| { |
| "best_metric": null, |
| "best_model_checkpoint": null, |
| "epoch": 25.356576862123614, |
| "global_step": 2000, |
| "is_hyper_param_search": false, |
| "is_local_process_zero": true, |
| "is_world_process_zero": true, |
| "log_history": [ |
| { |
| "epoch": 0.01, |
| "learning_rate": 2.5e-06, |
| "loss": 9.043, |
| "step": 1 |
| }, |
| { |
| "epoch": 0.03, |
| "learning_rate": 5e-06, |
| "loss": 8.9062, |
| "step": 2 |
| }, |
| { |
| "epoch": 0.04, |
| "learning_rate": 7.5e-06, |
| "loss": 9.1133, |
| "step": 3 |
| }, |
| { |
| "epoch": 0.05, |
| "learning_rate": 1e-05, |
| "loss": 9.0039, |
| "step": 4 |
| }, |
| { |
| "epoch": 0.06, |
| "learning_rate": 1.25e-05, |
| "loss": 8.6875, |
| "step": 5 |
| }, |
| { |
| "epoch": 0.08, |
| "learning_rate": 1.5e-05, |
| "loss": 8.2402, |
| "step": 6 |
| }, |
| { |
| "epoch": 0.09, |
| "learning_rate": 1.75e-05, |
| "loss": 7.2227, |
| "step": 7 |
| }, |
| { |
| "epoch": 0.1, |
| "learning_rate": 2e-05, |
| "loss": 6.6484, |
| "step": 8 |
| }, |
| { |
| "epoch": 0.11, |
| "learning_rate": 2.25e-05, |
| "loss": 5.1504, |
| "step": 9 |
| }, |
| { |
| "epoch": 0.13, |
| "learning_rate": 2.5e-05, |
| "loss": 4.666, |
| "step": 10 |
| }, |
| { |
| "epoch": 0.14, |
| "learning_rate": 2.7500000000000004e-05, |
| "loss": 4.5039, |
| "step": 11 |
| }, |
| { |
| "epoch": 0.15, |
| "learning_rate": 3e-05, |
| "loss": 4.3262, |
| "step": 12 |
| }, |
| { |
| "epoch": 0.16, |
| "learning_rate": 3.2500000000000004e-05, |
| "loss": 3.9326, |
| "step": 13 |
| }, |
| { |
| "epoch": 0.18, |
| "learning_rate": 3.5e-05, |
| "loss": 4.0342, |
| "step": 14 |
| }, |
| { |
| "epoch": 0.19, |
| "learning_rate": 3.7500000000000003e-05, |
| "loss": 3.4961, |
| "step": 15 |
| }, |
| { |
| "epoch": 0.2, |
| "learning_rate": 4e-05, |
| "loss": 3.6494, |
| "step": 16 |
| }, |
| { |
| "epoch": 0.22, |
| "learning_rate": 4.25e-05, |
| "loss": 3.7373, |
| "step": 17 |
| }, |
| { |
| "epoch": 0.23, |
| "learning_rate": 4.5e-05, |
| "loss": 3.3711, |
| "step": 18 |
| }, |
| { |
| "epoch": 0.24, |
| "learning_rate": 4.75e-05, |
| "loss": 3.3633, |
| "step": 19 |
| }, |
| { |
| "epoch": 0.25, |
| "learning_rate": 5e-05, |
| "loss": 3.3066, |
| "step": 20 |
| }, |
| { |
| "epoch": 0.27, |
| "learning_rate": 5e-05, |
| "loss": 3.3828, |
| "step": 21 |
| }, |
| { |
| "epoch": 0.28, |
| "learning_rate": 5e-05, |
| "loss": 3.2881, |
| "step": 22 |
| }, |
| { |
| "epoch": 0.29, |
| "learning_rate": 5e-05, |
| "loss": 3.4941, |
| "step": 23 |
| }, |
| { |
| "epoch": 0.3, |
| "learning_rate": 5e-05, |
| "loss": 3.3477, |
| "step": 24 |
| }, |
| { |
| "epoch": 0.32, |
| "learning_rate": 5e-05, |
| "loss": 3.2959, |
| "step": 25 |
| }, |
| { |
| "epoch": 0.33, |
| "learning_rate": 5e-05, |
| "loss": 3.1514, |
| "step": 26 |
| }, |
| { |
| "epoch": 0.34, |
| "learning_rate": 5e-05, |
| "loss": 3.2627, |
| "step": 27 |
| }, |
| { |
| "epoch": 0.35, |
| "learning_rate": 5e-05, |
| "loss": 3.042, |
| "step": 28 |
| }, |
| { |
| "epoch": 0.37, |
| "learning_rate": 5e-05, |
| "loss": 3.083, |
| "step": 29 |
| }, |
| { |
| "epoch": 0.38, |
| "learning_rate": 5e-05, |
| "loss": 3.252, |
| "step": 30 |
| }, |
| { |
| "epoch": 0.39, |
| "learning_rate": 5e-05, |
| "loss": 3.0361, |
| "step": 31 |
| }, |
| { |
| "epoch": 0.41, |
| "learning_rate": 5e-05, |
| "loss": 2.8525, |
| "step": 32 |
| }, |
| { |
| "epoch": 0.42, |
| "learning_rate": 5e-05, |
| "loss": 2.832, |
| "step": 33 |
| }, |
| { |
| "epoch": 0.43, |
| "learning_rate": 5e-05, |
| "loss": 2.5557, |
| "step": 34 |
| }, |
| { |
| "epoch": 0.44, |
| "learning_rate": 5e-05, |
| "loss": 3.0713, |
| "step": 35 |
| }, |
| { |
| "epoch": 0.46, |
| "learning_rate": 5e-05, |
| "loss": 3.1123, |
| "step": 36 |
| }, |
| { |
| "epoch": 0.47, |
| "learning_rate": 5e-05, |
| "loss": 2.7788, |
| "step": 37 |
| }, |
| { |
| "epoch": 0.48, |
| "learning_rate": 5e-05, |
| "loss": 2.3345, |
| "step": 38 |
| }, |
| { |
| "epoch": 0.49, |
| "learning_rate": 5e-05, |
| "loss": 3.0747, |
| "step": 39 |
| }, |
| { |
| "epoch": 0.51, |
| "learning_rate": 5e-05, |
| "loss": 2.1865, |
| "step": 40 |
| }, |
| { |
| "epoch": 0.52, |
| "learning_rate": 5e-05, |
| "loss": 2.1064, |
| "step": 41 |
| }, |
| { |
| "epoch": 0.53, |
| "learning_rate": 5e-05, |
| "loss": 3.1128, |
| "step": 42 |
| }, |
| { |
| "epoch": 0.55, |
| "learning_rate": 5e-05, |
| "loss": 1.9976, |
| "step": 43 |
| }, |
| { |
| "epoch": 0.56, |
| "learning_rate": 5e-05, |
| "loss": 2.0723, |
| "step": 44 |
| }, |
| { |
| "epoch": 0.57, |
| "learning_rate": 5e-05, |
| "loss": 2.3511, |
| "step": 45 |
| }, |
| { |
| "epoch": 0.58, |
| "learning_rate": 5e-05, |
| "loss": 2.4839, |
| "step": 46 |
| }, |
| { |
| "epoch": 0.6, |
| "learning_rate": 5e-05, |
| "loss": 2.1694, |
| "step": 47 |
| }, |
| { |
| "epoch": 0.61, |
| "learning_rate": 5e-05, |
| "loss": 2.0386, |
| "step": 48 |
| }, |
| { |
| "epoch": 0.62, |
| "learning_rate": 5e-05, |
| "loss": 2.856, |
| "step": 49 |
| }, |
| { |
| "epoch": 0.63, |
| "learning_rate": 5e-05, |
| "loss": 2.564, |
| "step": 50 |
| }, |
| { |
| "epoch": 0.65, |
| "learning_rate": 5e-05, |
| "loss": 2.2476, |
| "step": 51 |
| }, |
| { |
| "epoch": 0.66, |
| "learning_rate": 5e-05, |
| "loss": 2.4429, |
| "step": 52 |
| }, |
| { |
| "epoch": 0.67, |
| "learning_rate": 5e-05, |
| "loss": 2.4146, |
| "step": 53 |
| }, |
| { |
| "epoch": 0.68, |
| "learning_rate": 5e-05, |
| "loss": 2.2959, |
| "step": 54 |
| }, |
| { |
| "epoch": 0.7, |
| "learning_rate": 5e-05, |
| "loss": 2.3281, |
| "step": 55 |
| }, |
| { |
| "epoch": 0.71, |
| "learning_rate": 5e-05, |
| "loss": 2.0942, |
| "step": 56 |
| }, |
| { |
| "epoch": 0.72, |
| "learning_rate": 5e-05, |
| "loss": 2.1743, |
| "step": 57 |
| }, |
| { |
| "epoch": 0.74, |
| "learning_rate": 5e-05, |
| "loss": 2.0464, |
| "step": 58 |
| }, |
| { |
| "epoch": 0.75, |
| "learning_rate": 5e-05, |
| "loss": 2.0469, |
| "step": 59 |
| }, |
| { |
| "epoch": 0.76, |
| "learning_rate": 5e-05, |
| "loss": 2.0874, |
| "step": 60 |
| }, |
| { |
| "epoch": 0.77, |
| "learning_rate": 5e-05, |
| "loss": 2.3115, |
| "step": 61 |
| }, |
| { |
| "epoch": 0.79, |
| "learning_rate": 5e-05, |
| "loss": 1.5884, |
| "step": 62 |
| }, |
| { |
| "epoch": 0.8, |
| "learning_rate": 5e-05, |
| "loss": 2.1963, |
| "step": 63 |
| }, |
| { |
| "epoch": 0.81, |
| "learning_rate": 5e-05, |
| "loss": 1.8401, |
| "step": 64 |
| }, |
| { |
| "epoch": 0.82, |
| "learning_rate": 5e-05, |
| "loss": 1.9893, |
| "step": 65 |
| }, |
| { |
| "epoch": 0.84, |
| "learning_rate": 5e-05, |
| "loss": 2.5669, |
| "step": 66 |
| }, |
| { |
| "epoch": 0.85, |
| "learning_rate": 5e-05, |
| "loss": 2.2998, |
| "step": 67 |
| }, |
| { |
| "epoch": 0.86, |
| "learning_rate": 5e-05, |
| "loss": 2.0239, |
| "step": 68 |
| }, |
| { |
| "epoch": 0.87, |
| "learning_rate": 5e-05, |
| "loss": 2.2612, |
| "step": 69 |
| }, |
| { |
| "epoch": 0.89, |
| "learning_rate": 5e-05, |
| "loss": 1.856, |
| "step": 70 |
| }, |
| { |
| "epoch": 0.9, |
| "learning_rate": 5e-05, |
| "loss": 1.7812, |
| "step": 71 |
| }, |
| { |
| "epoch": 0.91, |
| "learning_rate": 5e-05, |
| "loss": 2.0645, |
| "step": 72 |
| }, |
| { |
| "epoch": 0.93, |
| "learning_rate": 5e-05, |
| "loss": 2.1143, |
| "step": 73 |
| }, |
| { |
| "epoch": 0.94, |
| "learning_rate": 5e-05, |
| "loss": 1.7314, |
| "step": 74 |
| }, |
| { |
| "epoch": 0.95, |
| "learning_rate": 5e-05, |
| "loss": 1.8833, |
| "step": 75 |
| }, |
| { |
| "epoch": 0.96, |
| "learning_rate": 5e-05, |
| "loss": 1.8752, |
| "step": 76 |
| }, |
| { |
| "epoch": 0.98, |
| "learning_rate": 5e-05, |
| "loss": 1.5498, |
| "step": 77 |
| }, |
| { |
| "epoch": 0.99, |
| "learning_rate": 5e-05, |
| "loss": 2.1572, |
| "step": 78 |
| }, |
| { |
| "epoch": 1.0, |
| "learning_rate": 5e-05, |
| "loss": 1.7842, |
| "step": 79 |
| }, |
| { |
| "epoch": 1.01, |
| "learning_rate": 5e-05, |
| "loss": 1.9199, |
| "step": 80 |
| }, |
| { |
| "epoch": 1.03, |
| "learning_rate": 5e-05, |
| "loss": 1.8677, |
| "step": 81 |
| }, |
| { |
| "epoch": 1.04, |
| "learning_rate": 5e-05, |
| "loss": 1.866, |
| "step": 82 |
| }, |
| { |
| "epoch": 1.05, |
| "learning_rate": 5e-05, |
| "loss": 1.7832, |
| "step": 83 |
| }, |
| { |
| "epoch": 1.06, |
| "learning_rate": 5e-05, |
| "loss": 2.4292, |
| "step": 84 |
| }, |
| { |
| "epoch": 1.08, |
| "learning_rate": 5e-05, |
| "loss": 2.3906, |
| "step": 85 |
| }, |
| { |
| "epoch": 1.09, |
| "learning_rate": 5e-05, |
| "loss": 2.1914, |
| "step": 86 |
| }, |
| { |
| "epoch": 1.1, |
| "learning_rate": 5e-05, |
| "loss": 1.8486, |
| "step": 87 |
| }, |
| { |
| "epoch": 1.12, |
| "learning_rate": 5e-05, |
| "loss": 1.8213, |
| "step": 88 |
| }, |
| { |
| "epoch": 1.13, |
| "learning_rate": 5e-05, |
| "loss": 1.9126, |
| "step": 89 |
| }, |
| { |
| "epoch": 1.14, |
| "learning_rate": 5e-05, |
| "loss": 2.1084, |
| "step": 90 |
| }, |
| { |
| "epoch": 1.15, |
| "learning_rate": 5e-05, |
| "loss": 1.4355, |
| "step": 91 |
| }, |
| { |
| "epoch": 1.17, |
| "learning_rate": 5e-05, |
| "loss": 2.6309, |
| "step": 92 |
| }, |
| { |
| "epoch": 1.18, |
| "learning_rate": 5e-05, |
| "loss": 1.7529, |
| "step": 93 |
| }, |
| { |
| "epoch": 1.19, |
| "learning_rate": 5e-05, |
| "loss": 1.4246, |
| "step": 94 |
| }, |
| { |
| "epoch": 1.2, |
| "learning_rate": 5e-05, |
| "loss": 2.0332, |
| "step": 95 |
| }, |
| { |
| "epoch": 1.22, |
| "learning_rate": 5e-05, |
| "loss": 1.7593, |
| "step": 96 |
| }, |
| { |
| "epoch": 1.23, |
| "learning_rate": 5e-05, |
| "loss": 2.0186, |
| "step": 97 |
| }, |
| { |
| "epoch": 1.24, |
| "learning_rate": 5e-05, |
| "loss": 1.7654, |
| "step": 98 |
| }, |
| { |
| "epoch": 1.26, |
| "learning_rate": 5e-05, |
| "loss": 1.8352, |
| "step": 99 |
| }, |
| { |
| "epoch": 1.27, |
| "learning_rate": 5e-05, |
| "loss": 1.9382, |
| "step": 100 |
| }, |
| { |
| "epoch": 1.28, |
| "learning_rate": 5e-05, |
| "loss": 1.9531, |
| "step": 101 |
| }, |
| { |
| "epoch": 1.29, |
| "learning_rate": 5e-05, |
| "loss": 1.7942, |
| "step": 102 |
| }, |
| { |
| "epoch": 1.31, |
| "learning_rate": 5e-05, |
| "loss": 1.6328, |
| "step": 103 |
| }, |
| { |
| "epoch": 1.32, |
| "learning_rate": 5e-05, |
| "loss": 1.5959, |
| "step": 104 |
| }, |
| { |
| "epoch": 1.33, |
| "learning_rate": 5e-05, |
| "loss": 1.644, |
| "step": 105 |
| }, |
| { |
| "epoch": 1.34, |
| "learning_rate": 5e-05, |
| "loss": 1.9714, |
| "step": 106 |
| }, |
| { |
| "epoch": 1.36, |
| "learning_rate": 5e-05, |
| "loss": 1.7971, |
| "step": 107 |
| }, |
| { |
| "epoch": 1.37, |
| "learning_rate": 5e-05, |
| "loss": 1.7383, |
| "step": 108 |
| }, |
| { |
| "epoch": 1.38, |
| "learning_rate": 5e-05, |
| "loss": 1.5474, |
| "step": 109 |
| }, |
| { |
| "epoch": 1.39, |
| "learning_rate": 5e-05, |
| "loss": 1.9756, |
| "step": 110 |
| }, |
| { |
| "epoch": 1.41, |
| "learning_rate": 5e-05, |
| "loss": 1.834, |
| "step": 111 |
| }, |
| { |
| "epoch": 1.42, |
| "learning_rate": 5e-05, |
| "loss": 2.0366, |
| "step": 112 |
| }, |
| { |
| "epoch": 1.43, |
| "learning_rate": 5e-05, |
| "loss": 1.5786, |
| "step": 113 |
| }, |
| { |
| "epoch": 1.45, |
| "learning_rate": 5e-05, |
| "loss": 1.8486, |
| "step": 114 |
| }, |
| { |
| "epoch": 1.46, |
| "learning_rate": 5e-05, |
| "loss": 1.8472, |
| "step": 115 |
| }, |
| { |
| "epoch": 1.47, |
| "learning_rate": 5e-05, |
| "loss": 1.7871, |
| "step": 116 |
| }, |
| { |
| "epoch": 1.48, |
| "learning_rate": 5e-05, |
| "loss": 1.7839, |
| "step": 117 |
| }, |
| { |
| "epoch": 1.5, |
| "learning_rate": 5e-05, |
| "loss": 1.6704, |
| "step": 118 |
| }, |
| { |
| "epoch": 1.51, |
| "learning_rate": 5e-05, |
| "loss": 1.9126, |
| "step": 119 |
| }, |
| { |
| "epoch": 1.52, |
| "learning_rate": 5e-05, |
| "loss": 1.8105, |
| "step": 120 |
| }, |
| { |
| "epoch": 1.53, |
| "learning_rate": 5e-05, |
| "loss": 1.8293, |
| "step": 121 |
| }, |
| { |
| "epoch": 1.55, |
| "learning_rate": 5e-05, |
| "loss": 1.6152, |
| "step": 122 |
| }, |
| { |
| "epoch": 1.56, |
| "learning_rate": 5e-05, |
| "loss": 1.6809, |
| "step": 123 |
| }, |
| { |
| "epoch": 1.57, |
| "learning_rate": 5e-05, |
| "loss": 1.592, |
| "step": 124 |
| }, |
| { |
| "epoch": 1.58, |
| "learning_rate": 5e-05, |
| "loss": 1.9656, |
| "step": 125 |
| }, |
| { |
| "epoch": 1.6, |
| "learning_rate": 5e-05, |
| "loss": 2.1069, |
| "step": 126 |
| }, |
| { |
| "epoch": 1.61, |
| "learning_rate": 5e-05, |
| "loss": 1.4539, |
| "step": 127 |
| }, |
| { |
| "epoch": 1.62, |
| "learning_rate": 5e-05, |
| "loss": 1.8301, |
| "step": 128 |
| }, |
| { |
| "epoch": 1.64, |
| "learning_rate": 5e-05, |
| "loss": 1.6951, |
| "step": 129 |
| }, |
| { |
| "epoch": 1.65, |
| "learning_rate": 5e-05, |
| "loss": 1.9124, |
| "step": 130 |
| }, |
| { |
| "epoch": 1.66, |
| "learning_rate": 5e-05, |
| "loss": 1.6172, |
| "step": 131 |
| }, |
| { |
| "epoch": 1.67, |
| "learning_rate": 5e-05, |
| "loss": 1.7207, |
| "step": 132 |
| }, |
| { |
| "epoch": 1.69, |
| "learning_rate": 5e-05, |
| "loss": 1.541, |
| "step": 133 |
| }, |
| { |
| "epoch": 1.7, |
| "learning_rate": 5e-05, |
| "loss": 1.9668, |
| "step": 134 |
| }, |
| { |
| "epoch": 1.71, |
| "learning_rate": 5e-05, |
| "loss": 1.4446, |
| "step": 135 |
| }, |
| { |
| "epoch": 1.72, |
| "learning_rate": 5e-05, |
| "loss": 2.1938, |
| "step": 136 |
| }, |
| { |
| "epoch": 1.74, |
| "learning_rate": 5e-05, |
| "loss": 1.6897, |
| "step": 137 |
| }, |
| { |
| "epoch": 1.75, |
| "learning_rate": 5e-05, |
| "loss": 1.8652, |
| "step": 138 |
| }, |
| { |
| "epoch": 1.76, |
| "learning_rate": 5e-05, |
| "loss": 1.9941, |
| "step": 139 |
| }, |
| { |
| "epoch": 1.77, |
| "learning_rate": 5e-05, |
| "loss": 1.645, |
| "step": 140 |
| }, |
| { |
| "epoch": 1.79, |
| "learning_rate": 5e-05, |
| "loss": 2.1062, |
| "step": 141 |
| }, |
| { |
| "epoch": 1.8, |
| "learning_rate": 5e-05, |
| "loss": 1.696, |
| "step": 142 |
| }, |
| { |
| "epoch": 1.81, |
| "learning_rate": 5e-05, |
| "loss": 1.748, |
| "step": 143 |
| }, |
| { |
| "epoch": 1.83, |
| "learning_rate": 5e-05, |
| "loss": 1.8682, |
| "step": 144 |
| }, |
| { |
| "epoch": 1.84, |
| "learning_rate": 5e-05, |
| "loss": 1.8108, |
| "step": 145 |
| }, |
| { |
| "epoch": 1.85, |
| "learning_rate": 5e-05, |
| "loss": 1.5439, |
| "step": 146 |
| }, |
| { |
| "epoch": 1.86, |
| "learning_rate": 5e-05, |
| "loss": 1.7905, |
| "step": 147 |
| }, |
| { |
| "epoch": 1.88, |
| "learning_rate": 5e-05, |
| "loss": 2.0469, |
| "step": 148 |
| }, |
| { |
| "epoch": 1.89, |
| "learning_rate": 5e-05, |
| "loss": 1.9272, |
| "step": 149 |
| }, |
| { |
| "epoch": 1.9, |
| "learning_rate": 5e-05, |
| "loss": 1.5505, |
| "step": 150 |
| }, |
| { |
| "epoch": 1.91, |
| "learning_rate": 5e-05, |
| "loss": 1.9673, |
| "step": 151 |
| }, |
| { |
| "epoch": 1.93, |
| "learning_rate": 5e-05, |
| "loss": 1.9258, |
| "step": 152 |
| }, |
| { |
| "epoch": 1.94, |
| "learning_rate": 5e-05, |
| "loss": 1.77, |
| "step": 153 |
| }, |
| { |
| "epoch": 1.95, |
| "learning_rate": 5e-05, |
| "loss": 1.9231, |
| "step": 154 |
| }, |
| { |
| "epoch": 1.97, |
| "learning_rate": 5e-05, |
| "loss": 1.855, |
| "step": 155 |
| }, |
| { |
| "epoch": 1.98, |
| "learning_rate": 5e-05, |
| "loss": 1.6724, |
| "step": 156 |
| }, |
| { |
| "epoch": 1.99, |
| "learning_rate": 5e-05, |
| "loss": 1.6978, |
| "step": 157 |
| }, |
| { |
| "epoch": 2.0, |
| "learning_rate": 5e-05, |
| "loss": 1.6162, |
| "step": 158 |
| }, |
| { |
| "epoch": 2.02, |
| "learning_rate": 5e-05, |
| "loss": 1.668, |
| "step": 159 |
| }, |
| { |
| "epoch": 2.03, |
| "learning_rate": 5e-05, |
| "loss": 1.6396, |
| "step": 160 |
| }, |
| { |
| "epoch": 2.04, |
| "learning_rate": 5e-05, |
| "loss": 1.4229, |
| "step": 161 |
| }, |
| { |
| "epoch": 2.05, |
| "learning_rate": 5e-05, |
| "loss": 1.4958, |
| "step": 162 |
| }, |
| { |
| "epoch": 2.07, |
| "learning_rate": 5e-05, |
| "loss": 1.7268, |
| "step": 163 |
| }, |
| { |
| "epoch": 2.08, |
| "learning_rate": 5e-05, |
| "loss": 1.3389, |
| "step": 164 |
| }, |
| { |
| "epoch": 2.09, |
| "learning_rate": 5e-05, |
| "loss": 2.2952, |
| "step": 165 |
| }, |
| { |
| "epoch": 2.1, |
| "learning_rate": 5e-05, |
| "loss": 1.616, |
| "step": 166 |
| }, |
| { |
| "epoch": 2.12, |
| "learning_rate": 5e-05, |
| "loss": 1.8743, |
| "step": 167 |
| }, |
| { |
| "epoch": 2.13, |
| "learning_rate": 5e-05, |
| "loss": 1.6609, |
| "step": 168 |
| }, |
| { |
| "epoch": 2.14, |
| "learning_rate": 5e-05, |
| "loss": 1.6138, |
| "step": 169 |
| }, |
| { |
| "epoch": 2.16, |
| "learning_rate": 5e-05, |
| "loss": 1.6819, |
| "step": 170 |
| }, |
| { |
| "epoch": 2.17, |
| "learning_rate": 5e-05, |
| "loss": 1.6897, |
| "step": 171 |
| }, |
| { |
| "epoch": 2.18, |
| "learning_rate": 5e-05, |
| "loss": 1.7712, |
| "step": 172 |
| }, |
| { |
| "epoch": 2.19, |
| "learning_rate": 5e-05, |
| "loss": 1.4998, |
| "step": 173 |
| }, |
| { |
| "epoch": 2.21, |
| "learning_rate": 5e-05, |
| "loss": 1.5398, |
| "step": 174 |
| }, |
| { |
| "epoch": 2.22, |
| "learning_rate": 5e-05, |
| "loss": 2.0825, |
| "step": 175 |
| }, |
| { |
| "epoch": 2.23, |
| "learning_rate": 5e-05, |
| "loss": 1.6025, |
| "step": 176 |
| }, |
| { |
| "epoch": 2.24, |
| "learning_rate": 5e-05, |
| "loss": 1.802, |
| "step": 177 |
| }, |
| { |
| "epoch": 2.26, |
| "learning_rate": 5e-05, |
| "loss": 1.7847, |
| "step": 178 |
| }, |
| { |
| "epoch": 2.27, |
| "learning_rate": 5e-05, |
| "loss": 1.6587, |
| "step": 179 |
| }, |
| { |
| "epoch": 2.28, |
| "learning_rate": 5e-05, |
| "loss": 1.6548, |
| "step": 180 |
| }, |
| { |
| "epoch": 2.29, |
| "learning_rate": 5e-05, |
| "loss": 1.9868, |
| "step": 181 |
| }, |
| { |
| "epoch": 2.31, |
| "learning_rate": 5e-05, |
| "loss": 1.5156, |
| "step": 182 |
| }, |
| { |
| "epoch": 2.32, |
| "learning_rate": 5e-05, |
| "loss": 1.4768, |
| "step": 183 |
| }, |
| { |
| "epoch": 2.33, |
| "learning_rate": 5e-05, |
| "loss": 1.8108, |
| "step": 184 |
| }, |
| { |
| "epoch": 2.35, |
| "learning_rate": 5e-05, |
| "loss": 1.4958, |
| "step": 185 |
| }, |
| { |
| "epoch": 2.36, |
| "learning_rate": 5e-05, |
| "loss": 1.8848, |
| "step": 186 |
| }, |
| { |
| "epoch": 2.37, |
| "learning_rate": 5e-05, |
| "loss": 1.4009, |
| "step": 187 |
| }, |
| { |
| "epoch": 2.38, |
| "learning_rate": 5e-05, |
| "loss": 1.4219, |
| "step": 188 |
| }, |
| { |
| "epoch": 2.4, |
| "learning_rate": 5e-05, |
| "loss": 1.334, |
| "step": 189 |
| }, |
| { |
| "epoch": 2.41, |
| "learning_rate": 5e-05, |
| "loss": 1.5735, |
| "step": 190 |
| }, |
| { |
| "epoch": 2.42, |
| "learning_rate": 5e-05, |
| "loss": 1.7261, |
| "step": 191 |
| }, |
| { |
| "epoch": 2.43, |
| "learning_rate": 5e-05, |
| "loss": 1.4009, |
| "step": 192 |
| }, |
| { |
| "epoch": 2.45, |
| "learning_rate": 5e-05, |
| "loss": 1.2161, |
| "step": 193 |
| }, |
| { |
| "epoch": 2.46, |
| "learning_rate": 5e-05, |
| "loss": 2.0112, |
| "step": 194 |
| }, |
| { |
| "epoch": 2.47, |
| "learning_rate": 5e-05, |
| "loss": 1.4658, |
| "step": 195 |
| }, |
| { |
| "epoch": 2.48, |
| "learning_rate": 5e-05, |
| "loss": 1.8738, |
| "step": 196 |
| }, |
| { |
| "epoch": 2.5, |
| "learning_rate": 5e-05, |
| "loss": 1.9304, |
| "step": 197 |
| }, |
| { |
| "epoch": 2.51, |
| "learning_rate": 5e-05, |
| "loss": 1.7585, |
| "step": 198 |
| }, |
| { |
| "epoch": 2.52, |
| "learning_rate": 5e-05, |
| "loss": 1.7458, |
| "step": 199 |
| }, |
| { |
| "epoch": 2.54, |
| "learning_rate": 5e-05, |
| "loss": 1.5288, |
| "step": 200 |
| }, |
| { |
| "epoch": 2.55, |
| "learning_rate": 5e-05, |
| "loss": 1.4287, |
| "step": 201 |
| }, |
| { |
| "epoch": 2.56, |
| "learning_rate": 5e-05, |
| "loss": 1.6968, |
| "step": 202 |
| }, |
| { |
| "epoch": 2.57, |
| "learning_rate": 5e-05, |
| "loss": 1.7329, |
| "step": 203 |
| }, |
| { |
| "epoch": 2.59, |
| "learning_rate": 5e-05, |
| "loss": 1.585, |
| "step": 204 |
| }, |
| { |
| "epoch": 2.6, |
| "learning_rate": 5e-05, |
| "loss": 2.0723, |
| "step": 205 |
| }, |
| { |
| "epoch": 2.61, |
| "learning_rate": 5e-05, |
| "loss": 1.5513, |
| "step": 206 |
| }, |
| { |
| "epoch": 2.62, |
| "learning_rate": 5e-05, |
| "loss": 1.6028, |
| "step": 207 |
| }, |
| { |
| "epoch": 2.64, |
| "learning_rate": 5e-05, |
| "loss": 1.6194, |
| "step": 208 |
| }, |
| { |
| "epoch": 2.65, |
| "learning_rate": 5e-05, |
| "loss": 1.8682, |
| "step": 209 |
| }, |
| { |
| "epoch": 2.66, |
| "learning_rate": 5e-05, |
| "loss": 1.7119, |
| "step": 210 |
| }, |
| { |
| "epoch": 2.68, |
| "learning_rate": 5e-05, |
| "loss": 1.561, |
| "step": 211 |
| }, |
| { |
| "epoch": 2.69, |
| "learning_rate": 5e-05, |
| "loss": 1.6775, |
| "step": 212 |
| }, |
| { |
| "epoch": 2.7, |
| "learning_rate": 5e-05, |
| "loss": 1.7085, |
| "step": 213 |
| }, |
| { |
| "epoch": 2.71, |
| "learning_rate": 5e-05, |
| "loss": 1.8184, |
| "step": 214 |
| }, |
| { |
| "epoch": 2.73, |
| "learning_rate": 5e-05, |
| "loss": 1.9097, |
| "step": 215 |
| }, |
| { |
| "epoch": 2.74, |
| "learning_rate": 5e-05, |
| "loss": 1.5383, |
| "step": 216 |
| }, |
| { |
| "epoch": 2.75, |
| "learning_rate": 5e-05, |
| "loss": 1.7207, |
| "step": 217 |
| }, |
| { |
| "epoch": 2.76, |
| "learning_rate": 5e-05, |
| "loss": 1.6553, |
| "step": 218 |
| }, |
| { |
| "epoch": 2.78, |
| "learning_rate": 5e-05, |
| "loss": 1.1951, |
| "step": 219 |
| }, |
| { |
| "epoch": 2.79, |
| "learning_rate": 5e-05, |
| "loss": 1.4666, |
| "step": 220 |
| }, |
| { |
| "epoch": 2.8, |
| "learning_rate": 5e-05, |
| "loss": 1.5291, |
| "step": 221 |
| }, |
| { |
| "epoch": 2.81, |
| "learning_rate": 5e-05, |
| "loss": 1.2676, |
| "step": 222 |
| }, |
| { |
| "epoch": 2.83, |
| "learning_rate": 5e-05, |
| "loss": 1.7742, |
| "step": 223 |
| }, |
| { |
| "epoch": 2.84, |
| "learning_rate": 5e-05, |
| "loss": 1.8682, |
| "step": 224 |
| }, |
| { |
| "epoch": 2.85, |
| "learning_rate": 5e-05, |
| "loss": 1.4187, |
| "step": 225 |
| }, |
| { |
| "epoch": 2.87, |
| "learning_rate": 5e-05, |
| "loss": 1.8169, |
| "step": 226 |
| }, |
| { |
| "epoch": 2.88, |
| "learning_rate": 5e-05, |
| "loss": 1.8091, |
| "step": 227 |
| }, |
| { |
| "epoch": 2.89, |
| "learning_rate": 5e-05, |
| "loss": 1.6699, |
| "step": 228 |
| }, |
| { |
| "epoch": 2.9, |
| "learning_rate": 5e-05, |
| "loss": 1.6289, |
| "step": 229 |
| }, |
| { |
| "epoch": 2.92, |
| "learning_rate": 5e-05, |
| "loss": 1.7617, |
| "step": 230 |
| }, |
| { |
| "epoch": 2.93, |
| "learning_rate": 5e-05, |
| "loss": 1.6008, |
| "step": 231 |
| }, |
| { |
| "epoch": 2.94, |
| "learning_rate": 5e-05, |
| "loss": 1.6143, |
| "step": 232 |
| }, |
| { |
| "epoch": 2.95, |
| "learning_rate": 5e-05, |
| "loss": 1.6729, |
| "step": 233 |
| }, |
| { |
| "epoch": 2.97, |
| "learning_rate": 5e-05, |
| "loss": 1.553, |
| "step": 234 |
| }, |
| { |
| "epoch": 2.98, |
| "learning_rate": 5e-05, |
| "loss": 1.3635, |
| "step": 235 |
| }, |
| { |
| "epoch": 2.99, |
| "learning_rate": 5e-05, |
| "loss": 1.5159, |
| "step": 236 |
| }, |
| { |
| "epoch": 3.0, |
| "learning_rate": 5e-05, |
| "loss": 1.4727, |
| "step": 237 |
| }, |
| { |
| "epoch": 3.02, |
| "learning_rate": 5e-05, |
| "loss": 1.4629, |
| "step": 238 |
| }, |
| { |
| "epoch": 3.03, |
| "learning_rate": 5e-05, |
| "loss": 1.373, |
| "step": 239 |
| }, |
| { |
| "epoch": 3.04, |
| "learning_rate": 5e-05, |
| "loss": 1.6309, |
| "step": 240 |
| }, |
| { |
| "epoch": 3.06, |
| "learning_rate": 5e-05, |
| "loss": 1.4705, |
| "step": 241 |
| }, |
| { |
| "epoch": 3.07, |
| "learning_rate": 5e-05, |
| "loss": 1.2559, |
| "step": 242 |
| }, |
| { |
| "epoch": 3.08, |
| "learning_rate": 5e-05, |
| "loss": 1.2842, |
| "step": 243 |
| }, |
| { |
| "epoch": 3.09, |
| "learning_rate": 5e-05, |
| "loss": 1.5884, |
| "step": 244 |
| }, |
| { |
| "epoch": 3.11, |
| "learning_rate": 5e-05, |
| "loss": 1.6172, |
| "step": 245 |
| }, |
| { |
| "epoch": 3.12, |
| "learning_rate": 5e-05, |
| "loss": 1.8439, |
| "step": 246 |
| }, |
| { |
| "epoch": 3.13, |
| "learning_rate": 5e-05, |
| "loss": 1.4701, |
| "step": 247 |
| }, |
| { |
| "epoch": 3.14, |
| "learning_rate": 5e-05, |
| "loss": 1.4697, |
| "step": 248 |
| }, |
| { |
| "epoch": 3.16, |
| "learning_rate": 5e-05, |
| "loss": 1.4583, |
| "step": 249 |
| }, |
| { |
| "epoch": 3.17, |
| "learning_rate": 5e-05, |
| "loss": 1.2957, |
| "step": 250 |
| }, |
| { |
| "epoch": 3.18, |
| "learning_rate": 5e-05, |
| "loss": 1.5305, |
| "step": 251 |
| }, |
| { |
