| { | |
| "best_metric": null, | |
| "best_model_checkpoint": null, | |
| "epoch": 0.9443985361822689, | |
| "eval_steps": 500, | |
| "global_step": 2000, | |
| "is_hyper_param_search": false, | |
| "is_local_process_zero": true, | |
| "is_world_process_zero": true, | |
| "log_history": [ | |
| { | |
| "epoch": 0.0, | |
| "learning_rate": 4.000000000000001e-06, | |
| "loss": 1.8444, | |
| "step": 1 | |
| }, | |
| { | |
| "epoch": 0.0, | |
| "learning_rate": 8.000000000000001e-06, | |
| "loss": 1.9183, | |
| "step": 2 | |
| }, | |
| { | |
| "epoch": 0.0, | |
| "learning_rate": 1.2e-05, | |
| "loss": 2.0161, | |
| "step": 3 | |
| }, | |
| { | |
| "epoch": 0.0, | |
| "learning_rate": 1.6000000000000003e-05, | |
| "loss": 2.0216, | |
| "step": 4 | |
| }, | |
| { | |
| "epoch": 0.0, | |
| "learning_rate": 2e-05, | |
| "loss": 1.9427, | |
| "step": 5 | |
| }, | |
| { | |
| "epoch": 0.0, | |
| "learning_rate": 2.4e-05, | |
| "loss": 1.9254, | |
| "step": 6 | |
| }, | |
| { | |
| "epoch": 0.0, | |
| "learning_rate": 2.8000000000000003e-05, | |
| "loss": 1.6781, | |
| "step": 7 | |
| }, | |
| { | |
| "epoch": 0.0, | |
| "learning_rate": 3.2000000000000005e-05, | |
| "loss": 1.703, | |
| "step": 8 | |
| }, | |
| { | |
| "epoch": 0.0, | |
| "learning_rate": 3.6e-05, | |
| "loss": 2.0178, | |
| "step": 9 | |
| }, | |
| { | |
| "epoch": 0.0, | |
| "learning_rate": 4e-05, | |
| "loss": 1.8788, | |
| "step": 10 | |
| }, | |
| { | |
| "epoch": 0.01, | |
| "learning_rate": 4.4000000000000006e-05, | |
| "loss": 2.0176, | |
| "step": 11 | |
| }, | |
| { | |
| "epoch": 0.01, | |
| "learning_rate": 4.8e-05, | |
| "loss": 1.659, | |
| "step": 12 | |
| }, | |
| { | |
| "epoch": 0.01, | |
| "learning_rate": 5.2000000000000004e-05, | |
| "loss": 1.4496, | |
| "step": 13 | |
| }, | |
| { | |
| "epoch": 0.01, | |
| "learning_rate": 5.6000000000000006e-05, | |
| "loss": 1.3411, | |
| "step": 14 | |
| }, | |
| { | |
| "epoch": 0.01, | |
| "learning_rate": 6e-05, | |
| "loss": 1.3288, | |
| "step": 15 | |
| }, | |
| { | |
| "epoch": 0.01, | |
| "learning_rate": 6.400000000000001e-05, | |
| "loss": 1.7064, | |
| "step": 16 | |
| }, | |
| { | |
| "epoch": 0.01, | |
| "learning_rate": 6.800000000000001e-05, | |
| "loss": 1.8597, | |
| "step": 17 | |
| }, | |
| { | |
| "epoch": 0.01, | |
| "learning_rate": 7.2e-05, | |
| "loss": 1.5418, | |
| "step": 18 | |
| }, | |
| { | |
| "epoch": 0.01, | |
| "learning_rate": 7.6e-05, | |
| "loss": 1.6277, | |
| "step": 19 | |
| }, | |
| { | |
| "epoch": 0.01, | |
| "learning_rate": 8e-05, | |
| "loss": 1.4583, | |
| "step": 20 | |
| }, | |
| { | |
| "epoch": 0.01, | |
| "learning_rate": 8.4e-05, | |
| "loss": 1.466, | |
| "step": 21 | |
| }, | |
| { | |
| "epoch": 0.01, | |
| "learning_rate": 8.800000000000001e-05, | |
| "loss": 1.7423, | |
| "step": 22 | |
| }, | |
| { | |
| "epoch": 0.01, | |
| "learning_rate": 9.200000000000001e-05, | |
| "loss": 1.5443, | |
| "step": 23 | |
| }, | |
| { | |
| "epoch": 0.01, | |
| "learning_rate": 9.6e-05, | |
| "loss": 1.751, | |
| "step": 24 | |
| }, | |
| { | |
| "epoch": 0.01, | |
| "learning_rate": 0.0001, | |
| "loss": 1.7061, | |
| "step": 25 | |
| }, | |
| { | |
| "epoch": 0.01, | |
| "learning_rate": 0.00010400000000000001, | |
| "loss": 1.5738, | |
| "step": 26 | |
| }, | |
| { | |
| "epoch": 0.01, | |
| "learning_rate": 0.00010800000000000001, | |
| "loss": 1.4467, | |
| "step": 27 | |
| }, | |
| { | |
| "epoch": 0.01, | |
| "learning_rate": 0.00011200000000000001, | |
| "loss": 1.4368, | |
| "step": 28 | |
| }, | |
| { | |
| "epoch": 0.01, | |
| "learning_rate": 0.000116, | |
| "loss": 1.3751, | |
| "step": 29 | |
| }, | |
| { | |
| "epoch": 0.01, | |
| "learning_rate": 0.00012, | |
| "loss": 1.2682, | |
| "step": 30 | |
| }, | |
| { | |
| "epoch": 0.01, | |
| "learning_rate": 0.000124, | |
| "loss": 1.685, | |
| "step": 31 | |
| }, | |
| { | |
| "epoch": 0.02, | |
| "learning_rate": 0.00012800000000000002, | |
| "loss": 1.3104, | |
| "step": 32 | |
| }, | |
| { | |
| "epoch": 0.02, | |
| "learning_rate": 0.000132, | |
| "loss": 1.21, | |
| "step": 33 | |
| }, | |
| { | |
| "epoch": 0.02, | |
| "learning_rate": 0.00013600000000000003, | |
| "loss": 1.7759, | |
| "step": 34 | |
| }, | |
| { | |
| "epoch": 0.02, | |
| "learning_rate": 0.00014, | |
| "loss": 1.4145, | |
| "step": 35 | |
| }, | |
| { | |
| "epoch": 0.02, | |
| "learning_rate": 0.000144, | |
| "loss": 1.5398, | |
| "step": 36 | |
| }, | |
| { | |
| "epoch": 0.02, | |
| "learning_rate": 0.000148, | |
| "loss": 1.6678, | |
| "step": 37 | |
| }, | |
| { | |
| "epoch": 0.02, | |
| "learning_rate": 0.000152, | |
| "loss": 1.5502, | |
| "step": 38 | |
| }, | |
| { | |
| "epoch": 0.02, | |
| "learning_rate": 0.00015600000000000002, | |
| "loss": 1.3757, | |
| "step": 39 | |
| }, | |
| { | |
| "epoch": 0.02, | |
| "learning_rate": 0.00016, | |
| "loss": 1.5165, | |
| "step": 40 | |
| }, | |
| { | |
| "epoch": 0.02, | |
| "learning_rate": 0.000164, | |
| "loss": 1.1687, | |
| "step": 41 | |
| }, | |
| { | |
| "epoch": 0.02, | |
| "learning_rate": 0.000168, | |
| "loss": 1.8161, | |
| "step": 42 | |
| }, | |
| { | |
| "epoch": 0.02, | |
| "learning_rate": 0.000172, | |
| "loss": 1.4016, | |
| "step": 43 | |
| }, | |
| { | |
| "epoch": 0.02, | |
| "learning_rate": 0.00017600000000000002, | |
| "loss": 1.4039, | |
| "step": 44 | |
| }, | |
| { | |
| "epoch": 0.02, | |
| "learning_rate": 0.00018, | |
| "loss": 1.4823, | |
| "step": 45 | |
| }, | |
| { | |
| "epoch": 0.02, | |
| "learning_rate": 0.00018400000000000003, | |
| "loss": 1.3028, | |
| "step": 46 | |
| }, | |
| { | |
| "epoch": 0.02, | |
| "learning_rate": 0.000188, | |
| "loss": 1.3179, | |
| "step": 47 | |
| }, | |
| { | |
| "epoch": 0.02, | |
| "learning_rate": 0.000192, | |
| "loss": 1.465, | |
| "step": 48 | |
| }, | |
| { | |
| "epoch": 0.02, | |
| "learning_rate": 0.000196, | |
| "loss": 1.3618, | |
| "step": 49 | |
| }, | |
| { | |
| "epoch": 0.02, | |
| "learning_rate": 0.0002, | |
| "loss": 1.4782, | |
| "step": 50 | |
| }, | |
| { | |
| "epoch": 0.02, | |
| "learning_rate": 0.00020400000000000003, | |
| "loss": 1.1542, | |
| "step": 51 | |
| }, | |
| { | |
| "epoch": 0.02, | |
| "learning_rate": 0.00020800000000000001, | |
| "loss": 1.3877, | |
| "step": 52 | |
| }, | |
| { | |
| "epoch": 0.03, | |
| "learning_rate": 0.00021200000000000003, | |
| "loss": 1.6806, | |
| "step": 53 | |
| }, | |
| { | |
| "epoch": 0.03, | |
| "learning_rate": 0.00021600000000000002, | |
| "loss": 1.4566, | |
| "step": 54 | |
| }, | |
| { | |
| "epoch": 0.03, | |
| "learning_rate": 0.00022000000000000003, | |
| "loss": 1.4927, | |
| "step": 55 | |
| }, | |
| { | |
| "epoch": 0.03, | |
| "learning_rate": 0.00022400000000000002, | |
| "loss": 1.7024, | |
| "step": 56 | |
| }, | |
| { | |
| "epoch": 0.03, | |
| "learning_rate": 0.00022799999999999999, | |
| "loss": 1.5621, | |
| "step": 57 | |
| }, | |
| { | |
| "epoch": 0.03, | |
| "learning_rate": 0.000232, | |
| "loss": 1.5714, | |
| "step": 58 | |
| }, | |
| { | |
| "epoch": 0.03, | |
| "learning_rate": 0.000236, | |
| "loss": 1.2905, | |
| "step": 59 | |
| }, | |
| { | |
| "epoch": 0.03, | |
| "learning_rate": 0.00024, | |
| "loss": 1.332, | |
| "step": 60 | |
| }, | |
| { | |
| "epoch": 0.03, | |
| "learning_rate": 0.000244, | |
| "loss": 1.5307, | |
| "step": 61 | |
| }, | |
| { | |
| "epoch": 0.03, | |
| "learning_rate": 0.000248, | |
| "loss": 1.4634, | |
| "step": 62 | |
| }, | |
| { | |
| "epoch": 0.03, | |
| "learning_rate": 0.000252, | |
| "loss": 1.3234, | |
| "step": 63 | |
| }, | |
| { | |
| "epoch": 0.03, | |
| "learning_rate": 0.00025600000000000004, | |
| "loss": 1.3457, | |
| "step": 64 | |
| }, | |
| { | |
| "epoch": 0.03, | |
| "learning_rate": 0.00026000000000000003, | |
| "loss": 1.6265, | |
| "step": 65 | |
| }, | |
| { | |
| "epoch": 0.03, | |
| "learning_rate": 0.000264, | |
| "loss": 1.4168, | |
| "step": 66 | |
| }, | |
| { | |
| "epoch": 0.03, | |
| "learning_rate": 0.000268, | |
| "loss": 1.6415, | |
| "step": 67 | |
| }, | |
| { | |
| "epoch": 0.03, | |
| "learning_rate": 0.00027200000000000005, | |
| "loss": 1.5993, | |
| "step": 68 | |
| }, | |
| { | |
| "epoch": 0.03, | |
| "learning_rate": 0.000276, | |
| "loss": 1.5743, | |
| "step": 69 | |
| }, | |
| { | |
| "epoch": 0.03, | |
| "learning_rate": 0.00028, | |
| "loss": 1.1849, | |
| "step": 70 | |
| }, | |
| { | |
| "epoch": 0.03, | |
| "learning_rate": 0.000284, | |
| "loss": 1.4831, | |
| "step": 71 | |
| }, | |
| { | |
| "epoch": 0.03, | |
| "learning_rate": 0.000288, | |
| "loss": 1.3906, | |
| "step": 72 | |
| }, | |
| { | |
| "epoch": 0.03, | |
| "learning_rate": 0.000292, | |
| "loss": 1.2823, | |
| "step": 73 | |
| }, | |
| { | |
| "epoch": 0.03, | |
| "learning_rate": 0.000296, | |
| "loss": 1.3114, | |
| "step": 74 | |
| }, | |
| { | |
| "epoch": 0.04, | |
| "learning_rate": 0.00030000000000000003, | |
| "loss": 1.8664, | |
| "step": 75 | |
| }, | |
| { | |
| "epoch": 0.04, | |
| "learning_rate": 0.000304, | |
| "loss": 1.3636, | |
| "step": 76 | |
| }, | |
| { | |
| "epoch": 0.04, | |
| "learning_rate": 0.000308, | |
| "loss": 1.5702, | |
| "step": 77 | |
| }, | |
| { | |
| "epoch": 0.04, | |
| "learning_rate": 0.00031200000000000005, | |
| "loss": 1.5398, | |
| "step": 78 | |
| }, | |
| { | |
| "epoch": 0.04, | |
| "learning_rate": 0.00031600000000000004, | |
| "loss": 1.3273, | |
| "step": 79 | |
| }, | |
| { | |
| "epoch": 0.04, | |
| "learning_rate": 0.00032, | |
| "loss": 1.6198, | |
| "step": 80 | |
| }, | |
| { | |
| "epoch": 0.04, | |
| "learning_rate": 0.000324, | |
| "loss": 1.3984, | |
| "step": 81 | |
| }, | |
| { | |
| "epoch": 0.04, | |
| "learning_rate": 0.000328, | |
| "loss": 1.4277, | |
| "step": 82 | |
| }, | |
| { | |
| "epoch": 0.04, | |
| "learning_rate": 0.000332, | |
| "loss": 1.3518, | |
| "step": 83 | |
| }, | |
| { | |
| "epoch": 0.04, | |
| "learning_rate": 0.000336, | |
| "loss": 1.6, | |
| "step": 84 | |
| }, | |
| { | |
| "epoch": 0.04, | |
| "learning_rate": 0.00034, | |
| "loss": 1.6434, | |
| "step": 85 | |
| }, | |
| { | |
| "epoch": 0.04, | |
| "learning_rate": 0.000344, | |
| "loss": 1.4813, | |
| "step": 86 | |
| }, | |
| { | |
| "epoch": 0.04, | |
| "learning_rate": 0.000348, | |
| "loss": 1.7128, | |
| "step": 87 | |
| }, | |
| { | |
| "epoch": 0.04, | |
| "learning_rate": 0.00035200000000000005, | |
| "loss": 1.1899, | |
| "step": 88 | |
| }, | |
| { | |
| "epoch": 0.04, | |
| "learning_rate": 0.00035600000000000003, | |
| "loss": 1.6948, | |
| "step": 89 | |
| }, | |
| { | |
| "epoch": 0.04, | |
| "learning_rate": 0.00036, | |
| "loss": 1.3608, | |
| "step": 90 | |
| }, | |
| { | |
| "epoch": 0.04, | |
| "learning_rate": 0.000364, | |
| "loss": 1.4245, | |
| "step": 91 | |
| }, | |
| { | |
| "epoch": 0.04, | |
| "learning_rate": 0.00036800000000000005, | |
| "loss": 1.4884, | |
| "step": 92 | |
| }, | |
| { | |
| "epoch": 0.04, | |
| "learning_rate": 0.00037200000000000004, | |
| "loss": 1.4071, | |
| "step": 93 | |
| }, | |
| { | |
| "epoch": 0.04, | |
| "learning_rate": 0.000376, | |
| "loss": 1.4908, | |
| "step": 94 | |
| }, | |
| { | |
| "epoch": 0.04, | |
| "learning_rate": 0.00038, | |
| "loss": 1.564, | |
| "step": 95 | |
| }, | |
| { | |
| "epoch": 0.05, | |
| "learning_rate": 0.000384, | |
| "loss": 1.3262, | |
| "step": 96 | |
| }, | |
| { | |
| "epoch": 0.05, | |
| "learning_rate": 0.000388, | |
| "loss": 1.4975, | |
| "step": 97 | |
| }, | |
| { | |
| "epoch": 0.05, | |
| "learning_rate": 0.000392, | |
| "loss": 1.1834, | |
| "step": 98 | |
| }, | |
| { | |
| "epoch": 0.05, | |
| "learning_rate": 0.00039600000000000003, | |
| "loss": 1.4296, | |
| "step": 99 | |
| }, | |
| { | |
| "epoch": 0.05, | |
| "learning_rate": 0.0004, | |
| "loss": 1.5633, | |
| "step": 100 | |
| }, | |
| { | |
| "epoch": 0.05, | |
| "learning_rate": 0.00039999975740163973, | |
| "loss": 1.5139, | |
| "step": 101 | |
| }, | |
| { | |
| "epoch": 0.05, | |
| "learning_rate": 0.0003999990296071474, | |
| "loss": 1.3188, | |
| "step": 102 | |
| }, | |
| { | |
| "epoch": 0.05, | |
| "learning_rate": 0.0003999978166182887, | |
| "loss": 1.495, | |
| "step": 103 | |
| }, | |
| { | |
| "epoch": 0.05, | |
| "learning_rate": 0.0003999961184380062, | |
| "loss": 1.6301, | |
| "step": 104 | |
| }, | |
| { | |
| "epoch": 0.05, | |
| "learning_rate": 0.0003999939350704197, | |
| "loss": 1.5732, | |
| "step": 105 | |
| }, | |
| { | |
| "epoch": 0.05, | |
| "learning_rate": 0.0003999912665208261, | |
| "loss": 1.5114, | |
| "step": 106 | |
| }, | |
| { | |
| "epoch": 0.05, | |
| "learning_rate": 0.0003999881127956992, | |
| "loss": 1.3505, | |
| "step": 107 | |
| }, | |
| { | |
| "epoch": 0.05, | |
| "learning_rate": 0.0003999844739026899, | |
| "loss": 1.4632, | |
| "step": 108 | |
| }, | |
| { | |
| "epoch": 0.05, | |
| "learning_rate": 0.00039998034985062604, | |
| "loss": 1.5271, | |
| "step": 109 | |
| }, | |
| { | |
| "epoch": 0.05, | |
| "learning_rate": 0.0003999757406495126, | |
| "loss": 1.3881, | |
| "step": 110 | |
| }, | |
| { | |
| "epoch": 0.05, | |
| "learning_rate": 0.00039997064631053133, | |
| "loss": 1.2799, | |
| "step": 111 | |
| }, | |
| { | |
| "epoch": 0.05, | |
| "learning_rate": 0.0003999650668460411, | |
| "loss": 1.5678, | |
| "step": 112 | |
| }, | |
| { | |
| "epoch": 0.05, | |
| "learning_rate": 0.00039995900226957745, | |
| "loss": 1.1719, | |
| "step": 113 | |
| }, | |
| { | |
| "epoch": 0.05, | |
| "learning_rate": 0.0003999524525958531, | |
| "loss": 1.4742, | |
| "step": 114 | |
| }, | |
| { | |
| "epoch": 0.05, | |
| "learning_rate": 0.00039994541784075746, | |
| "loss": 1.3277, | |
| "step": 115 | |
| }, | |
| { | |
| "epoch": 0.05, | |
| "learning_rate": 0.00039993789802135655, | |
| "loss": 1.2949, | |
| "step": 116 | |
| }, | |
| { | |
| "epoch": 0.06, | |
| "learning_rate": 0.0003999298931558936, | |
| "loss": 1.7286, | |
| "step": 117 | |
| }, | |
| { | |
| "epoch": 0.06, | |
| "learning_rate": 0.00039992140326378794, | |
| "loss": 1.7081, | |
| "step": 118 | |
| }, | |
| { | |
| "epoch": 0.06, | |
| "learning_rate": 0.0003999124283656362, | |
| "loss": 1.4714, | |
| "step": 119 | |
| }, | |
| { | |
| "epoch": 0.06, | |
| "learning_rate": 0.0003999029684832112, | |
| "loss": 1.3633, | |
| "step": 120 | |
| }, | |
| { | |
| "epoch": 0.06, | |
| "learning_rate": 0.0003998930236394625, | |
| "loss": 1.3825, | |
| "step": 121 | |
| }, | |
| { | |
| "epoch": 0.06, | |
| "learning_rate": 0.0003998825938585162, | |
| "loss": 1.5665, | |
| "step": 122 | |
| }, | |
| { | |
| "epoch": 0.06, | |
| "learning_rate": 0.00039987167916567454, | |
| "loss": 1.2232, | |
| "step": 123 | |
| }, | |
| { | |
| "epoch": 0.06, | |
| "learning_rate": 0.0003998602795874167, | |
| "loss": 1.5348, | |
| "step": 124 | |
| }, | |
| { | |
| "epoch": 0.06, | |
| "learning_rate": 0.00039984839515139767, | |
| "loss": 1.4528, | |
| "step": 125 | |
| }, | |
| { | |
| "epoch": 0.06, | |
| "learning_rate": 0.00039983602588644884, | |
| "loss": 1.4307, | |
| "step": 126 | |
| }, | |
| { | |
| "epoch": 0.06, | |
| "learning_rate": 0.000399823171822578, | |
| "loss": 1.246, | |
| "step": 127 | |
| }, | |
| { | |
| "epoch": 0.06, | |
| "learning_rate": 0.0003998098329909689, | |
| "loss": 1.1974, | |
| "step": 128 | |
| }, | |
| { | |
| "epoch": 0.06, | |
| "learning_rate": 0.0003997960094239812, | |
| "loss": 1.3772, | |
| "step": 129 | |
| }, | |
| { | |
| "epoch": 0.06, | |
| "learning_rate": 0.0003997817011551508, | |
| "loss": 1.5958, | |
| "step": 130 | |
| }, | |
| { | |
| "epoch": 0.06, | |
| "learning_rate": 0.00039976690821918917, | |
| "loss": 1.2651, | |
| "step": 131 | |
| }, | |
| { | |
| "epoch": 0.06, | |
| "learning_rate": 0.00039975163065198376, | |
| "loss": 1.4289, | |
| "step": 132 | |
| }, | |
| { | |
| "epoch": 0.06, | |
| "learning_rate": 0.0003997358684905979, | |
| "loss": 1.3508, | |
| "step": 133 | |
| }, | |
| { | |
| "epoch": 0.06, | |
| "learning_rate": 0.00039971962177326996, | |
| "loss": 1.3393, | |
| "step": 134 | |
| }, | |
| { | |
| "epoch": 0.06, | |
| "learning_rate": 0.0003997028905394146, | |
| "loss": 1.547, | |
| "step": 135 | |
| }, | |
| { | |
| "epoch": 0.06, | |
| "learning_rate": 0.0003996856748296212, | |
| "loss": 1.2613, | |
| "step": 136 | |
| }, | |
| { | |
| "epoch": 0.06, | |
| "learning_rate": 0.00039966797468565503, | |
| "loss": 1.6868, | |
| "step": 137 | |
| }, | |
| { | |
| "epoch": 0.07, | |
| "learning_rate": 0.00039964979015045626, | |
| "loss": 1.5064, | |
| "step": 138 | |
| }, | |
| { | |
| "epoch": 0.07, | |
| "learning_rate": 0.0003996311212681402, | |
| "loss": 1.2082, | |
| "step": 139 | |
| }, | |
| { | |
| "epoch": 0.07, | |
| "learning_rate": 0.0003996119680839973, | |
| "loss": 1.4344, | |
| "step": 140 | |
| }, | |
| { | |
| "epoch": 0.07, | |
| "learning_rate": 0.000399592330644493, | |
| "loss": 1.4068, | |
| "step": 141 | |
| }, | |
| { | |
| "epoch": 0.07, | |
| "learning_rate": 0.0003995722089972672, | |
| "loss": 1.0907, | |
| "step": 142 | |
| }, | |
| { | |
| "epoch": 0.07, | |
| "learning_rate": 0.00039955160319113484, | |
| "loss": 1.4394, | |
| "step": 143 | |
| }, | |
| { | |
| "epoch": 0.07, | |
| "learning_rate": 0.00039953051327608515, | |
| "loss": 1.2697, | |
| "step": 144 | |
| }, | |
| { | |
| "epoch": 0.07, | |
| "learning_rate": 0.00039950893930328204, | |
| "loss": 1.6047, | |
| "step": 145 | |
| }, | |
| { | |
| "epoch": 0.07, | |
| "learning_rate": 0.00039948688132506355, | |
| "loss": 1.5132, | |
| "step": 146 | |
| }, | |
| { | |
| "epoch": 0.07, | |
| "learning_rate": 0.0003994643393949419, | |
| "loss": 1.6163, | |
| "step": 147 | |
| }, | |
| { | |
| "epoch": 0.07, | |
| "learning_rate": 0.0003994413135676036, | |
| "loss": 1.5018, | |
| "step": 148 | |
| }, | |
| { | |
| "epoch": 0.07, | |
| "learning_rate": 0.0003994178038989088, | |
| "loss": 1.2945, | |
| "step": 149 | |
| }, | |
| { | |
| "epoch": 0.07, | |
| "learning_rate": 0.00039939381044589166, | |
| "loss": 1.7071, | |
| "step": 150 | |
| }, | |
| { | |
| "epoch": 0.07, | |
| "learning_rate": 0.0003993693332667598, | |
| "loss": 1.5062, | |
| "step": 151 | |
| }, | |
| { | |
| "epoch": 0.07, | |
| "learning_rate": 0.00039934437242089457, | |
| "loss": 1.3906, | |
| "step": 152 | |
| }, | |
| { | |
| "epoch": 0.07, | |
| "learning_rate": 0.0003993189279688505, | |
| "loss": 1.2134, | |
| "step": 153 | |
| }, | |
| { | |
| "epoch": 0.07, | |
| "learning_rate": 0.0003992929999723554, | |
| "loss": 1.3035, | |
| "step": 154 | |
| }, | |
| { | |
| "epoch": 0.07, | |
| "learning_rate": 0.0003992665884943102, | |
| "loss": 1.6205, | |
| "step": 155 | |
| }, | |
| { | |
| "epoch": 0.07, | |
| "learning_rate": 0.00039923969359878877, | |
| "loss": 1.6062, | |
| "step": 156 | |
| }, | |
| { | |
| "epoch": 0.07, | |
| "learning_rate": 0.00039921231535103757, | |
| "loss": 1.4719, | |
| "step": 157 | |
| }, | |
| { | |
| "epoch": 0.07, | |
| "learning_rate": 0.00039918445381747586, | |
| "loss": 1.4253, | |
| "step": 158 | |
| }, | |
| { | |
| "epoch": 0.08, | |
| "learning_rate": 0.00039915610906569523, | |
| "loss": 1.4565, | |
| "step": 159 | |
| }, | |
| { | |
| "epoch": 0.08, | |
| "learning_rate": 0.0003991272811644596, | |
| "loss": 1.4769, | |
| "step": 160 | |
| }, | |
| { | |
| "epoch": 0.08, | |
| "learning_rate": 0.00039909797018370497, | |
| "loss": 1.32, | |
| "step": 161 | |
| }, | |
| { | |
| "epoch": 0.08, | |
| "learning_rate": 0.00039906817619453935, | |
| "loss": 1.3666, | |
| "step": 162 | |
| }, | |
| { | |
| "epoch": 0.08, | |
| "learning_rate": 0.0003990378992692424, | |
| "loss": 1.4335, | |
| "step": 163 | |
| }, | |
| { | |
| "epoch": 0.08, | |
| "learning_rate": 0.0003990071394812654, | |
| "loss": 1.3242, | |
| "step": 164 | |
| }, | |
| { | |
| "epoch": 0.08, | |
| "learning_rate": 0.00039897589690523127, | |
| "loss": 1.3204, | |
| "step": 165 | |
| }, | |
| { | |
| "epoch": 0.08, | |
| "learning_rate": 0.00039894417161693386, | |
| "loss": 1.3507, | |
| "step": 166 | |
| }, | |
| { | |
| "epoch": 0.08, | |
| "learning_rate": 0.00039891196369333813, | |
| "loss": 1.6852, | |
| "step": 167 | |
| }, | |
| { | |
| "epoch": 0.08, | |
| "learning_rate": 0.0003988792732125801, | |
| "loss": 1.5667, | |
| "step": 168 | |
| }, | |
| { | |
| "epoch": 0.08, | |
| "learning_rate": 0.0003988461002539664, | |
| "loss": 1.4404, | |
| "step": 169 | |
| }, | |
| { | |
| "epoch": 0.08, | |
| "learning_rate": 0.0003988124448979739, | |
| "loss": 1.4452, | |
| "step": 170 | |
| }, | |
| { | |
| "epoch": 0.08, | |
| "learning_rate": 0.00039877830722625, | |
| "loss": 1.6917, | |
| "step": 171 | |
| }, | |
| { | |
| "epoch": 0.08, | |
| "learning_rate": 0.0003987436873216122, | |
| "loss": 1.7008, | |
| "step": 172 | |
| }, | |
| { | |
| "epoch": 0.08, | |
| "learning_rate": 0.0003987085852680478, | |
| "loss": 1.5154, | |
| "step": 173 | |
| }, | |
| { | |
| "epoch": 0.08, | |
| "learning_rate": 0.00039867300115071374, | |
| "loss": 1.3697, | |
| "step": 174 | |
| }, | |
| { | |
| "epoch": 0.08, | |
| "learning_rate": 0.00039863693505593655, | |
| "loss": 1.3099, | |
| "step": 175 | |
| }, | |
| { | |
| "epoch": 0.08, | |
| "learning_rate": 0.00039860038707121204, | |
| "loss": 1.3247, | |
| "step": 176 | |
| }, | |
| { | |
| "epoch": 0.08, | |
| "learning_rate": 0.00039856335728520495, | |
| "loss": 1.4999, | |
| "step": 177 | |
| }, | |
| { | |
| "epoch": 0.08, | |
| "learning_rate": 0.0003985258457877489, | |
| "loss": 1.4268, | |
| "step": 178 | |
| }, | |
| { | |
| "epoch": 0.08, | |
| "learning_rate": 0.0003984878526698463, | |
| "loss": 1.7714, | |
| "step": 179 | |
| }, | |
| { | |
| "epoch": 0.08, | |
| "learning_rate": 0.0003984493780236677, | |
| "loss": 1.356, | |
| "step": 180 | |
| }, | |
| { | |
| "epoch": 0.09, | |
| "learning_rate": 0.00039841042194255205, | |
| "loss": 1.2773, | |
| "step": 181 | |
| }, | |
| { | |
| "epoch": 0.09, | |
| "learning_rate": 0.0003983709845210061, | |
| "loss": 1.0558, | |
| "step": 182 | |
| }, | |
| { | |
| "epoch": 0.09, | |
| "learning_rate": 0.0003983310658547044, | |
| "loss": 1.7372, | |
| "step": 183 | |
| }, | |
| { | |
| "epoch": 0.09, | |
| "learning_rate": 0.000398290666040489, | |
| "loss": 1.5983, | |
| "step": 184 | |
| }, | |
| { | |
| "epoch": 0.09, | |
| "learning_rate": 0.0003982497851763692, | |
| "loss": 1.437, | |
| "step": 185 | |
| }, | |
| { | |
| "epoch": 0.09, | |
| "learning_rate": 0.0003982084233615213, | |
| "loss": 1.3883, | |
| "step": 186 | |
| }, | |
| { | |
| "epoch": 0.09, | |
| "learning_rate": 0.00039816658069628835, | |
| "loss": 1.5226, | |
| "step": 187 | |
| }, | |
| { | |
| "epoch": 0.09, | |
| "learning_rate": 0.00039812425728218, | |
| "loss": 1.4093, | |
| "step": 188 | |
| }, | |
| { | |
| "epoch": 0.09, | |
| "learning_rate": 0.00039808145322187214, | |
| "loss": 1.3972, | |
| "step": 189 | |
| }, | |
| { | |
| "epoch": 0.09, | |
| "learning_rate": 0.00039803816861920673, | |
| "loss": 1.1755, | |
| "step": 190 | |
| }, | |
| { | |
| "epoch": 0.09, | |
| "learning_rate": 0.00039799440357919153, | |
| "loss": 1.606, | |
| "step": 191 | |
| }, | |
| { | |
| "epoch": 0.09, | |
| "learning_rate": 0.0003979501582079997, | |
| "loss": 1.5801, | |
| "step": 192 | |
| }, | |
| { | |
| "epoch": 0.09, | |
| "learning_rate": 0.00039790543261296996, | |
| "loss": 1.4668, | |
| "step": 193 | |
| }, | |
| { | |
| "epoch": 0.09, | |
| "learning_rate": 0.0003978602269026057, | |
| "loss": 1.5086, | |
| "step": 194 | |
| }, | |
| { | |
| "epoch": 0.09, | |
| "learning_rate": 0.0003978145411865754, | |
| "loss": 1.3564, | |
| "step": 195 | |
| }, | |
| { | |
| "epoch": 0.09, | |
| "learning_rate": 0.00039776837557571176, | |
| "loss": 1.53, | |
| "step": 196 | |
| }, | |
| { | |
| "epoch": 0.09, | |
| "learning_rate": 0.0003977217301820117, | |
| "loss": 1.5908, | |
| "step": 197 | |
| }, | |
| { | |
| "epoch": 0.09, | |
| "learning_rate": 0.0003976746051186364, | |
| "loss": 1.1649, | |
| "step": 198 | |
| }, | |
| { | |
| "epoch": 0.09, | |
| "learning_rate": 0.0003976270004999103, | |
| "loss": 1.4091, | |
| "step": 199 | |
| }, | |
| { | |
| "epoch": 0.09, | |
| "learning_rate": 0.00039757891644132153, | |
| "loss": 1.5316, | |
| "step": 200 | |
| }, | |
| { | |
| "epoch": 0.09, | |
| "learning_rate": 0.0003975303530595212, | |
| "loss": 1.5858, | |
| "step": 201 | |
| }, | |
| { | |
| "epoch": 0.1, | |
| "learning_rate": 0.0003974813104723233, | |
| "loss": 1.6776, | |
| "step": 202 | |
| }, | |
| { | |
| "epoch": 0.1, | |
| "learning_rate": 0.00039743178879870426, | |
| "loss": 1.4326, | |
| "step": 203 | |
| }, | |
| { | |
| "epoch": 0.1, | |
| "learning_rate": 0.00039738178815880296, | |
| "loss": 1.2465, | |
| "step": 204 | |
| }, | |
| { | |
| "epoch": 0.1, | |
| "learning_rate": 0.00039733130867392004, | |
| "loss": 1.374, | |
| "step": 205 | |
| }, | |
| { | |
| "epoch": 0.1, | |
| "learning_rate": 0.00039728035046651797, | |
| "loss": 1.5635, | |
| "step": 206 | |
| }, | |
| { | |
| "epoch": 0.1, | |
| "learning_rate": 0.0003972289136602204, | |
| "loss": 1.3614, | |
| "step": 207 | |
| }, | |
| { | |
| "epoch": 0.1, | |
| "learning_rate": 0.0003971769983798124, | |
| "loss": 1.2695, | |
| "step": 208 | |
| }, | |
| { | |
| "epoch": 0.1, | |
| "learning_rate": 0.00039712460475123943, | |
| "loss": 1.4433, | |
| "step": 209 | |
| }, | |
| { | |
| "epoch": 0.1, | |
| "learning_rate": 0.00039707173290160766, | |
| "loss": 1.4809, | |
| "step": 210 | |
| }, | |
| { | |
| "epoch": 0.1, | |
| "learning_rate": 0.00039701838295918315, | |
| "loss": 1.3501, | |
| "step": 211 | |
| }, | |
| { | |
| "epoch": 0.1, | |
| "learning_rate": 0.0003969645550533922, | |
| "loss": 1.3736, | |
| "step": 212 | |
| }, | |
| { | |
| "epoch": 0.1, | |
| "learning_rate": 0.0003969102493148203, | |
| "loss": 1.1709, | |
| "step": 213 | |
| }, | |
| { | |
| "epoch": 0.1, | |
| "learning_rate": 0.00039685546587521237, | |
| "loss": 1.3236, | |
| "step": 214 | |
| }, | |
| { | |
| "epoch": 0.1, | |
| "learning_rate": 0.00039680020486747205, | |
| "loss": 1.2384, | |
| "step": 215 | |
| }, | |
| { | |
| "epoch": 0.1, | |
| "learning_rate": 0.00039674446642566175, | |
| "loss": 1.3349, | |
| "step": 216 | |
| }, | |
| { | |
| "epoch": 0.1, | |
| "learning_rate": 0.0003966882506850018, | |
| "loss": 1.3416, | |
| "step": 217 | |
| }, | |
| { | |
| "epoch": 0.1, | |
| "learning_rate": 0.0003966315577818709, | |
| "loss": 1.132, | |
| "step": 218 | |
| }, | |
| { | |
| "epoch": 0.1, | |
| "learning_rate": 0.000396574387853805, | |
| "loss": 1.4271, | |
| "step": 219 | |
| }, | |
| { | |
| "epoch": 0.1, | |
| "learning_rate": 0.0003965167410394975, | |
| "loss": 1.5336, | |
| "step": 220 | |
| }, | |
| { | |
| "epoch": 0.1, | |
| "learning_rate": 0.00039645861747879843, | |
| "loss": 1.1742, | |
| "step": 221 | |
| }, | |
| { | |
| "epoch": 0.1, | |
| "learning_rate": 0.0003964000173127148, | |
| "loss": 1.2881, | |
| "step": 222 | |
| }, | |
| { | |
| "epoch": 0.11, | |
| "learning_rate": 0.0003963409406834095, | |
| "loss": 1.3789, | |
| "step": 223 | |
| }, | |
| { | |
| "epoch": 0.11, | |
| "learning_rate": 0.00039628138773420147, | |
| "loss": 1.5167, | |
| "step": 224 | |
| }, | |
| { | |
| "epoch": 0.11, | |
| "learning_rate": 0.0003962213586095653, | |
| "loss": 1.4953, | |
| "step": 225 | |
| }, | |
| { | |
| "epoch": 0.11, | |
| "learning_rate": 0.0003961608534551307, | |
| "loss": 1.7612, | |
| "step": 226 | |
| }, | |
| { | |
| "epoch": 0.11, | |
| "learning_rate": 0.000396099872417682, | |
| "loss": 1.2886, | |
| "step": 227 | |
| }, | |
| { | |
| "epoch": 0.11, | |
| "learning_rate": 0.0003960384156451583, | |
| "loss": 1.3367, | |
| "step": 228 | |
| }, | |
| { | |
| "epoch": 0.11, | |
| "learning_rate": 0.0003959764832866527, | |
| "loss": 1.6526, | |
| "step": 229 | |
| }, | |
| { | |
| "epoch": 0.11, | |
| "learning_rate": 0.0003959140754924121, | |
| "loss": 1.3393, | |
| "step": 230 | |
| }, | |
| { | |
| "epoch": 0.11, | |
| "learning_rate": 0.0003958511924138368, | |
| "loss": 1.6188, | |
| "step": 231 | |
| }, | |
| { | |
| "epoch": 0.11, | |
| "learning_rate": 0.0003957878342034801, | |
| "loss": 1.163, | |
| "step": 232 | |
| }, | |
| { | |
| "epoch": 0.11, | |
| "learning_rate": 0.00039572400101504793, | |
| "loss": 0.9372, | |
| "step": 233 | |
| }, | |
| { | |
| "epoch": 0.11, | |
| "learning_rate": 0.00039565969300339867, | |
| "loss": 1.3556, | |
| "step": 234 | |
| }, | |
| { | |
| "epoch": 0.11, | |
| "learning_rate": 0.0003955949103245424, | |
| "loss": 1.3058, | |
| "step": 235 | |
| }, | |
| { | |
| "epoch": 0.11, | |
| "learning_rate": 0.00039552965313564083, | |
| "loss": 1.3413, | |
| "step": 236 | |
| }, | |
| { | |
| "epoch": 0.11, | |
| "learning_rate": 0.000395463921595007, | |
| "loss": 1.6436, | |
| "step": 237 | |
| }, | |
| { | |
| "epoch": 0.11, | |
| "learning_rate": 0.0003953977158621043, | |
| "loss": 1.3746, | |
| "step": 238 | |
| }, | |
| { | |
| "epoch": 0.11, | |
| "learning_rate": 0.00039533103609754695, | |
| "loss": 1.4178, | |
| "step": 239 | |
| }, | |
| { | |
| "epoch": 0.11, | |
| "learning_rate": 0.0003952638824630989, | |
| "loss": 1.5444, | |
| "step": 240 | |
| }, | |
| { | |
| "epoch": 0.11, | |
| "learning_rate": 0.00039519625512167373, | |
| "loss": 1.6739, | |
| "step": 241 | |
| }, | |
| { | |
| "epoch": 0.11, | |
| "learning_rate": 0.0003951281542373343, | |
| "loss": 1.3911, | |
| "step": 242 | |
| }, | |
| { | |
| "epoch": 0.11, | |
| "learning_rate": 0.0003950595799752922, | |
| "loss": 1.2932, | |
| "step": 243 | |
| }, | |
| { | |
| "epoch": 0.12, | |
| "learning_rate": 0.00039499053250190754, | |
| "loss": 1.5476, | |
| "step": 244 | |
| }, | |
| { | |
| "epoch": 0.12, | |
| "learning_rate": 0.00039492101198468826, | |
| "loss": 1.5088, | |
| "step": 245 | |
| }, | |
| { | |
| "epoch": 0.12, | |
| "learning_rate": 0.0003948510185922901, | |
| "loss": 1.4024, | |
| "step": 246 | |
| }, | |
| { | |
| "epoch": 0.12, | |
| "learning_rate": 0.00039478055249451575, | |
| "loss": 1.4, | |
| "step": 247 | |
| }, | |
| { | |
| "epoch": 0.12, | |
| "learning_rate": 0.00039470961386231496, | |
| "loss": 1.2615, | |
| "step": 248 | |
| }, | |
| { | |
| "epoch": 0.12, | |
| "learning_rate": 0.0003946382028677836, | |
| "loss": 1.4472, | |
| "step": 249 | |
| }, | |
| { | |
| "epoch": 0.12, | |
| "learning_rate": 0.00039456631968416364, | |
| "loss": 1.4282, | |
| "step": 250 | |
| }, | |
| { | |
| "epoch": 0.12, | |
| "learning_rate": 0.00039449396448584237, | |
| "loss": 1.2253, | |
| "step": 251 | |
| }, | |
| { | |
| "epoch": 0.12, | |
| "learning_rate": 0.0003944211374483524, | |
| "loss": 1.6736, | |
| "step": 252 | |
| }, | |
| { | |
| "epoch": 0.12, | |
| "learning_rate": 0.00039434783874837097, | |
| "loss": 1.5025, | |
| "step": 253 | |
| }, | |
| { | |
| "epoch": 0.12, | |
| "learning_rate": 0.00039427406856371944, | |
| "loss": 1.5902, | |
| "step": 254 | |
| }, | |
| { | |
| "epoch": 0.12, | |
| "learning_rate": 0.0003941998270733631, | |
| "loss": 1.1451, | |
| "step": 255 | |
| }, | |
| { | |
| "epoch": 0.12, | |
| "learning_rate": 0.00039412511445741067, | |
| "loss": 1.5581, | |
| "step": 256 | |
| }, | |
| { | |
