| { | |
| "best_metric": null, | |
| "best_model_checkpoint": null, | |
| "epoch": 0.9998934772573944, | |
| "eval_steps": 500, | |
| "global_step": 1760, | |
| "is_hyper_param_search": false, | |
| "is_local_process_zero": true, | |
| "is_world_process_zero": true, | |
| "log_history": [ | |
| { | |
| "epoch": 0.0, | |
| "learning_rate": 0.0, | |
| "loss": 5.2482, | |
| "step": 1 | |
| }, | |
| { | |
| "epoch": 0.0, | |
| "learning_rate": 0.0, | |
| "loss": 5.0467, | |
| "step": 2 | |
| }, | |
| { | |
| "epoch": 0.0, | |
| "learning_rate": 1.8867924528301888e-05, | |
| "loss": 4.9564, | |
| "step": 3 | |
| }, | |
| { | |
| "epoch": 0.0, | |
| "learning_rate": 1.8867924528301888e-05, | |
| "loss": 5.1658, | |
| "step": 4 | |
| }, | |
| { | |
| "epoch": 0.0, | |
| "learning_rate": 1.8867924528301888e-05, | |
| "loss": 4.9956, | |
| "step": 5 | |
| }, | |
| { | |
| "epoch": 0.0, | |
| "learning_rate": 3.7735849056603776e-05, | |
| "loss": 4.9814, | |
| "step": 6 | |
| }, | |
| { | |
| "epoch": 0.0, | |
| "learning_rate": 5.660377358490566e-05, | |
| "loss": 4.2625, | |
| "step": 7 | |
| }, | |
| { | |
| "epoch": 0.0, | |
| "learning_rate": 7.547169811320755e-05, | |
| "loss": 3.9414, | |
| "step": 8 | |
| }, | |
| { | |
| "epoch": 0.01, | |
| "learning_rate": 9.433962264150944e-05, | |
| "loss": 3.7966, | |
| "step": 9 | |
| }, | |
| { | |
| "epoch": 0.01, | |
| "learning_rate": 0.00011320754716981132, | |
| "loss": 3.6802, | |
| "step": 10 | |
| }, | |
| { | |
| "epoch": 0.01, | |
| "learning_rate": 0.0001320754716981132, | |
| "loss": 3.6004, | |
| "step": 11 | |
| }, | |
| { | |
| "epoch": 0.01, | |
| "learning_rate": 0.0001509433962264151, | |
| "loss": 3.5606, | |
| "step": 12 | |
| }, | |
| { | |
| "epoch": 0.01, | |
| "learning_rate": 0.00016981132075471697, | |
| "loss": 3.5021, | |
| "step": 13 | |
| }, | |
| { | |
| "epoch": 0.01, | |
| "learning_rate": 0.00018867924528301889, | |
| "loss": 3.5367, | |
| "step": 14 | |
| }, | |
| { | |
| "epoch": 0.01, | |
| "learning_rate": 0.00020754716981132078, | |
| "loss": 3.4328, | |
| "step": 15 | |
| }, | |
| { | |
| "epoch": 0.01, | |
| "learning_rate": 0.00022641509433962264, | |
| "loss": 3.4319, | |
| "step": 16 | |
| }, | |
| { | |
| "epoch": 0.01, | |
| "learning_rate": 0.00024528301886792453, | |
| "loss": 3.4415, | |
| "step": 17 | |
| }, | |
| { | |
| "epoch": 0.01, | |
| "learning_rate": 0.0002641509433962264, | |
| "loss": 3.4164, | |
| "step": 18 | |
| }, | |
| { | |
| "epoch": 0.01, | |
| "learning_rate": 0.0002830188679245283, | |
| "loss": 3.5059, | |
| "step": 19 | |
| }, | |
| { | |
| "epoch": 0.01, | |
| "learning_rate": 0.0003018867924528302, | |
| "loss": 3.3674, | |
| "step": 20 | |
| }, | |
| { | |
| "epoch": 0.01, | |
| "learning_rate": 0.00032075471698113204, | |
| "loss": 3.4163, | |
| "step": 21 | |
| }, | |
| { | |
| "epoch": 0.01, | |
| "learning_rate": 0.00033962264150943393, | |
| "loss": 3.3629, | |
| "step": 22 | |
| }, | |
| { | |
| "epoch": 0.01, | |
| "learning_rate": 0.0003584905660377358, | |
| "loss": 3.4262, | |
| "step": 23 | |
| }, | |
| { | |
| "epoch": 0.01, | |
| "learning_rate": 0.00037735849056603777, | |
| "loss": 3.4067, | |
| "step": 24 | |
| }, | |
| { | |
| "epoch": 0.01, | |
| "learning_rate": 0.00039622641509433966, | |
| "loss": 3.2488, | |
| "step": 25 | |
| }, | |
| { | |
| "epoch": 0.01, | |
| "learning_rate": 0.00041509433962264155, | |
| "loss": 3.4274, | |
| "step": 26 | |
| }, | |
| { | |
| "epoch": 0.02, | |
| "learning_rate": 0.00043396226415094345, | |
| "loss": 3.2567, | |
| "step": 27 | |
| }, | |
| { | |
| "epoch": 0.02, | |
| "learning_rate": 0.0004528301886792453, | |
| "loss": 3.2996, | |
| "step": 28 | |
| }, | |
| { | |
| "epoch": 0.02, | |
| "learning_rate": 0.0004716981132075472, | |
| "loss": 3.2675, | |
| "step": 29 | |
| }, | |
| { | |
| "epoch": 0.02, | |
| "learning_rate": 0.0004905660377358491, | |
| "loss": 3.166, | |
| "step": 30 | |
| }, | |
| { | |
| "epoch": 0.02, | |
| "learning_rate": 0.000509433962264151, | |
| "loss": 3.1844, | |
| "step": 31 | |
| }, | |
| { | |
| "epoch": 0.02, | |
| "learning_rate": 0.0005283018867924528, | |
| "loss": 3.1868, | |
| "step": 32 | |
| }, | |
| { | |
| "epoch": 0.02, | |
| "learning_rate": 0.0005471698113207547, | |
| "loss": 3.2863, | |
| "step": 33 | |
| }, | |
| { | |
| "epoch": 0.02, | |
| "learning_rate": 0.0005660377358490566, | |
| "loss": 3.1815, | |
| "step": 34 | |
| }, | |
| { | |
| "epoch": 0.02, | |
| "learning_rate": 0.0005849056603773585, | |
| "loss": 3.2916, | |
| "step": 35 | |
| }, | |
| { | |
| "epoch": 0.02, | |
| "learning_rate": 0.0006037735849056604, | |
| "loss": 3.1167, | |
| "step": 36 | |
| }, | |
| { | |
| "epoch": 0.02, | |
| "learning_rate": 0.0006226415094339623, | |
| "loss": 3.1742, | |
| "step": 37 | |
| }, | |
| { | |
| "epoch": 0.02, | |
| "learning_rate": 0.0006415094339622641, | |
| "loss": 3.1693, | |
| "step": 38 | |
| }, | |
| { | |
| "epoch": 0.02, | |
| "learning_rate": 0.000660377358490566, | |
| "loss": 3.1958, | |
| "step": 39 | |
| }, | |
| { | |
| "epoch": 0.02, | |
| "learning_rate": 0.0006792452830188679, | |
| "loss": 3.1499, | |
| "step": 40 | |
| }, | |
| { | |
| "epoch": 0.02, | |
| "learning_rate": 0.0006981132075471698, | |
| "loss": 3.1234, | |
| "step": 41 | |
| }, | |
| { | |
| "epoch": 0.02, | |
| "learning_rate": 0.0007169811320754717, | |
| "loss": 3.0571, | |
| "step": 42 | |
| }, | |
| { | |
| "epoch": 0.02, | |
| "learning_rate": 0.0007358490566037735, | |
| "loss": 3.0833, | |
| "step": 43 | |
| }, | |
| { | |
| "epoch": 0.02, | |
| "learning_rate": 0.0007547169811320755, | |
| "loss": 3.1011, | |
| "step": 44 | |
| }, | |
| { | |
| "epoch": 0.03, | |
| "learning_rate": 0.0007735849056603774, | |
| "loss": 3.0498, | |
| "step": 45 | |
| }, | |
| { | |
| "epoch": 0.03, | |
| "learning_rate": 0.0007924528301886793, | |
| "loss": 3.1238, | |
| "step": 46 | |
| }, | |
| { | |
| "epoch": 0.03, | |
| "learning_rate": 0.0008113207547169812, | |
| "loss": 3.061, | |
| "step": 47 | |
| }, | |
| { | |
| "epoch": 0.03, | |
| "learning_rate": 0.0008301886792452831, | |
| "loss": 3.0054, | |
| "step": 48 | |
| }, | |
| { | |
| "epoch": 0.03, | |
| "learning_rate": 0.000849056603773585, | |
| "loss": 3.0881, | |
| "step": 49 | |
| }, | |
| { | |
| "epoch": 0.03, | |
| "learning_rate": 0.0008679245283018869, | |
| "loss": 3.1222, | |
| "step": 50 | |
| }, | |
| { | |
| "epoch": 0.03, | |
| "learning_rate": 0.0008867924528301887, | |
| "loss": 3.1099, | |
| "step": 51 | |
| }, | |
| { | |
| "epoch": 0.03, | |
| "learning_rate": 0.0009056603773584906, | |
| "loss": 2.9995, | |
| "step": 52 | |
| }, | |
| { | |
| "epoch": 0.03, | |
| "learning_rate": 0.0009245283018867925, | |
| "loss": 3.0292, | |
| "step": 53 | |
| }, | |
| { | |
| "epoch": 0.03, | |
| "learning_rate": 0.0009433962264150943, | |
| "loss": 3.1197, | |
| "step": 54 | |
| }, | |
| { | |
| "epoch": 0.03, | |
| "learning_rate": 0.0009622641509433962, | |
| "loss": 3.0159, | |
| "step": 55 | |
| }, | |
| { | |
| "epoch": 0.03, | |
| "learning_rate": 0.0009811320754716981, | |
| "loss": 2.93, | |
| "step": 56 | |
| }, | |
| { | |
| "epoch": 0.03, | |
| "learning_rate": 0.001, | |
| "loss": 2.9495, | |
| "step": 57 | |
| }, | |
| { | |
| "epoch": 0.03, | |
| "learning_rate": 0.0009999991532161057, | |
| "loss": 2.9419, | |
| "step": 58 | |
| }, | |
| { | |
| "epoch": 0.03, | |
| "learning_rate": 0.0009999966128672907, | |
| "loss": 3.027, | |
| "step": 59 | |
| }, | |
| { | |
| "epoch": 0.03, | |
| "learning_rate": 0.0009999923789621597, | |
| "loss": 3.0178, | |
| "step": 60 | |
| }, | |
| { | |
| "epoch": 0.03, | |
| "learning_rate": 0.0009999864515150534, | |
| "loss": 2.9705, | |
| "step": 61 | |
| }, | |
| { | |
| "epoch": 0.04, | |
| "learning_rate": 0.0009999788305460491, | |
| "loss": 2.927, | |
| "step": 62 | |
| }, | |
| { | |
| "epoch": 0.04, | |
| "learning_rate": 0.0009999695160809597, | |
| "loss": 3.0041, | |
| "step": 63 | |
| }, | |
| { | |
| "epoch": 0.04, | |
| "learning_rate": 0.0009999585081513348, | |
| "loss": 3.0589, | |
| "step": 64 | |
| }, | |
| { | |
| "epoch": 0.04, | |
| "learning_rate": 0.0009999458067944597, | |
| "loss": 2.9795, | |
| "step": 65 | |
| }, | |
| { | |
| "epoch": 0.04, | |
| "learning_rate": 0.0009999314120533555, | |
| "loss": 2.8439, | |
| "step": 66 | |
| }, | |
| { | |
| "epoch": 0.04, | |
| "learning_rate": 0.0009999153239767793, | |
| "loss": 3.0651, | |
| "step": 67 | |
| }, | |
| { | |
| "epoch": 0.04, | |
| "learning_rate": 0.0009998975426192237, | |
| "loss": 3.0304, | |
| "step": 68 | |
| }, | |
| { | |
| "epoch": 0.04, | |
| "learning_rate": 0.000999878068040916, | |
| "loss": 2.9077, | |
| "step": 69 | |
| }, | |
| { | |
| "epoch": 0.04, | |
| "learning_rate": 0.00099985690030782, | |
| "loss": 2.9534, | |
| "step": 70 | |
| }, | |
| { | |
| "epoch": 0.04, | |
| "learning_rate": 0.0009998340394916333, | |
| "loss": 2.9447, | |
| "step": 71 | |
| }, | |
| { | |
| "epoch": 0.04, | |
| "learning_rate": 0.0009998094856697884, | |
| "loss": 2.8214, | |
| "step": 72 | |
| }, | |
| { | |
| "epoch": 0.04, | |
| "learning_rate": 0.0009997832389254527, | |
| "loss": 3.0377, | |
| "step": 73 | |
| }, | |
| { | |
| "epoch": 0.04, | |
| "learning_rate": 0.0009997552993475274, | |
| "loss": 3.0556, | |
| "step": 74 | |
| }, | |
| { | |
| "epoch": 0.04, | |
| "learning_rate": 0.0009997256670306479, | |
| "loss": 2.9563, | |
| "step": 75 | |
| }, | |
| { | |
| "epoch": 0.04, | |
| "learning_rate": 0.0009996943420751824, | |
| "loss": 2.94, | |
| "step": 76 | |
| }, | |
| { | |
| "epoch": 0.04, | |
| "learning_rate": 0.000999661324587233, | |
| "loss": 2.9925, | |
| "step": 77 | |
| }, | |
| { | |
| "epoch": 0.04, | |
| "learning_rate": 0.0009996266146786344, | |
| "loss": 2.9317, | |
| "step": 78 | |
| }, | |
| { | |
| "epoch": 0.04, | |
| "learning_rate": 0.0009995902124669538, | |
| "loss": 3.021, | |
| "step": 79 | |
| }, | |
| { | |
| "epoch": 0.05, | |
| "learning_rate": 0.0009995521180754905, | |
| "loss": 2.9763, | |
| "step": 80 | |
| }, | |
| { | |
| "epoch": 0.05, | |
| "learning_rate": 0.000999512331633275, | |
| "loss": 2.9475, | |
| "step": 81 | |
| }, | |
| { | |
| "epoch": 0.05, | |
| "learning_rate": 0.0009994708532750699, | |
| "loss": 2.9488, | |
| "step": 82 | |
| }, | |
| { | |
| "epoch": 0.05, | |
| "learning_rate": 0.0009994276831413675, | |
| "loss": 2.7804, | |
| "step": 83 | |
| }, | |
| { | |
| "epoch": 0.05, | |
| "learning_rate": 0.0009993828213783913, | |
| "loss": 2.93, | |
| "step": 84 | |
| }, | |
| { | |
| "epoch": 0.05, | |
| "learning_rate": 0.0009993362681380942, | |
| "loss": 2.8958, | |
| "step": 85 | |
| }, | |
| { | |
| "epoch": 0.05, | |
| "learning_rate": 0.000999288023578158, | |
| "loss": 2.9788, | |
| "step": 86 | |
| }, | |
| { | |
| "epoch": 0.05, | |
| "learning_rate": 0.0009992380878619937, | |
| "loss": 2.9375, | |
| "step": 87 | |
| }, | |
| { | |
| "epoch": 0.05, | |
| "learning_rate": 0.0009991864611587405, | |
| "loss": 2.9048, | |
| "step": 88 | |
| }, | |
| { | |
| "epoch": 0.05, | |
| "learning_rate": 0.0009991331436432647, | |
| "loss": 2.8672, | |
| "step": 89 | |
| }, | |
| { | |
| "epoch": 0.05, | |
| "learning_rate": 0.0009990781354961605, | |
| "loss": 2.8791, | |
| "step": 90 | |
| }, | |
| { | |
| "epoch": 0.05, | |
| "learning_rate": 0.0009990214369037474, | |
| "loss": 2.8839, | |
| "step": 91 | |
| }, | |
| { | |
| "epoch": 0.05, | |
| "learning_rate": 0.0009989630480580713, | |
| "loss": 2.988, | |
| "step": 92 | |
| }, | |
| { | |
| "epoch": 0.05, | |
| "learning_rate": 0.0009989029691569037, | |
| "loss": 2.8999, | |
| "step": 93 | |
| }, | |
| { | |
| "epoch": 0.05, | |
| "learning_rate": 0.000998841200403739, | |
| "loss": 2.9038, | |
| "step": 94 | |
| }, | |
| { | |
| "epoch": 0.05, | |
| "learning_rate": 0.0009987777420077974, | |
| "loss": 2.9074, | |
| "step": 95 | |
| }, | |
| { | |
| "epoch": 0.05, | |
| "learning_rate": 0.0009987125941840205, | |
| "loss": 2.8319, | |
| "step": 96 | |
| }, | |
| { | |
| "epoch": 0.06, | |
| "learning_rate": 0.0009986457571530727, | |
| "loss": 2.9062, | |
| "step": 97 | |
| }, | |
| { | |
| "epoch": 0.06, | |
| "learning_rate": 0.0009985772311413403, | |
| "loss": 2.8477, | |
| "step": 98 | |
| }, | |
| { | |
| "epoch": 0.06, | |
| "learning_rate": 0.0009985070163809305, | |
| "loss": 2.8691, | |
| "step": 99 | |
| }, | |
| { | |
| "epoch": 0.06, | |
| "learning_rate": 0.0009984351131096696, | |
| "loss": 2.9212, | |
| "step": 100 | |
| }, | |
| { | |
| "epoch": 0.06, | |
| "learning_rate": 0.0009983615215711041, | |
| "loss": 2.9209, | |
| "step": 101 | |
| }, | |
| { | |
| "epoch": 0.06, | |
| "learning_rate": 0.0009982862420144984, | |
| "loss": 2.9573, | |
| "step": 102 | |
| }, | |
| { | |
| "epoch": 0.06, | |
| "learning_rate": 0.0009982092746948347, | |
| "loss": 2.8668, | |
| "step": 103 | |
| }, | |
| { | |
| "epoch": 0.06, | |
| "learning_rate": 0.0009981306198728116, | |
| "loss": 2.8374, | |
| "step": 104 | |
| }, | |
| { | |
| "epoch": 0.06, | |
| "learning_rate": 0.0009980502778148437, | |
| "loss": 2.8702, | |
| "step": 105 | |
| }, | |
| { | |
| "epoch": 0.06, | |
| "learning_rate": 0.0009979682487930605, | |
| "loss": 2.799, | |
| "step": 106 | |
| }, | |
| { | |
| "epoch": 0.06, | |
| "learning_rate": 0.0009978845330853054, | |
| "loss": 2.8704, | |
| "step": 107 | |
| }, | |
| { | |
| "epoch": 0.06, | |
| "learning_rate": 0.0009977991309751346, | |
| "loss": 2.8714, | |
| "step": 108 | |
| }, | |
| { | |
| "epoch": 0.06, | |
| "learning_rate": 0.000997712042751817, | |
| "loss": 2.8934, | |
| "step": 109 | |
| }, | |
| { | |
| "epoch": 0.06, | |
| "learning_rate": 0.0009976232687103322, | |
| "loss": 2.8024, | |
| "step": 110 | |
| }, | |
| { | |
| "epoch": 0.06, | |
| "learning_rate": 0.0009975328091513696, | |
| "loss": 2.8448, | |
| "step": 111 | |
| }, | |
| { | |
| "epoch": 0.06, | |
| "learning_rate": 0.000997440664381328, | |
| "loss": 2.8474, | |
| "step": 112 | |
| }, | |
| { | |
| "epoch": 0.06, | |
| "learning_rate": 0.0009973468347123148, | |
| "loss": 2.8874, | |
| "step": 113 | |
| }, | |
| { | |
| "epoch": 0.06, | |
| "learning_rate": 0.000997251320462143, | |
| "loss": 2.8774, | |
| "step": 114 | |
| }, | |
| { | |
| "epoch": 0.07, | |
| "learning_rate": 0.000997154121954333, | |
| "loss": 2.8746, | |
| "step": 115 | |
| }, | |
| { | |
| "epoch": 0.07, | |
| "learning_rate": 0.0009970552395181089, | |
| "loss": 2.9059, | |
| "step": 116 | |
| }, | |
| { | |
| "epoch": 0.07, | |
| "learning_rate": 0.000996954673488399, | |
| "loss": 2.8659, | |
| "step": 117 | |
| }, | |
| { | |
| "epoch": 0.07, | |
| "learning_rate": 0.0009968524242058342, | |
| "loss": 2.8242, | |
| "step": 118 | |
| }, | |
| { | |
| "epoch": 0.07, | |
| "learning_rate": 0.0009967484920167466, | |
| "loss": 2.7469, | |
| "step": 119 | |
| }, | |
| { | |
| "epoch": 0.07, | |
| "learning_rate": 0.0009966428772731689, | |
| "loss": 2.9072, | |
| "step": 120 | |
| }, | |
| { | |
| "epoch": 0.07, | |
| "learning_rate": 0.000996535580332832, | |
| "loss": 2.8385, | |
| "step": 121 | |
| }, | |
| { | |
| "epoch": 0.07, | |
| "learning_rate": 0.0009964266015591655, | |
| "loss": 2.8843, | |
| "step": 122 | |
| }, | |
| { | |
| "epoch": 0.07, | |
| "learning_rate": 0.0009963159413212952, | |
| "loss": 2.8877, | |
| "step": 123 | |
| }, | |
| { | |
| "epoch": 0.07, | |
| "learning_rate": 0.0009962035999940425, | |
| "loss": 2.8456, | |
| "step": 124 | |
| }, | |
| { | |
| "epoch": 0.07, | |
| "learning_rate": 0.0009960895779579228, | |
| "loss": 2.8527, | |
| "step": 125 | |
| }, | |
| { | |
| "epoch": 0.07, | |
| "learning_rate": 0.0009959738755991436, | |
| "loss": 2.8255, | |
| "step": 126 | |
| }, | |
| { | |
| "epoch": 0.07, | |
| "learning_rate": 0.000995856493309605, | |
| "loss": 2.8471, | |
| "step": 127 | |
| }, | |
| { | |
| "epoch": 0.07, | |
| "learning_rate": 0.0009957374314868966, | |
| "loss": 2.847, | |
| "step": 128 | |
| }, | |
| { | |
| "epoch": 0.07, | |
| "learning_rate": 0.000995616690534297, | |
| "loss": 2.7666, | |
| "step": 129 | |
| }, | |
| { | |
| "epoch": 0.07, | |
| "learning_rate": 0.000995494270860772, | |
| "loss": 2.855, | |
| "step": 130 | |
| }, | |
| { | |
| "epoch": 0.07, | |
| "learning_rate": 0.0009953701728809736, | |
| "loss": 2.871, | |
| "step": 131 | |
| }, | |
| { | |
| "epoch": 0.07, | |
| "learning_rate": 0.000995244397015239, | |
| "loss": 2.7651, | |
| "step": 132 | |
| }, | |
| { | |
| "epoch": 0.08, | |
| "learning_rate": 0.0009951169436895875, | |
| "loss": 2.8782, | |
| "step": 133 | |
| }, | |
| { | |
| "epoch": 0.08, | |
| "learning_rate": 0.000994987813335721, | |
| "loss": 2.7868, | |
| "step": 134 | |
| }, | |
| { | |
| "epoch": 0.08, | |
| "learning_rate": 0.0009948570063910214, | |
| "loss": 2.821, | |
| "step": 135 | |
| }, | |
| { | |
| "epoch": 0.08, | |
| "learning_rate": 0.00099472452329855, | |
| "loss": 2.7883, | |
| "step": 136 | |
| }, | |
| { | |
| "epoch": 0.08, | |
| "learning_rate": 0.0009945903645070446, | |
| "loss": 2.8112, | |
| "step": 137 | |
| }, | |
| { | |
| "epoch": 0.08, | |
| "learning_rate": 0.0009944545304709192, | |
| "loss": 2.8404, | |
| "step": 138 | |
| }, | |
| { | |
| "epoch": 0.08, | |
| "learning_rate": 0.000994317021650262, | |
| "loss": 2.8906, | |
| "step": 139 | |
| }, | |
| { | |
| "epoch": 0.08, | |
| "learning_rate": 0.0009941778385108347, | |
| "loss": 2.8831, | |
| "step": 140 | |
| }, | |
| { | |
| "epoch": 0.08, | |
| "learning_rate": 0.0009940369815240688, | |
| "loss": 2.7353, | |
| "step": 141 | |
| }, | |
| { | |
| "epoch": 0.08, | |
| "learning_rate": 0.000993894451167066, | |
| "loss": 2.8234, | |
| "step": 142 | |
| }, | |
| { | |
| "epoch": 0.08, | |
| "learning_rate": 0.0009937502479225964, | |
| "loss": 2.7929, | |
| "step": 143 | |
| }, | |
| { | |
| "epoch": 0.08, | |
| "learning_rate": 0.0009936043722790955, | |
| "loss": 2.7202, | |
| "step": 144 | |
| }, | |
| { | |
| "epoch": 0.08, | |
| "learning_rate": 0.0009934568247306642, | |
| "loss": 2.8444, | |
| "step": 145 | |
| }, | |
| { | |
| "epoch": 0.08, | |
| "learning_rate": 0.0009933076057770658, | |
| "loss": 2.8307, | |
| "step": 146 | |
| }, | |
| { | |
| "epoch": 0.08, | |
| "learning_rate": 0.000993156715923725, | |
| "loss": 2.8203, | |
| "step": 147 | |
| }, | |
| { | |
| "epoch": 0.08, | |
| "learning_rate": 0.0009930041556817269, | |
| "loss": 2.7556, | |
| "step": 148 | |
| }, | |
| { | |
| "epoch": 0.08, | |
| "learning_rate": 0.000992849925567813, | |
| "loss": 2.8758, | |
| "step": 149 | |
| }, | |
| { | |
| "epoch": 0.09, | |
| "learning_rate": 0.000992694026104382, | |
| "loss": 2.7549, | |
| "step": 150 | |
| }, | |
| { | |
| "epoch": 0.09, | |
| "learning_rate": 0.0009925364578194861, | |
| "loss": 2.8526, | |
| "step": 151 | |
| }, | |
| { | |
| "epoch": 0.09, | |
| "learning_rate": 0.000992377221246831, | |
| "loss": 2.781, | |
| "step": 152 | |
| }, | |
| { | |
| "epoch": 0.09, | |
| "learning_rate": 0.000992216316925772, | |
| "loss": 2.8254, | |
| "step": 153 | |
| }, | |
| { | |
| "epoch": 0.09, | |
| "learning_rate": 0.0009920537454013144, | |
| "loss": 2.7487, | |
| "step": 154 | |
| }, | |
| { | |
| "epoch": 0.09, | |
| "learning_rate": 0.0009918895072241096, | |
| "loss": 2.7887, | |
| "step": 155 | |
| }, | |
| { | |
| "epoch": 0.09, | |
| "learning_rate": 0.0009917236029504547, | |
| "loss": 2.8059, | |
| "step": 156 | |
| }, | |
| { | |
| "epoch": 0.09, | |
| "learning_rate": 0.00099155603314229, | |
| "loss": 2.8325, | |
| "step": 157 | |
| }, | |
| { | |
| "epoch": 0.09, | |
| "learning_rate": 0.000991386798367197, | |
| "loss": 2.8345, | |
| "step": 158 | |
| }, | |
| { | |
| "epoch": 0.09, | |
| "learning_rate": 0.000991215899198397, | |
| "loss": 2.8227, | |
| "step": 159 | |
| }, | |
| { | |
| "epoch": 0.09, | |
| "learning_rate": 0.0009910433362147488, | |
| "loss": 2.8391, | |
| "step": 160 | |
| }, | |
| { | |
| "epoch": 0.09, | |
| "learning_rate": 0.0009908691100007463, | |
| "loss": 2.8131, | |
| "step": 161 | |
| }, | |
| { | |
| "epoch": 0.09, | |
| "learning_rate": 0.0009906932211465173, | |
| "loss": 2.7959, | |
| "step": 162 | |
| }, | |
| { | |
| "epoch": 0.09, | |
| "learning_rate": 0.0009905156702478214, | |
| "loss": 2.8631, | |
| "step": 163 | |
| }, | |
| { | |
| "epoch": 0.09, | |
| "learning_rate": 0.0009903364579060476, | |
| "loss": 2.7267, | |
| "step": 164 | |
| }, | |
| { | |
| "epoch": 0.09, | |
| "learning_rate": 0.0009901555847282122, | |
| "loss": 2.7919, | |
| "step": 165 | |
| }, | |
| { | |
| "epoch": 0.09, | |
| "learning_rate": 0.0009899730513269573, | |
| "loss": 2.711, | |
| "step": 166 | |
| }, | |
| { | |
| "epoch": 0.09, | |
| "learning_rate": 0.0009897888583205482, | |
| "loss": 2.8072, | |
| "step": 167 | |
| }, | |
| { | |
| "epoch": 0.1, | |
| "learning_rate": 0.0009896030063328717, | |
| "loss": 2.7619, | |
| "step": 168 | |
| }, | |
| { | |
| "epoch": 0.1, | |
| "learning_rate": 0.0009894154959934336, | |
| "loss": 2.7956, | |
| "step": 169 | |
| }, | |
| { | |
| "epoch": 0.1, | |
| "learning_rate": 0.0009892263279373567, | |
| "loss": 2.7723, | |
| "step": 170 | |
| }, | |
| { | |
| "epoch": 0.1, | |
| "learning_rate": 0.0009890355028053792, | |
| "loss": 2.7653, | |
| "step": 171 | |
| }, | |
| { | |
| "epoch": 0.1, | |
| "learning_rate": 0.0009888430212438514, | |
| "loss": 2.7335, | |
| "step": 172 | |
| }, | |
| { | |
| "epoch": 0.1, | |
| "learning_rate": 0.0009886488839047347, | |
| "loss": 2.7963, | |
| "step": 173 | |
| }, | |
| { | |
| "epoch": 0.1, | |
| "learning_rate": 0.0009884530914455983, | |
| "loss": 2.8076, | |
| "step": 174 | |
| }, | |
| { | |
| "epoch": 0.1, | |
| "learning_rate": 0.000988255644529618, | |
| "loss": 2.7262, | |
| "step": 175 | |
| }, | |
| { | |
| "epoch": 0.1, | |
| "learning_rate": 0.0009880565438255733, | |
| "loss": 2.8753, | |
| "step": 176 | |
| }, | |
| { | |
| "epoch": 0.1, | |
| "learning_rate": 0.0009878557900078449, | |
| "loss": 2.7464, | |
| "step": 177 | |
| }, | |
| { | |
| "epoch": 0.1, | |
| "learning_rate": 0.0009876533837564138, | |
| "loss": 2.7399, | |
| "step": 178 | |
| }, | |
| { | |
| "epoch": 0.1, | |
| "learning_rate": 0.0009874493257568569, | |
| "loss": 2.7399, | |
| "step": 179 | |
| }, | |
| { | |
| "epoch": 0.1, | |
| "learning_rate": 0.0009872436167003467, | |
| "loss": 2.7471, | |
| "step": 180 | |
| }, | |
| { | |
| "epoch": 0.1, | |
| "learning_rate": 0.0009870362572836472, | |
| "loss": 2.7681, | |
| "step": 181 | |
| }, | |
| { | |
| "epoch": 0.1, | |
| "learning_rate": 0.0009868272482091134, | |
| "loss": 2.8374, | |
| "step": 182 | |
| }, | |
| { | |
| "epoch": 0.1, | |
| "learning_rate": 0.000986616590184687, | |
| "loss": 2.7107, | |
| "step": 183 | |
| }, | |
| { | |
| "epoch": 0.1, | |
| "learning_rate": 0.0009864042839238954, | |
| "loss": 2.8556, | |
| "step": 184 | |
| }, | |
| { | |
| "epoch": 0.11, | |
| "learning_rate": 0.0009861903301458487, | |
| "loss": 2.7352, | |
| "step": 185 | |
| }, | |
| { | |
| "epoch": 0.11, | |
| "learning_rate": 0.0009859747295752372, | |
| "loss": 2.8007, | |
| "step": 186 | |
| }, | |
| { | |
| "epoch": 0.11, | |
| "learning_rate": 0.0009857574829423297, | |
| "loss": 2.7395, | |
| "step": 187 | |
| }, | |
| { | |
| "epoch": 0.11, | |
| "learning_rate": 0.0009855385909829696, | |
| "loss": 2.8226, | |
| "step": 188 | |
| }, | |
| { | |
| "epoch": 0.11, | |
| "learning_rate": 0.0009853180544385737, | |
| "loss": 2.7514, | |
| "step": 189 | |
| }, | |
| { | |
| "epoch": 0.11, | |
| "learning_rate": 0.000985095874056129, | |
| "loss": 2.7679, | |
| "step": 190 | |
| }, | |
| { | |
| "epoch": 0.11, | |
| "learning_rate": 0.0009848720505881909, | |
| "loss": 2.805, | |
| "step": 191 | |
| }, | |
| { | |
| "epoch": 0.11, | |
| "learning_rate": 0.00098464658479288, | |
| "loss": 2.7142, | |
| "step": 192 | |
| }, | |
| { | |
| "epoch": 0.11, | |
| "learning_rate": 0.0009844194774338792, | |
| "loss": 2.7042, | |
| "step": 193 | |
| }, | |
| { | |
| "epoch": 0.11, | |
| "learning_rate": 0.0009841907292804316, | |
| "loss": 2.7734, | |
| "step": 194 | |
| }, | |
| { | |
| "epoch": 0.11, | |
| "learning_rate": 0.0009839603411073389, | |
| "loss": 2.7753, | |
| "step": 195 | |
| }, | |
| { | |
| "epoch": 0.11, | |
| "learning_rate": 0.0009837283136949566, | |
| "loss": 2.7156, | |
| "step": 196 | |
| }, | |
| { | |
| "epoch": 0.11, | |
| "learning_rate": 0.0009834946478291932, | |
| "loss": 2.6806, | |
| "step": 197 | |
| }, | |
| { | |
| "epoch": 0.11, | |
| "learning_rate": 0.0009832593443015066, | |
| "loss": 2.7494, | |
| "step": 198 | |
| }, | |
| { | |
| "epoch": 0.11, | |
| "learning_rate": 0.000983022403908902, | |
| "loss": 2.6956, | |
| "step": 199 | |
| }, | |
| { | |
| "epoch": 0.11, | |
| "learning_rate": 0.000982783827453928, | |
| "loss": 2.7452, | |
| "step": 200 | |
| }, | |
| { | |
| "epoch": 0.11, | |
| "learning_rate": 0.0009825436157446762, | |
| "loss": 2.6981, | |
| "step": 201 | |
| }, | |
| { | |
| "epoch": 0.11, | |
| "learning_rate": 0.0009823017695947756, | |
| "loss": 2.8186, | |
| "step": 202 | |
| }, | |
| { | |
| "epoch": 0.12, | |
| "learning_rate": 0.0009820582898233921, | |
| "loss": 2.6943, | |
| "step": 203 | |
| }, | |
| { | |
| "epoch": 0.12, | |
| "learning_rate": 0.0009818131772552248, | |
| "loss": 2.7198, | |
| "step": 204 | |
| }, | |
| { | |
| "epoch": 0.12, | |
| "learning_rate": 0.0009815664327205032, | |
| "loss": 2.7194, | |
| "step": 205 | |
| }, | |
| { | |
| "epoch": 0.12, | |
| "learning_rate": 0.0009813180570549841, | |
| "loss": 2.834, | |
| "step": 206 | |
| }, | |
| { | |
| "epoch": 0.12, | |
| "learning_rate": 0.0009810680510999503, | |
| "loss": 2.7756, | |
| "step": 207 | |
| }, | |
| { | |
| "epoch": 0.12, | |
| "learning_rate": 0.0009808164157022053, | |
| "loss": 2.7234, | |
| "step": 208 | |
| }, | |
| { | |
| "epoch": 0.12, | |
| "learning_rate": 0.0009805631517140723, | |
| "loss": 2.7346, | |
| "step": 209 | |
| }, | |
| { | |
| "epoch": 0.12, | |
| "learning_rate": 0.000980308259993391, | |
| "loss": 2.6711, | |
| "step": 210 | |
| }, | |
| { | |
| "epoch": 0.12, | |
| "learning_rate": 0.0009800517414035141, | |
| "loss": 2.7579, | |
| "step": 211 | |
| }, | |
| { | |
| "epoch": 0.12, | |
| "learning_rate": 0.000979793596813305, | |
| "loss": 2.7716, | |
| "step": 212 | |
| }, | |
| { | |
| "epoch": 0.12, | |
| "learning_rate": 0.000979533827097134, | |
| "loss": 2.7547, | |
| "step": 213 | |
| }, | |
| { | |
| "epoch": 0.12, | |
| "learning_rate": 0.0009792724331348767, | |
| "loss": 2.7107, | |
| "step": 214 | |
| }, | |
| { | |
| "epoch": 0.12, | |
| "learning_rate": 0.00097900941581191, | |
| "loss": 2.7257, | |
| "step": 215 | |
| }, | |
| { | |
| "epoch": 0.12, | |
| "learning_rate": 0.0009787447760191091, | |
| "loss": 2.7289, | |
| "step": 216 | |
| }, | |
| { | |
| "epoch": 0.12, | |
| "learning_rate": 0.0009784785146528447, | |
| "loss": 2.6947, | |
| "step": 217 | |
| }, | |
| { | |
| "epoch": 0.12, | |
| "learning_rate": 0.0009782106326149801, | |
| "loss": 2.769, | |
| "step": 218 | |
| }, | |
| { | |
| "epoch": 0.12, | |
| "learning_rate": 0.0009779411308128685, | |
| "loss": 2.7123, | |
| "step": 219 | |
| }, | |
| { | |
| "epoch": 0.12, | |
| "learning_rate": 0.0009776700101593488, | |
| "loss": 2.7408, | |
| "step": 220 | |
| }, | |
| { | |
| "epoch": 0.13, | |
| "learning_rate": 0.0009773972715727433, | |
| "loss": 2.7125, | |
| "step": 221 | |
