch": 9.93, + "learning_rate": 3.408346078217173e-08, + "loss": 2.581, + "step": 45508 + }, + { + "epoch": 9.93, + "learning_rate": 3.397421892069041e-08, + "loss": 0.3046, + "step": 45509 + }, + { + "epoch": 9.93, + "learning_rate": 3.3864977059209094e-08, + "loss": 1.9454, + "step": 45510 + }, + { + "epoch": 9.93, + "learning_rate": 3.375573519772777e-08, + "loss": 1.0267, + "step": 45511 + }, + { + "epoch": 9.93, + "learning_rate": 3.364649333624645e-08, + "loss": 0.2382, + "step": 45512 + }, + { + "epoch": 9.93, + "learning_rate": 3.353725147476513e-08, + "loss": 0.5753, + "step": 45513 + }, + { + "epoch": 9.93, + "learning_rate": 3.3428009613283816e-08, + "loss": 0.1612, + "step": 45514 + }, + { + "epoch": 9.93, + "learning_rate": 3.331876775180249e-08, + "loss": 0.9468, + "step": 45515 + }, + { + "epoch": 9.93, + "learning_rate": 3.3209525890321177e-08, + "loss": 0.1118, + "step": 45516 + }, + { + "epoch": 9.93, + "learning_rate": 3.3100284028839854e-08, + "loss": 0.4663, + "step": 45517 + }, + { + "epoch": 9.93, + "learning_rate": 3.299104216735853e-08, + "loss": 1.0882, + "step": 45518 + }, + { + "epoch": 9.93, + "learning_rate": 3.2881800305877215e-08, + "loss": 1.0506, + "step": 45519 + }, + { + "epoch": 9.93, + "learning_rate": 3.277255844439589e-08, + "loss": 1.6351, + "step": 45520 + }, + { + "epoch": 9.93, + "learning_rate": 3.2663316582914576e-08, + "loss": 2.1299, + "step": 45521 + }, + { + "epoch": 9.93, + "learning_rate": 3.255407472143326e-08, + "loss": 2.4062, + "step": 45522 + }, + { + "epoch": 9.94, + "learning_rate": 3.2444832859951937e-08, + "loss": 0.941, + "step": 45523 + }, + { + "epoch": 9.94, + "learning_rate": 3.233559099847062e-08, + "loss": 1.0672, + "step": 45524 + }, + { + "epoch": 9.94, + "learning_rate": 3.22263491369893e-08, + "loss": 1.0508, + "step": 45525 + }, + { + "epoch": 9.94, + "learning_rate": 3.2117107275507975e-08, + "loss": 1.1816, + "step": 45526 + }, + { + "epoch": 9.94, + "learning_rate": 3.200786541402666e-08, + "loss": 1.4555, + "step": 45527 + }, + { + "epoch": 9.94, + "learning_rate": 3.1898623552545336e-08, + "loss": 1.02, + "step": 45528 + }, + { + "epoch": 9.94, + "learning_rate": 3.178938169106402e-08, + "loss": 1.7153, + "step": 45529 + }, + { + "epoch": 9.94, + "learning_rate": 3.16801398295827e-08, + "loss": 1.6425, + "step": 45530 + }, + { + "epoch": 9.94, + "learning_rate": 3.157089796810138e-08, + "loss": 0.5287, + "step": 45531 + }, + { + "epoch": 9.94, + "learning_rate": 3.146165610662006e-08, + "loss": 0.9462, + "step": 45532 + }, + { + "epoch": 9.94, + "learning_rate": 3.135241424513874e-08, + "loss": 0.7433, + "step": 45533 + }, + { + "epoch": 9.94, + "learning_rate": 3.124317238365742e-08, + "loss": 1.0197, + "step": 45534 + }, + { + "epoch": 9.94, + "learning_rate": 3.11339305221761e-08, + "loss": 1.604, + "step": 45535 + }, + { + "epoch": 9.94, + "learning_rate": 3.102468866069478e-08, + "loss": 0.459, + "step": 45536 + }, + { + "epoch": 9.94, + "learning_rate": 3.0915446799213456e-08, + "loss": 0.1014, + "step": 45537 + }, + { + "epoch": 9.94, + "learning_rate": 3.080620493773214e-08, + "loss": 0.4634, + "step": 45538 + }, + { + "epoch": 9.94, + "learning_rate": 3.0696963076250824e-08, + "loss": 0.4203, + "step": 45539 + }, + { + "epoch": 9.94, + "learning_rate": 3.05877212147695e-08, + "loss": 1.4407, + "step": 45540 + }, + { + "epoch": 9.94, + "learning_rate": 3.0478479353288185e-08, + "loss": 0.4862, + "step": 45541 + }, + { + "epoch": 9.94, + "learning_rate": 3.036923749180686e-08, + "loss": 1.1114, + "step": 45542 + }, + { + "epoch": 9.94, + "learning_rate": 3.025999563032554e-08, + "loss": 0.7887, + "step": 45543 + }, + { + "epoch": 9.94, + "learning_rate": 3.015075376884422e-08, + "loss": 1.5868, + "step": 45544 + }, + { + "epoch": 9.94, + "learning_rate": 3.00415119073629e-08, + "loss": 0.6627, + "step": 45545 + }, + { + "epoch": 9.94, + "learning_rate": 2.9932270045881584e-08, + "loss": 0.0996, + "step": 45546 + }, + { + "epoch": 9.94, + "learning_rate": 2.982302818440027e-08, + "loss": 