| "epoch": 0.12, | |
| "learning_rate": 0.0003940499308971136, | |
| "loss": 1.5489, | |
| "step": 257 | |
| }, | |
| { | |
| "epoch": 0.12, | |
| "learning_rate": 0.00039397427657486604, | |
| "loss": 1.5395, | |
| "step": 258 | |
| }, | |
| { | |
| "epoch": 0.12, | |
| "learning_rate": 0.0003938981516742041, | |
| "loss": 1.5695, | |
| "step": 259 | |
| }, | |
| { | |
| "epoch": 0.12, | |
| "learning_rate": 0.0003938215563798057, | |
| "loss": 1.4453, | |
| "step": 260 | |
| }, | |
| { | |
| "epoch": 0.12, | |
| "learning_rate": 0.0003937444908774895, | |
| "loss": 1.505, | |
| "step": 261 | |
| }, | |
| { | |
| "epoch": 0.12, | |
| "learning_rate": 0.0003936669553542153, | |
| "loss": 1.4602, | |
| "step": 262 | |
| }, | |
| { | |
| "epoch": 0.12, | |
| "learning_rate": 0.0003935889499980831, | |
| "loss": 1.4522, | |
| "step": 263 | |
| }, | |
| { | |
| "epoch": 0.12, | |
| "learning_rate": 0.00039351047499833235, | |
| "loss": 1.4439, | |
| "step": 264 | |
| }, | |
| { | |
| "epoch": 0.13, | |
| "learning_rate": 0.0003934315305453423, | |
| "loss": 1.6039, | |
| "step": 265 | |
| }, | |
| { | |
| "epoch": 0.13, | |
| "learning_rate": 0.0003933521168306309, | |
| "loss": 1.2475, | |
| "step": 266 | |
| }, | |
| { | |
| "epoch": 0.13, | |
| "learning_rate": 0.0003932722340468545, | |
| "loss": 1.4151, | |
| "step": 267 | |
| }, | |
| { | |
| "epoch": 0.13, | |
| "learning_rate": 0.0003931918823878074, | |
| "loss": 1.3802, | |
| "step": 268 | |
| }, | |
| { | |
| "epoch": 0.13, | |
| "learning_rate": 0.0003931110620484215, | |
| "loss": 1.4119, | |
| "step": 269 | |
| }, | |
| { | |
| "epoch": 0.13, | |
| "learning_rate": 0.0003930297732247654, | |
| "loss": 1.2728, | |
| "step": 270 | |
| }, | |
| { | |
| "epoch": 0.13, | |
| "learning_rate": 0.0003929480161140446, | |
| "loss": 1.4114, | |
| "step": 271 | |
| }, | |
| { | |
| "epoch": 0.13, | |
| "learning_rate": 0.0003928657909146006, | |
| "loss": 1.4245, | |
| "step": 272 | |
| }, | |
| { | |
| "epoch": 0.13, | |
| "learning_rate": 0.00039278309782591014, | |
| "loss": 1.5001, | |
| "step": 273 | |
| }, | |
| { | |
| "epoch": 0.13, | |
| "learning_rate": 0.0003926999370485856, | |
| "loss": 1.2046, | |
| "step": 274 | |
| }, | |
| { | |
| "epoch": 0.13, | |
| "learning_rate": 0.00039261630878437343, | |
| "loss": 1.409, | |
| "step": 275 | |
| }, | |
| { | |
| "epoch": 0.13, | |
| "learning_rate": 0.00039253221323615444, | |
| "loss": 1.4574, | |
| "step": 276 | |
| }, | |
| { | |
| "epoch": 0.13, | |
| "learning_rate": 0.00039244765060794315, | |
| "loss": 1.5181, | |
| "step": 277 | |
| }, | |
| { | |
| "epoch": 0.13, | |
| "learning_rate": 0.0003923626211048871, | |
| "loss": 1.2098, | |
| "step": 278 | |
| }, | |
| { | |
| "epoch": 0.13, | |
| "learning_rate": 0.00039227712493326644, | |
| "loss": 1.5624, | |
| "step": 279 | |
| }, | |
| { | |
| "epoch": 0.13, | |
| "learning_rate": 0.0003921911623004935, | |
| "loss": 1.5203, | |
| "step": 280 | |
| }, | |
| { | |
| "epoch": 0.13, | |
| "learning_rate": 0.0003921047334151121, | |
| "loss": 1.2721, | |
| "step": 281 | |
| }, | |
| { | |
| "epoch": 0.13, | |
| "learning_rate": 0.0003920178384867975, | |
| "loss": 1.2674, | |
| "step": 282 | |
| }, | |
| { | |
| "epoch": 0.13, | |
| "learning_rate": 0.0003919304777263553, | |
| "loss": 1.574, | |
| "step": 283 | |
| }, | |
| { | |
| "epoch": 0.13, | |
| "learning_rate": 0.0003918426513457211, | |
| "loss": 1.5655, | |
| "step": 284 | |
| }, | |
| { | |
| "epoch": 0.13, | |
| "learning_rate": 0.0003917543595579605, | |
| "loss": 1.4375, | |
| "step": 285 | |
| }, | |
| { | |
| "epoch": 0.14, | |
| "learning_rate": 0.0003916656025772678, | |
| "loss": 1.4467, | |
| "step": 286 | |
| }, | |
| { | |
| "epoch": 0.14, | |
| "learning_rate": 0.000391576380618966, | |
| "loss": 1.1648, | |
| "step": 287 | |
| }, | |
| { | |
| "epoch": 0.14, | |
| "learning_rate": 0.0003914866938995062, | |
| "loss": 1.4204, | |
| "step": 288 | |
| }, | |
| { | |
| "epoch": 0.14, | |
| "learning_rate": 0.00039139654263646676, | |
| "loss": 1.0229, | |
| "step": 289 | |
| }, | |
| { | |
| "epoch": 0.14, | |
| "learning_rate": 0.00039130592704855326, | |
| "loss": 1.3691, | |
| "step": 290 | |
| }, | |
| { | |
| "epoch": 0.14, | |
| "learning_rate": 0.00039121484735559755, | |
| "loss": 1.4937, | |
| "step": 291 | |
| }, | |
| { | |
| "epoch": 0.14, | |
| "learning_rate": 0.0003911233037785576, | |
| "loss": 1.2192, | |
| "step": 292 | |
| }, | |
| { | |
| "epoch": 0.14, | |
| "learning_rate": 0.00039103129653951654, | |
| "loss": 1.1896, | |
| "step": 293 | |
| }, | |
| { | |
| "epoch": 0.14, | |
| "learning_rate": 0.0003909388258616824, | |
| "loss": 1.1806, | |
| "step": 294 | |
| }, | |
| { | |
| "epoch": 0.14, | |
| "learning_rate": 0.00039084589196938753, | |
| "loss": 1.4472, | |
| "step": 295 | |
| }, | |
| { | |
| "epoch": 0.14, | |
| "learning_rate": 0.0003907524950880881, | |
| "loss": 1.2835, | |
| "step": 296 | |
| }, | |
| { | |
| "epoch": 0.14, | |
| "learning_rate": 0.00039065863544436347, | |
| "loss": 1.6074, | |
| "step": 297 | |
| }, | |
| { | |
| "epoch": 0.14, | |
| "learning_rate": 0.0003905643132659154, | |
| "loss": 1.6836, | |
| "step": 298 | |
| }, | |
| { | |
| "epoch": 0.14, | |
| "learning_rate": 0.0003904695287815681, | |
| "loss": 1.1548, | |
| "step": 299 | |
| }, | |
| { | |
| "epoch": 0.14, | |
| "learning_rate": 0.0003903742822212671, | |
| "loss": 1.3484, | |
| "step": 300 | |
| }, | |
| { | |
| "epoch": 0.14, | |
| "learning_rate": 0.000390278573816079, | |
| "loss": 1.568, | |
| "step": 301 | |
| }, | |
| { | |
| "epoch": 0.14, | |
| "learning_rate": 0.0003901824037981909, | |
| "loss": 1.4065, | |
| "step": 302 | |
| }, | |
| { | |
| "epoch": 0.14, | |
| "learning_rate": 0.0003900857724009096, | |
| "loss": 1.4226, | |
| "step": 303 | |
| }, | |
| { | |
| "epoch": 0.14, | |
| "learning_rate": 0.0003899886798586613, | |
| "loss": 1.4137, | |
| "step": 304 | |
| }, | |
| { | |
| "epoch": 0.14, | |
| "learning_rate": 0.00038989112640699095, | |
| "loss": 1.2969, | |
| "step": 305 | |
| }, | |
| { | |
| "epoch": 0.14, | |
| "learning_rate": 0.0003897931122825616, | |
| "loss": 1.4033, | |
| "step": 306 | |
| }, | |
| { | |
| "epoch": 0.14, | |
| "learning_rate": 0.0003896946377231539, | |
| "loss": 1.5774, | |
| "step": 307 | |
| }, | |
| { | |
| "epoch": 0.15, | |
| "learning_rate": 0.00038959570296766554, | |
| "loss": 1.4064, | |
| "step": 308 | |
| }, | |
| { | |
| "epoch": 0.15, | |
| "learning_rate": 0.0003894963082561106, | |
| "loss": 1.6194, | |
| "step": 309 | |
| }, | |
| { | |
| "epoch": 0.15, | |
| "learning_rate": 0.000389396453829619, | |
| "loss": 1.3165, | |
| "step": 310 | |
| }, | |
| { | |
| "epoch": 0.15, | |
| "learning_rate": 0.000389296139930436, | |
| "loss": 1.5399, | |
| "step": 311 | |
| }, | |
| { | |
| "epoch": 0.15, | |
| "learning_rate": 0.0003891953668019214, | |
| "loss": 1.3126, | |
| "step": 312 | |
| }, | |
| { | |
| "epoch": 0.15, | |
| "learning_rate": 0.00038909413468854925, | |
| "loss": 1.4554, | |
| "step": 313 | |
| }, | |
| { | |
| "epoch": 0.15, | |
| "learning_rate": 0.0003889924438359069, | |
| "loss": 1.1586, | |
| "step": 314 | |
| }, | |
| { | |
| "epoch": 0.15, | |
| "learning_rate": 0.0003888902944906947, | |
| "loss": 1.0335, | |
| "step": 315 | |
| }, | |
| { | |
| "epoch": 0.15, | |
| "learning_rate": 0.0003887876869007254, | |
| "loss": 1.5808, | |
| "step": 316 | |
| }, | |
| { | |
| "epoch": 0.15, | |
| "learning_rate": 0.00038868462131492326, | |
| "loss": 1.3421, | |
| "step": 317 | |
| }, | |
| { | |
| "epoch": 0.15, | |
| "learning_rate": 0.00038858109798332364, | |
| "loss": 1.6144, | |
| "step": 318 | |
| }, | |
| { | |
| "epoch": 0.15, | |
| "learning_rate": 0.00038847711715707254, | |
| "loss": 1.5857, | |
| "step": 319 | |
| }, | |
| { | |
| "epoch": 0.15, | |
| "learning_rate": 0.00038837267908842563, | |
| "loss": 1.3044, | |
| "step": 320 | |
| }, | |
| { | |
| "epoch": 0.15, | |
| "learning_rate": 0.00038826778403074807, | |
| "loss": 1.3907, | |
| "step": 321 | |
| }, | |
| { | |
| "epoch": 0.15, | |
| "learning_rate": 0.00038816243223851354, | |
| "loss": 1.3099, | |
| "step": 322 | |
| }, | |
| { | |
| "epoch": 0.15, | |
| "learning_rate": 0.0003880566239673036, | |
| "loss": 1.4291, | |
| "step": 323 | |
| }, | |
| { | |
| "epoch": 0.15, | |
| "learning_rate": 0.0003879503594738076, | |
| "loss": 1.3069, | |
| "step": 324 | |
| }, | |
| { | |
| "epoch": 0.15, | |
| "learning_rate": 0.00038784363901582134, | |
| "loss": 1.3345, | |
| "step": 325 | |
| }, | |
| { | |
| "epoch": 0.15, | |
| "learning_rate": 0.00038773646285224694, | |
| "loss": 1.3937, | |
| "step": 326 | |
| }, | |
| { | |
| "epoch": 0.15, | |
| "learning_rate": 0.000387628831243092, | |
| "loss": 1.2333, | |
| "step": 327 | |
| }, | |
| { | |
| "epoch": 0.15, | |
| "learning_rate": 0.00038752074444946895, | |
| "loss": 1.2367, | |
| "step": 328 | |
| }, | |
| { | |
| "epoch": 0.16, | |
| "learning_rate": 0.0003874122027335948, | |
| "loss": 1.4973, | |
| "step": 329 | |
| }, | |
| { | |
| "epoch": 0.16, | |
| "learning_rate": 0.00038730320635878973, | |
| "loss": 1.2619, | |
| "step": 330 | |
| }, | |
| { | |
| "epoch": 0.16, | |
| "learning_rate": 0.0003871937555894773, | |
| "loss": 1.3699, | |
| "step": 331 | |
| }, | |
| { | |
| "epoch": 0.16, | |
| "learning_rate": 0.0003870838506911832, | |
| "loss": 1.2068, | |
| "step": 332 | |
| }, | |
| { | |
| "epoch": 0.16, | |
| "learning_rate": 0.000386973491930535, | |
| "loss": 1.4411, | |
| "step": 333 | |
| }, | |
| { | |
| "epoch": 0.16, | |
| "learning_rate": 0.0003868626795752612, | |
| "loss": 1.4592, | |
| "step": 334 | |
| }, | |
| { | |
| "epoch": 0.16, | |
| "learning_rate": 0.0003867514138941908, | |
| "loss": 1.4266, | |
| "step": 335 | |
| }, | |
| { | |
| "epoch": 0.16, | |
| "learning_rate": 0.00038663969515725244, | |
| "loss": 1.2687, | |
| "step": 336 | |
| }, | |
| { | |
| "epoch": 0.16, | |
| "learning_rate": 0.000386527523635474, | |
| "loss": 1.5515, | |
| "step": 337 | |
| }, | |
| { | |
| "epoch": 0.16, | |
| "learning_rate": 0.0003864148996009817, | |
| "loss": 1.4121, | |
| "step": 338 | |
| }, | |
| { | |
| "epoch": 0.16, | |
| "learning_rate": 0.00038630182332699965, | |
| "loss": 1.4067, | |
| "step": 339 | |
| }, | |
| { | |
| "epoch": 0.16, | |
| "learning_rate": 0.00038618829508784905, | |
| "loss": 1.3161, | |
| "step": 340 | |
| }, | |
| { | |
| "epoch": 0.16, | |
| "learning_rate": 0.00038607431515894747, | |
| "loss": 1.3532, | |
| "step": 341 | |
| }, | |
| { | |
| "epoch": 0.16, | |
| "learning_rate": 0.00038595988381680843, | |
| "loss": 1.5935, | |
| "step": 342 | |
| }, | |
| { | |
| "epoch": 0.16, | |
| "learning_rate": 0.0003858450013390404, | |
| "loss": 1.5683, | |
| "step": 343 | |
| }, | |
| { | |
| "epoch": 0.16, | |
| "learning_rate": 0.0003857296680043465, | |
| "loss": 1.2895, | |
| "step": 344 | |
| }, | |
| { | |
| "epoch": 0.16, | |
| "learning_rate": 0.00038561388409252346, | |
| "loss": 1.4423, | |
| "step": 345 | |
| }, | |
| { | |
| "epoch": 0.16, | |
| "learning_rate": 0.00038549764988446113, | |
| "loss": 1.5754, | |
| "step": 346 | |
| }, | |
| { | |
| "epoch": 0.16, | |
| "learning_rate": 0.0003853809656621418, | |
| "loss": 1.4581, | |
| "step": 347 | |
| }, | |
| { | |
| "epoch": 0.16, | |
| "learning_rate": 0.00038526383170863946, | |
| "loss": 1.4425, | |
| "step": 348 | |
| }, | |
| { | |
| "epoch": 0.16, | |
| "learning_rate": 0.00038514624830811924, | |
| "loss": 1.0757, | |
| "step": 349 | |
| }, | |
| { | |
| "epoch": 0.17, | |
| "learning_rate": 0.0003850282157458365, | |
| "loss": 1.5043, | |
| "step": 350 | |
| }, | |
| { | |
| "epoch": 0.17, | |
| "learning_rate": 0.0003849097343081362, | |
| "loss": 1.4347, | |
| "step": 351 | |
| }, | |
| { | |
| "epoch": 0.17, | |
| "learning_rate": 0.00038479080428245247, | |
| "loss": 1.4784, | |
| "step": 352 | |
| }, | |
| { | |
| "epoch": 0.17, | |
| "learning_rate": 0.0003846714259573077, | |
| "loss": 1.2782, | |
| "step": 353 | |
| }, | |
| { | |
| "epoch": 0.17, | |
| "learning_rate": 0.00038455159962231156, | |
| "loss": 1.3142, | |
| "step": 354 | |
| }, | |
| { | |
| "epoch": 0.17, | |
| "learning_rate": 0.00038443132556816083, | |
| "loss": 1.4472, | |
| "step": 355 | |
| }, | |
| { | |
| "epoch": 0.17, | |
| "learning_rate": 0.0003843106040866384, | |
| "loss": 1.5361, | |
| "step": 356 | |
| }, | |
| { | |
| "epoch": 0.17, | |
| "learning_rate": 0.0003841894354706127, | |
| "loss": 1.345, | |
| "step": 357 | |
| }, | |
| { | |
| "epoch": 0.17, | |
| "learning_rate": 0.00038406782001403663, | |
| "loss": 1.4903, | |
| "step": 358 | |
| }, | |
| { | |
| "epoch": 0.17, | |
| "learning_rate": 0.00038394575801194744, | |
| "loss": 1.2769, | |
| "step": 359 | |
| }, | |
| { | |
| "epoch": 0.17, | |
| "learning_rate": 0.0003838232497604655, | |
| "loss": 1.3184, | |
| "step": 360 | |
| }, | |
| { | |
| "epoch": 0.17, | |
| "learning_rate": 0.0003837002955567938, | |
| "loss": 1.338, | |
| "step": 361 | |
| }, | |
| { | |
| "epoch": 0.17, | |
| "learning_rate": 0.00038357689569921716, | |
| "loss": 1.5696, | |
| "step": 362 | |
| }, | |
| { | |
| "epoch": 0.17, | |
| "learning_rate": 0.0003834530504871018, | |
| "loss": 1.0948, | |
| "step": 363 | |
| }, | |
| { | |
| "epoch": 0.17, | |
| "learning_rate": 0.000383328760220894, | |
| "loss": 1.2939, | |
| "step": 364 | |
| }, | |
| { | |
| "epoch": 0.17, | |
| "learning_rate": 0.00038320402520211996, | |
| "loss": 1.5023, | |
| "step": 365 | |
| }, | |
| { | |
| "epoch": 0.17, | |
| "learning_rate": 0.0003830788457333848, | |
| "loss": 1.272, | |
| "step": 366 | |
| }, | |
| { | |
| "epoch": 0.17, | |
| "learning_rate": 0.0003829532221183719, | |
| "loss": 1.2972, | |
| "step": 367 | |
| }, | |
| { | |
| "epoch": 0.17, | |
| "learning_rate": 0.00038282715466184206, | |
| "loss": 1.3987, | |
| "step": 368 | |
| }, | |
| { | |
| "epoch": 0.17, | |
| "learning_rate": 0.0003827006436696328, | |
| "loss": 1.3171, | |
| "step": 369 | |
| }, | |
| { | |
| "epoch": 0.17, | |
| "learning_rate": 0.00038257368944865776, | |
| "loss": 1.2435, | |
| "step": 370 | |
| }, | |
| { | |
| "epoch": 0.18, | |
| "learning_rate": 0.00038244629230690583, | |
| "loss": 1.4025, | |
| "step": 371 | |
| }, | |
| { | |
| "epoch": 0.18, | |
| "learning_rate": 0.00038231845255344034, | |
| "loss": 1.378, | |
| "step": 372 | |
| }, | |
| { | |
| "epoch": 0.18, | |
| "learning_rate": 0.0003821901704983984, | |
| "loss": 1.5586, | |
| "step": 373 | |
| }, | |
| { | |
| "epoch": 0.18, | |
| "learning_rate": 0.00038206144645299025, | |
| "loss": 1.2798, | |
| "step": 374 | |
| }, | |
| { | |
| "epoch": 0.18, | |
| "learning_rate": 0.0003819322807294984, | |
| "loss": 1.4547, | |
| "step": 375 | |
| }, | |
| { | |
| "epoch": 0.18, | |
| "learning_rate": 0.00038180267364127654, | |
| "loss": 1.3677, | |
| "step": 376 | |
| }, | |
| { | |
| "epoch": 0.18, | |
| "learning_rate": 0.0003816726255027494, | |
| "loss": 1.4257, | |
| "step": 377 | |
| }, | |
| { | |
| "epoch": 0.18, | |
| "learning_rate": 0.00038154213662941185, | |
| "loss": 1.3593, | |
| "step": 378 | |
| }, | |
| { | |
| "epoch": 0.18, | |
| "learning_rate": 0.0003814112073378276, | |
| "loss": 1.3271, | |
| "step": 379 | |
| }, | |
| { | |
| "epoch": 0.18, | |
| "learning_rate": 0.0003812798379456289, | |
| "loss": 1.5296, | |
| "step": 380 | |
| }, | |
| { | |
| "epoch": 0.18, | |
| "learning_rate": 0.0003811480287715158, | |
| "loss": 1.5515, | |
| "step": 381 | |
| }, | |
| { | |
| "epoch": 0.18, | |
| "learning_rate": 0.0003810157801352553, | |
| "loss": 1.4145, | |
| "step": 382 | |
| }, | |
| { | |
| "epoch": 0.18, | |
| "learning_rate": 0.00038088309235768025, | |
| "loss": 1.2181, | |
| "step": 383 | |
| }, | |
| { | |
| "epoch": 0.18, | |
| "learning_rate": 0.0003807499657606891, | |
| "loss": 1.4692, | |
| "step": 384 | |
| }, | |
| { | |
| "epoch": 0.18, | |
| "learning_rate": 0.00038061640066724473, | |
| "loss": 1.3907, | |
| "step": 385 | |
| }, | |
| { | |
| "epoch": 0.18, | |
| "learning_rate": 0.0003804823974013741, | |
| "loss": 1.2358, | |
| "step": 386 | |
| }, | |
| { | |
| "epoch": 0.18, | |
| "learning_rate": 0.0003803479562881666, | |
| "loss": 1.5738, | |
| "step": 387 | |
| }, | |
| { | |
| "epoch": 0.18, | |
| "learning_rate": 0.00038021307765377443, | |
| "loss": 1.3619, | |
| "step": 388 | |
| }, | |
| { | |
| "epoch": 0.18, | |
| "learning_rate": 0.0003800777618254108, | |
| "loss": 1.4637, | |
| "step": 389 | |
| }, | |
| { | |
| "epoch": 0.18, | |
| "learning_rate": 0.0003799420091313498, | |
| "loss": 1.3023, | |
| "step": 390 | |
| }, | |
| { | |
| "epoch": 0.18, | |
| "learning_rate": 0.00037980581990092513, | |
| "loss": 1.3733, | |
| "step": 391 | |
| }, | |
| { | |
| "epoch": 0.19, | |
| "learning_rate": 0.0003796691944645296, | |
| "loss": 1.5788, | |
| "step": 392 | |
| }, | |
| { | |
| "epoch": 0.19, | |
| "learning_rate": 0.0003795321331536145, | |
| "loss": 1.4521, | |
| "step": 393 | |
| }, | |
| { | |
| "epoch": 0.19, | |
| "learning_rate": 0.0003793946363006881, | |
| "loss": 1.3638, | |
| "step": 394 | |
| }, | |
| { | |
| "epoch": 0.19, | |
| "learning_rate": 0.0003792567042393156, | |
| "loss": 1.3267, | |
| "step": 395 | |
| }, | |
| { | |
| "epoch": 0.19, | |
| "learning_rate": 0.0003791183373041179, | |
| "loss": 1.5984, | |
| "step": 396 | |
| }, | |
| { | |
| "epoch": 0.19, | |
| "learning_rate": 0.00037897953583077097, | |
| "loss": 1.678, | |
| "step": 397 | |
| }, | |
| { | |
| "epoch": 0.19, | |
| "learning_rate": 0.0003788403001560048, | |
| "loss": 1.284, | |
| "step": 398 | |
| }, | |
| { | |
| "epoch": 0.19, | |
| "learning_rate": 0.00037870063061760294, | |
| "loss": 1.3238, | |
| "step": 399 | |
| }, | |
| { | |
| "epoch": 0.19, | |
| "learning_rate": 0.0003785605275544014, | |
| "loss": 1.4447, | |
| "step": 400 | |
| }, | |
| { | |
| "epoch": 0.19, | |
| "learning_rate": 0.0003784199913062879, | |
| "loss": 1.4148, | |
| "step": 401 | |
| }, | |
| { | |
| "epoch": 0.19, | |
| "learning_rate": 0.000378279022214201, | |
| "loss": 1.3677, | |
| "step": 402 | |
| }, | |
| { | |
| "epoch": 0.19, | |
| "learning_rate": 0.0003781376206201295, | |
| "loss": 1.3583, | |
| "step": 403 | |
| }, | |
| { | |
| "epoch": 0.19, | |
| "learning_rate": 0.00037799578686711135, | |
| "loss": 1.3407, | |
| "step": 404 | |
| }, | |
| { | |
| "epoch": 0.19, | |
| "learning_rate": 0.00037785352129923283, | |
| "loss": 1.4658, | |
| "step": 405 | |
| }, | |
| { | |
| "epoch": 0.19, | |
| "learning_rate": 0.00037771082426162793, | |
| "loss": 1.4504, | |
| "step": 406 | |
| }, | |
| { | |
| "epoch": 0.19, | |
| "learning_rate": 0.0003775676961004774, | |
| "loss": 1.2846, | |
| "step": 407 | |
| }, | |
| { | |
| "epoch": 0.19, | |
| "learning_rate": 0.00037742413716300763, | |
| "loss": 1.4006, | |
| "step": 408 | |
| }, | |
| { | |
| "epoch": 0.19, | |
| "learning_rate": 0.00037728014779749036, | |
| "loss": 1.3782, | |
| "step": 409 | |
| }, | |
| { | |
| "epoch": 0.19, | |
| "learning_rate": 0.00037713572835324145, | |
| "loss": 1.332, | |
| "step": 410 | |
| }, | |
| { | |
| "epoch": 0.19, | |
| "learning_rate": 0.00037699087918062, | |
| "loss": 1.4053, | |
| "step": 411 | |
| }, | |
| { | |
| "epoch": 0.19, | |
| "learning_rate": 0.0003768456006310279, | |
| "loss": 1.3307, | |
| "step": 412 | |
| }, | |
| { | |
| "epoch": 0.2, | |
| "learning_rate": 0.00037669989305690835, | |
| "loss": 1.5948, | |
| "step": 413 | |
| }, | |
| { | |
| "epoch": 0.2, | |
| "learning_rate": 0.00037655375681174566, | |
| "loss": 1.6806, | |
| "step": 414 | |
| }, | |
| { | |
| "epoch": 0.2, | |
| "learning_rate": 0.0003764071922500639, | |
| "loss": 1.3207, | |
| "step": 415 | |
| }, | |
| { | |
| "epoch": 0.2, | |
| "learning_rate": 0.0003762601997274263, | |
| "loss": 1.326, | |
| "step": 416 | |
| }, | |
| { | |
| "epoch": 0.2, | |
| "learning_rate": 0.0003761127796004343, | |
| "loss": 1.3651, | |
| "step": 417 | |
| }, | |
| { | |
| "epoch": 0.2, | |
| "learning_rate": 0.00037596493222672676, | |
| "loss": 1.2577, | |
| "step": 418 | |
| }, | |
| { | |
| "epoch": 0.2, | |
| "learning_rate": 0.00037581665796497895, | |
| "loss": 1.4533, | |
| "step": 419 | |
| }, | |
| { | |
| "epoch": 0.2, | |
| "learning_rate": 0.0003756679571749018, | |
| "loss": 1.2807, | |
| "step": 420 | |
| }, | |
| { | |
| "epoch": 0.2, | |
| "learning_rate": 0.00037551883021724096, | |
| "loss": 1.3671, | |
| "step": 421 | |
| }, | |
| { | |
| "epoch": 0.2, | |
| "learning_rate": 0.00037536927745377606, | |
| "loss": 1.4806, | |
| "step": 422 | |
| }, | |
| { | |
| "epoch": 0.2, | |
| "learning_rate": 0.0003752192992473196, | |
| "loss": 1.4572, | |
| "step": 423 | |
| }, | |
| { | |
| "epoch": 0.2, | |
| "learning_rate": 0.00037506889596171627, | |
| "loss": 1.4408, | |
| "step": 424 | |
| }, | |
| { | |
| "epoch": 0.2, | |
| "learning_rate": 0.00037491806796184196, | |
| "loss": 1.1899, | |
| "step": 425 | |
| }, | |
| { | |
| "epoch": 0.2, | |
| "learning_rate": 0.00037476681561360287, | |
| "loss": 1.2646, | |
| "step": 426 | |
| }, | |
| { | |
| "epoch": 0.2, | |
| "learning_rate": 0.0003746151392839349, | |
| "loss": 1.0926, | |
| "step": 427 | |
| }, | |
| { | |
| "epoch": 0.2, | |
| "learning_rate": 0.0003744630393408021, | |
| "loss": 1.4036, | |
| "step": 428 | |
| }, | |
| { | |
| "epoch": 0.2, | |
| "learning_rate": 0.00037431051615319656, | |
| "loss": 1.384, | |
| "step": 429 | |
| }, | |
| { | |
| "epoch": 0.2, | |
| "learning_rate": 0.000374157570091137, | |
| "loss": 1.3298, | |
| "step": 430 | |
| }, | |
| { | |
| "epoch": 0.2, | |
| "learning_rate": 0.0003740042015256681, | |
| "loss": 1.4192, | |
| "step": 431 | |
| }, | |
| { | |
| "epoch": 0.2, | |
| "learning_rate": 0.0003738504108288595, | |
| "loss": 1.241, | |
| "step": 432 | |
| }, | |
| { | |
| "epoch": 0.2, | |
| "learning_rate": 0.00037369619837380487, | |
| "loss": 1.6417, | |
| "step": 433 | |
| }, | |
| { | |
| "epoch": 0.2, | |
| "learning_rate": 0.0003735415645346211, | |
| "loss": 1.0922, | |
| "step": 434 | |
| }, | |
| { | |
| "epoch": 0.21, | |
| "learning_rate": 0.00037338650968644736, | |
| "loss": 1.331, | |
| "step": 435 | |
| }, | |
| { | |
| "epoch": 0.21, | |
| "learning_rate": 0.00037323103420544416, | |
| "loss": 1.4702, | |
| "step": 436 | |
| }, | |
| { | |
| "epoch": 0.21, | |
| "learning_rate": 0.0003730751384687924, | |
| "loss": 1.1575, | |
| "step": 437 | |
| }, | |
| { | |
| "epoch": 0.21, | |
| "learning_rate": 0.0003729188228546927, | |
| "loss": 1.466, | |
| "step": 438 | |
| }, | |
| { | |
| "epoch": 0.21, | |
| "learning_rate": 0.00037276208774236416, | |
| "loss": 1.1979, | |
| "step": 439 | |
| }, | |
| { | |
| "epoch": 0.21, | |
| "learning_rate": 0.0003726049335120435, | |
| "loss": 1.2145, | |
| "step": 440 | |
| }, | |
| { | |
| "epoch": 0.21, | |
| "learning_rate": 0.0003724473605449844, | |
| "loss": 1.3891, | |
| "step": 441 | |
| }, | |
| { | |
| "epoch": 0.21, | |
| "learning_rate": 0.0003722893692234562, | |
| "loss": 1.1398, | |
| "step": 442 | |
| }, | |
| { | |
| "epoch": 0.21, | |
| "learning_rate": 0.0003721309599307434, | |
| "loss": 1.3276, | |
| "step": 443 | |
| }, | |
| { | |
| "epoch": 0.21, | |
| "learning_rate": 0.00037197213305114414, | |
| "loss": 1.3792, | |
| "step": 444 | |
| }, | |
| { | |
| "epoch": 0.21, | |
| "learning_rate": 0.00037181288896997, | |
| "loss": 1.3913, | |
| "step": 445 | |
| }, | |
| { | |
| "epoch": 0.21, | |
| "learning_rate": 0.0003716532280735445, | |
| "loss": 1.5605, | |
| "step": 446 | |
| }, | |
| { | |
| "epoch": 0.21, | |
| "learning_rate": 0.00037149315074920227, | |
| "loss": 1.3072, | |
| "step": 447 | |
| }, | |
| { | |
| "epoch": 0.21, | |
| "learning_rate": 0.0003713326573852883, | |
| "loss": 1.4979, | |
| "step": 448 | |
| }, | |
| { | |
| "epoch": 0.21, | |
| "learning_rate": 0.0003711717483711569, | |
| "loss": 1.5968, | |
| "step": 449 | |
| }, | |
| { | |
| "epoch": 0.21, | |
| "learning_rate": 0.0003710104240971707, | |
| "loss": 1.7091, | |
| "step": 450 | |
| }, | |
| { | |
| "epoch": 0.21, | |
| "learning_rate": 0.0003708486849546997, | |
| "loss": 1.4488, | |
| "step": 451 | |
| }, | |
| { | |
| "epoch": 0.21, | |
| "learning_rate": 0.00037068653133612045, | |
| "loss": 1.097, | |
| "step": 452 | |
| }, | |
| { | |
| "epoch": 0.21, | |
| "learning_rate": 0.0003705239636348149, | |
| "loss": 1.5878, | |
| "step": 453 | |
| }, | |
| { | |
| "epoch": 0.21, | |
| "learning_rate": 0.00037036098224516967, | |
| "loss": 1.5458, | |
| "step": 454 | |
| }, | |
| { | |
| "epoch": 0.21, | |
| "learning_rate": 0.00037019758756257494, | |
| "loss": 1.1706, | |
| "step": 455 | |
| }, | |
| { | |
| "epoch": 0.22, | |
| "learning_rate": 0.0003700337799834236, | |
| "loss": 1.5924, | |
| "step": 456 | |
| }, | |
| { | |
| "epoch": 0.22, | |
| "learning_rate": 0.00036986955990511, | |
| "loss": 1.15, | |
| "step": 457 | |
| }, | |
| { | |
| "epoch": 0.22, | |
| "learning_rate": 0.00036970492772602946, | |
| "loss": 1.485, | |
| "step": 458 | |
| }, | |
| { | |
| "epoch": 0.22, | |
| "learning_rate": 0.0003695398838455769, | |
| "loss": 1.3393, | |
| "step": 459 | |
| }, | |
| { | |
| "epoch": 0.22, | |
| "learning_rate": 0.0003693744286641462, | |
| "loss": 1.5324, | |
| "step": 460 | |
| }, | |
| { | |
| "epoch": 0.22, | |
| "learning_rate": 0.00036920856258312873, | |
| "loss": 1.4272, | |
| "step": 461 | |
| }, | |
| { | |
| "epoch": 0.22, | |
| "learning_rate": 0.000369042286004913, | |
| "loss": 1.3321, | |
| "step": 462 | |
| }, | |
| { | |
| "epoch": 0.22, | |
| "learning_rate": 0.0003688755993328832, | |
| "loss": 1.423, | |
| "step": 463 | |
| }, | |
| { | |
| "epoch": 0.22, | |
| "learning_rate": 0.00036870850297141846, | |
| "loss": 1.6408, | |
| "step": 464 | |
| }, | |
| { | |
| "epoch": 0.22, | |
| "learning_rate": 0.0003685409973258919, | |
| "loss": 1.4626, | |
| "step": 465 | |
| }, | |
| { | |
| "epoch": 0.22, | |
| "learning_rate": 0.0003683730828026694, | |
| "loss": 1.5387, | |
| "step": 466 | |
| }, | |
| { | |
| "epoch": 0.22, | |
| "learning_rate": 0.0003682047598091089, | |
| "loss": 1.1369, | |
| "step": 467 | |
| }, | |
| { | |
| "epoch": 0.22, | |
| "learning_rate": 0.00036803602875355913, | |
| "loss": 1.553, | |
| "step": 468 | |
| }, | |
| { | |
| "epoch": 0.22, | |
| "learning_rate": 0.00036786689004535887, | |
| "loss": 1.6675, | |
| "step": 469 | |
| }, | |
| { | |
| "epoch": 0.22, | |
| "learning_rate": 0.00036769734409483596, | |
| "loss": 1.4323, | |
| "step": 470 | |
| }, | |
| { | |
| "epoch": 0.22, | |
| "learning_rate": 0.00036752739131330596, | |
| "loss": 1.4708, | |
| "step": 471 | |
| }, | |
| { | |
| "epoch": 0.22, | |
| "learning_rate": 0.00036735703211307166, | |
| "loss": 1.2672, | |
| "step": 472 | |
| }, | |
| { | |
| "epoch": 0.22, | |
| "learning_rate": 0.0003671862669074216, | |
| "loss": 1.6278, | |
| "step": 473 | |
| }, | |
| { | |
| "epoch": 0.22, | |
| "learning_rate": 0.0003670150961106294, | |
| "loss": 1.4408, | |
| "step": 474 | |
| }, | |
| { | |
| "epoch": 0.22, | |
| "learning_rate": 0.0003668435201379526, | |
| "loss": 1.2482, | |
| "step": 475 | |
| }, | |
| { | |
| "epoch": 0.22, | |
| "learning_rate": 0.00036667153940563176, | |
| "loss": 1.3008, | |
| "step": 476 | |
| }, | |
| { | |
| "epoch": 0.23, | |
| "learning_rate": 0.0003664991543308892, | |
| "loss": 1.2442, | |
| "step": 477 | |
| }, | |
| { | |
| "epoch": 0.23, | |
| "learning_rate": 0.0003663263653319283, | |
| "loss": 1.4983, | |
| "step": 478 | |
| }, | |
| { | |
| "epoch": 0.23, | |
| "learning_rate": 0.00036615317282793244, | |
| "loss": 1.4815, | |
| "step": 479 | |
| }, | |
| { | |
| "epoch": 0.23, | |
| "learning_rate": 0.0003659795772390637, | |
| "loss": 1.516, | |
| "step": 480 | |
| }, | |
| { | |
| "epoch": 0.23, | |
| "learning_rate": 0.00036580557898646215, | |
| "loss": 1.5424, | |
| "step": 481 | |
| }, | |
| { | |
| "epoch": 0.23, | |
| "learning_rate": 0.00036563117849224474, | |
| "loss": 1.3857, | |
| "step": 482 | |
| }, | |
| { | |
| "epoch": 0.23, | |
| "learning_rate": 0.00036545637617950416, | |
| "loss": 1.5148, | |
| "step": 483 | |
| }, | |
| { | |
| "epoch": 0.23, | |
| "learning_rate": 0.000365281172472308, | |
| "loss": 1.1913, | |
| "step": 484 | |
| }, | |
| { | |
| "epoch": 0.23, | |
| "learning_rate": 0.00036510556779569757, | |
| "loss": 1.4367, | |
| "step": 485 | |
| }, | |
| { | |
| "epoch": 0.23, | |
| "learning_rate": 0.0003649295625756869, | |
| "loss": 1.334, | |
| "step": 486 | |
| }, | |
| { | |
| "epoch": 0.23, | |
| "learning_rate": 0.00036475315723926175, | |
| "loss": 1.4054, | |
| "step": 487 | |
| }, | |
| { | |
| "epoch": 0.23, | |
| "learning_rate": 0.0003645763522143787, | |
| "loss": 1.317, | |
| "step": 488 | |
| }, | |
| { | |
| "epoch": 0.23, | |
| "learning_rate": 0.00036439914792996364, | |
| "loss": 1.2957, | |
| "step": 489 | |
| }, | |
| { | |
| "epoch": 0.23, | |
| "learning_rate": 0.0003642215448159115, | |
| "loss": 1.2054, | |
| "step": 490 | |
| }, | |
| { | |
| "epoch": 0.23, | |
| "learning_rate": 0.0003640435433030843, | |
| "loss": 1.3806, | |
| "step": 491 | |
| }, | |
| { | |
| "epoch": 0.23, | |
| "learning_rate": 0.0003638651438233109, | |
| "loss": 1.3092, | |
| "step": 492 | |
| }, | |
| { | |
| "epoch": 0.23, | |
| "learning_rate": 0.0003636863468093855, | |
| "loss": 1.3545, | |
| "step": 493 | |
| }, | |
| { | |
| "epoch": 0.23, | |
| "learning_rate": 0.0003635071526950668, | |
| "loss": 1.3809, | |
| "step": 494 | |
| }, | |
| { | |
| "epoch": 0.23, | |
| "learning_rate": 0.0003633275619150767, | |
| "loss": 0.981, | |
| "step": 495 | |
| }, | |
| { | |
| "epoch": 0.23, | |
| "learning_rate": 0.0003631475749050994, | |
| "loss": 1.4125, | |
| "step": 496 | |
| }, | |
| { | |
| "epoch": 0.23, | |
| "learning_rate": 0.00036296719210178056, | |
| "loss": 1.3213, | |
| "step": 497 | |
| }, | |
| { | |
| "epoch": 0.24, | |
| "learning_rate": 0.0003627864139427259, | |
| "loss": 1.5648, | |
| "step": 498 | |
| }, | |
| { | |
| "epoch": 0.24, | |
| "learning_rate": 0.00036260524086650025, | |
| "loss": 1.274, | |
| "step": 499 | |
| }, | |
| { | |
| "epoch": 0.24, | |
| "learning_rate": 0.0003624236733126265, | |
| "loss": 1.3447, | |
| "step": 500 | |
| }, | |
| { | |
| "epoch": 0.24, | |
| "learning_rate": 0.00036224171172158457, | |
| "loss": 1.4144, | |
| "step": 501 | |
| }, | |
| { | |
| "epoch": 0.24, | |
| "learning_rate": 0.0003620593565348103, | |
| "loss": 1.4745, | |
| "step": 502 | |
| }, | |
| { | |
| "epoch": 0.24, | |
| "learning_rate": 0.00036187660819469433, | |
| "loss": 1.4349, | |
| "step": 503 | |
| }, | |
| { | |
| "epoch": 0.24, | |
| "learning_rate": 0.0003616934671445813, | |
| "loss": 1.3624, | |
| "step": 504 | |
| }, | |
| { | |
| "epoch": 0.24, | |
| "learning_rate": 0.00036150993382876826, | |
| "loss": 1.1218, | |
| "step": 505 | |
| }, | |
| { | |
| "epoch": 0.24, | |
| "learning_rate": 0.000361326008692504, | |
| "loss": 1.1525, | |
| "step": 506 | |
| }, | |
| { | |
| "epoch": 0.24, | |
| "learning_rate": 0.00036114169218198793, | |
| "loss": 1.557, | |