| }, | |
| { | |
| "epoch": 0.13, | |
| "learning_rate": 0.0009771229159768547, | |
| "loss": 2.6776, | |
| "step": 222 | |
| }, | |
| { | |
| "epoch": 0.13, | |
| "learning_rate": 0.0009768469443009625, | |
| "loss": 2.7475, | |
| "step": 223 | |
| }, | |
| { | |
| "epoch": 0.13, | |
| "learning_rate": 0.0009765693574798203, | |
| "loss": 2.7532, | |
| "step": 224 | |
| }, | |
| { | |
| "epoch": 0.13, | |
| "learning_rate": 0.0009762901564536521, | |
| "loss": 2.6687, | |
| "step": 225 | |
| }, | |
| { | |
| "epoch": 0.13, | |
| "learning_rate": 0.00097600934216815, | |
| "loss": 2.783, | |
| "step": 226 | |
| }, | |
| { | |
| "epoch": 0.13, | |
| "learning_rate": 0.0009757269155744696, | |
| "loss": 2.7581, | |
| "step": 227 | |
| }, | |
| { | |
| "epoch": 0.13, | |
| "learning_rate": 0.0009754428776292286, | |
| "loss": 2.7305, | |
| "step": 228 | |
| }, | |
| { | |
| "epoch": 0.13, | |
| "learning_rate": 0.0009751572292945015, | |
| "loss": 2.7487, | |
| "step": 229 | |
| }, | |
| { | |
| "epoch": 0.13, | |
| "learning_rate": 0.0009748699715378182, | |
| "loss": 2.7499, | |
| "step": 230 | |
| }, | |
| { | |
| "epoch": 0.13, | |
| "learning_rate": 0.0009745811053321596, | |
| "loss": 2.6904, | |
| "step": 231 | |
| }, | |
| { | |
| "epoch": 0.13, | |
| "learning_rate": 0.0009742906316559548, | |
| "loss": 2.7082, | |
| "step": 232 | |
| }, | |
| { | |
| "epoch": 0.13, | |
| "learning_rate": 0.0009739985514930775, | |
| "loss": 2.7261, | |
| "step": 233 | |
| }, | |
| { | |
| "epoch": 0.13, | |
| "learning_rate": 0.0009737048658328426, | |
| "loss": 2.6158, | |
| "step": 234 | |
| }, | |
| { | |
| "epoch": 0.13, | |
| "learning_rate": 0.0009734095756700037, | |
| "loss": 2.6389, | |
| "step": 235 | |
| }, | |
| { | |
| "epoch": 0.13, | |
| "learning_rate": 0.0009731126820047481, | |
| "loss": 2.6846, | |
| "step": 236 | |
| }, | |
| { | |
| "epoch": 0.13, | |
| "learning_rate": 0.0009728141858426952, | |
| "loss": 2.7424, | |
| "step": 237 | |
| }, | |
| { | |
| "epoch": 0.14, | |
| "learning_rate": 0.000972514088194892, | |
| "loss": 2.7509, | |
| "step": 238 | |
| }, | |
| { | |
| "epoch": 0.14, | |
| "learning_rate": 0.0009722123900778096, | |
| "loss": 2.722, | |
| "step": 239 | |
| }, | |
| { | |
| "epoch": 0.14, | |
| "learning_rate": 0.0009719090925133407, | |
| "loss": 2.6854, | |
| "step": 240 | |
| }, | |
| { | |
| "epoch": 0.14, | |
| "learning_rate": 0.0009716041965287953, | |
| "loss": 2.7199, | |
| "step": 241 | |
| }, | |
| { | |
| "epoch": 0.14, | |
| "learning_rate": 0.0009712977031568973, | |
| "loss": 2.6969, | |
| "step": 242 | |
| }, | |
| { | |
| "epoch": 0.14, | |
| "learning_rate": 0.0009709896134357815, | |
| "loss": 2.7208, | |
| "step": 243 | |
| }, | |
| { | |
| "epoch": 0.14, | |
| "learning_rate": 0.0009706799284089891, | |
| "loss": 2.647, | |
| "step": 244 | |
| }, | |
| { | |
| "epoch": 0.14, | |
| "learning_rate": 0.0009703686491254658, | |
| "loss": 2.6342, | |
| "step": 245 | |
| }, | |
| { | |
| "epoch": 0.14, | |
| "learning_rate": 0.0009700557766395565, | |
| "loss": 2.7401, | |
| "step": 246 | |
| }, | |
| { | |
| "epoch": 0.14, | |
| "learning_rate": 0.0009697413120110028, | |
| "loss": 2.701, | |
| "step": 247 | |
| }, | |
| { | |
| "epoch": 0.14, | |
| "learning_rate": 0.000969425256304939, | |
| "loss": 2.6749, | |
| "step": 248 | |
| }, | |
| { | |
| "epoch": 0.14, | |
| "learning_rate": 0.0009691076105918884, | |
| "loss": 2.7221, | |
| "step": 249 | |
| }, | |
| { | |
| "epoch": 0.14, | |
| "learning_rate": 0.0009687883759477605, | |
| "loss": 2.7285, | |
| "step": 250 | |
| }, | |
| { | |
| "epoch": 0.14, | |
| "learning_rate": 0.0009684675534538461, | |
| "loss": 2.704, | |
| "step": 251 | |
| }, | |
| { | |
| "epoch": 0.14, | |
| "learning_rate": 0.0009681451441968143, | |
| "loss": 2.7219, | |
| "step": 252 | |
| }, | |
| { | |
| "epoch": 0.14, | |
| "learning_rate": 0.0009678211492687095, | |
| "loss": 2.6817, | |
| "step": 253 | |
| }, | |
| { | |
| "epoch": 0.14, | |
| "learning_rate": 0.0009674955697669458, | |
| "loss": 2.6934, | |
| "step": 254 | |
| }, | |
| { | |
| "epoch": 0.14, | |
| "learning_rate": 0.0009671684067943056, | |
| "loss": 2.6851, | |
| "step": 255 | |
| }, | |
| { | |
| "epoch": 0.15, | |
| "learning_rate": 0.0009668396614589341, | |
| "loss": 2.6716, | |
| "step": 256 | |
| }, | |
| { | |
| "epoch": 0.15, | |
| "learning_rate": 0.0009665093348743361, | |
| "loss": 2.7255, | |
| "step": 257 | |
| }, | |
| { | |
| "epoch": 0.15, | |
| "learning_rate": 0.0009661774281593729, | |
| "loss": 2.659, | |
| "step": 258 | |
| }, | |
| { | |
| "epoch": 0.15, | |
| "learning_rate": 0.0009658439424382575, | |
| "loss": 2.6577, | |
| "step": 259 | |
| }, | |
| { | |
| "epoch": 0.15, | |
| "learning_rate": 0.0009655088788405508, | |
| "loss": 2.7046, | |
| "step": 260 | |
| }, | |
| { | |
| "epoch": 0.15, | |
| "learning_rate": 0.0009651722385011592, | |
| "loss": 2.6396, | |
| "step": 261 | |
| }, | |
| { | |
| "epoch": 0.15, | |
| "learning_rate": 0.0009648340225603287, | |
| "loss": 2.7347, | |
| "step": 262 | |
| }, | |
| { | |
| "epoch": 0.15, | |
| "learning_rate": 0.000964494232163643, | |
| "loss": 2.6952, | |
| "step": 263 | |
| }, | |
| { | |
| "epoch": 0.15, | |
| "learning_rate": 0.0009641528684620179, | |
| "loss": 2.7599, | |
| "step": 264 | |
| }, | |
| { | |
| "epoch": 0.15, | |
| "learning_rate": 0.0009638099326116987, | |
| "loss": 2.7123, | |
| "step": 265 | |
| }, | |
| { | |
| "epoch": 0.15, | |
| "learning_rate": 0.0009634654257742554, | |
| "loss": 2.6239, | |
| "step": 266 | |
| }, | |
| { | |
| "epoch": 0.15, | |
| "learning_rate": 0.0009631193491165797, | |
| "loss": 2.6331, | |
| "step": 267 | |
| }, | |
| { | |
| "epoch": 0.15, | |
| "learning_rate": 0.0009627717038108799, | |
| "loss": 2.6387, | |
| "step": 268 | |
| }, | |
| { | |
| "epoch": 0.15, | |
| "learning_rate": 0.000962422491034678, | |
| "loss": 2.7134, | |
| "step": 269 | |
| }, | |
| { | |
| "epoch": 0.15, | |
| "learning_rate": 0.0009620717119708047, | |
| "loss": 2.6042, | |
| "step": 270 | |
| }, | |
| { | |
| "epoch": 0.15, | |
| "learning_rate": 0.0009617193678073965, | |
| "loss": 2.7081, | |
| "step": 271 | |
| }, | |
| { | |
| "epoch": 0.15, | |
| "learning_rate": 0.0009613654597378909, | |
| "loss": 2.6387, | |
| "step": 272 | |
| }, | |
| { | |
| "epoch": 0.16, | |
| "learning_rate": 0.0009610099889610223, | |
| "loss": 2.6194, | |
| "step": 273 | |
| }, | |
| { | |
| "epoch": 0.16, | |
| "learning_rate": 0.0009606529566808186, | |
| "loss": 2.6575, | |
| "step": 274 | |
| }, | |
| { | |
| "epoch": 0.16, | |
| "learning_rate": 0.0009602943641065964, | |
| "loss": 2.6457, | |
| "step": 275 | |
| }, | |
| { | |
| "epoch": 0.16, | |
| "learning_rate": 0.0009599342124529575, | |
| "loss": 2.5716, | |
| "step": 276 | |
| }, | |
| { | |
| "epoch": 0.16, | |
| "learning_rate": 0.0009595725029397841, | |
| "loss": 2.6692, | |
| "step": 277 | |
| }, | |
| { | |
| "epoch": 0.16, | |
| "learning_rate": 0.0009592092367922358, | |
| "loss": 2.6708, | |
| "step": 278 | |
| }, | |
| { | |
| "epoch": 0.16, | |
| "learning_rate": 0.0009588444152407441, | |
| "loss": 2.686, | |
| "step": 279 | |
| }, | |
| { | |
| "epoch": 0.16, | |
| "learning_rate": 0.0009584780395210088, | |
| "loss": 2.5811, | |
| "step": 280 | |
| }, | |
| { | |
| "epoch": 0.16, | |
| "learning_rate": 0.0009581101108739944, | |
| "loss": 2.5678, | |
| "step": 281 | |
| }, | |
| { | |
| "epoch": 0.16, | |
| "learning_rate": 0.000957740630545925, | |
| "loss": 2.6341, | |
| "step": 282 | |
| }, | |
| { | |
| "epoch": 0.16, | |
| "learning_rate": 0.0009573695997882808, | |
| "loss": 2.6191, | |
| "step": 283 | |
| }, | |
| { | |
| "epoch": 0.16, | |
| "learning_rate": 0.000956997019857793, | |
| "loss": 2.7044, | |
| "step": 284 | |
| }, | |
| { | |
| "epoch": 0.16, | |
| "learning_rate": 0.0009566228920164405, | |
| "loss": 2.6047, | |
| "step": 285 | |
| }, | |
| { | |
| "epoch": 0.16, | |
| "learning_rate": 0.0009562472175314449, | |
| "loss": 2.6376, | |
| "step": 286 | |
| }, | |
| { | |
| "epoch": 0.16, | |
| "learning_rate": 0.0009558699976752668, | |
| "loss": 2.686, | |
| "step": 287 | |
| }, | |
| { | |
| "epoch": 0.16, | |
| "learning_rate": 0.0009554912337256007, | |
| "loss": 2.6306, | |
| "step": 288 | |
| }, | |
| { | |
| "epoch": 0.16, | |
| "learning_rate": 0.0009551109269653718, | |
| "loss": 2.6262, | |
| "step": 289 | |
| }, | |
| { | |
| "epoch": 0.16, | |
| "learning_rate": 0.0009547290786827303, | |
| "loss": 2.6516, | |
| "step": 290 | |
| }, | |
| { | |
| "epoch": 0.17, | |
| "learning_rate": 0.0009543456901710483, | |
| "loss": 2.6135, | |
| "step": 291 | |
| }, | |
| { | |
| "epoch": 0.17, | |
| "learning_rate": 0.0009539607627289146, | |
| "loss": 2.7414, | |
| "step": 292 | |
| }, | |
| { | |
| "epoch": 0.17, | |
| "learning_rate": 0.0009535742976601309, | |
| "loss": 2.6305, | |
| "step": 293 | |
| }, | |
| { | |
| "epoch": 0.17, | |
| "learning_rate": 0.0009531862962737065, | |
| "loss": 2.6181, | |
| "step": 294 | |
| }, | |
| { | |
| "epoch": 0.17, | |
| "learning_rate": 0.0009527967598838547, | |
| "loss": 2.6371, | |
| "step": 295 | |
| }, | |
| { | |
| "epoch": 0.17, | |
| "learning_rate": 0.0009524056898099881, | |
| "loss": 2.6824, | |
| "step": 296 | |
| }, | |
| { | |
| "epoch": 0.17, | |
| "learning_rate": 0.0009520130873767141, | |
| "loss": 2.5707, | |
| "step": 297 | |
| }, | |
| { | |
| "epoch": 0.17, | |
| "learning_rate": 0.0009516189539138305, | |
| "loss": 2.5983, | |
| "step": 298 | |
| }, | |
| { | |
| "epoch": 0.17, | |
| "learning_rate": 0.0009512232907563206, | |
| "loss": 2.6824, | |
| "step": 299 | |
| }, | |
| { | |
| "epoch": 0.17, | |
| "learning_rate": 0.0009508260992443492, | |
| "loss": 2.6618, | |
| "step": 300 | |
| }, | |
| { | |
| "epoch": 0.17, | |
| "learning_rate": 0.0009504273807232576, | |
| "loss": 2.6094, | |
| "step": 301 | |
| }, | |
| { | |
| "epoch": 0.17, | |
| "learning_rate": 0.0009500271365435599, | |
| "loss": 2.5858, | |
| "step": 302 | |
| }, | |
| { | |
| "epoch": 0.17, | |
| "learning_rate": 0.000949625368060937, | |
| "loss": 2.6308, | |
| "step": 303 | |
| }, | |
| { | |
| "epoch": 0.17, | |
| "learning_rate": 0.0009492220766362336, | |
| "loss": 2.5783, | |
| "step": 304 | |
| }, | |
| { | |
| "epoch": 0.17, | |
| "learning_rate": 0.0009488172636354521, | |
| "loss": 2.6703, | |
| "step": 305 | |
| }, | |
| { | |
| "epoch": 0.17, | |
| "learning_rate": 0.0009484109304297492, | |
| "loss": 2.6737, | |
| "step": 306 | |
| }, | |
| { | |
| "epoch": 0.17, | |
| "learning_rate": 0.0009480030783954306, | |
| "loss": 2.5912, | |
| "step": 307 | |
| }, | |
| { | |
| "epoch": 0.17, | |
| "learning_rate": 0.0009475937089139463, | |
| "loss": 2.5694, | |
| "step": 308 | |
| }, | |
| { | |
| "epoch": 0.18, | |
| "learning_rate": 0.0009471828233718863, | |
| "loss": 2.6827, | |
| "step": 309 | |
| }, | |
| { | |
| "epoch": 0.18, | |
| "learning_rate": 0.0009467704231609755, | |
| "loss": 2.6466, | |
| "step": 310 | |
| }, | |
| { | |
| "epoch": 0.18, | |
| "learning_rate": 0.0009463565096780695, | |
| "loss": 2.6016, | |
| "step": 311 | |
| }, | |
| { | |
| "epoch": 0.18, | |
| "learning_rate": 0.0009459410843251495, | |
| "loss": 2.5613, | |
| "step": 312 | |
| }, | |
| { | |
| "epoch": 0.18, | |
| "learning_rate": 0.0009455241485093171, | |
| "loss": 2.5848, | |
| "step": 313 | |
| }, | |
| { | |
| "epoch": 0.18, | |
| "learning_rate": 0.0009451057036427909, | |
| "loss": 2.6192, | |
| "step": 314 | |
| }, | |
| { | |
| "epoch": 0.18, | |
| "learning_rate": 0.0009446857511428998, | |
| "loss": 2.6521, | |
| "step": 315 | |
| }, | |
| { | |
| "epoch": 0.18, | |
| "learning_rate": 0.0009442642924320806, | |
| "loss": 2.6224, | |
| "step": 316 | |
| }, | |
| { | |
| "epoch": 0.18, | |
| "learning_rate": 0.0009438413289378705, | |
| "loss": 2.5611, | |
| "step": 317 | |
| }, | |
| { | |
| "epoch": 0.18, | |
| "learning_rate": 0.0009434168620929044, | |
| "loss": 2.602, | |
| "step": 318 | |
| }, | |
| { | |
| "epoch": 0.18, | |
| "learning_rate": 0.0009429908933349091, | |
| "loss": 2.5021, | |
| "step": 319 | |
| }, | |
| { | |
| "epoch": 0.18, | |
| "learning_rate": 0.0009425634241066985, | |
| "loss": 2.5542, | |
| "step": 320 | |
| }, | |
| { | |
| "epoch": 0.18, | |
| "learning_rate": 0.0009421344558561687, | |
| "loss": 2.5592, | |
| "step": 321 | |
| }, | |
| { | |
| "epoch": 0.18, | |
| "learning_rate": 0.0009417039900362936, | |
| "loss": 2.6289, | |
| "step": 322 | |
| }, | |
| { | |
| "epoch": 0.18, | |
| "learning_rate": 0.000941272028105119, | |
| "loss": 2.5938, | |
| "step": 323 | |
| }, | |
| { | |
| "epoch": 0.18, | |
| "learning_rate": 0.0009408385715257588, | |
| "loss": 2.5195, | |
| "step": 324 | |
| }, | |
| { | |
| "epoch": 0.18, | |
| "learning_rate": 0.000940403621766389, | |
| "loss": 2.6641, | |
| "step": 325 | |
| }, | |
| { | |
| "epoch": 0.19, | |
| "learning_rate": 0.0009399671803002434, | |
| "loss": 2.5733, | |
| "step": 326 | |
| }, | |
| { | |
| "epoch": 0.19, | |
| "learning_rate": 0.0009395292486056086, | |
| "loss": 2.5534, | |
| "step": 327 | |
| }, | |
| { | |
| "epoch": 0.19, | |
| "learning_rate": 0.0009390898281658184, | |
| "loss": 2.6198, | |
| "step": 328 | |
| }, | |
| { | |
| "epoch": 0.19, | |
| "learning_rate": 0.0009386489204692497, | |
| "loss": 2.6807, | |
| "step": 329 | |
| }, | |
| { | |
| "epoch": 0.19, | |
| "learning_rate": 0.0009382065270093163, | |
| "loss": 2.5644, | |
| "step": 330 | |
| }, | |
| { | |
| "epoch": 0.19, | |
| "learning_rate": 0.0009377626492844649, | |
| "loss": 2.6815, | |
| "step": 331 | |
| }, | |
| { | |
| "epoch": 0.19, | |
| "learning_rate": 0.0009373172887981699, | |
| "loss": 2.5921, | |
| "step": 332 | |
| }, | |
| { | |
| "epoch": 0.19, | |
| "learning_rate": 0.0009368704470589271, | |
| "loss": 2.5429, | |
| "step": 333 | |
| }, | |
| { | |
| "epoch": 0.19, | |
| "learning_rate": 0.0009364221255802504, | |
| "loss": 2.5444, | |
| "step": 334 | |
| }, | |
| { | |
| "epoch": 0.19, | |
| "learning_rate": 0.0009359723258806654, | |
| "loss": 2.5487, | |
| "step": 335 | |
| }, | |
| { | |
| "epoch": 0.19, | |
| "learning_rate": 0.0009355210494837045, | |
| "loss": 2.5282, | |
| "step": 336 | |
| }, | |
| { | |
| "epoch": 0.19, | |
| "learning_rate": 0.0009350682979179024, | |
| "loss": 2.6438, | |
| "step": 337 | |
| }, | |
| { | |
| "epoch": 0.19, | |
| "learning_rate": 0.0009346140727167896, | |
| "loss": 2.5692, | |
| "step": 338 | |
| }, | |
| { | |
| "epoch": 0.19, | |
| "learning_rate": 0.0009341583754188887, | |
| "loss": 2.5261, | |
| "step": 339 | |
| }, | |
| { | |
| "epoch": 0.19, | |
| "learning_rate": 0.000933701207567708, | |
| "loss": 2.6031, | |
| "step": 340 | |
| }, | |
| { | |
| "epoch": 0.19, | |
| "learning_rate": 0.0009332425707117373, | |
| "loss": 2.5488, | |
| "step": 341 | |
| }, | |
| { | |
| "epoch": 0.19, | |
| "learning_rate": 0.0009327824664044417, | |
| "loss": 2.6051, | |
| "step": 342 | |
| }, | |
| { | |
| "epoch": 0.19, | |
| "learning_rate": 0.0009323208962042569, | |
| "loss": 2.6468, | |
| "step": 343 | |
| }, | |
| { | |
| "epoch": 0.2, | |
| "learning_rate": 0.0009318578616745834, | |
| "loss": 2.4675, | |
| "step": 344 | |
| }, | |
| { | |
| "epoch": 0.2, | |
| "learning_rate": 0.0009313933643837825, | |
| "loss": 2.6519, | |
| "step": 345 | |
| }, | |
| { | |
| "epoch": 0.2, | |
| "learning_rate": 0.0009309274059051691, | |
| "loss": 2.5038, | |
| "step": 346 | |
| }, | |
| { | |
| "epoch": 0.2, | |
| "learning_rate": 0.0009304599878170078, | |
| "loss": 2.6157, | |
| "step": 347 | |
| }, | |
| { | |
| "epoch": 0.2, | |
| "learning_rate": 0.0009299911117025071, | |
| "loss": 2.6254, | |
| "step": 348 | |
| }, | |
| { | |
| "epoch": 0.2, | |
| "learning_rate": 0.000929520779149814, | |
| "loss": 2.5462, | |
| "step": 349 | |
| }, | |
| { | |
| "epoch": 0.2, | |
| "learning_rate": 0.0009290489917520087, | |
| "loss": 2.5864, | |
| "step": 350 | |
| }, | |
| { | |
| "epoch": 0.2, | |
| "learning_rate": 0.0009285757511070987, | |
| "loss": 2.5915, | |
| "step": 351 | |
| }, | |
| { | |
| "epoch": 0.2, | |
| "learning_rate": 0.0009281010588180146, | |
| "loss": 2.5718, | |
| "step": 352 | |
| }, | |
| { | |
| "epoch": 0.2, | |
| "learning_rate": 0.0009276249164926034, | |
| "loss": 2.6059, | |
| "step": 353 | |
| }, | |
| { | |
| "epoch": 0.2, | |
| "learning_rate": 0.0009271473257436238, | |
| "loss": 2.5459, | |
| "step": 354 | |
| }, | |
| { | |
| "epoch": 0.2, | |
| "learning_rate": 0.0009266682881887403, | |
| "loss": 2.5749, | |
| "step": 355 | |
| }, | |
| { | |
| "epoch": 0.2, | |
| "learning_rate": 0.0009261878054505181, | |
| "loss": 2.5975, | |
| "step": 356 | |
| }, | |
| { | |
| "epoch": 0.2, | |
| "learning_rate": 0.0009257058791564173, | |
| "loss": 2.5799, | |
| "step": 357 | |
| }, | |
| { | |
| "epoch": 0.2, | |
| "learning_rate": 0.0009252225109387879, | |
| "loss": 2.5351, | |
| "step": 358 | |
| }, | |
| { | |
| "epoch": 0.2, | |
| "learning_rate": 0.0009247377024348631, | |
| "loss": 2.6105, | |
| "step": 359 | |
| }, | |
| { | |
| "epoch": 0.2, | |
| "learning_rate": 0.0009242514552867555, | |
| "loss": 2.6401, | |
| "step": 360 | |
| }, | |
| { | |
| "epoch": 0.21, | |
| "learning_rate": 0.0009237637711414497, | |
| "loss": 2.5392, | |
| "step": 361 | |
| }, | |
| { | |
| "epoch": 0.21, | |
| "learning_rate": 0.0009232746516507984, | |
| "loss": 2.589, | |
| "step": 362 | |
| }, | |
| { | |
| "epoch": 0.21, | |
| "learning_rate": 0.0009227840984715153, | |
| "loss": 2.5614, | |
| "step": 363 | |
| }, | |
| { | |
| "epoch": 0.21, | |
| "learning_rate": 0.0009222921132651708, | |
| "loss": 2.5614, | |
| "step": 364 | |
| }, | |
| { | |
| "epoch": 0.21, | |
| "learning_rate": 0.0009217986976981854, | |
| "loss": 2.5427, | |
| "step": 365 | |
| }, | |
| { | |
| "epoch": 0.21, | |
| "learning_rate": 0.0009213038534418243, | |
| "loss": 2.5695, | |
| "step": 366 | |
| }, | |
| { | |
| "epoch": 0.21, | |
| "learning_rate": 0.0009208075821721926, | |
| "loss": 2.5442, | |
| "step": 367 | |
| }, | |
| { | |
| "epoch": 0.21, | |
| "learning_rate": 0.0009203098855702277, | |
| "loss": 2.6164, | |
| "step": 368 | |
| }, | |
| { | |
| "epoch": 0.21, | |
| "learning_rate": 0.0009198107653216961, | |
| "loss": 2.5373, | |
| "step": 369 | |
| }, | |
| { | |
| "epoch": 0.21, | |
| "learning_rate": 0.0009193102231171852, | |
| "loss": 2.5858, | |
| "step": 370 | |
| }, | |
| { | |
| "epoch": 0.21, | |
| "learning_rate": 0.0009188082606520997, | |
| "loss": 2.5471, | |
| "step": 371 | |
| }, | |
| { | |
| "epoch": 0.21, | |
| "learning_rate": 0.0009183048796266546, | |
| "loss": 2.5069, | |
| "step": 372 | |
| }, | |
| { | |
| "epoch": 0.21, | |
| "learning_rate": 0.0009178000817458694, | |
| "loss": 2.5816, | |
| "step": 373 | |
| }, | |
| { | |
| "epoch": 0.21, | |
| "learning_rate": 0.0009172938687195629, | |
| "loss": 2.4701, | |
| "step": 374 | |
| }, | |
| { | |
| "epoch": 0.21, | |
| "learning_rate": 0.0009167862422623474, | |
| "loss": 2.6111, | |
| "step": 375 | |
| }, | |
| { | |
| "epoch": 0.21, | |
| "learning_rate": 0.0009162772040936227, | |
| "loss": 2.6355, | |
| "step": 376 | |
| }, | |
| { | |
| "epoch": 0.21, | |
| "learning_rate": 0.0009157667559375697, | |
| "loss": 2.5845, | |
| "step": 377 | |
| }, | |
| { | |
| "epoch": 0.21, | |
| "learning_rate": 0.000915254899523146, | |
| "loss": 2.6657, | |
| "step": 378 | |
| }, | |
| { | |
| "epoch": 0.22, | |
| "learning_rate": 0.0009147416365840783, | |
| "loss": 2.5875, | |
| "step": 379 | |
| }, | |
| { | |
| "epoch": 0.22, | |
| "learning_rate": 0.0009142269688588578, | |
| "loss": 2.5423, | |
| "step": 380 | |
| }, | |
| { | |
| "epoch": 0.22, | |
| "learning_rate": 0.000913710898090734, | |
| "loss": 2.5814, | |
| "step": 381 | |
| }, | |
| { | |
| "epoch": 0.22, | |
| "learning_rate": 0.0009131934260277084, | |
| "loss": 2.5892, | |
| "step": 382 | |
| }, | |
| { | |
| "epoch": 0.22, | |
| "learning_rate": 0.0009126745544225292, | |
| "loss": 2.6292, | |
| "step": 383 | |
| }, | |
| { | |
| "epoch": 0.22, | |
| "learning_rate": 0.0009121542850326848, | |
| "loss": 2.5795, | |
| "step": 384 | |
| }, | |
| { | |
| "epoch": 0.22, | |
| "learning_rate": 0.0009116326196203981, | |
| "loss": 2.5144, | |
| "step": 385 | |
| }, | |
| { | |
| "epoch": 0.22, | |
| "learning_rate": 0.0009111095599526206, | |
| "loss": 2.5801, | |
| "step": 386 | |
| }, | |
| { | |
| "epoch": 0.22, | |
| "learning_rate": 0.0009105851078010266, | |
| "loss": 2.5736, | |
| "step": 387 | |
| }, | |
| { | |
| "epoch": 0.22, | |
| "learning_rate": 0.0009100592649420061, | |
| "loss": 2.56, | |
| "step": 388 | |
| }, | |
| { | |
| "epoch": 0.22, | |
| "learning_rate": 0.0009095320331566606, | |
| "loss": 2.5927, | |
| "step": 389 | |
| }, | |
| { | |
| "epoch": 0.22, | |
| "learning_rate": 0.0009090034142307954, | |
| "loss": 2.5959, | |
| "step": 390 | |
| }, | |
| { | |
| "epoch": 0.22, | |
| "learning_rate": 0.0009084734099549146, | |
| "loss": 2.5392, | |
| "step": 391 | |
| }, | |
| { | |
| "epoch": 0.22, | |
| "learning_rate": 0.0009079420221242145, | |
| "loss": 2.6631, | |
| "step": 392 | |
| }, | |
| { | |
| "epoch": 0.22, | |
| "learning_rate": 0.0009074092525385777, | |
| "loss": 2.5413, | |
| "step": 393 | |
| }, | |
| { | |
| "epoch": 0.22, | |
| "learning_rate": 0.000906875103002567, | |
| "loss": 2.5866, | |
| "step": 394 | |
| }, | |
| { | |
| "epoch": 0.22, | |
| "learning_rate": 0.0009063395753254193, | |
| "loss": 2.5817, | |
| "step": 395 | |
| }, | |
| { | |
| "epoch": 0.22, | |
| "learning_rate": 0.0009058026713210397, | |
| "loss": 2.6183, | |
| "step": 396 | |
| }, | |
| { | |
| "epoch": 0.23, | |
| "learning_rate": 0.0009052643928079945, | |
| "loss": 2.5762, | |
| "step": 397 | |
| }, | |
| { | |
| "epoch": 0.23, | |
| "learning_rate": 0.0009047247416095059, | |
| "loss": 2.5657, | |
| "step": 398 | |
| }, | |
| { | |
| "epoch": 0.23, | |
| "learning_rate": 0.0009041837195534462, | |
| "loss": 2.4247, | |
| "step": 399 | |
| }, | |
| { | |
| "epoch": 0.23, | |
| "learning_rate": 0.0009036413284723301, | |
| "loss": 2.6515, | |
| "step": 400 | |
| }, | |
| { | |
| "epoch": 0.23, | |
| "learning_rate": 0.0009030975702033098, | |
| "loss": 2.5671, | |
| "step": 401 | |
| }, | |
| { | |
| "epoch": 0.23, | |
| "learning_rate": 0.0009025524465881683, | |
| "loss": 2.5696, | |
| "step": 402 | |
| }, | |
| { | |
| "epoch": 0.23, | |
| "learning_rate": 0.000902005959473313, | |
| "loss": 2.5915, | |
| "step": 403 | |
| }, | |
| { | |
| "epoch": 0.23, | |
| "learning_rate": 0.0009014581107097702, | |
| "loss": 2.5173, | |
| "step": 404 | |
| }, | |
| { | |
| "epoch": 0.23, | |
| "learning_rate": 0.0009009089021531776, | |
| "loss": 2.5242, | |
| "step": 405 | |
| }, | |
| { | |
| "epoch": 0.23, | |
| "learning_rate": 0.0009003583356637793, | |
| "loss": 2.5752, | |
| "step": 406 | |
| }, | |
| { | |
| "epoch": 0.23, | |
| "learning_rate": 0.0008998064131064185, | |
| "loss": 2.4775, | |
| "step": 407 | |
| }, | |
| { | |
| "epoch": 0.23, | |
| "learning_rate": 0.0008992531363505318, | |
| "loss": 2.5739, | |
| "step": 408 | |
| }, | |
| { | |
| "epoch": 0.23, | |
| "learning_rate": 0.0008986985072701425, | |
| "loss": 2.4938, | |
| "step": 409 | |
| }, | |
| { | |
| "epoch": 0.23, | |
| "learning_rate": 0.0008981425277438546, | |
| "loss": 2.4801, | |
| "step": 410 | |
| }, | |
| { | |
| "epoch": 0.23, | |
| "learning_rate": 0.0008975851996548459, | |
| "loss": 2.4974, | |
| "step": 411 | |
| }, | |
| { | |
| "epoch": 0.23, | |
| "learning_rate": 0.0008970265248908626, | |
| "loss": 2.5897, | |
| "step": 412 | |
| }, | |
| { | |
| "epoch": 0.23, | |
| "learning_rate": 0.0008964665053442116, | |
| "loss": 2.5122, | |
| "step": 413 | |
| }, | |
| { | |
| "epoch": 0.24, | |
| "learning_rate": 0.000895905142911755, | |
| "loss": 2.5803, | |
| "step": 414 | |
| }, | |
| { | |
| "epoch": 0.24, | |
| "learning_rate": 0.0008953424394949035, | |
| "loss": 2.6401, | |
| "step": 415 | |
| }, | |
| { | |
| "epoch": 0.24, | |
| "learning_rate": 0.00089477839699961, | |
| "loss": 2.561, | |
| "step": 416 | |
| }, | |
| { | |
| "epoch": 0.24, | |
| "learning_rate": 0.0008942130173363627, | |
| "loss": 2.6039, | |
| "step": 417 | |
| }, | |
| { | |
| "epoch": 0.24, | |
| "learning_rate": 0.0008936463024201794, | |
| "loss": 2.5148, | |
| "step": 418 | |
| }, | |
| { | |
| "epoch": 0.24, | |
| "learning_rate": 0.0008930782541706002, | |
| "loss": 2.5421, | |
| "step": 419 | |
| }, | |
| { | |
| "epoch": 0.24, | |
| "learning_rate": 0.0008925088745116816, | |
| "loss": 2.4854, | |
| "step": 420 | |
| }, | |
| { | |
| "epoch": 0.24, | |
| "learning_rate": 0.0008919381653719895, | |
| "loss": 2.5535, | |
| "step": 421 | |
| }, | |
| { | |
| "epoch": 0.24, | |
| "learning_rate": 0.0008913661286845936, | |
| "loss": 2.576, | |
| "step": 422 | |
| }, | |
| { | |
| "epoch": 0.24, | |
| "learning_rate": 0.0008907927663870592, | |
| "loss": 2.6122, | |
| "step": 423 | |
| }, | |
| { | |
| "epoch": 0.24, | |
| "learning_rate": 0.0008902180804214423, | |
| "loss": 2.6166, | |
| "step": 424 | |
| }, | |
| { | |
| "epoch": 0.24, | |
| "learning_rate": 0.0008896420727342823, | |
| "loss": 2.5119, | |
| "step": 425 | |
| }, | |
| { | |
| "epoch": 0.24, | |
| "learning_rate": 0.0008890647452765953, | |
| "loss": 2.5047, | |
| "step": 426 | |
| }, | |
| { | |
| "epoch": 0.24, | |
| "learning_rate": 0.0008884861000038676, | |
| "loss": 2.5731, | |
| "step": 427 | |
| }, | |
| { | |
| "epoch": 0.24, | |
| "learning_rate": 0.0008879061388760492, | |
| "loss": 2.576, | |
| "step": 428 | |
| }, | |
| { | |
| "epoch": 0.24, | |
| "learning_rate": 0.0008873248638575471, | |
| "loss": 2.5827, | |
| "step": 429 | |
| }, | |
| { | |
| "epoch": 0.24, | |
| "learning_rate": 0.0008867422769172187, | |
| "loss": 2.5291, | |
| "step": 430 | |
| }, | |
| { | |
| "epoch": 0.24, | |
| "learning_rate": 0.0008861583800283646, | |
| "loss": 2.4942, | |
| "step": 431 | |
| }, | |
| { | |
| "epoch": 0.25, | |
| "learning_rate": 0.0008855731751687232, | |
| "loss": 2.5566, | |
| "step": 432 | |
| }, | |
| { | |
| "epoch": 0.25, | |
| "learning_rate": 0.0008849866643204625, | |
| "loss": 2.5337, | |
| "step": 433 | |
| }, | |
| { | |
| "epoch": 0.25, | |
| "learning_rate": 0.000884398849470174, | |
| "loss": 2.5165, | |
| "step": 434 | |
| }, | |
| { | |
| "epoch": 0.25, | |
| "learning_rate": 0.0008838097326088666, | |
| "loss": 2.6361, | |
| "step": 435 | |
| }, | |
| { | |
| "epoch": 0.25, | |
| "learning_rate": 0.000883219315731959, | |
| "loss": 2.5504, | |
| "step": 436 | |
| }, | |
| { | |
| "epoch": 0.25, | |
| "learning_rate": 0.0008826276008392731, | |
| "loss": 2.5573, | |
| "step": 437 | |
| }, | |
| { | |
| "epoch": 0.25, | |
| "learning_rate": 0.0008820345899350273, | |
| "loss": 2.562, | |
| "step": 438 | |
| }, | |
| { | |
| "epoch": 0.25, | |
| "learning_rate": 0.0008814402850278303, | |
| "loss": 2.4775, | |
| "step": 439 | |
| }, | |
| { | |
| "epoch": 0.25, | |
| "learning_rate": 0.0008808446881306731, | |
| "loss": 2.5607, | |
| "step": 440 | |
| }, | |
| { | |
| "epoch": 0.25, | |
| "learning_rate": 0.0008802478012609234, | |
| "loss": 2.5578, | |
| "step": 441 | |
| }, | |
| { | |
| "epoch": 0.25, | |
| "learning_rate": 0.0008796496264403175, | |
| "loss": 2.5214, | |
| "step": 442 | |
| }, | |
| { | |
| "epoch": 0.25, | |
| "learning_rate": 0.0008790501656949552, | |