| "epoch": 3.19, |
| "learning_rate": 5e-05, |
| "loss": 1.6843, |
| "step": 252 |
| }, |
| { |
| "epoch": 3.21, |
| "learning_rate": 5e-05, |
| "loss": 1.6582, |
| "step": 253 |
| }, |
| { |
| "epoch": 3.22, |
| "learning_rate": 5e-05, |
| "loss": 1.4143, |
| "step": 254 |
| }, |
| { |
| "epoch": 3.23, |
| "learning_rate": 5e-05, |
| "loss": 1.7776, |
| "step": 255 |
| }, |
| { |
| "epoch": 3.25, |
| "learning_rate": 5e-05, |
| "loss": 1.3401, |
| "step": 256 |
| }, |
| { |
| "epoch": 3.26, |
| "learning_rate": 5e-05, |
| "loss": 1.3796, |
| "step": 257 |
| }, |
| { |
| "epoch": 3.27, |
| "learning_rate": 5e-05, |
| "loss": 1.748, |
| "step": 258 |
| }, |
| { |
| "epoch": 3.28, |
| "learning_rate": 5e-05, |
| "loss": 1.6765, |
| "step": 259 |
| }, |
| { |
| "epoch": 3.3, |
| "learning_rate": 5e-05, |
| "loss": 1.2087, |
| "step": 260 |
| }, |
| { |
| "epoch": 3.31, |
| "learning_rate": 5e-05, |
| "loss": 1.2397, |
| "step": 261 |
| }, |
| { |
| "epoch": 3.32, |
| "learning_rate": 5e-05, |
| "loss": 1.5796, |
| "step": 262 |
| }, |
| { |
| "epoch": 3.33, |
| "learning_rate": 5e-05, |
| "loss": 1.5437, |
| "step": 263 |
| }, |
| { |
| "epoch": 3.35, |
| "learning_rate": 5e-05, |
| "loss": 1.2051, |
| "step": 264 |
| }, |
| { |
| "epoch": 3.36, |
| "learning_rate": 5e-05, |
| "loss": 1.4612, |
| "step": 265 |
| }, |
| { |
| "epoch": 3.37, |
| "learning_rate": 5e-05, |
| "loss": 1.3948, |
| "step": 266 |
| }, |
| { |
| "epoch": 3.39, |
| "learning_rate": 5e-05, |
| "loss": 1.3843, |
| "step": 267 |
| }, |
| { |
| "epoch": 3.4, |
| "learning_rate": 5e-05, |
| "loss": 1.2458, |
| "step": 268 |
| }, |
| { |
| "epoch": 3.41, |
| "learning_rate": 5e-05, |
| "loss": 1.4883, |
| "step": 269 |
| }, |
| { |
| "epoch": 3.42, |
| "learning_rate": 5e-05, |
| "loss": 1.3499, |
| "step": 270 |
| }, |
| { |
| "epoch": 3.44, |
| "learning_rate": 5e-05, |
| "loss": 1.5173, |
| "step": 271 |
| }, |
| { |
| "epoch": 3.45, |
| "learning_rate": 5e-05, |
| "loss": 1.198, |
| "step": 272 |
| }, |
| { |
| "epoch": 3.46, |
| "learning_rate": 5e-05, |
| "loss": 1.5085, |
| "step": 273 |
| }, |
| { |
| "epoch": 3.47, |
| "learning_rate": 5e-05, |
| "loss": 1.3892, |
| "step": 274 |
| }, |
| { |
| "epoch": 3.49, |
| "learning_rate": 5e-05, |
| "loss": 1.7917, |
| "step": 275 |
| }, |
| { |
| "epoch": 3.5, |
| "learning_rate": 5e-05, |
| "loss": 1.509, |
| "step": 276 |
| }, |
| { |
| "epoch": 3.51, |
| "learning_rate": 5e-05, |
| "loss": 1.8389, |
| "step": 277 |
| }, |
| { |
| "epoch": 3.52, |
| "learning_rate": 5e-05, |
| "loss": 1.2966, |
| "step": 278 |
| }, |
| { |
| "epoch": 3.54, |
| "learning_rate": 5e-05, |
| "loss": 1.6704, |
| "step": 279 |
| }, |
| { |
| "epoch": 3.55, |
| "learning_rate": 5e-05, |
| "loss": 1.3864, |
| "step": 280 |
| }, |
| { |
| "epoch": 3.56, |
| "learning_rate": 5e-05, |
| "loss": 1.7454, |
| "step": 281 |
| }, |
| { |
| "epoch": 3.58, |
| "learning_rate": 5e-05, |
| "loss": 1.3401, |
| "step": 282 |
| }, |
| { |
| "epoch": 3.59, |
| "learning_rate": 5e-05, |
| "loss": 1.6721, |
| "step": 283 |
| }, |
| { |
| "epoch": 3.6, |
| "learning_rate": 5e-05, |
| "loss": 1.3755, |
| "step": 284 |
| }, |
| { |
| "epoch": 3.61, |
| "learning_rate": 5e-05, |
| "loss": 1.4204, |
| "step": 285 |
| }, |
| { |
| "epoch": 3.63, |
| "learning_rate": 5e-05, |
| "loss": 1.5044, |
| "step": 286 |
| }, |
| { |
| "epoch": 3.64, |
| "learning_rate": 5e-05, |
| "loss": 1.3423, |
| "step": 287 |
| }, |
| { |
| "epoch": 3.65, |
| "learning_rate": 5e-05, |
| "loss": 1.3213, |
| "step": 288 |
| }, |
| { |
| "epoch": 3.66, |
| "learning_rate": 5e-05, |
| "loss": 1.4752, |
| "step": 289 |
| }, |
| { |
| "epoch": 3.68, |
| "learning_rate": 5e-05, |
| "loss": 1.3779, |
| "step": 290 |
| }, |
| { |
| "epoch": 3.69, |
| "learning_rate": 5e-05, |
| "loss": 1.3022, |
| "step": 291 |
| }, |
| { |
| "epoch": 3.7, |
| "learning_rate": 5e-05, |
| "loss": 1.5862, |
| "step": 292 |
| }, |
| { |
| "epoch": 3.71, |
| "learning_rate": 5e-05, |
| "loss": 1.3879, |
| "step": 293 |
| }, |
| { |
| "epoch": 3.73, |
| "learning_rate": 5e-05, |
| "loss": 1.6495, |
| "step": 294 |
| }, |
| { |
| "epoch": 3.74, |
| "learning_rate": 5e-05, |
| "loss": 1.3733, |
| "step": 295 |
| }, |
| { |
| "epoch": 3.75, |
| "learning_rate": 5e-05, |
| "loss": 1.6184, |
| "step": 296 |
| }, |
| { |
| "epoch": 3.77, |
| "learning_rate": 5e-05, |
| "loss": 1.3611, |
| "step": 297 |
| }, |
| { |
| "epoch": 3.78, |
| "learning_rate": 5e-05, |
| "loss": 1.3248, |
| "step": 298 |
| }, |
| { |
| "epoch": 3.79, |
| "learning_rate": 5e-05, |
| "loss": 1.4023, |
| "step": 299 |
| }, |
| { |
| "epoch": 3.8, |
| "learning_rate": 5e-05, |
| "loss": 1.4833, |
| "step": 300 |
| }, |
| { |
| "epoch": 3.82, |
| "learning_rate": 5e-05, |
| "loss": 1.3478, |
| "step": 301 |
| }, |
| { |
| "epoch": 3.83, |
| "learning_rate": 5e-05, |
| "loss": 1.4612, |
| "step": 302 |
| }, |
| { |
| "epoch": 3.84, |
| "learning_rate": 5e-05, |
| "loss": 1.2048, |
| "step": 303 |
| }, |
| { |
| "epoch": 3.85, |
| "learning_rate": 5e-05, |
| "loss": 1.5564, |
| "step": 304 |
| }, |
| { |
| "epoch": 3.87, |
| "learning_rate": 5e-05, |
| "loss": 1.7966, |
| "step": 305 |
| }, |
| { |
| "epoch": 3.88, |
| "learning_rate": 5e-05, |
| "loss": 1.4214, |
| "step": 306 |
| }, |
| { |
| "epoch": 3.89, |
| "learning_rate": 5e-05, |
| "loss": 1.7578, |
| "step": 307 |
| }, |
| { |
| "epoch": 3.9, |
| "learning_rate": 5e-05, |
| "loss": 1.3619, |
| "step": 308 |
| }, |
| { |
| "epoch": 3.92, |
| "learning_rate": 5e-05, |
| "loss": 1.6614, |
| "step": 309 |
| }, |
| { |
| "epoch": 3.93, |
| "learning_rate": 5e-05, |
| "loss": 1.2427, |
| "step": 310 |
| }, |
| { |
| "epoch": 3.94, |
| "learning_rate": 5e-05, |
| "loss": 1.9487, |
| "step": 311 |
| }, |
| { |
| "epoch": 3.96, |
| "learning_rate": 5e-05, |
| "loss": 1.6064, |
| "step": 312 |
| }, |
| { |
| "epoch": 3.97, |
| "learning_rate": 5e-05, |
| "loss": 1.6147, |
| "step": 313 |
| }, |
| { |
| "epoch": 3.98, |
| "learning_rate": 5e-05, |
| "loss": 1.3103, |
| "step": 314 |
| }, |
| { |
| "epoch": 3.99, |
| "learning_rate": 5e-05, |
| "loss": 1.4141, |
| "step": 315 |
| }, |
| { |
| "epoch": 4.01, |
| "learning_rate": 5e-05, |
| "loss": 1.6216, |
| "step": 316 |
| }, |
| { |
| "epoch": 4.02, |
| "learning_rate": 5e-05, |
| "loss": 1.3606, |
| "step": 317 |
| }, |
| { |
| "epoch": 4.03, |
| "learning_rate": 5e-05, |
| "loss": 1.3246, |
| "step": 318 |
| }, |
| { |
| "epoch": 4.04, |
| "learning_rate": 5e-05, |
| "loss": 1.3274, |
| "step": 319 |
| }, |
| { |
| "epoch": 4.06, |
| "learning_rate": 5e-05, |
| "loss": 1.3033, |
| "step": 320 |
| }, |
| { |
| "epoch": 4.07, |
| "learning_rate": 5e-05, |
| "loss": 1.4629, |
| "step": 321 |
| }, |
| { |
| "epoch": 4.08, |
| "learning_rate": 5e-05, |
| "loss": 1.368, |
| "step": 322 |
| }, |
| { |
| "epoch": 4.1, |
| "learning_rate": 5e-05, |
| "loss": 1.2075, |
| "step": 323 |
| }, |
| { |
| "epoch": 4.11, |
| "learning_rate": 5e-05, |
| "loss": 1.4302, |
| "step": 324 |
| }, |
| { |
| "epoch": 4.12, |
| "learning_rate": 5e-05, |
| "loss": 1.1218, |
| "step": 325 |
| }, |
| { |
| "epoch": 4.13, |
| "learning_rate": 5e-05, |
| "loss": 1.6167, |
| "step": 326 |
| }, |
| { |
| "epoch": 4.15, |
| "learning_rate": 5e-05, |
| "loss": 1.355, |
| "step": 327 |
| }, |
| { |
| "epoch": 4.16, |
| "learning_rate": 5e-05, |
| "loss": 1.3278, |
| "step": 328 |
| }, |
| { |
| "epoch": 4.17, |
| "learning_rate": 5e-05, |
| "loss": 1.2373, |
| "step": 329 |
| }, |
| { |
| "epoch": 4.18, |
| "learning_rate": 5e-05, |
| "loss": 1.2812, |
| "step": 330 |
| }, |
| { |
| "epoch": 4.2, |
| "learning_rate": 5e-05, |
| "loss": 1.4514, |
| "step": 331 |
| }, |
| { |
| "epoch": 4.21, |
| "learning_rate": 5e-05, |
| "loss": 1.3529, |
| "step": 332 |
| }, |
| { |
| "epoch": 4.22, |
| "learning_rate": 5e-05, |
| "loss": 1.4648, |
| "step": 333 |
| }, |
| { |
| "epoch": 4.23, |
| "learning_rate": 5e-05, |
| "loss": 0.9886, |
| "step": 334 |
| }, |
| { |
| "epoch": 4.25, |
| "learning_rate": 5e-05, |
| "loss": 1.3233, |
| "step": 335 |
| }, |
| { |
| "epoch": 4.26, |
| "learning_rate": 5e-05, |
| "loss": 1.4673, |
| "step": 336 |
| }, |
| { |
| "epoch": 4.27, |
| "learning_rate": 5e-05, |
| "loss": 1.1567, |
| "step": 337 |
| }, |
| { |
| "epoch": 4.29, |
| "learning_rate": 5e-05, |
| "loss": 0.9458, |
| "step": 338 |
| }, |
| { |
| "epoch": 4.3, |
| "learning_rate": 5e-05, |
| "loss": 1.3669, |
| "step": 339 |
| }, |
| { |
| "epoch": 4.31, |
| "learning_rate": 5e-05, |
| "loss": 1.3571, |
| "step": 340 |
| }, |
| { |
| "epoch": 4.32, |
| "learning_rate": 5e-05, |
| "loss": 1.187, |
| "step": 341 |
| }, |
| { |
| "epoch": 4.34, |
| "learning_rate": 5e-05, |
| "loss": 1.434, |
| "step": 342 |
| }, |
| { |
| "epoch": 4.35, |
| "learning_rate": 5e-05, |
| "loss": 1.1541, |
| "step": 343 |
| }, |
| { |
| "epoch": 4.36, |
| "learning_rate": 5e-05, |
| "loss": 1.2866, |
| "step": 344 |
| }, |
| { |
| "epoch": 4.37, |
| "learning_rate": 5e-05, |
| "loss": 1.2463, |
| "step": 345 |
| }, |
| { |
| "epoch": 4.39, |
| "learning_rate": 5e-05, |
| "loss": 1.0688, |
| "step": 346 |
| }, |
| { |
| "epoch": 4.4, |
| "learning_rate": 5e-05, |
| "loss": 1.4114, |
| "step": 347 |
| }, |
| { |
| "epoch": 4.41, |
| "learning_rate": 5e-05, |
| "loss": 1.3093, |
| "step": 348 |
| }, |
| { |
| "epoch": 4.42, |
| "learning_rate": 5e-05, |
| "loss": 1.3925, |
| "step": 349 |
| }, |
| { |
| "epoch": 4.44, |
| "learning_rate": 5e-05, |
| "loss": 1.3931, |
| "step": 350 |
| }, |
| { |
| "epoch": 4.45, |
| "learning_rate": 5e-05, |
| "loss": 1.332, |
| "step": 351 |
| }, |
| { |
| "epoch": 4.46, |
| "learning_rate": 5e-05, |
| "loss": 1.3589, |
| "step": 352 |
| }, |
| { |
| "epoch": 4.48, |
| "learning_rate": 5e-05, |
| "loss": 1.3401, |
| "step": 353 |
| }, |
| { |
| "epoch": 4.49, |
| "learning_rate": 5e-05, |
| "loss": 1.2417, |
| "step": 354 |
| }, |
| { |
| "epoch": 4.5, |
| "learning_rate": 5e-05, |
| "loss": 1.2009, |
| "step": 355 |
| }, |
| { |
| "epoch": 4.51, |
| "learning_rate": 5e-05, |
| "loss": 1.834, |
| "step": 356 |
| }, |
| { |
| "epoch": 4.53, |
| "learning_rate": 5e-05, |
| "loss": 1.146, |
| "step": 357 |
| }, |
| { |
| "epoch": 4.54, |
| "learning_rate": 5e-05, |
| "loss": 1.4711, |
| "step": 358 |
| }, |
| { |
| "epoch": 4.55, |
| "learning_rate": 5e-05, |
| "loss": 1.3916, |
| "step": 359 |
| }, |
| { |
| "epoch": 4.56, |
| "learning_rate": 5e-05, |
| "loss": 1.0986, |
| "step": 360 |
| }, |
| { |
| "epoch": 4.58, |
| "learning_rate": 5e-05, |
| "loss": 1.0406, |
| "step": 361 |
| }, |
| { |
| "epoch": 4.59, |
| "learning_rate": 5e-05, |
| "loss": 1.1685, |
| "step": 362 |
| }, |
| { |
| "epoch": 4.6, |
| "learning_rate": 5e-05, |
| "loss": 1.4353, |
| "step": 363 |
| }, |
| { |
| "epoch": 4.61, |
| "learning_rate": 5e-05, |
| "loss": 1.2482, |
| "step": 364 |
| }, |
| { |
| "epoch": 4.63, |
| "learning_rate": 5e-05, |
| "loss": 1.1086, |
| "step": 365 |
| }, |
| { |
| "epoch": 4.64, |
| "learning_rate": 5e-05, |
| "loss": 1.3533, |
| "step": 366 |
| }, |
| { |
| "epoch": 4.65, |
| "learning_rate": 5e-05, |
| "loss": 1.2775, |
| "step": 367 |
| }, |
| { |
| "epoch": 4.67, |
| "learning_rate": 5e-05, |
| "loss": 1.251, |
| "step": 368 |
| }, |
| { |
| "epoch": 4.68, |
| "learning_rate": 5e-05, |
| "loss": 1.2703, |
| "step": 369 |
| }, |
| { |
| "epoch": 4.69, |
| "learning_rate": 5e-05, |
| "loss": 1.5557, |
| "step": 370 |
| }, |
| { |
| "epoch": 4.7, |
| "learning_rate": 5e-05, |
| "loss": 1.3955, |
| "step": 371 |
| }, |
| { |
| "epoch": 4.72, |
| "learning_rate": 5e-05, |
| "loss": 1.405, |
| "step": 372 |
| }, |
| { |
| "epoch": 4.73, |
| "learning_rate": 5e-05, |
| "loss": 1.4375, |
| "step": 373 |
| }, |
| { |
| "epoch": 4.74, |
| "learning_rate": 5e-05, |
| "loss": 1.1005, |
| "step": 374 |
| }, |
| { |
| "epoch": 4.75, |
| "learning_rate": 5e-05, |
| "loss": 1.3351, |
| "step": 375 |
| }, |
| { |
| "epoch": 4.77, |
| "learning_rate": 5e-05, |
| "loss": 1.3782, |
| "step": 376 |
| }, |
| { |
| "epoch": 4.78, |
| "learning_rate": 5e-05, |
| "loss": 1.4373, |
| "step": 377 |
| }, |
| { |
| "epoch": 4.79, |
| "learning_rate": 5e-05, |
| "loss": 1.271, |
| "step": 378 |
| }, |
| { |
| "epoch": 4.81, |
| "learning_rate": 5e-05, |
| "loss": 1.3931, |
| "step": 379 |
| }, |
| { |
| "epoch": 4.82, |
| "learning_rate": 5e-05, |
| "loss": 1.418, |
| "step": 380 |
| }, |
| { |
| "epoch": 4.83, |
| "learning_rate": 5e-05, |
| "loss": 1.448, |
| "step": 381 |
| }, |
| { |
| "epoch": 4.84, |
| "learning_rate": 5e-05, |
| "loss": 1.3296, |
| "step": 382 |
| }, |
| { |
| "epoch": 4.86, |
| "learning_rate": 5e-05, |
| "loss": 1.2343, |
| "step": 383 |
| }, |
| { |
| "epoch": 4.87, |
| "learning_rate": 5e-05, |
| "loss": 1.2234, |
| "step": 384 |
| }, |
| { |
| "epoch": 4.88, |
| "learning_rate": 5e-05, |
| "loss": 1.12, |
| "step": 385 |
| }, |
| { |
| "epoch": 4.89, |
| "learning_rate": 5e-05, |
| "loss": 1.2791, |
| "step": 386 |
| }, |
| { |
| "epoch": 4.91, |
| "learning_rate": 5e-05, |
| "loss": 1.5156, |
| "step": 387 |
| }, |
| { |
| "epoch": 4.92, |
| "learning_rate": 5e-05, |
| "loss": 1.4131, |
| "step": 388 |
| }, |
| { |
| "epoch": 4.93, |
| "learning_rate": 5e-05, |
| "loss": 1.0717, |
| "step": 389 |
| }, |
| { |
| "epoch": 4.94, |
| "learning_rate": 5e-05, |
| "loss": 1.3638, |
| "step": 390 |
| }, |
| { |
| "epoch": 4.96, |
| "learning_rate": 5e-05, |
| "loss": 1.2627, |
| "step": 391 |
| }, |
| { |
| "epoch": 4.97, |
| "learning_rate": 5e-05, |
| "loss": 1.2263, |
| "step": 392 |
| }, |
| { |
| "epoch": 4.98, |
| "learning_rate": 5e-05, |
| "loss": 1.1666, |
| "step": 393 |
| }, |
| { |
| "epoch": 5.0, |
| "learning_rate": 5e-05, |
| "loss": 1.1797, |
| "step": 394 |
| }, |
| { |
| "epoch": 5.01, |
| "learning_rate": 5e-05, |
| "loss": 1.0352, |
| "step": 395 |
| }, |
| { |
| "epoch": 5.02, |
| "learning_rate": 5e-05, |
| "loss": 1.1233, |
| "step": 396 |
| }, |
| { |
| "epoch": 5.03, |
| "learning_rate": 5e-05, |
| "loss": 1.0109, |
| "step": 397 |
| }, |
| { |
| "epoch": 5.05, |
| "learning_rate": 5e-05, |
| "loss": 1.2725, |
| "step": 398 |
| }, |
| { |
| "epoch": 5.06, |
| "learning_rate": 5e-05, |
| "loss": 1.0253, |
| "step": 399 |
| }, |
| { |
| "epoch": 5.07, |
| "learning_rate": 5e-05, |
| "loss": 1.5437, |
| "step": 400 |
| }, |
| { |
| "epoch": 5.08, |
| "learning_rate": 5e-05, |
| "loss": 1.0244, |
| "step": 401 |
| }, |
| { |
| "epoch": 5.1, |
| "learning_rate": 5e-05, |
| "loss": 1.1301, |
| "step": 402 |
| }, |
| { |
| "epoch": 5.11, |
| "learning_rate": 5e-05, |
| "loss": 1.1255, |
| "step": 403 |
| }, |
| { |
| "epoch": 5.12, |
| "learning_rate": 5e-05, |
| "loss": 1.1585, |
| "step": 404 |
| }, |
| { |
| "epoch": 5.13, |
| "learning_rate": 5e-05, |
| "loss": 0.9051, |
| "step": 405 |
| }, |
| { |
| "epoch": 5.15, |
| "learning_rate": 5e-05, |
| "loss": 1.0485, |
| "step": 406 |
| }, |
| { |
| "epoch": 5.16, |
| "learning_rate": 5e-05, |
| "loss": 1.0323, |
| "step": 407 |
| }, |
| { |
| "epoch": 5.17, |
| "learning_rate": 5e-05, |
| "loss": 1.0601, |
| "step": 408 |
| }, |
| { |
| "epoch": 5.19, |
| "learning_rate": 5e-05, |
| "loss": 1.0352, |
| "step": 409 |
| }, |
| { |
| "epoch": 5.2, |
| "learning_rate": 5e-05, |
| "loss": 1.1669, |
| "step": 410 |
| }, |
| { |
| "epoch": 5.21, |
| "learning_rate": 5e-05, |
| "loss": 1.1257, |
| "step": 411 |
| }, |
| { |
| "epoch": 5.22, |
| "learning_rate": 5e-05, |
| "loss": 1.2361, |
| "step": 412 |
| }, |
| { |
| "epoch": 5.24, |
| "learning_rate": 5e-05, |
| "loss": 1.0942, |
| "step": 413 |
| }, |
| { |
| "epoch": 5.25, |
| "learning_rate": 5e-05, |
| "loss": 1.0406, |
| "step": 414 |
| }, |
| { |
| "epoch": 5.26, |
| "learning_rate": 5e-05, |
| "loss": 1.1814, |
| "step": 415 |
| }, |
| { |
| "epoch": 5.27, |
| "learning_rate": 5e-05, |
| "loss": 1.2825, |
| "step": 416 |
| }, |
| { |
| "epoch": 5.29, |
| "learning_rate": 5e-05, |
| "loss": 0.9783, |
| "step": 417 |
| }, |
| { |
| "epoch": 5.3, |
| "learning_rate": 5e-05, |
| "loss": 1.059, |
| "step": 418 |
| }, |
| { |
| "epoch": 5.31, |
| "learning_rate": 5e-05, |
| "loss": 1.0726, |
| "step": 419 |
| }, |
| { |
| "epoch": 5.32, |
| "learning_rate": 5e-05, |
| "loss": 1.3308, |
| "step": 420 |
| }, |
| { |
| "epoch": 5.34, |
| "learning_rate": 5e-05, |
| "loss": 0.959, |
| "step": 421 |
| }, |
| { |
| "epoch": 5.35, |
| "learning_rate": 5e-05, |
| "loss": 0.8518, |
| "step": 422 |
| }, |
| { |
| "epoch": 5.36, |
| "learning_rate": 5e-05, |
| "loss": 1.0836, |
| "step": 423 |
| }, |
| { |
| "epoch": 5.38, |
| "learning_rate": 5e-05, |
| "loss": 0.8994, |
| "step": 424 |
| }, |
| { |
| "epoch": 5.39, |
| "learning_rate": 5e-05, |
| "loss": 0.9929, |
| "step": 425 |
| }, |
| { |
| "epoch": 5.4, |
| "learning_rate": 5e-05, |
| "loss": 1.0156, |
| "step": 426 |
| }, |
| { |
| "epoch": 5.41, |
| "learning_rate": 5e-05, |
| "loss": 0.9542, |
| "step": 427 |
| }, |
| { |
| "epoch": 5.43, |
| "learning_rate": 5e-05, |
| "loss": 1.0391, |
| "step": 428 |
| }, |
| { |
| "epoch": 5.44, |
| "learning_rate": 5e-05, |
| "loss": 1.3759, |
| "step": 429 |
| }, |
| { |
| "epoch": 5.45, |
| "learning_rate": 5e-05, |
| "loss": 1.0764, |
| "step": 430 |
| }, |
| { |
| "epoch": 5.46, |
| "learning_rate": 5e-05, |
| "loss": 1.415, |
| "step": 431 |
| }, |
| { |
| "epoch": 5.48, |
| "learning_rate": 5e-05, |
| "loss": 0.8115, |
| "step": 432 |
| }, |
| { |
| "epoch": 5.49, |
| "learning_rate": 5e-05, |
| "loss": 1.2021, |
| "step": 433 |
| }, |
| { |
| "epoch": 5.5, |
| "learning_rate": 5e-05, |
| "loss": 1.1346, |
| "step": 434 |
| }, |
| { |
| "epoch": 5.52, |
| "learning_rate": 5e-05, |
| "loss": 1.1128, |
| "step": 435 |
| }, |
| { |
| "epoch": 5.53, |
| "learning_rate": 5e-05, |
| "loss": 1.1196, |
| "step": 436 |
| }, |
| { |
| "epoch": 5.54, |
| "learning_rate": 5e-05, |
| "loss": 1.1588, |
| "step": 437 |
| }, |
| { |
| "epoch": 5.55, |
| "learning_rate": 5e-05, |
| "loss": 1.29, |
| "step": 438 |
| }, |
| { |
| "epoch": 5.57, |
| "learning_rate": 5e-05, |
| "loss": 1.0735, |
| "step": 439 |
| }, |
| { |
| "epoch": 5.58, |
| "learning_rate": 5e-05, |
| "loss": 0.9816, |
| "step": 440 |
| }, |
| { |
| "epoch": 5.59, |
| "learning_rate": 5e-05, |
| "loss": 0.9113, |
| "step": 441 |
| }, |
| { |
| "epoch": 5.6, |
| "learning_rate": 5e-05, |
| "loss": 1.2759, |
| "step": 442 |
| }, |
| { |
| "epoch": 5.62, |
| "learning_rate": 5e-05, |
| "loss": 1.2019, |
| "step": 443 |
| }, |
| { |
| "epoch": 5.63, |
| "learning_rate": 5e-05, |
| "loss": 1.3916, |
| "step": 444 |
| }, |
| { |
| "epoch": 5.64, |
| "learning_rate": 5e-05, |
| "loss": 1.1307, |
| "step": 445 |
| }, |
| { |
| "epoch": 5.65, |
| "learning_rate": 5e-05, |
| "loss": 0.8333, |
| "step": 446 |
| }, |
| { |
| "epoch": 5.67, |
| "learning_rate": 5e-05, |
| "loss": 1.2332, |
| "step": 447 |
| }, |
| { |
| "epoch": 5.68, |
| "learning_rate": 5e-05, |
| "loss": 1.3669, |
| "step": 448 |
| }, |
| { |
| "epoch": 5.69, |
| "learning_rate": 5e-05, |
| "loss": 1.0098, |
| "step": 449 |
| }, |
| { |
| "epoch": 5.71, |
| "learning_rate": 5e-05, |
| "loss": 1.1978, |
| "step": 450 |
| }, |
| { |
| "epoch": 5.72, |
| "learning_rate": 5e-05, |
| "loss": 1.1589, |
| "step": 451 |
| }, |
| { |
| "epoch": 5.73, |
| "learning_rate": 5e-05, |
| "loss": 1.0649, |
| "step": 452 |
| }, |
| { |
| "epoch": 5.74, |
| "learning_rate": 5e-05, |
| "loss": 1.4026, |
| "step": 453 |
| }, |
| { |
| "epoch": 5.76, |
| "learning_rate": 5e-05, |
| "loss": 1.259, |
| "step": 454 |
| }, |
| { |
| "epoch": 5.77, |
| "learning_rate": 5e-05, |
| "loss": 1.0715, |
| "step": 455 |
| }, |
| { |
| "epoch": 5.78, |
| "learning_rate": 5e-05, |
| "loss": 1.3926, |
| "step": 456 |
| }, |
| { |
| "epoch": 5.79, |
| "learning_rate": 5e-05, |
| "loss": 0.989, |
| "step": 457 |
| }, |
| { |
| "epoch": 5.81, |
| "learning_rate": 5e-05, |
| "loss": 1.3079, |
| "step": 458 |
| }, |
| { |
| "epoch": 5.82, |
| "learning_rate": 5e-05, |
| "loss": 0.9657, |
| "step": 459 |
| }, |
| { |
| "epoch": 5.83, |
| "learning_rate": 5e-05, |
| "loss": 1.0565, |
| "step": 460 |
| }, |
| { |
| "epoch": 5.84, |
| "learning_rate": 5e-05, |
| "loss": 1.2346, |
| "step": 461 |
| }, |
| { |
| "epoch": 5.86, |
| "learning_rate": 5e-05, |
| "loss": 1.0129, |
| "step": 462 |
| }, |
| { |
| "epoch": 5.87, |
| "learning_rate": 5e-05, |
| "loss": 1.3492, |
| "step": 463 |
| }, |
| { |
| "epoch": 5.88, |
| "learning_rate": 5e-05, |
| "loss": 0.9885, |
| "step": 464 |
| }, |
| { |
| "epoch": 5.9, |
| "learning_rate": 5e-05, |
| "loss": 0.9617, |
| "step": 465 |
| }, |
| { |
| "epoch": 5.91, |
| "learning_rate": 5e-05, |
| "loss": 1.1428, |
| "step": 466 |
| }, |
| { |
| "epoch": 5.92, |
| "learning_rate": 5e-05, |
| "loss": 1.0911, |
| "step": 467 |
| }, |
| { |
| "epoch": 5.93, |
| "learning_rate": 5e-05, |
| "loss": 0.98, |
| "step": 468 |
| }, |
| { |
| "epoch": 5.95, |
| "learning_rate": 5e-05, |
| "loss": 1.0477, |
| "step": 469 |
| }, |
| { |
| "epoch": 5.96, |
| "learning_rate": 5e-05, |
| "loss": 1.2196, |
| "step": 470 |
| }, |
| { |
| "epoch": 5.97, |
| "learning_rate": 5e-05, |
| "loss": 1.4609, |
| "step": 471 |
| }, |
| { |
| "epoch": 5.98, |
| "learning_rate": 5e-05, |
| "loss": 1.105, |
| "step": 472 |
| }, |
| { |
| "epoch": 6.0, |
| "learning_rate": 5e-05, |
| "loss": 1.0371, |
| "step": 473 |
| }, |
| { |
| "epoch": 6.01, |
| "learning_rate": 5e-05, |
| "loss": 1.2643, |
| "step": 474 |
| }, |
| { |
| "epoch": 6.02, |
| "learning_rate": 5e-05, |
| "loss": 1.27, |
| "step": 475 |
| }, |
| { |
| "epoch": 6.03, |
| "learning_rate": 5e-05, |
| "loss": 0.9491, |
| "step": 476 |
| }, |
| { |
| "epoch": 6.05, |
| "learning_rate": 5e-05, |
| "loss": 0.9374, |
| "step": 477 |
| }, |
| { |
| "epoch": 6.06, |
| "learning_rate": 5e-05, |
| "loss": 0.9541, |
| "step": 478 |
| }, |
| { |
| "epoch": 6.07, |
| "learning_rate": 5e-05, |
| "loss": 0.8734, |
| "step": 479 |
| }, |
| { |
| "epoch": 6.09, |
| "learning_rate": 5e-05, |
| "loss": 1.0566, |
| "step": 480 |
| }, |
| { |
| "epoch": 6.1, |
| "learning_rate": 5e-05, |
| "loss": 0.9071, |
| "step": 481 |
| }, |
| { |
| "epoch": 6.11, |
| "learning_rate": 5e-05, |
| "loss": 0.8359, |
| "step": 482 |
| }, |
| { |
| "epoch": 6.12, |
| "learning_rate": 5e-05, |
| "loss": 0.9695, |
| "step": 483 |
| }, |
| { |
| "epoch": 6.14, |
| "learning_rate": 5e-05, |
| "loss": 0.7808, |
| "step": 484 |
| }, |
| { |
| "epoch": 6.15, |
| "learning_rate": 5e-05, |
| "loss": 0.7584, |
| "step": 485 |
| }, |
| { |
| "epoch": 6.16, |
| "learning_rate": 5e-05, |
| "loss": 1.0564, |
| "step": 486 |
| }, |
| { |
| "epoch": 6.17, |
| "learning_rate": 5e-05, |
| "loss": 0.8324, |
| "step": 487 |
| }, |
| { |
| "epoch": 6.19, |
| "learning_rate": 5e-05, |
| "loss": 0.9714, |
| "step": 488 |
| }, |
| { |
| "epoch": 6.2, |
| "learning_rate": 5e-05, |
| "loss": 0.9785, |
| "step": 489 |
| }, |
| { |
| "epoch": 6.21, |
| "learning_rate": 5e-05, |
| "loss": 0.95, |
| "step": 490 |
| }, |
| { |
| "epoch": 6.23, |
| "learning_rate": 5e-05, |
| "loss": 0.806, |
| "step": 491 |
| }, |
| { |
| "epoch": 6.24, |
| "learning_rate": 5e-05, |
| "loss": 0.8848, |
| "step": 492 |
| }, |
| { |
| "epoch": 6.25, |
| "learning_rate": 5e-05, |
| "loss": 0.824, |
| "step": 493 |
| }, |
| { |
| "epoch": 6.26, |
| "learning_rate": 5e-05, |
| "loss": 0.969, |
| "step": 494 |
| }, |
| { |
| "epoch": 6.28, |
| "learning_rate": 5e-05, |
| "loss": 0.8861, |
| "step": 495 |
| }, |
| { |
| "epoch": 6.29, |
| "learning_rate": 5e-05, |
| "loss": 0.6434, |
| "step": 496 |
| }, |
| { |
| "epoch": 6.3, |
| "learning_rate": 5e-05, |
| "loss": 1.0431, |
| "step": 497 |
| }, |
| { |
| "epoch": 6.31, |
| "learning_rate": 5e-05, |
| "loss": 0.9127, |
| "step": 498 |
| }, |
| { |
| "epoch": 6.33, |
| "learning_rate": 5e-05, |
| "loss": 0.9074, |
| "step": 499 |
| }, |
| { |
| "epoch": 6.34, |
| "learning_rate": 5e-05, |
| "loss": 1.0694, |
| "step": 500 |
| }, |
| { |
| "epoch": 6.35, |
| "learning_rate": 5e-05, |
| "loss": 0.9895, |
| "step": 501 |
| }, |
| { |
| "epoch": 6.36, |
| "learning_rate": 5e-05, |
| "loss": 0.9509, |
| "step": 502 |
| }, |
| { |
| "epoch": 6.38, |
| "learning_rate": 5e-05, |
| "loss": 0.9293, |
| "step": 503 |
| }, |
| { |
| "epoch": 6.39, |
| "learning_rate": 5e-05, |
| "loss": 1.0732, |
| "step": 504 |
| }, |
| { |
| "epoch": 6.4, |
| "learning_rate": 5e-05, |