| "step": 507 | |
| }, | |
| { | |
| "epoch": 0.24, | |
| "learning_rate": 0.0003609569847443689, | |
| "loss": 1.421, | |
| "step": 508 | |
| }, | |
| { | |
| "epoch": 0.24, | |
| "learning_rate": 0.0003607718868277441, | |
| "loss": 1.3613, | |
| "step": 509 | |
| }, | |
| { | |
| "epoch": 0.24, | |
| "learning_rate": 0.0003605863988811581, | |
| "loss": 1.5302, | |
| "step": 510 | |
| }, | |
| { | |
| "epoch": 0.24, | |
| "learning_rate": 0.00036040052135460156, | |
| "loss": 1.2848, | |
| "step": 511 | |
| }, | |
| { | |
| "epoch": 0.24, | |
| "learning_rate": 0.0003602142546990103, | |
| "loss": 1.4686, | |
| "step": 512 | |
| }, | |
| { | |
| "epoch": 0.24, | |
| "learning_rate": 0.00036002759936626415, | |
| "loss": 1.3161, | |
| "step": 513 | |
| }, | |
| { | |
| "epoch": 0.24, | |
| "learning_rate": 0.00035984055580918595, | |
| "loss": 1.307, | |
| "step": 514 | |
| }, | |
| { | |
| "epoch": 0.24, | |
| "learning_rate": 0.00035965312448154034, | |
| "loss": 1.4552, | |
| "step": 515 | |
| }, | |
| { | |
| "epoch": 0.24, | |
| "learning_rate": 0.00035946530583803247, | |
| "loss": 1.2265, | |
| "step": 516 | |
| }, | |
| { | |
| "epoch": 0.24, | |
| "learning_rate": 0.0003592771003343075, | |
| "loss": 1.4014, | |
| "step": 517 | |
| }, | |
| { | |
| "epoch": 0.24, | |
| "learning_rate": 0.0003590885084269488, | |
| "loss": 1.5022, | |
| "step": 518 | |
| }, | |
| { | |
| "epoch": 0.25, | |
| "learning_rate": 0.0003588995305734772, | |
| "loss": 1.0336, | |
| "step": 519 | |
| }, | |
| { | |
| "epoch": 0.25, | |
| "learning_rate": 0.00035871016723234997, | |
| "loss": 1.1944, | |
| "step": 520 | |
| }, | |
| { | |
| "epoch": 0.25, | |
| "learning_rate": 0.00035852041886295943, | |
| "loss": 1.5178, | |
| "step": 521 | |
| }, | |
| { | |
| "epoch": 0.25, | |
| "learning_rate": 0.00035833028592563204, | |
| "loss": 1.4343, | |
| "step": 522 | |
| }, | |
| { | |
| "epoch": 0.25, | |
| "learning_rate": 0.0003581397688816271, | |
| "loss": 1.2634, | |
| "step": 523 | |
| }, | |
| { | |
| "epoch": 0.25, | |
| "learning_rate": 0.00035794886819313584, | |
| "loss": 1.3973, | |
| "step": 524 | |
| }, | |
| { | |
| "epoch": 0.25, | |
| "learning_rate": 0.00035775758432328035, | |
| "loss": 1.5337, | |
| "step": 525 | |
| }, | |
| { | |
| "epoch": 0.25, | |
| "learning_rate": 0.00035756591773611194, | |
| "loss": 1.5783, | |
| "step": 526 | |
| }, | |
| { | |
| "epoch": 0.25, | |
| "learning_rate": 0.00035737386889661077, | |
| "loss": 1.3674, | |
| "step": 527 | |
| }, | |
| { | |
| "epoch": 0.25, | |
| "learning_rate": 0.0003571814382706842, | |
| "loss": 1.2439, | |
| "step": 528 | |
| }, | |
| { | |
| "epoch": 0.25, | |
| "learning_rate": 0.0003569886263251657, | |
| "loss": 1.2478, | |
| "step": 529 | |
| }, | |
| { | |
| "epoch": 0.25, | |
| "learning_rate": 0.00035679543352781383, | |
| "loss": 1.6748, | |
| "step": 530 | |
| }, | |
| { | |
| "epoch": 0.25, | |
| "learning_rate": 0.00035660186034731125, | |
| "loss": 1.1766, | |
| "step": 531 | |
| }, | |
| { | |
| "epoch": 0.25, | |
| "learning_rate": 0.0003564079072532633, | |
| "loss": 1.4937, | |
| "step": 532 | |
| }, | |
| { | |
| "epoch": 0.25, | |
| "learning_rate": 0.00035621357471619697, | |
| "loss": 1.5937, | |
| "step": 533 | |
| }, | |
| { | |
| "epoch": 0.25, | |
| "learning_rate": 0.00035601886320755984, | |
| "loss": 1.3704, | |
| "step": 534 | |
| }, | |
| { | |
| "epoch": 0.25, | |
| "learning_rate": 0.00035582377319971884, | |
| "loss": 1.1386, | |
| "step": 535 | |
| }, | |
| { | |
| "epoch": 0.25, | |
| "learning_rate": 0.0003556283051659591, | |
| "loss": 1.4755, | |
| "step": 536 | |
| }, | |
| { | |
| "epoch": 0.25, | |
| "learning_rate": 0.0003554324595804829, | |
| "loss": 1.5553, | |
| "step": 537 | |
| }, | |
| { | |
| "epoch": 0.25, | |
| "learning_rate": 0.0003552362369184084, | |
| "loss": 1.5264, | |
| "step": 538 | |
| }, | |
| { | |
| "epoch": 0.25, | |
| "learning_rate": 0.00035503963765576855, | |
| "loss": 1.3787, | |
| "step": 539 | |
| }, | |
| { | |
| "epoch": 0.25, | |
| "learning_rate": 0.0003548426622695099, | |
| "loss": 1.5708, | |
| "step": 540 | |
| }, | |
| { | |
| "epoch": 0.26, | |
| "learning_rate": 0.0003546453112374916, | |
| "loss": 1.5532, | |
| "step": 541 | |
| }, | |
| { | |
| "epoch": 0.26, | |
| "learning_rate": 0.00035444758503848405, | |
| "loss": 1.2355, | |
| "step": 542 | |
| }, | |
| { | |
| "epoch": 0.26, | |
| "learning_rate": 0.00035424948415216763, | |
| "loss": 1.2705, | |
| "step": 543 | |
| }, | |
| { | |
| "epoch": 0.26, | |
| "learning_rate": 0.00035405100905913184, | |
| "loss": 1.3754, | |
| "step": 544 | |
| }, | |
| { | |
| "epoch": 0.26, | |
| "learning_rate": 0.0003538521602408741, | |
| "loss": 1.4639, | |
| "step": 545 | |
| }, | |
| { | |
| "epoch": 0.26, | |
| "learning_rate": 0.0003536529381797984, | |
| "loss": 1.3271, | |
| "step": 546 | |
| }, | |
| { | |
| "epoch": 0.26, | |
| "learning_rate": 0.0003534533433592141, | |
| "loss": 1.4256, | |
| "step": 547 | |
| }, | |
| { | |
| "epoch": 0.26, | |
| "learning_rate": 0.00035325337626333504, | |
| "loss": 1.4383, | |
| "step": 548 | |
| }, | |
| { | |
| "epoch": 0.26, | |
| "learning_rate": 0.00035305303737727807, | |
| "loss": 1.1625, | |
| "step": 549 | |
| }, | |
| { | |
| "epoch": 0.26, | |
| "learning_rate": 0.000352852327187062, | |
| "loss": 1.4411, | |
| "step": 550 | |
| }, | |
| { | |
| "epoch": 0.26, | |
| "learning_rate": 0.00035265124617960653, | |
| "loss": 1.2599, | |
| "step": 551 | |
| }, | |
| { | |
| "epoch": 0.26, | |
| "learning_rate": 0.0003524497948427309, | |
| "loss": 1.2186, | |
| "step": 552 | |
| }, | |
| { | |
| "epoch": 0.26, | |
| "learning_rate": 0.0003522479736651527, | |
| "loss": 1.5345, | |
| "step": 553 | |
| }, | |
| { | |
| "epoch": 0.26, | |
| "learning_rate": 0.0003520457831364868, | |
| "loss": 1.2944, | |
| "step": 554 | |
| }, | |
| { | |
| "epoch": 0.26, | |
| "learning_rate": 0.00035184322374724416, | |
| "loss": 1.3791, | |
| "step": 555 | |
| }, | |
| { | |
| "epoch": 0.26, | |
| "learning_rate": 0.0003516402959888304, | |
| "loss": 1.6277, | |
| "step": 556 | |
| }, | |
| { | |
| "epoch": 0.26, | |
| "learning_rate": 0.00035143700035354517, | |
| "loss": 1.4593, | |
| "step": 557 | |
| }, | |
| { | |
| "epoch": 0.26, | |
| "learning_rate": 0.00035123333733458014, | |
| "loss": 1.3568, | |
| "step": 558 | |
| }, | |
| { | |
| "epoch": 0.26, | |
| "learning_rate": 0.00035102930742601855, | |
| "loss": 1.2664, | |
| "step": 559 | |
| }, | |
| { | |
| "epoch": 0.26, | |
| "learning_rate": 0.0003508249111228336, | |
| "loss": 1.1526, | |
| "step": 560 | |
| }, | |
| { | |
| "epoch": 0.26, | |
| "learning_rate": 0.00035062014892088734, | |
| "loss": 1.568, | |
| "step": 561 | |
| }, | |
| { | |
| "epoch": 0.27, | |
| "learning_rate": 0.00035041502131692953, | |
| "loss": 1.3015, | |
| "step": 562 | |
| }, | |
| { | |
| "epoch": 0.27, | |
| "learning_rate": 0.0003502095288085964, | |
| "loss": 1.2436, | |
| "step": 563 | |
| }, | |
| { | |
| "epoch": 0.27, | |
| "learning_rate": 0.00035000367189440935, | |
| "loss": 1.246, | |
| "step": 564 | |
| }, | |
| { | |
| "epoch": 0.27, | |
| "learning_rate": 0.0003497974510737739, | |
| "loss": 1.2972, | |
| "step": 565 | |
| }, | |
| { | |
| "epoch": 0.27, | |
| "learning_rate": 0.00034959086684697847, | |
| "loss": 1.4416, | |
| "step": 566 | |
| }, | |
| { | |
| "epoch": 0.27, | |
| "learning_rate": 0.0003493839197151928, | |
| "loss": 1.4723, | |
| "step": 567 | |
| }, | |
| { | |
| "epoch": 0.27, | |
| "learning_rate": 0.00034917661018046755, | |
| "loss": 1.2364, | |
| "step": 568 | |
| }, | |
| { | |
| "epoch": 0.27, | |
| "learning_rate": 0.00034896893874573193, | |
| "loss": 0.9685, | |
| "step": 569 | |
| }, | |
| { | |
| "epoch": 0.27, | |
| "learning_rate": 0.0003487609059147936, | |
| "loss": 1.4989, | |
| "step": 570 | |
| }, | |
| { | |
| "epoch": 0.27, | |
| "learning_rate": 0.00034855251219233674, | |
| "loss": 1.331, | |
| "step": 571 | |
| }, | |
| { | |
| "epoch": 0.27, | |
| "learning_rate": 0.0003483437580839212, | |
| "loss": 1.282, | |
| "step": 572 | |
| }, | |
| { | |
| "epoch": 0.27, | |
| "learning_rate": 0.0003481346440959809, | |
| "loss": 1.4483, | |
| "step": 573 | |
| }, | |
| { | |
| "epoch": 0.27, | |
| "learning_rate": 0.00034792517073582306, | |
| "loss": 1.3028, | |
| "step": 574 | |
| }, | |
| { | |
| "epoch": 0.27, | |
| "learning_rate": 0.0003477153385116265, | |
| "loss": 1.2324, | |
| "step": 575 | |
| }, | |
| { | |
| "epoch": 0.27, | |
| "learning_rate": 0.00034750514793244083, | |
| "loss": 1.4206, | |
| "step": 576 | |
| }, | |
| { | |
| "epoch": 0.27, | |
| "learning_rate": 0.00034729459950818496, | |
| "loss": 1.5627, | |
| "step": 577 | |
| }, | |
| { | |
| "epoch": 0.27, | |
| "learning_rate": 0.00034708369374964593, | |
| "loss": 1.4441, | |
| "step": 578 | |
| }, | |
| { | |
| "epoch": 0.27, | |
| "learning_rate": 0.0003468724311684775, | |
| "loss": 1.2597, | |
| "step": 579 | |
| }, | |
| { | |
| "epoch": 0.27, | |
| "learning_rate": 0.00034666081227719943, | |
| "loss": 1.3732, | |
| "step": 580 | |
| }, | |
| { | |
| "epoch": 0.27, | |
| "learning_rate": 0.00034644883758919555, | |
| "loss": 1.4752, | |
| "step": 581 | |
| }, | |
| { | |
| "epoch": 0.27, | |
| "learning_rate": 0.00034623650761871305, | |
| "loss": 1.0906, | |
| "step": 582 | |
| }, | |
| { | |
| "epoch": 0.28, | |
| "learning_rate": 0.0003460238228808609, | |
| "loss": 1.3774, | |
| "step": 583 | |
| }, | |
| { | |
| "epoch": 0.28, | |
| "learning_rate": 0.0003458107838916088, | |
| "loss": 1.2825, | |
| "step": 584 | |
| }, | |
| { | |
| "epoch": 0.28, | |
| "learning_rate": 0.0003455973911677859, | |
| "loss": 1.442, | |
| "step": 585 | |
| }, | |
| { | |
| "epoch": 0.28, | |
| "learning_rate": 0.00034538364522707934, | |
| "loss": 1.3051, | |
| "step": 586 | |
| }, | |
| { | |
| "epoch": 0.28, | |
| "learning_rate": 0.00034516954658803334, | |
| "loss": 1.2828, | |
| "step": 587 | |
| }, | |
| { | |
| "epoch": 0.28, | |
| "learning_rate": 0.00034495509577004774, | |
| "loss": 1.112, | |
| "step": 588 | |
| }, | |
| { | |
| "epoch": 0.28, | |
| "learning_rate": 0.00034474029329337663, | |
| "loss": 1.707, | |
| "step": 589 | |
| }, | |
| { | |
| "epoch": 0.28, | |
| "learning_rate": 0.0003445251396791273, | |
| "loss": 1.403, | |
| "step": 590 | |
| }, | |
| { | |
| "epoch": 0.28, | |
| "learning_rate": 0.0003443096354492589, | |
| "loss": 1.2986, | |
| "step": 591 | |
| }, | |
| { | |
| "epoch": 0.28, | |
| "learning_rate": 0.00034409378112658113, | |
| "loss": 1.3418, | |
| "step": 592 | |
| }, | |
| { | |
| "epoch": 0.28, | |
| "learning_rate": 0.00034387757723475313, | |
| "loss": 1.2424, | |
| "step": 593 | |
| }, | |
| { | |
| "epoch": 0.28, | |
| "learning_rate": 0.0003436610242982819, | |
| "loss": 1.2661, | |
| "step": 594 | |
| }, | |
| { | |
| "epoch": 0.28, | |
| "learning_rate": 0.00034344412284252135, | |
| "loss": 1.2827, | |
| "step": 595 | |
| }, | |
| { | |
| "epoch": 0.28, | |
| "learning_rate": 0.00034322687339367083, | |
| "loss": 1.2011, | |
| "step": 596 | |
| }, | |
| { | |
| "epoch": 0.28, | |
| "learning_rate": 0.00034300927647877403, | |
| "loss": 1.2413, | |
| "step": 597 | |
| }, | |
| { | |
| "epoch": 0.28, | |
| "learning_rate": 0.00034279133262571734, | |
| "loss": 1.3173, | |
| "step": 598 | |
| }, | |
| { | |
| "epoch": 0.28, | |
| "learning_rate": 0.0003425730423632291, | |
| "loss": 1.2728, | |
| "step": 599 | |
| }, | |
| { | |
| "epoch": 0.28, | |
| "learning_rate": 0.0003423544062208779, | |
| "loss": 1.2539, | |
| "step": 600 | |
| }, | |
| { | |
| "epoch": 0.28, | |
| "learning_rate": 0.00034213542472907144, | |
| "loss": 1.2058, | |
| "step": 601 | |
| }, | |
| { | |
| "epoch": 0.28, | |
| "learning_rate": 0.0003419160984190552, | |
| "loss": 1.1629, | |
| "step": 602 | |
| }, | |
| { | |
| "epoch": 0.28, | |
| "learning_rate": 0.00034169642782291116, | |
| "loss": 1.1723, | |
| "step": 603 | |
| }, | |
| { | |
| "epoch": 0.29, | |
| "learning_rate": 0.0003414764134735566, | |
| "loss": 1.3751, | |
| "step": 604 | |
| }, | |
| { | |
| "epoch": 0.29, | |
| "learning_rate": 0.0003412560559047428, | |
| "loss": 1.3485, | |
| "step": 605 | |
| }, | |
| { | |
| "epoch": 0.29, | |
| "learning_rate": 0.0003410353556510536, | |
| "loss": 1.4167, | |
| "step": 606 | |
| }, | |
| { | |
| "epoch": 0.29, | |
| "learning_rate": 0.0003408143132479041, | |
| "loss": 1.4048, | |
| "step": 607 | |
| }, | |
| { | |
| "epoch": 0.29, | |
| "learning_rate": 0.0003405929292315397, | |
| "loss": 1.5166, | |
| "step": 608 | |
| }, | |
| { | |
| "epoch": 0.29, | |
| "learning_rate": 0.00034037120413903425, | |
| "loss": 1.6244, | |
| "step": 609 | |
| }, | |
| { | |
| "epoch": 0.29, | |
| "learning_rate": 0.0003401491385082892, | |
| "loss": 1.3791, | |
| "step": 610 | |
| }, | |
| { | |
| "epoch": 0.29, | |
| "learning_rate": 0.00033992673287803224, | |
| "loss": 1.086, | |
| "step": 611 | |
| }, | |
| { | |
| "epoch": 0.29, | |
| "learning_rate": 0.00033970398778781564, | |
| "loss": 1.5672, | |
| "step": 612 | |
| }, | |
| { | |
| "epoch": 0.29, | |
| "learning_rate": 0.00033948090377801545, | |
| "loss": 1.513, | |
| "step": 613 | |
| }, | |
| { | |
| "epoch": 0.29, | |
| "learning_rate": 0.0003392574813898298, | |
| "loss": 1.516, | |
| "step": 614 | |
| }, | |
| { | |
| "epoch": 0.29, | |
| "learning_rate": 0.0003390337211652777, | |
| "loss": 1.499, | |
| "step": 615 | |
| }, | |
| { | |
| "epoch": 0.29, | |
| "learning_rate": 0.0003388096236471978, | |
| "loss": 1.5168, | |
| "step": 616 | |
| }, | |
| { | |
| "epoch": 0.29, | |
| "learning_rate": 0.000338585189379247, | |
| "loss": 1.1873, | |
| "step": 617 | |
| }, | |
| { | |
| "epoch": 0.29, | |
| "learning_rate": 0.0003383604189058992, | |
| "loss": 1.3861, | |
| "step": 618 | |
| }, | |
| { | |
| "epoch": 0.29, | |
| "learning_rate": 0.00033813531277244384, | |
| "loss": 1.5166, | |
| "step": 619 | |
| }, | |
| { | |
| "epoch": 0.29, | |
| "learning_rate": 0.0003379098715249847, | |
| "loss": 1.0685, | |
| "step": 620 | |
| }, | |
| { | |
| "epoch": 0.29, | |
| "learning_rate": 0.0003376840957104385, | |
| "loss": 1.2126, | |
| "step": 621 | |
| }, | |
| { | |
| "epoch": 0.29, | |
| "learning_rate": 0.0003374579858765338, | |
| "loss": 1.201, | |
| "step": 622 | |
| }, | |
| { | |
| "epoch": 0.29, | |
| "learning_rate": 0.00033723154257180927, | |
| "loss": 1.2925, | |
| "step": 623 | |
| }, | |
| { | |
| "epoch": 0.29, | |
| "learning_rate": 0.0003370047663456127, | |
| "loss": 1.4116, | |
| "step": 624 | |
| }, | |
| { | |
| "epoch": 0.3, | |
| "learning_rate": 0.0003367776577480994, | |
| "loss": 1.2275, | |
| "step": 625 | |
| }, | |
| { | |
| "epoch": 0.3, | |
| "learning_rate": 0.00033655021733023114, | |
| "loss": 1.3411, | |
| "step": 626 | |
| }, | |
| { | |
| "epoch": 0.3, | |
| "learning_rate": 0.0003363224456437747, | |
| "loss": 1.4525, | |
| "step": 627 | |
| }, | |
| { | |
| "epoch": 0.3, | |
| "learning_rate": 0.00033609434324130045, | |
| "loss": 1.4977, | |
| "step": 628 | |
| }, | |
| { | |
| "epoch": 0.3, | |
| "learning_rate": 0.00033586591067618103, | |
| "loss": 1.2514, | |
| "step": 629 | |
| }, | |
| { | |
| "epoch": 0.3, | |
| "learning_rate": 0.0003356371485025901, | |
| "loss": 1.3123, | |
| "step": 630 | |
| }, | |
| { | |
| "epoch": 0.3, | |
| "learning_rate": 0.000335408057275501, | |
| "loss": 1.3881, | |
| "step": 631 | |
| }, | |
| { | |
| "epoch": 0.3, | |
| "learning_rate": 0.0003351786375506852, | |
| "loss": 1.3676, | |
| "step": 632 | |
| }, | |
| { | |
| "epoch": 0.3, | |
| "learning_rate": 0.0003349488898847113, | |
| "loss": 1.3726, | |
| "step": 633 | |
| }, | |
| { | |
| "epoch": 0.3, | |
| "learning_rate": 0.00033471881483494324, | |
| "loss": 1.5587, | |
| "step": 634 | |
| }, | |
| { | |
| "epoch": 0.3, | |
| "learning_rate": 0.0003344884129595395, | |
| "loss": 1.5089, | |
| "step": 635 | |
| }, | |
| { | |
| "epoch": 0.3, | |
| "learning_rate": 0.00033425768481745103, | |
| "loss": 1.1553, | |
| "step": 636 | |
| }, | |
| { | |
| "epoch": 0.3, | |
| "learning_rate": 0.0003340266309684207, | |
| "loss": 1.196, | |
| "step": 637 | |
| }, | |
| { | |
| "epoch": 0.3, | |
| "learning_rate": 0.0003337952519729813, | |
| "loss": 1.437, | |
| "step": 638 | |
| }, | |
| { | |
| "epoch": 0.3, | |
| "learning_rate": 0.00033356354839245443, | |
| "loss": 1.5862, | |
| "step": 639 | |
| }, | |
| { | |
| "epoch": 0.3, | |
| "learning_rate": 0.0003333315207889492, | |
| "loss": 1.2645, | |
| "step": 640 | |
| }, | |
| { | |
| "epoch": 0.3, | |
| "learning_rate": 0.0003330991697253608, | |
| "loss": 1.493, | |
| "step": 641 | |
| }, | |
| { | |
| "epoch": 0.3, | |
| "learning_rate": 0.00033286649576536915, | |
| "loss": 1.2698, | |
| "step": 642 | |
| }, | |
| { | |
| "epoch": 0.3, | |
| "learning_rate": 0.0003326334994734373, | |
| "loss": 1.6227, | |
| "step": 643 | |
| }, | |
| { | |
| "epoch": 0.3, | |
| "learning_rate": 0.00033240018141481064, | |
| "loss": 1.2429, | |
| "step": 644 | |
| }, | |
| { | |
| "epoch": 0.3, | |
| "learning_rate": 0.00033216654215551477, | |
| "loss": 1.4009, | |
| "step": 645 | |
| }, | |
| { | |
| "epoch": 0.31, | |
| "learning_rate": 0.0003319325822623548, | |
| "loss": 1.5027, | |
| "step": 646 | |
| }, | |
| { | |
| "epoch": 0.31, | |
| "learning_rate": 0.00033169830230291356, | |
| "loss": 1.4413, | |
| "step": 647 | |
| }, | |
| { | |
| "epoch": 0.31, | |
| "learning_rate": 0.00033146370284555047, | |
| "loss": 1.332, | |
| "step": 648 | |
| }, | |
| { | |
| "epoch": 0.31, | |
| "learning_rate": 0.0003312287844593998, | |
| "loss": 1.2462, | |
| "step": 649 | |
| }, | |
| { | |
| "epoch": 0.31, | |
| "learning_rate": 0.0003309935477143699, | |
| "loss": 1.3925, | |
| "step": 650 | |
| }, | |
| { | |
| "epoch": 0.31, | |
| "learning_rate": 0.00033075799318114116, | |
| "loss": 1.0344, | |
| "step": 651 | |
| }, | |
| { | |
| "epoch": 0.31, | |
| "learning_rate": 0.00033052212143116496, | |
| "loss": 1.3053, | |
| "step": 652 | |
| }, | |
| { | |
| "epoch": 0.31, | |
| "learning_rate": 0.00033028593303666235, | |
| "loss": 1.302, | |
| "step": 653 | |
| }, | |
| { | |
| "epoch": 0.31, | |
| "learning_rate": 0.0003300494285706226, | |
| "loss": 1.4378, | |
| "step": 654 | |
| }, | |
| { | |
| "epoch": 0.31, | |
| "learning_rate": 0.0003298126086068015, | |
| "loss": 1.281, | |
| "step": 655 | |
| }, | |
| { | |
| "epoch": 0.31, | |
| "learning_rate": 0.00032957547371972043, | |
| "loss": 1.1997, | |
| "step": 656 | |
| }, | |
| { | |
| "epoch": 0.31, | |
| "learning_rate": 0.0003293380244846648, | |
| "loss": 1.2488, | |
| "step": 657 | |
| }, | |
| { | |
| "epoch": 0.31, | |
| "learning_rate": 0.0003291002614776826, | |
| "loss": 1.616, | |
| "step": 658 | |
| }, | |
| { | |
| "epoch": 0.31, | |
| "learning_rate": 0.0003288621852755829, | |
| "loss": 0.9443, | |
| "step": 659 | |
| }, | |
| { | |
| "epoch": 0.31, | |
| "learning_rate": 0.0003286237964559346, | |
| "loss": 1.296, | |
| "step": 660 | |
| }, | |
| { | |
| "epoch": 0.31, | |
| "learning_rate": 0.0003283850955970653, | |
| "loss": 1.3768, | |
| "step": 661 | |
| }, | |
| { | |
| "epoch": 0.31, | |
| "learning_rate": 0.0003281460832780591, | |
| "loss": 1.5167, | |
| "step": 662 | |
| }, | |
| { | |
| "epoch": 0.31, | |
| "learning_rate": 0.00032790676007875627, | |
| "loss": 1.2445, | |
| "step": 663 | |
| }, | |
| { | |
| "epoch": 0.31, | |
| "learning_rate": 0.00032766712657975066, | |
| "loss": 0.9358, | |
| "step": 664 | |
| }, | |
| { | |
| "epoch": 0.31, | |
| "learning_rate": 0.0003274271833623893, | |
| "loss": 1.0435, | |
| "step": 665 | |
| }, | |
| { | |
| "epoch": 0.31, | |
| "learning_rate": 0.00032718693100877065, | |
| "loss": 1.1012, | |
| "step": 666 | |
| }, | |
| { | |
| "epoch": 0.31, | |
| "learning_rate": 0.0003269463701017428, | |
| "loss": 1.3027, | |
| "step": 667 | |
| }, | |
| { | |
| "epoch": 0.32, | |
| "learning_rate": 0.00032670550122490267, | |
| "loss": 1.5241, | |
| "step": 668 | |
| }, | |
| { | |
| "epoch": 0.32, | |
| "learning_rate": 0.00032646432496259416, | |
| "loss": 1.3093, | |
| "step": 669 | |
| }, | |
| { | |
| "epoch": 0.32, | |
| "learning_rate": 0.000326222841899907, | |
| "loss": 1.2715, | |
| "step": 670 | |
| }, | |
| { | |
| "epoch": 0.32, | |
| "learning_rate": 0.000325981052622675, | |
| "loss": 1.3346, | |
| "step": 671 | |
| }, | |
| { | |
| "epoch": 0.32, | |
| "learning_rate": 0.00032573895771747505, | |
| "loss": 1.2683, | |
| "step": 672 | |
| }, | |
| { | |
| "epoch": 0.32, | |
| "learning_rate": 0.0003254965577716255, | |
| "loss": 1.579, | |
| "step": 673 | |
| }, | |
| { | |
| "epoch": 0.32, | |
| "learning_rate": 0.0003252538533731846, | |
| "loss": 1.2409, | |
| "step": 674 | |
| }, | |
| { | |
| "epoch": 0.32, | |
| "learning_rate": 0.00032501084511094914, | |
| "loss": 1.3204, | |
| "step": 675 | |
| }, | |
| { | |
| "epoch": 0.32, | |
| "learning_rate": 0.0003247675335744533, | |
| "loss": 1.444, | |
| "step": 676 | |
| }, | |
| { | |
| "epoch": 0.32, | |
| "learning_rate": 0.0003245239193539668, | |
| "loss": 1.0368, | |
| "step": 677 | |
| }, | |
| { | |
| "epoch": 0.32, | |
| "learning_rate": 0.00032428000304049383, | |
| "loss": 1.4488, | |
| "step": 678 | |
| }, | |
| { | |
| "epoch": 0.32, | |
| "learning_rate": 0.0003240357852257713, | |
| "loss": 1.2406, | |
| "step": 679 | |
| }, | |
| { | |
| "epoch": 0.32, | |
| "learning_rate": 0.0003237912665022676, | |
| "loss": 1.4373, | |
| "step": 680 | |
| }, | |
| { | |
| "epoch": 0.32, | |
| "learning_rate": 0.0003235464474631813, | |
| "loss": 1.0534, | |
| "step": 681 | |
| }, | |
| { | |
| "epoch": 0.32, | |
| "learning_rate": 0.0003233013287024391, | |
| "loss": 1.3751, | |
| "step": 682 | |
| }, | |
| { | |
| "epoch": 0.32, | |
| "learning_rate": 0.0003230559108146953, | |
| "loss": 1.2604, | |
| "step": 683 | |
| }, | |
| { | |
| "epoch": 0.32, | |
| "learning_rate": 0.0003228101943953297, | |
| "loss": 1.1124, | |
| "step": 684 | |
| }, | |
| { | |
| "epoch": 0.32, | |
| "learning_rate": 0.00032256418004044616, | |
| "loss": 1.1776, | |
| "step": 685 | |
| }, | |
| { | |
| "epoch": 0.32, | |
| "learning_rate": 0.0003223178683468716, | |
| "loss": 1.0755, | |
| "step": 686 | |
| }, | |
| { | |
| "epoch": 0.32, | |
| "learning_rate": 0.0003220712599121541, | |
| "loss": 1.2169, | |
| "step": 687 | |
| }, | |
| { | |
| "epoch": 0.32, | |
| "learning_rate": 0.00032182435533456154, | |
| "loss": 1.4314, | |
| "step": 688 | |
| }, | |
| { | |
| "epoch": 0.33, | |
| "learning_rate": 0.00032157715521308064, | |
| "loss": 1.2822, | |
| "step": 689 | |
| }, | |
| { | |
| "epoch": 0.33, | |
| "learning_rate": 0.00032132966014741457, | |
| "loss": 1.4197, | |
| "step": 690 | |
| }, | |
| { | |
| "epoch": 0.33, | |
| "learning_rate": 0.0003210818707379826, | |
| "loss": 1.3165, | |
| "step": 691 | |
| }, | |
| { | |
| "epoch": 0.33, | |
| "learning_rate": 0.00032083378758591757, | |
| "loss": 1.2202, | |
| "step": 692 | |
| }, | |
| { | |
| "epoch": 0.33, | |
| "learning_rate": 0.0003205854112930651, | |
| "loss": 1.3977, | |
| "step": 693 | |
| }, | |
| { | |
| "epoch": 0.33, | |
| "learning_rate": 0.0003203367424619822, | |
| "loss": 1.4865, | |
| "step": 694 | |
| }, | |
| { | |
| "epoch": 0.33, | |
| "learning_rate": 0.0003200877816959352, | |
| "loss": 1.2872, | |
| "step": 695 | |
| }, | |
| { | |
| "epoch": 0.33, | |
| "learning_rate": 0.0003198385295988989, | |
| "loss": 1.3833, | |
| "step": 696 | |
| }, | |
| { | |
| "epoch": 0.33, | |
| "learning_rate": 0.00031958898677555476, | |
| "loss": 1.2206, | |
| "step": 697 | |
| }, | |
| { | |
| "epoch": 0.33, | |
| "learning_rate": 0.0003193391538312897, | |
| "loss": 1.2306, | |
| "step": 698 | |
| }, | |
| { | |
| "epoch": 0.33, | |
| "learning_rate": 0.00031908903137219417, | |
| "loss": 1.4376, | |
| "step": 699 | |
| }, | |
| { | |
| "epoch": 0.33, | |
| "learning_rate": 0.00031883862000506134, | |
| "loss": 1.5088, | |
| "step": 700 | |
| }, | |
| { | |
| "epoch": 0.33, | |
| "learning_rate": 0.0003185879203373849, | |
| "loss": 1.4149, | |
| "step": 701 | |
| }, | |
| { | |
| "epoch": 0.33, | |
| "learning_rate": 0.0003183369329773583, | |
| "loss": 1.1201, | |
| "step": 702 | |
| }, | |
| { | |
| "epoch": 0.33, | |
| "learning_rate": 0.00031808565853387263, | |
| "loss": 1.2864, | |
| "step": 703 | |
| }, | |
| { | |
| "epoch": 0.33, | |
| "learning_rate": 0.0003178340976165157, | |
| "loss": 1.3081, | |
| "step": 704 | |
| }, | |
| { | |
| "epoch": 0.33, | |
| "learning_rate": 0.00031758225083557, | |
| "loss": 1.223, | |
| "step": 705 | |
| }, | |
| { | |
| "epoch": 0.33, | |
| "learning_rate": 0.00031733011880201185, | |
| "loss": 1.3362, | |
| "step": 706 | |
| }, | |
| { | |
| "epoch": 0.33, | |
| "learning_rate": 0.0003170777021275093, | |
| "loss": 1.3054, | |
| "step": 707 | |
| }, | |
| { | |
| "epoch": 0.33, | |
| "learning_rate": 0.0003168250014244212, | |
| "loss": 1.1346, | |
| "step": 708 | |
| }, | |
| { | |
| "epoch": 0.33, | |
| "learning_rate": 0.00031657201730579513, | |
| "loss": 1.4264, | |
| "step": 709 | |
| }, | |
| { | |
| "epoch": 0.34, | |
| "learning_rate": 0.0003163187503853666, | |
| "loss": 1.2921, | |
| "step": 710 | |
| }, | |
| { | |
| "epoch": 0.34, | |
| "learning_rate": 0.0003160652012775569, | |
| "loss": 1.3265, | |
| "step": 711 | |
| }, | |
| { | |
| "epoch": 0.34, | |
| "learning_rate": 0.000315811370597472, | |
| "loss": 1.0747, | |
| "step": 712 | |
| }, | |
| { | |
| "epoch": 0.34, | |
| "learning_rate": 0.0003155572589609011, | |
| "loss": 1.3607, | |
| "step": 713 | |
| }, | |
| { | |
| "epoch": 0.34, | |
| "learning_rate": 0.00031530286698431474, | |
| "loss": 1.2866, | |
| "step": 714 | |
| }, | |
| { | |
| "epoch": 0.34, | |
| "learning_rate": 0.00031504819528486365, | |
| "loss": 1.4863, | |
| "step": 715 | |
| }, | |
| { | |
| "epoch": 0.34, | |
| "learning_rate": 0.00031479324448037725, | |
| "loss": 1.2086, | |
| "step": 716 | |
| }, | |
| { | |
| "epoch": 0.34, | |
| "learning_rate": 0.0003145380151893621, | |
| "loss": 1.4026, | |
| "step": 717 | |
| }, | |
| { | |
| "epoch": 0.34, | |
| "learning_rate": 0.00031428250803100017, | |
| "loss": 1.4166, | |
| "step": 718 | |
| }, | |
| { | |
| "epoch": 0.34, | |
| "learning_rate": 0.0003140267236251476, | |
| "loss": 1.6379, | |
| "step": 719 | |
| }, | |
| { | |
| "epoch": 0.34, | |
| "learning_rate": 0.0003137706625923333, | |
| "loss": 1.3528, | |
| "step": 720 | |
| }, | |
| { | |
| "epoch": 0.34, | |
| "learning_rate": 0.000313514325553757, | |
| "loss": 1.4633, | |
| "step": 721 | |
| }, | |
| { | |
| "epoch": 0.34, | |
| "learning_rate": 0.0003132577131312883, | |
| "loss": 1.3926, | |
| "step": 722 | |
| }, | |
| { | |
| "epoch": 0.34, | |
| "learning_rate": 0.0003130008259474646, | |
| "loss": 1.2951, | |
| "step": 723 | |
| }, | |
| { | |
| "epoch": 0.34, | |
| "learning_rate": 0.00031274366462549, | |
| "loss": 1.3975, | |
| "step": 724 | |
| }, | |
| { | |
| "epoch": 0.34, | |
| "learning_rate": 0.0003124862297892337, | |
| "loss": 1.7294, | |
| "step": 725 | |
| }, | |
| { | |
| "epoch": 0.34, | |
| "learning_rate": 0.0003122285220632284, | |
| "loss": 1.4895, | |
| "step": 726 | |
| }, | |
| { | |
| "epoch": 0.34, | |
| "learning_rate": 0.0003119705420726688, | |
| "loss": 1.3108, | |
| "step": 727 | |
| }, | |
| { | |
| "epoch": 0.34, | |
| "learning_rate": 0.0003117122904434101, | |
| "loss": 1.2647, | |
| "step": 728 | |
| }, | |
| { | |
| "epoch": 0.34, | |
| "learning_rate": 0.00031145376780196654, | |
| "loss": 1.133, | |
| "step": 729 | |
| }, | |
| { | |
| "epoch": 0.34, | |
| "learning_rate": 0.00031119497477550986, | |
| "loss": 1.3407, | |
| "step": 730 | |
| }, | |
| { | |
| "epoch": 0.35, | |
| "learning_rate": 0.0003109359119918676, | |
| "loss": 1.3305, | |
| "step": 731 | |
| }, | |
| { | |
| "epoch": 0.35, | |
| "learning_rate": 0.0003106765800795219, | |
| "loss": 1.3938, | |
| "step": 732 | |
| }, | |
| { | |
| "epoch": 0.35, | |
| "learning_rate": 0.00031041697966760777, | |
| "loss": 1.5855, | |
| "step": 733 | |
| }, | |
| { | |
| "epoch": 0.35, | |
| "learning_rate": 0.00031015711138591143, | |
| "loss": 1.347, | |
| "step": 734 | |
| }, | |
| { | |
| "epoch": 0.35, | |
| "learning_rate": 0.0003098969758648691, | |
| "loss": 1.5524, | |
| "step": 735 | |
| }, | |
| { | |
| "epoch": 0.35, | |
| "learning_rate": 0.00030963657373556526, | |
| "loss": 1.1343, | |
| "step": 736 | |
| }, | |
| { | |
| "epoch": 0.35, | |
| "learning_rate": 0.0003093759056297314, | |
| "loss": 1.2149, | |
| "step": 737 | |
| }, | |
| { | |
| "epoch": 0.35, | |
| "learning_rate": 0.00030911497217974386, | |
| "loss": 1.2507, | |
| "step": 738 | |
| }, | |
| { | |
| "epoch": 0.35, | |
| "learning_rate": 0.00030885377401862306, | |
| "loss": 1.5341, | |
| "step": 739 | |
| }, | |
| { | |
| "epoch": 0.35, | |
| "learning_rate": 0.0003085923117800312, | |
| "loss": 1.3615, | |
| "step": 740 | |
| }, | |
| { | |
| "epoch": 0.35, | |
| "learning_rate": 0.00030833058609827175, | |
| "loss": 1.3651, | |
| "step": 741 | |
| }, | |
| { | |
| "epoch": 0.35, | |
| "learning_rate": 0.00030806859760828663, | |
| "loss": 1.2132, | |
| "step": 742 | |
| }, | |
| { | |
| "epoch": 0.35, | |
| "learning_rate": 0.00030780634694565573, | |
| "loss": 1.166, | |
| "step": 743 | |
| }, | |
| { | |
| "epoch": 0.35, | |
| "learning_rate": 0.0003075438347465949, | |
| "loss": 1.4359, | |
| "step": 744 | |
| }, | |
| { | |
| "epoch": 0.35, | |
| "learning_rate": 0.00030728106164795433, | |
| "loss": 1.5718, | |
| "step": 745 | |
| }, | |
| { | |
| "epoch": 0.35, | |
| "learning_rate": 0.00030701802828721734, | |
| "loss": 1.4244, | |
| "step": 746 | |
| }, | |
| { | |
| "epoch": 0.35, | |
| "learning_rate": 0.0003067547353024985, | |
| "loss": 1.208, | |
| "step": 747 | |
| }, | |
| { | |
| "epoch": 0.35, | |
| "learning_rate": 0.0003064911833325423, | |
| "loss": 1.2651, | |
| "step": 748 | |
| }, | |
| { | |
| "epoch": 0.35, | |
| "learning_rate": 0.0003062273730167215, | |
| "loss": 1.2595, | |
| "step": 749 | |
| }, | |
| { | |
| "epoch": 0.35, | |
| "learning_rate": 0.00030596330499503555, | |
| "loss": 1.6153, | |
| "step": 750 | |
| }, | |
| { | |
| "epoch": 0.35, | |
| "learning_rate": 0.00030569897990810916, | |
| "loss": 1.3482, | |
| "step": 751 | |
| }, | |
| { | |
| "epoch": 0.36, | |
| "learning_rate": 0.0003054343983971907, | |
| "loss": 1.1848, | |
| "step": 752 | |
| }, | |
| { | |
| "epoch": 0.36, | |
| "learning_rate": 0.0003051695611041505, | |
| "loss": 1.5675, | |
| "step": 753 | |
| }, | |
| { | |
| "epoch": 0.36, | |
| "learning_rate": 0.0003049044686714796, | |
| "loss": 1.2735, | |
| "step": 754 | |
| }, | |
| { | |
| "epoch": 0.36, | |
| "learning_rate": 0.0003046391217422878, | |
| "loss": 1.5186, | |
| "step": 755 | |
| }, | |
| { | |
| "epoch": 0.36, | |
| "learning_rate": 0.0003043735209603025, | |
| "loss": 1.3627, | |
| "step": 756 | |
| }, | |
| { | |
| "epoch": 0.36, | |
| "learning_rate": 0.00030410766696986667, | |
| "loss": 1.3874, | |
| "step": 757 | |
| }, | |
| { | |
| "epoch": 0.36, | |