| "loss": 2.519, | |
| "step": 443 | |
| }, | |
| { | |
| "epoch": 0.25, | |
| "learning_rate": 0.0008784494210552909, | |
| "loss": 2.5888, | |
| "step": 444 | |
| }, | |
| { | |
| "epoch": 0.25, | |
| "learning_rate": 0.0008778473945561283, | |
| "loss": 2.487, | |
| "step": 445 | |
| }, | |
| { | |
| "epoch": 0.25, | |
| "learning_rate": 0.0008772440882366126, | |
| "loss": 2.5737, | |
| "step": 446 | |
| }, | |
| { | |
| "epoch": 0.25, | |
| "learning_rate": 0.0008766395041402244, | |
| "loss": 2.5556, | |
| "step": 447 | |
| }, | |
| { | |
| "epoch": 0.25, | |
| "learning_rate": 0.0008760336443147718, | |
| "loss": 2.5803, | |
| "step": 448 | |
| }, | |
| { | |
| "epoch": 0.26, | |
| "learning_rate": 0.0008754265108123843, | |
| "loss": 2.5664, | |
| "step": 449 | |
| }, | |
| { | |
| "epoch": 0.26, | |
| "learning_rate": 0.0008748181056895051, | |
| "loss": 2.5422, | |
| "step": 450 | |
| }, | |
| { | |
| "epoch": 0.26, | |
| "learning_rate": 0.0008742084310068851, | |
| "loss": 2.5137, | |
| "step": 451 | |
| }, | |
| { | |
| "epoch": 0.26, | |
| "learning_rate": 0.0008735974888295753, | |
| "loss": 2.4922, | |
| "step": 452 | |
| }, | |
| { | |
| "epoch": 0.26, | |
| "learning_rate": 0.0008729852812269191, | |
| "loss": 2.5818, | |
| "step": 453 | |
| }, | |
| { | |
| "epoch": 0.26, | |
| "learning_rate": 0.0008723718102725471, | |
| "loss": 2.5574, | |
| "step": 454 | |
| }, | |
| { | |
| "epoch": 0.26, | |
| "learning_rate": 0.0008717570780443685, | |
| "loss": 2.582, | |
| "step": 455 | |
| }, | |
| { | |
| "epoch": 0.26, | |
| "learning_rate": 0.0008711410866245647, | |
| "loss": 2.4976, | |
| "step": 456 | |
| }, | |
| { | |
| "epoch": 0.26, | |
| "learning_rate": 0.0008705238380995821, | |
| "loss": 2.5415, | |
| "step": 457 | |
| }, | |
| { | |
| "epoch": 0.26, | |
| "learning_rate": 0.0008699053345601252, | |
| "loss": 2.548, | |
| "step": 458 | |
| }, | |
| { | |
| "epoch": 0.26, | |
| "learning_rate": 0.0008692855781011493, | |
| "loss": 2.4905, | |
| "step": 459 | |
| }, | |
| { | |
| "epoch": 0.26, | |
| "learning_rate": 0.0008686645708218535, | |
| "loss": 2.5447, | |
| "step": 460 | |
| }, | |
| { | |
| "epoch": 0.26, | |
| "learning_rate": 0.0008680423148256737, | |
| "loss": 2.5469, | |
| "step": 461 | |
| }, | |
| { | |
| "epoch": 0.26, | |
| "learning_rate": 0.0008674188122202755, | |
| "loss": 2.5667, | |
| "step": 462 | |
| }, | |
| { | |
| "epoch": 0.26, | |
| "learning_rate": 0.0008667940651175465, | |
| "loss": 2.5645, | |
| "step": 463 | |
| }, | |
| { | |
| "epoch": 0.26, | |
| "learning_rate": 0.00086616807563359, | |
| "loss": 2.4549, | |
| "step": 464 | |
| }, | |
| { | |
| "epoch": 0.26, | |
| "learning_rate": 0.0008655408458887171, | |
| "loss": 2.4675, | |
| "step": 465 | |
| }, | |
| { | |
| "epoch": 0.26, | |
| "learning_rate": 0.0008649123780074401, | |
| "loss": 2.4899, | |
| "step": 466 | |
| }, | |
| { | |
| "epoch": 0.27, | |
| "learning_rate": 0.000864282674118465, | |
| "loss": 2.4976, | |
| "step": 467 | |
| }, | |
| { | |
| "epoch": 0.27, | |
| "learning_rate": 0.0008636517363546838, | |
| "loss": 2.5422, | |
| "step": 468 | |
| }, | |
| { | |
| "epoch": 0.27, | |
| "learning_rate": 0.0008630195668531689, | |
| "loss": 2.5581, | |
| "step": 469 | |
| }, | |
| { | |
| "epoch": 0.27, | |
| "learning_rate": 0.0008623861677551637, | |
| "loss": 2.5215, | |
| "step": 470 | |
| }, | |
| { | |
| "epoch": 0.27, | |
| "learning_rate": 0.000861751541206077, | |
| "loss": 2.583, | |
| "step": 471 | |
| }, | |
| { | |
| "epoch": 0.27, | |
| "learning_rate": 0.0008611156893554748, | |
| "loss": 2.5575, | |
| "step": 472 | |
| }, | |
| { | |
| "epoch": 0.27, | |
| "learning_rate": 0.0008604786143570735, | |
| "loss": 2.4754, | |
| "step": 473 | |
| }, | |
| { | |
| "epoch": 0.27, | |
| "learning_rate": 0.0008598403183687328, | |
| "loss": 2.5841, | |
| "step": 474 | |
| }, | |
| { | |
| "epoch": 0.27, | |
| "learning_rate": 0.0008592008035524474, | |
| "loss": 2.5482, | |
| "step": 475 | |
| }, | |
| { | |
| "epoch": 0.27, | |
| "learning_rate": 0.0008585600720743409, | |
| "loss": 2.4925, | |
| "step": 476 | |
| }, | |
| { | |
| "epoch": 0.27, | |
| "learning_rate": 0.0008579181261046576, | |
| "loss": 2.5339, | |
| "step": 477 | |
| }, | |
| { | |
| "epoch": 0.27, | |
| "learning_rate": 0.0008572749678177555, | |
| "loss": 2.5182, | |
| "step": 478 | |
| }, | |
| { | |
| "epoch": 0.27, | |
| "learning_rate": 0.000856630599392099, | |
| "loss": 2.4739, | |
| "step": 479 | |
| }, | |
| { | |
| "epoch": 0.27, | |
| "learning_rate": 0.0008559850230102512, | |
| "loss": 2.5473, | |
| "step": 480 | |
| }, | |
| { | |
| "epoch": 0.27, | |
| "learning_rate": 0.000855338240858867, | |
| "loss": 2.5086, | |
| "step": 481 | |
| }, | |
| { | |
| "epoch": 0.27, | |
| "learning_rate": 0.0008546902551286853, | |
| "loss": 2.5512, | |
| "step": 482 | |
| }, | |
| { | |
| "epoch": 0.27, | |
| "learning_rate": 0.0008540410680145211, | |
| "loss": 2.4758, | |
| "step": 483 | |
| }, | |
| { | |
| "epoch": 0.27, | |
| "learning_rate": 0.0008533906817152599, | |
| "loss": 2.5077, | |
| "step": 484 | |
| }, | |
| { | |
| "epoch": 0.28, | |
| "learning_rate": 0.0008527390984338477, | |
| "loss": 2.5646, | |
| "step": 485 | |
| }, | |
| { | |
| "epoch": 0.28, | |
| "learning_rate": 0.0008520863203772857, | |
| "loss": 2.555, | |
| "step": 486 | |
| }, | |
| { | |
| "epoch": 0.28, | |
| "learning_rate": 0.0008514323497566216, | |
| "loss": 2.4723, | |
| "step": 487 | |
| }, | |
| { | |
| "epoch": 0.28, | |
| "learning_rate": 0.0008507771887869426, | |
| "loss": 2.5635, | |
| "step": 488 | |
| }, | |
| { | |
| "epoch": 0.28, | |
| "learning_rate": 0.0008501208396873676, | |
| "loss": 2.503, | |
| "step": 489 | |
| }, | |
| { | |
| "epoch": 0.28, | |
| "learning_rate": 0.0008494633046810402, | |
| "loss": 2.4948, | |
| "step": 490 | |
| }, | |
| { | |
| "epoch": 0.28, | |
| "learning_rate": 0.0008488045859951206, | |
| "loss": 2.5435, | |
| "step": 491 | |
| }, | |
| { | |
| "epoch": 0.28, | |
| "learning_rate": 0.000848144685860778, | |
| "loss": 2.5013, | |
| "step": 492 | |
| }, | |
| { | |
| "epoch": 0.28, | |
| "learning_rate": 0.0008474836065131838, | |
| "loss": 2.5691, | |
| "step": 493 | |
| }, | |
| { | |
| "epoch": 0.28, | |
| "learning_rate": 0.0008468213501915034, | |
| "loss": 2.5194, | |
| "step": 494 | |
| }, | |
| { | |
| "epoch": 0.28, | |
| "learning_rate": 0.0008461579191388889, | |
| "loss": 2.5856, | |
| "step": 495 | |
| }, | |
| { | |
| "epoch": 0.28, | |
| "learning_rate": 0.0008454933156024709, | |
| "loss": 2.6139, | |
| "step": 496 | |
| }, | |
| { | |
| "epoch": 0.28, | |
| "learning_rate": 0.0008448275418333518, | |
| "loss": 2.4525, | |
| "step": 497 | |
| }, | |
| { | |
| "epoch": 0.28, | |
| "learning_rate": 0.0008441606000865978, | |
| "loss": 2.5658, | |
| "step": 498 | |
| }, | |
| { | |
| "epoch": 0.28, | |
| "learning_rate": 0.0008434924926212307, | |
| "loss": 2.4961, | |
| "step": 499 | |
| }, | |
| { | |
| "epoch": 0.28, | |
| "learning_rate": 0.0008428232217002214, | |
| "loss": 2.546, | |
| "step": 500 | |
| }, | |
| { | |
| "epoch": 0.28, | |
| "learning_rate": 0.0008421527895904811, | |
| "loss": 2.496, | |
| "step": 501 | |
| }, | |
| { | |
| "epoch": 0.29, | |
| "learning_rate": 0.0008414811985628544, | |
| "loss": 2.4998, | |
| "step": 502 | |
| }, | |
| { | |
| "epoch": 0.29, | |
| "learning_rate": 0.0008408084508921106, | |
| "loss": 2.4938, | |
| "step": 503 | |
| }, | |
| { | |
| "epoch": 0.29, | |
| "learning_rate": 0.000840134548856938, | |
| "loss": 2.5078, | |
| "step": 504 | |
| }, | |
| { | |
| "epoch": 0.29, | |
| "learning_rate": 0.0008394594947399337, | |
| "loss": 2.5869, | |
| "step": 505 | |
| }, | |
| { | |
| "epoch": 0.29, | |
| "learning_rate": 0.0008387832908275977, | |
| "loss": 2.5255, | |
| "step": 506 | |
| }, | |
| { | |
| "epoch": 0.29, | |
| "learning_rate": 0.0008381059394103243, | |
| "loss": 2.5458, | |
| "step": 507 | |
| }, | |
| { | |
| "epoch": 0.29, | |
| "learning_rate": 0.0008374274427823946, | |
| "loss": 2.5222, | |
| "step": 508 | |
| }, | |
| { | |
| "epoch": 0.29, | |
| "learning_rate": 0.0008367478032419686, | |
| "loss": 2.5068, | |
| "step": 509 | |
| }, | |
| { | |
| "epoch": 0.29, | |
| "learning_rate": 0.0008360670230910777, | |
| "loss": 2.511, | |
| "step": 510 | |
| }, | |
| { | |
| "epoch": 0.29, | |
| "learning_rate": 0.0008353851046356163, | |
| "loss": 2.4518, | |
| "step": 511 | |
| }, | |
| { | |
| "epoch": 0.29, | |
| "learning_rate": 0.0008347020501853348, | |
| "loss": 2.404, | |
| "step": 512 | |
| }, | |
| { | |
| "epoch": 0.29, | |
| "learning_rate": 0.0008340178620538315, | |
| "loss": 2.5601, | |
| "step": 513 | |
| }, | |
| { | |
| "epoch": 0.29, | |
| "learning_rate": 0.0008333325425585439, | |
| "loss": 2.4297, | |
| "step": 514 | |
| }, | |
| { | |
| "epoch": 0.29, | |
| "learning_rate": 0.0008326460940207423, | |
| "loss": 2.5419, | |
| "step": 515 | |
| }, | |
| { | |
| "epoch": 0.29, | |
| "learning_rate": 0.000831958518765521, | |
| "loss": 2.5175, | |
| "step": 516 | |
| }, | |
| { | |
| "epoch": 0.29, | |
| "learning_rate": 0.0008312698191217906, | |
| "loss": 2.5704, | |
| "step": 517 | |
| }, | |
| { | |
| "epoch": 0.29, | |
| "learning_rate": 0.0008305799974222701, | |
| "loss": 2.5524, | |
| "step": 518 | |
| }, | |
| { | |
| "epoch": 0.29, | |
| "learning_rate": 0.0008298890560034791, | |
| "loss": 2.5907, | |
| "step": 519 | |
| }, | |
| { | |
| "epoch": 0.3, | |
| "learning_rate": 0.0008291969972057298, | |
| "loss": 2.5519, | |
| "step": 520 | |
| }, | |
| { | |
| "epoch": 0.3, | |
| "learning_rate": 0.0008285038233731193, | |
| "loss": 2.3889, | |
| "step": 521 | |
| }, | |
| { | |
| "epoch": 0.3, | |
| "learning_rate": 0.0008278095368535215, | |
| "loss": 2.4414, | |
| "step": 522 | |
| }, | |
| { | |
| "epoch": 0.3, | |
| "learning_rate": 0.0008271141399985787, | |
| "loss": 2.4874, | |
| "step": 523 | |
| }, | |
| { | |
| "epoch": 0.3, | |
| "learning_rate": 0.0008264176351636943, | |
| "loss": 2.4667, | |
| "step": 524 | |
| }, | |
| { | |
| "epoch": 0.3, | |
| "learning_rate": 0.0008257200247080247, | |
| "loss": 2.4855, | |
| "step": 525 | |
| }, | |
| { | |
| "epoch": 0.3, | |
| "learning_rate": 0.0008250213109944712, | |
| "loss": 2.4921, | |
| "step": 526 | |
| }, | |
| { | |
| "epoch": 0.3, | |
| "learning_rate": 0.0008243214963896718, | |
| "loss": 2.4261, | |
| "step": 527 | |
| }, | |
| { | |
| "epoch": 0.3, | |
| "learning_rate": 0.0008236205832639935, | |
| "loss": 2.4202, | |
| "step": 528 | |
| }, | |
| { | |
| "epoch": 0.3, | |
| "learning_rate": 0.000822918573991524, | |
| "loss": 2.5184, | |
| "step": 529 | |
| }, | |
| { | |
| "epoch": 0.3, | |
| "learning_rate": 0.0008222154709500637, | |
| "loss": 2.5094, | |
| "step": 530 | |
| }, | |
| { | |
| "epoch": 0.3, | |
| "learning_rate": 0.0008215112765211184, | |
| "loss": 2.4613, | |
| "step": 531 | |
| }, | |
| { | |
| "epoch": 0.3, | |
| "learning_rate": 0.0008208059930898898, | |
| "loss": 2.4444, | |
| "step": 532 | |
| }, | |
| { | |
| "epoch": 0.3, | |
| "learning_rate": 0.0008200996230452684, | |
| "loss": 2.5802, | |
| "step": 533 | |
| }, | |
| { | |
| "epoch": 0.3, | |
| "learning_rate": 0.0008193921687798256, | |
| "loss": 2.5125, | |
| "step": 534 | |
| }, | |
| { | |
| "epoch": 0.3, | |
| "learning_rate": 0.0008186836326898047, | |
| "loss": 2.4846, | |
| "step": 535 | |
| }, | |
| { | |
| "epoch": 0.3, | |
| "learning_rate": 0.0008179740171751135, | |
| "loss": 2.4259, | |
| "step": 536 | |
| }, | |
| { | |
| "epoch": 0.31, | |
| "learning_rate": 0.000817263324639316, | |
| "loss": 2.4611, | |
| "step": 537 | |
| }, | |
| { | |
| "epoch": 0.31, | |
| "learning_rate": 0.0008165515574896243, | |
| "loss": 2.5396, | |
| "step": 538 | |
| }, | |
| { | |
| "epoch": 0.31, | |
| "learning_rate": 0.0008158387181368901, | |
| "loss": 2.5483, | |
| "step": 539 | |
| }, | |
| { | |
| "epoch": 0.31, | |
| "learning_rate": 0.0008151248089955969, | |
| "loss": 2.4982, | |
| "step": 540 | |
| }, | |
| { | |
| "epoch": 0.31, | |
| "learning_rate": 0.0008144098324838518, | |
| "loss": 2.4895, | |
| "step": 541 | |
| }, | |
| { | |
| "epoch": 0.31, | |
| "learning_rate": 0.0008136937910233772, | |
| "loss": 2.4413, | |
| "step": 542 | |
| }, | |
| { | |
| "epoch": 0.31, | |
| "learning_rate": 0.0008129766870395026, | |
| "loss": 2.5752, | |
| "step": 543 | |
| }, | |
| { | |
| "epoch": 0.31, | |
| "learning_rate": 0.0008122585229611565, | |
| "loss": 2.4546, | |
| "step": 544 | |
| }, | |
| { | |
| "epoch": 0.31, | |
| "learning_rate": 0.0008115393012208578, | |
| "loss": 2.475, | |
| "step": 545 | |
| }, | |
| { | |
| "epoch": 0.31, | |
| "learning_rate": 0.0008108190242547082, | |
| "loss": 2.4423, | |
| "step": 546 | |
| }, | |
| { | |
| "epoch": 0.31, | |
| "learning_rate": 0.0008100976945023834, | |
| "loss": 2.5039, | |
| "step": 547 | |
| }, | |
| { | |
| "epoch": 0.31, | |
| "learning_rate": 0.000809375314407125, | |
| "loss": 2.4312, | |
| "step": 548 | |
| }, | |
| { | |
| "epoch": 0.31, | |
| "learning_rate": 0.0008086518864157324, | |
| "loss": 2.4996, | |
| "step": 549 | |
| }, | |
| { | |
| "epoch": 0.31, | |
| "learning_rate": 0.0008079274129785543, | |
| "loss": 2.5001, | |
| "step": 550 | |
| }, | |
| { | |
| "epoch": 0.31, | |
| "learning_rate": 0.0008072018965494804, | |
| "loss": 2.5265, | |
| "step": 551 | |
| }, | |
| { | |
| "epoch": 0.31, | |
| "learning_rate": 0.0008064753395859332, | |
| "loss": 2.4214, | |
| "step": 552 | |
| }, | |
| { | |
| "epoch": 0.31, | |
| "learning_rate": 0.0008057477445488597, | |
| "loss": 2.4731, | |
| "step": 553 | |
| }, | |
| { | |
| "epoch": 0.31, | |
| "learning_rate": 0.000805019113902723, | |
| "loss": 2.4975, | |
| "step": 554 | |
| }, | |
| { | |
| "epoch": 0.32, | |
| "learning_rate": 0.0008042894501154937, | |
| "loss": 2.4966, | |
| "step": 555 | |
| }, | |
| { | |
| "epoch": 0.32, | |
| "learning_rate": 0.0008035587556586421, | |
| "loss": 2.4733, | |
| "step": 556 | |
| }, | |
| { | |
| "epoch": 0.32, | |
| "learning_rate": 0.0008028270330071292, | |
| "loss": 2.5611, | |
| "step": 557 | |
| }, | |
| { | |
| "epoch": 0.32, | |
| "learning_rate": 0.0008020942846393992, | |
| "loss": 2.4482, | |
| "step": 558 | |
| }, | |
| { | |
| "epoch": 0.32, | |
| "learning_rate": 0.0008013605130373697, | |
| "loss": 2.4835, | |
| "step": 559 | |
| }, | |
| { | |
| "epoch": 0.32, | |
| "learning_rate": 0.0008006257206864252, | |
| "loss": 2.5224, | |
| "step": 560 | |
| }, | |
| { | |
| "epoch": 0.32, | |
| "learning_rate": 0.0007998899100754064, | |
| "loss": 2.5073, | |
| "step": 561 | |
| }, | |
| { | |
| "epoch": 0.32, | |
| "learning_rate": 0.0007991530836966039, | |
| "loss": 2.3737, | |
| "step": 562 | |
| }, | |
| { | |
| "epoch": 0.32, | |
| "learning_rate": 0.0007984152440457485, | |
| "loss": 2.5418, | |
| "step": 563 | |
| }, | |
| { | |
| "epoch": 0.32, | |
| "learning_rate": 0.000797676393622003, | |
| "loss": 2.4813, | |
| "step": 564 | |
| }, | |
| { | |
| "epoch": 0.32, | |
| "learning_rate": 0.0007969365349279544, | |
| "loss": 2.515, | |
| "step": 565 | |
| }, | |
| { | |
| "epoch": 0.32, | |
| "learning_rate": 0.0007961956704696039, | |
| "loss": 2.5198, | |
| "step": 566 | |
| }, | |
| { | |
| "epoch": 0.32, | |
| "learning_rate": 0.0007954538027563601, | |
| "loss": 2.5001, | |
| "step": 567 | |
| }, | |
| { | |
| "epoch": 0.32, | |
| "learning_rate": 0.0007947109343010295, | |
| "loss": 2.4378, | |
| "step": 568 | |
| }, | |
| { | |
| "epoch": 0.32, | |
| "learning_rate": 0.0007939670676198083, | |
| "loss": 2.4959, | |
| "step": 569 | |
| }, | |
| { | |
| "epoch": 0.32, | |
| "learning_rate": 0.0007932222052322737, | |
| "loss": 2.493, | |
| "step": 570 | |
| }, | |
| { | |
| "epoch": 0.32, | |
| "learning_rate": 0.0007924763496613756, | |
| "loss": 2.4409, | |
| "step": 571 | |
| }, | |
| { | |
| "epoch": 0.32, | |
| "learning_rate": 0.000791729503433428, | |
| "loss": 2.5195, | |
| "step": 572 | |
| }, | |
| { | |
| "epoch": 0.33, | |
| "learning_rate": 0.0007909816690781004, | |
| "loss": 2.5452, | |
| "step": 573 | |
| }, | |
| { | |
| "epoch": 0.33, | |
| "learning_rate": 0.000790232849128409, | |
| "loss": 2.49, | |
| "step": 574 | |
| }, | |
| { | |
| "epoch": 0.33, | |
| "learning_rate": 0.0007894830461207085, | |
| "loss": 2.5596, | |
| "step": 575 | |
| }, | |
| { | |
| "epoch": 0.33, | |
| "learning_rate": 0.0007887322625946835, | |
| "loss": 2.503, | |
| "step": 576 | |
| }, | |
| { | |
| "epoch": 0.33, | |
| "learning_rate": 0.0007879805010933396, | |
| "loss": 2.5275, | |
| "step": 577 | |
| }, | |
| { | |
| "epoch": 0.33, | |
| "learning_rate": 0.0007872277641629948, | |
| "loss": 2.4316, | |
| "step": 578 | |
| }, | |
| { | |
| "epoch": 0.33, | |
| "learning_rate": 0.0007864740543532711, | |
| "loss": 2.4647, | |
| "step": 579 | |
| }, | |
| { | |
| "epoch": 0.33, | |
| "learning_rate": 0.0007857193742170859, | |
| "loss": 2.4628, | |
| "step": 580 | |
| }, | |
| { | |
| "epoch": 0.33, | |
| "learning_rate": 0.0007849637263106431, | |
| "loss": 2.4359, | |
| "step": 581 | |
| }, | |
| { | |
| "epoch": 0.33, | |
| "learning_rate": 0.0007842071131934246, | |
| "loss": 2.3986, | |
| "step": 582 | |
| }, | |
| { | |
| "epoch": 0.33, | |
| "learning_rate": 0.0007834495374281816, | |
| "loss": 2.459, | |
| "step": 583 | |
| }, | |
| { | |
| "epoch": 0.33, | |
| "learning_rate": 0.0007826910015809261, | |
| "loss": 2.4605, | |
| "step": 584 | |
| }, | |
| { | |
| "epoch": 0.33, | |
| "learning_rate": 0.0007819315082209217, | |
| "loss": 2.4607, | |
| "step": 585 | |
| }, | |
| { | |
| "epoch": 0.33, | |
| "learning_rate": 0.0007811710599206752, | |
| "loss": 2.4803, | |
| "step": 586 | |
| }, | |
| { | |
| "epoch": 0.33, | |
| "learning_rate": 0.0007804096592559284, | |
| "loss": 2.5248, | |
| "step": 587 | |
| }, | |
| { | |
| "epoch": 0.33, | |
| "learning_rate": 0.0007796473088056487, | |
| "loss": 2.4919, | |
| "step": 588 | |
| }, | |
| { | |
| "epoch": 0.33, | |
| "learning_rate": 0.0007788840111520197, | |
| "loss": 2.5144, | |
| "step": 589 | |
| }, | |
| { | |
| "epoch": 0.34, | |
| "learning_rate": 0.0007781197688804349, | |
| "loss": 2.513, | |
| "step": 590 | |
| }, | |
| { | |
| "epoch": 0.34, | |
| "learning_rate": 0.000777354584579486, | |
| "loss": 2.4596, | |
| "step": 591 | |
| }, | |
| { | |
| "epoch": 0.34, | |
| "learning_rate": 0.0007765884608409561, | |
| "loss": 2.4682, | |
| "step": 592 | |
| }, | |
| { | |
| "epoch": 0.34, | |
| "learning_rate": 0.0007758214002598103, | |
| "loss": 2.3991, | |
| "step": 593 | |
| }, | |
| { | |
| "epoch": 0.34, | |
| "learning_rate": 0.0007750534054341866, | |
| "loss": 2.4929, | |
| "step": 594 | |
| }, | |
| { | |
| "epoch": 0.34, | |
| "learning_rate": 0.0007742844789653875, | |
| "loss": 2.4513, | |
| "step": 595 | |
| }, | |
| { | |
| "epoch": 0.34, | |
| "learning_rate": 0.0007735146234578716, | |
| "loss": 2.49, | |
| "step": 596 | |
| }, | |
| { | |
| "epoch": 0.34, | |
| "learning_rate": 0.0007727438415192434, | |
| "loss": 2.4734, | |
| "step": 597 | |
| }, | |
| { | |
| "epoch": 0.34, | |
| "learning_rate": 0.0007719721357602461, | |
| "loss": 2.4681, | |
| "step": 598 | |
| }, | |
| { | |
| "epoch": 0.34, | |
| "learning_rate": 0.0007711995087947517, | |
| "loss": 2.4802, | |
| "step": 599 | |
| }, | |
| { | |
| "epoch": 0.34, | |
| "learning_rate": 0.0007704259632397525, | |
| "loss": 2.5405, | |
| "step": 600 | |
| }, | |
| { | |
| "epoch": 0.34, | |
| "learning_rate": 0.0007696515017153522, | |
| "loss": 2.5218, | |
| "step": 601 | |
| }, | |
| { | |
| "epoch": 0.34, | |
| "learning_rate": 0.0007688761268447569, | |
| "loss": 2.52, | |
| "step": 602 | |
| }, | |
| { | |
| "epoch": 0.34, | |
| "learning_rate": 0.0007680998412542664, | |
| "loss": 2.5314, | |
| "step": 603 | |
| }, | |
| { | |
| "epoch": 0.34, | |
| "learning_rate": 0.0007673226475732652, | |
| "loss": 2.3808, | |
| "step": 604 | |
| }, | |
| { | |
| "epoch": 0.34, | |
| "learning_rate": 0.0007665445484342137, | |
| "loss": 2.4448, | |
| "step": 605 | |
| }, | |
| { | |
| "epoch": 0.34, | |
| "learning_rate": 0.0007657655464726394, | |
| "loss": 2.4814, | |
| "step": 606 | |
| }, | |
| { | |
| "epoch": 0.34, | |
| "learning_rate": 0.0007649856443271273, | |
| "loss": 2.4774, | |
| "step": 607 | |
| }, | |
| { | |
| "epoch": 0.35, | |
| "learning_rate": 0.0007642048446393117, | |
| "loss": 2.4356, | |
| "step": 608 | |
| }, | |
| { | |
| "epoch": 0.35, | |
| "learning_rate": 0.0007634231500538671, | |
| "loss": 2.4701, | |
| "step": 609 | |
| }, | |
| { | |
| "epoch": 0.35, | |
| "learning_rate": 0.0007626405632184992, | |
| "loss": 2.4459, | |
| "step": 610 | |
| }, | |
| { | |
| "epoch": 0.35, | |
| "learning_rate": 0.0007618570867839354, | |
| "loss": 2.5487, | |
| "step": 611 | |
| }, | |
| { | |
| "epoch": 0.35, | |
| "learning_rate": 0.0007610727234039167, | |
| "loss": 2.4338, | |
| "step": 612 | |
| }, | |
| { | |
| "epoch": 0.35, | |
| "learning_rate": 0.0007602874757351883, | |
| "loss": 2.4479, | |
| "step": 613 | |
| }, | |
| { | |
| "epoch": 0.35, | |
| "learning_rate": 0.0007595013464374904, | |
| "loss": 2.4928, | |
| "step": 614 | |
| }, | |
| { | |
| "epoch": 0.35, | |
| "learning_rate": 0.0007587143381735498, | |
| "loss": 2.423, | |
| "step": 615 | |
| }, | |
| { | |
| "epoch": 0.35, | |
| "learning_rate": 0.0007579264536090697, | |
| "loss": 2.4827, | |
| "step": 616 | |
| }, | |
| { | |
| "epoch": 0.35, | |
| "learning_rate": 0.0007571376954127223, | |
| "loss": 2.5261, | |
| "step": 617 | |
| }, | |
| { | |
| "epoch": 0.35, | |
| "learning_rate": 0.0007563480662561386, | |
| "loss": 2.47, | |
| "step": 618 | |
| }, | |
| { | |
| "epoch": 0.35, | |
| "learning_rate": 0.0007555575688138993, | |
| "loss": 2.4221, | |
| "step": 619 | |
| }, | |
| { | |
| "epoch": 0.35, | |
| "learning_rate": 0.0007547662057635266, | |
| "loss": 2.4059, | |
| "step": 620 | |
| }, | |
| { | |
| "epoch": 0.35, | |
| "learning_rate": 0.0007539739797854746, | |
| "loss": 2.381, | |
| "step": 621 | |
| }, | |
| { | |
| "epoch": 0.35, | |
| "learning_rate": 0.0007531808935631197, | |
| "loss": 2.5162, | |
| "step": 622 | |
| }, | |
| { | |
| "epoch": 0.35, | |
| "learning_rate": 0.0007523869497827527, | |
| "loss": 2.4406, | |
| "step": 623 | |
| }, | |
| { | |
| "epoch": 0.35, | |
| "learning_rate": 0.0007515921511335688, | |
| "loss": 2.4417, | |
| "step": 624 | |
| }, | |
| { | |
| "epoch": 0.36, | |
| "learning_rate": 0.0007507965003076589, | |
| "loss": 2.4449, | |
| "step": 625 | |
| }, | |
| { | |
| "epoch": 0.36, | |
| "learning_rate": 0.00075, | |
| "loss": 2.4725, | |
| "step": 626 | |
| }, | |
| { | |
| "epoch": 0.36, | |
| "learning_rate": 0.0007492026529084468, | |
| "loss": 2.4522, | |
| "step": 627 | |
| }, | |
| { | |
| "epoch": 0.36, | |
| "learning_rate": 0.000748404461733722, | |
| "loss": 2.5874, | |
| "step": 628 | |
| }, | |
| { | |
| "epoch": 0.36, | |
| "learning_rate": 0.000747605429179407, | |
| "loss": 2.5147, | |
| "step": 629 | |
| }, | |
| { | |
| "epoch": 0.36, | |
| "learning_rate": 0.0007468055579519338, | |
| "loss": 2.4721, | |
| "step": 630 | |
| }, | |
| { | |
| "epoch": 0.36, | |
| "learning_rate": 0.0007460048507605746, | |
| "loss": 2.4702, | |
| "step": 631 | |
| }, | |
| { | |
| "epoch": 0.36, | |
| "learning_rate": 0.0007452033103174332, | |
| "loss": 2.3927, | |
| "step": 632 | |
| }, | |
| { | |
| "epoch": 0.36, | |
| "learning_rate": 0.0007444009393374356, | |
| "loss": 2.4879, | |
| "step": 633 | |
| }, | |
| { | |
| "epoch": 0.36, | |
| "learning_rate": 0.0007435977405383212, | |
| "loss": 2.4123, | |
| "step": 634 | |
| }, | |
| { | |
| "epoch": 0.36, | |
| "learning_rate": 0.0007427937166406332, | |
| "loss": 2.4447, | |
| "step": 635 | |
| }, | |
| { | |
| "epoch": 0.36, | |
| "learning_rate": 0.0007419888703677097, | |
| "loss": 2.4302, | |
| "step": 636 | |
| }, | |
| { | |
| "epoch": 0.36, | |
| "learning_rate": 0.0007411832044456739, | |
| "loss": 2.4721, | |
| "step": 637 | |
| }, | |
| { | |
| "epoch": 0.36, | |
| "learning_rate": 0.0007403767216034257, | |
| "loss": 2.4133, | |
| "step": 638 | |
| }, | |
| { | |
| "epoch": 0.36, | |
| "learning_rate": 0.0007395694245726318, | |
| "loss": 2.3523, | |
| "step": 639 | |
| }, | |
| { | |
| "epoch": 0.36, | |
| "learning_rate": 0.0007387613160877165, | |
| "loss": 2.4691, | |
| "step": 640 | |
| }, | |
| { | |
| "epoch": 0.36, | |
| "learning_rate": 0.0007379523988858532, | |
| "loss": 2.447, | |
| "step": 641 | |
| }, | |
| { | |
| "epoch": 0.36, | |
| "learning_rate": 0.0007371426757069537, | |
| "loss": 2.5446, | |
| "step": 642 | |
| }, | |
| { | |
| "epoch": 0.37, | |
| "learning_rate": 0.0007363321492936604, | |
| "loss": 2.4753, | |
| "step": 643 | |
| }, | |
| { | |
| "epoch": 0.37, | |
| "learning_rate": 0.0007355208223913362, | |
| "loss": 2.4594, | |
| "step": 644 | |
| }, | |
| { | |
| "epoch": 0.37, | |
| "learning_rate": 0.0007347086977480551, | |
| "loss": 2.4352, | |
| "step": 645 | |
| }, | |
| { | |
| "epoch": 0.37, | |
| "learning_rate": 0.0007338957781145936, | |
| "loss": 2.442, | |
| "step": 646 | |
| }, | |
| { | |
| "epoch": 0.37, | |
| "learning_rate": 0.0007330820662444207, | |
| "loss": 2.3776, | |
| "step": 647 | |
| }, | |
| { | |
| "epoch": 0.37, | |
| "learning_rate": 0.0007322675648936887, | |
| "loss": 2.3646, | |
| "step": 648 | |
| }, | |
| { | |
| "epoch": 0.37, | |
| "learning_rate": 0.0007314522768212241, | |
| "loss": 2.4233, | |
| "step": 649 | |
| }, | |
| { | |
| "epoch": 0.37, | |
| "learning_rate": 0.0007306362047885182, | |
| "loss": 2.4675, | |
| "step": 650 | |
| }, | |
| { | |
| "epoch": 0.37, | |
| "learning_rate": 0.0007298193515597177, | |
| "loss": 2.5046, | |
| "step": 651 | |
| }, | |
| { | |
| "epoch": 0.37, | |
| "learning_rate": 0.0007290017199016152, | |
| "loss": 2.548, | |
| "step": 652 | |
| }, | |
| { | |
| "epoch": 0.37, | |
| "learning_rate": 0.0007281833125836397, | |
| "loss": 2.4224, | |
| "step": 653 | |
| }, | |
| { | |
| "epoch": 0.37, | |
| "learning_rate": 0.0007273641323778481, | |
| "loss": 2.4324, | |
| "step": 654 | |
| }, | |
| { | |
| "epoch": 0.37, | |
| "learning_rate": 0.0007265441820589145, | |
| "loss": 2.4377, | |
| "step": 655 | |
| }, | |
| { | |
| "epoch": 0.37, | |
| "learning_rate": 0.0007257234644041222, | |
| "loss": 2.4007, | |
| "step": 656 | |
| }, | |
| { | |
| "epoch": 0.37, | |
| "learning_rate": 0.0007249019821933529, | |
| "loss": 2.4604, | |
| "step": 657 | |
| }, | |
| { | |
| "epoch": 0.37, | |
| "learning_rate": 0.0007240797382090783, | |
| "loss": 2.483, | |
| "step": 658 | |
| }, | |
| { | |
| "epoch": 0.37, | |
| "learning_rate": 0.0007232567352363501, | |
| "loss": 2.4878, | |
| "step": 659 | |
| }, | |
| { | |
| "epoch": 0.37, | |
| "learning_rate": 0.0007224329760627911, | |
| "loss": 2.5135, | |
| "step": 660 | |
| }, | |
| { | |
| "epoch": 0.38, | |
| "learning_rate": 0.0007216084634785854, | |
| "loss": 2.495, | |
| "step": 661 | |
| }, | |
| { | |
| "epoch": 0.38, | |
| "learning_rate": 0.0007207832002764687, | |
| "loss": 2.4743, | |
| "step": 662 | |
| }, | |
| { | |
| "epoch": 0.38, | |
| "learning_rate": 0.0007199571892517193, | |
| "loss": 2.4359, | |
| "step": 663 | |
| }, | |
| { | |
| "epoch": 0.38, | |
| "learning_rate": 0.0007191304332021489, | |
| "loss": 2.4334, | |
| "step": 664 | |
| }, | |
| { | |
| "epoch": 0.38, | |
| "learning_rate": 0.0007183029349280921, | |