| "loss": 1.2473, |
| "step": 505 |
| }, |
| { |
| "epoch": 6.42, |
| "learning_rate": 5e-05, |
| "loss": 1.0963, |
| "step": 506 |
| }, |
| { |
| "epoch": 6.43, |
| "learning_rate": 5e-05, |
| "loss": 1.0005, |
| "step": 507 |
| }, |
| { |
| "epoch": 6.44, |
| "learning_rate": 5e-05, |
| "loss": 0.8832, |
| "step": 508 |
| }, |
| { |
| "epoch": 6.45, |
| "learning_rate": 5e-05, |
| "loss": 0.983, |
| "step": 509 |
| }, |
| { |
| "epoch": 6.47, |
| "learning_rate": 5e-05, |
| "loss": 1.0771, |
| "step": 510 |
| }, |
| { |
| "epoch": 6.48, |
| "learning_rate": 5e-05, |
| "loss": 0.851, |
| "step": 511 |
| }, |
| { |
| "epoch": 6.49, |
| "learning_rate": 5e-05, |
| "loss": 1.0426, |
| "step": 512 |
| }, |
| { |
| "epoch": 6.5, |
| "learning_rate": 5e-05, |
| "loss": 0.7806, |
| "step": 513 |
| }, |
| { |
| "epoch": 6.52, |
| "learning_rate": 5e-05, |
| "loss": 0.8573, |
| "step": 514 |
| }, |
| { |
| "epoch": 6.53, |
| "learning_rate": 5e-05, |
| "loss": 0.9017, |
| "step": 515 |
| }, |
| { |
| "epoch": 6.54, |
| "learning_rate": 5e-05, |
| "loss": 1.0674, |
| "step": 516 |
| }, |
| { |
| "epoch": 6.55, |
| "learning_rate": 5e-05, |
| "loss": 0.8411, |
| "step": 517 |
| }, |
| { |
| "epoch": 6.57, |
| "learning_rate": 5e-05, |
| "loss": 0.7642, |
| "step": 518 |
| }, |
| { |
| "epoch": 6.58, |
| "learning_rate": 5e-05, |
| "loss": 0.7722, |
| "step": 519 |
| }, |
| { |
| "epoch": 6.59, |
| "learning_rate": 5e-05, |
| "loss": 1.0848, |
| "step": 520 |
| }, |
| { |
| "epoch": 6.61, |
| "learning_rate": 5e-05, |
| "loss": 1.1509, |
| "step": 521 |
| }, |
| { |
| "epoch": 6.62, |
| "learning_rate": 5e-05, |
| "loss": 0.9004, |
| "step": 522 |
| }, |
| { |
| "epoch": 6.63, |
| "learning_rate": 5e-05, |
| "loss": 1.0056, |
| "step": 523 |
| }, |
| { |
| "epoch": 6.64, |
| "learning_rate": 5e-05, |
| "loss": 0.7882, |
| "step": 524 |
| }, |
| { |
| "epoch": 6.66, |
| "learning_rate": 5e-05, |
| "loss": 1.0292, |
| "step": 525 |
| }, |
| { |
| "epoch": 6.67, |
| "learning_rate": 5e-05, |
| "loss": 0.8975, |
| "step": 526 |
| }, |
| { |
| "epoch": 6.68, |
| "learning_rate": 5e-05, |
| "loss": 0.7839, |
| "step": 527 |
| }, |
| { |
| "epoch": 6.69, |
| "learning_rate": 5e-05, |
| "loss": 0.6125, |
| "step": 528 |
| }, |
| { |
| "epoch": 6.71, |
| "learning_rate": 5e-05, |
| "loss": 0.8376, |
| "step": 529 |
| }, |
| { |
| "epoch": 6.72, |
| "learning_rate": 5e-05, |
| "loss": 1.2618, |
| "step": 530 |
| }, |
| { |
| "epoch": 6.73, |
| "learning_rate": 5e-05, |
| "loss": 0.9618, |
| "step": 531 |
| }, |
| { |
| "epoch": 6.74, |
| "learning_rate": 5e-05, |
| "loss": 0.8689, |
| "step": 532 |
| }, |
| { |
| "epoch": 6.76, |
| "learning_rate": 5e-05, |
| "loss": 0.9085, |
| "step": 533 |
| }, |
| { |
| "epoch": 6.77, |
| "learning_rate": 5e-05, |
| "loss": 1.0693, |
| "step": 534 |
| }, |
| { |
| "epoch": 6.78, |
| "learning_rate": 5e-05, |
| "loss": 0.8101, |
| "step": 535 |
| }, |
| { |
| "epoch": 6.8, |
| "learning_rate": 5e-05, |
| "loss": 0.723, |
| "step": 536 |
| }, |
| { |
| "epoch": 6.81, |
| "learning_rate": 5e-05, |
| "loss": 1.139, |
| "step": 537 |
| }, |
| { |
| "epoch": 6.82, |
| "learning_rate": 5e-05, |
| "loss": 0.8303, |
| "step": 538 |
| }, |
| { |
| "epoch": 6.83, |
| "learning_rate": 5e-05, |
| "loss": 0.8186, |
| "step": 539 |
| }, |
| { |
| "epoch": 6.85, |
| "learning_rate": 5e-05, |
| "loss": 0.9393, |
| "step": 540 |
| }, |
| { |
| "epoch": 6.86, |
| "learning_rate": 5e-05, |
| "loss": 0.876, |
| "step": 541 |
| }, |
| { |
| "epoch": 6.87, |
| "learning_rate": 5e-05, |
| "loss": 0.898, |
| "step": 542 |
| }, |
| { |
| "epoch": 6.88, |
| "learning_rate": 5e-05, |
| "loss": 0.7925, |
| "step": 543 |
| }, |
| { |
| "epoch": 6.9, |
| "learning_rate": 5e-05, |
| "loss": 0.832, |
| "step": 544 |
| }, |
| { |
| "epoch": 6.91, |
| "learning_rate": 5e-05, |
| "loss": 0.9486, |
| "step": 545 |
| }, |
| { |
| "epoch": 6.92, |
| "learning_rate": 5e-05, |
| "loss": 0.895, |
| "step": 546 |
| }, |
| { |
| "epoch": 6.94, |
| "learning_rate": 5e-05, |
| "loss": 0.905, |
| "step": 547 |
| }, |
| { |
| "epoch": 6.95, |
| "learning_rate": 5e-05, |
| "loss": 0.875, |
| "step": 548 |
| }, |
| { |
| "epoch": 6.96, |
| "learning_rate": 5e-05, |
| "loss": 0.9828, |
| "step": 549 |
| }, |
| { |
| "epoch": 6.97, |
| "learning_rate": 5e-05, |
| "loss": 0.8679, |
| "step": 550 |
| }, |
| { |
| "epoch": 6.99, |
| "learning_rate": 5e-05, |
| "loss": 0.9418, |
| "step": 551 |
| }, |
| { |
| "epoch": 7.0, |
| "learning_rate": 5e-05, |
| "loss": 0.9807, |
| "step": 552 |
| }, |
| { |
| "epoch": 7.01, |
| "learning_rate": 5e-05, |
| "loss": 0.8455, |
| "step": 553 |
| }, |
| { |
| "epoch": 7.02, |
| "learning_rate": 5e-05, |
| "loss": 1.0291, |
| "step": 554 |
| }, |
| { |
| "epoch": 7.04, |
| "learning_rate": 5e-05, |
| "loss": 0.7048, |
| "step": 555 |
| }, |
| { |
| "epoch": 7.05, |
| "learning_rate": 5e-05, |
| "loss": 0.7972, |
| "step": 556 |
| }, |
| { |
| "epoch": 7.06, |
| "learning_rate": 5e-05, |
| "loss": 0.6609, |
| "step": 557 |
| }, |
| { |
| "epoch": 7.07, |
| "learning_rate": 5e-05, |
| "loss": 0.5593, |
| "step": 558 |
| }, |
| { |
| "epoch": 7.09, |
| "learning_rate": 5e-05, |
| "loss": 0.757, |
| "step": 559 |
| }, |
| { |
| "epoch": 7.1, |
| "learning_rate": 5e-05, |
| "loss": 0.9493, |
| "step": 560 |
| }, |
| { |
| "epoch": 7.11, |
| "learning_rate": 5e-05, |
| "loss": 0.7764, |
| "step": 561 |
| }, |
| { |
| "epoch": 7.13, |
| "learning_rate": 5e-05, |
| "loss": 0.7507, |
| "step": 562 |
| }, |
| { |
| "epoch": 7.14, |
| "learning_rate": 5e-05, |
| "loss": 0.6658, |
| "step": 563 |
| }, |
| { |
| "epoch": 7.15, |
| "learning_rate": 5e-05, |
| "loss": 0.8028, |
| "step": 564 |
| }, |
| { |
| "epoch": 7.16, |
| "learning_rate": 5e-05, |
| "loss": 0.8905, |
| "step": 565 |
| }, |
| { |
| "epoch": 7.18, |
| "learning_rate": 5e-05, |
| "loss": 0.8983, |
| "step": 566 |
| }, |
| { |
| "epoch": 7.19, |
| "learning_rate": 5e-05, |
| "loss": 0.8674, |
| "step": 567 |
| }, |
| { |
| "epoch": 7.2, |
| "learning_rate": 5e-05, |
| "loss": 0.8704, |
| "step": 568 |
| }, |
| { |
| "epoch": 7.21, |
| "learning_rate": 5e-05, |
| "loss": 0.6869, |
| "step": 569 |
| }, |
| { |
| "epoch": 7.23, |
| "learning_rate": 5e-05, |
| "loss": 0.6305, |
| "step": 570 |
| }, |
| { |
| "epoch": 7.24, |
| "learning_rate": 5e-05, |
| "loss": 0.5664, |
| "step": 571 |
| }, |
| { |
| "epoch": 7.25, |
| "learning_rate": 5e-05, |
| "loss": 1.0355, |
| "step": 572 |
| }, |
| { |
| "epoch": 7.26, |
| "learning_rate": 5e-05, |
| "loss": 0.6752, |
| "step": 573 |
| }, |
| { |
| "epoch": 7.28, |
| "learning_rate": 5e-05, |
| "loss": 0.8231, |
| "step": 574 |
| }, |
| { |
| "epoch": 7.29, |
| "learning_rate": 5e-05, |
| "loss": 0.7209, |
| "step": 575 |
| }, |
| { |
| "epoch": 7.3, |
| "learning_rate": 5e-05, |
| "loss": 0.7499, |
| "step": 576 |
| }, |
| { |
| "epoch": 7.32, |
| "learning_rate": 5e-05, |
| "loss": 0.649, |
| "step": 577 |
| }, |
| { |
| "epoch": 7.33, |
| "learning_rate": 5e-05, |
| "loss": 0.6372, |
| "step": 578 |
| }, |
| { |
| "epoch": 7.34, |
| "learning_rate": 5e-05, |
| "loss": 0.8981, |
| "step": 579 |
| }, |
| { |
| "epoch": 7.35, |
| "learning_rate": 5e-05, |
| "loss": 0.5923, |
| "step": 580 |
| }, |
| { |
| "epoch": 7.37, |
| "learning_rate": 5e-05, |
| "loss": 0.7766, |
| "step": 581 |
| }, |
| { |
| "epoch": 7.38, |
| "learning_rate": 5e-05, |
| "loss": 0.8104, |
| "step": 582 |
| }, |
| { |
| "epoch": 7.39, |
| "learning_rate": 5e-05, |
| "loss": 0.5765, |
| "step": 583 |
| }, |
| { |
| "epoch": 7.4, |
| "learning_rate": 5e-05, |
| "loss": 0.6414, |
| "step": 584 |
| }, |
| { |
| "epoch": 7.42, |
| "learning_rate": 5e-05, |
| "loss": 0.7383, |
| "step": 585 |
| }, |
| { |
| "epoch": 7.43, |
| "learning_rate": 5e-05, |
| "loss": 0.7659, |
| "step": 586 |
| }, |
| { |
| "epoch": 7.44, |
| "learning_rate": 5e-05, |
| "loss": 0.9116, |
| "step": 587 |
| }, |
| { |
| "epoch": 7.45, |
| "learning_rate": 5e-05, |
| "loss": 0.8367, |
| "step": 588 |
| }, |
| { |
| "epoch": 7.47, |
| "learning_rate": 5e-05, |
| "loss": 0.692, |
| "step": 589 |
| }, |
| { |
| "epoch": 7.48, |
| "learning_rate": 5e-05, |
| "loss": 0.5653, |
| "step": 590 |
| }, |
| { |
| "epoch": 7.49, |
| "learning_rate": 5e-05, |
| "loss": 0.6509, |
| "step": 591 |
| }, |
| { |
| "epoch": 7.51, |
| "learning_rate": 5e-05, |
| "loss": 0.8573, |
| "step": 592 |
| }, |
| { |
| "epoch": 7.52, |
| "learning_rate": 5e-05, |
| "loss": 0.5039, |
| "step": 593 |
| }, |
| { |
| "epoch": 7.53, |
| "learning_rate": 5e-05, |
| "loss": 0.5755, |
| "step": 594 |
| }, |
| { |
| "epoch": 7.54, |
| "learning_rate": 5e-05, |
| "loss": 0.6116, |
| "step": 595 |
| }, |
| { |
| "epoch": 7.56, |
| "learning_rate": 5e-05, |
| "loss": 0.82, |
| "step": 596 |
| }, |
| { |
| "epoch": 7.57, |
| "learning_rate": 5e-05, |
| "loss": 0.7148, |
| "step": 597 |
| }, |
| { |
| "epoch": 7.58, |
| "learning_rate": 5e-05, |
| "loss": 0.9592, |
| "step": 598 |
| }, |
| { |
| "epoch": 7.59, |
| "learning_rate": 5e-05, |
| "loss": 0.6801, |
| "step": 599 |
| }, |
| { |
| "epoch": 7.61, |
| "learning_rate": 5e-05, |
| "loss": 0.762, |
| "step": 600 |
| }, |
| { |
| "epoch": 7.62, |
| "learning_rate": 5e-05, |
| "loss": 0.7289, |
| "step": 601 |
| }, |
| { |
| "epoch": 7.63, |
| "learning_rate": 5e-05, |
| "loss": 0.808, |
| "step": 602 |
| }, |
| { |
| "epoch": 7.65, |
| "learning_rate": 5e-05, |
| "loss": 0.5103, |
| "step": 603 |
| }, |
| { |
| "epoch": 7.66, |
| "learning_rate": 5e-05, |
| "loss": 0.7641, |
| "step": 604 |
| }, |
| { |
| "epoch": 7.67, |
| "learning_rate": 5e-05, |
| "loss": 0.8445, |
| "step": 605 |
| }, |
| { |
| "epoch": 7.68, |
| "learning_rate": 5e-05, |
| "loss": 0.7042, |
| "step": 606 |
| }, |
| { |
| "epoch": 7.7, |
| "learning_rate": 5e-05, |
| "loss": 0.6248, |
| "step": 607 |
| }, |
| { |
| "epoch": 7.71, |
| "learning_rate": 5e-05, |
| "loss": 0.6974, |
| "step": 608 |
| }, |
| { |
| "epoch": 7.72, |
| "learning_rate": 5e-05, |
| "loss": 0.7323, |
| "step": 609 |
| }, |
| { |
| "epoch": 7.73, |
| "learning_rate": 5e-05, |
| "loss": 0.6591, |
| "step": 610 |
| }, |
| { |
| "epoch": 7.75, |
| "learning_rate": 5e-05, |
| "loss": 0.8257, |
| "step": 611 |
| }, |
| { |
| "epoch": 7.76, |
| "learning_rate": 5e-05, |
| "loss": 0.7774, |
| "step": 612 |
| }, |
| { |
| "epoch": 7.77, |
| "learning_rate": 5e-05, |
| "loss": 0.9021, |
| "step": 613 |
| }, |
| { |
| "epoch": 7.78, |
| "learning_rate": 5e-05, |
| "loss": 0.6187, |
| "step": 614 |
| }, |
| { |
| "epoch": 7.8, |
| "learning_rate": 5e-05, |
| "loss": 0.5164, |
| "step": 615 |
| }, |
| { |
| "epoch": 7.81, |
| "learning_rate": 5e-05, |
| "loss": 0.605, |
| "step": 616 |
| }, |
| { |
| "epoch": 7.82, |
| "learning_rate": 5e-05, |
| "loss": 0.8193, |
| "step": 617 |
| }, |
| { |
| "epoch": 7.84, |
| "learning_rate": 5e-05, |
| "loss": 0.7986, |
| "step": 618 |
| }, |
| { |
| "epoch": 7.85, |
| "learning_rate": 5e-05, |
| "loss": 0.8107, |
| "step": 619 |
| }, |
| { |
| "epoch": 7.86, |
| "learning_rate": 5e-05, |
| "loss": 0.6984, |
| "step": 620 |
| }, |
| { |
| "epoch": 7.87, |
| "learning_rate": 5e-05, |
| "loss": 0.8126, |
| "step": 621 |
| }, |
| { |
| "epoch": 7.89, |
| "learning_rate": 5e-05, |
| "loss": 0.6204, |
| "step": 622 |
| }, |
| { |
| "epoch": 7.9, |
| "learning_rate": 5e-05, |
| "loss": 0.6644, |
| "step": 623 |
| }, |
| { |
| "epoch": 7.91, |
| "learning_rate": 5e-05, |
| "loss": 0.9012, |
| "step": 624 |
| }, |
| { |
| "epoch": 7.92, |
| "learning_rate": 5e-05, |
| "loss": 0.7433, |
| "step": 625 |
| }, |
| { |
| "epoch": 7.94, |
| "learning_rate": 5e-05, |
| "loss": 0.5367, |
| "step": 626 |
| }, |
| { |
| "epoch": 7.95, |
| "learning_rate": 5e-05, |
| "loss": 0.7661, |
| "step": 627 |
| }, |
| { |
| "epoch": 7.96, |
| "learning_rate": 5e-05, |
| "loss": 0.6797, |
| "step": 628 |
| }, |
| { |
| "epoch": 7.97, |
| "learning_rate": 5e-05, |
| "loss": 0.8271, |
| "step": 629 |
| }, |
| { |
| "epoch": 7.99, |
| "learning_rate": 5e-05, |
| "loss": 0.6868, |
| "step": 630 |
| }, |
| { |
| "epoch": 8.0, |
| "learning_rate": 5e-05, |
| "loss": 0.7044, |
| "step": 631 |
| }, |
| { |
| "epoch": 8.01, |
| "learning_rate": 5e-05, |
| "loss": 0.6336, |
| "step": 632 |
| }, |
| { |
| "epoch": 8.03, |
| "learning_rate": 5e-05, |
| "loss": 0.7318, |
| "step": 633 |
| }, |
| { |
| "epoch": 8.04, |
| "learning_rate": 5e-05, |
| "loss": 0.6107, |
| "step": 634 |
| }, |
| { |
| "epoch": 8.05, |
| "learning_rate": 5e-05, |
| "loss": 0.6628, |
| "step": 635 |
| }, |
| { |
| "epoch": 8.06, |
| "learning_rate": 5e-05, |
| "loss": 0.7153, |
| "step": 636 |
| }, |
| { |
| "epoch": 8.08, |
| "learning_rate": 5e-05, |
| "loss": 0.6462, |
| "step": 637 |
| }, |
| { |
| "epoch": 8.09, |
| "learning_rate": 5e-05, |
| "loss": 0.6727, |
| "step": 638 |
| }, |
| { |
| "epoch": 8.1, |
| "learning_rate": 5e-05, |
| "loss": 0.5637, |
| "step": 639 |
| }, |
| { |
| "epoch": 8.11, |
| "learning_rate": 5e-05, |
| "loss": 0.5032, |
| "step": 640 |
| }, |
| { |
| "epoch": 8.13, |
| "learning_rate": 5e-05, |
| "loss": 0.8464, |
| "step": 641 |
| }, |
| { |
| "epoch": 8.14, |
| "learning_rate": 5e-05, |
| "loss": 0.6292, |
| "step": 642 |
| }, |
| { |
| "epoch": 8.15, |
| "learning_rate": 5e-05, |
| "loss": 0.5012, |
| "step": 643 |
| }, |
| { |
| "epoch": 8.16, |
| "learning_rate": 5e-05, |
| "loss": 0.7806, |
| "step": 644 |
| }, |
| { |
| "epoch": 8.18, |
| "learning_rate": 5e-05, |
| "loss": 0.7304, |
| "step": 645 |
| }, |
| { |
| "epoch": 8.19, |
| "learning_rate": 5e-05, |
| "loss": 0.6582, |
| "step": 646 |
| }, |
| { |
| "epoch": 8.2, |
| "learning_rate": 5e-05, |
| "loss": 0.7004, |
| "step": 647 |
| }, |
| { |
| "epoch": 8.22, |
| "learning_rate": 5e-05, |
| "loss": 0.6578, |
| "step": 648 |
| }, |
| { |
| "epoch": 8.23, |
| "learning_rate": 5e-05, |
| "loss": 0.7305, |
| "step": 649 |
| }, |
| { |
| "epoch": 8.24, |
| "learning_rate": 5e-05, |
| "loss": 0.5677, |
| "step": 650 |
| }, |
| { |
| "epoch": 8.25, |
| "learning_rate": 5e-05, |
| "loss": 0.7588, |
| "step": 651 |
| }, |
| { |
| "epoch": 8.27, |
| "learning_rate": 5e-05, |
| "loss": 0.6982, |
| "step": 652 |
| }, |
| { |
| "epoch": 8.28, |
| "learning_rate": 5e-05, |
| "loss": 0.5543, |
| "step": 653 |
| }, |
| { |
| "epoch": 8.29, |
| "learning_rate": 5e-05, |
| "loss": 0.5165, |
| "step": 654 |
| }, |
| { |
| "epoch": 8.3, |
| "learning_rate": 5e-05, |
| "loss": 0.4101, |
| "step": 655 |
| }, |
| { |
| "epoch": 8.32, |
| "learning_rate": 5e-05, |
| "loss": 0.6526, |
| "step": 656 |
| }, |
| { |
| "epoch": 8.33, |
| "learning_rate": 5e-05, |
| "loss": 0.4875, |
| "step": 657 |
| }, |
| { |
| "epoch": 8.34, |
| "learning_rate": 5e-05, |
| "loss": 0.6135, |
| "step": 658 |
| }, |
| { |
| "epoch": 8.35, |
| "learning_rate": 5e-05, |
| "loss": 0.5615, |
| "step": 659 |
| }, |
| { |
| "epoch": 8.37, |
| "learning_rate": 5e-05, |
| "loss": 0.8262, |
| "step": 660 |
| }, |
| { |
| "epoch": 8.38, |
| "learning_rate": 5e-05, |
| "loss": 0.5992, |
| "step": 661 |
| }, |
| { |
| "epoch": 8.39, |
| "learning_rate": 5e-05, |
| "loss": 0.7123, |
| "step": 662 |
| }, |
| { |
| "epoch": 8.41, |
| "learning_rate": 5e-05, |
| "loss": 0.6752, |
| "step": 663 |
| }, |
| { |
| "epoch": 8.42, |
| "learning_rate": 5e-05, |
| "loss": 0.7463, |
| "step": 664 |
| }, |
| { |
| "epoch": 8.43, |
| "learning_rate": 5e-05, |
| "loss": 0.5771, |
| "step": 665 |
| }, |
| { |
| "epoch": 8.44, |
| "learning_rate": 5e-05, |
| "loss": 0.5872, |
| "step": 666 |
| }, |
| { |
| "epoch": 8.46, |
| "learning_rate": 5e-05, |
| "loss": 0.6154, |
| "step": 667 |
| }, |
| { |
| "epoch": 8.47, |
| "learning_rate": 5e-05, |
| "loss": 0.7467, |
| "step": 668 |
| }, |
| { |
| "epoch": 8.48, |
| "learning_rate": 5e-05, |
| "loss": 0.6267, |
| "step": 669 |
| }, |
| { |
| "epoch": 8.49, |
| "learning_rate": 5e-05, |
| "loss": 0.5279, |
| "step": 670 |
| }, |
| { |
| "epoch": 8.51, |
| "learning_rate": 5e-05, |
| "loss": 0.5716, |
| "step": 671 |
| }, |
| { |
| "epoch": 8.52, |
| "learning_rate": 5e-05, |
| "loss": 0.6625, |
| "step": 672 |
| }, |
| { |
| "epoch": 8.53, |
| "learning_rate": 5e-05, |
| "loss": 0.6882, |
| "step": 673 |
| }, |
| { |
| "epoch": 8.55, |
| "learning_rate": 5e-05, |
| "loss": 0.7455, |
| "step": 674 |
| }, |
| { |
| "epoch": 8.56, |
| "learning_rate": 5e-05, |
| "loss": 0.6188, |
| "step": 675 |
| }, |
| { |
| "epoch": 8.57, |
| "learning_rate": 5e-05, |
| "loss": 0.5472, |
| "step": 676 |
| }, |
| { |
| "epoch": 8.58, |
| "learning_rate": 5e-05, |
| "loss": 0.5699, |
| "step": 677 |
| }, |
| { |
| "epoch": 8.6, |
| "learning_rate": 5e-05, |
| "loss": 0.5967, |
| "step": 678 |
| }, |
| { |
| "epoch": 8.61, |
| "learning_rate": 5e-05, |
| "loss": 0.666, |
| "step": 679 |
| }, |
| { |
| "epoch": 8.62, |
| "learning_rate": 5e-05, |
| "loss": 0.6593, |
| "step": 680 |
| }, |
| { |
| "epoch": 8.63, |
| "learning_rate": 5e-05, |
| "loss": 0.5339, |
| "step": 681 |
| }, |
| { |
| "epoch": 8.65, |
| "learning_rate": 5e-05, |
| "loss": 0.5736, |
| "step": 682 |
| }, |
| { |
| "epoch": 8.66, |
| "learning_rate": 5e-05, |
| "loss": 0.594, |
| "step": 683 |
| }, |
| { |
| "epoch": 8.67, |
| "learning_rate": 5e-05, |
| "loss": 0.679, |
| "step": 684 |
| }, |
| { |
| "epoch": 8.68, |
| "learning_rate": 5e-05, |
| "loss": 0.847, |
| "step": 685 |
| }, |
| { |
| "epoch": 8.7, |
| "learning_rate": 5e-05, |
| "loss": 0.6812, |
| "step": 686 |
| }, |
| { |
| "epoch": 8.71, |
| "learning_rate": 5e-05, |
| "loss": 0.6936, |
| "step": 687 |
| }, |
| { |
| "epoch": 8.72, |
| "learning_rate": 5e-05, |
| "loss": 0.6807, |
| "step": 688 |
| }, |
| { |
| "epoch": 8.74, |
| "learning_rate": 5e-05, |
| "loss": 0.4848, |
| "step": 689 |
| }, |
| { |
| "epoch": 8.75, |
| "learning_rate": 5e-05, |
| "loss": 0.6636, |
| "step": 690 |
| }, |
| { |
| "epoch": 8.76, |
| "learning_rate": 5e-05, |
| "loss": 0.6176, |
| "step": 691 |
| }, |
| { |
| "epoch": 8.77, |
| "learning_rate": 5e-05, |
| "loss": 0.6465, |
| "step": 692 |
| }, |
| { |
| "epoch": 8.79, |
| "learning_rate": 5e-05, |
| "loss": 0.6425, |
| "step": 693 |
| }, |
| { |
| "epoch": 8.8, |
| "learning_rate": 5e-05, |
| "loss": 0.5508, |
| "step": 694 |
| }, |
| { |
| "epoch": 8.81, |
| "learning_rate": 5e-05, |
| "loss": 0.6239, |
| "step": 695 |
| }, |
| { |
| "epoch": 8.82, |
| "learning_rate": 5e-05, |
| "loss": 0.7543, |
| "step": 696 |
| }, |
| { |
| "epoch": 8.84, |
| "learning_rate": 5e-05, |
| "loss": 0.6024, |
| "step": 697 |
| }, |
| { |
| "epoch": 8.85, |
| "learning_rate": 5e-05, |
| "loss": 0.73, |
| "step": 698 |
| }, |
| { |
| "epoch": 8.86, |
| "learning_rate": 5e-05, |
| "loss": 0.6318, |
| "step": 699 |
| }, |
| { |
| "epoch": 8.87, |
| "learning_rate": 5e-05, |
| "loss": 0.6396, |
| "step": 700 |
| }, |
| { |
| "epoch": 8.89, |
| "learning_rate": 5e-05, |
| "loss": 0.5022, |
| "step": 701 |
| }, |
| { |
| "epoch": 8.9, |
| "learning_rate": 5e-05, |
| "loss": 0.5851, |
| "step": 702 |
| }, |
| { |
| "epoch": 8.91, |
| "learning_rate": 5e-05, |
| "loss": 0.6386, |
| "step": 703 |
| }, |
| { |
| "epoch": 8.93, |
| "learning_rate": 5e-05, |
| "loss": 0.5703, |
| "step": 704 |
| }, |
| { |
| "epoch": 8.94, |
| "learning_rate": 5e-05, |
| "loss": 0.6477, |
| "step": 705 |
| }, |
| { |
| "epoch": 8.95, |
| "learning_rate": 5e-05, |
| "loss": 0.6033, |
| "step": 706 |
| }, |
| { |
| "epoch": 8.96, |
| "learning_rate": 5e-05, |
| "loss": 0.7205, |
| "step": 707 |
| }, |
| { |
| "epoch": 8.98, |
| "learning_rate": 5e-05, |
| "loss": 0.6237, |
| "step": 708 |
| }, |
| { |
| "epoch": 8.99, |
| "learning_rate": 5e-05, |
| "loss": 0.6492, |
| "step": 709 |
| }, |
| { |
| "epoch": 9.0, |
| "learning_rate": 5e-05, |
| "loss": 0.3812, |
| "step": 710 |
| }, |
| { |
| "epoch": 9.01, |
| "learning_rate": 5e-05, |
| "loss": 0.5149, |
| "step": 711 |
| }, |
| { |
| "epoch": 9.03, |
| "learning_rate": 5e-05, |
| "loss": 0.5416, |
| "step": 712 |
| }, |
| { |
| "epoch": 9.04, |
| "learning_rate": 5e-05, |
| "loss": 0.6983, |
| "step": 713 |
| }, |
| { |
| "epoch": 9.05, |
| "learning_rate": 5e-05, |
| "loss": 0.5732, |
| "step": 714 |
| }, |
| { |
| "epoch": 9.06, |
| "learning_rate": 5e-05, |
| "loss": 0.603, |
| "step": 715 |
| }, |
| { |
| "epoch": 9.08, |
| "learning_rate": 5e-05, |
| "loss": 0.5342, |
| "step": 716 |
| }, |
| { |
| "epoch": 9.09, |
| "learning_rate": 5e-05, |
| "loss": 0.4529, |
| "step": 717 |
| }, |
| { |
| "epoch": 9.1, |
| "learning_rate": 5e-05, |
| "loss": 0.5396, |
| "step": 718 |
| }, |
| { |
| "epoch": 9.12, |
| "learning_rate": 5e-05, |
| "loss": 0.5394, |
| "step": 719 |
| }, |
| { |
| "epoch": 9.13, |
| "learning_rate": 5e-05, |
| "loss": 0.6228, |
| "step": 720 |
| }, |
| { |
| "epoch": 9.14, |
| "learning_rate": 5e-05, |
| "loss": 0.618, |
| "step": 721 |
| }, |
| { |
| "epoch": 9.15, |
| "learning_rate": 5e-05, |
| "loss": 0.5387, |
| "step": 722 |
| }, |
| { |
| "epoch": 9.17, |
| "learning_rate": 5e-05, |
| "loss": 0.6755, |
| "step": 723 |
| }, |
| { |
| "epoch": 9.18, |
| "learning_rate": 5e-05, |
| "loss": 0.5405, |
| "step": 724 |
| }, |
| { |
| "epoch": 9.19, |
| "learning_rate": 5e-05, |
| "loss": 0.5071, |
| "step": 725 |
| }, |
| { |
| "epoch": 9.2, |
| "learning_rate": 5e-05, |
| "loss": 0.452, |
| "step": 726 |
| }, |
| { |
| "epoch": 9.22, |
| "learning_rate": 5e-05, |
| "loss": 0.4978, |
| "step": 727 |
| }, |
| { |
| "epoch": 9.23, |
| "learning_rate": 5e-05, |
| "loss": 0.5898, |
| "step": 728 |
| }, |
| { |
| "epoch": 9.24, |
| "learning_rate": 5e-05, |
| "loss": 0.4939, |
| "step": 729 |
| }, |
| { |
| "epoch": 9.26, |
| "learning_rate": 5e-05, |
| "loss": 0.6207, |
| "step": 730 |
| }, |
| { |
| "epoch": 9.27, |
| "learning_rate": 5e-05, |
| "loss": 0.5715, |
| "step": 731 |
| }, |
| { |
| "epoch": 9.28, |
| "learning_rate": 5e-05, |
| "loss": 0.5505, |
| "step": 732 |
| }, |
| { |
| "epoch": 9.29, |
| "learning_rate": 5e-05, |
| "loss": 0.5795, |
| "step": 733 |
| }, |
| { |
| "epoch": 9.31, |
| "learning_rate": 5e-05, |
| "loss": 0.606, |
| "step": 734 |
| }, |
| { |
| "epoch": 9.32, |
| "learning_rate": 5e-05, |
| "loss": 0.5188, |
| "step": 735 |
| }, |
| { |
| "epoch": 9.33, |
| "learning_rate": 5e-05, |
| "loss": 0.5559, |
| "step": 736 |
| }, |
| { |
| "epoch": 9.34, |
| "learning_rate": 5e-05, |
| "loss": 0.5515, |
| "step": 737 |
| }, |
| { |
| "epoch": 9.36, |
| "learning_rate": 5e-05, |
| "loss": 0.5927, |
| "step": 738 |
| }, |
| { |
| "epoch": 9.37, |
| "learning_rate": 5e-05, |
| "loss": 0.4682, |
| "step": 739 |
| }, |
| { |
| "epoch": 9.38, |
| "learning_rate": 5e-05, |
| "loss": 0.4223, |
| "step": 740 |
| }, |
| { |
| "epoch": 9.39, |
| "learning_rate": 5e-05, |
| "loss": 0.6277, |
| "step": 741 |
| }, |
| { |
| "epoch": 9.41, |
| "learning_rate": 5e-05, |
| "loss": 0.6748, |
| "step": 742 |
| }, |
| { |
| "epoch": 9.42, |
| "learning_rate": 5e-05, |
| "loss": 0.6284, |
| "step": 743 |
| }, |
| { |
| "epoch": 9.43, |
| "learning_rate": 5e-05, |
| "loss": 0.6585, |
| "step": 744 |
| }, |
| { |
| "epoch": 9.45, |
| "learning_rate": 5e-05, |
| "loss": 0.5709, |
| "step": 745 |
| }, |
| { |
| "epoch": 9.46, |
| "learning_rate": 5e-05, |
| "loss": 0.6085, |
| "step": 746 |
| }, |
| { |
| "epoch": 9.47, |
| "learning_rate": 5e-05, |
| "loss": 0.6371, |
| "step": 747 |
| }, |
| { |
| "epoch": 9.48, |
| "learning_rate": 5e-05, |
| "loss": 0.6223, |
| "step": 748 |
| }, |
| { |
| "epoch": 9.5, |
| "learning_rate": 5e-05, |
| "loss": 0.5335, |
| "step": 749 |
| }, |
| { |
| "epoch": 9.51, |
| "learning_rate": 5e-05, |
| "loss": 0.6281, |
| "step": 750 |
| }, |
| { |
| "epoch": 9.52, |
| "learning_rate": 5e-05, |
| "loss": 0.5711, |
| "step": 751 |
| }, |
| { |
| "epoch": 9.53, |
| "learning_rate": 5e-05, |
| "loss": 0.4897, |
| "step": 752 |
| }, |
| { |
| "epoch": 9.55, |
| "learning_rate": 5e-05, |
| "loss": 0.6711, |
| "step": 753 |
| }, |
| { |
| "epoch": 9.56, |
| "learning_rate": 5e-05, |
| "loss": 0.4947, |
| "step": 754 |
| }, |
| { |
| "epoch": 9.57, |
| "learning_rate": 5e-05, |
| "loss": 0.3919, |
| "step": 755 |
| }, |
| { |
| "epoch": 9.58, |
| "learning_rate": 5e-05, |
| "loss": 0.6627, |
| "step": 756 |
| }, |
| { |
| "epoch": 9.6, |
| "learning_rate": 5e-05, |
| "loss": 0.5762, |
| "step": 757 |
| }, |
| { |
| "epoch": 9.61, |
| "learning_rate": 5e-05, |
| "loss": 0.6124, |