| "learning_rate": 0.00030384156041593795, | |
| "loss": 1.3036, | |
| "step": 758 | |
| }, | |
| { | |
| "epoch": 0.36, | |
| "learning_rate": 0.0003035752019440863, | |
| "loss": 1.2132, | |
| "step": 759 | |
| }, | |
| { | |
| "epoch": 0.36, | |
| "learning_rate": 0.00030330859220049303, | |
| "loss": 1.299, | |
| "step": 760 | |
| }, | |
| { | |
| "epoch": 0.36, | |
| "learning_rate": 0.0003030417318319491, | |
| "loss": 1.2785, | |
| "step": 761 | |
| }, | |
| { | |
| "epoch": 0.36, | |
| "learning_rate": 0.00030277462148585324, | |
| "loss": 1.1849, | |
| "step": 762 | |
| }, | |
| { | |
| "epoch": 0.36, | |
| "learning_rate": 0.00030250726181021097, | |
| "loss": 1.3676, | |
| "step": 763 | |
| }, | |
| { | |
| "epoch": 0.36, | |
| "learning_rate": 0.00030223965345363234, | |
| "loss": 1.5585, | |
| "step": 764 | |
| }, | |
| { | |
| "epoch": 0.36, | |
| "learning_rate": 0.0003019717970653309, | |
| "loss": 1.5943, | |
| "step": 765 | |
| }, | |
| { | |
| "epoch": 0.36, | |
| "learning_rate": 0.00030170369329512173, | |
| "loss": 1.0816, | |
| "step": 766 | |
| }, | |
| { | |
| "epoch": 0.36, | |
| "learning_rate": 0.00030143534279342035, | |
| "loss": 1.5398, | |
| "step": 767 | |
| }, | |
| { | |
| "epoch": 0.36, | |
| "learning_rate": 0.0003011667462112406, | |
| "loss": 1.2893, | |
| "step": 768 | |
| }, | |
| { | |
| "epoch": 0.36, | |
| "learning_rate": 0.00030089790420019335, | |
| "loss": 1.3099, | |
| "step": 769 | |
| }, | |
| { | |
| "epoch": 0.36, | |
| "learning_rate": 0.0003006288174124849, | |
| "loss": 1.5297, | |
| "step": 770 | |
| }, | |
| { | |
| "epoch": 0.36, | |
| "learning_rate": 0.00030035948650091555, | |
| "loss": 1.2783, | |
| "step": 771 | |
| }, | |
| { | |
| "epoch": 0.36, | |
| "learning_rate": 0.0003000899121188775, | |
| "loss": 1.3238, | |
| "step": 772 | |
| }, | |
| { | |
| "epoch": 0.37, | |
| "learning_rate": 0.0002998200949203538, | |
| "loss": 1.6145, | |
| "step": 773 | |
| }, | |
| { | |
| "epoch": 0.37, | |
| "learning_rate": 0.0002995500355599167, | |
| "loss": 1.2022, | |
| "step": 774 | |
| }, | |
| { | |
| "epoch": 0.37, | |
| "learning_rate": 0.00029927973469272555, | |
| "loss": 1.2806, | |
| "step": 775 | |
| }, | |
| { | |
| "epoch": 0.37, | |
| "learning_rate": 0.000299009192974526, | |
| "loss": 1.424, | |
| "step": 776 | |
| }, | |
| { | |
| "epoch": 0.37, | |
| "learning_rate": 0.00029873841106164766, | |
| "loss": 1.177, | |
| "step": 777 | |
| }, | |
| { | |
| "epoch": 0.37, | |
| "learning_rate": 0.0002984673896110032, | |
| "loss": 1.3138, | |
| "step": 778 | |
| }, | |
| { | |
| "epoch": 0.37, | |
| "learning_rate": 0.00029819612928008607, | |
| "loss": 1.3856, | |
| "step": 779 | |
| }, | |
| { | |
| "epoch": 0.37, | |
| "learning_rate": 0.0002979246307269695, | |
| "loss": 1.2792, | |
| "step": 780 | |
| }, | |
| { | |
| "epoch": 0.37, | |
| "learning_rate": 0.00029765289461030443, | |
| "loss": 1.2964, | |
| "step": 781 | |
| }, | |
| { | |
| "epoch": 0.37, | |
| "learning_rate": 0.0002973809215893183, | |
| "loss": 1.1395, | |
| "step": 782 | |
| }, | |
| { | |
| "epoch": 0.37, | |
| "learning_rate": 0.0002971087123238131, | |
| "loss": 1.3927, | |
| "step": 783 | |
| }, | |
| { | |
| "epoch": 0.37, | |
| "learning_rate": 0.00029683626747416416, | |
| "loss": 1.241, | |
| "step": 784 | |
| }, | |
| { | |
| "epoch": 0.37, | |
| "learning_rate": 0.0002965635877013181, | |
| "loss": 1.5876, | |
| "step": 785 | |
| }, | |
| { | |
| "epoch": 0.37, | |
| "learning_rate": 0.00029629067366679173, | |
| "loss": 1.5482, | |
| "step": 786 | |
| }, | |
| { | |
| "epoch": 0.37, | |
| "learning_rate": 0.0002960175260326699, | |
| "loss": 1.4041, | |
| "step": 787 | |
| }, | |
| { | |
| "epoch": 0.37, | |
| "learning_rate": 0.00029574414546160426, | |
| "loss": 1.2867, | |
| "step": 788 | |
| }, | |
| { | |
| "epoch": 0.37, | |
| "learning_rate": 0.0002954705326168117, | |
| "loss": 1.1729, | |
| "step": 789 | |
| }, | |
| { | |
| "epoch": 0.37, | |
| "learning_rate": 0.0002951966881620725, | |
| "loss": 1.4974, | |
| "step": 790 | |
| }, | |
| { | |
| "epoch": 0.37, | |
| "learning_rate": 0.0002949226127617287, | |
| "loss": 1.4537, | |
| "step": 791 | |
| }, | |
| { | |
| "epoch": 0.37, | |
| "learning_rate": 0.00029464830708068277, | |
| "loss": 1.2313, | |
| "step": 792 | |
| }, | |
| { | |
| "epoch": 0.37, | |
| "learning_rate": 0.0002943737717843959, | |
| "loss": 1.4326, | |
| "step": 793 | |
| }, | |
| { | |
| "epoch": 0.37, | |
| "learning_rate": 0.0002940990075388861, | |
| "loss": 1.249, | |
| "step": 794 | |
| }, | |
| { | |
| "epoch": 0.38, | |
| "learning_rate": 0.000293824015010727, | |
| "loss": 1.3426, | |
| "step": 795 | |
| }, | |
| { | |
| "epoch": 0.38, | |
| "learning_rate": 0.0002935487948670459, | |
| "loss": 1.35, | |
| "step": 796 | |
| }, | |
| { | |
| "epoch": 0.38, | |
| "learning_rate": 0.00029327334777552245, | |
| "loss": 1.1759, | |
| "step": 797 | |
| }, | |
| { | |
| "epoch": 0.38, | |
| "learning_rate": 0.0002929976744043866, | |
| "loss": 1.181, | |
| "step": 798 | |
| }, | |
| { | |
| "epoch": 0.38, | |
| "learning_rate": 0.00029272177542241765, | |
| "loss": 1.1735, | |
| "step": 799 | |
| }, | |
| { | |
| "epoch": 0.38, | |
| "learning_rate": 0.00029244565149894183, | |
| "loss": 0.9487, | |
| "step": 800 | |
| }, | |
| { | |
| "epoch": 0.38, | |
| "learning_rate": 0.00029216930330383127, | |
| "loss": 1.4527, | |
| "step": 801 | |
| }, | |
| { | |
| "epoch": 0.38, | |
| "learning_rate": 0.0002918927315075023, | |
| "loss": 1.4078, | |
| "step": 802 | |
| }, | |
| { | |
| "epoch": 0.38, | |
| "learning_rate": 0.00029161593678091337, | |
| "loss": 1.2437, | |
| "step": 803 | |
| }, | |
| { | |
| "epoch": 0.38, | |
| "learning_rate": 0.00029133891979556413, | |
| "loss": 1.2497, | |
| "step": 804 | |
| }, | |
| { | |
| "epoch": 0.38, | |
| "learning_rate": 0.0002910616812234931, | |
| "loss": 1.1957, | |
| "step": 805 | |
| }, | |
| { | |
| "epoch": 0.38, | |
| "learning_rate": 0.00029078422173727664, | |
| "loss": 1.3722, | |
| "step": 806 | |
| }, | |
| { | |
| "epoch": 0.38, | |
| "learning_rate": 0.0002905065420100267, | |
| "loss": 1.2857, | |
| "step": 807 | |
| }, | |
| { | |
| "epoch": 0.38, | |
| "learning_rate": 0.00029022864271539004, | |
| "loss": 1.099, | |
| "step": 808 | |
| }, | |
| { | |
| "epoch": 0.38, | |
| "learning_rate": 0.0002899505245275456, | |
| "loss": 1.4149, | |
| "step": 809 | |
| }, | |
| { | |
| "epoch": 0.38, | |
| "learning_rate": 0.0002896721881212036, | |
| "loss": 1.3387, | |
| "step": 810 | |
| }, | |
| { | |
| "epoch": 0.38, | |
| "learning_rate": 0.0002893936341716035, | |
| "loss": 1.6126, | |
| "step": 811 | |
| }, | |
| { | |
| "epoch": 0.38, | |
| "learning_rate": 0.0002891148633545128, | |
| "loss": 1.2203, | |
| "step": 812 | |
| }, | |
| { | |
| "epoch": 0.38, | |
| "learning_rate": 0.0002888358763462248, | |
| "loss": 1.3316, | |
| "step": 813 | |
| }, | |
| { | |
| "epoch": 0.38, | |
| "learning_rate": 0.0002885566738235574, | |
| "loss": 1.2368, | |
| "step": 814 | |
| }, | |
| { | |
| "epoch": 0.38, | |
| "learning_rate": 0.0002882772564638515, | |
| "loss": 1.2352, | |
| "step": 815 | |
| }, | |
| { | |
| "epoch": 0.39, | |
| "learning_rate": 0.00028799762494496873, | |
| "loss": 1.2296, | |
| "step": 816 | |
| }, | |
| { | |
| "epoch": 0.39, | |
| "learning_rate": 0.00028771777994529086, | |
| "loss": 1.3651, | |
| "step": 817 | |
| }, | |
| { | |
| "epoch": 0.39, | |
| "learning_rate": 0.00028743772214371715, | |
| "loss": 1.5197, | |
| "step": 818 | |
| }, | |
| { | |
| "epoch": 0.39, | |
| "learning_rate": 0.00028715745221966326, | |
| "loss": 1.5716, | |
| "step": 819 | |
| }, | |
| { | |
| "epoch": 0.39, | |
| "learning_rate": 0.00028687697085305933, | |
| "loss": 1.2013, | |
| "step": 820 | |
| }, | |
| { | |
| "epoch": 0.39, | |
| "learning_rate": 0.0002865962787243487, | |
| "loss": 1.4137, | |
| "step": 821 | |
| }, | |
| { | |
| "epoch": 0.39, | |
| "learning_rate": 0.0002863153765144858, | |
| "loss": 1.2877, | |
| "step": 822 | |
| }, | |
| { | |
| "epoch": 0.39, | |
| "learning_rate": 0.00028603426490493485, | |
| "loss": 1.4059, | |
| "step": 823 | |
| }, | |
| { | |
| "epoch": 0.39, | |
| "learning_rate": 0.00028575294457766794, | |
| "loss": 1.1746, | |
| "step": 824 | |
| }, | |
| { | |
| "epoch": 0.39, | |
| "learning_rate": 0.00028547141621516355, | |
| "loss": 1.673, | |
| "step": 825 | |
| }, | |
| { | |
| "epoch": 0.39, | |
| "learning_rate": 0.0002851896805004049, | |
| "loss": 1.4219, | |
| "step": 826 | |
| }, | |
| { | |
| "epoch": 0.39, | |
| "learning_rate": 0.00028490773811687826, | |
| "loss": 1.4337, | |
| "step": 827 | |
| }, | |
| { | |
| "epoch": 0.39, | |
| "learning_rate": 0.0002846255897485712, | |
| "loss": 1.1064, | |
| "step": 828 | |
| }, | |
| { | |
| "epoch": 0.39, | |
| "learning_rate": 0.000284343236079971, | |
| "loss": 1.2186, | |
| "step": 829 | |
| }, | |
| { | |
| "epoch": 0.39, | |
| "learning_rate": 0.000284060677796063, | |
| "loss": 1.2138, | |
| "step": 830 | |
| }, | |
| { | |
| "epoch": 0.39, | |
| "learning_rate": 0.0002837779155823291, | |
| "loss": 1.156, | |
| "step": 831 | |
| }, | |
| { | |
| "epoch": 0.39, | |
| "learning_rate": 0.0002834949501247457, | |
| "loss": 1.3292, | |
| "step": 832 | |
| }, | |
| { | |
| "epoch": 0.39, | |
| "learning_rate": 0.00028321178210978233, | |
| "loss": 1.472, | |
| "step": 833 | |
| }, | |
| { | |
| "epoch": 0.39, | |
| "learning_rate": 0.0002829284122244, | |
| "loss": 1.4211, | |
| "step": 834 | |
| }, | |
| { | |
| "epoch": 0.39, | |
| "learning_rate": 0.0002826448411560494, | |
| "loss": 1.0677, | |
| "step": 835 | |
| }, | |
| { | |
| "epoch": 0.39, | |
| "learning_rate": 0.0002823610695926694, | |
| "loss": 1.5694, | |
| "step": 836 | |
| }, | |
| { | |
| "epoch": 0.4, | |
| "learning_rate": 0.0002820770982226849, | |
| "loss": 1.1922, | |
| "step": 837 | |
| }, | |
| { | |
| "epoch": 0.4, | |
| "learning_rate": 0.00028179292773500606, | |
| "loss": 1.379, | |
| "step": 838 | |
| }, | |
| { | |
| "epoch": 0.4, | |
| "learning_rate": 0.00028150855881902564, | |
| "loss": 1.3802, | |
| "step": 839 | |
| }, | |
| { | |
| "epoch": 0.4, | |
| "learning_rate": 0.0002812239921646181, | |
| "loss": 1.3687, | |
| "step": 840 | |
| }, | |
| { | |
| "epoch": 0.4, | |
| "learning_rate": 0.00028093922846213735, | |
| "loss": 1.3385, | |
| "step": 841 | |
| }, | |
| { | |
| "epoch": 0.4, | |
| "learning_rate": 0.00028065426840241545, | |
| "loss": 1.0829, | |
| "step": 842 | |
| }, | |
| { | |
| "epoch": 0.4, | |
| "learning_rate": 0.00028036911267676103, | |
| "loss": 1.3422, | |
| "step": 843 | |
| }, | |
| { | |
| "epoch": 0.4, | |
| "learning_rate": 0.00028008376197695694, | |
| "loss": 1.3711, | |
| "step": 844 | |
| }, | |
| { | |
| "epoch": 0.4, | |
| "learning_rate": 0.0002797982169952596, | |
| "loss": 1.6293, | |
| "step": 845 | |
| }, | |
| { | |
| "epoch": 0.4, | |
| "learning_rate": 0.00027951247842439617, | |
| "loss": 1.2781, | |
| "step": 846 | |
| }, | |
| { | |
| "epoch": 0.4, | |
| "learning_rate": 0.0002792265469575639, | |
| "loss": 1.4301, | |
| "step": 847 | |
| }, | |
| { | |
| "epoch": 0.4, | |
| "learning_rate": 0.00027894042328842773, | |
| "loss": 1.2093, | |
| "step": 848 | |
| }, | |
| { | |
| "epoch": 0.4, | |
| "learning_rate": 0.00027865410811111916, | |
| "loss": 1.5235, | |
| "step": 849 | |
| }, | |
| { | |
| "epoch": 0.4, | |
| "learning_rate": 0.000278367602120234, | |
| "loss": 1.0747, | |
| "step": 850 | |
| }, | |
| { | |
| "epoch": 0.4, | |
| "learning_rate": 0.0002780809060108311, | |
| "loss": 1.5449, | |
| "step": 851 | |
| }, | |
| { | |
| "epoch": 0.4, | |
| "learning_rate": 0.00027779402047843046, | |
| "loss": 1.5861, | |
| "step": 852 | |
| }, | |
| { | |
| "epoch": 0.4, | |
| "learning_rate": 0.0002775069462190118, | |
| "loss": 1.4148, | |
| "step": 853 | |
| }, | |
| { | |
| "epoch": 0.4, | |
| "learning_rate": 0.0002772196839290124, | |
| "loss": 1.1843, | |
| "step": 854 | |
| }, | |
| { | |
| "epoch": 0.4, | |
| "learning_rate": 0.0002769322343053261, | |
| "loss": 1.3968, | |
| "step": 855 | |
| }, | |
| { | |
| "epoch": 0.4, | |
| "learning_rate": 0.00027664459804530077, | |
| "loss": 1.517, | |
| "step": 856 | |
| }, | |
| { | |
| "epoch": 0.4, | |
| "learning_rate": 0.00027635677584673734, | |
| "loss": 1.1604, | |
| "step": 857 | |
| }, | |
| { | |
| "epoch": 0.41, | |
| "learning_rate": 0.0002760687684078877, | |
| "loss": 1.0792, | |
| "step": 858 | |
| }, | |
| { | |
| "epoch": 0.41, | |
| "learning_rate": 0.0002757805764274533, | |
| "loss": 1.1854, | |
| "step": 859 | |
| }, | |
| { | |
| "epoch": 0.41, | |
| "learning_rate": 0.00027549220060458303, | |
| "loss": 1.099, | |
| "step": 860 | |
| }, | |
| { | |
| "epoch": 0.41, | |
| "learning_rate": 0.00027520364163887193, | |
| "loss": 1.3248, | |
| "step": 861 | |
| }, | |
| { | |
| "epoch": 0.41, | |
| "learning_rate": 0.00027491490023035935, | |
| "loss": 1.3282, | |
| "step": 862 | |
| }, | |
| { | |
| "epoch": 0.41, | |
| "learning_rate": 0.00027462597707952717, | |
| "loss": 1.305, | |
| "step": 863 | |
| }, | |
| { | |
| "epoch": 0.41, | |
| "learning_rate": 0.0002743368728872983, | |
| "loss": 1.4731, | |
| "step": 864 | |
| }, | |
| { | |
| "epoch": 0.41, | |
| "learning_rate": 0.0002740475883550347, | |
| "loss": 1.1555, | |
| "step": 865 | |
| }, | |
| { | |
| "epoch": 0.41, | |
| "learning_rate": 0.00027375812418453584, | |
| "loss": 1.2658, | |
| "step": 866 | |
| }, | |
| { | |
| "epoch": 0.41, | |
| "learning_rate": 0.00027346848107803723, | |
| "loss": 1.1447, | |
| "step": 867 | |
| }, | |
| { | |
| "epoch": 0.41, | |
| "learning_rate": 0.0002731786597382082, | |
| "loss": 1.3056, | |
| "step": 868 | |
| }, | |
| { | |
| "epoch": 0.41, | |
| "learning_rate": 0.0002728886608681505, | |
| "loss": 1.2856, | |
| "step": 869 | |
| }, | |
| { | |
| "epoch": 0.41, | |
| "learning_rate": 0.0002725984851713967, | |
| "loss": 1.4002, | |
| "step": 870 | |
| }, | |
| { | |
| "epoch": 0.41, | |
| "learning_rate": 0.0002723081333519083, | |
| "loss": 1.0462, | |
| "step": 871 | |
| }, | |
| { | |
| "epoch": 0.41, | |
| "learning_rate": 0.0002720176061140741, | |
| "loss": 1.3978, | |
| "step": 872 | |
| }, | |
| { | |
| "epoch": 0.41, | |
| "learning_rate": 0.0002717269041627082, | |
| "loss": 1.3281, | |
| "step": 873 | |
| }, | |
| { | |
| "epoch": 0.41, | |
| "learning_rate": 0.000271436028203049, | |
| "loss": 1.3246, | |
| "step": 874 | |
| }, | |
| { | |
| "epoch": 0.41, | |
| "learning_rate": 0.00027114497894075676, | |
| "loss": 1.0361, | |
| "step": 875 | |
| }, | |
| { | |
| "epoch": 0.41, | |
| "learning_rate": 0.00027085375708191207, | |
| "loss": 1.1194, | |
| "step": 876 | |
| }, | |
| { | |
| "epoch": 0.41, | |
| "learning_rate": 0.0002705623633330146, | |
| "loss": 1.2326, | |
| "step": 877 | |
| }, | |
| { | |
| "epoch": 0.41, | |
| "learning_rate": 0.00027027079840098066, | |
| "loss": 1.324, | |
| "step": 878 | |
| }, | |
| { | |
| "epoch": 0.42, | |
| "learning_rate": 0.0002699790629931421, | |
| "loss": 1.2296, | |
| "step": 879 | |
| }, | |
| { | |
| "epoch": 0.42, | |
| "learning_rate": 0.00026968715781724404, | |
| "loss": 1.1764, | |
| "step": 880 | |
| }, | |
| { | |
| "epoch": 0.42, | |
| "learning_rate": 0.000269395083581444, | |
| "loss": 1.2608, | |
| "step": 881 | |
| }, | |
| { | |
| "epoch": 0.42, | |
| "learning_rate": 0.0002691028409943089, | |
| "loss": 1.3877, | |
| "step": 882 | |
| }, | |
| { | |
| "epoch": 0.42, | |
| "learning_rate": 0.00026881043076481463, | |
| "loss": 1.1516, | |
| "step": 883 | |
| }, | |
| { | |
| "epoch": 0.42, | |
| "learning_rate": 0.0002685178536023437, | |
| "loss": 1.4622, | |
| "step": 884 | |
| }, | |
| { | |
| "epoch": 0.42, | |
| "learning_rate": 0.0002682251102166833, | |
| "loss": 1.3839, | |
| "step": 885 | |
| }, | |
| { | |
| "epoch": 0.42, | |
| "learning_rate": 0.0002679322013180244, | |
| "loss": 1.3128, | |
| "step": 886 | |
| }, | |
| { | |
| "epoch": 0.42, | |
| "learning_rate": 0.00026763912761695885, | |
| "loss": 1.2344, | |
| "step": 887 | |
| }, | |
| { | |
| "epoch": 0.42, | |
| "learning_rate": 0.00026734588982447885, | |
| "loss": 1.4078, | |
| "step": 888 | |
| }, | |
| { | |
| "epoch": 0.42, | |
| "learning_rate": 0.00026705248865197434, | |
| "loss": 1.0956, | |
| "step": 889 | |
| }, | |
| { | |
| "epoch": 0.42, | |
| "learning_rate": 0.0002667589248112319, | |
| "loss": 1.5464, | |
| "step": 890 | |
| }, | |
| { | |
| "epoch": 0.42, | |
| "learning_rate": 0.0002664651990144325, | |
| "loss": 1.227, | |
| "step": 891 | |
| }, | |
| { | |
| "epoch": 0.42, | |
| "learning_rate": 0.0002661713119741502, | |
| "loss": 1.2573, | |
| "step": 892 | |
| }, | |
| { | |
| "epoch": 0.42, | |
| "learning_rate": 0.00026587726440334994, | |
| "loss": 1.3705, | |
| "step": 893 | |
| }, | |
| { | |
| "epoch": 0.42, | |
| "learning_rate": 0.00026558305701538645, | |
| "loss": 1.6353, | |
| "step": 894 | |
| }, | |
| { | |
| "epoch": 0.42, | |
| "learning_rate": 0.0002652886905240021, | |
| "loss": 1.439, | |
| "step": 895 | |
| }, | |
| { | |
| "epoch": 0.42, | |
| "learning_rate": 0.00026499416564332507, | |
| "loss": 1.6142, | |
| "step": 896 | |
| }, | |
| { | |
| "epoch": 0.42, | |
| "learning_rate": 0.00026469948308786785, | |
| "loss": 1.4789, | |
| "step": 897 | |
| }, | |
| { | |
| "epoch": 0.42, | |
| "learning_rate": 0.00026440464357252555, | |
| "loss": 1.3731, | |
| "step": 898 | |
| }, | |
| { | |
| "epoch": 0.42, | |
| "learning_rate": 0.00026410964781257404, | |
| "loss": 1.2835, | |
| "step": 899 | |
| }, | |
| { | |
| "epoch": 0.42, | |
| "learning_rate": 0.0002638144965236681, | |
| "loss": 1.4461, | |
| "step": 900 | |
| }, | |
| { | |
| "epoch": 0.43, | |
| "learning_rate": 0.00026351919042184006, | |
| "loss": 1.3314, | |
| "step": 901 | |
| }, | |
| { | |
| "epoch": 0.43, | |
| "learning_rate": 0.0002632237302234975, | |
| "loss": 1.4887, | |
| "step": 902 | |
| }, | |
| { | |
| "epoch": 0.43, | |
| "learning_rate": 0.0002629281166454221, | |
| "loss": 1.474, | |
| "step": 903 | |
| }, | |
| { | |
| "epoch": 0.43, | |
| "learning_rate": 0.0002626323504047676, | |
| "loss": 1.6467, | |
| "step": 904 | |
| }, | |
| { | |
| "epoch": 0.43, | |
| "learning_rate": 0.000262336432219058, | |
| "loss": 1.4272, | |
| "step": 905 | |
| }, | |
| { | |
| "epoch": 0.43, | |
| "learning_rate": 0.0002620403628061859, | |
| "loss": 1.3052, | |
| "step": 906 | |
| }, | |
| { | |
| "epoch": 0.43, | |
| "learning_rate": 0.000261744142884411, | |
| "loss": 1.1621, | |
| "step": 907 | |
| }, | |
| { | |
| "epoch": 0.43, | |
| "learning_rate": 0.0002614477731723579, | |
| "loss": 1.289, | |
| "step": 908 | |
| }, | |
| { | |
| "epoch": 0.43, | |
| "learning_rate": 0.0002611512543890146, | |
| "loss": 0.7957, | |
| "step": 909 | |
| }, | |
| { | |
| "epoch": 0.43, | |
| "learning_rate": 0.0002608545872537309, | |
| "loss": 1.1423, | |
| "step": 910 | |
| }, | |
| { | |
| "epoch": 0.43, | |
| "learning_rate": 0.0002605577724862163, | |
| "loss": 1.3153, | |
| "step": 911 | |
| }, | |
| { | |
| "epoch": 0.43, | |
| "learning_rate": 0.0002602608108065387, | |
| "loss": 1.1333, | |
| "step": 912 | |
| }, | |
| { | |
| "epoch": 0.43, | |
| "learning_rate": 0.00025996370293512216, | |
| "loss": 1.1941, | |
| "step": 913 | |
| }, | |
| { | |
| "epoch": 0.43, | |
| "learning_rate": 0.00025966644959274555, | |
| "loss": 1.2343, | |
| "step": 914 | |
| }, | |
| { | |
| "epoch": 0.43, | |
| "learning_rate": 0.00025936905150054045, | |
| "loss": 1.3379, | |
| "step": 915 | |
| }, | |
| { | |
| "epoch": 0.43, | |
| "learning_rate": 0.00025907150937999005, | |
| "loss": 1.3146, | |
| "step": 916 | |
| }, | |
| { | |
| "epoch": 0.43, | |
| "learning_rate": 0.0002587738239529264, | |
| "loss": 1.4962, | |
| "step": 917 | |
| }, | |
| { | |
| "epoch": 0.43, | |
| "learning_rate": 0.0002584759959415297, | |
| "loss": 1.4088, | |
| "step": 918 | |
| }, | |
| { | |
| "epoch": 0.43, | |
| "learning_rate": 0.00025817802606832547, | |
| "loss": 1.2505, | |
| "step": 919 | |
| }, | |
| { | |
| "epoch": 0.43, | |
| "learning_rate": 0.0002578799150561841, | |
| "loss": 1.0342, | |
| "step": 920 | |
| }, | |
| { | |
| "epoch": 0.43, | |
| "learning_rate": 0.00025758166362831777, | |
| "loss": 1.2603, | |
| "step": 921 | |
| }, | |
| { | |
| "epoch": 0.44, | |
| "learning_rate": 0.0002572832725082798, | |
| "loss": 1.2303, | |
| "step": 922 | |
| }, | |
| { | |
| "epoch": 0.44, | |
| "learning_rate": 0.0002569847424199619, | |
| "loss": 1.351, | |
| "step": 923 | |
| }, | |
| { | |
| "epoch": 0.44, | |
| "learning_rate": 0.00025668607408759327, | |
| "loss": 1.1902, | |
| "step": 924 | |
| }, | |
| { | |
| "epoch": 0.44, | |
| "learning_rate": 0.00025638726823573843, | |
| "loss": 1.5928, | |
| "step": 925 | |
| }, | |
| { | |
| "epoch": 0.44, | |
| "learning_rate": 0.0002560883255892954, | |
| "loss": 1.1507, | |
| "step": 926 | |
| }, | |
| { | |
| "epoch": 0.44, | |
| "learning_rate": 0.0002557892468734942, | |
| "loss": 1.2932, | |
| "step": 927 | |
| }, | |
| { | |
| "epoch": 0.44, | |
| "learning_rate": 0.0002554900328138949, | |
| "loss": 1.4027, | |
| "step": 928 | |
| }, | |
| { | |
| "epoch": 0.44, | |
| "learning_rate": 0.00025519068413638586, | |
| "loss": 1.2834, | |
| "step": 929 | |
| }, | |
| { | |
| "epoch": 0.44, | |
| "learning_rate": 0.000254891201567182, | |
| "loss": 1.1498, | |
| "step": 930 | |
| }, | |
| { | |
| "epoch": 0.44, | |
| "learning_rate": 0.00025459158583282325, | |
| "loss": 1.2244, | |
| "step": 931 | |
| }, | |
| { | |
| "epoch": 0.44, | |
| "learning_rate": 0.0002542918376601724, | |
| "loss": 1.3832, | |
| "step": 932 | |
| }, | |
| { | |
| "epoch": 0.44, | |
| "learning_rate": 0.00025399195777641364, | |
| "loss": 1.3634, | |
| "step": 933 | |
| }, | |
| { | |
| "epoch": 0.44, | |
| "learning_rate": 0.00025369194690905063, | |
| "loss": 1.243, | |
| "step": 934 | |
| }, | |
| { | |
| "epoch": 0.44, | |
| "learning_rate": 0.0002533918057859048, | |
| "loss": 0.9249, | |
| "step": 935 | |
| }, | |
| { | |
| "epoch": 0.44, | |
| "learning_rate": 0.0002530915351351136, | |
| "loss": 1.4711, | |
| "step": 936 | |
| }, | |
| { | |
| "epoch": 0.44, | |
| "learning_rate": 0.0002527911356851287, | |
| "loss": 1.2048, | |
| "step": 937 | |
| }, | |
| { | |
| "epoch": 0.44, | |
| "learning_rate": 0.0002524906081647143, | |
| "loss": 1.2095, | |
| "step": 938 | |
| }, | |
| { | |
| "epoch": 0.44, | |
| "learning_rate": 0.0002521899533029451, | |
| "loss": 1.3093, | |
| "step": 939 | |
| }, | |
| { | |
| "epoch": 0.44, | |
| "learning_rate": 0.000251889171829205, | |
| "loss": 1.6384, | |
| "step": 940 | |
| }, | |
| { | |
| "epoch": 0.44, | |
| "learning_rate": 0.0002515882644731849, | |
| "loss": 1.1837, | |
| "step": 941 | |
| }, | |
| { | |
| "epoch": 0.44, | |
| "learning_rate": 0.00025128723196488105, | |
| "loss": 1.2502, | |
| "step": 942 | |
| }, | |
| { | |
| "epoch": 0.45, | |
| "learning_rate": 0.00025098607503459336, | |
| "loss": 1.3219, | |
| "step": 943 | |
| }, | |
| { | |
| "epoch": 0.45, | |
| "learning_rate": 0.0002506847944129238, | |
| "loss": 1.4386, | |
| "step": 944 | |
| }, | |
| { | |
| "epoch": 0.45, | |
| "learning_rate": 0.000250383390830774, | |
| "loss": 1.2891, | |
| "step": 945 | |
| }, | |
| { | |
| "epoch": 0.45, | |
| "learning_rate": 0.0002500818650193442, | |
| "loss": 1.1171, | |
| "step": 946 | |
| }, | |
| { | |
| "epoch": 0.45, | |
| "learning_rate": 0.000249780217710131, | |
| "loss": 1.112, | |
| "step": 947 | |
| }, | |
| { | |
| "epoch": 0.45, | |
| "learning_rate": 0.00024947844963492594, | |
| "loss": 1.0929, | |
| "step": 948 | |
| }, | |
| { | |
| "epoch": 0.45, | |
| "learning_rate": 0.0002491765615258134, | |
| "loss": 1.4064, | |
| "step": 949 | |
| }, | |
| { | |
| "epoch": 0.45, | |
| "learning_rate": 0.00024887455411516895, | |
| "loss": 1.2253, | |
| "step": 950 | |
| }, | |
| { | |
| "epoch": 0.45, | |
| "learning_rate": 0.0002485724281356576, | |
| "loss": 1.1958, | |
| "step": 951 | |
| }, | |
| { | |
| "epoch": 0.45, | |
| "learning_rate": 0.00024827018432023196, | |
| "loss": 1.3716, | |
| "step": 952 | |
| }, | |
| { | |
| "epoch": 0.45, | |
| "learning_rate": 0.00024796782340213075, | |
| "loss": 1.1603, | |
| "step": 953 | |
| }, | |
| { | |
| "epoch": 0.45, | |
| "learning_rate": 0.0002476653461148765, | |
| "loss": 1.2685, | |
| "step": 954 | |
| }, | |
| { | |
| "epoch": 0.45, | |
| "learning_rate": 0.0002473627531922741, | |
| "loss": 1.4317, | |
| "step": 955 | |
| }, | |
| { | |
| "epoch": 0.45, | |
| "learning_rate": 0.00024706004536840907, | |
| "loss": 1.3646, | |
| "step": 956 | |
| }, | |
| { | |
| "epoch": 0.45, | |
| "learning_rate": 0.00024675722337764565, | |
| "loss": 1.2738, | |
| "step": 957 | |
| }, | |
| { | |
| "epoch": 0.45, | |
| "learning_rate": 0.000246454287954625, | |
| "loss": 1.3745, | |
| "step": 958 | |
| }, | |
| { | |
| "epoch": 0.45, | |
| "learning_rate": 0.0002461512398342635, | |
| "loss": 1.2735, | |
| "step": 959 | |
| }, | |
| { | |
| "epoch": 0.45, | |
| "learning_rate": 0.0002458480797517509, | |
| "loss": 1.3064, | |
| "step": 960 | |
| }, | |
| { | |
| "epoch": 0.45, | |
| "learning_rate": 0.0002455448084425486, | |
| "loss": 1.5011, | |
| "step": 961 | |
| }, | |
| { | |
| "epoch": 0.45, | |
| "learning_rate": 0.00024524142664238775, | |
| "loss": 1.3956, | |
| "step": 962 | |
| }, | |
| { | |
| "epoch": 0.45, | |
| "learning_rate": 0.0002449379350872678, | |
| "loss": 0.9758, | |
| "step": 963 | |
| }, | |
| { | |
| "epoch": 0.46, | |
| "learning_rate": 0.00024463433451345413, | |
| "loss": 1.2169, | |
| "step": 964 | |
| }, | |
| { | |
| "epoch": 0.46, | |
| "learning_rate": 0.0002443306256574768, | |
| "loss": 1.5301, | |
| "step": 965 | |
| }, | |
| { | |
| "epoch": 0.46, | |
| "learning_rate": 0.00024402680925612855, | |
| "loss": 1.2099, | |
| "step": 966 | |
| }, | |
| { | |
| "epoch": 0.46, | |
| "learning_rate": 0.0002437228860464629, | |
| "loss": 1.4701, | |
| "step": 967 | |
| }, | |
| { | |
| "epoch": 0.46, | |
| "learning_rate": 0.00024341885676579272, | |
| "loss": 1.3601, | |
| "step": 968 | |
| }, | |
| { | |
| "epoch": 0.46, | |
| "learning_rate": 0.0002431147221516879, | |
| "loss": 1.1317, | |
| "step": 969 | |
| }, | |
| { | |
| "epoch": 0.46, | |
| "learning_rate": 0.00024281048294197413, | |
| "loss": 1.1306, | |
| "step": 970 | |
| }, | |
| { | |
| "epoch": 0.46, | |
| "learning_rate": 0.00024250613987473067, | |
| "loss": 1.3336, | |
| "step": 971 | |
| }, | |
| { | |
| "epoch": 0.46, | |
| "learning_rate": 0.00024220169368828888, | |
| "loss": 1.2368, | |
| "step": 972 | |
| }, | |
| { | |
| "epoch": 0.46, | |
| "learning_rate": 0.0002418971451212302, | |
| "loss": 1.256, | |
| "step": 973 | |
| }, | |
| { | |
| "epoch": 0.46, | |
| "learning_rate": 0.00024159249491238446, | |
| "loss": 1.3063, | |
| "step": 974 | |
| }, | |
| { | |
| "epoch": 0.46, | |
| "learning_rate": 0.000241287743800828, | |
| "loss": 1.2309, | |
| "step": 975 | |
| }, | |
| { | |
| "epoch": 0.46, | |
| "learning_rate": 0.00024098289252588208, | |
| "loss": 1.2604, | |
| "step": 976 | |
| }, | |
| { | |
| "epoch": 0.46, | |
| "learning_rate": 0.00024067794182711097, | |
| "loss": 1.4222, | |
| "step": 977 | |
| }, | |
| { | |
| "epoch": 0.46, | |
| "learning_rate": 0.00024037289244431995, | |
| "loss": 1.5747, | |
| "step": 978 | |
| }, | |
| { | |
| "epoch": 0.46, | |
| "learning_rate": 0.00024006774511755387, | |
| "loss": 1.3778, | |
| "step": 979 | |
| }, | |
| { | |
| "epoch": 0.46, | |
| "learning_rate": 0.00023976250058709505, | |
| "loss": 1.463, | |
| "step": 980 | |
| }, | |
| { | |
| "epoch": 0.46, | |
| "learning_rate": 0.00023945715959346192, | |
| "loss": 1.4231, | |
| "step": 981 | |
| }, | |
| { | |
| "epoch": 0.46, | |
| "learning_rate": 0.00023915172287740654, | |
| "loss": 1.3177, | |
| "step": 982 | |
| }, | |
| { | |
| "epoch": 0.46, | |
| "learning_rate": 0.00023884619117991345, | |
| "loss": 1.2004, | |
| "step": 983 | |
| }, | |
| { | |
| "epoch": 0.46, | |
| "learning_rate": 0.00023854056524219748, | |
| "loss": 1.1497, | |
| "step": 984 | |
| }, | |
| { | |
| "epoch": 0.47, | |
| "learning_rate": 0.0002382348458057022, | |
| "loss": 1.4359, | |
| "step": 985 | |
| }, | |
| { | |
| "epoch": 0.47, | |
| "learning_rate": 0.00023792903361209797, | |
| "loss": 1.169, | |
| "step": 986 | |
| }, | |
| { | |
| "epoch": 0.47, | |
| "learning_rate": 0.00023762312940328012, | |
| "loss": 0.9957, | |
| "step": 987 | |
| }, | |
| { | |
| "epoch": 0.47, | |
| "learning_rate": 0.0002373171339213672, | |
| "loss": 1.4934, | |
| "step": 988 | |
| }, | |
| { | |
| "epoch": 0.47, | |
| "learning_rate": 0.00023701104790869933, | |
| "loss": 1.2416, | |
| "step": 989 | |
| }, | |
| { | |
| "epoch": 0.47, | |
| "learning_rate": 0.0002367048721078361, | |
| "loss": 1.0249, | |
| "step": 990 | |
| }, | |
| { | |
| "epoch": 0.47, | |
| "learning_rate": 0.00023639860726155497, | |
| "loss": 1.4124, | |
| "step": 991 | |
| }, | |
| { | |
| "epoch": 0.47, | |
| "learning_rate": 0.0002360922541128495, | |
| "loss": 1.2547, | |
| "step": 992 | |
| }, | |
| { | |
| "epoch": 0.47, | |
| "learning_rate": 0.00023578581340492728, | |
| "loss": 1.3123, | |
| "step": 993 | |
| }, | |
| { | |
| "epoch": 0.47, | |
| "learning_rate": 0.00023547928588120863, | |
| "loss": 1.0849, | |
| "step": 994 | |
| }, | |
| { | |
| "epoch": 0.47, | |
| "learning_rate": 0.0002351726722853241, | |
| "loss": 1.3818, | |
| "step": 995 | |
| }, | |
| { | |
| "epoch": 0.47, | |
| "learning_rate": 0.00023486597336111337, | |
| "loss": 1.3689, | |
| "step": 996 | |
| }, | |
| { | |
| "epoch": 0.47, | |
| "learning_rate": 0.00023455918985262298, | |
| "loss": 1.5367, | |
| "step": 997 | |
| }, | |
| { | |
| "epoch": 0.47, | |
| "learning_rate": 0.00023425232250410467, | |
| "loss": 1.3149, | |
| "step": 998 | |
| }, | |
| { | |
| "epoch": 0.47, | |
| "learning_rate": 0.0002339453720600136, | |
| "loss": 1.4009, | |
| "step": 999 | |
| }, | |
| { | |
| "epoch": 0.47, | |
| "learning_rate": 0.0002336383392650065, | |
| "loss": 1.5073, | |
| "step": 1000 | |
| }, | |
| { | |
| "epoch": 0.47, | |
| "learning_rate": 0.00023333122486393992, | |
| "loss": 1.2735, | |
| "step": 1001 | |
| }, | |
| { | |
| "epoch": 0.47, | |
| "learning_rate": 0.00023302402960186837, | |
| "loss": 1.2892, | |
| "step": 1002 | |
| }, | |
| { | |
| "epoch": 0.47, | |
| "learning_rate": 0.00023271675422404249, | |
| "loss": 1.376, | |
| "step": 1003 | |
| }, | |
| { | |
| "epoch": 0.47, | |
| "learning_rate": 0.0002324093994759074, | |
| "loss": 1.4852, | |
| "step": 1004 | |
| }, | |
| { | |
| "epoch": 0.47, | |
| "learning_rate": 0.00023210196610310048, | |
| "loss": 1.4539, | |
| "step": 1005 | |
| }, | |
| { | |
| "epoch": 0.48, | |
| "learning_rate": 0.00023179445485145027, | |
| "loss": 1.1888, | |
| "step": 1006 | |
| }, | |
| { | |
| "epoch": 0.48, | |
| "learning_rate": 0.00023148686646697387, | |
| "loss": 1.109, | |
| "step": 1007 | |
| }, | |
| { | |
| "epoch": 0.48, | |
| "learning_rate": 0.00023117920169587572, | |
| "loss": 1.4764, | |