| "loss": 2.4952, | |
| "step": 665 | |
| }, | |
| { | |
| "epoch": 0.38, | |
| "learning_rate": 0.0007174746972323975, | |
| "loss": 2.439, | |
| "step": 666 | |
| }, | |
| { | |
| "epoch": 0.38, | |
| "learning_rate": 0.0007166457229204189, | |
| "loss": 2.5289, | |
| "step": 667 | |
| }, | |
| { | |
| "epoch": 0.38, | |
| "learning_rate": 0.0007158160148000044, | |
| "loss": 2.4211, | |
| "step": 668 | |
| }, | |
| { | |
| "epoch": 0.38, | |
| "learning_rate": 0.000714985575681488, | |
| "loss": 2.4422, | |
| "step": 669 | |
| }, | |
| { | |
| "epoch": 0.38, | |
| "learning_rate": 0.0007141544083776795, | |
| "loss": 2.4207, | |
| "step": 670 | |
| }, | |
| { | |
| "epoch": 0.38, | |
| "learning_rate": 0.0007133225157038554, | |
| "loss": 2.4256, | |
| "step": 671 | |
| }, | |
| { | |
| "epoch": 0.38, | |
| "learning_rate": 0.0007124899004777489, | |
| "loss": 2.4774, | |
| "step": 672 | |
| }, | |
| { | |
| "epoch": 0.38, | |
| "learning_rate": 0.0007116565655195406, | |
| "loss": 2.3843, | |
| "step": 673 | |
| }, | |
| { | |
| "epoch": 0.38, | |
| "learning_rate": 0.000710822513651849, | |
| "loss": 2.5185, | |
| "step": 674 | |
| }, | |
| { | |
| "epoch": 0.38, | |
| "learning_rate": 0.000709987747699721, | |
| "loss": 2.4379, | |
| "step": 675 | |
| }, | |
| { | |
| "epoch": 0.38, | |
| "learning_rate": 0.0007091522704906218, | |
| "loss": 2.4733, | |
| "step": 676 | |
| }, | |
| { | |
| "epoch": 0.38, | |
| "learning_rate": 0.0007083160848544261, | |
| "loss": 2.4462, | |
| "step": 677 | |
| }, | |
| { | |
| "epoch": 0.39, | |
| "learning_rate": 0.0007074791936234082, | |
| "loss": 2.3959, | |
| "step": 678 | |
| }, | |
| { | |
| "epoch": 0.39, | |
| "learning_rate": 0.0007066415996322318, | |
| "loss": 2.4609, | |
| "step": 679 | |
| }, | |
| { | |
| "epoch": 0.39, | |
| "learning_rate": 0.0007058033057179415, | |
| "loss": 2.4024, | |
| "step": 680 | |
| }, | |
| { | |
| "epoch": 0.39, | |
| "learning_rate": 0.0007049643147199524, | |
| "loss": 2.5052, | |
| "step": 681 | |
| }, | |
| { | |
| "epoch": 0.39, | |
| "learning_rate": 0.0007041246294800411, | |
| "loss": 2.4512, | |
| "step": 682 | |
| }, | |
| { | |
| "epoch": 0.39, | |
| "learning_rate": 0.0007032842528423348, | |
| "loss": 2.4629, | |
| "step": 683 | |
| }, | |
| { | |
| "epoch": 0.39, | |
| "learning_rate": 0.0007024431876533035, | |
| "loss": 2.3917, | |
| "step": 684 | |
| }, | |
| { | |
| "epoch": 0.39, | |
| "learning_rate": 0.0007016014367617487, | |
| "loss": 2.4708, | |
| "step": 685 | |
| }, | |
| { | |
| "epoch": 0.39, | |
| "learning_rate": 0.0007007590030187952, | |
| "loss": 2.4183, | |
| "step": 686 | |
| }, | |
| { | |
| "epoch": 0.39, | |
| "learning_rate": 0.0006999158892778798, | |
| "loss": 2.4237, | |
| "step": 687 | |
| }, | |
| { | |
| "epoch": 0.39, | |
| "learning_rate": 0.0006990720983947435, | |
| "loss": 2.45, | |
| "step": 688 | |
| }, | |
| { | |
| "epoch": 0.39, | |
| "learning_rate": 0.0006982276332274202, | |
| "loss": 2.4202, | |
| "step": 689 | |
| }, | |
| { | |
| "epoch": 0.39, | |
| "learning_rate": 0.0006973824966362281, | |
| "loss": 2.4131, | |
| "step": 690 | |
| }, | |
| { | |
| "epoch": 0.39, | |
| "learning_rate": 0.0006965366914837591, | |
| "loss": 2.4219, | |
| "step": 691 | |
| }, | |
| { | |
| "epoch": 0.39, | |
| "learning_rate": 0.0006956902206348702, | |
| "loss": 2.4663, | |
| "step": 692 | |
| }, | |
| { | |
| "epoch": 0.39, | |
| "learning_rate": 0.0006948430869566728, | |
| "loss": 2.5585, | |
| "step": 693 | |
| }, | |
| { | |
| "epoch": 0.39, | |
| "learning_rate": 0.0006939952933185234, | |
| "loss": 2.4702, | |
| "step": 694 | |
| }, | |
| { | |
| "epoch": 0.39, | |
| "learning_rate": 0.0006931468425920142, | |
| "loss": 2.4307, | |
| "step": 695 | |
| }, | |
| { | |
| "epoch": 0.4, | |
| "learning_rate": 0.0006922977376509629, | |
| "loss": 2.4803, | |
| "step": 696 | |
| }, | |
| { | |
| "epoch": 0.4, | |
| "learning_rate": 0.000691447981371403, | |
| "loss": 2.4817, | |
| "step": 697 | |
| }, | |
| { | |
| "epoch": 0.4, | |
| "learning_rate": 0.0006905975766315739, | |
| "loss": 2.4975, | |
| "step": 698 | |
| }, | |
| { | |
| "epoch": 0.4, | |
| "learning_rate": 0.0006897465263119122, | |
| "loss": 2.5327, | |
| "step": 699 | |
| }, | |
| { | |
| "epoch": 0.4, | |
| "learning_rate": 0.0006888948332950405, | |
| "loss": 2.4044, | |
| "step": 700 | |
| }, | |
| { | |
| "epoch": 0.4, | |
| "learning_rate": 0.0006880425004657586, | |
| "loss": 2.5197, | |
| "step": 701 | |
| }, | |
| { | |
| "epoch": 0.4, | |
| "learning_rate": 0.0006871895307110332, | |
| "loss": 2.4461, | |
| "step": 702 | |
| }, | |
| { | |
| "epoch": 0.4, | |
| "learning_rate": 0.0006863359269199886, | |
| "loss": 2.4614, | |
| "step": 703 | |
| }, | |
| { | |
| "epoch": 0.4, | |
| "learning_rate": 0.0006854816919838967, | |
| "loss": 2.3401, | |
| "step": 704 | |
| }, | |
| { | |
| "epoch": 0.4, | |
| "learning_rate": 0.0006846268287961667, | |
| "loss": 2.4542, | |
| "step": 705 | |
| }, | |
| { | |
| "epoch": 0.4, | |
| "learning_rate": 0.0006837713402523364, | |
| "loss": 2.4437, | |
| "step": 706 | |
| }, | |
| { | |
| "epoch": 0.4, | |
| "learning_rate": 0.0006829152292500613, | |
| "loss": 2.4147, | |
| "step": 707 | |
| }, | |
| { | |
| "epoch": 0.4, | |
| "learning_rate": 0.0006820584986891058, | |
| "loss": 2.512, | |
| "step": 708 | |
| }, | |
| { | |
| "epoch": 0.4, | |
| "learning_rate": 0.0006812011514713321, | |
| "loss": 2.4996, | |
| "step": 709 | |
| }, | |
| { | |
| "epoch": 0.4, | |
| "learning_rate": 0.0006803431905006916, | |
| "loss": 2.4634, | |
| "step": 710 | |
| }, | |
| { | |
| "epoch": 0.4, | |
| "learning_rate": 0.0006794846186832143, | |
| "loss": 2.4315, | |
| "step": 711 | |
| }, | |
| { | |
| "epoch": 0.4, | |
| "learning_rate": 0.0006786254389269994, | |
| "loss": 2.4492, | |
| "step": 712 | |
| }, | |
| { | |
| "epoch": 0.41, | |
| "learning_rate": 0.0006777656541422055, | |
| "loss": 2.4084, | |
| "step": 713 | |
| }, | |
| { | |
| "epoch": 0.41, | |
| "learning_rate": 0.0006769052672410398, | |
| "loss": 2.4165, | |
| "step": 714 | |
| }, | |
| { | |
| "epoch": 0.41, | |
| "learning_rate": 0.0006760442811377499, | |
| "loss": 2.4669, | |
| "step": 715 | |
| }, | |
| { | |
| "epoch": 0.41, | |
| "learning_rate": 0.0006751826987486118, | |
| "loss": 2.3723, | |
| "step": 716 | |
| }, | |
| { | |
| "epoch": 0.41, | |
| "learning_rate": 0.0006743205229919224, | |
| "loss": 2.4004, | |
| "step": 717 | |
| }, | |
| { | |
| "epoch": 0.41, | |
| "learning_rate": 0.0006734577567879876, | |
| "loss": 2.4222, | |
| "step": 718 | |
| }, | |
| { | |
| "epoch": 0.41, | |
| "learning_rate": 0.0006725944030591135, | |
| "loss": 2.4204, | |
| "step": 719 | |
| }, | |
| { | |
| "epoch": 0.41, | |
| "learning_rate": 0.0006717304647295964, | |
| "loss": 2.4191, | |
| "step": 720 | |
| }, | |
| { | |
| "epoch": 0.41, | |
| "learning_rate": 0.0006708659447257126, | |
| "loss": 2.4486, | |
| "step": 721 | |
| }, | |
| { | |
| "epoch": 0.41, | |
| "learning_rate": 0.0006700008459757083, | |
| "loss": 2.3486, | |
| "step": 722 | |
| }, | |
| { | |
| "epoch": 0.41, | |
| "learning_rate": 0.0006691351714097905, | |
| "loss": 2.3805, | |
| "step": 723 | |
| }, | |
| { | |
| "epoch": 0.41, | |
| "learning_rate": 0.0006682689239601161, | |
| "loss": 2.402, | |
| "step": 724 | |
| }, | |
| { | |
| "epoch": 0.41, | |
| "learning_rate": 0.0006674021065607829, | |
| "loss": 2.367, | |
| "step": 725 | |
| }, | |
| { | |
| "epoch": 0.41, | |
| "learning_rate": 0.0006665347221478187, | |
| "loss": 2.4574, | |
| "step": 726 | |
| }, | |
| { | |
| "epoch": 0.41, | |
| "learning_rate": 0.0006656667736591724, | |
| "loss": 2.4072, | |
| "step": 727 | |
| }, | |
| { | |
| "epoch": 0.41, | |
| "learning_rate": 0.0006647982640347029, | |
| "loss": 2.4673, | |
| "step": 728 | |
| }, | |
| { | |
| "epoch": 0.41, | |
| "learning_rate": 0.0006639291962161702, | |
| "loss": 2.4676, | |
| "step": 729 | |
| }, | |
| { | |
| "epoch": 0.41, | |
| "learning_rate": 0.0006630595731472249, | |
| "loss": 2.4421, | |
| "step": 730 | |
| }, | |
| { | |
| "epoch": 0.42, | |
| "learning_rate": 0.000662189397773398, | |
| "loss": 2.4604, | |
| "step": 731 | |
| }, | |
| { | |
| "epoch": 0.42, | |
| "learning_rate": 0.0006613186730420917, | |
| "loss": 2.4199, | |
| "step": 732 | |
| }, | |
| { | |
| "epoch": 0.42, | |
| "learning_rate": 0.0006604474019025686, | |
| "loss": 2.4471, | |
| "step": 733 | |
| }, | |
| { | |
| "epoch": 0.42, | |
| "learning_rate": 0.0006595755873059422, | |
| "loss": 2.4323, | |
| "step": 734 | |
| }, | |
| { | |
| "epoch": 0.42, | |
| "learning_rate": 0.0006587032322051667, | |
| "loss": 2.3491, | |
| "step": 735 | |
| }, | |
| { | |
| "epoch": 0.42, | |
| "learning_rate": 0.0006578303395550271, | |
| "loss": 2.4445, | |
| "step": 736 | |
| }, | |
| { | |
| "epoch": 0.42, | |
| "learning_rate": 0.0006569569123121294, | |
| "loss": 2.5522, | |
| "step": 737 | |
| }, | |
| { | |
| "epoch": 0.42, | |
| "learning_rate": 0.0006560829534348897, | |
| "loss": 2.393, | |
| "step": 738 | |
| }, | |
| { | |
| "epoch": 0.42, | |
| "learning_rate": 0.0006552084658835254, | |
| "loss": 2.4933, | |
| "step": 739 | |
| }, | |
| { | |
| "epoch": 0.42, | |
| "learning_rate": 0.0006543334526200445, | |
| "loss": 2.4027, | |
| "step": 740 | |
| }, | |
| { | |
| "epoch": 0.42, | |
| "learning_rate": 0.0006534579166082354, | |
| "loss": 2.4971, | |
| "step": 741 | |
| }, | |
| { | |
| "epoch": 0.42, | |
| "learning_rate": 0.0006525818608136572, | |
| "loss": 2.4315, | |
| "step": 742 | |
| }, | |
| { | |
| "epoch": 0.42, | |
| "learning_rate": 0.0006517052882036298, | |
| "loss": 2.4696, | |
| "step": 743 | |
| }, | |
| { | |
| "epoch": 0.42, | |
| "learning_rate": 0.0006508282017472235, | |
| "loss": 2.4609, | |
| "step": 744 | |
| }, | |
| { | |
| "epoch": 0.42, | |
| "learning_rate": 0.000649950604415249, | |
| "loss": 2.4614, | |
| "step": 745 | |
| }, | |
| { | |
| "epoch": 0.42, | |
| "learning_rate": 0.0006490724991802474, | |
| "loss": 2.4132, | |
| "step": 746 | |
| }, | |
| { | |
| "epoch": 0.42, | |
| "learning_rate": 0.00064819388901648, | |
| "loss": 2.5059, | |
| "step": 747 | |
| }, | |
| { | |
| "epoch": 0.42, | |
| "learning_rate": 0.000647314776899919, | |
| "loss": 2.4084, | |
| "step": 748 | |
| }, | |
| { | |
| "epoch": 0.43, | |
| "learning_rate": 0.000646435165808236, | |
| "loss": 2.4937, | |
| "step": 749 | |
| }, | |
| { | |
| "epoch": 0.43, | |
| "learning_rate": 0.000645555058720793, | |
| "loss": 2.5148, | |
| "step": 750 | |
| }, | |
| { | |
| "epoch": 0.43, | |
| "learning_rate": 0.0006446744586186322, | |
| "loss": 2.4094, | |
| "step": 751 | |
| }, | |
| { | |
| "epoch": 0.43, | |
| "learning_rate": 0.0006437933684844655, | |
| "loss": 2.4952, | |
| "step": 752 | |
| }, | |
| { | |
| "epoch": 0.43, | |
| "learning_rate": 0.0006429117913026646, | |
| "loss": 2.4557, | |
| "step": 753 | |
| }, | |
| { | |
| "epoch": 0.43, | |
| "learning_rate": 0.000642029730059251, | |
| "loss": 2.5387, | |
| "step": 754 | |
| }, | |
| { | |
| "epoch": 0.43, | |
| "learning_rate": 0.0006411471877418855, | |
| "loss": 2.4937, | |
| "step": 755 | |
| }, | |
| { | |
| "epoch": 0.43, | |
| "learning_rate": 0.000640264167339859, | |
| "loss": 2.4658, | |
| "step": 756 | |
| }, | |
| { | |
| "epoch": 0.43, | |
| "learning_rate": 0.0006393806718440806, | |
| "loss": 2.4209, | |
| "step": 757 | |
| }, | |
| { | |
| "epoch": 0.43, | |
| "learning_rate": 0.0006384967042470702, | |
| "loss": 2.4134, | |
| "step": 758 | |
| }, | |
| { | |
| "epoch": 0.43, | |
| "learning_rate": 0.0006376122675429452, | |
| "loss": 2.4574, | |
| "step": 759 | |
| }, | |
| { | |
| "epoch": 0.43, | |
| "learning_rate": 0.0006367273647274131, | |
| "loss": 2.3942, | |
| "step": 760 | |
| }, | |
| { | |
| "epoch": 0.43, | |
| "learning_rate": 0.0006358419987977595, | |
| "loss": 2.4388, | |
| "step": 761 | |
| }, | |
| { | |
| "epoch": 0.43, | |
| "learning_rate": 0.0006349561727528388, | |
| "loss": 2.3744, | |
| "step": 762 | |
| }, | |
| { | |
| "epoch": 0.43, | |
| "learning_rate": 0.0006340698895930638, | |
| "loss": 2.4443, | |
| "step": 763 | |
| }, | |
| { | |
| "epoch": 0.43, | |
| "learning_rate": 0.0006331831523203963, | |
| "loss": 2.4567, | |
| "step": 764 | |
| }, | |
| { | |
| "epoch": 0.43, | |
| "learning_rate": 0.000632295963938335, | |
| "loss": 2.498, | |
| "step": 765 | |
| }, | |
| { | |
| "epoch": 0.44, | |
| "learning_rate": 0.0006314083274519076, | |
| "loss": 2.5225, | |
| "step": 766 | |
| }, | |
| { | |
| "epoch": 0.44, | |
| "learning_rate": 0.000630520245867659, | |
| "loss": 2.4169, | |
| "step": 767 | |
| }, | |
| { | |
| "epoch": 0.44, | |
| "learning_rate": 0.0006296317221936421, | |
| "loss": 2.4793, | |
| "step": 768 | |
| }, | |
| { | |
| "epoch": 0.44, | |
| "learning_rate": 0.0006287427594394069, | |
| "loss": 2.4272, | |
| "step": 769 | |
| }, | |
| { | |
| "epoch": 0.44, | |
| "learning_rate": 0.000627853360615991, | |
| "loss": 2.5425, | |
| "step": 770 | |
| }, | |
| { | |
| "epoch": 0.44, | |
| "learning_rate": 0.0006269635287359086, | |
| "loss": 2.3694, | |
| "step": 771 | |
| }, | |
| { | |
| "epoch": 0.44, | |
| "learning_rate": 0.0006260732668131409, | |
| "loss": 2.4553, | |
| "step": 772 | |
| }, | |
| { | |
| "epoch": 0.44, | |
| "learning_rate": 0.0006251825778631258, | |
| "loss": 2.4705, | |
| "step": 773 | |
| }, | |
| { | |
| "epoch": 0.44, | |
| "learning_rate": 0.0006242914649027476, | |
| "loss": 2.416, | |
| "step": 774 | |
| }, | |
| { | |
| "epoch": 0.44, | |
| "learning_rate": 0.0006233999309503267, | |
| "loss": 2.3962, | |
| "step": 775 | |
| }, | |
| { | |
| "epoch": 0.44, | |
| "learning_rate": 0.0006225079790256094, | |
| "loss": 2.4481, | |
| "step": 776 | |
| }, | |
| { | |
| "epoch": 0.44, | |
| "learning_rate": 0.0006216156121497578, | |
| "loss": 2.4229, | |
| "step": 777 | |
| }, | |
| { | |
| "epoch": 0.44, | |
| "learning_rate": 0.0006207228333453396, | |
| "loss": 2.3921, | |
| "step": 778 | |
| }, | |
| { | |
| "epoch": 0.44, | |
| "learning_rate": 0.0006198296456363174, | |
| "loss": 2.4722, | |
| "step": 779 | |
| }, | |
| { | |
| "epoch": 0.44, | |
| "learning_rate": 0.0006189360520480393, | |
| "loss": 2.4749, | |
| "step": 780 | |
| }, | |
| { | |
| "epoch": 0.44, | |
| "learning_rate": 0.0006180420556072279, | |
| "loss": 2.482, | |
| "step": 781 | |
| }, | |
| { | |
| "epoch": 0.44, | |
| "learning_rate": 0.0006171476593419703, | |
| "loss": 2.4231, | |
| "step": 782 | |
| }, | |
| { | |
| "epoch": 0.44, | |
| "learning_rate": 0.000616252866281708, | |
| "loss": 2.4011, | |
| "step": 783 | |
| }, | |
| { | |
| "epoch": 0.45, | |
| "learning_rate": 0.0006153576794572262, | |
| "loss": 2.4231, | |
| "step": 784 | |
| }, | |
| { | |
| "epoch": 0.45, | |
| "learning_rate": 0.0006144621019006443, | |
| "loss": 2.478, | |
| "step": 785 | |
| }, | |
| { | |
| "epoch": 0.45, | |
| "learning_rate": 0.0006135661366454045, | |
| "loss": 2.4443, | |
| "step": 786 | |
| }, | |
| { | |
| "epoch": 0.45, | |
| "learning_rate": 0.0006126697867262632, | |
| "loss": 2.4758, | |
| "step": 787 | |
| }, | |
| { | |
| "epoch": 0.45, | |
| "learning_rate": 0.0006117730551792786, | |
| "loss": 2.4348, | |
| "step": 788 | |
| }, | |
| { | |
| "epoch": 0.45, | |
| "learning_rate": 0.0006108759450418022, | |
| "loss": 2.4317, | |
| "step": 789 | |
| }, | |
| { | |
| "epoch": 0.45, | |
| "learning_rate": 0.0006099784593524678, | |
| "loss": 2.4133, | |
| "step": 790 | |
| }, | |
| { | |
| "epoch": 0.45, | |
| "learning_rate": 0.0006090806011511808, | |
| "loss": 2.4218, | |
| "step": 791 | |
| }, | |
| { | |
| "epoch": 0.45, | |
| "learning_rate": 0.0006081823734791089, | |
| "loss": 2.3959, | |
| "step": 792 | |
| }, | |
| { | |
| "epoch": 0.45, | |
| "learning_rate": 0.000607283779378671, | |
| "loss": 2.436, | |
| "step": 793 | |
| }, | |
| { | |
| "epoch": 0.45, | |
| "learning_rate": 0.000606384821893527, | |
| "loss": 2.4626, | |
| "step": 794 | |
| }, | |
| { | |
| "epoch": 0.45, | |
| "learning_rate": 0.0006054855040685679, | |
| "loss": 2.4465, | |
| "step": 795 | |
| }, | |
| { | |
| "epoch": 0.45, | |
| "learning_rate": 0.000604585828949905, | |
| "loss": 2.4273, | |
| "step": 796 | |
| }, | |
| { | |
| "epoch": 0.45, | |
| "learning_rate": 0.00060368579958486, | |
| "loss": 2.4659, | |
| "step": 797 | |
| }, | |
| { | |
| "epoch": 0.45, | |
| "learning_rate": 0.0006027854190219544, | |
| "loss": 2.3485, | |
| "step": 798 | |
| }, | |
| { | |
| "epoch": 0.45, | |
| "learning_rate": 0.0006018846903108991, | |
| "loss": 2.4722, | |
| "step": 799 | |
| }, | |
| { | |
| "epoch": 0.45, | |
| "learning_rate": 0.0006009836165025845, | |
| "loss": 2.372, | |
| "step": 800 | |
| }, | |
| { | |
| "epoch": 0.46, | |
| "learning_rate": 0.0006000822006490696, | |
| "loss": 2.4584, | |
| "step": 801 | |
| }, | |
| { | |
| "epoch": 0.46, | |
| "learning_rate": 0.0005991804458035724, | |
| "loss": 2.5084, | |
| "step": 802 | |
| }, | |
| { | |
| "epoch": 0.46, | |
| "learning_rate": 0.0005982783550204585, | |
| "loss": 2.3997, | |
| "step": 803 | |
| }, | |
| { | |
| "epoch": 0.46, | |
| "learning_rate": 0.0005973759313552318, | |
| "loss": 2.4404, | |
| "step": 804 | |
| }, | |
| { | |
| "epoch": 0.46, | |
| "learning_rate": 0.0005964731778645237, | |
| "loss": 2.4222, | |
| "step": 805 | |
| }, | |
| { | |
| "epoch": 0.46, | |
| "learning_rate": 0.0005955700976060826, | |
| "loss": 2.4262, | |
| "step": 806 | |
| }, | |
| { | |
| "epoch": 0.46, | |
| "learning_rate": 0.0005946666936387637, | |
| "loss": 2.4388, | |
| "step": 807 | |
| }, | |
| { | |
| "epoch": 0.46, | |
| "learning_rate": 0.0005937629690225188, | |
| "loss": 2.4265, | |
| "step": 808 | |
| }, | |
| { | |
| "epoch": 0.46, | |
| "learning_rate": 0.0005928589268183856, | |
| "loss": 2.3676, | |
| "step": 809 | |
| }, | |
| { | |
| "epoch": 0.46, | |
| "learning_rate": 0.0005919545700884778, | |
| "loss": 2.3852, | |
| "step": 810 | |
| }, | |
| { | |
| "epoch": 0.46, | |
| "learning_rate": 0.0005910499018959741, | |
| "loss": 2.3914, | |
| "step": 811 | |
| }, | |
| { | |
| "epoch": 0.46, | |
| "learning_rate": 0.0005901449253051084, | |
| "loss": 2.4019, | |
| "step": 812 | |
| }, | |
| { | |
| "epoch": 0.46, | |
| "learning_rate": 0.0005892396433811589, | |
| "loss": 2.4193, | |
| "step": 813 | |
| }, | |
| { | |
| "epoch": 0.46, | |
| "learning_rate": 0.0005883340591904385, | |
| "loss": 2.343, | |
| "step": 814 | |
| }, | |
| { | |
| "epoch": 0.46, | |
| "learning_rate": 0.0005874281758002835, | |
| "loss": 2.4708, | |
| "step": 815 | |
| }, | |
| { | |
| "epoch": 0.46, | |
| "learning_rate": 0.0005865219962790438, | |
| "loss": 2.4664, | |
| "step": 816 | |
| }, | |
| { | |
| "epoch": 0.46, | |
| "learning_rate": 0.0005856155236960723, | |
| "loss": 2.3792, | |
| "step": 817 | |
| }, | |
| { | |
| "epoch": 0.46, | |
| "learning_rate": 0.0005847087611217145, | |
| "loss": 2.4771, | |
| "step": 818 | |
| }, | |
| { | |
| "epoch": 0.47, | |
| "learning_rate": 0.0005838017116272981, | |
| "loss": 2.3939, | |
| "step": 819 | |
| }, | |
| { | |
| "epoch": 0.47, | |
| "learning_rate": 0.0005828943782851228, | |
| "loss": 2.4681, | |
| "step": 820 | |
| }, | |
| { | |
| "epoch": 0.47, | |
| "learning_rate": 0.0005819867641684497, | |
| "loss": 2.5123, | |
| "step": 821 | |
| }, | |
| { | |
| "epoch": 0.47, | |
| "learning_rate": 0.0005810788723514908, | |
| "loss": 2.478, | |
| "step": 822 | |
| }, | |
| { | |
| "epoch": 0.47, | |
| "learning_rate": 0.0005801707059093988, | |
| "loss": 2.4284, | |
| "step": 823 | |
| }, | |
| { | |
| "epoch": 0.47, | |
| "learning_rate": 0.0005792622679182565, | |
| "loss": 2.3805, | |
| "step": 824 | |
| }, | |
| { | |
| "epoch": 0.47, | |
| "learning_rate": 0.0005783535614550666, | |
| "loss": 2.4595, | |
| "step": 825 | |
| }, | |
| { | |
| "epoch": 0.47, | |
| "learning_rate": 0.000577444589597741, | |
| "loss": 2.4835, | |
| "step": 826 | |
| }, | |
| { | |
| "epoch": 0.47, | |
| "learning_rate": 0.0005765353554250907, | |
| "loss": 2.4444, | |
| "step": 827 | |
| }, | |
| { | |
| "epoch": 0.47, | |
| "learning_rate": 0.0005756258620168151, | |
| "loss": 2.3644, | |
| "step": 828 | |
| }, | |
| { | |
| "epoch": 0.47, | |
| "learning_rate": 0.0005747161124534918, | |
| "loss": 2.3951, | |
| "step": 829 | |
| }, | |
| { | |
| "epoch": 0.47, | |
| "learning_rate": 0.0005738061098165657, | |
| "loss": 2.3674, | |
| "step": 830 | |
| }, | |
| { | |
| "epoch": 0.47, | |
| "learning_rate": 0.0005728958571883392, | |
| "loss": 2.3816, | |
| "step": 831 | |
| }, | |
| { | |
| "epoch": 0.47, | |
| "learning_rate": 0.0005719853576519614, | |
| "loss": 2.4328, | |
| "step": 832 | |
| }, | |
| { | |
| "epoch": 0.47, | |
| "learning_rate": 0.0005710746142914177, | |
| "loss": 2.4554, | |
| "step": 833 | |
| }, | |
| { | |
| "epoch": 0.47, | |
| "learning_rate": 0.0005701636301915191, | |
| "loss": 2.4233, | |
| "step": 834 | |
| }, | |
| { | |
| "epoch": 0.47, | |
| "learning_rate": 0.0005692524084378926, | |
| "loss": 2.3548, | |
| "step": 835 | |
| }, | |
| { | |
| "epoch": 0.47, | |
| "learning_rate": 0.0005683409521169697, | |
| "loss": 2.3324, | |
| "step": 836 | |
| }, | |
| { | |
| "epoch": 0.48, | |
| "learning_rate": 0.0005674292643159764, | |
| "loss": 2.4403, | |
| "step": 837 | |
| }, | |
| { | |
| "epoch": 0.48, | |
| "learning_rate": 0.000566517348122923, | |
| "loss": 2.4564, | |
| "step": 838 | |
| }, | |
| { | |
| "epoch": 0.48, | |
| "learning_rate": 0.0005656052066265931, | |
| "loss": 2.3844, | |
| "step": 839 | |
| }, | |
| { | |
| "epoch": 0.48, | |
| "learning_rate": 0.0005646928429165339, | |
| "loss": 2.3731, | |
| "step": 840 | |
| }, | |
| { | |
| "epoch": 0.48, | |
| "learning_rate": 0.0005637802600830449, | |
| "loss": 2.4234, | |
| "step": 841 | |
| }, | |
| { | |
| "epoch": 0.48, | |
| "learning_rate": 0.0005628674612171679, | |
| "loss": 2.3916, | |
| "step": 842 | |
| }, | |
| { | |
| "epoch": 0.48, | |
| "learning_rate": 0.0005619544494106761, | |
| "loss": 2.3402, | |
| "step": 843 | |
| }, | |
| { | |
| "epoch": 0.48, | |
| "learning_rate": 0.0005610412277560646, | |
| "loss": 2.5129, | |
| "step": 844 | |
| }, | |
| { | |
| "epoch": 0.48, | |
| "learning_rate": 0.000560127799346539, | |
| "loss": 2.3983, | |
| "step": 845 | |
| }, | |
| { | |
| "epoch": 0.48, | |
| "learning_rate": 0.0005592141672760049, | |
| "loss": 2.4103, | |
| "step": 846 | |
| }, | |
| { | |
| "epoch": 0.48, | |
| "learning_rate": 0.0005583003346390582, | |
| "loss": 2.3486, | |
| "step": 847 | |
| }, | |
| { | |
| "epoch": 0.48, | |
| "learning_rate": 0.0005573863045309738, | |
| "loss": 2.4649, | |
| "step": 848 | |
| }, | |
| { | |
| "epoch": 0.48, | |
| "learning_rate": 0.0005564720800476958, | |
| "loss": 2.4571, | |
| "step": 849 | |
| }, | |
| { | |
| "epoch": 0.48, | |
| "learning_rate": 0.0005555576642858263, | |
| "loss": 2.3613, | |
| "step": 850 | |
| }, | |
| { | |
| "epoch": 0.48, | |
| "learning_rate": 0.0005546430603426154, | |
| "loss": 2.4394, | |
| "step": 851 | |
| }, | |
| { | |
| "epoch": 0.48, | |
| "learning_rate": 0.0005537282713159507, | |
| "loss": 2.4288, | |
| "step": 852 | |
| }, | |
| { | |
| "epoch": 0.48, | |
| "learning_rate": 0.0005528133003043468, | |
| "loss": 2.4392, | |
| "step": 853 | |
| }, | |
| { | |
| "epoch": 0.49, | |
| "learning_rate": 0.0005518981504069345, | |
| "loss": 2.4756, | |
| "step": 854 | |
| }, | |
| { | |
| "epoch": 0.49, | |
| "learning_rate": 0.0005509828247234505, | |
| "loss": 2.4553, | |
| "step": 855 | |
| }, | |
| { | |
| "epoch": 0.49, | |
| "learning_rate": 0.0005500673263542272, | |
| "loss": 2.4203, | |
| "step": 856 | |
| }, | |
| { | |
| "epoch": 0.49, | |
| "learning_rate": 0.0005491516584001814, | |
| "loss": 2.3956, | |
| "step": 857 | |
| }, | |
| { | |
| "epoch": 0.49, | |
| "learning_rate": 0.0005482358239628047, | |
| "loss": 2.3655, | |
| "step": 858 | |
| }, | |
| { | |
| "epoch": 0.49, | |
| "learning_rate": 0.0005473198261441526, | |
| "loss": 2.4502, | |
| "step": 859 | |
| }, | |
| { | |
| "epoch": 0.49, | |
| "learning_rate": 0.0005464036680468339, | |
| "loss": 2.346, | |
| "step": 860 | |
| }, | |
| { | |
| "epoch": 0.49, | |
| "learning_rate": 0.0005454873527740002, | |
| "loss": 2.4168, | |
| "step": 861 | |
| }, | |
| { | |
| "epoch": 0.49, | |
| "learning_rate": 0.0005445708834293356, | |
| "loss": 2.4072, | |
| "step": 862 | |
| }, | |
| { | |
| "epoch": 0.49, | |
| "learning_rate": 0.0005436542631170459, | |
| "loss": 2.4012, | |
| "step": 863 | |
| }, | |
| { | |
| "epoch": 0.49, | |
| "learning_rate": 0.0005427374949418486, | |
| "loss": 2.4058, | |
| "step": 864 | |
| }, | |
| { | |
| "epoch": 0.49, | |
| "learning_rate": 0.0005418205820089616, | |
| "loss": 2.4099, | |
| "step": 865 | |
| }, | |
| { | |
| "epoch": 0.49, | |
| "learning_rate": 0.0005409035274240935, | |
| "loss": 2.4391, | |
| "step": 866 | |
| }, | |
| { | |
| "epoch": 0.49, | |
| "learning_rate": 0.0005399863342934323, | |
| "loss": 2.412, | |
| "step": 867 | |
| }, | |
| { | |
| "epoch": 0.49, | |
| "learning_rate": 0.0005390690057236359, | |
| "loss": 2.3586, | |
| "step": 868 | |
| }, | |
| { | |
| "epoch": 0.49, | |
| "learning_rate": 0.00053815154482182, | |
| "loss": 2.4768, | |
| "step": 869 | |
| }, | |
| { | |
| "epoch": 0.49, | |
| "learning_rate": 0.0005372339546955492, | |
| "loss": 2.3788, | |
| "step": 870 | |
| }, | |
| { | |
| "epoch": 0.49, | |
| "learning_rate": 0.000536316238452826, | |
| "loss": 2.263, | |
| "step": 871 | |
| }, | |
| { | |
| "epoch": 0.5, | |
| "learning_rate": 0.0005353983992020794, | |
| "loss": 2.4776, | |
| "step": 872 | |
| }, | |
| { | |
| "epoch": 0.5, | |
| "learning_rate": 0.0005344804400521554, | |
| "loss": 2.4101, | |
| "step": 873 | |
| }, | |
| { | |
| "epoch": 0.5, | |
| "learning_rate": 0.0005335623641123062, | |
| "loss": 2.3999, | |
| "step": 874 | |
| }, | |
| { | |
| "epoch": 0.5, | |
| "learning_rate": 0.0005326441744921795, | |
| "loss": 2.4281, | |
| "step": 875 | |
| }, | |
| { | |
| "epoch": 0.5, | |
| "learning_rate": 0.000531725874301808, | |
| "loss": 2.443, | |
| "step": 876 | |
| }, | |
| { | |
| "epoch": 0.5, | |
| "learning_rate": 0.0005308074666515989, | |
| "loss": 2.3822, | |
| "step": 877 | |
| }, | |
| { | |
| "epoch": 0.5, | |
| "learning_rate": 0.0005298889546523233, | |
| "loss": 2.4589, | |
| "step": 878 | |
| }, | |
| { | |
| "epoch": 0.5, | |
| "learning_rate": 0.0005289703414151062, | |
| "loss": 2.4423, | |
| "step": 879 | |
| }, | |
| { | |
| "epoch": 0.5, | |
| "learning_rate": 0.000528051630051415, | |
| "loss": 2.4234, | |
| "step": 880 | |
| }, | |
| { | |
| "epoch": 0.5, | |
| "learning_rate": 0.0005271328236730496, | |
| "loss": 2.4683, | |
| "step": 881 | |
| }, | |
| { | |
| "epoch": 0.5, | |
| "learning_rate": 0.0005262139253921318, | |
| "loss": 2.3599, | |
| "step": 882 | |
| }, | |
| { | |
| "epoch": 0.5, | |
| "learning_rate": 0.0005252949383210948, | |
| "loss": 2.4579, | |
| "step": 883 | |
| }, | |
| { | |
| "epoch": 0.5, | |
| "learning_rate": 0.0005243758655726721, | |
| "loss": 2.3679, | |
| "step": 884 | |
| }, | |
| { | |
| "epoch": 0.5, | |
| "learning_rate": 0.0005234567102598881, | |
| "loss": 2.3544, | |
| "step": 885 | |
| }, | |
| { | |
| "epoch": 0.5, | |
| "learning_rate": 0.0005225374754960462, | |
| "loss": 2.3712, | |
| "step": 886 | |
| }, | |
| { | |
| "epoch": 0.5, | |
| "learning_rate": 0.0005216181643947191, | |