2.7665, + "step": 45547 + }, + { + "epoch": 9.94, + "learning_rate": 2.9713786322918945e-08, + "loss": 1.1346, + "step": 45548 + }, + { + "epoch": 9.94, + "learning_rate": 2.9604544461437622e-08, + "loss": 1.7105, + "step": 45549 + }, + { + "epoch": 9.94, + "learning_rate": 2.9495302599956303e-08, + "loss": 0.0895, + "step": 45550 + }, + { + "epoch": 9.94, + "learning_rate": 2.9386060738474986e-08, + "loss": 0.4143, + "step": 45551 + }, + { + "epoch": 9.94, + "learning_rate": 2.9276818876993667e-08, + "loss": 1.0703, + "step": 45552 + }, + { + "epoch": 9.94, + "learning_rate": 2.9167577015512347e-08, + "loss": 1.5679, + "step": 45553 + }, + { + "epoch": 9.94, + "learning_rate": 2.9058335154031028e-08, + "loss": 1.3891, + "step": 45554 + }, + { + "epoch": 9.94, + "learning_rate": 2.8949093292549708e-08, + "loss": 1.4552, + "step": 45555 + }, + { + "epoch": 9.94, + "learning_rate": 2.8839851431068385e-08, + "loss": 0.5902, + "step": 45556 + }, + { + "epoch": 9.94, + "learning_rate": 2.8730609569587066e-08, + "loss": 0.099, + "step": 45557 + }, + { + "epoch": 9.94, + "learning_rate": 2.862136770810575e-08, + "loss": 1.4544, + "step": 45558 + }, + { + "epoch": 9.94, + "learning_rate": 2.851212584662443e-08, + "loss": 3.7249, + "step": 45559 + }, + { + "epoch": 9.94, + "learning_rate": 2.840288398514311e-08, + "loss": 0.8406, + "step": 45560 + }, + { + "epoch": 9.94, + "learning_rate": 2.829364212366179e-08, + "loss": 0.0578, + "step": 45561 + }, + { + "epoch": 9.94, + "learning_rate": 2.8184400262180468e-08, + "loss": 0.9933, + "step": 45562 + }, + { + "epoch": 9.94, + "learning_rate": 2.807515840069915e-08, + "loss": 1.6915, + "step": 45563 + }, + { + "epoch": 9.94, + "learning_rate": 2.796591653921783e-08, + "loss": 0.7075, + "step": 45564 + }, + { + "epoch": 9.94, + "learning_rate": 2.785667467773651e-08, + "loss": 0.9868, + "step": 45565 + }, + { + "epoch": 9.94, + "learning_rate": 2.7747432816255193e-08, + "loss": 1.4862, + "step": 45566 + }, + { + "epoch": 9.94, + "learning_rate": 2.7638190954773874e-08, + "loss": 1.1134, + "step": 45567 + }, + { + "epoch": 9.95, + "learning_rate": 2.752894909329255e-08, + "loss": 2.4732, + "step": 45568 + }, + { + "epoch": 9.95, + "learning_rate": 2.741970723181123e-08, + "loss": 2.2933, + "step": 45569 + }, + { + "epoch": 9.95, + "learning_rate": 2.7310465370329912e-08, + "loss": 0.5876, + "step": 45570 + }, + { + "epoch": 9.95, + "learning_rate": 2.7201223508848593e-08, + "loss": 0.7213, + "step": 45571 + }, + { + "epoch": 9.95, + "learning_rate": 2.7091981647367273e-08, + "loss": 0.8641, + "step": 45572 + }, + { + "epoch": 9.95, + "learning_rate": 2.6982739785885957e-08, + "loss": 1.5049, + "step": 45573 + }, + { + "epoch": 9.95, + "learning_rate": 2.687349792440463e-08, + "loss": 0.6161, + "step": 45574 + }, + { + "epoch": 9.95, + "learning_rate": 2.6764256062923314e-08, + "loss": 0.2415, + "step": 45575 + }, + { + "epoch": 9.95, + "learning_rate": 2.6655014201441995e-08, + "loss": 0.727, + "step": 45576 + }, + { + "epoch": 9.95, + "learning_rate": 2.6545772339960675e-08, + "loss": 0.1869, + "step": 45577 + }, + { + "epoch": 9.95, + "learning_rate": 2.6436530478479356e-08, + "loss": 0.2841, + "step": 45578 + }, + { + "epoch": 9.95, + "learning_rate": 2.6327288616998036e-08, + "loss": 0.7506, + "step": 45579 + }, + { + "epoch": 9.95, + "learning_rate": 2.6218046755516717e-08, + "loss": 0.411, + "step": 45580 + }, + { + "epoch": 9.95, + "learning_rate": 2.6108804894035394e-08, + "loss": 0.7654, + "step": 45581 + }, + { + "epoch": 9.95, + "learning_rate": 2.5999563032554074e-08, + "loss": 1.161, + "step": 45582 + }, + { + "epoch": 9.95, + "learning_rate": 2.5890321171072758e-08, + "loss": 0.9202, + "step": 45583 + }, + { + "epoch": 9.95, + "learning_rate": 2.578107930959144e-08, + "loss": 1.2991, + "step": 45584 + }, + { + "epoch": 9.95, + "learning_rate": 2.567183744811012e-08, + "loss": 0.9905, + "step": 45585 + }, + { + "epoch": 9.95, + "learning_rate": 2.55625955866288e-08, + "loss": 0.5959, + "step": 45586 + }, + { + "epoch": 