| "step": 1008 | |
| }, | |
| { | |
| "epoch": 0.48, | |
| "learning_rate": 0.00023087146128454556, | |
| "loss": 1.1469, | |
| "step": 1009 | |
| }, | |
| { | |
| "epoch": 0.48, | |
| "learning_rate": 0.00023056364597955648, | |
| "loss": 1.1258, | |
| "step": 1010 | |
| }, | |
| { | |
| "epoch": 0.48, | |
| "learning_rate": 0.00023025575652766341, | |
| "loss": 1.0952, | |
| "step": 1011 | |
| }, | |
| { | |
| "epoch": 0.48, | |
| "learning_rate": 0.0002299477936758011, | |
| "loss": 1.5308, | |
| "step": 1012 | |
| }, | |
| { | |
| "epoch": 0.48, | |
| "learning_rate": 0.00022963975817108238, | |
| "loss": 1.4326, | |
| "step": 1013 | |
| }, | |
| { | |
| "epoch": 0.48, | |
| "learning_rate": 0.00022933165076079642, | |
| "loss": 1.203, | |
| "step": 1014 | |
| }, | |
| { | |
| "epoch": 0.48, | |
| "learning_rate": 0.00022902347219240664, | |
| "loss": 1.1518, | |
| "step": 1015 | |
| }, | |
| { | |
| "epoch": 0.48, | |
| "learning_rate": 0.00022871522321354912, | |
| "loss": 1.365, | |
| "step": 1016 | |
| }, | |
| { | |
| "epoch": 0.48, | |
| "learning_rate": 0.00022840690457203106, | |
| "loss": 1.1922, | |
| "step": 1017 | |
| }, | |
| { | |
| "epoch": 0.48, | |
| "learning_rate": 0.00022809851701582825, | |
| "loss": 1.3392, | |
| "step": 1018 | |
| }, | |
| { | |
| "epoch": 0.48, | |
| "learning_rate": 0.0002277900612930839, | |
| "loss": 1.1993, | |
| "step": 1019 | |
| }, | |
| { | |
| "epoch": 0.48, | |
| "learning_rate": 0.00022748153815210654, | |
| "loss": 1.3798, | |
| "step": 1020 | |
| }, | |
| { | |
| "epoch": 0.48, | |
| "learning_rate": 0.00022717294834136818, | |
| "loss": 1.5262, | |
| "step": 1021 | |
| }, | |
| { | |
| "epoch": 0.48, | |
| "learning_rate": 0.00022686429260950278, | |
| "loss": 1.206, | |
| "step": 1022 | |
| }, | |
| { | |
| "epoch": 0.48, | |
| "learning_rate": 0.00022655557170530398, | |
| "loss": 1.1834, | |
| "step": 1023 | |
| }, | |
| { | |
| "epoch": 0.48, | |
| "learning_rate": 0.00022624678637772364, | |
| "loss": 1.2479, | |
| "step": 1024 | |
| }, | |
| { | |
| "epoch": 0.48, | |
| "learning_rate": 0.00022593793737586988, | |
| "loss": 1.0998, | |
| "step": 1025 | |
| }, | |
| { | |
| "epoch": 0.48, | |
| "learning_rate": 0.00022562902544900543, | |
| "loss": 1.1673, | |
| "step": 1026 | |
| }, | |
| { | |
| "epoch": 0.48, | |
| "learning_rate": 0.00022532005134654547, | |
| "loss": 1.1091, | |
| "step": 1027 | |
| }, | |
| { | |
| "epoch": 0.49, | |
| "learning_rate": 0.00022501101581805615, | |
| "loss": 1.4556, | |
| "step": 1028 | |
| }, | |
| { | |
| "epoch": 0.49, | |
| "learning_rate": 0.00022470191961325244, | |
| "loss": 1.4475, | |
| "step": 1029 | |
| }, | |
| { | |
| "epoch": 0.49, | |
| "learning_rate": 0.0002243927634819969, | |
| "loss": 1.4791, | |
| "step": 1030 | |
| }, | |
| { | |
| "epoch": 0.49, | |
| "learning_rate": 0.0002240835481742971, | |
| "loss": 1.3507, | |
| "step": 1031 | |
| }, | |
| { | |
| "epoch": 0.49, | |
| "learning_rate": 0.00022377427444030432, | |
| "loss": 1.2467, | |
| "step": 1032 | |
| }, | |
| { | |
| "epoch": 0.49, | |
| "learning_rate": 0.00022346494303031157, | |
| "loss": 1.0791, | |
| "step": 1033 | |
| }, | |
| { | |
| "epoch": 0.49, | |
| "learning_rate": 0.00022315555469475177, | |
| "loss": 1.345, | |
| "step": 1034 | |
| }, | |
| { | |
| "epoch": 0.49, | |
| "learning_rate": 0.000222846110184196, | |
| "loss": 1.3105, | |
| "step": 1035 | |
| }, | |
| { | |
| "epoch": 0.49, | |
| "learning_rate": 0.00022253661024935158, | |
| "loss": 1.3515, | |
| "step": 1036 | |
| }, | |
| { | |
| "epoch": 0.49, | |
| "learning_rate": 0.00022222705564106017, | |
| "loss": 1.2521, | |
| "step": 1037 | |
| }, | |
| { | |
| "epoch": 0.49, | |
| "learning_rate": 0.00022191744711029618, | |
| "loss": 1.0644, | |
| "step": 1038 | |
| }, | |
| { | |
| "epoch": 0.49, | |
| "learning_rate": 0.00022160778540816499, | |
| "loss": 1.4489, | |
| "step": 1039 | |
| }, | |
| { | |
| "epoch": 0.49, | |
| "learning_rate": 0.00022129807128590067, | |
| "loss": 1.3382, | |
| "step": 1040 | |
| }, | |
| { | |
| "epoch": 0.49, | |
| "learning_rate": 0.00022098830549486463, | |
| "loss": 1.2667, | |
| "step": 1041 | |
| }, | |
| { | |
| "epoch": 0.49, | |
| "learning_rate": 0.00022067848878654352, | |
| "loss": 1.2325, | |
| "step": 1042 | |
| }, | |
| { | |
| "epoch": 0.49, | |
| "learning_rate": 0.0002203686219125478, | |
| "loss": 1.255, | |
| "step": 1043 | |
| }, | |
| { | |
| "epoch": 0.49, | |
| "learning_rate": 0.00022005870562460925, | |
| "loss": 1.4091, | |
| "step": 1044 | |
| }, | |
| { | |
| "epoch": 0.49, | |
| "learning_rate": 0.00021974874067457983, | |
| "loss": 1.2677, | |
| "step": 1045 | |
| }, | |
| { | |
| "epoch": 0.49, | |
| "learning_rate": 0.0002194387278144293, | |
| "loss": 1.2642, | |
| "step": 1046 | |
| }, | |
| { | |
| "epoch": 0.49, | |
| "learning_rate": 0.0002191286677962438, | |
| "loss": 1.3356, | |
| "step": 1047 | |
| }, | |
| { | |
| "epoch": 0.49, | |
| "learning_rate": 0.00021881856137222395, | |
| "loss": 1.1519, | |
| "step": 1048 | |
| }, | |
| { | |
| "epoch": 0.5, | |
| "learning_rate": 0.00021850840929468272, | |
| "loss": 1.3606, | |
| "step": 1049 | |
| }, | |
| { | |
| "epoch": 0.5, | |
| "learning_rate": 0.00021819821231604413, | |
| "loss": 1.1905, | |
| "step": 1050 | |
| }, | |
| { | |
| "epoch": 0.5, | |
| "learning_rate": 0.00021788797118884072, | |
| "loss": 1.219, | |
| "step": 1051 | |
| }, | |
| { | |
| "epoch": 0.5, | |
| "learning_rate": 0.00021757768666571262, | |
| "loss": 1.3163, | |
| "step": 1052 | |
| }, | |
| { | |
| "epoch": 0.5, | |
| "learning_rate": 0.0002172673594994048, | |
| "loss": 1.2363, | |
| "step": 1053 | |
| }, | |
| { | |
| "epoch": 0.5, | |
| "learning_rate": 0.000216956990442766, | |
| "loss": 1.3136, | |
| "step": 1054 | |
| }, | |
| { | |
| "epoch": 0.5, | |
| "learning_rate": 0.00021664658024874646, | |
| "loss": 1.0558, | |
| "step": 1055 | |
| }, | |
| { | |
| "epoch": 0.5, | |
| "learning_rate": 0.00021633612967039622, | |
| "loss": 1.2747, | |
| "step": 1056 | |
| }, | |
| { | |
| "epoch": 0.5, | |
| "learning_rate": 0.0002160256394608632, | |
| "loss": 1.5154, | |
| "step": 1057 | |
| }, | |
| { | |
| "epoch": 0.5, | |
| "learning_rate": 0.00021571511037339164, | |
| "loss": 1.4203, | |
| "step": 1058 | |
| }, | |
| { | |
| "epoch": 0.5, | |
| "learning_rate": 0.00021540454316132003, | |
| "loss": 1.1171, | |
| "step": 1059 | |
| }, | |
| { | |
| "epoch": 0.5, | |
| "learning_rate": 0.0002150939385780793, | |
| "loss": 1.47, | |
| "step": 1060 | |
| }, | |
| { | |
| "epoch": 0.5, | |
| "learning_rate": 0.0002147832973771911, | |
| "loss": 1.5328, | |
| "step": 1061 | |
| }, | |
| { | |
| "epoch": 0.5, | |
| "learning_rate": 0.00021447262031226578, | |
| "loss": 1.2424, | |
| "step": 1062 | |
| }, | |
| { | |
| "epoch": 0.5, | |
| "learning_rate": 0.00021416190813700094, | |
| "loss": 1.4171, | |
| "step": 1063 | |
| }, | |
| { | |
| "epoch": 0.5, | |
| "learning_rate": 0.00021385116160517917, | |
| "loss": 1.1544, | |
| "step": 1064 | |
| }, | |
| { | |
| "epoch": 0.5, | |
| "learning_rate": 0.00021354038147066646, | |
| "loss": 1.2143, | |
| "step": 1065 | |
| }, | |
| { | |
| "epoch": 0.5, | |
| "learning_rate": 0.00021322956848741033, | |
| "loss": 1.3098, | |
| "step": 1066 | |
| }, | |
| { | |
| "epoch": 0.5, | |
| "learning_rate": 0.000212918723409438, | |
| "loss": 1.1003, | |
| "step": 1067 | |
| }, | |
| { | |
| "epoch": 0.5, | |
| "learning_rate": 0.00021260784699085448, | |
| "loss": 1.434, | |
| "step": 1068 | |
| }, | |
| { | |
| "epoch": 0.5, | |
| "learning_rate": 0.0002122969399858409, | |
| "loss": 1.1671, | |
| "step": 1069 | |
| }, | |
| { | |
| "epoch": 0.51, | |
| "learning_rate": 0.0002119860031486525, | |
| "loss": 1.2743, | |
| "step": 1070 | |
| }, | |
| { | |
| "epoch": 0.51, | |
| "learning_rate": 0.00021167503723361705, | |
| "loss": 1.4703, | |
| "step": 1071 | |
| }, | |
| { | |
| "epoch": 0.51, | |
| "learning_rate": 0.00021136404299513273, | |
| "loss": 1.2531, | |
| "step": 1072 | |
| }, | |
| { | |
| "epoch": 0.51, | |
| "learning_rate": 0.00021105302118766642, | |
| "loss": 1.289, | |
| "step": 1073 | |
| }, | |
| { | |
| "epoch": 0.51, | |
| "learning_rate": 0.0002107419725657519, | |
| "loss": 1.406, | |
| "step": 1074 | |
| }, | |
| { | |
| "epoch": 0.51, | |
| "learning_rate": 0.00021043089788398814, | |
| "loss": 1.2649, | |
| "step": 1075 | |
| }, | |
| { | |
| "epoch": 0.51, | |
| "learning_rate": 0.0002101197978970371, | |
| "loss": 1.5223, | |
| "step": 1076 | |
| }, | |
| { | |
| "epoch": 0.51, | |
| "learning_rate": 0.00020980867335962234, | |
| "loss": 1.3938, | |
| "step": 1077 | |
| }, | |
| { | |
| "epoch": 0.51, | |
| "learning_rate": 0.00020949752502652684, | |
| "loss": 1.0398, | |
| "step": 1078 | |
| }, | |
| { | |
| "epoch": 0.51, | |
| "learning_rate": 0.00020918635365259133, | |
| "loss": 1.5071, | |
| "step": 1079 | |
| }, | |
| { | |
| "epoch": 0.51, | |
| "learning_rate": 0.0002088751599927125, | |
| "loss": 1.333, | |
| "step": 1080 | |
| }, | |
| { | |
| "epoch": 0.51, | |
| "learning_rate": 0.00020856394480184105, | |
| "loss": 1.4037, | |
| "step": 1081 | |
| }, | |
| { | |
| "epoch": 0.51, | |
| "learning_rate": 0.00020825270883497994, | |
| "loss": 1.301, | |
| "step": 1082 | |
| }, | |
| { | |
| "epoch": 0.51, | |
| "learning_rate": 0.0002079414528471825, | |
| "loss": 1.092, | |
| "step": 1083 | |
| }, | |
| { | |
| "epoch": 0.51, | |
| "learning_rate": 0.0002076301775935507, | |
| "loss": 1.347, | |
| "step": 1084 | |
| }, | |
| { | |
| "epoch": 0.51, | |
| "learning_rate": 0.00020731888382923312, | |
| "loss": 1.2909, | |
| "step": 1085 | |
| }, | |
| { | |
| "epoch": 0.51, | |
| "learning_rate": 0.00020700757230942345, | |
| "loss": 1.3255, | |
| "step": 1086 | |
| }, | |
| { | |
| "epoch": 0.51, | |
| "learning_rate": 0.00020669624378935822, | |
| "loss": 1.0013, | |
| "step": 1087 | |
| }, | |
| { | |
| "epoch": 0.51, | |
| "learning_rate": 0.00020638489902431536, | |
| "loss": 1.1933, | |
| "step": 1088 | |
| }, | |
| { | |
| "epoch": 0.51, | |
| "learning_rate": 0.00020607353876961218, | |
| "loss": 1.5111, | |
| "step": 1089 | |
| }, | |
| { | |
| "epoch": 0.51, | |
| "learning_rate": 0.00020576216378060349, | |
| "loss": 1.3786, | |
| "step": 1090 | |
| }, | |
| { | |
| "epoch": 0.52, | |
| "learning_rate": 0.00020545077481268003, | |
| "loss": 0.8628, | |
| "step": 1091 | |
| }, | |
| { | |
| "epoch": 0.52, | |
| "learning_rate": 0.00020513937262126616, | |
| "loss": 1.3142, | |
| "step": 1092 | |
| }, | |
| { | |
| "epoch": 0.52, | |
| "learning_rate": 0.00020482795796181863, | |
| "loss": 1.247, | |
| "step": 1093 | |
| }, | |
| { | |
| "epoch": 0.52, | |
| "learning_rate": 0.00020451653158982418, | |
| "loss": 1.3025, | |
| "step": 1094 | |
| }, | |
| { | |
| "epoch": 0.52, | |
| "learning_rate": 0.00020420509426079825, | |
| "loss": 1.3176, | |
| "step": 1095 | |
| }, | |
| { | |
| "epoch": 0.52, | |
| "learning_rate": 0.00020389364673028255, | |
| "loss": 1.2238, | |
| "step": 1096 | |
| }, | |
| { | |
| "epoch": 0.52, | |
| "learning_rate": 0.00020389364673028255, | |
| "loss": 1.3724, | |
| "step": 1097 | |
| }, | |
| { | |
| "epoch": 0.52, | |
| "learning_rate": 0.00020358218975384374, | |
| "loss": 1.2739, | |
| "step": 1098 | |
| }, | |
| { | |
| "epoch": 0.52, | |
| "learning_rate": 0.0002032707240870713, | |
| "loss": 1.3877, | |
| "step": 1099 | |
| }, | |
| { | |
| "epoch": 0.52, | |
| "learning_rate": 0.00020295925048557586, | |
| "loss": 1.2701, | |
| "step": 1100 | |
| }, | |
| { | |
| "epoch": 0.52, | |
| "learning_rate": 0.0002026477697049873, | |
| "loss": 1.3203, | |
| "step": 1101 | |
| }, | |
| { | |
| "epoch": 0.52, | |
| "learning_rate": 0.00020233628250095284, | |
| "loss": 1.314, | |
| "step": 1102 | |
| }, | |
| { | |
| "epoch": 0.52, | |
| "learning_rate": 0.00020202478962913536, | |
| "loss": 1.6357, | |
| "step": 1103 | |
| }, | |
| { | |
| "epoch": 0.52, | |
| "learning_rate": 0.00020171329184521137, | |
| "loss": 1.1124, | |
| "step": 1104 | |
| }, | |
| { | |
| "epoch": 0.52, | |
| "learning_rate": 0.00020140178990486952, | |
| "loss": 1.4138, | |
| "step": 1105 | |
| }, | |
| { | |
| "epoch": 0.52, | |
| "learning_rate": 0.0002010902845638083, | |
| "loss": 1.3425, | |
| "step": 1106 | |
| }, | |
| { | |
| "epoch": 0.52, | |
| "learning_rate": 0.00020077877657773466, | |
| "loss": 1.2216, | |
| "step": 1107 | |
| }, | |
| { | |
| "epoch": 0.52, | |
| "learning_rate": 0.00020046726670236176, | |
| "loss": 1.519, | |
| "step": 1108 | |
| }, | |
| { | |
| "epoch": 0.52, | |
| "learning_rate": 0.0002001557556934075, | |
| "loss": 1.3377, | |
| "step": 1109 | |
| }, | |
| { | |
| "epoch": 0.52, | |
| "learning_rate": 0.00019984424430659253, | |
| "loss": 1.6655, | |
| "step": 1110 | |
| }, | |
| { | |
| "epoch": 0.52, | |
| "learning_rate": 0.0001995327332976383, | |
| "loss": 1.4196, | |
| "step": 1111 | |
| }, | |
| { | |
| "epoch": 0.53, | |
| "learning_rate": 0.00019922122342226538, | |
| "loss": 1.2147, | |
| "step": 1112 | |
| }, | |
| { | |
| "epoch": 0.53, | |
| "learning_rate": 0.00019890971543619167, | |
| "loss": 1.1667, | |
| "step": 1113 | |
| }, | |
| { | |
| "epoch": 0.53, | |
| "learning_rate": 0.00019859821009513055, | |
| "loss": 1.1459, | |
| "step": 1114 | |
| }, | |
| { | |
| "epoch": 0.53, | |
| "learning_rate": 0.00019828670815478868, | |
| "loss": 1.3117, | |
| "step": 1115 | |
| }, | |
| { | |
| "epoch": 0.53, | |
| "learning_rate": 0.00019797521037086472, | |
| "loss": 1.1508, | |
| "step": 1116 | |
| }, | |
| { | |
| "epoch": 0.53, | |
| "learning_rate": 0.00019766371749904723, | |
| "loss": 1.0721, | |
| "step": 1117 | |
| }, | |
| { | |
| "epoch": 0.53, | |
| "learning_rate": 0.00019735223029501272, | |
| "loss": 1.1657, | |
| "step": 1118 | |
| }, | |
| { | |
| "epoch": 0.53, | |
| "learning_rate": 0.00019704074951442418, | |
| "loss": 1.0856, | |
| "step": 1119 | |
| }, | |
| { | |
| "epoch": 0.53, | |
| "learning_rate": 0.00019672927591292871, | |
| "loss": 1.1536, | |
| "step": 1120 | |
| }, | |
| { | |
| "epoch": 0.53, | |
| "learning_rate": 0.00019641781024615633, | |
| "loss": 1.3234, | |
| "step": 1121 | |
| }, | |
| { | |
| "epoch": 0.53, | |
| "learning_rate": 0.0001961063532697175, | |
| "loss": 1.446, | |
| "step": 1122 | |
| }, | |
| { | |
| "epoch": 0.53, | |
| "learning_rate": 0.0001957949057392018, | |
| "loss": 1.359, | |
| "step": 1123 | |
| }, | |
| { | |
| "epoch": 0.53, | |
| "learning_rate": 0.00019548346841017578, | |
| "loss": 1.2901, | |
| "step": 1124 | |
| }, | |
| { | |
| "epoch": 0.53, | |
| "learning_rate": 0.00019517204203818147, | |
| "loss": 1.3227, | |
| "step": 1125 | |
| }, | |
| { | |
| "epoch": 0.53, | |
| "learning_rate": 0.00019486062737873389, | |
| "loss": 1.5337, | |
| "step": 1126 | |
| }, | |
| { | |
| "epoch": 0.53, | |
| "learning_rate": 0.00019454922518732004, | |
| "loss": 1.0848, | |
| "step": 1127 | |
| }, | |
| { | |
| "epoch": 0.53, | |
| "learning_rate": 0.00019423783621939653, | |
| "loss": 1.2892, | |
| "step": 1128 | |
| }, | |
| { | |
| "epoch": 0.53, | |
| "learning_rate": 0.00019392646123038792, | |
| "loss": 1.1257, | |
| "step": 1129 | |
| }, | |
| { | |
| "epoch": 0.53, | |
| "learning_rate": 0.0001936151009756847, | |
| "loss": 1.3003, | |
| "step": 1130 | |
| }, | |
| { | |
| "epoch": 0.53, | |
| "learning_rate": 0.0001933037562106418, | |
| "loss": 1.1776, | |
| "step": 1131 | |
| }, | |
| { | |
| "epoch": 0.53, | |
| "learning_rate": 0.00019299242769057665, | |
| "loss": 1.1835, | |
| "step": 1132 | |
| }, | |
| { | |
| "epoch": 0.54, | |
| "learning_rate": 0.00019268111617076693, | |
| "loss": 1.5712, | |
| "step": 1133 | |
| }, | |
| { | |
| "epoch": 0.54, | |
| "learning_rate": 0.00019236982240644935, | |
| "loss": 0.9642, | |
| "step": 1134 | |
| }, | |
| { | |
| "epoch": 0.54, | |
| "learning_rate": 0.0001920585471528175, | |
| "loss": 1.3297, | |
| "step": 1135 | |
| }, | |
| { | |
| "epoch": 0.54, | |
| "learning_rate": 0.00019174729116502016, | |
| "loss": 1.3761, | |
| "step": 1136 | |
| }, | |
| { | |
| "epoch": 0.54, | |
| "learning_rate": 0.000191436055198159, | |
| "loss": 1.4211, | |
| "step": 1137 | |
| }, | |
| { | |
| "epoch": 0.54, | |
| "learning_rate": 0.00019112484000728756, | |
| "loss": 1.3953, | |
| "step": 1138 | |
| }, | |
| { | |
| "epoch": 0.54, | |
| "learning_rate": 0.0001908136463474087, | |
| "loss": 1.3674, | |
| "step": 1139 | |
| }, | |
| { | |
| "epoch": 0.54, | |
| "learning_rate": 0.00019050247497347323, | |
| "loss": 1.2079, | |
| "step": 1140 | |
| }, | |
| { | |
| "epoch": 0.54, | |
| "learning_rate": 0.00019019132664037768, | |
| "loss": 1.5356, | |
| "step": 1141 | |
| }, | |
| { | |
| "epoch": 0.54, | |
| "learning_rate": 0.00018988020210296294, | |
| "loss": 0.9021, | |
| "step": 1142 | |
| }, | |
| { | |
| "epoch": 0.54, | |
| "learning_rate": 0.00018956910211601185, | |
| "loss": 1.4613, | |
| "step": 1143 | |
| }, | |
| { | |
| "epoch": 0.54, | |
| "learning_rate": 0.00018925802743424813, | |
| "loss": 1.3332, | |
| "step": 1144 | |
| }, | |
| { | |
| "epoch": 0.54, | |
| "learning_rate": 0.00018894697881233362, | |
| "loss": 1.3994, | |
| "step": 1145 | |
| }, | |
| { | |
| "epoch": 0.54, | |
| "learning_rate": 0.00018863595700486732, | |
| "loss": 1.0774, | |
| "step": 1146 | |
| }, | |
| { | |
| "epoch": 0.54, | |
| "learning_rate": 0.00018832496276638302, | |
| "loss": 1.3694, | |
| "step": 1147 | |
| }, | |
| { | |
| "epoch": 0.54, | |
| "learning_rate": 0.00018801399685134754, | |
| "loss": 1.4133, | |
| "step": 1148 | |
| }, | |
| { | |
| "epoch": 0.54, | |
| "learning_rate": 0.00018770306001415917, | |
| "loss": 1.3533, | |
| "step": 1149 | |
| }, | |
| { | |
| "epoch": 0.54, | |
| "learning_rate": 0.0001873921530091456, | |
| "loss": 1.1007, | |
| "step": 1150 | |
| }, | |
| { | |
| "epoch": 0.54, | |
| "learning_rate": 0.00018708127659056208, | |
| "loss": 1.1953, | |
| "step": 1151 | |
| }, | |
| { | |
| "epoch": 0.54, | |
| "learning_rate": 0.00018677043151258971, | |
| "loss": 1.2118, | |
| "step": 1152 | |
| }, | |
| { | |
| "epoch": 0.54, | |
| "learning_rate": 0.00018645961852933356, | |
| "loss": 1.4276, | |
| "step": 1153 | |
| }, | |
| { | |
| "epoch": 0.54, | |
| "learning_rate": 0.00018614883839482082, | |
| "loss": 1.4432, | |
| "step": 1154 | |
| }, | |
| { | |
| "epoch": 0.55, | |
| "learning_rate": 0.00018583809186299913, | |
| "loss": 1.1725, | |
| "step": 1155 | |
| }, | |
| { | |
| "epoch": 0.55, | |
| "learning_rate": 0.00018552737968773427, | |
| "loss": 1.3901, | |
| "step": 1156 | |
| }, | |
| { | |
| "epoch": 0.55, | |
| "learning_rate": 0.00018521670262280896, | |
| "loss": 1.471, | |
| "step": 1157 | |
| }, | |
| { | |
| "epoch": 0.55, | |
| "learning_rate": 0.00018490606142192078, | |
| "loss": 1.7013, | |
| "step": 1158 | |
| }, | |
| { | |
| "epoch": 0.55, | |
| "learning_rate": 0.00018459545683868, | |
| "loss": 1.3411, | |
| "step": 1159 | |
| }, | |
| { | |
| "epoch": 0.55, | |
| "learning_rate": 0.00018428488962660838, | |
| "loss": 1.1998, | |
| "step": 1160 | |
| }, | |
| { | |
| "epoch": 0.55, | |
| "learning_rate": 0.0001839743605391368, | |
| "loss": 1.174, | |
| "step": 1161 | |
| }, | |
| { | |
| "epoch": 0.55, | |
| "learning_rate": 0.00018366387032960386, | |
| "loss": 1.2417, | |
| "step": 1162 | |
| }, | |
| { | |
| "epoch": 0.55, | |
| "learning_rate": 0.00018335341975125356, | |
| "loss": 1.1483, | |
| "step": 1163 | |
| }, | |
| { | |
| "epoch": 0.55, | |
| "learning_rate": 0.00018304300955723402, | |
| "loss": 1.2208, | |
| "step": 1164 | |
| }, | |
| { | |
| "epoch": 0.55, | |
| "learning_rate": 0.00018273264050059523, | |
| "loss": 1.4443, | |
| "step": 1165 | |
| }, | |
| { | |
| "epoch": 0.55, | |
| "learning_rate": 0.0001824223133342875, | |
| "loss": 1.1945, | |
| "step": 1166 | |
| }, | |
| { | |
| "epoch": 0.55, | |
| "learning_rate": 0.00018211202881115933, | |
| "loss": 1.2849, | |
| "step": 1167 | |
| }, | |
| { | |
| "epoch": 0.55, | |
| "learning_rate": 0.00018180178768395595, | |
| "loss": 1.0626, | |
| "step": 1168 | |
| }, | |
| { | |
| "epoch": 0.55, | |
| "learning_rate": 0.0001814915907053173, | |
| "loss": 1.1795, | |
| "step": 1169 | |
| }, | |
| { | |
| "epoch": 0.55, | |
| "learning_rate": 0.00018118143862777615, | |
| "loss": 1.1829, | |
| "step": 1170 | |
| }, | |
| { | |
| "epoch": 0.55, | |
| "learning_rate": 0.00018087133220375624, | |
| "loss": 1.1555, | |
| "step": 1171 | |
| }, | |
| { | |
| "epoch": 0.55, | |
| "learning_rate": 0.00018056127218557072, | |
| "loss": 1.3981, | |
| "step": 1172 | |
| }, | |
| { | |
| "epoch": 0.55, | |
| "learning_rate": 0.00018025125932542027, | |
| "loss": 1.2743, | |
| "step": 1173 | |
| }, | |
| { | |
| "epoch": 0.55, | |
| "learning_rate": 0.0001799412943753908, | |
| "loss": 1.1822, | |
| "step": 1174 | |
| }, | |
| { | |
| "epoch": 0.55, | |
| "learning_rate": 0.00017963137808745222, | |
| "loss": 1.1138, | |
| "step": 1175 | |
| }, | |
| { | |
| "epoch": 0.56, | |
| "learning_rate": 0.00017932151121345644, | |
| "loss": 1.1169, | |
| "step": 1176 | |
| }, | |
| { | |
| "epoch": 0.56, | |
| "learning_rate": 0.00017901169450513547, | |
| "loss": 1.2054, | |
| "step": 1177 | |
| }, | |
| { | |
| "epoch": 0.56, | |
| "learning_rate": 0.00017870192871409938, | |
| "loss": 1.1044, | |
| "step": 1178 | |
| }, | |
| { | |
| "epoch": 0.56, | |
| "learning_rate": 0.00017839221459183506, | |
| "loss": 1.4161, | |
| "step": 1179 | |
| }, | |
| { | |
| "epoch": 0.56, | |
| "learning_rate": 0.00017808255288970378, | |
| "loss": 1.322, | |
| "step": 1180 | |
| }, | |
| { | |
| "epoch": 0.56, | |
| "learning_rate": 0.00017777294435893993, | |
| "loss": 1.1476, | |
| "step": 1181 | |
| }, | |
| { | |
| "epoch": 0.56, | |
| "learning_rate": 0.0001774633897506485, | |
| "loss": 1.3483, | |
| "step": 1182 | |
| }, | |
| { | |
| "epoch": 0.56, | |
| "learning_rate": 0.000177153889815804, | |
| "loss": 1.4175, | |
| "step": 1183 | |
| }, | |
| { | |
| "epoch": 0.56, | |
| "learning_rate": 0.0001768444453052482, | |
| "loss": 1.277, | |
| "step": 1184 | |
| }, | |
| { | |
| "epoch": 0.56, | |
| "learning_rate": 0.00017653505696968848, | |
| "loss": 1.3087, | |
| "step": 1185 | |
| }, | |
| { | |
| "epoch": 0.56, | |
| "learning_rate": 0.0001762257255596957, | |
| "loss": 1.225, | |
| "step": 1186 | |
| }, | |
| { | |
| "epoch": 0.56, | |
| "learning_rate": 0.00017591645182570294, | |
| "loss": 1.2801, | |
| "step": 1187 | |
| }, | |
| { | |
| "epoch": 0.56, | |
| "learning_rate": 0.0001756072365180032, | |
| "loss": 1.601, | |
| "step": 1188 | |
| }, | |
| { | |
| "epoch": 0.56, | |
| "learning_rate": 0.00017529808038674758, | |
| "loss": 1.3707, | |
| "step": 1189 | |
| }, | |
| { | |
| "epoch": 0.56, | |
| "learning_rate": 0.00017498898418194392, | |
| "loss": 1.0633, | |
| "step": 1190 | |
| }, | |
| { | |
| "epoch": 0.56, | |
| "learning_rate": 0.00017467994865345455, | |
| "loss": 1.3597, | |
| "step": 1191 | |
| }, | |
| { | |
| "epoch": 0.56, | |
| "learning_rate": 0.0001743709745509946, | |
| "loss": 1.1676, | |
| "step": 1192 | |
| }, | |
| { | |
| "epoch": 0.56, | |
| "learning_rate": 0.00017406206262413014, | |
| "loss": 1.1692, | |
| "step": 1193 | |
| }, | |
| { | |
| "epoch": 0.56, | |
| "learning_rate": 0.00017375321362227638, | |
| "loss": 1.294, | |
| "step": 1194 | |
| }, | |
| { | |
| "epoch": 0.56, | |
| "learning_rate": 0.00017344442829469604, | |
| "loss": 1.3103, | |
| "step": 1195 | |
| }, | |
| { | |
| "epoch": 0.56, | |
| "learning_rate": 0.0001731357073904973, | |
| "loss": 1.1713, | |
| "step": 1196 | |
| }, | |
| { | |
| "epoch": 0.57, | |
| "learning_rate": 0.00017282705165863187, | |
| "loss": 1.2887, | |
| "step": 1197 | |
| }, | |
| { | |
| "epoch": 0.57, | |
| "learning_rate": 0.0001725184618478935, | |
| "loss": 1.2084, | |
| "step": 1198 | |
| }, | |
| { | |
| "epoch": 0.57, | |
| "learning_rate": 0.0001722099387069162, | |
| "loss": 1.4899, | |
| "step": 1199 | |
| }, | |
| { | |
| "epoch": 0.57, | |
| "learning_rate": 0.0001719014829841718, | |
| "loss": 1.3522, | |
| "step": 1200 | |
| }, | |
| { | |
| "epoch": 0.57, | |
| "learning_rate": 0.00017159309542796902, | |
| "loss": 1.2895, | |
| "step": 1201 | |
| }, | |
| { | |
| "epoch": 0.57, | |
| "learning_rate": 0.00017128477678645087, | |
| "loss": 1.1823, | |
| "step": 1202 | |
| }, | |
| { | |
| "epoch": 0.57, | |
| "learning_rate": 0.00017097652780759349, | |
| "loss": 1.4265, | |
| "step": 1203 | |
| }, | |
| { | |
| "epoch": 0.57, | |
| "learning_rate": 0.00017066834923920365, | |
| "loss": 1.148, | |
| "step": 1204 | |
| }, | |
| { | |
| "epoch": 0.57, | |
| "learning_rate": 0.00017036024182891766, | |
| "loss": 1.2976, | |
| "step": 1205 | |
| }, | |
| { | |
| "epoch": 0.57, | |
| "learning_rate": 0.00017005220632419893, | |
| "loss": 1.0658, | |
| "step": 1206 | |
| }, | |
| { | |
| "epoch": 0.57, | |
| "learning_rate": 0.00016974424347233669, | |
| "loss": 1.3412, | |
| "step": 1207 | |
| }, | |
| { | |
| "epoch": 0.57, | |
| "learning_rate": 0.00016943635402044357, | |
| "loss": 1.2008, | |
| "step": 1208 | |
| }, | |
| { | |
| "epoch": 0.57, | |
| "learning_rate": 0.00016912853871545446, | |
| "loss": 1.2512, | |
| "step": 1209 | |
| }, | |
| { | |
| "epoch": 0.57, | |
| "learning_rate": 0.0001688207983041243, | |
| "loss": 1.4035, | |
| "step": 1210 | |
| }, | |
| { | |
| "epoch": 0.57, | |
| "learning_rate": 0.0001685131335330262, | |
| "loss": 1.2508, | |
| "step": 1211 | |
| }, | |
| { | |
| "epoch": 0.57, | |
| "learning_rate": 0.0001682055451485498, | |
| "loss": 1.2797, | |
| "step": 1212 | |
| }, | |
| { | |
| "epoch": 0.57, | |
| "learning_rate": 0.0001678980338968995, | |
| "loss": 1.1797, | |
| "step": 1213 | |
| }, | |
| { | |
| "epoch": 0.57, | |
| "learning_rate": 0.0001675906005240927, | |
| "loss": 1.1574, | |
| "step": 1214 | |
| }, | |
| { | |
| "epoch": 0.57, | |
| "learning_rate": 0.00016728324577595756, | |
| "loss": 1.3345, | |
| "step": 1215 | |
| }, | |
| { | |
| "epoch": 0.57, | |
| "learning_rate": 0.00016697597039813168, | |
| "loss": 1.6107, | |
| "step": 1216 | |
| }, | |
| { | |
| "epoch": 0.57, | |
| "learning_rate": 0.00016666877513606007, | |
| "loss": 1.3272, | |
| "step": 1217 | |
| }, | |
| { | |
| "epoch": 0.58, | |
| "learning_rate": 0.00016636166073499356, | |
| "loss": 1.2153, | |
| "step": 1218 | |
| }, | |
| { | |
| "epoch": 0.58, | |
| "learning_rate": 0.00016605462793998645, | |
| "loss": 1.1083, | |
| "step": 1219 | |
| }, | |
| { | |
| "epoch": 0.58, | |
| "learning_rate": 0.0001657476774958954, | |
| "loss": 1.3484, | |
| "step": 1220 | |
| }, | |
| { | |
| "epoch": 0.58, | |
| "learning_rate": 0.00016544081014737704, | |
| "loss": 1.3568, | |
| "step": 1221 | |
| }, | |
| { | |
| "epoch": 0.58, | |
| "learning_rate": 0.00016513402663888668, | |
| "loss": 1.2882, | |
| "step": 1222 | |
| }, | |
| { | |
| "epoch": 0.58, | |
| "learning_rate": 0.00016482732771467594, | |
| "loss": 1.293, | |
| "step": 1223 | |
| }, | |
| { | |
| "epoch": 0.58, | |
| "learning_rate": 0.00016452071411879145, | |
| "loss": 1.4508, | |
| "step": 1224 | |
| }, | |
| { | |
| "epoch": 0.58, | |
| "learning_rate": 0.0001642141865950727, | |
| "loss": 1.2862, | |
| "step": 1225 | |
| }, | |
| { | |
| "epoch": 0.58, | |
| "learning_rate": 0.00016390774588715057, | |
| "loss": 1.2247, | |
| "step": 1226 | |
| }, | |
| { | |
| "epoch": 0.58, | |
| "learning_rate": 0.00016360139273844504, | |
| "loss": 1.2461, | |
| "step": 1227 | |
| }, | |
| { | |
| "epoch": 0.58, | |
| "learning_rate": 0.00016329512789216394, | |
| "loss": 1.4069, | |
| "step": 1228 | |
| }, | |
| { | |
| "epoch": 0.58, | |
| "learning_rate": 0.00016298895209130077, | |
| "loss": 1.3323, | |
| "step": 1229 | |
| }, | |
| { | |
| "epoch": 0.58, | |
| "learning_rate": 0.00016268286607863284, | |
| "loss": 1.4585, | |
| "step": 1230 | |
| }, | |
| { | |
| "epoch": 0.58, | |
| "learning_rate": 0.0001623768705967199, | |
| "loss": 1.3159, | |
| "step": 1231 | |
| }, | |
| { | |
| "epoch": 0.58, | |
| "learning_rate": 0.00016207096638790207, | |
| "loss": 1.1522, | |
| "step": 1232 | |
| }, | |
| { | |
| "epoch": 0.58, | |
| "learning_rate": 0.00016176515419429784, | |
| "loss": 1.3848, | |
| "step": 1233 | |
| }, | |
| { | |
| "epoch": 0.58, | |
| "learning_rate": 0.0001614594347578026, | |
| "loss": 1.5621, | |
| "step": 1234 | |
| }, | |
| { | |
| "epoch": 0.58, | |
| "learning_rate": 0.00016115380882008662, | |
| "loss": 1.3241, | |
| "step": 1235 | |
| }, | |
| { | |
| "epoch": 0.58, | |
| "learning_rate": 0.00016084827712259348, | |
| "loss": 1.6407, | |
| "step": 1236 | |
| }, | |
| { | |
| "epoch": 0.58, | |
| "learning_rate": 0.00016054284040653815, | |
| "loss": 1.2676, | |
| "step": 1237 | |
| }, | |
| { | |
| "epoch": 0.58, | |
| "learning_rate": 0.00016023749941290497, | |
| "loss": 1.3582, | |
| "step": 1238 | |
| }, | |
| { | |
| "epoch": 0.59, | |
| "learning_rate": 0.00015993225488244618, | |
| "loss": 1.3799, | |
| "step": 1239 | |
| }, | |
| { | |
| "epoch": 0.59, | |
| "learning_rate": 0.00015962710755568012, | |
| "loss": 1.4649, | |
| "step": 1240 | |
| }, | |
| { | |
| "epoch": 0.59, | |
| "learning_rate": 0.00015932205817288908, | |
| "loss": 1.0993, | |
| "step": 1241 | |
| }, | |
| { | |
| "epoch": 0.59, | |
| "learning_rate": 0.00015901710747411794, | |
| "loss": 1.4141, | |
| "step": 1242 | |
| }, | |
| { | |
| "epoch": 0.59, | |
| "learning_rate": 0.00015871225619917202, | |
| "loss": 1.4413, | |
| "step": 1243 | |
| }, | |
| { | |
| "epoch": 0.59, | |
| "learning_rate": 0.00015840750508761567, | |
| "loss": 1.1094, | |
| "step": 1244 | |
| }, | |
| { | |
| "epoch": 0.59, | |
| "learning_rate": 0.00015810285487876985, | |
| "loss": 0.9049, | |
| "step": 1245 | |
| }, | |
| { | |
| "epoch": 0.59, | |
| "learning_rate": 0.00015779830631171117, | |
| "loss": 1.2571, | |
| "step": 1246 | |
| }, | |
| { | |
| "epoch": 0.59, | |
| "learning_rate": 0.00015749386012526935, | |
| "loss": 1.59, | |
| "step": 1247 | |
| }, | |
| { | |
| "epoch": 0.59, | |
| "learning_rate": 0.00015718951705802597, | |
| "loss": 1.2153, | |
| "step": 1248 | |
| }, | |
| { | |
| "epoch": 0.59, | |
| "learning_rate": 0.00015688527784831215, | |
| "loss": 1.2648, | |
| "step": 1249 | |
| }, | |
| { | |
| "epoch": 0.59, | |
| "learning_rate": 0.00015658114323420733, | |
| "loss": 1.1044, | |
| "step": 1250 | |
| }, | |
| { | |
| "epoch": 0.59, | |
| "learning_rate": 0.0001562771139535371, | |
| "loss": 1.3576, | |
| "step": 1251 | |
| }, | |
| { | |
| "epoch": 0.59, | |
| "learning_rate": 0.00015597319074387152, | |
| "loss": 1.2877, | |
| "step": 1252 | |
| }, | |
| { | |
| "epoch": 0.59, | |
| "learning_rate": 0.00015566937434252325, | |
| "loss": 1.3409, | |
| "step": 1253 | |
| }, | |
| { | |
| "epoch": 0.59, | |
| "learning_rate": 0.0001553656654865459, | |
| "loss": 1.1914, | |
| "step": 1254 | |
| }, | |
| { | |
| "epoch": 0.59, | |
| "learning_rate": 0.00015506206491273228, | |
| "loss": 1.4526, | |
| "step": 1255 | |
| }, | |
| { | |
| "epoch": 0.59, | |
| "learning_rate": 0.0001547585733576123, | |
| "loss": 1.3002, | |