| "step": 758 |
| }, |
| { |
| "epoch": 9.62, |
| "learning_rate": 5e-05, |
| "loss": 0.6042, |
| "step": 759 |
| }, |
| { |
| "epoch": 9.64, |
| "learning_rate": 5e-05, |
| "loss": 0.6709, |
| "step": 760 |
| }, |
| { |
| "epoch": 9.65, |
| "learning_rate": 5e-05, |
| "loss": 0.5055, |
| "step": 761 |
| }, |
| { |
| "epoch": 9.66, |
| "learning_rate": 5e-05, |
| "loss": 0.6379, |
| "step": 762 |
| }, |
| { |
| "epoch": 9.67, |
| "learning_rate": 5e-05, |
| "loss": 0.5237, |
| "step": 763 |
| }, |
| { |
| "epoch": 9.69, |
| "learning_rate": 5e-05, |
| "loss": 0.6292, |
| "step": 764 |
| }, |
| { |
| "epoch": 9.7, |
| "learning_rate": 5e-05, |
| "loss": 0.4818, |
| "step": 765 |
| }, |
| { |
| "epoch": 9.71, |
| "learning_rate": 5e-05, |
| "loss": 0.6838, |
| "step": 766 |
| }, |
| { |
| "epoch": 9.72, |
| "learning_rate": 5e-05, |
| "loss": 0.651, |
| "step": 767 |
| }, |
| { |
| "epoch": 9.74, |
| "learning_rate": 5e-05, |
| "loss": 0.531, |
| "step": 768 |
| }, |
| { |
| "epoch": 9.75, |
| "learning_rate": 5e-05, |
| "loss": 0.7369, |
| "step": 769 |
| }, |
| { |
| "epoch": 9.76, |
| "learning_rate": 5e-05, |
| "loss": 0.6574, |
| "step": 770 |
| }, |
| { |
| "epoch": 9.77, |
| "learning_rate": 5e-05, |
| "loss": 0.5632, |
| "step": 771 |
| }, |
| { |
| "epoch": 9.79, |
| "learning_rate": 5e-05, |
| "loss": 0.5956, |
| "step": 772 |
| }, |
| { |
| "epoch": 9.8, |
| "learning_rate": 5e-05, |
| "loss": 0.7866, |
| "step": 773 |
| }, |
| { |
| "epoch": 9.81, |
| "learning_rate": 5e-05, |
| "loss": 0.587, |
| "step": 774 |
| }, |
| { |
| "epoch": 9.83, |
| "learning_rate": 5e-05, |
| "loss": 0.6711, |
| "step": 775 |
| }, |
| { |
| "epoch": 9.84, |
| "learning_rate": 5e-05, |
| "loss": 0.6587, |
| "step": 776 |
| }, |
| { |
| "epoch": 9.85, |
| "learning_rate": 5e-05, |
| "loss": 0.7639, |
| "step": 777 |
| }, |
| { |
| "epoch": 9.86, |
| "learning_rate": 5e-05, |
| "loss": 0.5661, |
| "step": 778 |
| }, |
| { |
| "epoch": 9.88, |
| "learning_rate": 5e-05, |
| "loss": 0.6012, |
| "step": 779 |
| }, |
| { |
| "epoch": 9.89, |
| "learning_rate": 5e-05, |
| "loss": 0.4672, |
| "step": 780 |
| }, |
| { |
| "epoch": 9.9, |
| "learning_rate": 5e-05, |
| "loss": 0.5804, |
| "step": 781 |
| }, |
| { |
| "epoch": 9.91, |
| "learning_rate": 5e-05, |
| "loss": 0.5762, |
| "step": 782 |
| }, |
| { |
| "epoch": 9.93, |
| "learning_rate": 5e-05, |
| "loss": 0.5283, |
| "step": 783 |
| }, |
| { |
| "epoch": 9.94, |
| "learning_rate": 5e-05, |
| "loss": 0.5313, |
| "step": 784 |
| }, |
| { |
| "epoch": 9.95, |
| "learning_rate": 5e-05, |
| "loss": 0.3967, |
| "step": 785 |
| }, |
| { |
| "epoch": 9.97, |
| "learning_rate": 5e-05, |
| "loss": 0.712, |
| "step": 786 |
| }, |
| { |
| "epoch": 9.98, |
| "learning_rate": 5e-05, |
| "loss": 0.645, |
| "step": 787 |
| }, |
| { |
| "epoch": 9.99, |
| "learning_rate": 5e-05, |
| "loss": 0.606, |
| "step": 788 |
| }, |
| { |
| "epoch": 10.0, |
| "learning_rate": 5e-05, |
| "loss": 0.5231, |
| "step": 789 |
| }, |
| { |
| "epoch": 10.02, |
| "learning_rate": 5e-05, |
| "loss": 0.5255, |
| "step": 790 |
| }, |
| { |
| "epoch": 10.03, |
| "learning_rate": 5e-05, |
| "loss": 0.5819, |
| "step": 791 |
| }, |
| { |
| "epoch": 10.04, |
| "learning_rate": 5e-05, |
| "loss": 0.6257, |
| "step": 792 |
| }, |
| { |
| "epoch": 10.05, |
| "learning_rate": 5e-05, |
| "loss": 0.6124, |
| "step": 793 |
| }, |
| { |
| "epoch": 10.07, |
| "learning_rate": 5e-05, |
| "loss": 0.4842, |
| "step": 794 |
| }, |
| { |
| "epoch": 10.08, |
| "learning_rate": 5e-05, |
| "loss": 0.4978, |
| "step": 795 |
| }, |
| { |
| "epoch": 10.09, |
| "learning_rate": 5e-05, |
| "loss": 0.5709, |
| "step": 796 |
| }, |
| { |
| "epoch": 10.1, |
| "learning_rate": 5e-05, |
| "loss": 0.6196, |
| "step": 797 |
| }, |
| { |
| "epoch": 10.12, |
| "learning_rate": 5e-05, |
| "loss": 0.5189, |
| "step": 798 |
| }, |
| { |
| "epoch": 10.13, |
| "learning_rate": 5e-05, |
| "loss": 0.5959, |
| "step": 799 |
| }, |
| { |
| "epoch": 10.14, |
| "learning_rate": 5e-05, |
| "loss": 0.5035, |
| "step": 800 |
| }, |
| { |
| "epoch": 10.16, |
| "learning_rate": 5e-05, |
| "loss": 0.6525, |
| "step": 801 |
| }, |
| { |
| "epoch": 10.17, |
| "learning_rate": 5e-05, |
| "loss": 0.6511, |
| "step": 802 |
| }, |
| { |
| "epoch": 10.18, |
| "learning_rate": 5e-05, |
| "loss": 0.432, |
| "step": 803 |
| }, |
| { |
| "epoch": 10.19, |
| "learning_rate": 5e-05, |
| "loss": 0.4594, |
| "step": 804 |
| }, |
| { |
| "epoch": 10.21, |
| "learning_rate": 5e-05, |
| "loss": 0.4679, |
| "step": 805 |
| }, |
| { |
| "epoch": 10.22, |
| "learning_rate": 5e-05, |
| "loss": 0.5258, |
| "step": 806 |
| }, |
| { |
| "epoch": 10.23, |
| "learning_rate": 5e-05, |
| "loss": 0.7523, |
| "step": 807 |
| }, |
| { |
| "epoch": 10.24, |
| "learning_rate": 5e-05, |
| "loss": 0.629, |
| "step": 808 |
| }, |
| { |
| "epoch": 10.26, |
| "learning_rate": 5e-05, |
| "loss": 0.5258, |
| "step": 809 |
| }, |
| { |
| "epoch": 10.27, |
| "learning_rate": 5e-05, |
| "loss": 0.5151, |
| "step": 810 |
| }, |
| { |
| "epoch": 10.28, |
| "learning_rate": 5e-05, |
| "loss": 0.6367, |
| "step": 811 |
| }, |
| { |
| "epoch": 10.29, |
| "learning_rate": 5e-05, |
| "loss": 0.5675, |
| "step": 812 |
| }, |
| { |
| "epoch": 10.31, |
| "learning_rate": 5e-05, |
| "loss": 0.6341, |
| "step": 813 |
| }, |
| { |
| "epoch": 10.32, |
| "learning_rate": 5e-05, |
| "loss": 0.5514, |
| "step": 814 |
| }, |
| { |
| "epoch": 10.33, |
| "learning_rate": 5e-05, |
| "loss": 0.5146, |
| "step": 815 |
| }, |
| { |
| "epoch": 10.35, |
| "learning_rate": 5e-05, |
| "loss": 0.4999, |
| "step": 816 |
| }, |
| { |
| "epoch": 10.36, |
| "learning_rate": 5e-05, |
| "loss": 0.399, |
| "step": 817 |
| }, |
| { |
| "epoch": 10.37, |
| "learning_rate": 5e-05, |
| "loss": 0.5131, |
| "step": 818 |
| }, |
| { |
| "epoch": 10.38, |
| "learning_rate": 5e-05, |
| "loss": 0.6276, |
| "step": 819 |
| }, |
| { |
| "epoch": 10.4, |
| "learning_rate": 5e-05, |
| "loss": 0.6068, |
| "step": 820 |
| }, |
| { |
| "epoch": 10.41, |
| "learning_rate": 5e-05, |
| "loss": 0.5765, |
| "step": 821 |
| }, |
| { |
| "epoch": 10.42, |
| "learning_rate": 5e-05, |
| "loss": 0.4395, |
| "step": 822 |
| }, |
| { |
| "epoch": 10.43, |
| "learning_rate": 5e-05, |
| "loss": 0.5028, |
| "step": 823 |
| }, |
| { |
| "epoch": 10.45, |
| "learning_rate": 5e-05, |
| "loss": 0.4159, |
| "step": 824 |
| }, |
| { |
| "epoch": 10.46, |
| "learning_rate": 5e-05, |
| "loss": 0.5623, |
| "step": 825 |
| }, |
| { |
| "epoch": 10.47, |
| "learning_rate": 5e-05, |
| "loss": 0.6216, |
| "step": 826 |
| }, |
| { |
| "epoch": 10.48, |
| "learning_rate": 5e-05, |
| "loss": 0.5637, |
| "step": 827 |
| }, |
| { |
| "epoch": 10.5, |
| "learning_rate": 5e-05, |
| "loss": 0.4251, |
| "step": 828 |
| }, |
| { |
| "epoch": 10.51, |
| "learning_rate": 5e-05, |
| "loss": 0.6422, |
| "step": 829 |
| }, |
| { |
| "epoch": 10.52, |
| "learning_rate": 5e-05, |
| "loss": 0.421, |
| "step": 830 |
| }, |
| { |
| "epoch": 10.54, |
| "learning_rate": 5e-05, |
| "loss": 0.5367, |
| "step": 831 |
| }, |
| { |
| "epoch": 10.55, |
| "learning_rate": 5e-05, |
| "loss": 0.4775, |
| "step": 832 |
| }, |
| { |
| "epoch": 10.56, |
| "learning_rate": 5e-05, |
| "loss": 0.7163, |
| "step": 833 |
| }, |
| { |
| "epoch": 10.57, |
| "learning_rate": 5e-05, |
| "loss": 0.5123, |
| "step": 834 |
| }, |
| { |
| "epoch": 10.59, |
| "learning_rate": 5e-05, |
| "loss": 0.4816, |
| "step": 835 |
| }, |
| { |
| "epoch": 10.6, |
| "learning_rate": 5e-05, |
| "loss": 0.5764, |
| "step": 836 |
| }, |
| { |
| "epoch": 10.61, |
| "learning_rate": 5e-05, |
| "loss": 0.4854, |
| "step": 837 |
| }, |
| { |
| "epoch": 10.62, |
| "learning_rate": 5e-05, |
| "loss": 0.5264, |
| "step": 838 |
| }, |
| { |
| "epoch": 10.64, |
| "learning_rate": 5e-05, |
| "loss": 0.4726, |
| "step": 839 |
| }, |
| { |
| "epoch": 10.65, |
| "learning_rate": 5e-05, |
| "loss": 0.4714, |
| "step": 840 |
| }, |
| { |
| "epoch": 10.66, |
| "learning_rate": 5e-05, |
| "loss": 0.647, |
| "step": 841 |
| }, |
| { |
| "epoch": 10.68, |
| "learning_rate": 5e-05, |
| "loss": 0.6007, |
| "step": 842 |
| }, |
| { |
| "epoch": 10.69, |
| "learning_rate": 5e-05, |
| "loss": 0.57, |
| "step": 843 |
| }, |
| { |
| "epoch": 10.7, |
| "learning_rate": 5e-05, |
| "loss": 0.5966, |
| "step": 844 |
| }, |
| { |
| "epoch": 10.71, |
| "learning_rate": 5e-05, |
| "loss": 0.4865, |
| "step": 845 |
| }, |
| { |
| "epoch": 10.73, |
| "learning_rate": 5e-05, |
| "loss": 0.5159, |
| "step": 846 |
| }, |
| { |
| "epoch": 10.74, |
| "learning_rate": 5e-05, |
| "loss": 0.458, |
| "step": 847 |
| }, |
| { |
| "epoch": 10.75, |
| "learning_rate": 5e-05, |
| "loss": 0.431, |
| "step": 848 |
| }, |
| { |
| "epoch": 10.76, |
| "learning_rate": 5e-05, |
| "loss": 0.4846, |
| "step": 849 |
| }, |
| { |
| "epoch": 10.78, |
| "learning_rate": 5e-05, |
| "loss": 0.4295, |
| "step": 850 |
| }, |
| { |
| "epoch": 10.79, |
| "learning_rate": 5e-05, |
| "loss": 0.7443, |
| "step": 851 |
| }, |
| { |
| "epoch": 10.8, |
| "learning_rate": 5e-05, |
| "loss": 0.446, |
| "step": 852 |
| }, |
| { |
| "epoch": 10.81, |
| "learning_rate": 5e-05, |
| "loss": 0.5116, |
| "step": 853 |
| }, |
| { |
| "epoch": 10.83, |
| "learning_rate": 5e-05, |
| "loss": 0.5221, |
| "step": 854 |
| }, |
| { |
| "epoch": 10.84, |
| "learning_rate": 5e-05, |
| "loss": 0.5675, |
| "step": 855 |
| }, |
| { |
| "epoch": 10.85, |
| "learning_rate": 5e-05, |
| "loss": 0.5102, |
| "step": 856 |
| }, |
| { |
| "epoch": 10.87, |
| "learning_rate": 5e-05, |
| "loss": 0.6235, |
| "step": 857 |
| }, |
| { |
| "epoch": 10.88, |
| "learning_rate": 5e-05, |
| "loss": 0.53, |
| "step": 858 |
| }, |
| { |
| "epoch": 10.89, |
| "learning_rate": 5e-05, |
| "loss": 0.4925, |
| "step": 859 |
| }, |
| { |
| "epoch": 10.9, |
| "learning_rate": 5e-05, |
| "loss": 0.6401, |
| "step": 860 |
| }, |
| { |
| "epoch": 10.92, |
| "learning_rate": 5e-05, |
| "loss": 0.5017, |
| "step": 861 |
| }, |
| { |
| "epoch": 10.93, |
| "learning_rate": 5e-05, |
| "loss": 0.4983, |
| "step": 862 |
| }, |
| { |
| "epoch": 10.94, |
| "learning_rate": 5e-05, |
| "loss": 0.6534, |
| "step": 863 |
| }, |
| { |
| "epoch": 10.95, |
| "learning_rate": 5e-05, |
| "loss": 0.5763, |
| "step": 864 |
| }, |
| { |
| "epoch": 10.97, |
| "learning_rate": 5e-05, |
| "loss": 0.5763, |
| "step": 865 |
| }, |
| { |
| "epoch": 10.98, |
| "learning_rate": 5e-05, |
| "loss": 0.4669, |
| "step": 866 |
| }, |
| { |
| "epoch": 10.99, |
| "learning_rate": 5e-05, |
| "loss": 0.6519, |
| "step": 867 |
| }, |
| { |
| "epoch": 11.0, |
| "learning_rate": 5e-05, |
| "loss": 0.5764, |
| "step": 868 |
| }, |
| { |
| "epoch": 11.02, |
| "learning_rate": 5e-05, |
| "loss": 0.6319, |
| "step": 869 |
| }, |
| { |
| "epoch": 11.03, |
| "learning_rate": 5e-05, |
| "loss": 0.4754, |
| "step": 870 |
| }, |
| { |
| "epoch": 11.04, |
| "learning_rate": 5e-05, |
| "loss": 0.4764, |
| "step": 871 |
| }, |
| { |
| "epoch": 11.06, |
| "learning_rate": 5e-05, |
| "loss": 0.5356, |
| "step": 872 |
| }, |
| { |
| "epoch": 11.07, |
| "learning_rate": 5e-05, |
| "loss": 0.5807, |
| "step": 873 |
| }, |
| { |
| "epoch": 11.08, |
| "learning_rate": 5e-05, |
| "loss": 0.4799, |
| "step": 874 |
| }, |
| { |
| "epoch": 11.09, |
| "learning_rate": 5e-05, |
| "loss": 0.6319, |
| "step": 875 |
| }, |
| { |
| "epoch": 11.11, |
| "learning_rate": 5e-05, |
| "loss": 0.4283, |
| "step": 876 |
| }, |
| { |
| "epoch": 11.12, |
| "learning_rate": 5e-05, |
| "loss": 0.4271, |
| "step": 877 |
| }, |
| { |
| "epoch": 11.13, |
| "learning_rate": 5e-05, |
| "loss": 0.4639, |
| "step": 878 |
| }, |
| { |
| "epoch": 11.14, |
| "learning_rate": 5e-05, |
| "loss": 0.5665, |
| "step": 879 |
| }, |
| { |
| "epoch": 11.16, |
| "learning_rate": 5e-05, |
| "loss": 0.5105, |
| "step": 880 |
| }, |
| { |
| "epoch": 11.17, |
| "learning_rate": 5e-05, |
| "loss": 0.4494, |
| "step": 881 |
| }, |
| { |
| "epoch": 11.18, |
| "learning_rate": 5e-05, |
| "loss": 0.5205, |
| "step": 882 |
| }, |
| { |
| "epoch": 11.19, |
| "learning_rate": 5e-05, |
| "loss": 0.5538, |
| "step": 883 |
| }, |
| { |
| "epoch": 11.21, |
| "learning_rate": 5e-05, |
| "loss": 0.5935, |
| "step": 884 |
| }, |
| { |
| "epoch": 11.22, |
| "learning_rate": 5e-05, |
| "loss": 0.4685, |
| "step": 885 |
| }, |
| { |
| "epoch": 11.23, |
| "learning_rate": 5e-05, |
| "loss": 0.5145, |
| "step": 886 |
| }, |
| { |
| "epoch": 11.25, |
| "learning_rate": 5e-05, |
| "loss": 0.585, |
| "step": 887 |
| }, |
| { |
| "epoch": 11.26, |
| "learning_rate": 5e-05, |
| "loss": 0.3645, |
| "step": 888 |
| }, |
| { |
| "epoch": 11.27, |
| "learning_rate": 5e-05, |
| "loss": 0.502, |
| "step": 889 |
| }, |
| { |
| "epoch": 11.28, |
| "learning_rate": 5e-05, |
| "loss": 0.5752, |
| "step": 890 |
| }, |
| { |
| "epoch": 11.3, |
| "learning_rate": 5e-05, |
| "loss": 0.5229, |
| "step": 891 |
| }, |
| { |
| "epoch": 11.31, |
| "learning_rate": 5e-05, |
| "loss": 0.5768, |
| "step": 892 |
| }, |
| { |
| "epoch": 11.32, |
| "learning_rate": 5e-05, |
| "loss": 0.5751, |
| "step": 893 |
| }, |
| { |
| "epoch": 11.33, |
| "learning_rate": 5e-05, |
| "loss": 0.4743, |
| "step": 894 |
| }, |
| { |
| "epoch": 11.35, |
| "learning_rate": 5e-05, |
| "loss": 0.5179, |
| "step": 895 |
| }, |
| { |
| "epoch": 11.36, |
| "learning_rate": 5e-05, |
| "loss": 0.4811, |
| "step": 896 |
| }, |
| { |
| "epoch": 11.37, |
| "learning_rate": 5e-05, |
| "loss": 0.375, |
| "step": 897 |
| }, |
| { |
| "epoch": 11.39, |
| "learning_rate": 5e-05, |
| "loss": 0.4885, |
| "step": 898 |
| }, |
| { |
| "epoch": 11.4, |
| "learning_rate": 5e-05, |
| "loss": 0.5356, |
| "step": 899 |
| }, |
| { |
| "epoch": 11.41, |
| "learning_rate": 5e-05, |
| "loss": 0.524, |
| "step": 900 |
| }, |
| { |
| "epoch": 11.42, |
| "learning_rate": 5e-05, |
| "loss": 0.439, |
| "step": 901 |
| }, |
| { |
| "epoch": 11.44, |
| "learning_rate": 5e-05, |
| "loss": 0.4821, |
| "step": 902 |
| }, |
| { |
| "epoch": 11.45, |
| "learning_rate": 5e-05, |
| "loss": 0.4556, |
| "step": 903 |
| }, |
| { |
| "epoch": 11.46, |
| "learning_rate": 5e-05, |
| "loss": 0.4192, |
| "step": 904 |
| }, |
| { |
| "epoch": 11.47, |
| "learning_rate": 5e-05, |
| "loss": 0.4317, |
| "step": 905 |
| }, |
| { |
| "epoch": 11.49, |
| "learning_rate": 5e-05, |
| "loss": 0.5406, |
| "step": 906 |
| }, |
| { |
| "epoch": 11.5, |
| "learning_rate": 5e-05, |
| "loss": 0.4377, |
| "step": 907 |
| }, |
| { |
| "epoch": 11.51, |
| "learning_rate": 5e-05, |
| "loss": 0.5314, |
| "step": 908 |
| }, |
| { |
| "epoch": 11.52, |
| "learning_rate": 5e-05, |
| "loss": 0.5076, |
| "step": 909 |
| }, |
| { |
| "epoch": 11.54, |
| "learning_rate": 5e-05, |
| "loss": 0.5106, |
| "step": 910 |
| }, |
| { |
| "epoch": 11.55, |
| "learning_rate": 5e-05, |
| "loss": 0.6748, |
| "step": 911 |
| }, |
| { |
| "epoch": 11.56, |
| "learning_rate": 5e-05, |
| "loss": 0.5719, |
| "step": 912 |
| }, |
| { |
| "epoch": 11.58, |
| "learning_rate": 5e-05, |
| "loss": 0.5189, |
| "step": 913 |
| }, |
| { |
| "epoch": 11.59, |
| "learning_rate": 5e-05, |
| "loss": 0.5477, |
| "step": 914 |
| }, |
| { |
| "epoch": 11.6, |
| "learning_rate": 5e-05, |
| "loss": 0.5941, |
| "step": 915 |
| }, |
| { |
| "epoch": 11.61, |
| "learning_rate": 5e-05, |
| "loss": 0.48, |
| "step": 916 |
| }, |
| { |
| "epoch": 11.63, |
| "learning_rate": 5e-05, |
| "loss": 0.4149, |
| "step": 917 |
| }, |
| { |
| "epoch": 11.64, |
| "learning_rate": 5e-05, |
| "loss": 0.4771, |
| "step": 918 |
| }, |
| { |
| "epoch": 11.65, |
| "learning_rate": 5e-05, |
| "loss": 0.4871, |
| "step": 919 |
| }, |
| { |
| "epoch": 11.66, |
| "learning_rate": 5e-05, |
| "loss": 0.4545, |
| "step": 920 |
| }, |
| { |
| "epoch": 11.68, |
| "learning_rate": 5e-05, |
| "loss": 0.4596, |
| "step": 921 |
| }, |
| { |
| "epoch": 11.69, |
| "learning_rate": 5e-05, |
| "loss": 0.4525, |
| "step": 922 |
| }, |
| { |
| "epoch": 11.7, |
| "learning_rate": 5e-05, |
| "loss": 0.437, |
| "step": 923 |
| }, |
| { |
| "epoch": 11.71, |
| "learning_rate": 5e-05, |
| "loss": 0.5391, |
| "step": 924 |
| }, |
| { |
| "epoch": 11.73, |
| "learning_rate": 5e-05, |
| "loss": 0.4782, |
| "step": 925 |
| }, |
| { |
| "epoch": 11.74, |
| "learning_rate": 5e-05, |
| "loss": 0.4787, |
| "step": 926 |
| }, |
| { |
| "epoch": 11.75, |
| "learning_rate": 5e-05, |
| "loss": 0.541, |
| "step": 927 |
| }, |
| { |
| "epoch": 11.77, |
| "learning_rate": 5e-05, |
| "loss": 0.5323, |
| "step": 928 |
| }, |
| { |
| "epoch": 11.78, |
| "learning_rate": 5e-05, |
| "loss": 0.5033, |
| "step": 929 |
| }, |
| { |
| "epoch": 11.79, |
| "learning_rate": 5e-05, |
| "loss": 0.5184, |
| "step": 930 |
| }, |
| { |
| "epoch": 11.8, |
| "learning_rate": 5e-05, |
| "loss": 0.5864, |
| "step": 931 |
| }, |
| { |
| "epoch": 11.82, |
| "learning_rate": 5e-05, |
| "loss": 0.4332, |
| "step": 932 |
| }, |
| { |
| "epoch": 11.83, |
| "learning_rate": 5e-05, |
| "loss": 0.5447, |
| "step": 933 |
| }, |
| { |
| "epoch": 11.84, |
| "learning_rate": 5e-05, |
| "loss": 0.4963, |
| "step": 934 |
| }, |
| { |
| "epoch": 11.85, |
| "learning_rate": 5e-05, |
| "loss": 0.5365, |
| "step": 935 |
| }, |
| { |
| "epoch": 11.87, |
| "learning_rate": 5e-05, |
| "loss": 0.4441, |
| "step": 936 |
| }, |
| { |
| "epoch": 11.88, |
| "learning_rate": 5e-05, |
| "loss": 0.6383, |
| "step": 937 |
| }, |
| { |
| "epoch": 11.89, |
| "learning_rate": 5e-05, |
| "loss": 0.4645, |
| "step": 938 |
| }, |
| { |
| "epoch": 11.9, |
| "learning_rate": 5e-05, |
| "loss": 0.5156, |
| "step": 939 |
| }, |
| { |
| "epoch": 11.92, |
| "learning_rate": 5e-05, |
| "loss": 0.5111, |
| "step": 940 |
| }, |
| { |
| "epoch": 11.93, |
| "learning_rate": 5e-05, |
| "loss": 0.4814, |
| "step": 941 |
| }, |
| { |
| "epoch": 11.94, |
| "learning_rate": 5e-05, |
| "loss": 0.6128, |
| "step": 942 |
| }, |
| { |
| "epoch": 11.96, |
| "learning_rate": 5e-05, |
| "loss": 0.5208, |
| "step": 943 |
| }, |
| { |
| "epoch": 11.97, |
| "learning_rate": 5e-05, |
| "loss": 0.4339, |
| "step": 944 |
| }, |
| { |
| "epoch": 11.98, |
| "learning_rate": 5e-05, |
| "loss": 0.5546, |
| "step": 945 |
| }, |
| { |
| "epoch": 11.99, |
| "learning_rate": 5e-05, |
| "loss": 0.7077, |
| "step": 946 |
| }, |
| { |
| "epoch": 12.01, |
| "learning_rate": 5e-05, |
| "loss": 0.5787, |
| "step": 947 |
| }, |
| { |
| "epoch": 12.02, |
| "learning_rate": 5e-05, |
| "loss": 0.567, |
| "step": 948 |
| }, |
| { |
| "epoch": 12.03, |
| "learning_rate": 5e-05, |
| "loss": 0.4588, |
| "step": 949 |
| }, |
| { |
| "epoch": 12.04, |
| "learning_rate": 5e-05, |
| "loss": 0.4945, |
| "step": 950 |
| }, |
| { |
| "epoch": 12.06, |
| "learning_rate": 5e-05, |
| "loss": 0.4169, |
| "step": 951 |
| }, |
| { |
| "epoch": 12.07, |
| "learning_rate": 5e-05, |
| "loss": 0.4474, |
| "step": 952 |
| }, |
| { |
| "epoch": 12.08, |
| "learning_rate": 5e-05, |
| "loss": 0.4415, |
| "step": 953 |
| }, |
| { |
| "epoch": 12.1, |
| "learning_rate": 5e-05, |
| "loss": 0.4644, |
| "step": 954 |
| }, |
| { |
| "epoch": 12.11, |
| "learning_rate": 5e-05, |
| "loss": 0.4588, |
| "step": 955 |
| }, |
| { |
| "epoch": 12.12, |
| "learning_rate": 5e-05, |
| "loss": 0.3986, |
| "step": 956 |
| }, |
| { |
| "epoch": 12.13, |
| "learning_rate": 5e-05, |
| "loss": 0.4631, |
| "step": 957 |
| }, |
| { |
| "epoch": 12.15, |
| "learning_rate": 5e-05, |
| "loss": 0.4516, |
| "step": 958 |
| }, |
| { |
| "epoch": 12.16, |
| "learning_rate": 5e-05, |
| "loss": 0.5047, |
| "step": 959 |
| }, |
| { |
| "epoch": 12.17, |
| "learning_rate": 5e-05, |
| "loss": 0.4009, |
| "step": 960 |
| }, |
| { |
| "epoch": 12.18, |
| "learning_rate": 5e-05, |
| "loss": 0.4851, |
| "step": 961 |
| }, |
| { |
| "epoch": 12.2, |
| "learning_rate": 5e-05, |
| "loss": 0.4445, |
| "step": 962 |
| }, |
| { |
| "epoch": 12.21, |
| "learning_rate": 5e-05, |
| "loss": 0.4717, |
| "step": 963 |
| }, |
| { |
| "epoch": 12.22, |
| "learning_rate": 5e-05, |
| "loss": 0.4124, |
| "step": 964 |
| }, |
| { |
| "epoch": 12.23, |
| "learning_rate": 5e-05, |
| "loss": 0.4289, |
| "step": 965 |
| }, |
| { |
| "epoch": 12.25, |
| "learning_rate": 5e-05, |
| "loss": 0.5673, |
| "step": 966 |
| }, |
| { |
| "epoch": 12.26, |
| "learning_rate": 5e-05, |
| "loss": 0.4995, |
| "step": 967 |
| }, |
| { |
| "epoch": 12.27, |
| "learning_rate": 5e-05, |
| "loss": 0.5652, |
| "step": 968 |
| }, |
| { |
| "epoch": 12.29, |
| "learning_rate": 5e-05, |
| "loss": 0.5102, |
| "step": 969 |
| }, |
| { |
| "epoch": 12.3, |
| "learning_rate": 5e-05, |
| "loss": 0.527, |
| "step": 970 |
| }, |
| { |
| "epoch": 12.31, |
| "learning_rate": 5e-05, |
| "loss": 0.4239, |
| "step": 971 |
| }, |
| { |
| "epoch": 12.32, |
| "learning_rate": 5e-05, |
| "loss": 0.4427, |
| "step": 972 |
| }, |
| { |
| "epoch": 12.34, |
| "learning_rate": 5e-05, |
| "loss": 0.5881, |
| "step": 973 |
| }, |
| { |
| "epoch": 12.35, |
| "learning_rate": 5e-05, |
| "loss": 0.509, |
| "step": 974 |
| }, |
| { |
| "epoch": 12.36, |
| "learning_rate": 5e-05, |
| "loss": 0.4929, |
| "step": 975 |
| }, |
| { |
| "epoch": 12.37, |
| "learning_rate": 5e-05, |
| "loss": 0.4273, |
| "step": 976 |
| }, |
| { |
| "epoch": 12.39, |
| "learning_rate": 5e-05, |
| "loss": 0.5216, |
| "step": 977 |
| }, |
| { |
| "epoch": 12.4, |
| "learning_rate": 5e-05, |
| "loss": 0.4013, |
| "step": 978 |
| }, |
| { |
| "epoch": 12.41, |
| "learning_rate": 5e-05, |
| "loss": 0.5181, |
| "step": 979 |
| }, |
| { |
| "epoch": 12.42, |
| "learning_rate": 5e-05, |
| "loss": 0.4188, |
| "step": 980 |
| }, |
| { |
| "epoch": 12.44, |
| "learning_rate": 5e-05, |
| "loss": 0.4627, |
| "step": 981 |
| }, |
| { |
| "epoch": 12.45, |
| "learning_rate": 5e-05, |
| "loss": 0.3966, |
| "step": 982 |
| }, |
| { |
| "epoch": 12.46, |
| "learning_rate": 5e-05, |
| "loss": 0.4508, |
| "step": 983 |
| }, |
| { |
| "epoch": 12.48, |
| "learning_rate": 5e-05, |
| "loss": 0.4853, |
| "step": 984 |
| }, |
| { |
| "epoch": 12.49, |
| "learning_rate": 5e-05, |
| "loss": 0.6032, |
| "step": 985 |
| }, |
| { |
| "epoch": 12.5, |
| "learning_rate": 5e-05, |
| "loss": 0.489, |
| "step": 986 |
| }, |
| { |
| "epoch": 12.51, |
| "learning_rate": 5e-05, |
| "loss": 0.499, |
| "step": 987 |
| }, |
| { |
| "epoch": 12.53, |
| "learning_rate": 5e-05, |
| "loss": 0.4286, |
| "step": 988 |
| }, |
| { |
| "epoch": 12.54, |
| "learning_rate": 5e-05, |
| "loss": 0.3587, |
| "step": 989 |
| }, |
| { |
| "epoch": 12.55, |
| "learning_rate": 5e-05, |
| "loss": 0.5621, |
| "step": 990 |
| }, |
| { |
| "epoch": 12.56, |
| "learning_rate": 5e-05, |
| "loss": 0.4773, |
| "step": 991 |
| }, |
| { |
| "epoch": 12.58, |
| "learning_rate": 5e-05, |
| "loss": 0.4381, |
| "step": 992 |
| }, |
| { |
| "epoch": 12.59, |
| "learning_rate": 5e-05, |
| "loss": 0.3809, |
| "step": 993 |
| }, |
| { |
| "epoch": 12.6, |
| "learning_rate": 5e-05, |
| "loss": 0.4647, |
| "step": 994 |
| }, |
| { |
| "epoch": 12.61, |
| "learning_rate": 5e-05, |
| "loss": 0.435, |
| "step": 995 |
| }, |
| { |
| "epoch": 12.63, |
| "learning_rate": 5e-05, |
| "loss": 0.5375, |
| "step": 996 |
| }, |
| { |
| "epoch": 12.64, |
| "learning_rate": 5e-05, |
| "loss": 0.4973, |
| "step": 997 |
| }, |
| { |
| "epoch": 12.65, |
| "learning_rate": 5e-05, |
| "loss": 0.4236, |
| "step": 998 |
| }, |
| { |
| "epoch": 12.67, |
| "learning_rate": 5e-05, |
| "loss": 0.4384, |
| "step": 999 |
| }, |
| { |
| "epoch": 12.68, |
| "learning_rate": 5e-05, |
| "loss": 0.3213, |
| "step": 1000 |
| }, |
| { |
| "epoch": 12.69, |
| "learning_rate": 5e-05, |
| "loss": 0.5907, |
| "step": 1001 |
| }, |
| { |
| "epoch": 12.7, |
| "learning_rate": 5e-05, |
| "loss": 0.485, |
| "step": 1002 |
| }, |
| { |
| "epoch": 12.72, |
| "learning_rate": 5e-05, |
| "loss": 0.4278, |
| "step": 1003 |
| }, |
| { |
| "epoch": 12.73, |
| "learning_rate": 5e-05, |
| "loss": 0.4992, |
| "step": 1004 |
| }, |
| { |
| "epoch": 12.74, |
| "learning_rate": 5e-05, |
| "loss": 0.6232, |
| "step": 1005 |
| }, |
| { |
| "epoch": 12.75, |
| "learning_rate": 5e-05, |
| "loss": 0.4068, |
| "step": 1006 |
| }, |
| { |
| "epoch": 12.77, |
| "learning_rate": 5e-05, |
| "loss": 0.3576, |
| "step": 1007 |
| }, |
| { |
| "epoch": 12.78, |
| "learning_rate": 5e-05, |
| "loss": 0.599, |
| "step": 1008 |
| }, |
| { |
| "epoch": 12.79, |
| "learning_rate": 5e-05, |
| "loss": 0.4189, |