| "loss": 2.4384, | |
| "step": 887 | |
| }, | |
| { | |
| "epoch": 0.5, | |
| "learning_rate": 0.0005206987800697383, | |
| "loss": 2.3886, | |
| "step": 888 | |
| }, | |
| { | |
| "epoch": 0.51, | |
| "learning_rate": 0.0005197793256351832, | |
| "loss": 2.4095, | |
| "step": 889 | |
| }, | |
| { | |
| "epoch": 0.51, | |
| "learning_rate": 0.0005188598042053704, | |
| "loss": 2.404, | |
| "step": 890 | |
| }, | |
| { | |
| "epoch": 0.51, | |
| "learning_rate": 0.0005179402188948438, | |
| "loss": 2.4051, | |
| "step": 891 | |
| }, | |
| { | |
| "epoch": 0.51, | |
| "learning_rate": 0.0005170205728183636, | |
| "loss": 2.4318, | |
| "step": 892 | |
| }, | |
| { | |
| "epoch": 0.51, | |
| "learning_rate": 0.0005161008690908956, | |
| "loss": 2.3231, | |
| "step": 893 | |
| }, | |
| { | |
| "epoch": 0.51, | |
| "learning_rate": 0.0005151811108276011, | |
| "loss": 2.4394, | |
| "step": 894 | |
| }, | |
| { | |
| "epoch": 0.51, | |
| "learning_rate": 0.000514261301143826, | |
| "loss": 2.3758, | |
| "step": 895 | |
| }, | |
| { | |
| "epoch": 0.51, | |
| "learning_rate": 0.0005133414431550905, | |
| "loss": 2.3559, | |
| "step": 896 | |
| }, | |
| { | |
| "epoch": 0.51, | |
| "learning_rate": 0.0005124215399770782, | |
| "loss": 2.4378, | |
| "step": 897 | |
| }, | |
| { | |
| "epoch": 0.51, | |
| "learning_rate": 0.0005115015947256259, | |
| "loss": 2.3401, | |
| "step": 898 | |
| }, | |
| { | |
| "epoch": 0.51, | |
| "learning_rate": 0.0005105816105167129, | |
| "loss": 2.4664, | |
| "step": 899 | |
| }, | |
| { | |
| "epoch": 0.51, | |
| "learning_rate": 0.0005096615904664505, | |
| "loss": 2.308, | |
| "step": 900 | |
| }, | |
| { | |
| "epoch": 0.51, | |
| "learning_rate": 0.0005087415376910712, | |
| "loss": 2.316, | |
| "step": 901 | |
| }, | |
| { | |
| "epoch": 0.51, | |
| "learning_rate": 0.0005078214553069186, | |
| "loss": 2.4194, | |
| "step": 902 | |
| }, | |
| { | |
| "epoch": 0.51, | |
| "learning_rate": 0.0005069013464304365, | |
| "loss": 2.3844, | |
| "step": 903 | |
| }, | |
| { | |
| "epoch": 0.51, | |
| "learning_rate": 0.0005059812141781584, | |
| "loss": 2.4863, | |
| "step": 904 | |
| }, | |
| { | |
| "epoch": 0.51, | |
| "learning_rate": 0.0005050610616666968, | |
| "loss": 2.3552, | |
| "step": 905 | |
| }, | |
| { | |
| "epoch": 0.51, | |
| "learning_rate": 0.0005041408920127332, | |
| "loss": 2.4535, | |
| "step": 906 | |
| }, | |
| { | |
| "epoch": 0.52, | |
| "learning_rate": 0.0005032207083330071, | |
| "loss": 2.3841, | |
| "step": 907 | |
| }, | |
| { | |
| "epoch": 0.52, | |
| "learning_rate": 0.0005023005137443051, | |
| "loss": 2.4558, | |
| "step": 908 | |
| }, | |
| { | |
| "epoch": 0.52, | |
| "learning_rate": 0.0005013803113634514, | |
| "loss": 2.4689, | |
| "step": 909 | |
| }, | |
| { | |
| "epoch": 0.52, | |
| "learning_rate": 0.0005004601043072958, | |
| "loss": 2.4366, | |
| "step": 910 | |
| }, | |
| { | |
| "epoch": 0.52, | |
| "learning_rate": 0.0004995398956927044, | |
| "loss": 2.3713, | |
| "step": 911 | |
| }, | |
| { | |
| "epoch": 0.52, | |
| "learning_rate": 0.0004986196886365487, | |
| "loss": 2.4485, | |
| "step": 912 | |
| }, | |
| { | |
| "epoch": 0.52, | |
| "learning_rate": 0.0004976994862556949, | |
| "loss": 2.41, | |
| "step": 913 | |
| }, | |
| { | |
| "epoch": 0.52, | |
| "learning_rate": 0.0004967792916669929, | |
| "loss": 2.4275, | |
| "step": 914 | |
| }, | |
| { | |
| "epoch": 0.52, | |
| "learning_rate": 0.0004958591079872668, | |
| "loss": 2.4909, | |
| "step": 915 | |
| }, | |
| { | |
| "epoch": 0.52, | |
| "learning_rate": 0.0004949389383333032, | |
| "loss": 2.4397, | |
| "step": 916 | |
| }, | |
| { | |
| "epoch": 0.52, | |
| "learning_rate": 0.0004940187858218417, | |
| "loss": 2.495, | |
| "step": 917 | |
| }, | |
| { | |
| "epoch": 0.52, | |
| "learning_rate": 0.0004930986535695636, | |
| "loss": 2.3796, | |
| "step": 918 | |
| }, | |
| { | |
| "epoch": 0.52, | |
| "learning_rate": 0.0004921785446930815, | |
| "loss": 2.4329, | |
| "step": 919 | |
| }, | |
| { | |
| "epoch": 0.52, | |
| "learning_rate": 0.0004912584623089289, | |
| "loss": 2.3956, | |
| "step": 920 | |
| }, | |
| { | |
| "epoch": 0.52, | |
| "learning_rate": 0.0004903384095335496, | |
| "loss": 2.4484, | |
| "step": 921 | |
| }, | |
| { | |
| "epoch": 0.52, | |
| "learning_rate": 0.0004894183894832872, | |
| "loss": 2.3731, | |
| "step": 922 | |
| }, | |
| { | |
| "epoch": 0.52, | |
| "learning_rate": 0.0004884984052743741, | |
| "loss": 2.3831, | |
| "step": 923 | |
| }, | |
| { | |
| "epoch": 0.52, | |
| "learning_rate": 0.0004875784600229219, | |
| "loss": 2.4037, | |
| "step": 924 | |
| }, | |
| { | |
| "epoch": 0.53, | |
| "learning_rate": 0.00048665855684490954, | |
| "loss": 2.4096, | |
| "step": 925 | |
| }, | |
| { | |
| "epoch": 0.53, | |
| "learning_rate": 0.000485738698856174, | |
| "loss": 2.3367, | |
| "step": 926 | |
| }, | |
| { | |
| "epoch": 0.53, | |
| "learning_rate": 0.000484818889172399, | |
| "loss": 2.4187, | |
| "step": 927 | |
| }, | |
| { | |
| "epoch": 0.53, | |
| "learning_rate": 0.0004838991309091045, | |
| "loss": 2.4407, | |
| "step": 928 | |
| }, | |
| { | |
| "epoch": 0.53, | |
| "learning_rate": 0.0004829794271816365, | |
| "loss": 2.4333, | |
| "step": 929 | |
| }, | |
| { | |
| "epoch": 0.53, | |
| "learning_rate": 0.0004820597811051563, | |
| "loss": 2.4117, | |
| "step": 930 | |
| }, | |
| { | |
| "epoch": 0.53, | |
| "learning_rate": 0.00048114019579462977, | |
| "loss": 2.4389, | |
| "step": 931 | |
| }, | |
| { | |
| "epoch": 0.53, | |
| "learning_rate": 0.00048022067436481703, | |
| "loss": 2.4473, | |
| "step": 932 | |
| }, | |
| { | |
| "epoch": 0.53, | |
| "learning_rate": 0.0004793012199302619, | |
| "loss": 2.3529, | |
| "step": 933 | |
| }, | |
| { | |
| "epoch": 0.53, | |
| "learning_rate": 0.00047838183560528115, | |
| "loss": 2.4897, | |
| "step": 934 | |
| }, | |
| { | |
| "epoch": 0.53, | |
| "learning_rate": 0.0004774625245039541, | |
| "loss": 2.3568, | |
| "step": 935 | |
| }, | |
| { | |
| "epoch": 0.53, | |
| "learning_rate": 0.0004765432897401121, | |
| "loss": 2.4079, | |
| "step": 936 | |
| }, | |
| { | |
| "epoch": 0.53, | |
| "learning_rate": 0.00047562413442732784, | |
| "loss": 2.4042, | |
| "step": 937 | |
| }, | |
| { | |
| "epoch": 0.53, | |
| "learning_rate": 0.0004747050616789052, | |
| "loss": 2.3854, | |
| "step": 938 | |
| }, | |
| { | |
| "epoch": 0.53, | |
| "learning_rate": 0.00047378607460786814, | |
| "loss": 2.3854, | |
| "step": 939 | |
| }, | |
| { | |
| "epoch": 0.53, | |
| "learning_rate": 0.00047286717632695035, | |
| "loss": 2.4135, | |
| "step": 940 | |
| }, | |
| { | |
| "epoch": 0.53, | |
| "learning_rate": 0.00047194836994858503, | |
| "loss": 2.4586, | |
| "step": 941 | |
| }, | |
| { | |
| "epoch": 0.54, | |
| "learning_rate": 0.0004710296585848938, | |
| "loss": 2.4829, | |
| "step": 942 | |
| }, | |
| { | |
| "epoch": 0.54, | |
| "learning_rate": 0.0004701110453476767, | |
| "loss": 2.4407, | |
| "step": 943 | |
| }, | |
| { | |
| "epoch": 0.54, | |
| "learning_rate": 0.00046919253334840125, | |
| "loss": 2.3109, | |
| "step": 944 | |
| }, | |
| { | |
| "epoch": 0.54, | |
| "learning_rate": 0.00046827412569819217, | |
| "loss": 2.3722, | |
| "step": 945 | |
| }, | |
| { | |
| "epoch": 0.54, | |
| "learning_rate": 0.00046735582550782066, | |
| "loss": 2.3676, | |
| "step": 946 | |
| }, | |
| { | |
| "epoch": 0.54, | |
| "learning_rate": 0.0004664376358876939, | |
| "loss": 2.4991, | |
| "step": 947 | |
| }, | |
| { | |
| "epoch": 0.54, | |
| "learning_rate": 0.0004655195599478448, | |
| "loss": 2.3987, | |
| "step": 948 | |
| }, | |
| { | |
| "epoch": 0.54, | |
| "learning_rate": 0.0004646016007979207, | |
| "loss": 2.3807, | |
| "step": 949 | |
| }, | |
| { | |
| "epoch": 0.54, | |
| "learning_rate": 0.0004636837615471741, | |
| "loss": 2.3651, | |
| "step": 950 | |
| }, | |
| { | |
| "epoch": 0.54, | |
| "learning_rate": 0.0004627660453044508, | |
| "loss": 2.4073, | |
| "step": 951 | |
| }, | |
| { | |
| "epoch": 0.54, | |
| "learning_rate": 0.00046184845517818014, | |
| "loss": 2.4485, | |
| "step": 952 | |
| }, | |
| { | |
| "epoch": 0.54, | |
| "learning_rate": 0.00046093099427636433, | |
| "loss": 2.4144, | |
| "step": 953 | |
| }, | |
| { | |
| "epoch": 0.54, | |
| "learning_rate": 0.0004600136657065677, | |
| "loss": 2.3495, | |
| "step": 954 | |
| }, | |
| { | |
| "epoch": 0.54, | |
| "learning_rate": 0.0004590964725759066, | |
| "loss": 2.4432, | |
| "step": 955 | |
| }, | |
| { | |
| "epoch": 0.54, | |
| "learning_rate": 0.00045817941799103845, | |
| "loss": 2.3367, | |
| "step": 956 | |
| }, | |
| { | |
| "epoch": 0.54, | |
| "learning_rate": 0.0004572625050581516, | |
| "loss": 2.4404, | |
| "step": 957 | |
| }, | |
| { | |
| "epoch": 0.54, | |
| "learning_rate": 0.0004563457368829542, | |
| "loss": 2.4057, | |
| "step": 958 | |
| }, | |
| { | |
| "epoch": 0.54, | |
| "learning_rate": 0.00045542911657066467, | |
| "loss": 2.4284, | |
| "step": 959 | |
| }, | |
| { | |
| "epoch": 0.55, | |
| "learning_rate": 0.0004545126472260001, | |
| "loss": 2.3815, | |
| "step": 960 | |
| }, | |
| { | |
| "epoch": 0.55, | |
| "learning_rate": 0.0004535963319531663, | |
| "loss": 2.3614, | |
| "step": 961 | |
| }, | |
| { | |
| "epoch": 0.55, | |
| "learning_rate": 0.0004526801738558476, | |
| "loss": 2.4519, | |
| "step": 962 | |
| }, | |
| { | |
| "epoch": 0.55, | |
| "learning_rate": 0.0004517641760371955, | |
| "loss": 2.3831, | |
| "step": 963 | |
| }, | |
| { | |
| "epoch": 0.55, | |
| "learning_rate": 0.00045084834159981864, | |
| "loss": 2.346, | |
| "step": 964 | |
| }, | |
| { | |
| "epoch": 0.55, | |
| "learning_rate": 0.00044993267364577285, | |
| "loss": 2.4427, | |
| "step": 965 | |
| }, | |
| { | |
| "epoch": 0.55, | |
| "learning_rate": 0.0004490171752765494, | |
| "loss": 2.391, | |
| "step": 966 | |
| }, | |
| { | |
| "epoch": 0.55, | |
| "learning_rate": 0.00044810184959306545, | |
| "loss": 2.451, | |
| "step": 967 | |
| }, | |
| { | |
| "epoch": 0.55, | |
| "learning_rate": 0.00044718669969565317, | |
| "loss": 2.3838, | |
| "step": 968 | |
| }, | |
| { | |
| "epoch": 0.55, | |
| "learning_rate": 0.00044627172868404923, | |
| "loss": 2.4119, | |
| "step": 969 | |
| }, | |
| { | |
| "epoch": 0.55, | |
| "learning_rate": 0.00044535693965738467, | |
| "loss": 2.4274, | |
| "step": 970 | |
| }, | |
| { | |
| "epoch": 0.55, | |
| "learning_rate": 0.0004444423357141738, | |
| "loss": 2.3029, | |
| "step": 971 | |
| }, | |
| { | |
| "epoch": 0.55, | |
| "learning_rate": 0.0004435279199523043, | |
| "loss": 2.3777, | |
| "step": 972 | |
| }, | |
| { | |
| "epoch": 0.55, | |
| "learning_rate": 0.0004426136954690262, | |
| "loss": 2.3987, | |
| "step": 973 | |
| }, | |
| { | |
| "epoch": 0.55, | |
| "learning_rate": 0.00044169966536094184, | |
| "loss": 2.3581, | |
| "step": 974 | |
| }, | |
| { | |
| "epoch": 0.55, | |
| "learning_rate": 0.00044078583272399513, | |
| "loss": 2.3907, | |
| "step": 975 | |
| }, | |
| { | |
| "epoch": 0.55, | |
| "learning_rate": 0.00043987220065346114, | |
| "loss": 2.4248, | |
| "step": 976 | |
| }, | |
| { | |
| "epoch": 0.56, | |
| "learning_rate": 0.00043895877224393545, | |
| "loss": 2.3192, | |
| "step": 977 | |
| }, | |
| { | |
| "epoch": 0.56, | |
| "learning_rate": 0.000438045550589324, | |
| "loss": 2.358, | |
| "step": 978 | |
| }, | |
| { | |
| "epoch": 0.56, | |
| "learning_rate": 0.00043713253878283233, | |
| "loss": 2.391, | |
| "step": 979 | |
| }, | |
| { | |
| "epoch": 0.56, | |
| "learning_rate": 0.00043621973991695524, | |
| "loss": 2.4907, | |
| "step": 980 | |
| }, | |
| { | |
| "epoch": 0.56, | |
| "learning_rate": 0.0004353071570834662, | |
| "loss": 2.4146, | |
| "step": 981 | |
| }, | |
| { | |
| "epoch": 0.56, | |
| "learning_rate": 0.00043439479337340696, | |
| "loss": 2.3712, | |
| "step": 982 | |
| }, | |
| { | |
| "epoch": 0.56, | |
| "learning_rate": 0.0004334826518770771, | |
| "loss": 2.4449, | |
| "step": 983 | |
| }, | |
| { | |
| "epoch": 0.56, | |
| "learning_rate": 0.0004325707356840237, | |
| "loss": 2.3753, | |
| "step": 984 | |
| }, | |
| { | |
| "epoch": 0.56, | |
| "learning_rate": 0.0004316590478830304, | |
| "loss": 2.3247, | |
| "step": 985 | |
| }, | |
| { | |
| "epoch": 0.56, | |
| "learning_rate": 0.0004307475915621074, | |
| "loss": 2.3548, | |
| "step": 986 | |
| }, | |
| { | |
| "epoch": 0.56, | |
| "learning_rate": 0.0004298363698084809, | |
| "loss": 2.3644, | |
| "step": 987 | |
| }, | |
| { | |
| "epoch": 0.56, | |
| "learning_rate": 0.0004289253857085824, | |
| "loss": 2.4098, | |
| "step": 988 | |
| }, | |
| { | |
| "epoch": 0.56, | |
| "learning_rate": 0.00042801464234803864, | |
| "loss": 2.3517, | |
| "step": 989 | |
| }, | |
| { | |
| "epoch": 0.56, | |
| "learning_rate": 0.00042710414281166075, | |
| "loss": 2.439, | |
| "step": 990 | |
| }, | |
| { | |
| "epoch": 0.56, | |
| "learning_rate": 0.00042619389018343435, | |
| "loss": 2.4223, | |
| "step": 991 | |
| }, | |
| { | |
| "epoch": 0.56, | |
| "learning_rate": 0.0004252838875465083, | |
| "loss": 2.3211, | |
| "step": 992 | |
| }, | |
| { | |
| "epoch": 0.56, | |
| "learning_rate": 0.0004243741379831848, | |
| "loss": 2.401, | |
| "step": 993 | |
| }, | |
| { | |
| "epoch": 0.56, | |
| "learning_rate": 0.0004234646445749093, | |
| "loss": 2.4459, | |
| "step": 994 | |
| }, | |
| { | |
| "epoch": 0.57, | |
| "learning_rate": 0.000422555410402259, | |
| "loss": 2.2829, | |
| "step": 995 | |
| }, | |
| { | |
| "epoch": 0.57, | |
| "learning_rate": 0.0004216464385449335, | |
| "loss": 2.3647, | |
| "step": 996 | |
| }, | |
| { | |
| "epoch": 0.57, | |
| "learning_rate": 0.0004207377320817436, | |
| "loss": 2.4144, | |
| "step": 997 | |
| }, | |
| { | |
| "epoch": 0.57, | |
| "learning_rate": 0.00041982929409060135, | |
| "loss": 2.3446, | |
| "step": 998 | |
| }, | |
| { | |
| "epoch": 0.57, | |
| "learning_rate": 0.00041892112764850927, | |
| "loss": 2.3324, | |
| "step": 999 | |
| }, | |
| { | |
| "epoch": 0.57, | |
| "learning_rate": 0.00041801323583155037, | |
| "loss": 2.2845, | |
| "step": 1000 | |
| }, | |
| { | |
| "epoch": 0.57, | |
| "learning_rate": 0.00041710562171487725, | |
| "loss": 2.3375, | |
| "step": 1001 | |
| }, | |
| { | |
| "epoch": 0.57, | |
| "learning_rate": 0.000416198288372702, | |
| "loss": 2.3582, | |
| "step": 1002 | |
| }, | |
| { | |
| "epoch": 0.57, | |
| "learning_rate": 0.0004152912388782856, | |
| "loss": 2.2841, | |
| "step": 1003 | |
| }, | |
| { | |
| "epoch": 0.57, | |
| "learning_rate": 0.00041438447630392784, | |
| "loss": 2.3542, | |
| "step": 1004 | |
| }, | |
| { | |
| "epoch": 0.57, | |
| "learning_rate": 0.00041347800372095627, | |
| "loss": 2.4362, | |
| "step": 1005 | |
| }, | |
| { | |
| "epoch": 0.57, | |
| "learning_rate": 0.00041257182419971654, | |
| "loss": 2.4016, | |
| "step": 1006 | |
| }, | |
| { | |
| "epoch": 0.57, | |
| "learning_rate": 0.0004116659408095616, | |
| "loss": 2.4134, | |
| "step": 1007 | |
| }, | |
| { | |
| "epoch": 0.57, | |
| "learning_rate": 0.0004107603566188412, | |
| "loss": 2.3195, | |
| "step": 1008 | |
| }, | |
| { | |
| "epoch": 0.57, | |
| "learning_rate": 0.0004098550746948918, | |
| "loss": 2.3902, | |
| "step": 1009 | |
| }, | |
| { | |
| "epoch": 0.57, | |
| "learning_rate": 0.0004089500981040261, | |
| "loss": 2.4356, | |
| "step": 1010 | |
| }, | |
| { | |
| "epoch": 0.57, | |
| "learning_rate": 0.0004080454299115224, | |
| "loss": 2.4718, | |
| "step": 1011 | |
| }, | |
| { | |
| "epoch": 0.57, | |
| "learning_rate": 0.00040714107318161455, | |
| "loss": 2.3165, | |
| "step": 1012 | |
| }, | |
| { | |
| "epoch": 0.58, | |
| "learning_rate": 0.0004062370309774814, | |
| "loss": 2.3676, | |
| "step": 1013 | |
| }, | |
| { | |
| "epoch": 0.58, | |
| "learning_rate": 0.00040533330636123647, | |
| "loss": 2.334, | |
| "step": 1014 | |
| }, | |
| { | |
| "epoch": 0.58, | |
| "learning_rate": 0.00040442990239391766, | |
| "loss": 2.4078, | |
| "step": 1015 | |
| }, | |
| { | |
| "epoch": 0.58, | |
| "learning_rate": 0.00040352682213547626, | |
| "loss": 2.3833, | |
| "step": 1016 | |
| }, | |
| { | |
| "epoch": 0.58, | |
| "learning_rate": 0.0004026240686447681, | |
| "loss": 2.3571, | |
| "step": 1017 | |
| }, | |
| { | |
| "epoch": 0.58, | |
| "learning_rate": 0.0004017216449795415, | |
| "loss": 2.3641, | |
| "step": 1018 | |
| }, | |
| { | |
| "epoch": 0.58, | |
| "learning_rate": 0.00040081955419642764, | |
| "loss": 2.367, | |
| "step": 1019 | |
| }, | |
| { | |
| "epoch": 0.58, | |
| "learning_rate": 0.0003999177993509303, | |
| "loss": 2.3368, | |
| "step": 1020 | |
| }, | |
| { | |
| "epoch": 0.58, | |
| "learning_rate": 0.00039901638349741555, | |
| "loss": 2.4514, | |
| "step": 1021 | |
| }, | |
| { | |
| "epoch": 0.58, | |
| "learning_rate": 0.0003981153096891009, | |
| "loss": 2.3297, | |
| "step": 1022 | |
| }, | |
| { | |
| "epoch": 0.58, | |
| "learning_rate": 0.0003972145809780457, | |
| "loss": 2.3843, | |
| "step": 1023 | |
| }, | |
| { | |
| "epoch": 0.58, | |
| "learning_rate": 0.0003963142004151401, | |
| "loss": 2.4125, | |
| "step": 1024 | |
| }, | |
| { | |
| "epoch": 0.58, | |
| "learning_rate": 0.00039541417105009506, | |
| "loss": 2.3168, | |
| "step": 1025 | |
| }, | |
| { | |
| "epoch": 0.58, | |
| "learning_rate": 0.00039451449593143215, | |
| "loss": 2.4065, | |
| "step": 1026 | |
| }, | |
| { | |
| "epoch": 0.58, | |
| "learning_rate": 0.0003936151781064731, | |
| "loss": 2.3478, | |
| "step": 1027 | |
| }, | |
| { | |
| "epoch": 0.58, | |
| "learning_rate": 0.0003927162206213291, | |
| "loss": 2.3989, | |
| "step": 1028 | |
| }, | |
| { | |
| "epoch": 0.58, | |
| "learning_rate": 0.00039181762652089114, | |
| "loss": 2.4372, | |
| "step": 1029 | |
| }, | |
| { | |
| "epoch": 0.59, | |
| "learning_rate": 0.0003909193988488192, | |
| "loss": 2.3873, | |
| "step": 1030 | |
| }, | |
| { | |
| "epoch": 0.59, | |
| "learning_rate": 0.0003900215406475324, | |
| "loss": 2.4046, | |
| "step": 1031 | |
| }, | |
| { | |
| "epoch": 0.59, | |
| "learning_rate": 0.00038912405495819785, | |
| "loss": 2.3524, | |
| "step": 1032 | |
| }, | |
| { | |
| "epoch": 0.59, | |
| "learning_rate": 0.00038822694482072153, | |
| "loss": 2.3564, | |
| "step": 1033 | |
| }, | |
| { | |
| "epoch": 0.59, | |
| "learning_rate": 0.00038733021327373696, | |
| "loss": 2.369, | |
| "step": 1034 | |
| }, | |
| { | |
| "epoch": 0.59, | |
| "learning_rate": 0.0003864338633545956, | |
| "loss": 2.3578, | |
| "step": 1035 | |
| }, | |
| { | |
| "epoch": 0.59, | |
| "learning_rate": 0.000385537898099356, | |
| "loss": 2.3602, | |
| "step": 1036 | |
| }, | |
| { | |
| "epoch": 0.59, | |
| "learning_rate": 0.000384642320542774, | |
| "loss": 2.4313, | |
| "step": 1037 | |
| }, | |
| { | |
| "epoch": 0.59, | |
| "learning_rate": 0.0003837471337182923, | |
| "loss": 2.3291, | |
| "step": 1038 | |
| }, | |
| { | |
| "epoch": 0.59, | |
| "learning_rate": 0.0003828523406580299, | |
| "loss": 2.3555, | |
| "step": 1039 | |
| }, | |
| { | |
| "epoch": 0.59, | |
| "learning_rate": 0.00038195794439277225, | |
| "loss": 2.3163, | |
| "step": 1040 | |
| }, | |
| { | |
| "epoch": 0.59, | |
| "learning_rate": 0.00038106394795196086, | |
| "loss": 2.3385, | |
| "step": 1041 | |
| }, | |
| { | |
| "epoch": 0.59, | |
| "learning_rate": 0.00038017035436368254, | |
| "loss": 2.4187, | |
| "step": 1042 | |
| }, | |
| { | |
| "epoch": 0.59, | |
| "learning_rate": 0.00037927716665466047, | |
| "loss": 2.4404, | |
| "step": 1043 | |
| }, | |
| { | |
| "epoch": 0.59, | |
| "learning_rate": 0.00037838438785024216, | |
| "loss": 2.397, | |
| "step": 1044 | |
| }, | |
| { | |
| "epoch": 0.59, | |
| "learning_rate": 0.00037749202097439057, | |
| "loss": 2.3734, | |
| "step": 1045 | |
| }, | |
| { | |
| "epoch": 0.59, | |
| "learning_rate": 0.00037660006904967325, | |
| "loss": 2.4009, | |
| "step": 1046 | |
| }, | |
| { | |
| "epoch": 0.59, | |
| "learning_rate": 0.00037570853509725236, | |
| "loss": 2.4276, | |
| "step": 1047 | |
| }, | |
| { | |
| "epoch": 0.6, | |
| "learning_rate": 0.0003748174221368742, | |
| "loss": 2.3821, | |
| "step": 1048 | |
| }, | |
| { | |
| "epoch": 0.6, | |
| "learning_rate": 0.00037392673318685916, | |
| "loss": 2.321, | |
| "step": 1049 | |
| }, | |
| { | |
| "epoch": 0.6, | |
| "learning_rate": 0.00037303647126409154, | |
| "loss": 2.3743, | |
| "step": 1050 | |
| }, | |
| { | |
| "epoch": 0.6, | |
| "learning_rate": 0.00037214663938400913, | |
| "loss": 2.3692, | |
| "step": 1051 | |
| }, | |
| { | |
| "epoch": 0.6, | |
| "learning_rate": 0.0003712572405605932, | |
| "loss": 2.383, | |
| "step": 1052 | |
| }, | |
| { | |
| "epoch": 0.6, | |
| "learning_rate": 0.0003703682778063581, | |
| "loss": 2.3707, | |
| "step": 1053 | |
| }, | |
| { | |
| "epoch": 0.6, | |
| "learning_rate": 0.00036947975413234113, | |
| "loss": 2.3864, | |
| "step": 1054 | |
| }, | |
| { | |
| "epoch": 0.6, | |
| "learning_rate": 0.00036859167254809256, | |
| "loss": 2.3399, | |
| "step": 1055 | |
| }, | |
| { | |
| "epoch": 0.6, | |
| "learning_rate": 0.00036770403606166516, | |
| "loss": 2.469, | |
| "step": 1056 | |
| }, | |
| { | |
| "epoch": 0.6, | |
| "learning_rate": 0.00036681684767960387, | |
| "loss": 2.4053, | |
| "step": 1057 | |
| }, | |
| { | |
| "epoch": 0.6, | |
| "learning_rate": 0.00036593011040693613, | |
| "loss": 2.4008, | |
| "step": 1058 | |
| }, | |
| { | |
| "epoch": 0.6, | |
| "learning_rate": 0.00036504382724716135, | |
| "loss": 2.3744, | |
| "step": 1059 | |
| }, | |
| { | |
| "epoch": 0.6, | |
| "learning_rate": 0.00036415800120224057, | |
| "loss": 2.3632, | |
| "step": 1060 | |
| }, | |
| { | |
| "epoch": 0.6, | |
| "learning_rate": 0.0003632726352725869, | |
| "loss": 2.3286, | |
| "step": 1061 | |
| }, | |
| { | |
| "epoch": 0.6, | |
| "learning_rate": 0.0003623877324570548, | |
| "loss": 2.4072, | |
| "step": 1062 | |
| }, | |
| { | |
| "epoch": 0.6, | |
| "learning_rate": 0.00036150329575292996, | |
| "loss": 2.2996, | |
| "step": 1063 | |
| }, | |
| { | |
| "epoch": 0.6, | |
| "learning_rate": 0.0003606193281559194, | |
| "loss": 2.4821, | |
| "step": 1064 | |
| }, | |
| { | |
| "epoch": 0.61, | |
| "learning_rate": 0.00035973583266014133, | |
| "loss": 2.3883, | |
| "step": 1065 | |
| }, | |
| { | |
| "epoch": 0.61, | |
| "learning_rate": 0.0003588528122581146, | |
| "loss": 2.3579, | |
| "step": 1066 | |
| }, | |
| { | |
| "epoch": 0.61, | |
| "learning_rate": 0.0003579702699407492, | |
| "loss": 2.3023, | |
| "step": 1067 | |
| }, | |
| { | |
| "epoch": 0.61, | |
| "learning_rate": 0.0003570882086973355, | |
| "loss": 2.4152, | |
| "step": 1068 | |
| }, | |
| { | |
| "epoch": 0.61, | |
| "learning_rate": 0.0003562066315155345, | |
| "loss": 2.3282, | |
| "step": 1069 | |
| }, | |
| { | |
| "epoch": 0.61, | |
| "learning_rate": 0.0003553255413813676, | |
| "loss": 2.2976, | |
| "step": 1070 | |
| }, | |
| { | |
| "epoch": 0.61, | |
| "learning_rate": 0.0003544449412792069, | |
| "loss": 2.4261, | |
| "step": 1071 | |
| }, | |
| { | |
| "epoch": 0.61, | |
| "learning_rate": 0.000353564834191764, | |
| "loss": 2.4305, | |
| "step": 1072 | |
| }, | |
| { | |
| "epoch": 0.61, | |
| "learning_rate": 0.000352685223100081, | |
| "loss": 2.3566, | |
| "step": 1073 | |
| }, | |
| { | |
| "epoch": 0.61, | |
| "learning_rate": 0.00035180611098351987, | |
| "loss": 2.3931, | |
| "step": 1074 | |
| }, | |
| { | |
| "epoch": 0.61, | |
| "learning_rate": 0.00035092750081975265, | |
| "loss": 2.4228, | |
| "step": 1075 | |
| }, | |
| { | |
| "epoch": 0.61, | |
| "learning_rate": 0.00035004939558475105, | |
| "loss": 2.3614, | |
| "step": 1076 | |
| }, | |
| { | |
| "epoch": 0.61, | |
| "learning_rate": 0.0003491717982527765, | |
| "loss": 2.3207, | |
| "step": 1077 | |
| }, | |
| { | |
| "epoch": 0.61, | |
| "learning_rate": 0.0003482947117963702, | |
| "loss": 2.3927, | |
| "step": 1078 | |
| }, | |
| { | |
| "epoch": 0.61, | |
| "learning_rate": 0.0003474181391863428, | |
| "loss": 2.3219, | |
| "step": 1079 | |
| }, | |
| { | |
| "epoch": 0.61, | |
| "learning_rate": 0.0003465420833917647, | |
| "loss": 2.4248, | |
| "step": 1080 | |
| }, | |
| { | |
| "epoch": 0.61, | |
| "learning_rate": 0.00034566654737995555, | |
| "loss": 2.393, | |
| "step": 1081 | |
| }, | |
| { | |
| "epoch": 0.61, | |
| "learning_rate": 0.00034479153411647457, | |
| "loss": 2.3521, | |
| "step": 1082 | |
| }, | |
| { | |
| "epoch": 0.62, | |
| "learning_rate": 0.00034391704656511034, | |
| "loss": 2.3803, | |
| "step": 1083 | |
| }, | |
| { | |
| "epoch": 0.62, | |
| "learning_rate": 0.00034304308768787076, | |
| "loss": 2.3621, | |
| "step": 1084 | |
| }, | |
| { | |
| "epoch": 0.62, | |
| "learning_rate": 0.0003421696604449729, | |
| "loss": 2.365, | |
| "step": 1085 | |
| }, | |
| { | |
| "epoch": 0.62, | |
| "learning_rate": 0.00034129676779483346, | |
| "loss": 2.3525, | |
| "step": 1086 | |
| }, | |
| { | |
| "epoch": 0.62, | |
| "learning_rate": 0.000340424412694058, | |
| "loss": 2.3504, | |
| "step": 1087 | |
| }, | |
| { | |
| "epoch": 0.62, | |
| "learning_rate": 0.00033955259809743157, | |
| "loss": 2.3886, | |
| "step": 1088 | |
| }, | |
| { | |
| "epoch": 0.62, | |
| "learning_rate": 0.0003386813269579085, | |
| "loss": 2.4235, | |
| "step": 1089 | |
| }, | |
| { | |
| "epoch": 0.62, | |
| "learning_rate": 0.0003378106022266022, | |
| "loss": 2.3589, | |
| "step": 1090 | |
| }, | |
| { | |
| "epoch": 0.62, | |
| "learning_rate": 0.0003369404268527754, | |
| "loss": 2.3473, | |
| "step": 1091 | |
| }, | |
| { | |
| "epoch": 0.62, | |
| "learning_rate": 0.00033607080378383, | |
| "loss": 2.3863, | |
| "step": 1092 | |
| }, | |
| { | |
| "epoch": 0.62, | |
| "learning_rate": 0.00033520173596529725, | |
| "loss": 2.3328, | |
| "step": 1093 | |
| }, | |
| { | |
| "epoch": 0.62, | |
| "learning_rate": 0.0003343332263408277, | |
| "loss": 2.3727, | |
| "step": 1094 | |
| }, | |
| { | |
| "epoch": 0.62, | |
| "learning_rate": 0.0003334652778521813, | |
| "loss": 2.3475, | |
| "step": 1095 | |
| }, | |
| { | |
| "epoch": 0.62, | |
| "learning_rate": 0.0003325978934392171, | |
| "loss": 2.3711, | |
| "step": 1096 | |
| }, | |
| { | |
| "epoch": 0.62, | |
| "learning_rate": 0.0003317310760398839, | |
| "loss": 2.2961, | |
| "step": 1097 | |
| }, | |
| { | |
| "epoch": 0.62, | |
| "learning_rate": 0.00033086482859020957, | |
| "loss": 2.3632, | |
| "step": 1098 | |
| }, | |
| { | |
| "epoch": 0.62, | |
| "learning_rate": 0.00032999915402429173, | |
| "loss": 2.3933, | |
| "step": 1099 | |
| }, | |
| { | |
| "epoch": 0.62, | |
| "learning_rate": 0.0003291340552742875, | |
| "loss": 2.3845, | |
| "step": 1100 | |
| }, | |
| { | |
| "epoch": 0.63, | |
| "learning_rate": 0.0003282695352704036, | |
| "loss": 2.3837, | |
| "step": 1101 | |
| }, | |
| { | |
| "epoch": 0.63, | |
| "learning_rate": 0.00032740559694088655, | |
| "loss": 2.3652, | |
| "step": 1102 | |
| }, | |
| { | |
| "epoch": 0.63, | |
| "learning_rate": 0.0003265422432120125, | |
| "loss": 2.5405, | |
| "step": 1103 | |
| }, | |
| { | |
| "epoch": 0.63, | |
| "learning_rate": 0.0003256794770080778, | |
| "loss": 2.3815, | |
| "step": 1104 | |
| }, | |
| { | |
| "epoch": 0.63, | |
| "learning_rate": 0.00032481730125138823, | |
| "loss": 2.4564, | |
| "step": 1105 | |
| }, | |
| { | |
| "epoch": 0.63, | |
| "learning_rate": 0.0003239557188622503, | |
| "loss": 2.367, | |
| "step": 1106 | |
| }, | |
| { | |
| "epoch": 0.63, | |
| "learning_rate": 0.0003230947327589602, | |
| "loss": 2.4146, | |