| "step": 1256 | |
| }, | |
| { | |
| "epoch": 0.59, | |
| "learning_rate": 0.00015445519155745147, | |
| "loss": 1.0568, | |
| "step": 1257 | |
| }, | |
| { | |
| "epoch": 0.59, | |
| "learning_rate": 0.00015415192024824913, | |
| "loss": 1.2317, | |
| "step": 1258 | |
| }, | |
| { | |
| "epoch": 0.59, | |
| "learning_rate": 0.00015384876016573656, | |
| "loss": 1.2007, | |
| "step": 1259 | |
| }, | |
| { | |
| "epoch": 0.59, | |
| "learning_rate": 0.00015354571204537502, | |
| "loss": 1.2982, | |
| "step": 1260 | |
| }, | |
| { | |
| "epoch": 0.6, | |
| "learning_rate": 0.00015324277662235436, | |
| "loss": 1.3678, | |
| "step": 1261 | |
| }, | |
| { | |
| "epoch": 0.6, | |
| "learning_rate": 0.00015293995463159092, | |
| "loss": 1.4411, | |
| "step": 1262 | |
| }, | |
| { | |
| "epoch": 0.6, | |
| "learning_rate": 0.00015263724680772596, | |
| "loss": 1.3717, | |
| "step": 1263 | |
| }, | |
| { | |
| "epoch": 0.6, | |
| "learning_rate": 0.00015233465388512355, | |
| "loss": 1.1964, | |
| "step": 1264 | |
| }, | |
| { | |
| "epoch": 0.6, | |
| "learning_rate": 0.00015203217659786927, | |
| "loss": 1.2942, | |
| "step": 1265 | |
| }, | |
| { | |
| "epoch": 0.6, | |
| "learning_rate": 0.000151729815679768, | |
| "loss": 1.1961, | |
| "step": 1266 | |
| }, | |
| { | |
| "epoch": 0.6, | |
| "learning_rate": 0.00015142757186434249, | |
| "loss": 1.5537, | |
| "step": 1267 | |
| }, | |
| { | |
| "epoch": 0.6, | |
| "learning_rate": 0.00015112544588483112, | |
| "loss": 1.1772, | |
| "step": 1268 | |
| }, | |
| { | |
| "epoch": 0.6, | |
| "learning_rate": 0.00015082343847418664, | |
| "loss": 1.4643, | |
| "step": 1269 | |
| }, | |
| { | |
| "epoch": 0.6, | |
| "learning_rate": 0.0001505215503650741, | |
| "loss": 0.9425, | |
| "step": 1270 | |
| }, | |
| { | |
| "epoch": 0.6, | |
| "learning_rate": 0.00015021978228986903, | |
| "loss": 1.2285, | |
| "step": 1271 | |
| }, | |
| { | |
| "epoch": 0.6, | |
| "learning_rate": 0.00014991813498065585, | |
| "loss": 1.5057, | |
| "step": 1272 | |
| }, | |
| { | |
| "epoch": 0.6, | |
| "learning_rate": 0.00014961660916922606, | |
| "loss": 1.1536, | |
| "step": 1273 | |
| }, | |
| { | |
| "epoch": 0.6, | |
| "learning_rate": 0.00014931520558707627, | |
| "loss": 1.2474, | |
| "step": 1274 | |
| }, | |
| { | |
| "epoch": 0.6, | |
| "learning_rate": 0.00014901392496540665, | |
| "loss": 1.145, | |
| "step": 1275 | |
| }, | |
| { | |
| "epoch": 0.6, | |
| "learning_rate": 0.00014871276803511896, | |
| "loss": 1.1519, | |
| "step": 1276 | |
| }, | |
| { | |
| "epoch": 0.6, | |
| "learning_rate": 0.00014841173552681514, | |
| "loss": 1.1699, | |
| "step": 1277 | |
| }, | |
| { | |
| "epoch": 0.6, | |
| "learning_rate": 0.00014811082817079503, | |
| "loss": 1.2612, | |
| "step": 1278 | |
| }, | |
| { | |
| "epoch": 0.6, | |
| "learning_rate": 0.00014781004669705495, | |
| "loss": 1.4188, | |
| "step": 1279 | |
| }, | |
| { | |
| "epoch": 0.6, | |
| "learning_rate": 0.00014750939183528575, | |
| "loss": 1.4395, | |
| "step": 1280 | |
| }, | |
| { | |
| "epoch": 0.6, | |
| "learning_rate": 0.0001472088643148713, | |
| "loss": 1.4652, | |
| "step": 1281 | |
| }, | |
| { | |
| "epoch": 0.61, | |
| "learning_rate": 0.00014690846486488645, | |
| "loss": 1.0477, | |
| "step": 1282 | |
| }, | |
| { | |
| "epoch": 0.61, | |
| "learning_rate": 0.00014660819421409528, | |
| "loss": 1.1328, | |
| "step": 1283 | |
| }, | |
| { | |
| "epoch": 0.61, | |
| "learning_rate": 0.00014630805309094938, | |
| "loss": 1.0984, | |
| "step": 1284 | |
| }, | |
| { | |
| "epoch": 0.61, | |
| "learning_rate": 0.0001460080422235864, | |
| "loss": 1.2177, | |
| "step": 1285 | |
| }, | |
| { | |
| "epoch": 0.61, | |
| "learning_rate": 0.00014570816233982762, | |
| "loss": 1.4792, | |
| "step": 1286 | |
| }, | |
| { | |
| "epoch": 0.61, | |
| "learning_rate": 0.0001454084141671768, | |
| "loss": 1.317, | |
| "step": 1287 | |
| }, | |
| { | |
| "epoch": 0.61, | |
| "learning_rate": 0.000145108798432818, | |
| "loss": 1.4423, | |
| "step": 1288 | |
| }, | |
| { | |
| "epoch": 0.61, | |
| "learning_rate": 0.00014480931586361426, | |
| "loss": 1.0848, | |
| "step": 1289 | |
| }, | |
| { | |
| "epoch": 0.61, | |
| "learning_rate": 0.00014450996718610515, | |
| "loss": 1.3578, | |
| "step": 1290 | |
| }, | |
| { | |
| "epoch": 0.61, | |
| "learning_rate": 0.00014421075312650577, | |
| "loss": 1.4258, | |
| "step": 1291 | |
| }, | |
| { | |
| "epoch": 0.61, | |
| "learning_rate": 0.0001439116744107046, | |
| "loss": 1.2835, | |
| "step": 1292 | |
| }, | |
| { | |
| "epoch": 0.61, | |
| "learning_rate": 0.00014361273176426165, | |
| "loss": 1.1916, | |
| "step": 1293 | |
| }, | |
| { | |
| "epoch": 0.61, | |
| "learning_rate": 0.00014331392591240675, | |
| "loss": 1.3486, | |
| "step": 1294 | |
| }, | |
| { | |
| "epoch": 0.61, | |
| "learning_rate": 0.0001430152575800381, | |
| "loss": 1.3556, | |
| "step": 1295 | |
| }, | |
| { | |
| "epoch": 0.61, | |
| "learning_rate": 0.00014271672749172027, | |
| "loss": 1.2307, | |
| "step": 1296 | |
| }, | |
| { | |
| "epoch": 0.61, | |
| "learning_rate": 0.00014241833637168222, | |
| "loss": 1.4525, | |
| "step": 1297 | |
| }, | |
| { | |
| "epoch": 0.61, | |
| "learning_rate": 0.00014212008494381593, | |
| "loss": 1.2207, | |
| "step": 1298 | |
| }, | |
| { | |
| "epoch": 0.61, | |
| "learning_rate": 0.0001418219739316745, | |
| "loss": 1.3679, | |
| "step": 1299 | |
| }, | |
| { | |
| "epoch": 0.61, | |
| "learning_rate": 0.00014152400405847043, | |
| "loss": 0.8475, | |
| "step": 1300 | |
| }, | |
| { | |
| "epoch": 0.61, | |
| "learning_rate": 0.00014122617604707364, | |
| "loss": 1.2213, | |
| "step": 1301 | |
| }, | |
| { | |
| "epoch": 0.61, | |
| "learning_rate": 0.00014092849062001, | |
| "loss": 1.1242, | |
| "step": 1302 | |
| }, | |
| { | |
| "epoch": 0.62, | |
| "learning_rate": 0.0001406309484994595, | |
| "loss": 1.3462, | |
| "step": 1303 | |
| }, | |
| { | |
| "epoch": 0.62, | |
| "learning_rate": 0.00014033355040725458, | |
| "loss": 1.4102, | |
| "step": 1304 | |
| }, | |
| { | |
| "epoch": 0.62, | |
| "learning_rate": 0.00014003629706487792, | |
| "loss": 1.3565, | |
| "step": 1305 | |
| }, | |
| { | |
| "epoch": 0.62, | |
| "learning_rate": 0.00013973918919346135, | |
| "loss": 1.2443, | |
| "step": 1306 | |
| }, | |
| { | |
| "epoch": 0.62, | |
| "learning_rate": 0.00013944222751378368, | |
| "loss": 1.4912, | |
| "step": 1307 | |
| }, | |
| { | |
| "epoch": 0.62, | |
| "learning_rate": 0.00013914541274626915, | |
| "loss": 1.4366, | |
| "step": 1308 | |
| }, | |
| { | |
| "epoch": 0.62, | |
| "learning_rate": 0.0001388487456109854, | |
| "loss": 1.2756, | |
| "step": 1309 | |
| }, | |
| { | |
| "epoch": 0.62, | |
| "learning_rate": 0.00013855222682764216, | |
| "loss": 1.2778, | |
| "step": 1310 | |
| }, | |
| { | |
| "epoch": 0.62, | |
| "learning_rate": 0.00013825585711558906, | |
| "loss": 1.4932, | |
| "step": 1311 | |
| }, | |
| { | |
| "epoch": 0.62, | |
| "learning_rate": 0.0001379596371938141, | |
| "loss": 1.2772, | |
| "step": 1312 | |
| }, | |
| { | |
| "epoch": 0.62, | |
| "learning_rate": 0.00013766356778094205, | |
| "loss": 1.2692, | |
| "step": 1313 | |
| }, | |
| { | |
| "epoch": 0.62, | |
| "learning_rate": 0.00013736764959523242, | |
| "loss": 1.1895, | |
| "step": 1314 | |
| }, | |
| { | |
| "epoch": 0.62, | |
| "learning_rate": 0.0001370718833545779, | |
| "loss": 1.3382, | |
| "step": 1315 | |
| }, | |
| { | |
| "epoch": 0.62, | |
| "learning_rate": 0.00013677626977650256, | |
| "loss": 1.337, | |
| "step": 1316 | |
| }, | |
| { | |
| "epoch": 0.62, | |
| "learning_rate": 0.00013648080957816, | |
| "loss": 1.1632, | |
| "step": 1317 | |
| }, | |
| { | |
| "epoch": 0.62, | |
| "learning_rate": 0.00013618550347633187, | |
| "loss": 1.1321, | |
| "step": 1318 | |
| }, | |
| { | |
| "epoch": 0.62, | |
| "learning_rate": 0.000135890352187426, | |
| "loss": 1.1655, | |
| "step": 1319 | |
| }, | |
| { | |
| "epoch": 0.62, | |
| "learning_rate": 0.00013559535642747447, | |
| "loss": 1.3333, | |
| "step": 1320 | |
| }, | |
| { | |
| "epoch": 0.62, | |
| "learning_rate": 0.00013530051691213217, | |
| "loss": 1.3118, | |
| "step": 1321 | |
| }, | |
| { | |
| "epoch": 0.62, | |
| "learning_rate": 0.00013500583435667495, | |
| "loss": 1.4974, | |
| "step": 1322 | |
| }, | |
| { | |
| "epoch": 0.62, | |
| "learning_rate": 0.00013471130947599792, | |
| "loss": 1.2056, | |
| "step": 1323 | |
| }, | |
| { | |
| "epoch": 0.63, | |
| "learning_rate": 0.00013441694298461357, | |
| "loss": 1.1723, | |
| "step": 1324 | |
| }, | |
| { | |
| "epoch": 0.63, | |
| "learning_rate": 0.00013412273559665008, | |
| "loss": 1.6037, | |
| "step": 1325 | |
| }, | |
| { | |
| "epoch": 0.63, | |
| "learning_rate": 0.00013382868802584992, | |
| "loss": 1.0672, | |
| "step": 1326 | |
| }, | |
| { | |
| "epoch": 0.63, | |
| "learning_rate": 0.00013353480098556752, | |
| "loss": 1.2062, | |
| "step": 1327 | |
| }, | |
| { | |
| "epoch": 0.63, | |
| "learning_rate": 0.00013324107518876814, | |
| "loss": 1.3481, | |
| "step": 1328 | |
| }, | |
| { | |
| "epoch": 0.63, | |
| "learning_rate": 0.00013294751134802565, | |
| "loss": 1.2264, | |
| "step": 1329 | |
| }, | |
| { | |
| "epoch": 0.63, | |
| "learning_rate": 0.00013265411017552125, | |
| "loss": 1.36, | |
| "step": 1330 | |
| }, | |
| { | |
| "epoch": 0.63, | |
| "learning_rate": 0.00013236087238304117, | |
| "loss": 1.0622, | |
| "step": 1331 | |
| }, | |
| { | |
| "epoch": 0.63, | |
| "learning_rate": 0.00013206779868197565, | |
| "loss": 1.3958, | |
| "step": 1332 | |
| }, | |
| { | |
| "epoch": 0.63, | |
| "learning_rate": 0.00013177488978331664, | |
| "loss": 1.2677, | |
| "step": 1333 | |
| }, | |
| { | |
| "epoch": 0.63, | |
| "learning_rate": 0.00013148214639765638, | |
| "loss": 1.283, | |
| "step": 1334 | |
| }, | |
| { | |
| "epoch": 0.63, | |
| "learning_rate": 0.0001311895692351854, | |
| "loss": 1.5159, | |
| "step": 1335 | |
| }, | |
| { | |
| "epoch": 0.63, | |
| "learning_rate": 0.00013089715900569115, | |
| "loss": 1.3298, | |
| "step": 1336 | |
| }, | |
| { | |
| "epoch": 0.63, | |
| "learning_rate": 0.00013060491641855612, | |
| "loss": 0.898, | |
| "step": 1337 | |
| }, | |
| { | |
| "epoch": 0.63, | |
| "learning_rate": 0.00013031284218275596, | |
| "loss": 1.2589, | |
| "step": 1338 | |
| }, | |
| { | |
| "epoch": 0.63, | |
| "learning_rate": 0.00013002093700685796, | |
| "loss": 1.077, | |
| "step": 1339 | |
| }, | |
| { | |
| "epoch": 0.63, | |
| "learning_rate": 0.0001297292015990193, | |
| "loss": 1.2823, | |
| "step": 1340 | |
| }, | |
| { | |
| "epoch": 0.63, | |
| "learning_rate": 0.00012943763666698542, | |
| "loss": 1.4288, | |
| "step": 1341 | |
| }, | |
| { | |
| "epoch": 0.63, | |
| "learning_rate": 0.00012914624291808795, | |
| "loss": 1.2173, | |
| "step": 1342 | |
| }, | |
| { | |
| "epoch": 0.63, | |
| "learning_rate": 0.0001288550210592433, | |
| "loss": 0.9642, | |
| "step": 1343 | |
| }, | |
| { | |
| "epoch": 0.63, | |
| "learning_rate": 0.00012856397179695098, | |
| "loss": 1.3113, | |
| "step": 1344 | |
| }, | |
| { | |
| "epoch": 0.64, | |
| "learning_rate": 0.00012827309583729183, | |
| "loss": 1.3067, | |
| "step": 1345 | |
| }, | |
| { | |
| "epoch": 0.64, | |
| "learning_rate": 0.000127982393885926, | |
| "loss": 1.3371, | |
| "step": 1346 | |
| }, | |
| { | |
| "epoch": 0.64, | |
| "learning_rate": 0.00012769186664809173, | |
| "loss": 1.5906, | |
| "step": 1347 | |
| }, | |
| { | |
| "epoch": 0.64, | |
| "learning_rate": 0.00012740151482860333, | |
| "loss": 1.3334, | |
| "step": 1348 | |
| }, | |
| { | |
| "epoch": 0.64, | |
| "learning_rate": 0.00012711133913184957, | |
| "loss": 1.2184, | |
| "step": 1349 | |
| }, | |
| { | |
| "epoch": 0.64, | |
| "learning_rate": 0.00012682134026179186, | |
| "loss": 1.0252, | |
| "step": 1350 | |
| }, | |
| { | |
| "epoch": 0.64, | |
| "learning_rate": 0.0001265315189219628, | |
| "loss": 1.3463, | |
| "step": 1351 | |
| }, | |
| { | |
| "epoch": 0.64, | |
| "learning_rate": 0.0001262418758154642, | |
| "loss": 1.4166, | |
| "step": 1352 | |
| }, | |
| { | |
| "epoch": 0.64, | |
| "learning_rate": 0.00012595241164496538, | |
| "loss": 1.7178, | |
| "step": 1353 | |
| }, | |
| { | |
| "epoch": 0.64, | |
| "learning_rate": 0.00012566312711270175, | |
| "loss": 1.3719, | |
| "step": 1354 | |
| }, | |
| { | |
| "epoch": 0.64, | |
| "learning_rate": 0.00012537402292047288, | |
| "loss": 1.2142, | |
| "step": 1355 | |
| }, | |
| { | |
| "epoch": 0.64, | |
| "learning_rate": 0.00012508509976964072, | |
| "loss": 1.1583, | |
| "step": 1356 | |
| }, | |
| { | |
| "epoch": 0.64, | |
| "learning_rate": 0.00012479635836112814, | |
| "loss": 1.2868, | |
| "step": 1357 | |
| }, | |
| { | |
| "epoch": 0.64, | |
| "learning_rate": 0.00012450779939541701, | |
| "loss": 1.1618, | |
| "step": 1358 | |
| }, | |
| { | |
| "epoch": 0.64, | |
| "learning_rate": 0.0001242194235725467, | |
| "loss": 1.1537, | |
| "step": 1359 | |
| }, | |
| { | |
| "epoch": 0.64, | |
| "learning_rate": 0.0001239312315921123, | |
| "loss": 1.4446, | |
| "step": 1360 | |
| }, | |
| { | |
| "epoch": 0.64, | |
| "learning_rate": 0.0001236432241532627, | |
| "loss": 1.0411, | |
| "step": 1361 | |
| }, | |
| { | |
| "epoch": 0.64, | |
| "learning_rate": 0.00012335540195469925, | |
| "loss": 1.2367, | |
| "step": 1362 | |
| }, | |
| { | |
| "epoch": 0.64, | |
| "learning_rate": 0.0001230677656946739, | |
| "loss": 1.4935, | |
| "step": 1363 | |
| }, | |
| { | |
| "epoch": 0.64, | |
| "learning_rate": 0.0001227803160709876, | |
| "loss": 1.3403, | |
| "step": 1364 | |
| }, | |
| { | |
| "epoch": 0.64, | |
| "learning_rate": 0.00012249305378098828, | |
| "loss": 1.22, | |
| "step": 1365 | |
| }, | |
| { | |
| "epoch": 0.65, | |
| "learning_rate": 0.00012220597952156955, | |
| "loss": 1.4497, | |
| "step": 1366 | |
| }, | |
| { | |
| "epoch": 0.65, | |
| "learning_rate": 0.00012191909398916901, | |
| "loss": 1.4069, | |
| "step": 1367 | |
| }, | |
| { | |
| "epoch": 0.65, | |
| "learning_rate": 0.00012163239787976603, | |
| "loss": 1.2646, | |
| "step": 1368 | |
| }, | |
| { | |
| "epoch": 0.65, | |
| "learning_rate": 0.00012134589188888088, | |
| "loss": 1.499, | |
| "step": 1369 | |
| }, | |
| { | |
| "epoch": 0.65, | |
| "learning_rate": 0.00012105957671157222, | |
| "loss": 1.3049, | |
| "step": 1370 | |
| }, | |
| { | |
| "epoch": 0.65, | |
| "learning_rate": 0.00012077345304243617, | |
| "loss": 1.5111, | |
| "step": 1371 | |
| }, | |
| { | |
| "epoch": 0.65, | |
| "learning_rate": 0.00012048752157560387, | |
| "loss": 1.2329, | |
| "step": 1372 | |
| }, | |
| { | |
| "epoch": 0.65, | |
| "learning_rate": 0.00012020178300474043, | |
| "loss": 1.3798, | |
| "step": 1373 | |
| }, | |
| { | |
| "epoch": 0.65, | |
| "learning_rate": 0.00011991623802304302, | |
| "loss": 1.2576, | |
| "step": 1374 | |
| }, | |
| { | |
| "epoch": 0.65, | |
| "learning_rate": 0.00011963088732323906, | |
| "loss": 1.1369, | |
| "step": 1375 | |
| }, | |
| { | |
| "epoch": 0.65, | |
| "learning_rate": 0.00011934573159758456, | |
| "loss": 1.3211, | |
| "step": 1376 | |
| }, | |
| { | |
| "epoch": 0.65, | |
| "learning_rate": 0.00011906077153786271, | |
| "loss": 1.4205, | |
| "step": 1377 | |
| }, | |
| { | |
| "epoch": 0.65, | |
| "learning_rate": 0.000118776007835382, | |
| "loss": 1.3222, | |
| "step": 1378 | |
| }, | |
| { | |
| "epoch": 0.65, | |
| "learning_rate": 0.00011849144118097442, | |
| "loss": 1.2918, | |
| "step": 1379 | |
| }, | |
| { | |
| "epoch": 0.65, | |
| "learning_rate": 0.00011820707226499399, | |
| "loss": 1.217, | |
| "step": 1380 | |
| }, | |
| { | |
| "epoch": 0.65, | |
| "learning_rate": 0.00011792290177731504, | |
| "loss": 1.5136, | |
| "step": 1381 | |
| }, | |
| { | |
| "epoch": 0.65, | |
| "learning_rate": 0.00011763893040733068, | |
| "loss": 1.3777, | |
| "step": 1382 | |
| }, | |
| { | |
| "epoch": 0.65, | |
| "learning_rate": 0.00011735515884395061, | |
| "loss": 1.3681, | |
| "step": 1383 | |
| }, | |
| { | |
| "epoch": 0.65, | |
| "learning_rate": 0.00011707158777559999, | |
| "loss": 0.8953, | |
| "step": 1384 | |
| }, | |
| { | |
| "epoch": 0.65, | |
| "learning_rate": 0.00011678821789021768, | |
| "loss": 1.2376, | |
| "step": 1385 | |
| }, | |
| { | |
| "epoch": 0.65, | |
| "learning_rate": 0.00011650504987525441, | |
| "loss": 0.9827, | |
| "step": 1386 | |
| }, | |
| { | |
| "epoch": 0.65, | |
| "learning_rate": 0.00011622208441767093, | |
| "loss": 1.3995, | |
| "step": 1387 | |
| }, | |
| { | |
| "epoch": 0.66, | |
| "learning_rate": 0.00011593932220393702, | |
| "loss": 1.4319, | |
| "step": 1388 | |
| }, | |
| { | |
| "epoch": 0.66, | |
| "learning_rate": 0.00011565676392002909, | |
| "loss": 1.5205, | |
| "step": 1389 | |
| }, | |
| { | |
| "epoch": 0.66, | |
| "learning_rate": 0.00011537441025142888, | |
| "loss": 1.1702, | |
| "step": 1390 | |
| }, | |
| { | |
| "epoch": 0.66, | |
| "learning_rate": 0.00011509226188312178, | |
| "loss": 1.2451, | |
| "step": 1391 | |
| }, | |
| { | |
| "epoch": 0.66, | |
| "learning_rate": 0.00011481031949959508, | |
| "loss": 1.4108, | |
| "step": 1392 | |
| }, | |
| { | |
| "epoch": 0.66, | |
| "learning_rate": 0.00011452858378483652, | |
| "loss": 1.235, | |
| "step": 1393 | |
| }, | |
| { | |
| "epoch": 0.66, | |
| "learning_rate": 0.0001142470554223321, | |
| "loss": 1.3461, | |
| "step": 1394 | |
| }, | |
| { | |
| "epoch": 0.66, | |
| "learning_rate": 0.00011396573509506517, | |
| "loss": 1.1677, | |
| "step": 1395 | |
| }, | |
| { | |
| "epoch": 0.66, | |
| "learning_rate": 0.00011368462348551418, | |
| "loss": 1.2003, | |
| "step": 1396 | |
| }, | |
| { | |
| "epoch": 0.66, | |
| "learning_rate": 0.0001134037212756513, | |
| "loss": 1.3054, | |
| "step": 1397 | |
| }, | |
| { | |
| "epoch": 0.66, | |
| "learning_rate": 0.00011312302914694069, | |
| "loss": 1.7106, | |
| "step": 1398 | |
| }, | |
| { | |
| "epoch": 0.66, | |
| "learning_rate": 0.00011284254778033685, | |
| "loss": 1.1096, | |
| "step": 1399 | |
| }, | |
| { | |
| "epoch": 0.66, | |
| "learning_rate": 0.00011256227785628283, | |
| "loss": 1.4595, | |
| "step": 1400 | |
| }, | |
| { | |
| "epoch": 0.66, | |
| "learning_rate": 0.0001122822200547092, | |
| "loss": 0.9698, | |
| "step": 1401 | |
| }, | |
| { | |
| "epoch": 0.66, | |
| "learning_rate": 0.00011200237505503124, | |
| "loss": 1.2934, | |
| "step": 1402 | |
| }, | |
| { | |
| "epoch": 0.66, | |
| "learning_rate": 0.00011172274353614858, | |
| "loss": 1.1179, | |
| "step": 1403 | |
| }, | |
| { | |
| "epoch": 0.66, | |
| "learning_rate": 0.0001114433261764426, | |
| "loss": 1.5195, | |
| "step": 1404 | |
| }, | |
| { | |
| "epoch": 0.66, | |
| "learning_rate": 0.00011116412365377522, | |
| "loss": 1.2414, | |
| "step": 1405 | |
| }, | |
| { | |
| "epoch": 0.66, | |
| "learning_rate": 0.00011088513664548724, | |
| "loss": 1.1524, | |
| "step": 1406 | |
| }, | |
| { | |
| "epoch": 0.66, | |
| "learning_rate": 0.0001106063658283964, | |
| "loss": 1.2442, | |
| "step": 1407 | |
| }, | |
| { | |
| "epoch": 0.66, | |
| "learning_rate": 0.00011032781187879648, | |
| "loss": 1.0068, | |
| "step": 1408 | |
| }, | |
| { | |
| "epoch": 0.67, | |
| "learning_rate": 0.00011004947547245446, | |
| "loss": 1.2076, | |
| "step": 1409 | |
| }, | |
| { | |
| "epoch": 0.67, | |
| "learning_rate": 0.00010977135728460997, | |
| "loss": 1.0662, | |
| "step": 1410 | |
| }, | |
| { | |
| "epoch": 0.67, | |
| "learning_rate": 0.00010949345798997326, | |
| "loss": 1.3129, | |
| "step": 1411 | |
| }, | |
| { | |
| "epoch": 0.67, | |
| "learning_rate": 0.00010921577826272345, | |
| "loss": 1.1221, | |
| "step": 1412 | |
| }, | |
| { | |
| "epoch": 0.67, | |
| "learning_rate": 0.00010893831877650695, | |
| "loss": 1.2171, | |
| "step": 1413 | |
| }, | |
| { | |
| "epoch": 0.67, | |
| "learning_rate": 0.00010866108020443596, | |
| "loss": 1.4596, | |
| "step": 1414 | |
| }, | |
| { | |
| "epoch": 0.67, | |
| "learning_rate": 0.0001083840632190866, | |
| "loss": 1.3664, | |
| "step": 1415 | |
| }, | |
| { | |
| "epoch": 0.67, | |
| "learning_rate": 0.00010810726849249782, | |
| "loss": 1.0676, | |
| "step": 1416 | |
| }, | |
| { | |
| "epoch": 0.67, | |
| "learning_rate": 0.0001078306966961688, | |
| "loss": 1.1301, | |
| "step": 1417 | |
| }, | |
| { | |
| "epoch": 0.67, | |
| "learning_rate": 0.00010755434850105823, | |
| "loss": 1.2053, | |
| "step": 1418 | |
| }, | |
| { | |
| "epoch": 0.67, | |
| "learning_rate": 0.0001072782245775825, | |
| "loss": 1.066, | |
| "step": 1419 | |
| }, | |
| { | |
| "epoch": 0.67, | |
| "learning_rate": 0.00010700232559561341, | |
| "loss": 1.1748, | |
| "step": 1420 | |
| }, | |
| { | |
| "epoch": 0.67, | |
| "learning_rate": 0.00010672665222447764, | |
| "loss": 1.3408, | |
| "step": 1421 | |
| }, | |
| { | |
| "epoch": 0.67, | |
| "learning_rate": 0.00010645120513295412, | |
| "loss": 1.1208, | |
| "step": 1422 | |
| }, | |
| { | |
| "epoch": 0.67, | |
| "learning_rate": 0.00010617598498927304, | |
| "loss": 1.4331, | |
| "step": 1423 | |
| }, | |
| { | |
| "epoch": 0.67, | |
| "learning_rate": 0.00010590099246111394, | |
| "loss": 1.4032, | |
| "step": 1424 | |
| }, | |
| { | |
| "epoch": 0.67, | |
| "learning_rate": 0.00010562622821560406, | |
| "loss": 1.164, | |
| "step": 1425 | |
| }, | |
| { | |
| "epoch": 0.67, | |
| "learning_rate": 0.0001053516929193172, | |
| "loss": 1.0663, | |
| "step": 1426 | |
| }, | |
| { | |
| "epoch": 0.67, | |
| "learning_rate": 0.00010507738723827141, | |
| "loss": 1.1152, | |
| "step": 1427 | |
| }, | |
| { | |
| "epoch": 0.67, | |
| "learning_rate": 0.00010480331183792756, | |
| "loss": 1.0993, | |
| "step": 1428 | |
| }, | |
| { | |
| "epoch": 0.67, | |
| "learning_rate": 0.00010452946738318835, | |
| "loss": 1.1678, | |
| "step": 1429 | |
| }, | |
| { | |
| "epoch": 0.68, | |
| "learning_rate": 0.0001042558545383958, | |
| "loss": 1.4099, | |
| "step": 1430 | |
| }, | |
| { | |
| "epoch": 0.68, | |
| "learning_rate": 0.00010398247396733021, | |
| "loss": 1.1813, | |
| "step": 1431 | |
| }, | |
| { | |
| "epoch": 0.68, | |
| "learning_rate": 0.00010370932633320837, | |
| "loss": 1.0752, | |
| "step": 1432 | |
| }, | |
| { | |
| "epoch": 0.68, | |
| "learning_rate": 0.00010343641229868188, | |
| "loss": 1.4635, | |
| "step": 1433 | |
| }, | |
| { | |
| "epoch": 0.68, | |
| "learning_rate": 0.00010316373252583593, | |
| "loss": 1.197, | |
| "step": 1434 | |
| }, | |
| { | |
| "epoch": 0.68, | |
| "learning_rate": 0.0001028912876761869, | |
| "loss": 1.1872, | |
| "step": 1435 | |
| }, | |
| { | |
| "epoch": 0.68, | |
| "learning_rate": 0.00010261907841068174, | |
| "loss": 1.1227, | |
| "step": 1436 | |
| }, | |
| { | |
| "epoch": 0.68, | |
| "learning_rate": 0.0001023471053896956, | |
| "loss": 1.2541, | |
| "step": 1437 | |
| }, | |
| { | |
| "epoch": 0.68, | |
| "learning_rate": 0.00010207536927303056, | |
| "loss": 1.0352, | |
| "step": 1438 | |
| }, | |
| { | |
| "epoch": 0.68, | |
| "learning_rate": 0.00010180387071991395, | |
| "loss": 1.1582, | |
| "step": 1439 | |
| }, | |
| { | |
| "epoch": 0.68, | |
| "learning_rate": 0.00010153261038899686, | |
| "loss": 1.1927, | |
| "step": 1440 | |
| }, | |
| { | |
| "epoch": 0.68, | |
| "learning_rate": 0.00010126158893835231, | |
| "loss": 1.1228, | |
| "step": 1441 | |
| }, | |
| { | |
| "epoch": 0.68, | |
| "learning_rate": 0.00010099080702547414, | |
| "loss": 1.2856, | |
| "step": 1442 | |
| }, | |
| { | |
| "epoch": 0.68, | |
| "learning_rate": 0.00010072026530727448, | |
| "loss": 1.4384, | |
| "step": 1443 | |
| }, | |
| { | |
| "epoch": 0.68, | |
| "learning_rate": 0.00010044996444008336, | |
| "loss": 1.1022, | |
| "step": 1444 | |
| }, | |
| { | |
| "epoch": 0.68, | |
| "learning_rate": 0.00010017990507964619, | |
| "loss": 1.3684, | |
| "step": 1445 | |
| }, | |
| { | |
| "epoch": 0.68, | |
| "learning_rate": 9.991008788112253e-05, | |
| "loss": 1.18, | |
| "step": 1446 | |
| }, | |
| { | |
| "epoch": 0.68, | |
| "learning_rate": 9.96405134990845e-05, | |
| "loss": 1.4392, | |
| "step": 1447 | |
| }, | |
| { | |
| "epoch": 0.68, | |
| "learning_rate": 9.937118258751501e-05, | |
| "loss": 1.2864, | |
| "step": 1448 | |
| }, | |
| { | |
| "epoch": 0.68, | |
| "learning_rate": 9.910209579980671e-05, | |
| "loss": 1.5951, | |
| "step": 1449 | |
| }, | |
| { | |
| "epoch": 0.68, | |
| "learning_rate": 9.883325378875949e-05, | |
| "loss": 1.237, | |
| "step": 1450 | |
| }, | |
| { | |
| "epoch": 0.69, | |
| "learning_rate": 9.856465720657966e-05, | |
| "loss": 1.25, | |
| "step": 1451 | |
| }, | |
| { | |
| "epoch": 0.69, | |
| "learning_rate": 9.829630670487826e-05, | |
| "loss": 1.0986, | |
| "step": 1452 | |
| }, | |
| { | |
| "epoch": 0.69, | |
| "learning_rate": 9.802820293466917e-05, | |
| "loss": 1.3593, | |
| "step": 1453 | |
| }, | |
| { | |
| "epoch": 0.69, | |
| "learning_rate": 9.776034654636769e-05, | |
| "loss": 1.6713, | |
| "step": 1454 | |
| }, | |
| { | |
| "epoch": 0.69, | |
| "learning_rate": 9.749273818978908e-05, | |
| "loss": 1.1223, | |
| "step": 1455 | |
| }, | |
| { | |
| "epoch": 0.69, | |
| "learning_rate": 9.722537851414671e-05, | |
| "loss": 1.2637, | |
| "step": 1456 | |
| }, | |
| { | |
| "epoch": 0.69, | |
| "learning_rate": 9.6958268168051e-05, | |
| "loss": 1.3128, | |
| "step": 1457 | |
| }, | |
| { | |
| "epoch": 0.69, | |
| "learning_rate": 9.669140779950707e-05, | |
| "loss": 1.3434, | |
| "step": 1458 | |
| }, | |
| { | |
| "epoch": 0.69, | |
| "learning_rate": 9.642479805591378e-05, | |
| "loss": 1.2366, | |
| "step": 1459 | |
| }, | |
| { | |
| "epoch": 0.69, | |
| "learning_rate": 9.615843958406221e-05, | |
| "loss": 1.3496, | |
| "step": 1460 | |
| }, | |
| { | |
| "epoch": 0.69, | |
| "learning_rate": 9.589233303013335e-05, | |
| "loss": 1.1488, | |
| "step": 1461 | |
| }, | |
| { | |
| "epoch": 0.69, | |
| "learning_rate": 9.56264790396976e-05, | |
| "loss": 1.3925, | |
| "step": 1462 | |
| }, | |
| { | |
| "epoch": 0.69, | |
| "learning_rate": 9.536087825771222e-05, | |
| "loss": 1.4061, | |
| "step": 1463 | |
| }, | |
| { | |
| "epoch": 0.69, | |
| "learning_rate": 9.509553132852044e-05, | |
| "loss": 1.0716, | |
| "step": 1464 | |
| }, | |
| { | |
| "epoch": 0.69, | |
| "learning_rate": 9.483043889584955e-05, | |
| "loss": 1.3894, | |
| "step": 1465 | |
| }, | |
| { | |
| "epoch": 0.69, | |
| "learning_rate": 9.456560160280932e-05, | |
| "loss": 1.1413, | |
| "step": 1466 | |
| }, | |
| { | |
| "epoch": 0.69, | |
| "learning_rate": 9.430102009189086e-05, | |
| "loss": 1.1573, | |
| "step": 1467 | |
| }, | |
| { | |
| "epoch": 0.69, | |
| "learning_rate": 9.403669500496457e-05, | |
| "loss": 1.2565, | |
| "step": 1468 | |
| }, | |
| { | |
| "epoch": 0.69, | |
| "learning_rate": 9.377262698327855e-05, | |
| "loss": 1.3642, | |
| "step": 1469 | |
| }, | |
| { | |
| "epoch": 0.69, | |
| "learning_rate": 9.350881666745774e-05, | |
| "loss": 1.2612, | |
| "step": 1470 | |
| }, | |
| { | |
| "epoch": 0.69, | |
| "learning_rate": 9.324526469750154e-05, | |
| "loss": 1.1751, | |
| "step": 1471 | |
| }, | |
| { | |
| "epoch": 0.7, | |
| "learning_rate": 9.298197171278272e-05, | |
| "loss": 1.3304, | |
| "step": 1472 | |
| }, | |
| { | |
| "epoch": 0.7, | |
| "learning_rate": 9.271893835204573e-05, | |
| "loss": 1.3528, | |
| "step": 1473 | |
| }, | |
| { | |
| "epoch": 0.7, | |
| "learning_rate": 9.245616525340513e-05, | |
| "loss": 1.1737, | |
| "step": 1474 | |
| }, | |
| { | |
| "epoch": 0.7, | |
| "learning_rate": 9.219365305434436e-05, | |
| "loss": 1.5289, | |
| "step": 1475 | |
| }, | |
| { | |
| "epoch": 0.7, | |
| "learning_rate": 9.19314023917134e-05, | |
| "loss": 1.2544, | |
| "step": 1476 | |
| }, | |
| { | |
| "epoch": 0.7, | |
| "learning_rate": 9.166941390172831e-05, | |
| "loss": 1.4722, | |
| "step": 1477 | |
| }, | |
| { | |
| "epoch": 0.7, | |
| "learning_rate": 9.14076882199688e-05, | |
| "loss": 1.4618, | |
| "step": 1478 | |
| }, | |
| { | |
| "epoch": 0.7, | |
| "learning_rate": 9.114622598137706e-05, | |
| "loss": 1.2061, | |
| "step": 1479 | |
| }, | |
| { | |
| "epoch": 0.7, | |
| "learning_rate": 9.088502782025618e-05, | |
| "loss": 1.5329, | |
| "step": 1480 | |
| }, | |
| { | |
| "epoch": 0.7, | |
| "learning_rate": 9.062409437026866e-05, | |
| "loss": 1.1748, | |
| "step": 1481 | |
| }, | |
| { | |
| "epoch": 0.7, | |
| "learning_rate": 9.036342626443468e-05, | |
| "loss": 1.6027, | |
| "step": 1482 | |
| }, | |
| { | |
| "epoch": 0.7, | |
| "learning_rate": 9.010302413513102e-05, | |
| "loss": 1.1159, | |
| "step": 1483 | |
| }, | |
| { | |
| "epoch": 0.7, | |
| "learning_rate": 8.984288861408863e-05, | |
| "loss": 1.3043, | |
| "step": 1484 | |
| }, | |
| { | |
| "epoch": 0.7, | |
| "learning_rate": 8.95830203323923e-05, | |
| "loss": 1.1157, | |
| "step": 1485 | |
| }, | |
| { | |
| "epoch": 0.7, | |
| "learning_rate": 8.932341992047811e-05, | |
| "loss": 1.0479, | |
| "step": 1486 | |
| }, | |
| { | |
| "epoch": 0.7, | |
| "learning_rate": 8.906408800813242e-05, | |
| "loss": 1.1356, | |
| "step": 1487 | |
| }, | |
| { | |
| "epoch": 0.7, | |
| "learning_rate": 8.88050252244902e-05, | |
| "loss": 1.2109, | |
| "step": 1488 | |
| }, | |
| { | |
| "epoch": 0.7, | |
| "learning_rate": 8.854623219803341e-05, | |
| "loss": 1.2692, | |
| "step": 1489 | |
| }, | |
| { | |
| "epoch": 0.7, | |
| "learning_rate": 8.828770955658999e-05, | |
| "loss": 1.4603, | |
| "step": 1490 | |
| }, | |
| { | |
| "epoch": 0.7, | |
| "learning_rate": 8.80294579273313e-05, | |
| "loss": 1.3165, | |
| "step": 1491 | |
| }, | |
| { | |
| "epoch": 0.7, | |
| "learning_rate": 8.777147793677163e-05, | |
| "loss": 1.1158, | |
| "step": 1492 | |
| }, | |
| { | |
| "epoch": 0.7, | |
| "learning_rate": 8.751377021076634e-05, | |
| "loss": 1.3774, | |
| "step": 1493 | |
| }, | |
| { | |
| "epoch": 0.71, | |
| "learning_rate": 8.725633537451006e-05, | |
| "loss": 1.4457, | |
| "step": 1494 | |
| }, | |
| { | |
| "epoch": 0.71, | |
| "learning_rate": 8.69991740525355e-05, | |
| "loss": 1.2375, | |
| "step": 1495 | |
| }, | |
| { | |
| "epoch": 0.71, | |
| "learning_rate": 8.67422868687118e-05, | |
| "loss": 1.4202, | |
| "step": 1496 | |
| }, | |
| { | |
| "epoch": 0.71, | |
| "learning_rate": 8.648567444624296e-05, | |
| "loss": 1.3173, | |
| "step": 1497 | |
| }, | |
| { | |
| "epoch": 0.71, | |
| "learning_rate": 8.622933740766679e-05, | |
| "loss": 1.2608, | |
| "step": 1498 | |
| }, | |
| { | |
| "epoch": 0.71, | |
| "learning_rate": 8.597327637485246e-05, | |
| "loss": 1.4031, | |
| "step": 1499 | |
| }, | |
| { | |
| "epoch": 0.71, | |
| "learning_rate": 8.571749196899987e-05, | |
| "loss": 1.272, | |
| "step": 1500 | |
| }, | |
| { | |
| "epoch": 0.71, | |
| "learning_rate": 8.546198481063801e-05, | |
| "loss": 1.3425, | |
| "step": 1501 | |
| }, | |
| { | |
| "epoch": 0.71, | |
| "learning_rate": 8.520675551962275e-05, | |
| "loss": 1.1886, | |
| "step": 1502 | |
| }, | |
| { | |
| "epoch": 0.71, | |
| "learning_rate": 8.49518047151364e-05, | |
| "loss": 1.0558, | |
| "step": 1503 | |
| }, | |
| { | |
| "epoch": 0.71, | |
| "learning_rate": 8.469713301568533e-05, | |
| "loss": 1.328, | |
| "step": 1504 | |
| }, | |
| { | |
| "epoch": 0.71, | |