9.95, + "learning_rate": 2.5453353725147477e-08, + "loss": 2.2855, + "step": 45587 + }, + { + "epoch": 9.95, + "learning_rate": 2.5344111863666157e-08, + "loss": 1.4436, + "step": 45588 + }, + { + "epoch": 9.95, + "learning_rate": 2.5234870002184838e-08, + "loss": 1.5825, + "step": 45589 + }, + { + "epoch": 9.95, + "learning_rate": 2.512562814070352e-08, + "loss": 0.6817, + "step": 45590 + }, + { + "epoch": 9.95, + "learning_rate": 2.5016386279222202e-08, + "loss": 0.8328, + "step": 45591 + }, + { + "epoch": 9.95, + "learning_rate": 2.4907144417740882e-08, + "loss": 0.7877, + "step": 45592 + }, + { + "epoch": 9.95, + "learning_rate": 2.479790255625956e-08, + "loss": 1.9461, + "step": 45593 + }, + { + "epoch": 9.95, + "learning_rate": 2.468866069477824e-08, + "loss": 1.0751, + "step": 45594 + }, + { + "epoch": 9.95, + "learning_rate": 2.457941883329692e-08, + "loss": 0.3626, + "step": 45595 + }, + { + "epoch": 9.95, + "learning_rate": 2.44701769718156e-08, + "loss": 0.4967, + "step": 45596 + }, + { + "epoch": 9.95, + "learning_rate": 2.436093511033428e-08, + "loss": 0.7182, + "step": 45597 + }, + { + "epoch": 9.95, + "learning_rate": 2.4251693248852965e-08, + "loss": 2.2053, + "step": 45598 + }, + { + "epoch": 9.95, + "learning_rate": 2.4142451387371642e-08, + "loss": 0.1202, + "step": 45599 + }, + { + "epoch": 9.95, + "learning_rate": 2.4033209525890323e-08, + "loss": 0.9406, + "step": 45600 + }, + { + "epoch": 9.95, + "learning_rate": 2.3923967664409003e-08, + "loss": 0.1029, + "step": 45601 + }, + { + "epoch": 9.95, + "learning_rate": 2.3814725802927684e-08, + "loss": 3.5061, + "step": 45602 + }, + { + "epoch": 9.95, + "learning_rate": 2.3705483941446364e-08, + "loss": 1.4382, + "step": 45603 + }, + { + "epoch": 9.95, + "learning_rate": 2.3596242079965045e-08, + "loss": 0.8877, + "step": 45604 + }, + { + "epoch": 9.95, + "learning_rate": 2.348700021848373e-08, + "loss": 0.6969, + "step": 45605 + }, + { + "epoch": 9.95, + "learning_rate": 2.3377758357002402e-08, + "loss": 1.8175, + "step": 45606 + }, + { + "epoch": 9.95, + "learning_rate": 2.3268516495521086e-08, + "loss": 2.5119, + "step": 45607 + }, + { + "epoch": 9.95, + "learning_rate": 2.3159274634039767e-08, + "loss": 1.8218, + "step": 45608 + }, + { + "epoch": 9.95, + "learning_rate": 2.3050032772558447e-08, + "loss": 0.0797, + "step": 45609 + }, + { + "epoch": 9.95, + "learning_rate": 2.2940790911077128e-08, + "loss": 0.7364, + "step": 45610 + }, + { + "epoch": 9.95, + "learning_rate": 2.2831549049595808e-08, + "loss": 0.253, + "step": 45611 + }, + { + "epoch": 9.95, + "learning_rate": 2.2722307188114485e-08, + "loss": 0.6477, + "step": 45612 + }, + { + "epoch": 9.95, + "learning_rate": 2.2613065326633166e-08, + "loss": 0.4005, + "step": 45613 + }, + { + "epoch": 9.96, + "learning_rate": 2.2503823465151846e-08, + "loss": 1.277, + "step": 45614 + }, + { + "epoch": 9.96, + "learning_rate": 2.239458160367053e-08, + "loss": 0.6169, + "step": 45615 + }, + { + "epoch": 9.96, + "learning_rate": 2.228533974218921e-08, + "loss": 0.3458, + "step": 45616 + }, + { + "epoch": 9.96, + "learning_rate": 2.217609788070789e-08, + "loss": 1.3732, + "step": 45617 + }, + { + "epoch": 9.96, + "learning_rate": 2.2066856019226568e-08, + "loss": 2.1282, + "step": 45618 + }, + { + "epoch": 9.96, + "learning_rate": 2.195761415774525e-08, + "loss": 1.4745, + "step": 45619 + }, + { + "epoch": 9.96, + "learning_rate": 2.184837229626393e-08, + "loss": 0.6769, + "step": 45620 + }, + { + "epoch": 9.96, + "learning_rate": 2.173913043478261e-08, + "loss": 1.1172, + "step": 45621 + }, + { + "epoch": 9.96, + "learning_rate": 2.1629888573301293e-08, + "loss": 0.3563, + "step": 45622 + }, + { + "epoch": 9.96, + "learning_rate": 2.1520646711819974e-08, + "loss": 0.2843, + "step": 45623 + }, + { + "epoch": 9.96, + "learning_rate": 2.141140485033865e-08, + "loss": 2.0701, + "step": 45624 + }, + { + "epoch": 9.96, + "learning_rate": 2.130216298885733e-08, + "loss": 1.4701, + "step": 45625 + }, + { + "epoch": 9.96, + "learning_rate": 2.1192921127376012e-08, + "loss": 