| "step": 1107 | |
| }, | |
| { | |
| "epoch": 0.63, | |
| "learning_rate": 0.0003222343458577945, | |
| "loss": 2.3144, | |
| "step": 1108 | |
| }, | |
| { | |
| "epoch": 0.63, | |
| "learning_rate": 0.00032137456107300054, | |
| "loss": 2.3881, | |
| "step": 1109 | |
| }, | |
| { | |
| "epoch": 0.63, | |
| "learning_rate": 0.0003205153813167858, | |
| "loss": 2.3861, | |
| "step": 1110 | |
| }, | |
| { | |
| "epoch": 0.63, | |
| "learning_rate": 0.00031965680949930854, | |
| "loss": 2.292, | |
| "step": 1111 | |
| }, | |
| { | |
| "epoch": 0.63, | |
| "learning_rate": 0.00031879884852866803, | |
| "loss": 2.418, | |
| "step": 1112 | |
| }, | |
| { | |
| "epoch": 0.63, | |
| "learning_rate": 0.00031794150131089435, | |
| "loss": 2.4622, | |
| "step": 1113 | |
| }, | |
| { | |
| "epoch": 0.63, | |
| "learning_rate": 0.0003170847707499387, | |
| "loss": 2.3496, | |
| "step": 1114 | |
| }, | |
| { | |
| "epoch": 0.63, | |
| "learning_rate": 0.0003162286597476638, | |
| "loss": 2.3934, | |
| "step": 1115 | |
| }, | |
| { | |
| "epoch": 0.63, | |
| "learning_rate": 0.0003153731712038335, | |
| "loss": 2.3758, | |
| "step": 1116 | |
| }, | |
| { | |
| "epoch": 0.63, | |
| "learning_rate": 0.0003145183080161036, | |
| "loss": 2.3755, | |
| "step": 1117 | |
| }, | |
| { | |
| "epoch": 0.64, | |
| "learning_rate": 0.0003136640730800116, | |
| "loss": 2.339, | |
| "step": 1118 | |
| }, | |
| { | |
| "epoch": 0.64, | |
| "learning_rate": 0.000312810469288967, | |
| "loss": 2.3153, | |
| "step": 1119 | |
| }, | |
| { | |
| "epoch": 0.64, | |
| "learning_rate": 0.00031195749953424163, | |
| "loss": 2.4115, | |
| "step": 1120 | |
| }, | |
| { | |
| "epoch": 0.64, | |
| "learning_rate": 0.00031110516670495946, | |
| "loss": 2.3596, | |
| "step": 1121 | |
| }, | |
| { | |
| "epoch": 0.64, | |
| "learning_rate": 0.00031025347368808775, | |
| "loss": 2.3859, | |
| "step": 1122 | |
| }, | |
| { | |
| "epoch": 0.64, | |
| "learning_rate": 0.000309402423368426, | |
| "loss": 2.3421, | |
| "step": 1123 | |
| }, | |
| { | |
| "epoch": 0.64, | |
| "learning_rate": 0.00030855201862859706, | |
| "loss": 2.4224, | |
| "step": 1124 | |
| }, | |
| { | |
| "epoch": 0.64, | |
| "learning_rate": 0.0003077022623490371, | |
| "loss": 2.4121, | |
| "step": 1125 | |
| }, | |
| { | |
| "epoch": 0.64, | |
| "learning_rate": 0.0003068531574079857, | |
| "loss": 2.3836, | |
| "step": 1126 | |
| }, | |
| { | |
| "epoch": 0.64, | |
| "learning_rate": 0.00030600470668147654, | |
| "loss": 2.3476, | |
| "step": 1127 | |
| }, | |
| { | |
| "epoch": 0.64, | |
| "learning_rate": 0.00030515691304332725, | |
| "loss": 2.3695, | |
| "step": 1128 | |
| }, | |
| { | |
| "epoch": 0.64, | |
| "learning_rate": 0.0003043097793651299, | |
| "loss": 2.4237, | |
| "step": 1129 | |
| }, | |
| { | |
| "epoch": 0.64, | |
| "learning_rate": 0.00030346330851624095, | |
| "loss": 2.3247, | |
| "step": 1130 | |
| }, | |
| { | |
| "epoch": 0.64, | |
| "learning_rate": 0.00030261750336377203, | |
| "loss": 2.2692, | |
| "step": 1131 | |
| }, | |
| { | |
| "epoch": 0.64, | |
| "learning_rate": 0.0003017723667725798, | |
| "loss": 2.3942, | |
| "step": 1132 | |
| }, | |
| { | |
| "epoch": 0.64, | |
| "learning_rate": 0.00030092790160525653, | |
| "loss": 2.4238, | |
| "step": 1133 | |
| }, | |
| { | |
| "epoch": 0.64, | |
| "learning_rate": 0.00030008411072212027, | |
| "loss": 2.3459, | |
| "step": 1134 | |
| }, | |
| { | |
| "epoch": 0.64, | |
| "learning_rate": 0.000299240996981205, | |
| "loss": 2.3385, | |
| "step": 1135 | |
| }, | |
| { | |
| "epoch": 0.65, | |
| "learning_rate": 0.0002983985632382514, | |
| "loss": 2.4455, | |
| "step": 1136 | |
| }, | |
| { | |
| "epoch": 0.65, | |
| "learning_rate": 0.0002975568123466966, | |
| "loss": 2.3829, | |
| "step": 1137 | |
| }, | |
| { | |
| "epoch": 0.65, | |
| "learning_rate": 0.00029671574715766524, | |
| "loss": 2.2501, | |
| "step": 1138 | |
| }, | |
| { | |
| "epoch": 0.65, | |
| "learning_rate": 0.000295875370519959, | |
| "loss": 2.3536, | |
| "step": 1139 | |
| }, | |
| { | |
| "epoch": 0.65, | |
| "learning_rate": 0.0002950356852800475, | |
| "loss": 2.2893, | |
| "step": 1140 | |
| }, | |
| { | |
| "epoch": 0.65, | |
| "learning_rate": 0.00029419669428205853, | |
| "loss": 2.3131, | |
| "step": 1141 | |
| }, | |
| { | |
| "epoch": 0.65, | |
| "learning_rate": 0.00029335840036776836, | |
| "loss": 2.3992, | |
| "step": 1142 | |
| }, | |
| { | |
| "epoch": 0.65, | |
| "learning_rate": 0.000292520806376592, | |
| "loss": 2.4069, | |
| "step": 1143 | |
| }, | |
| { | |
| "epoch": 0.65, | |
| "learning_rate": 0.000291683915145574, | |
| "loss": 2.3701, | |
| "step": 1144 | |
| }, | |
| { | |
| "epoch": 0.65, | |
| "learning_rate": 0.0002908477295093784, | |
| "loss": 2.4106, | |
| "step": 1145 | |
| }, | |
| { | |
| "epoch": 0.65, | |
| "learning_rate": 0.0002900122523002792, | |
| "loss": 2.3763, | |
| "step": 1146 | |
| }, | |
| { | |
| "epoch": 0.65, | |
| "learning_rate": 0.000289177486348151, | |
| "loss": 2.385, | |
| "step": 1147 | |
| }, | |
| { | |
| "epoch": 0.65, | |
| "learning_rate": 0.00028834343448045936, | |
| "loss": 2.433, | |
| "step": 1148 | |
| }, | |
| { | |
| "epoch": 0.65, | |
| "learning_rate": 0.00028751009952225116, | |
| "loss": 2.338, | |
| "step": 1149 | |
| }, | |
| { | |
| "epoch": 0.65, | |
| "learning_rate": 0.0002866774842961445, | |
| "loss": 2.3869, | |
| "step": 1150 | |
| }, | |
| { | |
| "epoch": 0.65, | |
| "learning_rate": 0.00028584559162232054, | |
| "loss": 2.4237, | |
| "step": 1151 | |
| }, | |
| { | |
| "epoch": 0.65, | |
| "learning_rate": 0.000285014424318512, | |
| "loss": 2.4219, | |
| "step": 1152 | |
| }, | |
| { | |
| "epoch": 0.66, | |
| "learning_rate": 0.0002841839851999958, | |
| "loss": 2.3598, | |
| "step": 1153 | |
| }, | |
| { | |
| "epoch": 0.66, | |
| "learning_rate": 0.00028335427707958116, | |
| "loss": 2.3515, | |
| "step": 1154 | |
| }, | |
| { | |
| "epoch": 0.66, | |
| "learning_rate": 0.0002825253027676026, | |
| "loss": 2.37, | |
| "step": 1155 | |
| }, | |
| { | |
| "epoch": 0.66, | |
| "learning_rate": 0.00028169706507190806, | |
| "loss": 2.3696, | |
| "step": 1156 | |
| }, | |
| { | |
| "epoch": 0.66, | |
| "learning_rate": 0.00028086956679785127, | |
| "loss": 2.396, | |
| "step": 1157 | |
| }, | |
| { | |
| "epoch": 0.66, | |
| "learning_rate": 0.0002800428107482806, | |
| "loss": 2.4012, | |
| "step": 1158 | |
| }, | |
| { | |
| "epoch": 0.66, | |
| "learning_rate": 0.0002792167997235313, | |
| "loss": 2.3801, | |
| "step": 1159 | |
| }, | |
| { | |
| "epoch": 0.66, | |
| "learning_rate": 0.0002783915365214147, | |
| "loss": 2.3561, | |
| "step": 1160 | |
| }, | |
| { | |
| "epoch": 0.66, | |
| "learning_rate": 0.00027756702393720876, | |
| "loss": 2.3585, | |
| "step": 1161 | |
| }, | |
| { | |
| "epoch": 0.66, | |
| "learning_rate": 0.00027674326476364995, | |
| "loss": 2.394, | |
| "step": 1162 | |
| }, | |
| { | |
| "epoch": 0.66, | |
| "learning_rate": 0.00027592026179092176, | |
| "loss": 2.3956, | |
| "step": 1163 | |
| }, | |
| { | |
| "epoch": 0.66, | |
| "learning_rate": 0.0002750980178066472, | |
| "loss": 2.3338, | |
| "step": 1164 | |
| }, | |
| { | |
| "epoch": 0.66, | |
| "learning_rate": 0.00027427653559587776, | |
| "loss": 2.4041, | |
| "step": 1165 | |
| }, | |
| { | |
| "epoch": 0.66, | |
| "learning_rate": 0.00027345581794108555, | |
| "loss": 2.2587, | |
| "step": 1166 | |
| }, | |
| { | |
| "epoch": 0.66, | |
| "learning_rate": 0.000272635867622152, | |
| "loss": 2.3868, | |
| "step": 1167 | |
| }, | |
| { | |
| "epoch": 0.66, | |
| "learning_rate": 0.00027181668741636046, | |
| "loss": 2.354, | |
| "step": 1168 | |
| }, | |
| { | |
| "epoch": 0.66, | |
| "learning_rate": 0.000270998280098385, | |
| "loss": 2.2642, | |
| "step": 1169 | |
| }, | |
| { | |
| "epoch": 0.66, | |
| "learning_rate": 0.00027018064844028244, | |
| "loss": 2.4138, | |
| "step": 1170 | |
| }, | |
| { | |
| "epoch": 0.67, | |
| "learning_rate": 0.00026936379521148184, | |
| "loss": 2.3258, | |
| "step": 1171 | |
| }, | |
| { | |
| "epoch": 0.67, | |
| "learning_rate": 0.0002685477231787761, | |
| "loss": 2.4207, | |
| "step": 1172 | |
| }, | |
| { | |
| "epoch": 0.67, | |
| "learning_rate": 0.00026773243510631146, | |
| "loss": 2.3677, | |
| "step": 1173 | |
| }, | |
| { | |
| "epoch": 0.67, | |
| "learning_rate": 0.0002669179337555794, | |
| "loss": 2.3905, | |
| "step": 1174 | |
| }, | |
| { | |
| "epoch": 0.67, | |
| "learning_rate": 0.0002661042218854063, | |
| "loss": 2.3724, | |
| "step": 1175 | |
| }, | |
| { | |
| "epoch": 0.67, | |
| "learning_rate": 0.0002652913022519449, | |
| "loss": 2.3516, | |
| "step": 1176 | |
| }, | |
| { | |
| "epoch": 0.67, | |
| "learning_rate": 0.0002644791776086638, | |
| "loss": 2.3496, | |
| "step": 1177 | |
| }, | |
| { | |
| "epoch": 0.67, | |
| "learning_rate": 0.0002636678507063397, | |
| "loss": 2.307, | |
| "step": 1178 | |
| }, | |
| { | |
| "epoch": 0.67, | |
| "learning_rate": 0.0002628573242930463, | |
| "loss": 2.402, | |
| "step": 1179 | |
| }, | |
| { | |
| "epoch": 0.67, | |
| "learning_rate": 0.000262047601114147, | |
| "loss": 2.3445, | |
| "step": 1180 | |
| }, | |
| { | |
| "epoch": 0.67, | |
| "learning_rate": 0.0002612386839122834, | |
| "loss": 2.3665, | |
| "step": 1181 | |
| }, | |
| { | |
| "epoch": 0.67, | |
| "learning_rate": 0.0002604305754273684, | |
| "loss": 2.3387, | |
| "step": 1182 | |
| }, | |
| { | |
| "epoch": 0.67, | |
| "learning_rate": 0.00025962327839657435, | |
| "loss": 2.3526, | |
| "step": 1183 | |
| }, | |
| { | |
| "epoch": 0.67, | |
| "learning_rate": 0.00025881679555432623, | |
| "loss": 2.4376, | |
| "step": 1184 | |
| }, | |
| { | |
| "epoch": 0.67, | |
| "learning_rate": 0.0002580111296322904, | |
| "loss": 2.4212, | |
| "step": 1185 | |
| }, | |
| { | |
| "epoch": 0.67, | |
| "learning_rate": 0.0002572062833593669, | |
| "loss": 2.3326, | |
| "step": 1186 | |
| }, | |
| { | |
| "epoch": 0.67, | |
| "learning_rate": 0.00025640225946167895, | |
| "loss": 2.3383, | |
| "step": 1187 | |
| }, | |
| { | |
| "epoch": 0.67, | |
| "learning_rate": 0.00025559906066256467, | |
| "loss": 2.4154, | |
| "step": 1188 | |
| }, | |
| { | |
| "epoch": 0.68, | |
| "learning_rate": 0.000254796689682567, | |
| "loss": 2.3926, | |
| "step": 1189 | |
| }, | |
| { | |
| "epoch": 0.68, | |
| "learning_rate": 0.0002539951492394256, | |
| "loss": 2.3123, | |
| "step": 1190 | |
| }, | |
| { | |
| "epoch": 0.68, | |
| "learning_rate": 0.0002531944420480662, | |
| "loss": 2.2899, | |
| "step": 1191 | |
| }, | |
| { | |
| "epoch": 0.68, | |
| "learning_rate": 0.00025239457082059297, | |
| "loss": 2.3844, | |
| "step": 1192 | |
| }, | |
| { | |
| "epoch": 0.68, | |
| "learning_rate": 0.00025159553826627825, | |
| "loss": 2.3151, | |
| "step": 1193 | |
| }, | |
| { | |
| "epoch": 0.68, | |
| "learning_rate": 0.00025079734709155323, | |
| "loss": 2.4088, | |
| "step": 1194 | |
| }, | |
| { | |
| "epoch": 0.68, | |
| "learning_rate": 0.0002500000000000001, | |
| "loss": 2.4768, | |
| "step": 1195 | |
| }, | |
| { | |
| "epoch": 0.68, | |
| "learning_rate": 0.00024920349969234114, | |
| "loss": 2.3978, | |
| "step": 1196 | |
| }, | |
| { | |
| "epoch": 0.68, | |
| "learning_rate": 0.0002484078488664313, | |
| "loss": 2.3491, | |
| "step": 1197 | |
| }, | |
| { | |
| "epoch": 0.68, | |
| "learning_rate": 0.00024761305021724734, | |
| "loss": 2.3269, | |
| "step": 1198 | |
| }, | |
| { | |
| "epoch": 0.68, | |
| "learning_rate": 0.0002468191064368805, | |
| "loss": 2.3558, | |
| "step": 1199 | |
| }, | |
| { | |
| "epoch": 0.68, | |
| "learning_rate": 0.0002460260202145256, | |
| "loss": 2.3755, | |
| "step": 1200 | |
| }, | |
| { | |
| "epoch": 0.68, | |
| "learning_rate": 0.00024523379423647333, | |
| "loss": 2.3858, | |
| "step": 1201 | |
| }, | |
| { | |
| "epoch": 0.68, | |
| "learning_rate": 0.0002444424311861006, | |
| "loss": 2.3476, | |
| "step": 1202 | |
| }, | |
| { | |
| "epoch": 0.68, | |
| "learning_rate": 0.00024365193374386148, | |
| "loss": 2.3317, | |
| "step": 1203 | |
| }, | |
| { | |
| "epoch": 0.68, | |
| "learning_rate": 0.00024286230458727754, | |
| "loss": 2.369, | |
| "step": 1204 | |
| }, | |
| { | |
| "epoch": 0.68, | |
| "learning_rate": 0.00024207354639093026, | |
| "loss": 2.277, | |
| "step": 1205 | |
| }, | |
| { | |
| "epoch": 0.69, | |
| "learning_rate": 0.00024128566182645024, | |
| "loss": 2.3561, | |
| "step": 1206 | |
| }, | |
| { | |
| "epoch": 0.69, | |
| "learning_rate": 0.00024049865356250955, | |
| "loss": 2.3518, | |
| "step": 1207 | |
| }, | |
| { | |
| "epoch": 0.69, | |
| "learning_rate": 0.00023971252426481166, | |
| "loss": 2.3853, | |
| "step": 1208 | |
| }, | |
| { | |
| "epoch": 0.69, | |
| "learning_rate": 0.00023892727659608338, | |
| "loss": 2.3574, | |
| "step": 1209 | |
| }, | |
| { | |
| "epoch": 0.69, | |
| "learning_rate": 0.00023814291321606462, | |
| "loss": 2.2977, | |
| "step": 1210 | |
| }, | |
| { | |
| "epoch": 0.69, | |
| "learning_rate": 0.00023735943678150097, | |
| "loss": 2.4343, | |
| "step": 1211 | |
| }, | |
| { | |
| "epoch": 0.69, | |
| "learning_rate": 0.00023657684994613282, | |
| "loss": 2.3817, | |
| "step": 1212 | |
| }, | |
| { | |
| "epoch": 0.69, | |
| "learning_rate": 0.00023579515536068836, | |
| "loss": 2.4122, | |
| "step": 1213 | |
| }, | |
| { | |
| "epoch": 0.69, | |
| "learning_rate": 0.00023501435567287277, | |
| "loss": 2.3791, | |
| "step": 1214 | |
| }, | |
| { | |
| "epoch": 0.69, | |
| "learning_rate": 0.0002342344535273608, | |
| "loss": 2.3868, | |
| "step": 1215 | |
| }, | |
| { | |
| "epoch": 0.69, | |
| "learning_rate": 0.0002334554515657863, | |
| "loss": 2.4097, | |
| "step": 1216 | |
| }, | |
| { | |
| "epoch": 0.69, | |
| "learning_rate": 0.000232677352426735, | |
| "loss": 2.316, | |
| "step": 1217 | |
| }, | |
| { | |
| "epoch": 0.69, | |
| "learning_rate": 0.00023190015874573373, | |
| "loss": 2.3906, | |
| "step": 1218 | |
| }, | |
| { | |
| "epoch": 0.69, | |
| "learning_rate": 0.00023112387315524336, | |
| "loss": 2.3863, | |
| "step": 1219 | |
| }, | |
| { | |
| "epoch": 0.69, | |
| "learning_rate": 0.00023034849828464787, | |
| "loss": 2.3612, | |
| "step": 1220 | |
| }, | |
| { | |
| "epoch": 0.69, | |
| "learning_rate": 0.00022957403676024762, | |
| "loss": 2.3062, | |
| "step": 1221 | |
| }, | |
| { | |
| "epoch": 0.69, | |
| "learning_rate": 0.00022880049120524833, | |
| "loss": 2.3007, | |
| "step": 1222 | |
| }, | |
| { | |
| "epoch": 0.69, | |
| "learning_rate": 0.0002280278642397541, | |
| "loss": 2.2701, | |
| "step": 1223 | |
| }, | |
| { | |
| "epoch": 0.7, | |
| "learning_rate": 0.0002272561584807567, | |
| "loss": 2.4067, | |
| "step": 1224 | |
| }, | |
| { | |
| "epoch": 0.7, | |
| "learning_rate": 0.0002264853765421287, | |
| "loss": 2.339, | |
| "step": 1225 | |
| }, | |
| { | |
| "epoch": 0.7, | |
| "learning_rate": 0.0002257155210346124, | |
| "loss": 2.314, | |
| "step": 1226 | |
| }, | |
| { | |
| "epoch": 0.7, | |
| "learning_rate": 0.0002249465945658135, | |
| "loss": 2.3595, | |
| "step": 1227 | |
| }, | |
| { | |
| "epoch": 0.7, | |
| "learning_rate": 0.00022417859974018967, | |
| "loss": 2.2792, | |
| "step": 1228 | |
| }, | |
| { | |
| "epoch": 0.7, | |
| "learning_rate": 0.00022341153915904387, | |
| "loss": 2.3722, | |
| "step": 1229 | |
| }, | |
| { | |
| "epoch": 0.7, | |
| "learning_rate": 0.00022264541542051398, | |
| "loss": 2.3774, | |
| "step": 1230 | |
| }, | |
| { | |
| "epoch": 0.7, | |
| "learning_rate": 0.00022188023111956517, | |
| "loss": 2.2806, | |
| "step": 1231 | |
| }, | |
| { | |
| "epoch": 0.7, | |
| "learning_rate": 0.00022111598884798022, | |
| "loss": 2.3556, | |
| "step": 1232 | |
| }, | |
| { | |
| "epoch": 0.7, | |
| "learning_rate": 0.00022035269119435164, | |
| "loss": 2.3964, | |
| "step": 1233 | |
| }, | |
| { | |
| "epoch": 0.7, | |
| "learning_rate": 0.00021959034074407163, | |
| "loss": 2.3563, | |
| "step": 1234 | |
| }, | |
| { | |
| "epoch": 0.7, | |
| "learning_rate": 0.00021882894007932492, | |
| "loss": 2.3775, | |
| "step": 1235 | |
| }, | |
| { | |
| "epoch": 0.7, | |
| "learning_rate": 0.00021806849177907845, | |
| "loss": 2.3476, | |
| "step": 1236 | |
| }, | |
| { | |
| "epoch": 0.7, | |
| "learning_rate": 0.0002173089984190739, | |
| "loss": 2.3872, | |
| "step": 1237 | |
| }, | |
| { | |
| "epoch": 0.7, | |
| "learning_rate": 0.00021655046257181842, | |
| "loss": 2.2585, | |
| "step": 1238 | |
| }, | |
| { | |
| "epoch": 0.7, | |
| "learning_rate": 0.00021579288680657538, | |
| "loss": 2.3441, | |
| "step": 1239 | |
| }, | |
| { | |
| "epoch": 0.7, | |
| "learning_rate": 0.00021503627368935703, | |
| "loss": 2.3755, | |
| "step": 1240 | |
| }, | |
| { | |
| "epoch": 0.71, | |
| "learning_rate": 0.00021428062578291414, | |
| "loss": 2.3433, | |
| "step": 1241 | |
| }, | |
| { | |
| "epoch": 0.71, | |
| "learning_rate": 0.00021352594564672905, | |
| "loss": 2.3281, | |
| "step": 1242 | |
| }, | |
| { | |
| "epoch": 0.71, | |
| "learning_rate": 0.00021277223583700527, | |
| "loss": 2.3531, | |
| "step": 1243 | |
| }, | |
| { | |
| "epoch": 0.71, | |
| "learning_rate": 0.0002120194989066605, | |
| "loss": 2.3857, | |
| "step": 1244 | |
| }, | |
| { | |
| "epoch": 0.71, | |
| "learning_rate": 0.0002112677374053164, | |
| "loss": 2.3441, | |
| "step": 1245 | |
| }, | |
| { | |
| "epoch": 0.71, | |
| "learning_rate": 0.00021051695387929153, | |
| "loss": 2.3564, | |
| "step": 1246 | |
| }, | |
| { | |
| "epoch": 0.71, | |
| "learning_rate": 0.00020976715087159104, | |
| "loss": 2.396, | |
| "step": 1247 | |
| }, | |
| { | |
| "epoch": 0.71, | |
| "learning_rate": 0.00020901833092189982, | |
| "loss": 2.3913, | |
| "step": 1248 | |
| }, | |
| { | |
| "epoch": 0.71, | |
| "learning_rate": 0.0002082704965665721, | |
| "loss": 2.3628, | |
| "step": 1249 | |
| }, | |
| { | |
| "epoch": 0.71, | |
| "learning_rate": 0.00020752365033862463, | |
| "loss": 2.288, | |
| "step": 1250 | |
| }, | |
| { | |
| "epoch": 0.71, | |
| "learning_rate": 0.00020677779476772651, | |
| "loss": 2.3476, | |
| "step": 1251 | |
| }, | |
| { | |
| "epoch": 0.71, | |
| "learning_rate": 0.00020603293238019183, | |
| "loss": 2.3533, | |
| "step": 1252 | |
| }, | |
| { | |
| "epoch": 0.71, | |
| "learning_rate": 0.00020528906569897043, | |
| "loss": 2.281, | |
| "step": 1253 | |
| }, | |
| { | |
| "epoch": 0.71, | |
| "learning_rate": 0.00020454619724363994, | |
| "loss": 2.4064, | |
| "step": 1254 | |
| }, | |
| { | |
| "epoch": 0.71, | |
| "learning_rate": 0.00020380432953039603, | |
| "loss": 2.4079, | |
| "step": 1255 | |
| }, | |
| { | |
| "epoch": 0.71, | |
| "learning_rate": 0.00020306346507204564, | |
| "loss": 2.3546, | |
| "step": 1256 | |
| }, | |
| { | |
| "epoch": 0.71, | |
| "learning_rate": 0.00020232360637799685, | |
| "loss": 2.3198, | |
| "step": 1257 | |
| }, | |
| { | |
| "epoch": 0.71, | |
| "learning_rate": 0.00020158475595425158, | |
| "loss": 2.3119, | |
| "step": 1258 | |
| }, | |
| { | |
| "epoch": 0.72, | |
| "learning_rate": 0.00020084691630339614, | |
| "loss": 2.3479, | |
| "step": 1259 | |
| }, | |
| { | |
| "epoch": 0.72, | |
| "learning_rate": 0.00020011008992459374, | |
| "loss": 2.3278, | |
| "step": 1260 | |
| }, | |
| { | |
| "epoch": 0.72, | |
| "learning_rate": 0.0001993742793135749, | |
| "loss": 2.3611, | |
| "step": 1261 | |
| }, | |
| { | |
| "epoch": 0.72, | |
| "learning_rate": 0.0001986394869626303, | |
| "loss": 2.2788, | |
| "step": 1262 | |
| }, | |
| { | |
| "epoch": 0.72, | |
| "learning_rate": 0.00019790571536060087, | |
| "loss": 2.3928, | |
| "step": 1263 | |
| }, | |
| { | |
| "epoch": 0.72, | |
| "learning_rate": 0.00019717296699287086, | |
| "loss": 2.3541, | |
| "step": 1264 | |
| }, | |
| { | |
| "epoch": 0.72, | |
| "learning_rate": 0.00019644124434135802, | |
| "loss": 2.3972, | |
| "step": 1265 | |
| }, | |
| { | |
| "epoch": 0.72, | |
| "learning_rate": 0.00019571054988450648, | |
| "loss": 2.4321, | |
| "step": 1266 | |
| }, | |
| { | |
| "epoch": 0.72, | |
| "learning_rate": 0.0001949808860972771, | |
| "loss": 2.3616, | |
| "step": 1267 | |
| }, | |
| { | |
| "epoch": 0.72, | |
| "learning_rate": 0.0001942522554511404, | |
| "loss": 2.3426, | |
| "step": 1268 | |
| }, | |
| { | |
| "epoch": 0.72, | |
| "learning_rate": 0.00019352466041406684, | |
| "loss": 2.309, | |
| "step": 1269 | |
| }, | |
| { | |
| "epoch": 0.72, | |
| "learning_rate": 0.00019279810345051957, | |
| "loss": 2.3185, | |
| "step": 1270 | |
| }, | |
| { | |
| "epoch": 0.72, | |
| "learning_rate": 0.0001920725870214458, | |
| "loss": 2.3605, | |
| "step": 1271 | |
| }, | |
| { | |
| "epoch": 0.72, | |
| "learning_rate": 0.00019134811358426757, | |
| "loss": 2.2825, | |
| "step": 1272 | |
| }, | |
| { | |
| "epoch": 0.72, | |
| "learning_rate": 0.0001906246855928751, | |
| "loss": 2.3287, | |
| "step": 1273 | |
| }, | |
| { | |
| "epoch": 0.72, | |
| "learning_rate": 0.00018990230549761666, | |
| "loss": 2.3759, | |
| "step": 1274 | |
| }, | |
| { | |
| "epoch": 0.72, | |
| "learning_rate": 0.0001891809757452919, | |
| "loss": 2.3883, | |
| "step": 1275 | |
| }, | |
| { | |
| "epoch": 0.72, | |
| "learning_rate": 0.00018846069877914224, | |
| "loss": 2.3734, | |
| "step": 1276 | |
| }, | |
| { | |
| "epoch": 0.73, | |
| "learning_rate": 0.00018774147703884365, | |
| "loss": 2.3386, | |
| "step": 1277 | |
| }, | |
| { | |
| "epoch": 0.73, | |
| "learning_rate": 0.00018702331296049742, | |
| "loss": 2.3961, | |
| "step": 1278 | |
| }, | |
| { | |
| "epoch": 0.73, | |
| "learning_rate": 0.00018630620897662275, | |
| "loss": 2.305, | |
| "step": 1279 | |
| }, | |
| { | |
| "epoch": 0.73, | |
| "learning_rate": 0.0001855901675161481, | |
| "loss": 2.2302, | |
| "step": 1280 | |
| }, | |
| { | |
| "epoch": 0.73, | |
| "learning_rate": 0.00018487519100440314, | |
| "loss": 2.3607, | |
| "step": 1281 | |
| }, | |
| { | |
| "epoch": 0.73, | |
| "learning_rate": 0.00018416128186310988, | |
| "loss": 2.405, | |
| "step": 1282 | |
| }, | |
| { | |
| "epoch": 0.73, | |
| "learning_rate": 0.00018344844251037574, | |
| "loss": 2.3226, | |
| "step": 1283 | |
| }, | |
| { | |
| "epoch": 0.73, | |
| "learning_rate": 0.0001827366753606839, | |
| "loss": 2.4195, | |
| "step": 1284 | |
| }, | |
| { | |
| "epoch": 0.73, | |
| "learning_rate": 0.00018202598282488652, | |
| "loss": 2.3763, | |
| "step": 1285 | |
| }, | |
| { | |
| "epoch": 0.73, | |
| "learning_rate": 0.0001813163673101953, | |
| "loss": 2.3062, | |
| "step": 1286 | |
| }, | |
| { | |
| "epoch": 0.73, | |
| "learning_rate": 0.00018060783122017448, | |
| "loss": 2.3832, | |
| "step": 1287 | |
| }, | |
| { | |
| "epoch": 0.73, | |
| "learning_rate": 0.00017990037695473154, | |
| "loss": 2.2896, | |
| "step": 1288 | |
| }, | |
| { | |
| "epoch": 0.73, | |
| "learning_rate": 0.00017919400691011035, | |
| "loss": 2.3387, | |
| "step": 1289 | |
| }, | |
| { | |
| "epoch": 0.73, | |
| "learning_rate": 0.00017848872347888162, | |
| "loss": 2.357, | |
| "step": 1290 | |
| }, | |
| { | |
| "epoch": 0.73, | |
| "learning_rate": 0.0001777845290499363, | |
| "loss": 2.2424, | |
| "step": 1291 | |
| }, | |
| { | |
| "epoch": 0.73, | |
| "learning_rate": 0.0001770814260084761, | |
| "loss": 2.4304, | |
| "step": 1292 | |
| }, | |
| { | |
| "epoch": 0.73, | |
| "learning_rate": 0.00017637941673600665, | |
| "loss": 2.2946, | |
| "step": 1293 | |
| }, | |
| { | |
| "epoch": 0.74, | |
| "learning_rate": 0.0001756785036103282, | |
| "loss": 2.3486, | |
| "step": 1294 | |
| }, | |
| { | |
| "epoch": 0.74, | |
| "learning_rate": 0.00017497868900552888, | |
| "loss": 2.3464, | |
| "step": 1295 | |
| }, | |
| { | |
| "epoch": 0.74, | |
| "learning_rate": 0.0001742799752919753, | |
| "loss": 2.3703, | |
| "step": 1296 | |
| }, | |
| { | |
| "epoch": 0.74, | |
| "learning_rate": 0.0001735823648363059, | |
| "loss": 2.3988, | |
| "step": 1297 | |
| }, | |
| { | |
| "epoch": 0.74, | |
| "learning_rate": 0.00017288586000142149, | |
| "loss": 2.4077, | |
| "step": 1298 | |
| }, | |
| { | |
| "epoch": 0.74, | |
| "learning_rate": 0.00017219046314647873, | |
| "loss": 2.2285, | |
| "step": 1299 | |
| }, | |
| { | |
| "epoch": 0.74, | |
| "learning_rate": 0.00017149617662688072, | |
| "loss": 2.3668, | |
| "step": 1300 | |
| }, | |
| { | |
| "epoch": 0.74, | |
| "learning_rate": 0.00017080300279427035, | |
| "loss": 2.3386, | |
| "step": 1301 | |
| }, | |
| { | |
| "epoch": 0.74, | |
| "learning_rate": 0.00017011094399652104, | |
| "loss": 2.3591, | |
| "step": 1302 | |
| }, | |
| { | |
| "epoch": 0.74, | |
| "learning_rate": 0.00016942000257772998, | |
| "loss": 2.2742, | |
| "step": 1303 | |
| }, | |
| { | |
| "epoch": 0.74, | |
| "learning_rate": 0.0001687301808782095, | |
| "loss": 2.2877, | |
| "step": 1304 | |
| }, | |
| { | |
| "epoch": 0.74, | |
| "learning_rate": 0.00016804148123447898, | |
| "loss": 2.3301, | |
| "step": 1305 | |
| }, | |
| { | |
| "epoch": 0.74, | |
| "learning_rate": 0.00016735390597925765, | |
| "loss": 2.2378, | |
| "step": 1306 | |
| }, | |
| { | |
| "epoch": 0.74, | |
| "learning_rate": 0.00016666745744145617, | |
| "loss": 2.3547, | |
| "step": 1307 | |
| }, | |
| { | |
| "epoch": 0.74, | |
| "learning_rate": 0.00016598213794616863, | |
| "loss": 2.3288, | |
| "step": 1308 | |
| }, | |
| { | |
| "epoch": 0.74, | |
| "learning_rate": 0.00016529794981466527, | |
| "loss": 2.3687, | |
| "step": 1309 | |
| }, | |
| { | |
| "epoch": 0.74, | |
| "learning_rate": 0.00016461489536438378, | |
| "loss": 2.3908, | |
| "step": 1310 | |
| }, | |
| { | |
| "epoch": 0.74, | |
| "learning_rate": 0.00016393297690892256, | |
| "loss": 2.4032, | |
| "step": 1311 | |
| }, | |
| { | |
| "epoch": 0.75, | |
| "learning_rate": 0.0001632521967580315, | |
| "loss": 2.3021, | |
| "step": 1312 | |
| }, | |
| { | |
| "epoch": 0.75, | |
| "learning_rate": 0.00016257255721760538, | |
| "loss": 2.2437, | |
| "step": 1313 | |
| }, | |
| { | |
| "epoch": 0.75, | |
| "learning_rate": 0.00016189406058967577, | |
| "loss": 2.4009, | |
| "step": 1314 | |
| }, | |
| { | |
| "epoch": 0.75, | |
| "learning_rate": 0.00016121670917240228, | |
| "loss": 2.4062, | |
| "step": 1315 | |
| }, | |
| { | |
| "epoch": 0.75, | |
| "learning_rate": 0.00016054050526006635, | |
| "loss": 2.2743, | |
| "step": 1316 | |
| }, | |
| { | |
| "epoch": 0.75, | |
| "learning_rate": 0.000159865451143062, | |
| "loss": 2.3105, | |
| "step": 1317 | |
| }, | |
| { | |
| "epoch": 0.75, | |
| "learning_rate": 0.0001591915491078894, | |
| "loss": 2.369, | |
| "step": 1318 | |
| }, | |
| { | |
| "epoch": 0.75, | |
| "learning_rate": 0.00015851880143714575, | |
| "loss": 2.3745, | |
| "step": 1319 | |
| }, | |
| { | |
| "epoch": 0.75, | |
| "learning_rate": 0.00015784721040951899, | |
| "loss": 2.3623, | |
| "step": 1320 | |
| }, | |
| { | |
| "epoch": 0.75, | |
| "learning_rate": 0.0001571767782997786, | |
| "loss": 2.3515, | |
| "step": 1321 | |
| }, | |
| { | |
| "epoch": 0.75, | |
| "learning_rate": 0.00015650750737876935, | |
| "loss": 2.3739, | |
| "step": 1322 | |
| }, | |
| { | |
| "epoch": 0.75, | |
| "learning_rate": 0.0001558393999134023, | |
| "loss": 2.371, | |
| "step": 1323 | |
| }, | |
| { | |
| "epoch": 0.75, | |
| "learning_rate": 0.00015517245816664834, | |
| "loss": 2.2585, | |
| "step": 1324 | |
| }, | |
| { | |
| "epoch": 0.75, | |
| "learning_rate": 0.00015450668439752924, | |
| "loss": 2.3359, | |
| "step": 1325 | |
| }, | |
| { | |
| "epoch": 0.75, | |
| "learning_rate": 0.0001538420808611114, | |
| "loss": 2.4117, | |