| "learning_rate": 8.444274103909897e-05, | |
| "loss": 1.1842, | |
| "step": 1505 | |
| }, | |
| { | |
| "epoch": 0.71, | |
| "learning_rate": 8.418862940252804e-05, | |
| "loss": 1.1098, | |
| "step": 1506 | |
| }, | |
| { | |
| "epoch": 0.71, | |
| "learning_rate": 8.39347987224431e-05, | |
| "loss": 1.2275, | |
| "step": 1507 | |
| }, | |
| { | |
| "epoch": 0.71, | |
| "learning_rate": 8.368124961463342e-05, | |
| "loss": 1.2985, | |
| "step": 1508 | |
| }, | |
| { | |
| "epoch": 0.71, | |
| "learning_rate": 8.342798269420495e-05, | |
| "loss": 1.5345, | |
| "step": 1509 | |
| }, | |
| { | |
| "epoch": 0.71, | |
| "learning_rate": 8.317499857557886e-05, | |
| "loss": 1.3177, | |
| "step": 1510 | |
| }, | |
| { | |
| "epoch": 0.71, | |
| "learning_rate": 8.292229787249073e-05, | |
| "loss": 1.1131, | |
| "step": 1511 | |
| }, | |
| { | |
| "epoch": 0.71, | |
| "learning_rate": 8.266988119798821e-05, | |
| "loss": 1.1952, | |
| "step": 1512 | |
| }, | |
| { | |
| "epoch": 0.71, | |
| "learning_rate": 8.241774916443003e-05, | |
| "loss": 1.5274, | |
| "step": 1513 | |
| }, | |
| { | |
| "epoch": 0.71, | |
| "learning_rate": 8.216590238348438e-05, | |
| "loss": 1.1378, | |
| "step": 1514 | |
| }, | |
| { | |
| "epoch": 0.72, | |
| "learning_rate": 8.191434146612733e-05, | |
| "loss": 1.2205, | |
| "step": 1515 | |
| }, | |
| { | |
| "epoch": 0.72, | |
| "learning_rate": 8.166306702264177e-05, | |
| "loss": 0.9346, | |
| "step": 1516 | |
| }, | |
| { | |
| "epoch": 0.72, | |
| "learning_rate": 8.141207966261509e-05, | |
| "loss": 1.1095, | |
| "step": 1517 | |
| }, | |
| { | |
| "epoch": 0.72, | |
| "learning_rate": 8.116137999493869e-05, | |
| "loss": 1.2342, | |
| "step": 1518 | |
| }, | |
| { | |
| "epoch": 0.72, | |
| "learning_rate": 8.091096862780582e-05, | |
| "loss": 1.2845, | |
| "step": 1519 | |
| }, | |
| { | |
| "epoch": 0.72, | |
| "learning_rate": 8.066084616871037e-05, | |
| "loss": 1.0934, | |
| "step": 1520 | |
| }, | |
| { | |
| "epoch": 0.72, | |
| "learning_rate": 8.041101322444526e-05, | |
| "loss": 1.1626, | |
| "step": 1521 | |
| }, | |
| { | |
| "epoch": 0.72, | |
| "learning_rate": 8.016147040110118e-05, | |
| "loss": 1.033, | |
| "step": 1522 | |
| }, | |
| { | |
| "epoch": 0.72, | |
| "learning_rate": 7.991221830406481e-05, | |
| "loss": 1.3303, | |
| "step": 1523 | |
| }, | |
| { | |
| "epoch": 0.72, | |
| "learning_rate": 7.966325753801792e-05, | |
| "loss": 1.0324, | |
| "step": 1524 | |
| }, | |
| { | |
| "epoch": 0.72, | |
| "learning_rate": 7.941458870693488e-05, | |
| "loss": 1.3455, | |
| "step": 1525 | |
| }, | |
| { | |
| "epoch": 0.72, | |
| "learning_rate": 7.916621241408249e-05, | |
| "loss": 1.2513, | |
| "step": 1526 | |
| }, | |
| { | |
| "epoch": 0.72, | |
| "learning_rate": 7.891812926201742e-05, | |
| "loss": 1.4107, | |
| "step": 1527 | |
| }, | |
| { | |
| "epoch": 0.72, | |
| "learning_rate": 7.867033985258541e-05, | |
| "loss": 1.3377, | |
| "step": 1528 | |
| }, | |
| { | |
| "epoch": 0.72, | |
| "learning_rate": 7.842284478691943e-05, | |
| "loss": 1.0802, | |
| "step": 1529 | |
| }, | |
| { | |
| "epoch": 0.72, | |
| "learning_rate": 7.817564466543843e-05, | |
| "loss": 1.2704, | |
| "step": 1530 | |
| }, | |
| { | |
| "epoch": 0.72, | |
| "learning_rate": 7.792874008784603e-05, | |
| "loss": 1.1902, | |
| "step": 1531 | |
| }, | |
| { | |
| "epoch": 0.72, | |
| "learning_rate": 7.76821316531285e-05, | |
| "loss": 1.0903, | |
| "step": 1532 | |
| }, | |
| { | |
| "epoch": 0.72, | |
| "learning_rate": 7.743581995955385e-05, | |
| "loss": 1.303, | |
| "step": 1533 | |
| }, | |
| { | |
| "epoch": 0.72, | |
| "learning_rate": 7.718980560467033e-05, | |
| "loss": 1.4708, | |
| "step": 1534 | |
| }, | |
| { | |
| "epoch": 0.72, | |
| "learning_rate": 7.694408918530469e-05, | |
| "loss": 1.3564, | |
| "step": 1535 | |
| }, | |
| { | |
| "epoch": 0.73, | |
| "learning_rate": 7.669867129756095e-05, | |
| "loss": 1.1663, | |
| "step": 1536 | |
| }, | |
| { | |
| "epoch": 0.73, | |
| "learning_rate": 7.645355253681882e-05, | |
| "loss": 1.1493, | |
| "step": 1537 | |
| }, | |
| { | |
| "epoch": 0.73, | |
| "learning_rate": 7.620873349773239e-05, | |
| "loss": 1.243, | |
| "step": 1538 | |
| }, | |
| { | |
| "epoch": 0.73, | |
| "learning_rate": 7.596421477422879e-05, | |
| "loss": 1.332, | |
| "step": 1539 | |
| }, | |
| { | |
| "epoch": 0.73, | |
| "learning_rate": 7.571999695950627e-05, | |
| "loss": 1.4258, | |
| "step": 1540 | |
| }, | |
| { | |
| "epoch": 0.73, | |
| "learning_rate": 7.54760806460332e-05, | |
| "loss": 1.2923, | |
| "step": 1541 | |
| }, | |
| { | |
| "epoch": 0.73, | |
| "learning_rate": 7.52324664255468e-05, | |
| "loss": 1.253, | |
| "step": 1542 | |
| }, | |
| { | |
| "epoch": 0.73, | |
| "learning_rate": 7.49891548890509e-05, | |
| "loss": 1.2166, | |
| "step": 1543 | |
| }, | |
| { | |
| "epoch": 0.73, | |
| "learning_rate": 7.474614662681547e-05, | |
| "loss": 1.3181, | |
| "step": 1544 | |
| }, | |
| { | |
| "epoch": 0.73, | |
| "learning_rate": 7.450344222837453e-05, | |
| "loss": 1.272, | |
| "step": 1545 | |
| }, | |
| { | |
| "epoch": 0.73, | |
| "learning_rate": 7.426104228252497e-05, | |
| "loss": 1.1434, | |
| "step": 1546 | |
| }, | |
| { | |
| "epoch": 0.73, | |
| "learning_rate": 7.401894737732506e-05, | |
| "loss": 1.1643, | |
| "step": 1547 | |
| }, | |
| { | |
| "epoch": 0.73, | |
| "learning_rate": 7.377715810009303e-05, | |
| "loss": 1.2468, | |
| "step": 1548 | |
| }, | |
| { | |
| "epoch": 0.73, | |
| "learning_rate": 7.353567503740582e-05, | |
| "loss": 1.2365, | |
| "step": 1549 | |
| }, | |
| { | |
| "epoch": 0.73, | |
| "learning_rate": 7.32944987750974e-05, | |
| "loss": 1.3432, | |
| "step": 1550 | |
| }, | |
| { | |
| "epoch": 0.73, | |
| "learning_rate": 7.305362989825719e-05, | |
| "loss": 1.4599, | |
| "step": 1551 | |
| }, | |
| { | |
| "epoch": 0.73, | |
| "learning_rate": 7.281306899122939e-05, | |
| "loss": 1.102, | |
| "step": 1552 | |
| }, | |
| { | |
| "epoch": 0.73, | |
| "learning_rate": 7.257281663761068e-05, | |
| "loss": 1.1645, | |
| "step": 1553 | |
| }, | |
| { | |
| "epoch": 0.73, | |
| "learning_rate": 7.233287342024941e-05, | |
| "loss": 1.1368, | |
| "step": 1554 | |
| }, | |
| { | |
| "epoch": 0.73, | |
| "learning_rate": 7.209323992124382e-05, | |
| "loss": 1.1735, | |
| "step": 1555 | |
| }, | |
| { | |
| "epoch": 0.73, | |
| "learning_rate": 7.185391672194082e-05, | |
| "loss": 1.1274, | |
| "step": 1556 | |
| }, | |
| { | |
| "epoch": 0.74, | |
| "learning_rate": 7.161490440293477e-05, | |
| "loss": 0.9405, | |
| "step": 1557 | |
| }, | |
| { | |
| "epoch": 0.74, | |
| "learning_rate": 7.137620354406535e-05, | |
| "loss": 0.9062, | |
| "step": 1558 | |
| }, | |
| { | |
| "epoch": 0.74, | |
| "learning_rate": 7.113781472441714e-05, | |
| "loss": 1.5528, | |
| "step": 1559 | |
| }, | |
| { | |
| "epoch": 0.74, | |
| "learning_rate": 7.089973852231744e-05, | |
| "loss": 1.3515, | |
| "step": 1560 | |
| }, | |
| { | |
| "epoch": 0.74, | |
| "learning_rate": 7.066197551533522e-05, | |
| "loss": 1.5267, | |
| "step": 1561 | |
| }, | |
| { | |
| "epoch": 0.74, | |
| "learning_rate": 7.042452628027963e-05, | |
| "loss": 1.195, | |
| "step": 1562 | |
| }, | |
| { | |
| "epoch": 0.74, | |
| "learning_rate": 7.018739139319861e-05, | |
| "loss": 1.2459, | |
| "step": 1563 | |
| }, | |
| { | |
| "epoch": 0.74, | |
| "learning_rate": 6.995057142937747e-05, | |
| "loss": 1.3211, | |
| "step": 1564 | |
| }, | |
| { | |
| "epoch": 0.74, | |
| "learning_rate": 6.971406696333771e-05, | |
| "loss": 1.3303, | |
| "step": 1565 | |
| }, | |
| { | |
| "epoch": 0.74, | |
| "learning_rate": 6.947787856883507e-05, | |
| "loss": 1.235, | |
| "step": 1566 | |
| }, | |
| { | |
| "epoch": 0.74, | |
| "learning_rate": 6.924200681885891e-05, | |
| "loss": 1.0975, | |
| "step": 1567 | |
| }, | |
| { | |
| "epoch": 0.74, | |
| "learning_rate": 6.900645228563014e-05, | |
| "loss": 1.0561, | |
| "step": 1568 | |
| }, | |
| { | |
| "epoch": 0.74, | |
| "learning_rate": 6.87712155406002e-05, | |
| "loss": 1.0033, | |
| "step": 1569 | |
| }, | |
| { | |
| "epoch": 0.74, | |
| "learning_rate": 6.85362971544496e-05, | |
| "loss": 1.0609, | |
| "step": 1570 | |
| }, | |
| { | |
| "epoch": 0.74, | |
| "learning_rate": 6.83016976970864e-05, | |
| "loss": 1.22, | |
| "step": 1571 | |
| }, | |
| { | |
| "epoch": 0.74, | |
| "learning_rate": 6.806741773764527e-05, | |
| "loss": 1.2394, | |
| "step": 1572 | |
| }, | |
| { | |
| "epoch": 0.74, | |
| "learning_rate": 6.783345784448532e-05, | |
| "loss": 1.1023, | |
| "step": 1573 | |
| }, | |
| { | |
| "epoch": 0.74, | |
| "learning_rate": 6.759981858518941e-05, | |
| "loss": 1.2294, | |
| "step": 1574 | |
| }, | |
| { | |
| "epoch": 0.74, | |
| "learning_rate": 6.736650052656271e-05, | |
| "loss": 1.0494, | |
| "step": 1575 | |
| }, | |
| { | |
| "epoch": 0.74, | |
| "learning_rate": 6.713350423463092e-05, | |
| "loss": 1.2432, | |
| "step": 1576 | |
| }, | |
| { | |
| "epoch": 0.74, | |
| "learning_rate": 6.690083027463923e-05, | |
| "loss": 1.0197, | |
| "step": 1577 | |
| }, | |
| { | |
| "epoch": 0.75, | |
| "learning_rate": 6.666847921105084e-05, | |
| "loss": 1.1944, | |
| "step": 1578 | |
| }, | |
| { | |
| "epoch": 0.75, | |
| "learning_rate": 6.643645160754559e-05, | |
| "loss": 0.9204, | |
| "step": 1579 | |
| }, | |
| { | |
| "epoch": 0.75, | |
| "learning_rate": 6.620474802701879e-05, | |
| "loss": 1.4313, | |
| "step": 1580 | |
| }, | |
| { | |
| "epoch": 0.75, | |
| "learning_rate": 6.597336903157938e-05, | |
| "loss": 1.2751, | |
| "step": 1581 | |
| }, | |
| { | |
| "epoch": 0.75, | |
| "learning_rate": 6.574231518254896e-05, | |
| "loss": 1.2733, | |
| "step": 1582 | |
| }, | |
| { | |
| "epoch": 0.75, | |
| "learning_rate": 6.551158704046061e-05, | |
| "loss": 1.3241, | |
| "step": 1583 | |
| }, | |
| { | |
| "epoch": 0.75, | |
| "learning_rate": 6.528118516505675e-05, | |
| "loss": 1.2363, | |
| "step": 1584 | |
| }, | |
| { | |
| "epoch": 0.75, | |
| "learning_rate": 6.505111011528872e-05, | |
| "loss": 1.5814, | |
| "step": 1585 | |
| }, | |
| { | |
| "epoch": 0.75, | |
| "learning_rate": 6.482136244931483e-05, | |
| "loss": 1.023, | |
| "step": 1586 | |
| }, | |
| { | |
| "epoch": 0.75, | |
| "learning_rate": 6.459194272449904e-05, | |
| "loss": 1.241, | |
| "step": 1587 | |
| }, | |
| { | |
| "epoch": 0.75, | |
| "learning_rate": 6.436285149740995e-05, | |
| "loss": 1.208, | |
| "step": 1588 | |
| }, | |
| { | |
| "epoch": 0.75, | |
| "learning_rate": 6.413408932381896e-05, | |
| "loss": 1.3445, | |
| "step": 1589 | |
| }, | |
| { | |
| "epoch": 0.75, | |
| "learning_rate": 6.390565675869955e-05, | |
| "loss": 1.2188, | |
| "step": 1590 | |
| }, | |
| { | |
| "epoch": 0.75, | |
| "learning_rate": 6.367755435622536e-05, | |
| "loss": 1.0361, | |
| "step": 1591 | |
| }, | |
| { | |
| "epoch": 0.75, | |
| "learning_rate": 6.344978266976886e-05, | |
| "loss": 1.2409, | |
| "step": 1592 | |
| }, | |
| { | |
| "epoch": 0.75, | |
| "learning_rate": 6.322234225190065e-05, | |
| "loss": 1.1411, | |
| "step": 1593 | |
| }, | |
| { | |
| "epoch": 0.75, | |
| "learning_rate": 6.299523365438737e-05, | |
| "loss": 1.1396, | |
| "step": 1594 | |
| }, | |
| { | |
| "epoch": 0.75, | |
| "learning_rate": 6.276845742819078e-05, | |
| "loss": 1.1535, | |
| "step": 1595 | |
| }, | |
| { | |
| "epoch": 0.75, | |
| "learning_rate": 6.254201412346626e-05, | |
| "loss": 1.3295, | |
| "step": 1596 | |
| }, | |
| { | |
| "epoch": 0.75, | |
| "learning_rate": 6.231590428956149e-05, | |
| "loss": 1.3898, | |
| "step": 1597 | |
| }, | |
| { | |
| "epoch": 0.75, | |
| "learning_rate": 6.209012847501541e-05, | |
| "loss": 1.2843, | |
| "step": 1598 | |
| }, | |
| { | |
| "epoch": 0.76, | |
| "learning_rate": 6.186468722755621e-05, | |
| "loss": 1.4253, | |
| "step": 1599 | |
| }, | |
| { | |
| "epoch": 0.76, | |
| "learning_rate": 6.163958109410084e-05, | |
| "loss": 1.1143, | |
| "step": 1600 | |
| }, | |
| { | |
| "epoch": 0.76, | |
| "learning_rate": 6.141481062075302e-05, | |
| "loss": 1.1324, | |
| "step": 1601 | |
| }, | |
| { | |
| "epoch": 0.76, | |
| "learning_rate": 6.119037635280225e-05, | |
| "loss": 1.176, | |
| "step": 1602 | |
| }, | |
| { | |
| "epoch": 0.76, | |
| "learning_rate": 6.0966278834722346e-05, | |
| "loss": 1.0477, | |
| "step": 1603 | |
| }, | |
| { | |
| "epoch": 0.76, | |
| "learning_rate": 6.0742518610170266e-05, | |
| "loss": 1.205, | |
| "step": 1604 | |
| }, | |
| { | |
| "epoch": 0.76, | |
| "learning_rate": 6.051909622198453e-05, | |
| "loss": 1.3148, | |
| "step": 1605 | |
| }, | |
| { | |
| "epoch": 0.76, | |
| "learning_rate": 6.0296012212184416e-05, | |
| "loss": 1.2212, | |
| "step": 1606 | |
| }, | |
| { | |
| "epoch": 0.76, | |
| "learning_rate": 6.0073267121967794e-05, | |
| "loss": 1.1882, | |
| "step": 1607 | |
| }, | |
| { | |
| "epoch": 0.76, | |
| "learning_rate": 5.985086149171082e-05, | |
| "loss": 1.2734, | |
| "step": 1608 | |
| }, | |
| { | |
| "epoch": 0.76, | |
| "learning_rate": 5.962879586096581e-05, | |
| "loss": 1.1069, | |
| "step": 1609 | |
| }, | |
| { | |
| "epoch": 0.76, | |
| "learning_rate": 5.940707076846035e-05, | |
| "loss": 1.1915, | |
| "step": 1610 | |
| }, | |
| { | |
| "epoch": 0.76, | |
| "learning_rate": 5.91856867520959e-05, | |
| "loss": 1.0885, | |
| "step": 1611 | |
| }, | |
| { | |
| "epoch": 0.76, | |
| "learning_rate": 5.89646443489464e-05, | |
| "loss": 1.2153, | |
| "step": 1612 | |
| }, | |
| { | |
| "epoch": 0.76, | |
| "learning_rate": 5.874394409525723e-05, | |
| "loss": 1.1803, | |
| "step": 1613 | |
| }, | |
| { | |
| "epoch": 0.76, | |
| "learning_rate": 5.852358652644343e-05, | |
| "loss": 1.3352, | |
| "step": 1614 | |
| }, | |
| { | |
| "epoch": 0.76, | |
| "learning_rate": 5.830357217708886e-05, | |
| "loss": 1.1369, | |
| "step": 1615 | |
| }, | |
| { | |
| "epoch": 0.76, | |
| "learning_rate": 5.8083901580944856e-05, | |
| "loss": 1.3195, | |
| "step": 1616 | |
| }, | |
| { | |
| "epoch": 0.76, | |
| "learning_rate": 5.786457527092859e-05, | |
| "loss": 1.2551, | |
| "step": 1617 | |
| }, | |
| { | |
| "epoch": 0.76, | |
| "learning_rate": 5.7645593779122084e-05, | |
| "loss": 1.3172, | |
| "step": 1618 | |
| }, | |
| { | |
| "epoch": 0.76, | |
| "learning_rate": 5.742695763677088e-05, | |
| "loss": 1.3458, | |
| "step": 1619 | |
| }, | |
| { | |
| "epoch": 0.76, | |
| "learning_rate": 5.720866737428261e-05, | |
| "loss": 1.1934, | |
| "step": 1620 | |
| }, | |
| { | |
| "epoch": 0.77, | |
| "learning_rate": 5.699072352122605e-05, | |
| "loss": 0.9409, | |
| "step": 1621 | |
| }, | |
| { | |
| "epoch": 0.77, | |
| "learning_rate": 5.677312660632921e-05, | |
| "loss": 1.162, | |
| "step": 1622 | |
| }, | |
| { | |
| "epoch": 0.77, | |
| "learning_rate": 5.6555877157478674e-05, | |
| "loss": 1.3535, | |
| "step": 1623 | |
| }, | |
| { | |
| "epoch": 0.77, | |
| "learning_rate": 5.633897570171813e-05, | |
| "loss": 1.2635, | |
| "step": 1624 | |
| }, | |
| { | |
| "epoch": 0.77, | |
| "learning_rate": 5.6122422765246905e-05, | |
| "loss": 1.5661, | |
| "step": 1625 | |
| }, | |
| { | |
| "epoch": 0.77, | |
| "learning_rate": 5.590621887341889e-05, | |
| "loss": 1.1086, | |
| "step": 1626 | |
| }, | |
| { | |
| "epoch": 0.77, | |
| "learning_rate": 5.569036455074117e-05, | |
| "loss": 1.0895, | |
| "step": 1627 | |
| }, | |
| { | |
| "epoch": 0.77, | |
| "learning_rate": 5.547486032087277e-05, | |
| "loss": 1.3535, | |
| "step": 1628 | |
| }, | |
| { | |
| "epoch": 0.77, | |
| "learning_rate": 5.525970670662344e-05, | |
| "loss": 1.2312, | |
| "step": 1629 | |
| }, | |
| { | |
| "epoch": 0.77, | |
| "learning_rate": 5.504490422995226e-05, | |
| "loss": 1.1607, | |
| "step": 1630 | |
| }, | |
| { | |
| "epoch": 0.77, | |
| "learning_rate": 5.483045341196662e-05, | |
| "loss": 1.2972, | |
| "step": 1631 | |
| }, | |
| { | |
| "epoch": 0.77, | |
| "learning_rate": 5.461635477292071e-05, | |
| "loss": 1.3073, | |
| "step": 1632 | |
| }, | |
| { | |
| "epoch": 0.77, | |
| "learning_rate": 5.440260883221413e-05, | |
| "loss": 1.3124, | |
| "step": 1633 | |
| }, | |
| { | |
| "epoch": 0.77, | |
| "learning_rate": 5.4189216108391205e-05, | |
| "loss": 1.2518, | |
| "step": 1634 | |
| }, | |
| { | |
| "epoch": 0.77, | |
| "learning_rate": 5.397617711913914e-05, | |
| "loss": 0.8451, | |
| "step": 1635 | |
| }, | |
| { | |
| "epoch": 0.77, | |
| "learning_rate": 5.3763492381286995e-05, | |
| "loss": 1.5449, | |
| "step": 1636 | |
| }, | |
| { | |
| "epoch": 0.77, | |
| "learning_rate": 5.355116241080449e-05, | |
| "loss": 1.1825, | |
| "step": 1637 | |
| }, | |
| { | |
| "epoch": 0.77, | |
| "learning_rate": 5.333918772280055e-05, | |
| "loss": 1.2807, | |
| "step": 1638 | |
| }, | |
| { | |
| "epoch": 0.77, | |
| "learning_rate": 5.312756883152252e-05, | |
| "loss": 1.4253, | |
| "step": 1639 | |
| }, | |
| { | |
| "epoch": 0.77, | |
| "learning_rate": 5.291630625035411e-05, | |
| "loss": 1.1767, | |
| "step": 1640 | |
| }, | |
| { | |
| "epoch": 0.77, | |
| "learning_rate": 5.270540049181505e-05, | |
| "loss": 1.1725, | |
| "step": 1641 | |
| }, | |
| { | |
| "epoch": 0.78, | |
| "learning_rate": 5.249485206755917e-05, | |
| "loss": 1.0447, | |
| "step": 1642 | |
| }, | |
| { | |
| "epoch": 0.78, | |
| "learning_rate": 5.228466148837352e-05, | |
| "loss": 0.9761, | |
| "step": 1643 | |
| }, | |
| { | |
| "epoch": 0.78, | |
| "learning_rate": 5.207482926417702e-05, | |
| "loss": 1.2413, | |
| "step": 1644 | |
| }, | |
| { | |
| "epoch": 0.78, | |
| "learning_rate": 5.186535590401915e-05, | |
| "loss": 1.1953, | |
| "step": 1645 | |
| }, | |
| { | |
| "epoch": 0.78, | |
| "learning_rate": 5.165624191607883e-05, | |
| "loss": 1.2767, | |
| "step": 1646 | |
| }, | |
| { | |
| "epoch": 0.78, | |
| "learning_rate": 5.1447487807663306e-05, | |
| "loss": 0.9696, | |
| "step": 1647 | |
| }, | |
| { | |
| "epoch": 0.78, | |
| "learning_rate": 5.123909408520644e-05, | |
| "loss": 1.145, | |
| "step": 1648 | |
| }, | |
| { | |
| "epoch": 0.78, | |
| "learning_rate": 5.103106125426811e-05, | |
| "loss": 1.4862, | |
| "step": 1649 | |
| }, | |
| { | |
| "epoch": 0.78, | |
| "learning_rate": 5.082338981953254e-05, | |
| "loss": 0.9743, | |
| "step": 1650 | |
| }, | |
| { | |
| "epoch": 0.78, | |
| "learning_rate": 5.0616080284807175e-05, | |
| "loss": 1.2537, | |
| "step": 1651 | |
| }, | |
| { | |
| "epoch": 0.78, | |
| "learning_rate": 5.0409133153021606e-05, | |
| "loss": 1.4724, | |
| "step": 1652 | |
| }, | |
| { | |
| "epoch": 0.78, | |
| "learning_rate": 5.0202548926226066e-05, | |
| "loss": 1.4174, | |
| "step": 1653 | |
| }, | |
| { | |
| "epoch": 0.78, | |
| "learning_rate": 4.99963281055907e-05, | |
| "loss": 1.5319, | |
| "step": 1654 | |
| }, | |
| { | |
| "epoch": 0.78, | |
| "learning_rate": 4.9790471191403674e-05, | |
| "loss": 1.5159, | |
| "step": 1655 | |
| }, | |
| { | |
| "epoch": 0.78, | |
| "learning_rate": 4.958497868307048e-05, | |
| "loss": 0.9962, | |
| "step": 1656 | |
| }, | |
| { | |
| "epoch": 0.78, | |
| "learning_rate": 4.937985107911267e-05, | |
| "loss": 1.3318, | |
| "step": 1657 | |
| }, | |
| { | |
| "epoch": 0.78, | |
| "learning_rate": 4.917508887716642e-05, | |
| "loss": 1.3187, | |
| "step": 1658 | |
| }, | |
| { | |
| "epoch": 0.78, | |
| "learning_rate": 4.8970692573981455e-05, | |
| "loss": 1.0479, | |
| "step": 1659 | |
| }, | |
| { | |
| "epoch": 0.78, | |
| "learning_rate": 4.876666266541987e-05, | |
| "loss": 1.1313, | |
| "step": 1660 | |
| }, | |
| { | |
| "epoch": 0.78, | |
| "learning_rate": 4.856299964645481e-05, | |
| "loss": 1.1031, | |
| "step": 1661 | |
| }, | |
| { | |
| "epoch": 0.78, | |
| "learning_rate": 4.835970401116958e-05, | |
| "loss": 0.955, | |
| "step": 1662 | |
| }, | |
| { | |
| "epoch": 0.79, | |
| "learning_rate": 4.815677625275592e-05, | |
| "loss": 1.2066, | |
| "step": 1663 | |
| }, | |
| { | |
| "epoch": 0.79, | |
| "learning_rate": 4.79542168635132e-05, | |
| "loss": 1.1384, | |
| "step": 1664 | |
| }, | |
| { | |
| "epoch": 0.79, | |
| "learning_rate": 4.775202633484732e-05, | |
| "loss": 1.3048, | |
| "step": 1665 | |
| }, | |
| { | |
| "epoch": 0.79, | |
| "learning_rate": 4.7550205157269134e-05, | |
| "loss": 1.223, | |
| "step": 1666 | |
| }, | |
| { | |
| "epoch": 0.79, | |
| "learning_rate": 4.73487538203935e-05, | |
| "loss": 1.2406, | |
| "step": 1667 | |
| }, | |
| { | |
| "epoch": 0.79, | |
| "learning_rate": 4.7147672812938034e-05, | |
| "loss": 1.0293, | |
| "step": 1668 | |
| }, | |
| { | |
| "epoch": 0.79, | |
| "learning_rate": 4.6946962622722e-05, | |
| "loss": 1.1029, | |
| "step": 1669 | |
| }, | |
| { | |
| "epoch": 0.79, | |
| "learning_rate": 4.674662373666503e-05, | |
| "loss": 1.3801, | |
| "step": 1670 | |
| }, | |
| { | |
| "epoch": 0.79, | |
| "learning_rate": 4.654665664078588e-05, | |
| "loss": 1.3824, | |
| "step": 1671 | |
| }, | |
| { | |
| "epoch": 0.79, | |
| "learning_rate": 4.634706182020161e-05, | |
| "loss": 1.3021, | |
| "step": 1672 | |
| }, | |
| { | |
| "epoch": 0.79, | |
| "learning_rate": 4.6147839759125936e-05, | |
| "loss": 1.2043, | |
| "step": 1673 | |
| }, | |
| { | |
| "epoch": 0.79, | |
| "learning_rate": 4.5948990940868175e-05, | |
| "loss": 1.1829, | |
| "step": 1674 | |
| }, | |
| { | |
| "epoch": 0.79, | |
| "learning_rate": 4.5750515847832434e-05, | |
| "loss": 1.1432, | |
| "step": 1675 | |
| }, | |
| { | |
| "epoch": 0.79, | |
| "learning_rate": 4.555241496151603e-05, | |
| "loss": 1.2856, | |
| "step": 1676 | |
| }, | |
| { | |
| "epoch": 0.79, | |
| "learning_rate": 4.535468876250841e-05, | |
| "loss": 1.1847, | |
| "step": 1677 | |
| }, | |
| { | |
| "epoch": 0.79, | |
| "learning_rate": 4.5157337730490135e-05, | |
| "loss": 0.8359, | |
| "step": 1678 | |
| }, | |
| { | |
| "epoch": 0.79, | |
| "learning_rate": 4.4960362344231466e-05, | |
| "loss": 1.1477, | |
| "step": 1679 | |
| }, | |
| { | |
| "epoch": 0.79, | |
| "learning_rate": 4.4763763081591694e-05, | |
| "loss": 1.1412, | |
| "step": 1680 | |
| }, | |
| { | |
| "epoch": 0.79, | |
| "learning_rate": 4.4567540419517115e-05, | |
| "loss": 1.3056, | |
| "step": 1681 | |
| }, | |
| { | |
| "epoch": 0.79, | |
| "learning_rate": 4.437169483404091e-05, | |
| "loss": 1.3224, | |
| "step": 1682 | |
| }, | |
| { | |
| "epoch": 0.79, | |
| "learning_rate": 4.41762268002812e-05, | |
| "loss": 1.2936, | |
| "step": 1683 | |
| }, | |
| { | |
| "epoch": 0.8, | |
| "learning_rate": 4.398113679244018e-05, | |
| "loss": 0.9675, | |
| "step": 1684 | |
| }, | |
| { | |
| "epoch": 0.8, | |
| "learning_rate": 4.378642528380306e-05, | |
| "loss": 1.3418, | |
| "step": 1685 | |
| }, | |
| { | |
| "epoch": 0.8, | |
| "learning_rate": 4.359209274673675e-05, | |
| "loss": 1.1064, | |
| "step": 1686 | |
| }, | |
| { | |
| "epoch": 0.8, | |
| "learning_rate": 4.339813965268873e-05, | |
| "loss": 1.1713, | |
| "step": 1687 | |
| }, | |
| { | |
| "epoch": 0.8, | |
| "learning_rate": 4.3204566472186204e-05, | |
| "loss": 1.2932, | |
| "step": 1688 | |
| }, | |
| { | |
| "epoch": 0.8, | |
| "learning_rate": 4.301137367483432e-05, | |
| "loss": 1.3301, | |
| "step": 1689 | |
| }, | |
| { | |
| "epoch": 0.8, | |
| "learning_rate": 4.2818561729315795e-05, | |
| "loss": 0.9956, | |
| "step": 1690 | |
| }, | |
| { | |
| "epoch": 0.8, | |
| "learning_rate": 4.26261311033892e-05, | |
| "loss": 1.5543, | |
| "step": 1691 | |
| }, | |
| { | |
| "epoch": 0.8, | |
| "learning_rate": 4.243408226388808e-05, | |
| "loss": 1.2308, | |
| "step": 1692 | |
| }, | |
| { | |
| "epoch": 0.8, | |
| "learning_rate": 4.2242415676719736e-05, | |
| "loss": 1.2872, | |
| "step": 1693 | |
| }, | |
| { | |
| "epoch": 0.8, | |
| "learning_rate": 4.205113180686415e-05, | |
| "loss": 1.3242, | |
| "step": 1694 | |
| }, | |
| { | |
| "epoch": 0.8, | |
| "learning_rate": 4.1860231118373004e-05, | |
| "loss": 1.2402, | |
| "step": 1695 | |
| }, | |
| { | |
| "epoch": 0.8, | |
| "learning_rate": 4.166971407436806e-05, | |
| "loss": 1.3378, | |
| "step": 1696 | |
| }, | |
| { | |
| "epoch": 0.8, | |
| "learning_rate": 4.147958113704058e-05, | |
| "loss": 1.103, | |
| "step": 1697 | |
| }, | |
| { | |
| "epoch": 0.8, | |
| "learning_rate": 4.1289832767650015e-05, | |
| "loss": 1.1262, | |
| "step": 1698 | |
| }, | |
| { | |
| "epoch": 0.8, | |
| "learning_rate": 4.110046942652279e-05, | |
| "loss": 1.3142, | |
| "step": 1699 | |
| }, | |
| { | |
| "epoch": 0.8, | |
| "learning_rate": 4.091149157305123e-05, | |
| "loss": 1.1188, | |
| "step": 1700 | |
| }, | |
| { | |
| "epoch": 0.8, | |
| "learning_rate": 4.072289966569252e-05, | |
| "loss": 1.1387, | |
| "step": 1701 | |
| }, | |
| { | |
| "epoch": 0.8, | |
| "learning_rate": 4.0534694161967465e-05, | |
| "loss": 1.2965, | |
| "step": 1702 | |
| }, | |
| { | |
| "epoch": 0.8, | |
| "learning_rate": 4.034687551845975e-05, | |
| "loss": 1.4459, | |
| "step": 1703 | |
| }, | |
| { | |
| "epoch": 0.8, | |
| "learning_rate": 4.01594441908141e-05, | |
| "loss": 1.072, | |
| "step": 1704 | |
| }, | |
| { | |
| "epoch": 0.81, | |
| "learning_rate": 3.9972400633735865e-05, | |
| "loss": 0.9118, | |
| "step": 1705 | |
| }, | |
| { | |
| "epoch": 0.81, | |
| "learning_rate": 3.9785745300989753e-05, | |
| "loss": 1.1985, | |
| "step": 1706 | |
| }, | |
| { | |
| "epoch": 0.81, | |
| "learning_rate": 3.95994786453985e-05, | |
| "loss": 1.3719, | |
| "step": 1707 | |
| }, | |
| { | |
| "epoch": 0.81, | |
| "learning_rate": 3.9413601118841935e-05, | |
| "loss": 1.3824, | |
| "step": 1708 | |
| }, | |
| { | |
| "epoch": 0.81, | |
| "learning_rate": 3.92281131722559e-05, | |
| "loss": 1.2938, | |
| "step": 1709 | |
| }, | |
| { | |
| "epoch": 0.81, | |
| "learning_rate": 3.904301525563112e-05, | |
| "loss": 0.9944, | |
| "step": 1710 | |
| }, | |
| { | |
| "epoch": 0.81, | |
| "learning_rate": 3.885830781801212e-05, | |
| "loss": 1.3555, | |
| "step": 1711 | |
| }, | |
| { | |
| "epoch": 0.81, | |
| "learning_rate": 3.8673991307496004e-05, | |
| "loss": 1.1441, | |
| "step": 1712 | |
| }, | |
| { | |
| "epoch": 0.81, | |
| "learning_rate": 3.849006617123179e-05, | |
| "loss": 1.3109, | |
| "step": 1713 | |
| }, | |
| { | |
| "epoch": 0.81, | |
| "learning_rate": 3.830653285541877e-05, | |
| "loss": 1.0028, | |
| "step": 1714 | |
| }, | |
| { | |
| "epoch": 0.81, | |
| "learning_rate": 3.812339180530564e-05, | |
| "loss": 1.1562, | |
| "step": 1715 | |
| }, | |
| { | |
| "epoch": 0.81, | |
| "learning_rate": 3.794064346518973e-05, | |
| "loss": 1.0405, | |
| "step": 1716 | |
| }, | |
| { | |
| "epoch": 0.81, | |
| "learning_rate": 3.7758288278415455e-05, | |
| "loss": 1.2983, | |
| "step": 1717 | |
| }, | |
| { | |
| "epoch": 0.81, | |
| "learning_rate": 3.7576326687373545e-05, | |
| "loss": 1.2136, | |
| "step": 1718 | |
| }, | |
| { | |
| "epoch": 0.81, | |
| "learning_rate": 3.739475913349981e-05, | |
| "loss": 1.2544, | |
| "step": 1719 | |
| }, | |
| { | |
| "epoch": 0.81, | |
| "learning_rate": 3.72135860572741e-05, | |
| "loss": 1.1806, | |
| "step": 1720 | |
| }, | |
| { | |
| "epoch": 0.81, | |
| "learning_rate": 3.703280789821948e-05, | |
| "loss": 1.5005, | |
| "step": 1721 | |
| }, | |
| { | |
| "epoch": 0.81, | |
| "learning_rate": 3.685242509490061e-05, | |
| "loss": 1.3666, | |
| "step": 1722 | |
| }, | |
| { | |
| "epoch": 0.81, | |
| "learning_rate": 3.6672438084923376e-05, | |
| "loss": 1.2283, | |
| "step": 1723 | |
| }, | |
| { | |
| "epoch": 0.81, | |
| "learning_rate": 3.649284730493323e-05, | |
| "loss": 1.3949, | |
| "step": 1724 | |
| }, | |
| { | |
| "epoch": 0.81, | |
| "learning_rate": 3.631365319061448e-05, | |
| "loss": 1.0868, | |
| "step": 1725 | |
| }, | |
| { | |
| "epoch": 0.82, | |
| "learning_rate": 3.6134856176689126e-05, | |
| "loss": 1.089, | |
| "step": 1726 | |
| }, | |
| { | |
| "epoch": 0.82, | |
| "learning_rate": 3.5956456696915765e-05, | |
| "loss": 1.2464, | |
| "step": 1727 | |
| }, | |
| { | |
| "epoch": 0.82, | |
| "learning_rate": 3.577845518408856e-05, | |
| "loss": 0.941, | |
| "step": 1728 | |
| }, | |
| { | |
| "epoch": 0.82, | |
| "learning_rate": 3.5600852070036406e-05, | |
| "loss": 1.198, | |
| "step": 1729 | |
| }, | |
| { | |
| "epoch": 0.82, | |
| "learning_rate": 3.542364778562137e-05, | |
| "loss": 1.1854, | |
| "step": 1730 | |
| }, | |
| { | |
| "epoch": 0.82, | |
| "learning_rate": 3.5246842760738265e-05, | |
| "loss": 1.2015, | |
| "step": 1731 | |
| }, | |
| { | |
| "epoch": 0.82, | |
| "learning_rate": 3.5070437424313176e-05, | |
| "loss": 1.1797, | |
| "step": 1732 | |
| }, | |
| { | |
| "epoch": 0.82, | |
| "learning_rate": 3.489443220430251e-05, | |
| "loss": 1.3385, | |
| "step": 1733 | |
| }, | |
| { | |
| "epoch": 0.82, | |
| "learning_rate": 3.4718827527692044e-05, | |
| "loss": 1.3971, | |
| "step": 1734 | |
| }, | |
| { | |
| "epoch": 0.82, | |
| "learning_rate": 3.454362382049583e-05, | |
| "loss": 1.0747, | |
| "step": 1735 | |
| }, | |
| { | |
| "epoch": 0.82, | |
| "learning_rate": 3.4368821507755336e-05, | |
| "loss": 1.0842, | |
| "step": 1736 | |
| }, | |
| { | |
| "epoch": 0.82, | |
| "learning_rate": 3.4194421013537914e-05, | |
| "loss": 1.1598, | |
| "step": 1737 | |
| }, | |
| { | |
| "epoch": 0.82, | |
| "learning_rate": 3.4020422760936335e-05, | |
| "loss": 1.2907, | |
| "step": 1738 | |
| }, | |
| { | |
| "epoch": 0.82, | |
| "learning_rate": 3.38468271720676e-05, | |
| "loss": 1.3287, | |
| "step": 1739 | |
| }, | |
| { | |
| "epoch": 0.82, | |
| "learning_rate": 3.367363466807172e-05, | |
| "loss": 1.0782, | |
| "step": 1740 | |
| }, | |
| { | |
| "epoch": 0.82, | |
| "learning_rate": 3.350084566911087e-05, | |
| "loss": 1.4371, | |
| "step": 1741 | |
| }, | |
| { | |
| "epoch": 0.82, | |
| "learning_rate": 3.332846059436831e-05, | |
| "loss": 1.1154, | |
| "step": 1742 | |
| }, | |
| { | |
| "epoch": 0.82, | |
| "learning_rate": 3.3156479862047375e-05, | |
| "loss": 1.1617, | |
| "step": 1743 | |
| }, | |
| { | |
| "epoch": 0.82, | |
| "learning_rate": 3.2984903889370634e-05, | |
| "loss": 1.1814, | |
| "step": 1744 | |
| }, | |
| { | |
| "epoch": 0.82, | |
| "learning_rate": 3.281373309257842e-05, | |
| "loss": 1.3911, | |
| "step": 1745 | |
| }, | |
| { | |
| "epoch": 0.82, | |
| "learning_rate": 3.264296788692833e-05, | |
| "loss": 1.2625, | |
| "step": 1746 | |
| }, | |
| { | |
| "epoch": 0.82, | |
| "learning_rate": 3.2472608686694015e-05, | |
| "loss": 1.1116, | |
| "step": 1747 | |
| }, | |
| { | |
| "epoch": 0.83, | |
| "learning_rate": 3.230265590516408e-05, | |
| "loss": 1.3004, | |
| "step": 1748 | |
| }, | |
| { | |
| "epoch": 0.83, | |
| "learning_rate": 3.213310995464114e-05, | |
| "loss": 1.0188, | |
| "step": 1749 | |
| }, | |
| { | |
| "epoch": 0.83, | |
| "learning_rate": 3.1963971246440924e-05, | |
| "loss": 1.3164, | |
| "step": 1750 | |
| }, | |
| { | |
| "epoch": 0.83, | |
| "learning_rate": 3.179524019089117e-05, | |
| "loss": 1.0501, | |
| "step": 1751 | |
| }, | |
| { | |
| "epoch": 0.83, | |
| "learning_rate": 3.1626917197330616e-05, | |
| "loss": 1.2515, | |
| "step": 1752 | |
| }, | |
| { | |
| "epoch": 0.83, | |
| "learning_rate": 3.1459002674108085e-05, | |
| "loss": 1.2192, | |
| "step": 1753 | |
| }, | |
| { | |
| "epoch": 0.83, | |