| "step": 1009 |
| }, |
| { |
| "epoch": 12.81, |
| "learning_rate": 5e-05, |
| "loss": 0.4901, |
| "step": 1010 |
| }, |
| { |
| "epoch": 12.82, |
| "learning_rate": 5e-05, |
| "loss": 0.5568, |
| "step": 1011 |
| }, |
| { |
| "epoch": 12.83, |
| "learning_rate": 5e-05, |
| "loss": 0.5632, |
| "step": 1012 |
| }, |
| { |
| "epoch": 12.84, |
| "learning_rate": 5e-05, |
| "loss": 0.6059, |
| "step": 1013 |
| }, |
| { |
| "epoch": 12.86, |
| "learning_rate": 5e-05, |
| "loss": 0.4592, |
| "step": 1014 |
| }, |
| { |
| "epoch": 12.87, |
| "learning_rate": 5e-05, |
| "loss": 0.478, |
| "step": 1015 |
| }, |
| { |
| "epoch": 12.88, |
| "learning_rate": 5e-05, |
| "loss": 0.5699, |
| "step": 1016 |
| }, |
| { |
| "epoch": 12.89, |
| "learning_rate": 5e-05, |
| "loss": 0.3658, |
| "step": 1017 |
| }, |
| { |
| "epoch": 12.91, |
| "learning_rate": 5e-05, |
| "loss": 0.5251, |
| "step": 1018 |
| }, |
| { |
| "epoch": 12.92, |
| "learning_rate": 5e-05, |
| "loss": 0.4586, |
| "step": 1019 |
| }, |
| { |
| "epoch": 12.93, |
| "learning_rate": 5e-05, |
| "loss": 0.5217, |
| "step": 1020 |
| }, |
| { |
| "epoch": 12.94, |
| "learning_rate": 5e-05, |
| "loss": 0.5115, |
| "step": 1021 |
| }, |
| { |
| "epoch": 12.96, |
| "learning_rate": 5e-05, |
| "loss": 0.4744, |
| "step": 1022 |
| }, |
| { |
| "epoch": 12.97, |
| "learning_rate": 5e-05, |
| "loss": 0.4892, |
| "step": 1023 |
| }, |
| { |
| "epoch": 12.98, |
| "learning_rate": 5e-05, |
| "loss": 0.6366, |
| "step": 1024 |
| }, |
| { |
| "epoch": 13.0, |
| "learning_rate": 5e-05, |
| "loss": 0.4354, |
| "step": 1025 |
| }, |
| { |
| "epoch": 13.01, |
| "learning_rate": 5e-05, |
| "loss": 0.447, |
| "step": 1026 |
| }, |
| { |
| "epoch": 13.02, |
| "learning_rate": 5e-05, |
| "loss": 0.3776, |
| "step": 1027 |
| }, |
| { |
| "epoch": 13.03, |
| "learning_rate": 5e-05, |
| "loss": 0.401, |
| "step": 1028 |
| }, |
| { |
| "epoch": 13.05, |
| "learning_rate": 5e-05, |
| "loss": 0.4575, |
| "step": 1029 |
| }, |
| { |
| "epoch": 13.06, |
| "learning_rate": 5e-05, |
| "loss": 0.4085, |
| "step": 1030 |
| }, |
| { |
| "epoch": 13.07, |
| "learning_rate": 5e-05, |
| "loss": 0.4752, |
| "step": 1031 |
| }, |
| { |
| "epoch": 13.08, |
| "learning_rate": 5e-05, |
| "loss": 0.3523, |
| "step": 1032 |
| }, |
| { |
| "epoch": 13.1, |
| "learning_rate": 5e-05, |
| "loss": 0.5191, |
| "step": 1033 |
| }, |
| { |
| "epoch": 13.11, |
| "learning_rate": 5e-05, |
| "loss": 0.3837, |
| "step": 1034 |
| }, |
| { |
| "epoch": 13.12, |
| "learning_rate": 5e-05, |
| "loss": 0.4144, |
| "step": 1035 |
| }, |
| { |
| "epoch": 13.13, |
| "learning_rate": 5e-05, |
| "loss": 0.3981, |
| "step": 1036 |
| }, |
| { |
| "epoch": 13.15, |
| "learning_rate": 5e-05, |
| "loss": 0.3832, |
| "step": 1037 |
| }, |
| { |
| "epoch": 13.16, |
| "learning_rate": 5e-05, |
| "loss": 0.6293, |
| "step": 1038 |
| }, |
| { |
| "epoch": 13.17, |
| "learning_rate": 5e-05, |
| "loss": 0.4895, |
| "step": 1039 |
| }, |
| { |
| "epoch": 13.19, |
| "learning_rate": 5e-05, |
| "loss": 0.4003, |
| "step": 1040 |
| }, |
| { |
| "epoch": 13.2, |
| "learning_rate": 5e-05, |
| "loss": 0.5658, |
| "step": 1041 |
| }, |
| { |
| "epoch": 13.21, |
| "learning_rate": 5e-05, |
| "loss": 0.4552, |
| "step": 1042 |
| }, |
| { |
| "epoch": 13.22, |
| "learning_rate": 5e-05, |
| "loss": 0.4279, |
| "step": 1043 |
| }, |
| { |
| "epoch": 13.24, |
| "learning_rate": 5e-05, |
| "loss": 0.3837, |
| "step": 1044 |
| }, |
| { |
| "epoch": 13.25, |
| "learning_rate": 5e-05, |
| "loss": 0.4808, |
| "step": 1045 |
| }, |
| { |
| "epoch": 13.26, |
| "learning_rate": 5e-05, |
| "loss": 0.4669, |
| "step": 1046 |
| }, |
| { |
| "epoch": 13.27, |
| "learning_rate": 5e-05, |
| "loss": 0.3425, |
| "step": 1047 |
| }, |
| { |
| "epoch": 13.29, |
| "learning_rate": 5e-05, |
| "loss": 0.372, |
| "step": 1048 |
| }, |
| { |
| "epoch": 13.3, |
| "learning_rate": 5e-05, |
| "loss": 0.4972, |
| "step": 1049 |
| }, |
| { |
| "epoch": 13.31, |
| "learning_rate": 5e-05, |
| "loss": 0.4734, |
| "step": 1050 |
| }, |
| { |
| "epoch": 13.32, |
| "learning_rate": 5e-05, |
| "loss": 0.444, |
| "step": 1051 |
| }, |
| { |
| "epoch": 13.34, |
| "learning_rate": 5e-05, |
| "loss": 0.4908, |
| "step": 1052 |
| }, |
| { |
| "epoch": 13.35, |
| "learning_rate": 5e-05, |
| "loss": 0.5162, |
| "step": 1053 |
| }, |
| { |
| "epoch": 13.36, |
| "learning_rate": 5e-05, |
| "loss": 0.3781, |
| "step": 1054 |
| }, |
| { |
| "epoch": 13.38, |
| "learning_rate": 5e-05, |
| "loss": 0.4343, |
| "step": 1055 |
| }, |
| { |
| "epoch": 13.39, |
| "learning_rate": 5e-05, |
| "loss": 0.3756, |
| "step": 1056 |
| }, |
| { |
| "epoch": 13.4, |
| "learning_rate": 5e-05, |
| "loss": 0.5252, |
| "step": 1057 |
| }, |
| { |
| "epoch": 13.41, |
| "learning_rate": 5e-05, |
| "loss": 0.4383, |
| "step": 1058 |
| }, |
| { |
| "epoch": 13.43, |
| "learning_rate": 5e-05, |
| "loss": 0.4851, |
| "step": 1059 |
| }, |
| { |
| "epoch": 13.44, |
| "learning_rate": 5e-05, |
| "loss": 0.4927, |
| "step": 1060 |
| }, |
| { |
| "epoch": 13.45, |
| "learning_rate": 5e-05, |
| "loss": 0.3646, |
| "step": 1061 |
| }, |
| { |
| "epoch": 13.46, |
| "learning_rate": 5e-05, |
| "loss": 0.5748, |
| "step": 1062 |
| }, |
| { |
| "epoch": 13.48, |
| "learning_rate": 5e-05, |
| "loss": 0.4436, |
| "step": 1063 |
| }, |
| { |
| "epoch": 13.49, |
| "learning_rate": 5e-05, |
| "loss": 0.4607, |
| "step": 1064 |
| }, |
| { |
| "epoch": 13.5, |
| "learning_rate": 5e-05, |
| "loss": 0.4718, |
| "step": 1065 |
| }, |
| { |
| "epoch": 13.52, |
| "learning_rate": 5e-05, |
| "loss": 0.3709, |
| "step": 1066 |
| }, |
| { |
| "epoch": 13.53, |
| "learning_rate": 5e-05, |
| "loss": 0.4008, |
| "step": 1067 |
| }, |
| { |
| "epoch": 13.54, |
| "learning_rate": 5e-05, |
| "loss": 0.4224, |
| "step": 1068 |
| }, |
| { |
| "epoch": 13.55, |
| "learning_rate": 5e-05, |
| "loss": 0.4314, |
| "step": 1069 |
| }, |
| { |
| "epoch": 13.57, |
| "learning_rate": 5e-05, |
| "loss": 0.5238, |
| "step": 1070 |
| }, |
| { |
| "epoch": 13.58, |
| "learning_rate": 5e-05, |
| "loss": 0.6027, |
| "step": 1071 |
| }, |
| { |
| "epoch": 13.59, |
| "learning_rate": 5e-05, |
| "loss": 0.4447, |
| "step": 1072 |
| }, |
| { |
| "epoch": 13.6, |
| "learning_rate": 5e-05, |
| "loss": 0.4908, |
| "step": 1073 |
| }, |
| { |
| "epoch": 13.62, |
| "learning_rate": 5e-05, |
| "loss": 0.4072, |
| "step": 1074 |
| }, |
| { |
| "epoch": 13.63, |
| "learning_rate": 5e-05, |
| "loss": 0.2766, |
| "step": 1075 |
| }, |
| { |
| "epoch": 13.64, |
| "learning_rate": 5e-05, |
| "loss": 0.5846, |
| "step": 1076 |
| }, |
| { |
| "epoch": 13.65, |
| "learning_rate": 5e-05, |
| "loss": 0.5032, |
| "step": 1077 |
| }, |
| { |
| "epoch": 13.67, |
| "learning_rate": 5e-05, |
| "loss": 0.4969, |
| "step": 1078 |
| }, |
| { |
| "epoch": 13.68, |
| "learning_rate": 5e-05, |
| "loss": 0.4611, |
| "step": 1079 |
| }, |
| { |
| "epoch": 13.69, |
| "learning_rate": 5e-05, |
| "loss": 0.4005, |
| "step": 1080 |
| }, |
| { |
| "epoch": 13.71, |
| "learning_rate": 5e-05, |
| "loss": 0.4703, |
| "step": 1081 |
| }, |
| { |
| "epoch": 13.72, |
| "learning_rate": 5e-05, |
| "loss": 0.5163, |
| "step": 1082 |
| }, |
| { |
| "epoch": 13.73, |
| "learning_rate": 5e-05, |
| "loss": 0.4598, |
| "step": 1083 |
| }, |
| { |
| "epoch": 13.74, |
| "learning_rate": 5e-05, |
| "loss": 0.4114, |
| "step": 1084 |
| }, |
| { |
| "epoch": 13.76, |
| "learning_rate": 5e-05, |
| "loss": 0.3698, |
| "step": 1085 |
| }, |
| { |
| "epoch": 13.77, |
| "learning_rate": 5e-05, |
| "loss": 0.3678, |
| "step": 1086 |
| }, |
| { |
| "epoch": 13.78, |
| "learning_rate": 5e-05, |
| "loss": 0.5077, |
| "step": 1087 |
| }, |
| { |
| "epoch": 13.79, |
| "learning_rate": 5e-05, |
| "loss": 0.4167, |
| "step": 1088 |
| }, |
| { |
| "epoch": 13.81, |
| "learning_rate": 5e-05, |
| "loss": 0.4406, |
| "step": 1089 |
| }, |
| { |
| "epoch": 13.82, |
| "learning_rate": 5e-05, |
| "loss": 0.4691, |
| "step": 1090 |
| }, |
| { |
| "epoch": 13.83, |
| "learning_rate": 5e-05, |
| "loss": 0.5019, |
| "step": 1091 |
| }, |
| { |
| "epoch": 13.84, |
| "learning_rate": 5e-05, |
| "loss": 0.4387, |
| "step": 1092 |
| }, |
| { |
| "epoch": 13.86, |
| "learning_rate": 5e-05, |
| "loss": 0.5331, |
| "step": 1093 |
| }, |
| { |
| "epoch": 13.87, |
| "learning_rate": 5e-05, |
| "loss": 0.4941, |
| "step": 1094 |
| }, |
| { |
| "epoch": 13.88, |
| "learning_rate": 5e-05, |
| "loss": 0.4659, |
| "step": 1095 |
| }, |
| { |
| "epoch": 13.9, |
| "learning_rate": 5e-05, |
| "loss": 0.4521, |
| "step": 1096 |
| }, |
| { |
| "epoch": 13.91, |
| "learning_rate": 5e-05, |
| "loss": 0.4481, |
| "step": 1097 |
| }, |
| { |
| "epoch": 13.92, |
| "learning_rate": 5e-05, |
| "loss": 0.499, |
| "step": 1098 |
| }, |
| { |
| "epoch": 13.93, |
| "learning_rate": 5e-05, |
| "loss": 0.4409, |
| "step": 1099 |
| }, |
| { |
| "epoch": 13.95, |
| "learning_rate": 5e-05, |
| "loss": 0.4139, |
| "step": 1100 |
| }, |
| { |
| "epoch": 13.96, |
| "learning_rate": 5e-05, |
| "loss": 0.5755, |
| "step": 1101 |
| }, |
| { |
| "epoch": 13.97, |
| "learning_rate": 5e-05, |
| "loss": 0.4096, |
| "step": 1102 |
| }, |
| { |
| "epoch": 13.98, |
| "learning_rate": 5e-05, |
| "loss": 0.4603, |
| "step": 1103 |
| }, |
| { |
| "epoch": 14.0, |
| "learning_rate": 5e-05, |
| "loss": 0.398, |
| "step": 1104 |
| }, |
| { |
| "epoch": 14.01, |
| "learning_rate": 5e-05, |
| "loss": 0.4625, |
| "step": 1105 |
| }, |
| { |
| "epoch": 14.02, |
| "learning_rate": 5e-05, |
| "loss": 0.4837, |
| "step": 1106 |
| }, |
| { |
| "epoch": 14.03, |
| "learning_rate": 5e-05, |
| "loss": 0.4783, |
| "step": 1107 |
| }, |
| { |
| "epoch": 14.05, |
| "learning_rate": 5e-05, |
| "loss": 0.3399, |
| "step": 1108 |
| }, |
| { |
| "epoch": 14.06, |
| "learning_rate": 5e-05, |
| "loss": 0.5306, |
| "step": 1109 |
| }, |
| { |
| "epoch": 14.07, |
| "learning_rate": 5e-05, |
| "loss": 0.3891, |
| "step": 1110 |
| }, |
| { |
| "epoch": 14.09, |
| "learning_rate": 5e-05, |
| "loss": 0.4706, |
| "step": 1111 |
| }, |
| { |
| "epoch": 14.1, |
| "learning_rate": 5e-05, |
| "loss": 0.3787, |
| "step": 1112 |
| }, |
| { |
| "epoch": 14.11, |
| "learning_rate": 5e-05, |
| "loss": 0.3711, |
| "step": 1113 |
| }, |
| { |
| "epoch": 14.12, |
| "learning_rate": 5e-05, |
| "loss": 0.4506, |
| "step": 1114 |
| }, |
| { |
| "epoch": 14.14, |
| "learning_rate": 5e-05, |
| "loss": 0.4384, |
| "step": 1115 |
| }, |
| { |
| "epoch": 14.15, |
| "learning_rate": 5e-05, |
| "loss": 0.3459, |
| "step": 1116 |
| }, |
| { |
| "epoch": 14.16, |
| "learning_rate": 5e-05, |
| "loss": 0.286, |
| "step": 1117 |
| }, |
| { |
| "epoch": 14.17, |
| "learning_rate": 5e-05, |
| "loss": 0.4046, |
| "step": 1118 |
| }, |
| { |
| "epoch": 14.19, |
| "learning_rate": 5e-05, |
| "loss": 0.3749, |
| "step": 1119 |
| }, |
| { |
| "epoch": 14.2, |
| "learning_rate": 5e-05, |
| "loss": 0.4883, |
| "step": 1120 |
| }, |
| { |
| "epoch": 14.21, |
| "learning_rate": 5e-05, |
| "loss": 0.4656, |
| "step": 1121 |
| }, |
| { |
| "epoch": 14.23, |
| "learning_rate": 5e-05, |
| "loss": 0.5804, |
| "step": 1122 |
| }, |
| { |
| "epoch": 14.24, |
| "learning_rate": 5e-05, |
| "loss": 0.4372, |
| "step": 1123 |
| }, |
| { |
| "epoch": 14.25, |
| "learning_rate": 5e-05, |
| "loss": 0.3496, |
| "step": 1124 |
| }, |
| { |
| "epoch": 14.26, |
| "learning_rate": 5e-05, |
| "loss": 0.3694, |
| "step": 1125 |
| }, |
| { |
| "epoch": 14.28, |
| "learning_rate": 5e-05, |
| "loss": 0.4647, |
| "step": 1126 |
| }, |
| { |
| "epoch": 14.29, |
| "learning_rate": 5e-05, |
| "loss": 0.2855, |
| "step": 1127 |
| }, |
| { |
| "epoch": 14.3, |
| "learning_rate": 5e-05, |
| "loss": 0.4518, |
| "step": 1128 |
| }, |
| { |
| "epoch": 14.31, |
| "learning_rate": 5e-05, |
| "loss": 0.5026, |
| "step": 1129 |
| }, |
| { |
| "epoch": 14.33, |
| "learning_rate": 5e-05, |
| "loss": 0.3875, |
| "step": 1130 |
| }, |
| { |
| "epoch": 14.34, |
| "learning_rate": 5e-05, |
| "loss": 0.4012, |
| "step": 1131 |
| }, |
| { |
| "epoch": 14.35, |
| "learning_rate": 5e-05, |
| "loss": 0.3976, |
| "step": 1132 |
| }, |
| { |
| "epoch": 14.36, |
| "learning_rate": 5e-05, |
| "loss": 0.4651, |
| "step": 1133 |
| }, |
| { |
| "epoch": 14.38, |
| "learning_rate": 5e-05, |
| "loss": 0.4469, |
| "step": 1134 |
| }, |
| { |
| "epoch": 14.39, |
| "learning_rate": 5e-05, |
| "loss": 0.2851, |
| "step": 1135 |
| }, |
| { |
| "epoch": 14.4, |
| "learning_rate": 5e-05, |
| "loss": 0.3696, |
| "step": 1136 |
| }, |
| { |
| "epoch": 14.42, |
| "learning_rate": 5e-05, |
| "loss": 0.4549, |
| "step": 1137 |
| }, |
| { |
| "epoch": 14.43, |
| "learning_rate": 5e-05, |
| "loss": 0.3546, |
| "step": 1138 |
| }, |
| { |
| "epoch": 14.44, |
| "learning_rate": 5e-05, |
| "loss": 0.4238, |
| "step": 1139 |
| }, |
| { |
| "epoch": 14.45, |
| "learning_rate": 5e-05, |
| "loss": 0.4918, |
| "step": 1140 |
| }, |
| { |
| "epoch": 14.47, |
| "learning_rate": 5e-05, |
| "loss": 0.4539, |
| "step": 1141 |
| }, |
| { |
| "epoch": 14.48, |
| "learning_rate": 5e-05, |
| "loss": 0.5216, |
| "step": 1142 |
| }, |
| { |
| "epoch": 14.49, |
| "learning_rate": 5e-05, |
| "loss": 0.4762, |
| "step": 1143 |
| }, |
| { |
| "epoch": 14.5, |
| "learning_rate": 5e-05, |
| "loss": 0.3103, |
| "step": 1144 |
| }, |
| { |
| "epoch": 14.52, |
| "learning_rate": 5e-05, |
| "loss": 0.4122, |
| "step": 1145 |
| }, |
| { |
| "epoch": 14.53, |
| "learning_rate": 5e-05, |
| "loss": 0.3763, |
| "step": 1146 |
| }, |
| { |
| "epoch": 14.54, |
| "learning_rate": 5e-05, |
| "loss": 0.339, |
| "step": 1147 |
| }, |
| { |
| "epoch": 14.55, |
| "learning_rate": 5e-05, |
| "loss": 0.4981, |
| "step": 1148 |
| }, |
| { |
| "epoch": 14.57, |
| "learning_rate": 5e-05, |
| "loss": 0.5468, |
| "step": 1149 |
| }, |
| { |
| "epoch": 14.58, |
| "learning_rate": 5e-05, |
| "loss": 0.3848, |
| "step": 1150 |
| }, |
| { |
| "epoch": 14.59, |
| "learning_rate": 5e-05, |
| "loss": 0.3347, |
| "step": 1151 |
| }, |
| { |
| "epoch": 14.61, |
| "learning_rate": 5e-05, |
| "loss": 0.4657, |
| "step": 1152 |
| }, |
| { |
| "epoch": 14.62, |
| "learning_rate": 5e-05, |
| "loss": 0.3211, |
| "step": 1153 |
| }, |
| { |
| "epoch": 14.63, |
| "learning_rate": 5e-05, |
| "loss": 0.4712, |
| "step": 1154 |
| }, |
| { |
| "epoch": 14.64, |
| "learning_rate": 5e-05, |
| "loss": 0.4557, |
| "step": 1155 |
| }, |
| { |
| "epoch": 14.66, |
| "learning_rate": 5e-05, |
| "loss": 0.4152, |
| "step": 1156 |
| }, |
| { |
| "epoch": 14.67, |
| "learning_rate": 5e-05, |
| "loss": 0.3626, |
| "step": 1157 |
| }, |
| { |
| "epoch": 14.68, |
| "learning_rate": 5e-05, |
| "loss": 0.4295, |
| "step": 1158 |
| }, |
| { |
| "epoch": 14.69, |
| "learning_rate": 5e-05, |
| "loss": 0.5018, |
| "step": 1159 |
| }, |
| { |
| "epoch": 14.71, |
| "learning_rate": 5e-05, |
| "loss": 0.3661, |
| "step": 1160 |
| }, |
| { |
| "epoch": 14.72, |
| "learning_rate": 5e-05, |
| "loss": 0.3922, |
| "step": 1161 |
| }, |
| { |
| "epoch": 14.73, |
| "learning_rate": 5e-05, |
| "loss": 0.3854, |
| "step": 1162 |
| }, |
| { |
| "epoch": 14.74, |
| "learning_rate": 5e-05, |
| "loss": 0.4796, |
| "step": 1163 |
| }, |
| { |
| "epoch": 14.76, |
| "learning_rate": 5e-05, |
| "loss": 0.4918, |
| "step": 1164 |
| }, |
| { |
| "epoch": 14.77, |
| "learning_rate": 5e-05, |
| "loss": 0.3889, |
| "step": 1165 |
| }, |
| { |
| "epoch": 14.78, |
| "learning_rate": 5e-05, |
| "loss": 0.454, |
| "step": 1166 |
| }, |
| { |
| "epoch": 14.8, |
| "learning_rate": 5e-05, |
| "loss": 0.4411, |
| "step": 1167 |
| }, |
| { |
| "epoch": 14.81, |
| "learning_rate": 5e-05, |
| "loss": 0.4061, |
| "step": 1168 |
| }, |
| { |
| "epoch": 14.82, |
| "learning_rate": 5e-05, |
| "loss": 0.456, |
| "step": 1169 |
| }, |
| { |
| "epoch": 14.83, |
| "learning_rate": 5e-05, |
| "loss": 0.522, |
| "step": 1170 |
| }, |
| { |
| "epoch": 14.85, |
| "learning_rate": 5e-05, |
| "loss": 0.4226, |
| "step": 1171 |
| }, |
| { |
| "epoch": 14.86, |
| "learning_rate": 5e-05, |
| "loss": 0.4414, |
| "step": 1172 |
| }, |
| { |
| "epoch": 14.87, |
| "learning_rate": 5e-05, |
| "loss": 0.5507, |
| "step": 1173 |
| }, |
| { |
| "epoch": 14.88, |
| "learning_rate": 5e-05, |
| "loss": 0.4131, |
| "step": 1174 |
| }, |
| { |
| "epoch": 14.9, |
| "learning_rate": 5e-05, |
| "loss": 0.4105, |
| "step": 1175 |
| }, |
| { |
| "epoch": 14.91, |
| "learning_rate": 5e-05, |
| "loss": 0.5714, |
| "step": 1176 |
| }, |
| { |
| "epoch": 14.92, |
| "learning_rate": 5e-05, |
| "loss": 0.4392, |
| "step": 1177 |
| }, |
| { |
| "epoch": 14.94, |
| "learning_rate": 5e-05, |
| "loss": 0.2773, |
| "step": 1178 |
| }, |
| { |
| "epoch": 14.95, |
| "learning_rate": 5e-05, |
| "loss": 0.4836, |
| "step": 1179 |
| }, |
| { |
| "epoch": 14.96, |
| "learning_rate": 5e-05, |
| "loss": 0.3461, |
| "step": 1180 |
| }, |
| { |
| "epoch": 14.97, |
| "learning_rate": 5e-05, |
| "loss": 0.5077, |
| "step": 1181 |
| }, |
| { |
| "epoch": 14.99, |
| "learning_rate": 5e-05, |
| "loss": 0.3795, |
| "step": 1182 |
| }, |
| { |
| "epoch": 15.0, |
| "learning_rate": 5e-05, |
| "loss": 0.4307, |
| "step": 1183 |
| }, |
| { |
| "epoch": 15.01, |
| "learning_rate": 5e-05, |
| "loss": 0.366, |
| "step": 1184 |
| }, |
| { |
| "epoch": 15.02, |
| "learning_rate": 5e-05, |
| "loss": 0.4216, |
| "step": 1185 |
| }, |
| { |
| "epoch": 15.04, |
| "learning_rate": 5e-05, |
| "loss": 0.3915, |
| "step": 1186 |
| }, |
| { |
| "epoch": 15.05, |
| "learning_rate": 5e-05, |
| "loss": 0.4084, |
| "step": 1187 |
| }, |
| { |
| "epoch": 15.06, |
| "learning_rate": 5e-05, |
| "loss": 0.3265, |
| "step": 1188 |
| }, |
| { |
| "epoch": 15.07, |
| "learning_rate": 5e-05, |
| "loss": 0.4896, |
| "step": 1189 |
| }, |
| { |
| "epoch": 15.09, |
| "learning_rate": 5e-05, |
| "loss": 0.3416, |
| "step": 1190 |
| }, |
| { |
| "epoch": 15.1, |
| "learning_rate": 5e-05, |
| "loss": 0.4238, |
| "step": 1191 |
| }, |
| { |
| "epoch": 15.11, |
| "learning_rate": 5e-05, |
| "loss": 0.4288, |
| "step": 1192 |
| }, |
| { |
| "epoch": 15.13, |
| "learning_rate": 5e-05, |
| "loss": 0.4676, |
| "step": 1193 |
| }, |
| { |
| "epoch": 15.14, |
| "learning_rate": 5e-05, |
| "loss": 0.3879, |
| "step": 1194 |
| }, |
| { |
| "epoch": 15.15, |
| "learning_rate": 5e-05, |
| "loss": 0.3602, |
| "step": 1195 |
| }, |
| { |
| "epoch": 15.16, |
| "learning_rate": 5e-05, |
| "loss": 0.37, |
| "step": 1196 |
| }, |
| { |
| "epoch": 15.18, |
| "learning_rate": 5e-05, |
| "loss": 0.36, |
| "step": 1197 |
| }, |
| { |
| "epoch": 15.19, |
| "learning_rate": 5e-05, |
| "loss": 0.3246, |
| "step": 1198 |
| }, |
| { |
| "epoch": 15.2, |
| "learning_rate": 5e-05, |
| "loss": 0.3777, |
| "step": 1199 |
| }, |
| { |
| "epoch": 15.21, |
| "learning_rate": 5e-05, |
| "loss": 0.334, |
| "step": 1200 |
| }, |
| { |
| "epoch": 15.23, |
| "learning_rate": 5e-05, |
| "loss": 0.3737, |
| "step": 1201 |
| }, |
| { |
| "epoch": 15.24, |
| "learning_rate": 5e-05, |
| "loss": 0.4866, |
| "step": 1202 |
| }, |
| { |
| "epoch": 15.25, |
| "learning_rate": 5e-05, |
| "loss": 0.4264, |
| "step": 1203 |
| }, |
| { |
| "epoch": 15.26, |
| "learning_rate": 5e-05, |
| "loss": 0.3854, |
| "step": 1204 |
| }, |
| { |
| "epoch": 15.28, |
| "learning_rate": 5e-05, |
| "loss": 0.3793, |
| "step": 1205 |
| }, |
| { |
| "epoch": 15.29, |
| "learning_rate": 5e-05, |
| "loss": 0.3911, |
| "step": 1206 |
| }, |
| { |
| "epoch": 15.3, |
| "learning_rate": 5e-05, |
| "loss": 0.4112, |
| "step": 1207 |
| }, |
| { |
| "epoch": 15.32, |
| "learning_rate": 5e-05, |
| "loss": 0.4049, |
| "step": 1208 |
| }, |
| { |
| "epoch": 15.33, |
| "learning_rate": 5e-05, |
| "loss": 0.3316, |
| "step": 1209 |
| }, |
| { |
| "epoch": 15.34, |
| "learning_rate": 5e-05, |
| "loss": 0.3827, |
| "step": 1210 |
| }, |
| { |
| "epoch": 15.35, |
| "learning_rate": 5e-05, |
| "loss": 0.3905, |
| "step": 1211 |
| }, |
| { |
| "epoch": 15.37, |
| "learning_rate": 5e-05, |
| "loss": 0.4163, |
| "step": 1212 |
| }, |
| { |
| "epoch": 15.38, |
| "learning_rate": 5e-05, |
| "loss": 0.3945, |
| "step": 1213 |
| }, |
| { |
| "epoch": 15.39, |
| "learning_rate": 5e-05, |
| "loss": 0.5054, |
| "step": 1214 |
| }, |
| { |
| "epoch": 15.4, |
| "learning_rate": 5e-05, |
| "loss": 0.3017, |
| "step": 1215 |
| }, |
| { |
| "epoch": 15.42, |
| "learning_rate": 5e-05, |
| "loss": 0.347, |
| "step": 1216 |
| }, |
| { |
| "epoch": 15.43, |
| "learning_rate": 5e-05, |
| "loss": 0.3823, |
| "step": 1217 |
| }, |
| { |
| "epoch": 15.44, |
| "learning_rate": 5e-05, |
| "loss": 0.4482, |
| "step": 1218 |
| }, |
| { |
| "epoch": 15.45, |
| "learning_rate": 5e-05, |
| "loss": 0.4606, |
| "step": 1219 |
| }, |
| { |
| "epoch": 15.47, |
| "learning_rate": 5e-05, |
| "loss": 0.3733, |
| "step": 1220 |
| }, |
| { |
| "epoch": 15.48, |
| "learning_rate": 5e-05, |
| "loss": 0.4246, |
| "step": 1221 |
| }, |
| { |
| "epoch": 15.49, |
| "learning_rate": 5e-05, |
| "loss": 0.4252, |
| "step": 1222 |
| }, |
| { |
| "epoch": 15.51, |
| "learning_rate": 5e-05, |
| "loss": 0.2923, |
| "step": 1223 |
| }, |
| { |
| "epoch": 15.52, |
| "learning_rate": 5e-05, |
| "loss": 0.3988, |
| "step": 1224 |
| }, |
| { |
| "epoch": 15.53, |
| "learning_rate": 5e-05, |
| "loss": 0.4172, |
| "step": 1225 |
| }, |
| { |
| "epoch": 15.54, |
| "learning_rate": 5e-05, |
| "loss": 0.459, |
| "step": 1226 |
| }, |
| { |
| "epoch": 15.56, |
| "learning_rate": 5e-05, |
| "loss": 0.4096, |
| "step": 1227 |
| }, |
| { |
| "epoch": 15.57, |
| "learning_rate": 5e-05, |
| "loss": 0.3934, |
| "step": 1228 |
| }, |
| { |
| "epoch": 15.58, |
| "learning_rate": 5e-05, |
| "loss": 0.3579, |
| "step": 1229 |
| }, |
| { |
| "epoch": 15.59, |
| "learning_rate": 5e-05, |
| "loss": 0.3139, |
| "step": 1230 |
| }, |
| { |
| "epoch": 15.61, |
| "learning_rate": 5e-05, |
| "loss": 0.4081, |
| "step": 1231 |
| }, |
| { |
| "epoch": 15.62, |
| "learning_rate": 5e-05, |
| "loss": 0.4139, |
| "step": 1232 |
| }, |
| { |
| "epoch": 15.63, |
| "learning_rate": 5e-05, |
| "loss": 0.4492, |
| "step": 1233 |
| }, |
| { |
| "epoch": 15.65, |
| "learning_rate": 5e-05, |
| "loss": 0.3481, |
| "step": 1234 |
| }, |
| { |
| "epoch": 15.66, |
| "learning_rate": 5e-05, |
| "loss": 0.4285, |
| "step": 1235 |
| }, |
| { |
| "epoch": 15.67, |
| "learning_rate": 5e-05, |
| "loss": 0.4883, |
| "step": 1236 |
| }, |
| { |
| "epoch": 15.68, |
| "learning_rate": 5e-05, |
| "loss": 0.4801, |
| "step": 1237 |
| }, |
| { |
| "epoch": 15.7, |
| "learning_rate": 5e-05, |
| "loss": 0.3815, |
| "step": 1238 |
| }, |
| { |
| "epoch": 15.71, |
| "learning_rate": 5e-05, |
| "loss": 0.3642, |
| "step": 1239 |
| }, |
| { |
| "epoch": 15.72, |
| "learning_rate": 5e-05, |
| "loss": 0.4413, |
| "step": 1240 |
| }, |
| { |
| "epoch": 15.73, |
| "learning_rate": 5e-05, |
| "loss": 0.3482, |
| "step": 1241 |
| }, |
| { |
| "epoch": 15.75, |
| "learning_rate": 5e-05, |
| "loss": 0.3517, |
| "step": 1242 |
| }, |
| { |
| "epoch": 15.76, |
| "learning_rate": 5e-05, |
| "loss": 0.3798, |
| "step": 1243 |
| }, |
| { |
| "epoch": 15.77, |
| "learning_rate": 5e-05, |
| "loss": 0.3234, |
| "step": 1244 |
| }, |
| { |
| "epoch": 15.78, |
| "learning_rate": 5e-05, |
| "loss": 0.4371, |
| "step": 1245 |
| }, |
| { |
| "epoch": 15.8, |
| "learning_rate": 5e-05, |
| "loss": 0.4482, |
| "step": 1246 |
| }, |
| { |
| "epoch": 15.81, |
| "learning_rate": 5e-05, |
| "loss": 0.327, |
| "step": 1247 |
| }, |
| { |
| "epoch": 15.82, |
| "learning_rate": 5e-05, |
| "loss": 0.4378, |
| "step": 1248 |
| }, |
| { |
| "epoch": 15.84, |
| "learning_rate": 5e-05, |
| "loss": 0.4441, |
| "step": 1249 |
| }, |
| { |
| "epoch": 15.85, |
| "learning_rate": 5e-05, |
| "loss": 0.4397, |
| "step": 1250 |
| }, |
| { |
| "epoch": 15.86, |
| "learning_rate": 5e-05, |
| "loss": 0.5376, |
| "step": 1251 |
| }, |
| { |
| "epoch": 15.87, |
| "learning_rate": 5e-05, |
| "loss": 0.4697, |
| "step": 1252 |
| }, |
| { |
| "epoch": 15.89, |
| "learning_rate": 5e-05, |
| "loss": 0.4065, |
| "step": 1253 |
| }, |
| { |
| "epoch": 15.9, |
| "learning_rate": 5e-05, |
| "loss": 0.3276, |
| "step": 1254 |
| }, |
| { |
| "epoch": 15.91, |
| "learning_rate": 5e-05, |
| "loss": 0.4976, |
| "step": 1255 |
| }, |
| { |
| "epoch": 15.92, |
| "learning_rate": 5e-05, |
| "loss": 0.4434, |
| "step": 1256 |
| }, |
| { |
| "epoch": 15.94, |
| "learning_rate": 5e-05, |
| "loss": 0.3519, |
| "step": 1257 |