| "step": 1326 | |
| }, | |
| { | |
| "epoch": 0.75, | |
| "learning_rate": 0.00015317864980849666, | |
| "loss": 2.3056, | |
| "step": 1327 | |
| }, | |
| { | |
| "epoch": 0.75, | |
| "learning_rate": 0.0001525163934868164, | |
| "loss": 2.3315, | |
| "step": 1328 | |
| }, | |
| { | |
| "epoch": 0.76, | |
| "learning_rate": 0.00015185531413922215, | |
| "loss": 2.3256, | |
| "step": 1329 | |
| }, | |
| { | |
| "epoch": 0.76, | |
| "learning_rate": 0.0001511954140048797, | |
| "loss": 2.3463, | |
| "step": 1330 | |
| }, | |
| { | |
| "epoch": 0.76, | |
| "learning_rate": 0.0001505366953189597, | |
| "loss": 2.3142, | |
| "step": 1331 | |
| }, | |
| { | |
| "epoch": 0.76, | |
| "learning_rate": 0.00014987916031263232, | |
| "loss": 2.317, | |
| "step": 1332 | |
| }, | |
| { | |
| "epoch": 0.76, | |
| "learning_rate": 0.00014922281121305736, | |
| "loss": 2.4335, | |
| "step": 1333 | |
| }, | |
| { | |
| "epoch": 0.76, | |
| "learning_rate": 0.00014856765024337843, | |
| "loss": 2.3708, | |
| "step": 1334 | |
| }, | |
| { | |
| "epoch": 0.76, | |
| "learning_rate": 0.00014791367962271423, | |
| "loss": 2.299, | |
| "step": 1335 | |
| }, | |
| { | |
| "epoch": 0.76, | |
| "learning_rate": 0.00014726090156615235, | |
| "loss": 2.3126, | |
| "step": 1336 | |
| }, | |
| { | |
| "epoch": 0.76, | |
| "learning_rate": 0.0001466093182847401, | |
| "loss": 2.3929, | |
| "step": 1337 | |
| }, | |
| { | |
| "epoch": 0.76, | |
| "learning_rate": 0.00014595893198547888, | |
| "loss": 2.3254, | |
| "step": 1338 | |
| }, | |
| { | |
| "epoch": 0.76, | |
| "learning_rate": 0.00014530974487131482, | |
| "loss": 2.3279, | |
| "step": 1339 | |
| }, | |
| { | |
| "epoch": 0.76, | |
| "learning_rate": 0.00014466175914113304, | |
| "loss": 2.3223, | |
| "step": 1340 | |
| }, | |
| { | |
| "epoch": 0.76, | |
| "learning_rate": 0.00014401497698974874, | |
| "loss": 2.3211, | |
| "step": 1341 | |
| }, | |
| { | |
| "epoch": 0.76, | |
| "learning_rate": 0.00014336940060790117, | |
| "loss": 2.3453, | |
| "step": 1342 | |
| }, | |
| { | |
| "epoch": 0.76, | |
| "learning_rate": 0.00014272503218224454, | |
| "loss": 2.4381, | |
| "step": 1343 | |
| }, | |
| { | |
| "epoch": 0.76, | |
| "learning_rate": 0.00014208187389534255, | |
| "loss": 2.3102, | |
| "step": 1344 | |
| }, | |
| { | |
| "epoch": 0.76, | |
| "learning_rate": 0.00014143992792565917, | |
| "loss": 2.3664, | |
| "step": 1345 | |
| }, | |
| { | |
| "epoch": 0.76, | |
| "learning_rate": 0.00014079919644755258, | |
| "loss": 2.3132, | |
| "step": 1346 | |
| }, | |
| { | |
| "epoch": 0.77, | |
| "learning_rate": 0.0001401596816312673, | |
| "loss": 2.3598, | |
| "step": 1347 | |
| }, | |
| { | |
| "epoch": 0.77, | |
| "learning_rate": 0.00013952138564292643, | |
| "loss": 2.404, | |
| "step": 1348 | |
| }, | |
| { | |
| "epoch": 0.77, | |
| "learning_rate": 0.00013888431064452527, | |
| "loss": 2.4006, | |
| "step": 1349 | |
| }, | |
| { | |
| "epoch": 0.77, | |
| "learning_rate": 0.00013824845879392301, | |
| "loss": 2.3513, | |
| "step": 1350 | |
| }, | |
| { | |
| "epoch": 0.77, | |
| "learning_rate": 0.0001376138322448363, | |
| "loss": 2.2951, | |
| "step": 1351 | |
| }, | |
| { | |
| "epoch": 0.77, | |
| "learning_rate": 0.00013698043314683107, | |
| "loss": 2.3619, | |
| "step": 1352 | |
| }, | |
| { | |
| "epoch": 0.77, | |
| "learning_rate": 0.00013634826364531617, | |
| "loss": 2.3567, | |
| "step": 1353 | |
| }, | |
| { | |
| "epoch": 0.77, | |
| "learning_rate": 0.00013571732588153512, | |
| "loss": 2.324, | |
| "step": 1354 | |
| }, | |
| { | |
| "epoch": 0.77, | |
| "learning_rate": 0.00013508762199256004, | |
| "loss": 2.2771, | |
| "step": 1355 | |
| }, | |
| { | |
| "epoch": 0.77, | |
| "learning_rate": 0.00013445915411128295, | |
| "loss": 2.3974, | |
| "step": 1356 | |
| }, | |
| { | |
| "epoch": 0.77, | |
| "learning_rate": 0.00013383192436641, | |
| "loss": 2.4326, | |
| "step": 1357 | |
| }, | |
| { | |
| "epoch": 0.77, | |
| "learning_rate": 0.00013320593488245336, | |
| "loss": 2.3954, | |
| "step": 1358 | |
| }, | |
| { | |
| "epoch": 0.77, | |
| "learning_rate": 0.0001325811877797245, | |
| "loss": 2.3697, | |
| "step": 1359 | |
| }, | |
| { | |
| "epoch": 0.77, | |
| "learning_rate": 0.00013195768517432611, | |
| "loss": 2.4528, | |
| "step": 1360 | |
| }, | |
| { | |
| "epoch": 0.77, | |
| "learning_rate": 0.0001313354291781465, | |
| "loss": 2.2912, | |
| "step": 1361 | |
| }, | |
| { | |
| "epoch": 0.77, | |
| "learning_rate": 0.0001307144218988507, | |
| "loss": 2.2383, | |
| "step": 1362 | |
| }, | |
| { | |
| "epoch": 0.77, | |
| "learning_rate": 0.00013009466543987487, | |
| "loss": 2.3279, | |
| "step": 1363 | |
| }, | |
| { | |
| "epoch": 0.77, | |
| "learning_rate": 0.00012947616190041783, | |
| "loss": 2.3716, | |
| "step": 1364 | |
| }, | |
| { | |
| "epoch": 0.78, | |
| "learning_rate": 0.00012885891337543538, | |
| "loss": 2.3219, | |
| "step": 1365 | |
| }, | |
| { | |
| "epoch": 0.78, | |
| "learning_rate": 0.00012824292195563148, | |
| "loss": 2.4057, | |
| "step": 1366 | |
| }, | |
| { | |
| "epoch": 0.78, | |
| "learning_rate": 0.00012762818972745292, | |
| "loss": 2.3609, | |
| "step": 1367 | |
| }, | |
| { | |
| "epoch": 0.78, | |
| "learning_rate": 0.0001270147187730809, | |
| "loss": 2.3141, | |
| "step": 1368 | |
| }, | |
| { | |
| "epoch": 0.78, | |
| "learning_rate": 0.0001264025111704249, | |
| "loss": 2.3393, | |
| "step": 1369 | |
| }, | |
| { | |
| "epoch": 0.78, | |
| "learning_rate": 0.00012579156899311485, | |
| "loss": 2.291, | |
| "step": 1370 | |
| }, | |
| { | |
| "epoch": 0.78, | |
| "learning_rate": 0.00012518189431049497, | |
| "loss": 2.3709, | |
| "step": 1371 | |
| }, | |
| { | |
| "epoch": 0.78, | |
| "learning_rate": 0.00012457348918761585, | |
| "loss": 2.2204, | |
| "step": 1372 | |
| }, | |
| { | |
| "epoch": 0.78, | |
| "learning_rate": 0.00012396635568522835, | |
| "loss": 2.3543, | |
| "step": 1373 | |
| }, | |
| { | |
| "epoch": 0.78, | |
| "learning_rate": 0.00012336049585977567, | |
| "loss": 2.3487, | |
| "step": 1374 | |
| }, | |
| { | |
| "epoch": 0.78, | |
| "learning_rate": 0.00012275591176338753, | |
| "loss": 2.3581, | |
| "step": 1375 | |
| }, | |
| { | |
| "epoch": 0.78, | |
| "learning_rate": 0.00012215260544387187, | |
| "loss": 2.3577, | |
| "step": 1376 | |
| }, | |
| { | |
| "epoch": 0.78, | |
| "learning_rate": 0.00012155057894470928, | |
| "loss": 2.3512, | |
| "step": 1377 | |
| }, | |
| { | |
| "epoch": 0.78, | |
| "learning_rate": 0.00012094983430504491, | |
| "loss": 2.2911, | |
| "step": 1378 | |
| }, | |
| { | |
| "epoch": 0.78, | |
| "learning_rate": 0.00012035037355968259, | |
| "loss": 2.3251, | |
| "step": 1379 | |
| }, | |
| { | |
| "epoch": 0.78, | |
| "learning_rate": 0.00011975219873907678, | |
| "loss": 2.2616, | |
| "step": 1380 | |
| }, | |
| { | |
| "epoch": 0.78, | |
| "learning_rate": 0.00011915531186932688, | |
| "loss": 2.3803, | |
| "step": 1381 | |
| }, | |
| { | |
| "epoch": 0.79, | |
| "learning_rate": 0.00011855971497216983, | |
| "loss": 2.3612, | |
| "step": 1382 | |
| }, | |
| { | |
| "epoch": 0.79, | |
| "learning_rate": 0.00011796541006497269, | |
| "loss": 2.3583, | |
| "step": 1383 | |
| }, | |
| { | |
| "epoch": 0.79, | |
| "learning_rate": 0.00011737239916072695, | |
| "loss": 2.3789, | |
| "step": 1384 | |
| }, | |
| { | |
| "epoch": 0.79, | |
| "learning_rate": 0.00011678068426804106, | |
| "loss": 2.457, | |
| "step": 1385 | |
| }, | |
| { | |
| "epoch": 0.79, | |
| "learning_rate": 0.00011619026739113331, | |
| "loss": 2.2434, | |
| "step": 1386 | |
| }, | |
| { | |
| "epoch": 0.79, | |
| "learning_rate": 0.00011560115052982606, | |
| "loss": 2.2616, | |
| "step": 1387 | |
| }, | |
| { | |
| "epoch": 0.79, | |
| "learning_rate": 0.00011501333567953758, | |
| "loss": 2.3304, | |
| "step": 1388 | |
| }, | |
| { | |
| "epoch": 0.79, | |
| "learning_rate": 0.00011442682483127687, | |
| "loss": 2.3475, | |
| "step": 1389 | |
| }, | |
| { | |
| "epoch": 0.79, | |
| "learning_rate": 0.00011384161997163533, | |
| "loss": 2.2778, | |
| "step": 1390 | |
| }, | |
| { | |
| "epoch": 0.79, | |
| "learning_rate": 0.00011325772308278132, | |
| "loss": 2.3904, | |
| "step": 1391 | |
| }, | |
| { | |
| "epoch": 0.79, | |
| "learning_rate": 0.00011267513614245289, | |
| "loss": 2.3907, | |
| "step": 1392 | |
| }, | |
| { | |
| "epoch": 0.79, | |
| "learning_rate": 0.00011209386112395076, | |
| "loss": 2.3336, | |
| "step": 1393 | |
| }, | |
| { | |
| "epoch": 0.79, | |
| "learning_rate": 0.00011151389999613248, | |
| "loss": 2.3793, | |
| "step": 1394 | |
| }, | |
| { | |
| "epoch": 0.79, | |
| "learning_rate": 0.00011093525472340472, | |
| "loss": 2.3641, | |
| "step": 1395 | |
| }, | |
| { | |
| "epoch": 0.79, | |
| "learning_rate": 0.00011035792726571776, | |
| "loss": 2.3915, | |
| "step": 1396 | |
| }, | |
| { | |
| "epoch": 0.79, | |
| "learning_rate": 0.00010978191957855771, | |
| "loss": 2.3244, | |
| "step": 1397 | |
| }, | |
| { | |
| "epoch": 0.79, | |
| "learning_rate": 0.00010920723361294099, | |
| "loss": 2.2922, | |
| "step": 1398 | |
| }, | |
| { | |
| "epoch": 0.79, | |
| "learning_rate": 0.00010863387131540653, | |
| "loss": 2.4013, | |
| "step": 1399 | |
| }, | |
| { | |
| "epoch": 0.8, | |
| "learning_rate": 0.0001080618346280105, | |
| "loss": 2.3676, | |
| "step": 1400 | |
| }, | |
| { | |
| "epoch": 0.8, | |
| "learning_rate": 0.00010749112548831846, | |
| "loss": 2.3596, | |
| "step": 1401 | |
| }, | |
| { | |
| "epoch": 0.8, | |
| "learning_rate": 0.0001069217458293999, | |
| "loss": 2.3525, | |
| "step": 1402 | |
| }, | |
| { | |
| "epoch": 0.8, | |
| "learning_rate": 0.00010635369757982061, | |
| "loss": 2.3417, | |
| "step": 1403 | |
| }, | |
| { | |
| "epoch": 0.8, | |
| "learning_rate": 0.00010578698266363734, | |
| "loss": 2.358, | |
| "step": 1404 | |
| }, | |
| { | |
| "epoch": 0.8, | |
| "learning_rate": 0.00010522160300039007, | |
| "loss": 2.3709, | |
| "step": 1405 | |
| }, | |
| { | |
| "epoch": 0.8, | |
| "learning_rate": 0.00010465756050509661, | |
| "loss": 2.4521, | |
| "step": 1406 | |
| }, | |
| { | |
| "epoch": 0.8, | |
| "learning_rate": 0.00010409485708824506, | |
| "loss": 2.4123, | |
| "step": 1407 | |
| }, | |
| { | |
| "epoch": 0.8, | |
| "learning_rate": 0.00010353349465578859, | |
| "loss": 2.3052, | |
| "step": 1408 | |
| }, | |
| { | |
| "epoch": 0.8, | |
| "learning_rate": 0.00010297347510913746, | |
| "loss": 2.3366, | |
| "step": 1409 | |
| }, | |
| { | |
| "epoch": 0.8, | |
| "learning_rate": 0.00010241480034515404, | |
| "loss": 2.3183, | |
| "step": 1410 | |
| }, | |
| { | |
| "epoch": 0.8, | |
| "learning_rate": 0.00010185747225614539, | |
| "loss": 2.3082, | |
| "step": 1411 | |
| }, | |
| { | |
| "epoch": 0.8, | |
| "learning_rate": 0.00010130149272985756, | |
| "loss": 2.37, | |
| "step": 1412 | |
| }, | |
| { | |
| "epoch": 0.8, | |
| "learning_rate": 0.00010074686364946823, | |
| "loss": 2.399, | |
| "step": 1413 | |
| }, | |
| { | |
| "epoch": 0.8, | |
| "learning_rate": 0.00010019358689358154, | |
| "loss": 2.4144, | |
| "step": 1414 | |
| }, | |
| { | |
| "epoch": 0.8, | |
| "learning_rate": 9.964166433622068e-05, | |
| "loss": 2.3796, | |
| "step": 1415 | |
| }, | |
| { | |
| "epoch": 0.8, | |
| "learning_rate": 9.90910978468224e-05, | |
| "loss": 2.3717, | |
| "step": 1416 | |
| }, | |
| { | |
| "epoch": 0.81, | |
| "learning_rate": 9.854188929022984e-05, | |
| "loss": 2.3963, | |
| "step": 1417 | |
| }, | |
| { | |
| "epoch": 0.81, | |
| "learning_rate": 9.799404052668704e-05, | |
| "loss": 2.3167, | |
| "step": 1418 | |
| }, | |
| { | |
| "epoch": 0.81, | |
| "learning_rate": 9.744755341183181e-05, | |
| "loss": 2.4056, | |
| "step": 1419 | |
| }, | |
| { | |
| "epoch": 0.81, | |
| "learning_rate": 9.690242979669035e-05, | |
| "loss": 2.3068, | |
| "step": 1420 | |
| }, | |
| { | |
| "epoch": 0.81, | |
| "learning_rate": 9.635867152766997e-05, | |
| "loss": 2.3482, | |
| "step": 1421 | |
| }, | |
| { | |
| "epoch": 0.81, | |
| "learning_rate": 9.581628044655394e-05, | |
| "loss": 2.3084, | |
| "step": 1422 | |
| }, | |
| { | |
| "epoch": 0.81, | |
| "learning_rate": 9.527525839049406e-05, | |
| "loss": 2.3281, | |
| "step": 1423 | |
| }, | |
| { | |
| "epoch": 0.81, | |
| "learning_rate": 9.473560719200564e-05, | |
| "loss": 2.3111, | |
| "step": 1424 | |
| }, | |
| { | |
| "epoch": 0.81, | |
| "learning_rate": 9.419732867896047e-05, | |
| "loss": 2.3346, | |
| "step": 1425 | |
| }, | |
| { | |
| "epoch": 0.81, | |
| "learning_rate": 9.366042467458064e-05, | |
| "loss": 2.3117, | |
| "step": 1426 | |
| }, | |
| { | |
| "epoch": 0.81, | |
| "learning_rate": 9.312489699743309e-05, | |
| "loss": 2.3598, | |
| "step": 1427 | |
| }, | |
| { | |
| "epoch": 0.81, | |
| "learning_rate": 9.259074746142238e-05, | |
| "loss": 2.3327, | |
| "step": 1428 | |
| }, | |
| { | |
| "epoch": 0.81, | |
| "learning_rate": 9.205797787578563e-05, | |
| "loss": 2.3157, | |
| "step": 1429 | |
| }, | |
| { | |
| "epoch": 0.81, | |
| "learning_rate": 9.152659004508545e-05, | |
| "loss": 2.4291, | |
| "step": 1430 | |
| }, | |
| { | |
| "epoch": 0.81, | |
| "learning_rate": 9.099658576920466e-05, | |
| "loss": 2.3179, | |
| "step": 1431 | |
| }, | |
| { | |
| "epoch": 0.81, | |
| "learning_rate": 9.046796684333947e-05, | |
| "loss": 2.3465, | |
| "step": 1432 | |
| }, | |
| { | |
| "epoch": 0.81, | |
| "learning_rate": 8.994073505799399e-05, | |
| "loss": 2.395, | |
| "step": 1433 | |
| }, | |
| { | |
| "epoch": 0.81, | |
| "learning_rate": 8.941489219897354e-05, | |
| "loss": 2.3563, | |
| "step": 1434 | |
| }, | |
| { | |
| "epoch": 0.82, | |
| "learning_rate": 8.889044004737939e-05, | |
| "loss": 2.3259, | |
| "step": 1435 | |
| }, | |
| { | |
| "epoch": 0.82, | |
| "learning_rate": 8.836738037960179e-05, | |
| "loss": 2.3736, | |
| "step": 1436 | |
| }, | |
| { | |
| "epoch": 0.82, | |
| "learning_rate": 8.78457149673152e-05, | |
| "loss": 2.3209, | |
| "step": 1437 | |
| }, | |
| { | |
| "epoch": 0.82, | |
| "learning_rate": 8.732544557747073e-05, | |
| "loss": 2.3401, | |
| "step": 1438 | |
| }, | |
| { | |
| "epoch": 0.82, | |
| "learning_rate": 8.680657397229158e-05, | |
| "loss": 2.3871, | |
| "step": 1439 | |
| }, | |
| { | |
| "epoch": 0.82, | |
| "learning_rate": 8.628910190926598e-05, | |
| "loss": 2.3419, | |
| "step": 1440 | |
| }, | |
| { | |
| "epoch": 0.82, | |
| "learning_rate": 8.577303114114227e-05, | |
| "loss": 2.3418, | |
| "step": 1441 | |
| }, | |
| { | |
| "epoch": 0.82, | |
| "learning_rate": 8.525836341592175e-05, | |
| "loss": 2.4486, | |
| "step": 1442 | |
| }, | |
| { | |
| "epoch": 0.82, | |
| "learning_rate": 8.474510047685408e-05, | |
| "loss": 2.3244, | |
| "step": 1443 | |
| }, | |
| { | |
| "epoch": 0.82, | |
| "learning_rate": 8.423324406243016e-05, | |
| "loss": 2.3141, | |
| "step": 1444 | |
| }, | |
| { | |
| "epoch": 0.82, | |
| "learning_rate": 8.372279590637738e-05, | |
| "loss": 2.321, | |
| "step": 1445 | |
| }, | |
| { | |
| "epoch": 0.82, | |
| "learning_rate": 8.321375773765255e-05, | |
| "loss": 2.2767, | |
| "step": 1446 | |
| }, | |
| { | |
| "epoch": 0.82, | |
| "learning_rate": 8.27061312804372e-05, | |
| "loss": 2.3465, | |
| "step": 1447 | |
| }, | |
| { | |
| "epoch": 0.82, | |
| "learning_rate": 8.21999182541307e-05, | |
| "loss": 2.3883, | |
| "step": 1448 | |
| }, | |
| { | |
| "epoch": 0.82, | |
| "learning_rate": 8.169512037334554e-05, | |
| "loss": 2.3339, | |
| "step": 1449 | |
| }, | |
| { | |
| "epoch": 0.82, | |
| "learning_rate": 8.119173934790019e-05, | |
| "loss": 2.3466, | |
| "step": 1450 | |
| }, | |
| { | |
| "epoch": 0.82, | |
| "learning_rate": 8.068977688281482e-05, | |
| "loss": 2.3427, | |
| "step": 1451 | |
| }, | |
| { | |
| "epoch": 0.82, | |
| "learning_rate": 8.018923467830403e-05, | |
| "loss": 2.3627, | |
| "step": 1452 | |
| }, | |
| { | |
| "epoch": 0.83, | |
| "learning_rate": 7.969011442977237e-05, | |
| "loss": 2.3335, | |
| "step": 1453 | |
| }, | |
| { | |
| "epoch": 0.83, | |
| "learning_rate": 7.919241782780756e-05, | |
| "loss": 2.2868, | |
| "step": 1454 | |
| }, | |
| { | |
| "epoch": 0.83, | |
| "learning_rate": 7.869614655817576e-05, | |
| "loss": 2.3564, | |
| "step": 1455 | |
| }, | |
| { | |
| "epoch": 0.83, | |
| "learning_rate": 7.820130230181472e-05, | |
| "loss": 2.333, | |
| "step": 1456 | |
| }, | |
| { | |
| "epoch": 0.83, | |
| "learning_rate": 7.770788673482921e-05, | |
| "loss": 2.2877, | |
| "step": 1457 | |
| }, | |
| { | |
| "epoch": 0.83, | |
| "learning_rate": 7.721590152848473e-05, | |
| "loss": 2.3721, | |
| "step": 1458 | |
| }, | |
| { | |
| "epoch": 0.83, | |
| "learning_rate": 7.672534834920164e-05, | |
| "loss": 2.3642, | |
| "step": 1459 | |
| }, | |
| { | |
| "epoch": 0.83, | |
| "learning_rate": 7.623622885855036e-05, | |
| "loss": 2.3208, | |
| "step": 1460 | |
| }, | |
| { | |
| "epoch": 0.83, | |
| "learning_rate": 7.574854471324461e-05, | |
| "loss": 2.3572, | |
| "step": 1461 | |
| }, | |
| { | |
| "epoch": 0.83, | |
| "learning_rate": 7.526229756513686e-05, | |
| "loss": 2.3688, | |
| "step": 1462 | |
| }, | |
| { | |
| "epoch": 0.83, | |
| "learning_rate": 7.477748906121223e-05, | |
| "loss": 2.3688, | |
| "step": 1463 | |
| }, | |
| { | |
| "epoch": 0.83, | |
| "learning_rate": 7.429412084358261e-05, | |
| "loss": 2.319, | |
| "step": 1464 | |
| }, | |
| { | |
| "epoch": 0.83, | |
| "learning_rate": 7.381219454948196e-05, | |
| "loss": 2.3633, | |
| "step": 1465 | |
| }, | |
| { | |
| "epoch": 0.83, | |
| "learning_rate": 7.333171181125975e-05, | |
| "loss": 2.3446, | |
| "step": 1466 | |
| }, | |
| { | |
| "epoch": 0.83, | |
| "learning_rate": 7.285267425637621e-05, | |
| "loss": 2.295, | |
| "step": 1467 | |
| }, | |
| { | |
| "epoch": 0.83, | |
| "learning_rate": 7.237508350739663e-05, | |
| "loss": 2.3123, | |
| "step": 1468 | |
| }, | |
| { | |
| "epoch": 0.83, | |
| "learning_rate": 7.189894118198542e-05, | |
| "loss": 2.3276, | |
| "step": 1469 | |
| }, | |
| { | |
| "epoch": 0.84, | |
| "learning_rate": 7.142424889290139e-05, | |
| "loss": 2.3237, | |
| "step": 1470 | |
| }, | |
| { | |
| "epoch": 0.84, | |
| "learning_rate": 7.095100824799145e-05, | |
| "loss": 2.3258, | |
| "step": 1471 | |
| }, | |
| { | |
| "epoch": 0.84, | |
| "learning_rate": 7.047922085018604e-05, | |
| "loss": 2.3153, | |
| "step": 1472 | |
| }, | |
| { | |
| "epoch": 0.84, | |
| "learning_rate": 7.00088882974929e-05, | |
| "loss": 2.3657, | |
| "step": 1473 | |
| }, | |
| { | |
| "epoch": 0.84, | |
| "learning_rate": 6.954001218299227e-05, | |
| "loss": 2.3485, | |
| "step": 1474 | |
| }, | |
| { | |
| "epoch": 0.84, | |
| "learning_rate": 6.907259409483097e-05, | |
| "loss": 2.3716, | |
| "step": 1475 | |
| }, | |
| { | |
| "epoch": 0.84, | |
| "learning_rate": 6.860663561621766e-05, | |
| "loss": 2.3926, | |
| "step": 1476 | |
| }, | |
| { | |
| "epoch": 0.84, | |
| "learning_rate": 6.814213832541655e-05, | |
| "loss": 2.3702, | |
| "step": 1477 | |
| }, | |
| { | |
| "epoch": 0.84, | |
| "learning_rate": 6.767910379574332e-05, | |
| "loss": 2.3756, | |
| "step": 1478 | |
| }, | |
| { | |
| "epoch": 0.84, | |
| "learning_rate": 6.721753359555833e-05, | |
| "loss": 2.3763, | |
| "step": 1479 | |
| }, | |
| { | |
| "epoch": 0.84, | |
| "learning_rate": 6.675742928826273e-05, | |
| "loss": 2.373, | |
| "step": 1480 | |
| }, | |
| { | |
| "epoch": 0.84, | |
| "learning_rate": 6.629879243229198e-05, | |
| "loss": 2.2572, | |
| "step": 1481 | |
| }, | |
| { | |
| "epoch": 0.84, | |
| "learning_rate": 6.584162458111148e-05, | |
| "loss": 2.3411, | |
| "step": 1482 | |
| }, | |
| { | |
| "epoch": 0.84, | |
| "learning_rate": 6.538592728321053e-05, | |
| "loss": 2.3552, | |
| "step": 1483 | |
| }, | |
| { | |
| "epoch": 0.84, | |
| "learning_rate": 6.493170208209781e-05, | |
| "loss": 2.4024, | |
| "step": 1484 | |
| }, | |
| { | |
| "epoch": 0.84, | |
| "learning_rate": 6.447895051629549e-05, | |
| "loss": 2.3138, | |
| "step": 1485 | |
| }, | |
| { | |
| "epoch": 0.84, | |
| "learning_rate": 6.402767411933475e-05, | |
| "loss": 2.3135, | |
| "step": 1486 | |
| }, | |
| { | |
| "epoch": 0.84, | |
| "learning_rate": 6.357787441974967e-05, | |
| "loss": 2.271, | |
| "step": 1487 | |
| }, | |
| { | |
| "epoch": 0.85, | |
| "learning_rate": 6.312955294107303e-05, | |
| "loss": 2.3715, | |
| "step": 1488 | |
| }, | |
| { | |
| "epoch": 0.85, | |
| "learning_rate": 6.268271120183022e-05, | |
| "loss": 2.3508, | |
| "step": 1489 | |
| }, | |
| { | |
| "epoch": 0.85, | |
| "learning_rate": 6.22373507155351e-05, | |
| "loss": 2.3682, | |
| "step": 1490 | |
| }, | |
| { | |
| "epoch": 0.85, | |
| "learning_rate": 6.179347299068378e-05, | |
| "loss": 2.361, | |
| "step": 1491 | |
| }, | |
| { | |
| "epoch": 0.85, | |
| "learning_rate": 6.135107953075048e-05, | |
| "loss": 2.2785, | |
| "step": 1492 | |
| }, | |
| { | |
| "epoch": 0.85, | |
| "learning_rate": 6.091017183418157e-05, | |
| "loss": 2.3612, | |
| "step": 1493 | |
| }, | |
| { | |
| "epoch": 0.85, | |
| "learning_rate": 6.047075139439151e-05, | |
| "loss": 2.3511, | |
| "step": 1494 | |
| }, | |
| { | |
| "epoch": 0.85, | |
| "learning_rate": 6.00328196997566e-05, | |
| "loss": 2.3548, | |
| "step": 1495 | |
| }, | |
| { | |
| "epoch": 0.85, | |
| "learning_rate": 5.9596378233611135e-05, | |
| "loss": 2.257, | |
| "step": 1496 | |
| }, | |
| { | |
| "epoch": 0.85, | |
| "learning_rate": 5.9161428474241274e-05, | |
| "loss": 2.2737, | |
| "step": 1497 | |
| }, | |
| { | |
| "epoch": 0.85, | |
| "learning_rate": 5.872797189488105e-05, | |
| "loss": 2.3026, | |
| "step": 1498 | |
| }, | |
| { | |
| "epoch": 0.85, | |
| "learning_rate": 5.829600996370649e-05, | |
| "loss": 2.2523, | |
| "step": 1499 | |
| }, | |
| { | |
| "epoch": 0.85, | |
| "learning_rate": 5.7865544143831274e-05, | |
| "loss": 2.2814, | |
| "step": 1500 | |
| }, | |
| { | |
| "epoch": 0.85, | |
| "learning_rate": 5.743657589330159e-05, | |
| "loss": 2.2883, | |
| "step": 1501 | |
| }, | |
| { | |
| "epoch": 0.85, | |
| "learning_rate": 5.700910666509096e-05, | |
| "loss": 2.2735, | |
| "step": 1502 | |
| }, | |
| { | |
| "epoch": 0.85, | |
| "learning_rate": 5.6583137907095686e-05, | |
| "loss": 2.3183, | |
| "step": 1503 | |
| }, | |
| { | |
| "epoch": 0.85, | |
| "learning_rate": 5.615867106212957e-05, | |
| "loss": 2.3637, | |
| "step": 1504 | |
| }, | |
| { | |
| "epoch": 0.86, | |
| "learning_rate": 5.573570756791957e-05, | |
| "loss": 2.2044, | |
| "step": 1505 | |
| }, | |
| { | |
| "epoch": 0.86, | |
| "learning_rate": 5.531424885710012e-05, | |
| "loss": 2.3988, | |
| "step": 1506 | |
| }, | |
| { | |
| "epoch": 0.86, | |
| "learning_rate": 5.4894296357209274e-05, | |
| "loss": 2.3839, | |
| "step": 1507 | |
| }, | |
| { | |
| "epoch": 0.86, | |
| "learning_rate": 5.4475851490682904e-05, | |
| "loss": 2.3843, | |
| "step": 1508 | |
| }, | |
| { | |
| "epoch": 0.86, | |
| "learning_rate": 5.4058915674850664e-05, | |
| "loss": 2.3209, | |
| "step": 1509 | |
| }, | |
| { | |
| "epoch": 0.86, | |
| "learning_rate": 5.36434903219305e-05, | |
| "loss": 2.3541, | |
| "step": 1510 | |
| }, | |
| { | |
| "epoch": 0.86, | |
| "learning_rate": 5.322957683902457e-05, | |
| "loss": 2.3996, | |
| "step": 1511 | |
| }, | |
| { | |
| "epoch": 0.86, | |
| "learning_rate": 5.2817176628113807e-05, | |
| "loss": 2.3179, | |
| "step": 1512 | |
| }, | |
| { | |
| "epoch": 0.86, | |
| "learning_rate": 5.2406291086053826e-05, | |
| "loss": 2.3432, | |
| "step": 1513 | |
| }, | |
| { | |
| "epoch": 0.86, | |
| "learning_rate": 5.199692160456948e-05, | |
| "loss": 2.367, | |
| "step": 1514 | |
| }, | |
| { | |
| "epoch": 0.86, | |
| "learning_rate": 5.158906957025078e-05, | |
| "loss": 2.368, | |
| "step": 1515 | |
| }, | |
| { | |
| "epoch": 0.86, | |
| "learning_rate": 5.118273636454784e-05, | |
| "loss": 2.3652, | |
| "step": 1516 | |
| }, | |
| { | |
| "epoch": 0.86, | |
| "learning_rate": 5.077792336376647e-05, | |
| "loss": 2.3421, | |
| "step": 1517 | |
| }, | |
| { | |
| "epoch": 0.86, | |
| "learning_rate": 5.0374631939062943e-05, | |
| "loss": 2.3421, | |
| "step": 1518 | |
| }, | |
| { | |
| "epoch": 0.86, | |
| "learning_rate": 4.99728634564402e-05, | |
| "loss": 2.2835, | |
| "step": 1519 | |
| }, | |
| { | |
| "epoch": 0.86, | |
| "learning_rate": 4.9572619276742393e-05, | |
| "loss": 2.3785, | |
| "step": 1520 | |
| }, | |
| { | |
| "epoch": 0.86, | |
| "learning_rate": 4.917390075565098e-05, | |
| "loss": 2.3257, | |
| "step": 1521 | |
| }, | |
| { | |
| "epoch": 0.86, | |
| "learning_rate": 4.8776709243679495e-05, | |
| "loss": 2.3212, | |
| "step": 1522 | |
| }, | |
| { | |
| "epoch": 0.87, | |
| "learning_rate": 4.838104608616961e-05, | |
| "loss": 2.3864, | |
| "step": 1523 | |
| }, | |
| { | |
| "epoch": 0.87, | |
| "learning_rate": 4.798691262328586e-05, | |
| "loss": 2.3235, | |
| "step": 1524 | |
| }, | |
| { | |
| "epoch": 0.87, | |
| "learning_rate": 4.759431019001198e-05, | |
| "loss": 2.3822, | |
| "step": 1525 | |
| }, | |
| { | |
| "epoch": 0.87, | |
| "learning_rate": 4.720324011614535e-05, | |
| "loss": 2.3212, | |
| "step": 1526 | |
| }, | |
| { | |
| "epoch": 0.87, | |
| "learning_rate": 4.6813703726293676e-05, | |
| "loss": 2.451, | |
| "step": 1527 | |
| }, | |
| { | |
| "epoch": 0.87, | |
| "learning_rate": 4.642570233986915e-05, | |
| "loss": 2.3598, | |
| "step": 1528 | |
| }, | |
| { | |
| "epoch": 0.87, | |
| "learning_rate": 4.6039237271085386e-05, | |
| "loss": 2.2875, | |
| "step": 1529 | |
| }, | |
| { | |
| "epoch": 0.87, | |
| "learning_rate": 4.565430982895175e-05, | |
| "loss": 2.4021, | |
| "step": 1530 | |
| }, | |
| { | |
| "epoch": 0.87, | |
| "learning_rate": 4.52709213172699e-05, | |
| "loss": 2.3524, | |
| "step": 1531 | |
| }, | |
| { | |
| "epoch": 0.87, | |
| "learning_rate": 4.488907303462841e-05, | |
| "loss": 2.3779, | |
| "step": 1532 | |
| }, | |
| { | |
| "epoch": 0.87, | |
| "learning_rate": 4.4508766274399424e-05, | |
| "loss": 2.3639, | |
| "step": 1533 | |
| }, | |
| { | |
| "epoch": 0.87, | |
| "learning_rate": 4.4130002324733355e-05, | |
| "loss": 2.3627, | |
| "step": 1534 | |
| }, | |
| { | |
| "epoch": 0.87, | |
| "learning_rate": 4.375278246855513e-05, | |
| "loss": 2.3944, | |
| "step": 1535 | |
| }, | |
| { | |
| "epoch": 0.87, | |
| "learning_rate": 4.3377107983559596e-05, | |
| "loss": 2.2929, | |
| "step": 1536 | |
| }, | |
| { | |
| "epoch": 0.87, | |
| "learning_rate": 4.300298014220699e-05, | |
| "loss": 2.3835, | |
| "step": 1537 | |
| }, | |
| { | |
| "epoch": 0.87, | |
| "learning_rate": 4.2630400211719264e-05, | |
| "loss": 2.2726, | |
| "step": 1538 | |
| }, | |
| { | |
| "epoch": 0.87, | |
| "learning_rate": 4.2259369454074935e-05, | |
| "loss": 2.312, | |
| "step": 1539 | |
| }, | |
| { | |
| "epoch": 0.87, | |
| "learning_rate": 4.188988912600566e-05, | |
| "loss": 2.3564, | |
| "step": 1540 | |
| }, | |
| { | |
| "epoch": 0.88, | |
| "learning_rate": 4.152196047899126e-05, | |
| "loss": 2.4055, | |
| "step": 1541 | |
| }, | |
| { | |
| "epoch": 0.88, | |
| "learning_rate": 4.115558475925601e-05, | |
| "loss": 2.3959, | |
| "step": 1542 | |
| }, | |
| { | |
| "epoch": 0.88, | |
| "learning_rate": 4.079076320776426e-05, | |
| "loss": 2.3602, | |
| "step": 1543 | |
| }, | |
| { | |
| "epoch": 0.88, | |
| "learning_rate": 4.042749706021581e-05, | |
| "loss": 2.362, | |
| "step": 1544 | |
| }, | |
| { | |
| "epoch": 0.88, | |
| "learning_rate": 4.0065787547042544e-05, | |
| "loss": 2.2404, | |
| "step": 1545 | |
| }, | |