0.6983, + "step": 45626 + }, + { + "epoch": 9.96, + "learning_rate": 2.1083679265894692e-08, + "loss": 2.0715, + "step": 45627 + }, + { + "epoch": 9.96, + "learning_rate": 2.0974437404413373e-08, + "loss": 0.4933, + "step": 45628 + }, + { + "epoch": 9.96, + "learning_rate": 2.0865195542932053e-08, + "loss": 2.8619, + "step": 45629 + }, + { + "epoch": 9.96, + "learning_rate": 2.075595368145073e-08, + "loss": 1.0813, + "step": 45630 + }, + { + "epoch": 9.96, + "learning_rate": 2.0646711819969414e-08, + "loss": 0.7099, + "step": 45631 + }, + { + "epoch": 9.96, + "learning_rate": 2.0537469958488095e-08, + "loss": 1.1908, + "step": 45632 + }, + { + "epoch": 9.96, + "learning_rate": 2.0428228097006775e-08, + "loss": 1.9425, + "step": 45633 + }, + { + "epoch": 9.96, + "learning_rate": 2.0318986235525456e-08, + "loss": 2.2084, + "step": 45634 + }, + { + "epoch": 9.96, + "learning_rate": 2.0209744374044136e-08, + "loss": 0.1251, + "step": 45635 + }, + { + "epoch": 9.96, + "learning_rate": 2.0100502512562817e-08, + "loss": 1.5951, + "step": 45636 + }, + { + "epoch": 9.96, + "learning_rate": 1.9991260651081494e-08, + "loss": 0.839, + "step": 45637 + }, + { + "epoch": 9.96, + "learning_rate": 1.9882018789600174e-08, + "loss": 0.2593, + "step": 45638 + }, + { + "epoch": 9.96, + "learning_rate": 1.9772776928118858e-08, + "loss": 1.0658, + "step": 45639 + }, + { + "epoch": 9.96, + "learning_rate": 1.966353506663754e-08, + "loss": 1.1165, + "step": 45640 + }, + { + "epoch": 9.96, + "learning_rate": 1.955429320515622e-08, + "loss": 2.8804, + "step": 45641 + }, + { + "epoch": 9.96, + "learning_rate": 1.9445051343674896e-08, + "loss": 0.6573, + "step": 45642 + }, + { + "epoch": 9.96, + "learning_rate": 1.933580948219358e-08, + "loss": 0.1355, + "step": 45643 + }, + { + "epoch": 9.96, + "learning_rate": 1.9226567620712257e-08, + "loss": 0.9306, + "step": 45644 + }, + { + "epoch": 9.96, + "learning_rate": 1.9117325759230937e-08, + "loss": 0.161, + "step": 45645 + }, + { + "epoch": 9.96, + "learning_rate": 1.9008083897749618e-08, + "loss": 1.5056, + "step": 45646 + }, + { + "epoch": 9.96, + "learning_rate": 1.8898842036268302e-08, + "loss": 1.6664, + "step": 45647 + }, + { + "epoch": 9.96, + "learning_rate": 1.878960017478698e-08, + "loss": 0.6523, + "step": 45648 + }, + { + "epoch": 9.96, + "learning_rate": 1.868035831330566e-08, + "loss": 2.4489, + "step": 45649 + }, + { + "epoch": 9.96, + "learning_rate": 1.8571116451824343e-08, + "loss": 1.2756, + "step": 45650 + }, + { + "epoch": 9.96, + "learning_rate": 1.846187459034302e-08, + "loss": 1.1212, + "step": 45651 + }, + { + "epoch": 9.96, + "learning_rate": 1.83526327288617e-08, + "loss": 1.6424, + "step": 45652 + }, + { + "epoch": 9.96, + "learning_rate": 1.824339086738038e-08, + "loss": 0.3192, + "step": 45653 + }, + { + "epoch": 9.96, + "learning_rate": 1.8134149005899062e-08, + "loss": 1.2523, + "step": 45654 + }, + { + "epoch": 9.96, + "learning_rate": 1.8024907144417742e-08, + "loss": 0.743, + "step": 45655 + }, + { + "epoch": 9.96, + "learning_rate": 1.7915665282936423e-08, + "loss": 2.184, + "step": 45656 + }, + { + "epoch": 9.96, + "learning_rate": 1.7806423421455103e-08, + "loss": 0.1001, + "step": 45657 + }, + { + "epoch": 9.96, + "learning_rate": 1.7697181559973784e-08, + "loss": 0.5594, + "step": 45658 + }, + { + "epoch": 9.96, + "learning_rate": 1.7587939698492464e-08, + "loss": 0.9353, + "step": 45659 + }, + { + "epoch": 9.97, + "learning_rate": 1.7478697837011144e-08, + "loss": 1.1278, + "step": 45660 + }, + { + "epoch": 9.97, + "learning_rate": 1.7369455975529825e-08, + "loss": 1.4409, + "step": 45661 + }, + { + "epoch": 9.97, + "learning_rate": 1.7260214114048505e-08, + "loss": 1.3829, + "step": 45662 + }, + { + "epoch": 9.97, + "learning_rate": 1.7150972252567186e-08, + "loss": 1.0432, + "step": 45663 + }, + { + "epoch": 9.97, + "learning_rate": 1.7041730391085866e-08, + "loss": 1.2301, + "step": 45664 + }, + { + "epoch": 9.97, + "learning_rate": 1.6932488529604547e-08, + "loss": 0.8242, + "step": 45665 + }, + { + "epoch": 9.97, + "learning_rate": 1.6823246668123224e-08, + "loss": 0.3199, + "step": 45666 + }, + { + "epoch": 9.97, + "learning_rate": 1.6714004806641908e-08, + "loss": 1.3663, + "step": 45667 + }, + { + "epoch": 9.97, + "learning_rate": 1.6604762945160588e-08, + "loss": 2.2006, + "step": 45668 + }, + { + "epoch": 9.97, + "learning_rate": 1.6495521083679265e-08, + "loss": 1.939, + "step": 45669 + }, + { + "epoch": 9.97, + "learning_rate": 1.6386279222197946e-08, + "loss": 2.843, + "step": 45670 + }, + { + "epoch": 9.97, + "learning_rate": 1.627703736071663e-08, + "loss": 0.6879, + "step": 45671 + }, + { + "epoch": 9.97, + "learning_rate": 1.616779549923531e-08, + "loss": 0.7562, + "step": 45672 + }, + { + "epoch": 9.97, + "learning_rate": 1.6058553637753987e-08, + "loss": 2.9315, + "step": 45673 + }, + { + "epoch": 9.97, + "learning_rate": 1.5949311776272668e-08, + "loss": 2.1336, + "step": 45674 + }, + { + "epoch": 9.97, + "learning_rate": 1.584006991479135e-08, + "loss": 0.7935, + "step": 45675 + }, + { + "epoch": 9.97, + "learning_rate": 1.573082805331003e-08, + "loss": 1.8318, + "step": 45676 + }, + { + "epoch": 9.97, + "learning_rate": 1.562158619182871e-08, + "loss": 0.6523, + "step": 45677 + }, + { + "epoch": 9.97, + "learning_rate": 1.551234433034739e-08, + "loss": 0.1321, + "step": 45678 + }, + { + "epoch": 9.97, + "learning_rate": 1.540310246886607e-08, + "loss": 1.6777, + "step": 45679 + }, + { + "epoch": 9.97, + "learning_rate": 1.529386060738475e-08, + "loss": 1.3188, + "step": 45680 + }, + { + "epoch": 9.97, + "learning_rate": 1.518461874590343e-08, + "loss": 1.1682, + "step": 45681 + }, + { + "epoch": 9.97, + "learning_rate": 1.507537688442211e-08, + "loss": 0.2107, + "step": 45682 + }, + { + "epoch": 9.97, + "learning_rate": 1.4966135022940792e-08, + "loss": 1.7704, + "step": 45683 + }, + { + "epoch": 9.97, + "learning_rate": 1.4856893161459472e-08, + "loss": 0.366, + "step": 45684 + }, + { + "epoch": 9.97, + "learning_rate": 1.4747651299978151e-08, + "loss": 1.366, + "step": 45685 + }, + { + "epoch": 9.97, + "learning_rate": 1.4638409438496833e-08, + "loss": 1.3265, + "step": 45686 + }, + { + "epoch": 9.97, + "learning_rate": 1.4529167577015514e-08, + "loss": 0.624, + "step": 45687 + }, + { + "epoch": 9.97, + "learning_rate": 1.4419925715534193e-08, + "loss": 1.9281, + "step": 45688 + }, + { + "epoch": 9.97, + "learning_rate": 1.4310683854052875e-08, + "loss": 0.3958, + "step": 45689 + }, + { + "epoch": 9.97, + "learning_rate": 1.4201441992571555e-08, + "loss": 0.997, + "step": 45690 + }, + { + "epoch": 9.97, + "learning_rate": 1.4092200131090234e-08, + "loss": 1.3238, + "step": 45691 + }, + { + "epoch": 9.97, + "learning_rate": 1.3982958269608915e-08, + "loss": 1.999, + "step": 45692 + }, + { + "epoch": 9.97, + "learning_rate": 1.3873716408127597e-08, + "loss": 1.1337, + "step": 45693 + }, + { + "epoch": 9.97, + "learning_rate": 1.3764474546646276e-08, + "loss": 0.6909, + "step": 45694 + }, + { + "epoch": 9.97, + "learning_rate": 1.3655232685164956e-08, + "loss": 0.8646, + "step": 45695 + }, + { + "epoch": 9.97, + "learning_rate": 1.3545990823683636e-08, + "loss": 0.7384, + "step": 45696 + }, + { + "epoch": 9.97, + "learning_rate": 1.3436748962202315e-08, + "loss": 1.5426, + "step": 45697 + }, + { + "epoch": 9.97, + "learning_rate": 1.3327507100720997e-08, + "loss": 1.2831, + "step": 45698 + }, + { + "epoch": 9.97, + "learning_rate": 1.3218265239239678e-08, + "loss": 4.1188, + "step": 45699 + }, + { + "epoch": 9.97, + "learning_rate": 1.3109023377758358e-08, + "loss": 1.066, + "step": 45700 + }, + { + "epoch": 9.97, + "learning_rate": 1.2999781516277037e-08, + "loss": 1.437, + "step": 45701 + }, + { + "epoch": 9.97, + "learning_rate": 1.289053965479572e-08, + "loss": 0.565, + "step": 45702 + }, + { + "epoch": 9.97, + "learning_rate": 1.27812977933144e-08, + "loss": 0.9649, + "step": 45703 + }, + { + "epoch": 9.97, + "learning_rate": 1.2672055931833079e-08, + "loss": 0.4677, + "step": 45704 + }, + { + "epoch": 9.97, + "learning_rate": 