| }, |
| { |
| "epoch": 15.95, |
| "learning_rate": 5e-05, |
| "loss": 0.2603, |
| "step": 1258 |
| }, |
| { |
| "epoch": 15.96, |
| "learning_rate": 5e-05, |
| "loss": 0.465, |
| "step": 1259 |
| }, |
| { |
| "epoch": 15.97, |
| "learning_rate": 5e-05, |
| "loss": 0.4661, |
| "step": 1260 |
| }, |
| { |
| "epoch": 15.99, |
| "learning_rate": 5e-05, |
| "loss": 0.5712, |
| "step": 1261 |
| }, |
| { |
| "epoch": 16.0, |
| "learning_rate": 5e-05, |
| "loss": 0.3966, |
| "step": 1262 |
| }, |
| { |
| "epoch": 16.01, |
| "learning_rate": 5e-05, |
| "loss": 0.3904, |
| "step": 1263 |
| }, |
| { |
| "epoch": 16.03, |
| "learning_rate": 5e-05, |
| "loss": 0.3482, |
| "step": 1264 |
| }, |
| { |
| "epoch": 16.04, |
| "learning_rate": 5e-05, |
| "loss": 0.3397, |
| "step": 1265 |
| }, |
| { |
| "epoch": 16.05, |
| "learning_rate": 5e-05, |
| "loss": 0.3387, |
| "step": 1266 |
| }, |
| { |
| "epoch": 16.06, |
| "learning_rate": 5e-05, |
| "loss": 0.4651, |
| "step": 1267 |
| }, |
| { |
| "epoch": 16.08, |
| "learning_rate": 5e-05, |
| "loss": 0.3306, |
| "step": 1268 |
| }, |
| { |
| "epoch": 16.09, |
| "learning_rate": 5e-05, |
| "loss": 0.4278, |
| "step": 1269 |
| }, |
| { |
| "epoch": 16.1, |
| "learning_rate": 5e-05, |
| "loss": 0.3961, |
| "step": 1270 |
| }, |
| { |
| "epoch": 16.11, |
| "learning_rate": 5e-05, |
| "loss": 0.3782, |
| "step": 1271 |
| }, |
| { |
| "epoch": 16.13, |
| "learning_rate": 5e-05, |
| "loss": 0.387, |
| "step": 1272 |
| }, |
| { |
| "epoch": 16.14, |
| "learning_rate": 5e-05, |
| "loss": 0.3287, |
| "step": 1273 |
| }, |
| { |
| "epoch": 16.15, |
| "learning_rate": 5e-05, |
| "loss": 0.4089, |
| "step": 1274 |
| }, |
| { |
| "epoch": 16.16, |
| "learning_rate": 5e-05, |
| "loss": 0.4018, |
| "step": 1275 |
| }, |
| { |
| "epoch": 16.18, |
| "learning_rate": 5e-05, |
| "loss": 0.4225, |
| "step": 1276 |
| }, |
| { |
| "epoch": 16.19, |
| "learning_rate": 5e-05, |
| "loss": 0.3874, |
| "step": 1277 |
| }, |
| { |
| "epoch": 16.2, |
| "learning_rate": 5e-05, |
| "loss": 0.4767, |
| "step": 1278 |
| }, |
| { |
| "epoch": 16.22, |
| "learning_rate": 5e-05, |
| "loss": 0.3558, |
| "step": 1279 |
| }, |
| { |
| "epoch": 16.23, |
| "learning_rate": 5e-05, |
| "loss": 0.3216, |
| "step": 1280 |
| }, |
| { |
| "epoch": 16.24, |
| "learning_rate": 5e-05, |
| "loss": 0.3813, |
| "step": 1281 |
| }, |
| { |
| "epoch": 16.25, |
| "learning_rate": 5e-05, |
| "loss": 0.4318, |
| "step": 1282 |
| }, |
| { |
| "epoch": 16.27, |
| "learning_rate": 5e-05, |
| "loss": 0.402, |
| "step": 1283 |
| }, |
| { |
| "epoch": 16.28, |
| "learning_rate": 5e-05, |
| "loss": 0.3895, |
| "step": 1284 |
| }, |
| { |
| "epoch": 16.29, |
| "learning_rate": 5e-05, |
| "loss": 0.3442, |
| "step": 1285 |
| }, |
| { |
| "epoch": 16.3, |
| "learning_rate": 5e-05, |
| "loss": 0.2807, |
| "step": 1286 |
| }, |
| { |
| "epoch": 16.32, |
| "learning_rate": 5e-05, |
| "loss": 0.3422, |
| "step": 1287 |
| }, |
| { |
| "epoch": 16.33, |
| "learning_rate": 5e-05, |
| "loss": 0.312, |
| "step": 1288 |
| }, |
| { |
| "epoch": 16.34, |
| "learning_rate": 5e-05, |
| "loss": 0.4428, |
| "step": 1289 |
| }, |
| { |
| "epoch": 16.35, |
| "learning_rate": 5e-05, |
| "loss": 0.5253, |
| "step": 1290 |
| }, |
| { |
| "epoch": 16.37, |
| "learning_rate": 5e-05, |
| "loss": 0.3727, |
| "step": 1291 |
| }, |
| { |
| "epoch": 16.38, |
| "learning_rate": 5e-05, |
| "loss": 0.4031, |
| "step": 1292 |
| }, |
| { |
| "epoch": 16.39, |
| "learning_rate": 5e-05, |
| "loss": 0.2861, |
| "step": 1293 |
| }, |
| { |
| "epoch": 16.41, |
| "learning_rate": 5e-05, |
| "loss": 0.4453, |
| "step": 1294 |
| }, |
| { |
| "epoch": 16.42, |
| "learning_rate": 5e-05, |
| "loss": 0.3882, |
| "step": 1295 |
| }, |
| { |
| "epoch": 16.43, |
| "learning_rate": 5e-05, |
| "loss": 0.4962, |
| "step": 1296 |
| }, |
| { |
| "epoch": 16.44, |
| "learning_rate": 5e-05, |
| "loss": 0.296, |
| "step": 1297 |
| }, |
| { |
| "epoch": 16.46, |
| "learning_rate": 5e-05, |
| "loss": 0.3031, |
| "step": 1298 |
| }, |
| { |
| "epoch": 16.47, |
| "learning_rate": 5e-05, |
| "loss": 0.3729, |
| "step": 1299 |
| }, |
| { |
| "epoch": 16.48, |
| "learning_rate": 5e-05, |
| "loss": 0.3406, |
| "step": 1300 |
| }, |
| { |
| "epoch": 16.49, |
| "learning_rate": 5e-05, |
| "loss": 0.3255, |
| "step": 1301 |
| }, |
| { |
| "epoch": 16.51, |
| "learning_rate": 5e-05, |
| "loss": 0.3931, |
| "step": 1302 |
| }, |
| { |
| "epoch": 16.52, |
| "learning_rate": 5e-05, |
| "loss": 0.3113, |
| "step": 1303 |
| }, |
| { |
| "epoch": 16.53, |
| "learning_rate": 5e-05, |
| "loss": 0.4105, |
| "step": 1304 |
| }, |
| { |
| "epoch": 16.55, |
| "learning_rate": 5e-05, |
| "loss": 0.3323, |
| "step": 1305 |
| }, |
| { |
| "epoch": 16.56, |
| "learning_rate": 5e-05, |
| "loss": 0.4222, |
| "step": 1306 |
| }, |
| { |
| "epoch": 16.57, |
| "learning_rate": 5e-05, |
| "loss": 0.3945, |
| "step": 1307 |
| }, |
| { |
| "epoch": 16.58, |
| "learning_rate": 5e-05, |
| "loss": 0.2936, |
| "step": 1308 |
| }, |
| { |
| "epoch": 16.6, |
| "learning_rate": 5e-05, |
| "loss": 0.2422, |
| "step": 1309 |
| }, |
| { |
| "epoch": 16.61, |
| "learning_rate": 5e-05, |
| "loss": 0.3322, |
| "step": 1310 |
| }, |
| { |
| "epoch": 16.62, |
| "learning_rate": 5e-05, |
| "loss": 0.3817, |
| "step": 1311 |
| }, |
| { |
| "epoch": 16.63, |
| "learning_rate": 5e-05, |
| "loss": 0.4071, |
| "step": 1312 |
| }, |
| { |
| "epoch": 16.65, |
| "learning_rate": 5e-05, |
| "loss": 0.3631, |
| "step": 1313 |
| }, |
| { |
| "epoch": 16.66, |
| "learning_rate": 5e-05, |
| "loss": 0.3675, |
| "step": 1314 |
| }, |
| { |
| "epoch": 16.67, |
| "learning_rate": 5e-05, |
| "loss": 0.3589, |
| "step": 1315 |
| }, |
| { |
| "epoch": 16.68, |
| "learning_rate": 5e-05, |
| "loss": 0.3211, |
| "step": 1316 |
| }, |
| { |
| "epoch": 16.7, |
| "learning_rate": 5e-05, |
| "loss": 0.3132, |
| "step": 1317 |
| }, |
| { |
| "epoch": 16.71, |
| "learning_rate": 5e-05, |
| "loss": 0.3756, |
| "step": 1318 |
| }, |
| { |
| "epoch": 16.72, |
| "learning_rate": 5e-05, |
| "loss": 0.3201, |
| "step": 1319 |
| }, |
| { |
| "epoch": 16.74, |
| "learning_rate": 5e-05, |
| "loss": 0.4067, |
| "step": 1320 |
| }, |
| { |
| "epoch": 16.75, |
| "learning_rate": 5e-05, |
| "loss": 0.4163, |
| "step": 1321 |
| }, |
| { |
| "epoch": 16.76, |
| "learning_rate": 5e-05, |
| "loss": 0.4683, |
| "step": 1322 |
| }, |
| { |
| "epoch": 16.77, |
| "learning_rate": 5e-05, |
| "loss": 0.3919, |
| "step": 1323 |
| }, |
| { |
| "epoch": 16.79, |
| "learning_rate": 5e-05, |
| "loss": 0.3264, |
| "step": 1324 |
| }, |
| { |
| "epoch": 16.8, |
| "learning_rate": 5e-05, |
| "loss": 0.4432, |
| "step": 1325 |
| }, |
| { |
| "epoch": 16.81, |
| "learning_rate": 5e-05, |
| "loss": 0.3984, |
| "step": 1326 |
| }, |
| { |
| "epoch": 16.82, |
| "learning_rate": 5e-05, |
| "loss": 0.2963, |
| "step": 1327 |
| }, |
| { |
| "epoch": 16.84, |
| "learning_rate": 5e-05, |
| "loss": 0.502, |
| "step": 1328 |
| }, |
| { |
| "epoch": 16.85, |
| "learning_rate": 5e-05, |
| "loss": 0.3015, |
| "step": 1329 |
| }, |
| { |
| "epoch": 16.86, |
| "learning_rate": 5e-05, |
| "loss": 0.3797, |
| "step": 1330 |
| }, |
| { |
| "epoch": 16.87, |
| "learning_rate": 5e-05, |
| "loss": 0.3689, |
| "step": 1331 |
| }, |
| { |
| "epoch": 16.89, |
| "learning_rate": 5e-05, |
| "loss": 0.4154, |
| "step": 1332 |
| }, |
| { |
| "epoch": 16.9, |
| "learning_rate": 5e-05, |
| "loss": 0.4495, |
| "step": 1333 |
| }, |
| { |
| "epoch": 16.91, |
| "learning_rate": 5e-05, |
| "loss": 0.5139, |
| "step": 1334 |
| }, |
| { |
| "epoch": 16.93, |
| "learning_rate": 5e-05, |
| "loss": 0.3568, |
| "step": 1335 |
| }, |
| { |
| "epoch": 16.94, |
| "learning_rate": 5e-05, |
| "loss": 0.4269, |
| "step": 1336 |
| }, |
| { |
| "epoch": 16.95, |
| "learning_rate": 5e-05, |
| "loss": 0.3886, |
| "step": 1337 |
| }, |
| { |
| "epoch": 16.96, |
| "learning_rate": 5e-05, |
| "loss": 0.4385, |
| "step": 1338 |
| }, |
| { |
| "epoch": 16.98, |
| "learning_rate": 5e-05, |
| "loss": 0.4062, |
| "step": 1339 |
| }, |
| { |
| "epoch": 16.99, |
| "learning_rate": 5e-05, |
| "loss": 0.4084, |
| "step": 1340 |
| }, |
| { |
| "epoch": 17.0, |
| "learning_rate": 5e-05, |
| "loss": 0.299, |
| "step": 1341 |
| }, |
| { |
| "epoch": 17.01, |
| "learning_rate": 5e-05, |
| "loss": 0.3695, |
| "step": 1342 |
| }, |
| { |
| "epoch": 17.03, |
| "learning_rate": 5e-05, |
| "loss": 0.3462, |
| "step": 1343 |
| }, |
| { |
| "epoch": 17.04, |
| "learning_rate": 5e-05, |
| "loss": 0.3537, |
| "step": 1344 |
| }, |
| { |
| "epoch": 17.05, |
| "learning_rate": 5e-05, |
| "loss": 0.3013, |
| "step": 1345 |
| }, |
| { |
| "epoch": 17.06, |
| "learning_rate": 5e-05, |
| "loss": 0.2961, |
| "step": 1346 |
| }, |
| { |
| "epoch": 17.08, |
| "learning_rate": 5e-05, |
| "loss": 0.3329, |
| "step": 1347 |
| }, |
| { |
| "epoch": 17.09, |
| "learning_rate": 5e-05, |
| "loss": 0.3648, |
| "step": 1348 |
| }, |
| { |
| "epoch": 17.1, |
| "learning_rate": 5e-05, |
| "loss": 0.2547, |
| "step": 1349 |
| }, |
| { |
| "epoch": 17.12, |
| "learning_rate": 5e-05, |
| "loss": 0.3743, |
| "step": 1350 |
| }, |
| { |
| "epoch": 17.13, |
| "learning_rate": 5e-05, |
| "loss": 0.34, |
| "step": 1351 |
| }, |
| { |
| "epoch": 17.14, |
| "learning_rate": 5e-05, |
| "loss": 0.4066, |
| "step": 1352 |
| }, |
| { |
| "epoch": 17.15, |
| "learning_rate": 5e-05, |
| "loss": 0.2486, |
| "step": 1353 |
| }, |
| { |
| "epoch": 17.17, |
| "learning_rate": 5e-05, |
| "loss": 0.3372, |
| "step": 1354 |
| }, |
| { |
| "epoch": 17.18, |
| "learning_rate": 5e-05, |
| "loss": 0.3085, |
| "step": 1355 |
| }, |
| { |
| "epoch": 17.19, |
| "learning_rate": 5e-05, |
| "loss": 0.4211, |
| "step": 1356 |
| }, |
| { |
| "epoch": 17.2, |
| "learning_rate": 5e-05, |
| "loss": 0.3603, |
| "step": 1357 |
| }, |
| { |
| "epoch": 17.22, |
| "learning_rate": 5e-05, |
| "loss": 0.3341, |
| "step": 1358 |
| }, |
| { |
| "epoch": 17.23, |
| "learning_rate": 5e-05, |
| "loss": 0.41, |
| "step": 1359 |
| }, |
| { |
| "epoch": 17.24, |
| "learning_rate": 5e-05, |
| "loss": 0.3679, |
| "step": 1360 |
| }, |
| { |
| "epoch": 17.26, |
| "learning_rate": 5e-05, |
| "loss": 0.3394, |
| "step": 1361 |
| }, |
| { |
| "epoch": 17.27, |
| "learning_rate": 5e-05, |
| "loss": 0.3449, |
| "step": 1362 |
| }, |
| { |
| "epoch": 17.28, |
| "learning_rate": 5e-05, |
| "loss": 0.4752, |
| "step": 1363 |
| }, |
| { |
| "epoch": 17.29, |
| "learning_rate": 5e-05, |
| "loss": 0.3759, |
| "step": 1364 |
| }, |
| { |
| "epoch": 17.31, |
| "learning_rate": 5e-05, |
| "loss": 0.246, |
| "step": 1365 |
| }, |
| { |
| "epoch": 17.32, |
| "learning_rate": 5e-05, |
| "loss": 0.4163, |
| "step": 1366 |
| }, |
| { |
| "epoch": 17.33, |
| "learning_rate": 5e-05, |
| "loss": 0.3298, |
| "step": 1367 |
| }, |
| { |
| "epoch": 17.34, |
| "learning_rate": 5e-05, |
| "loss": 0.4717, |
| "step": 1368 |
| }, |
| { |
| "epoch": 17.36, |
| "learning_rate": 5e-05, |
| "loss": 0.345, |
| "step": 1369 |
| }, |
| { |
| "epoch": 17.37, |
| "learning_rate": 5e-05, |
| "loss": 0.3001, |
| "step": 1370 |
| }, |
| { |
| "epoch": 17.38, |
| "learning_rate": 5e-05, |
| "loss": 0.3764, |
| "step": 1371 |
| }, |
| { |
| "epoch": 17.39, |
| "learning_rate": 5e-05, |
| "loss": 0.2827, |
| "step": 1372 |
| }, |
| { |
| "epoch": 17.41, |
| "learning_rate": 5e-05, |
| "loss": 0.3118, |
| "step": 1373 |
| }, |
| { |
| "epoch": 17.42, |
| "learning_rate": 5e-05, |
| "loss": 0.3234, |
| "step": 1374 |
| }, |
| { |
| "epoch": 17.43, |
| "learning_rate": 5e-05, |
| "loss": 0.3275, |
| "step": 1375 |
| }, |
| { |
| "epoch": 17.45, |
| "learning_rate": 5e-05, |
| "loss": 0.3916, |
| "step": 1376 |
| }, |
| { |
| "epoch": 17.46, |
| "learning_rate": 5e-05, |
| "loss": 0.3566, |
| "step": 1377 |
| }, |
| { |
| "epoch": 17.47, |
| "learning_rate": 5e-05, |
| "loss": 0.3116, |
| "step": 1378 |
| }, |
| { |
| "epoch": 17.48, |
| "learning_rate": 5e-05, |
| "loss": 0.3905, |
| "step": 1379 |
| }, |
| { |
| "epoch": 17.5, |
| "learning_rate": 5e-05, |
| "loss": 0.3094, |
| "step": 1380 |
| }, |
| { |
| "epoch": 17.51, |
| "learning_rate": 5e-05, |
| "loss": 0.3269, |
| "step": 1381 |
| }, |
| { |
| "epoch": 17.52, |
| "learning_rate": 5e-05, |
| "loss": 0.3602, |
| "step": 1382 |
| }, |
| { |
| "epoch": 17.53, |
| "learning_rate": 5e-05, |
| "loss": 0.391, |
| "step": 1383 |
| }, |
| { |
| "epoch": 17.55, |
| "learning_rate": 5e-05, |
| "loss": 0.3651, |
| "step": 1384 |
| }, |
| { |
| "epoch": 17.56, |
| "learning_rate": 5e-05, |
| "loss": 0.3134, |
| "step": 1385 |
| }, |
| { |
| "epoch": 17.57, |
| "learning_rate": 5e-05, |
| "loss": 0.449, |
| "step": 1386 |
| }, |
| { |
| "epoch": 17.58, |
| "learning_rate": 5e-05, |
| "loss": 0.3735, |
| "step": 1387 |
| }, |
| { |
| "epoch": 17.6, |
| "learning_rate": 5e-05, |
| "loss": 0.3729, |
| "step": 1388 |
| }, |
| { |
| "epoch": 17.61, |
| "learning_rate": 5e-05, |
| "loss": 0.2881, |
| "step": 1389 |
| }, |
| { |
| "epoch": 17.62, |
| "learning_rate": 5e-05, |
| "loss": 0.385, |
| "step": 1390 |
| }, |
| { |
| "epoch": 17.64, |
| "learning_rate": 5e-05, |
| "loss": 0.506, |
| "step": 1391 |
| }, |
| { |
| "epoch": 17.65, |
| "learning_rate": 5e-05, |
| "loss": 0.2769, |
| "step": 1392 |
| }, |
| { |
| "epoch": 17.66, |
| "learning_rate": 5e-05, |
| "loss": 0.3641, |
| "step": 1393 |
| }, |
| { |
| "epoch": 17.67, |
| "learning_rate": 5e-05, |
| "loss": 0.2921, |
| "step": 1394 |
| }, |
| { |
| "epoch": 17.69, |
| "learning_rate": 5e-05, |
| "loss": 0.3031, |
| "step": 1395 |
| }, |
| { |
| "epoch": 17.7, |
| "learning_rate": 5e-05, |
| "loss": 0.3121, |
| "step": 1396 |
| }, |
| { |
| "epoch": 17.71, |
| "learning_rate": 5e-05, |
| "loss": 0.3884, |
| "step": 1397 |
| }, |
| { |
| "epoch": 17.72, |
| "learning_rate": 5e-05, |
| "loss": 0.4109, |
| "step": 1398 |
| }, |
| { |
| "epoch": 17.74, |
| "learning_rate": 5e-05, |
| "loss": 0.3388, |
| "step": 1399 |
| }, |
| { |
| "epoch": 17.75, |
| "learning_rate": 5e-05, |
| "loss": 0.3442, |
| "step": 1400 |
| }, |
| { |
| "epoch": 17.76, |
| "learning_rate": 5e-05, |
| "loss": 0.4407, |
| "step": 1401 |
| }, |
| { |
| "epoch": 17.77, |
| "learning_rate": 5e-05, |
| "loss": 0.4583, |
| "step": 1402 |
| }, |
| { |
| "epoch": 17.79, |
| "learning_rate": 5e-05, |
| "loss": 0.2997, |
| "step": 1403 |
| }, |
| { |
| "epoch": 17.8, |
| "learning_rate": 5e-05, |
| "loss": 0.3579, |
| "step": 1404 |
| }, |
| { |
| "epoch": 17.81, |
| "learning_rate": 5e-05, |
| "loss": 0.4163, |
| "step": 1405 |
| }, |
| { |
| "epoch": 17.83, |
| "learning_rate": 5e-05, |
| "loss": 0.3174, |
| "step": 1406 |
| }, |
| { |
| "epoch": 17.84, |
| "learning_rate": 5e-05, |
| "loss": 0.4205, |
| "step": 1407 |
| }, |
| { |
| "epoch": 17.85, |
| "learning_rate": 5e-05, |
| "loss": 0.318, |
| "step": 1408 |
| }, |
| { |
| "epoch": 17.86, |
| "learning_rate": 5e-05, |
| "loss": 0.3175, |
| "step": 1409 |
| }, |
| { |
| "epoch": 17.88, |
| "learning_rate": 5e-05, |
| "loss": 0.3885, |
| "step": 1410 |
| }, |
| { |
| "epoch": 17.89, |
| "learning_rate": 5e-05, |
| "loss": 0.3758, |
| "step": 1411 |
| }, |
| { |
| "epoch": 17.9, |
| "learning_rate": 5e-05, |
| "loss": 0.3824, |
| "step": 1412 |
| }, |
| { |
| "epoch": 17.91, |
| "learning_rate": 5e-05, |
| "loss": 0.3653, |
| "step": 1413 |
| }, |
| { |
| "epoch": 17.93, |
| "learning_rate": 5e-05, |
| "loss": 0.3568, |
| "step": 1414 |
| }, |
| { |
| "epoch": 17.94, |
| "learning_rate": 5e-05, |
| "loss": 0.3318, |
| "step": 1415 |
| }, |
| { |
| "epoch": 17.95, |
| "learning_rate": 5e-05, |
| "loss": 0.301, |
| "step": 1416 |
| }, |
| { |
| "epoch": 17.97, |
| "learning_rate": 5e-05, |
| "loss": 0.4334, |
| "step": 1417 |
| }, |
| { |
| "epoch": 17.98, |
| "learning_rate": 5e-05, |
| "loss": 0.4528, |
| "step": 1418 |
| }, |
| { |
| "epoch": 17.99, |
| "learning_rate": 5e-05, |
| "loss": 0.4518, |
| "step": 1419 |
| }, |
| { |
| "epoch": 18.0, |
| "learning_rate": 5e-05, |
| "loss": 0.4263, |
| "step": 1420 |
| }, |
| { |
| "epoch": 18.02, |
| "learning_rate": 5e-05, |
| "loss": 0.2745, |
| "step": 1421 |
| }, |
| { |
| "epoch": 18.03, |
| "learning_rate": 5e-05, |
| "loss": 0.3926, |
| "step": 1422 |
| }, |
| { |
| "epoch": 18.04, |
| "learning_rate": 5e-05, |
| "loss": 0.3136, |
| "step": 1423 |
| }, |
| { |
| "epoch": 18.05, |
| "learning_rate": 5e-05, |
| "loss": 0.2218, |
| "step": 1424 |
| }, |
| { |
| "epoch": 18.07, |
| "learning_rate": 5e-05, |
| "loss": 0.3466, |
| "step": 1425 |
| }, |
| { |
| "epoch": 18.08, |
| "learning_rate": 5e-05, |
| "loss": 0.3007, |
| "step": 1426 |
| }, |
| { |
| "epoch": 18.09, |
| "learning_rate": 5e-05, |
| "loss": 0.3591, |
| "step": 1427 |
| }, |
| { |
| "epoch": 18.1, |
| "learning_rate": 5e-05, |
| "loss": 0.3561, |
| "step": 1428 |
| }, |
| { |
| "epoch": 18.12, |
| "learning_rate": 5e-05, |
| "loss": 0.2682, |
| "step": 1429 |
| }, |
| { |
| "epoch": 18.13, |
| "learning_rate": 5e-05, |
| "loss": 0.4256, |
| "step": 1430 |
| }, |
| { |
| "epoch": 18.14, |
| "learning_rate": 5e-05, |
| "loss": 0.3125, |
| "step": 1431 |
| }, |
| { |
| "epoch": 18.16, |
| "learning_rate": 5e-05, |
| "loss": 0.2886, |
| "step": 1432 |
| }, |
| { |
| "epoch": 18.17, |
| "learning_rate": 5e-05, |
| "loss": 0.2678, |
| "step": 1433 |
| }, |
| { |
| "epoch": 18.18, |
| "learning_rate": 5e-05, |
| "loss": 0.2858, |
| "step": 1434 |
| }, |
| { |
| "epoch": 18.19, |
| "learning_rate": 5e-05, |
| "loss": 0.322, |
| "step": 1435 |
| }, |
| { |
| "epoch": 18.21, |
| "learning_rate": 5e-05, |
| "loss": 0.2567, |
| "step": 1436 |
| }, |
| { |
| "epoch": 18.22, |
| "learning_rate": 5e-05, |
| "loss": 0.4295, |
| "step": 1437 |
| }, |
| { |
| "epoch": 18.23, |
| "learning_rate": 5e-05, |
| "loss": 0.3125, |
| "step": 1438 |
| }, |
| { |
| "epoch": 18.24, |
| "learning_rate": 5e-05, |
| "loss": 0.2675, |
| "step": 1439 |
| }, |
| { |
| "epoch": 18.26, |
| "learning_rate": 5e-05, |
| "loss": 0.3444, |
| "step": 1440 |
| }, |
| { |
| "epoch": 18.27, |
| "learning_rate": 5e-05, |
| "loss": 0.2257, |
| "step": 1441 |
| }, |
| { |
| "epoch": 18.28, |
| "learning_rate": 5e-05, |
| "loss": 0.3068, |
| "step": 1442 |
| }, |
| { |
| "epoch": 18.29, |
| "learning_rate": 5e-05, |
| "loss": 0.3322, |
| "step": 1443 |
| }, |
| { |
| "epoch": 18.31, |
| "learning_rate": 5e-05, |
| "loss": 0.3652, |
| "step": 1444 |
| }, |
| { |
| "epoch": 18.32, |
| "learning_rate": 5e-05, |
| "loss": 0.3433, |
| "step": 1445 |
| }, |
| { |
| "epoch": 18.33, |
| "learning_rate": 5e-05, |
| "loss": 0.3204, |
| "step": 1446 |
| }, |
| { |
| "epoch": 18.35, |
| "learning_rate": 5e-05, |
| "loss": 0.3976, |
| "step": 1447 |
| }, |
| { |
| "epoch": 18.36, |
| "learning_rate": 5e-05, |
| "loss": 0.27, |
| "step": 1448 |
| }, |
| { |
| "epoch": 18.37, |
| "learning_rate": 5e-05, |
| "loss": 0.457, |
| "step": 1449 |
| }, |
| { |
| "epoch": 18.38, |
| "learning_rate": 5e-05, |
| "loss": 0.356, |
| "step": 1450 |
| }, |
| { |
| "epoch": 18.4, |
| "learning_rate": 5e-05, |
| "loss": 0.3641, |
| "step": 1451 |
| }, |
| { |
| "epoch": 18.41, |
| "learning_rate": 5e-05, |
| "loss": 0.3934, |
| "step": 1452 |
| }, |
| { |
| "epoch": 18.42, |
| "learning_rate": 5e-05, |
| "loss": 0.2828, |
| "step": 1453 |
| }, |
| { |
| "epoch": 18.43, |
| "learning_rate": 5e-05, |
| "loss": 0.342, |
| "step": 1454 |
| }, |
| { |
| "epoch": 18.45, |
| "learning_rate": 5e-05, |
| "loss": 0.2701, |
| "step": 1455 |
| }, |
| { |
| "epoch": 18.46, |
| "learning_rate": 5e-05, |
| "loss": 0.3704, |
| "step": 1456 |
| }, |
| { |
| "epoch": 18.47, |
| "learning_rate": 5e-05, |
| "loss": 0.3083, |
| "step": 1457 |
| }, |
| { |
| "epoch": 18.48, |
| "learning_rate": 5e-05, |
| "loss": 0.3224, |
| "step": 1458 |
| }, |
| { |
| "epoch": 18.5, |
| "learning_rate": 5e-05, |
| "loss": 0.4321, |
| "step": 1459 |
| }, |
| { |
| "epoch": 18.51, |
| "learning_rate": 5e-05, |
| "loss": 0.4678, |
| "step": 1460 |
| }, |
| { |
| "epoch": 18.52, |
| "learning_rate": 5e-05, |
| "loss": 0.3911, |
| "step": 1461 |
| }, |
| { |
| "epoch": 18.54, |
| "learning_rate": 5e-05, |
| "loss": 0.2843, |
| "step": 1462 |
| }, |
| { |
| "epoch": 18.55, |
| "learning_rate": 5e-05, |
| "loss": 0.3723, |
| "step": 1463 |
| }, |
| { |
| "epoch": 18.56, |
| "learning_rate": 5e-05, |
| "loss": 0.3211, |
| "step": 1464 |
| }, |
| { |
| "epoch": 18.57, |
| "learning_rate": 5e-05, |
| "loss": 0.3418, |
| "step": 1465 |
| }, |
| { |
| "epoch": 18.59, |
| "learning_rate": 5e-05, |
| "loss": 0.3697, |
| "step": 1466 |
| }, |
| { |
| "epoch": 18.6, |
| "learning_rate": 5e-05, |
| "loss": 0.3075, |
| "step": 1467 |
| }, |
| { |
| "epoch": 18.61, |
| "learning_rate": 5e-05, |
| "loss": 0.3072, |
| "step": 1468 |
| }, |
| { |
| "epoch": 18.62, |
| "learning_rate": 5e-05, |
| "loss": 0.3318, |
| "step": 1469 |
| }, |
| { |
| "epoch": 18.64, |
| "learning_rate": 5e-05, |
| "loss": 0.2997, |
| "step": 1470 |
| }, |
| { |
| "epoch": 18.65, |
| "learning_rate": 5e-05, |
| "loss": 0.2487, |
| "step": 1471 |
| }, |
| { |
| "epoch": 18.66, |
| "learning_rate": 5e-05, |
| "loss": 0.3958, |
| "step": 1472 |
| }, |
| { |
| "epoch": 18.68, |
| "learning_rate": 5e-05, |
| "loss": 0.3189, |
| "step": 1473 |
| }, |
| { |
| "epoch": 18.69, |
| "learning_rate": 5e-05, |
| "loss": 0.3298, |
| "step": 1474 |
| }, |
| { |
| "epoch": 18.7, |
| "learning_rate": 5e-05, |
| "loss": 0.3234, |
| "step": 1475 |
| }, |
| { |
| "epoch": 18.71, |
| "learning_rate": 5e-05, |
| "loss": 0.3622, |
| "step": 1476 |
| }, |
| { |
| "epoch": 18.73, |
| "learning_rate": 5e-05, |
| "loss": 0.3127, |
| "step": 1477 |
| }, |
| { |
| "epoch": 18.74, |
| "learning_rate": 5e-05, |
| "loss": 0.3277, |
| "step": 1478 |
| }, |
| { |
| "epoch": 18.75, |
| "learning_rate": 5e-05, |
| "loss": 0.3103, |
| "step": 1479 |
| }, |
| { |
| "epoch": 18.76, |
| "learning_rate": 5e-05, |
| "loss": 0.3804, |
| "step": 1480 |
| }, |
| { |
| "epoch": 18.78, |
| "learning_rate": 5e-05, |
| "loss": 0.2954, |
| "step": 1481 |
| }, |
| { |
| "epoch": 18.79, |
| "learning_rate": 5e-05, |
| "loss": 0.2949, |
| "step": 1482 |
| }, |
| { |
| "epoch": 18.8, |
| "learning_rate": 5e-05, |
| "loss": 0.3734, |
| "step": 1483 |
| }, |
| { |
| "epoch": 18.81, |
| "learning_rate": 5e-05, |
| "loss": 0.2556, |
| "step": 1484 |
| }, |
| { |
| "epoch": 18.83, |
| "learning_rate": 5e-05, |
| "loss": 0.3064, |
| "step": 1485 |
| }, |
| { |
| "epoch": 18.84, |
| "learning_rate": 5e-05, |
| "loss": 0.2772, |
| "step": 1486 |
| }, |
| { |
| "epoch": 18.85, |
| "learning_rate": 5e-05, |
| "loss": 0.3371, |
| "step": 1487 |
| }, |
| { |
| "epoch": 18.87, |
| "learning_rate": 5e-05, |
| "loss": 0.3357, |
| "step": 1488 |
| }, |
| { |
| "epoch": 18.88, |
| "learning_rate": 5e-05, |
| "loss": 0.3161, |
| "step": 1489 |
| }, |
| { |
| "epoch": 18.89, |
| "learning_rate": 5e-05, |
| "loss": 0.3534, |
| "step": 1490 |
| }, |
| { |
| "epoch": 18.9, |
| "learning_rate": 5e-05, |
| "loss": 0.2867, |
| "step": 1491 |
| }, |
| { |
| "epoch": 18.92, |
| "learning_rate": 5e-05, |
| "loss": 0.3899, |
| "step": 1492 |
| }, |
| { |
| "epoch": 18.93, |
| "learning_rate": 5e-05, |
| "loss": 0.3096, |
| "step": 1493 |
| }, |
| { |
| "epoch": 18.94, |
| "learning_rate": 5e-05, |
| "loss": 0.3989, |
| "step": 1494 |
| }, |
| { |
| "epoch": 18.95, |
| "learning_rate": 5e-05, |
| "loss": 0.3664, |
| "step": 1495 |
| }, |
| { |
| "epoch": 18.97, |
| "learning_rate": 5e-05, |
| "loss": 0.3084, |
| "step": 1496 |
| }, |
| { |
| "epoch": 18.98, |
| "learning_rate": 5e-05, |
| "loss": 0.3496, |
| "step": 1497 |
| }, |
| { |
| "epoch": 18.99, |
| "learning_rate": 5e-05, |
| "loss": 0.3769, |
| "step": 1498 |
| }, |
| { |
| "epoch": 19.0, |
| "learning_rate": 5e-05, |
| "loss": 0.2682, |
| "step": 1499 |
| }, |
| { |
| "epoch": 19.02, |
| "learning_rate": 5e-05, |
| "loss": 0.3359, |
| "step": 1500 |
| }, |
| { |
| "epoch": 19.03, |
| "learning_rate": 5e-05, |
| "loss": 0.3165, |
| "step": 1501 |
| }, |
| { |
| "epoch": 19.04, |
| "learning_rate": 5e-05, |
| "loss": 0.2487, |
| "step": 1502 |
| }, |
| { |
| "epoch": 19.06, |
| "learning_rate": 5e-05, |
| "loss": 0.307, |
| "step": 1503 |
| }, |
| { |
| "epoch": 19.07, |
| "learning_rate": 5e-05, |
| "loss": 0.2739, |
| "step": 1504 |
| }, |
| { |
| "epoch": 19.08, |
| "learning_rate": 5e-05, |
| "loss": 0.2974, |
| "step": 1505 |
| }, |
| { |