| { | |
| "epoch": 0.88, | |
| "learning_rate": 3.9705635893403594e-05, | |
| "loss": 2.3495, | |
| "step": 1546 | |
| }, | |
| { | |
| "epoch": 0.88, | |
| "learning_rate": 3.934704331918137e-05, | |
| "loss": 2.3613, | |
| "step": 1547 | |
| }, | |
| { | |
| "epoch": 0.88, | |
| "learning_rate": 3.89900110389777e-05, | |
| "loss": 2.255, | |
| "step": 1548 | |
| }, | |
| { | |
| "epoch": 0.88, | |
| "learning_rate": 3.863454026210911e-05, | |
| "loss": 2.3231, | |
| "step": 1549 | |
| }, | |
| { | |
| "epoch": 0.88, | |
| "learning_rate": 3.828063219260347e-05, | |
| "loss": 2.3597, | |
| "step": 1550 | |
| }, | |
| { | |
| "epoch": 0.88, | |
| "learning_rate": 3.792828802919529e-05, | |
| "loss": 2.4335, | |
| "step": 1551 | |
| }, | |
| { | |
| "epoch": 0.88, | |
| "learning_rate": 3.757750896532214e-05, | |
| "loss": 2.3403, | |
| "step": 1552 | |
| }, | |
| { | |
| "epoch": 0.88, | |
| "learning_rate": 3.72282961891201e-05, | |
| "loss": 2.4333, | |
| "step": 1553 | |
| }, | |
| { | |
| "epoch": 0.88, | |
| "learning_rate": 3.688065088342041e-05, | |
| "loss": 2.377, | |
| "step": 1554 | |
| }, | |
| { | |
| "epoch": 0.88, | |
| "learning_rate": 3.653457422574458e-05, | |
| "loss": 2.3642, | |
| "step": 1555 | |
| }, | |
| { | |
| "epoch": 0.88, | |
| "learning_rate": 3.619006738830144e-05, | |
| "loss": 2.4349, | |
| "step": 1556 | |
| }, | |
| { | |
| "epoch": 0.88, | |
| "learning_rate": 3.584713153798214e-05, | |
| "loss": 2.3163, | |
| "step": 1557 | |
| }, | |
| { | |
| "epoch": 0.89, | |
| "learning_rate": 3.550576783635712e-05, | |
| "loss": 2.3527, | |
| "step": 1558 | |
| }, | |
| { | |
| "epoch": 0.89, | |
| "learning_rate": 3.516597743967126e-05, | |
| "loss": 2.3009, | |
| "step": 1559 | |
| }, | |
| { | |
| "epoch": 0.89, | |
| "learning_rate": 3.4827761498840924e-05, | |
| "loss": 2.3623, | |
| "step": 1560 | |
| }, | |
| { | |
| "epoch": 0.89, | |
| "learning_rate": 3.449112115944919e-05, | |
| "loss": 2.2798, | |
| "step": 1561 | |
| }, | |
| { | |
| "epoch": 0.89, | |
| "learning_rate": 3.415605756174267e-05, | |
| "loss": 2.3023, | |
| "step": 1562 | |
| }, | |
| { | |
| "epoch": 0.89, | |
| "learning_rate": 3.382257184062709e-05, | |
| "loss": 2.4368, | |
| "step": 1563 | |
| }, | |
| { | |
| "epoch": 0.89, | |
| "learning_rate": 3.34906651256639e-05, | |
| "loss": 2.3222, | |
| "step": 1564 | |
| }, | |
| { | |
| "epoch": 0.89, | |
| "learning_rate": 3.316033854106604e-05, | |
| "loss": 2.2483, | |
| "step": 1565 | |
| }, | |
| { | |
| "epoch": 0.89, | |
| "learning_rate": 3.283159320569451e-05, | |
| "loss": 2.2451, | |
| "step": 1566 | |
| }, | |
| { | |
| "epoch": 0.89, | |
| "learning_rate": 3.2504430233054274e-05, | |
| "loss": 2.3724, | |
| "step": 1567 | |
| }, | |
| { | |
| "epoch": 0.89, | |
| "learning_rate": 3.217885073129067e-05, | |
| "loss": 2.3633, | |
| "step": 1568 | |
| }, | |
| { | |
| "epoch": 0.89, | |
| "learning_rate": 3.1854855803185645e-05, | |
| "loss": 2.408, | |
| "step": 1569 | |
| }, | |
| { | |
| "epoch": 0.89, | |
| "learning_rate": 3.153244654615406e-05, | |
| "loss": 2.3808, | |
| "step": 1570 | |
| }, | |
| { | |
| "epoch": 0.89, | |
| "learning_rate": 3.121162405223954e-05, | |
| "loss": 2.4206, | |
| "step": 1571 | |
| }, | |
| { | |
| "epoch": 0.89, | |
| "learning_rate": 3.089238940811162e-05, | |
| "loss": 2.3333, | |
| "step": 1572 | |
| }, | |
| { | |
| "epoch": 0.89, | |
| "learning_rate": 3.057474369506108e-05, | |
| "loss": 2.306, | |
| "step": 1573 | |
| }, | |
| { | |
| "epoch": 0.89, | |
| "learning_rate": 3.025868798899728e-05, | |
| "loss": 2.2793, | |
| "step": 1574 | |
| }, | |
| { | |
| "epoch": 0.89, | |
| "learning_rate": 2.9944223360443446e-05, | |
| "loss": 2.4081, | |
| "step": 1575 | |
| }, | |
| { | |
| "epoch": 0.9, | |
| "learning_rate": 2.9631350874534225e-05, | |
| "loss": 2.3331, | |
| "step": 1576 | |
| }, | |
| { | |
| "epoch": 0.9, | |
| "learning_rate": 2.9320071591010854e-05, | |
| "loss": 2.3505, | |
| "step": 1577 | |
| }, | |
| { | |
| "epoch": 0.9, | |
| "learning_rate": 2.9010386564218615e-05, | |
| "loss": 2.2429, | |
| "step": 1578 | |
| }, | |
| { | |
| "epoch": 0.9, | |
| "learning_rate": 2.8702296843102715e-05, | |
| "loss": 2.2545, | |
| "step": 1579 | |
| }, | |
| { | |
| "epoch": 0.9, | |
| "learning_rate": 2.8395803471204684e-05, | |
| "loss": 2.3952, | |
| "step": 1580 | |
| }, | |
| { | |
| "epoch": 0.9, | |
| "learning_rate": 2.8090907486659322e-05, | |
| "loss": 2.3529, | |
| "step": 1581 | |
| }, | |
| { | |
| "epoch": 0.9, | |
| "learning_rate": 2.7787609922190427e-05, | |
| "loss": 2.2856, | |
| "step": 1582 | |
| }, | |
| { | |
| "epoch": 0.9, | |
| "learning_rate": 2.7485911805108178e-05, | |
| "loss": 2.3722, | |
| "step": 1583 | |
| }, | |
| { | |
| "epoch": 0.9, | |
| "learning_rate": 2.718581415730481e-05, | |
| "loss": 2.3342, | |
| "step": 1584 | |
| }, | |
| { | |
| "epoch": 0.9, | |
| "learning_rate": 2.688731799525196e-05, | |
| "loss": 2.3142, | |
| "step": 1585 | |
| }, | |
| { | |
| "epoch": 0.9, | |
| "learning_rate": 2.6590424329996367e-05, | |
| "loss": 2.2973, | |
| "step": 1586 | |
| }, | |
| { | |
| "epoch": 0.9, | |
| "learning_rate": 2.6295134167157342e-05, | |
| "loss": 2.3004, | |
| "step": 1587 | |
| }, | |
| { | |
| "epoch": 0.9, | |
| "learning_rate": 2.6001448506922544e-05, | |
| "loss": 2.3562, | |
| "step": 1588 | |
| }, | |
| { | |
| "epoch": 0.9, | |
| "learning_rate": 2.5709368344045248e-05, | |
| "loss": 2.2182, | |
| "step": 1589 | |
| }, | |
| { | |
| "epoch": 0.9, | |
| "learning_rate": 2.5418894667840364e-05, | |
| "loss": 2.2891, | |
| "step": 1590 | |
| }, | |
| { | |
| "epoch": 0.9, | |
| "learning_rate": 2.5130028462181876e-05, | |
| "loss": 2.2727, | |
| "step": 1591 | |
| }, | |
| { | |
| "epoch": 0.9, | |
| "learning_rate": 2.4842770705498563e-05, | |
| "loss": 2.3455, | |
| "step": 1592 | |
| }, | |
| { | |
| "epoch": 0.91, | |
| "learning_rate": 2.455712237077157e-05, | |
| "loss": 2.3612, | |
| "step": 1593 | |
| }, | |
| { | |
| "epoch": 0.91, | |
| "learning_rate": 2.427308442553028e-05, | |
| "loss": 2.2578, | |
| "step": 1594 | |
| }, | |
| { | |
| "epoch": 0.91, | |
| "learning_rate": 2.399065783185006e-05, | |
| "loss": 2.2582, | |
| "step": 1595 | |
| }, | |
| { | |
| "epoch": 0.91, | |
| "learning_rate": 2.3709843546347864e-05, | |
| "loss": 2.3917, | |
| "step": 1596 | |
| }, | |
| { | |
| "epoch": 0.91, | |
| "learning_rate": 2.3430642520179847e-05, | |
| "loss": 2.2371, | |
| "step": 1597 | |
| }, | |
| { | |
| "epoch": 0.91, | |
| "learning_rate": 2.3153055699037583e-05, | |
| "loss": 2.3884, | |
| "step": 1598 | |
| }, | |
| { | |
| "epoch": 0.91, | |
| "learning_rate": 2.2877084023145424e-05, | |
| "loss": 2.2932, | |
| "step": 1599 | |
| }, | |
| { | |
| "epoch": 0.91, | |
| "learning_rate": 2.26027284272568e-05, | |
| "loss": 2.3436, | |
| "step": 1600 | |
| }, | |
| { | |
| "epoch": 0.91, | |
| "learning_rate": 2.232998984065132e-05, | |
| "loss": 2.3251, | |
| "step": 1601 | |
| }, | |
| { | |
| "epoch": 0.91, | |
| "learning_rate": 2.2058869187131515e-05, | |
| "loss": 2.3943, | |
| "step": 1602 | |
| }, | |
| { | |
| "epoch": 0.91, | |
| "learning_rate": 2.1789367385019854e-05, | |
| "loss": 2.3229, | |
| "step": 1603 | |
| }, | |
| { | |
| "epoch": 0.91, | |
| "learning_rate": 2.152148534715537e-05, | |
| "loss": 2.3256, | |
| "step": 1604 | |
| }, | |
| { | |
| "epoch": 0.91, | |
| "learning_rate": 2.1255223980891027e-05, | |
| "loss": 2.3357, | |
| "step": 1605 | |
| }, | |
| { | |
| "epoch": 0.91, | |
| "learning_rate": 2.0990584188089966e-05, | |
| "loss": 2.36, | |
| "step": 1606 | |
| }, | |
| { | |
| "epoch": 0.91, | |
| "learning_rate": 2.0727566865123228e-05, | |
| "loss": 2.4062, | |
| "step": 1607 | |
| }, | |
| { | |
| "epoch": 0.91, | |
| "learning_rate": 2.046617290286601e-05, | |
| "loss": 2.3019, | |
| "step": 1608 | |
| }, | |
| { | |
| "epoch": 0.91, | |
| "learning_rate": 2.0206403186695198e-05, | |
| "loss": 2.3427, | |
| "step": 1609 | |
| }, | |
| { | |
| "epoch": 0.91, | |
| "learning_rate": 1.994825859648597e-05, | |
| "loss": 2.3014, | |
| "step": 1610 | |
| }, | |
| { | |
| "epoch": 0.92, | |
| "learning_rate": 1.969174000660906e-05, | |
| "loss": 2.3167, | |
| "step": 1611 | |
| }, | |
| { | |
| "epoch": 0.92, | |
| "learning_rate": 1.9436848285927787e-05, | |
| "loss": 2.2969, | |
| "step": 1612 | |
| }, | |
| { | |
| "epoch": 0.92, | |
| "learning_rate": 1.918358429779482e-05, | |
| "loss": 2.3046, | |
| "step": 1613 | |
| }, | |
| { | |
| "epoch": 0.92, | |
| "learning_rate": 1.893194890004979e-05, | |
| "loss": 2.3401, | |
| "step": 1614 | |
| }, | |
| { | |
| "epoch": 0.92, | |
| "learning_rate": 1.868194294501574e-05, | |
| "loss": 2.2342, | |
| "step": 1615 | |
| }, | |
| { | |
| "epoch": 0.92, | |
| "learning_rate": 1.843356727949691e-05, | |
| "loss": 2.3481, | |
| "step": 1616 | |
| }, | |
| { | |
| "epoch": 0.92, | |
| "learning_rate": 1.8186822744775234e-05, | |
| "loss": 2.2885, | |
| "step": 1617 | |
| }, | |
| { | |
| "epoch": 0.92, | |
| "learning_rate": 1.794171017660795e-05, | |
| "loss": 2.3167, | |
| "step": 1618 | |
| }, | |
| { | |
| "epoch": 0.92, | |
| "learning_rate": 1.7698230405224504e-05, | |
| "loss": 2.3262, | |
| "step": 1619 | |
| }, | |
| { | |
| "epoch": 0.92, | |
| "learning_rate": 1.745638425532392e-05, | |
| "loss": 2.358, | |
| "step": 1620 | |
| }, | |
| { | |
| "epoch": 0.92, | |
| "learning_rate": 1.7216172546071885e-05, | |
| "loss": 2.3507, | |
| "step": 1621 | |
| }, | |
| { | |
| "epoch": 0.92, | |
| "learning_rate": 1.6977596091098113e-05, | |
| "loss": 2.3153, | |
| "step": 1622 | |
| }, | |
| { | |
| "epoch": 0.92, | |
| "learning_rate": 1.6740655698493313e-05, | |
| "loss": 2.2792, | |
| "step": 1623 | |
| }, | |
| { | |
| "epoch": 0.92, | |
| "learning_rate": 1.6505352170806787e-05, | |
| "loss": 2.2654, | |
| "step": 1624 | |
| }, | |
| { | |
| "epoch": 0.92, | |
| "learning_rate": 1.627168630504339e-05, | |
| "loss": 2.3947, | |
| "step": 1625 | |
| }, | |
| { | |
| "epoch": 0.92, | |
| "learning_rate": 1.603965889266118e-05, | |
| "loss": 2.3774, | |
| "step": 1626 | |
| }, | |
| { | |
| "epoch": 0.92, | |
| "learning_rate": 1.5809270719568335e-05, | |
| "loss": 2.3866, | |
| "step": 1627 | |
| }, | |
| { | |
| "epoch": 0.92, | |
| "learning_rate": 1.5580522566121024e-05, | |
| "loss": 2.3637, | |
| "step": 1628 | |
| }, | |
| { | |
| "epoch": 0.93, | |
| "learning_rate": 1.535341520712008e-05, | |
| "loss": 2.3164, | |
| "step": 1629 | |
| }, | |
| { | |
| "epoch": 0.93, | |
| "learning_rate": 1.512794941180906e-05, | |
| "loss": 2.279, | |
| "step": 1630 | |
| }, | |
| { | |
| "epoch": 0.93, | |
| "learning_rate": 1.4904125943871028e-05, | |
| "loss": 2.3181, | |
| "step": 1631 | |
| }, | |
| { | |
| "epoch": 0.93, | |
| "learning_rate": 1.4681945561426547e-05, | |
| "loss": 2.3144, | |
| "step": 1632 | |
| }, | |
| { | |
| "epoch": 0.93, | |
| "learning_rate": 1.446140901703058e-05, | |
| "loss": 2.4334, | |
| "step": 1633 | |
| }, | |
| { | |
| "epoch": 0.93, | |
| "learning_rate": 1.4242517057670434e-05, | |
| "loss": 2.364, | |
| "step": 1634 | |
| }, | |
| { | |
| "epoch": 0.93, | |
| "learning_rate": 1.402527042476276e-05, | |
| "loss": 2.3259, | |
| "step": 1635 | |
| }, | |
| { | |
| "epoch": 0.93, | |
| "learning_rate": 1.380966985415144e-05, | |
| "loss": 2.3583, | |
| "step": 1636 | |
| }, | |
| { | |
| "epoch": 0.93, | |
| "learning_rate": 1.3595716076104658e-05, | |
| "loss": 2.3553, | |
| "step": 1637 | |
| }, | |
| { | |
| "epoch": 0.93, | |
| "learning_rate": 1.3383409815313108e-05, | |
| "loss": 2.2715, | |
| "step": 1638 | |
| }, | |
| { | |
| "epoch": 0.93, | |
| "learning_rate": 1.3172751790886672e-05, | |
| "loss": 2.3486, | |
| "step": 1639 | |
| }, | |
| { | |
| "epoch": 0.93, | |
| "learning_rate": 1.2963742716352812e-05, | |
| "loss": 2.3027, | |
| "step": 1640 | |
| }, | |
| { | |
| "epoch": 0.93, | |
| "learning_rate": 1.2756383299653452e-05, | |
| "loss": 2.3052, | |
| "step": 1641 | |
| }, | |
| { | |
| "epoch": 0.93, | |
| "learning_rate": 1.2550674243143156e-05, | |
| "loss": 2.3633, | |
| "step": 1642 | |
| }, | |
| { | |
| "epoch": 0.93, | |
| "learning_rate": 1.2346616243586294e-05, | |
| "loss": 2.3606, | |
| "step": 1643 | |
| }, | |
| { | |
| "epoch": 0.93, | |
| "learning_rate": 1.2144209992155042e-05, | |
| "loss": 2.3093, | |
| "step": 1644 | |
| }, | |
| { | |
| "epoch": 0.93, | |
| "learning_rate": 1.1943456174426825e-05, | |
| "loss": 2.364, | |
| "step": 1645 | |
| }, | |
| { | |
| "epoch": 0.94, | |
| "learning_rate": 1.1744355470381996e-05, | |
| "loss": 2.4326, | |
| "step": 1646 | |
| }, | |
| { | |
| "epoch": 0.94, | |
| "learning_rate": 1.1546908554401658e-05, | |
| "loss": 2.3876, | |
| "step": 1647 | |
| }, | |
| { | |
| "epoch": 0.94, | |
| "learning_rate": 1.135111609526529e-05, | |
| "loss": 2.3325, | |
| "step": 1648 | |
| }, | |
| { | |
| "epoch": 0.94, | |
| "learning_rate": 1.1156978756148462e-05, | |
| "loss": 2.3742, | |
| "step": 1649 | |
| }, | |
| { | |
| "epoch": 0.94, | |
| "learning_rate": 1.0964497194620726e-05, | |
| "loss": 2.3834, | |
| "step": 1650 | |
| }, | |
| { | |
| "epoch": 0.94, | |
| "learning_rate": 1.077367206264318e-05, | |
| "loss": 2.3848, | |
| "step": 1651 | |
| }, | |
| { | |
| "epoch": 0.94, | |
| "learning_rate": 1.0584504006566465e-05, | |
| "loss": 2.3192, | |
| "step": 1652 | |
| }, | |
| { | |
| "epoch": 0.94, | |
| "learning_rate": 1.0396993667128318e-05, | |
| "loss": 2.4382, | |
| "step": 1653 | |
| }, | |
| { | |
| "epoch": 0.94, | |
| "learning_rate": 1.0211141679451753e-05, | |
| "loss": 2.2944, | |
| "step": 1654 | |
| }, | |
| { | |
| "epoch": 0.94, | |
| "learning_rate": 1.0026948673042768e-05, | |
| "loss": 2.3364, | |
| "step": 1655 | |
| }, | |
| { | |
| "epoch": 0.94, | |
| "learning_rate": 9.84441527178781e-06, | |
| "loss": 2.368, | |
| "step": 1656 | |
| }, | |
| { | |
| "epoch": 0.94, | |
| "learning_rate": 9.663542093952483e-06, | |
| "loss": 2.3153, | |
| "step": 1657 | |
| }, | |
| { | |
| "epoch": 0.94, | |
| "learning_rate": 9.484329752178555e-06, | |
| "loss": 2.3558, | |
| "step": 1658 | |
| }, | |
| { | |
| "epoch": 0.94, | |
| "learning_rate": 9.306778853482744e-06, | |
| "loss": 2.348, | |
| "step": 1659 | |
| }, | |
| { | |
| "epoch": 0.94, | |
| "learning_rate": 9.130889999253822e-06, | |
| "loss": 2.3848, | |
| "step": 1660 | |
| }, | |
| { | |
| "epoch": 0.94, | |
| "learning_rate": 8.95666378525134e-06, | |
| "loss": 2.3773, | |
| "step": 1661 | |
| }, | |
| { | |
| "epoch": 0.94, | |
| "learning_rate": 8.784100801602912e-06, | |
| "loss": 2.3582, | |
| "step": 1662 | |
| }, | |
| { | |
| "epoch": 0.94, | |
| "learning_rate": 8.613201632802992e-06, | |
| "loss": 2.3574, | |
| "step": 1663 | |
| }, | |
| { | |
| "epoch": 0.95, | |
| "learning_rate": 8.443966857710095e-06, | |
| "loss": 2.3734, | |
| "step": 1664 | |
| }, | |
| { | |
| "epoch": 0.95, | |
| "learning_rate": 8.276397049545359e-06, | |
| "loss": 2.291, | |
| "step": 1665 | |
| }, | |
| { | |
| "epoch": 0.95, | |
| "learning_rate": 8.110492775890433e-06, | |
| "loss": 2.3514, | |
| "step": 1666 | |
| }, | |
| { | |
| "epoch": 0.95, | |
| "learning_rate": 7.946254598685643e-06, | |
| "loss": 2.3549, | |
| "step": 1667 | |
| }, | |
| { | |
| "epoch": 0.95, | |
| "learning_rate": 7.783683074227943e-06, | |
| "loss": 2.3524, | |
| "step": 1668 | |
| }, | |
| { | |
| "epoch": 0.95, | |
| "learning_rate": 7.622778753169079e-06, | |
| "loss": 2.3111, | |
| "step": 1669 | |
| }, | |
| { | |
| "epoch": 0.95, | |
| "learning_rate": 7.463542180513871e-06, | |
| "loss": 2.323, | |
| "step": 1670 | |
| }, | |
| { | |
| "epoch": 0.95, | |
| "learning_rate": 7.305973895618157e-06, | |
| "loss": 2.3242, | |
| "step": 1671 | |
| }, | |
| { | |
| "epoch": 0.95, | |
| "learning_rate": 7.150074432187015e-06, | |
| "loss": 2.3144, | |
| "step": 1672 | |
| }, | |
| { | |
| "epoch": 0.95, | |
| "learning_rate": 6.995844318273104e-06, | |
| "loss": 2.3182, | |
| "step": 1673 | |
| }, | |
| { | |
| "epoch": 0.95, | |
| "learning_rate": 6.8432840762747695e-06, | |
| "loss": 2.3319, | |
| "step": 1674 | |
| }, | |
| { | |
| "epoch": 0.95, | |
| "learning_rate": 6.692394222934272e-06, | |
| "loss": 2.3763, | |
| "step": 1675 | |
| }, | |
| { | |
| "epoch": 0.95, | |
| "learning_rate": 6.543175269335844e-06, | |
| "loss": 2.3976, | |
| "step": 1676 | |
| }, | |
| { | |
| "epoch": 0.95, | |
| "learning_rate": 6.395627720904518e-06, | |
| "loss": 2.341, | |
| "step": 1677 | |
| }, | |
| { | |
| "epoch": 0.95, | |
| "learning_rate": 6.249752077403636e-06, | |
| "loss": 2.3106, | |
| "step": 1678 | |
| }, | |
| { | |
| "epoch": 0.95, | |
| "learning_rate": 6.10554883293396e-06, | |
| "loss": 2.3996, | |
| "step": 1679 | |
| }, | |
| { | |
| "epoch": 0.95, | |
| "learning_rate": 5.963018475931281e-06, | |
| "loss": 2.3723, | |
| "step": 1680 | |
| }, | |
| { | |
| "epoch": 0.96, | |
| "learning_rate": 5.8221614891653115e-06, | |
| "loss": 2.3411, | |
| "step": 1681 | |
| }, | |
| { | |
| "epoch": 0.96, | |
| "learning_rate": 5.682978349737799e-06, | |
| "loss": 2.2581, | |
| "step": 1682 | |
| }, | |
| { | |
| "epoch": 0.96, | |
| "learning_rate": 5.54546952908086e-06, | |
| "loss": 2.3132, | |
| "step": 1683 | |
| }, | |
| { | |
| "epoch": 0.96, | |
| "learning_rate": 5.409635492955478e-06, | |
| "loss": 2.2714, | |
| "step": 1684 | |
| }, | |
| { | |
| "epoch": 0.96, | |
| "learning_rate": 5.275476701450066e-06, | |
| "loss": 2.3827, | |
| "step": 1685 | |
| }, | |
| { | |
| "epoch": 0.96, | |
| "learning_rate": 5.142993608978519e-06, | |
| "loss": 2.3675, | |
| "step": 1686 | |
| }, | |
| { | |
| "epoch": 0.96, | |
| "learning_rate": 5.012186664279106e-06, | |
| "loss": 2.3512, | |
| "step": 1687 | |
| }, | |
| { | |
| "epoch": 0.96, | |
| "learning_rate": 4.883056310412581e-06, | |
| "loss": 2.2707, | |
| "step": 1688 | |
| }, | |
| { | |
| "epoch": 0.96, | |
| "learning_rate": 4.755602984761132e-06, | |
| "loss": 2.3824, | |
| "step": 1689 | |
| }, | |
| { | |
| "epoch": 0.96, | |
| "learning_rate": 4.62982711902632e-06, | |
| "loss": 2.307, | |
| "step": 1690 | |
| }, | |
| { | |
| "epoch": 0.96, | |
| "learning_rate": 4.505729139228087e-06, | |
| "loss": 2.403, | |
| "step": 1691 | |
| }, | |
| { | |
| "epoch": 0.96, | |
| "learning_rate": 4.383309465703145e-06, | |
| "loss": 2.2792, | |
| "step": 1692 | |
| }, | |
| { | |
| "epoch": 0.96, | |
| "learning_rate": 4.262568513103415e-06, | |
| "loss": 2.2435, | |
| "step": 1693 | |
| }, | |
| { | |
| "epoch": 0.96, | |
| "learning_rate": 4.143506690395038e-06, | |
| "loss": 2.3489, | |
| "step": 1694 | |
| }, | |
| { | |
| "epoch": 0.96, | |
| "learning_rate": 4.026124400856479e-06, | |
| "loss": 2.3703, | |
| "step": 1695 | |
| }, | |
| { | |
| "epoch": 0.96, | |
| "learning_rate": 3.910422042077421e-06, | |
| "loss": 2.3224, | |
| "step": 1696 | |
| }, | |
| { | |
| "epoch": 0.96, | |
| "learning_rate": 3.79640000595749e-06, | |
| "loss": 2.4044, | |
| "step": 1697 | |
| }, | |
| { | |
| "epoch": 0.96, | |
| "learning_rate": 3.6840586787048045e-06, | |
| "loss": 2.3773, | |
| "step": 1698 | |
| }, | |
| { | |
| "epoch": 0.97, | |
| "learning_rate": 3.573398440834541e-06, | |
| "loss": 2.3467, | |
| "step": 1699 | |
| }, | |
| { | |
| "epoch": 0.97, | |
| "learning_rate": 3.464419667168095e-06, | |
| "loss": 2.3637, | |
| "step": 1700 | |
| }, | |
| { | |
| "epoch": 0.97, | |
| "learning_rate": 3.357122726831252e-06, | |
| "loss": 2.3537, | |
| "step": 1701 | |
| }, | |
| { | |
| "epoch": 0.97, | |
| "learning_rate": 3.2515079832533543e-06, | |
| "loss": 2.3654, | |
| "step": 1702 | |
| }, | |
| { | |
| "epoch": 0.97, | |
| "learning_rate": 3.1475757941658023e-06, | |
| "loss": 2.329, | |
| "step": 1703 | |
| }, | |
| { | |
| "epoch": 0.97, | |
| "learning_rate": 3.0453265116009987e-06, | |
| "loss": 2.3691, | |
| "step": 1704 | |
| }, | |
| { | |
| "epoch": 0.97, | |
| "learning_rate": 2.9447604818911844e-06, | |
| "loss": 2.3862, | |
| "step": 1705 | |
| }, | |
| { | |
| "epoch": 0.97, | |
| "learning_rate": 2.845878045667161e-06, | |
| "loss": 2.3251, | |
| "step": 1706 | |
| }, | |
| { | |
| "epoch": 0.97, | |
| "learning_rate": 2.748679537857013e-06, | |
| "loss": 2.3459, | |
| "step": 1707 | |
| }, | |
| { | |
| "epoch": 0.97, | |
| "learning_rate": 2.6531652876853884e-06, | |
| "loss": 2.2902, | |
| "step": 1708 | |
| }, | |
| { | |
| "epoch": 0.97, | |
| "learning_rate": 2.559335618671943e-06, | |
| "loss": 2.3358, | |
| "step": 1709 | |
| }, | |
| { | |
| "epoch": 0.97, | |
| "learning_rate": 2.4671908486305073e-06, | |
| "loss": 2.4531, | |
| "step": 1710 | |
| }, | |
| { | |
| "epoch": 0.97, | |
| "learning_rate": 2.3767312896679214e-06, | |
| "loss": 2.3742, | |
| "step": 1711 | |
| }, | |
| { | |
| "epoch": 0.97, | |
| "learning_rate": 2.2879572481829814e-06, | |
| "loss": 2.4129, | |
| "step": 1712 | |
| }, | |
| { | |
| "epoch": 0.97, | |
| "learning_rate": 2.200869024865326e-06, | |
| "loss": 2.2343, | |
| "step": 1713 | |
| }, | |
| { | |
| "epoch": 0.97, | |
| "learning_rate": 2.115466914694719e-06, | |
| "loss": 2.414, | |
| "step": 1714 | |
| }, | |
| { | |
| "epoch": 0.97, | |
| "learning_rate": 2.0317512069394917e-06, | |
| "loss": 2.3059, | |
| "step": 1715 | |
| }, | |
| { | |
| "epoch": 0.97, | |
| "learning_rate": 1.949722185156266e-06, | |
| "loss": 2.2761, | |
| "step": 1716 | |
| }, | |
| { | |
| "epoch": 0.98, | |
| "learning_rate": 1.8693801271884026e-06, | |
| "loss": 2.3328, | |
| "step": 1717 | |
| }, | |
| { | |
| "epoch": 0.98, | |
| "learning_rate": 1.7907253051653305e-06, | |
| "loss": 2.3489, | |
| "step": 1718 | |
| }, | |
| { | |
| "epoch": 0.98, | |
| "learning_rate": 1.7137579855016072e-06, | |
| "loss": 2.2278, | |
| "step": 1719 | |
| }, | |
| { | |
| "epoch": 0.98, | |
| "learning_rate": 1.6384784288959175e-06, | |
| "loss": 2.2491, | |
| "step": 1720 | |
| }, | |
| { | |
| "epoch": 0.98, | |
| "learning_rate": 1.5648868903304637e-06, | |
| "loss": 2.3181, | |
| "step": 1721 | |
| }, | |
| { | |
| "epoch": 0.98, | |
| "learning_rate": 1.4929836190696322e-06, | |
| "loss": 2.3198, | |
| "step": 1722 | |
| }, | |
| { | |
| "epoch": 0.98, | |
| "learning_rate": 1.4227688586596066e-06, | |
| "loss": 2.3803, | |
| "step": 1723 | |
| }, | |
| { | |
| "epoch": 0.98, | |
| "learning_rate": 1.3542428469273116e-06, | |
| "loss": 2.3766, | |
| "step": 1724 | |
| }, | |
| { | |
| "epoch": 0.98, | |
| "learning_rate": 1.2874058159796366e-06, | |
| "loss": 2.4288, | |
| "step": 1725 | |
| }, | |
| { | |
| "epoch": 0.98, | |
| "learning_rate": 1.2222579922026577e-06, | |
| "loss": 2.4122, | |
| "step": 1726 | |
| }, | |
| { | |
| "epoch": 0.98, | |
| "learning_rate": 1.1587995962608623e-06, | |
| "loss": 2.3623, | |
| "step": 1727 | |
| }, | |
| { | |
| "epoch": 0.98, | |
| "learning_rate": 1.0970308430964805e-06, | |
| "loss": 2.381, | |
| "step": 1728 | |
| }, | |
| { | |
| "epoch": 0.98, | |
| "learning_rate": 1.0369519419286543e-06, | |
| "loss": 2.4029, | |
| "step": 1729 | |
| }, | |
| { | |
| "epoch": 0.98, | |
| "learning_rate": 9.785630962527159e-07, | |
| "loss": 2.3803, | |
| "step": 1730 | |
| }, | |
| { | |
| "epoch": 0.98, | |
| "learning_rate": 9.218645038396312e-07, | |
| "loss": 2.3646, | |
| "step": 1731 | |
| }, | |
| { | |
| "epoch": 0.98, | |
| "learning_rate": 8.668563567352238e-07, | |
| "loss": 2.2831, | |
| "step": 1732 | |
| }, | |
| { | |
| "epoch": 0.98, | |
| "learning_rate": 8.135388412595646e-07, | |
| "loss": 2.3445, | |
| "step": 1733 | |
| }, | |
| { | |
| "epoch": 0.99, | |
| "learning_rate": 7.619121380063043e-07, | |
| "loss": 2.3601, | |
| "step": 1734 | |
| }, | |
| { | |
| "epoch": 0.99, | |
| "learning_rate": 7.119764218420644e-07, | |
| "loss": 2.299, | |
| "step": 1735 | |
| }, | |
| { | |
| "epoch": 0.99, | |
| "learning_rate": 6.637318619058807e-07, | |
| "loss": 2.3585, | |
| "step": 1736 | |
| }, | |
| { | |
| "epoch": 0.99, | |
| "learning_rate": 6.17178621608594e-07, | |
| "loss": 2.2829, | |
| "step": 1737 | |
| }, | |
| { | |
| "epoch": 0.99, | |
| "learning_rate": 5.723168586324046e-07, | |
| "loss": 2.3443, | |
| "step": 1738 | |
| }, | |
| { | |
| "epoch": 0.99, | |
| "learning_rate": 5.291467249302074e-07, | |
| "loss": 2.2769, | |
| "step": 1739 | |
| }, | |
| { | |
| "epoch": 0.99, | |
| "learning_rate": 4.876683667249804e-07, | |
| "loss": 2.31, | |
| "step": 1740 | |
| }, | |
| { | |
| "epoch": 0.99, | |
| "learning_rate": 4.4788192450961886e-07, | |
| "loss": 2.335, | |
| "step": 1741 | |
| }, | |
| { | |
| "epoch": 0.99, | |
| "learning_rate": 4.097875330462131e-07, | |
| "loss": 2.3004, | |
| "step": 1742 | |
| }, | |
| { | |
| "epoch": 0.99, | |
| "learning_rate": 3.733853213656047e-07, | |
| "loss": 2.3632, | |
| "step": 1743 | |
| }, | |
| { | |
| "epoch": 0.99, | |
| "learning_rate": 3.386754127670533e-07, | |
| "loss": 2.3078, | |
| "step": 1744 | |
| }, | |
| { | |
| "epoch": 0.99, | |
| "learning_rate": 3.056579248176816e-07, | |
| "loss": 2.3236, | |
| "step": 1745 | |
| }, | |
| { | |
| "epoch": 0.99, | |
| "learning_rate": 2.7433296935219785e-07, | |
| "loss": 2.3353, | |
| "step": 1746 | |
| }, | |
| { | |
| "epoch": 0.99, | |
| "learning_rate": 2.4470065247250706e-07, | |
| "loss": 2.4458, | |
| "step": 1747 | |
| }, | |
| { | |
| "epoch": 0.99, | |
| "learning_rate": 2.16761074547267e-07, | |
| "loss": 2.3492, | |
| "step": 1748 | |
| }, | |
| { | |
| "epoch": 0.99, | |
| "learning_rate": 1.905143302116108e-07, | |
| "loss": 2.3494, | |
| "step": 1749 | |
| }, | |
| { | |
| "epoch": 0.99, | |
| "learning_rate": 1.6596050836681365e-07, | |
| "loss": 2.2391, | |
| "step": 1750 | |
| }, | |
| { | |
| "epoch": 0.99, | |
| "learning_rate": 1.4309969218001538e-07, | |
| "loss": 2.3532, | |
| "step": 1751 | |
| }, | |
| { | |
| "epoch": 1.0, | |
| "learning_rate": 1.2193195908388744e-07, | |
| "loss": 2.294, | |
| "step": 1752 | |
| }, | |
| { | |
| "epoch": 1.0, | |
| "learning_rate": 1.0245738077641065e-07, | |
| "loss": 2.2277, | |
| "step": 1753 | |
| }, | |
| { | |
| "epoch": 1.0, | |
| "learning_rate": 8.467602322065338e-08, | |
| "loss": 2.3152, | |
| "step": 1754 | |
| }, | |
| { | |
| "epoch": 1.0, | |
| "learning_rate": 6.858794664449386e-08, | |
| "loss": 2.2876, | |
| "step": 1755 | |
| }, | |
| { | |
| "epoch": 1.0, | |
| "learning_rate": 5.419320554034268e-08, | |
| "loss": 2.3495, | |
| "step": 1756 | |
| }, | |
| { | |
| "epoch": 1.0, | |
| "learning_rate": 4.149184866519828e-08, | |
| "loss": 2.2936, | |
| "step": 1757 | |
| }, | |
| { | |
| "epoch": 1.0, | |
| "learning_rate": 3.048391904031389e-08, | |
| "loss": 2.2783, | |
| "step": 1758 | |
| }, | |
| { | |
| "epoch": 1.0, | |
| "learning_rate": 2.1169453950975472e-08, | |
| "loss": 2.3406, | |
| "step": 1759 | |
| }, | |
| { | |
| "epoch": 1.0, | |
| "learning_rate": 1.3548484946501737e-08, | |
| "loss": 2.3872, | |
| "step": 1760 | |
| }, | |
| { | |
| "epoch": 1.0, | |
| "step": 1760, | |
| "total_flos": 1.4654118649502106e+17, | |
| "train_loss": 2.4962065879594197, | |
| "train_runtime": 20089.3142, | |
| "train_samples_per_second": 5.607, | |
| "train_steps_per_second": 0.088 | |
| } | |
| ], | |
| "logging_steps": 1.0, | |
| "max_steps": 1760, | |
| "num_input_tokens_seen": 0, | |
| "num_train_epochs": 1, | |
| "save_steps": 20, | |
| "total_flos": 1.4654118649502106e+17, | |
| "train_batch_size": 2, | |
| "trial_name": null, | |
| "trial_params": null | |
| } | |