1.256281407035176e-08, + "loss": 1.2467, + "step": 45705 + }, + { + "epoch": 9.98, + "learning_rate": 1.2453572208870441e-08, + "loss": 0.3687, + "step": 45706 + }, + { + "epoch": 9.98, + "learning_rate": 1.234433034738912e-08, + "loss": 1.0813, + "step": 45707 + }, + { + "epoch": 9.98, + "learning_rate": 1.22350884859078e-08, + "loss": 1.651, + "step": 45708 + }, + { + "epoch": 9.98, + "learning_rate": 1.2125846624426483e-08, + "loss": 1.4094, + "step": 45709 + }, + { + "epoch": 9.98, + "learning_rate": 1.2016604762945161e-08, + "loss": 1.345, + "step": 45710 + }, + { + "epoch": 9.98, + "learning_rate": 1.1907362901463842e-08, + "loss": 0.3768, + "step": 45711 + }, + { + "epoch": 9.98, + "learning_rate": 1.1798121039982522e-08, + "loss": 2.3225, + "step": 45712 + }, + { + "epoch": 9.98, + "learning_rate": 1.1688879178501201e-08, + "loss": 0.8733, + "step": 45713 + }, + { + "epoch": 9.98, + "learning_rate": 1.1579637317019883e-08, + "loss": 1.5773, + "step": 45714 + }, + { + "epoch": 9.98, + "learning_rate": 1.1470395455538564e-08, + "loss": 1.0695, + "step": 45715 + }, + { + "epoch": 9.98, + "learning_rate": 1.1361153594057243e-08, + "loss": 0.443, + "step": 45716 + }, + { + "epoch": 9.98, + "learning_rate": 1.1251911732575923e-08, + "loss": 0.2784, + "step": 45717 + }, + { + "epoch": 9.98, + "learning_rate": 1.1142669871094605e-08, + "loss": 0.702, + "step": 45718 + }, + { + "epoch": 9.98, + "learning_rate": 1.1033428009613284e-08, + "loss": 0.876, + "step": 45719 + }, + { + "epoch": 9.98, + "learning_rate": 1.0924186148131964e-08, + "loss": 0.9917, + "step": 45720 + }, + { + "epoch": 9.98, + "learning_rate": 1.0814944286650647e-08, + "loss": 0.0898, + "step": 45721 + }, + { + "epoch": 9.98, + "learning_rate": 1.0705702425169325e-08, + "loss": 1.1988, + "step": 45722 + }, + { + "epoch": 9.98, + "learning_rate": 1.0596460563688006e-08, + "loss": 1.2891, + "step": 45723 + }, + { + "epoch": 9.98, + "learning_rate": 1.0487218702206686e-08, + "loss": 2.5742, + "step": 45724 + }, + { + "epoch": 9.98, + "learning_rate": 1.0377976840725365e-08, + "loss": 0.8309, + "step": 45725 + }, + { + "epoch": 9.98, + "learning_rate": 1.0268734979244047e-08, + "loss": 0.8127, + "step": 45726 + }, + { + "epoch": 9.98, + "learning_rate": 1.0159493117762728e-08, + "loss": 0.6226, + "step": 45727 + }, + { + "epoch": 9.98, + "learning_rate": 1.0050251256281408e-08, + "loss": 1.1312, + "step": 45728 + }, + { + "epoch": 9.98, + "learning_rate": 9.941009394800087e-09, + "loss": 1.2101, + "step": 45729 + }, + { + "epoch": 9.98, + "learning_rate": 9.83176753331877e-09, + "loss": 0.9342, + "step": 45730 + }, + { + "epoch": 9.98, + "learning_rate": 9.722525671837448e-09, + "loss": 0.1602, + "step": 45731 + }, + { + "epoch": 9.98, + "learning_rate": 9.613283810356128e-09, + "loss": 1.395, + "step": 45732 + }, + { + "epoch": 9.98, + "learning_rate": 9.504041948874809e-09, + "loss": 1.1011, + "step": 45733 + }, + { + "epoch": 9.98, + "learning_rate": 9.39480008739349e-09, + "loss": 0.4123, + "step": 45734 + }, + { + "epoch": 9.98, + "learning_rate": 9.285558225912172e-09, + "loss": 0.8069, + "step": 45735 + }, + { + "epoch": 9.98, + "learning_rate": 9.17631636443085e-09, + "loss": 0.0607, + "step": 45736 + }, + { + "epoch": 9.98, + "learning_rate": 9.067074502949531e-09, + "loss": 0.1873, + "step": 45737 + }, + { + "epoch": 9.98, + "learning_rate": 8.957832641468211e-09, + "loss": 3.3837, + "step": 45738 + }, + { + "epoch": 9.98, + "learning_rate": 8.848590779986892e-09, + "loss": 1.0421, + "step": 45739 + }, + { + "epoch": 9.98, + "learning_rate": 8.739348918505572e-09, + "loss": 0.2975, + "step": 45740 + }, + { + "epoch": 9.98, + "learning_rate": 8.630107057024253e-09, + "loss": 0.4104, + "step": 45741 + }, + { + "epoch": 9.98, + "learning_rate": 8.520865195542933e-09, + "loss": 1.3922, + "step": 45742 + }, + { + "epoch": 9.98, + "learning_rate": 8.411623334061612e-09, + "loss": 0.3222, + "step": 45743 + }, + { + "epoch": 9.98, + "learning_rate": 8.302381472580294e-09, + "loss": 2.3047, + "step": 45744 + }, + { + "epoch": 