| "learning_rate": 3.1291497028581516e-05, | |
| "loss": 1.1618, | |
| "step": 1754 | |
| }, | |
| { | |
| "epoch": 0.83, | |
| "learning_rate": 3.1124400667116884e-05, | |
| "loss": 1.1596, | |
| "step": 1755 | |
| }, | |
| { | |
| "epoch": 0.83, | |
| "learning_rate": 3.095771399508705e-05, | |
| "loss": 1.2991, | |
| "step": 1756 | |
| }, | |
| { | |
| "epoch": 0.83, | |
| "learning_rate": 3.0791437416871316e-05, | |
| "loss": 1.1086, | |
| "step": 1757 | |
| }, | |
| { | |
| "epoch": 0.83, | |
| "learning_rate": 3.0625571335853864e-05, | |
| "loss": 1.3552, | |
| "step": 1758 | |
| }, | |
| { | |
| "epoch": 0.83, | |
| "learning_rate": 3.0460116154423102e-05, | |
| "loss": 1.1716, | |
| "step": 1759 | |
| }, | |
| { | |
| "epoch": 0.83, | |
| "learning_rate": 3.0295072273970592e-05, | |
| "loss": 1.1182, | |
| "step": 1760 | |
| }, | |
| { | |
| "epoch": 0.83, | |
| "learning_rate": 3.0130440094890033e-05, | |
| "loss": 1.2788, | |
| "step": 1761 | |
| }, | |
| { | |
| "epoch": 0.83, | |
| "learning_rate": 2.996622001657652e-05, | |
| "loss": 1.5593, | |
| "step": 1762 | |
| }, | |
| { | |
| "epoch": 0.83, | |
| "learning_rate": 2.9802412437425077e-05, | |
| "loss": 1.216, | |
| "step": 1763 | |
| }, | |
| { | |
| "epoch": 0.83, | |
| "learning_rate": 2.9639017754830377e-05, | |
| "loss": 1.5077, | |
| "step": 1764 | |
| }, | |
| { | |
| "epoch": 0.83, | |
| "learning_rate": 2.9476036365185166e-05, | |
| "loss": 1.362, | |
| "step": 1765 | |
| }, | |
| { | |
| "epoch": 0.83, | |
| "learning_rate": 2.931346866387963e-05, | |
| "loss": 1.3215, | |
| "step": 1766 | |
| }, | |
| { | |
| "epoch": 0.83, | |
| "learning_rate": 2.9151315045300353e-05, | |
| "loss": 1.2743, | |
| "step": 1767 | |
| }, | |
| { | |
| "epoch": 0.83, | |
| "learning_rate": 2.8989575902829334e-05, | |
| "loss": 1.3791, | |
| "step": 1768 | |
| }, | |
| { | |
| "epoch": 0.84, | |
| "learning_rate": 2.8828251628843083e-05, | |
| "loss": 0.9143, | |
| "step": 1769 | |
| }, | |
| { | |
| "epoch": 0.84, | |
| "learning_rate": 2.8667342614711733e-05, | |
| "loss": 1.1611, | |
| "step": 1770 | |
| }, | |
| { | |
| "epoch": 0.84, | |
| "learning_rate": 2.850684925079774e-05, | |
| "loss": 1.0213, | |
| "step": 1771 | |
| }, | |
| { | |
| "epoch": 0.84, | |
| "learning_rate": 2.834677192645552e-05, | |
| "loss": 1.2423, | |
| "step": 1772 | |
| }, | |
| { | |
| "epoch": 0.84, | |
| "learning_rate": 2.818711103003e-05, | |
| "loss": 1.2826, | |
| "step": 1773 | |
| }, | |
| { | |
| "epoch": 0.84, | |
| "learning_rate": 2.8027866948855862e-05, | |
| "loss": 1.0059, | |
| "step": 1774 | |
| }, | |
| { | |
| "epoch": 0.84, | |
| "learning_rate": 2.786904006925668e-05, | |
| "loss": 1.2449, | |
| "step": 1775 | |
| }, | |
| { | |
| "epoch": 0.84, | |
| "learning_rate": 2.771063077654383e-05, | |
| "loss": 1.4577, | |
| "step": 1776 | |
| }, | |
| { | |
| "epoch": 0.84, | |
| "learning_rate": 2.7552639455015672e-05, | |
| "loss": 1.354, | |
| "step": 1777 | |
| }, | |
| { | |
| "epoch": 0.84, | |
| "learning_rate": 2.7395066487956557e-05, | |
| "loss": 1.3385, | |
| "step": 1778 | |
| }, | |
| { | |
| "epoch": 0.84, | |
| "learning_rate": 2.7237912257635877e-05, | |
| "loss": 1.1573, | |
| "step": 1779 | |
| }, | |
| { | |
| "epoch": 0.84, | |
| "learning_rate": 2.7081177145307314e-05, | |
| "loss": 1.2943, | |
| "step": 1780 | |
| }, | |
| { | |
| "epoch": 0.84, | |
| "learning_rate": 2.6924861531207613e-05, | |
| "loss": 1.3373, | |
| "step": 1781 | |
| }, | |
| { | |
| "epoch": 0.84, | |
| "learning_rate": 2.67689657945559e-05, | |
| "loss": 1.4295, | |
| "step": 1782 | |
| }, | |
| { | |
| "epoch": 0.84, | |
| "learning_rate": 2.6613490313552693e-05, | |
| "loss": 1.5029, | |
| "step": 1783 | |
| }, | |
| { | |
| "epoch": 0.84, | |
| "learning_rate": 2.6458435465378895e-05, | |
| "loss": 1.4016, | |
| "step": 1784 | |
| }, | |
| { | |
| "epoch": 0.84, | |
| "learning_rate": 2.6303801626195168e-05, | |
| "loss": 1.1433, | |
| "step": 1785 | |
| }, | |
| { | |
| "epoch": 0.84, | |
| "learning_rate": 2.614958917114052e-05, | |
| "loss": 1.0045, | |
| "step": 1786 | |
| }, | |
| { | |
| "epoch": 0.84, | |
| "learning_rate": 2.5995798474331868e-05, | |
| "loss": 1.2273, | |
| "step": 1787 | |
| }, | |
| { | |
| "epoch": 0.84, | |
| "learning_rate": 2.5842429908862986e-05, | |
| "loss": 0.8235, | |
| "step": 1788 | |
| }, | |
| { | |
| "epoch": 0.84, | |
| "learning_rate": 2.5689483846803475e-05, | |
| "loss": 1.2355, | |
| "step": 1789 | |
| }, | |
| { | |
| "epoch": 0.85, | |
| "learning_rate": 2.553696065919795e-05, | |
| "loss": 1.1482, | |
| "step": 1790 | |
| }, | |
| { | |
| "epoch": 0.85, | |
| "learning_rate": 2.5384860716065183e-05, | |
| "loss": 1.0946, | |
| "step": 1791 | |
| }, | |
| { | |
| "epoch": 0.85, | |
| "learning_rate": 2.5233184386397146e-05, | |
| "loss": 1.2652, | |
| "step": 1792 | |
| }, | |
| { | |
| "epoch": 0.85, | |
| "learning_rate": 2.5081932038158094e-05, | |
| "loss": 0.9681, | |
| "step": 1793 | |
| }, | |
| { | |
| "epoch": 0.85, | |
| "learning_rate": 2.4931104038283738e-05, | |
| "loss": 1.1351, | |
| "step": 1794 | |
| }, | |
| { | |
| "epoch": 0.85, | |
| "learning_rate": 2.4780700752680397e-05, | |
| "loss": 1.0974, | |
| "step": 1795 | |
| }, | |
| { | |
| "epoch": 0.85, | |
| "learning_rate": 2.4630722546223983e-05, | |
| "loss": 0.9365, | |
| "step": 1796 | |
| }, | |
| { | |
| "epoch": 0.85, | |
| "learning_rate": 2.448116978275905e-05, | |
| "loss": 1.2426, | |
| "step": 1797 | |
| }, | |
| { | |
| "epoch": 0.85, | |
| "learning_rate": 2.4332042825098245e-05, | |
| "loss": 1.2849, | |
| "step": 1798 | |
| }, | |
| { | |
| "epoch": 0.85, | |
| "learning_rate": 2.4183342035021107e-05, | |
| "loss": 1.3706, | |
| "step": 1799 | |
| }, | |
| { | |
| "epoch": 0.85, | |
| "learning_rate": 2.403506777327329e-05, | |
| "loss": 1.1806, | |
| "step": 1800 | |
| }, | |
| { | |
| "epoch": 0.85, | |
| "learning_rate": 2.3887220399565745e-05, | |
| "loss": 1.3546, | |
| "step": 1801 | |
| }, | |
| { | |
| "epoch": 0.85, | |
| "learning_rate": 2.373980027257374e-05, | |
| "loss": 1.2827, | |
| "step": 1802 | |
| }, | |
| { | |
| "epoch": 0.85, | |
| "learning_rate": 2.3592807749936196e-05, | |
| "loss": 1.2513, | |
| "step": 1803 | |
| }, | |
| { | |
| "epoch": 0.85, | |
| "learning_rate": 2.344624318825439e-05, | |
| "loss": 1.3511, | |
| "step": 1804 | |
| }, | |
| { | |
| "epoch": 0.85, | |
| "learning_rate": 2.330010694309168e-05, | |
| "loss": 1.1903, | |
| "step": 1805 | |
| }, | |
| { | |
| "epoch": 0.85, | |
| "learning_rate": 2.3154399368972167e-05, | |
| "loss": 1.1434, | |
| "step": 1806 | |
| }, | |
| { | |
| "epoch": 0.85, | |
| "learning_rate": 2.3009120819380026e-05, | |
| "loss": 1.0491, | |
| "step": 1807 | |
| }, | |
| { | |
| "epoch": 0.85, | |
| "learning_rate": 2.286427164675862e-05, | |
| "loss": 1.1442, | |
| "step": 1808 | |
| }, | |
| { | |
| "epoch": 0.85, | |
| "learning_rate": 2.2719852202509695e-05, | |
| "loss": 1.2043, | |
| "step": 1809 | |
| }, | |
| { | |
| "epoch": 0.85, | |
| "learning_rate": 2.25758628369924e-05, | |
| "loss": 1.289, | |
| "step": 1810 | |
| }, | |
| { | |
| "epoch": 0.86, | |
| "learning_rate": 2.2432303899522712e-05, | |
| "loss": 1.1428, | |
| "step": 1811 | |
| }, | |
| { | |
| "epoch": 0.86, | |
| "learning_rate": 2.2289175738372083e-05, | |
| "loss": 0.963, | |
| "step": 1812 | |
| }, | |
| { | |
| "epoch": 0.86, | |
| "learning_rate": 2.2146478700767205e-05, | |
| "loss": 1.487, | |
| "step": 1813 | |
| }, | |
| { | |
| "epoch": 0.86, | |
| "learning_rate": 2.2004213132888696e-05, | |
| "loss": 1.1215, | |
| "step": 1814 | |
| }, | |
| { | |
| "epoch": 0.86, | |
| "learning_rate": 2.1862379379870525e-05, | |
| "loss": 1.2184, | |
| "step": 1815 | |
| }, | |
| { | |
| "epoch": 0.86, | |
| "learning_rate": 2.1720977785799046e-05, | |
| "loss": 1.0798, | |
| "step": 1816 | |
| }, | |
| { | |
| "epoch": 0.86, | |
| "learning_rate": 2.1580008693712173e-05, | |
| "loss": 1.3635, | |
| "step": 1817 | |
| }, | |
| { | |
| "epoch": 0.86, | |
| "learning_rate": 2.1439472445598653e-05, | |
| "loss": 1.087, | |
| "step": 1818 | |
| }, | |
| { | |
| "epoch": 0.86, | |
| "learning_rate": 2.1299369382397093e-05, | |
| "loss": 1.2251, | |
| "step": 1819 | |
| }, | |
| { | |
| "epoch": 0.86, | |
| "learning_rate": 2.1159699843995217e-05, | |
| "loss": 1.1589, | |
| "step": 1820 | |
| }, | |
| { | |
| "epoch": 0.86, | |
| "learning_rate": 2.102046416922907e-05, | |
| "loss": 1.2955, | |
| "step": 1821 | |
| }, | |
| { | |
| "epoch": 0.86, | |
| "learning_rate": 2.088166269588212e-05, | |
| "loss": 1.5373, | |
| "step": 1822 | |
| }, | |
| { | |
| "epoch": 0.86, | |
| "learning_rate": 2.0743295760684435e-05, | |
| "loss": 1.3092, | |
| "step": 1823 | |
| }, | |
| { | |
| "epoch": 0.86, | |
| "learning_rate": 2.0605363699311943e-05, | |
| "loss": 1.17, | |
| "step": 1824 | |
| }, | |
| { | |
| "epoch": 0.86, | |
| "learning_rate": 2.0467866846385507e-05, | |
| "loss": 1.191, | |
| "step": 1825 | |
| }, | |
| { | |
| "epoch": 0.86, | |
| "learning_rate": 2.03308055354704e-05, | |
| "loss": 1.0686, | |
| "step": 1826 | |
| }, | |
| { | |
| "epoch": 0.86, | |
| "learning_rate": 2.0194180099074943e-05, | |
| "loss": 1.1097, | |
| "step": 1827 | |
| }, | |
| { | |
| "epoch": 0.86, | |
| "learning_rate": 2.0057990868650234e-05, | |
| "loss": 1.1788, | |
| "step": 1828 | |
| }, | |
| { | |
| "epoch": 0.86, | |
| "learning_rate": 1.9922238174589202e-05, | |
| "loss": 1.2758, | |
| "step": 1829 | |
| }, | |
| { | |
| "epoch": 0.86, | |
| "learning_rate": 1.978692234622559e-05, | |
| "loss": 1.5553, | |
| "step": 1830 | |
| }, | |
| { | |
| "epoch": 0.86, | |
| "learning_rate": 1.9652043711833402e-05, | |
| "loss": 1.0434, | |
| "step": 1831 | |
| }, | |
| { | |
| "epoch": 0.87, | |
| "learning_rate": 1.9517602598625983e-05, | |
| "loss": 1.3425, | |
| "step": 1832 | |
| }, | |
| { | |
| "epoch": 0.87, | |
| "learning_rate": 1.9383599332755266e-05, | |
| "loss": 1.4005, | |
| "step": 1833 | |
| }, | |
| { | |
| "epoch": 0.87, | |
| "learning_rate": 1.9250034239310975e-05, | |
| "loss": 1.1019, | |
| "step": 1834 | |
| }, | |
| { | |
| "epoch": 0.87, | |
| "learning_rate": 1.911690764231979e-05, | |
| "loss": 1.5482, | |
| "step": 1835 | |
| }, | |
| { | |
| "epoch": 0.87, | |
| "learning_rate": 1.8984219864744768e-05, | |
| "loss": 1.1322, | |
| "step": 1836 | |
| }, | |
| { | |
| "epoch": 0.87, | |
| "learning_rate": 1.8851971228484233e-05, | |
| "loss": 1.3651, | |
| "step": 1837 | |
| }, | |
| { | |
| "epoch": 0.87, | |
| "learning_rate": 1.8720162054371126e-05, | |
| "loss": 1.5033, | |
| "step": 1838 | |
| }, | |
| { | |
| "epoch": 0.87, | |
| "learning_rate": 1.858879266217246e-05, | |
| "loss": 1.1742, | |
| "step": 1839 | |
| }, | |
| { | |
| "epoch": 0.87, | |
| "learning_rate": 1.845786337058817e-05, | |
| "loss": 1.2404, | |
| "step": 1840 | |
| }, | |
| { | |
| "epoch": 0.87, | |
| "learning_rate": 1.8327374497250572e-05, | |
| "loss": 1.2239, | |
| "step": 1841 | |
| }, | |
| { | |
| "epoch": 0.87, | |
| "learning_rate": 1.8197326358723533e-05, | |
| "loss": 1.4019, | |
| "step": 1842 | |
| }, | |
| { | |
| "epoch": 0.87, | |
| "learning_rate": 1.806771927050166e-05, | |
| "loss": 0.9853, | |
| "step": 1843 | |
| }, | |
| { | |
| "epoch": 0.87, | |
| "learning_rate": 1.7938553547009774e-05, | |
| "loss": 0.8759, | |
| "step": 1844 | |
| }, | |
| { | |
| "epoch": 0.87, | |
| "learning_rate": 1.7809829501601594e-05, | |
| "loss": 1.1722, | |
| "step": 1845 | |
| }, | |
| { | |
| "epoch": 0.87, | |
| "learning_rate": 1.7681547446559722e-05, | |
| "loss": 1.3178, | |
| "step": 1846 | |
| }, | |
| { | |
| "epoch": 0.87, | |
| "learning_rate": 1.7553707693094213e-05, | |
| "loss": 1.494, | |
| "step": 1847 | |
| }, | |
| { | |
| "epoch": 0.87, | |
| "learning_rate": 1.7426310551342273e-05, | |
| "loss": 1.1439, | |
| "step": 1848 | |
| }, | |
| { | |
| "epoch": 0.87, | |
| "learning_rate": 1.7299356330367256e-05, | |
| "loss": 1.1022, | |
| "step": 1849 | |
| }, | |
| { | |
| "epoch": 0.87, | |
| "learning_rate": 1.7172845338157995e-05, | |
| "loss": 1.3888, | |
| "step": 1850 | |
| }, | |
| { | |
| "epoch": 0.87, | |
| "learning_rate": 1.7046777881628095e-05, | |
| "loss": 1.177, | |
| "step": 1851 | |
| }, | |
| { | |
| "epoch": 0.87, | |
| "learning_rate": 1.69211542666152e-05, | |
| "loss": 1.0688, | |
| "step": 1852 | |
| }, | |
| { | |
| "epoch": 0.87, | |
| "learning_rate": 1.679597479788002e-05, | |
| "loss": 1.4828, | |
| "step": 1853 | |
| }, | |
| { | |
| "epoch": 0.88, | |
| "learning_rate": 1.6671239779106008e-05, | |
| "loss": 1.4438, | |
| "step": 1854 | |
| }, | |
| { | |
| "epoch": 0.88, | |
| "learning_rate": 1.6546949512898236e-05, | |
| "loss": 1.5251, | |
| "step": 1855 | |
| }, | |
| { | |
| "epoch": 0.88, | |
| "learning_rate": 1.6423104300782822e-05, | |
| "loss": 1.1282, | |
| "step": 1856 | |
| }, | |
| { | |
| "epoch": 0.88, | |
| "learning_rate": 1.6299704443206253e-05, | |
| "loss": 1.0136, | |
| "step": 1857 | |
| }, | |
| { | |
| "epoch": 0.88, | |
| "learning_rate": 1.6176750239534555e-05, | |
| "loss": 1.2888, | |
| "step": 1858 | |
| }, | |
| { | |
| "epoch": 0.88, | |
| "learning_rate": 1.6054241988052608e-05, | |
| "loss": 1.0351, | |
| "step": 1859 | |
| }, | |
| { | |
| "epoch": 0.88, | |
| "learning_rate": 1.5932179985963412e-05, | |
| "loss": 1.3194, | |
| "step": 1860 | |
| }, | |
| { | |
| "epoch": 0.88, | |
| "learning_rate": 1.581056452938736e-05, | |
| "loss": 1.1395, | |
| "step": 1861 | |
| }, | |
| { | |
| "epoch": 0.88, | |
| "learning_rate": 1.5689395913361603e-05, | |
| "loss": 1.1652, | |
| "step": 1862 | |
| }, | |
| { | |
| "epoch": 0.88, | |
| "learning_rate": 1.5568674431839224e-05, | |
| "loss": 1.2574, | |
| "step": 1863 | |
| }, | |
| { | |
| "epoch": 0.88, | |
| "learning_rate": 1.544840037768851e-05, | |
| "loss": 1.3462, | |
| "step": 1864 | |
| }, | |
| { | |
| "epoch": 0.88, | |
| "learning_rate": 1.5328574042692368e-05, | |
| "loss": 1.5134, | |
| "step": 1865 | |
| }, | |
| { | |
| "epoch": 0.88, | |
| "learning_rate": 1.5209195717547486e-05, | |
| "loss": 1.0249, | |
| "step": 1866 | |
| }, | |
| { | |
| "epoch": 0.88, | |
| "learning_rate": 1.5090265691863848e-05, | |
| "loss": 1.2759, | |
| "step": 1867 | |
| }, | |
| { | |
| "epoch": 0.88, | |
| "learning_rate": 1.497178425416359e-05, | |
| "loss": 1.3408, | |
| "step": 1868 | |
| }, | |
| { | |
| "epoch": 0.88, | |
| "learning_rate": 1.485375169188079e-05, | |
| "loss": 1.2896, | |
| "step": 1869 | |
| }, | |
| { | |
| "epoch": 0.88, | |
| "learning_rate": 1.4736168291360552e-05, | |
| "loss": 1.3258, | |
| "step": 1870 | |
| }, | |
| { | |
| "epoch": 0.88, | |
| "learning_rate": 1.4619034337858229e-05, | |
| "loss": 1.3167, | |
| "step": 1871 | |
| }, | |
| { | |
| "epoch": 0.88, | |
| "learning_rate": 1.450235011553891e-05, | |
| "loss": 1.3734, | |
| "step": 1872 | |
| }, | |
| { | |
| "epoch": 0.88, | |
| "learning_rate": 1.438611590747656e-05, | |
| "loss": 1.1074, | |
| "step": 1873 | |
| }, | |
| { | |
| "epoch": 0.88, | |
| "learning_rate": 1.4270331995653507e-05, | |
| "loss": 1.6006, | |
| "step": 1874 | |
| }, | |
| { | |
| "epoch": 0.89, | |
| "learning_rate": 1.4154998660959596e-05, | |
| "loss": 1.131, | |
| "step": 1875 | |
| }, | |
| { | |
| "epoch": 0.89, | |
| "learning_rate": 1.4040116183191587e-05, | |
| "loss": 1.4408, | |
| "step": 1876 | |
| }, | |
| { | |
| "epoch": 0.89, | |
| "learning_rate": 1.3925684841052545e-05, | |
| "loss": 1.2286, | |
| "step": 1877 | |
| }, | |
| { | |
| "epoch": 0.89, | |
| "learning_rate": 1.3811704912151024e-05, | |
| "loss": 1.1798, | |
| "step": 1878 | |
| }, | |
| { | |
| "epoch": 0.89, | |
| "learning_rate": 1.3698176673000351e-05, | |
| "loss": 1.0013, | |
| "step": 1879 | |
| }, | |
| { | |
| "epoch": 0.89, | |
| "learning_rate": 1.3585100399018324e-05, | |
| "loss": 1.1529, | |
| "step": 1880 | |
| }, | |
| { | |
| "epoch": 0.89, | |
| "learning_rate": 1.3472476364526043e-05, | |
| "loss": 1.1891, | |
| "step": 1881 | |
| }, | |
| { | |
| "epoch": 0.89, | |
| "learning_rate": 1.3360304842747595e-05, | |
| "loss": 1.3186, | |
| "step": 1882 | |
| }, | |
| { | |
| "epoch": 0.89, | |
| "learning_rate": 1.3248586105809235e-05, | |
| "loss": 1.0345, | |
| "step": 1883 | |
| }, | |
| { | |
| "epoch": 0.89, | |
| "learning_rate": 1.3137320424738764e-05, | |
| "loss": 1.1464, | |
| "step": 1884 | |
| }, | |
| { | |
| "epoch": 0.89, | |
| "learning_rate": 1.3026508069465016e-05, | |
| "loss": 0.9625, | |
| "step": 1885 | |
| }, | |
| { | |
| "epoch": 0.89, | |
| "learning_rate": 1.2916149308816793e-05, | |
| "loss": 1.3932, | |
| "step": 1886 | |
| }, | |
| { | |
| "epoch": 0.89, | |
| "learning_rate": 1.2806244410522739e-05, | |
| "loss": 1.0426, | |
| "step": 1887 | |
| }, | |
| { | |
| "epoch": 0.89, | |
| "learning_rate": 1.2696793641210303e-05, | |
| "loss": 1.2517, | |
| "step": 1888 | |
| }, | |
| { | |
| "epoch": 0.89, | |
| "learning_rate": 1.2587797266405265e-05, | |
| "loss": 1.0455, | |
| "step": 1889 | |
| }, | |
| { | |
| "epoch": 0.89, | |
| "learning_rate": 1.2479255550531044e-05, | |
| "loss": 1.3104, | |
| "step": 1890 | |
| }, | |
| { | |
| "epoch": 0.89, | |
| "learning_rate": 1.237116875690807e-05, | |
| "loss": 1.1541, | |
| "step": 1891 | |
| }, | |
| { | |
| "epoch": 0.89, | |
| "learning_rate": 1.226353714775308e-05, | |
| "loss": 1.0212, | |
| "step": 1892 | |
| }, | |
| { | |
| "epoch": 0.89, | |
| "learning_rate": 1.2156360984178672e-05, | |
| "loss": 1.1716, | |
| "step": 1893 | |
| }, | |
| { | |
| "epoch": 0.89, | |
| "learning_rate": 1.2049640526192396e-05, | |
| "loss": 1.1511, | |
| "step": 1894 | |
| }, | |
| { | |
| "epoch": 0.89, | |
| "learning_rate": 1.1943376032696374e-05, | |
| "loss": 1.4378, | |
| "step": 1895 | |
| }, | |
| { | |
| "epoch": 0.9, | |
| "learning_rate": 1.18375677614865e-05, | |
| "loss": 1.1137, | |
| "step": 1896 | |
| }, | |
| { | |
| "epoch": 0.9, | |
| "learning_rate": 1.1732215969251936e-05, | |
| "loss": 1.2647, | |
| "step": 1897 | |
| }, | |
| { | |
| "epoch": 0.9, | |
| "learning_rate": 1.1627320911574369e-05, | |
| "loss": 1.2192, | |
| "step": 1898 | |
| }, | |
| { | |
| "epoch": 0.9, | |
| "learning_rate": 1.1522882842927507e-05, | |
| "loss": 1.2002, | |
| "step": 1899 | |
| }, | |
| { | |
| "epoch": 0.9, | |
| "learning_rate": 1.1418902016676392e-05, | |
| "loss": 1.12, | |
| "step": 1900 | |
| }, | |
| { | |
| "epoch": 0.9, | |
| "learning_rate": 1.1315378685076793e-05, | |
| "loss": 1.2486, | |
| "step": 1901 | |
| }, | |
| { | |
| "epoch": 0.9, | |
| "learning_rate": 1.1212313099274596e-05, | |
| "loss": 1.2119, | |
| "step": 1902 | |
| }, | |
| { | |
| "epoch": 0.9, | |
| "learning_rate": 1.1109705509305256e-05, | |
| "loss": 1.2259, | |
| "step": 1903 | |
| }, | |
| { | |
| "epoch": 0.9, | |
| "learning_rate": 1.1007556164093125e-05, | |
| "loss": 1.1457, | |
| "step": 1904 | |
| }, | |
| { | |
| "epoch": 0.9, | |
| "learning_rate": 1.0905865311450792e-05, | |
| "loss": 1.1518, | |
| "step": 1905 | |
| }, | |
| { | |
| "epoch": 0.9, | |
| "learning_rate": 1.0804633198078606e-05, | |
| "loss": 1.0608, | |
| "step": 1906 | |
| }, | |
| { | |
| "epoch": 0.9, | |
| "learning_rate": 1.0703860069564009e-05, | |
| "loss": 1.1139, | |
| "step": 1907 | |
| }, | |
| { | |
| "epoch": 0.9, | |
| "learning_rate": 1.0603546170381018e-05, | |
| "loss": 1.2249, | |
| "step": 1908 | |
| }, | |
| { | |
| "epoch": 0.9, | |
| "learning_rate": 1.0503691743889455e-05, | |
| "loss": 1.4376, | |
| "step": 1909 | |
| }, | |
| { | |
| "epoch": 0.9, | |
| "learning_rate": 1.0404297032334476e-05, | |
| "loss": 1.115, | |
| "step": 1910 | |
| }, | |
| { | |
| "epoch": 0.9, | |
| "learning_rate": 1.0305362276846131e-05, | |
| "loss": 1.1447, | |
| "step": 1911 | |
| }, | |
| { | |
| "epoch": 0.9, | |
| "learning_rate": 1.020688771743843e-05, | |
| "loss": 1.3071, | |
| "step": 1912 | |
| }, | |
| { | |
| "epoch": 0.9, | |
| "learning_rate": 1.0108873593009071e-05, | |
| "loss": 1.4344, | |
| "step": 1913 | |
| }, | |
| { | |
| "epoch": 0.9, | |
| "learning_rate": 1.0011320141338721e-05, | |
| "loss": 1.3076, | |
| "step": 1914 | |
| }, | |
| { | |
| "epoch": 0.9, | |
| "learning_rate": 9.914227599090442e-06, | |
| "loss": 1.2095, | |
| "step": 1915 | |
| }, | |
| { | |
| "epoch": 0.9, | |
| "learning_rate": 9.817596201809132e-06, | |
| "loss": 1.0488, | |
| "step": 1916 | |
| }, | |
| { | |
| "epoch": 0.91, | |
| "learning_rate": 9.72142618392098e-06, | |
| "loss": 1.0581, | |
| "step": 1917 | |
| }, | |
| { | |
| "epoch": 0.91, | |
| "learning_rate": 9.625717778732913e-06, | |
| "loss": 1.5264, | |
| "step": 1918 | |
| }, | |
| { | |
| "epoch": 0.91, | |
| "learning_rate": 9.530471218431957e-06, | |
| "loss": 0.7503, | |
| "step": 1919 | |
| }, | |
| { | |
| "epoch": 0.91, | |
| "learning_rate": 9.435686734084614e-06, | |
| "loss": 1.1666, | |
| "step": 1920 | |
| }, | |
| { | |
| "epoch": 0.91, | |
| "learning_rate": 9.341364555636568e-06, | |
| "loss": 1.1909, | |
| "step": 1921 | |
| }, | |
| { | |
| "epoch": 0.91, | |
| "learning_rate": 9.247504911911865e-06, | |
| "loss": 1.2914, | |
| "step": 1922 | |
| }, | |
| { | |
| "epoch": 0.91, | |
| "learning_rate": 9.154108030612451e-06, | |
| "loss": 1.0331, | |
| "step": 1923 | |
| }, | |
| { | |
| "epoch": 0.91, | |
| "learning_rate": 9.06117413831764e-06, | |
| "loss": 1.2396, | |
| "step": 1924 | |
| }, | |
| { | |
| "epoch": 0.91, | |
| "learning_rate": 8.968703460483508e-06, | |
| "loss": 1.3323, | |
| "step": 1925 | |
| }, | |
| { | |
| "epoch": 0.91, | |
| "learning_rate": 8.87669622144245e-06, | |
| "loss": 1.4465, | |
| "step": 1926 | |
| }, | |
| { | |
| "epoch": 0.91, | |
| "learning_rate": 8.78515264440245e-06, | |
| "loss": 1.2658, | |
| "step": 1927 | |
| }, | |
| { | |
| "epoch": 0.91, | |
| "learning_rate": 8.694072951446775e-06, | |
| "loss": 1.1626, | |
| "step": 1928 | |
| }, | |
| { | |
| "epoch": 0.91, | |
| "learning_rate": 8.603457363533274e-06, | |
| "loss": 1.1752, | |
| "step": 1929 | |
| }, | |
| { | |
| "epoch": 0.91, | |
| "learning_rate": 8.513306100493834e-06, | |
| "loss": 1.1018, | |
| "step": 1930 | |
| }, | |
| { | |
| "epoch": 0.91, | |
| "learning_rate": 8.423619381033997e-06, | |
| "loss": 1.2596, | |
| "step": 1931 | |
| }, | |
| { | |
| "epoch": 0.91, | |
| "learning_rate": 8.33439742273221e-06, | |
| "loss": 1.0342, | |
| "step": 1932 | |
| }, | |
| { | |
| "epoch": 0.91, | |
| "learning_rate": 8.245640442039504e-06, | |
| "loss": 1.3798, | |
| "step": 1933 | |
| }, | |
| { | |
| "epoch": 0.91, | |
| "learning_rate": 8.157348654278906e-06, | |
| "loss": 1.1456, | |
| "step": 1934 | |
| }, | |
| { | |
| "epoch": 0.91, | |
| "learning_rate": 8.06952227364477e-06, | |
| "loss": 1.2221, | |
| "step": 1935 | |
| }, | |
| { | |
| "epoch": 0.91, | |
| "learning_rate": 7.982161513202502e-06, | |
| "loss": 1.2754, | |
| "step": 1936 | |
| }, | |
| { | |
| "epoch": 0.91, | |
| "learning_rate": 7.895266584887884e-06, | |
| "loss": 0.9985, | |
| "step": 1937 | |
| }, | |
| { | |
| "epoch": 0.92, | |
| "learning_rate": 7.808837699506556e-06, | |
| "loss": 1.544, | |
| "step": 1938 | |
| }, | |
| { | |
| "epoch": 0.92, | |
| "learning_rate": 7.722875066733593e-06, | |
| "loss": 1.1587, | |
| "step": 1939 | |
| }, | |
| { | |
| "epoch": 0.92, | |
| "learning_rate": 7.637378895112911e-06, | |
| "loss": 1.3164, | |
| "step": 1940 | |
| }, | |
| { | |
| "epoch": 0.92, | |
| "learning_rate": 7.552349392056867e-06, | |
| "loss": 1.1158, | |
| "step": 1941 | |
| }, | |
| { | |
| "epoch": 0.92, | |
| "learning_rate": 7.467786763845586e-06, | |
| "loss": 0.9729, | |
| "step": 1942 | |
| }, | |
| { | |
| "epoch": 0.92, | |
| "learning_rate": 7.383691215626632e-06, | |
| "loss": 1.238, | |
| "step": 1943 | |
| }, | |
| { | |
| "epoch": 0.92, | |
| "learning_rate": 7.300062951414455e-06, | |
| "loss": 1.0044, | |
| "step": 1944 | |
| }, | |
| { | |
| "epoch": 0.92, | |
| "learning_rate": 7.2169021740898346e-06, | |
| "loss": 1.0541, | |
| "step": 1945 | |
| }, | |
| { | |
| "epoch": 0.92, | |
| "learning_rate": 7.134209085399457e-06, | |
| "loss": 1.2909, | |
| "step": 1946 | |
| }, | |
| { | |
| "epoch": 0.92, | |
| "learning_rate": 7.051983885955404e-06, | |
| "loss": 1.2217, | |
| "step": 1947 | |
| }, | |
| { | |
| "epoch": 0.92, | |
| "learning_rate": 6.970226775234623e-06, | |
| "loss": 1.309, | |
| "step": 1948 | |
| }, | |
| { | |
| "epoch": 0.92, | |
| "learning_rate": 6.888937951578589e-06, | |
| "loss": 1.2958, | |
| "step": 1949 | |
| }, | |
| { | |
| "epoch": 0.92, | |
| "learning_rate": 6.808117612192599e-06, | |
| "loss": 0.9312, | |
| "step": 1950 | |
| }, | |
| { | |
| "epoch": 0.92, | |
| "learning_rate": 6.727765953145482e-06, | |
| "loss": 1.1902, | |
| "step": 1951 | |
| }, | |
| { | |
| "epoch": 0.92, | |
| "learning_rate": 6.647883169369085e-06, | |
| "loss": 1.0265, | |
| "step": 1952 | |
| }, | |
| { | |
| "epoch": 0.92, | |
| "learning_rate": 6.568469454657677e-06, | |
| "loss": 1.1769, | |
| "step": 1953 | |
| }, | |
| { | |
| "epoch": 0.92, | |
| "learning_rate": 6.4895250016676625e-06, | |
| "loss": 1.3831, | |
| "step": 1954 | |
| }, | |
| { | |
| "epoch": 0.92, | |
| "learning_rate": 6.411050001916996e-06, | |
| "loss": 1.3066, | |
| "step": 1955 | |
| }, | |
| { | |
| "epoch": 0.92, | |
| "learning_rate": 6.333044645784703e-06, | |
| "loss": 1.2549, | |
| "step": 1956 | |
| }, | |
| { | |
| "epoch": 0.92, | |
| "learning_rate": 6.255509122510539e-06, | |
| "loss": 1.1581, | |
| "step": 1957 | |
| }, | |
| { | |
| "epoch": 0.92, | |
| "learning_rate": 6.178443620194352e-06, | |
| "loss": 1.0347, | |
| "step": 1958 | |
| }, | |
| { | |
| "epoch": 0.93, | |
| "learning_rate": 6.101848325795856e-06, | |
| "loss": 1.249, | |
| "step": 1959 | |
| }, | |
| { | |
| "epoch": 0.93, | |
| "learning_rate": 6.025723425133989e-06, | |
| "loss": 1.1094, | |
| "step": 1960 | |
| }, | |
| { | |
| "epoch": 0.93, | |
| "learning_rate": 5.950069102886402e-06, | |
| "loss": 1.363, | |
| "step": 1961 | |
| }, | |
| { | |
| "epoch": 0.93, | |
| "learning_rate": 5.8748855425893475e-06, | |
| "loss": 1.0292, | |
| "step": 1962 | |
| }, | |
| { | |
| "epoch": 0.93, | |
| "learning_rate": 5.800172926636882e-06, | |
| "loss": 1.0867, | |
| "step": 1963 | |
| }, | |
| { | |
| "epoch": 0.93, | |
| "learning_rate": 5.7259314362805735e-06, | |
| "loss": 1.2136, | |
| "step": 1964 | |
| }, | |
| { | |
| "epoch": 0.93, | |
| "learning_rate": 5.652161251629085e-06, | |
| "loss": 1.3425, | |
| "step": 1965 | |
| }, | |
| { | |
| "epoch": 0.93, | |
| "learning_rate": 5.578862551647612e-06, | |
| "loss": 1.28, | |
| "step": 1966 | |
| }, | |
| { | |
| "epoch": 0.93, | |
| "learning_rate": 5.50603551415767e-06, | |
| "loss": 1.2319, | |
| "step": 1967 | |
| }, | |
| { | |
| "epoch": 0.93, | |
| "learning_rate": 5.4336803158364205e-06, | |
| "loss": 1.3405, | |
| "step": 1968 | |
| }, | |
| { | |
| "epoch": 0.93, | |
| "learning_rate": 5.361797132216406e-06, | |
| "loss": 1.2899, | |
| "step": 1969 | |
| }, | |
| { | |
| "epoch": 0.93, | |
| "learning_rate": 5.290386137685066e-06, | |
| "loss": 1.3527, | |
| "step": 1970 | |
| }, | |
| { | |
| "epoch": 0.93, | |
| "learning_rate": 5.219447505484243e-06, | |
| "loss": 1.4975, | |
| "step": 1971 | |
| }, | |
| { | |
| "epoch": 0.93, | |
| "learning_rate": 5.148981407709963e-06, | |
| "loss": 1.2036, | |
| "step": 1972 | |
| }, | |
| { | |
| "epoch": 0.93, | |
| "learning_rate": 5.078988015311769e-06, | |
| "loss": 1.0778, | |
| "step": 1973 | |
| }, | |
| { | |
| "epoch": 0.93, | |
| "learning_rate": 5.009467498092502e-06, | |
| "loss": 1.5373, | |
| "step": 1974 | |
| }, | |
| { | |
| "epoch": 0.93, | |
| "learning_rate": 4.9404200247078485e-06, | |
| "loss": 1.3212, | |
| "step": 1975 | |
| }, | |
| { | |
| "epoch": 0.93, | |
| "learning_rate": 4.8718457626657496e-06, | |
| "loss": 1.0493, | |
| "step": 1976 | |
| }, | |
| { | |
| "epoch": 0.93, | |
| "learning_rate": 4.803744878326311e-06, | |
| "loss": 1.2744, | |
| "step": 1977 | |
| }, | |
| { | |
| "epoch": 0.93, | |
| "learning_rate": 4.736117536901152e-06, | |
| "loss": 1.421, | |
| "step": 1978 | |
| }, | |
| { | |
| "epoch": 0.93, | |
| "learning_rate": 4.66896390245306e-06, | |
| "loss": 0.9075, | |
| "step": 1979 | |
| }, | |
| { | |
| "epoch": 0.93, | |
| "learning_rate": 4.602284137895696e-06, | |
| "loss": 1.3903, | |
| "step": 1980 | |
| }, | |
| { | |
| "epoch": 0.94, | |
| "learning_rate": 4.53607840499306e-06, | |
| "loss": 1.1848, | |
| "step": 1981 | |
| }, | |
| { | |
| "epoch": 0.94, | |
| "learning_rate": 4.470346864359165e-06, | |
| "loss": 1.0173, | |
| "step": 1982 | |
| }, | |
| { | |
| "epoch": 0.94, | |
| "learning_rate": 4.40508967545763e-06, | |
| "loss": 1.3709, | |
| "step": 1983 | |
| }, | |
| { | |
| "epoch": 0.94, | |
| "learning_rate": 4.340306996601351e-06, | |
| "loss": 1.0662, | |
| "step": 1984 | |
| }, | |
| { | |
| "epoch": 0.94, | |
| "learning_rate": 4.2759989849520785e-06, | |
| "loss": 1.4046, | |
| "step": 1985 | |
| }, | |
| { | |
| "epoch": 0.94, | |
| "learning_rate": 4.212165796519929e-06, | |
| "loss": 1.1181, | |
| "step": 1986 | |
| }, | |
| { | |
| "epoch": 0.94, | |
| "learning_rate": 4.148807586163228e-06, | |
| "loss": 1.2173, | |
| "step": 1987 | |
| }, | |
| { | |
| "epoch": 0.94, | |
| "learning_rate": 4.085924507587913e-06, | |
| "loss": 1.4524, | |
| "step": 1988 | |
| }, | |
| { | |
| "epoch": 0.94, | |
| "learning_rate": 4.0235167133473086e-06, | |
| "loss": 1.2748, | |
| "step": 1989 | |
| }, | |
| { | |
| "epoch": 0.94, | |
| "learning_rate": 3.96158435484173e-06, | |
| "loss": 1.0592, | |
| "step": 1990 | |
| }, | |
| { | |
| "epoch": 0.94, | |
| "learning_rate": 3.900127582318036e-06, | |
| "loss": 1.3902, | |
| "step": 1991 | |
| }, | |
| { | |
| "epoch": 0.94, | |
| "learning_rate": 3.839146544869321e-06, | |
| "loss": 1.1882, | |
| "step": 1992 | |
| }, | |
| { | |
| "epoch": 0.94, | |
| "learning_rate": 3.7786413904346674e-06, | |
| "loss": 1.1128, | |
| "step": 1993 | |
| }, | |
| { | |
| "epoch": 0.94, | |
| "learning_rate": 3.7186122657985045e-06, | |
| "loss": 1.2136, | |
| "step": 1994 | |
| }, | |
| { | |
| "epoch": 0.94, | |
| "learning_rate": 3.659059316590541e-06, | |
| "loss": 1.0755, | |
| "step": 1995 | |
| }, | |
| { | |
| "epoch": 0.94, | |
| "learning_rate": 3.599982687285275e-06, | |
| "loss": 1.3191, | |
| "step": 1996 | |
| }, | |
| { | |
| "epoch": 0.94, | |
| "learning_rate": 3.5413825212015973e-06, | |
| "loss": 0.9365, | |
| "step": 1997 | |
| }, | |
| { | |
| "epoch": 0.94, | |
| "learning_rate": 3.4832589605025666e-06, | |
| "loss": 1.3355, | |
| "step": 1998 | |
| }, | |
| { | |
| "epoch": 0.94, | |
| "learning_rate": 3.4256121461949676e-06, | |
| "loss": 1.4349, | |
| "step": 1999 | |
| }, | |
| { | |
| "epoch": 0.94, | |
| "learning_rate": 3.3684422181290864e-06, | |
| "loss": 1.1091, | |
| "step": 2000 | |
| } | |
| ], | |
| "logging_steps": 1, | |
| "max_steps": 2117, | |
| "num_train_epochs": 1, | |
| "save_steps": 1000, | |
| "total_flos": 3.327487298555085e+17, | |
| "trial_name": null, | |
| "trial_params": null | |
| } | |