| "epoch": 19.09, |
| "learning_rate": 5e-05, |
| "loss": 0.2712, |
| "step": 1506 |
| }, |
| { |
| "epoch": 19.11, |
| "learning_rate": 5e-05, |
| "loss": 0.284, |
| "step": 1507 |
| }, |
| { |
| "epoch": 19.12, |
| "learning_rate": 5e-05, |
| "loss": 0.3309, |
| "step": 1508 |
| }, |
| { |
| "epoch": 19.13, |
| "learning_rate": 5e-05, |
| "loss": 0.3223, |
| "step": 1509 |
| }, |
| { |
| "epoch": 19.14, |
| "learning_rate": 5e-05, |
| "loss": 0.3366, |
| "step": 1510 |
| }, |
| { |
| "epoch": 19.16, |
| "learning_rate": 5e-05, |
| "loss": 0.3226, |
| "step": 1511 |
| }, |
| { |
| "epoch": 19.17, |
| "learning_rate": 5e-05, |
| "loss": 0.3629, |
| "step": 1512 |
| }, |
| { |
| "epoch": 19.18, |
| "learning_rate": 5e-05, |
| "loss": 0.2342, |
| "step": 1513 |
| }, |
| { |
| "epoch": 19.19, |
| "learning_rate": 5e-05, |
| "loss": 0.3137, |
| "step": 1514 |
| }, |
| { |
| "epoch": 19.21, |
| "learning_rate": 5e-05, |
| "loss": 0.2988, |
| "step": 1515 |
| }, |
| { |
| "epoch": 19.22, |
| "learning_rate": 5e-05, |
| "loss": 0.3877, |
| "step": 1516 |
| }, |
| { |
| "epoch": 19.23, |
| "learning_rate": 5e-05, |
| "loss": 0.286, |
| "step": 1517 |
| }, |
| { |
| "epoch": 19.25, |
| "learning_rate": 5e-05, |
| "loss": 0.3368, |
| "step": 1518 |
| }, |
| { |
| "epoch": 19.26, |
| "learning_rate": 5e-05, |
| "loss": 0.2309, |
| "step": 1519 |
| }, |
| { |
| "epoch": 19.27, |
| "learning_rate": 5e-05, |
| "loss": 0.265, |
| "step": 1520 |
| }, |
| { |
| "epoch": 19.28, |
| "learning_rate": 5e-05, |
| "loss": 0.2512, |
| "step": 1521 |
| }, |
| { |
| "epoch": 19.3, |
| "learning_rate": 5e-05, |
| "loss": 0.3631, |
| "step": 1522 |
| }, |
| { |
| "epoch": 19.31, |
| "learning_rate": 5e-05, |
| "loss": 0.3513, |
| "step": 1523 |
| }, |
| { |
| "epoch": 19.32, |
| "learning_rate": 5e-05, |
| "loss": 0.3052, |
| "step": 1524 |
| }, |
| { |
| "epoch": 19.33, |
| "learning_rate": 5e-05, |
| "loss": 0.2587, |
| "step": 1525 |
| }, |
| { |
| "epoch": 19.35, |
| "learning_rate": 5e-05, |
| "loss": 0.3261, |
| "step": 1526 |
| }, |
| { |
| "epoch": 19.36, |
| "learning_rate": 5e-05, |
| "loss": 0.2936, |
| "step": 1527 |
| }, |
| { |
| "epoch": 19.37, |
| "learning_rate": 5e-05, |
| "loss": 0.3511, |
| "step": 1528 |
| }, |
| { |
| "epoch": 19.39, |
| "learning_rate": 5e-05, |
| "loss": 0.3138, |
| "step": 1529 |
| }, |
| { |
| "epoch": 19.4, |
| "learning_rate": 5e-05, |
| "loss": 0.3452, |
| "step": 1530 |
| }, |
| { |
| "epoch": 19.41, |
| "learning_rate": 5e-05, |
| "loss": 0.2474, |
| "step": 1531 |
| }, |
| { |
| "epoch": 19.42, |
| "learning_rate": 5e-05, |
| "loss": 0.3069, |
| "step": 1532 |
| }, |
| { |
| "epoch": 19.44, |
| "learning_rate": 5e-05, |
| "loss": 0.3045, |
| "step": 1533 |
| }, |
| { |
| "epoch": 19.45, |
| "learning_rate": 5e-05, |
| "loss": 0.2757, |
| "step": 1534 |
| }, |
| { |
| "epoch": 19.46, |
| "learning_rate": 5e-05, |
| "loss": 0.288, |
| "step": 1535 |
| }, |
| { |
| "epoch": 19.47, |
| "learning_rate": 5e-05, |
| "loss": 0.2452, |
| "step": 1536 |
| }, |
| { |
| "epoch": 19.49, |
| "learning_rate": 5e-05, |
| "loss": 0.3897, |
| "step": 1537 |
| }, |
| { |
| "epoch": 19.5, |
| "learning_rate": 5e-05, |
| "loss": 0.1806, |
| "step": 1538 |
| }, |
| { |
| "epoch": 19.51, |
| "learning_rate": 5e-05, |
| "loss": 0.2861, |
| "step": 1539 |
| }, |
| { |
| "epoch": 19.52, |
| "learning_rate": 5e-05, |
| "loss": 0.2763, |
| "step": 1540 |
| }, |
| { |
| "epoch": 19.54, |
| "learning_rate": 5e-05, |
| "loss": 0.2857, |
| "step": 1541 |
| }, |
| { |
| "epoch": 19.55, |
| "learning_rate": 5e-05, |
| "loss": 0.2454, |
| "step": 1542 |
| }, |
| { |
| "epoch": 19.56, |
| "learning_rate": 5e-05, |
| "loss": 0.293, |
| "step": 1543 |
| }, |
| { |
| "epoch": 19.58, |
| "learning_rate": 5e-05, |
| "loss": 0.317, |
| "step": 1544 |
| }, |
| { |
| "epoch": 19.59, |
| "learning_rate": 5e-05, |
| "loss": 0.3071, |
| "step": 1545 |
| }, |
| { |
| "epoch": 19.6, |
| "learning_rate": 5e-05, |
| "loss": 0.3351, |
| "step": 1546 |
| }, |
| { |
| "epoch": 19.61, |
| "learning_rate": 5e-05, |
| "loss": 0.2296, |
| "step": 1547 |
| }, |
| { |
| "epoch": 19.63, |
| "learning_rate": 5e-05, |
| "loss": 0.4136, |
| "step": 1548 |
| }, |
| { |
| "epoch": 19.64, |
| "learning_rate": 5e-05, |
| "loss": 0.4726, |
| "step": 1549 |
| }, |
| { |
| "epoch": 19.65, |
| "learning_rate": 5e-05, |
| "loss": 0.373, |
| "step": 1550 |
| }, |
| { |
| "epoch": 19.66, |
| "learning_rate": 5e-05, |
| "loss": 0.3009, |
| "step": 1551 |
| }, |
| { |
| "epoch": 19.68, |
| "learning_rate": 5e-05, |
| "loss": 0.3487, |
| "step": 1552 |
| }, |
| { |
| "epoch": 19.69, |
| "learning_rate": 5e-05, |
| "loss": 0.3236, |
| "step": 1553 |
| }, |
| { |
| "epoch": 19.7, |
| "learning_rate": 5e-05, |
| "loss": 0.1818, |
| "step": 1554 |
| }, |
| { |
| "epoch": 19.71, |
| "learning_rate": 5e-05, |
| "loss": 0.2883, |
| "step": 1555 |
| }, |
| { |
| "epoch": 19.73, |
| "learning_rate": 5e-05, |
| "loss": 0.2588, |
| "step": 1556 |
| }, |
| { |
| "epoch": 19.74, |
| "learning_rate": 5e-05, |
| "loss": 0.3389, |
| "step": 1557 |
| }, |
| { |
| "epoch": 19.75, |
| "learning_rate": 5e-05, |
| "loss": 0.2814, |
| "step": 1558 |
| }, |
| { |
| "epoch": 19.77, |
| "learning_rate": 5e-05, |
| "loss": 0.3465, |
| "step": 1559 |
| }, |
| { |
| "epoch": 19.78, |
| "learning_rate": 5e-05, |
| "loss": 0.3318, |
| "step": 1560 |
| }, |
| { |
| "epoch": 19.79, |
| "learning_rate": 5e-05, |
| "loss": 0.3589, |
| "step": 1561 |
| }, |
| { |
| "epoch": 19.8, |
| "learning_rate": 5e-05, |
| "loss": 0.3196, |
| "step": 1562 |
| }, |
| { |
| "epoch": 19.82, |
| "learning_rate": 5e-05, |
| "loss": 0.4132, |
| "step": 1563 |
| }, |
| { |
| "epoch": 19.83, |
| "learning_rate": 5e-05, |
| "loss": 0.3853, |
| "step": 1564 |
| }, |
| { |
| "epoch": 19.84, |
| "learning_rate": 5e-05, |
| "loss": 0.2056, |
| "step": 1565 |
| }, |
| { |
| "epoch": 19.85, |
| "learning_rate": 5e-05, |
| "loss": 0.296, |
| "step": 1566 |
| }, |
| { |
| "epoch": 19.87, |
| "learning_rate": 5e-05, |
| "loss": 0.2892, |
| "step": 1567 |
| }, |
| { |
| "epoch": 19.88, |
| "learning_rate": 5e-05, |
| "loss": 0.3101, |
| "step": 1568 |
| }, |
| { |
| "epoch": 19.89, |
| "learning_rate": 5e-05, |
| "loss": 0.4169, |
| "step": 1569 |
| }, |
| { |
| "epoch": 19.9, |
| "learning_rate": 5e-05, |
| "loss": 0.4244, |
| "step": 1570 |
| }, |
| { |
| "epoch": 19.92, |
| "learning_rate": 5e-05, |
| "loss": 0.306, |
| "step": 1571 |
| }, |
| { |
| "epoch": 19.93, |
| "learning_rate": 5e-05, |
| "loss": 0.3005, |
| "step": 1572 |
| }, |
| { |
| "epoch": 19.94, |
| "learning_rate": 5e-05, |
| "loss": 0.3005, |
| "step": 1573 |
| }, |
| { |
| "epoch": 19.96, |
| "learning_rate": 5e-05, |
| "loss": 0.2365, |
| "step": 1574 |
| }, |
| { |
| "epoch": 19.97, |
| "learning_rate": 5e-05, |
| "loss": 0.3382, |
| "step": 1575 |
| }, |
| { |
| "epoch": 19.98, |
| "learning_rate": 5e-05, |
| "loss": 0.3506, |
| "step": 1576 |
| }, |
| { |
| "epoch": 19.99, |
| "learning_rate": 5e-05, |
| "loss": 0.3173, |
| "step": 1577 |
| }, |
| { |
| "epoch": 20.01, |
| "learning_rate": 5e-05, |
| "loss": 0.2177, |
| "step": 1578 |
| }, |
| { |
| "epoch": 20.02, |
| "learning_rate": 5e-05, |
| "loss": 0.2115, |
| "step": 1579 |
| }, |
| { |
| "epoch": 20.03, |
| "learning_rate": 5e-05, |
| "loss": 0.2513, |
| "step": 1580 |
| }, |
| { |
| "epoch": 20.04, |
| "learning_rate": 5e-05, |
| "loss": 0.2082, |
| "step": 1581 |
| }, |
| { |
| "epoch": 20.06, |
| "learning_rate": 5e-05, |
| "loss": 0.3109, |
| "step": 1582 |
| }, |
| { |
| "epoch": 20.07, |
| "learning_rate": 5e-05, |
| "loss": 0.3577, |
| "step": 1583 |
| }, |
| { |
| "epoch": 20.08, |
| "learning_rate": 5e-05, |
| "loss": 0.1344, |
| "step": 1584 |
| }, |
| { |
| "epoch": 20.1, |
| "learning_rate": 5e-05, |
| "loss": 0.1899, |
| "step": 1585 |
| }, |
| { |
| "epoch": 20.11, |
| "learning_rate": 5e-05, |
| "loss": 0.257, |
| "step": 1586 |
| }, |
| { |
| "epoch": 20.12, |
| "learning_rate": 5e-05, |
| "loss": 0.2409, |
| "step": 1587 |
| }, |
| { |
| "epoch": 20.13, |
| "learning_rate": 5e-05, |
| "loss": 0.2555, |
| "step": 1588 |
| }, |
| { |
| "epoch": 20.15, |
| "learning_rate": 5e-05, |
| "loss": 0.2755, |
| "step": 1589 |
| }, |
| { |
| "epoch": 20.16, |
| "learning_rate": 5e-05, |
| "loss": 0.3005, |
| "step": 1590 |
| }, |
| { |
| "epoch": 20.17, |
| "learning_rate": 5e-05, |
| "loss": 0.2756, |
| "step": 1591 |
| }, |
| { |
| "epoch": 20.18, |
| "learning_rate": 5e-05, |
| "loss": 0.295, |
| "step": 1592 |
| }, |
| { |
| "epoch": 20.2, |
| "learning_rate": 5e-05, |
| "loss": 0.3251, |
| "step": 1593 |
| }, |
| { |
| "epoch": 20.21, |
| "learning_rate": 5e-05, |
| "loss": 0.2807, |
| "step": 1594 |
| }, |
| { |
| "epoch": 20.22, |
| "learning_rate": 5e-05, |
| "loss": 0.3818, |
| "step": 1595 |
| }, |
| { |
| "epoch": 20.23, |
| "learning_rate": 5e-05, |
| "loss": 0.21, |
| "step": 1596 |
| }, |
| { |
| "epoch": 20.25, |
| "learning_rate": 5e-05, |
| "loss": 0.1937, |
| "step": 1597 |
| }, |
| { |
| "epoch": 20.26, |
| "learning_rate": 5e-05, |
| "loss": 0.2028, |
| "step": 1598 |
| }, |
| { |
| "epoch": 20.27, |
| "learning_rate": 5e-05, |
| "loss": 0.3418, |
| "step": 1599 |
| }, |
| { |
| "epoch": 20.29, |
| "learning_rate": 5e-05, |
| "loss": 0.2817, |
| "step": 1600 |
| }, |
| { |
| "epoch": 20.3, |
| "learning_rate": 5e-05, |
| "loss": 0.2752, |
| "step": 1601 |
| }, |
| { |
| "epoch": 20.31, |
| "learning_rate": 5e-05, |
| "loss": 0.3664, |
| "step": 1602 |
| }, |
| { |
| "epoch": 20.32, |
| "learning_rate": 5e-05, |
| "loss": 0.3485, |
| "step": 1603 |
| }, |
| { |
| "epoch": 20.34, |
| "learning_rate": 5e-05, |
| "loss": 0.2378, |
| "step": 1604 |
| }, |
| { |
| "epoch": 20.35, |
| "learning_rate": 5e-05, |
| "loss": 0.3606, |
| "step": 1605 |
| }, |
| { |
| "epoch": 20.36, |
| "learning_rate": 5e-05, |
| "loss": 0.2387, |
| "step": 1606 |
| }, |
| { |
| "epoch": 20.37, |
| "learning_rate": 5e-05, |
| "loss": 0.3161, |
| "step": 1607 |
| }, |
| { |
| "epoch": 20.39, |
| "learning_rate": 5e-05, |
| "loss": 0.256, |
| "step": 1608 |
| }, |
| { |
| "epoch": 20.4, |
| "learning_rate": 5e-05, |
| "loss": 0.1815, |
| "step": 1609 |
| }, |
| { |
| "epoch": 20.41, |
| "learning_rate": 5e-05, |
| "loss": 0.2504, |
| "step": 1610 |
| }, |
| { |
| "epoch": 20.42, |
| "learning_rate": 5e-05, |
| "loss": 0.27, |
| "step": 1611 |
| }, |
| { |
| "epoch": 20.44, |
| "learning_rate": 5e-05, |
| "loss": 0.3529, |
| "step": 1612 |
| }, |
| { |
| "epoch": 20.45, |
| "learning_rate": 5e-05, |
| "loss": 0.2674, |
| "step": 1613 |
| }, |
| { |
| "epoch": 20.46, |
| "learning_rate": 5e-05, |
| "loss": 0.2587, |
| "step": 1614 |
| }, |
| { |
| "epoch": 20.48, |
| "learning_rate": 5e-05, |
| "loss": 0.331, |
| "step": 1615 |
| }, |
| { |
| "epoch": 20.49, |
| "learning_rate": 5e-05, |
| "loss": 0.241, |
| "step": 1616 |
| }, |
| { |
| "epoch": 20.5, |
| "learning_rate": 5e-05, |
| "loss": 0.3235, |
| "step": 1617 |
| }, |
| { |
| "epoch": 20.51, |
| "learning_rate": 5e-05, |
| "loss": 0.3152, |
| "step": 1618 |
| }, |
| { |
| "epoch": 20.53, |
| "learning_rate": 5e-05, |
| "loss": 0.2636, |
| "step": 1619 |
| }, |
| { |
| "epoch": 20.54, |
| "learning_rate": 5e-05, |
| "loss": 0.227, |
| "step": 1620 |
| }, |
| { |
| "epoch": 20.55, |
| "learning_rate": 5e-05, |
| "loss": 0.2316, |
| "step": 1621 |
| }, |
| { |
| "epoch": 20.56, |
| "learning_rate": 5e-05, |
| "loss": 0.3619, |
| "step": 1622 |
| }, |
| { |
| "epoch": 20.58, |
| "learning_rate": 5e-05, |
| "loss": 0.2473, |
| "step": 1623 |
| }, |
| { |
| "epoch": 20.59, |
| "learning_rate": 5e-05, |
| "loss": 0.3582, |
| "step": 1624 |
| }, |
| { |
| "epoch": 20.6, |
| "learning_rate": 5e-05, |
| "loss": 0.4274, |
| "step": 1625 |
| }, |
| { |
| "epoch": 20.61, |
| "learning_rate": 5e-05, |
| "loss": 0.2947, |
| "step": 1626 |
| }, |
| { |
| "epoch": 20.63, |
| "learning_rate": 5e-05, |
| "loss": 0.3496, |
| "step": 1627 |
| }, |
| { |
| "epoch": 20.64, |
| "learning_rate": 5e-05, |
| "loss": 0.2986, |
| "step": 1628 |
| }, |
| { |
| "epoch": 20.65, |
| "learning_rate": 5e-05, |
| "loss": 0.2653, |
| "step": 1629 |
| }, |
| { |
| "epoch": 20.67, |
| "learning_rate": 5e-05, |
| "loss": 0.2526, |
| "step": 1630 |
| }, |
| { |
| "epoch": 20.68, |
| "learning_rate": 5e-05, |
| "loss": 0.3163, |
| "step": 1631 |
| }, |
| { |
| "epoch": 20.69, |
| "learning_rate": 5e-05, |
| "loss": 0.356, |
| "step": 1632 |
| }, |
| { |
| "epoch": 20.7, |
| "learning_rate": 5e-05, |
| "loss": 0.2878, |
| "step": 1633 |
| }, |
| { |
| "epoch": 20.72, |
| "learning_rate": 5e-05, |
| "loss": 0.2441, |
| "step": 1634 |
| }, |
| { |
| "epoch": 20.73, |
| "learning_rate": 5e-05, |
| "loss": 0.1991, |
| "step": 1635 |
| }, |
| { |
| "epoch": 20.74, |
| "learning_rate": 5e-05, |
| "loss": 0.3432, |
| "step": 1636 |
| }, |
| { |
| "epoch": 20.75, |
| "learning_rate": 5e-05, |
| "loss": 0.2455, |
| "step": 1637 |
| }, |
| { |
| "epoch": 20.77, |
| "learning_rate": 5e-05, |
| "loss": 0.2698, |
| "step": 1638 |
| }, |
| { |
| "epoch": 20.78, |
| "learning_rate": 5e-05, |
| "loss": 0.3536, |
| "step": 1639 |
| }, |
| { |
| "epoch": 20.79, |
| "learning_rate": 5e-05, |
| "loss": 0.2574, |
| "step": 1640 |
| }, |
| { |
| "epoch": 20.81, |
| "learning_rate": 5e-05, |
| "loss": 0.342, |
| "step": 1641 |
| }, |
| { |
| "epoch": 20.82, |
| "learning_rate": 5e-05, |
| "loss": 0.2633, |
| "step": 1642 |
| }, |
| { |
| "epoch": 20.83, |
| "learning_rate": 5e-05, |
| "loss": 0.2968, |
| "step": 1643 |
| }, |
| { |
| "epoch": 20.84, |
| "learning_rate": 5e-05, |
| "loss": 0.3373, |
| "step": 1644 |
| }, |
| { |
| "epoch": 20.86, |
| "learning_rate": 5e-05, |
| "loss": 0.2767, |
| "step": 1645 |
| }, |
| { |
| "epoch": 20.87, |
| "learning_rate": 5e-05, |
| "loss": 0.3579, |
| "step": 1646 |
| }, |
| { |
| "epoch": 20.88, |
| "learning_rate": 5e-05, |
| "loss": 0.3292, |
| "step": 1647 |
| }, |
| { |
| "epoch": 20.89, |
| "learning_rate": 5e-05, |
| "loss": 0.2639, |
| "step": 1648 |
| }, |
| { |
| "epoch": 20.91, |
| "learning_rate": 5e-05, |
| "loss": 0.347, |
| "step": 1649 |
| }, |
| { |
| "epoch": 20.92, |
| "learning_rate": 5e-05, |
| "loss": 0.2889, |
| "step": 1650 |
| }, |
| { |
| "epoch": 20.93, |
| "learning_rate": 5e-05, |
| "loss": 0.2848, |
| "step": 1651 |
| }, |
| { |
| "epoch": 20.94, |
| "learning_rate": 5e-05, |
| "loss": 0.4394, |
| "step": 1652 |
| }, |
| { |
| "epoch": 20.96, |
| "learning_rate": 5e-05, |
| "loss": 0.2353, |
| "step": 1653 |
| }, |
| { |
| "epoch": 20.97, |
| "learning_rate": 5e-05, |
| "loss": 0.323, |
| "step": 1654 |
| }, |
| { |
| "epoch": 20.98, |
| "learning_rate": 5e-05, |
| "loss": 0.2626, |
| "step": 1655 |
| }, |
| { |
| "epoch": 21.0, |
| "learning_rate": 5e-05, |
| "loss": 0.3107, |
| "step": 1656 |
| }, |
| { |
| "epoch": 21.01, |
| "learning_rate": 5e-05, |
| "loss": 0.307, |
| "step": 1657 |
| }, |
| { |
| "epoch": 21.02, |
| "learning_rate": 5e-05, |
| "loss": 0.1955, |
| "step": 1658 |
| }, |
| { |
| "epoch": 21.03, |
| "learning_rate": 5e-05, |
| "loss": 0.2498, |
| "step": 1659 |
| }, |
| { |
| "epoch": 21.05, |
| "learning_rate": 5e-05, |
| "loss": 0.2701, |
| "step": 1660 |
| }, |
| { |
| "epoch": 21.06, |
| "learning_rate": 5e-05, |
| "loss": 0.2194, |
| "step": 1661 |
| }, |
| { |
| "epoch": 21.07, |
| "learning_rate": 5e-05, |
| "loss": 0.2837, |
| "step": 1662 |
| }, |
| { |
| "epoch": 21.08, |
| "learning_rate": 5e-05, |
| "loss": 0.231, |
| "step": 1663 |
| }, |
| { |
| "epoch": 21.1, |
| "learning_rate": 5e-05, |
| "loss": 0.3166, |
| "step": 1664 |
| }, |
| { |
| "epoch": 21.11, |
| "learning_rate": 5e-05, |
| "loss": 0.3352, |
| "step": 1665 |
| }, |
| { |
| "epoch": 21.12, |
| "learning_rate": 5e-05, |
| "loss": 0.269, |
| "step": 1666 |
| }, |
| { |
| "epoch": 21.13, |
| "learning_rate": 5e-05, |
| "loss": 0.2293, |
| "step": 1667 |
| }, |
| { |
| "epoch": 21.15, |
| "learning_rate": 5e-05, |
| "loss": 0.2385, |
| "step": 1668 |
| }, |
| { |
| "epoch": 21.16, |
| "learning_rate": 5e-05, |
| "loss": 0.2956, |
| "step": 1669 |
| }, |
| { |
| "epoch": 21.17, |
| "learning_rate": 5e-05, |
| "loss": 0.2407, |
| "step": 1670 |
| }, |
| { |
| "epoch": 21.19, |
| "learning_rate": 5e-05, |
| "loss": 0.2597, |
| "step": 1671 |
| }, |
| { |
| "epoch": 21.2, |
| "learning_rate": 5e-05, |
| "loss": 0.1771, |
| "step": 1672 |
| }, |
| { |
| "epoch": 21.21, |
| "learning_rate": 5e-05, |
| "loss": 0.3541, |
| "step": 1673 |
| }, |
| { |
| "epoch": 21.22, |
| "learning_rate": 5e-05, |
| "loss": 0.2445, |
| "step": 1674 |
| }, |
| { |
| "epoch": 21.24, |
| "learning_rate": 5e-05, |
| "loss": 0.2524, |
| "step": 1675 |
| }, |
| { |
| "epoch": 21.25, |
| "learning_rate": 5e-05, |
| "loss": 0.2739, |
| "step": 1676 |
| }, |
| { |
| "epoch": 21.26, |
| "learning_rate": 5e-05, |
| "loss": 0.223, |
| "step": 1677 |
| }, |
| { |
| "epoch": 21.27, |
| "learning_rate": 5e-05, |
| "loss": 0.248, |
| "step": 1678 |
| }, |
| { |
| "epoch": 21.29, |
| "learning_rate": 5e-05, |
| "loss": 0.2244, |
| "step": 1679 |
| }, |
| { |
| "epoch": 21.3, |
| "learning_rate": 5e-05, |
| "loss": 0.1994, |
| "step": 1680 |
| }, |
| { |
| "epoch": 21.31, |
| "learning_rate": 5e-05, |
| "loss": 0.3105, |
| "step": 1681 |
| }, |
| { |
| "epoch": 21.32, |
| "learning_rate": 5e-05, |
| "loss": 0.2394, |
| "step": 1682 |
| }, |
| { |
| "epoch": 21.34, |
| "learning_rate": 5e-05, |
| "loss": 0.2846, |
| "step": 1683 |
| }, |
| { |
| "epoch": 21.35, |
| "learning_rate": 5e-05, |
| "loss": 0.2885, |
| "step": 1684 |
| }, |
| { |
| "epoch": 21.36, |
| "learning_rate": 5e-05, |
| "loss": 0.2815, |
| "step": 1685 |
| }, |
| { |
| "epoch": 21.38, |
| "learning_rate": 5e-05, |
| "loss": 0.1875, |
| "step": 1686 |
| }, |
| { |
| "epoch": 21.39, |
| "learning_rate": 5e-05, |
| "loss": 0.2611, |
| "step": 1687 |
| }, |
| { |
| "epoch": 21.4, |
| "learning_rate": 5e-05, |
| "loss": 0.2415, |
| "step": 1688 |
| }, |
| { |
| "epoch": 21.41, |
| "learning_rate": 5e-05, |
| "loss": 0.1967, |
| "step": 1689 |
| }, |
| { |
| "epoch": 21.43, |
| "learning_rate": 5e-05, |
| "loss": 0.2404, |
| "step": 1690 |
| }, |
| { |
| "epoch": 21.44, |
| "learning_rate": 5e-05, |
| "loss": 0.2552, |
| "step": 1691 |
| }, |
| { |
| "epoch": 21.45, |
| "learning_rate": 5e-05, |
| "loss": 0.3171, |
| "step": 1692 |
| }, |
| { |
| "epoch": 21.46, |
| "learning_rate": 5e-05, |
| "loss": 0.3392, |
| "step": 1693 |
| }, |
| { |
| "epoch": 21.48, |
| "learning_rate": 5e-05, |
| "loss": 0.1975, |
| "step": 1694 |
| }, |
| { |
| "epoch": 21.49, |
| "learning_rate": 5e-05, |
| "loss": 0.2263, |
| "step": 1695 |
| }, |
| { |
| "epoch": 21.5, |
| "learning_rate": 5e-05, |
| "loss": 0.2086, |
| "step": 1696 |
| }, |
| { |
| "epoch": 21.52, |
| "learning_rate": 5e-05, |
| "loss": 0.2221, |
| "step": 1697 |
| }, |
| { |
| "epoch": 21.53, |
| "learning_rate": 5e-05, |
| "loss": 0.296, |
| "step": 1698 |
| }, |
| { |
| "epoch": 21.54, |
| "learning_rate": 5e-05, |
| "loss": 0.2925, |
| "step": 1699 |
| }, |
| { |
| "epoch": 21.55, |
| "learning_rate": 5e-05, |
| "loss": 0.27, |
| "step": 1700 |
| }, |
| { |
| "epoch": 21.57, |
| "learning_rate": 5e-05, |
| "loss": 0.2856, |
| "step": 1701 |
| }, |
| { |
| "epoch": 21.58, |
| "learning_rate": 5e-05, |
| "loss": 0.3015, |
| "step": 1702 |
| }, |
| { |
| "epoch": 21.59, |
| "learning_rate": 5e-05, |
| "loss": 0.2346, |
| "step": 1703 |
| }, |
| { |
| "epoch": 21.6, |
| "learning_rate": 5e-05, |
| "loss": 0.2778, |
| "step": 1704 |
| }, |
| { |
| "epoch": 21.62, |
| "learning_rate": 5e-05, |
| "loss": 0.2442, |
| "step": 1705 |
| }, |
| { |
| "epoch": 21.63, |
| "learning_rate": 5e-05, |
| "loss": 0.2728, |
| "step": 1706 |
| }, |
| { |
| "epoch": 21.64, |
| "learning_rate": 5e-05, |
| "loss": 0.2566, |
| "step": 1707 |
| }, |
| { |
| "epoch": 21.65, |
| "learning_rate": 5e-05, |
| "loss": 0.2868, |
| "step": 1708 |
| }, |
| { |
| "epoch": 21.67, |
| "learning_rate": 5e-05, |
| "loss": 0.2958, |
| "step": 1709 |
| }, |
| { |
| "epoch": 21.68, |
| "learning_rate": 5e-05, |
| "loss": 0.2633, |
| "step": 1710 |
| }, |
| { |
| "epoch": 21.69, |
| "learning_rate": 5e-05, |
| "loss": 0.2519, |
| "step": 1711 |
| }, |
| { |
| "epoch": 21.71, |
| "learning_rate": 5e-05, |
| "loss": 0.2986, |
| "step": 1712 |
| }, |
| { |
| "epoch": 21.72, |
| "learning_rate": 5e-05, |
| "loss": 0.3082, |
| "step": 1713 |
| }, |
| { |
| "epoch": 21.73, |
| "learning_rate": 5e-05, |
| "loss": 0.2418, |
| "step": 1714 |
| }, |
| { |
| "epoch": 21.74, |
| "learning_rate": 5e-05, |
| "loss": 0.237, |
| "step": 1715 |
| }, |
| { |
| "epoch": 21.76, |
| "learning_rate": 5e-05, |
| "loss": 0.3279, |
| "step": 1716 |
| }, |
| { |
| "epoch": 21.77, |
| "learning_rate": 5e-05, |
| "loss": 0.2513, |
| "step": 1717 |
| }, |
| { |
| "epoch": 21.78, |
| "learning_rate": 5e-05, |
| "loss": 0.1656, |
| "step": 1718 |
| }, |
| { |
| "epoch": 21.79, |
| "learning_rate": 5e-05, |
| "loss": 0.3231, |
| "step": 1719 |
| }, |
| { |
| "epoch": 21.81, |
| "learning_rate": 5e-05, |
| "loss": 0.2182, |
| "step": 1720 |
| }, |
| { |
| "epoch": 21.82, |
| "learning_rate": 5e-05, |
| "loss": 0.2204, |
| "step": 1721 |
| }, |
| { |
| "epoch": 21.83, |
| "learning_rate": 5e-05, |
| "loss": 0.2752, |
| "step": 1722 |
| }, |
| { |
| "epoch": 21.84, |
| "learning_rate": 5e-05, |
| "loss": 0.3171, |
| "step": 1723 |
| }, |
| { |
| "epoch": 21.86, |
| "learning_rate": 5e-05, |
| "loss": 0.2618, |
| "step": 1724 |
| }, |
| { |
| "epoch": 21.87, |
| "learning_rate": 5e-05, |
| "loss": 0.2199, |
| "step": 1725 |
| }, |
| { |
| "epoch": 21.88, |
| "learning_rate": 5e-05, |
| "loss": 0.3857, |
| "step": 1726 |
| }, |
| { |
| "epoch": 21.9, |
| "learning_rate": 5e-05, |
| "loss": 0.2059, |
| "step": 1727 |
| }, |
| { |
| "epoch": 21.91, |
| "learning_rate": 5e-05, |
| "loss": 0.2264, |
| "step": 1728 |
| }, |
| { |
| "epoch": 21.92, |
| "learning_rate": 5e-05, |
| "loss": 0.3572, |
| "step": 1729 |
| }, |
| { |
| "epoch": 21.93, |
| "learning_rate": 5e-05, |
| "loss": 0.3099, |
| "step": 1730 |
| }, |
| { |
| "epoch": 21.95, |
| "learning_rate": 5e-05, |
| "loss": 0.2686, |
| "step": 1731 |
| }, |
| { |
| "epoch": 21.96, |
| "learning_rate": 5e-05, |
| "loss": 0.4122, |
| "step": 1732 |
| }, |
| { |
| "epoch": 21.97, |
| "learning_rate": 5e-05, |
| "loss": 0.2984, |
| "step": 1733 |
| }, |
| { |
| "epoch": 21.98, |
| "learning_rate": 5e-05, |
| "loss": 0.1762, |
| "step": 1734 |
| }, |
| { |
| "epoch": 22.0, |
| "learning_rate": 5e-05, |
| "loss": 0.344, |
| "step": 1735 |
| }, |
| { |
| "epoch": 22.01, |
| "learning_rate": 5e-05, |
| "loss": 0.3043, |
| "step": 1736 |
| }, |
| { |
| "epoch": 22.02, |
| "learning_rate": 5e-05, |
| "loss": 0.2189, |
| "step": 1737 |
| }, |
| { |
| "epoch": 22.03, |
| "learning_rate": 5e-05, |
| "loss": 0.1782, |
| "step": 1738 |
| }, |
| { |
| "epoch": 22.05, |
| "learning_rate": 5e-05, |
| "loss": 0.2156, |
| "step": 1739 |
| }, |
| { |
| "epoch": 22.06, |
| "learning_rate": 5e-05, |
| "loss": 0.1929, |
| "step": 1740 |
| }, |
| { |
| "epoch": 22.07, |
| "learning_rate": 5e-05, |
| "loss": 0.2193, |
| "step": 1741 |
| }, |
| { |
| "epoch": 22.09, |
| "learning_rate": 5e-05, |
| "loss": 0.1927, |
| "step": 1742 |
| }, |
| { |
| "epoch": 22.1, |
| "learning_rate": 5e-05, |
| "loss": 0.2603, |
| "step": 1743 |
| }, |
| { |
| "epoch": 22.11, |
| "learning_rate": 5e-05, |
| "loss": 0.2099, |
| "step": 1744 |
| }, |
| { |
| "epoch": 22.12, |
| "learning_rate": 5e-05, |
| "loss": 0.1699, |
| "step": 1745 |
| }, |
| { |
| "epoch": 22.14, |
| "learning_rate": 5e-05, |
| "loss": 0.2327, |
| "step": 1746 |
| }, |
| { |
| "epoch": 22.15, |
| "learning_rate": 5e-05, |
| "loss": 0.209, |
| "step": 1747 |
| }, |
| { |
| "epoch": 22.16, |
| "learning_rate": 5e-05, |
| "loss": 0.2258, |
| "step": 1748 |
| }, |
| { |
| "epoch": 22.17, |
| "learning_rate": 5e-05, |
| "loss": 0.2243, |
| "step": 1749 |
| }, |
| { |
| "epoch": 22.19, |
| "learning_rate": 5e-05, |
| "loss": 0.223, |
| "step": 1750 |
| }, |
| { |
| "epoch": 22.2, |
| "learning_rate": 5e-05, |
| "loss": 0.2288, |
| "step": 1751 |
| }, |
| { |
| "epoch": 22.21, |
| "learning_rate": 5e-05, |
| "loss": 0.3273, |
| "step": 1752 |
| }, |
| { |
| "epoch": 22.23, |
| "learning_rate": 5e-05, |
| "loss": 0.2858, |
| "step": 1753 |
| }, |
| { |
| "epoch": 22.24, |