9.98, + "learning_rate": 8.193139611098973e-09, + "loss": 0.2524, + "step": 45745 + }, + { + "epoch": 9.98, + "learning_rate": 8.083897749617655e-09, + "loss": 0.3386, + "step": 45746 + }, + { + "epoch": 9.98, + "learning_rate": 7.974655888136334e-09, + "loss": 1.1636, + "step": 45747 + }, + { + "epoch": 9.98, + "learning_rate": 7.865414026655014e-09, + "loss": 0.8658, + "step": 45748 + }, + { + "epoch": 9.98, + "learning_rate": 7.756172165173695e-09, + "loss": 0.2091, + "step": 45749 + }, + { + "epoch": 9.98, + "learning_rate": 7.646930303692375e-09, + "loss": 1.0256, + "step": 45750 + }, + { + "epoch": 9.98, + "learning_rate": 7.537688442211056e-09, + "loss": 3.8332, + "step": 45751 + }, + { + "epoch": 9.99, + "learning_rate": 7.428446580729736e-09, + "loss": 0.7348, + "step": 45752 + }, + { + "epoch": 9.99, + "learning_rate": 7.319204719248417e-09, + "loss": 1.4733, + "step": 45753 + }, + { + "epoch": 9.99, + "learning_rate": 7.209962857767096e-09, + "loss": 2.6207, + "step": 45754 + }, + { + "epoch": 9.99, + "learning_rate": 7.100720996285778e-09, + "loss": 0.5916, + "step": 45755 + }, + { + "epoch": 9.99, + "learning_rate": 6.991479134804457e-09, + "loss": 1.6902, + "step": 45756 + }, + { + "epoch": 9.99, + "learning_rate": 6.882237273323138e-09, + "loss": 1.9254, + "step": 45757 + }, + { + "epoch": 9.99, + "learning_rate": 6.772995411841818e-09, + "loss": 0.4537, + "step": 45758 + }, + { + "epoch": 9.99, + "learning_rate": 6.663753550360499e-09, + "loss": 4.4236, + "step": 45759 + }, + { + "epoch": 9.99, + "learning_rate": 6.554511688879179e-09, + "loss": 1.7531, + "step": 45760 + }, + { + "epoch": 9.99, + "learning_rate": 6.44526982739786e-09, + "loss": 1.1796, + "step": 45761 + }, + { + "epoch": 9.99, + "learning_rate": 6.336027965916539e-09, + "loss": 1.841, + "step": 45762 + }, + { + "epoch": 9.99, + "learning_rate": 6.226786104435221e-09, + "loss": 2.3306, + "step": 45763 + }, + { + "epoch": 9.99, + "learning_rate": 6.1175442429539e-09, + "loss": 0.3504, + "step": 45764 + }, + { + "epoch": 9.99, + "learning_rate": 6.008302381472581e-09, + "loss": 2.1172, + "step": 45765 + }, + { + "epoch": 9.99, + "learning_rate": 5.899060519991261e-09, + "loss": 1.3567, + "step": 45766 + }, + { + "epoch": 9.99, + "learning_rate": 5.789818658509942e-09, + "loss": 0.9352, + "step": 45767 + }, + { + "epoch": 9.99, + "learning_rate": 5.680576797028621e-09, + "loss": 0.2577, + "step": 45768 + }, + { + "epoch": 9.99, + "learning_rate": 5.571334935547303e-09, + "loss": 0.8813, + "step": 45769 + }, + { + "epoch": 9.99, + "learning_rate": 5.462093074065982e-09, + "loss": 1.9246, + "step": 45770 + }, + { + "epoch": 9.99, + "learning_rate": 5.352851212584663e-09, + "loss": 0.9947, + "step": 45771 + }, + { + "epoch": 9.99, + "learning_rate": 5.243609351103343e-09, + "loss": 1.1326, + "step": 45772 + }, + { + "epoch": 9.99, + "learning_rate": 5.134367489622024e-09, + "loss": 1.7984, + "step": 45773 + }, + { + "epoch": 9.99, + "learning_rate": 5.025125628140704e-09, + "loss": 1.9244, + "step": 45774 + }, + { + "epoch": 9.99, + "learning_rate": 4.915883766659385e-09, + "loss": 1.2761, + "step": 45775 + }, + { + "epoch": 9.99, + "learning_rate": 4.806641905178064e-09, + "loss": 0.5841, + "step": 45776 + }, + { + "epoch": 9.99, + "learning_rate": 4.697400043696745e-09, + "loss": 1.8193, + "step": 45777 + }, + { + "epoch": 9.99, + "learning_rate": 4.588158182215425e-09, + "loss": 0.9196, + "step": 45778 + }, + { + "epoch": 9.99, + "learning_rate": 4.478916320734106e-09, + "loss": 0.7131, + "step": 45779 + }, + { + "epoch": 9.99, + "learning_rate": 4.369674459252786e-09, + "loss": 1.5906, + "step": 45780 + }, + { + "epoch": 9.99, + "learning_rate": 4.260432597771467e-09, + "loss": 1.5054, + "step": 45781 + }, + { + "epoch": 9.99, + "learning_rate": 4.151190736290147e-09, + "loss": 1.4688, + "step": 45782 + }, + { + "epoch": 9.99, + "learning_rate": 4.0419488748088275e-09, + "loss": 0.5625, + "step": 45783 + }, + { + "epoch": 9.99, + "learning_rate": 3.932707013327507e-09, + "loss": 