| "learning_rate": 5e-05, |
| "loss": 0.1705, |
| "step": 1754 |
| }, |
| { |
| "epoch": 22.25, |
| "learning_rate": 5e-05, |
| "loss": 0.2788, |
| "step": 1755 |
| }, |
| { |
| "epoch": 22.26, |
| "learning_rate": 5e-05, |
| "loss": 0.2295, |
| "step": 1756 |
| }, |
| { |
| "epoch": 22.28, |
| "learning_rate": 5e-05, |
| "loss": 0.2232, |
| "step": 1757 |
| }, |
| { |
| "epoch": 22.29, |
| "learning_rate": 5e-05, |
| "loss": 0.1828, |
| "step": 1758 |
| }, |
| { |
| "epoch": 22.3, |
| "learning_rate": 5e-05, |
| "loss": 0.2698, |
| "step": 1759 |
| }, |
| { |
| "epoch": 22.31, |
| "learning_rate": 5e-05, |
| "loss": 0.2377, |
| "step": 1760 |
| }, |
| { |
| "epoch": 22.33, |
| "learning_rate": 5e-05, |
| "loss": 0.247, |
| "step": 1761 |
| }, |
| { |
| "epoch": 22.34, |
| "learning_rate": 5e-05, |
| "loss": 0.2077, |
| "step": 1762 |
| }, |
| { |
| "epoch": 22.35, |
| "learning_rate": 5e-05, |
| "loss": 0.1855, |
| "step": 1763 |
| }, |
| { |
| "epoch": 22.36, |
| "learning_rate": 5e-05, |
| "loss": 0.2424, |
| "step": 1764 |
| }, |
| { |
| "epoch": 22.38, |
| "learning_rate": 5e-05, |
| "loss": 0.2727, |
| "step": 1765 |
| }, |
| { |
| "epoch": 22.39, |
| "learning_rate": 5e-05, |
| "loss": 0.2838, |
| "step": 1766 |
| }, |
| { |
| "epoch": 22.4, |
| "learning_rate": 5e-05, |
| "loss": 0.2172, |
| "step": 1767 |
| }, |
| { |
| "epoch": 22.42, |
| "learning_rate": 5e-05, |
| "loss": 0.2351, |
| "step": 1768 |
| }, |
| { |
| "epoch": 22.43, |
| "learning_rate": 5e-05, |
| "loss": 0.2396, |
| "step": 1769 |
| }, |
| { |
| "epoch": 22.44, |
| "learning_rate": 5e-05, |
| "loss": 0.2574, |
| "step": 1770 |
| }, |
| { |
| "epoch": 22.45, |
| "learning_rate": 5e-05, |
| "loss": 0.2789, |
| "step": 1771 |
| }, |
| { |
| "epoch": 22.47, |
| "learning_rate": 5e-05, |
| "loss": 0.2207, |
| "step": 1772 |
| }, |
| { |
| "epoch": 22.48, |
| "learning_rate": 5e-05, |
| "loss": 0.192, |
| "step": 1773 |
| }, |
| { |
| "epoch": 22.49, |
| "learning_rate": 5e-05, |
| "loss": 0.2551, |
| "step": 1774 |
| }, |
| { |
| "epoch": 22.5, |
| "learning_rate": 5e-05, |
| "loss": 0.2921, |
| "step": 1775 |
| }, |
| { |
| "epoch": 22.52, |
| "learning_rate": 5e-05, |
| "loss": 0.1525, |
| "step": 1776 |
| }, |
| { |
| "epoch": 22.53, |
| "learning_rate": 5e-05, |
| "loss": 0.1736, |
| "step": 1777 |
| }, |
| { |
| "epoch": 22.54, |
| "learning_rate": 5e-05, |
| "loss": 0.2211, |
| "step": 1778 |
| }, |
| { |
| "epoch": 22.55, |
| "learning_rate": 5e-05, |
| "loss": 0.2406, |
| "step": 1779 |
| }, |
| { |
| "epoch": 22.57, |
| "learning_rate": 5e-05, |
| "loss": 0.2435, |
| "step": 1780 |
| }, |
| { |
| "epoch": 22.58, |
| "learning_rate": 5e-05, |
| "loss": 0.2758, |
| "step": 1781 |
| }, |
| { |
| "epoch": 22.59, |
| "learning_rate": 5e-05, |
| "loss": 0.1668, |
| "step": 1782 |
| }, |
| { |
| "epoch": 22.61, |
| "learning_rate": 5e-05, |
| "loss": 0.2235, |
| "step": 1783 |
| }, |
| { |
| "epoch": 22.62, |
| "learning_rate": 5e-05, |
| "loss": 0.2238, |
| "step": 1784 |
| }, |
| { |
| "epoch": 22.63, |
| "learning_rate": 5e-05, |
| "loss": 0.2744, |
| "step": 1785 |
| }, |
| { |
| "epoch": 22.64, |
| "learning_rate": 5e-05, |
| "loss": 0.2734, |
| "step": 1786 |
| }, |
| { |
| "epoch": 22.66, |
| "learning_rate": 5e-05, |
| "loss": 0.2212, |
| "step": 1787 |
| }, |
| { |
| "epoch": 22.67, |
| "learning_rate": 5e-05, |
| "loss": 0.2489, |
| "step": 1788 |
| }, |
| { |
| "epoch": 22.68, |
| "learning_rate": 5e-05, |
| "loss": 0.2407, |
| "step": 1789 |
| }, |
| { |
| "epoch": 22.69, |
| "learning_rate": 5e-05, |
| "loss": 0.1802, |
| "step": 1790 |
| }, |
| { |
| "epoch": 22.71, |
| "learning_rate": 5e-05, |
| "loss": 0.2195, |
| "step": 1791 |
| }, |
| { |
| "epoch": 22.72, |
| "learning_rate": 5e-05, |
| "loss": 0.2827, |
| "step": 1792 |
| }, |
| { |
| "epoch": 22.73, |
| "learning_rate": 5e-05, |
| "loss": 0.253, |
| "step": 1793 |
| }, |
| { |
| "epoch": 22.74, |
| "learning_rate": 5e-05, |
| "loss": 0.3167, |
| "step": 1794 |
| }, |
| { |
| "epoch": 22.76, |
| "learning_rate": 5e-05, |
| "loss": 0.329, |
| "step": 1795 |
| }, |
| { |
| "epoch": 22.77, |
| "learning_rate": 5e-05, |
| "loss": 0.2291, |
| "step": 1796 |
| }, |
| { |
| "epoch": 22.78, |
| "learning_rate": 5e-05, |
| "loss": 0.2607, |
| "step": 1797 |
| }, |
| { |
| "epoch": 22.8, |
| "learning_rate": 5e-05, |
| "loss": 0.2761, |
| "step": 1798 |
| }, |
| { |
| "epoch": 22.81, |
| "learning_rate": 5e-05, |
| "loss": 0.3297, |
| "step": 1799 |
| }, |
| { |
| "epoch": 22.82, |
| "learning_rate": 5e-05, |
| "loss": 0.3904, |
| "step": 1800 |
| }, |
| { |
| "epoch": 22.83, |
| "learning_rate": 5e-05, |
| "loss": 0.2468, |
| "step": 1801 |
| }, |
| { |
| "epoch": 22.85, |
| "learning_rate": 5e-05, |
| "loss": 0.251, |
| "step": 1802 |
| }, |
| { |
| "epoch": 22.86, |
| "learning_rate": 5e-05, |
| "loss": 0.2622, |
| "step": 1803 |
| }, |
| { |
| "epoch": 22.87, |
| "learning_rate": 5e-05, |
| "loss": 0.1888, |
| "step": 1804 |
| }, |
| { |
| "epoch": 22.88, |
| "learning_rate": 5e-05, |
| "loss": 0.2433, |
| "step": 1805 |
| }, |
| { |
| "epoch": 22.9, |
| "learning_rate": 5e-05, |
| "loss": 0.2025, |
| "step": 1806 |
| }, |
| { |
| "epoch": 22.91, |
| "learning_rate": 5e-05, |
| "loss": 0.2392, |
| "step": 1807 |
| }, |
| { |
| "epoch": 22.92, |
| "learning_rate": 5e-05, |
| "loss": 0.2675, |
| "step": 1808 |
| }, |
| { |
| "epoch": 22.94, |
| "learning_rate": 5e-05, |
| "loss": 0.2866, |
| "step": 1809 |
| }, |
| { |
| "epoch": 22.95, |
| "learning_rate": 5e-05, |
| "loss": 0.2996, |
| "step": 1810 |
| }, |
| { |
| "epoch": 22.96, |
| "learning_rate": 5e-05, |
| "loss": 0.274, |
| "step": 1811 |
| }, |
| { |
| "epoch": 22.97, |
| "learning_rate": 5e-05, |
| "loss": 0.2888, |
| "step": 1812 |
| }, |
| { |
| "epoch": 22.99, |
| "learning_rate": 5e-05, |
| "loss": 0.2864, |
| "step": 1813 |
| }, |
| { |
| "epoch": 23.0, |
| "learning_rate": 5e-05, |
| "loss": 0.2985, |
| "step": 1814 |
| }, |
| { |
| "epoch": 23.01, |
| "learning_rate": 5e-05, |
| "loss": 0.1968, |
| "step": 1815 |
| }, |
| { |
| "epoch": 23.02, |
| "learning_rate": 5e-05, |
| "loss": 0.2475, |
| "step": 1816 |
| }, |
| { |
| "epoch": 23.04, |
| "learning_rate": 5e-05, |
| "loss": 0.1946, |
| "step": 1817 |
| }, |
| { |
| "epoch": 23.05, |
| "learning_rate": 5e-05, |
| "loss": 0.2792, |
| "step": 1818 |
| }, |
| { |
| "epoch": 23.06, |
| "learning_rate": 5e-05, |
| "loss": 0.1875, |
| "step": 1819 |
| }, |
| { |
| "epoch": 23.07, |
| "learning_rate": 5e-05, |
| "loss": 0.2401, |
| "step": 1820 |
| }, |
| { |
| "epoch": 23.09, |
| "learning_rate": 5e-05, |
| "loss": 0.2543, |
| "step": 1821 |
| }, |
| { |
| "epoch": 23.1, |
| "learning_rate": 5e-05, |
| "loss": 0.2835, |
| "step": 1822 |
| }, |
| { |
| "epoch": 23.11, |
| "learning_rate": 5e-05, |
| "loss": 0.1681, |
| "step": 1823 |
| }, |
| { |
| "epoch": 23.13, |
| "learning_rate": 5e-05, |
| "loss": 0.1869, |
| "step": 1824 |
| }, |
| { |
| "epoch": 23.14, |
| "learning_rate": 5e-05, |
| "loss": 0.1942, |
| "step": 1825 |
| }, |
| { |
| "epoch": 23.15, |
| "learning_rate": 5e-05, |
| "loss": 0.1804, |
| "step": 1826 |
| }, |
| { |
| "epoch": 23.16, |
| "learning_rate": 5e-05, |
| "loss": 0.2466, |
| "step": 1827 |
| }, |
| { |
| "epoch": 23.18, |
| "learning_rate": 5e-05, |
| "loss": 0.1974, |
| "step": 1828 |
| }, |
| { |
| "epoch": 23.19, |
| "learning_rate": 5e-05, |
| "loss": 0.2014, |
| "step": 1829 |
| }, |
| { |
| "epoch": 23.2, |
| "learning_rate": 5e-05, |
| "loss": 0.1476, |
| "step": 1830 |
| }, |
| { |
| "epoch": 23.21, |
| "learning_rate": 5e-05, |
| "loss": 0.1681, |
| "step": 1831 |
| }, |
| { |
| "epoch": 23.23, |
| "learning_rate": 5e-05, |
| "loss": 0.1854, |
| "step": 1832 |
| }, |
| { |
| "epoch": 23.24, |
| "learning_rate": 5e-05, |
| "loss": 0.2518, |
| "step": 1833 |
| }, |
| { |
| "epoch": 23.25, |
| "learning_rate": 5e-05, |
| "loss": 0.2063, |
| "step": 1834 |
| }, |
| { |
| "epoch": 23.26, |
| "learning_rate": 5e-05, |
| "loss": 0.3458, |
| "step": 1835 |
| }, |
| { |
| "epoch": 23.28, |
| "learning_rate": 5e-05, |
| "loss": 0.1478, |
| "step": 1836 |
| }, |
| { |
| "epoch": 23.29, |
| "learning_rate": 5e-05, |
| "loss": 0.2636, |
| "step": 1837 |
| }, |
| { |
| "epoch": 23.3, |
| "learning_rate": 5e-05, |
| "loss": 0.1188, |
| "step": 1838 |
| }, |
| { |
| "epoch": 23.32, |
| "learning_rate": 5e-05, |
| "loss": 0.1628, |
| "step": 1839 |
| }, |
| { |
| "epoch": 23.33, |
| "learning_rate": 5e-05, |
| "loss": 0.1736, |
| "step": 1840 |
| }, |
| { |
| "epoch": 23.34, |
| "learning_rate": 5e-05, |
| "loss": 0.1637, |
| "step": 1841 |
| }, |
| { |
| "epoch": 23.35, |
| "learning_rate": 5e-05, |
| "loss": 0.1631, |
| "step": 1842 |
| }, |
| { |
| "epoch": 23.37, |
| "learning_rate": 5e-05, |
| "loss": 0.1542, |
| "step": 1843 |
| }, |
| { |
| "epoch": 23.38, |
| "learning_rate": 5e-05, |
| "loss": 0.2349, |
| "step": 1844 |
| }, |
| { |
| "epoch": 23.39, |
| "learning_rate": 5e-05, |
| "loss": 0.1763, |
| "step": 1845 |
| }, |
| { |
| "epoch": 23.4, |
| "learning_rate": 5e-05, |
| "loss": 0.2482, |
| "step": 1846 |
| }, |
| { |
| "epoch": 23.42, |
| "learning_rate": 5e-05, |
| "loss": 0.2525, |
| "step": 1847 |
| }, |
| { |
| "epoch": 23.43, |
| "learning_rate": 5e-05, |
| "loss": 0.2543, |
| "step": 1848 |
| }, |
| { |
| "epoch": 23.44, |
| "learning_rate": 5e-05, |
| "loss": 0.2391, |
| "step": 1849 |
| }, |
| { |
| "epoch": 23.45, |
| "learning_rate": 5e-05, |
| "loss": 0.2077, |
| "step": 1850 |
| }, |
| { |
| "epoch": 23.47, |
| "learning_rate": 5e-05, |
| "loss": 0.2366, |
| "step": 1851 |
| }, |
| { |
| "epoch": 23.48, |
| "learning_rate": 5e-05, |
| "loss": 0.1794, |
| "step": 1852 |
| }, |
| { |
| "epoch": 23.49, |
| "learning_rate": 5e-05, |
| "loss": 0.2902, |
| "step": 1853 |
| }, |
| { |
| "epoch": 23.51, |
| "learning_rate": 5e-05, |
| "loss": 0.2629, |
| "step": 1854 |
| }, |
| { |
| "epoch": 23.52, |
| "learning_rate": 5e-05, |
| "loss": 0.2424, |
| "step": 1855 |
| }, |
| { |
| "epoch": 23.53, |
| "learning_rate": 5e-05, |
| "loss": 0.1952, |
| "step": 1856 |
| }, |
| { |
| "epoch": 23.54, |
| "learning_rate": 5e-05, |
| "loss": 0.177, |
| "step": 1857 |
| }, |
| { |
| "epoch": 23.56, |
| "learning_rate": 5e-05, |
| "loss": 0.2585, |
| "step": 1858 |
| }, |
| { |
| "epoch": 23.57, |
| "learning_rate": 5e-05, |
| "loss": 0.2679, |
| "step": 1859 |
| }, |
| { |
| "epoch": 23.58, |
| "learning_rate": 5e-05, |
| "loss": 0.2726, |
| "step": 1860 |
| }, |
| { |
| "epoch": 23.59, |
| "learning_rate": 5e-05, |
| "loss": 0.2509, |
| "step": 1861 |
| }, |
| { |
| "epoch": 23.61, |
| "learning_rate": 5e-05, |
| "loss": 0.2273, |
| "step": 1862 |
| }, |
| { |
| "epoch": 23.62, |
| "learning_rate": 5e-05, |
| "loss": 0.2023, |
| "step": 1863 |
| }, |
| { |
| "epoch": 23.63, |
| "learning_rate": 5e-05, |
| "loss": 0.3105, |
| "step": 1864 |
| }, |
| { |
| "epoch": 23.65, |
| "learning_rate": 5e-05, |
| "loss": 0.1716, |
| "step": 1865 |
| }, |
| { |
| "epoch": 23.66, |
| "learning_rate": 5e-05, |
| "loss": 0.1618, |
| "step": 1866 |
| }, |
| { |
| "epoch": 23.67, |
| "learning_rate": 5e-05, |
| "loss": 0.2648, |
| "step": 1867 |
| }, |
| { |
| "epoch": 23.68, |
| "learning_rate": 5e-05, |
| "loss": 0.2047, |
| "step": 1868 |
| }, |
| { |
| "epoch": 23.7, |
| "learning_rate": 5e-05, |
| "loss": 0.2024, |
| "step": 1869 |
| }, |
| { |
| "epoch": 23.71, |
| "learning_rate": 5e-05, |
| "loss": 0.2609, |
| "step": 1870 |
| }, |
| { |
| "epoch": 23.72, |
| "learning_rate": 5e-05, |
| "loss": 0.272, |
| "step": 1871 |
| }, |
| { |
| "epoch": 23.73, |
| "learning_rate": 5e-05, |
| "loss": 0.1999, |
| "step": 1872 |
| }, |
| { |
| "epoch": 23.75, |
| "learning_rate": 5e-05, |
| "loss": 0.2445, |
| "step": 1873 |
| }, |
| { |
| "epoch": 23.76, |
| "learning_rate": 5e-05, |
| "loss": 0.2999, |
| "step": 1874 |
| }, |
| { |
| "epoch": 23.77, |
| "learning_rate": 5e-05, |
| "loss": 0.2779, |
| "step": 1875 |
| }, |
| { |
| "epoch": 23.78, |
| "learning_rate": 5e-05, |
| "loss": 0.2032, |
| "step": 1876 |
| }, |
| { |
| "epoch": 23.8, |
| "learning_rate": 5e-05, |
| "loss": 0.1973, |
| "step": 1877 |
| }, |
| { |
| "epoch": 23.81, |
| "learning_rate": 5e-05, |
| "loss": 0.2449, |
| "step": 1878 |
| }, |
| { |
| "epoch": 23.82, |
| "learning_rate": 5e-05, |
| "loss": 0.2927, |
| "step": 1879 |
| }, |
| { |
| "epoch": 23.84, |
| "learning_rate": 5e-05, |
| "loss": 0.1448, |
| "step": 1880 |
| }, |
| { |
| "epoch": 23.85, |
| "learning_rate": 5e-05, |
| "loss": 0.1592, |
| "step": 1881 |
| }, |
| { |
| "epoch": 23.86, |
| "learning_rate": 5e-05, |
| "loss": 0.2308, |
| "step": 1882 |
| }, |
| { |
| "epoch": 23.87, |
| "learning_rate": 5e-05, |
| "loss": 0.1851, |
| "step": 1883 |
| }, |
| { |
| "epoch": 23.89, |
| "learning_rate": 5e-05, |
| "loss": 0.3323, |
| "step": 1884 |
| }, |
| { |
| "epoch": 23.9, |
| "learning_rate": 5e-05, |
| "loss": 0.1954, |
| "step": 1885 |
| }, |
| { |
| "epoch": 23.91, |
| "learning_rate": 5e-05, |
| "loss": 0.1804, |
| "step": 1886 |
| }, |
| { |
| "epoch": 23.92, |
| "learning_rate": 5e-05, |
| "loss": 0.2445, |
| "step": 1887 |
| }, |
| { |
| "epoch": 23.94, |
| "learning_rate": 5e-05, |
| "loss": 0.2911, |
| "step": 1888 |
| }, |
| { |
| "epoch": 23.95, |
| "learning_rate": 5e-05, |
| "loss": 0.217, |
| "step": 1889 |
| }, |
| { |
| "epoch": 23.96, |
| "learning_rate": 5e-05, |
| "loss": 0.3145, |
| "step": 1890 |
| }, |
| { |
| "epoch": 23.97, |
| "learning_rate": 5e-05, |
| "loss": 0.1775, |
| "step": 1891 |
| }, |
| { |
| "epoch": 23.99, |
| "learning_rate": 5e-05, |
| "loss": 0.2446, |
| "step": 1892 |
| }, |
| { |
| "epoch": 24.0, |
| "learning_rate": 5e-05, |
| "loss": 0.2406, |
| "step": 1893 |
| }, |
| { |
| "epoch": 24.01, |
| "learning_rate": 5e-05, |
| "loss": 0.153, |
| "step": 1894 |
| }, |
| { |
| "epoch": 24.03, |
| "learning_rate": 5e-05, |
| "loss": 0.2006, |
| "step": 1895 |
| }, |
| { |
| "epoch": 24.04, |
| "learning_rate": 5e-05, |
| "loss": 0.2094, |
| "step": 1896 |
| }, |
| { |
| "epoch": 24.05, |
| "learning_rate": 5e-05, |
| "loss": 0.1435, |
| "step": 1897 |
| }, |
| { |
| "epoch": 24.06, |
| "learning_rate": 5e-05, |
| "loss": 0.2003, |
| "step": 1898 |
| }, |
| { |
| "epoch": 24.08, |
| "learning_rate": 5e-05, |
| "loss": 0.2608, |
| "step": 1899 |
| }, |
| { |
| "epoch": 24.09, |
| "learning_rate": 5e-05, |
| "loss": 0.1494, |
| "step": 1900 |
| }, |
| { |
| "epoch": 24.1, |
| "learning_rate": 5e-05, |
| "loss": 0.222, |
| "step": 1901 |
| }, |
| { |
| "epoch": 24.11, |
| "learning_rate": 5e-05, |
| "loss": 0.1691, |
| "step": 1902 |
| }, |
| { |
| "epoch": 24.13, |
| "learning_rate": 5e-05, |
| "loss": 0.2058, |
| "step": 1903 |
| }, |
| { |
| "epoch": 24.14, |
| "learning_rate": 5e-05, |
| "loss": 0.1036, |
| "step": 1904 |
| }, |
| { |
| "epoch": 24.15, |
| "learning_rate": 5e-05, |
| "loss": 0.1803, |
| "step": 1905 |
| }, |
| { |
| "epoch": 24.16, |
| "learning_rate": 5e-05, |
| "loss": 0.1212, |
| "step": 1906 |
| }, |
| { |
| "epoch": 24.18, |
| "learning_rate": 5e-05, |
| "loss": 0.1841, |
| "step": 1907 |
| }, |
| { |
| "epoch": 24.19, |
| "learning_rate": 5e-05, |
| "loss": 0.2012, |
| "step": 1908 |
| }, |
| { |
| "epoch": 24.2, |
| "learning_rate": 5e-05, |
| "loss": 0.2818, |
| "step": 1909 |
| }, |
| { |
| "epoch": 24.22, |
| "learning_rate": 5e-05, |
| "loss": 0.2581, |
| "step": 1910 |
| }, |
| { |
| "epoch": 24.23, |
| "learning_rate": 5e-05, |
| "loss": 0.2849, |
| "step": 1911 |
| }, |
| { |
| "epoch": 24.24, |
| "learning_rate": 5e-05, |
| "loss": 0.1937, |
| "step": 1912 |
| }, |
| { |
| "epoch": 24.25, |
| "learning_rate": 5e-05, |
| "loss": 0.1882, |
| "step": 1913 |
| }, |
| { |
| "epoch": 24.27, |
| "learning_rate": 5e-05, |
| "loss": 0.1698, |
| "step": 1914 |
| }, |
| { |
| "epoch": 24.28, |
| "learning_rate": 5e-05, |
| "loss": 0.1488, |
| "step": 1915 |
| }, |
| { |
| "epoch": 24.29, |
| "learning_rate": 5e-05, |
| "loss": 0.1961, |
| "step": 1916 |
| }, |
| { |
| "epoch": 24.3, |
| "learning_rate": 5e-05, |
| "loss": 0.1676, |
| "step": 1917 |
| }, |
| { |
| "epoch": 24.32, |
| "learning_rate": 5e-05, |
| "loss": 0.1514, |
| "step": 1918 |
| }, |
| { |
| "epoch": 24.33, |
| "learning_rate": 5e-05, |
| "loss": 0.1978, |
| "step": 1919 |
| }, |
| { |
| "epoch": 24.34, |
| "learning_rate": 5e-05, |
| "loss": 0.2534, |
| "step": 1920 |
| }, |
| { |
| "epoch": 24.35, |
| "learning_rate": 5e-05, |
| "loss": 0.1005, |
| "step": 1921 |
| }, |
| { |
| "epoch": 24.37, |
| "learning_rate": 5e-05, |
| "loss": 0.1986, |
| "step": 1922 |
| }, |
| { |
| "epoch": 24.38, |
| "learning_rate": 5e-05, |
| "loss": 0.2066, |
| "step": 1923 |
| }, |
| { |
| "epoch": 24.39, |
| "learning_rate": 5e-05, |
| "loss": 0.2759, |
| "step": 1924 |
| }, |
| { |
| "epoch": 24.41, |
| "learning_rate": 5e-05, |
| "loss": 0.204, |
| "step": 1925 |
| }, |
| { |
| "epoch": 24.42, |
| "learning_rate": 5e-05, |
| "loss": 0.2113, |
| "step": 1926 |
| }, |
| { |
| "epoch": 24.43, |
| "learning_rate": 5e-05, |
| "loss": 0.2124, |
| "step": 1927 |
| }, |
| { |
| "epoch": 24.44, |
| "learning_rate": 5e-05, |
| "loss": 0.2739, |
| "step": 1928 |
| }, |
| { |
| "epoch": 24.46, |
| "learning_rate": 5e-05, |
| "loss": 0.1746, |
| "step": 1929 |
| }, |
| { |
| "epoch": 24.47, |
| "learning_rate": 5e-05, |
| "loss": 0.2456, |
| "step": 1930 |
| }, |
| { |
| "epoch": 24.48, |
| "learning_rate": 5e-05, |
| "loss": 0.1732, |
| "step": 1931 |
| }, |
| { |
| "epoch": 24.49, |
| "learning_rate": 5e-05, |
| "loss": 0.231, |
| "step": 1932 |
| }, |
| { |
| "epoch": 24.51, |
| "learning_rate": 5e-05, |
| "loss": 0.2006, |
| "step": 1933 |
| }, |
| { |
| "epoch": 24.52, |
| "learning_rate": 5e-05, |
| "loss": 0.1644, |
| "step": 1934 |
| }, |
| { |
| "epoch": 24.53, |
| "learning_rate": 5e-05, |
| "loss": 0.2304, |
| "step": 1935 |
| }, |
| { |
| "epoch": 24.55, |
| "learning_rate": 5e-05, |
| "loss": 0.2037, |
| "step": 1936 |
| }, |
| { |
| "epoch": 24.56, |
| "learning_rate": 5e-05, |
| "loss": 0.1601, |
| "step": 1937 |
| }, |
| { |
| "epoch": 24.57, |
| "learning_rate": 5e-05, |
| "loss": 0.1968, |
| "step": 1938 |
| }, |
| { |
| "epoch": 24.58, |
| "learning_rate": 5e-05, |
| "loss": 0.1912, |
| "step": 1939 |
| }, |
| { |
| "epoch": 24.6, |
| "learning_rate": 5e-05, |
| "loss": 0.2069, |
| "step": 1940 |
| }, |
| { |
| "epoch": 24.61, |
| "learning_rate": 5e-05, |
| "loss": 0.1518, |
| "step": 1941 |
| }, |
| { |
| "epoch": 24.62, |
| "learning_rate": 5e-05, |
| "loss": 0.2153, |
| "step": 1942 |
| }, |
| { |
| "epoch": 24.63, |
| "learning_rate": 5e-05, |
| "loss": 0.2978, |
| "step": 1943 |
| }, |
| { |
| "epoch": 24.65, |
| "learning_rate": 5e-05, |
| "loss": 0.1987, |
| "step": 1944 |
| }, |
| { |
| "epoch": 24.66, |
| "learning_rate": 5e-05, |
| "loss": 0.2692, |
| "step": 1945 |
| }, |
| { |
| "epoch": 24.67, |
| "learning_rate": 5e-05, |
| "loss": 0.1771, |
| "step": 1946 |
| }, |
| { |
| "epoch": 24.68, |
| "learning_rate": 5e-05, |
| "loss": 0.2638, |
| "step": 1947 |
| }, |
| { |
| "epoch": 24.7, |
| "learning_rate": 5e-05, |
| "loss": 0.1715, |
| "step": 1948 |
| }, |
| { |
| "epoch": 24.71, |
| "learning_rate": 5e-05, |
| "loss": 0.2392, |
| "step": 1949 |
| }, |
| { |
| "epoch": 24.72, |
| "learning_rate": 5e-05, |
| "loss": 0.2175, |
| "step": 1950 |
| }, |
| { |
| "epoch": 24.74, |
| "learning_rate": 5e-05, |
| "loss": 0.1938, |
| "step": 1951 |
| }, |
| { |
| "epoch": 24.75, |
| "learning_rate": 5e-05, |
| "loss": 0.1901, |
| "step": 1952 |
| }, |
| { |
| "epoch": 24.76, |
| "learning_rate": 5e-05, |
| "loss": 0.1909, |
| "step": 1953 |
| }, |
| { |
| "epoch": 24.77, |
| "learning_rate": 5e-05, |
| "loss": 0.217, |
| "step": 1954 |
| }, |
| { |
| "epoch": 24.79, |
| "learning_rate": 5e-05, |
| "loss": 0.2872, |
| "step": 1955 |
| }, |
| { |
| "epoch": 24.8, |
| "learning_rate": 5e-05, |
| "loss": 0.2159, |
| "step": 1956 |
| }, |
| { |
| "epoch": 24.81, |
| "learning_rate": 5e-05, |
| "loss": 0.2022, |
| "step": 1957 |
| }, |
| { |
| "epoch": 24.82, |
| "learning_rate": 5e-05, |
| "loss": 0.1672, |
| "step": 1958 |
| }, |
| { |
| "epoch": 24.84, |
| "learning_rate": 5e-05, |
| "loss": 0.2629, |
| "step": 1959 |
| }, |
| { |
| "epoch": 24.85, |
| "learning_rate": 5e-05, |
| "loss": 0.1835, |
| "step": 1960 |
| }, |
| { |
| "epoch": 24.86, |
| "learning_rate": 5e-05, |
| "loss": 0.1562, |
| "step": 1961 |
| }, |
| { |
| "epoch": 24.87, |
| "learning_rate": 5e-05, |
| "loss": 0.2913, |
| "step": 1962 |
| }, |
| { |
| "epoch": 24.89, |
| "learning_rate": 5e-05, |
| "loss": 0.1545, |
| "step": 1963 |
| }, |
| { |
| "epoch": 24.9, |
| "learning_rate": 5e-05, |
| "loss": 0.2616, |
| "step": 1964 |
| }, |
| { |
| "epoch": 24.91, |
| "learning_rate": 5e-05, |
| "loss": 0.1081, |
| "step": 1965 |
| }, |
| { |
| "epoch": 24.93, |
| "learning_rate": 5e-05, |
| "loss": 0.237, |
| "step": 1966 |
| }, |
| { |
| "epoch": 24.94, |
| "learning_rate": 5e-05, |
| "loss": 0.2414, |
| "step": 1967 |
| }, |
| { |
| "epoch": 24.95, |
| "learning_rate": 5e-05, |
| "loss": 0.2651, |
| "step": 1968 |
| }, |
| { |
| "epoch": 24.96, |
| "learning_rate": 5e-05, |
| "loss": 0.158, |
| "step": 1969 |
| }, |
| { |
| "epoch": 24.98, |
| "learning_rate": 5e-05, |
| "loss": 0.2477, |
| "step": 1970 |
| }, |
| { |
| "epoch": 24.99, |
| "learning_rate": 5e-05, |
| "loss": 0.1836, |
| "step": 1971 |
| }, |
| { |
| "epoch": 25.0, |
| "learning_rate": 5e-05, |
| "loss": 0.1485, |
| "step": 1972 |
| }, |
| { |
| "epoch": 25.01, |
| "learning_rate": 5e-05, |
| "loss": 0.1532, |
| "step": 1973 |
| }, |
| { |
| "epoch": 25.03, |
| "learning_rate": 5e-05, |
| "loss": 0.1454, |
| "step": 1974 |
| }, |
| { |
| "epoch": 25.04, |
| "learning_rate": 5e-05, |
| "loss": 0.1787, |
| "step": 1975 |
| }, |
| { |
| "epoch": 25.05, |
| "learning_rate": 5e-05, |
| "loss": 0.1433, |
| "step": 1976 |
| }, |
| { |
| "epoch": 25.06, |
| "learning_rate": 5e-05, |
| "loss": 0.1635, |
| "step": 1977 |
| }, |
| { |
| "epoch": 25.08, |
| "learning_rate": 5e-05, |
| "loss": 0.1169, |
| "step": 1978 |
| }, |
| { |
| "epoch": 25.09, |
| "learning_rate": 5e-05, |
| "loss": 0.1417, |
| "step": 1979 |
| }, |
| { |
| "epoch": 25.1, |
| "learning_rate": 5e-05, |
| "loss": 0.2019, |
| "step": 1980 |
| }, |
| { |
| "epoch": 25.12, |
| "learning_rate": 5e-05, |
| "loss": 0.1088, |
| "step": 1981 |
| }, |
| { |
| "epoch": 25.13, |
| "learning_rate": 5e-05, |
| "loss": 0.1885, |
| "step": 1982 |
| }, |
| { |
| "epoch": 25.14, |
| "learning_rate": 5e-05, |
| "loss": 0.1823, |
| "step": 1983 |
| }, |
| { |
| "epoch": 25.15, |
| "learning_rate": 5e-05, |
| "loss": 0.1741, |
| "step": 1984 |
| }, |
| { |
| "epoch": 25.17, |
| "learning_rate": 5e-05, |
| "loss": 0.1966, |
| "step": 1985 |
| }, |
| { |
| "epoch": 25.18, |
| "learning_rate": 5e-05, |
| "loss": 0.1877, |
| "step": 1986 |
| }, |
| { |
| "epoch": 25.19, |
| "learning_rate": 5e-05, |
| "loss": 0.1433, |
| "step": 1987 |
| }, |
| { |
| "epoch": 25.2, |
| "learning_rate": 5e-05, |
| "loss": 0.1397, |
| "step": 1988 |
| }, |
| { |
| "epoch": 25.22, |
| "learning_rate": 5e-05, |
| "loss": 0.2733, |
| "step": 1989 |
| }, |
| { |
| "epoch": 25.23, |
| "learning_rate": 5e-05, |
| "loss": 0.1591, |
| "step": 1990 |
| }, |
| { |
| "epoch": 25.24, |
| "learning_rate": 5e-05, |
| "loss": 0.1155, |
| "step": 1991 |
| }, |
| { |
| "epoch": 25.26, |
| "learning_rate": 5e-05, |
| "loss": 0.22, |
| "step": 1992 |
| }, |
| { |
| "epoch": 25.27, |
| "learning_rate": 5e-05, |
| "loss": 0.1074, |
| "step": 1993 |
| }, |
| { |
| "epoch": 25.28, |
| "learning_rate": 5e-05, |
| "loss": 0.1704, |
| "step": 1994 |
| }, |
| { |
| "epoch": 25.29, |
| "learning_rate": 5e-05, |
| "loss": 0.2257, |
| "step": 1995 |
| }, |
| { |
| "epoch": 25.31, |
| "learning_rate": 5e-05, |
| "loss": 0.1605, |
| "step": 1996 |
| }, |
| { |
| "epoch": 25.32, |
| "learning_rate": 5e-05, |
| "loss": 0.2086, |
| "step": 1997 |
| }, |
| { |
| "epoch": 25.33, |
| "learning_rate": 5e-05, |
| "loss": 0.1974, |
| "step": 1998 |
| }, |
| { |
| "epoch": 25.34, |
| "learning_rate": 5e-05, |
| "loss": 0.2178, |
| "step": 1999 |
| }, |
| { |
| "epoch": 25.36, |
| "learning_rate": 5e-05, |
| "loss": 0.1402, |
| "step": 2000 |
| } |
| ], |
| "max_steps": 3900, |
| "num_train_epochs": 50, |
| "total_flos": 0.0, |
| "trial_name": null, |
| "trial_params": null |
| } |
|
|