1.2002, + "step": 45784 + }, + { + "epoch": 9.99, + "learning_rate": 3.823465151846188e-09, + "loss": 1.3552, + "step": 45785 + }, + { + "epoch": 9.99, + "learning_rate": 3.714223290364868e-09, + "loss": 1.1774, + "step": 45786 + }, + { + "epoch": 9.99, + "learning_rate": 3.604981428883548e-09, + "loss": 1.4942, + "step": 45787 + }, + { + "epoch": 9.99, + "learning_rate": 3.4957395674022287e-09, + "loss": 0.8564, + "step": 45788 + }, + { + "epoch": 9.99, + "learning_rate": 3.386497705920909e-09, + "loss": 0.3316, + "step": 45789 + }, + { + "epoch": 9.99, + "learning_rate": 3.2772558444395896e-09, + "loss": 0.9836, + "step": 45790 + }, + { + "epoch": 9.99, + "learning_rate": 3.1680139829582697e-09, + "loss": 0.9875, + "step": 45791 + }, + { + "epoch": 9.99, + "learning_rate": 3.05877212147695e-09, + "loss": 1.4722, + "step": 45792 + }, + { + "epoch": 9.99, + "learning_rate": 2.9495302599956306e-09, + "loss": 3.3142, + "step": 45793 + }, + { + "epoch": 9.99, + "learning_rate": 2.8402883985143106e-09, + "loss": 1.3094, + "step": 45794 + }, + { + "epoch": 9.99, + "learning_rate": 2.731046537032991e-09, + "loss": 0.0378, + "step": 45795 + }, + { + "epoch": 9.99, + "learning_rate": 2.6218046755516716e-09, + "loss": 0.5498, + "step": 45796 + }, + { + "epoch": 9.99, + "learning_rate": 2.512562814070352e-09, + "loss": 0.6295, + "step": 45797 + }, + { + "epoch": 10.0, + "learning_rate": 2.403320952589032e-09, + "loss": 0.691, + "step": 45798 + }, + { + "epoch": 10.0, + "learning_rate": 2.2940790911077126e-09, + "loss": 0.5615, + "step": 45799 + }, + { + "epoch": 10.0, + "learning_rate": 2.184837229626393e-09, + "loss": 0.0991, + "step": 45800 + }, + { + "epoch": 10.0, + "learning_rate": 2.0755953681450735e-09, + "loss": 0.1378, + "step": 45801 + }, + { + "epoch": 10.0, + "learning_rate": 1.9663535066637536e-09, + "loss": 1.3845, + "step": 45802 + }, + { + "epoch": 10.0, + "learning_rate": 1.857111645182434e-09, + "loss": 0.4832, + "step": 45803 + }, + { + "epoch": 10.0, + "learning_rate": 1.7478697837011143e-09, + "loss": 0.9414, + "step": 45804 + }, + { + "epoch": 10.0, + "learning_rate": 1.6386279222197948e-09, + "loss": 0.4453, + "step": 45805 + }, + { + "epoch": 10.0, + "learning_rate": 1.529386060738475e-09, + "loss": 0.4495, + "step": 45806 + }, + { + "epoch": 10.0, + "learning_rate": 1.4201441992571553e-09, + "loss": 1.4628, + "step": 45807 + }, + { + "epoch": 10.0, + "learning_rate": 1.3109023377758358e-09, + "loss": 3.0021, + "step": 45808 + }, + { + "epoch": 10.0, + "learning_rate": 1.201660476294516e-09, + "loss": 0.9733, + "step": 45809 + }, + { + "epoch": 10.0, + "learning_rate": 1.0924186148131965e-09, + "loss": 0.107, + "step": 45810 + }, + { + "epoch": 10.0, + "learning_rate": 9.831767533318768e-10, + "loss": 0.9754, + "step": 45811 + }, + { + "epoch": 10.0, + "learning_rate": 8.739348918505572e-10, + "loss": 0.6583, + "step": 45812 + }, + { + "epoch": 10.0, + "learning_rate": 7.646930303692375e-10, + "loss": 1.1163, + "step": 45813 + }, + { + "epoch": 10.0, + "learning_rate": 6.554511688879179e-10, + "loss": 0.1137, + "step": 45814 + }, + { + "epoch": 10.0, + "learning_rate": 5.462093074065983e-10, + "loss": 1.0645, + "step": 45815 + }, + { + "epoch": 10.0, + "learning_rate": 4.369674459252786e-10, + "loss": 0.6755, + "step": 45816 + }, + { + "epoch": 10.0, + "learning_rate": 3.2772558444395895e-10, + "loss": 0.6988, + "step": 45817 + }, + { + "epoch": 10.0, + "learning_rate": 2.184837229626393e-10, + "loss": 0.6089, + "step": 45818 + }, + { + "epoch": 10.0, + "learning_rate": 1.0924186148131965e-10, + "loss": 0.2331, + "step": 45819 + }, + { + "epoch": 10.0, + "learning_rate": 0.0, + "loss": 0.5512, + "step": 45820 + }, + { + "epoch": 10.0, + "eval_accuracy": 0.7222585924713584, + "eval_loss": 1.1250884532928467, + "eval_runtime": 339.2494, + "eval_samples_per_second": 18.01, + "eval_steps_per_second": 4.504, + "step": 45820 + } + ], + "max_steps": 45820, + "num_train_epochs": 10, + "total_flos": 6.332427435798505e+18, + "trial_name": null, + "trial_params": null +}